Aspects of the biology and growth of three species of (: ).

by Valerie Kathleen BROWN, B.Sc.(Lond.), A.R.C.S.

Thesis submitted for the Degree of Doctor of Philosophy

November 1969 Imperial College of Science and Technology, Silwood Park, Sunninghill, ASCOT, Berkshire. -1-

ABSTRACT

This thesis concerns three species in the genus Ectobius Stephens which occur in Britain. The basic life histories of the species are clarified and several biological topics are considered in more detail. Aspects of the oviposition behaviour in mated and unmated females and the extent of parthenogenesis in the species are investigated. The oothecae pass the winter in a state of dormancy which has been confirmed as a diapause in Ectobiuslayponicus. (Linnaeus). Oothecae are subject to attack by the Evaniid parasite, Brachygaster minutus (Olivier); the life history of this species is considered in relation to that of the host. The overwintering behaviour of the nymphal instars of E. lapponicus and Ectobius 1,allidus (Olivier) in a range of intermediate instars has been found to involve a diapause in the former species. The relationship between the proportion of nymphs entering the winter in each instar and the nature of the adult emergence the following summer is discussed. The post-embryonic growth of two species with different life cycles, E. lapponicus and Ectobius panzeri Stephens, is considered mainly from an analytical standpoint. A means of determining the sex of the nymphal instars is described and thus permits a more detailed study. The post-embryonic development is analysed by several techniques, each of which is applied to a large number of characters. Dyar's Law (1890) and its extension by Richards (1949) are assessed. A detailed appraisal of simple allometry of growth, including alternative methods of deriving the constants and a full range of significance tests, is made. However, emphasis is focused on the more sensitive multivariate techniques which have only recently been applied to growth. In this work, for the first time, several of the available techniques, including a generalisation of the allometry equation, are applied to the same body of data, thereby enabling an evaluation of these methods in the study of growth.

-2-

TABLE OF CONTENTS.

Page ABSTRACT 1 TABLE OF CONTENTS 2 GENERAL INTRODUCTION 7 GENERAL MATERIALS AND METHODS (i) Collecting Methods 8 (ii)Culture Methods 10

SECTION A: THE BIOLOGY OF ECTOBIUS SPECIES Introduction and Review of Literature 17 Introduction to the Biology of the Species (i) General a. Life History 23 b. Habits 26 (ii) Habitat Preference a. List of Collecting Sites 29 b. Description of Habitats 31 c. Overwintering Sites 33

SELECTED BIOLOGICAL TOPICS

I. REPRODUCTIVE BIOLOGY 36 (i) Aspects of Oviposition in Mated and Unmated Females under Natural and Controlled Conditions and the Hatching of the Resultant Eggs a. The Number of Oothecae Deposited 39 b. The Preoviposition Period 41 c. The Interval between the Deposition of Oothecae 43 d. The Number of Eggs per Ootheca 44 e. The Weight of the Ootheca 46 -3-

Page f. The Size of the Ootheca (i) Length 48 (ii)Width 50 g. The Period of Retention of the Ootheca 51 h. The Direction of Rotation of the Ootheca 53 i. The Method of Deposition of the Ootheca 57 j. The Percentage of Oothecae which Hatch and the Number of Nymphs which Emerge from an Ootheca 58 (ii) The Longevity of Mated Males and Females 62

II. THE EGG STAGE (i) Description of the Ootheca 64 Key to the Oothecae of the British Species of Ectobius 70 (ii)Determination of Diapause in the Oothecae of E. lapponicus 71 (iii)Water Relations of the Oothecae 86 (iv)Hatching and Description of Pronymph 93

III. OVERWINTERING OF THE NYMPHAL INSTARS 98 (i) Field Collections of Nymphal Instars 99 (ii)Overwintering Behaviour of Nymphs in an Outdoor Insectary 100 (iii)The Occurrence of an Abnormal Instar under Experimental Conditions 102 (iv)Determination of Diapause in the Nymphs of E...lapponicus 103 (v) The Effect of a Short Day Regime on the Nymphs of E. lapponicus 110 (vi)Relationship between the Overwintering Stages and Adult Emergence 113 Page IV. BIOLOGY OF THE EGG PARASITE, BRACHYGASTER MINUTUS (OLIVIER) Materials and Methods a. Collecting Methods 119 b. Culture Methods 119 (i) Life History and Habits 120 (ii)Mating 123 (ii) Oviposition 124 (iv)Emergence 126

GENERAL DISCUSSION 130

SUMMARY OF THE BIOLOGY SECTION 137

SECTION B: THE GROWTH OF ECTOBIUS SPECIES Introduction and Review of Literature 140 Introduction to the Growth of the Species (i) Key to the Nymphal Instars of the British Species of Ectobius 154 (ii)The Development of the Genital Segments and their use in Distinguishing the Sexes 168

ANALYSIS OF GROWTH

Materials and Methods 188

I. SUMMARY AND STORAGE OF DATA 193

II. DYAR'S LAW

(i) Application of Dyar's Law 194 (ii)Application of Richards' Extension of

Dyar's Law 198

(iii)Application of Przibram's Rule 203 -5- Page III. ALLOMETRIC ANALYSIS (i) The Choice of a Reference Dimension 208 (ii) Fitting the Allometric Equation 208 (iii) Significance Tests 220 a. The Significance of the Slope of the Line 221 b. The Significance of Deviations from a equal to Unity 221 c. The Significance of Deviations from Linearity 225 (iv) Growth Gradients 231 a. Growth Gradient in the Mid-Dorsal Line 232 b. Growth Gradient in the Mid-Ventral Line 234 c. Growth Gradient in the Legs 236 d. Growth Gradient in the Antennae 243 (v) Growth Contours 246

IV. MULTIVARIATE ANALYSIS (i) Eigenanalysis of Correlation Matrices Introduction to the Technique 257 Outline of Statistical Methods and Details of Programs utilised 258 a. The Correlation Matrix 258 b. The Latent Roots and Vectors of the Correlation Matrix 263 (ii) Principal Component Analysis Introduction to the Technique 276 Outline of Statistical Methods and Details of Programs utilised 277 a. Correlation Matrix 278 b. Log. Covariance Matrix 297 c. The Generalised Allometry Parameter 309 (iii) Factor Analysis 316 -6- Page (iv) Multiple Discriminant Analysis Introduction to the Technique 319 Test for Homogeneity of Covariance Matrices 320 Outline of Statistical Methods and Details of Programs utilised 321 a. Untransformed Data 322 b. Logarithmically Transformed Data 329

GENERAL DISCUSSION 346 SUMMARY OF THE GROWTH SECTION, including an Evaluation of the MethodS used in the Analysis of the

Growth of Two Species of Ectobius 352

ACKNOWLEDGMENTS 356

REFERENCES 357

STATISTICAL APPENDIX 374 -7-

GENERAL INTRODUCTION.

Lucas (1920) in his monograph of British Orthoptera recorded three species in the genus Ectobius Stephens. These native species, (Linnaeus), the Dusky , (Olivier), the Tawny Cockroach and Ectobius panzeri Stephens, the Lesser Cockroach are the subject of this thesis. The thesis is presented in two sections. The first part is concerned with the general biology of the species. Very little is known to date of the biology of native, field-dwelling , the wealth of literature on the Dictyoptera being centred around the cosmopolitan, domestic species. The second part of the thesis deals with the post-embryonic growth of two of the species, E. lapponicus and E. panzeri. Dyar's Law and the Law of Simple Allometry have in the past constituted the only analytical methods to be applied to growth data, and these were usually restricted to a limited range of characters in a single species. In the present work emphasis has been placed on the application of multivariate analyses to the growth of the species; a field which seems very well suited to these more sensitive techniques. This is, to some extent, an exploratory study in which several separate analyses are applied to a wide range of characters measured on two closely related species. -8-

GENERAL MATERIALS AND METHODS.

(i) Collecting Methods.

The collection of specimens proved to be both time-consuming and tedious. Many collecting methods were attempted but only one or two were utilised when large numbers of were required. Generally E. lapponicus and E. pallidus were collected by similar methods, but different methods had to be adopted for E. panzeri.

Method 1. All developmental stages of both E. lapponicus and E. pallidus were collected in substantial numbers by lifting the dead, fallen fronds of bracken, Pteridium aquilinum (Linnaeus) Klihn, and inserting a metre square beating tray under them. The fronds were then gently shaken, and the debris on the beating tray sorted. The bracken had to be dry for this method to be efficient. This means of collection became totally impractical once the young bracken shoots commenced the new season's growth. Method 2. An equally good procedure was the beating of low-lying branches onto a beating tray. This was particularly effective when grass blades mingled with the lowermost branches of trees and shrubs. Numerous specimens of E. pallidus were collected in this way from the branches of Viburnum lantana Linnaeus, and in particular Picea abies (Linnaeus)Karst, the Norway Spruce, yielded vast numbers of E. lapponicus. Moderate numbers of both.species were found on young branches of Quercus Linnaeus species, and all stages of E. lapponicus were found on young bushes of broom Sarothamnus scoparius (Linnaeus) Wimmer. Method 3. The most rewarding method for the collection of E. lapponicus and E. pallidus involved the "preparation" of the habitat and the revisiting of the site for collecting several days later. The preparation entailed lifting the dead grass from the cores of large tussocks. This could be efficiently performed by -9-

thrusting a garden fork through the base of the tussock, just above ground level and drawing it vertically upwards, thus freeing the considerable quantity of dry, dead grass. The latter was condensed into a clump and placed back on the surface of the tussock. After a few, warm, dry days very large numbers of all developmental stages would accumulate in the clump of dry grass. This could be lifted onto a beating tray intact with a minimal amount of disturbance enabling a much higher percentage of the cockroaches to be collected than in the two former methods, where disturbance of the vegetation prior to placing it on the beating tray undoubtedly caused many cockroaches to be lost, especially adults and last instar nymphs. Method 4. Male adults of E. lapponicus were collected in numbers by sweeping through the flower heads of various grass species on a hot, sunny afternoon, with a sweep net. Adult males could also be induced to climb into a 7.6 x 2.5 cns. glass tube placed over a grass stem. A few adult females may be collected by this method also. Method 5. The sorting of oak and pine litter on a beating tray proved relatively ineffective for the collection of most stages of E. lapponicus and E. pallidus but adult females of the former species were often abundant in pine litter. Young instars were gathered from the beating tray by using an aspirator; this method was found to be impractical for late instars and adults, which were collected in 7.6 x 2.5 cms. glass tubes to prevent injury. It was noted that both nymphs and adults of E. lapponicus assembled in moderate numbers at the base of a Robinson light trap. However, this method was not adopted, since it was difficult to collect the cockroaches from under the trap. Method 6. For collections of E. lapponicus and E. pallidus during the winter months the most effective method was found to be the cutting of grass tussocks near ground level with a sharp knife, or preferably a saw. The whole tussock was brought into the laboratory -10- in a sealed polythene bag. After an hour at room temperature many of the insects, including the cockroach nymphs, became active and could be shaken from the tussock onto a beating tray. Method 7. No adequate method was found for collecting large numbers of E. panzeri. By far the best means of collecting all stages of this species was the careful combing of the habitat for individual specimens. All instars with the exception of the adult could be collected from the ground by means of an aspirator. Adults were collected singly in 7.6 x 2.5 ems. glass tubes. Adults, predominately males, were collected on occasions in small numbers by sweeping patches of heather, Erica cinerea Linnaeus or ling, Calluna vulgaris (Linnaeus) Hull. Method 8. Cutting the growing stems of bracken and shaking these onto a beating tray yielded a few young instars of E. panzeri, but the numbers obtained did not warrant the labour entailed. It was found throughout this study that the efficiency of any collecting method for any species was greatly effected by weather conditions and the time of day. The vegetation should preferably be dry, and conditions hot and sunny. The late afternoon and early evening proved to be the most favourable collecting times. It was vital to exercise considerable care in the collection of these insects, as they were very easily damaged.

(ii) Culture Methods.

Much difficulty was initially encountered in the rearing of these native field-dwelling species of cockroaches. Roth (personal communication) also expressed this difficulty with species of the genus. Many preliminary experiments were performed before the species could be cultured satisfactorily in the laboratory. The nature of the food given was found to be of paramount importance in each species. Eventually two diets were provided; these were mixed together in equal quantities and were used throughout the -11-

course of the work. One diet offered was that suggested by Gordon (1959) for Blatella germanica (Linnaeus) as a standard "natural" diet. This diet consisted of:- 75% wheat germ 20% dried skimmed milk 5% dried brewer's yeast and was supplemented by a combination of two brands of dog biscuits (equal parts by weight of Winalot and Spiller's Shapes) which were finely ground. These provided the additional vitamins and minerals which were found necessary for B. germanica by Gordon (1959). The other basic requirement in the culture of these species was found to be a supply of readily available water. Folded sheets of filter paper increased the available space and provided moulting sites for the nymphs. It was also vital to prevent the insects escaping from the cage. Vaseline was used initially for this purpose but was later superceded by a dispersion of polytetrafluoroethylene (P.T.F.E.) known commercially as Fluon. This was painted in a thin film around the top of the container. Fluon was far more satisfactory than vaseline, since younger instars had a tendency to become trapped in the latter. Normal variations in humidity appeared to be of little importance, and this factor was not controlled. All containers were furnished with some means of ventilation, which prevented the otherwise rapid growth of mould on the food.

Different containers were used as cages for the cockroaches, according to the numbers involved. Method 1. For small numbers of adults or very young instars, round transparent plastic cages 10 cms. in diameter and 4 cms. high were used (Fig. 1a). Food and water were contAned in polythene caps 1.5 cms. in diameter. Absorbent dental rolls (size 4) were divided into three equal parts; each of which, fitted into one of the caps, could retain distilled water for several days. Folded strips of filter paper afforded suitable substitutes for natural -12- shelters. A circular hole 2 ems. in diameter and covered with terylene net (11.5 meshes per cm.) provided the required ventilation. It was necessary to renew or replenish the food and water every three or four days in this type of cage. Method 2. For larger numbers of insects or for long-term experiments when frequent renewal of food and water had to be avoided, transparent, rectangular plastic cages 13.5 ems. long, 7.5 ems. wide and 6.0 ems. high were used, with three, 2 ems. diameter net covered holes in the lid (Fig. lb). Food was contained in the top of a plastic pill-box 5.3 cms. in diameter, and a wad of wet cotton wool in the base provided a supply of water. Strips of folded filter paper were again scattered at the base of the container. Water was replenished at weekly intervals and food renewed after two weeks. Method 3. For very large collections of nymphs or adults glass tanks of.length 36 ems. and a width and height of 23 cnis. were utilised (Plate 1). These were fitted with a muslin (32 meshes per cm.) lid having a central sleeve to facilitate access to the interior of the vessel. The food and water were contained in several plastic pill-boxes as in the previous method. The floor of the tank was covered with folded strips of filter paper. Method 1+. Nymphs reared under outdoor conditions during the winter months required a more protective culture method, and were kept successfully in Watkins and Doncaster cages, in the base of which a small grass tussock was planted in a mixture of peat and sand (Plate 2). The tussock remained in good condition throughout the winter and required only periodic watering. A small container of food was provided. Nymphs were not reared successfully during the winter in dry bracken litter; this was presumably due to a lack of moisture.

Oothecae were kept in 5.0 x 1.0 ems. glass tubes. A thick layer of fine white sand filled the base of each tube, and a wad of teased non-absorbent cotton wool was positioned in the top

-1 3 -

net covered aperture

film of fluon round plastic container

food cotton wool 111111Ma roll filter paper 1111r1111111111111111111111711A .1111/ polythene cap

5cms.

FIG la

net covered aperture

film of fluon rectangular cotton plastic box wool

food plastic pill box filter paper

5cms

FIG lb. Cages used for Rearing Cockroaches for Experimental Purposes. Plate 1 Cage for rearing large numbers of cockroaches (Method 3).

Plate 2 Watkins & Doncaster cage (Method 4). -15-

(Fig. 2). The tube and contents were sterilised, using an autoclave (20 lb, pressure for 20 - 30 minutes). The sand was watered initially with a 0.1 w/w solution of Nipagin M (Methyl parahydroxy- benzoate). This retarded nearly all fungal growth and the sand was subsequently kept damp only by the addition of sterile water. The oothecae themselves were also painted with a solution of this fungicide. The tubes were stored in honeycomb in the rectangular plastic boxes, previously described (Plate 3). Each box stored a maximum of ninety tubes. This method was very efficient since the hatching of any oothecae was immediately apparent when the box was viewed laterally, as nymphs of the three species have an innate ability to climb minutes after leaving the ootheca. Oothecae were inspected for fungal growth at three-weekly intervals and sterile water added to the sand if it had become dry. teased cotton wool

glass tube

— fine sand 3cms

FIG 2 Tube used for Keeping Oothecae until Hatching

Plate 3 Container for storing overwintering oothecae. -17-

SECTION A: THE BIOLOGY OF ECTOBIUS SPECIES.

Introduction and Review of Literature.

The three British species of Ectobius are restricted in distribution to the south of the British Isles (Kevan, 1961; Ragge, 1965); but are widely distributed in Europe (Lucas, 1920), whilst E. liallidus has recently been introduced into the United States of America (Flint, 1951). Ragge (1965) in the most recent monograph of British Orthopteroid insects, referring to the British species of Ectobius, states *the life histories of our three native species remain a mystery, particularly the question of how they pass the winter". The over- wintering stage of E. lapponicus and E. pallidus was thought to be the adult by Lucas (1920), although Killington (1927), Lucas (1928) and Haines (1936) confirmed that E. lapponicus overwinters as a nymphal instar. The latter author suggested the presence of two broods a year to explain the combined occurrence of young nymphs and adults. No hibernating nymphs or adults of E. panzeri were found, although several intact oothecae were recorded in the field during the winter (Brown E., 1952). E. pallidus is known to overwinter in the United States as oothecae (Roth & Willis, 1957) although some nymphs may also overwinter (Gurney, 1953) and these either die or hibernate (Roth & Willis, 1957). The oothecae of E. lapponicus were thought to hatch soon after deposition (Lucas, 1928). The life histories of the three British species of Ectobius have been considered in detail, with a more intensive study of several aspects of their biology. Little.is known of the bionomics of native species of cockroaches, although there is an extensive literature involving the familiar domestic species. The biology of pensylvanica (De Geer) and Panchlora nivea (Linnaeus) have been investigated by -18-

Rau (1940) and Roth & Willis (1958a) respectively. (Poda) and E. lapponicus were included in an account of the biology of several species of cockroaches due to Harz (1960). The bionomics of E. pallidus have been considered in slightly more detail (Roth & Willis, 1957), and limited observations on E. panzeri were recorded by Brown (1952). The British species of Ectobius are not known to be economically important; however, E. lapponicus has been reported as a pest of dried fish in Lappland although no evidence for this was found by Gaunitz (1936). A key to the adults of the British species of Ectobius has been constructed by Blair (1934), and keys to the immature stages and oothecae are given in this thesis. By limiting the investigation of the biology of the three species of Ectobius to four main topics a more detailed experimental analysis was possible; this yielded more conclusive results than a general approach to the whole biology of the species. The reproductive biology of many species of cockroaches was reviewed by Roth & Willis (1954a). Cockroaches may be classified with reference to their oviposition behaviour into oviparous, ovoviviparous and viviparous species (Shelford, 1906); E. panzeri and E. pallidus being quoted as examples of oviparous species (Roth & Willis, 1955c). The reproduction and development of a selection of species from each of the above groups have been compared by Willie, Riser & Roth (1958), whilst the effect of temperature on the fecundity and longevity of Periplaneta americana (Linnaeus) has been analysed by Griffiths & Tauber (1942a) The long duration of adult life in this species enabled.the effect of mating, at various stages in the life of the female, on the pattern of oviposition to be determined. Parthenogenesis has been found to occur in cockroaches; but successful reproduction by this method is not common (Roth & Willis, 1956). Several aspects of the fecundity, oviposition and longevity of mated and virgin females of the three species of Ectobius under natural and controlled conditions are examined. Notes on the oviposition of E. panzeri were made by -19 -

Brown (1952), although hi.s observations were confined to females collected at the end of the season and these have since been found to behave atypically. There are few detailed descriptions of cockroach oothecae. Shelford (1912) made observations on the external features of oothecae of several species. Lawson (1951) described the oothecae of six common oviparous. species, and found a considerable range in complexity of structure, particularly with regard to the dorsal seam or keel. The species he studied could be divided into two groups, one having a complicated keel arrangement, e.g. P. pensylvanica, whilst the other was structurally simple e.g. orientalis Linnaeus. In later papers (1952, 1953 & 1951+) he gives lucid accounts of the oothecae of lutea lutea (Saussure & Zehntner), Eurycotis floridana (Walker) and Parcoblatta uhleriana (Saussure) with emphasis on the diversity of keel structure. Roth (1968a), using the classification of McKittrick (1964), described oothecal structure in relation to the evolution of oviparity and viviparity, a change which necessitated the reduction of the ootheca. Eggs enclosed in the oothecae of most cockroach species commence their development immediately after ootheca formation and hatch in five or six weeks (Willis, Riser & Roth, 1958). The embryos of the three species of Ectobius were thought to enter a "resting phase" after their formation (Ragge, 1965). However, the presence of a diapause has not yet been confirmed in the oothecae of these species. Roth & Willis (1957) suggested a dormant period or diapause in the oothecae of E. pallidus, since hatching was only induced after the oothecae had been subjected to a period of dryness. The oothecae of E. panzeri which were deposited by females in the laboratory and also those collected from the field during the winter did not hatch until the following year (Brown, 1952). A diapause has not yet been confirmed in the oothecae of the other native species, although overwintering egg capsules of P. pensylvanica and Parcoblatta virginica (Brunner) have been recorded by -20-

Edmunds (1952b). The water content of the oothecae of many species of cockroaches, together with the changes which occur during embryogenesis, have been considered in relation to the evolution of ovoviviparity and viviparity (Roth, 1967b). It was found that the oothecae of E. pallidus only hatched after the uptake of water (Rolander, unpublished observations in Roth & Willis, 1957). The water content of the oothecae of this species was found to be approxi- mately 40% after formation but increased to 75% immediately before hatching (Roth & Willis, 1958a), the water being absorbed over a short period. The water relations of the main oviposition types were discussed by Roth & Willis (1955c), although no mention was made of the group to which Ectobius belonged. The determination of a diapause and the water relations of the oothecae of E. lapponicus are considered in some detail in this thesis. The wood roaches P._Lensylvanica and P. virinica overwinter as nymphal instars of varying sizes (Edmunds, 1952b). Nymphs of these species are inactive in the field during the winter, but become agile when warmed and probably move and feed on warm days in the winter (Holmquist, 1926). E. lapponicus and E._pallidus also overwinter as nymphs in the British Isles, a range of instars entering the winter. The occurrence of diapause in more than one instar is not common in insects. However, a few examples among Orthopteroid insects include the following:- the alpine wetas, Hemideina maori (Pictet et Saussure) in New Zealand which over- winter in three or four different stages (Sutherland, 1964), Nemobius yezoensis Shiraki (= Pteronemobius nitidus (Bolivar)) which overwinters in a range of late instars (Masaki & Oyama, 1963), Acheta (= Gryllus) veletis Alexander & Bigelow which also diapauses in one of several late nymphal instars (Bigelow, 1960), whilst in Chortophaga viridifasciata (De Geer) the winter is passed as either a third or fourth instar (Halliburton & Alexander, 1964). The occurrence of a two year life cycle similar to that of E. lapponicus and E. pallidus with a period of dormancy in the egg and in the -21- nymph has been recorded in two species of Acrididae in central Saskatchewan, Pardalophora apiculata (Harris) and Xanthippus corallipes latefasciatus (Scudder) (Pickford, 1953). The overwintering of the nymphs of Anax imperator Leach in two instars resulted in a bimodal adult emergence of the species the following spring (Corbet, 1955). The relationship between the proportion of nymphs overwintering in each stage of Ectobius and the emergence pattern of the adults the following year is discussed in this thesis. The effect of photoperiod on the induction and termination of diapause is considered in these species. The daily increase in photoperiod was thought to induce diapause in the nymphs of A. imperDtor (Corbet, 1955, 1956). On the other hand, increasing photoperiods were responsible for the termination of diapause in the nymphs of C. viridifasciata (Halliburton & Alexander, 1964). The numerous parasites, commensals and symbionts of cockroaches have been reviewed by Roth & Willis (1960). The only parasite which has been considered in this thesis is a member of a small group of parasitic Hymenoptera, the Evanioidea. The morphology and of the British species were examined in detail by Crosskey (1951), who referred to the extreme paucity of information on the habits and life histories of these species. Only two species of occur in Britain and both are parasites of cockroach egg capsules: firstly Evania appendigaster (Linnaeus) which has a world-wide distribution, although no localities for its capture are known in this country (Crosskey, 1951) and no specimens were found during the course of this study, and secondly Brachygaster minutus (Olivier) which has a widespread distribution in Europe, although few British records of its occurrence exist (Crossley, 1951). Remarkably little is known of the biology of B. minutus; however, several closely related species have been studied, Cameron (1957) described the developmental stages and biology of E. appendigaster, and Haber (1920) made observations on its oviposition, Edmunds (1952aj 1954) and Cros (1942) investigated the -22- oviposition behaviour and biology of Prosevania punctata (Brune), and Genieys (1924) studied the biology of Zeuxevania splendidula Costa. The percentage parasitism of the oothecae of P. americana by E. appendigaster averaged 25 - 29% (Cameron, 1957), the author suggesting the use of the parasite in the control of the cockroach. Each Evaniid larva is capable of destroying 16 - 40 cockroach eggs (Edmunds, 1954). Edmunds (1952b) recorded that 12.4% of the overwintering egg capsules of P. pensylvanica and P. virginica were parasitised, 6.7% by Evaniids, this percentage varied in different years. Species of the genus Hyptia Illiger are parasites of the native wood roaches.in North America, and these produce only one annual generation, the last larval instar or pupa over- wintering in the cockroach egg capsule (Edmunds, 1954). The life history and biology of B. minutus, which has been found to parasitise the oothecae of the three British species of Ectobius, has been examined in some detail in this thesis. -23-

Introduction to the Biology of the Species.

(i) General.

t. Life History.

The species of the genus Ectobius occurring in the British Isles differ in the duration of their life cycles. E. panzeri is univoltine, whereas E. lapponicus and E. pallidus require two years for the completion of their development. A diagrammatic representation of the life histories of the three species is given in Fig. 3. The post-embryonic development of E. lapponicus and E. panzeri is characterised by having five nymphal instars in each sex. However, both sexes of E. pallidus have an additional sixth nymphal instar, before becoming adult. The eggs are enclosed in tough, compact oothecae, and the number present in each ootheca varies according to the species. Prior to hatching the oothecae of all species increase visibly in size and assume a much paler colour. A fissure along the length of the keel of the ootheca permits the escape of the pronymphs. Hatching occurs over a limited period only. In E. lapponicus and E. panzeri the oothecae hatch during early June, whilst in E. pallidus hatching takes place a little later in the season from mid June to the beginning of July. These periods fluctuate slightly in different seasons. The pronymphal ecdysis occurs very shortly after hatching. The nymphs of E. panzeri attain maturity in seven to twelve weeks. The first adults are noticed in the field at the end of July, and few final instars are to be found later than mid August. This species exhibits a more rapid nymphal development than the other two species. The adult emergence of males and females coincides, but the greater longevity of the females results in a larger number of

E panzeri

E. lapponicus Oothecae

Nymphs

E pallidus Adults

Jan Mar May Jul Sep Nov Jan Mar May Jul Sep Nov

First Year Second Year

FIG 3 Diagrammatic Representation of the Life Histories of the British Species of Ectobius. -25- females at the end of the season though few persist beyond the middle of October. The female produces a small number of oothecae, which are rotated through 900 prior to deposition. The female carries an ootheca for one or two days. The oothecae overwinter in a state of diapause, and hatch the following year. In contrast, the development of the nymphs of E. lapponicus and E. pallidus is still incomplete by the end of the first season. The nymphs overwinter and complete their growth the following year. Hibernation occurs in a variety of instars, but never in the first or final instar. In E. lapponicus the greatest percentage of nymphs cease development as third or fourth instars, but some nymphs of E. pallidus reach the penultimate or fifth nymphal instar. The proportion of each instar varies according to seasonal conditions. The nymphs resume activity in April of the following year and undergo their final ecdyses. The first adults of E. pallidus emerge by mid June and these are usually males; the remainder become adult during the next two months. Adults of E. lapponicus are found in the field from the end of May and continue to emerge until the beginning of August. As in E. pallidus males predominate in the early part of the season. Adults of both species decline in numbers during the early part of October and few remain after the end of the month. The production of oothecae by females of E. lapponicus and E. pallidus is similar in that both carry their oothecae for one or two days before deposition. Rotation occurs as in E. panzeri. The winter months are similarly passed in a state of dormancy. All three species are capable of producing oothecae partheno- genetically in the laboratory, but the importance of this in field populations is uncertain. Only one entomophagous parasite was found, namely B. minutus, an Evaniid which parasitises the eggs of the three species. However, nymphs and adults were liable to attack by a species of fungus of the genus Entomopthora (probably Entomopthora grylli (Fresenius)). -26-

Victims of the fungus were found with their ventral surfaces closely adherent to structures such as blades of grass. It was particularly common for adult males of E. lapponicus to be seen cemented to the flower heads of Holcus Linnaeus species, (Plate 4), or for nymphs to be attached to particles of bracken debris, (Plate 5). b. Habits.

The nymphs and adults of all species are extremely active, and all stages are capable of very rapid movement. These species exhibit a marked periodicity in activity. Early in the day all stages are relatively sluggish, but as the day proceeds their activity increases and reaches a peak in the late afternoon and early evening. At this time nymphs scurry from place to place, and climb nearby blades of grass with considerable agility. By this means they reach the lower branches of trees and shrubs, where they may collect in large numbers when the weather is favourable. Adults of E. pallidus and E. panzeri which often frequent sand dunes, move.between tufts of marram grass, Ammophila arenaria (Linnaeus) Link, with such rapidity that they become covered with fine sand and are only noticed by careful observation. The nymphal instars have a tendency to remain within the boundaries of the large tufts. The sparsely covered ground where E. panzeri nymphs are commonly to be found is often a scene of intense activity, the young instars continually traversing the bare patches of ground. Nymphs and adults normally move with their antennae held forward; however, when actively pursued (for example when collecting), the antennae are brought under the body and the insect adopts a very rapid crawl, with its ventral surface in contact with the substratum, making it very difficult to be seen. The wings of the adult males of each species are fully developed. However, the males seem incapable of sustained flight and merely fly intermittently for short distances. The females of Plate 4 Adult Male of E. lapponicus attacked by Entpmuthoragrylli.

Plate 5 Nymphal Instar of E. panzeri attacked by EntomaploraLrilli. -28-

E. pallidus also rossess well developed wings and can fly, but they seem to resort to this type of movement only very rarely. The reduced wings of the females of E. lapponicus and E. panzeri are obviously inadequate for flight. Females frequent the litter layers for most of their lives, and, unlike the nymphs and adult males, seldom climb. In the afternoon and particularly at dusk large numbers of males can be seen at the top of stems of grass or bracken. These remain stationary for several hours but drop to the ground if disturbed. It is possible that this may be a type of "basking.' behaviour. Perhaps one of the most interesting yet unexplored habits of these species is their obvious gregarious behaviour. Both in experimental cages and in the field aggregations of nymphs are often to be found. This habit appears to be particularly prevalent in the younger instars, and is apparently non-existent in the adults. Collecting records furnished adequate proof of the gregarious tendencies of these species. A feature common to each species is the production of cereal threads. When jarred suddenly or shaken from vegetation a long viscous thread is produced from the point of contact of each cercus with the object. The two threads coalesce and the weight of the suspended insect apparently causes the lengthening of the thread which sometimes reaches 25 ems. in length, but is usually much shorter. These threads are readily produced when the nymphs are anaethetised in a killing bottle, but are often observed when collecting. The production of cereal threads appears to prevent damage to the insect when it falls to the ground. However, Stock & O'Farrell (1954) noticed the production of similar threads in B.Aermanica and suggested that these threads may be produced by the young instars to facilitate the maintenance of the loose aggregations found in this species. Mating occurs during the day, but it seems that dusk is perhaps the more favourable time. The sequence of events involved in the courtship and mating of E. pallidus has been described and

-29- illustrated by Roth & Willis (1957). The process was observed in E. lapponicus and E. panzeri and found to be identical with their account, and has therefore not been described. Mating pairs are often seen on vertical structures, and may remain in copula for several hours. It is probable that these native species of Ectobius are omnivorous. Smith (1966) reported that adults of E. pallidus were capable of consuming four young Acyrthosiphon spartii (Koch) daily. Harz (1960) suggested that E. lapponicus and E. sylvestris probably feed exclusively on vegetable material under natural conditions, feeding on moss, algae, pollen and other detritus, and that moist vegetation is used for the intake of water. Observation confirmed that algal coverings were grazed upon by these species but detailed work in this field was not included.

(ii) Habitat Preference. a. List of Collecting Sites.

Numerous localities were examined as possible collecting sites. Yost specimens utilised during the course of this work were collected from the following areas:-

Locality. National Grid Reference.

E. nanzeri.

Thorney Hill Holms, Hants. • • • SU213002

Canford Heath, Nr. Wimbourne Minster, Dorset. • • • SZ024955 -30-

Studland Bay, Isle of Purbeck, Dorset. 004 SZ037850

Hengistbury Head, Mr. Christchurch, Hants. 000 SZ175908

E. pallidus.

Hengistbury Head, Nr. Christchurch, Hants. SZ175908

North Downs, Kent. N675623

Canford Heath, Nr, Wimbourne Minster, Dorset. ... SZ024955

Studland Bay, Isle of Purbeck, Dorset. SZ037850

E. lapnonicus.

Imperial College Field Station, Silwood Park, Berks. The most profitable collecting areas being:-

(i) South and West margin of Heyes Wood. 000 su946689 (ii) North Gravel. SU947688 (iii) South-east facing slope between Water Meadow and Pound Hill. SU940693

Crown Estate Plantations, Mr. Bracknell, Berks. ... su881664 su884662

Forestry Commission Plantations, Nr. Bracknell,

Berks. ••• su856662

This species was abundant in the New Forest area, but its exploitation was unnecessary since adequate material was available from local sites. -31-

b. Description of Habitats.

It was found that E. pallidus and E. panzeri often occupy the same habitat, but one species is usually present in greater numbers. On no occasion was E. lapponicus collected with either of the above species. The habitats described below are those which were found to be good collecting areas. It is possible that the species may occur in equal abundance in other habitats not suitable for the collection of material, or that in other parts of the country not visited, species may abound in a different type of situation.

E. panz.c27i. Two distinct types of habitat are preferred by E. panzeri. The sand dunes near the coast support large numbers of this species. The dunes rise from the open sand of the beach and are covered with tufts of marram grass. This is the dominant plant on the seaward side of the dunes, but only a few insects venture to this side, as the sand continually- drifts between the marram clumps. The landward side of the dunes is characterised by larger areas of marram grass, small patches of stunted ling, C. vulgaris, moss and lichen, and provides a more stable habitat. It is the centre of these tufts which this species occupies. The interiors of individual tufts are not dense, and it is the narrow channels which exist between the single stems that the cockroaches frequent in surprisingly large numbers (Plate 6). An equally typical situation for this species is one best described as heather scrub where sparse patches of stunted C. vulgaris are separated by large areas of bare sandy soil with an occasional patch of bracken. The leaf debris formed from the previous years growth of this bracken appears to be a particularly favourable niche, (Plate 7). The species was noticeably absent from land which had recently been burned.

E. pallidus. This was the most elusive of the three species. Small numbers

-32-

prviAT; 'lir) !,„,, 7;i;41 .1;1,...--/: l, ;lir i! Pr ii; ,InizgrAr ,401.1,.; iri. i g - i - . g , .11 ifi 1.;* , ..ICUr 6rdisiti, i Ni• p,,,glyyt 14,.),-,„ fopn , ,,, Jwv

- ' 41 44

P1 6 Studland Bay, Dorset.

Plate 7 Canford Heath, Dorset.

Typical sites for the collection of E. loanzeri.

-33- were found in the two types of habitat described for E. panzeri. It was noted that the sand hills further from the shore are frequented by this species. Typical chalk grasslands, invaded by numerous V. lantana saplings support moderate numbers of this species. However, the most frequented habitat is that where patches of bracken merge with a zone of grass. In this species and in E. lapyonicus it is the periphery of an area of bracken where the previous years growth has fallen on the grass boundary that yields abundant specimens of all stages. The interior of vast areas of bracken support only small numbers of specimens of E. pallidus or E. lapponicus. The former type of habitat can occur near to the sea, and it was such a site at Hengistbury Head that was the most successful collecting area for E. Lallidus (Plate 8 ).

ELLaRP2114-211P- This species commonly occurs in marginal zones of bracken as described above. The fringe of grass under trees of the Norway Spruce, P. abies, and Pinus sylvestris Linnaeus maintains a 'wealth of specimens. The favourite habitat for this species was found to be the narrow zones of herbage, at the margins of pine plantations, which slope down to shortly cut grass drives (Plate 9). Large tufts of Molinia caerulea (Linnaeus) Moench cover this area, with isolated plants of C. vulgaris and Rubus Linnaeus species. These tufts support incredible numbers of this species and provided a constant supply of material throughout the course of the work. This species remains confined to this restricted area and is not to be found in the litter under the trees. c. Overwintering Sites.

The ootheca is an overwintering stage common to the three species, while in E. lapponicus and E. pallidus the nymphal instars also overwinter. These nymphs do not move far during the adverse season but remain concealed in a sheltered niche. During this time -34-

Plate 8 Hengistbury Head, Hants. - Typical site for the collection of E. iaallidus.

Plate 9 Crown Estate Plantations, Berks. - Typical site for the collection of E.laroponicus. -35- nymphs of E. pallidus can be located in the core of large marram grass tufts, particularly on dunes some distance from the shore. Alternatively, nymphs are found deeply submerged in bracken litter. The nymphal instars of E. pallidus and also those of E. lapponicus seem to favour grass tussocks as permanent overwintering sites. Field observations have indicated that nymphs of E. lapponicus move actively into grass tussocks at the beginning of winter. Nymphs are found in numbers at the base of tussocks of Dactylis glomerata Linnaeus and Festuca rubra Linnaeus at this time, although few are to be found in similar situations during the summer months. Nymphs of E. lapponicus can also be found in pine or bracken litter several inches deep. The tufts of M. caerulea form ideal sites for the winter dormancy of this species. Tussocks which remain wet for much of the winter tend to be avoided. -36 -

SELECTED BIOLOGICAL TOPICS.

I. REPRODUCTIVE BIOLOGY.

This section of the investigation had two principal aims:- (i) To study various aspects of oviposition in mated and unmated females under natural and controlled conditions, and the hatching of the resultant eggs. (ii) To investigate the longevity of mated and unmated males and females under natural and controlled conditions.

Materials and Methods.

Final instar nymphs of E. lapponicus, E. pallidus and E. panzeri were collected from the field. The nymphs of each species were reared in separate glass tanks in an outdoor insectary. A few days before the adult moult, male and female nymphs were separated to ensure that mating did not occur before the beginning of the experiment. The tanks were fitted with a muslin lid and sleeve to facilitate the removal of the freshly emerged adults (Plate 1 ). Nymphs which moulted into adults were removed daily, and were therefore less than 24 hours old at the beginning of the experiment. The single pairs or individual cockroaches were kept in round plastic cages with food and water (Page 11; Fig. la). Observations were made daily at approximately the same time. The experiment was carried out in an outdoor insectary, to give the adults conditions comparable with those experienced in the field, and also in a controlled environment to facilitate comparisons between species and between mated and unmated females. The controlled conditions were a 16 hr. per 24 hr. light regime and a temperature of 20°C.; the humidity was uncontrolled. For the purposes of this investigation virgin males and -37-

females are those which were isolated for their entire adult life, whereas mated adults are those which were kept with an individual of the opposite sex from emergence until the death of one or other of the pair. Mating occurred reacJily in these pairs and was confirmed by keeping another series of pairs under the same conditions and establishing the presence of spermatozoa in the spermothecae of the females after two weeks, i.e. before the deposition of the first ootheca. It is therefore justifiable to conclude that a high percentage of the pairs did actually mate. Twenty replicates of isolated females and male-female pairs of each species were subjected to the controlled conditions, but under natural conditions ten replicates were used for E. lapponicus and E. panzeri, and twenty for E. pallidus.

(i) The following factors were considered in the study of oviposition and hatching in the three species:- a. The number of oothecae deposited. b. The preoviposition period. c. The interval between the deposition of oothecae. d. The number of eggs per ootheca. e. The weight of the ootheca. f. The size of the ootheca:- (i) Length (ii)Width g. The period of retention of the ootheca. h. The direction of rotation of the ootheca. i. The method of deposition of the ootheca. j. The percentage of oothecae which hatch and the number of nymphs which emerge from an ootheca.

The preoviposition period was taken as the time in days from adult emergence to the formation and deposition of the first ootheca. The interval between laying oothecae refers to the time in days from the deposition of one ootheca to that of the next. -38-

The number of eggs in each ootheca was determined by placing the ootheca laterally on a glass slide and counting the respiratory tubes in the keel; these are clearly visible when viewed under a microscope by transmitted light. Repeated dissections of oothecae had previously revealed that the number of respiratory tubes and the number of eggs were consistently equal. Oothecae were weighed and measured on the day of deposition. Each egg case was weighed to an accuracy of 0.001 mgs. using a Cahn Gram Electrobalance. The length and width Of the oothecae were measured to the nearest 0.1 mms. with a monocular compound microscope fitted with an eye-piece graticule. The length refers to the greatest length of the ootheca, and the width to the greatest width when the ootheca is resting on its basal surface with the keel uppermost. The period of retention of each ootheca was recorded in days. The first noticeable indication that ootheca formation has begun is the lowering of the distal part of the subgenital plate and the slight extrusion of the intersternal membrane. As the observations were made at the same time each day it was possible to record whether an ootheca had been carried for a period less than a known number of days. Females rotate their oothecae through 900 prior to deposition. The direction of this rotation, whether to the right or left was recorded. This necessitated frequent observations as rotation only preceded deposition by a few hours. Female cockroaches either carefully bury their oothecae.in the food material or merely deposit them on the base of the cage, without any attempt to cover them. These two methods were recorded as "concealed" or "not concealed". The oothecae were kept for the winter months in sterile 5.0 x 1.0 ems. glass tubes (Fig. 2), until hatching. The date of hatching and the number of nymphs emerging was recorded. Some pronymphs fail to free themselves from their exuviae although they escape from the ootheca, such nymphs were included in the number -39- which hatched.

(ii) The longevity of virgin and mated males and females was recorded as the time in days from the emergence of the adults until their death.

Standard errors of the means for each series of results were calculated, and differences between mated and unmated females compared by means of a 't' test.

Results and Discussion. a. The Number of Oothecae Deposited.

Virgin females of the three species deposit oothecae readily. These oothecae appear normal in gross structure. In each species mated females lay more oothecae than those which have not mated (Table la). This has also been recorded in P. americana (Roth & Willis, 1956). The maximum number of oothecae laid by E. yanzeri is 2, a few E. lapponicus females deposit 3 or 4, while in E. pallidus this number is quite common, and as many as 5 oothecae were recorded from one female (Table lb). In the outdoor insectary both mated and unmated females deposit more oothecae than at 20°C.; this is presumably related to the greater longevity under these conditions. Only in E. panzeri were some of the oothecae, laid by virgin females, obviously deformed at deposition. In this species although the difference between the total number of oothecae laid by mated and unmated females is not significant, the level of significance is much increased when only perfectly formed oothecae are considered. (Figures are given in parentheses in Table la). This may well be correlated with the fact that E. panzeri has a comparatively sliort adult season, th6 emergence period for both sexes being about 5 weeks, from the end of July to the end of -4o-

Table la The Number of Oothecae Deposited.

Number of Oothecae Species Deposited (Mean ± S.E.*) p Value Mated Unmated

20°C. E. lapponicus 1.8 ± 0.1 1.7 ± 0.2 0.7 - 0.6 E. pallidus 2.1 4 0.2 1.9 0.2 0.5 - 0.4 E. panzeri 1.4 t 0.1 1.1 0.2 0.2 - 0.1 (0.8 ± 0.1) (

Outdoors E. lapponicus 2.3 t 0.2 2.1 0.1 0.6 - 0.5 E. pallidus 2.7 t 0.2 2.4 ± 0.3 0.5 - 0.4 E. panzeri 1.7 ± 0.2 1.3 t 0.2 0.2 - 0.1 (1.1 t 0.2) (0.05 - 0.02)

= Standard Error.

August (Fig. 20, Page 118). During this period both sexes are equally abundant in the field. The non-significant p value obtained in E. lapponicus suggests that unmated females of this species will readily lay perfectly. formed oothecae; this may well occur in the field. E. lapponicus has a far longer adult season; adults being found in the field from the end of May until the beginning of October. The males emerge early in the season and reach an abundance peak at the end of June; the numbers then decline rapidly and adult males are only rarely found after the middle of July. However, adult females are still emerging at this time (Fig. 18, Page 114), and have little chance of being mated, since the longevity of an adult male is, on average, only twenty days in the field. Table lb The Proportions of Females Depositing 0 - 5 Oothecae.

Percentage of Females Depositing the Following Number of Oothecae Species Mated Unmated 0 1 2 3 0 1 2 3 4 5 i 20°C. IE. lapponicus 0 30 65 5 10 20 65 5 E. pallidus 0 25 45 30 ! 5 35 35 20 5 !E. panzeri 5 50 45 !20 50 30

Outdoors E. lapponicus 0 10 60 20 10 0 10 70 20 E. pallidus 0 10 30 45 15 5 10 45 25 10 1E. panzeri 0 30 70 110 50 40

It is probable then that the three species are facultatively parthenogenetic, though the need for this seldom arises in E. panzeri. E. pallidus appears to lie in an intermediate position with a shorter adult season than that of E. lapponicus, but considerably longer than that of E. panzeri (Fig. 19, Page 116). However, healthy oothecae are readily produced by virgin females of this species. The success of parthenogenesis in these species is revealed by the number of the oothecae which hatch successfully. This will be discussed in detail later. b. The Preoviposition Period.

Under field conditions and at 20°C. the preoviposition period in each of the three species is considerably longer in virgin than in mated females. This is highly significant at 20°C. in -42-

E. lapponicus and E. pallidus (p less than 0.01), and very highly significant in E. panzeri (p less than 0.001), as shown in Table 2. In the outside insectary this period is subject to considerable variation, and significant differences are not so marked. The preoviposition period was found to be significantly longer in virgin females of P. americana (Roth & Willis, 1956).

Table 2 The Preoviposition Period.

Preoviposition Period in Species days (Mean t S.E.*) Value Mated Unmated

20°C. E. lapponicus 17.1 ± 0.5 21.0 ±1.0 0.01 - 0.001 E. pallidus 20.9 ±0.7 23.5 0.6 0.01 - 0.001 E. panzeri 18..5 ± 0.4 22.7 0.8 <0.001

Outdoors E. lapponicus 24.3 ± 1.4 26.2 1.2 0.4 - 0.3 E. pallidus 21.8 ± 0.6 23.8 0.8 0.05 - 0.02 E. panzeri 21.8 ± 0.8 24.6 4- 1.9 0.05 - 0.02

= Standard Error.

The absence of a male definitely delays ootheca formation, although formation will begin a few days after the normal time for this period if mating has still not taken place. During the preoviposition period many females which had males present were observed in the act of mating, but it was not possible to note this in every replicate. In fed Leuco;haea maderae,(Fabricius) and Nauphoeta cinerea -43-

(Olivier) females there appears to be a synergistic action of nutrition and mating in controlling the rate of oocyte development. Mating and feeding stimuli are both usually required for activating the corpora allata to their fullest extent so that the oocytes mature at their maximum rate (Roth, 1964). A preliminary experiment showed that the quality of food has a marked effect on oocyte development in E. lapponicus. In the virgin Ectobius females the feeding stimulus is present, but the mating stimulus is absent, and it is probably this which causes a reduction in the rate of oocyte maturation and accounts for the delay in oviposition compared with mated females where both stimuli are present. c. The Interval between the Deposition of Oothecae.

Virgin females require a longer period between the deposition of successive oothecae than mated females. This is especially significant in E. pallidus and E. panzeri, both at 20°C. and under outdoor conditions, but is less obvious in E. lapponicus (Table 3). The interval is reduced when the females are kept constantly at 20°C. This period is shortest in E. pallidus, since more oothecae are deposited by each female without an accompanying increase in longevity. Some females were observed to mate for a second time after the deposition of,the first ootheca; however, this would appear to be unnecessary, since the spermothecae of females were found to be still full of live spermatozoa after the deposition of the first or second ootheca. Generally the interval increases after the second ootheca has been deposited, but records of subsequent depositions are few, and variation between females considerable. -44-

Table 3 The Interval between the Deposition of Oothecae.

Interval between Deposition of Species Oothecae in Days (Mean ± S.E.*), p Value Mated Unmated

20°C. E. lapponicus 12.7 ± 0.5 15.1 J.- 0.6 0.01 - 0.001 E. pallidus 10.5 ± 0.5 14.7 ± 0.9 <-0.001 E. panzeri 11.7 ± 0.5 20.0 -1 1.0 <0.001

Outdoors E. lapponicus 16.9 ± 0.8 18.8 ± 1.0 0.2 - 0.1 E. pallidus 11.7 ± 0.5 15.5 ± 0.8 ,,0.001 E. panzeri 15.6 0.8 23.8 t 1.5 0.01 - 0.001

= Standard Error.

d. The Number of Eggs per Ootheca.

The number of eggs in an ootheca varies according to the species. E. panzeri oothecae contain fewer eggs than E. pallidus or E. lapponicus. It has been recorded that the maximum number of eggs deposited at one time is dependent on the number of ovarioles comprising the ovaries (Roth & Willis, 1954a). Dissections of females of the three species has indicated that the ovaries are composed of two groups of eight ovarioles, although occasionally one or more ovarioles may be deformed. It is interesting that a high percentage of oothecae in E. lapponicus and E. „pallidus enclosed more than 16 eggs, and a small percentage contained as many as 20 eggs. This suggests -45-

either that the number of ovarioles is more variable within the females of a species than the dissections indicated, or that some ovarioles may contribute more than one egg to an ootheca. The latter seems more feasible, since the number of eggs hatching from an ootheca seldom exceeds 16 or 18, the remainder of the eggs being either undeveloped or partially developed. It is possible that these were the second eggs contributed by some of the ovarioles. The unhatched eggs are usually positioned at the anterior end of the ootheca, i.e. the end which was formed last. Roth (1968b) found that the two posterior eggs in each ovariole of E. lapponicus contain yolk at the time of ovulation. It is possible that this is true of E. pallidus also and that this enables a double contribution of eggs by a small but variable number of ovarioles. Further support for this idea is found in the fact that the oothecae which any one female lays do not contain the same number of eggs, and this variation is quite marked. There is no obvious tendency for the number of eggs in successive depositions to increase or decrease. E. panzeri very rarely deposits more than 16 eggs in an ootheca. Brown (1952) recorded from a total of 11 oothecae a mean of 12 eggs per ootheca with a range from 9 to 13. However, his observations were confined to specimens collected at the end of the season. The present work has shown that the number of eggs per ootheca declines markedly at this time. The proportion of oothecae with less than the normal number of eggs is greatest in E. panzeri. A possible explanation for this may be found in the work of Gier (1947) on P. americana, in which he showed that those oothecae with less than the normal number of eggs were laid by females with either disease or some abnormality of one or more ovarioles. This may be the explanation for the fact that E. panzeri often lays fewer eggs. The oothecae laid by mated females contain more eggs than those of females which have been isolated from males for their adult life. This difference is significant in each species (Table 4).

-46 -

Table 4 The Number of Eggs per Ootheca.

Number of Eggs per Ootheca 4 Species (Mean "" S.E.*) p Mated Unmated Value

2000. E. lapponicus 17.7 ± 0.2 16.8 ± 0.3 0.02 - 0.01 E. pallidus 17.2 0.2 16.6 ± 0.2 0.05 - 0.02 E. panzeri 14.6 -I- 0.2 13.5 ± 0.4 0.05 - 0.02

Outdoors E. lapponicus 17.5 4- 0.3 16.2 0.3 0.01 - 0.001 E. pallidus 18.0 ± 0.2 17.2 ± 0.3 0.05 - 0.02 E. panzeri 14.9 "1" 0.3 13.8 ± 0.5 0.1 - 0.05

= Standard Error.

It should be pointed out that the number of eggs in an ootheca laid by a mated or unmated female differs significantly (p less than 0.001) in E. panzeri from both E. pallidus and E. lapponicus; the latter species, however, do not differ significantly. e. The Weight of the Ootheca.

The weight of oothecae laid by mated and unmated females differs significantly. This, however, would be expected, since oothecae laid by virgin females contain fewer eggs. The most interesting feature emerging from this investigation is that although the oothecae of E. lapponicus and E. pallidus contain a similar number of eggs, the oothecae of the latter species weigh significantly less (p less than 0.001). Although nothing is known of the weight of the individual eggs of the three species, a comparison of the size of four characters of the first instar nymphs (Table 5) reveals that the nymphs of E. pallidus are much smaller than those of E. lapponicus. It is this which accounts for the difference in the weight of the ootheca. An extra instar in the life cycle of E. pallidus compensates for the smaller first instar; since the adults of the two species are similar in size.

Table 5 To Compare the Size of 1st. Instar Nymphs of the Three Species.

Pronotum Hind Leg

Species Length Breadth Length i Length in in of Femur 1 of Tibia mms. mms. in mms. in mms. L-.

Mole. E. lapponicus o.63 1.11 0.71 0.75 E. pallidus 0.55 0.95 0.59 0.63 E. panzeri 0.57 0.89 0.63 0.67

Female. E. lapponicus 0.62 0.72 0.77 E. pallidus 0.55 0.95 0.60 0.64 E. panzeri 0.58 0.89 0.63 0.69

Each entry represents the mean of 10 replicates.

The highly significant difference between the weights of oothecae deposited by mated and virgin females of E. panzeri (Table 6) is apparently due to the fact that the keel of some -48-

Table 6 The Weight of the Oothecae.

Weight of Oothecae in mgs. (Mean S.E.*) Species p Value Mated Unmated

20°C. E. lapponicus 6.25 I 0.11 6.00 ± 0.14 0.2 - 0.1 E. pallidus 447 1- 0.07 4.25 ± 0.08 0.05 - 0.02 E. panzeri 3.61 ± 0.08 3.08 ± 0.09 <0.001

Outdoors E. lapponicus 6.07 0.12 5.62 -1 0.13 0.02 - 0.01 E. pallidus 4.70 ± o.o8 4.38 -±0.10 0.02 - 0.01 E. panzeri 3.66 ± 0.08 3.28 I 0.08 0.01 - 0.001

Standard Error. oothecae deposited by virgin females is often deformed. This causes water loss in the oothecae as they are often retained for several days. Some such oothecae have a shrivelled appearance prior to deposition and subsequently fail to hatch. In general there is no marked decline in the weight of oothecae deposited later in the life of the female. f. The Size of the Ootheca:- (i) Length.

There is a considerable difference in the shape of oothecae laid by mated and virgin females. Although fewer eggs are deposited in oothecae laid by unmated females of the three species, these oothecae are significantly longer than those deposited by mated females (Table 7). The keel of the latter is appreciably different, the respiratory tubes being closer together and the denticles more distinct.

Table 7 The Size of the Oothecae (i) Length.

Length of Oothecae in mms. Species (Mean ± S.E.*) p Value Mated Unmated i

20°C. ...._E.3.46I 0.04 3.62 ± 0.05 0.01 - 0.001 E. pallidus 2.81 ± 0.04 3.05 ± 0.04 <0.001 E. panzeri 2.57 ± 0.03 2.76 ± 0.06 0.02 - 0.01

Outdoors E. lapponicus 3.43 ± 0.06 3.58 ± 0.04 0.05 - 0.02 E. pallidus 2.88 ± 0.03 3.02 ± 0.04 0.01 - 0.001 E. panzeri 2.60 + 0.03 2.77 ± 0.06 0.02 - 0.01

= Standard Error.

Oothecae deposited by females of E. lapponicus are more elongate than those of E. panzeri or E. pallidus, although the number of eggs per ootheca is similar to the latter species. Oothecae resembling the extra long oothecae deposited by virgin females of E. lapponicus can occasionally be found in the field, one such ootheca kept until hatching, produced only female nymphs suggesting the occurrence of.partheno2enesis in the field. The shape of the ootheca is constant for a species, whether it has been laid outdoors or at 200C. -50-

(ii) Width.

The width of oothecae deposited by mated and virgin females differs significantly within each species studied (p is less than 0.001). The width of oothecae laid by mated females is always greater (Table 8).

Table 8 The size of the Oothecae (ii) Width.

Width of Oothecae in rams. Species (Mean t S.E.*) p Mated Unmated Value

20°C. E. lapponicus 1.61 I 0.01 1.52 + 0.02 -0.001 E. pallidus 1.49 ± 0.01 1.38 ± 0.01 :::0.001 E. panzeri 1.42 1- 0.01 1.30 t 0.01 '7.0.001

Outdoors E. lapponicus 1.57 4- 0.01 1.49 t 0.02 <0.001 E. pallidus 1.50 t 0.01 1.43 ± 0.01 .0.001 E. panzeri 1.43 t 0.01 1.29 t 0.01 '! ...0.001

Standard Error.

The difference in width must in some way be associated with mating during the preoviposition period. Mating is known to activate the corpora allata which produce the gonadotropic hormone and this promotes the deposition of yolk in the oocytes (Roth & Stay, 1959). Since the oocyte maturation rate is increased by - mating, it is possible that fertilised eggs contain more yolk at -51- the time of ootheca formation. This would increase the width of each egg and hence that of the ootheca. However, the difference in yolk content cannot be very marked since the weights of oothecae laid by mated and virgin females do not differ markedly. It is perhaps more probable that the lack of mating or the absence of spermatozoa in the spermothecae in some way changes the fashion in which the ootheca is moulded. This would result in a differently shaped ootheca; but these ideas require further investigation before conclusive remarks can be made. g. The Period of Retention of the Ootheca.

Various workers have classified species of cockroaches with reference to the time for which the female retains the ootheca and the type of ootheca formed. Shelford (1906) uses the following classification:- (i) Oviparous species. Eggs are enclosed in a chitinous ootheca and are carried by the female for a short time only. (ii) Ovoviviparous species. Eggs may either be enclosed in a semichitinous or transparent membranous capsule and are carried by the female protruding from her abdomen for most of the embryonic period. (iii) Viviparous species. Eggs are enclosed in a chitinous ootheca or transparent membrane which is retained in the brood pouch of the mother.

E. lapponicus, E..pallidus and E. panzeri are therefore oviparous species. However, Roth & Willis (1955c) subdivide oviparous cockroaches into three oviposition types:- Oviposition Type 1. Ootheca is extruded and carried by the female for only a short time, then deposited and abandoned. Oviposition Type 2. Ootheca is extruded and carried for a longer period than type 1, but eventually dropped long before the eggs hatch. -52-

Oviposition Type 3. Ootheca is extruded and carried externally by the female until, or shortly before the eggs hatch.

E. panzeri and E. pallidus are given as the only examples of type 2. E. panzeri was recorded to retain its ootheca for 16 days (Brown, 1952), and E. pallidus also for 16 days (personal communication from P. Rolander in Roth & Willis, 1955c). These observations are contrary to the results obtained in this study. Generally, E. pallidus females retain their oothecae for less than one day at 20°C. or outdoors, which is in agreement with a later statement by Roth & Willis (1957). It is interesting that virgin females of E. lapponicus and E. pallidus deposit their oothecae in a shorter time than mated females (Table 9).

Table 9 The Period of Retention of the Ootheca.

Mean Period of Retention of Oothecae in Days Species p Value Mated Unmated

20°C. E. lapponicus 1.9 1.6 0.1 - 0.05 E. pallidus 1.1 1.0 0.5 - 0.6 E. panzeri 2.5 4.1 <0.001

Outdoors E. lapponicus 2.0 1.8 0.7 - 0.6 E. pallidus 1.3 1.2 0.4 - 0.3 E. panzeri 2.6 4.3 0.01 - 0.001 -53-

This is almost certainly due to the presence in the cage of a male, which disturbs the female and prevents her from settling to deposit her ootheca. A female may carry a completely formed ootheca for many hours attempting to deposit it, but if at this stage she is separated from the male she will release it without delay. After the death of the male (which is usually prior to the deposition of the second ootheca), the period of retention is reduced in most females. E. panzeri females retain their oothecae for a greater period than the other two species; virgin females retaining their oothecae for a significantly longer period than mated females. Oothecae produced by virgin females are often deformed and females seem unable to release them, and frequently die while still in possession of an ootheca. This also happens occasionally with mated females and probably explains the records of prolonged retention in this species (Brown, 1952), coupled with the fact that these observations were restricted to the end of the season when this behaviour is prevalent. A similar situation also occurs to a lesser extent in E. lapponicus and E. pallidus. From these results there seems little justification for the oviposition type 2 advocated by Roth & Willis (1955c), unless other species are subsequently found to fall into this category. h. The Direction of Rotation of the Ootheca.

All cockroach species extrude their oothecae with the keel orientated dorsally. Some retain the ootheca in this position until deposition, while others rotate their oothecae through 900 prior to deposition (Roth, 1967a). The three British species of the genus Ectobius all rotate their oothecae before deposition (Plate 10). Oothecae laid by unmated females are rotated to the right (dextrorotatory) and left (levorotatory) in almost equal numbers, but the proportion rotated to the right is always slightly larger. a. E. lalppnicus b. E. nallidus

c. E. panzeri

Plate 10 Rotation of the ootheca prior to deposition.

-55-

(Table 10). In mated females rotation occurs predominately in one direction. In E. lapponicus and E. pallidus rotation occurs mainly to the right. Roth & Willis (1957) recorded that 19 oothecae deposited by females of E. pallidus were all rotated to the right. The present work indicates that a number (15 - 24%) are also rotated to the left prior to deposition. Harz (1960) observed that ootheca rotation in E. lapponicus is dextrorotatory, but he does not state the number of observations involved. In E. panzeri, however, the majority of oothecae were rotated to the left. Details of the total number and percentage of oothecae rotated in each direction by mated and virgin females are given in Table 10.

Table 10 The Direction of Rotation of the Ootheca.

Direction of Rotation of the Ootheca Species (5' in parenthesis) hated Unmated —1 Right Left Right Left

20°C. E. luponicus 32 3 18 15 (91.4) ( 8.6) (54.5) (45.5) E. pallidus 31 10 20 17 (75.6) (24.4) (54.1) (45.9) E. panzeri 5 23 14 8 (17.9) (82.1) (63.6) (36.4)

Outdoors E. lapponicus 18 5 12 9 (78.3) (21.7) (57.1) (42.9) E. pallidus 45 8 26 22 (84.9) (15.1) (54.2) (45.8) E. panzeri 2 15 7 6 (11.8) i (88.2) (53.8) (46.2) -56-

Oothecae which had a deformed keel were not always rotated through a full 90°, but remained merely inclined to the right or left until deposition or the death of the female. This was more frequent in unmated females of E. panzeri. It would appear that either the act of copulation or the presence of spermatozoa in the spermothecae as a result of mating seemed to effect the direction of rotation of the ootheca. An attempt to correlate the direction of rotation of the ootheca with the asymmetry of the male genitalia may well prove interesting. In E. lapponicus and E. pallidus the left side of the subgenital plate is elongated and bears a single terminal style, and the left phallomere forms a long retractable hook, whereas in E. panzeri it is the right side of the subgenital plate which is extended and the right phallomere which constitutes the retractable hook. The male genitalia may disarrange the ovipositor valves during mating so that rotation in one direction is facilitated. However, much further work on the mechanism involved in the rotation of cockroach oothecae is needed. Roth (1967a) concluded that the direction of rotation of the ootheca is an inherited character in B. germanica. The results obtained for Ectobius in the present work contradict this hypothesis, since the direction of rotation is not constant within a species or within the oothecae deposited by a single female. Brown (1952) suggested that E. panzeri females rotate their oothecae towards the side containing the most eggs. However, no evidence to support this idea was found in the three species of Ectobius or in B. germanica (Roth,1967a). According to Roth (1967a) the position in which the ootheca is carried at the time of deposition is significant taxonomically, and also played an important role in the evolution of oviparity and viviparity. Roth, using the classification of McKittrick (1964), considers that the rotation found in the Blattellidae (Blattellinae, Ectobiinae and Nyctiborinae) to be an advanced type, since the anterior eggs of the ootheca are in close contact with the tissues

-57-

of the female vestibulum; a prerequisite for the evolution of oviparity and viviparity.

i. The Method of Deposition of the Ootheca.

Mated and unmated females are similar in their oviposition behaviour. E. lapponicus and E. panzeri merely deposit their oothecae on the surface of the substrate or in a suitable crevice and seldom attempt to bury or conceal them. In comparison, E. pallidus usually buries its ootheca after deposition (Table 11), the complete behavioural sequence often taking several hours.

Table 11 The Method of Deposition of the Ootheca.

Method of Deposition of Ootheca (% in parenthesis)

Species Mated Unmated

Not Not Concealed Concealed Concealed iConcealed

20°C. E. lapponicus 11 24 7 26 (31.4) (68.6) (21.2) i (78.8) E. pallidus 4o 35 2 (97.6 ) ( 2.4) (94.6) ( 5.4) E. panzeri 6 22 0 1 22 (21.4) (78.6) ( 0.0) (100.0)

Outdoors E. lapponicus 6 17 6 15 (26.1) (73.9) (28.6) 1 (71.4) E. pallidus 48 5 44 4 (90.6) ( 9.4) (91.7) 8.3) 1 ( E. panzeri 3 14 o 1 13 (17.6) (82.4) ( 0.0) I(100.0) -58-

The female uses her fore legs to dig a cavity and the hind legs to remove the excavated material. She deposits the ootheca in the hole which is then filled again principally by the fore legs. Digging with the legs evolved from pushing excavated debris away from the hole edge in primitive cockroaches which carve a hole for the ootheca with their mouthparts, and has become highly evolved in the Blattellidae (McKittrick, 1964). This digging is well exhibited in E. pallidus, but appears to have become secondarily lost in E. lapponicus and E. panzeri. A few females of E. panzeri stuck their oothecae to the side of the container. All the females which died while carrying their oothecae, had completed ootheca formation at the time of death, and this apparently does not prevent the development of the eggs, which hatch the following summer. Many such oothecae were easily detached from the female, and this presumably happens in the field during the winter.

j. The Percentage of Oothecae which Hatch and the Number of Nymphs which Emerge from an Ootheca.

In E. panzeri only oothecae deposited by mated females hatched, whereas in the other two species a relatively small proportion of the oothecae laid by unmated females also hatched (Table 12). The percentage of oothecae which hatched was higher when the females were kept under outdoor conditions. A high proportion of eggs in oothecae laid by mated females of the three species developed and hatched (Table 13). Those eggs which failed to hatch were frequently situated at the ends of the oothecae and were often undeveloped. It is interesting that in both E. lapponicus and E. pallidus more than 16 eggs may develop and hatch, indicating that the additional eggs discussed on Page 44. must be viable and contain enough yolk for complete development. The mean number of nymphs recorded from oothecae of E. pallidus (Roth & Willis, 1957) and E. panzeri (Brown, 1952) is less than -59-

Table 12 The Percentage of Oothecae which Hatch.

20°C Outdoors

Species % of Oothecae % of Oothecae which Hatch which Hatch

Mated Unmated Mated Unmated

E. lapponicus 71.4 27.3 78.3 33.3

E. pallidus 68.3 16.2 77.4 16.7

E. panzeri 60.7 0.0 76..5 0.0

that recorded in this study. However, in the case of E. panzeri these previous observations were restricted to oothecae deposited late in the season when the number of eggs enclosed is lower, whilst in E. pallidus hatching was induced artificially. The percentage of nymphs which emerged from a hatched ootheca deposited by a virgin female was far less in E. lapponicus and E. pallidus, ranging from only 6 to.13. Oothecae laid by unmated females produced only female nymphs; parthenogenesis in Ectobius is therefore thelyotokous. However, few nymphs developed beyond the second instar. The oothecae of each species hatch over a relatively short period of time. There is no difference in the time of hatching of oothecae deposited by mated or virgin females, or between oothecae laid by females which were kept under outdoor or controlled conditions (Table 14). This restricted hatching period is suggestive of a diapause in the oothecae of the three species, which is discussed in detail later (Pages 71 - 86). -6o-

Table 13 Number and Percentage of the Eggs in an Ootheca which Develop and Hatch.

% of eggs Number of Eggs per Range in Nos. Species which hatch Ootheca which hatch hatching per (Mean ± S.E,*) Ootheca

MATED 20°C. 1 E. lapponicus 85.4 16.4 t 0.4 12 - 19 1 E. pallidus 93.3 15.9 t 0.2 14 - 18 E. panzeri 92.7 13.3 ± 0.4 10 - 16 Outdoors E. lapponicus 94.3 16.7 ± 0.4 13 - 18 E. pallidus 93.6 17.0 ± 0.2 14 - 19 E. panzeri 90.1 13.3 ± 0.6 10 - 16

UNMATED 20°C. E. lapponicus 64.7 11.0 ± 1.1 6 - 15 E: pallidus 50.5 8.5 t 0.8 7 - 12 E. panzeri 0.0 0.0 - Outdoors E. lapponicus 59.3 9.6 I 1.0 6 - 12 E. pallidus 50.7 9.3 1.o 6 - 13 E. panzeri 0.0 0.0 -

= Standard Error.

It would appear that both E. lapponicus and E. pallidus are facultatively parthenogenetic, but that this mode of reproduction results in a marked decline in the fertility of the species. It is unlikely that the British species of Ectobius reproduce parthenogenetically for more than one generation, since the viability of the nymphs seems to be reduced. However, the ability of these species to reproduce by this method is of adaptive value since the last females of the season may emerge after the males Table 14 Hatching Period of Oothecae (Summer 1963).

....P...... , Species Mated i Unmated

2pyc, i E. lapponicus 14th - 25th June 1 16th - 28th June E. pallidus 21st - 29th June 23rd - 28th June E. panzeri 14th - 25th June -

Outdoors E. lapponicus 12th - 24th June 14th - 26th June E...p.allidus 20th - 30th June 21st - 30th June . panzeri 16th - 24th June -

have declined in numbers (Figs. 18&19, Pages 114&116). There are few conclusive records of parthenogenesis in cockroaches, and these are confined to the cosmopolitan, domestic species. Roth & Willis (1956) found considerable evidence of thelyotokous parthenogenesis in P. americana, though as in E. lapponicus and E. pallidus the number of unfertilised eggs which hatch was much reduced. Parthenogenesis has also been recorded in B. orientalis and N. cinerea to a lesser extent, whilst in B. $ermanica and Supella supellectilium (Serville) (= S. longipalpa (Fabricius)) a few unfertilised eggs developed but did not hatch (Roth & Willis, 1956). Pycnoscelus surinamensis (Linnaeus), however, exists in two strains, a bisexual strain from Hawaii and a parthenogenetic strain from Florida (Roth & Willis, 1961). Parthenogenesis appears tc, be a succesful mode of reproduction in P. americana since it was carried through two filial generations. However, as in the British species of Ectobius the mortality among parthenogenetically produced nymphs, was much higher and some nymphs were visibly deformed on hatching (Roth & Willis, 1956). -62-

It should be stressed that any estimate of the extent of parthenogenesis in the oothecae of cockroaches is biased, since an ootheca will only hatch if a high percentage of the enclosed eggs are fully developed. The oothecae laid by unmated females often contain only a few fully developed nymphs, and therefore fail to hatch. Oothecae of the three species deposited by unmated females were found to contain developing nymphs, although these were seldom (or never in E. panzeri) present in sufficient numbers to result in a successful emergence from the ootheca.

(ii) The Longevity of Virgin and Mated Males and Females.

The adults of these three species of Ectobius are relatively short lived; since the longevity of the common domestic species often extends to several months. In each species the females lived longer than the males. The longevity of both sexes was reduced when the adults were kept at 20°C. Much variation in longevity occurred when the adults were subjected to outdoor conditions. Virgin females lived longer than mated females (Table 15a), though the number of oothecae produced was lower (Table la). A similar increase in the longevity of virgin females has been recorded in P. amcricana (Griffiths & Tauber, 1942a)and in this species and B. orientalis by Roth & Willis (1956). The longevity of E. pallidus cited by Roth & Willis (1957) is less than that found in this investigation. They do not, however, state the exact temperature at which the adults were kept, which may explain the discrepancy. The lorigovij;y of aduzl; ua1ea was similar whether or not mating occurred ( Table 1I-Jb). -63-

Table 17a The Longevit of Adult Females.

4111.111.1.0* Duration of Adult Stage in Days Species (Mean ± S.E.*)

Mated Unmated Value

2000. E. lapponicus 36.6 ± 1.5 45.5 ± 1.6 <0.001 E. pallidus 42.4 ± 2.7 48.1 ± 2.9 0.2 - 0.1 E. panzeri 30.5 ± 2.2 39.5 ± 3.6 0.05 - 0.02

Outdoors E. lapponicus 59.1 ± 4.9 64.9 4.o 0.4 - 0.3 E. pallidus 54.5 ± 2.8 60.3 ± 2.8 0.2 - 0.1 E.Lanzeri. 45.3 ± 2.6 48.5 ± 2.9 0.5 - 0.4

Table 15b The Longevity of Adult Males.

Duration of Adult Stage in Days Species (Mean t S.E.*) Mated j Unmated

20°C. E. lapponicus 22.5 ± 1.0 22.0 ± 1.0 E. pallidus 29.9 ± 1.8 22.7 ± 1.2 E. panzeri 23.2 ± 1.8 24.0 ± 1.2

Outdoors E. lapponicus 28.8 ± 1.3 27.3 ± 1.6 E. pallidus 30.4 ± 1.4 37.0 ± 2.7 E. panzeri 26.7 ± 1.5 26.2 ± 2.4

= Standard Error. -64-

II. THE EGG STAGE.

(i) Description of the Oothecae.

The oothecae produced by the British species of the genus Ectobius are essentially oval in shape and have rigid walls. The base of the ootheca is short and the lateral margins spread out from this to the keel-bearing surface. Oothecae of E. lapponicus are pale brown at deposition but soon assume a darker reddish brown colour. Those of E. pallidus and E. panzeri are a very dark brown when laid and do not deepen in colour with age. The dorsal surface of the ootheca is convex and the keel is continuous over this surface. The two ends of the ootheca differ in profile in each species. The distal end (i.e. the end formed first) is more rounded and curves gradually to the base of the ootheca. The keel extends further ventrally at this end (Fig. 4). The proximal end terminates abruptly and this prevents the keel reaching beyond the dorsal surface. This end is often characterised further by bearing small irregular flanges, which are merely cuticular extensions of the oothecal wall (Fig. 5). A darkened line which thickens ventrally continues from both ends of the keel; this demarcates the end of the seam, and also the limits of the fissure caused by the hatching of the nymphs. Each ootheca contains two longitudinal rows of eggs. The number and position of the enclosed eggs are indicated externally by lines extending from beneath the keel to the base of the ootheca. Each line demarcates the lateral extent of an individual egg chamber and is accompanied by a shallow invagination. In E. lapponicus these limitations appear as darkened lines which are easily seen, whilst in E. panzeri and E. pallidus only the slight indentations are discernible. When the ootheca is viewed laterally each line has the shape of a letter "Y", with very short arms and a long straight stem which extends to the ventral surface. The

a. E. lapponicus

keel egg termination

egg chamber

b. E. pallidus c. E. panzeri

cuticular ridge

1 mm. 1 1 FIG 4 Lateral View of Oothecae of the British Species of Ectobius.

-66-

terminal egg chamber end cuticular seam flange

ventral lmm. groove

a. Proximal End - E. lapponicus b. Distal End - E lapponicus

hollow cell upper lamina

"white body" lumen lower lamina chorion

0.1mm

c. Transverse Section of Keel - E. pallidus

FIG 5 -67- space between two such stems indicates a single egg space. The two rows of eggs are arranged alternately, such that the space between two eggs on one side is partially filled by an egg from the other. The space between the arms of the "Y" - shaped line is occupied by the end of an egg from the opposite row (Fig. 4). It is the extension of each egg beyond the median keel which is responsible for the characteristic "Y" - shaped zone beneath the keel, found in the three species. This area enclosed by the arms of the "Y" is considerably larger and more marked in E. lapponicus than the other two species. Each end of the ootheca is only occupied by a single egg space (Figs. 5a & b). The wall of the ootheca in E. lapponicus and E. panzeri is devoid of any ridges or striations, though that of E. pallidus has a pronounced ridge parallel but ventral to the keel on each side of the capsule. Other prominent ridges are often visible but these usually terminate without completely crossing the ootheca (Fig. 4b). The walls of the oothecae of the three species are rigid; all are impregnated with crystals of calcium oxalate. Stay, King & Roth (1960) by microscopic and X - ray diffraction methods surveyed 27 species of cockroaches for the presence of calcium oxalate crystals embedded in the walls of their oothecae. Non-quadratic crystals of the monohydrate salt were found in E. pallidus, all other species previously examined had the salt present in its dihydrate form. The presence of calcium oxalate was also confirmed in the oothecae of E. panzeri and stated to be present in E. lapponicus also (Roth, 1968a). Stay et al. (1960) discuss the possible functions of these crystals, the most plausible being that the oxalate crystals replace the chitin, which is absent from the oothecae of P. americana (Campbell, 1929), and probably those of other species, and serve a structural and protective role. The slightly concave base of the ootheca is about one half the length of the keel. This surface bears the terminations of the eggs which are clearly visible in E. lapponicus but much less so in E. pallidus and E. panzeri. This surface also displays the -68-- internal arrangement of the eggs. Characteristic of the three species is the deep ventral groove which persists until a short while before hatching. This groove is formed by each side of the oothecal wall being moulded into a ridge ventrally, the median ventral part of the ootheca being completely invaginated (Figs. 5a & b). A short while before hatching the ootheca increases in width, and the ventral surface becomes rounded leaving no trace of the groove. Roth (1968a) suggests that this groove allows the stretching of the oothecal wall to occur when water is taken up. Dissections of the ootheca during embryogenesis reveal that all the embryos are orientated with their ventral surface facing inwards and their heads towards the keel of the ootheca. Each egg is surrounded by its own chorion. This together with the chorion of the next egg forms a doubled wall partition between the two eggs. Similarly a double layer exists between the two rows of eggs, although only a single layer separates the egg from the outer capsule wall and dorsally from the keel. These partitions remain in the ootheca after hatching, and become dry and brittle. The size (i.e. the length and width) of the oothecae and the number of eggs enclosed varies between species and also between oothecae laid by mated and virgin females. Details of these criteria can be seen by reference to Tables 7, 8 & 4 respectively.

The Keel. (Fig. 5c).

The oothecal walls are thin but thicken where they unite dorsally to form the keel. Each dorsal edge divides into an upper and lower lamina. The lower or inner lamina of each side extends medially to form the base of the keel. The upper laminae curve upwards and meet also in the mid-line, thereby enclosing a space, the lumen, of the keel. The lumen so formed is continuous along each side of the keel, but is reduced apically by an inflexion of the upper laminae where they meet. The lumen is divided medially by a row of "white bodies" the name given by Lawson (1951) to the -69- white structures found in the keel of the oothecae of B. germanica and P. pensylvanica. These are clearly visible when the keel is viewed in strong transmitted light, as a white band situated in the middle of the keel. The "white bodies" meet the inflexion of the upper layers of the keel and thus the lumen is reduced to two circular canals. These vacuolated structures are an extension of the chorion of the individual eggs (Wigglesworth & Beament, 1950). The top of the keel is produced into a regular series of rounded elevations or denticles. Each denticle slopes more gradually towards the rounded end of the ootheca. The number of denticles corresponds exactly to the number of eggs enclosed in the ootheca in each species. The denticles are often more crowded at the rounded end of the ootheca. Each denticle contains a single pore which opens on the surface of the denticle and passes into the lumen of the keel, via an empty cavity or cell. All the pores curve in the same direction and do not open on the apex of the denticle but on its lateral margin, directed away from the rounded end. These openings act as respiratory channels which convey air to the "white bodies" and hence to the developing eggs (Wigglesworth & Beament, 1950). Roth (1968a) suggests that the primitive ootheca had a raised keel which was devoid of respiratory openings. Respiratory tubules evolved, but were at first present in greater numbers than the enclosed eggs. A reduction in their number followed until the number of tubules corresponded to the number of eggs. This state is well exhibited in the three species of Ectobius under study. McKittrick (1961+) considers the reduction in the number of respiratory channels as a refinement restricting the loss of water.

Oothecae of the three species undergo gross external changes during the few weeks before hatching, and it should be pointed out that the above descriptions and the subsequent key are only valid for oothecae which are not near to hatching.

-70-

Key to the Oothecae of the British Species of Ectobius.

1. Dark brown, keel dark. Segmentation present but not obvious. Length 2.5 - 3.5 mms.

Reddish brown, keel pale. Segmentation distinct.

Length 3.2 - 4.1 mms. ••• ...... E. lapponicus

2. Cuticular ridge parallel to keel, at base of "Y" shaped zone, (Fig. kb). Often other cuticular ridges present. Keel with 16 - 20 teeth or denticles. 0441 ••• •• . E. pallidus

No cuticular ridges present. Keel with 12 - 15 teeth or denticles. (Fig. 4c). ••• • • • • • • E. panzeri (ii) Determination of Diapause in the Oothecae of E. lapponicus.

Materials and Methods.

Two large cultures of field collected adult females were reared in glass tanks (Plate 1 ) as described on Page 12. One tank contained the first females to become adult and the other one females which attained maturity much later in the season. The oothecae deposited by these females were collected daily. A total of 175 oothecae were deposited by the former group of females from 1st - 17th July and these were divided into seven groups of 25. The later females laid fewer oothecae, a total of 90 being collected from 5th - 28th August, and these were subsequently divided into six groups of 15. Only average-sized, healthy oothecae were used, and these were allocated at random to the groups. Each ootheca was kept in an individual sterilised 5.0 x 1.0 cms. glass tube (Page 12; Fig. 2 ). The separate groups of oothecae were stored in plastic boxes (Plate 3). The oothecae were kept in an outdoor insectary, and transferred to a constant temperature of 20°C. and a 16 hr. per 24 hr. light regime at monthly intervals throughout the winter and following spring. The first group of oothecae deposited by the early and late females were put at 20°C. immediately. The oothecae laid by the two sets of females were transferred to 20°C. on alternate months so that the oothecae deposited by early females were transferred in July, September, November, January, March and May, and those laid by late females in August, October, December, February and April. A group of oothecae laid by females from the two cultures were kept in an outdoor insectary until hatching. The oothecae were watered with sterile water at intervals of three weeks throughout the experiment. Hatching was recorded daily, and the number of nymphs which emerged noted. The oothecae were dissected after hatching and the number of eggs failing to hatch recorded. Pronymphs which emerged from the oothecae but -72-

failed to undergo their first ecdysis were included in the total number which hatched. The embryonic stage at which diapause occurs was determined by fixing the oothecae in Alcoholic Bouin's Fixative. The individual embryos were dissected and cleared in a mixture of benzyl benzoate and 2 - ethoxy-ethanol (Peters, 1961).

Results and Discussion.

The development of the particular phase in the life cycle when diapause occurs can be regarded as two distinct processes, morphogenesis or morphological development and physiogonesis or physiological development (Andrewartha & Birch, 1954). The diapause stage is the stage in the life cycle when morphogenesis is at a minimum, while diapause development is the physiogenetic process which proceeds during the diapause stage and is a prerequisite to the active resumption of morphogenesis. The oothecae were subjected to varying periods of temperatures similar to those experienced in the field and then incubated to hatching at 20°C. Thus, the later the transfer to 200C. was made the longer was the duration of exposure to low outdoor temperatures. The presence of a diapause in the oothecae of E. lapponicus was determined by a consideration of the following criteria:- a. The proportion of oothecae which hatched after transfer to 20°C. b. The duration of the incubation period at 20°C. required for hatching. c. The variability in the time at 20°C. before hatching. These three aspects are discussed separately. Hatching of the oothecae was taken as an indication that diapause had been completed successfully. However, this criterion has several disadvantages, which are discussed fully by Browning (1952). -73- a. The Proportion of Oothecae which Hatched after Transfer to 20°C.

The number of oothecae of E. lapponicus, expressed as a percentage, which hatched after transfer to 20°C. at intervals throughout the winter increased the later the oothecae were put at 20°C. (i.e. the longer the exposure to low temperatures) with only a few minor deviations from this trend (Fig. 6). Oothecae laid both early and late in the season showed this tendency. It is interesting that approximately 30% of the oothecae deposited by early or late females hatch without exposure to temperatures below 20°C. This suggests that diapause development can proceed at 20°C.; but the low percentage hatch indicates that in most cases its completion is unsuccessful. A sudden rise in the percentage hatch therefore indicates the successful completion of diapause. The number of oothecae hatching increases after two (late oothecae) or four (early oothecae) months exposure to outdoor temperatures. There is very little difference in the percentage hatch of oothecae kept.under outdoor conditions and those put at 20°C. in April and May, indicating that morphogenesis proceeds successfully at a constant temperature of 20°C. The lower percentage hatch exhibited by oothecae transferred to 20°C. earlier in the year must therefore be due to the unsuccessful completion of diapause development. For an oothecae to hatch, a high percentage of the enclosed eggs must be at the same stage in development. This probably introduces some bias into the number of oothecae hatching, a problem which is not normally experienced in the study of diapauce. The actual number of eggs which fail to develop in an ootheca which hatches successfully is reduced the longer the exposure to low temperatures (Tables 16&17). However, the numbers are rather small to be conclusive, particularly in the case of oothecae laid late in the season. 100

80

60 hing tc 50% Hatch Ha

40 hecae t Oothecae deposited

Oo • 1:7:67 - 17:7:67 % Oothecae deposited 20 0 5:8:67 - 28:8:67

0 1 1 1 1 1 1 1 1 1 1 1 Jul Sep Nov Jan Mar May i C1 C2

Month of Transfer to 20°C.

FIG 6 Proportion of Oothecae Hatching after Transfer to 20°C.

-75-

Table .16 Oothecae Deposited Early in the Season (1st July - 17th July)

Month of r% Hatch % of Eggs Mean Number Vari- ! Coeff. Transfer of failing of Days at ante of to 20°C. Oothecae to Hatch 20°C. "4" S.E. Variation

July 32.0 6.2 222.3 t 12.3 :1201.1 15.6 Sept. 36.o 13.3 170.2 t 16.2 X2368.4 28.6 Nov. 40.0 6.0 85.o ± 14.o 11958.4 52.0 Jan. 60.o 6.1 36.0 ± 1.5 i 32.7 15.9 Mar. 44.o 2.2 33.5 ± 1.4 i 21.1 13.7 May 68.o 2.7 20.5 ± 1.0 15.5 19.2

Outdoors 72.0 3.0 2.7 (C1)

Table 17 Oothecae Deposited Late in the Season (6th Aug. - 28th Aug.)

Month of % Hatch i% of Eggs Mean Number Vari- Coeff. Transfer of failing of Days at ante of to 20°C. Oothecae to Hatch 20°C. + S.E.* Variation

Aug. 26.7 4.7 201.3 ± 8.7 304.3 8.7 Oct. 20.0 2.0 175.0 I 4.5 61.0 4.5 Dec. 46.7 10.9 59.3 ± 0.4 1.2 1.9 Feb. 40.0 12.2 35.8 ± 0.6 2.2 4.1 Apr. 86.7 5.1 27.6 + 0.5 2.9 6.2

Outdoors 86.7 4.1 1.0 (C2) _..

* = Standard Error. -76-

b. the Duration of the Incubation Period et 20°C. Required for Hatching.

The time required for hatching at a high temperature after a period of exposure to a low temperature indicates the extent of diapause development (Browning, 1952). Thus a sudden reduction in the mean incubation period at 20°C. suggests that diapause development has been completed successfully. Such a reduction in this period occurs in the oothecae of E. lapponicus (Fig. 7); in oothecae deposited early in the season it occurs between the groups transferred to 20°C. in September and November with a further proportionally greater reduction between the latter and those transferred in January, whilst in oothecae laid later in the season the sudden reduction occurs between the oothecae subjected to outdoor conditions until October and December. This reduction in the duration of the incubation period coincides exactly with the increase in the percentage hatch which further emphasises the near completion of diapause development by December or January. There is no further decrease in the duration of the incubation period at 20°C. in the oothecae transferred after January or February in the case of early or late oothecae respectively. This suggests that diapause development is complete by this time, although the arrest of growth probably continues in the field until the temperatures are above the threshold for morphogenesis. The standard error of the mean is greatly reduced for both early and late oothecae in the groups moved to a higher temperature in January or December respectively, and shows no further decrease for the remaining months of transfer. Oothecae which had not been exposed to a substantial period of winter outdoor conditions again show that although diapause development can be completed at 20°C. it takes much longer and is seldom successful. A further reduction in the duration of the incubation period at 20°C. in the oothecae transferred to this temperature in May, and to a lesser extent in those transferred in FIG 7Duration ofIncubationPeriod at20°C.Requiredfor Hatching. Mean Time at20 °C. (Days ) 200 250 100 150 50 0 Jul Sep Month ofTransferto20°C. Nov 0 • T Jan 1:7:67 -17:7:67 Standard Error 5:8:67 -28:8:67 Oothecae deposited Oothecae deposited Mar May -78-

April suggests that post-diapause development may well begin under field conditions by this time and thus results in a further reduction in the period at 20°C. prior to hatching. It would appear that post-diapause morphogenesis in this species requires approximately 30 - 35 days at 20°C. for its completion (Figs. 8 & 9). c. The Variability in the Time at 20°C. before Hatching.

The restriction of hatching to a very short period after incubation at a high temperature again suggests a satisfactory completion of diapause development. The spread of hatching over a considerable period indicates that diapause development is at varying stages of completion when the transfer to a high ',.cmperature occurs. Since diapause development proceeds only slowly at high temperatures, this results in a much larger range of hatching dates. Thus a reduction in the spread of hatching is an indication of the completion of diapause development. Figs. 8 & 9 show the distribution of hatching times after transfer to 20°C. at various intervals throughout the winter. It can be seen that of the oothecae laid early in the season those put at 20°C. immediately after laying took a long time to hatch, and hatched over a considerable period of time. The oothecae transferred to 20°C. in September, with the exception of a single ootheca which showed a marked reduction in the incubation period, had a similar variation in their hatching dates. The group transferred to 20°C. during November again showed a wide spread in their hatching dates. This was apparently due to some oothecae having more or less completed their diapause development by this time and hatching over a limited period, whereas the remainder hatch over a far longer period and require greater incubation periods. Oothecae subjected to outdoor temperatures from the time of laying until January exhibit a drastic reduction in the variation of their hatching times. This trend becomes even more marked in the oothecae transferred to 20°C. in March and May. In these groups diapause 51- JULY

0 n nn n n n 51 SEPTEMBER

0 n nn nn

NOVEMBER

0 n n

JANUARY 5 Cr)

(.) C3 0 10 cb

a, MARCH o

gg 0 10 [

MAY 5

A 0 50 100 150 200 250 300 Month of Transfer Incubation Period at 20°C. (Days) to 20°C.

FIG 8 Number of Oothecae, laid Early in the Season, Hatching after Transfer to 20°C. 51 AUGUST

0 nnn n 5i OCTOBER

0 nn ng hi

DECEMBER tc Ha e heca t FEBRUARY 5 Oo ber

m 111 0 Nu

10

APRIL 5

• 0 50 100 150 200 250 300 Month of Transfer Incubation Period at 20°C. (Days) to 20°C

FIG 9 Number of Oothecae, laid Late in the Season, Hatching after Transfer to 20°C. -81- development had been successfully completed in all oothecae by this time and morphological development was resumed immediately on transfer to 20°C. The results obtained from the oothecae laid late in the season are rather inconclusive, since the numbers involved are low. However, a similar reduction in the spread of the hatching times is apparent (Fig. 9). The group of oothecae transferred to 20°C. in December shows a decrease in the variation of the incubation period required at 20°C. This trend is continued in the groups transferred in February and April, the incubation periods them- selves also being reduced. Oothecae kept continuously under field conditions from the time of deposition, hatched over a period of four or seven days depending on whether they were laid late or early in the season. Oothecae kept under these conditions hatch over the same few days irrespective of the time in the season at which they were laid. This is yet further evidence of the existence of a diapause in the oothecae of this species and emphasises the ecological role of diapause in synchronising development. A number of measures of the variability of hatching dates have been employed by different authors. Rakshpal (1962) used the range or hatching period which refers to the difference in time between the hatching of the first and last eggs in a group; this has an obvious disadvantage in that it takes no account of the distribution of the hatching dates. However, as an absolute measure of the total spread, considered with a knowledge of the actual distributions of the incubation periods at 20°C. (Figs. 8 & 9 ) the hatching period does show clearly the sudden decrease in the spread of hatching dates at 20°C., which occurs between the groups transferred in November and January (Fig. 10). The large hatching period in the two groups moved to a high temperature in September and November can be explained by the fact that the oothecae can be divided into two groups, those which have almost completed diapause development, and those which have not, Oothecae deposited

1: 7:67 - 17: 7:67 ) s Oothecae deposited

Day 0 5:8:67 - 28:8:67 d ( io Per ing h tc Ha

Jul Sep Nov Jan Mar May C1 C2

Month of Transfer to 20°C.

FIG 10 Hatching Period after Transfer to 20°C. thus giving a wide range in the times of hatching (Fig. 8). In comparison the oothecae laid later in the season have much reduced hatching periods throughout the experiment. This may well be due to the low numbers involved. However, the reduction in the hatching period in the group transferred to 20°C. in December is still noticeable. In the oothecae subjected to outdoor conditions after December or January no further reduction in the hatching period occurs, which again indicates that diapause development is probably more or less complete by this time. The variance is a preferable measure of the variation in the incubation period at 20°C. (Tables 16 & 17), but in this experiment the mean duration of the incubation periods are so variable that its use is precluded. Consequently, the coefficient of variation is used as a measure of the dispersion (Fig. 11) of the incubation time at 20°C. before hatching. The explanation of the very high coefficient of variation in the groups of oothecae moved to 200C. in September and November probably concerns the distinction between those oothecae which have or have not completed diapause development at the time of transfer as previously described. The most striking feature emerging from the calculation of the coefficient of variation is the vast difference between the oothecae deposited early and late in the season; this requires further work. However, it is possible that oothecae deposited early in the season when the temperatures are still high may enter diapause more firmly than those laid later when the temperatures have started to fall, Browning (1952) found that the eggs of Gryllulus (= Acheta) (= Teleogryllus) commodus Walker have a more intense diapause if they are initially subject to a high temperature before the low temperature treatment, and Hogan (1960) also found that diapause is weakened in the eggs of this species as the temperatures decline. However, Way (1959) found that initial exposure to high temperatures did not intensify the diapause in Lutohylemzia coarctata Fall6n.

50

Oothecae deposited

1:7:67 - 17:7:67 40 Oothecae deposited 0 5:8:67- 28 .8:67.

30

tion ia r

f Va 20 t o n ie ffic e

Co 10

0

Jul Sep NOV Jan Mar May

Month of Transfer to 20°C.

FIG 11 Variation in the Hatching Time after Transfer to 20°C. -85-

Oothecae of E. lapponicus were subjected to varying periods at conditions simulating those occurring in the field during winter, and then incubated at 20°C. The presence of a diapause was confirmed by examination of three separate criteria; it was found that the longer the exposure to low temperatures:- a. The greater the percentage hatch after transfer to 20°C. b. The shorter is the period of incubation at 20°C. c. The smaller is the variation in the hatching dates. Since all oothecae undergo an arrest in morphological growth, diapause in this species is obligatory. There was no evidence of diapause lasting more than one winter. It is suggested that diapause development, which follows a period of morphogenesis known as pre-diapause development, is complete in oothecae laid early in the season by January and in the later oothecae by February. Howeveri post-diapause morphogenesis is not resumed until later in spring when the temperatures are rising. The results of this work show that post-diapause development has just commenced by April in the oothecae laid late in the season. The possible effect of photoperiod on diapause was not given detailed consideration in this work; however, a group of oothecae transferred from outdoor conditions to a 10 hr. per 24 hr. light regime at 20°C. in January gave very similar results to an identical group of oothecae subjected to a long day regime. Hatching of this species and E. pallidus and E. panzeri occurs in the field over a very restricted period which suggests that the oothecae of the two latter species also pass the winter in a state of diapause. Preparations of embryos of E. lapponicus and Eipallidus fixed at various intervals during the winter show that the arrest of morphogenesis occurs at the post-anatrepsis stage prior to the revolution of the embryo (blastokinesis) when the body segments and appendages have differentiated. The occurrence of diapause at this stage has been recorded in several Orthopteroid species cited by Rakshpal (1962:188). -86-

The occurrence of an obligate diapause in the oothecae of E. lapponicus is of considerable adaptive value, since it ensures that the nymphs emerge when conditions are favourable for their growth. It is possible that the difference noticed between oothecae deposited at different times during the season may ensure that no oothecae hatch while temperatures are still favourable in the autumn, since winter conditions would probably kill the newly hatched nymphs.

(iii) Water Relations of the Oothecae.

Materials and Methods.

Twenty healthy oothecae of E. lapponicus, deposited over a period of three days in August, were stored for their entire developmental period in an outdoor insectary, in individual 5.0 x 1.0 cms. glass tubes (Page 12; Fig. 2). These were weighed and measured at intervals of two months until the oothecae visibly increased in size, and then at three-weekly intervals until hatching. The oothecae were weighed, using a Cahn Electrobalance, to an accuracy of 0.001 mgs. The length and width of the oothecae were measured, correct to 0.1 mms., with a monocular compound microscope fitted with an eye-piece graticule. The following year 130 oothecae of E. lapponicus deposited from 16th - 22nd July were divided randomly into eleven groups of ten and one of twenty oothecae. These were kept in separate tubes in an outdoor insectary as before. The group of twenty oothecae was weighed immediately after deposition to give an estimate of the water content of the oothecae of this species immediately after formation. The remainder were weighed at monthly intervals from August until May of the following year. After weighing each group of oothecae was air-oven dried at 80°C. to constant weight. One group was kept in the insectary until hatching to determine the exact time of hatching. -87-

Results and Discussion.

a. Changes in the Weight and Shape of Oothecae of E. lapponicus during Development.

Oothecae of E. lapponicus increase in weight very gradually from the time of deposition until February. For the remainder of the developmental period changes in weight become more pronounced; during a three week period from the beginning of May the oothecae almost double their initial weight (Fig. 12a). The rise in the mean weight recorded in April and early May is the result of an advanced increase in a few oothecae, illustrated by the larger standard error of the means for these periods (Table 11))).

Table 18 Changes in the Weight and Shape of Oothecae of E. lapponicus during Development.

Time during Mean weight of oothecae Mean width of oothecae Development in mgs. ± S.E.* in mms. S.E.*

August 6.46 ± 0.23 1.59 ± 0.01 October 6.6o ± 0.22 1.60 ±0.01 December 6.72 ± 0.21 1.61 1. 0.01 February 6.80 ± 0.22 1.65 ± 0.01 April 7.08 ± 0,42 1.71 ± 0.04 Early May 8.59 ±0.54 1.92 ± 0.05 Late May 12.20 t 0.27 2.24 ±0.02

= Standard Error. 13

12 a. Weight b. Width.

23 11 -, --, tri H Hatching vi o) E H Hatching ....E 10 i Standard Error E 2.1

cu Q) o 0 o (.) q, 9 (i) -c 1 -c 1.9 co o co c)o c)

17 3 7 c c cs _ o a) a) Z 6 1.5

Aug Oct Dec Feb Ap May Aug Oct Dec Feb Ap May Month Month

FIG 12 Changes in the Obthecae of E. lapponicus during Development. 89

The increase in weight was found to be the result of a sudden uptake of water by the oothecae of this species. The oothecae of the three British species of Ectobius vary in size during development. Oothecae which contain normally developing eggs swell rapidly prior to hatching; this is accompanied by a change in colour of the wall of the ootheca from dark brown to pale grey-brown. The length of the ootheca remains unchanged, but the width gradually increases from deposition until February, after which a more marked change occurs (Fig. 12b). This size increase, caused by an expansion of the individual eggs, is brought about by the flattening of the median ventral groove which extends the full length of the ootheca (Fig. 5c, Page 66). Several species which require additional water during development have similar devices to accommodate the expanding eggs (Roth, 1967b). However, the individual eggs of B. orientalis and P. americana extend further into the spongy inner layer of the ootheca as they increase in size, there being no change in the gross size of the ootheca (Roth & Willis, 1955c). Hatching occurred two weeks after the final measurements were made. It is therefore possible that a further increase in weight and size may take place immediately prior to hatching. b. Changes in the Water Content of Oothecae of E. lapponicus during Development.

The water content of twenty oothecae of E. lapponicus at deposition was found to be 37.11 t 0.30%. The proportion of water in the oothecae remains remarkably constant during much of the developmental period; however, there is a very slight increase throughout the winter until March, when a more marked increase occurs. This is followed by a substantial 20% rise over a four week period from the end of April (Fig. 13). The increase in the mean percentage of water recorded in March and particularly April is the result of a few oothecae which commence their uptake of 100 • Water Content o Dry Matter 7

v 80 H Hatching nek u

6 10 4. theca f Standard Error 60 - 941 Oo 5p er d 1q p

r 5 Je te

40. 00 Wa e ooet4 tag 0

en 4

(.0 20 cn Perc

Mean 3

0 ' Jul Sep Nov Jan Mar May

Time

FIG 13 Changes in the Water Content and Dry Matter of Oathecae of F. lapponicus during Development. -91-

water in advance of the majority. This is clearly illustrated by a consideration of the standard errors of the means which remain consistently small during most of embryogenesis, become slightly larger in March, reach a maximum in April and decline again in May when most of the oothecae have absorbed the required amount of water. Two weeks before hatching the water content of oothecae of E. lapponicus is 68.92 t 0.64%. It is interesting that Roth & Willis (1958a) recorded the water content of a small number of newly formed oothecae of E. pallidus to be 40.0 ± 0.61%, whilst oothecae collected from the field in April contained 41.0 t 1.1% water. At the time of hatching they found the oothecae contained about 75% water. It seems that the water relations of this species are similar to E. lapponicus, there being no appreciable uptake of water for a long period, followed by a sharp rise in the water content before hatching. No accurate measure of the dry weight of the oothecae of E. lapponicus was possible, since time did not permit a quantitative study of individual eggs. However, a preliminary estimate of the mean dry weight of the oothecae at monthly intervals during embryogenesis reveals a slight decrease in the quantity of dry matter (Fig. 13). A similar loss in dry matter has been recorded in several oviparous species (Roth & Willis, 1955c).

Hatching occurs in the oothecae of most species of cockroaches when the water content is in the order of 65 - 75% (Roth & Willis, 1955a, b, c and 1958a). The initial water content of the oothecae varies with the species. Some contain adequate water for development immediately after formation whilst others need to acquire additional water during embryogenesis. The oothecae of B. orientalis and P. americana contain sufficient water at deposition for their complete development. However, the individual eggs within the ootheca pick up water from the moist inner surface of the ootheca, which consequently becomes progressively drier as development proceeds (Roth & Willis, -92-

1955c). This fluid surrounding the eggs in a freshly formed ootheca was found to be a dilute solution of protocatechuic acid responsible for the hardening of the ootheca (Pryor et al., 1946). Many species deposit oothecae with a water content of less than 50%, and it is these which take up water from the surroundings (Roth, 1967b). If the oothecae of Ectobius are kept dry throughout their normal developmental period none will hatch. In E. lapponicus most of this water is taken up at the end of embryogenesis, after the resumption of post-diapause development in April. The embryonic stage at which water uptake begins is not known. Moreover, its initiation varies in time between different oothecae; by April 30% of the oothecae had considerably increased water contents. In this way E. lapponicus differs from other field-dwelling species e.g. P. virginica and Cariblatta lutea minima Hebard, which do not diapause as oothecae and therefore gradually take up water throughout development (Roth & Willis, 1958a). It is interesting that the oothecae of E. lapponicus contain a very similar percentage of water after formation to those of ovoviviparous species e.g. P. nivea, the oothecae initially containing 37% water (Roth, 1967b); such oothecae acquire additional water from the body of the female during embryogenesis. The site of water uptake in cockroach oothecae is uncertain (Roth & Willis, 1955c). Since the respiratory tubes in the keel of oothecae allow the passage of air, it is possible that the eggs are capable of picking up water from a saturated atmosphere. The chorionic expansion at the base of the tubule would facilitate this (Fig. 5c, Page 66). Roth & Willis (1957) found the eggs of E. pallidus would not hatch if kept continually moist during the winter. However, a large percentage of the oothecae of the three species of Ectobius hatched when, moistened at regular intervals throughout development (Table12, Page59); a situation which they would normally experience in the field. Only oothecae which were allowed to become dry after water uptake had commenced (indicated by the increase in size of the ootheca) failed to hatch. -93-

(iv) Hatching and Description of Pronjmph.

Hatching of the ootheca is initiated by the internal pressure exerted by the fully developed nymphs, which causes the dehiscence of the keel in the central region. This split extends laterally along the entire length of the keel (Plate 11a). The two halves of the ootheca separate, and the chorion of each egg is ruptured dorsally. The nymphs are arranged in two parallel rows, their dorsal surface adjacent to the wall of the ootheca, and their head towards the seam (Plate 11a).The further separation of the two halves of the keel is brought about by the expansion of the nymphs which seem to swallow air. A series of convulsive movements of the anterior part of the body of each nymph enables them to escape from their egg membranes and emerge from the ootheca by crawling between the fissure created by the splitting of the keel (Plate 11b). Hatching is essentially similar in the three species. All the nymphs may emerge simultaneously or more frequently one or two may escape before the others which facilitates the emergence of the remainder (Plate 11b). The duration of hatching is variable; but is usually complete in 10 - 15 minutes. After the emergence of the nymphs the two halves of the ootheca come together again and enclose the dried persistent remnants of the egg membranes. The number of nymphs which normally emerge from oothecae of the three species is given in Table 13, Page 60. In an ootheca containing an average number of eggs it is not infrequent that a few may fail to hatch. By the dissection of oothecae from which some nymphs have emerged the main reasons for this inability of some eggs to hatch have been determined as follows:- a. Nymphs may be fully developed, but fail to emerge from the ootheca with the majority of the nymphs. Their subsequent emergence is prevented by the closure of the two halves of the ootheca. b. Alternatively nymphs may become fully developed, but -94-

a. b.

r-

. d.

Plate 11 Hatching sequence in an ootheca of E. lap.00nicus. -95.-

due to the initial orientation of the egg in the ootheca are inverted so that their abdomen is directed towards the keel, and are therefore unable to escape under natural conditions. c. The development of certain embryos may be retarded or inhibited completely; such embryos become desiccated after the emergence of most of the nymphs and thus fail to develop further. Such eggs are normally situated at the extreme ends of the ootheca, or under a portion of the keel which has not been perfectly formed.

In each species the freshly emerged nymphs or pronymphs, a term preferred by Qadri (1938), are enclosed in a thin transparent membrane. This membrane is continuous and surrounds the legs, antennae and cerci, which are held close to the body (Fig. 14). During the later stages of hatching, or after the nymphs have left the ootheca, the first ecdysis, often referred to as the pronymphal ecdysis" occurs. The pronymphal membrane is ruptured dorsally and is gradually shed towards the posterior end of the body (Plate 11c). The fore, mid and hind legs and antennae are freed in turn and finally the wrinkled membrane is removed from the abdominal cerci by the legs. These cast skins which shrivel into a small whitish mass either remain near or attached to the hatched ootheca (Plate 11d), but were not observed to he eaten. Uvarov (1966) considers the function of the pronymph or "vermiform larva" in Orthopteroid insects is to assist the insect to emerge on the surface. Brown (1952) noted that nymphs emerging from oothecae of E. panzeri only moulted when they reached the surface. However, it appeared more usual for the pronymphal ecdysis to occur as the nymphs left the oothecae. Since the pronymphal membrane is now regarded as a true cuticle (Uvarov, 1966), the pronymph is really the first nymphal instar. However, the term nymphal instar in this work refers only to the free- lmm.

FIG 14 Lateral View Pronymph - E. lapponicus. -97-

living immature stages. The freshly emerged first instar nymphs are at first elongated and inflated by the absorption of air during hatching, although after a few minutes the abdomen assumes its usual flattened shape and the tanning of the cuticle commences. -98-

III. OVERWINTERING OF THE NYMPHAL INSTARS.

Materials and Methods.

All nymphal instars of E. lapponicus were collected during the winter months by cutting grass tussocks (Method 6, Page 9). The limited number of collections of nymphs of E. pallidus during this time were made by lifting fallen bracken fronds onto a beating tray (Method 1, Page 8). The nymphal instars were successfully reared under outdoor conditions during the winter in Watkins & Doncaster cages (Method 4, Page 12; Plate 2). This provided a relatively convenient means of re-collecting the nymphs at intervals to assess mortality. By rearing the individual instars in separate cages the number of nymphs moulting was easily determined. The nymphs of E. lapponicus subjected to experimental conditions at intervals during the winter were collected from the field at five-weekly intervals from September to April of the following year, and immediately transferred to 20°C. Two different light regimes were employed with photoperiods of 10 and 16 hr. per 24 hr. The nymphs of each instar were reared in separate cages of the type described on Page 12, Fig. it. Between fifteen and twenty nymphs of the third and fourth instar were used in each treatment. Observations were made at weekly intervals, and the number of nymphs moulting recorded. Due to the difficulty in collecting large numbers of nymphs during the winter, the data for two consecutive years were pooled. To investigate the nature of the adult emergence, nymphs of the three species were collected in May before the final ecdysis occurred. The collections of E. lapponicus and E. pallidus were made from the same localities as those made during the winter months. These nymphs were kept in an outdoor insectary in glass tanks (Page 12; Plate 1). Daily records of the adult emergence of the two sexes of the three species were kept. -99-

Results and Discussion.

(i) Field Collections of Nymphal Instars.

At five-weekly intervals, from October to April of the following year, fifty large grass tussocks were cut from North Gravel, Silwood Park, and a1.1 E. lapponicus nymphs collected.

Table 19 Number of Each Instar Collected from the Field (Expressed as a Percentage of the Total).

J.110 UdV Month of Total Collection Collected 2nd. 3rd, 4th,

October 2.6 33.5 63.9 155

November 3.4 28.1 68.5 146

Dccouber 3.7 34.9 61.5 109

January 1.7 40.9 57.4 115

March 11.3 45.3 43.4 106

L,lril 0.0 29.4 70.6 17

May 0.0 11.1 88.9 9 -100-

The majority of nymphs of E. lapponicus overwinter as third or fourth instars. The first and final instars are not found during the winter in this locality. There is also no trace of adult insects. Some mortality is indicated from December onwards, which is particularly noticeable in the fourth instar; however, the fall in totals per collection from December to March may be due to differences' in ease of sampling, since collection is more difficult when the ground is frozen. The sharp decline in numbers in April and May is probably a result of the dispersal of the nymphs from the tussocks with the onset of warmer weather. Nymphs are commonly active in the field at this time. The restricted distribution of E. pallidus made regular collections throughout the winter impractical. However, a single collection in February showed that this species may overwinter in any one of four instars. No trace of the first or final instar or adult were found. The vast majority of nymphs pass the winter as fourth instars (83%), with only 8% as third and ttem.tCi 4% as second oze fifth instars.

(ii) Overwintering Behaviour of Nymphs in an Outdoor Insectary.

Nymphs of E. lapponicus, reared in Watkins & Doncaster cages, were carefully observed throughout the winter, and a monthly check on the insects made. The cages were kept in an outdoor insectary where conditions simulated those in the field. The mortality occurring in the immature stages of IF,. la-vonicus is relatively low under insectary conditions. It is probable that a higher mortality may exist in the field. The nymphs remain inactive during the cold months of the year, concealed in the interior of the grass tussocks, or under the roots. They assume a characteristic posture; the body is flexed ventrally and the legs and antennae are held close to the body. -101 -

Table 20 Mortality of Nymphal Instars of E. lapponicus during the Winter Months.

Number of Each Instar Total Date of Number of Observation 2nd. 3rd. 4th. Insects

18th Oct. 16 32 12 6o

14th Nov. 16 31 12 59

12th Dec. 16 31 11 58

9th Jan. 16 31 11 58

6th Feb. 16 31 11 58

6th Mar. 15 30 11 56

3rd April 15 30 10 55

1st May 14 30 10 54

15th May Moulting Commenced. 54

However, on warm days in April and May many nymphs of the three overwintering instars become active and climb the grass stems where they may remain for several hours.. The nymphs of this species apparently feed during the winter, and in spring -102- many nymphs feed actively on the artificial diet provided. No moulting occurs during the winter; the nymphs moult last at the end of September, and moulting is resumed during the first week of May. This behaviour has been observed over three winters. -A less detailed study of the overwintering behaviour of E. pallidus reveals a close similarity with E. lapponicus.

(iii) The Occurrence of an Abnormal Instar under Experimental Conditions.

If first instar nymphs of E. lapponicus are kept at 20°C. in a 16 hr. per 24 hr. light regime the first three moults occur normally, and the instars have similar durations to those quoted on Page 199. However, the duration of the fourth instar is extended to 36.2 ± 1.0 days, and the resultant fifth instar has a totally different appearance to the normal fifth instar. This abnormal moult occurs in both sexes, and in a large proportion of individuals; the remainder fail to moult. In size the instar approaches that of a normal fifth instar, but the genitalia and wing pads show little further differentiation from the previous instar, indicating some degree of metathetely. The body is pale in colour, even in the males. Abnormal instars may live for considerable periods, usually without undergoing a further moult. However, a few specimens moult again and produce another slightly larger abnormal instar. Additional instars in the life cycle of certain Blattidae, recorded by Zabinski (1936), were caused by "depressed metabolism" as a result of low temperatures or an inadequate diet. In Locusta nigratoria migratorioides (Reiche & Fairmaire) an additional instar in some females is due to an inherited factor (Key, 1936), whereas Albrecht (1955) found that the number of instars in Nomadacris septemfasciata (Serville) and Schistocerca gregaria (Forskal) was correlated with the size of -103-

the hatchlings, the smaller hatchlings having a slower rate of growth and therefore requiring an additional instar. However, the abnormal instars in E. lapponicus are not merely additional instars in the development of the species, since they never moult into a normal fifth instar or adult. No record has been found of the production of a completely abnormal instar resulting .in the death of the insect before maturity is reached. No abnormal instars have been found in the extensive collections from the field of this species. The production of such an instar is apparently a response by the insect to conditions which are not normally experienced in the field by the fourth instar, and are therefore unfavourable to its normal growth. E. pallidus produces a similar abnormal instar at the fifth ecdysis (i.e. an abnormal sixth instar) when nymphs are reared under the same conditions.

(iv) Determination of Diapause in the Nymphs of E. lapponicus.

Nymphs collected from the field at five-weekly intervals during the winter months, from September to April of the following year were transferred immediately to 20°C. and a 16 hr. per 24 hr. light regime. These nymphs were therefore subjected to varying periods under winter field conditions followed by a period at a high temperature. The existence of a diapause in the overwintering instars was then investigated by a consideration of the following:- a. The percentage of nymphs moulting after transfer to 20°C. b. The duration of the period at 20°C. before the first ecdysis. c. The relative proportions of normal and abnormal instars produced after transfer to 20°C. -104-

Fourth Instar. a. The Percentage of Nymphs Moulting after Transfer to 20°C.

The number of nymphs which moult after transfer to 20°C. is first considered irrespective of the nature of the instar produced. There is a steady rise in the percentage of nymphs moulting which is correlated with an increase in the duration of exposure to field conditions (Fig. 15a). Most of the nymphs (95%) entering the winter as fourth instars moult when the nymphs are moved to 20°C. in April. The rise in the percentage of nymphs capable of ecdysis after increasing periods of exposure to field conditions suggests the completion of diapause development in these nymphs. b. The Duration of the Period at 20°C. before the First Ecdysis.

The duration of a fourth instar of E. lapponicus at 20°C. and a 16 hr. per 24 hr. light regime is normally 20.7 ± 0.3 days. However, when fourth instars are transferred from field conditions to a constant temperature of.20°C. in September the duration of the period is extended to 32.3 i 3.1 days, indicating an arrest of growth at this stage. This pre-moult period is gradually reduced the longer the nymphs are exposed to field conditions (Fig. 15b);nymphs moved to 20°C. in April require a mean of only 30 days at this temperature before ecdysis. The gradual completion of diapause development results in a decrease in the period at 20°C. before ecdysis. c. The Relative Proportions of Normal and Abnormal Instars Produced after Transfer to 20°C.

The abnormal fifth instar in this species is apparently produced in response to conditions which are not normally experienced at this particular stage in ontogeny. Since normal fifth instars are readily produced at 20°C. and a 16 hr. per 24 hr. photoperiod, -105- 100

75

ing lt 50 Mou

hs 0 3rd lnstar mp

Ny 25 • 4th lnstar %

Sep Oct Nov Jan Feb Apr Month of Transfer to 20°C.

FIG 15a Nymphs Moulting after Transfer to 20°C. -16hr. Day.

100-

0 3rd Instar 75 ) • 4th Instar s Day (

is s 50 dy Ec

e for 25 be Time

Sep Oct Nov Jan Feb Apr Month of Transfer to 20°C.

FIG 15b Duration at 20°C. before First Ecdysis. -1o6- the production of abnormal fifth instars indicates the unsuccessful completion of diapause development. Nymphs transferred to 20°C. in September, October and November all produced abnormal fifth instars when moulting occurred (Fig.16a),whereas of those transferred in January 67% of the nymphs which moulted produced normal fifth instars which moulted again and became adult, indicating the successful completion of diapause development in these nymphs by this time (Fig. 16b). By February over 90% of the nymphs have completed diapause development. Although diapause development appears to be completed by this time, the moult into the final instar does not occur in the field until the beginning of May when conditions are presumably more suitable for the active resumption of morphogenesis.

Third Instar. a. The Percentage of Nymphs Moulting after Transfer to 20°C.

Unlike the fourth instar nymphs, a high percentage of the preceding instar undergo at least one ecdysis when transferred from field conditions to 20°C., irrespective of the duration of the former period (Fig. 15a). A lower percentage of nymphs moulted when moved to a higher temperature from November to February, but this appears to have no special biological significance, since all the nymphs used in these treatments for one year moulted readily, although in the following year some of these nymphs exhibited an inexplicable mortality soon after transfer to 20°C. b. The Duration of the Period at 20°C. before the First Ecdysis.

The duration of this instar at 20°C. and a 16 hr. per 24 hr. light regime is normally 25.3 ± 0.2 days. The mean time at 20°C. before the first ecdysis remains more or less constant in this FIG 16bMoulting NymphsProducing Normal5thlnstars. FIG 16aMoultingNymphsProducingAbnormal5thInstars.

% Normal 5th Instar %Abnorma l 5th Instar 100 50 75 25 Sep OctNovJanFebApr Sep OctNov JanFebApr • o 3rdInstar 4th Instar Month ofTransferto20°C. Month ofTransferto 20°C. -107- -108- instar (varying from 29.2 - 41.2 days), regardless of the duration of the previous period under field conditions (Fig. 15b). There is no obvious trend showing this period to decrease as the period of exposure to outdoor temperatures increases.

c. The Relative Proportions of Normal and Abnormal Instars Produced after Transfer to 20°C.

The two former considerations suggest that diapause may be confined to the fourth instar and that the third instar may merely exhibit quiescence, defined as an arrest of development caused by unfavourable environmental conditions and terminated as soon as favourable conditions return (Way, 1962). Thus on moving a quiescent third instar to 20°C. and a long day regime one would expect normal development to be resumed and maturity attained by nymphs collected from the field at any time during the winter. No abnormal fifth instars would therefore be produced. However, this is not the case and the existence of a diapause is suggested in the third instar also. The number of insects which moult and produce an abnormal fifth instar decreases the longer the nymphs are subjected to field conditions (Fig. 16a). Mortality in the fourth instar causes the percentage of abnormal instars to be lower in the nymphs transferred to 20°C. in September and October. Diapause development takes longer to be completed in the third instar than in the fourth; 67% of the fourth instar nymphs have completed diapause development by January, whereas in only 14% of the third instar nymphs is it complete by this time. It is not until April that diapause development is complete in almost 90% of the third instar nymphs.

Second Instar.

The proportion of nymphs of E. lapponicus overwintering as a second instar is small. It was therefore, not possible to collect -109-

enough specimens for a detailed investigation of the overwintering behaviour of this instar. However, all specimens collected at intervals during the winter were subjected to the same experimental conditions as the third and fourth instars. The mode of behaviour of the second instar nymphs shows a close similarity with that of the third instar. The moult into a third and fourth instar occurs after a short period at 20°C. which does not vary with the previous duration of exposure to field conditions. However, the moult into a fifth instar is delayed. Only abnormal fifth instars are produced in nymphs transferred to 20°C. before February. However, after this time a few normal instars are produced.

From this study of the overwintering behaviour of E. lapponicus it would appear that a diapause in the nymphs of this species is obligate and may occur in any one of the three overwintering instars. The first instars hatching from oothecae moult readily throughout the summer until they reach the fourth instar and at this stage development is halted until the following spring. Many nymphs reach the fourth instar by the beginning of August when conditions are still favourable for growth and differentiation. The earlier instars continue to moult until the end of September, or until the fourth instar is reached. Insect species differ in the intensity of their diapause (Andrewartha & Birch, 1954). It is apparent that the diapause occurring in the fourth instar of E. Iapponicus is of a more intense nature than that in the preceding instars. Seeing that no further arrest of growth occurs in the fifth instar, the intense diapause in the fourth instar is of great adaptive importance to the species; since any chance of adults being produced during the unfavourable winter months is thereby precluded. The existence of a diapause in the second or third instar causes diapause in the fourth instar to be averted. Similarly Ludwig (1932) found that if diapause had been completed in the -110- second instar of Popillia japonica Newman it did not recur in the third instar; but if absent from the second instar diapause was automatically induced in the third instar. However, Readio (1931) found that nymphs of Reduvius personatus (Linnaeus) diapausing in the third instar also diapause in the fifth instar, whereas those which diapause as fourth instars do not undergo this second period of arrest. The presence of a diapause in one instar seems to prevent its recurrence in the next instar. In E. lapponicus diapause in one of the earlier instars results in the omission of a further diapause in the remaining instars.

(v) The Effect of a Short Day Regime on the Nymphs of E. lapponicus.

As in the previous section nymphs collected from the field at five-weekly intervals throughout the winter were transferred to 20°C. and subjected to a light regime of 10 hr. per 24 hr.

Fourth Instar.

Moulting in this instar is completely inhibited when nymphs are introduced to these conditions from September to January (Fig. 17). It has been concluded that diapause development is complete in the field by February; however, only a small percentage of the nymphs moved to 20°C. after this time moulted. Since most fourth instar nymphs moulted when subjected to 20°C. and a long day regime, this cessation of moulting must be a direct result of the shorter daylength. Only abaormal fifth instars were produced from nymphs transferred to 20°C. in February, although some normal instars were produced when the transfer was made in April; however, these nymphs died before becoming adult. o 3rd Instar • 4th Instar

lting Mou hs mp Ny %

Sep Oct Nov Jan Feb Mar

Month of Transfer to 20°C.

FIG 17 Nymphs Moulting after Transfer to 20°C. - 10hr. Day. -112-

Third Instar.

Moulting in this instar is also retarded under conditions of a high temperature and a short day, but not to the same extent as the previous instar. This suggests that the response to daylength is less marked in this instar. When subjected to a long day regime at 20°C., third instar nymphs moulted readily into the next instar. However, a short day regime prevented moulting in the majority of nymphs moved to 20°C. from October to January. It is interesting that a number of third instar nymphs (60%) subjected to these conditions from September moulted into a fourth instar, but died without moulting again (Fig. 17). It has already been noted that in the field moulting in this instar does not cease until the end of September, when presumably diapause intervenes. It is probable therefore that nymphs moved to 200C. at this time may moult once and enter diapause as a fourth instar, which undergoes no further moults. The sudden increase in the number of nymphs which moult after transfer to 20°C. and a short day regime in April is difficult to explain. It is possible that by April nymphs in the field are ready to moult and that a temperature of 20°C. causes moulting to occur. However, no nymphs moult beyond the fourth instar.

Daylength is obviously an important factor in the development of the nymphs of E. lapponicus. Moulting is retarded in all instars, although in some individuals a single moult may occur. Oothecae hatch in a 10 hr. per 24 hr. photoperiod, but the nymphs fail to develop beyond the first instar. In no case was an adult produced in a short day. A daily light regime of 10 hr. per 24 hr. appears too short for active growth in this species even if the temperature is favourable. -113-

(vi) Relationship between the OverwinteringStaa.es and Adult Emer1ence.

Adults of E. lapponicus emerge from the end of May to the middle of August. The dates when a collection of 210 nymphs became adult in an outdoor insectary were recorded and. the frequency distribution plotted as shown (Fig. 18). The emergence too the form of two peaks (A & B) and a long tail (C). It seems .orobable that this adult emergence pattern is related to the way in which the species overwinters in more than one instar. Nymphs collected from the field in February were assumed to have survived the winter conditions successfully. The percentages of nymphs overwintering in the three instars were compared with the percentage of the total population of adults in the three emergence peaks evident in the following summer.

Table 21 Relationship between the Overwintering and Adult Emergence in E. lapponicus.

Instars collected in field (February 1968)

4th. 3rd. 2nd.

Percentage of nymphs 43.4 45.3 11.3 collected

Percentage of adult 42.9 44.8 12.4 emergence

A B C

Emergence peaks of adults (Summer 1968)

8

7 )

ion t

la 6 u O Male

l Pop Female ta 5 ❑ To % ( 4 C) CI)

•a) 3

• 2

O O

0 1 R 25 30 4 9 14 19 24 29 4 9 14 19 24 29 3 8 13

May June July August

4 A 1.4

FIG 18 Adult Emergence of a Population of E. laprackc_u3 - Summer 1968. -115-

These two sets of percentages correspond well, suggesting that when moulting is resumed in early May, nymphs overwintering as a fourth instar undergo the two remaining ecdyses and become the first adults of the season (Fig. 18: A). Third instar nymphs apparently undergo three more moults without further interruption and become adult from the middle of June to the beginning of July (Fig. 18: B). Nymphs which survive the winter as second instars require another four moults before attaining maturity. These moults are obviously influenced by environmental factors with the result that nymphs overwintering as second instars become adult over a longer period (Fig. 18: C). The proportion of nymphs overwintering in each instar probably varies from year to year, thus directly influencing the number of adults in each emergence peak.

A similar pattern is apparent in E. pallidus. The emergence period in this species extends for two months from the end of June until the end of August. Fig. 19 shows the frequency distribution of emergence dates for 177 adults. This species overwinters in four different instars; however, the relationship between the overwintering stage and the time of adult emergence is less clear. This may be due to the relatively small numbers of nymphs which were collected in February. The adult emergence of this species is characterised by one main peak (Fig. 19: B); this is the result of the high percentage of nymphs which overwinter as fourth instars. Only a very small proportion of nymphs hatch and reach fifth: ' instars before the onset of winter, and thus the first emergence peak is not clearly defined. E (i) a, 0) U 0 ( %To talPop ulation) FIG 19Adult Emergence ofaPopulation ofE.QgilLc___ 4- 9 - 6 5 8 - 3 7 - 2 OTir 1 - 4-A -4.4 22 27271217 2227161116212631 June

11 -

rl

B July

(11 —.4-- C-40

r lus Summer 1968. 11 ❑ n r August Female Male [1 111111 r -117-

Table 22 Relationship between the Overwintering Stage and Adult Emergence in E. pallidus.

Instars collected in field (February 1968)

5th. 4th. 3rd. 2nd.

Percentage of nymphs 4.4 82.6 8.7 4.4 collected

Percentage of adult 3.4 78.0 14.1 4.5 emergence

A B C D

Emergence peaks.of adults (Summer 1968)

It is interesting to note that the adult emergence of E. panzeri is more or less equally spread over a period from the end of July until the end of August without any obvious emergence peaks (Fig. 20). This is almost certainly due to the fact that the oothecae of this species hatch over a limited period and become adult without a nymphal arrest during development. 8

7

.0

6

❑ Male • 5 ❑ Female

4 nce e rg 3 Eme lt

Adu 2

ily Da 1

11 0 r r 21 26 31 5 10 15 20 25 30 4 9 July August Sept.

FIG 20 Adult Emergence of a Population of ff_p_aazi Summer 1968. -119-

IV. BIOLOGY OF THE EGG PARASITE, BRACHYGASTER MINUTUS (OLIVIER)

Materials and Methods. a. Collecting Methods.

Brachygaster minutus can either be collected as an immature stage within the oothecae of Ectobius, or as a free-living adult. The collection of parasitised oothecae was found to be very difficult. Samples of sand to a depth of 30 cms. were taken during the winter and early spring from the dunes at Studland Bay, Dorset. This sand was dried for 3 - 4 days- and later sifted through terylene net (11.5 meshes per cm.). The debris was sorted for the oothecae of E. panzeri and E. pallidus, some of which proved to be parasitised. This method yielded relatively few parasitised oothecae and was extremely laborious. Oothecae of E. lapponicus and E. pallidus were found in bracken litter and at the base of grass tussocks during the winter. The oothecae were kept until the adult parasites emerged in the summer. A more efficient method is the collection of adults from the field. Since adults could not be collected by sweeping low herbage, the only satisfactory method was the careful examination of the habitat for individual specimens which were collected from the ground by an aspirator. The most suitable habitat for this type of collection was found to be the interior of a large tuft of marram grass (A. arenaria). The afternoon or early evening is preferable for collecting, since the adults become more active at this time.

b. Culture Methods.

Parasitised oothecae were kept for their entire developmental period in 5.0 x 1.0 cms. glass tubes under identical conditions -120-

to unparasitised oothecae (Page 12; Fig. 2 ). Oothecae known be parasitised were fixed immediately after oviposition and at monthly intervals during the winter and spring using Bouin's fixative (Alcohlic). Adult parasites were kept in round, plastic cages (Fig. la), with a layer of fine sand covering the base. A moistened dental roll provided a supply of water, and several moistened raisins were given as food material (Peterson, 1953); these were replaced every few days. Folded filter paper afforded resting sites and shelter for the adults. Oviposition by the parasites was easily induced in the laboratory. Single oothecae were offered at intervals to mated females; after oviposition, which occurred readily, they were removed. This method resulted in a far higher percentage of parasitised oothecae than the alternative method, in which many oothecae were introduced into the cage at one time and were only removed when the female died. The latter method resulted in repeated oviposition by the female in a small number of oothecae. Roth & Willis (1954c) found that females of Tetrastichus hagenowii Ratzeburg behaved similarly. In this work adequate supplies of the parasite were maintained by introducing one ootheca per day to a female parasite. After mating the females were isolated, since other females attempt to prevent oviposition.

(1) Life History and Habits.

The parasite B. minutus has been recorded in this study from the oothecae of the three British species of Ectobius. It is solitary in habit; only one adult emerging from a parasitised ootheca. British records of this species are limited (Crosskey, 1951). However, specimens have been collected from the following areas in moderate numbers:- -121-

Locality. National Grfzl. Reference

Studland Bay, Isle of Purbeck, Dorset. • • • SZ037850

Canford Heath, Nr. Wimbourne Minster, Dorset. • • • SZ024955

Hengistbury Head, Nr. Christchurch, Hants. • • • SZ175908

Imperial College Field Station Silwood Park,

• • • sU947688 Berks. (North Gravel).

Crown Estate Plantations, Nr. Bracknell,

Berks. • • • su881664 su884662

From the above list it would appear that B. minutus is probably present in most localities which support large populations of a species of Ectobius. The paucity of records for the species may well be explained by the difficulties experienced in its collection. Adult parasites emerge from the oothecae during the summer months when the adult cockroaches are abundant. The adults of both sexes are agile creatures which become increasingly active in the afternoon and early evening, when mating and oviposition frequently occur. Their usual method of locomotion is to run swiftly over the substrate, but occasionally they climb nearby vegetation. The males fly sporadically for very short distances. Both sexes move with extreme rapidity when pursued and it is this which makes their collection so difficult. However, they become far less active towards the end of their adult life. The food consumed during the larval period is apparently adequate, since mating and oviposition will take place without the adult feeding. However, the longevity of both sexes in an outdoor insectary is reduced to a few days if no food is provided; -122-

whereas with food and water males live for an average of 21.2 2.4 days and females for 38.9 ± 2.7 days. The feeding habits of this species in the field are not known. However, related species have been reported to feed on certain flowers (Edmunds, 1954). B. minutus has been recorded on Asparagus officinalis Linnaeus (Crosskey, 1951). Fixation of parasitised oothecae at monthly intervals from oviposition until the emergence of the adult during the following summer revealed that the hatching of the egg and the early larval stages proceed rapidly; since at the end of one month a mature last instar larva has been formed. By this time the entire contents of the ootheca have been consumed. The parasite overwinters in this stage and does not undergo any further morphological change until a prepupa is formed in May or June.. The exarate pupa becomes darker in colour prior to emergence, the abdomen and antennae are the last structures to darken. Edmunds (1954) records three distinct stages in the development of the pupa of P. punctata, each based on the degree of pigmentation. Parasitised oothecae containing late larval instars can easily be distinguished from unparasitised oothecae when viewed by transmitted light. The outline of the larva or pupa can be seen, slightly withdrawn from the periphery of the ootheca unlike the developing cockroach embryos. Edmunds (1952a) observed a similar difference in oothecae parasitised by P. punctata. The oxygen uptake of oothecae parasitised by T. hagenowii was found to be lower than that recorded for oothecae containing normally developing cockroach embryos during the intermediate developmental stages (Watanbe (1953), unpublished observations in Roth & Willis (1954c)). An additional difference is apparent later in development, since before hatching normal oothecae of Ectobius increase in size and become pale in colour as a result of the uptake of water; whereas, parasitised oothecae do not change in shape or colour at this time. -123-

(ii) Mating.

The adults of the two sexes are similar, but can be readily distinguished by the colour of the front and mid tibiae which are black or dark brown in the male, but yellow in the female (Crosskey, 1951). Both sexes are capable of mating immediately after emergence from the ootheca, and may mate more than once. several males are confined in a cage with a single female, the males fight and often become damaged or even killed. Mating is usually preceded by the active flight of both sexes. The males and females then become stationary and pass their antennae in turn through their mouth parts. The male rubs the side of his abdomen with the hind legs, often raising his wings. The fore and mid, and mid and hind pairs of legs may also be rubbed together. The male actively pursues the female for about 30 seconds, the female eventually stops, usually on the roof of the cage, and turns to face the male. Antennal fencing begins and finally the male runs behind the female and climbs onto her back with his wings raised. The female abdomen is raised dorsally with the genitalia extruded, and the exposed male genitalia grips those of the female. The act of copulation lasts from 13 - 20 seconds, during which the male continually strokes the antennae of the female. The male frequently loses contact with the roof of the cage and hangs suspended from the female. At the completion of mating the male regains his hold on the roof and walks away. After mating the female often becomes aggressive towards the male. In the field mating probably takes place on the vegetation, since in every case observed in the laboratory (approximately 30) mating occurred on the roof of. the co3e. Virgin females of B. minutus readily oviposit when oothecae are available. The eggs develop and adults emerge the following year, all the progeny are males. Edmunds (1954) found P. punctata could reproduce parthenogenetically and that aimilarly all the offspring were males (arrhenotoky). -124-

(iii) Oviposition.

Females of B. minutus oviposit readily in oothecae of the three species of Ectobius without showing any obvious preference. Oviposition takes place immediately after emergence if the females are unmated, or after mating has occurred. They oviposit in either freshly deposited oothecae or in oothecae which have become completely hardened. However, the parasites show no interest in oothecae which are still being carried by a female cockroach. Oothecae which have been buried by a cockroach after deposition are frequently uncovered by the parasite, thus suggesting that the location of an ootheca may be in response to an olfactory stimulus. The female parasite approaches the ootheca with rapidly vibrating antennae (Plate 12a); this is followed' by a thorough exploration of the ootheca using the antennae which is continued for several minutes, the female often walking away and returning to the ootheca. The parasite then climbs onto the ootheca, with antennae still repeatedly vibrating (Plate 12b), and walks over the ootheca several times in an attempt to balance it (Plate 12c). Finally the female assumes a position parallel to the longitudinal axis of the ootheca, the head is bent forwards and the antennae are flexed over the end of the ootheca and slowly vibrated. The legs are apparently used to anchor the female onto the ootheca, the fore legs are held over the end of the ootheca. The left legs are braced against the keel or seam, whilst the right legs merely rest on the surface of the ootheca. The abdomen is then lowered and the ovipositor extruded; after some difficulty this penetrates the oothecal wall and the chorion of the cockroach egg. The female remains in this position for 20 - 1+5 minutes (Plate 12d) and then withdraws her ovipositor and walks away (Plate 12e). During oviposition the ootheca may move slightly; however, oviposition continues without interruption, the female often lying on her side. In nearly every case the female returns to the ootheca and

-125-

a. b.

i

C. d.

e. f. Plate 12 Oviposition by the parasite, Brachygaster minutus. -126-

attempts to bury it in the sand (Plate 12f). The fore and mid are used to scrape the sand over the ootheca whilst the hind legs often remain on the ootheca to steady it. The female frequently changes her position so that all sides of the ootheca are equally covered. The sand is finally built up into a dome 2 - 3 ems. in diameter with the ootheca in the centre; this may take upto 15 minutes. The orientation of the ootheca before oviposition is importanc— A female will make continual efforts to roll an ootheca, which has been placed with the keel upwards, onto its side. On one particular occasion a female parasite carried an ootheca for about 3 cms. between its legs before depositing it on its side and commencing oviposition. The head is often used to reorientate an ootheca prior to oviposition, or a female may run repeatedly over an ootheca which frequently rocks it into a more suitable position. If several females are enclosed in a cage in which a female is engaged in oviposition, the other females will make repeated attempts to disturb her and often move the ootheca by using their heads. A single spindle-shaped egg is laid within a cockroach egg at each oviposition, and is orientated to lie parallel to the longitudinal axis of the latter. The large size of the egg may well account for the long duration of oviposition. One female is capable of parasitising many oothecae. Two females occasionally oviposit into different eggs in the same ootheca, but since only one larva develops additional eggs are presumably consumed together with the cockroach eggs.

(iv) Emergence.

The solitary adult parasites emerge from the ootheca by chewing single emergence holes in.the oothecal wall. Prior to emergence the movements of the adult can be seen within the ootheca. -127-

The ootheca is punctured by the mouth parts and a small quantity of moisture is released, the small hole is gradually enlarged by the parasite biting pieces from the margins of the aperture. It is possible that the secretion may contain enzymes which soften the wall to facilitate this. The head and thorax of the parasite can be seen as the hole becomes larger, and frequent attempts are made to escape through the hole, by the entire body being expanded and pushed forwards. The systematic enlargement of the hole is continued until the fore legs emerge, followed by the rest of the body. On emergence the body of the parasite is dry and deeply pigmented and the wings are fully expanded. The emergence hole is approximately 1.25 cms. in diameter and is always situated at one end of the ootheca in a ventro- lateral position under the keel. The interior of an ootheca from which a parasite has emerged is devoid of the internal lattice structure formed by the individual membranes of the cockroach eggs. All that remains after emergence are the cast larval and pupal skins and a mass of grey material in the ventral part of the ootheca, thought by Edmunds (1954) to be an accumulation of excretory material from the larval stages. In some cases one or two cockroach embryos are not consumed by the parasite, but these always fail to hatch. This is presumably due to the fact that the hatching of an ootheca is dependent upon the combined effort of all the embryos to rupture the seam. The parasites of the three species emerge at different times during the season when the adult cockroaches are in the field. The relationship between the emergence of the parasite and that of the adult cockroaches can be seen from Table 23 overleaf. -128-

Table 23 Relationship between the emergence of the wasite and that of the adult cockroach.

Host Emergence period Emergence period Species of B. minutus of adult cockroaches (Figs. 18-20)

E. lapponicus 22nd June - 10th Aug. 24th Nay - 10th Aug.

E. pallidus 7th July - 29th July 30th June - 26th Aug.

E. panzeri 5th Aug. - 23rd Aug. 20th July - 6th Sept.

The eotergence of the parasite from the oothecae of the three species is completely synchronised with the occurrence of the adult cockroaches in the field. The first parasites emerge after the early adults have been in the field for some time and have commenced deposition of oothecae. The small number of parasitised oothecae of E. pallidus and E. panzeri involved may possibly present a reduced period for the emergence of these species. The mechanism involved in this synchronisation is not certain, but the results are absolute; if parasites collected at,the end of the season oviposit in oothecae of the three species, the emergence of the adults the following summer will invariably fall within the range for that species given in Table 23. From a total of 44 parasites which emerged in one season the ratio of males to females was exactly equal. A considerable range in the size of the adult parasite was noted. Generally males tend to be slightly larger than females. -129-

However, the major size difference seems to be affected by the type of ootheca from which it emerged. Oothecae of E. panzeri consistently produce smaller parasites than the other two species; the largest adults emerging from oothecae of E. lapponicus. The size of the adults of Anastatus floridanus Roth & Willis, parasites of the oothecae of E. floridana and T. hagenowii a parasite of several cockroach species was found.to be correlated with the available food material (Roth & Willis, 1954 b & c). The oothecae of the three British species of Ectobius vary in size (Table 7 ) and since the entire contents of the oothecae are consumed by the developing parasite larvae, the quantity of available food material is therefore different. It is probably this difference in the amount of food material coupled with the size of the ootheca itself which governs the size of the parasites on emergence. -130-

GENERAL DISCUSSION.

This may be organised around six main points:-

(i) The Occurrence of a Nymphal Diapause as a Means of Synchronisi.lg the Life Cycle.

Species which have life histories extending over a long period require some means whereby the emergence of the reproductive stages is synchronised. The total duration of the life cycle of E. lapponicus and E. pallidus is two years, of which one year is spent as a nymph. The intervention of a diapause in the period of nymphal development prevents extensive variability in the emergence of the adults, which is, as a result, synchronised with the onset of favourable conditions. The nymphal diapause, which has been confirmed in E. lapponicus, causes development to cease at the fourth nymphal instar which then overw:Lnters in a state of diapause. Diapause in this instar is facultative, since it is induced in the summer and autumn, but averted in spring. Diapause in the fourth instar may intervene relatively early in the season. However, earlier instars continue to moult after many fourth instar nymphs have terminated morphogenesis, thus increasing the proportion of nymphs at the penultimate juvenile stage. Much of the variation accumulated during the period of nymphal development is thereby reduced. A more uniform adult emergence would be achieved if the majority of nymphs overwintered as fourth instars. Since the number of nymphs reaching this stage varies according to the season, it would be of interest to investigate the overwintering behaviour and adult emergence of this species in other localities. E. pallidus has a more precisely synchronised adult emergence; W.() of the season's adults emerging over a period of three weeks, a direct result of the high proportion of nymphs overwintering in a single -131-

instar. It is interesting that Corbet (1955) similarly found that the adult emergence of A. imperator was bimodal and was a direct result of the overwintering of this species in two instars.

(ii) The Possible Influence of Changing Photoperiods on the Induction and Termination of Nymphal Diapause.

The possible influence of photoperiod on the induction and termination of nymphal diapause in E. lapponicus is a subject which warrants further study. Much of the work to date in this field has been confined to multivoltine species. However, photoperiod has been found to have a marked effect on the growth and moulting of this species and E. pallidus. Nymphs of E. lapponicus may stop moulting in early August when the temperatures are similar to those which enhance active growth in the spring. Thus decreasing photo- period may have an influence on the induction of diapause. It is feasible that the same mechanism recorded in A. imperator (Corbet, 1956) is important in the nymphs of E. lapponicus, whereby the gradual decrease in the daily photoperiod is the factor which induces diapause, at least in the fourth instar. Results indicate that the response to photoperiod of the third and fourth instar varies. It may well be that the induction of diapause in the three overwintering instars may be brought about by different mechanisms, the effect of photoperiod being relevant only to the fourth instar. It is also possible that increasing photoperiod is responsible for the resumption of active growth and moulting in the spring. The onset of growth after diapause is governed by photoperiod in N. yezoensis (Masaki & Oyama, 1963), and is also the major factor in the termination of diapause in the nymphs of C. viridifasciata (Halliburton & Alexander, 1964). The real significance of photoperiod in E. lapponicus is not entirely clear, but since diapause development is apparently complete by February the effect of the short photoperiod probably prevents the nymphs -132-

moulting as a result of sporadic rises in temperature. Moulting may be resumed in response to increasing photoperiods over a certain threshold. Increasing photoperiods may therefore cause diapause to be averted in nymphs becoming fourth instars in the spring, i.e. those which overwinter as second or third instar nymphs. Thus changing photoperiod may act in E. lapponicus as a seasonal indicator, which affords a more reliable signal than absolute values of photoperiod or other environmental factors, such as temperature, which are less closely correlated with time of year.

(iii)The Occurrence of a Limited Hatchills Period as a Result of an Obligatory Diapause in the Egg.

The presence of an obligatory diapause in the oothecae of E. lapponicus results in a very restricted hatching period in the field, when conditions are favourable for the growth of the young instars. Thus the effect of a long period of ootheca deposition in the preceding summer is completely erased. The limited hatching period recorded in E. pallidus and E. panzeri strongly suggests a similar obligate diapause in the oothecae of these species1 however, this requires confirmation.

(iv) The Grouping of Cockroach Species according to the Water Content of the Oothecae at Oviposition and their subsequent Water Relations during Embryogenesis.

The grouping of cockroaches according to their ovipositional behaviour was first considered by Shelford (1906). This, however, was extended by Roth & Willis (1955c), who subdivided the oviparous species into three types dependent on the period of retention of the ootheca by the female. The second oviposition type, referring to species which retain their oothecae for periods in excess of one or two days, had E. pallidus and E. panzeri as the only examples. By a consideration of the period of retention of the oothecae of -133--

these species no evidence for this separate oviposition type was found. However, it appears that a preferable subdivision of the oviparous cockroaches is based on the water relations of the oothecae (Roth & Willis, 1958b). Some oothecae contain enough water at the time of deposition for their complete development, and require no additional water from outside the ootheca, although water may be transferred from the moist inside wall of the ootheca to the eggs during embryogenesis, eig. B. orientalis, S. supellectilium and P. americana (Roth & Willis, 1955c). Other species such as Blatella vaga Hebard and B. germanica carry their oothecae externally until just before hatching, and these acquire additional water from the body of the female during embryonic development, although the oothecae of the latter species contain sufficient water after formation to permit complete embryogenesis. Oothecae of these species contain approximately 60 - 65% water immediately after formation (Roth & Willis, 1955c). However, the oothecae of other species e.g. Agmoblatta thaxteri (Hebard), Amazonina sp., Lophoblatta sp. and three species of Cariblatta contain a relatively low percentage of water at deposition (approximately 40%), (Roth, 1967b). Included in this group are the oothecae of E. lapponicus with an initial water content of 37% (and E. pallidus which contains 40% water at deposition (Roth & Willis, 1958a)). The oothecae of all these species require water from the surroundings for the completion of development. If the oothecae of E. lapponicus are kept dry they do not hatch, but eventually die before embryogenesis is complete. These and similar species, which require additional water from the surroundings during development,constitute a separate ovipositional type. There are two separate methods of water uptake in these species. The former group of oothecae are deposited in moist surroundings, and acquire water gradually during development until hatching when they contain approximately 70% water (Roth, 1967b). -134-

The diapausing oothecae of Ectobius have a similar water content before hatching but acquire it suddenly after the completion of diapause development. Even if the oothecae of E. lapponicus are kept moist throughout the winter they absorb little water until a short period preceding hatching. The situations where oothecae of the three species of Ectobius are deposited ensure that moisture is available when required. Although the structure of only a few oothecae has been examined in detail, it is of interest that the chorionic expansions of the eggs, referred to as "white bodies" by Lawson (1951) have only been recorded in oothecae found by Roth & Willis (1958b) to contain insufficient water at oviposition e.g. C. lutea lutea (Lawson, 1952), P. uhleriana (Lawson, 1954) and P. pensylvanica and B. germanica (Lawson, 1951) and the three species of Ectobius considered in this work. All these species, with the exception of B. germanica, obtain additional water from the surroundings. However, the oothecae of B. orientalis, P. americana and S. supellectilium do not have "white bodies" (Lawson, 1951) neither do they require additional water from outside the ootheca after deposition (Roth & Willis, 1955c). It would be of value to investigate any further correlation between the presence of "white bodies" and the uptake of water after oviposition by oothecae and to examine the fine structure and functional significance of these structures.

(v) The Absolute Synchronisation of the Life Histories of the Parasite and the Host Species.

Cameron (195?) suggests that all Evaniids breed more rapidly than their hosts. This is apparently true for the species which parasitise the domestic cockroaches; P. punctata a parasite of the oothecae of B. orientalis and P. americana has three generations a year (Edmunds, 1952a),whereas two or three annual generations have been recorded by Cros (1942) from the oothecae of the former species. E. appendigaster,a parasite of the oothecae of -135-

P. americana, has three or sometimes four generations a year (Cameron, 1957). However, Hyptia species, which parasitise the oothecae of the native wood roaches in North America, have only a single generation a year (Edmunds, 1954). B. minutus, which has been found to parasitise the egg capsules of the three British species of Ectobius is also restricted to a single generation a year. In domestic species of cockroaches there is a constant supply of adults and thus the synchronisation of the life cycle of the parasite with that of the host, brought about by the reduction in the number of generations a year, is not necessary. The overwintering of the parasite at a fixed developmental stage results in the production of only a single generation a year. B. minutus overwinters as the last larval instar in the oothecae of the three species of Ectobius. It is interesting that Hyptia thoracica (Blanchard) and aptia harwroides Bradley overwinter in the last larval instar in North America in the oothecae of Epensylvanica and T. virginica respectively (Edmunds, 1954). Diapause in the host is frequently accompanied by a corresponding arrest in growth of the parasite (Andrewartha & Birch, 1954). The development of the parasite is controlled by the host; the parasite remaining dormant until reactivated by a specific physiological event. Such a stimulus may occur during the post-diapause growth of the egg (Lees, 1955), as recorded in Scelio chortoicetes Froggatt, a parasite of the eggs of Austroicetes cruciata Saussure (Birch, 1945). The mechanism ensures the synchronisation of the life cycles of the parasite and its host. However, in B. minutus the arrest in development occurs after larval nutrition is complete and all the cockroach eggs have been consumed. It thus seems unlikely that the host can control the development of the parasite. The overwintering behaviour of this species requires much further work. This may well serve to clarify the interesting synchronisation in the time of emergence which occurs when the species is a parasite of the oothecae of different species of Ectobius. This feature is a valuable adaptation in the life -136- cycle of the parasite, since most habitats support populations of only one host species.

(vi) Factors to be considered in the Extension of this Work.

An examination of certain aspects of the biology of the three British species of the genus Ectobius has revealed many fields where further work is required. Some of the difficulties experienced in a study of these native field-dwelling species should perhaps be emphasised in this context. The culture of all species proved difficult initially, but this was later overcome and all stages were successfully maintained on a specially developed artificial diet. The collection of large numbers of any developmental stage proved to be very time consuming, especially in the winter months. No suitable method of trapping live insects has become obvious during this work. The optimum conditions for diapause development in the nymphs of E. lapponicus and E._pallidus and the oothecae of the three species need to be established before the continuous culture of these species in the laboratory is possible, a feature which would obviate the necessity for repeated field collections. The study of diapause in the oothecae of the three species requires vast numbers of adult females; since so few oothecae are deposited by a single female. However, several topics show such great potential for rewarding future work, that the difficulties experienced in the collection and culture of the species are undoubtedly worthwhile. -137-

ST2.!ARY OF THE BIOLOGY SECTION.

1. The life cycles of the three British species of Ectobius differ; one species, E. panzeri, is univoltine whereas the other two species, E. lapponicus and E. pallidus, require two years to complete their development.

2. Both sexes of E. lapponicus and E. panzeri have five nymphal instars, an additional instar occurs in both sexes of E. pallidus.

3. E. lapponicus and E. pallidus are both facultatively partheno- genetic, although this mode of reproduction results in a reduction in the fertility of the species. Parthenogenesis is thelyotokous. Some unfertilised eggs of E. panzeri develop but do not hatch.

4. Differences in certain aspects of the oviposition of the three species occur between mated and virgin females. The preoviposition period and the interval between the deposition of oothecae is extended in unmated females, whilst the period of retention of oothecae is increased by the presence of a male.

5. Oothecae laid by mated females contain more eggs and are heavier than those laid by unmated females. The shape of the oothecae also varies; those deposited by mated females are shorter and wider. Mating apparently affects the direction of rotation of the oothecae. A species difference is evident in the method of deposition of the oothecae.

6. The longevity of virgin females is greater than that of mated females although fewer oothecae are deposited.

7. The oothecae of the three species are basically similar in structure; however, a key based on several external features -138- enables them to be distinguished.

8. The oothecae overwinter and hatch the following spring. An obligate diapause exists in the oothecae of E. lapponicus. Morphogenesis is halted at the post-anatrepsis stage. The restricted hatching period also suggests the presence of a diapause in the other two species.

9. The oothecae of these three species of Ectobius increase markedly in size prior to hatching. This is the result of water uptake during the final stages of embryogenesis.

10. Nymphs of E. lapponicus and E. pallidus overwinter in a range of instars, although post-embryonic development always ceases before the moult into the last instar. This dormancy, which is a diapause in E. lapponicus, differs in intensity in different instars. The effect of photoperiod is apparently an important factor in the induction and termination of diapause in E. lapponicus.

11. An abnormal fifth instar is produced in E. lapponicus when nymphs are subjected to conditions of a high temperature and a 16 hr. per 24 hr. photoperiod. These abnormal nymphs never attain maturity.

12. Moulting of all the nymphal instars is inhibited by a 10 hr. per 24 hr. photoperiod, although oothecae hatch readily under these conditions.

13. The emergence pattern of the adults of E. lapponicus and E. pallidus is correlated with the relative proportions of nymphs overwintering in each instar.

14. An Evaniid, B. minutus, parasitises the oothecae of the three species. A female parasite lays a single egg within a cockroach -139- egg. During development the larva consumes the entire contents of the ootheca. The parasite overwinters in the ootheca as a last larval instar; the emergence of the adult the following summer is synchronised with the occu.7.rence in the field of the adult cockroaches. -14o-

SECTION B: THE GROWTH OF ECTOBIUS SPECIES.

Introduction and Review of Literature.

There are few detailed morphological accounts of the nymphal instars of cockroaches. The immature stages of B. orientalis and P. americana have been generally described by Qadri (1938) and Griffiths & Tauber (1942b) respectively, and developmental changes in the genital segments of a few species by Gupta (1948) and Qadri (1940).

Dyar's Law. The first truly quantitative assessment of growth in the Insecta was made by Dyar (1890) who discovered a relationship between the head capsule width of successive instars of Lepidopterous larvae. In 28 species he found that the width of the larval head capsule followed a regular geometric progression; the ratio of the width in an instar to that in the following instar being constant throughout development. This observation has proved to be valid for many species of Lepidoptera and is known as "Dyar's Law" or "Dyar's Rule". It was suggested that any deviation from the geometric progression might indicate an overlooked stage in a developmental study. Przibram & Megusar (1912), weighing the exuviae and the bodies of freshly moulted specimens of Sphodromantis bioculata (Burmeister) (= S. viridis (Forskal) concluded that both the cast skins and the bodies doubled their weight from one moult to the next. They also found that the increase in length of the pronotum was constant and equal to the cube root of two, and that the area increased as the square root of two. These factors for the increase in weight, length and area were interpreted as being caused by a single division of cells. This suggestion was confirmed by Sztern (1914), -141-

who counted the number of cells in the epidermis of S. bioculata and found that they doubled in number between two successive moults. Much variation was found to occur in these growth coefficients or progression factors, and Bodenheimer (1927) suggested that when the rate of increase exceeded the standard value, more than one cell division may have occurred during the instar. Such additional divisions were termed "latent divisions". According to Bodenheimer (1933) the number of latent divisions increases in holometabolous insects, but although this may obscure, it does not invalidate, the progression factors suggested by Przibram & Megusar (1912). The importance of such constant progression factors has been disputed by many authors (Titschack (1926); Calvert (1929); Ludwig (1934) and Harries & Henderson (1938)). However, Duarte (1938) applying Przibram's Rule with Bodenheimer's modification, to the two phases of L. migratoria migratorioides, found that it was of value in describing the growth in length of various structures, and that latent cell divisions did occur in this species. An attempt has been made in the present work to use latent divisions to explain cases when the progression factor is above the standard value suggested by Przibram. Much of the early work on Dyar's Law was confined to Lepidopterous larvae. However, Taylor (1931) considered that Dyar's Rule may be applied successfully to other groups of insects, provided that the same growth phenomena exist, i.e. that size increases only at ecdyses and remains static between them. Miles (1931) and Taylor (1931) have considered the application of this law to Tenthredinid larvae and have found, that with the exception of the non-feeding prepupal instar, the head capsule widths do follow a geometric progression. Sen & Das Gupta (1958) found that the larval instars of Culex fatigans Wiedemann followed a very precise geometric series, and that the law was invaluable in separating the larval instars. Metcalfe (1932), however, working on the Anobiid beetle, Sitodrepa (= Stegobium) paniceum Linnaeus, failed to find any close approximation to the law. The law has not -142- been used to any extent in Orthopteroid insects in the past; this is probably due to the relative ease with which the instars can be distinguished. Bednarz (1955) reported a regular geometric pro- gression throughout development in Tettigonia viridissima Linnaeus. In the Blattidae, Qadri (1938) noted that the width of the head capsule of B. orientalis follows a geometric progression only upto the third instar; however, more recently Manley (1969) found a constant growth rate existed in the linear measurement of three structures of Blaberus discoidalis Serville, although the ratio between structures varied. Initially, Dyar (1890) chose the head as his organ of study, since it was not subject to growth between ecdyses, and he selected the width as the easiest dimension to measure. Most of the subsequent work has been devoted solely to a consideration of the head capsule, however, Bodenheimer (1927) included other characters in his study. Duarte (1938) concluded that the length of the mid femur of L. migratoria migratorioides follows a regular geometric progression. In the present study, of two closely related species, ratios have been calculated for a total of 74 characters for each species, and the extent to which Dyar's Law is applicable has received consideration. Such an extensive analysis is made practicable by the use of an electronic digital computer. Many authors have discovered various irregularities in the series of ratios obtained in the development of a species. Perhaps the most common departure is the gradual decrease in the ratio during development. Edwards (1961+) adequately illustrates this decrease in a series of ratios obtained from the head widths of the twelve larval instars in Hepialus humuli (Linnaeus). A similar trend has been noted by Forbes (1934) in Agrotis ypsilon (von Rottenburg) and.by Gaines & Campbell (1935) in Heliothis obsoleta Fabricius (= H. armigera (Hubner)). Ghent (1956) points out that the non-feeding last instar should be omitted from growth studies of sawflies, since little growth occurs during this instar. Miles (1931) found that sex differentiationiin the later instars in -143- sawflies rendered larval growth irregular. The irregularities met by Metcalfe (1932) in the larval development of S. paniceum may well be explained by the early differentiation of the sexes causing two separate geometric progressions to be formed, thus giving a series of overlapping peaks in a frequency distribution. Keler (1934) concluded that Dyar's Law is applicable to males but not to females in Porthetria (= Lymantria) dispar Linnaeus. Slama & Janda (1960) and Ghent (1956), working with sawflies, list the dangers of the indiscriminate use of Dyar's Law, the most salient error being the application of Dyar's Law, involving a geometric progression, to data which form a linear progression. From work on sawflies by several authors, two forms of growth of the head capsule have become evident. One type confirms the validity of Dyar's Law (Miles (1931); Taylor (1931)), whereas the other indicates linear growth (Ghent (1956); Friend (1933)). The latter involves the addition of a constant amount at each ecdysis, and therefore gives a declining series of ratios during development. Growth in such species is more accurately described by a straight line than by an exponential curve, when a dimension is plotted against time. When the logarithms of the head widths are plotted against the instar number, the result is a straight line (Forbes, 1934), if growth is exponential. Any deviations from the line indicate a departure from the geometric progression. Ludwig & Abercrombie (1940) found that the growth ratios for the length and width of the head capsule of P. japonica, decrease at each stage, and concluded that the growth of the head can be more accurately described by a second degree polynomial of the form: log y = a bx cx2 than by the straight line: log y = a bx The presence of linear growth has been considered in this study in both species. It was Schedl (1934) who first suggested that irregularities -144- in larval growth in Neodiprion Rohwer species were probably due to the length of the instar and not to sexual differentiation. This idea was investigated further by Richards (1949), who noticed that a regular relation between the measurements of the instars did exist if account was taken of their different durations. He concluded that Dyar's Law only applies when all the instars are of the same duration, since more growth would occur in a longer instar causing a deviation from the regular series of increments. A suggestion is made that the principle is applicable to other groups. However, no attempt was made to compare the modification with the original form used by Dyar, the author merely drawing linear regression lines for various characters, against accumulated days of larval duration. A fuller assessment of Richards' modification has been made using the whole range of characters measured, and a means derived to test its validity in two soecies of Ectobius.

Allometric Analysis of Growth. All mature organisms are the result of differential growth during their development. This has been studied in many animal phyla. However, an attempt has been made to draw most of the examples from the Insecta in this account. The fundamental phenomenon of differential growth was illustrated by Thompson (1917) as a system of Cartesian Transformations which merely presented a qualitative picture, and neglected any gradual changes in relative proportions with absolute size. The first quantitative analysis of differential growth was made by Huxley (1924), who used the formula: y = bx to describe the relation of the size of a structure (y) to that of the whole body (x). This formula, often termed the "Law of Simple Lllometry", has been used extensively. The basic assumption in this law is that the ratio of the geometric growth rates of the two dimensions remains constant (Huxley, 1932). In this formula, x is the measurement of the whole body or standard organ, y is the measurement of the differentially growing organ and b and a are _145_

constants. The constant b represents the value of y when x is equal to unity, and is known as the initial growth index. The equilibrium constant a is the ratio of the geometric growth rates. The biological interpretation of the parameters has received much attention. The constant a is the most important term in the formula; it is a measurement of the ratio of the growth rates of x and y. It is a pure number without dimensions and its value can be compared directly in different samples (Kermack, 1954). The biological meaning of the initial growth index b is, however, not yet clear. The work on this subject is reviewed by Gould (1966), who concludes that the biological interpretation of this parameter is limited to instances when the values of a are equal; values of b can then be compared at a common value of x, e.g. at the onset of allometric growth. It is therefore a measure of the initial size of the organism, irrespective of changes in shape due to allometric growth (Kermack, 1954). The relationship between the equilibrium constant a and the initial growth index b has been studied mainly by Matsuda (1961a, b, c, 1961d, 1962a, 1962b, 1963b) who.proposed a series of hypotheses, which are tested for numerous species, mostly Hemiptera: Heteroptera. Earlier workers (e.g. Hersh, 1931), however, found that the relation between a and b is an inverse one, and that b is approximately a decreasing exponential function of a . Much confusion has arisen over the numerous different terminologies which have evolved to describe differential growth. However, Huxley & Teissier (1936) suggested a satisfactory terminology, which is now generally accepted, and has been used throughout this work. Two structures exhibit allometry if the growth of one structure is more, or less,rapid than that of the standard organ, and the ratio between the two dimensions a changes constantly in accordance with the law (Huxley, 1932). If the constant is more than unity, allometry is positive, if less than unity it is negative. If, however, a is equal to unity, the two structures will have the -.146- same growth rate and a constant ratio exists between them irrespective of their absolute size. In this case the structures are said to exhibit isometry. Teissier (1960) uses the terms allometry of growth, when the specimens compared belong to successive stages in development, and allometry of size, when the specimens compared are individuals of different size at some particular stage in development. Both allometry of growth and of size are to be considered in this work, but the emphasis is on allometry of growth. Matsuda (1963a) discusses the evolution of relative growth in the Arthropoda. The choice of a reference dimension is difficult. The total length of the body is one which is frequently used, since it has an obvious biological meaning. However, the use of this dimension is probably one of the main sources of error. Clarke (1957) showed that linear size increased,, both within a stadium and at ecdysis, in L. migratoria migratorioides if the structure was composed of sclerites and intersegmental membranes, whereas simple sclerotic plates increase in size only at ecdysis. Attempts to overcome this difficulty include the selection of the elytral length in the Cerambycid beetles, Monochamus notatus (Drury) and Monochamus scutellatus (Say) as a reference character (Gardiner, 1954), the head width in several species of Coreoidea (Kumar, 1966) and in two species of Orthotylus Fieber (Matsuda, 1963b),,the length of the hind femur in P. surinamensis (Matsuda, 1962c), whilst the total body length is used by many authors e.g. Buxton (1938), Clark & Hersh (1939) and Matsuda (1963a). Consideration is given to the selection of a suitable reference character in the description of the allometric growth in two species of Ectobius. The allometry formula can equally well be written in the form: log y = log b f a log x indicating that when the logarithms of x and y are plotted against each other, the points will lie on a straight line, if the law adequately describes the data. The constant a being given by the slope of the line. The line is either fitted by eye or is a -147- regression line calculated by the method of least squares (Feldstein & Hersh, 1935). Kermack (1954), Kermack & Haldane (1950) and Teissier (1948) discuss the inadequacies of these methods and introduce more satisfactory ones. In this study much consideration has been given to the means by which a and b can be calculated, and a comparison is made between two methods. Deviations from simple allometry are of frequent occurrence. There are three main forms of departure; a progressive change in growth ratio, an abrupt change in growth ratio at a certain stage in the life history, or a rhythmical fluctuation in ratio about an average trend (Reeve & Huxley, 1945). A gradual change in growth ratio can be shown by a graph of a against time. The growth in head length relative to body length in Notonecta undulata Say is an example of a gradual change in a (Clark & Hersh, 1939), illustrated in this manner by Simpson, Roe & Lewontin (1960). An abrupt change in the growth ratio may result in either the intersection or separation,by a large or small discontinuity, of the regression lines on a logarithmic plot. Needham (1937) records evidence of rhythmic fluctuations in the thoracic and abdominal segments of Asellus aquaticus Linnaeus, and suggests that they are due to cyclic phenomena in absolute growth. Deviations from a constant a in consecutive segments of the body or an appendage are usually related to each other in such a way that, when the values of a are plotted against a series of body segments, a growth gradient will result. The highest point of the gradient is known as the growth centre (Huxley, 1932). The gradient will be directly influenced by the values of a . Huxley (1932) uses the term "growth profile" for a plot of growth gradients along a structure in two planes e.g. dorsal and ventral. Such gradients have been cited by Needham (1937) in A. aquaticus and by Blackith, Davies & Moy (1963) in Dysdercus fasciatus Signoret. Growth gradients in the appendages also occur; Clark & Hersh (1939) record gradients in the legs of N. undulata, although Reid (1942) found no evidence of growth gradients in the antennae -148- of Laemophloeus turcicus Grouvelle. However, Clark & Hersh (1939) point out that the growth gradients plotted for the mean of a number of individuals may give a misleading picture, since insects examined separately give different gradients. Variation in the parameter a may occur throughout development, and Needham (1937) constructed growth contours in A. aquaticus to give an overall picture of the variation in a along the body and in time. Examination of growth gradients and growth contours for both sexes in the two species of Ectobius is included in this work. The study of allometry has diverged into many fields apart from the basic study of growth phenomena. It is probably in taxonomic problems that allometry is of most value. Species otherwise unidentifiable can be separated by a comparison of their equilibrium constant,a , (Gardiner, 1954). Buxton (1938) found that instars could be separated satisfactorily in Pediculus Linnaeus. Both Boratynski (1952) and Johnson (1939) illustrate how differential growth invalidates certain ratios used in taxonomy. The relation between allometry and geographical variation is discussed fully by Petersen (1952). The use of an electronic digital computer has made possible a far more comprehensive and detailed analysis of allometry than has been possible before. Broader and more useful comparisons than are usually made are included in this work, since data is available for two species and two sexes.

Multivariate Analysis of Growth. The application of multivariate statistical methods to the study of animal growth is still at a very early stage, though Gould (1966) in a survey of the various methods has emphasised their great potentialities in the analysis of growth data. So far these techniques have involved only a few insect species though they have been used to analyse the following: adult variation in Coccus hesperidum Linnaeus (Blair, Blackith & Boratyriski, 1964) and -14-9- in Pemphiqus populi-transversus Riley (Sokal, 1962); shape variation in the cocoons of Bombyx mori (Linnaeus) (Fraisse & Arnoux, 1954); adult polymorphism in Australian locusts and grasshoppers (Blackith, 1957), migratory locusts (Blackith, 1962), British grasshoppers.(Blackith & Roberts, 1958), the Red Locust (Blackith & Albrecht, 1959) and social wasps (Blackith, 1958); evolutionary patterns in Kalotermes Hagen (Stroud, 1953); taxonomic relations of the Hoplitis Klug complex of bees (Rohlf & Sokal, 1962) and of some other insect groups and allometry in Trepobates trepidus Drake & Harris (Matsuda & Pont, 1961). Most of these studies include only the adult stage, though Blackith (1960) suggests that.the growth patterns occurring in the adult are the direct result, and therefore expressions, of those patterns which were present during the immature stages. However, the only application of multivariate techniques to the successive developmental stages of an insect (D. fasciatus) is due to Blackith, Davies & Moy (1963). In no case has any attempt been made to apply several of the available methods to a wide selection of characters. The primary purpose of the work described here, is, therefore, to examine the extent to which four distinct methods of multivariate analysis can be applied to a wide range of characters (74 in this study) and to learn something of the behaviour of such data under analysis. Multivariate statistical methods are essentially concerned with the analysis of a large number of correlated variables (in this case dimensions of parts of the body). Their aim is to investigate the extent to which a pattern of correlation can be expressed in terms of a much smaller number of orthogonal, i.e. non-correlated, dimensions. These serve not only to simplify greatly the complex interrelationships of the original data, but also to define a set of independent patterns of growth or size variation which, in favourable oozes, may be identified in two distinct ways:- (i) By the extent to which some or all of the characters -150-

participate in determining the pattern. (ii) By the extent to which the different patterns are manifested by the various biological entities analysed (species, sexes and instars in this study). Such methods appear well suited to understanding growth, not as an unconnected set of changes in size, but as an integrated complex of developmental patterns. The available multivariate methods are described in varying degrees of detail by Rao (1952), Anderson (1958), Cooley & Lohnes (1962), Seal (1964) and Morrison (1967). In general the structure of the correlation or covariance matrices established between the different variables is analysed by the extraction of a series of latent roots and associated vectors, each representing an orthogonal component and each capable of being interpreted biologically. This interpretation follows three main lines:- (i) The relative magnitudes of the latent roots allows one to determine the proportion of the total variance accounted for by each orthogonal component or factor. (ii) The relative sizes of the elements of each vector (each element corresponding to a variable) indicates the weight to be attached to that variable in determining a linear function of the variables. (iii) The vectors may be used to compute some form of component or discriminant score which characterises the particular stage, sex and species from which the original dimensions were obtained. These scores may then be used to compare the stages, sex and species in terms of a few orthogonal components or factors. Four multivariate methods are considered in this thesis, which although new to the field of insect growth have been used previously to varying degrees in other biological or psychological fields where data involving many correlated variables need to be reduced in complexity and interpreted physically. -151-

(i) Eigenanalysis of Correlation Matrices. The analysis is applied to separate correlation matrices for each developmental stage, sex and species. This method has sometimes been referred to as an analysis of growth patterns (Blackith, 1960); but it is really an.analysis of size variation within a single developmental stage, e.g. the adult female of C. hesperidum (Blair, Blackith & Boratynski, 1964). The primary purpose of this analysis is to examine the extent to which common patterns of size variation can be detected throughout the successive developmental stages of the two sexes and species. Such intra-stadial patterns of size variation can then be given a true ontogenetic meaning. (ii) Principal Component Analysis. This is perhaps the best known method of multivariate analysis, and has been employed in many fields of biology. By its use Kraus & Choi (1958) analysed the prenatal growth of the human skeleton, Jolicoeur & Mosimann (1960) examined the size and shape variation in the Painted Turtle whilst Bailey (1956) assessed the genetic and environmental components of morphogenesis in mice, to mention but a few applications. A principal component analysis is here applied.to all the developmental stages of each sex and species in turn, and to a single analysis embracing all the stages, sexes and species simultaneously. In both cases it represents an analysis of a true ontogenetic growth process rather than of size variation within an instar. The generalisation of the allometry equation proposed by Teissier (1955, 1960), and used by Matsuda & Rohlf (1961) utilises the first principal component of the correlation matrix, whilst that preferred by Jolicoeur (1963 a & b) uses the same component based on a covariance matrix of logarithmically transformed data. Only the latter has received attention in this study. The generalisation of the allometry equation has received further attention from Hopkins (1966) and Sprent (1968); the latter author considers that the generalisation proposed by Jolicoeur (1963a) is inadequate unless the nature of the remaining components is -152-

considered, a feature which has been included in the present multivariate allometric analysis. (iii) Factor Analysis. This widely adopted technique, which utilises a different mathematical model from principal component analysis, has been applied to many different problems: to establish taxonomic relations- among species and higher categories in the Hoplitis complex of bees (Rohlf & Sokal, 1962); to assess different levels of variation in the gall-making aphid P. populi-transversus (Sokal, 1962) and to construct evolutionary patterns in the Pelycosaurian reptiles (Gould, 1967). Factor analysis may take a variety of forms, described by Harman (1960), at present opinion is divided on the best means to conduct a factor analysis, and for this reason it has not been possible to carry out a satisfactory evaluation of factor analysis in the present exploratory study. However, the data are still available for further statistical analysis. The one method of factor analysis studied makes use of a principal component analysis as a first solution, from which, by rotation of orthogonal axes, an attempt is made to impart a simpler and biologically more meaningful interpretation to the patterns of weights which make up the latent vectors. Factor analytical methods have been commended by Gould (1966) though they present greater difficulties of statistical theory than do other methods; rather surprisingly Gould did not attempt to examine any alternative multivariate techniques. (iv) Multiple Discriminant Analysis. This method is an application to more than two groups of Fisher's discriminant function, and is intended here to provide an efficient method of discriminating between the stages of each sex and species. It is perhaps the method which has most frequently been used in analysing insect variation, since it provides an effective means of discrimination between biologically similar entities. Thus it has been used to separate polymorphs, phases -153- and castes of insects (Blackith, 1957, 1962 & 1958, respectively). The first canonical variate being used as a measure of phase in S. gregaria (Symmons, 1969). However, it has only been employed in one previous study of insect growth (Blackith, Davies & Moy, 1963). Its value as a statistical technique for discrimination suggests that it deserves closer study, in the separation of taxonomic categories (Seal, 1964), and therefore as a means of analysing differences between the growth stages of a particular sex or species. The term "general growth factor" has acquired two separate uses in the analytical study of growth. The first principal component is often referred to as a general factor since it involves most characters; on the other hand Blackith, Davies & Moy (1963) refer to the first canonical variate as a general size vector as it ranks all stages. These two meanings are quite distinct and should be distinguished by some terminology, for clarity the term "general character growth factor" is suggested for the former and "general stage growth factor" for the latter case. Detailed consideration is given below to each of the four methods and an attempt has been made, where appropriate, to discuss the difficult question of significance tests in multivariate analysis, which are so infrequently applied. Inevitably, an exploratory investigation of this kind reveals some deficiencies in the planning of the.analyses and difficulties in the interpretation of the results. It is, however, important that these relatively recently developed statistical techniques should be applied actively in a field for which they seem so well suited. Their application being made possible by the availability of electronic computers.

-154-

Introduction to the Growth of the Species.

(i) ay to the Nymphal Instars of the British Species of Ectobius.

1. Body mainly black with distinct white markings on thoracic terga, the most prominent being a wide hori- zontal white band on the metanotum (Fig. 21). Head black. Cerci black. ... • •• E. panzeri 3

Body pale brown, never completely dark. No white markings on thoracic terga.

Head brown. Cerci brown. • •• ••• • •• 2

2, Body pale golden brown with conspicuous dark spots on vertex, thoracic and abdominal terga. These spots are particularly obvious on thorax

(Fig. 22). 000 ... 00* E. pallidus 7

Body brown with no regular markings on vertex or thoracic and abdominal terga (Fig. 23). ... • •• E. lapponicus 12 Fig. 21 Dorsal View of Thoracic Terga E. yanzeri.

a. First Instar

b. Second Instar

c. Third Instar

d. Fourth Instar Female

e. Fourth Instar Male

f. Fifth Instar Female

g. Fifth Instar Male

-156 -

a. b. c.

1 mm. I I

e.

f g

FIG 21. 1mm. Fig. 22 Dorsal View of Thoracic Terga - E. pallidus.

a, First Instar

b. Second Instar

c. Third Instar

d. Fourth Instar

e. Fifth Instar

-158-

a. b. C.

lmm. 1 1

d. e.

f

FIG 22. 1mm. Fig. 23 Dorsal View of Thoracic Terga - E. lampnicus.

a. First Instar

b. Second Instar

c. Third Instar

d. Fourth Instar Female

e. Fourth Instar Male

f. Fifth Instar Female

g. Fifth Instar Male

-160-

a. b. c.

lmm. 1mm. 1 1 d. e.

I 9.

imm. FIG 23.

-161-

E. panzeri:-

3. Pronotum without a central white zone (Fig. 21a). Cerci with 3 segments (Fig. 24g). ... • • • • • • 1st. Instar

Pronotum with a central white zone. Cerci with 6 or more segments. 4

4. Mesonotum without white lateral areas (Fig. 21b). Cerci

with 6 segments (Fig. 24h). ... • • • • • • 2nd. Instar

Mesonotum with distinct white lateral areas. Cerci with 7 or

more segments. • • • • • * 5

5. Mesonotum and metanotum not extended laterally, posterior margin straight (Fig. 21p). Vertex without horizontal white band between eyes. Cerci with

7 segments (Fig. 24i). • • • 3rd. Instar

Mesonotum and metanotum extended laterally, posterior margin not straight. Vertex with horizontal white band between eyes. Cerci with 8 or more segments. 6 Fig. 24 Lateral View of Cercus.

E. lapponicus

a. First Instar b. Second Instar c. Third Instar d. Fourth Instar e. Fifth Instar f. Adult

E. panzeri

g. First Instar h. Second Instar i. Third Instar j. Fourth Instar k. Fifth Instar 1. Adult

E. pallidus

m. First Instar n. Second Instar o. Third Instar p. Fourth Instar q. Fifth Instar r. Sixth Instar s. Adult -163- lrnm. FIG 24

E. lapponicus

a. c. e.

o2 E. panzeri

i.

E. pallidus

-164-

6. Cerci with 8 segments (Fig. 24j). Breadth of pro-

notum 1.75 - 2.05 mms. • • • • • • • • • 4th. Instar

Cerci with 9 segments (Fig. 24k). Wing pads prominent (Figs. 21f & 21g) extending to the 3rd. abdominal tergum in the male and the 2nd. in the female. Breadth

of pronotum more than 2.30 mms. • • • • • • 5th. Ins tar

E. pallidus:-

7. Cerci with 3 segments (Fig. 24m). Breadth of pro-

notum 0.85 - 1.00 mms. • • • • • • • • • 1st. Instar

Cerci with 6 or more segments. Breadth of pronotum more than

1.20 mms. • • • • • • 8

8. Cerci with 6 segments (Fig. 24n). Breadth of pro-

notum 1.20 - 1.30 mms. • • • • • • • • • 2nd. Instar

Cerci with 7 or more segments. Breadth of pronotum more than

1.50 mms. ... 400 000 9 -165--

9. Mesonotum and metanotum not extended laterally, posterior margin straight (Fig. 22c).

Cerci with 7 segments (Fig. 24o). 00f 3rd. Instar

Mesonotum and metanotum extended laterally. Posterior margin not straight. Cerci with 8 or more segments. Breadth of pronotum more than 1.90 mms. • • • 10

10. Cerci with 8 segments (Fig. 24p). Breadth of pro-

notum 1.90 - 2.20 mms. ••• ••• ••• 4th. Instar

Cerci with 9 or more segments. Breadth of pronotum more than

2.45 mms. 400 11

11. Cerci with 9 segments (Fig. 24q). Breadth of pro-

notum 2.45 - 2.65 mms. ••• ••• ... 5th. Instar

Cerci with 10 segments (Fig. 240. Wing pads well developed in both sexes (Fig. 220;extending to the 3rd. abdominal tergite. Breadth of pronotum more than 2.95 mms. ••• ... 6th. Instar -166-

E. lapponicus:-

12. Cerci with 3 segments (Fig. 24a). Breadth of pro-

notum 1.00 - 1.15 mms. ••• ••• ••• 1st. Instar

Cerci with 6 or more segments. Breadth of pronotum more than 1.35 mms. 13

13. Cerci with 6 segments (Fig. 24b). Breadth of pro-

notum 1.35 - 1.60 mms. ••• ...... 2nd Instar

Cerci with 7 or more segments. Breadth of pronotum more than 1.85 mms. 14

14. Mesonotum and metanotum not extended laterally, posterior margin straight (Fig. 23c). Cerci with 7

segments (Fig. 24c). 0 •• ••• Oef 3rd. Instar

Mesonotum and metanotum extended laterally, posterior margin not straight. Cerci with 8 or more segments. Breadth of pronotum more than 2.30 mms. 15 -167-

15. Cerci with 8 segments (Fig. 24d). Breadth of pro-

notum 2.30 - 2.55 mms. • • • • • • • • • 4th. Instar

Cerci with 9 segments (Fig. 240). Wing pads prominent (Figs. 231 & 23g); extending to the 3rd. abdominal tergite in the male and the 2nd. in the female. Breadth of pronotum more than

2.85 mms. • • • • • • 5th. Instar

There is some variation in the colour of the nymphal instars of E. pallidus and E. lapponicus, and generally the later instars become darker. The fifth instar male of E. lapponicus is usually very dark. Nymphs confined to continually damp habitats are usually darker in colour, than those found in drier locations. Although the appendages of the nymphs of E. panzeri are variable in colour, the thoracic markings are remarkably constant in different habitats and vary little between individuals, which makes the latter a useful and reliable character in distinguishing the instars. The cerci of the three species do occasionally become damaged during development and are capable of regeneration to varying degrees; care must therefore be exercised in the identification of nymphs by using the number of cereal segments when it is apparent that these structures have been damaged. -168-

(ii) The Development of the Genital Segments and their use in Distinguishing the Sexes.

A study of the growth of a species is enhanced by the recognition of sexual differences at an early stage in development. Few workers have mentioned distinguishing features in the early instars of cockroaches. However, Ross & Cochran (1960) and Lawson & Lawson (1965) found that it was possible to determine the sex of first instar nymphs of the larger household species by using characteristics exhibited by the posterior abdominal sternites. The nature of the caudal margin of the ninth sternum and the degree of development of the seventh sternum are reliable means of determining the sex of the three species, including even the first instar. Detailed descriptions are confined to E. laiponicus and E. panzeri but reference is also made to E. pallidus. The development of the genitalia is an important consideration in the nymphal growth of Dictyoptera, since the sternal region of three segments is intricately involved in the development of the female genitalia, namely the seventh, eighth and ninth abdominal sternites, whilst the ninth sternite in the male undergoes gross structural modification. The ovipositor is described using the terminology of Scudder (1961). No consideration has been given to the genital lobes or phallomexes, which develop adjacent to the opening of the ductus ejaculatorius in the male, since it has been suggested that these structures do not originate from the segmental appendages of the genital segments, but are merely outgrowths to the right and left of the genital chamber wall, and the median lobe is a secondary evagination of the mouth of the ductus ejaculatorius (Snodgrass, 1937).

1st. Nymphal Instar.

The seventh and eighth sternites are unmodified rectangular sclerites, which are similar in both sexes Figs. 25 & 26, 27 & 28 (a & b). A darkened narrow line, on the seventh sternum demarcates -169-

posteriorly, the exposed region of this segment and anteriorly, that part covered by the sixth sternum. This line persists throughout the development of the sternite. The ninth sternum in both sexes bears laterally two similar styles each of which has four long terminal setae. The posterior margin of this segment is more or less straight in the male, whilst in the female there is a distinct median notch, which is well sclerotised. Both sexes have a single pair of long setae, positioned equidistant from the middle of the sclerite. The anterior margin of this segment is produced laterally into two lobes, (Figs. 29 & 30(a & b)).

2nd. Nymphal Instar.

The seventh and eighth sternites remain unmodified in this instar (Figs. 25 & 261 27 & 28 (c & d)). However, in the female the eighth sternum is retracted further beneath the seventh sternum. The degree of retraction varies during the instar, and is not therefore a reliable character for the determination of the sex of the nymph. The length of the seventh sternum in the mid-line is only slightly greater in the female than in the male. The caudal margin of the ninth sternum remains devoid of any central indentation in male nymphs, whilst the median notch in female nymphs becomes relatively deeper and better defined. (Figs. 29 & 30 (c & d)).

3rd. Nymphal Instar.

The abdominal sternites undergo radical structural modifications at the ecdysis into the third instar. The seventh sternum in the male persists as a normal rectangular sclerite similar in shape to the preceding sternites (Figs. 25 & 26 (f)). However, in the female this sternite undergoes a marked increase in size, and accompanying change in shape (Figs. 25 & 26 (e)). The length in the ventral mid-line is approximately double that of the previous -170-

instar, there is also a slight increase in breadth. Laterally, the posterior margin is characterised by two ill-defined lobes. A greater number of setae are present, irregularly positioned over the ventral surface of the sclerite. In female nymphs of this instar the eighth sternun appears as two small lateral lobes, when the insect is viewed from its ventral aspect. The central portion of this sclerite is hidden by the overlying seventh sternum. The degree of scierotisation and the number of setae are reduced. The posterior margin of this sclerite differs from that of the male, in that the length in the mid-line is reduced and the central zone flattened, accentuating the lateral areas (Figs. 27 & 28 (e & f)). In the male the ninth sternum is an uniformly sclerotised structure. The asymmetry of this sclerite can now be discerned; the styles are of unequal length, the left style of E. lapponicus and E. pallidus is slightly longer than the right, and the anterior projections of the former Side are also lopzor. In E. penzeri this situation is reversed, the dimensions of the left style being greater (Figs. 29 & 30 (f)). This sclerite in female nymphs is covered by the seventh sternum with the exception of the styles. The styles exhibit varying degrees of reduction in this instar. The median notch is very well developed and the structure appears on casual inspection to be two separate lobes, since the central region is only very weakly sclerotised (Figs. 29 & 30 (e)).

4th. Nymphal Instar.

The trends initiated in the previous instar are continued. The seventh sternum in the female is very much enlarged; the length in the mid-line has increased considerably. Two symmetrical anteriorly projecting apodemes have developed on the anterior margin of the sclerite in both sexes. These are more clearly defined in E. panzeri. This sternite persists as an unmodified sclerite in the male; however, there is a slight increase in the number of setae (Figs. 25 & 26 (g & h)). -171-

A marked sexual difference exists in the eighth sternum. In the male the sternite is well sclerotised, and bears an increased number of setae. The anterior margin is produced into two apodemes which are asymmetrical; the left apodeme being longer in E. lapponicus and E. pallidus and the right apodeme in E. panzeri (Figs. 27 & 28 (h)). In the female this sclerite undergoes a marked transition to an almost unsclerotised structure. Zones of weak sclerotisation are only present in the lateral regions, and the number of setae is considerably reduced. The posterior margin is extended into two finger-like, unsclerotised outgrowths. These are symmetrical although they vary slightly in size between different insects, and will give rise to the first gonapophyses in the adult (Figs. 27 & 28 (g)). As in the male the anterior margin is produced into apodemes, but not to the same extent. The entire sternite is covered ventral?.y by the seventh sternum. . The asymmetry of the ninth sternum in male nymphs is far more marked, with the left or right apodeme increasing in length in E. lapponicus and E. pallidus or E. nanzeri respectively. Similarly the style of the corresponding side has increased in size, whilst that of the other side (i.e. the right in E. lapponicus and E. pallidus and the left in E. panzeri) has decreased. However, the variation in the size of the styles remains considerable in this instar (Figs. 29 & 30 (h)). In the female this sternum undergoes a complete structural change, losing nearly all evidence of sclerotisation, and showing a marked reduction in the number of setae. The separation of the two lobes is completed by the enlargement of the median notch. The lobes still bear their styles, but the latter undergo further degrees of reduction. Between the two style-bearing lobes lie a pair• of minute membranous projections; these are the rudiments of the second gonapophyses (Figs. 29 & 30 (0). This sternite is also completely hidden by the enlarged seventh sternum. -172-

5th. Nymphal Instar.

The seventh sternum in the male exhibits no major alteration in shape, the apodemes have become more evident and a greater number of setae are present (Figs. 25 & 26 (j)). However, in the female the increase in size has continued, and its shape now resembles that of the adult sclerite. The ventral surface is scattered with setae. (Figs. 25 & 26 (i)). The sclerite extends to the end of the abdomen enveloping both the eighth and ninth sternites. The posterior margin of the eighth sternite of the male is similar to that of the previous instar. However, the apodemes show an increase in length, and maintain the asymmetry initiated in earlier instars. A greater number of setae are also present (Figs. 27 & 28 (j)). In the female the sternite is membranous, with the exception of a small lateral patch of sclerotised cuticle, devoid of setae. The posterior outgrowths, or first gonapophyses have increased considerably in length, but are still completely membranous. The anteriorly directed apodemes are not as well developed as in the male (Figs. 27 & 28 (i)). The ninth sternum in the male is an evenly sclerotised structure, with a marked increase in the number of setae, particularly in E. panzeri, where they form two groups. The asymmetrical shape is more noticeable in this instar as the styles are now markedly unequal, the right style in E. lapponicus and E. pallidus is very reduced or absent. In E. panzeri it is the left style which undergoes reduction in size and in the number of setae. The asymmetry extends to the apodemes as in previous instars (Figs. 29 & 30 (j)). In the female this sclerite is totally membranous, devoid of setae, and possesses two pairs of ventral outgrowths. The median pair of outgrowths (second gonapophyses) are symmetrical and very fragile. The style-bearing pair which develop into the gonoplacs in the adult are similarly both delicate and membranous. Minute styles are present in most nymphs, but these are absent altogether in some cases. This pair -173- of valve rudiments curve towards the mid-line (Figs. 29 & 30 (i)).

6th. Nymphal Instar.

This instar only occurs in E. pallidus. The modifications exhibited in the previous instar are carried a little further, but no additional changes occur.

Adult.

The ovipositor of E. pallidus has been described by Marks & Lawson (1962). McKittrick (1964) briefly discussed the ovipositor and male genitalia of E. pallidus and E. lapponicus respectively. In the female, the seventh sternum is much enlarged and extends backwards to the end of the abdomen, thus enclosing the large vestibulum and smaller genital cavity in which the ovipositor is entirely hidden. The well sclerotised lateral margins of this sclerite are curved dorsally. Two apodemes are present on the anterior margin and these extend into segment six. In the male this sternite is narrow and unmodified. The two apodemes present on the anterior boundary of the sclerite have increased in size and assume an asymmetry similar to those of the ninth sternum. In both sexes this sclerite is densely covered with setae of equal length; in the female these are particularly concentrated in the mid-line (Figs. 25 & 26 (k & 1)). The eighth sternum of the adult male also has two asymmetrical anteriorly directed apodemes. In E. lapponicus and E. pallidus it is the left apodeme which extends further anteriorly, whilst in E. panzeri the converse is true (Figs. 27 & 28 (1)). In the female this complex sternite forms both the dorsal and ventral walls of the genital cavity (Figs. 27 & 28 00). The first gonapophyses and first gonocoxae are situated in the roof of the cavity. The first gonapophyses are the largest of the ovipositor valves. The two valves lie adjacent to each other in the mid-line. The -174- sclerotised bases of the valves are strongly divergent, and articulate laterally with the fused paratergites of the eighth and ninth terga. The first gonocoxae also articulate with the paratergites and gonapophyses. In E. panzeri these sclerites are similar and rectangular, their bases supporting the membranous genital papilla. In E. lapponicus the gonocoxae are more slender, but differ in shape anteriorly; the left gonocoxa terminates as in the former species, whilst the right gonocoxa continues as a transverse sclerotised bar from which the genital papilla extends. The arrangement of the gonocoxae in E. pallidus are similar to those in E. lapponicus, but in the former species it is the right gonocoxa which ends abruptly and the left which forms the transverse sclerite. The common oviduct and the two pairs of ducts from the four spermothecae all open into the genital papilla. The eighth sternum also gives rise to three other sclerites, which have been adequately described by McKittrick (1964). One of these, which constitutes the basivalvulae, com)letes the roof of the genital cavity. The basivalvulae are a pair of oval structures which diverge posteriorly, where they fuse with a large single sclerite, the laterosternal shelf. This composes the floor of the genital cavity, and is a three-lobed sclerite which is strongly convex ventrally. The floor of the genital chamber is raised slightly above the floor of the vestibular chamber and here the semi-circular vestibular sclerite is positioned, anteriorly. The spermothecal plate which in many species lies between the two basivalvulae is absent in the three species. The homologies of these sclerites have caused much concern in the past among workers in the field, but it would seem rational from the evidence provided that the above sclerites all originate from the eighth sternum. The laterosternites of the segment givi,ng rise to the laterosternal shelf, the basivalvulae develop from the median part of the segment anterior to the developing gonapophyses, and the vestibular sclerite as a secondary sclerotisation of sternum eight (McKittrick, 1964). -175-

The ninth sternum of the male forms the subgenital plate, which exhibits extreme asymmetry in all species (Figs. 29 & 30 (1)). Only one terminal style persists i.e. the left in E. lapponicus and E. pallidus and the right in E. panzeri. This style is much enlarged and bears a dense tuft of long hairs on its dorsal surface. The plate is greatly expanded medially, and the asymmetry is such that the style is positioned at the apex of a triangular sclerite, and can be seen when the insect is viewed dorsally. The apodemes undergo a marked elongation and extend anteriorly to the fourth and sixth segment on the left and right respectively in E. lapponicus and E. pallidus. In E. panzeri the arrangement is an exact mirror image of that found in the previous two species. The arrangement of the phallomeres is similarly reversed in E. panzeri. Numerous setae are present, but confined to the periphery of the subgenital plate. In the female the second gonocoxae are fused into a ring structure consisting of an anterior arch, which supports the second gonapophyses, and a pair of posterior lobes (Figs. 29 & 30 (k)). The second gonapophyses are small and contiguous. The notch separating the two gonapophyses is well sclerotised and demarcates the opening of the paired colleterial glands. The posterior lobes of the second gonocoxae are keyhole-shaped structures which lie dorsal to the gonoplacs. These are attached posteriorly to the gonoplacs which lie dorsal to the second gonapophyses, but extend further laterally; no trace of the styles persist. The gonangulum is a well sclerotised rectangular structure which has the three characteristic articulations, namely, a ventral articulation with the base of the first gonapophyses, a dorsal articulation with the posterior lobes of the second gonocoxa and a third with the paratergite of tergum nine. The gonangula are derived from the laterosternites of segment nine (McKittrick, 1964). Fig. 25 Ventral View of Seventh Sternum - E. laupnicus.

Female Male

a. First Instar b. First Instar

c. Second Instar d. Second Instar

e. Third Instar f. Third Instar

.• Fourth Instar h. Fourth Instar

i. Fifth Instar j. Fifth Instar

k. Adult 1. Adult

-177 -

a. b.

c. d.

lmm. 1 e. f

imm' 1 1 g. h.

i. J.

k I.

imm. FIG 25. Fig. 26 Ventral View of Seventh Sternum - E. panzeri.

Female Male

a. First Instar b. First Instar

c. Second Instar d. Second Instar

e. Third Instar f. Third Instar

g. Fourth Instar h. Fourth Instar

i. Fifth Instar j. Fifth Instar

k. Adult 1. Adult

-179-

a. b.

"N. c. d.

e. f

imm. 1 1 g. h.

i. .1

k I.

1mm. FIG 26 Fig. 27 Ventral View of Eighth Sternum - E. lapponicus.

Female Male

a. First Instar b. First Instar

c. Second Instar d. Second Instar

e. Third Instar f. Third Instar

g. Fourth Instar h. Fourth Instar

i. Fifth Instar j. Fifth Instar

k. Adult 1. Adult

bsv basivalvula c.o.op. common oviduct opening g.p. genital papilla lgpo first gonapophysis lgx first gonocoxa 1.ap. left apodeme lts.sh. laterosternal shelf p. 8 & 9 paratergites 8 and 9 sp.op. spermothecal opening v.scl. vestibular sclerite

F.L. fold line (roof of vestibulum deflected anteriorly) a.

C. d.

e.

1mm. g. h.

lgpo

I. J.

lgpo

v. sct k. bsv Its. sh. sp. op. g.p. E c. a op.

lgx p.8&9 lgpo

FIG 27 lmm. Fig. 28 Ventral View of Eighth Sternum - E. panzeri.

Female Male

a. First Instar b. First Instar

c. Second Instar d. Second Instar

e. Third Instar f. Third Instar

g. Fourth Instar h. Fourth Instar

i. Fifth Instar j. Fifth Instar

k. Adult 1. Adult

bsv basivalvula c.o.op. common oviduct opening g.p. genital papilla lgpo first gonapophysis lgx first gonocoxa l.ap. left apodeme lts.sh. laterosternal shelf p. 8 & 9 paratergites 8 and 9 sp.op. spermothecal opening v.scl. vestibular sclerite

F.L. fold line (roof of vestibulum deflected anteriorly) -183-

a. b.

C. d.

e.

1 1mm. g. h.

lgpo

I.

lgpo

k. I.

v scl.

Its. sh. sp. op. C. o. op

lgpo

FIG 28 imm. Fig. 29 Ventral View of Ninth Sternum -

Female Male

a. First In.-3tar b, First Instar

c. Second Instar d. Second Instar

e. Third Instar f. Third Instar

g. Fourth Instar h. Fourth Instar

i. Fifth Instar j. Fifth Instar

k. Adult 1. Adult

a.2gx anterior arch of second gonocoxa ac•..gl.op. accessory gland opening ga gonangulum gpl gonoplac 2gpo second gonapophysis l.ap. left apodeme l.st. left style n notch p1.2gx posterior lobe of second gonocoxa D. 8 & 9 paratergites 3 and 9 r.ap. right apodeme r.st. right style -185-

a. b.

c. d.

e. f

1171m. I I

Q. h.

2gpo gpl

i. J.

2gpo

gpl

k.

a 2gx ac. gi. op. Pl. 2gx 2gpo P. 889 ga gpl

imm, FIG 29 I I Fig. 30 Ventral View of Ninth Sternum - E. Ipanzeri.

Female Male

a. First Instar b. First Instar

c. Second Instar d. Second Instar

e. Third Instar f. Third Instar

g. Fourth Instar h. Fourth Instar

i. Fifth Instar . Fifth Instar

k. Adult 1. Adult

a.2gx. anterior arch of second gonocoxa ac.gl.op. accessory gland opening ga gonangulum gpl gonoplac 2gpo second gonapophysis 1.ap. left apodeme l.st. left style n notch 1.2gx posterior lobe of second gonocoxa p. 8 & 9 paratergites 8 and 9 r.ap. right apodeme r.st. right style -187-

a. b.

C. d.

e.

1mm.

g. h.

2gpo gpl

I.

2gpo gpl

k.

a.2gx pL 2g x ac. gl. op. 2gpo p,6&9 go gpl

FIG 30 imm. -188-

ANALYSIS OF GROWTH.

Materials and Methods.

All specimens used in the growth analyses were collected from the field at the required developmental stage. All stadia of each species were collected from a single locality. The insects were killed with ethyl acetate in a killing bottle. Each insect was then softened by warming in a 10% solution of potassium hydroxide for 5 - 10 minutes. This was followed by a thorough washing in distilled water and subsequent transfer to 50% and then 70% alcohol. The insects were completely dissected in 70% alcohol, and the separate sclerites stained. Chlorazol Black E proved to be a very satisfactory stain for chitin. A fresh unfiltered saturated solution of the stain in 70% alcohol, adequately stained all sclerites in 1 7. 2 minutes and differentiation was unnecessary. The sclerites were dehydrated in 90% and absolute alcohol. Terpineol was used both as a clearing agent and a semi-permanent mountant. Terpineol was also used for the storage of the dissected sclerites, but it was found that the intensity of the stain decreased if preserved in this medium for more than a few weeks. Each sclerite was flattened, and mounted on a glass slide with terpineol. The head capsule was mounted on a cavity slide to prevent distortion. It was possible to determine the sex of all instars as previously described (Pages 168 - 187). Therefore for both species, ten replicates of each sex of the five nymphal instars and the adult were examined. The pronymphal instar was excluded from the analysis. A total of 74 characters• for each insect was measured, involving a wide range of sclerotised structures. Measurements were made with a Beck Screw Micrometer Eye-piecc which on calibration with a stage micrometer yielded measurements with an accuracy of -189-

0.001 mms. The eye-piece was fitted to a Swift Monocular Compound Microscope, which was particularly suitable for this work, since it can provide a working distance suitable for very low power objectives. All lengths were recorded in the mid-line unless otherwise stated, and breadths refer to maximum values. Measurements of paired structures were always made on the left-hand member of the pair. The following characters were measured, and the number accompanying each character remains constant throughout this study.

Characters measured in E. lapponicus and E. panzer!.

Head:-

1. Breadth of frons between antennal sockets. 2. Breadth of submentum of labium. 3. Length of labial palp segment 3. 4. Length of maxillary palp segment 5. 5. Length of scape of antenna. 6. Length of pedicel of antenna. 7. Length of flagellar segment 1 of antenna. 8. Length of flagellar segment 2 of antenna. 9. Length of flagellar segment 3 of antenna. 10.Length of flagellar segment 4 of antenna. 11.Breadth of flagellar segment 1 of antenna.

Thorax:-

12.Length of pronotum. 13.Length of mesonotum. 14.Length of metanotum. 15.Breadth of pronotum. 16.Breadth of metanotum. 17.Lateral length of metanotum. -190-

18.Lateral length of mesonotum. 19.Length of femur of fore leg. 20. Length of tibia of fore leg. 21. Breadth of femur of fore leg. 22. Length of tarsal segment 1 of fore leg. 23. Length of tarsal segment 2 of fore leg. 24. Length of tarsal segment 3 of fore leg. 25. Length of tarsal segment 4 of fore leg. 26. Length of tarsal segment 5 of fore leg. 27. Length of pretarsal claw of fore leg. 28. Length of femur of mid leg. 29. Length of tibia of mid leg. 30. Breadth of femur of mid leg. 31. Length of tarsal segment 1 of mid leg. 32. Length of tarsal segment 2 of mid leg. 33. Length of tarsal segment 3 of mid leg. 34. Length of tarsal segment 4 of mid leg. 35. Length of tarsal segment 5 of mid leg. 36. Length of pretarsal claw of mid leg. 37. Length of femur of hind leg. 38. Length of tibia of hind leg. 39. Breadth of femur of hind leg. 40. Length of tarsal segment 1 of hind leg. 41. Length of tarsal segment 2 of hind leg. 42. Length of tarsal segment 3 of hind leg. 43. Length of tarsal segment 4 of hind leg. 44. Length of tarsal segment 5 of hind leg. 45. Length of pretarsal claw of hind leg.

Abdomen:-

46. Length of abdominal tergum 1. 47. Length of abdominal tergum 2. 48. Length of abdominal tergum 3. -191-

49. Length of abdominal tergun 4. 50. Length of abdominal tergum 5. 51. Length of abdominal tergum 6. 52. Length of abdominal tergum 7. 53. Length of abdominal tergum 8. 54. Length of abdominal tergum 9. 55. Length of abdominal tergum 10. 56. Breadth of abdominal tergum 4. 57. Greatest length of paraproct. 58. Length of abdominal sternum 1. 59. Length of abdominal sternum 2. 60. Length of abdominal sternum 3. 61. Length of abdominal sternum 4. 62. Length of abdominal sternum 5. 63. Length of abdominal sternum 6. 64. Length of abdominal sternum 7. 65. Length of abdominal sternum 8. 66. Length of abdominal sternum 9. 67. Breadth of abdominal sternum 4. 68. Breadth of abdominal sternum 7. 69. Length of right style of abdominal sternum 9. 70. Length of left style of abdominal sternum 9. 71. Greatest length of the right side of abdominal sternum 9. 72. Greatest length of the left side of abdominal sternum 9. 73. Length of the epiproct. 74. Breadth of the epiproct'.

The application. of Richards' (1949) modification to Dyar's Law necessitated an accurate knowledge of the duration of the nymphal instars of the two species. Nymphs of each instar were kept in an outdoor insectary in cages as shown in Fig. lb. Individual instars of both species being kept in separate cages. The nymphs were observed daily -192- and those which had freshly moulted were put into plastic cages (Fig. la). Nymphs hatching from oothecae in the insectary were used to determine the duration of the first nymphal instar. The instars were reared separately, with five or more insects in each daily group, as preliminary experiments revealed that the duration of the instar was increased when insects were reared separately. All instars were subjected to a temperature of 20°C., and a 16 hr. per 24 hr. light regime; the humidity was uncontrolled, but remained relatively constant. Insects under these conditions were observed daily at the same time, and nymphs which had moulted were removed from the cage. From knowledge of the exact dates of the two consecutive moults, the duration of the instar in days was calculated. It is unfortunate that the number of replicates for the fourth and fifth instars of E. lapponicus and all instars of E. panzeri are lower than desired, this was the result of events beyond the control of the experimenter. In the study of Dyar's Law the sexes were combined to increase the number of replicates and to reduce the sampling error. In work on the allometry of growth and the use of multivariate techniques the sexes have been treated separately. All analyses were carried out at the Centre for Computing and Automation, Imperial College, using the IBM 7090 and 7094 computers. All tne computer programs used were written in FORTRAN IV. -193-

I. SUMMARY AND STORAGE OF DATA.

The original data have been summarised by calculating the mean, standard error and range of each variable for all the species, sexes and developmental stages. These are included in the statistical appendix (Pages 374-397). The individual dimensions in millimetres, listed from the punched cards, and the matrices involved in the multivariate techniques, which are not given in full in the thesis, are available as computer output in the Zoology Department, Imperial College. -194-

II. DYAR'S LAW.

(i) Application of Dyar's Law.

Growth ratios, which are the reciprocals of those used by Dyar (1890) have been calculated for each character, i.e. the ratio of the mean size of one stage to that of the preceding stage. Thus for each character five ratios were obtained, one for each stage during which growth could occur. These ratios were tested for similarity by calculating the ratio of the highest growth ratio for a character to that of the lowest. A structure which exhibited complete regularity in growth throughout development would therefore give a ratio of 1.0. For each species the total number of characters with ratios of the highest value to the lowest of 1.00 - 1.05, and of 1.05 - 1.10 were calculated. The former group includes the characters in close agreement with Dyar's Law, which demands a constant ratio between successive instars. The latter group includes characters which approximate to the law but would not be considered valuable evidence in favour of the hypothesis. It is of interest that, with only a few exceptions, the characters obeying Dyar's Law to varying degrees are the same in both species, as can be seen from Table 24. The total number of characters.which are in close agreement with Dyar's Law is small (Table 25), particularly in E. panzeri where only two characters give a ratio in the range 1.00 - 1.05, when the highest growth ratio is compared with the lowest. Many authors have omitted the last growth ratio from an assessment of the validity of Dyar's Law (Miles, 1931; Taylor, 1931; Ghent, 1956). It was noted that the last growth ratio in the series for many characters in Ectobius deviated more from a constant series than any other ratio. The last ratio was therefore excluded and the ratio of the highest growth ratio to the lowest was again -195- Table 24 Application of I/yar's Law to the Growth of E. lapponicus and E. panzeri. -- — E. lapponicus E. panzeri Variable Growth Ratio Growth Ratio Growth Ratio Growth Ratio 1.00 - 1.05 1.05 - 1.10 1.00 - 1.05 1.05 - 1.10 Head 6 2 2 (2) 3 3 (5) 4 4 (11) 5 5 (1) 6 (1) (11) • Thorax 19 20 31 19 28 24 32 20 31 26 (12) 23 37 32 (13) 28 41 33 (14) 29 42 35 (22) 33 (14) 38 (40 35 (29) 4o 37 44 38 (13) 4o (15) 41 (16) 42 (22) (15) (45) (3o) (34) (39) Abdomen 51 46 (59) 46 59 47 (60) 47 (47) 68 (61) 49 (56) (48) (67) 5o (49) (68) 59 (50) 6o (67) 61 62 63 68 (48) (56) (58)

Variables in parenthesis indicate changes which occur when 5th ratio is omitted. Variables can be identified from list on Pages 189-191. -196-

Table 25 The Total Percentage of Characters showing Approximation to Dyar's Law.

E. lapponicus E. panzeri Growth Ratio 1.00 - 1.05 1.05 - 1.10 1.00 - 1.05 1.05 - 1.10

All Growth Ratios 12.16% 21.62% 2.70% 36.49%

Omitting Final Growth 21.62% 31.08% 16.22% 41.89% Ratio

calculated. Additional variables falling in the two categories are shown in parenthesis in Table 24. There is a considerable increase in the number of characters which show either a close or approximate agreement with Dyar's Law (Table 25). Few characters do in fact exhibit a regular geometric progression during development. When all stadia are included it is the appendages of the head and thorax which predominate among these characters, particularly the tarsal segments of the mid and hind legs. The abdominal tergites are the only body segments showing any approximation to the law. When.the last instar is omitted, in addition to the above characters, the thoracic and abdominal tergites tend towards a more regular progression, since the most radical changes in these structures, oteurrirtg at Vle final moult, are disregarded. Since much of the work on Dyar's Law, to date, has been confined to the head capsule, it is desirable to include this character in a study of the law. However, it was found impossible to measure the total breadth of the head capsule with any degree of accuracy, and it was for this reason that a measurement of the -197- frons was made. Miles (1931) found that the breadth of the frons or the entire capsule could be used equally well to substantiate Dyer's Law. The breadth of the frons in Ectobius, however, did not follow a completely regular progression, since the last ratio was considerably lower than the others (Table 26).

Table 26 The Growth in Breadth of the Frons of E. lapponicus and E. panzeri.

E. lapponicus E. panzeri Growth Ratio Instar Growth Ratio

2 1.231 2 1.116 3 1.147 3 1.154 4 1.179 1.226 5 1.153 5 1.133 Adult 1.042 Adult 1.095

111M111,1•11•. Alim.11.1•110 .

A similar trend has been recorded by several authors (Forbes, 1934; Gaines & Campbell, 1935; Edwards, 1964). Bliss & Beard (1954) working on the milkweed bug, Oncopeltus fasciatus (Dallas) found that if the same insects were measured throughout larval development a close approximation to Dyar's Rule was evident. This, however, was overlooked if measurements were taken from a random selection of insects at each stage, as the breadth of the head capsule changed slightly during a single instar causing a deviation from regularity. This presents an unavoidable source of error, since the desired degree of accuracy in this analysis necessitated the dissection of each insect. This could have been overcome by measuring the exuviae, but these are split -198-

along the median dorsal-line during ecdysis which renders them unsuitable for accurate measurement. Structures growing in accordance with Dyar's Law, require that a greater addition is involved at each moult, in order to maintain a constant ratio throughout development. However, linear growth results from the addition of a constant amount at each moult (Ghent, 1956). This results in a declining series of ratios. For each structure the series of ratios was examined for evidence of a decline during development, which would indicate that growth was in fact linear. However, such a growth phenomenon was totally absent from all structures of both species. Moreover, regular trends of increase or decrease in ratios throughout the growth period are discernible in only a few characters. The most prevalent tendency is for the final ratio to be lower.

(ii) Application of Richards' Extention of Dyar's,Law.

Richards (1949) suggested that Dyar's Law only holds when all the instars are of the same duration, and that irregularities in a geometric progression can be explained if the lengths of the stadia are considered. The duration of each instar of E. lapponicus and E. panzeri has been established (Table 27). Each growth ratio for every structure measured was adjusted for the length of the instar using the following formula:-

D R1 Rn ( t ) Dl.Nr where R1 = Adjusted Ratio Ro = Growth Ratio Dt = Total Duration of Instars D1 = Duration of Instar Nr = Number of Growth Ratios (Total Number of Instars -1) -199-

Table 27 Duration of Each Instar of E. lapponicus and E. panzeri at 20°C.

E. lapponicus E. panzeri Instar Duration in Duration in N* Days N* Days (Mean ± S.E.*) (Mean ± S.E.*)

1 268 20.7 ± 0.2 39 15.1 ± 0.2 2 262 20.7 ± 0.1 35 15.1 ± 0.2 3 178 25.3 ± 0.2 36 18.4 ± 0.2 4 6o 20.7 ± 0.3 6o 17.4 ± 0.3 5 52 24.3 ± 0.3 68 19.4 ± 0.3

N* = Total Number of Insects S.E.* = Standard Error

The ratio of the highest adjusted growth ratio to that of the lowest adjusted ratio was calculated. These were compared with the unadjusted values for these ratios. If Richards' hypothesis is correct the ratio obtained from the adjusted values will be lower, since the adjustment would have the effect of making the growth ratios for a character more constant. All structures for both species were tested in this way. However, only 9.46% and 16.22% of the characters in E. lapponicus and E. panzeri respectively showed any agreement with Richards' modification of Dyer's Law. It can be seen that very few characters bear a more constant ratio to each other after Richards' modification, and these are mostly obscure characters (e.g. Variables 8 - 10, 69 - 72 in both species and Variable 57 in E. panzeri). The only main sclerites which support Richards' hypothesis are the posterior abdominal segments and the lateral length of the meso- and metanota (Variables 53 - 54 in E. lapponicus and 52 - 53, 65 - 66 and 17 - 18 -200- in E. panzeri). These characters are all involved in sexual differentiation and it is unlikely that in a less detailed study such structures would even be considered. In his work, Richards found that the first instar often deviated more from a regular series than any other instar. He suggested that the first instar was relatively "too long", since within this instar there was an initial phase when no growth occurred. For this reason, better results were obtained by omitting the first instar. However, when the first growth ratio of gctobius is omitted and the ratio of the highest to the lowest growth ratio for unadjusted values compared with a similar ratio for adjusted values, there is no further agreement with Richards' modification of Dyar's Law (9.46% and 14.86% of the characters in E. lapponicus and E. panzeri respectively showing a closer agreement with the law after the adjustment). The characters involved are identical with the exception that the greatest length of the paraproct in E. panzeri is excluded when the first ratio is disregarded. In both species three characters which showed a close approximation to Dyar's Law were the lengths of the fore, mid and hind femur. If the mean length of the femur is plotted against the instar number, the points lie on a curve (Fig. 31a). However, when the log. mean femur length is plotted against the instar number the points then lie on a straight line (Fig. 31b). There is no indication of a curvilinear relationship as was found in H. obsoleta (Gaines & Campbell, 1935) and by Nagasawa (1965) in the "Shimuza" race of the Gypsy Moth, Lymantria dispar (Linnaeus). Richards' work implies that a closer fit to these curves or lines would be achieved if either the mean femur length or the log, mean femur length were plotted against accumulated days. This was not found to be the case, as Figs. 32a&b show; the points being far more scattered about the lines, indicating that the growth increment per day is not constant. Among Richards' examples of his modification to Dyar's Law it would seem that several of the data would be better described by -201- FIG 31a 2.2 Hind 2.0

Mid E 1.6

Fore

cc') 1.2 a)

Q' 0.8

0.4

I 2 3 4 5 Ad. Instar FIG 31b 3.0

2.0 Hind Mid 0) Fore szb /0

(13

0) 0.5 •••.1

1 2 3 4 5 Ad. Instar Application of Dyar's Law to the Growth of the Femur in E_panzeri. -202- FIG 32a 2.2 Hind 20

) Mid 1-6

(mms. Fore th 1.2 ng Le n 0.8 Mea

0.4

20 60 100 Accumulated Days FIG 32b 3.0 Hind h 2.0 t Mid ng Fore Le

n 1.0 a Me 0.5

20 60 100 Accumulated Days Application of Richards' Extension of Dyar's Law to the Growth of the Femur in E panzeri. -203- a curve than a straight line. The log. of the measurement plotted against accumulated days would probably give a better straight line and thus avoid omitting the first instar. Richards' supposition that the exten ion of Dyar's Law is widely applicable has to be doubted. There is very little evidence to suggest its value in the growth of either of these species, for a large range of characters.

(iii) Application of Przibram's Rule.

Przibram and Megusar (1912) suggested that the increase in length of a structure at each ecdysis was constant and equal to the cube root of 2 or 1.2599; this was thought to be due to a doubling of the number of cells at each moult. Many of the growth ratios obtained in this work approximated to this value and therefore the mean growth ratio for each of the 74 characters was calculated together with the mean growth ratio over all characters for each species:- Mean growth ratio for E. lapponicus = 1.286 Mean growth ratio for E. panzeri = 1.250 It is interesting that over such a wide range of structures a relatively close approximation to the theoretical value was obtained. Bodenheimer (1927) suggested that extra cell divisions or "latent divisions" may occur when the growth ratio exceeds 1.26. An attempt has been made to apply Przibram's Rule with Bodenheimer's modification to a small number of characters in both species. Using the method of Bodenheimer, the mean value of the last instar (adult) is divided successively by 1.26 until a quotient is obtained near to the mean value of the first instar. This produces a geometric progression with a progression factor of 1.26. This series is then compared with the actual series of mean values. Some terms of the calculated series correspond to the actual values. -204-

Quotients lying between these corresponding values indicate a latent division. The two sexes were examined separately in this case. The variables examined in this way were:- Variable 17 (Lateral length of the metanotum) Variable 40 (Length of tarsal segment 1 of hind leg) Variable 52 (Length of abdominal tergum 7) Variable 64 (Length of abdominal sternum 7) Using the term "latent cell division" purely to explain numerical relationships (and without reference to any observable cytological processes) one sees from Tables 28 Cc 29 that latent cell divisions might occur, but how adequately they explain deviations from a regular geometric progression is difficult to assess. Most of the latent divisions occur at the last growth stage. The growth ratios of the lateral length of the metanotum in males of E. lapponicus indicate one latent division between each ecdysis and five at the last moult. In the female six latent divisions occur, one between the first and second, and third and fourth instars and four between the fifth instar and adult. This sexual difference is obviously correlated with the size of the hind wings in the adult. It is interesting that the metanotum of apterous females of ._paneri is without latent divisions, whereas in the fully winged male seven such divisions occur. Very few characters exhibit an equally large number of additional cell divisions; the number varying in different parts of the body, e.g. in both sexes of the two species only one latent division occurs in the development of the first tarsal segment in the hind leg. The seventh abdominal tergum and sternum exhibit sexual differences. The tergum becoming enlarged in the adult male to bear the dorsal gland whilst the sternum of this segment forms the subgenital plate in the female. The increase in size of these sclerites is accompanied by a corresponding increase in the number of latent cell divisions (Tables 28 & 29). Przibram's Rule and his coefficients of increase in length are perhaps of little value for individual characters, but as an overall growth phenomenon they cannot be overlooked comuletely. -205-

Table 23 Application of Przibram's Rule to the Growth of E. lapponicus.

Variable 17 40 52 64 Instar Obs. Calc., Obs. Cale. Obs. Calc. Obs. Calc. MALE 1 0.387 0.348 0.302 0.337 0.179 0.160 0.169 0.164 0.438 0.201 0.207 2 0.583 0.552 0.418 0.424 0.237 0.254 10.263 0.261 0.695 3 0.849 0.876 0.549 0.535 0.299 0.320 0.354 0.329 1.104 4 1.251 1.391 0.719 0.674 0.402 0.403 0.449 0.414 1.752 0.849 0.508 0.522 5 2.267 2.208 0.988 1.069 0.649 0.639 0.646 0.658 2.782 0.806 0.829 3.505 4.417 5.565 7.012 Adult 8.835 1.347 1.015 1.044

FEMALE 1 0.386 0.375 0.307 0.332 0.174 0.176 0.161 0.188 0.472 0.237 2 0.576 0.595 0.410 0.418 0.228 0.222 0.282 0.299 0.376 3 0.823 0.750 0.538 0.526 0.291 0.280 0.444 0.474 0.945 0.598 4 1.116 1.191 0.688 0.663 0.380 0.353 0.712 0.753 0.445 0.949 5 1.688 1.500 0.943 0.836 0.524 0.560 1.180 1.196 1.891 1.053 1.506 2.382 3.001 3.782 Adult 4.765 1.327 0.706 1.898 ___J -206-

Table 29 Application of Przibram's Rule to the Growth of E. panzeri.

Variable 17 40 52 64 1 Instar Obs! Calc: Obs. Cale. Obs. Calc.1 Obs. Calc. MALE 0.346 0.335 0.288 0.307 0.189 0.181 0.193 0.191 2 0.471 0.422 0.382 0.386 0.218 0.228 0.234 0.241 0.531 3 0.675 0.669 0.509 0.487 0.274 0.287 0.297 0.303 0.843 4 0.959 1.062 0.655 0.613 0.372 0.362 0.385 0.382 1.339 0.772 0.456 5 1.597 1.686 0.880 0.973 0.574 0.575 0.531 0.481 2.125 0.724 0.606 2.677 3.373 4.249 Adult 5.354 1.226 0.912 0.764

FEMALE 1 0.349 0.386 0.295 0.301 0.181 0.162 0.191 0.191 2 0.475 0.487 0.388 0.379 0.218 0.203 0.263 0.241 0.304 3 0.643 0.613 0.501 0.477 0.268 0.256 0.412 0.383 0.482 4 0.834 0.772 0.661 0.602 0.357 0.323 0.647 0.608 0.758 0.407 0.766 5 1.100 0.973 0.886 0.955 0.478 0.513 0.990 0.964 1.215 Adult 1.226 1.203 0.646 1.531

Obs.* = Observed Value Cale.* = Calculated Value

Variables can be identified from list on Pages 139-191. -207-

Latent cell divisions do explain some of the deviations from a constant series of growth ratios, but the accuracy of the correspondence between observed and calculated values is such that this explanation cannot be accepted without reservation. The use of a term with obvious cytological implications is also open to objection in the absence of any direct proof that the different growth rates are related to cellular multiplication. -208-

III. ALLOMETRIC ANALYSIS.

(i) The Choice of a Reference Dimension.

The total length of the body is the most relevant reference dimension to use in the allometry formula. However, this has a definite disadvantage: in that the body length in Ectobius varies within an instar; this is particularly evident in the final instar during which the length is doubled. The use of a single sclerite as a reference character is not to be recommended, since this may vary in proportion during development and between species. Therefore a measure of the total body length has been constructed by summing the measurements for the three thoracic terga and the first eight abdominal terga in the mid-dorsal line. The head has not been included, since an accurate measurement of the length of the head capsule could not be obtained. This measure of total body length is artificial, but far more accurate than any normally used. Blackith, Davies & Moy (1963) used this method of estimating total body length in D. fasciatus. An ideal reference dimension would be one which involved all the different measurements simultaneously; this is made possible by the use of multivariate methods which will be considered in more detail later (Pages 309-313).

(ii) Fitting the Allometric Equation.

The law of simple allometry, derived by Huxley (1924) can be expressed by the equation y = bxcc y and x are the dimensions whose relative growth rates are to be compared. The value of y is the measurement of a part of the body, whilst x is the reference dimension, i.e. in this case the total body length in the mid-dorsal line of the thoracic terga -209- and of the first eight abdominal terga. The constants, b and a represent the initial growth index and the equilibrium constant respectively. The sexes of both species are treated separately. The five nymphal instars and the adult are all considered in the appraisal of the allometry of growth. Several methods of estimating the parameters a and b have been used by previous authors. The most common are graphical determination and the method of least squares to estimate the regression of y on x. The former method involves plotting the logarithms for the separate parts of the body, y, against the logarithms for the total body length, x, and fitting a straight line through the data by eye. This method is obviously of little use in accurate determinations of relative growth; but is useful for the preliminary inspection of data. The use of regression lines fitted by the method of least squares, although used by many authors, is not strictly applicable in the study of growth. This method minimises the sums of squares of one of the variables, known as the dependent variable (y). The assumption is that all the error variation occurs in the estimation of this variable. The other variable, x, known as the independent variable is considered to be without error. However, in the estimation of the allometric parameters neither x nor y can be considered as the dependent variable; both are subject to error in their measurement, and to biological variation between the individuals in a sample. Several attempts have been made to devise ways of analysing such data. Two such methods are discussed by Simpson, Roe & Lewontin (1960). One, the "major axis" method minimises the perpendicular distances from the observed points to the calculated line, but has the severe disadvantage that the line is not constant with change of scale. The other method is the "reduced major axis" method of Kermack & Haldane (1950). In this case the products of the deviations of x and y are minimised for each point on the line. The advantages of the "reduced major axis" are summarised by -210-

Kermack (1954). The most important being that neither variable is treated as dependent on the other, that the method is not affected by a change of scale between the variables, and that both the slope and y intercept of the line can be efficiently calculated. Kermack & Haldane (1950) introduced a method which reduced the computation involved in fitting this line. However, no confidence intervals or significance tests are given, which limits the use of the method, while the computational improvements are of little importance if the problem is programmed for a digital computer. An alternative method which gives an accurate estimation of the parameters a and b is Bartlett's "best fit" straight line (Bartlett, 1949). This method gives a line very similar to the "reduced major axis", and like it does not have the scale restriction of the "major axis" method (Simpson, Roe & Lewontin, 1960). It has the further advantage that Bartlett gives methods for carrying out the appropriate significance tests and estimating confidence intervals. The confidence interval of the slope of the line tends, of course, to be wider than that obtained by the method of least squares, since x is now also subject to error (Bartlett, 1949). To compare Bartlett's method with the least squares technique usually employed, the parameters a and b in the allometric equation were first calculated by the latter method for both sexes of the two species. All seventy four characters measured have been considered with the exception of the length of the right and left styles of sternite nine (variables 69 and 70) in the males and females of both species, since these structures disappear in the adult. In addition, by using Bartlett's "best fit" method, a second complete set of estimates of a and b have been obtained for both sexes and species. However, variables 69 and 70 have been omitted in each case. The values of a and b calculated by least squares and Bartlett's "best fit" method are given in Tables 30 - 33. The differences in the values of these parameters, when calculated -211-

Table 30 Values of Allometric Parameters in E. lapponicus Male calculated by two methods.

Variable Least Squares Method Bartlett's Best Fit Method a b a b

1 0.511 0.189 0.513 0.189 2 0.571 0.170 0.574 0.169 3 0.658 0.077 0.668 0.076 4 0.912 0.073 0.921 0.072 5 0.863 0.065 0.860 0.065 6 0.532 0.064 0.538 0.063 7 0.384 0.093 0.400 0.090 8 0.511 0.019 0.550 0.018 9 0.385 0.026 0.430 0.024 10 0.359 0.031 0.390 0.029 11 0.311 0.101 0.334 0.097 12 0.873 0.258 0.872 0.259 13 0.818 0.172 0.821 0.171 14 1.019 0.123 1.024 0.122 15 0.791 0.504 0.794 0.501 16 0.598 0.757 0.632 0.713 17 2.128 0.030 2.104 0.032 18 2.143 0.029 2.113 0.031 19 0.885 0.197 0.887 0.196 20 0.955 0.126 0.957 0.126 21 0.691 0.096 0.682 0.098 22 1.197 0.037 1.187 0.038 23 1.005 0.02? 1.001 0.027 24 0.831 0.030 0.829 0.030 25 0.571 0.045 0.575 0.044 26 0.758 0.058 0.762 0.058 27 0.715 0.034 0.73o 0.034 28 0.989 0.203 0.986 0.204 29 1.082 0.155 1.076 0.157 3o 0.866 0.085 0.852 o.o87 31 1.106 0.064 1.104 0.064 32 0.921 0.037 0.918 0.038 33 0.799 0.034 0.797 0.034 34 0.556 0.046 0.563 0.045 35 0.699 0.066 0.707 0.065 36 0.623 0.041 0.637 0.040 -212-

Table contd.

Least Squares Method Bsrtlett's Best Fit Method Variable a b a b

37 0.966 0.249 0.967 0.248 38 1.119 0.214 1.117 0.215 39 0.744 0.115 0.740 0.116 4o 1.109 0.090 1.106 0.090 41 0.820 0.058 0.819 0.058 42 0.706 0.047 0.706 0.047 43 0.455 0.056 0.459 o.o56 44 0.670 0.072 0.675 0.071 45 0.628 0.041 0.636 0.040 46 0.971 0.065 0.975 0.064 47 0.964 0.072 0.965 0.072 48 1.029 0.071 1.027 0.071 49 1.056 0.069 1.058 0.069 5o 1.057 0.066 1.052 0.067 51 1.087 0.060 1.084 0.061 52 1.285 0.040 1.288 0.040 53 1.187 0.035 1.183 0.035 54 1.167 0.027 1.158 0.028 55 0.917 0.050 0.919 0.050 56 0.757 0.563 0.765 0.555 57 1.036 0.042 1.030 0.042 58 1.068 0.038 1.073 0.038 59 0.909 0.056 0.912 0.056 6o 1.108 0.061 1.101 0.062 61 1.129 0.064 1.121 0.065 62 1.133 o.o6o 1.121 0.062 63 1.172 0.054 1.160 0.055 64 1.293 0.042 1.278 0.043 65 1.528 0.022 1.528 0.022 66 1.755 0.018 1.766 0.018 67 0.808 0.528 0.819 0.518 68 0.816 0.385 0.823 0.318 71 2.029 0.015 2.033 0.015 72 2.217 0.012 2.228 0.011 73 0.483 0.031 0.487 0.031 74 0.632 0.028 0.641 0.028 -213-

Table 31 Values of Allometric Parameters in E. lapponicus Female calculated by two methods.

MeV Variable Least Squares Method Bartlett's Best Fit a b a b

1 0.580 0.171 0.576 0.173 2 0.654 0.154 0.661 0.152 3 0.699 0.074 0.704 0.073 4 0.884 0.076 0.890 0.075 5 0.741 0.076 0.733 o.o77 6 0.422 0.076 0.422 0.076 7 0.366 0.094 0.360 0.095 8 0.382 0.022 0.422 0.021 9 0.321 0.029 0.358 0.027 10 0.218 0.037 0.264 0.034 11 0.170 0.121 0.183 0.119 12 0.964 0.228 0.957 0.231 13 o.861 0.162 0.863 0.161 14 1.011 0.124 1.015 0.124 15 0.880 0.450 0.884 0.447 16 0.592 0.763 0.611 0.738 17 1.766 0.051 1.749 0.052 18 1.951 0.039 1.928 0.040 19 0.900 0.192 0.903 0.191 20 0.921 0.133 0.926 0.131 21 0.779 0.083 0.774 0.084 22 1.141 0.041 1.141 0.041 23 0.964 0.029 0.952 0.030 24 0.835 0.030 0.833 0.030 25 0.643 0.040 0.641 0.040 26 0.733 0.062 0.742 0.061 27 0.683 0.037 0.700 0.036 28 0.944 0.219 0.944 0.219 29 0.994 0.178 0.995 0.178 3o 0.817 0.091 0.816 0.091 31 1.091 0.066 1.092 0.066 32 0.930 0.037 0.921 0.038 33 0.782 0.036 0.779 0.036 34 0.623 0.042 0.621 0.042 35 0.725 0.064 0.735 0.063 36 0.671 0.039 0.685 0.038 -214-

Table 31 contd.

Bartlett's Best Fit Method Variable Least Squares Method 4 a b a b

37 0.963 0.251 0.965 0.250 38 1.077 0.231 1.080 0.229 39 0.844 0.101 0.84o 0.102 4o 1.115 0490 1.118 0.090 41 0.856 o.055 0.853 0.055 42 0.714 0.047 0.709 0.048 43 0.571 0.048 o.586 0.047 44 0.673 0.073 0.678 0.072 45 0.633 0.042 0.648 0.041 46 0.950 0.066 0.956 0.066 47 0.957 0.073 0.958 0.073 48 1.044 0.068 1.043 0.069 49 1.060 0.069 1.063 0.069 5o 1.074 0.065 1.072 0.065 51 1.081 0.061 1.082 o.o6o 52 1.087 0.052 1.090 0.052 53 1.091 0.040 1.097 0.040 54 1.064 0.032 1.074 0.031 55 0.757 0.063 0.783 0.060 56 0.949 0.430 0.946 0.452 57 1.154 0.036 1.171 0.035 58 1.096 0;07 1.094 0.037 59 0.965 0.051 0.971 0.051 6o 1.111 0.061 1.107 0.062 61 1.125 0.066 1.121 0.066 62 1.129 0.061 1.127 0.061 63 1.225 0.050 1.215 0.051 64 1.901 0.021 1.901 0.021 65 1.031 0.033 0.980 0.036 66 0.898 0.042 0.937 0.039 67 0.950 0.430 0.923 0.451 68 0.946 0.320 0.916 0.336 71 1.422 0.036 1.410 0.037 72 1.427 0.035 1.414 0.035 73 0.759 0.023 0.776 0.022 74 0.931 0.021 0.936 0.021 -215-

Table 32 Values of Allometric Parameters in E. panzeri Male calculated by two methods.

Least Squares Method Bartlett's Best Fit Me'c-yDd Variable a b a b I 1 0.595 0.160 0.591 0.161 2 0.605 0.158 0.608 0.157 3 0.712 0.072 0.709 0.072 4 0.895 0.077 0.896 0.077 5 0.822 0.065 0.826 0.064 6 0.569 0.056 0.578 0.055 7 0.324 0.095 0.341 0.092 8 0.621 0.014 0.657 0.013 9 0.532 0.018 0.560 0.017 10 0.414 0.024 0.439 0.023 11 0.364 0.091 0.394 0.087 12 0.942 0.224 0.943 0.224 13 0.830 0.159 0.838 0.157 14 0.951 0.122 0.959 0.121 15 0.897 0.379 0.909 0.372 16 0.595 0.627 0.644 o.50 17 2.153 0.031 2.102 0.033 18 2.220 0.028 2.168 0.031 19 1.015 0.155 1.017 0.154 20 1.066 0.106 1.063 0.107 21 0.787 0.032 0.801 0.080 22 1.319 0.034 1.309 0.035 23 1.221 0.021 1.214 0.022 24 1.111 0.020 1.114 0.020 25 0.564 0.045 0.575 0.044 26 0.761 0.058 0.761 0.058 27 0.456 0.044 0.469 0.043 28 1.051 0.175 1.052 0.175 29 1.167 0.136 1.165 0.137 3o 0.795 0.086 0.308 0.084 31 1.209 0.059 1.201 0.059 32 1.103 0.031 1.092 0.031 33 1.026 0.025 1.018 0.026 34 0.516 0.048 0.521 0.047 35 0.715 0.063 0.717 0.062 36 0.445 0.046 0.452 0.045 -216-

Table 32 contd.

Least Squares Method Bartlett's Best Fit Method Variable

a b a b

37 1.068 0.205 1.071 0.204 38 1.240 0.183 1.243 0.184 39 0.826 0.096 0.834 0.094 4o 1.222 0.082 1.213 0.083 41 1.042 0.046 1.037 0.046 42 0.952 0.035 0.948 0.035 43 0.511 0.050 0.519 0.049 44 0.676 0.071 0.634 0.070 45 0.475 0.044 0.494 0.043 46 0.915 0.065 0.918 o.o65 47 0.895 0.082 0.900 0.081 43 0.937 0.085 0.941 0.084 49 0.963 0.085 0.966 0.085 50 0.983 0.081 0.986 o.o8o 51 1.042 0.070 1.041 0.070 52 1.361 0.040 1.349 0.041 53 1.418 0.025 1.389 0.026 54 1.347 0.021 1.315 0.022 55 0.935 0.044 0.931 0.044 56 0.729 0.520 0.746 0.506 57 1.065 0.039 1.055 0.040 58 1.022 0.041 1.026 0.041 59 0.899 o.o6o 0.900 0.060 6o 1.013 0.074 1.017 0.074 61 0.981 0.085 0.981 0.085 62 0.982 0.082 0.976 0.083 63 1.041 0.071 1.036 0.072 64 1.177 0.054 1.166 0.055 65 1.549 0.024 1.534 0.025 66 1.966 0.014 1.944 0.015 67 0.742 0.527 0.753 0.514 68 0.797 0.365 0.805 0.360 71 2.351 0.011 2.291 0.012 72 2.134 0.014 2.084 0.015 73 0.412 0.035 0.434 0.034 74 0.642 0.029 0.686 0.027 -217-

Table 33 Values of Allometric Parameters in. E. panzeri Female calculated by two methods.

Least Squares Method Bartlett's Best Fit Method Variable a b a b

1 0.632 0.153 0.634 0.152 2 0.675 0.145 0.677 0.145 3 0.770 0.066 0.781 0.064 4 0.846 0.081 0.855 0.080 5 0.721 0.074 0.722 0.074 6 o.466 0.063 0.475 0.062 7 0.277 0.100 0.300 0.096 8 0.347 0.020 0.372 0.019 9 0.308 0.025 0.344 0.024 10 0.209 0.033 0.257 0.031 11 0.262 0.101 0.287 0.097 12 0.998 0.210 1.002 0.208 13 0.895 0.147 0.903 0.145 14 1.000 0.115 1.004 0.114 15 1.008 0.331 1.016 0.326 16 0.738 0.487 0.807 0.437 17 1.121 0.116 1.133 0.114 18 1.563 0.065 1.536 0.067 19 0.988 0.160 0.993 0.159 20 0.961 0.122 0.965 0.121 21 0.825 0.079 0.837 0.078 22 1.252 0.037 1.254 0.037 23 1.164 0.022 1.171 0.022 24 1.014 0.023 1.013 0.023 25 0.587 0.044 0.599 0.043 26 0.740 0.059 0.753 0.058 27 0.582 0.038 0.593 0.037 28 1.029 0.179 1.037 0.177 29 1.079 0.152 1.085 0.151 3o 0.885 0.077 0.894 0.076 31 1.176 0.061 1.179 0.061 32 1.033 0.034 1.034 0.033 33 0.949 0.028 0.946 0.028 34 0.539 0.047 0.534 0.047 35 0.689 0.065 0.697 0.064 36 0.579 0.038 0.588 0.037 -218-

Table 33 contd.

Least Squares Method Bartlett's Best Fit Method Variable a b a b

37 1.040 0.214 1.044 0.212 38 1.171 0.201 1.176 0.200 39 0.922 0.085 0.930 0.084 40 1.203 0.084 1.205 0.085 41 1:009 0.047 1.017 0.047 42 0.882 0.039 0.891 0.038 43 0.512 0.050 0.508 0.051 44 0.678 0.070 0.681 0.070 45 0.554 0.040 0.571 0.039 46 0.899 0.066 0.899 0.067 47 0.930 0.078 0.931 0.078 48 0.969 0.081 0.963 0.082 49 1.004 0.080 1.002 0.081 50 1.032 0.076 1.029 0.076 51 1.072 0.067 1.067 0.067 52 1.121 0.054 1.114 o.o54 53 1.195 0.034 1.183 0.035 54 1.104 0.027 1.161 0.024 55 0.850 0.047 0.912 0.042 56 0.962 0.387 0.964 0.386 57 1.097 0.038 1.103 0.038 58 1.060 0.040 1.061 o.o4o 59 0.895 0.061 0.897 0.060 6o 1.057 0.070 1.047 0.071 61 1.047 0.078 1.042 0.079 62 1.054 0.075 1.046 0.076 63 1.127 0.064 1.121 0.064 64 1.837 0.027 1.843 0.027 65 1.160 0.032 1.039 0.038 66 0.647 0.061 0.632 0.062 67 1.011 0.373 1.012 0.372 68 1.016 0.275 1.003 0.281 71 1.414 0.034 1.394 0.035 72 1.402 0.035 1.382 0.037 73 0.953 0.016 0.931 0.017 74 1.059 0.017 1.081 0.017 L.

Variables can be identified from the list on Pages 189-191. -219- separately by the two methods, are much less than was expected. The estimates of b are remarkably similar, and in 96.2% of the cases the difference lies merely in the third decimal place. It is of considerable interest that those variables exhibiting differences to the second decimal place are all variables describing the width of the body, e.g. variable 16, a measurement of the breadth of the metanotum is consistently the character which shows the highest difference in both sexes and species. Other characters include the breadth of the fourth and seventh abdominal sternites (variables 67 and 68), in both sexes of E. lapponicus, but only variable 67 in the male of E. panzeri and variable 68 in the female. The breadth of the fourth abdominal tergite of the male of E. panzeri also shows a difference of a higher order. This suggests that to overlook the error variation in x does introduce a bias into the calculation of the parameters which should be avoided. It is the measurements of the width of certain structures, which tend to have a wider range of values, as can be seen in the statistical appendix (Pages 374 - 397) for the above characters. The estimates of the other parameter, a, indicate more differences between the use of the two methods. However, 66.7% of the variables only show a difference in the third decimal place. Characters which are more affected by the use of other methods are again similar for the species and sexes, these include the length and width of the antennal segments (variables 8 - 11), the lateral length of the mesothorax and metathcra (variables 17 and 18) and the length in the mid-line of the ninth abdominal sternite. As with parameter b, the width of the metanotum is one of the characters with the highest difference. The only difference of similar order to that found by Kermack & Haldane (1950) is in the estimate of a for variable 65 in E. panzeri (female). Kermack & Haldane (1950) summarise the values of a and b estimated by the use of the method of least squares for the -220- regression of y on x and x on y, and for the "reduced major axis" and "major axis" methods. They find of difference of 0.09 for the evaluation of a and 0.57 for that of b, when estimated from the regression of y on x (method of least squares) and the "reduced major axis" method. The latter gives a similar line to that obtained by Bartlett's "best fit" method, used in this study. Very few differences as high as these were recorded in the present work. The difference between a least squares regression of y on x and Bartlett's method depends on the error variation of x as compared with y. When this is small the estimates of the two methods approach each other, and this presumably accounts for the difference between the Ectobius data and that for Micraster coranvinum used by Kermack & Haldane (1950).

(iii) Significance Tests.

Having estimated the parameters a and b in the allometry equation, it is necessary to calculate significance levels for the following:- a. The significance of the slope of the line, i.e. to test the null h othesis that the equilibrium constant a is equal to zero. b. The significance of deviations from the null hypothesis that a is equal to unity, i.e., that growth is isometric. c. The significance of deviation from linearity i.e., to test the possibility that relative growth involves components other than allometry. It is worth emphasising that in all the extensive literature on allometric growth in insects, significance tests are rarely reported, though without them it is often impossible to interpret the results. Significance tests for a. and b. were performed for both methods of fitting the line described above, but a test for deviations from linearity was only possible when using the -221-

estimates obtained by Bartlett's "best Fit" method. a. The Significance of the Slope of the Line.

In order to test whether there is any association between x and y in the equation y = bxtx it is necessary to test whether the equilibrium constant a, given in the formula as the regression coefficient B, is significantly different from zero. If a is equal to zero x and y are unrelated. When a has been estimated by the.method of least squares, to test the null hypothesis that a = 0, the formula given in Bailey (1959 : 98) to compare a regression coefficient with a known standard was used, where (the known standard) represents the hypothetical value of a, in this case zero. The number of degrees of freedom is calculated as N - 2 (i.e. 58). The values of a obtained for each character measured in both sexes and species are all significantly different from zero (p less than 0.001). Using Bartlett's "best fit" method, the formula given by Bartlett (1949 : 210) and quoted by Simpson, Roe and Lewontin (1960 : 233) was employed where p is the hypothetical value of the regression coefficient i.e. zero. The number of degrees of freedom is N-3 (i.e.57). As in the former significance test all values of a are significantly different from zero (p less than 0.001). A high degree of associaticn between parts of the body and the total body length was to be expected. However, the absolute nature of this association is of interest. b. The Significance of Deviations from a equal to Unity.

When a is equal to unity, the part of the body and the total -222- body length have the same geometric growth rates. The ratio between the dimensions of the parts will be constant irrespective of their absolute size. Such organs are said to grow isometrically. Thus, to test the null hypothesis that a is equal to unity is merely a test for the presence of isometry in the growth of a variable. When a is estimated by the method of least squares, the same formula as that quoted for the significance of the slope of the line is used, but in this case p t the hypothetical value of a, is equal to unity. Many characters show very highly significant deviation from a = 1 (p less than 0.001). The probability level selected to indicate isometry is p greater than 0.05, that is a non-significant deviation from unity. Very few characters do in fact fall in this category, and these are summarised in Table 34. Using the values of a calculated by Bartlett's method, a similar test to that performed for the significance of the slope of the line was repeated with p equal to unity. A list of the characters showing an approximation to isometric growth with p greater than 0.05, is given in Table 35. From Tables 34 2: 35 it can be seen that very few characters exhibit isometric growth, the phenomenon being more common in E. panzeri than E. lapponicus. The results obtained from the two sets of data are very similar, which is to be expected from the small differences between the values of a and b calculated by the two methods. Various similarities between the species and the sexes exist. The length of the femur of the fore leg (variable 19) grows isometrically only in E. panzeri. The breadth of the fourth and seventh abdominal sternites (variable 67 and 68) and the length of the eighth abdominal sternite (variable 65) exhibit isometry in the females of both species only. The length of the metanotum (variable 14) and that of the ninth abdominal tergite (variable 54) also have equilibrium constants near to unity -223-

Table 34 Characters showing Isometric Growth ( a calculated by Least Squares Method).

Variable E. lapponicus E. panzeri Male Female Male Female

12 I 14 I I I 15 I 19 I I 23 I 24 I 28 I 29 I 33 I 41 I 49 I 50 I 54 I I 57 I I 58 I 6o I 61 I 62 I 65 I I 67 I I 68 I I 70 I 73 I 74 I

Indicates Isometric Growth (p greater than 0.05) Variables can be identified from the list on Pages 189-191. -224-

Table_15 Characters showing Isometric Growth ( a calculated by Bartlett's Best Fit Method).

Variable E. lapponicus E. panzeri Male Female Male Female

12 I 14 I 15 I 16 I 19 1 1 23 I 24 I 28 I 29 I 33 I 41 I 49 I 50 I 55 I 57 I I 58 I 60 I 61 I 62 I 65 I I 67 I I 68 I I 73 I 74 I

____,

I Indicates Isometric Growth (p greater than 0.05) Variables can be identified from the list on Pages 189-191. -225-

in the females of both species, but only when a is calculated by the method of least squares. This is an indication of the importance of the use of a method applicable to the data at hand. The greatest length of the paraproct (variable 57) only grows isometrically in the males of both species. In contrast to this special case when a is equal to unity, the majority of characters grow with constantly changing ratios; such characters are growing allometrically. If a is less than unity allometry is negative, indicating that the geometric growth rate is greater for x than for y, the organ under study; if a is greater than unity allometry is positive. Using only the values of a obtained by Bartlett's method, these have been classified according to the nature of their allometric growth, and can be seen in Table 36. The characters showing isometry have also been included. Details of the actual values of a can be seen from Tables 30-33. The type of allometry is remarkably constant for a character, in both sexes and species. However, a detailed discussion of thin will be deferred until a consideration of the deviations from simple allometry have been made.

c. The Significance of Deviations from Linearity.

An organism shows simple allometry of growth for two dimensions if the ratio of the geometric growth rates remains constant over the entire period of growth (Simpson, Roe & Lewontin 1960). Scatter diagrams of the raw data often show deviations from simple allometry, but a statistical analysis of the deviations from linearity has not previously been made. Significant deviations from linearity may be the result of curvilinear trends, critical points in the allometry line or rhythmical variations about an average trend (Reeve & Huxley, 19+5)• Tests of deviations from linearity are not possible for the line calculated by the method of least squares, when only -226-

Table 36 Types of Allometry in all Characters ( a calculated by Bartlett's Best Fit Method).

Variable E.Luaapicus E. panzeri Male Female Male Female

1 - - -* 2 - - -* -* 3 ..* - _* _* 4 - - _* -* 5 - - -* _* 6 -* _* _* -* 7 - - - - 8 _ _ _ _ 9 - - - - 10 - - - - 11 - - - - 12 - - - I 13 - - - - 14 - + - I 15 - - - I 16 - - - I 17 + + + + 18 + + + + 19 -* - I I 20 - - + -* 21 - _* - _* 22 + + +* +* 23 I* - +* +* 24 - - +* I 25 - - - - 26 - - - - 27 - - -* - 28 I - + + 29 + I +* +* 30 -* - - - 31 + + +* +* 32 _* - +* +* 33 - - I* -* 34 - - _* - 35 _ - - _* 36 -* - -* - 37 _* - + + 38 + + +* -0 39 - - - - 40 + + +* +* -227--

Table 36 Contd.

E. lapponicus i E. panzeri Variable -- Male Female Male Female

41 _* - +* +* 42 - _* _* - 43 - - _* - 44 _* _ _* - 45 -* - -* -* 46 - - _ _* 47 - -* -* - 48 +.* +* -* - 49 + + -* I 5o +* 4 I* + 51 + + + + 52 + + + + 53 + + + + 54 + + + + 55 - -* - 1 56 - _* _ - 57 I + I + 58 + + I + 59 ...* - -* -* 6o + +* + + 61 + +* I +* 62 +* + I* +* 63 +* + + + 64 + + + 4-* 65 + I + I 66 + - + - 67 - I* - I 68 - 1* - 1* 71 + + + + 72 + +* + + 73 - - - I 74 - - - +

I Indicates Isometric Growth + Indicates Positive Allometric Growth • Indicates Negative Allometric Growth • Character growing by simple allometry (non-significant deviations from linearity) Variables can be identified from the list on Pages 189-191. -228- one variable is subject to error and measurements are not replicated for each value of the independent variable. Therefore a test of linearity has only been made for Bartlett's "best fit" line. This test is given by Bartlett (1949 : 211), where the number of degrees of freedom is equal to N-3. The test has been repeated for all the characters of the two species and sexes. Characters which do not show significant deviation from linearity are those where p is greater than 0.05. These characters are marked with an asterisk in Table 36. Such characters are relatively few in number. Moreover, it is only these characters which obey the simple allometry equation that should be considered in an appraisal of allometry in a species. Many authors in the past have merely applied the allometry equation to data which would be better described by an equation of a higher degree, in the case of curvilinear traits or as two or more separate straight lines where critical points are encountered during development. Many of the characters studied by Matsuda show considerable irregular deviations from linearity, as indicated by his double logarithmic plots (1961 a, b, c; 1961d), whereas in his work on two species of Trepobates (Matsuda, 1962a) there is preponderance of curvilinear trends, which might be better described by an equation of a higher order. Unfortunately in neither case does one even know whether the deviations are significant statistically. Gould (1966) considers that the introduction of more parameters into an equation makes the biological interpretation far less clear and is not to be recommended. The simplicity of the allometry formula is probably the reason for its widespread use, while on the other hand it is obvious that by introducing enough parameters any set of data can be fitted adequately without increasing our biological understanding of the subject. Nevertheless the deviations from simple allometry seem to deserve further attention (Gould, 1966). In males and females of E. lapponicus only 22% and 18% of the characters respectively show non-significant deviations from -229- linearity. In E. panzeri the percentages are higher being 39% and 3.5%' in the males and females respectively. A wide selection of characters show simple allometry of growth. However, this relation is totally absent from all thoracic characters measured. Scatter diagrams of the raw data indicate, that although these characters grow according to the law of simple allometry in the early instars, the penultimate instar and adult deviate markedly from this. These deviations are caused by the formation of the wings in the adult. This indicates the hazards which are involved in selecting structures which are convenient to measure, such as the pronotum, in less detailed studies of allometry. The flagellar segments of the antennae also show marked deviations from linearity; these are, however, easily explicable when their adult form is considered, a marked elongation occurring in the last stadival. Other characters showing deviations from simple allometry are the posterior abdominal tergites. In the male of both species these are intricately involved in the development of the tergal gland and undergo subsequent lengthening in the later instars. In the female these tergites also become lengthened in the final instar. The length of the pedicel of the antenna is the only character which grows by simple allometry in each sex of the two species and is the only variable which could be used in a valid comparative account of simple allometry in the two species and sexes. The percentages of characters showing non-significant deviations from allometry are similar in the two sexes of both species. In E. panzeri many characters grow by simple allometry in both sexes, whereas in E. lapponicus there are only two such characters (variables 6 and 48). A. discussion of the presence of negative or positive allometry or isometry should be confined to characters which do not yield significant deviations from linearity. However, the distribution of negative and positive values among the variables is of an interesting nature and will be considered briefly. Most characters are specific in their type of allometry, although there are several species or sex differences. -230-

Characters which grow at a slower rate than the rest of the body in both species and sexes include all dimensions of the head and antennae which were considered, the width of the femur and the length of the fourth and fifth tarsal segments and pretarsal claw of each leg. The length of the first, second and tenth tergites and the second sternite also consistently show negative allometry. Positive allometry is shown by the first tarsal segment in each leg and the tibia of the hind leg. The posterior abdominal tergites, six, seven, eight and nine and sternites one, three and six also grow at a greater relative rate than the rest of the body in both sexes and species. The only consistent difference between the sexes is the length of abdominal sternite nine, which grows by positive allometry in the females and negative allometry in the males. Interspecific differences occur only in the appendage segments. In E. lapponicus the length of the mid and hind femur, the length of the third tarsal segment in the fore leg and the second tarsal segment in the mid and hind legs show negative allometry, whilst these characters grow by positive allometry in E. panzeri. However, the length of abdominal tergite three has a positive value for a in E. lapponicus and a negative value in E. panzeri. The exact differences between the sexes and species can only be seen by inspection of the values of a (Tables 30 - 33); but relative differences between the sexes and species in terms of some of these characters can be seen in the growth gradients to be discussed later. (Pages 231-246). No character gives a negative value for a (enantiometry), which would indicate a decrease in absolute size with time. It is interesting that the lengths of the styles of abdominal tergite nine are characters which grow in this way. In the females of both species, both styles show a gradual reduction in absolute size, whilst only one style shows such a reduction in the males. These styles are finally lost in the adult and for this reason were not included in the estimations of a and b. -231- (iv) Growth Gradients.

One of the best methods of studying the distribution of growth intensity in an organism is the consideration of growth patterns known as growth gradients. These were first envisaged by Huxley (1932) as such, although they were the basis of Thompson's Cartesian Transformations (1917). Variations in the equilibrium constant of consecutive parts of the body, have been found by many authors, to be related in a simple way, and this can be described in terms of a growth gradient. Teissier (1960), howeverl , points out that the construction of these gradients is largely arbitrary, and that they may have little theoretical significance. However, in the present work the use of growth gradients has been found helpful, and presents a satisfactory means of understanding the data. Where growth gradients have previously been constructed it seems to have been assumed that only simple allometric relationships were involved so that a expresses all that need be said about the intensity of differential growth. The results given above show that this is not generally true for Ectobius (and probably, if the appropriate analysis had been made, for many other examples cited in the literature). Nevertheless, we may continue to construct growth gradients involving a on the understanding that they represent the importance of the allometric component in the growth relationship. How far this relationship would be modified by the inclusion of higher-degree terms is outside the scope of this account. Growth gradients have been constructed for the following in males and females of both species:- a. Growth gradient in the mid-dorsal line. b. Growth gradient in the mid-ventral line. c. Growth gradient in the fore, mid and hind legs. d. Growth gradient in the antennal segments. The growth profiles of the dorsal and ventral body segments for both species and sexes were constructed, but have not been included, since they contribute little to the understanding of -232-- the distribution of growth intensity in the body, which is better explained by the simpler growth gradient. Teissier (1960) considers that from the study of growth gradients interesting interspecific comparisons may result. It is in this field that the value of such a study becomes apparent, since no previous comparison of both sexes of two closely related species has previously been made in anything like the present detail. The construction of a growth gradient involves the plotting of the values obtained for the equilibrium constant (by Bartlett's method) against the successive segments of the body or appendage, equally spaced on the abscissa. The line joining the points is the growth gradient. a. Growth Gradient in the Mid-Dorsal Line. (Figs. 33a & b)

This growth gradient includes the thoracic and abdominal tergites. The basic growth pattern in the two species is essentially similar. The value of a in E. panzeri is generally lower than that in E. lapponicus. The gradient for the two sexes is similar, the females having higher values of a, with the exception of the posterior segments of the abdomen. Both sexes are characterised by a growth centre, with highly positive allometry, in the posterior abdominal segments. This centre which is higher in E. panzeri, is probably associated with the development of the dorsal tergal gland of the seventh segment in the male. It is interesting that the eighth segment is also involved in this region of positive allometry and that the growth centre is actually positioned in this segment in E. panzeri. The posterior abdominal segments in the females of E. panzeri undergo a similar relative lengthening but to a lesser extent than in the male. The growth gradient rises steeply from the sixth segment, in the males of both species, indicating a marked change in the equilibrium constant. The gradient declines sharply from the ninth segment in the males and females of both species to a negative value for a -233-

1.4 a. E lapponicus

0

08 o Female

06 ' , . 1 2 3 1 2 3 4 5 6 7 8 9 10

Thorax Abdomen

1.4 b. E panzeri

• Male o Female

06 ' 1 I 1 2 3 1 2 3 4 5 6 7 8 9 10

Thorax Abdomen

FIG 33 Growth Gradient in the Mid Dorsal Line. in segment ten. In E. panzeri both sexes have a similar value for a in this segment whereas in E. lapponicus the sexes have a different value, that for the female being lower. There is a slight upward gradient from the second to sixth abdominal segments. The sexes have very similar values for a in E. lapponicus, but in E. panzeri the value of a in the female is consistently 0.5 or more higher in these segments than in the male, with the exception of segment one. The thorax in the two species is governed by a similar gradient, the growth centre for the thorax being positioned in the metathorax. The gradient from the prothorax to the mesothorax declines to the most deviant value from unity (omitting abdominal segment ten). It is in the metathorax that the most salient interspecific difference occurs. In E. lapponicus the values for a are similar in both sexes and are fractionally above unity. In E. panzeri, however, the female has a higher value for a . This may well be explained by the fact that in E. lapponicus both sexes have fully developed hind wings, and therefore have similar values for a , whereas in E. panzeri the female is apterous and the metathorax does not undergo the reduction in length in the mid-line, which accompanies wing development. b. Growth Gradient in the Mid-Ventral Line. (Figs. 34a & b).

Measurements in the mid-ventral line are confined to the nine abdominal segments only. The gradient in these segments is similar for both species, showing a general anterior-posterior trend, but there is a marked sexual difference which is associated with the development of the genitalia. The gradient in the females of both species.is characterised by a very obvious growth centre in segment seven, a for this segment showing a high degree of positive allometry, which is even more marked in E. lukonicus. This growth centre is associated with the development of the subgenital plate

b. E. panzeri 1-8 - a. E. lapponicus 1.8

1-6 - 1.6

1-4 - 1.4 d

f 1.2 o lue Va

• Male • Male

0.8 o Female 0-8 o Female

0.6 ' ' ' ' 1 2 3 4 5 6 7 8 I 9 06 1 2 3 4 5 6 7 8 9 Abdominal Sternites Abdominal Sternites FIG 34 Growth Gradient in the Mid Ventral Line. -236-

(Figs. 25&26, Pages 17n179). The gradient rises (!radually fr6m the fifth to the sixth segment and then very sharply to the seventh segment. The vast difference in the value for a in this segment and segment eight causes a steep decline which is continued in E. panzeri to segment nine. The values for a in this segment for the two species show a clear difference. The low values for the relative growth constant in these two segments is the result of the reduction in these segments and their incorporation in the female ovipositor (Figs. 29&30, Pages 185&187). .In the males there is a terminal growth centre in segment nine, E. panzeri having a higher value for a in this segment. The growth gradient rises posteriorly from segment six in both species. The difference in the growth rate of segments seven, eight and nine can be seen in Figs. 25 - 30, and is responsible for the high values of a . Two trends noticed in the dorsal abdominal segments are repeated in the ventral segments; the values of a in the two sexes are similar in E. lapponicus, but differ in E. panzeri which generally has a lower value of a . The equilibrium constant for the mid abdominal segments is slightly higher than in the dorsal segments. The zone of negative allometry in the second abdominal segment is of interest. This segment is reduced in the mid line to accommodate the laterally reduced first segment, and although the latter grows at a greater rate than the rest of the body, the'second segment grows at a slower rate (negative allometry). The reduction of this segment was noted as becoming relatively more marked as development proceeded. c. Growth Gradients in the Legs. (Figs. 35 -37a & b).

The femur, tibia, five tarsal segments and the pretarsal claw are considered in the construction of growth gradients in the fore, mid and hind legs of the two sexes and species. The measurements -237- were confined to the left leg. The resultant gradients for the two species bear a close resemblance to each other. In each leg the positively allometric segments are confined to the proximal part of the appendage. In the fore leg the growth centre for the males and females of both species is located in the first tarsal segment. The gradient declines distally to tarsal segment four, and then rises in segment five. In E. lapponicus the values of a in this latter segment and the claw remain similar, whereas in E. panzeri the gradient drops sharply again. The values of a in this leg for E. Ranzeri are consistently higher than those for E. lapponicus in the proximal segments. The males of both species are characterised by higher values of a as far as the third segment, in E. lapponicus, and the fourth in E. panzeri. In the mid leg the growth centre lies in the first tarsal segment, as in the fore leg, in both species and sexes. However, in this leg the values- of a for the femur and tibia are higher and thus the proximal part of the gradient to the growth centre is not as steep. From the growth centre there is a decline in the gradient distally to the fourth tarsal segment; the gradient beyond this segment is an exact replica of the pattern produced in the fore leg. The equilibrium constant is again higher in the proximal leg segments of E. panzeri, and in the males of both species as far as the fourth tarsal segment. In the hind leg the growth centre remains in the first tarsal segment in the females of both species, whilst in the males it is located in the tibia. However, the values of a for these segments are fairly similar, giving the effect of a growth centre in two segments. The distal part of the gradient assumes a similar form to that of the more anterior appendages. The values of a are again consistently higher in the proximal segments of E. panzeri. In this leg, however, the -238- a. E. lapponicus

1.2

el. =1

• Male

- a Female

0.6

F T 1 2 3 4 5 C b. E. panzeri

oc. = 1

F Femur T Tibia C Claw 1-5 Tarsus

FIG 35 Growth Gradient in the Fore Leg

- 23 9 - a. E lapponicus

1.2

7. 0

0

0.8

0.6

0.4 F Ti 2 3 4 5 C b. E panzeri

1.2

7.0 .1_=1 0 F Femur T Tibia 5 0.8 C Claw • Male 1-5 Tarsus

0.6 0 Female

0.4 ' F T 1 2 3 4 5 C

FIG 36 Growth Gradient in the Mid Leg

-24o- a. E. lapponicus

1.2

--1? PO 4._ 0 aJ 0 0.8

0.6

0.4 ' F T 1 2 3 4 5 C b. E panzeri

IMI.

1.2

1.0 .L. 1 ‘,.. o F Femur T Tibia -6 0.8 C Claw • Male 1-5 Tarsus

0.6 0 Female

0.4 FT 1 2 3 4 5 C

FIG 37 Growth Gradient in the Hind Leg -241-

equilibrium constant is consistently higher in the male of E. panzeri as far as the fifth tarsal segment, whereas in E. lapponicus it is the female which has the higher equilibrium constants from the first tarsal segment distally. The most outstanding features of the growth gradients of the legs are the resemblance in general pattern between the species for any one appendage, the basic similarity between the gradients for the three appendages and the close resemblance between the sexes. The equilibrium constants for the proximal segments of the appendages in E. panzeri are higher than those of E. lapponicus. This is of interest, since the former species had lower values of a in the body segments. It may be suggested that the limb segments in. E. panzeri need to grow at a greater rate than the rest of the body, in order to produce suitable appendages for the species life on sandy soils. The equilibrium constants for the fourth tarsal segment in both species are similar, the compensation occurring by means of a sharper decline in the gradient in E. panzeri from the higher values of c in the proximal segments to this segment. In E. lapponicus the decrease in the equilibrium constant is more or less equal from the first to the fourth tarsal segment. In E, panzeri the difference between the first and third tarsal segment is relatively small, but that between the third and fourth is very large, the latter being indicated by a very steep decline in the gradient between the segments. In the legs of E. panzeri the values of a for the fourth segment are similar in both sexes, whereas in E. lapponicus the equilibrium constant for the female is consistantly higher than that for the male. The gradient from the fifth tarsal segment to the claw exhibits a species difference; in E. lapponicus the equilibrium constant for the claw is only slightly lower than that for the fifth tarsal segment, but in E. panzeri the decrease is very marked, especially in the male. -242-

gfJgpizIaLc315 1.4- 1.4 1.4 FEMUR TIBIA TARSAL SEG. 1

1.2 1.2 1.2

l. a f o

lue = 1

Va 1.0 1.0 1.0

Male

08 08 08 Female

F M H F M H F M H b. E panzeri 1.4 1.4 1.4- FEMUR TIBIA TARSAL SEG. 1

1.2 1.2 1.2

= 1.0 1.0

F Fore M Mid 08 08 0.8 H Hind

F M H F M H F M H

FIG 38 Growth Gradients in the Femur, Tibia and First Tarsal Segment. -243-

The growth gradients in the femur, tibia and first tarsal segment from the fore leg to the hind leg have been plotted in Figs. 38 a & b. In the femur of E. panzeri there is a slight anterior - posterior gradient. A similar gradient exists in the females of E. lapponicus; however, in the males this is maaked by a sharp increase in the value of a in the mid leg. The most interesting trend is that manifested in the tibia. In both sexes and species there is a marked increase in the value of a from the fore to the hind leg. This gradient ends in the tibia of the hind leg, which becomes the growth centre in the male. The first tarsal segment of the fore leg has a much higher a value than the tibia in both species ( Figs. 38a & b). This value decreases in the mid and hind legs, although the growth centre of the mid leg and of the hind leg in the females is still located in this segment. The importance of this segment_ is masked by the increasing value of a for the tibia of the mid and hind legs. It is surprising that in these species the growth centre is located in the tarsus, since in previous work it has usually been the femur (Matsuda, 1961 a, b, c), or occasionally the tibia, e.g. in the fore leg of N. undulata (Clark & Hersh, 1939), which have the highest values of a . The difference in the behaviour of the sexes as adults, i.e. the males being extremely active whilst the females remain mostly undercover, may explain the higher values of a for the femur, tibia and first tarsal segments of the males, giving rise to appendages which are longer relative to the size of the body. It i5 unfortunate that no data are available for the coxes and trocanter; these were omitted since their measurement would have introduced an unavoidable source of error. d. Growth Gradients in the Antennae. (Figs. 39a &- b).

Growth gradients have been constructed for the scape, pedicel and first four flagellar segments. The equilibrium constants for a. E lapponicus b. E panzeri • 7 10 oC = 1-0

• Male S Scape 0 Female 08 08 P Pedicel 1- 4 Flagella,- Segments

06 06 4. 0 f o lue

Va 0.4 0.4

0.2 0.2 S P 1 2 3 4 S P 1 2 3 4

FIG 39 Growth Gradient in the Antenna. all the antennal segments are below unity i.e. they show negative allometry. The growth gradients for the two species show some degree of similarity. The terminal growth centre is situated in the scape. The gradients decline distally to a zone of extreme negative allometry in the first flagellar segment. This segment would be expected to have a low equilibrium constant, since the increase in the number of flagellar segments is brought about by the division of this segment. Thus the size of the segment relative to the length of the body would decrease during development. The gradient rises to the second flagellar segment and then declines to the fourth. The values of a for the males are higher than the corresponding values for the females in both species.

The study of growth gradients in these two species of Ectobius has indicated their prominence in parts of the body which are not obviously heterogonic, although Huxley (1932) confined many of his examples to such structures. Growth gradients are also useful means of comparing the sexes or the species, not only in terms of the overall growth pattern they produce but for the individual components of the gradient. It would in some ways have been preferable if the measurements had been taken from a series of individuals which had been followed through their development i.e., if a "longitudinal" type of study had been performed. Clark & Hersh (1939) pointed out that the numerical values of the allometric constants obtained in this way are then the true ontogenetic values of these constants. However, this approach had to be forfeited for the detail required. From such a longitudinal study of individual insects Clark & Hersh (1939) found perhaps the main disadvantages in the use of growth gradients. An anterior - posterior growth gradient for the values of a , where y was leg length and x the length of the body, resulted from a plot of the average values of a for a number of individuals. However, when gradients for separate insects were plotted, only half conformed to the average pattern, -246- the remainder having a different gradient with the growth centre in another segment. The distortion resulting from the consideration of mean values of a would cause a decrease in the steepness of the gradient, and would not therefore be a true picture of the gradient for the species. The study of growth gradients involves yet another consideration which is rooted in the equation itself. Clark 2: Hersh (1939) demonstrated that the values of a and b obtained for the separate segments are not simply related to those obtained for the leg as a whole, although in practice they found that the value of a for the entire appendage was intermediate between the values obtained for the separate segments. Simpson, Roe & Lewontin (1960) conclude from work on growth gradients that the body may be considered as being covered by a general field of growth potential, and that changes in body proportions are the result of varying intensities of this potential, the intensities being distributed in orderly growth patterns. While allometric growth gradients of the kind discussed above convey an approximate picture of the more important patterns, further analysis is needed if the true complexity of the growth processes are to be revealed in greater detail.

(v) Growth Contours.

The deviations from simple allometry discussed on pages 225 - 229 indicate that a , the ratio of the two geometric growth rates, does not remain constant during the development of many parts of the body. Simpson, Roe & Lewontin (1960) outline methods for the determination of a for allometry of growth and of size. In allometry of growth for each stage in development the value of a should remain constant, if the two dimensions are growing by simple allometry, and thus apart from any slight error variations a should show no trend. -247-

The value of a for each stage can be calculated by

a loge y2 - loge 1 y1 loge x2 - loge x1

a is then the tangent of the angle between the abscissa and the straight line joining any two adjacent points in a double-logarithmic plot of y against x. The pattern of deviations from simple allometric growth can be seen by plotting the values of a for each instar against the stage in development. However, a more complete picture of the continuous variation in the value of a in the structures of the body and in time is given by the "growth contour". This has been used to describe growth patterns in the abdomen of A. aquaticus (Needham, 1937) and D. fasciatus (Blackith, Davies & Moy, 1963), and in the legs of the centipede Lithobius forficatus (Linnaeus) where the contour lines have been drawn around the mean growth rates expressed as percentage increments (Needham, 1964). This pictorial representation gives a far more meaningful approach to growth gradients and to the centres of positive and negative a values. In composing the growth contour the values of a for each developmental stage are spaced equally in horizontal rows for each segment of the body, the latter being equally spaced on the vertical axis. Contour lines at intervals of a = 0.25 are then accurately constructed around these values, given in Tables 37&38. Values of a , the equilibrium constant, for each growth stage have been calculated for the length of the segments in the mid-dorsal line, using the above formula, where y represents the length of the body segment and x is the total length of the body as used in previous methods. Only the three thoracic and the first eight abdominal tergites have been used. Four growth contours have been constructed, one for each sex and species (Figs. 40 - 43). The most striking feature of these contours is the similarity between the two sexes of a species, whilst the species patterns are quite different. Within a species -248-

Table 37 Values used in the Construction of Growth Contours in E. paneri.

Length of Body Growth Stage Segment I II III IV V

Male Pronotum 1.232 1.084 0.964 0.801 0.636 Mesonotum 1.113 1.018 0.941 0.760 0.034 Metanotum 1.211 1.110 1.026 0.918 0.233 Abd. Tergum 1 0.895 0.862 0.919 1.017 0.781 Abd. Tergum 2 0.878 0.923 0.886 0.974 0.659 Abd. Tergum 3 0.951 0.958 0.939 1.031 0.613 Abd. Tergum 4 0.881 0.974 0.966 1.043 0.795 Abd. Tergum 5 0.857 0.999 0.982 1.060 0.865 Abd. Tergum 6 0.801 0.947 1.080 1.139 1.148 Abd. Tergum 7 0.705 0.952 1.327 1.547 2.557 Abd. Tergum 8 0.737 0.828 1.153 1.448 4.027

Female Pronotum 1.133 1.123 1.004 0.933 0.772 Mesonotum 1.091 1.047 0.929 0.864 0.441 Metanotum 1.155 1.085 1.023 0.959 0.748 Abd. Tergum 1 0.723 0.902 0.936 0.869 1.060 Abd. Tergum 2 0.755 0.890 0.930 0.990 1.055 Abd. Tergum 3 0.989 0.827 0.959 1.015 1.110 Abd. Tergum 4 0.978 0.949 0.964 1.060 1.076 Abd. Tergum 5 0.986 0.953 1.014 1.083 1.131 Abd. Tergum 6 0.923 0.966 1.081 1.124 1.251 Abd. Tergum 7 0.880 0.920 1.213 1.169 1.407 Abd. Tergum 8 0.879 0.995 0.992 1.247 2.156

...... __ ___

-249-

Table 38 Values used in the Construction Growth Contours in E. lapponicus.

Length of Body Growth Stage Segment I II III 1 IV V

Male Pronotum 1.038 1.039 0.934 0.719 0.467 Mesonotum 1.008 1.024 0.894 0.725 0.109 Metanotum 1.070 1.104 1.076 1.042 0.577 Abd. Tergum 1 0.767 0.884 1.084 1.036 1.060 Abd. Tergum 2 0.902 0.915 0.963 1.019 1.035 Abd. Tergum 3 1.059 1.009 1.002 1.023 1.103 Abd. Tergum 4 1.099 1.014 0.986 1.181 0.962 Abd. Tergum 5 1.110 1.029 1.049 1.004 1.170 Abd. Tergum 6 1.064 0.975 1.088 1.115 1.295 Abd. Tergum 7 0.964 0.856 1.130 1.705 2.122 Abd. Tergum 8 0.613 0.862 0.989 1.290 3.001

Female Pronotum 1.094 0.987 0.954 0.819 1.053 Mesonotum 0.967 1.032 0.903 0.810 0.295 Metanotum 1.027 1.106 1.037 1.007 0.719 Abd. Tergum 1 0.736 0.917 1.010 1.032 1.052 Abd. Tergum 2 0.923 0.913 1.025 1.007 0.820 Abd. Tergum 3 1.087 1.046 1.019 1.034 1.047 Abd. Tergum 4 1.119 1.014 1.014 1.202 0.859 Abd. Tergum 5 1.139 1.009 1.053 1.090 1.131 Abd. Tergum 6 1.029 1.035 1.014 1.140 1.309 Abd. Tergum 7 0.911 0.931 1.045 1.230 1.522 Abd. Tergum 8 0.592 0.828 1.141 1.238 2.087

----

- 2 50- GROWTH STAGE

2 3 4 5

1-u c:p

aL ▪ 0-0-0.5

El 0.5-1-0

1.0-1.5

▪ 1.5-2.0

▪ 2-0-3.0

• 30-4.0

FIG 40 Growth Contour - E. lapponicus Male. - 2 51 - GROWTH STAGE

1 2 3 4 5

1

2

3

1

2

3 111

4

0 CO 5

QC. 0-0-0.5 6 0.5 1.0

1-0-1.5 7 1.5-20

• 2.0-30 8

FIG 41 Growth Contour - E. lapponicus Female. -2 52 - GROWTH STAGE

1 2 3 4 5

aC ❑ 0 0-0 5 Ej 0 5 -1.0 1.0 -1.5 ▪ 1.5 -20

II 20-30 • 3.0-40 In 4.0-50

FIG 42 Growth Contour - E panzer/ Male -253- GROWTH STAGE

1 2 3 4 5

ti

2

c) cz) ca 5

❑ 0-0-0.5 6 O 0.5-10

1 0 -1.5 7 II 1.5-2.0

▪ 2.0-3-0 8

FIG 43 Growth Contour - E,panzeri Female. -254-

the differences are restricted to variations in the contours associated with adult structures. In both sexes of E. panzeri there is a zone of positive allometry, with values of a only slightly in excess of unity, in the three thoracic segments of the early stadia. In later stadia the values of a slowly decrease giving a zone of very negative allometry in the fourth and fifth growth stages. This region is more extensive in the males, the strongly negative values being confined to the mesonotum and metanotum. In the females the values of a are higher, the lowest occurring in the mesonotum of the final stadium. The sexual difference is almost certainly correlated with the development of the wings. The lateral extension of the meso- and metathoracic tergites is accompanied by a reduction in the length in the mid-line of these sclerites. The males of this species are fully winged and thus the meso- and metathorax undergo the decrease in length in the mid-line, giving rise to very low values of a. The females are apterous, although the mesothorax does show slight lateral extensions which may account for the negative a values in the mesothorax. In the abdomen of the females of this species a zone of positive allometry commences in the first abdominal tergite in the final growth stage. In more posterior segments this increase in the value of a is initiated in earlier instars; in segments six and seven the increase occurs between the second and third stadia. In the males there is a similar pattern, which is masked in the last stadium by a zone of negative allometry extending posteriorly to the fifth segment. In the first stadium there is a limited area of negative allometric growth in the females which is confined to the first two abdominal segments. In the males of this instar a similar zone occurs in the seventh and eighth abdominal sternites. The most notable features of the growth contours of this species are the regions of high positive allometric growth which develop at a late stage in the posterior abdominal segments of -255-

both sexes. The values of a increase from the third stadium in males and from the fourth in females. The increase in the value of a is far more gradual in the female, whereas in the male the gradient rises very steeply to a peak in the eighth segment of the final stadium, far in excess of that in the female. This region of intense positive allometry is probably correlated with the development of the tergal gland in the male. The gland itself is first visible in the final instar, but does not become fully developed until the adult stage is reached. However, it would appear that the lengthening of the segment to accommodate this gland is initiated in the third growth stage. Although the gland itself is positioned on the seventh segment, the highest value of a occurs in the eighth segment. This explains the lengthening and attenuation characteristic of the posterior abdominal segments in the adult male. In the female there is no tergal gland, but the lengthening of the abdomen in this region may be associated with the development of the genitalia in the ventral sclerites of these segments. The gradient in the final stadium is less steep in the female than in the male where a negative zone of allometric growth precedes the more posterior intense region of positive allometry. In E. lapponicus, as in the previous species, there is a zone of strongly negative allometry in the thorax common to both species. It is initiated in the second stadium and by the final stadium in the male all three thoracic segments are included, although the prothorax of the female in this stage assumes a positive value of. a . The main region of negative allometry is in the mesothorax, although this spreads to the metathorax in the male, as in E. panzeri. The negative area in the metathorax of the female is slightly less than would be expected, since the hind wings of this species are well developed. There is also a region of ne3ative allometry in the first two abdominal segments, in the first three stadia in the male and the first two in the female. A similar zone was recorded in E. panzeri -256- but in this species it was not restricted to the anterior abdominal segments. The anterior abdominal tergites show a decrease in the value of the equilibrium constant in the last stadium. This is evident in the female of this species, whereas in those of E. lapponicus it was these segments which showed an increase in the value of a A similar region of negative allometry is present in the two posterior abdominal segments, but this is confined to the first two stadia. The difference between the species therefore occurs in the mid-abdominal.tergites, which assume a positive value for a in E. lapponicus, but a negative value in E. panzeri. This difference was also shown by the study of the growth gradients along the body in the two species (Figs. 33a & b), As in E. panzeri the region with the highest value of a is that including the sixth to eighth abdominal tergites. The gradient is steeper in the male and is initiated between the third and fourth stadia, whereas in the female only the last stadium is involved. The gradient in both sexes is less steep than in E. panzeri. Essentially the contour pattern is constant within a species. There must be considerable sampling error in a mean of only ten measurements. This would be reduced by the pooling of the data for the two sexes; however, differences between the sexes would then be lost, which is undesirable since many of these are of biological importance. Such a pattern of growth contours extends our understanding of the allometry of growth in these species. Many of its features are biologically significant, and the comparison between the sexes and species helps to formulate more precisely an overall picture of growth. It would be of particular interest to extend this work further to embrace growth contours for the width of the body, where perhaps the most interesting sexual differences lie. -257-

IV. MULTIVARIATE ANALYSIS.

(i) Eigenanalysis of Correlation Matrices.

Introduction to the Technique.

The correlation matrices for the total number of characters measured have been computed separately for the second and fourth instar and the adult for both species and sexes. This selection of developmental stages permits comparisons between the adults, and two nymphal instars of two sexes and species. The individual correlation matrices can be examined to determine the parts of the body where positive and negative correlations occur, and to explore any consistency in the locations of such correlations in different developmental stages. From the correlation matrices the latent roots and associated latent vectors are extracted. The latent roots indicate the percentage of the total variance which is accounted for by each of the vectors. Each vector is orthogonal with respect to the others and is capable of an independent biological interpretation. The relative sizes of the vector elements represent the weights which must be attached to each variable to determine the required linear function of the variables. By considering these weights it is possible to identify the vector in terms of those variables which show unusually large positive or negative weights. These variables define the vector and might be interpreted biologically as contributing towards a distinct growth pattern. The eigenanalysis of correlation matrices is really a study of intrastadial variation; and the object of this type of analysis was to discover whether common growth patterns could be recognised among closely related growth stages, or whether each stage is associated with a different pattern. -258-

Outline of Statistical Methods and Details of Programs Utilised.

For each developmental stage of the two sexes of E. lapponicus and E. panzeri a matrix of between - character correlations was constructed in the usual way using a FORTRAN IV computer program written by R. G. Davies. The seven or ten largest eigenvalues and their associated eigenvectors were then extracted using either:- (i) An iterative procedure similar to that outlined by Kendall (1957) or by Cooley & Lohnes (1962) and forming part of the same program as that yielding the correlation matrix. or (ii) A library program (subroutine EIGEN from the IBM Scientific Subroutines Package) which extracted latent roots and vectors by Jacobi's method. In all cases the vectors were normalised so that the squared sums of 11 their elements were equal to unity.

a. The Correlation Matrix.

Each correlation coefficient in the matrix was derived from 'ten replicates for the variable. Initially the separate matrices for the two sexes of a developmental stage were obtained, but in order to increase the number of replicates a correlation matrix of the pooled data for the sexes was also constructed. The three matrices for each stage are considered. The complete matrices are not given since their inclusion would involve considerable space. However, their important features are summarised and discussed. In each half-matrix the number of positive and negative correlation coefficients significant at the five per cent level of probability was estimated (Table 39). For the sexes taken -259-

Table 39 Number and Percentage of Significant Off-Diagonal Correlation Coefficients in Half-Matrix.

Species, Stage Positive Negative and Sex Number Per Cent Number Per Cent

E. lapponicus

Adult Male 376 13.92 11 0.41 Adult Female 646 23.92 10 0.37 Male & Female pooled 981 26.32 432 16.73 4th. Instar Male 982 36.36 2 0.07 4th. Instar Female 728 26.95 2 0.07 Male & Female pooled 1390 51.46 40 1.48 2nd. Instar Male 1129 41.80 4 0.15 2nd. Instar Female 1364 50.50 1 0.04 Male & Female pooled 1622 60.05 0 0.00

E. panzeri

Adult Male 496 18.36 4 0.15 Adult Female 1017 37.65 0 0.00 Male & Female pooled 1122 41.54 683 25.29 4th. Instar Male 396 14.66 2 0.07 4th. Instar Female 433 16.03 18 0.67 Male & Female pooled 557 20.62 69 2.55 2nd. Instar Male 633 23.44 11 0.41 2nd. Instar Female 380 14.07 13 0.48 Male & Female pooled 750 27.77 12 0.44 -260- separately values greater than 0.632 are significant at p = 0.05 for eight degrees el' freedom. For the pooled data, with eighteen degrees of freedom, values greater than 0.444 are significant at this level of probability. For a 74 x 74 matrix there are 2701 off-diagonal correlation coefficients in the half-matrix. An outstanding feature of the matrices for the separate sexes is the relatively low number of significant positive correlations and the negligible number of negative ones. No one developmental stage is characterised by a high number of positive significant correlations, and there is in fact little regularity in the occurrence of high or low percentages of positive correlations. The two sexes of both species have similar percentages of significant positive correlations, the number is much increased when the data for the two sexes is pooled. The percentage of characters which are negatively correlated is very much increased in the adults of both species when the sexes are pooled. These significant negative correlations in the adult are not evident in the matrices when the sexes are treated separately. It is not easy to suggest a physical explanation of 11 the statistical association. Blair, Blackith & Boratyriski (1964) also found similar rather low levels of correlation between characters in adult females of C. hesperidum and therefore suggested that development involved several independent growth patterns. This may well explain the low degree of positive correlation present in these two species of Ectobius. However, the number of significant correlations might well be increased if the number of replicates was greater. This has already been suggested in the analysis of these data, since when those for the two sexes are pooled the percentage of significant correlations is increased. An attempt was made to determine the parts of the body which are significantly correlated. From the total number of significant positive and negative correlations in every row of the separate matrices, the mean number and percentage for the whole range of -261-

stadia was derived (Table 40). Variables which are the most consistently highly correlated include the length in the mid-line of the three thoracic terga, the lateral length of the meso- and metanota, the length of the femur of the fore, mid and hind legs, and the tibia and first tarsal segment of the fore and mid legs. The first eight abdominal tergites and abdominal sternites three to six are also highly correlated with the other variables. The structures least correlated include the antennal segments, the tarsal segments (excluding the first segment) and the lengths of the styles of abdominal tergum nine. There are very few significant negative correlations, more or less evenly distributed over the whole range of structures, with a slight increase in their occurrence among the posterior abdominal segments. Cursory observations of the correlation matrices showed no apparent consistency in the pattern of correlation coefficients for the various developmental stages of the two sexes and species. To analyse this further the mean value of the off-diagonal correlation coefficients was calculated for each row in the correlation half-matrix for the adult male of E. panzeri, the zero coefficients being omitted. The correlation coefficient varied considerably in this stadium about a mean of 0.341. Those variables with the highest and lowest values of r were compared in the adult males of the two species, the adult female of E. panzeri and the fourth instar male of E. panzeri. By this means a comparison between the following can be made:-, (i) One stage (adult male) of both species. (ii) Two sexes of a single stadium (adult male and female of E. panzeri). (iii) Two developmental stages of the same species and sex (adult male and fourth instar male of E. panzeri). The variables with the highest correlation coefficients are basically similar in the males of the two species and in the adult female of E. panzeri. However, in the fourth instar male of -262-

Table 40 Number and Percentage of Significant Positive Off- Diagonal Correlation Coefficients in each Row of the Matrix over all Stages for both Species.

Variable Number Per Cent Variable Number Per Cent;

1 229 17.43 38 246 37.96 2 282 21.76 39 152 24.13 3 259 20.27 40 165 26.96 4 356 28.25 41 118 19.87 5 428 34,,46 42 131 22.74 6 163 13.32 43 57 10.22 7 146 12.11 44 65 12.04 8 111 9.34 45 26 4.98 9 log 9.32 46 224 44.44 10 104 9.03 47 276 56.79 11 273 24.07 48 259 55.34 12 475 42.56 49 242 53.78 13 476 43.35 5o 227 52.55 14 534 49.44 51 205 49.52 15 400 37.66 52 182 45.96 16 362 34.67 53 178 47.09 17 471 45.91 54 136 37.78 18 463 45.93 55 119 34.80 19 475 47.98 56 103 31.79 20 422 43.42 57 113 36.93 21 246 25.79 58 93 32.29 22 387 41.35 59 97 35.93 23 293 31.92 60 142 56.35 24 33o 36.67 61 116 49.57 25 266 30.16 62 103 47.69 26 25o 28.94 63 81 40.91 27 194 22.93 64 56 31.11 28 389 46.98 65 55 33.95 29 331 40.86 66 50 34.72 3o 254 32.07 67 41 32.54 31 313 40.44 68 25 23.15 32 183 24.41 69 19 21.11 33 173 23.44 70 11 15.27 34 200 27.78 71 24 44.44 35 146 20.80 72 6 16.67 36 89 13.01 73 11 61.11 37 290 43.54

1.....m...... m.....,...... L - Variables can be identified from the list on Pages 189-191. -263-

E. zanzeri some of the variables have much lower values. This suggests that the degree of positive correlation between a certain range.of structures may be similar within a particular developmental stage, but may vary between different stadia (Table 41). The variables showing a low degree of correlation are essentially similar in the males of the two species and the fourth instar male of E. panzeri. However, many of these variables have relatively high correlation coefficients in the adult female of E. panzeri. The suggestions made above need to be investigated further in a more extensively replicated set of data. On balance, however, they do not suggest that a consistent pattern of correlations can be identified throughout the development of either sex of either species. This being so, it is arguable that the approach to a study of ontogenetic growth patterns via the separate investigation of size variation in each developmental stage is far less promising than was suggested by Blair, Blackith & Doratyriski (1964) in their work on C. hesperidum. The method must, however, be pursued through an eigenanalysis of the correlation matrix before any attempt at a definite evaluation.

b. The Latent Roots and Vectors of the Correlation Matrix.

For the correlation matrix for each developmental stage of both sexes and species the latent roots and associated vectors were derived. Blackith (1960) interpreted such vectors as patterns of growth which take place in an organism. Most of the variance in such analyses tends to be associated with the first few latent roots and it was found necessary to extract only the first seven to account for a high percentage of the total variance as indicated in Table 42. In most cases over 90% of the total variance is accounted for by the first seven latent roots (Table 43). However, the latent roots of the correlation matrices derived from the pooled data for the two sexes account for considerably lower percentages of the total variance. This is, -264-

Table 41 Values of Correlation Coefficients in Four Developmental Stages.

Variables with Highest Values of r in E. panzeri Species, Stage Adult Male and Sex 14 18 19 22 26 28 37 47 72 73

E. panzeri .545 .490 .559 .496 .469 .490 .48o .518 .486 .855 Adult Male

E. panzeri .654 .550 .617 .571 .429 .605 .532 .409 .076 .531 Adult Female

E. lapponicus .358 .225 .446 .433 .288 .489 .515 .605 .519 .835 Adult Male

E. panzeri .138 .269 .271 .358 .380 .340 .324 .417 .025 .618 4th Instar Male

Variables with Lowest Values of r in E. panzeri Adult Male 2 6 32 34 39 41 42 45 56 67

E. panzeri Adult Male .120 .050 .128 .128 .013 .070 .o46 .058 .087 .069

E. panzeri Adult Female .562 .346 .460 .560 .431 .559 .459 .233 .315 .343

E. lapponicus Adult Male .321 .258 .106 .124 .156-.137-.070 .161 .213 .291

E. panzeri .156 .012 .316 4th Instar Male .415 .135 .312 .185 .249 .109 .128 -265-

Table 42 Percentage of the Total Variance Associated with the Latent Roots of the Correlation Matrix.

Latent Root Species, Stage J and Sex 1 II III IV V VI VII ------I

E. lapponicus

Adult Male 35.9 15.0 11.4 8.9 8.3 7.3 5.1 Adult Female 46.4 13.5 10.1 8.9 5.8 5.0 4.2 Male & Female pooled 54.4 18.4 5.2 3.6 2.9 2.7 2.3 4th. Instar Male 53.3 11.4 7.7 6.3 5.5 4.6 4.3 4th. Instar Female 49.3 11.8 8.6 8.1 7.0 5.1 3.9 Male & Female pooled 50.3 11.6 7.1 4.2 4.0 3.3 3.2 2nd. Instar Male 57.1 10.1 8.4 6.9 5.5 5.0 2.9 2nd. Instar Female 62.2 11.7 7.5 5.1 4.7 2.9 2.5 Male & Female pooled 55.1 7.2 6.8 4.8 4.3 3.4 2.9

E. -UEKRE1

Adult Male 41.4 16.5 9.3 9.2 6.1 5.8 4.4 Adult Female 54.8 12.1 8.9 6.9 5.5 3.9 3.1 Male & Female pooled 60.6 19.3 3.3 3.0 2.7 1.8 1.5 4th. Instar Male 32.6 21.0 10.9 9.0 6.5 6.2 5.8 4th. Instar Female 39.2 16.5 10.2 8.3 6.3 5.8 5.7 Male & Female pooled 25.3 23.1 12.9 6.4 4.4 3.9 3.6 2nd. Instar Male 42.0 21.5 11.4 7.6 5.5 4.5 4.0 2nd. Instar Female 36.9 15.9 10.8 8.4 7.6 6.8 5,7 Male & Female pooled 34.0 16.1 10.2 7.3 4.6 4.0 3.5 -266-

Table 43 Cumulative Percentage of the Total Variance Absorbed by the Latent Roots of the Correlation Matrix.

• Species, Stage Latent Root and Sex I II III IV V VI VII

E. lapponicus

Adult Male 35.9 50.9 62.3 71.2 79.5 86.7 91.9 Adult Female 46.4 59.9 70.0 79.0 84.8 89.8 94.0 Male & Female pooled 54.4 72.7 77.9 81.5 84.4 87.1 89.3 4th. Instar Male 53.3 64.7 72.4 78.7 84.2 88.8 93.2 4th, Instar Female 49.3 61.2 69,8 77.9 84.8 89.9 93.8 Male & Female pooled 50.3 61.9 69.0 73.1 77.1 80.4 83.6 2nd. Instar Male 57.1 67.2 75.6 82.5 87.9 92.9 95.8 2nd. Instar Female 62.2 74.0 81.5 86.6 91.3 94.2 96.7 Male & Female pooled 55.1 62.3 69.1 73.9 78.3 81.7 84.6

E. panzeri

Adult Male 41.4 57.9 67.3 76.4 82.5 88.2 92.6 Adult Female 54.8 66.9 75.8 82.7 88.7 92.6 95.7 Male & Female pooled 60.6 79.9 83.7 86.7 89.4 91.2 92.7 4th. Instar Male 32.6 53.5 64.4 73.4 79.9 86.1 92.0 4th. Instar Female 39.2 55.8 66.0 74.3 80.6 86.4 92.1 Male & Female pooled 25.3 48.4 61.3 67.7 72.1 76.1 79.7 2nd. Instar Male 42.0 63.5 74.8 82.5 87.9 92.4 96.4 2nd. Instar Female 36.9 52.7 63.5 71.9 79.5 86.2 92.0 Male & Female pooled 34.0 50.1 60.2 67.5 72.1 76.1 79.6

-___- -267-

however, a form of statistical artifact since the number of non-zero roots must be greater in the latter case where the matrix is presumably of higher rank. The first seven latent roots are only given in full for the adult male of E. panzeri (Table 44);the important features of the other stadia are summarised. Blair, Blackith & Boratynski (1964) gave a biological interpretation of the first eight vectors in C. hesperidum. Blackith (1960) claimed that the growth patterns emerging in adult insects have participated throughout development and are indications of the means by which the adult stage is reached. It was therefore hoped to determine the extent to which the adult form was characteristic of the developmental stages by comparing the vectors identified at each instar in a single sex and species with the adult form. If the two nymphal instars examined (second and fourth) were characterised by a similar set of vectors to the adult, a constant series of growth patterns throughout development is indicated. If, however, each developmental stage is characterised by a different set of biologically identifiable vectors, an independent set of growth patterns will have been shown to occur at each stage, the adult itself being no guide to its mode of growth. Any similarities between the biologically interpretable vectors in the two sexes of the two closely related species were, of course, also to be sought. Initially the latent vectors derived for the adult male of E. panzeri were considered. For each of the seven vectors the elements with the ten largest weights were examined for any obvious biological characteristics by which the vector could be defined. Frequency histograms of the weights appropriate to each variable were constructed (Fig. 44). It was possible in this way to characterise the first seven vectors of this stage; each representing different growth patterns. -268-

Table 44 The Latent Roots and Vectors of the Correlation Matrix for the Adult Male of E. panzer!.

( The Latent Roots are given in Parentheses ) Variables can be identified from the list on Pages 189-191

Vari- T TI III IV V VI VII able

(30.22) (12.06) ( 6.81) ( 6.69) ( 4.42) ( 4.21) ( 3.19)

1 0.147 -0.012 -0.097 0.082 0.077 0.191 0.037 2 0.038 0.055 -0.208 -0.162 0.214 0.164 0.179 3 0.078 0.055 0.208 -0.132 -0.004 0.011 -0.00 4 0.10o o.o4o 0.053 -0.153 0.031 -0.172 0.077 5 0.137 0.067 -0.005 0.021 0.104 -0.079 -0.164 6 0.021 0.080 0.036 -0.187 0.163 0.163 -0.283 7 0.087 0.012 -0.030 0.030 -0.329 -0.055 -0.110 8 0.128 -0.061 0.111 0.032 -0.257 -o.006 -0.095 9 0.128 -0.058 0.099 0.064 -0.128 0.161 -0.076 10 0.064 -0.112 0.227 0.224 -0.033 -0.017 -0.018 11 0.119 -0.004 0.022 -0.030 -0.093 -0.025 -0.338 12 0.125 -0.023 -0.053 -0.092 -0.099 0.077 -0.313 13 0.124 0.033 0.018 -0.173 0.048 0.212 0.167 14 0.169 -0.001 0.038 0.026 -0.011 0.111 -0.064 15 0.121 0.021 -0.105 -0.090 -0.034 0.278 0.00 16 0.124 -0.076 -0.227 0.00 0.054 -0.088 -0.011 17 0.136 -0.057 -0.156 -0.093 -0.072 -0.116 -0.133 18 0.150 -0.043 -0.131 -0.143 0.024 0.017 -0.025 19 0.179 0.041 -0.010 -0.022 -0.040 -0.015 -0.029 20 0.152 0.007 -0.058 -0.173 0.093 -0.026 -0.032 21 0.124 0.113 -0.083 -0.022 -0.127 -0.068 -0.076 22 0.161 0.059 0.081 -0.123 0.066 0.021 0.017 23 0.119 0.131 0.179 -0.017 0.068 -0.013 0.102 24 0.148 0.092 -0.014 0.137 -0.019 0.081 0.128 25 0.093 0.102 -0.164 0.106 0.027 -0.062 0.160 26 0.164 0.071 -0.104 0.026 0.012 0.017 0.024 27 0.122 0.010 0.142 -0.001 -0.172 -0.167 0.160 28 0.173 0.070 -0.003 -0.028 -0.046 0.012 -0.004 29 0.160 -0.016 -0.058 -0.115 -0.053 -0.085 0.049 30 0.090 0.192 -0.109 0.031 -0.170 -0.114 -0.055 31 0.135 0.091 0.175 -0.046 -0.089 0.035 -0.040 32 0.084 0.227 0.103 -0.028 0.017 -0.051 -0.006

••••••••••••••••••*, -269-

Table 44 Contd.

Vari- I 11 III IV V VI able

33 0.132 0.165 0.060 0.063 0.058 -0.044 -0.098 34 0.053 0.059 -0.037 0.035 0.304 -0.273 -0.004 35 0.160 0.051 -0.077 0.062 0.004 -0.099 0.139 36 0.163 0.075 -0.041 0.061 -0.050 0.005 0.153 37 0.172 0.002 0.019 -0.096 -0.033 -0.066 0.073 38 0.126 -0.107 -0.055 -0.155 -0.064 -0.156 0.077 39 0.072 0.225 0.041 -0.154 0.037 -0.049 0.105 4o 0.114 0.050 0.100 -0.185 -0.207 0.072 -0.104 41 0.086 0.177 0.172 -0.052 -0.023 0.064 0.153 42 0.088 0.221 0.063 0.075 0.121 0.030 0.089 43 0.140 0.098 -0.094 -0.100 0.103 -0.136 -0.085 44 0.149 0.113 -0.112 0.032 -0.043 -0.117 -0.008 45 0.113 0.178 -0.102 -0.005 0.053 -0.013 0.137 46 0.049 -0.193 -0.119 -0.100 -0.194 -0.075 0.045 47 0.115 -0.152 0.027 0.203 0.055 0.009 0.019 48 0.132 -0.104 -0.144 0.096 -0.008 -0.166 0.077 49 0.144 -0.107 -0.149 0.140 0.068 -0.180 0.091 50 0.091 -0.200 0.163 0.007 0.049 -0.066 -0.068 51 0.054 -0.185 0.032 -0.192 0.100 -0.167 0.022 52 0.150 0.013 -0.141 0.095 0.124 0.075 -0.032 53 0.126 -0.099 0.131 0.163 0.018 0.022 0.142 54 0.128 -0.023 -0.033 0.055 0.106 0.301 -0.096 55 0.019 -0.137 0.183 -0.068 0.299 -0.006 -0.024 56 0.049 0.037 -0.185 0.158 -0.123 0.297 0.005 57 0.105 -0.093 0.074 0.240 -0.089 0.037 0.127 58 0.130 -0.090 0.107 0.193 0.07 4 -0.063 -0.063 59 0.097 -0.028 -0.117 0.153 0.186 -0.199 -0.199 6o 0.072 -0.197 -0.043 -0.178 0.050 -0.009 -0.009 61 0.114 -0.178 0.066 -0.086 0.143 -0.033 -0.033 62 0.088 -0.187 0.117 -0.131 0.133 -0.027 -0.027 63 0.121 -0.183 0.113 0.004 0.091 0.010 -0.010 64 0.080 -0.156 0.072 0.017 -0.214 0.151 0.151 65 0.082 -0.214 0.04; -0.083 0.053 -0.087 0.132 66 0.093 -0.191 -0.111 -0.091 0.074 0.145 -0.004 67 -0.003 -0.033 0.060 -0.204 -0.117 0.253 0.320 68 -0.069 -0.060 -0.039 -0.305 -0.188 -0.052 -0.041 69 -0.018 -o.o65 -0.033 0.041 -0.067 0.007 -0.121 70 0.000 0.000 0.000 0.000 0.000 0.000 0.000 71 0.087 -0.233 -0.079 0.078 -0.072 0.005 0.029 72 0.143 -0.153 0.078 0.075 -0.034 0.023 0.041 73 0.112 0.094 0.163 0.035 0.015 -0.071 -0.171 74 0.120 0.068 0.176 0.120 0.025 0.136 -0.121 a. Vector I d. Vector 117 10

10 e. Vector 10 0 b. Vector 11 101 f Vector 17 10

0 c. Vector g. Vector VH 10 10

0 171 0 rl -0.12 +0.12 -0-12 0 +d12

FIG 44 Frequency Histograms of the Vector Weights - E. panzeri Adult Male. -271-

Vector I.

This vector has a latent root of 30.2 and absorbs 41.4% of the total variability. It is a general size vector characterised by positive weights for virtually all the variables. By definition the average values of the weights must be -,-24-1 i.e., 0.116, and as can be seen from the frequency histogram for this vector (Fig. 44a) they cluster around this value and have a relatively restricted range. All the variables thus contribute an approximately constant amount to this vector, and no one selected group of variables define the associated growth pattern. This vector represents the general growth in size of the organism, and as in most published work it accounts for a high percentage of the total variation, indicating, as Blair, Blackith Boratyriski (1964) suggest, that variation in size is far more prominent than variation in shape.

Vector II.

The elements of this and subsequent vectors differ markedly from the first vector in clustering around zero (Figs. 44b-g). Most of the variables therefore contribute very little to the growth pattern and it is the more extreme positive and negative values which distinguish the vector. Vector II has large negative weights for the length of the abdominal tergites and sternites and equally high positive weights for the breadth of the femur of the fore, mid and hind legs and the length of some tarsal segments. It therefore appears to represent a mode of growth in which the dimensions of the abdominal sclerites contrast in their behaviour with those of the appendages. The absolute value of the elements (i.e., whether positive or negative) is of no significance since the latent vectors of a matrix are determined only down to an arbitrary constant, which may be -1. It is the contrast between elements of opposite sign which is important. -272-

Vectors with a small number of large positive or negative elements are commonly referred to as "bipolar factors" in faCtor anatyis.

Vector III.

This vector is strongly determined by the length of the right style of sternum nine, and the lateral length and breadth of the meso- and metanota respectively. However, this vector indicates clearly that the patterns of growth cannot be so easily defined as might be hoped. One would not have expected for example that the length of the tarsal segments would be divided randomly among several vectors, but rather that a single vector would be associated with their increase in length.

Vector IV.

This vector reflects changes in the breadth of the body, since the weights for the variables associated with breadth are unusually large, e.g. the breadth of the submentum, the fourth abdominal tergite and sternite and the seventh abdominal sternite.

Vector V.

This vector represents a pattern of growth in which the flagellar segments of the antennae elongate. However, an apparently fortuitous selection of other variables also have high weights and a more exact definition of the vector is difficult.

Vector VI.

This is hard to identify, but again seems to reflect a pattern of growth influencing the breadth of the body e.g. the pronotum and abdominal tergum four. The length of certain abdominal sclerites also have high weights. -273-

Vector VII.

Only 4.4% of the total variation is accounted for by this vector, and it is the last considered worth interpreting. It represents a pattern of growth influencing the length of the thoracic nota and the proximal antennal segments, and also the breadth of the first flagellar segment of the antenna.

The latent vectors for the female adult of E. panzeri, the male adult of E. lapponicus and the fourth instar male of E. panzeri were next examined in order to assess any consistency in the growth patterns described above. The elements with the largest positive and negative weights were also examined for any similarity between the sexes, species or developmental stages. Frequency histograms were plotted for the first seven vectors, and closely resembled those in Fig. 44 for the adult male of E. panzeri. In each case the first vector represents a general size vector, to which most of the elements contribute equally and which accounts for a high percentage of the total variance in each case (Table 42). The second vector represents a pattern of growth which influences the length of the abdominal segments. In the adult male of E. panzeri this vector is determined largely by the abdominal sternites in particular, whilst only the sternites have high positive weights in the adult female of E. panzeri. In the fourth instar.male of E. panzeri this vector reflects a growth pattern.which, however, involves both the tergites and sternites. However, in the adult male of E. lapponicus the second latent vector represents an entirely different growth pattern which involves the growth of the appendages. The abdominal sclerites all have weights near to zero. The remaining vectors are far more difficult to identify and show no consistent association with the sexes, species or develop- mental stages. The third vector which in the adult male of -274-

E. panzeri involves the thoracic terga appears similar in the fourth instar male of E. panzeri, whereas in the adult male of E. lapponicus it is apparently concerned with the breadth of the body. No similarity is discernible in the biological interpretation of the fourth vector between the stadia, sexes or species. One of the patterns of growth described by the fifth vector, common to the four stadia under consideration, is strongly influenced by the length of the antennal segments. The sixth vector is difficult to identify in all cases. The seventh vector is interesting; in the adults of both sexes of E. panzeri and the fourth instar male of this species the growth pattern seems to involve the thorax. The identity of this vector is different in the adult male of E. lapponicus and is not readily interpretable.

The weights of the elements of the latent vectors for the It remaining developmental stages of both sexes and species were examined in a similar way but yielded few common growth patterns. The variable nature of these vectors may be due to the relatively small number of replicates used for each stadium. However, when the sexes for a stage are pooled the consistent nature of the vectors becomes no more evident. This suggests that the number of replicates is not the only reason for this variability. It was not practical to increase the number of replicates, since for other analyses the wide range of characters was desirable. It is in some ways disappointing that hardly any of the vectors could be given a clear and consistent biological interpretation over the stages, species and sexes. It would seem likely on a priori grounds that the growth patterns for each developmental stage are similar and it is not easy to believe that the irregularities observed were due to a great diversity of independent growth patterns. A more detailed analysis of this nature between two closely related species, such as those used in this work, with more replicates and possibly fewer variables would be useful and would have to be made before any further definite conclusions can be made. -275-

Provisionally, however, one may doubt very seriously whether the pattern of size variation exemplified by each separate growth stage does reflect the overall processes of postembryonic growth in insects. -276-

(ii) Principal Component Analysis.

Introduction to the Technique.

A single principal component analysis was applied to all stages of both sexes and species, and also to all stages of each sex and species taken separately. A principal component analysis can be based on either a correlation or a covariance matrix. The correlation matrix is frequently selected since the principal components resolved are independent of the scale of the measurements of the variables. However, the use of a covariance matrix based on logarithmically transformed data has a similar effect. Two separate principal component analyses embracing all the species, stages and sexes are included using a correlation matrix and a log. covariance matrix. The latter has a distinct advantage in that the first principal component represents a generalisation of the allometry equation (Jolicoeur, 19630. This physical interpretation of the first component makes the use of the log. covariance matrix more desirable, and it is for this reason that four separate principal component analyses involving all the developmental stages of each sex and species have been carried out. The extent to which principal component analysis has been taken in this study enables an exploratory examination of the results using both a correlation and a log. covariance matrix. The separate sexes and species can also be compared, using the principal components derived from the log. covariance matrices. Principal component analysis involves the extraction, from a correlation or covariance matrix, of a set of latent roots and vectors, each of which is orthogonal with respect to the others. The vector elements correspond to the weights which are applied to each variable in the determinantal equation. By using these to compute the weighted sum of the original variables a set of principal component scores are derived for each individual stage of every sex and species. The latent roots determine the percentage -277--

of the total variance absorbed by each component, and the relative "importance" of the vectors established. The vectors are then capable of biological interpretation in two ways. The vector elements can be examined to determine which variables are characterised by high weights, such variables define the vector in biological terms. The growth patterns associated with the vector can then be described. Alternatively, the principal component scores can be considered, and direct comparisons can be made between stages, sexes and species by analysis of the patterns of growth defined by each vector. From the small amount of information available on growth data analysed in this way it seems that only the first few vectors can be identified biologically with any certainty. This method represents an analysis of growth throughout development, whereas the eigenanalysis of correlation matrices for each separate developmental stage (Pages 257-275), merely represents an analysis of size and shape variation within that stage. The generalisation of the allometry equation by the use of the first principal component of the log. covariance matrix gives a very real biological interpretation to this vector, and this has been considered in more detail.

Outline of Statistical Methods and Details of Programs Utilised.

Single covariance or correlation matrices involving all stages, sexes and species simultaneously were constructed, using a FORTRAN IV program written by R.G. Davies. This program can be adapted to compute either correlation or covariance matrices, the latter matrix being derived after logarithmic transformation of the primary data. The ten largest latent roots and associated eigenvectors were then extracted by the Jacobi method, using the subroutine EIGEN, a library program from the IBM Scientific -278-

Subroutines Package. The vectors were normalised (the sums of squares of the elements made equal to unity). Principal component scores were calculated from the vectors of both correlation and covariance matrices. When the two species and sexes were analysed separately, the same procedure was utilised, based on a log. covariance matrix. The extraction of the latent roots and vectors in this case were computed by an iterative method similar to that described by Kendall (1957) and Cooley & Lohnes (1962). Only the five largest latent roots and associated eigenvectors were extracted, and in this case the estimation of the principal component scores was omitted, since interest centred primarily on the allometric nature of the first latent vector.

a. Principal Cosmonent Analysis - Correlation Matrix.

A single analysis was applied to all the developmental stages, sexes and species simultaneously. Each correlation coefficient incorporates the measurements of 240 cases (2 species, 2 sexes, 6 stages and 10 replicates). The number and percentage of off- diagonal correlation coefficients, significant at the 0.1% level of probability, is given in Table 45, (for 238 degrees of freedom, r = 0.321).

Table 45 Number and Percentage of Significant Off-Diagonal Correlation Coefficients.

Species, Stage Positive Negative

and Sex Number Per Cent Number 1 Per Cent ..., All Species, Stages and 2623 96.37 5 0.19 Sexes

-279-

A very high percentage of the coefficients are significantly correlated. It is interesting that most of the non-significant coefficients relate to characters 69 and 70 (the length of the right and left styles of sternum nine), and are presumably due to the absence of one or both of these structures in the adult males and females of both species respectively. It will be recalled that the number of significant correlations in the matrices constructed for a single developmental stage of one sex of a species was comparatively low (Table 39, Page 259). Therefore by including all developmental stages, sexes and species the extent of correlation between the variables is very much increased. A high percentage of the total, variance is accounted for by the first ten latent roots (98.86%), (Table 46). It is unlikely that any latent vectors beyond the first five can be interpreted biologic- ally. The fifth latent root is only responsible for 1% of the

Table 46 Percentages and Cumulative Percentages of the Total Variance Associated with the Latent Roots of the Correlation Matrix.

Latent Percentage Variance Cumulative Percentage Root Variance

1 84.08 84.08

2 7.14 91.22

3 3.17 94.39

if 1.90 96.29

5 1.02 97.31

6 0.52 97.83

7 0.40 98.23

8 0.27 98.50

9 0.21 98.70

10 0.16 98.86 -280- total variance. Significance tests are available to categorise the distinct vectors; but such tests assume a multivariate normal distribution. The multivariate distribution present in this context is not normal, since six distinct growth stages are involved. This problem has been resolved by selecting only those vectors which have a tairly clear biological interpretation, in this case the first five. Moreover, the other vectors account for so low a proportion of the total variance that they can be safely ignored.

The Identity of the Vectors.

The latent vectors of the correlation matrix can be identified in two ways:- (i) By examination of each element of the vector to determine high negative and positive weights. Such elements characterise the vector in terms of the growth rates of the corresponding variables. (ii) By calculation of the component scores for each stage of each sex or species, and the subsequent comparison of these scores based on any two of the orthogonal vectors. These are then plotted against each other and the growth patterns described by the vectors are then revealed.

If the vector elements are considered, it is possible to give a biological interpretation to the growth patterns reflected by the first five vectors. The frequency distribution of the weights of the elements of the first five vectors is given in Table 47.

Vector I.

This vector has a latent root of 62.2 and absorbs 84,: of the total variance. It has positive elements for all variables, -281-

Table 47 Frequency Distribution of Elements of Vectors I - V - Correlation Matrix.

Class I II III IV V I 0.47 - 0.45 1 0.45 - 0.43 2 0.43 - 0.41 1 0.41 - 0.39 1 1 0.39 - 0.37 0.3? - 0.35 0.35 - 0.33 1 1 0.33 - 0.31 0.31 - 0.29 0.29 - 0.27 1 0.27 - 0.25 1 1 0.25 - 0.23 5 1 2 0.23 - 0.21 2 1 0.21 - 0.19 1 1 1 0.19 - 0.17 1 1 1 0.17 - 0.15 1 0.15 - 0.13 1 2 0.13 - 0.11 57 1 2 1 0.11 - 0.09 9 1 3 1 0.09 - 0.07 5 2 4 0.07 - 0.05 1 4 3 3 3 0.05 - 0.03 2 6 5 8 0.03 - 0.01 1 8 5 7 7 0.01 - -0.01 11 12 11 18 -0.01 - -0.03 1 10 11 7 9 -0.03 - -0.05 3 12 8 10 -0.05 - -0.07 7 7 1 7 -0.07 - -0.09 4 2 2 3 -0.09 - -0.11 3 2 2 1 -0.11 - -0.13 3 3 1 -0.13 - -0.15 1 1 -0.15 - -0.17 1 1 4 -0.17 - -0.19 2 1 -0.19 - -0.21 -0.21 - -0.23 1 1 -0.23 - -0.25 -0.25 - -0.27 2 -0.27 - -0.29 1 -0.29 - -0.31 1 -0.31 - -0.33 1 a. Vector I b. Vector H 60- 10

50 rri c. Vector UT 10 40 (f) 171 d. Vector iv E --42` 30 10

0

.2 20 e. Vector I- Z 10 10

rt r-t 0 ,rf 0 :11 1.-e -0.12 0 +0.12 -0.12 0 +0.12 FIG 45 Frequency Histograms of the Vector Weights (Correlation Matrix) -283- except 69. The elements must by definition have a mean of I /7 or 0.116, and it can be seen from the histogram (Fig. 45a) that the weights to be attached to all the variables cluster closely around this value, with the exception of variables 69 and 70 which have unusually low weights (-0.029 and +0.025 respectively). The close approximation to 0.116 by most of the elements shows that all the structures participate in the pattern of growth to an almost equal amount. This vector, as in other published cases, is therefore the general size vector in the process of growth. The vector elements with the ten highest and lowest weights are given in Table 48. The range of values in this vector is restricted, there being only slight differences in the weights between the

Table 48 Variables with the ten highest and lowest Weipjhts - Correlation Matrix.

Vector I.

10 Highest Values 10 Lowest Values Variable Weight Variable Weight

37 L. Hind Femur 0.127 7 L. Flagellar Seg. 1 0.097 19 L. Fore Femur 0.127 74 B. Epiproct 0.096 28 L. Mid Femur 0.126 8 L. Flagellar Seg. 2 0.092 47 L. Abd. Tergum 2 0.126 73 L. Epiproct 0.087 48 L. Abd. Tergum 3 0.126 9 L. Flagellar Seg. 3 0.086 20 L. Fore Tibia 0.126 16 B. Metanotum 0.084 26 L. Fore Tarsal Seg. 5 0.12.6 10 L. Flagellar Seg. 4 0.079 4 L. Max. Palp Seg. 5 0.126 11 B. Flagellar Seg. 1 0.062 49 L. Abd. Tergum 4 0.126 70 L. Lt Style Sternum 9 0.025 60 L. Abd. Sternum 3 0.126 69 L. Rt Style Sternum 9 -0.039 -284-

Table 49 Variables with Positive and Negative Weights absolutely greater than the Mean Value.

a. Vector II.

Positive Negative E Variable Weight i Variable Weight i 10 L. Flagellar Seg. 4 0.252 73 L. Epiproct -0.277 9 L. Flagellar Seg. 3 0.247i 16 B. Metanotum -0.264 8 L. Flagellar Seg. 2 0.2441 11 B. Flagellar Seg. 1 -0.261 71 L. Rt. Side Sternum 9 0.236 174 B. Epiproct -0.226 17 Lateral L. Metanotum 0.233 ;69 L. Rt. Style Sternum 9 -0.133 72 L. Lt. Side Sternum 9 0.231 156B. Abd. Tergum 4 -0.119 66 L. Abd. Sternum 9 0.225 1 15 B. Pronotum -0.117 18 Lateral L. Mesonotum 0.219 i 65 L. Abd. Sternum 8 0.190 1 7 L. Flagellar Seg. 1 0.157 54 L. Abd. Tergum 9 0.156 53 L. Abd. Tergum 8 0.142f 52 L. Abd. Tergum 7 0.142 i 1

b. Vector III.

Positive Negative Variable Weight Variable Weight

70 L. Lt Stylo Sternum 9 0.450 64 L. Abd. Sternum 7 -0.310 69 L. Rt Style Sternum 9 0.440 74 B. Epiproct -0.159 11 B. Flagellar Seg. 1 0.344 66 L. Abd. Sternum 9 0.279 16 B. Metanotum 0.233 7 L. Flagellar Seg. 1 0.198 72 L. Lt Side Sternum 9 0.178 71 L. Rt Side Sternum 9 0.143 6 L. Pedicel 0.133 -285-

Table 50 Variables with Positive and Negative Weights absolutely greater than the Mean Value. a. Vector IV.

Positive Negative Variable Weight Variable Weight

70 L. Lt Style Sternum 9 0.409 69 L. Pt Style Sternum 9 -0.310 27 L. Fore Tarsal Claw 0.253 42 L. Hind Tarsal Seg. 3 -0.211 10 L. Flagellar Seg. 4 0.248 40 L. Hind Tarsal Seg. 1 -0.190 36 L. Mid Tarsal Claw 0.246 41 L. Hind Tarsal Seg. 2 -0.190 45 L. Hind Tarsal Claw 0.227 24 L. Fore Tarsal Seg. 3 -0.180 9 L. Flagellar Seg. 3 0.192 32 L. Mid Tarsal Seg. 2 -0.161 8 L. Flagellar Seg. 2 0.183 23 L. Fore Tarsal Seg. 2 -0.161 71 L. Rt Side Sternum 9 -0.156 33 L. Mid Tarsal Seg. 3 -0.154 11 B. Flagellar Seg. 1 -0.147 66 L. Abd. Sternum 9 -0.129

b. Vector V.

Positive Negative Variable :Weight Variable Weight

69 L. Rt Style Sternum 9i 0.465 70 L. Lt Style Sternum 9 -0.371 10 L. Flagellar Seg. 4 0.424 72 L. Lt Side Sternum 9 -0.177 9 L. Flagellar Seg. 3 0.401 17 Lateral L. Metanotum -0.120 8 L. Flagellar Seg. 2 0.328 7 L. Flagellar Seg. 1 0.171

L = Length Rt = Right

B = Breadth Lt = Left -286-

variables. It is perhaps of interest that the variables with the highest weights are prominent sclerites e.g. femur, tibia and mid-abdominal tergites and sternites. These sclerites contribute slightly more to the estimate of general growth, whereas the variables with the lower weights are rather less prominent characters many of which are intimately concerned with sexual differentiation. Variables 69 and 70 contribute extremely little to this pattern of growth, obviously due to their absence in the adult.

Vector II.

The remaining vectors have latent roots which account for only relatively small fractions of the total variance, The distribution of the weights of this vector (Fig. 45b), and of the subsequent vectors (Figs. 45c-e),is of a quite different nature to the former vector. Most of the values cluster markedly around zero, indicating that the majority of variables do not participate to any great extent in this pattern of growth. However, a number of variables are characterised by having abnormally large positive or negative weights and it is such characters which define the vector. The distribution of these weights is such that almost equal numbers are of a positive or negative sign and this, coupled with the occurrence of especially high or low weights indicates what factor analysts have termed a "bipolar factor". This is a common situation for the second and subsequent vectors. For this and the remaining vectors, variables with weights exceeding in absolute value the mean of 0.116 were exaTlined, 11 and an interpretation of the growth pattern made, based on the contribution of these characters which are shown for this and the next three vectors in Tables 49&50. This vector represents a mode of growth affecting simultaneously the length of one set of structures (i.e., those which have positive weights: flagellar segments of antennae, posterior abdominal tergites and sternites, -287-

and thoracic terga) and the width of another set of characters (i.e., those which have negative weights: thoracic and abdominal terga and flagellar segment one). This vector would appear to reflect the shape of the body, the length and width of these characters behaving differently. This pattern of growth is, of course, entirely independent of that represented by the previous vector, or of any subsequent vectors.

Vector III.

The distribution of the elements of this vector is similar to the previous vector (Fig. 45c), and again shows the same "bipolar" properties. The variables shown in Table 49b are those characterised by unusually large positive or negative weights for this vector, and are therefore involved in this third pattern of growth. This pattern represents essentially a disproportionate increase in the length of abdominal sternum nine (all aspects measured) and the basal antennal segments relative to the rest of the body, whereas sternum seven shows a relative decrease.

Vector IV.

The "bipolar" nature of this vector is again evident. The characteristically large positive and negative weights are given in Table 50a. This vector represents a mode of growth which is clearly identifiable and which involves the more distal segments of the appendages (legs and antennae). The flagellar segments of the antennae and the pretarsal claws of the three legs contrast in behaviour with the first three tarsal segments (second and third only in mid and hind legs).

Vector V.

This is the last vector which can be identified biologically with any degree of certainty, its latent root accounting for only -288-

1% of the total variance. The variables characterising the vector are given in Table 50b. This vector reflects a mode of growth in which the flagellar segments and the right style of sternum nine increase in length relative to the rest of the body, whereas the left side (including style) of sternum nine exhibits a relative decrease in length.

The vectors extracted from the correlation matrix were used to compute the principal component scores. These scores characterise the particular stage, sex and species from which the original measurements of the variables were made. The average scores for the groups of ten individuals were calculated for each of the first five vectors (Table 51). Since each vector represents an orthogonal or independent growth pattern, the scores for any two vectors for each developmental stage of the two sexes or species can be plotted against each other and the type of growth pattern outlined by the scores for each vector can be identified. In this work only the scores of two vectors have been plotted at a time, although it is possible to consider three vectors in a single diagram (Blackith, Davies & Noy, 1963). All the possible combinations of the first five vectors have been plotted for an initial complete understanding of the role of each vector, although only three diagrams are really necessary to describe the growth patterns represented by these vectors (Figs. 46-50).

Component I. (Figs. 46, 47)

The developmental stages are more widely distributed along this component than the others. This indicates further that the vector is of general importance and that all the instars are influenced by it. However, the significance of this mode of growth appears to increase during development. Sexual dimorphism is apparent even in this vector, the males of both species being dispersed further along this axis than the females. This

-289-

Table 51 Principal Component Scores - Correlation Matrix.

Principal Component Species, Stage and Sex I II III IV V

E. lapponicus Male Adult 9.848 4.329 2.041 0.974 -2.983 5th. Instar 6.859 -0.729 0.890 0.857 -0.569 4th. Instar 5.101 -1.105 0.441 o.688 -0.177 3rd. Instar 3.964 -1.008 0.313 0.556 -0.057 2nd. Instar 3.054 -0.826 0.234 0.420 0.002 1st. Instar 2.322 -0.642 0.195 0.333 0.068

E. lapponicus Female Adult 8.780 0.723 -0.251 1.351 -1:330 5th. Instar 6.416 -1.279 0.147 0.919 -0.291 4th. Instar 4.93o -1.229 0.236 0.736 -0.108 3rd. Instar 3.879 -1.037 0.243 0.572 -0.043 2nd. Instar 3.033 -o.845 0.218 0.427 0.009 1st. Instar 2.317 -0.659 0.188 0.336 0.080

E. panzeri Male Adult 7.536 2.804 1.279 0.049 -1.768 5th. Instar 5.567 -0.582 0.531 0.380 -0.366 4th. Instar 4.227 -0.772 0.320 0.370 -0.179 3rd. Instar 3.347 -0.789 0.208 0.365 -0.077 2nd. Instar 2.593 -0.652 0.141 0.294 -0.032 1st. Instar 2.119 -0.521 0.110 0.254 0.022

E. panzeri Female Adult 6.699 -0.622 -0.521 0.591 -0.554 5th. Instar 5.378 -1.136 0.034 0.650 -0.264 4th. Instar 4.215 -0.983 0.081 0.479 -0.142 3rd. Instar 3.308 -0.846 0.091 0.407 -0.065 2nd. Instar 2.605 -0.645 0.077 0.307 -0.041 1st. Instar 2.132 -0.525 0.104 0.260 0.025 -290-

• E. lapponicus Male o E. lapponicus Female • E panzeri Male ▪ Epanzeri Female

A Adult

-2 10

-2 a. Vector T against if

b. Vector T against M

FIG 46 Growth Patterns revealed by Principal

Component Scores (Correlation Matrix). -291-

a. Vector I against IV • E lapponicus Male o F. lapponicus Female • E panzeri Male E. panzeri Female V

-2 10

-4 b. Vector T against

FIG 47 Growth Patterns revealed by Principal

Component Scores (Correlation Matrix). -292-

3-

A Adult

A -2 -1' 0 1 2 3 4 5

• A

a. Vector II against Ill • E lapponicus Male o E. lapponicus Female 2- • E. panzeri Male o E. panzeri Female

• A

• A

-2 0 1 2 3 4 5

b. Vector l l against IF

FIG 48 Growth Patterns revealed by Principal

Component Scores (Correlation Matrix). -293-

a. Vector II against • E. lapponicus Male O E lapponicus Female 15 • E panzeri Male E. panzeri Female

1 0

05

V

-10 -05 0 05 1.0 15 20

-05 b. Vector Ill against IV

FIG 49 Growth Patterns revealed by Principal

Component Scores (Correlation Matrix). -294-

a. Vector III against

V

'V

• E. lapponicus Male o E lapponicus Female

■ E. panzeri Male E_panzeri Female

A Adult -3

b. Vector IV against

FIG 50 Growth Patterns revealed by Principal

Component Scores (Correlation Matrix). -295- dimorphism is slight in the first four instars but becomes more obvious in the fifth instar and particularly in the adult. The growth pattern represented by this general size vector is of a similar nature in the two species, the displacement along this axis by E. panzeri is, however, less.

Component II. (Figs. 46a, 48, 49a)

This component is strikingly characterised by the great transition which occurs during the last growth period. As in the previous case the growth pattern represented by this component is similar in the two species. There is virtually no sexual dimorphism apparent on this component in the first three instars, but the fifth instar and adult undergo a complete reversal in their displacement along this axis. The males and females are similar in this respect but the males are distributed much further along the axis than the females. The mode of growth represented by the second vector is therefore principally a sexually dimorphic pattern. It is of interest to recall that this vector has previously been defined (by consideration of the sizes of the vector elements) as one which reflects patterns of growth influencing sexually dimorphic characters.

Component III. (Figs. 46b, 48a, 49b, 50a)

Unlike the previous vector, sexual differences are dispersed along this vector throughout development to a more or less equal extent, and the transition to the adult is not so marked. The first instar in E. panzeri and the first two in E. lapponicus are, however, similar. The two species are identical in the type of growth pattern defined by this vector, but the two sexes of both species are in fact displaced along this axis in opposite directions (particularly well illustrated in Fig. 49b). The growth -296- pattern already identified with this vector (Page 286) is one which includes all the measured characters of sternum nine. The sexual differences in this sclerite are gradually acquired throughout development, whereas those characters which identified Vector II (e.g. lateral length of meso- and metanota) are variables which undergo sudden structural modification at the final ecdysis. This may well explain the differences between the growth patterns defined by the two vectors.

Component IV. (Figs. 47a, 48b, 49b, 50b)

The two species are clearly distinguished through the growth patterns represented by this component. The early instars of the two species are only slightly different, but in the fourth instar sexual differences occur and the species become more separate. The sexes of the two species are displaced along this axis in the same direction, but the species are displaced in opposite directions. This displacement is gradually acquired throughout development (Fig. 47a). By consideration of the vector elements this pattern of growth involves the tarsal and flagellar segments.

Component V. (Figs. 47b, 1+9a, 50)

Sexually dimorphic growth patterns are again reflected in this component. The first three stages show little dimorphism, but this increases in the fourth and fifth instars and the last ecdysis is accompanied by a large displacement along this axis, though one that is not as marked as that characterising the second vector. The growth pattern indicated by the vector elements includes both the ninth sternum and the lateral length of the metanotum, both of which are intimately involved in the sexual dimorphism of the adult.

The most striking features of the growth patterns revealed -297-

by the principal component scores include:- (I) The marked similarity between the two species. (ii) The late onset of sexual dimorphism. All the developmental stages are ranked only along the first vector. (iii) The abrupt change at the final moult shown by all vectors. Such a marked discontinuity is perhaps unexpected in a hemimetabolous insect, though it is widely agreed that even in this case the transition to the adult is the most significant mQrphogenetic part of postembryonic development (Wigglesworth, 1954).

b. Principal Component Analysis - Log. Covariance Matrix.

In most published works a principal component analysis has been based on either a correlation or a covariance matrix. The repetition of the method on the two types of matrix in this work is of considerable value, since the two methods can be compared for similarity of resultant growth trends, and a conclusion may be reached as to the best matrix to use for the data in question. The logarithmic transformation was used to standardise the variances. This necessitated the omission of Variables 69 and 70 (with zero values), but it was thought that in such a large matrix the effect of their elimination would be negligible. An even higher percentage of the total variance is accounted for by the first ten latent roots (99.22%). Using the log. covariance matrix only 0.5% of the total variance is accounted for by the fifth latent root (Table 52). However, the associated latent vector has been examined for comparative purposes, as it was extracted from the correlation matrix. The vectors were examined in the two ways described previously. The elements of the first five vectors were considered in detail; their frequency distribution is given in Table 53; frequency histograms of the weights appropriate to each variable were similar to those constructed for the correlation matrix (Fig. 45). -298-

Table 52 Percentages and Cumulative Percentages of the Total Variance associated with the Latent Roots of the Log. Covariance Matrix.

Latent Percentage Variance Cumulative Percentage Root Variance

89.99 89.99 2 4.78 94.78 3 1.73 96.51 4 1.03 97.54 5 0.51 98.05 6 0.42 98.47 7 0.24 98.71 8 0.20 98.90 9 0.19 99.09 10 0.13 99.22

Vector I.

The contributions (weights) of the characters to this pattern of growth are more variable than the weights derived from a correlation matrix. However, there is still a marked tendency for the weights to cluster around the average value, which is by definition 7,7Th = 0.118; indicating that this vector is also of general importance and that most characters contribute to tho growth pattern to about the average extent. All the variables are characterised by positive weights. The ten variables with the highest and lowest weights (Table 54) are, with a few exceptions, completely different from the set of variables selected in a similar way from the first latent vector of the correlation matrix -299- I - V - Table 53 Frequency Distribution of Elements of Vectors Log. Covariance Matrix.

Class I I II III IV V 0.49 - 0.47 1 0.47 - 0.45 0.45 - 0.43 1 0.43 - 0.41 0.41 - 0.39 2 0.39 - 0.37 0.37 - 0.35 1 0.35 - 0.33 1 0.33 - 0.31 0.31 - 0.29 1 0.29 - 0.27 0.27 - 0.25 2 1 0.25 - 0.23 1 0.23 - 0.21 3 1 2 0.21 - 0.19 1 3 0.19 - 0.17 1 2 0.17 - 0.15 2 1 0.15 - 0.13 9 1 1 1 1 0.13 - 0.11 19 2 1 3 1 0.11 - 0.09 15 6 3 1 0.09 - 0.07 12 3 4 3 5 0.07 - 0.05 7 12 5 4 3 0.05 - 0.03 3 18 2 8 4 0.03 - 0.01 8 8 5 6 0.01 - -0.01 5 18 7 8 -0.01 - -0.03 1 15 3 14 -0.03 - -0.05 7 5 8 -0.05 - -0.07 1 3 4 -0.07 - -0.09 1 5 -0.09 - -0.11 6 3 -0.11 - -0.13 1 1 3 3 -0.13 - -0.15 1 1 2 -0.15 - -0.17 2 2 -0.17 - -0.19 2 1 -0.19 - -0.21 1 1 1 -0.21 - -0.23 1 2 -0.23 - -0.25 5 -0.25 - -0.27 1 -0.27 - -0.29 1 -0.29 - -0.31 -0.31 - -0.33 -0.33 - -0.35 1 -0.35 - -0.37 -0.37 - -0.39 -0.41 - -0.43 -0.43 - -0.45 1 -0.45 - -0.47 -0.47 - -0.49 1 -300--

Table 54 Variables with the Ten Highest and Lowest Weights - Log. Covariance Matrix.

Vector I.

10 Highest Values 10 Lowest Values --- Variable Weight Variable Weight

18 Lateral L. Mesonotum 0.242 25 L. Fore Tarsal Seg. 4 0.070 17 Lateral L. Metanotum 0.225 1 B. Frans 0.068 72 L. Lt Side Sternum 9 0.220 8 L. Flagellar Seg. 2 0.066 71 L. Rt Side Sternum 9 0.219 34 L. Flagellar Seg. 4 0.066 64 L. Abd. Sternum 7 0.176 6 L. Pedicel 0.061 66 L. Abd. Sternum 9 0.166 43 L. Hind Tarsal Seg. 4 0.061 65 L. Abd. Sternum 8 0.161 9 L. Flagellar Seg. 3 0.054 53 L. Abd. Tergum 8 0.146 10 L. Flagellar Seg. 4 0.044 52 L. Abd. Tergum 7 0.143 7 L. Flagellar Seg. 1 0.041 22 L. Fore Tarsal Seg. 1 0.143 11 B. Flagellar Seg. 1 0.031

(Table 48), with the exception of the antennal segments which have low weights in both cases.

Vector II.

The distribution of the weights of the second and subsequent vectors is again such that the majority of weights cluster around zero, with only a few elements with unusually large negative and positive weights, The variables with the largest positive and negative weights (Table 55a) are strikingly similar to those previously cited (Table 49a) based on the correlation matrix, though their accompanying signs are completely reversed (as is also true of the subsequent vectors). This fact is of no biological significance since the latent vectors are, by definition, arbitrarily standardised. The important point is that the same sets of -301-

characters are contrasted. As with the correlation matrix therefore, this vector represents a pattern of growth which contrasts the growth of the length and width of the body, and describes the general shape of the body.

Vector III.

The third vector has unusually large weights for the characters given in Table 55b. As in the previous vector, many of these characters exhibited equally high weights when this latent vector from the correlation matrix was considered (Table 49b). The pattern of growth reflected by the vector represents a disproportion- ate increase in the length of all measured characters of the ninth sternum and the antennal segments including in this case also the distal flagellar segments.

Vector IV.

The characters listed in Table 56a show unusually large positive or negative weights on this vector and are therefore the features implicated in this growth pattern. There are some differences in the characters which make significant contributions to the growth pattern based on the log. covariance as compared with the correlation matrix (Table 50a). However, once again the vector represents a mode of growth involving the appendages of the body; and flagellar segments and the pretarsal claw. The tarsal segments have smaller weights when the log. covariance matrix is used, and do not contribute to any extent in the pattern of growth.

Vector V.

Again this vector (Table 56b) shows some slight similarity with that extracted from the correlation matrix (Table 50b). The vector in the present case represents a pattern of growth which -302-

Table 55 Variables with Positive and Negative Weights absolutely greater than the Mean Value. a. Vector II.

Positive Negative Variable Weight Variable Weight

73 L. Epiproct 0.258 66 L. Abd. Sternum 9 -0.344 74 B. Epiproct 0.258 8 L. Flagellar Seg. 2 -0.255 64 L. Abd. Sternum 7 0.217 65 L. Abd. Sternum 8 -0.246 16 B. Metanotum 0.199 17 Lateral L. Metanotum -0.244 67 B. Abd. Sternum 4 0.131 71 L. Rt Side Sternum 9 -0.237 56 B. Abd. Tergum 4 0.129 9 L. Flagellar Seg. 3 -0.231 15 B. Pronotum 0.122 72 L. Lt Side Sternum 9 -0.231 10 L. Flagellar Seg. 4 -0.217 18 Lateral L. Mesonotum -0.203

b. Vector III.

Positive Negative Variable Weight Variable Weight

8 L. Flagellar Seg. 1 0.401 66 L. Abd. Sternum 9 -0.489 9 L. Flagellar Seg. 2 0.397 11 B. Flagellar Seg. 1 -0.187 10 L. Flagellar Seg. 3 0.362 16 B. Metanotum -0.166 64 L. Abd. Sternum 7 0.300 71 L. Rt Side Sternum 9 -0.153 5 L. Scape 0.144 72 L. Lt Side Sternum 9 -0.136 65 L. Abd. Sternum 8 -0.127 -303-

Table 26 Variables with Positive and Negative Weights absolutely greater than the Mean Value. a. Vector IV.

Positive Negative Variable Weight Variable Weight

66 L. Abd. Sternum 9 0.254 64 L. Abd. Sternum 7 -0.284 45 L. Hind Tarsal Claw 0.220 65 L. Abd. Sternum 8 -0.197 36 L. Mid Tarsal Claw 0.213 18 Lateral L. Mesonotum -0.131 27 L. Fore Tarsal Claw 0.210 41 L. Hind Tarsal Seg. 2 -0.128 8 L. Flagellar Seg. 2 0.204 40 L. Hind Tarsal Seg. 1 -0.118 10 L. Flagellar Seg. 4 0.197 11 B. Flagellar Seg. 1 0.183 9 L. Flagellar Seg. 3 0.175 7 L. Flagellar Seg. 1 0.157 14 L. Metanotum 0.140 55 L. Abd. Tergum 10 0.130 15 B. Pronotum 0.119

b. Vector V.

Positive 1 Negative Variable Weight Variable Weight

17 Lateral L. Metanotum 0.487 65 L. Abd. Sternum 8 -0.438 18 Lateral L. Mesonotum 0.333 8 L. Flagellar Seg. 2 -0.228 64 L. Abd. Sternum 7 0.136 9 L. Flagellar Seg. 3 -0.210 57 L. Paraproct 0.129 10 L. Flagellar Seg. 4 -0.210 66 L. Abd. Sternum 9 -0.170 73 L. Epiproct -0.165 74 B. Epiproct -0.150 41 L. Hind Tarsal Seg. 2 -0.144 42 L. Hind Tarsal Seg. 3 -0.141 24 L. Fore Tarsal Seg. 3 -0.126

L = Length Rt = Right

B = Breadth Lt = Left -304- involves the flagellar segments, sternum eight and nine, and the lateral length of the meso- and metanota which have very high positive weights. These characters contribute little to this mode of growth when the correlation matrix is used.

In general, the most interesting feature of the comparison is the relative similarity of the vectors. One would expect some differences between the vectors extracted from correlation and log. covariance matrices, but it is gratifying to see that many variables are consistently characterised by large weights in the two sets of vectors. Such a result indicates the efficiency of this type of biometric analysis on data of tiiis nature. The differences in the first vector are obviously due to the fact that all the characters contribute to a more or less equal extent to this growth pattern, so that the differences between the weights for the variables are at least partly due to sampling error. The second and third vectors are very similar in their biological identity. It is perhaps surprising to find any consistency in the fourth and fifth vectors as these account for such a very small percentage.of the total variance (less than 2% and 1% respectively); however, certain characters are common to the two sets of vectors.

Principal component scores were computed from the latent vectors extracted from the log. covariance matrix (Table 57). Each set of scores represents an independent pattern of growth throughout the developmental stages, and once again all possible combinations of.the first five vectors were plotted against each other. However, only a selection of these are illustrated (Figs. 51&52). The growth patterns revealed by these diagrams are at first sight less easy to interpret; but certain features are evident, and are discussed in comparison with those derived from the correlation matrix. -305-

Table 57 Principal Component Scores - Log. Covariance Matrix.

Principal Component Species, Stage and Sex I II III IV V

E. lapponicus Male Adult -0.417 -0.785 -1.489 -1.059 2.134 5th. Instar -1.621 0.050 -1.637 -0.829 1.804 4th. Instar -2.760 0.255 -1.661 -0.945 1.894 3rd. Instar -3.651 0.214 -1.613 -1.029 1.954 2nd. Instar -4.530 0.035 -1.508 -1.079 1.952 1st. Instar -5.466 -0.314 -1.258 -0.956 1.904

E. lapponicus Female Adult -0.892 0.142 -0.999 -1.234 2.067 5th. Instar -1.972 0.545 -1.286 -0.947 2.040 4th. Instar -2.933 0.483 -1.349 -1.028 2.069 3rd. Instar -3.758 0.347 -1.474 -1.086 2.101 2nd. Instar -4.573 0.134 -1.470 -1.111 2.038 1st. Instar -5.493 -0.235 -1.189 -0.953 1.889

E. panzeri Male Adult -1.174 -0.852 -1.600 -1.387 1.950 5th. Instar -2.263 0.004 -1.708 -1.102 1.724 4th. Instar -3.293 0.154 -1.820 -1.215 1.858 3rd. Instar -4.169 0.174 -1.741 -1.262 1.944 2nd. Instar -5.032 0.025 -1.646 -1.320 1.996 1st. Instar -5.654 -0.289 -1.290 -1.178 1.875

E. panzeri Female Adult -1.697 0.375 -1.063 -1.497 1.701 5th. Instar -2.540 0.509 -1.317 -1.192 1.887 4th. Instar -3.408 0.406 -1.640 -1.275 1.881 3rd. Instar -4.275 0.351 -1.532 -1.326 2.104 2nd. Instar -5.035 0.076 -1.511 -1.377 2.051 1st. Instar -5.645 -0.294 -1.232 -1.156 1.867

-306 -

2-

A Adult

-8 -2 0 2

a...... 6"••••■■09.2111---71 74 A T

-2-

-4- a. Vector T against IV

• E lapponicus Male 1 05- E. lapponicus Female • E. panzeri Male I V E panzeri Female

-2.0 -1.5 -05 0 05

111 -So

-05-

-1.0-

-1.5- b. Vector III against IV

FIG 51 Growth Patterns revealed by Principal

Component Scores (Log. Covariance Matrix). a. Vector if against Ill

• E laQponicus Male

1.5 o Ej_gpponicus Female

• E. panzeri Male

E panzeri Female

-1.0 -05 0 05 1.0

-0.5

b. Vector n against

FIG 52 Growth Patterns revealed by Principal

Component Scores (Log. Covariance Matrix). -308-

Component I. (Fig. 51a)

All the developmental stages are again dispersed along this component to a greater extent than along any of the other axes, confirming the general size identity of the vector. When the scores are derived from the log. covariance matrix, the instars are more evenly ranked along this vector, and there is a less marked transition prior to the adult stage. The species again reveal similar growth patterns and on this axis the sexual dimorphism is less marked though the males are displaced along it to a greater extent than the females.

Component II. (Fig. 52)

This represents a pattern of growth which is principally concerned with sexual dimorphism. When the component scores were based on the correlation matrix, the pattern of growth was completely characterised by the abrupt transition prior to the final ecdysis. This transition is very much smaller when the log. covariance matrix is used though the basic pattern includes a complete reversal in displacement along the axis. The males are again displaced further along this axis than the females.

Component III. (Figs. 51b, 52a)

The third component also reflects a pattern of growth principally concerned with sexual differentiation. A clear interpretation of this axis is more difficult though the sexes are dispersed along it in opposite directions (which was also the case when the analysis was based on the correlation matrix). As in the previous component, sexual dimorphism is not marked in the first three instars but becomes gradually more apparent in the later ones. -309-

Component IV. (Fig. 51)

As before, the scores derived from this vector clearly define species differences, which are made more obvious by using the log. covariance matrix, since they can now be recognised from the first instar. These differences continue with little change throughout development.

Component V. (Fig. 52b)

When the scores were computed from vectors extracted from the correlation matrix, the growth patterns defined by the fifth vector showed clear sexual dimorphism. When the log. covariance matrix is used this is far less obvious and may well be explained by the low percentage of the variance absorbed by this vector.

The growth patterns which can be identified from the principal component scores of the separate vectors are similar in nature whether a correlation or log. covariance matrix is used. However, the interpretations of the growth patterns identified by the use of the correlation matrix are far more rational, and suggest that for this particular type of data it is preferable to base a principal component analysis on a correlation matrix.

c. The Generalised Allometry Parameter.

Since the first principal component of the log. covariance matrix has been physically identified as representing the generalisation of the allometry equation (Jolicoeur, 1963a & b), it is of value to consider this component in more detail. The choice of a suitable reference dimension in the study of allometry has already been discussed (Page 20S),and it was pointed out that the best character would be one which includes all the other measured -310- variables simultaneously. Teissier (1960) suggested a general criterion based on the first principal component of the correlation matrix of logarithmically transformed data. However, the use.of the log. covariance matrix was advocated by Jolicoeur (1963b), since the physical interpretation is then much simpler. Separate log. covariance matrices were computed for each species and sex, and the latent roots and associated vectors extracted. From a study of the latent vectors, it was possible:- (i) To consider in detail the elements of the first latent vector from which the generalised allometry exponents can be derived. (ii) To compare the latent vectors for each sex and species and hence to determine whether the vectors associated with the single log. covariance matrix over all stages, species and sexes are comparable with those derived from the matrices for a single sex and species.

The generalised allometry exponents were computed by 1 dividing each element of the first latent vector by W. i.e. the '72 average weight for each element of that vector (0.118). These generalised exponents were calculated for each sex and species in turn (Table 58). Their values can then be compared directly with the values of a based on body length as a reference dimension. The values of a computed by using Bartlett's "best fit" method (Tables 30-33, Pages 211-218) are similar to the values derived from the generalisation of the allometry equation; the mean difference between the males and females of E. lapponicus and E. panzeri being 0.028, 0.035, 0.058 and 0.037 respectively. However, few characters show consistently high or low differences, in the two sexes and species. In the study of allometry of growth many characters showed significant deviations from linearity (Table 36, Page 226). It was thought that these characters may have large weights for the -311-

Table 58 Generalised Allometry Exponents.

Vari- E. lapponicus E. lapponicus E,...RIELen1 E. panzeril able Male Female Male Female r

- 0.552 0.657 N 0.486 0.596 0.564 0.703 0.550 0.679 Pc

\ 1 0.640 0.725 0.668 0.802 0.890 0.917 0.839 0.880 .r 1 0.848 0.769 0.770 0.750 M3 o.485 EN 0.518 0.434 0.533 0.288 -0 0.353 0.367 0.291 0 0.653 0.390 CY 0.566 0.448

N 0 0.442 0.376 0.564 0.346 v- 0.241 r 0.406 0.266 0.447 r

- 0.311 0.263 0.266 0.144 V- N 0.872 1.035 0.835 0.993 tr C-

\ If 0.927 0.777 0.878 0.761 r

- 0.981 1.038 0.876 1.037 c - \ 0.748 0.900 0.823 1.043 O 0.532 0.574 0.503 0.761 Cs r - -00 2.147 1.872 2.083 1.160 r

- 1.630 ON 2.160 2.075 2.148 v - 0.932 0.951 1.026 0 0.860 N 1.000 r 0.936 0.959 1.005 a JC - o.858 N 0.664 0.804 0.728 \IN

1.174 1.189 1.245 1.303 0.975 0.999 1.149 1.211 C \ IN 0.807 0.867 1.050 1.056 I A 0.525 0.613 N 0.554 0.669 N 0.737 0.763 0.717 0.771 RIC

- CO 0.697 0.711 0.424 0.607 1IN

O 0.968 0.978 0.984 1.070

N 1.036 1.099 1.123 0 1.067 Cr

\te 0.918 r 0.852 0.841 0.732 - % 1.079 1.134 1.136 1.223 rc N Isr

l N 0.892 0.963 1.036 1.075 \ %nte 0.777 0.814 0.965 0.989 0.538 0.644 0.481 0.562 1 -r \r w 0.681 0.753 0.671 0.718 ) \ 0.409 0.605 0.605 0.697 -312-

Table 58 Contd.

Vari- E. lapponicus E. lapponicus E. panzeri E. panzeri able Male Female Male Female

37 0.939 0.998 1.000 1.031 38 1.097 1.123 1.167 1.219 39 0.710 0.868 0.760 0.957 40 1.083 1.161 1.149 1.257 41 0.798 0.892 0.978 1.051 1+2 0.687 0.742 0.894 0.921 43 0.436 0.591 0.479 0.535 44 0.650 0.698 0.635 0.707 45 0.607 0.654 0.438 0.579 46 0.947 0.984 0.855 0.936 47 0.938 0.986 0.833 0.967 48 0.999 1.076 0.871 1.005 49 1.024 1.091 0.898 1.041 50 1.026 1.108 0.918 1.071 51 1.059 1.119 0.977 1.113 52 1.272 1.131 1.297 1.165 53 1.190 1.145 1.371 1.247 54 1.178 1.121 1.306 1.159 55 0.908 0.777 0.890 0.890 56 0.717 0.984 0.662 0.999 57 1.030 1.199 1.028 1.145 58 1.028 1.127 0.939 1.099 59 0.884 1.00o 0.839 0.931 6o 1.073 1.146 0.940 1.098 61 1.091 1.161 0.911 1.087 62 1.098 1.168 0.916 1.094 63 1.140 1.271 0.977 1.171 64 1.269 1.970 1.113 1.911 65 1.516 1.136 1.478 1.227 66 1.753 0.936 1.880 0.669 67 0.766 1.006 0.673 1.050 68 0.776 1.005 0.732 1.057 71 2.038 1.480 2.267 1.474 72 2.23o 1.486 2.052 1.461 73 0.413 0.763 0.343 0.991 74 0.570 0.958 0.553 1.098

Variables can be identified from the list on Pages 189-191. -313-

second and subsequent eigenvectors which principally reflect changes in shape of the body and that this may well provide an explanation for their deviation from simple allometry. With respect to the %1A-t, second vector their seems to be a definite connection between those variables which deviate significantly from linearity and those which contribute most to this pattern of growth, describing the general shape of the body. However, so few characters show non-significant deviations from linearity that this correlation is difficult to clarify with certainty. This connection is also discernible in the third vector, but in subsequent vectors it becomes less obvious. However, it will be recalled that only a very small proportion of the total variance is attributed to these vectors (Table 59). From these relatively small differences it would appear that the allometry constants calculated on the basis of an accurate measurement of body length (the total length of the thoracic and abdominal tergites in the mid-dorsal line) represent a satisfactory reference dimension. The calculation of the generalised allometry exponents apparently contribute little new "information" though it is a theoretically preferable method, and should be used whenever a true multivariate analysis of allometry in the ontogeny of a species has to be made.

The latent roots indicate that the first three account for a very high percentage of the total variance in each sex and species, and are perhaps the only vectors worthy of interpretation (Tables 59&60). However, for comparative purposes the first five vectors have been considered. The latent root of the first vector accounts for 93% to 95% of the total variance in each of the sexes and species. All the variables are characterised by positive coefficients for this vector. The ten variables with the highest and lowest weights are remarkably consistent in the two species and sexes. However, Variable 66 (length of sternum nine) only has a high weight in the males of the two species. -314-

For the remaining vectors the variables with unusually large weights for the second vector are very similar in each of the separate analyses, and resemble closely those listed in.Table 55b. when all the sexes and species are considered. However, some variables show a sexual difference in the nature of their weights e.g. Variable 65 (length of sternum eight) has large weights in the females of the two species, whereas Variables 66, 71 and 72 (length of sternum nine - three aspects) have large weights for the males only. Those variables with extreme values for their weights on the third vector show less consistency between the.two species and sexes, than the previous two vectors. However, this is to be expected since the vector absorbs such a small proportion of the total variance (0.48 - 1.25%). In the remaining two vectors there is some similarity in the identity of the growth patterns represented; but this is limited to a few variables only. The principal component analyses involving all the developmental stages of the two sexes and species in turn results in the identity of certain distinct growth patterns from a study of the elements of the principal vectors. Many of these growth patterns are common to both the sexes and species. The single principal component analysis which embra.ced all the developmental stages, sexes and species therefore adequately describes growth patterns which are in fact relevant to all the individual sexes and species, although certain sexual differences are masked in an examination of the vector weights. -315-

Table 59 Percentages of the Total Variance Associated with the Latent Roots of the Log. Covariance Matrix.

Species and Sex Latent Root

(all stages) I II III IV V

E. lapponicus Male 94.46 4.34 0.48 0.15 0.09

E. lapponicus Female 94.14 3.10 1.25 0.73 0.17

E. panzeri Male 93.54 4.77 0.83 0.17 0.13

E. panzeri Female 93.09 3.47 1,11 0.85 0.52

Table 60 Cumulative Percentages of the Total Variance Associated with the Latent Roots of the Log. Covariance Matrix.

Species and Sex Latent Root (all stages) I II III IV

E. laonicus Male 94.46 98.80 99.28 99.43 99.52

E. lapponicus Female 94.14 97.24 98.49 99.22 99.39

E. panzer! Male 93.54 98.31 99.14 99.31 99.44

E. panzeri Female 93.09 96.56 97.67 98.52 99.04 -316-

(iii) Factor Analysis.

Opinion is divided on the best means to conduct a factor analysis. The method of maximum likelihood due to Lawley & Maxwell (1963) is now generally thought to be the most satisfactory statistically, but it has been shown that their computational method does not result in satisfactory convergence in the iterative calculation of vectors and a workable alternative was not available as a computer program. Instead, an attempt was made to carry out a factor analysis of the association matrix, by an extension of the principal component solution. The latent roots and eigenvectors were extracted from a covariance matrix based on logarithmically transformed data from all developmental stages, sexes and species, as previously described (Page 277). The normalised vectors are then scaled so that the sums of the squares of their elements nre equal to the corresponding latent root; these vectors constitute the "unrotated factor matrix". A new set of axes is then formed by rotating these principal component axes. This rotation was performed by using the Varimax rotation developed by Kaiser (1958), which was available as an IBM program (S.S.P. VARMX). Unlike certain methods of oblique rotation advocated by Gould (1966), the Varimax rotation provides an orthogonal solution in which the reference vectors remain uncorrelated. In general it is often found that the new axes are "simpler" after Varimax rotation, since the large loadings are maximised whilst the small ones are minimised. This tends to produce factors dominated by a few variables with high loadings whilst the remainder have loadings near to zero. The rotated matrix was then examined in the hope that it would provide a clearer indication of which variables were most important in the biological interpretation of the various factors. For the first five scaled vectors of the unrotated factor matrix those variables with the twenty highest loadings were -317- compared with the variables which after rotation had the highest weights. It was hoped that after rotation the variables with the highest loadings for any vector prior to rotation would be stressed. However, this was not found to be the case, and after rotation some "new" variables emerged with high loadings in each of the vectors, whilst others which had high relative loadings in the unrotated factor matrix had loadings reduced in magnitude after rotation. Thus Varimax rotation of the unrotated factor matrix was not able to simplify the original axes to any significant degree, due presumably to the relative positioning of the variables with respect to the unrotated principal component axes. Cooley 2t Lohnes (1962) suggest that the unsatisfactory solution by the Varimax method may be due to the fact that during rotation any general factors are "destroyed" and the variance associated with them tends to be distributed to other factors. In the present work the first factor was found to be a general size factor. From the rotated factor matrix one would ideally have liked to compute a set of factor scores comparable with those derived by principal component analysis. Of the various methods used for this purpose that due to Thompson (1951) and also described ay Morrison (1967 : 291) seemed most suitable. This method depends on being able to calculate the inverse of the log. covariance matrix. However, it appears that large correlation or covariance matrices of the kind used in growth studies are often singular (i.e. have no inverse). The reasons for this are not entirely clear but the explanation appears to be that a correlation or covariance matrix based on n replicates of k variables must, when n is less than k as in the case here, be of rank not greater than n. This means that it has at least (k.- n) zero latent roots, and is therefore singular or will appear, through a rounding-off error, to have a very small determinant, Morrison (1967) considers only matrices of full rank in developing the theory of principal components. The whole question of matrices of lower rank seems to be ignored in standard accounts of multivariate analysis, yet if many variables -318- are to be investigated there are great practical difficulties in examining a sufficient number of replicates to ensure that the covariance or correlation matrices are of full rank. The matrices involved in this study proved to be singular, and thus a full factor analysis was not feasible. The subject, however, deserves further detailed study. -319-

(iv) Multiple Discriminant Analysis.

Introduction to the Technique.

Multiple discriminant analysis provides an efficient method of discriminating between the stages of each species and sex. Its aim is to find a linear combination of the variables which will maximise the ratio of the between-group to the within-group variance. In this case the groups are the six developmental stages of each sex and species (i.e. 24 in all) and the variables are the same 74 characters as before. Such a linear compound is an extension of the discriminant function between two groups as devised by Fisher (1938). In this case there are 23 separate discriminant functions, the coefficients being the elements of the 23 non-zero roots of the non-symmetric matrix C = W-1B (where W = pooled within-group variance and B = between-group variance). The extent of the resulting separation, the generalised distance of Mahlanobis (1936), associated with the discriminant functions in this study, is a measure of the difference in the growth patterns between the groups of insects, but it is more satisfactory to assess the differences in terms of those vectors associated with the largest four or five latent roots. Multiple discriminant analysis assumes equality of the within-group covariance matrices, which should therefore be tested for homogeneity prior to attempting the analysis. As in principal component analysis the magnitudes of the latent roots reflect the percentage of the total variance accounted for by each orthogonal axis, and are a direct measure of the extent to which the associated discriminant function distinguishes between the groups. The individual elements of each canonical axis can be examined to determine the contribution of the different variables to the discriminant function. The relative contributions of the variables to the discriminant function may be found by scaling the vectors (as in Factor analysis (Page 316)). -320-

For the first ten canonical variates, discriminant scores (group centroids) were computed. These can be compared by isolating pairs of orthogonal axes and plotting the discriminant scores for the developmental stages of the two sexes and species in the discriminant space. Such plots indicate the extent and the position in development where the greatest discrimination occurs. Two separate analyses were carried out, one using untransformed data, and another using logarithmically transformed data. The results of the two analyses are compared and discussed. A test of the significance of the discrimination obtained, Wilks' lambda test, determines the significance of differences between the group centroids, assuming the within-group covariance matrices to be homogeneous. This test has been carried out on both the untransformed and transformed data, but it is unfortunate, for reasons given on page 3171 that the homogeneity of the covariance. matrices could not be tested.

Test for Homogeneity of Covariance Matrices.

Few workers, other than mathematical statisticians, have considered the subject of significance tests in multivariate analyses. However, multiple discriminant analysis assumes that the within-group covariance matrices are homogeneous. Before commencing this type of analysis it is therefore important to test the equality of the group covariance matrices. This was attempted, using a method due to Box (1949) and also given in Cooley & Lohnes (1962:62) on both the untransformed and the logarithmically transformed data. However, the test is inapplicable when either the matrix is singular (i.e. has no inverse) or when the determinant of the matrix is negative. In this particular analysis, using both untransformed and logarithmically transformed data the pooled within-group matrix was singular and unfortunately the test could not be applied. -321-

The assumption has therefore been made that the within-group covariance matrices are homogeneous. The reasons for the inapplicability of the test appear to be the same as those discussed on page 317 in connectj.on with the attempt to calculate the factor scores.

Outline of Statistical Methods and Details of Programs Utilised.

A single multiple discriminant analysis included all the developmental stages, sexes and species. A FORTRAN IV program written by R.G.Davies computed the between- and within-groups covariance matrices. The ten largest latent roots and vectors of the product matrix W-1B were extracted, utilising the subroutine DIRNM of Cooley & Lohnes (1962). This subroutine calls subroutine EIGEN, used in previous analyses for the extraction of latent roots and vectors. The elements of the computed latent vectors are the coefficients of the discriminant functions. The vectors were normalised by setting the sums of squares of the elements equal to unity. The individual discriminant scores on the first five canonical variates were calculated and the mean for each stage derived. Wilks' lambda test was included in the main program to determine the significance of differences between the group centroids. The complete analysis was repeated using both untransformed and logarithmically transformed data. In the latter case the original variables 69 and 70 were omitted, since these variables were absent in the adult stage, which introduced a zero value into the original data. A FORTRAN IV computer program written by R.G.Davies tested the homogeneity of the covariance matrices of the untransformed and logarithmically transformed data by the method due to Box (1949). -322-

a. Multiple Discriminant Analysis - Untransformed Data.

A single analysis was applied to all the developmental stages, sexes and species simultaneously, using the original, untransformed measurements of the variables. The relative sizes of the latent roots indicate the extent to which the associated discriminant functions distinguish among the groups. The first ten latent roots account for over 99% of the total variance (Table 61).

Table 61 Percentages and Cumulative Percentages of the Total Variance Associated with the Latent Roots - Untransformed Data.

Latent Percentage Variance Cumulative Percentage Root Variance

1 49.18 49.18 2 28.45 77.63 3 7.88 85.51 4 7.01 92.52 5 3.22 95.73 6 1.69 97.42 7 0.95 98.37 8 0.56 98.93 9 0.29 99.22 10 0.24 99.47

4.4

However, this analysis differs from principal component analysis in the relatively large proportion of the total variance attributable to the second latent root (the third to fifth latent roots also account for a higher percentage of the total variance). -323-

In a multiple discriminant analysis of the development of D. fasciatus the second and third canonical variates were also found to account for relatively large proportions of the total variance (Blackith, Davies & Moy, 1963). It is possible that only the first five vectors or canonical variates are capable of biological interpretation, the remainder accounting for such relatively small proportions of the total variance that they can be justifiably ignored. The non-normal distribution of the original data precludes the use of significance tests to categorise the important vectors. The frequency distributions of the elements of the first five canonical variates are given in Fig. 53. It can be seen that the distribution of the weights of the vector elements is similar for each canonical variate. This is in contrast to the principal component analyses where the elements of the first vector consistently approximated to the mean value whereas the \PT4 subsequent vectors with a few exceptions had weights which generally clustered around zero, indicating that all but a select few of the variables made a negligible contribution to the growth pattern. The elements of the first five canonical variates cluster between +0.1 and -0.1 with a few variables having markedly large positive or negative weights. The scatter of the elements is greater than in Vectors II - V of the principal component analyses; but the principle is the same. The clustering of the majority of elements in this discriminatory analysis between 0.4 and 0.6 instead of around zero as in the principal component analyses may be due to the fact that all the characters are contributing a minute amount to the discriminant function. Those variables which have weights in excess of the mean value and therefore contributing most to the discriminant functions are cited in Tables 62 - 64. The canonical variates can in fact be interpreted biologically and the group differences thus examined by the same two methods used in establishing the identity of the principal components:-

NEGA TI VE POSI TI VE

V • ... 17 weig ht • • • • 1 I 7

- 01 - -03 -03 - .05 05 - .07 .07 - .09 09 - .11 -'- •11 13 ---- 13--15 • 15 --1 1 .17 197 .19 - 21 [] [ 23 - 25 [ .25 - 27 [ .27 - 29 ] 29 - 31 ] 31 -33 ] [ 33- 35 35- .37 .37 - .39 39 - .41 41 - .43 .43 - .45 .45 - .47

FIG 53 Frequency Distribution of Elements of Canonical Variates T-V ( Untransformed Data ). -325-

Table 62 Variables with Positive and 7.Tcp.tive Weights absolutel greater than the Mean Value - Untransformed Data. a. Canonical Variate I.

Positive Negative Variable Weight Variable 1Weight

34 L. Mid Tarsal Seg. 4 0.297 43 L. Hind Tarsal Seg. 41 -0.290 50 L. Abd. Tergum 5 0.254 33 L. Mid Tarsal Seg. 3 -0.268 24 L. Fore Tarsal Seg. 3 0.237 51 L. Abd. Tergum 6 -0.223 65 L. Abd. Sternum 8 0.216 1 36 L. Hind Tarsal Claw -0.175 32 L. Mid Tarsal Seg. 2 0.216 48 L. Abd. Tergum 3 -0.166 6 L. Pedicel 0.211 42 L. Hind Tarsal Seg. 3 -0.161 26 L. Fore Tarsal Seg. 5 0.197 22 L. Fore Tarsal Seg. 1 -0.149 3 L. Labial Palp Seg. 3 0.170 21 B. Fore Femur -0.138 29 L. Mid Tibia 0.147 40 L. Hind Tarsal Seg. 1 -0.129 53 L. Abd. Tergum 8 0.139 4 L. Max. Palp Seg. 5 -0.124 47 L. Abd. Tergum 2 0.125 23 L. Fore Tarsal Seg. 2 -0.123 14 L. Metanotum -0.121

V. Canonical Variate II.

Positive Negative Variable Weight Variable Weight

8 L. Flagellar Seg. 2 0.440 10 L. Flagellar Seg. 4 -0.347 64 L. Abd. Sternum 7 0.366 9 L. Flagellar Seg. 3 -0.266 35 L. Hind Tarsal Seg. 5 0.266 11 B. Flagellar Seg. 1 -0.176 36 L. Hind Tarsal Claw 0.193 30 B. Mid Femur -0.160 48 L. Abd. Tergum 3 0.127 33 L. Mid Tarsal Seg. 3 -0.157 50 L. Abd. Tergum 5 0.122 53 L. Abd. Tergum 8 -0.154 51 L. Abd. Tergum 6 -0.142 43 L. Hind Tarsal Seg. 4 -0.136 49 L. Abd. Tergum 4 -0.132 66 L. Abd. Sternum 9 -0.121 -326-

Table 63 Variables with Positive and Negative Weights absolutely greater than the Mean Value a. Canonical Variate III.

Positive Negative Variable Weight Variable Weight

53 L. Abd. Tergum 8 0.233 25 L. Fore Tarsal Seg. 4 -0.339 39 B. Hind Femur 0.225 14 L. Metanotum -0.314 3 L. Labial Paip Seg. 3 0.220 49 L. Abd. Tergum 4 -0.241 43 L. Hind Tarsal Seg. 4 0.207 30 E. Mid Femur -0.185 24 L. Fore Tarsal Seg. 3 0.202 5 L. Scape -0.181 36 L. Mid Tarsal Claw 0.178 72 Lt L. Side Sternum -0.170 54 L. Abd. Tergum 9 0.178 46 L. Abd. Tergum 1 -0.122 47 L. Abd. Tergum 2 0.170 42 L. Hind Tarsal Seg. 3 0.165 50 L. Abd. Tergum 5 0.163 33 L. Mid Tarsal Seg. 3 0.151 71 L. Rt Side Sternum 9 0.148

h. Canonical Variate IV.

Positive Negative Variable Weight Variable Weight

8 Flagellar Seg. 2 0.336 9 L. Flagellar Seg. 3 -0.484 44 L. Hind Tarsal Seg. 5 0.217 25 L. Fore Tarsal Seg. 4 -0.274 66 L. Abd. Sternum 9 0.186 30 B. Mid Femur -0.207 32 L. Mid Tarsal Seg. 2 0.167 50 L. Abd. Tergum 5 -0.197 51 L. Abd. Tergum 6 0.162 54 L. Abd. Tergum 9 -0.167 49 L. Abd. Tergum 4 0.152 73 L. Epiproct -0.155 45 L. Hind Tarsal Claw 0.143 27 L. Fore Tarsal Claw -0.152 23 L. Fore Tarsal Seg. 2 0.143 26 L. Fore Tarsal Seg. 5 -0.126 53 L. Abd. Tergum 8 0.140 39 B. Hind Femur 0.120 -327-

Table 64 Variables with Positive and Negative Weights absolutely greater than the Mean Value.

Canonical Variate V.

Positive Negative

Variable Weight Variable Weight

65 L. Abd. Sternum 8 0.314 36 L. Mid Tarsal Claw -0.447 9 L. Flagellar Seg. 3 0.279 8 L. Flagellar Seg. 2 -0.257 59 L. Abd. Sternum 2 0.219 54 L. Abd. Tergum 9 -0.215 52 L. Abd. Tergum 7 0.165 35 L. Mid Tarsal Seg. 5 -0.207 23 L. Fore Tarsal Seg. 2 0.162 6 L. Pedicel -0.192 44 L. Hind Tarsal Seg. 5 0.152 45 L. :Hind Tarsal Claw -0.176 72 L. Lt Side Sternum 9 0.137 34 L. Mid Tarsal Seg. 4 -0.132 26 L. Fore Tarsal Seg. 5 0.130 46 L. Abd. Tergum 1 -0.126

-__

L = Length Rt = Right

B = Breadth Lt = Left

(i) By examination of the vector elements to indicate those variables with the largest absolute weights. (ii) By computing and plotting the group centroids in terms of each canonical variate, thus relating the latter to the biological attributes of the groups.

From the first five canonical variates the group centroids were computed. These discriminant scores characterise the particular stage, sex and species from which the original measurements were taken. The average scores for the groups of ten individuals of each stage were calculated (Table 65). The scores on any two of the orthogonal canonical variates can be plotted, and the growth -328-

Table 65 Canonical Variates (Group Centroids) - Untransformed Data.

7 Canonical Variate Species, Stage and Sex I II III IV V

E. lapponicus Male Adult 0.993 -0.063 -0.190 0.140 -0.011 5th. Instar 0.138 0.026 -0.199 0.363 0.009 4th. Instar 0.040 0.039 -0.103 0.292 -0.046 3rd. Instar 0.026 0.038 -0.071 0.227 -0.053 2nd. Instar 0.014 0.026 -0.042 0.173 -0.045 1st. Instar 0.016 0.006 -0.017 0.130 -0.052

E. lapponicus Female Adult 0.463 0.517 0.020 0.143 -0.087 5th. Instar 0.056 0.278 -0.126 0.297 -0.147 4th. Instar 0.023 0.158 -0.081 0.250 -0.088 3rd. Instar 0.007 0.083 -0.066 0.213 -0.080 2nd. Instar 0.009 0.037 -0.040 0.171 -0.059 1st. Instar 0.012 0.007 -0.011 0.127 -0.053

E. panzeri Male Adult 0.470 -0.076 0.281 0.354 -0.073 5th. Instar 0.070 0.018 -0.022 0.326 0.028 4th. Instar 0.020 0.025 -0.031 0.253 0.004 3rd. Instar 0.013 0.031 -0.033 0.191 -0.011 2nd. Instar 0.013 0.030 -0.013 0.148 -0.015 1st. Instar 0.016 0.017 -0.004 0.111 -0.026

E. panzeri Female Adult 0.122 0.445 0.031 0.210 0.105 5th. Instar 0.025 0.246 -0.049 0.254 -0.034 kth. Instar 0.021 0.136 -0.052 0.241 0.099 3rd. Instar 0.007 0.087 -0.029 0.174 -0.035 2nd. Instar 0.011 0.045 -0.017 0.140 -0.023 1st. Instar 0.015 0.017 -0.005 0.107 -0.026 -329- patterns which they reflect can be identified biologically with some certainty. All ten combinations of the first five canonical variates have been plotted in this way to assess the discrimination achieved by this analysis. Figs. 548c55 are examples of these growth patterns. It is immediately obvious that the discrimination in the early instars is very poor; the second and fourth canonical variates are the only ones to rank all the instars to any extent. With the exception of the final nymphal instar and the adult there is little or no discrimination between the stages, sexes or species. This suggests that a linear discriminant function derived from the untransformed data, is not the best one and that a quadratic discriminant might be better. Such a situation does in fact arise when the within-group variances are not homogeneous. A quadratic discriminant function is obtained by transforming the original data logarithmically. The results derived using this transformation proved to be more acceptable, in that discrimination between the earlier instars was more marked. It is unfortunate that the tests for the homogeneity of the within-group covariance matrices were not possible, since their results might well have indicated the theoretical desirability of using logarithmically transformed data. However, because of the better discrimination shown after logarithmic transformation, all further detailed discussion has been confined to the analysis of logarithmically transformed data.

b. Multiple Discriminant Analysis - Logarithmically Transformed Data.

The total variance associated with the first ten latent roots (Table 66) closely resembles that absorbed by the first ten latent roots computed from the untransformed data. The percentage of the variance accounted for by each of the first five latent roots is similar, the second latent root again accounting for a much larger proportion of the total variance than in the principal component -330-

it

--2 1.0 A

a. Canonical Variate T against IT

• E. lapponicus Male

• E. lapponicus Female

■ E. panzeri Male

o E. panzer/ Female

A Adult

-.2

b. Canonical Variate T against

FIG 54 Growth Patterns revealed by Multiple

Discriminant Scores (Untransformed Data). -331 -

- • 1 0 • 1 -2 4 •5

-•1 a. Canonical Variate TT against IV

• E. lapponicus Male

o E. lapponicus Female

■ E. panzeri Male

A Epanzeri Female

A Adult

- 1 0 1 2 3 4

b. Canonical Variate 111 against R- FIG 55 Growth Patterns (Untransformed Data). Contd. -332-

Table 66 Percentages and Cumulative Percentages of the Total Variance Associated with the Latent Roots - Logarithmically Transformed Data.

Latent Percentage Variance Cumulative Percentage Root Variance

1 53.87 53.87 2 21.22 75.09 3 8.87 83.96 4 5.73 89.69 5 3.49 93.18 6 2.57 95.74 7 1.28 97.02 8 0.94 97.96 9 0.55 98.51 10 0.36 98.87

analyses, and therefore contributing substantially to the discrimination between the groups. The distributions of the size of the elements of canonical variates I - V (Fig. 56) are similar to those previously cited using untransformed data. Most of the variables contribute relatively little to the discriminant function (elements of 0.0 to ±0.1), and only a few variables have exceptionally large weights for a particular canonical variate. For the first five canonical variates, variables with positive and negative weights above the mean value are listed in Tables 67-69. It is of interest that on every canonical variate, many of the variables which contribute substantially to the discriminant function are isolated segments of the legs. It will be recalled that in the relative growth study NEGATIVE POSI TI VE . . . ..27 . weight TT III V 0 - .01 01 - .03 -03 - .05 .05 - .07 , .07 - .09 J 09 - .11 .11 _.13-- L -13 - .15 1[ /5 - -17 •17 - .19 E •19 - .21 21 - .23 ] 23 - 25 ] -25-.27 [ - - 27- 29 29 - 31 31 - 33 ^ ] .35 - 37 -37 - .39 39 - .41 -41 - -43 .43 - .45 .45 - -47 ] .4 7 - .49

FIG 56 Frequency Distribution of Elements of Canonical Variates T-7 (Log. Transformed Data). -334-

Table 67 Variables with Positive and Negative Wets absolutely greater than-the Mean Value.

a. Canonical Variate I.

Positive Negative Variable WeightI Variable Weight

17 Lateral L. Metanotum 0.493 14 L. Metanotum -0.256 63 L. Abd. Sternum 6 0.315 37 L. Hind Femur -0.239 18 Lateral L. Mesonotum 0.298 62 L. Abd. Sternum 5 -0.198 29 L. Mid Tibia 0.166 38 L. Hind Tibia -0.194 31 L. Mid Tarsal Seg. 1 0.158 51 L. Abd. Tergum 6 -0.135 44 L. Hind Tarsal Seg. 5 0.153 33 L. Mid Tarsal Seg. 3 -0.126

b. Canonical Variate II.

1 Positive Negative Variable Weight Variable Weight

64 L. Abd. Sternum 7 0.566 15 B. Pronotum -0.367 63 L. Abd. Sternum 6 0.354 19 L. Fore Femur -0.240 17 Lateral L. Metanotum 0.251 50 L. Abd. Tergum 5 -0.209 71 L. Rt Side Sternum 9 0.156 35 L. Mid Tarsal Seg. 5 -0.154 52 L. Abd. Tergum 7 0.138 14 L. Metanotum -0.135 30 B. Mid Femur 0.135 72 L. Lt Side Sternum 9 -0.129 -335-

Table 68 Variables with Positive and Negative Weights absolutely greater than the Mean Value. a. Canonical Variate III.

Positive Negative -I Variable Weight Variable Weight

18 Lateral L. Mesonotum 0.381 15 B. Pronotum -0.473 20 L. Fore Femur 0.166 14 L. Metanotum -0.318 48 L. Abd. Tergum 3 0.154 51 L. Abd. Tergum 6 -0.255 64 L. Abd. Sternum 7 0.152 66 L. Abd. Tergum 9 -0.235 65 L. Abd. Sternum 8 0,148 49 L. Abd. Tergum 4 -0.222 2 B. Submentum 0.129 17 Lateral L. Metanotum -0.164 47 L. Abd. Tergum 2 0.125 29 L. Mid Tibia -0.138 24 L. Fore Tarsal Seg. 3 -0.120

b. Canonical Variate IV.

Positive Negative Variable 1 Weight Variable 1 Weight --1

14 L. Metanotum 0.382 38 L. Hind Tibia -0.285 17 Lateral L. Metanotum 0.273 39 B. Hind Femur -0.214 54 L. Abd. Tergum 9 0.226 18 Lateral L. Mesonotum -0.196 30 B. Mid Femur 0.223 65 L. Abd. Sternum 8 -0.189 55 L. Abd. Tergum 10 0.180 48 L. Abd. Tergum 2 -0.183 49 L. Abd. Tergum 3 0.176 59 L. Abd. Sternum 2 -0.149 37 L. Hind Femur 0.174 31 L. Mid Tarsal Seg. 1 -0.148 64 L. Abd. Sternum 7 0.155 23 L. Fore Tarsal Seg. 2 -0.140 36 L. Mid Tarsal Claw 0.130 52 L. Abd. Tergum 7 -0.133 62 L. Abd. Sternum 5 -0.131 -336-

Table 69 Variables with Positive and Negative Weights absolutely greater than the Mean Value.

Canonical Variate V.

Positive Negative

Variable Weight Variable Weight

72 L. Lt Side Sternum 9 0.462 71 L. Rt Side Sternum 9 -0.393 14 L. Metanotum 0.312 17 Lateral L. Metanotum -0.341 65 L. Abd. Sternum 8 0.235 19 L. Fore Femur -0.247 18 Lateral L. Mesonotum 0.191 64 L. Abd. Sternum 7 -0.212 49 L. Abd. Tergum 4 0.179 66 L. Abd. Sternum 9 -0.130 28 L. Mid Femur 0.154 63 L. Abd. Sternum 6 0.154

L = Length Rt = Right

B = Breadth Lt = Left

(Pages 236-243)it was among the leg segments that the most significant differences between the species occurred. Characters with extremely high positive or negative weights are those of particular interest for the separation of the sexes or species. Blackith, Davies & Moy (1963) define three types of biologically important variables:- (i) A single variable associated mainly with one axis, and contributing a negligible amount to the others. (Variables 17, 18, 71 and 72). (ii) Two or more variables associated to a similar extent with a single axis, i.e. they are symbatic (Blackith & Albrecht, 1959). (Variables 17 and 63 - Canonical Variate I, Variables 15, 63 and 64 - Canonical Variate II, etc.). (iii) A single variable associated with several axes to a similar degree. (Variables 14, 15 and 63). -337-

Table 70 indicates that all three types of variables are included in this growth analysis of Ectobius.

Table 70 Variables characterised by excessively large weights for Canonical Variates I - V.

Canonical Variate Variable I II III IV V

17 Lateral L. Metanotum 0.493 - - - - 63 L. Abd. Sternum 6 0.315 0.354 - - - 64 L. Abd. Sternum 7 - 0.566 - - - 15 B. Pronotum - 0.367 -0.473 - - 18 Lateral L. Mesonotum - - 0.381 - - 14 L. Metanotum - - -0.318 0.382 0.312 72 L. Lt Side Sternum 9 - - - - 0.462 71 L. Rt Side Sternum 9 - - - - -0.393

J Abbreviations as previously cited.

The discriminant scores for each of the first five canonical variates have been computed (Table 71), and the scores for any two of the first five canonical axes plotted. The canonical variates can then be described in terms of the growth patterns which they represent.

Canonical Variate I. (Figs. 57, 58)

This axis is predominantly concerned with general development although there is greater change at the last moult. In the males of both species and the females of E. lapponicus this pattern -338-

Table 71 Canonical Variates (Group Centroids) - Logarithmically Transformed Data.

Canonical Variate Species, Stage and Sex I II III IV V

E. lapponicus Male Adult 0.717 0.115 -0.178 -0.090 0.042 5th. Instar 0.214 -0.086 -0.342 -0.214 0.071 4th. Instar 0.055 -0.117 -0.298 -0.232 0.031 3rd. Instar -0.019 -0.110 -0.241 -0.213 0.014 2nd. Instar -0.105 -0.076 -0.182 -0.191 0.003 1st. Instar -0.210 -0.018 -0.137 -0.130 -0.014

E. lapponicus Female Adult 0.656 -0.266 -0.035 -0.169 -0.019 5th. Instar 0.196 -0.338 -0.263 -0.095 -0.088 4th. Instar 0.102 -0.290 -0.219 -0.139 -0.043 3rd. Instar 0.006 -0.202 -0.211 -0.148 -0.035 2nd. Instar -0.088 -0.108 -0.170 -0.163 -0.021 1st. Instar -0.210 -0.018 -0.124 -0.133 -0.013

E. panzeri Male Adult 0.609 0.165 -0.124 -0.288 -0.125 5th. Instar 0.150 -0.030 -0.262 -0.320 -0.027 4th. Instar 0.022 -0.054 -0.239 -0.316 -0.026 3rd. Instar -0.055 -0.068 -0.182 -0.288 -0.014 2nd. Instar -0.118 -0.046 -0.117 -0.257 -0.022 1st. Instar -0.196 -0.007 -0.063 -0.187 -0.025

E. panzeri Female Adult 0.269 -0.373 0.016 -0.316 0.086 5th. Instar 0.098 -0.321 -0.190 -0.205 -0.071 4th. Instar 0.073 -0.250 -0.212 -0.337 -0.056 3rd. Instar -0.023 -0.191 -0.132 -0.224 -0.055 2nd. Instar -0.099 -0.094 -0.094 -0.224 -0.027 1st. Instar -0.195 -0.015 -0.056 -0.177 -0.015 -339- of growth is strongly displayed and shows a striking discontinuity at the transition to the adult stage. Even in the females of E. panzeri there is enhanced growth at this stage. E. lapponicus displays this mode of growth more strongly than E. panzeri. This vector ranks all the instars of each sex and species, which supports the idea of a general growth component. The first four nymphal instars are dispersed more or less equally along this axis; the fifth instar begins to show the transition to the adult stage, by an increased displacement along the axis. In this respect this canonical variate can be referred to as a "general stage growth factor" since it ranks all the developmental stages. The variables which contribute most to this pattern of growth (Table 67E) may explain to some extent the differences in the displacement along this axis of the two sexes and species. The lateral length of the meso- and metanota are characterised by very high weights. The males of the two species are fully winged and display this mode of growth to a similar degree. However, the females of E. lapponicus have reduced fore wings and those of E. panzeri are virtually apterous, thus accounting for the less evident discontinuity from the last instar to the adult, which is almost absent in the latter species.

Canonical Variate II. (Figs. 57a, 59, 60a)

This represents a pattern of growth reflecting sexual tendencies throughout development, but is primarily one in which the sexes become progressively separated along the axis. This mode of growth may be said to be female specific in that the females are displaced progressively along this axis during development, to a more or less equal extent at each instar. The males, however, exhibit a complete though relatively slight reversal in the direction of displacement along this axis after the third or fourth instars in E. panzeri and E. lapponicus respectively. The transition from the fifth instar to the adult is not specially -340- distinguished in the females of the two species; but is quite marked in the males. The adult females of E. lapponicus show a minor discontinuity in that the direction of displacement along this axis is reversed between the fifth instar and adult.

Canonical Variate III. (Figs. 57b, 59a, 60b)

Initially this mode of growth is species specific; in the first two instars the two species are quite separate in their displacement along this axis (a difference which is maintained in the following instars). However, later in development both sexual and specific differences come to play a part. These sexual differences become more marked as development proceeds in both species. The difference between the males of the two species remains constant throughout development; in the females, however, the two species progressively separate during development (with the exception of a slight irregularity in the fourth instar of E. panzeri). This mode of growth shows a well defined reversal at the transition to the adult in each sex and species.

Canonical Variate IV. (Figs. 58a, 59b)

This mode of growth mainly reflects a species difference, which increases slightly during development, the adult stage being the most distinct. Like the fifth variate, however, it represents a component of only minor importance. The separation of the females of the two species remains constant for the first three instars. The irregular displacement of the fourth instar of the female of E. panzeri along this axis has not emerged in any of the previous analyses, which precludes any chance of error in the original data, which has incidentally been thoroughly checked. Sexual differences are also displayed by this canonical variate, these are negligible in the first instar but become more important as development proceeds. -34-1-

-.4

a. Canonical Variate T against • E. lapponicus Male

o E lapponicus Female ■ L.panzeri Male o E. panzeri Female

A

.4

A Adult

b. Canonical Variate T against III

FIG 57 Growth Patterns revealed by Multiple

Discriminant Scores (Log. Transformed Data).

-342-

2- 1 lv

-•4 - 2 0 2 •4 •8

A -.4- A Adult

- •6 - a. Canonical Variate T against IV • E. lapponicus Male

O E. lapponicus Female

2- • E. panzeri Male o E. panzeri Female

V • A

-•4 •8

7 A

-.2-

-.4- b. Canonical Variate T against

FIG 58 Growth Patterns revealed by Multiple

Discriminant Scores (Log. Transformed Data). -343-

• E. lapponicus Male

• E lapponicus Female • E panzeri Male • E panzeri Female a. Canonical Variate II against Ill A Adult

b. Canonical Variate against IV

FIG 59 Growth Patterns revealed by Multiple

Discriminant Scores (Log. Transformed Data). -344-

--4

A Adult

a. Canonical Variate against

• E. lapponicus Male

E. lapponicus Female

■ E panzeri Male

o E panzeri Female

- • 4 •2

b. Canonical Variate against y

FIG 60 Growth Patterns revealed by Multiple

Discriminant Scores (Log. Transformed Data). _345-

Canonical Variate V. (Figs. 58b, 60)

Unlike the previous canonical variates, this is not simply a species or sexual specific pattern, though both these modes of growth are involved to some extent. This growth pattern appears to reflect a species difference which is mainly shown by the males. The difference is present to only a small degree in the first instar but develops gradually throughout the nymphal instars until the adult stage, where the greatest displacement occurs; this is particularly evident in E. panzeri. This pattern of growth is most clearly shown in Fig. 60b (Canonical Variate III against V), the males of the two species being displaced in opposite directions along this axis. The males of E. panzeri are displaced in the same direction as the females of the two species, which remain very similar throughout their nymphal development, but differ slightly in the adult stage. Wilks' lambda test of the discrimination achieved showed that the differences between the group centroids are very highly significant (p less than 0.001), using both the untransformed and the logarithmically transformed data, and assuming that the within-group covariance matrices are equal. -31+6—

GENERAL DISCUSSION.

This may be organised around eight main points:-

(i) The Relationship between Dyar's Law and the Law of Simple Allometry.

There is a distinct relationship between Dyar's Law and the simple allometry of growth, which is worth emphasising. Dyer's Law (1890) assumes a constant rate of size increase of a structure at each ecdysis, whereas growth by simple allometry (Huxley, 1932) requires that the ratio of the growth rates of the two dimensions, x and y, remain constant. Thus it can be seen that Dyar's Law is really a special case of allometry, when both x and y themselves are growing at a constant rate at each moult. Its requirements are such that its occurrence in insect growth is therefore infrequent.

(ii) The Expression of Allometry of Growth in an Organism.

The growth gradients demonstrated by Huxley (1932) are a useful way to describe the distribution of growth intensity in the body. However, few characters were found to grow by simple allometry and the use of growth gradients therefore involved disregarding deviations from simple allometry, which is obviously undesirable. Significant deviations from linearity indicate changes in the value of the equilibrium constant during development and a more complete understanding of the continuous variation in the value of at in the structures of the body and in time is made possible by the use of growth contours. These contours afforded the most accurate means of expressing the bivariate analysis of the allometry of growth in Ectobius, and the extension of this method to embrace the width of consecutive segments of the body and the length of the appendages in Ectobius species may well prove -347- interesting. The introduction of a time parameter into the simple allometry relation has been considered by Laird (1965), who advocates the use of Gompertz growth equations. An equation of this nature includes the specific growth rate and its proportional rate of decay as variables and is therefore directly applicable to the allometry equation due to Huxley (Laird, Barton & Tyler, 1968). Barton & Laird (1969) have shown that the allometry equation of Huxley (1924) is in fact insensitive to time and that more accurate conclusions can be drawn from the analysis of these Gompertz growth curves for the individual parts of the body. Growth gradients can then be reconsidered as representing the distribution of growth intensity in time. The multivariate generalisation of the allometry parameter by the use of the first principal component of a log. covariance matrix is obviously advantageous, since it avoids the selection of an arbitrary reference character. A consideration of the nature of the remaining components was included by Jolicoeur (1963b) and their importance stressed by Sprent (1968). The relation between the weights of elements for these components and the occurrence of deviations from simple allometry proved interesting, and suggests a subject for rewarding further work. The availability of electronic digital computers also makes the use of the generalised allometry parameter feasible for less detailed investigations.

(iii) Eigenanalysis of Correlation Matrices as a Method of Analysing Intra-Stadial Variation.

The analysis of size variation within a single stadium by the eigenanalysis of correlation matrices did not reflect in any way the overall growth process in Ectobius. This was, perhaps, a little surprising since relatively few distinct growth patterns emerged from other analyses of the post--embryonic growth of Ectobius by different multivariate techniques. However, this -348-

method, which has to date been used only to express.intra-stadial variation (e.g. Blair, Blackith & BoratyriSki, 1964), warrants further investigation. It is suggested that more meaningful results may be obtained by considerably increasing the number of replicates for each stage, a requirement which would probably necessitate a reduction in the number of variables considered. However, pooling the data for the two sexes of a stage in E. lepponicus and E. panzeri proved relatively ineffective.

(iv) Factor Analysis and the Need for Further Work.

The application of a factor analysis to the growth of Ectobius species requires further study, preferably using a maximum likelihood technique. In this thesis a principal component analysis formed the first solution; however, subsequent rotation did not in any way simplify the biological interpretation of the vectors, presumably because the complexity of a situation involving so many variables was too great to be clarified by simple rotation. The matrices used in this analysis were singular, and thus the computation of the factor scores was precluded; to date little is known of matrices which are not of full rank.

(v) The relative Simplicity of the Growth Process in Ectobius.

One of the most striking features of the multivariate techniques in both species of Ectobius is the relative simplicity of the general growth process. Only a few components or factors account for most of the total variation, which indicates that an equally small number of separate growth patterns exist to describe the total change in size and shape throughout development. Thus 97 or 98% of the total variance is accounted for by the first five latent roots when a principal component analysis is applied, using a correlation or covariance matrix based on the logarithms of the original data, respectively. A multiple discriminant analysis -349-

yields latent roots, the first five of which account for 93 or 96% of the variance depending on whether the data was logarithmically transformed or untransformed. The latent roots derived from the analysis of intra-stadial variation, by the eigenanalysis of separate correlation matrices for each stage, account for slightly lower percentages of the total variance. This drastic simplification in the potential number of growth patterns has similarly been found by Blackith, Davies & Moy (1963) and even more strikingly by Matsuda & Rohlf (1961) who found that 99.6% of the total variance was attributed to the first growth pattern in T. trepidus. Less marked reductions have been found in the analyses of variation between individuals of a particular stage in the life history (usually the adult); three or four latent roots were required to absorb the total variance in the Painted Turtle.(Jolicoeur & Mosimann, 1960) and Martes americana (Jolicoeur, 1963b) respectively. Blackith & Roberts (1958) suggest that this simplicity in the overall growth pattern of an insect may be due to the relatively simple hormonal mechanisms controlling insect growth.

(vi) The Consistency of the Growth Patterns revealed by Different Techniques.

Since each of the main multivariate analytical methods is based on a different mathematical model, one would expect some differences between the results obtained using the separate techniques. However, it is in some ways a rewarding feature of this study that similar vectors can be identified from each type of analysis. These growth patterns are most easily described by the computation of the appropriate discriminant or component scores. The characters contributing most to these particular growth patterns vary only slightly in the different techniques. The nature of the growth patterns suggested by the various methods of analysis for the two species of Ectobius are therefore confirmed by their consistent occurrence in two essentially distinct types -350-- of analysis. The first vector is always a general size factor, but the other vectors describe growth patterns principally reflecting either adult maturation and sexual or specific differences.

(vii) The Selection of Characters to Use in an Analysis of Growth.

In some respects these multivariate analyses have an arbitrary nature, since the choice of characters is of necessity merely a random selection of structures which are convenient to measure. Nothing is known a priori of the response of such characters under analysis, and it is conceivable that the admission of more characters may introduce differences into the identity of the vectors. However, such a wide selection of characters, far in excess of the numbers used in any previous analyses, reduces this risk to the minimum. It should perhaps be stressed that in many cases the less prominent characters are those which carry the greatest weights in the determinantal equations, indicating that such characters should not be neglected in a growth study of this nature.

(viii) The Relationship between the Biology of the Growth of Two Species of Ectobius.

E. lapponicus and E. panzeri were selected since they represent species in the genus Ectobius which have different life cycles. E. panzeri is a univoltine species, the oothecae hatch and the nymphs become adult over a period of a few months; in comparison E. lapponicus has a two-yearly life cycle, passing one winter as an immature instar. It was thought possible that these two species may exhibit differences in their growth processes which might be related in some way to their different mode of life. However, these species behaved similarly under analysis, and revealed few consistent differences. One interesting feature which has emerged from the bivariate -351-

analysis of allometry and later in the multivariate techniques is that the leg segments feature prominently among the specific differences. Values for the equilibrium constant, a , in the simple allometry equation showed the greatest specific differences with respect to the leg segments. The growth of the leg segments of E. panzeri having far higher values of a than E. lapponicus. In the principal component analyses, the vectors identified by growth patterns reflecting differences between the species are characterised by elements representing the leg segments, whilst in the multiple discriminant analyses, where differences are maximised, the leg segments again feature prominently in the identity of each vector. It may be suggested that this difference in the growth of the appendages is connected in some way with the difference in the nature of the habitat selected by the two species; E. panzeri is always restricted to sandy soils where relatively longer appendages would be a distinct advantage. The inclusion in these analyses of a third closely related species, E. pallidus, would be interesting as would the extension of this work to different genera in the Dictyoptera to determine at what taxonomic level major differences in post-embryonic growth patterns occur. -352-

SUMNARy OF T3E GROWTH SECTION including an Evaluation of the Methods used in the Analysis of the Growth of Two Species of Ectobius.

1. The nymphal instars of the three British species of Ectobius can be distinguished by the use of a key based on characters of the thoracic nota and abdominal cerci.

2. Detailed examination of the posterior abdominal sternites reveals differences between the sexes in all nymphal instars. The development of the genital segments in both sexes is included.

3. The post-embryonic growth of E. lap onicus and E. panzeri is investigated by several separate analytical techniques and involves the consideration of 74 characters of both sexes of all the developmental stages.

4. Dyar's Law is applicable to only a limited number of characters, which are common to both species. It is therefore of little value in a developmental study.

5. An investigation of Richards' extension of Dyar's Law proved that there is little evidence in favour of this hypothesis in the growth of Ectobius.

6. Bodenheimer's modification of Przibram's Rule, when applied to a selection of characters, might be held to indicate the occurrence of latent cell divisions during development. However, without cytological evidence of cell multiplication this method of explanation for the observed deviations from a regular geometric progression is not to be recommended.

7. The choice of a suitable reference dimension in the study of allometry is vital. The generalisation of the allometry equation, -353-

by the use of the first principal component of the covariance matrix of logarithmically transformed data, is obviously preferable, although the results obtained from a measurement of the total body length, constructed by the summation of the lengths of individual sclerites, differs only slightly from it and could be used in a less comprehensive study.

8. Two of the available methods for fitting the allometry equation are compared. Since both variables are subject to error, Bartlett's "best fit" method is more satisfactory on statistical grounds, and enables rather more accurate estimates of the allometric constants. A complete set of significance tests are also available for this line.

9. Significance tests have seldom been applied in studies of allometry in the past. Estimates were obtained of the significance of the slope of the line, the significance of deviations from a = 1 (i.e. growth is isometric) and the significance of deviations from linearity. Few cases of isometric growth exist in the growth of Ectobius, the majority of characters showing either positive or negative allometry. In only a few cases, however, is this a matter of simple allometry. A large number of characters deviate significantly from the simple allometry relation, and might therefore be held theoretically to involve an additional higher- order mode of growth.

10. The study of growth gradients in the dorsal and ventral mid- line of the body, the legs and antennae proved a useful way to describe and compare the distribution of growth intensity in the two sexes and species. However, deviations from linearity in the allometry equation were disregarded in preparing the growth constants.

11. The deviations from linearity in many characters suggest that the value of a varies significantly during development. The -354-

construction of growth contours illustrate this and have been used to summarise the growth of sclerites in the mid-dorsal line. From these growth contours centres of positive and negative allometry are revealed. This method affords the most accurate picture of allometry of growth in an insect.

12. Eigenanalysis of the separate correlation matrices for each species, sex and stage is merely an analysis of the size variation within a particular stage and did not reveal much of the developmental basis of growth. Very few growth patterns common to a range of stadia could be identified. This method of multi- variate analysis therefore proved to be of little value as a means of studying the growth of a species as a developmental process.

13. A principal component analysis simultaneously embracing all species, sexes and stages was applied to a correlation matrix and a covariance matrix based on the logarithms of the original data. These methods represent a true analysis of the developmental growth processes of a species. The nature of the growth patterns is similar when either matrix is used although the interpretation is far easier and more rational when the logarithmic covariance matrix is used. Further principal component analyses applied to all stages of each sex and species in turn revealed similar growth patterns in each case.

14. It was not possible to conduct a full factor analysis on this set of data. A principal component analysis formed the first solution and by rotation of the eigenvectors it was, hoped that a clearer biological interpretation would be possible, but this was not the case. The computation of the factor scores was not feasible, since the log. covariance matrix is singular.

15. The application of a multiple discriminant analysis to all the species, sexes and stages provides an efficient method for -355-

discriminating between the stages for each sex and species, and indicates the characters principally involved in this discrimination. The use of logarithmically transformed data results in a far better discrimination than follows from the use of untransformed data especially in the early stages. For a multiple discriminant analysis to be used the within-group covariance matrices should be homogeneous; a test for this was attempted but could not be applied, since the within-group covariance matrices are singular. -356-

ACKNOWLEDGMENTS.

This work has been carried out at Imperial College Field Station, Silwood Park and I would like to thank Professor 0.W. Richards and Professor T.R.E. Southwood for facilities at the field station. My sincere thanks are due to my supervisor, Mr. R.G. Davies, for his constant help and encouragement throughout the course of this work, and particularly for his help in analysing the growth data. My thanks are also due to the following:- Professor S. Gill, Director of the Centre for Computing and Automation at Imperial College, for the use of the Centre. Dr. D.R. Ragge for his help and interest in the project. Mr. A.J. Wise for suggesting several good collecting areas. Dr. M.J. Madelin for the identification of several species of fungi. Major A.W. Haig for permission to collect on Crown Estate Property. Mr. D.A. Mithen for permission to collect on Forestry Commission Land in the New Forest. Mrs. van Emden for translating German literature. My mother, Mrs. K.N. Brown, for typing the manuscript. Mr. C. Wall for assistance with the collection of specimens, photogra)hy and the duplication of the thesis. Finally, I wish to thank the Science Research Council for the award of a S.R.C. Studentship. -357-

REFERENCES.

ALBRECHT, F.O. 1955. La densit4 des populations et la croissance chez Schistocerca gregaria (Forsk.) et Nomadacris septemfasciata (Serv.); la mue d'adjustment. J. Agric. trop. Bot. appl. 2: 110-192. ANDERSON, T.W. 1958. An Introduction to Multivariate Statistical Analysis. Wiley, New York: 374 pp. ANDREWARTHA, H.G. & BIRCH, L.C. 1954. The Distribution and Abundance of . University Press, Chicago: 782 pp.

BAILEY, D.W. 1956. A comparison of genetic and environmental principal components of morphogenesis in mice. Growth 20: 63-74. BAILEY, N.T.J. 1959. Statistical Methods in Biology. English Universities Press, London: 200 pp. BARTLETT, M.S. 1949. Fitting a straight line when both variables are subject to error. Biometrics 5: 207-212. BARTON, A.D. & LAIRD, A.K. 1969. Analysis of allometric and non-allometric differential growth. Growth 3: 1-16.

BEDNARZ, S. 1955. La taille de la tete des larves Tetti onia viridissima L. (Saltatoria: Tettigoniidae et IThypothese de Dyar. (In Polish with French summary). Polskie Pismo ent. 23: 191-203.

BIGELOW, R.S. 1960. Developmental rates and diapause in Acheta pennsylvanicus (Burmeister) and Acheta veletis Alexander & Bigelow (Orthoptera: Gryllida6T: Can. J. Zool. 38: 973-988. BIRCH, L.C. 1945. Diapause in Scelio chortoicetes Frogg. (Scelionidae), a parasite of the eggs of Austroicetes cruciata Sauss. J. Aust. Inst. agric. Sci. 11: 189-190. BLACKITH, R.E. 1957. Polymorphism in some Australian locusts and grasshoppers. Biometrics 13: 183-196. -358-

An analysis of polymorphism in social wasps. Insectes soc. 2: 263-272. BLACKITH, R.E. 1960. A synthesis of multivariate techniques to distinguish patterns of growth in grasshoppers. Biometrics 16: 28-40. BLACKITH, R.E. 1962. L' identite des manifestations phasaires chez les Acridiens migrateurs. Colloques int. Cent. natn. Rech. scient. 114: 299-310. BLACKITH; R.E. & ALBRECHT, P.O. 1959. Morphometric differences between the eye stripe polymorphs of the red locust. Scient. Jl R. Coll. Sci. 27: 13-27. BLACKITH, R.E., DAVIES, R.G. & MOY, E.A. 1963. A biometric analysis of development in apdercus fasciatus Sign. (Hemiptera: Pyrrhocoridae). Growth 27: 317-334. BLACKITH, R.E. & ROBERTS, M.I. 1958. Farbenpolymorphismus bei einigen Feldheuschrecken. Z. VererbLehre 89: 328-337. BLAIR, C.A., BLACKITH, R.E. & BORATYNSKI, K. 1964. Variation in Coccus hesperidum L. (Homoptera: Coccidae). Proc. R. ent. Soc. Lond. (A) 22: 129-134. BLAIR, K.G. 1934. A note on the British species of Ectobius Steph. Entomologist's mon. Mag. 70: 157-159. BLISS, C.I. & BEARD, R.L. 1954. The growth of the head capsule in individual milkweed bugs. Ann. ent. Soc. Am. 47: 388-392. BODENHEIMER, F.S. 1927. Ueber Regelmassigkeiten in dem Wachstum von Insekten. I. Das Langenwachstum. Dt. ent. Z. 1927: 33-57. BODENHEIMER, F.S. 1933. The progression factor in insect growth. Q. Jl microsc. Sci. 8: 92-95. BORATYNSKI, K.L. 1952. Matsucoccus pini (Green, 1925) (Homoptera, Coccoidea: Margarodidae): Bionomics and external anatomy with reference to the variability of some taxonomic characters. Trans. R. ent. Soc. Lond. 103: 285-326. BOX, G.E.P. 1949. A general distribution theory for a class of likelihood criteria. Biometrika 317-346. -359-

BROWN, E.B. 1952. Observations on the life history of the cockroach Ectobius panzeri Stephens (Orth., Blattidae). Entomologist's mon.jig7148: 209-212. BROWNING, T.O. 1952. The influence of temperature on the completion of diapause in the eggs of Gryllulus commodus Walker. Aust. J. scient. Res. (B) 112-127. BUXTON, P.A. 1938. Studies on the growth of Pediculus (Anoplura). Parasitology 12: 65-84. CALVERT, P.P. 1929. Different rates of growth among animals with special reference to the Odonata. Proc. Ara.phil.t. Soc. 63: 227-274. CAMERON, E. 1957. On the parasites and predators of the cockroach. II. Evania appendigaster (L.). Bull. ent. Res. 8: 199-209. CAMPBELL, F.L. 1929. The detection and estimation of insect chitin; and the relation of "chitinization" to hardness and pigmentation of the cuticula of the American cockroach, Periplaneta americana L. Ann. ent. Soc. Am. 22: 401-426. CLARK, L.B. & HERSH, A.H. 1939. A study of relative growth in Notonecta undulata. Growth 3: 347-372. CLARKE, K.U. 1957. On the increase in linear size during growth in Locusta mi:cratoria L. Proc. R. ent. Soc. Lond. (A) E: 35-39. COOLEY, W.W. & LOHNES, P.R. 1962. Multivariate Procedures in the Behavioural Sciences. Wiley, New York: 211 pp. CORBET, P.S. 1955. A critical response to changing length of day in an insect. Nature, Lond. 175: 338-339. CORBET, P.S. 1956. Environmental factors influencing the induction and termination of diapause in the Emperor dragonfly, Anax imperator Leach (Odonata: Aeshnidae). J. exp. Biol. 1-14. GROS, A. 1942. Blatta orientalis et ses parasites. i. Evania mactata Brulle; ii. Eulophus sp. Etude biolugique. Eos, Madr. 18: 45-67. -360-

CROSSKEY, R.W. 1951. The morphology, taxonomy and biology of the British Evanioidea (Hymenoptera). Trans. R. ent. Soc. Lond. 102: 247-301. DUARTE, A.J. 1938. Problems of growth of the African migratory locust. Bull. ent. Res. 29: 425-456. DYAR, H.G. 1890. The number of molts of Lepidopterous larvae. Psyche, Camb. 5: 420-422. EDMUNDS, L.R. 1952a. The oviposition of Prosevania punctata (Brune). A Hymenopterous parasite of cockroach egg capsules. Ohio J. Sci. 52: 29-30. EDMUNDS, L.R. 1952b. Some notes on the habits and parasites of native wood-roaches in Ohio (Orthoptera: Blattidae). Ent. News 63: 141-145. EDMUNDS, L.R. 1954. A study of the biology and life history of Prosevania punctata (Brune) with notes on additional species (Hymenoptera:- Evaniidae). Ann. ent. Soc. Am. 47: 575-592. EDWARDS, C.A. 1964. The bionomics of swift moths I. The ghost swift moth, Hepialus humuli (L.). Bull. ent. Res. 55: 147-160. FELDSTEIN, M.J. & HERSH, A.H. 1935. The determination of genetic constants of relative growth. Am. Nat. 69: 344-353. FISHER, R.A. 1938. The statistical utilization of multiple measurements. Ann. Eugen. 8: 376-386. FLINT, 0.S. 1951. A new cockroach record from the United States. Bull. Brooklyn ent. Soc. 46: 53. FORBES, W.T.M. 1934. A note on Dyar's Law (Lepidoptera: larvae). Bull. Brooklyn ent. Soc. 29: 146-149. FRAISSE, R. & ARNOUX, J. 1954. Les caracteres biometriques du cocon chez Bombyx mori L. et leurs variation sur 1' influence de 1' alimentation. Revue Ver Sole 6: 43-62. FRIEND, R.B. 1933. The birch leaf-mining sawfly, Fenusa pumila Klug. Bull. Conn. agr. Exp. Stn 348: 293-364. -361-

GAINES, J.C. & CAMPBELL, F.L. 1935. Dyar's rule as related to the number of instars of the corn ear worm, Heliothis obsoleta (Fab.), collected in the field. Ann. ent. Soc. Am. 28: 445-461. GARDINER, L.M. 1954. Differential growth as evidence of the relationship of Monochamus notatus (Drury) and M. scutellatus (SayY (Coleoptera: Cerambycidae). Can. Ent.-77-465-470.

GAUNITZ, C.B. 1936. Ectobius lapponicus L. als Vorratsschadling in Lappland, eine alte sicher unrichtige Vermutung in neuer Beleuchtung. Konowia 15: 162-166. GENIEYS, P. 1924. Contribution a 1' etude des Evaniidae, Zeuxevania splendidula Costa. Bull. biol. Fr. Bela,. 58: 482-494. GHENT, A.W. 1956. Linear increment in width of the head capsule of two species of sawflies. Can. Ent. 88: 17-23. GIER, H.T. 1947. Growth rate in the cockroach Periplaneta americana (Linn.). Ann. ent. Soc. Am. 40: 303-317. GORDON, H.T. 1959. Minimal nutritional requirements of the German roach, Blatella germanica (L.). Ann. N.Y. Acad. Sci. 77: 290-351. GOULD, S.J. 1966. Allometry and size in ontogeny and phylogeny. Biol. Rev. 41: 587-640.

GOULD, S.J. 1967. Evolutionary patterns in the Pelycosaurian reptiles: a factor-analytic study. Evolution, Lancaster, Pa. 21: 385-401. GRIFFITHS, jr., J.T. & TAUBER, O.E. 1942a. Fecundity, longevity and parthenogenesis of the American roach, Periplaneta americana L. Physiol. Zobl. 15: 196-209. GRIFFITHS, J.T. & TAUBER, 0.E. 1942b. The nymphal development of the roach, Periplaneta americana L. Jl N.Y. ent. Soc. 50: 263-272. GUPTA, P.D. 1948. On the structure, development and homology of the female reproductive organs in Orthopteroid insects. Indian J. Ent. 10: 75-123. -362-

GURNEY, A.B. 1953. Distribution, general bionomics and recognition characters of two cockroaches recently established in the United States. Proc. U.S. natn. Mus. 103: 39-56.

HABER, V.R. 1920. Oviposition by an Evaniid, Evania appendiRaster Linn. Can. Ent. 52: 248.

HAINES, F.H. 1936. How does Ectobius (Orth.) pass the winter? J. Soc. Br. Ent. 1: 146.

HALLIBURTON, W.H. & ALEXANDER, G. 1964. Effect of photoperiod on molting of Chortophaga viridifasciata (De Geer) (Orthoptera: Aorididae). Ent. News 75: 133-137. HARMAN, H.H. 1960. Modern Factor Analysis. University Press, Chicago: 471 pp. HARRIES, F.H. & HENDERSON, C.F. 1938. Growth of insects with reference to progression factors for successive growth stages. Ann. ent. Soc. Am. 31: 557-572. HARZ, K. 1960. Ein Beitrag zur Biologie der Schaben. Abh. naturw. Ver. Wilrzbur Heft 3: 5-32. HERSH, A.H. 1931. Facet number and genetic growth constants in bar-eyed stocks of Drosophila. J. exp. Zool. 60: 213-248. HOGAN, T.W. 1960. Onset and duration of diapause in Acheta commodus (Walk.) (Orthoptera). Aust. J. biol. Sci. 13: 14-29. HOLMQUIST, A.M. 1926. Studies in hibernation. Ann. ent. Soc. Am. 12: 395-426. HOPKINS, J.W. 1966. Some considerations in multivariate allometry. Biometrics 22: 747-760. HUXLEY, J.S. 1924. Constant differential growth ratios and their significance. Nature Lond. 114: 895-896.

HUXLEY, J.S. 1932. Problems of Relative Growth. Methuen, London: 27 pp.

HUXLEY, J.S. & TEISSIER, G. 1936. Terminology of relative growth. Nature Lond. 137: 780-781. -363-

JOHNSON, C.G. 1939. Taxonomic characters, variability and relative growth in Cimex lectularius L. and C. columbarius Jenyns (Heteropt. Cimicidae). Trans. R. ent. Soc. Lond. 89: 543-568. JOLICOEUR, P. 1963a. The multivariate generalisation of the allometry equation. Biometrics 19: 497-499.

JOLICOEUR, P. 1963b. The degree of generality of robustness in Martes americana. Growth 27: 1-27.

JOLICOEUR, P. & MOSIMANN, J.E. 1960. Size and shape variation in the painted turtle. A principal component analysis. Growth 24: 339-354. KAISER, H.F. 1958. The varimax criterion for analytical rotation in factor analysis. Psychometrika 23: 187-200. KELER, S. 1934. Ueber die Kopf-indices der Larven and die Dyar'sche Hypothese. (In Polish with German summary). Polskie Pismo ent. 12 (1933): 173-180. KENDALL, M.G. 1957. A course in Multivariate Analysis. Griffin, London: 1g-5 pp.

KERMACK, K.A. 1954. A biometrical study of Micraster coranguinum and M. (Isomicraster) senonensis. Phil. Trans. R. Soc. B 237: 375-428. KERMACK, K.A. & HALDANE, J.B.S. 1950. Organic correlation and Biometrika 22: 30-41.

KEVAN, D.K.McE. 1961. A revised summary of the known distribution of British Orthopteroids. Trans. Soc. Br. Ent. 14: 187-205. KEY, K.H.L. 1936. Observations on rate of growth, coloration and the abnormal six-instar life-cycle in Locusta migratoria mi&ratorioides R.& F. Bull. ent. Res. 27: 77-85. KILLINGTON, F.J. 1927. List of the Orthoptera of Hampshire and the Isle of Wight. Entomologist's Rec. J. Var. 21: 1-10. -364-

KRAUS, B.S. & CHOI, S.C. 1958. A factorial analysis af the prenatal growth of the human skeleton. Growth 22: 231-242.

KUMAR, R. 1966. Studies on the biology, immature stages, and relative growth of some Australian bugs of the super- family Coreoidea (Hemiptera: Heteroptera). Aust. J. tool. 14: 895-991. LAIRD, A.K. 1965. Dynamics of relative growth. Growth 249-263. LAIRD, A.K., BARTON, A.D. & TYLER, S.A. 1968. Growth and time: an interpretation of allometry. Growth 32: 347-354.

LAWLEY, D.N. & MAXWELL, A.E. 1963. Factor Analysis as a Statistical Method. Butterworths, London: 117 pp.

LAWSON, F.A. 1951. Structural features of the oothecae of certain species of cockroaches (Orthoptera: Blattidae). Ann. ent. Soc. Am. 44: 269-285. LAWSON, F.A. 1952. Structural features of cockroach egg capsules. II.The ootheca of Cariblatta lutea lutea (Orthoptera: Blattidae). Ohio J. Sci. 296-300. LAWSON, F.A. 1953. Structural features of cockroach egg capsules. III.The ootheca of Eurjcotis floridana (Orthoptera: Blattidae). J. Tenn. Acad. Sci. 28: 28-33. LAWSON, F.A. 1954. Structural features of cockroach egg capsules. IV.The ootheca of Parcoblatta uhleriana (Orthoptera: Blattidae). J. Kans. ent. Soc. 27: 14-20. LAWSON, F.A. & LAWSON, E.Q. 1965. Sexing first instar cockroaches (Orthoptera: Blattidae). J. Kans. ent. Soc. 28; 408-41o. LEES, A.D. 1955. The Physiology of Diapause in . University Press, Cambridge: 151 pp.

LUCAS, W.J. 1920. A Monograph of the British Orthoptera. Ray Society, London: 246 pp. -365-

LUCAS, W.J. 1928. Notes on British Orthoptera including Dermaptera in 1927. Entomologist 61: 78-81.

LUDWIG, D. 1932. The effect of temperature on the growth curves of the Japanese beetle (pillia japonica Newman). Physiol. Zo61. 5: 431-447. LUDWIG, D. 1934. The progression factor in the growth of the Japanese beetle (popillia 1.aponica Newman) (Coleoptera: Scarahaeidae). Ent. News 45: 141-153.

LUDWIG, D. & ABERCROMBIE, W.F. 1940. The growth of the head capsule of the Japanese beetle larva. Ann. ent. Soc. Am. 33: 385-390. MAHALANOBIS, P.C. 1936. On the generalised distance in statistics. Proc. natn. Inst. Sci. India 2: 49-55. MANLEY, T.R. 1969. Sexual characters of the instars of Blaberus discoidalis and an analysis of growth in this cockroach (Orthoptera: ). Ann. ent. Soc. Am. 62: 734-737. MARKS, E.P. & LAWSON, P.A. 1962. A comparative study of the Dictyopteran ovipositor. J. Morph. 111: 139-171. MASAKI, S. & OYAMA, N. 1963. Photoperiodic control of growth and wing-form in Nemobiusezoensis Shiraki (Orthoptera: Gryllidae). Kontyli 31: 16-26.

MATS UDA, R. 1961a,b,c. Studies of relative growth in Gerridae (Hemiptera: Heteroptera). I - III. Ann. ent. Soc. Am. 54: 578-598.

MATSUDA, R. 1961d. Studies of relative growth in Gerridae (IV) (Hemiptera: Heteroptera). J. Kans. ent. Soc. 34: 5-17.

MATSUDik, R. 1962a. Studies of relative growth in Gerridae. VI. Comparison of two species of Trepobates (Hemiptera: Insecta). Kans. Univ. Sci. Bull. 43: 113-129. MATSUDA, R. 1962b. A study of relative growth of leg and antennal segments in some species of Heteroptera. Kontyii 30: 152-159. -366-

MATSUDA, R. 1962c. A study of relative growth in two strains of Pycnoscelus surinamensis (Linnaeus) (Panchloridae: Blattaria). Growth 26: 129-135. MATSUDA, R. 1963a. Evolution of relative growth in the Arthropoda. Z. wiss. Zool. 169: 64-81. MATSUDA, R. 1963b. A study of relative growth of leg and antennal segments in two species of Orthotylus (Heteroptera: Miridae). Proc. R. ent. Soc. Lond. (A) 38: 86-89. MATSUDA, R. & ROHLF, F.J. 1961. Studies of relative growth in Gerridae (V). Comparison of two populations (Heteroptera: Insecta). Growth 25: 211-217. McKITTRICK, P.A. 1964. Evolutionary studies of cockroaches. Mem. Cornell Univ. agric. Exp. Stn 389: 1-197. METCALFE, M.E. 1932. On a suggested method for determining the number of larval instars in Sitodrepa panicea L. Ann. appl. Biol. 19: 413-419. MILES, H.W. 1931. Growth in the larvae of Tenthredinidae. J. exp. Biol. 'q 355-364. MORRISON, D.F. 1967. Multivariate Statistical Methods. McGraw-Hill, New York: 338 pp. NAGASAWA, S. 1965. Number of larval moults and the growth of the head capsule in successive instars in the 'Shimuza' race of the small gypsy moth, Lymantria dispar L. Problems on the breeding of insects for biological assay of insecticides. Kontyil 33: 466-474. NEEDHAM, A.E. 1937. On relative growth in Asellus aquaticus. Proc. zool. Soc. Lond. (A) 107: 289-313. NEEDHAM, A.E. 1964. The Growth Process in Animals. Pitman, London: 522 pp.

PETERS, W. 1961. Methoden zur Herstellung von Aufhellungspraparaten. Zool. Anz. 167: 233-240. PETERSEN, B. 1952. Studies on geographic variation of allometry in some European Lepidoptera. Zool. Bidr.jps. 29: 1-38. -367-

PETERSON, A. 1953. A Manual of Entomological Techniques. Edwards, AnnArbor, Michigan: 3 7 pp. PICKFORD, R. 1953. A two-year life cycle in grasshoppers (Orthoptera: Acrididae) overwintering as eggs and nymphs. Can. Ent. 85: 9-14. PRYOR, M.G.M., RUSSELL, P.B. & TODD, A.R. 1946. Protocatechuic acid, the substance responsible for the hardening of the cockroach ootheca. Biochem. J. yO: 627-628.

PRZIBRAM, H. & MEGUSAR, F. 1912. Wachstumsmessungen an Sphodromantis bioculata Burm. I. Lange and Masse. Arch. EntwMech. Or,&t. 34: 680-741. QADRI, M.A.H. 1938. The life history and growth of the cockroach, Blatta orientalis, Linn. Bull. ent. Res. 21: 263-276. QADRI, M.A.H. 1940. On the development of the genitalia and their ducts of Orthopteroid insects. Trans. R. ent. Soc. Load. 22: 121-175. RAGGE, D.R. 1965. Grasshoppers, Crickets and Cockroaches. Warne, London: 299 pp.

RAKSHPAL, R. 1962. Diapause in the eggs of pryllus pennsylvanicus Burmeister (Orthoptera: Gryllidae). Can. J. Zool. 40: 179-194. RAO, C.R. 1952. Advanced Statistical Methods in Biometric Research. Wiley, New York: 390 pp. RAU, P. 1940. The life history of the wood-roach Parcoblatta 2nnsylvanica De Geer (Orthoptera: Blattidae). Ent. News 51: 4-9, 33-35. READIO, P.A. 1931. Dormancy in Reduvius2ersonatus (Linnaeus). Ann. ent. Soc. Am. 24: 19-39. REEVE, E.C.R. & HUXLEY, J.S. 1945. Some problems in the study of allometric growth. In Essays on Growth and Form, pp 121-156. (Eds. Le Gros Clark, W.E. & Medawar, P.B.) University Press, Oxford: 408 pp. REID, J.A. 1942. The relative sizes of different parts in beetles of the genus Laemophloeus (Coleopt., Cucujidae). Proof R. ent. Soc. Lond. (A) 17: 19--26. -368-

RICHARDS, 0.W. 1949. The relation between measurements of the successive instars of insects. Proc. R. ent. Soc. Lond. (A) 24: 8-10.

ROHLF, F.J. & SOKAL, R.R. 1962. The description of taxonomic relationship by factor analysis. Syst. Zool. 11: 1-16. ROSS, M.H. & COCHRAN, D.G. 1960. A simple method for sexing nymphal German cockroaches. Ann. ent. Soc. Am. 53: 550-551. ROTH, L.M. 1964. Control of reproduction in female cockroaches with special reference to Nauphoeta cinerea. I. First preoviposition period. J. Insect.Physiol. 10: 915-945. ROTH, L.M. 1967a. The evolutionary significance of rotation of the ootheca in the Blattaria. Psyche, Camb. 74: 85-109. ROTH, L.M. 1967b. Water changes in cockroach oothecae in relation to the evolution of ovoviviparity and viviparity. Ann. ent. Soc. Am. 60: 928-946. ROTH, L.M. 1968a. Oothecae of the Blattaria. Ann. ent. Soc. Am. 61: 83-111. ROTH, L.M. 1968b. Ovarioles of the Blattaria. Ann. ent. Soc. Am. 61: 132-140. ROTH, L.M. & STAY, B. 1959. Control of oocyte development in cockroaches. Science, N.Y. 130: 271-272. ROTH, L.M. & WILLIS, E.R. 1954a. The reproduction of cockroaches. Smithson. misc. Colins 122: 1-49. ROTH, L.M. & WILLIS, E.R. 1954b. Anastatus floridanus (Hymenoptera: Eupelmidae) a new parasite on the eggs of the cockroach Eurycotis_ floridana. Trans. Am. ent. Soc. 80: 29-41. ROTH, L.M. & WILLIS, E.R. 1954c. The biology of the cockroach egg parasite, Tetrastichus hagenowii (Hymenoptera: Eulophidae). Trans. Am. ent. Soc. 80: 53-72. ROTH, L.M. & WILLIS, E.R. 1955a. Water relations of cockroach oothecae. J. econ. Ent. 48: 33-36. -369-

ROTH, L.M. & WILLIS, E.R. 1955b. Relation of water loss to the hatching of eggs from detached oothecae of Blatella aermanica L. J. econ. Ent. 48: 57-60. ROTH, L.M. & WILLIS, E.R. 1955c. Water content of cockroach eggs during embryogenesis in relation to oviposition behaviour. J. ex.2.1_ool. 128: 489-509. ROTH, L.M. & WILLIS, E.R. 1956. Parthenogenesis in cockroaches. Ann. ent. Soc. Am. !.1-2: 195-204. ROTH, L.M. & WILLIS, E.R. 1957. Observations on the biology of Ectobius pallidus (Olivier) (Blattaria: Blattidae). Trans. Am. ent. Soc. fj: 31-37. ROTH, L.M. & WILLIS, E.R. 1958a. The biology of Panchlora nivea, with observations on the eggs of other Blattaria. Trans. Am. ent. Soc. 82(1957): 195-207. ROTH, L.M. & WILLIS, E.R. 1958b. An analysis of oviparity and viviparity in the Blattaria. Trans. Am. ent. Soc. §2(1957): 221-238. ROTH, L.M. & WILLIS, E.R. 1960. The biotic associations of cockroaches. Smithson. misc. Collns 141: 1-470. ROTH, L.M. & WILLIS, E.R. 1961. A study of bisexual and parthenogenetic strains of ycnoscelus surinamensis (Blattaria: Epilamprinae). Ann. ent. Soc. Am. 54: 12-25. SCHEDL, K.E. 1934. Statistische Untersuchungen Uber die Kopfkapselbreiten bei Blattwespen. Z. angew. Ent. 20: 449-460. SCUDDER, G.G.E. 1961. The comparative morphology of the insect ovipositor. Trans. R. ent. Soc. Lond. 113: 25-40. SEAL, H. 1964. Multivariate Statistical Analysis for Biologists. Methuen, London: 207 pp. SEN, P. & DAS GUPTA, S.K. 1958. "Dyar's Law" in the determination of larval instars in mosquitoes. Bull. Calcutta Sch. trop. Med. Hyg. 6: 69-70.

SHELFORD, R.W.C. 1906. Studies of the Blattidae. VI. Viviparity amongst the Blattidae. Trans. R. ent. Soc. Lond. 1906: 509-514. -370-

SHELFORD, R. 1912. The oothecae of Blattidae. Entomologist's Rec. J. Var. 24: 283-287.

SIMPSON, G.G., ROE, A. & LEWONTIN, R.C. 1960. quantitative Zoology (Revised El:motion). Harcourt, New York: 440 pp. SIIMA, K. & JANDA, jr., V. 1960. The changes in width of the head capsule during development of different sawfly species (Hym. Tenthredinioidea). (In Czech with English summary) Vest. dsl. zemed. Mus. 24: 7-15. SMITH, B.D. 1966. Effects of parasites and predators on a natural population of the aphid Acyrtosiphon spartii (Koch) on broom (Sarothamnus scoparius J. Anim. Ecol. 22: 253-267. SNODGRASS, R.E. 1937. The male genitalia of Orthopteroid insects. Smithson. misc. Colins 96: 1-107. SOKAL, R.R. 1962. Variation and covariation of characters of alate Pemphigus populi-transversus in eastern North America. Evolution, Lancaster, Pa. 16: 227-245. SPRENT, P. 1968. Linear relationships in growth and size studies. Biometrics 24: 639-656.

STAY, B., KING, A. & ROTH, L.M. 1960. Calcium oxalate in the oothecae of cockroaches. Ann. ent. Soc. Am. 53: 79-86. STOCK, A. & O'FARRELL, A.F. 1954. Cercal spinning glands in the cockroach, Blatella germanica L. Aust. J. Sci. 17: 64-66. STROUD, C.P. 1953. An application of factor analysis to the systematics of Kalotermes. Syst. Zool. 2: 76-92. SUTHERLAND, O.R.W. 1964. Alpine wetas in New Zealand. N.Z. Ent. 3: 16-17. SYMMONS, P.M. 1969. A morphometric measure of phase in the desert locust, Schistocerca regaria (Forsk.). Bull. ent. Res. 03-809. SZTERN, H. 1914. Wachstumsmessungen an Sphodromantis bioculata Burm. II. Lange, Breite and Hohe. Arch. EntwMech. Org. 40: 429-493.

-371-

TAYLOR, R.L. 1931. On Dyar's rule and its application to sawfly larvae. Ann. ent. Soc. Am. 24: 451-466. TEISSIER, G. 1948. La relation d' allometrie: sa signification statistique et biologique. Biometrics 14-53. TEISSIER, G. 1955. Allometrie de taille et variabilite chez Maia squinado. Archs Zool. exp. gen. 92: 221-264. TEISSIER, G. 1960. Relative growth. In The Physiology of Crustacea, Chat). 16. Vol. I. (Ed. Waterman T.H.) Academic Press, New York: 670 pp. THOMPSON, D'ARCY W. 1917. On Growth and Form. University Press, Cambridge: 794 pp. THOMPSON, G.H. 1951. The Factorial Analysis of Human Ability. University Press, London: 392 pp. TITSCHACK, E. 1926. Untersuchungen iiber das Wachstum, den Nahrungsverbrauch and die Eierzeugung. II. Tineola biselliella Hum. Gleichzeitig ein Beitrag zur Klarung der Insektenhautung. Z. wiss. Zool. 128: 509-569. UVAROV, B. 1966. Grasshoppers and Locusts. University Press, Cambridge: 01 pp. WAY, M.J. 1959. The effect of temperature, particularly during diapause, on the development of the egg of Iieptollasia coarctata Fallen (Diptera: Muscidae). Trans. R. ent. Soc. Lond. 111: 351-364. WAY, M.J. 1962. Definition of diapause. Ann. appl. Biol. 50: 595-596. WIGGLESWORTH, V.B. 1954. The Physiology of Insect Metamorphosis. University Press, Cambridge: 152 pp. WIGGLESWORTH, V.B. & BEAMENT, J.W.L. 1950. The respiratory mechanisms of some insect eggs. Q. Jl microsc. Sci. 91: 429-452. WILLIS, E.R., RISER, G.R. & ROTH, L.M. 1958. Observations on reproduction and development in cockroaches. Ann. ent. Soc. Am. 51: 53-69. -372-

ZABINSKI, J. 1936. Inconstancy of the number of moults during the post-embryonal development of certain Blattidae. Annls Mus. tool. pol. 11: 237-240. -373-

STATISTICAL APPENDIX.

Summary of Growth Data for E. lapponicus and E. panzeri.

Included in the summary are the following:- (1) Mean (ii) Standard Error (S.E.) (iii) Range

All dimensions are in millimetres

Each variable is represented by 10 replicates

Variables can be identified from the list on Pages 189-191. -374-

Summary of Growth Data.

E. lapponicus. 1st. Instar Male.

1 Vari- Vari- able Mean S.E. Range able Mean S.E. Range

1 0.317 0.0046 0.036 38 0.749 0.0072 0.074 2 0.312 0.0016 0.016 39 0.248 0.0035 0.032 3 0.163 0.0020 0.019 40 0.302 0.0067 0.075 4 0.204 0.0015 0.014 41 0.141 0.0020 0.017 5 0.169 0.0021 0.019 42 0.102 0.0007 0.006 6 0.116 0.0012 0.014 43 0.091 0.0018 0.016 7 0.130 0.0014 0.015 44 0.151 0.0018 0.015 8 0.049 0.0012 0.011 45 0.082 0.0005 0.006 9 0.059 0.0014 0.014 46 0.192 0.0026 0.022 10 0.065 0.0017 0,019 47 0.209 0.0039 0.036 11 0.131 0.0012 0.011 48 0.215 0.0035 0.031 12 0.633 0.0083 0.078 49 0.217 0.0036 0.033 13 0.393 0.0052 0.038 5o 0.207 0.0036 0.034 14 0.362 0.0059 0.047 51 0.199 0.0043 0.043 15 1.105 0.0128 0.130 52 0.179 0.0044 0.049 16 1.271 0.0127 0.140 53 0.146 0.0026 0.023 17 0.387 0.0054 0.046 54 0.114 0.0015 0.015 18 0.374 0.0056 0.048 55 0.146 0.0035 0.035 19 0.518 0.0049 0.054 56 1.208 0.0130 0.140 20 0.370 0.0031 0.031 57 0.138 0.0034 0.034 21 0.199 0.0030 0.031 58 0.116 0.0023 0.023 22 0.140 0.0029 0.034 59 0.153 0.0023 0.023 23 o.o8o 0.0015 0.017 6o 0.195 0.0034 0.033 24 0.073 0.0009 0.010 61 0.204 0.0028 0.031 25 0.081 0.0011 0.011 62 0.197 0.0040 0.043 26 0.136 0.0022 0.023 63 0.185 0.0049 0.052 27 0.079 0.0008 0.006 64 0.169 0.0040 0.040 28 0.606 0.0058 0.043 65 0.126 0.0023 0.022 29 0.524 0.0057 0.064 66 0.151 0.0033 0.038 30 0.223 0.0032 0.034 67 1.200 0.0214 0.240 31 0.214 0.0016 0.015 68 0.880 0.0081 0.078 32 0.100 0.0019 0.018 69 0.072 0.0014 0.014 33 0.082 0.0017 0.017 7o 0.074 0.0014 0.014 34 0.083 0.0010 0.009 71 0.170 0.0027 0.00 35 0.146 0.0016 0.017 72 0.173 0.0027 0.030 36 0.083 0.0007 0.007 73 0.042 0.0017 0.016 37 0.714 0.0076 0.068 74 0.047 0.0028 0.024 -375-

Summary. of Growth Data.

E. lapponicus. 1st. Instar Female.

Vari- Var±- able Mean S.E. Range able Mean S.E. Range

1 0.309 0.0025 0.032 38 0.767 0.0121 0.113 2 0.320 0.0030 0.027 39 0.249 0.0035 0.034 3 0.160 0.0020 0.022 40 0.307 0.0044 0.044 4 0.199 0.0025 0.029 41 0.141 0.0026 0.023 5 0.168 0.0029 0.029 42 0.103 0.0013 0.011 6 0.120 0.0009 0.009 43 0.091 0.0014 0.015 7 0.135 0.0017 0.018 44 0.156 0.0034 0.030 8 0.050 0.0007 0.005 45 0.086 0.0010 0.009 9 0.060 0.0016 0.015 46 0.190 0.0034 0.032 10 0.068 0.0009 0.009 47 0.203 0.0038 0.037 11 0.134 0.0013 0.012 48 0.207 0.0044 0.038 12 0.622 0.0088 0.079 49 0.212 0.0038 0.032 13 0.390 0.0047 0.040 50 0.202 0.0033 0.030 14 0.360 0.0047 0.040 51 0.195 0.0032 0.025 15 1.114 0.0159 0.140 52 0.174 0.0029 0:025 16 1.277 0.0198 0.160 53 0.145 0.0027 0.025 17 0.386 0.0075 0.057 54 0.114 0.0020 0.016 18 0.373 0.0074 0.053 55 0.145 0.0022 0.019 19 0.513 0.0081 0.080 56 1.194 0.0148 0.130 20 0.372 0.0058 0.060 57 0.129 0.0050 0.051 21 0.196 0.003? 0.042 58 0.115 0.0008 0.009 22 0.144 0.0031 0.029 59 0.149 0.0023 0.022 23 0.084 0.0015 0.014 60 0.196 0.0039 0.039 24 0.075 0.0010 0.008 61 0.211 0.0036 0.034 25 0.079 0.0009 0.009 62 0.201 0.0032 0.031 26 0.144 0.0029 0.028 63 0.182 0.0041 0.034 27 0.081 0.0008 o.008 64 0.161 0.0030 0.026 28 0.610 0.0105 0.119 65 0.118 0.0025 0.029 29 0.538 0.0077 0.077 66 0.118 0.0032 0.031 30 0.220 0.0049 0.050 67 1.192 0.0200 0.190 31 0.219 0.0041 0.035 68 0.886 0.0131 0.127 32 0.099 0.0020 0.019 69 0.076 0.0015 0.012 33 0.083 0.0011 0.011 70 0.074 0.0015 0.013 34 0.079 0.0014 0.011 71 0.163 0.0028 0.025 35 0.145 0.0031 0.026 72 0.159 0.0030 0.027 36 0.083 0.0011 0.013 73 0.049 0.0025 0.022 37 0.715 0.0111 0.106 74 0.055 0.0010 0.009 -376-

Summary of Growth Data.

E. Iapponicus. 2nd. Instar Male.

Vari- Vari- able Mean S.E. Range able Mean S.E. Range

1 0.384 0.0053 0.058 38 0.995 0.0169 0.150 2 0.369 0.0021 0.019 39 0.315 0.0034 0.037 3 0.183 0.0032 0.029 4o 0.418 0.0077 0.086 4 0.248 0.0044 0.040 41 0.181 0.0038 0.031 5 0.211 0.0018 0.016 42 0.125 0.0030 0.028 6 0.129 0.0018 0.018 43 0.108 0.0010 0.010 7 0.153 0.0030 0.025 44 0.177 0.0026 0.026 8 0.035 060019 0.018 45 0.095 0.0015 0.013 9 0.039 0.0013 0.014 46 0.240 0.0036 0.034 10 0.045 0.0016 0.017 47 0.271 0.0046 0.048 11 0.150 0.0035 0.039 48 0.292 0.0049 0.051 12 0.855 0.0106 0.093 49 0.299 0.0050 0.052 13 0.526 0.0064 0.068 50 0.286 0.0044 0.044 14 0.494 0.0059 0.061 51 0.270 0.0040 0.039 15 1:492 0.0169 0.180 52 0.237 0.0043 0,037 16 1.684 0.0201 0.220 53 0.175 0.0022 0.025 17 0.583 0.0088 0.091 54 0.132 0.0010 0.010 18 0.574 0.0090 0.102 55 0.171 0.0033 0.033 19 0.656 0.0085 0.092 56 1.581 0.0139 0.130 20 0.463 0.0056 0.054 57 0.172 0.0044 0.048 21 0.249 0.0041 0.042 58 0.166 0.0024 0.025 22 0.198 00030 0.031 59 0.191 0.0017 0.016 23 0.112 0.0020 0.020 6o 0.286 0.0070 0.067 24 0.093 0.0022 0.020 61 0.318 0.0061 o.o66 25 0.10 0.0011 0.009 62 0.301 0.0060 0.059 26 0.163 0.0025 0.025 63 0.286 0.0072 0.069 27 0.088 0.0010 0.010 64 0.263 0.0054 0.055 28 0.787 0.0111 0.119 65 0.182 0.0016 0.014 29 0.690 0.0081 0.078 66 0.190 0.0022 0.021 30 0.275 0.0037 0.037 67 1.580 0.0201 0.210 31 0.295 0.0042 0.044 68 1.170 0.0120 0.130 32 0.134 0.0032 0.027 69 0.089 0.0019 0.017 33 0.104 0.0014 0.011 7o 0.093 0.0022 0.020 34 0;104 0.0008 0.007 71 0.244 0.0051 0.053 35 0.167 0.0014 0.015 72 0.251 0.0047 0.049 36 0.092 0.0016 0.015 73 0.065 0.0039 0.040 37 0.925 0.0134 0.130 74 0.068 0.0019 0.017 -377-

Summary of Growth Data.

E. lapponicus. 2nd. Instar Female.

Vari- Vari- able Mean S.E. Range able Mean S.E. Range

1 0.386 0.0044 0.042 38 0.990 0.0154 0.149 2 0.369 0.0050 0.051 39 0.311 0.0071 0.061 3 0.189 0.0030 0.031 4o 0.410 0.0068 0.061 4 0.251 0.0067 0.068 41 0.175 0.0042 0.040 5 0.213 0.0032 0.037 42 0.125 0.0033 0.027 6 0.133 0.0011 0.010 43 0.107 0.0012 0.014 7 0.149 0.0026 0.024 44 0.180 0.0023 0.026 8 0.032 0.0015 0.016 45 0.096 o.0008 0.007 9 0.036 0.0016 0.019 46 0.236 0.0032 0.030 10 0.042 0.0016 0.021 47 0.267 0.0053 0.049 11 0.149 0.0020 0.021 48 0.286 o.0048 0.044 12 0.860 0.0127 0.123 49 0.295 0.0055 0.047 13 0.520 0.0067 0.075 50 0.283 0.0058 0.055 14 0.489 0.0073 0.073 51 0.264 0.0063 0.062 15 1.476 0.0187 0.190 52 0.228 0.0059 0.062 16 1.685 0.0218 0.200 53 0.173 0.0036 0.038 17 0.576 0.0098 0.102 54 0.130 0.0020 0.021 18 0.568 0.0093 0.100 55 0.168 0.0027 0.026 19 0.648 0.0110 0.093 56 1.566 0.0156 0.180 20 0.457 0.0077 0.069 57 0.178 0.0050 0.043 21 0.235 0.0063 0.062 58 0.164 0.0027 0.028 22 0.198 0.0023 0.020 59 0.185 0.0024 0.022 23 0.112 0.0024 0.021 60 0.284 0.0064 0.054 24 0.094 0.0020 0.020 61 0.318 0.0079 0.076 25 0.102 0.0013 0.011 62 0.297 0.0070 0.066 26 0.164 0.0029 0.030 63 0.286 0.0056 0.059 27 0.089 0.0018 0.019 64 0.282 0.0058 0.052 28 0.789 0.0110 0.096 65 0.145 0.0014 0.014 29 0.685 0.0082 0.065 66 0.146 0.0014 0.015 30 0.271 0.0041 0.041 67 1.577 0.0223 0.200 31 0.291 0.0043 0.039 68 1.173 0.0168 0.130 32 0.136 0.0031 0.032 69 0.095 0.0026 0.023 33 0.106 0.0013 0.016 70 0.092 0.0023 0.021 34 0.104 0.0010 0.010 71 0.257 0.0033 0.034 35 0.168 0.0023 0.023 72 0.250 0.0031 0.031 36 0.094 0.0014 0.010 73 0.066 0.0029 0.029 37 0.925 0.0134 0.120 74 0.075 0.0029 0.029 -378-

Summary of Growth Data.

E. lapponicus. 3rd. Instar Male.

Vari- Vari- able Mean S.E. Range able Mean S.E. Range

1 0.444 0.0049 0.050 38 1.302 0.0147 0.160 2 0.436 0.0031 0.037 39 0.398 0.0035 0.042 3 0.226 0.0020 0.021 4o 0.549 0.0066 0.071 4 0.320 0.0034 0.038 41 0.221 0.0035 0.040 5 0.262 0.0019 0.022 42 0.146 0.0018 0.016 6 0.152 0.0011 0.011 43 0.115 0.0012 0.012 7 0.194 0.0033 0.034 44 0.212 0.0012 0.012 8 0.033 0.0012 0.013 45 0.111 0.0014 0.013 9 0.036 0.0011 0.011 46 0.306 0.0033 0.031 10 0.041 0.0009 0.009 47 0.348 0.0040 0.044 11 0.178 0.0033 0.036 48 0.384 o.0048 0.057 12 1.134 0.0100 0.100 49 0.393 0.0046 0.054 13 0.695 0.0070 0.070 50 0.378 0.0046 0.047 14 0.667 0.0074 0.075 51 0.352 0.0042 0.038 15 1.935 0.0133 0.110 52 0.299 0.0054 0.043 16 2.182 0.0158 0.130 53 0.221 0.0025 0.025 17 0.849 0.0098 0.083 54 0.163 0.0021 0.016 18 0.866 0.0114 0.105 55 0.218 0.0034 0.034 19 o.84o 0.0068 0.074 56 2.025 0.0159 0.140 20 0.586 0.0045 0.044 57 0.233 0.0034 0.037 21 0.300 0.0034 0.028 58 0.233 0.0040 0.042 22 0.259 00036 0.035 59 0.249 0.0035 0.032 23 0.140 0.0031 0.033 6o 0.388 0.0039 0.043 24 0.112 0.0012 0.013 61 0.43 1 0.0051 0.059 25 0.114 0.0009 0.008 62 0.394 0.0068 0.073 26 0.199 0.0018 0.019 63 0.381 0.0032 0.033 27 0.107 o.0006 0.005 64 0.354 o.0046 0.051 28 1.03o 0.0086 0.090 65 0.247 0.0048 0.050 29 0.886 0.0069 0.069 66 0.274 0.0047 0.045 3o 0.344 0.0045 0.041 67 2.086 0.0224 0.240 31 0.391 0.0040 0.040 68 1.533 0.0145 0.150 32 0.171 0.0021 0.021 69 0.117 0.0012 0.013 33 0.124 0.0016 0.015 70 0.121 0.0010 0.010 34 0.113 0.0014 0.015 71 0.332 0.0053 0.056 35 0.201 0.0017 0.021 72 0.340 0.0058 0.058 36 0.113 0.0014 0.014 73 0.083 0.0016 0.015 37 1.217 0.0115 0.120 74 0.092 0.0020 0.019 -379-

summary of Growth Data.

E. lapponicus. 3rd. Instar Female.

Vari- Vari- able Mean S.E. Range able Mean S.E. Range

1 0.439 0.0031 0.030 38 1.277 0.0151 0.130 2 0.438 0.0028 0.031 39 0.396 0.0028 0.027 3 0.222 0.0024 0.026 4o 0.538 0.0061 0.056 4 0.314 0.0034 0.038 41 0.216 0.0034 0.037 5 0.251 0.0027 0.024 42 0.151 0.0023 0.022 6 0.148 0.0014 0.013 43 0.112 0.0009 0.009 7 0.181 0.0038 0.037 44 0.208 0.0023 0.026 8 0.032 0.0015 0.014 45 0.111 0.0006 0.006 9 0.036 0.0013 0.013 46 0.300 0.0043 0.045 10 0.040 0.0013 0.015 47 0.339 0.0039 0.036 11 0.169 0.0043 0.038 48 0.376 0.0059 0.060 12 1.114 0.0120 0.120 49 0.385 0.0059 0.058 13 0.681 0.0081 0.090 50 0.368 0.0042 0.045 14 0.653 0.0074 0;076 51 0.347 0.0048 0.047 15 1.936 0.0176 0.170 52 0.291 0.0038 0.036 16 2.150 0.0217 0.200 53 0.215 0.0028 0.034 17 0.823 0.0115 0.100 54 0.161 0.0022 0.019 18 0.826 0.0112 0.097 55 0.216 0.0026 0.028 19 0.817 o.0068 0.062 56 2.000 0.0223 0.240 20 0.572 0.0057 0.051 57 0.233 0.0048 0.048 21 0.285 0.0030 0.031 58 0.223 0.0057 0.058 22 0.254 0.0029 0.030 59 0.238 0.0033 0.032 23 0.135 0.0033 0.034 6o 0.383 0.0084 0.080 24 0.109 0.0022 0.022 61 0.418 0.0074 0.077 25 0.113 0.0017 0.015 62 0.386 0.0057 0.059 26 0.191 0.0025 0.025 63 0.369 0.0060 0.053 27 0.107 0.0015 0.015 64 0.444 0.0101 0.096 28 0.999 0.0102 0.104 65 0.152 0.0025 0.026 29 0.856 0.0090 0.089 66 0.170 0.0032 0.00 3o 0.338 0.0040 0.046 67 2.050 0.0236 0.200 31 0.381 0.0042 0.043 68 1.503 0.0138 0.130 32 0.165 0.0028 0.025 69 0.111 0.0013 0.013 33 0.122 0.0016 0.016 70 0.108 0.0014 0.014 34 0.113 0.0010 0.009 71 0.358 0.0064 0.061 35 0.199 0.0020 0.023 72 0.352 0.0063 0.063 36 0.111 0.0008 0.010 73 o.o84 0.0025 0.022 37 1.183 0.0169 0.160 74 0.099 0.0036 0.034 -380-

Summary of Growth Data.

E. lapponicus. 4th. Instar Male.

Vari- Vari- able Mean S.E. Range able Mean S.E. Range

1 0.525 0.0066 0.070 38 1.726 0.0178 0.190 2 0.517 0.0032 0.029 39 0.517 0.0042 0.043 3 0.275 0.0020 0.017 4o 0.719 0.0068 0.077 4 0.409 0.0061 0.064 41 0.275 0.0022 0.023 5 0.324 0.0041 0.037 42 0.180 0.0015 0.013 6 0.174 0.0027 0.029 43 0.134 0.0003 0.003 7 0.205 0.0029 0.023 44 0.261 0.0023 0.026 8 0.037 0.0012 0.011 45 0.138 0.0012 0.011 9 0.040 0.0012 0.011 46 0.406 0.0043 0.048 10 0.047 0.0008 0.008 47 0.448 0.0054 0.059 11 0.208 0.0020 0.018 48 0.499 0.0047 0.050 12 1.449 0.0113 0.110 49 0.510 0.0052 0.055 13 0.878 0.0070 0.082 5o 0.498 0.0061 0.063 14 0.885 0.0072 0.068 51 0.469 0.0076 0.074 15 2.473 0.0161 0.150 52 0.402 0.0062 0.075 16 2.757 0.0196 0.220 53 0.286 0.0046 0.043 17 1.251 0.0157 0.150 54 0.211 0.0035 0.040 18 1.270 0.0141 0.130 55 0.263 0.0017 0.019 19 1.058 0.0070 0.070 56 2.551 0.0191 0.200 20 0.750 0.0063 0.065 57 0.269 0.0042 0.040 21 0.385 0.0039 0.033 58 0.295 0.0046 0.041 22 0.352 0.0031 0.039 59 0.314 0.0034 0.035 23 0.184 0.0013 0.016 60 0.504 0.0073 0.081 24 0.144 0.0015 0.017 61 0.553 0.0077 0.092 25 0.124 0.0005 0.005 62 0.519 0.0077 0.093 26 0.247 0.0032 0.029 63 0.483 0.0073 0.081 27 0.133 0.0008 0.009 64 0.449 0.0052 0.052 28 1.306 0.0108 0.110 65 0.335 0.0037 0.041 29 1.145 0.0111 0.110 66 0.416 0.0059 0.056 3o 0.442 0.0031 0.032 67 2.634 0.0150 0.150 31 0.514 0.0050 0.055 68 1.952 0.0146 0.160 32 0.218 0.0037 0.034 69 0.124 0.0017 0.019 33 0.156 0.0024 0.020 70 0.151 0.0026 0.029 34 0.125 0.0008 0.007 71 0.496 0.0104 0.115 35 0.250 0.0026 0.025 72 0.529 0.0114 0.121 36 0.133 0.0008 0.009 73 0.098 0.0114 0.012 37 1.572 0.0116 0.130 74 0.109 0.0022 0.021 -381-

Summary of Growth Data.

E. lapponicus. 4th.Instar Female.

Vari- Vari- able Mean S.E. Range able Mean S.E. Range

1 0.517 0.0052 0.049 38 1.635 0.0141 0.130 2 0.517 0.0041 0.040 39 0.520 0.0028 0.024 3 0.270 0.0023 0.022 4o 0.688 0.0049 0.057 4 0.383 0.0055 0.056 41 0.261 0.0031 0.029 5 0.302 0.0023 0.024 42 0.177 0.0022 0.021 6 0.171 0.0008 0.007 43 0.131 0.0009 0.007 7 0.204 0.0040 0.047 44 0.254 0.0018 0.018 8 0.038 0.0009 0.008 45 0.134 0.0011 0.010 9 0.045 0.0011 0.012 46 0.389 0.0029 0.027 10 0.047 0.0013 0.014 47 0.441 0.0041 0.045 11 0.192 0.0035 0.030 48 0.488 0.0038 0.040 12 1.422 0.0150 0.160 49 0.499 0.0033 0.039 13 0.858 0.0055 0.049 50 0.482 0.0035 0.037 14 0.851 0.0069 0.061 51 0.449 0.0040 0.043 15 2.435 0.0121 0.140 52 0.380 0.0045 0.039 16 2.676 0.0135 0.150 53 0.288 0.0038 0.040 17 1.116 0.0067 0.070 54 0.212 0.0023 0.025 18 1.164 0.0072 0.080 55 0.257 0.0025 0.020 19 1.018 0.0078 0.080 56 2.525 0.0149 0.150 20 0.712 0.0055 0.051 57 0.282 0.0069 0.054 21 0.380 0.0046 0.045 58 0.295 0.0053 0.051 22 0.326 0.0041 0.038 59 0.310 0.0021 0.023 23 0.181 0.0026 0.028 60 0.491 0.0036 0.036 24 0.139 0.0016 0.012 61 0.523 0.0037 0.041 25 0.124 0.0011 0.009 62 o.484 0.0021 0.018 26 0.236 0.0028 0.030 63 0.466 0.0023 0.025 27 0.127 0.0008 0.007 64 0.712 0.0036 0.046 28 1.273 0.0084 0.080 65 0.171 0.0018 0.018 29 1.100 0.0094 o.o90 66 0.188 0.0023 0.029 3o 0.440 0.0033 0.032 67 2.593 0.0141 0.140 31 0.495 0.0062 0.050 68 1.878 0.0087 0.100 32 0.206 0.0030 0.029 69 0.125 0.0022 0.020 33 0.147 0.0016 0.015 7o 0.120 0.0023 0.019 34 0.126 0.0012 0.010 71 0.488 0.0041 0.046 35 0.240 0.0026 0.025 72 0.477 0.0035 0.031 36 0.131 0.0010 0.009 73 0.101 0.0017 0.016 37 1.509 0.0118 0.110 74 0.119 0.0024 0.023 -382-

Summary of Growth Data.

E. lapponicus. 5th. Instar Male.

Vari- Vari- able Mean S.E. Range able Mean S.E. Range

1 0.603 0.0062 0.062 38 2.382 0.0305 0.320 2 0.602 0.0040 0.039 39 0.593 0.0128 0.159 3 0.329 0.0019 0.015 4o 0.988 0.0174 0.181 4 0.539 0.0077 0.075 41 0.350 0.0048 0.044 5 0.408 0.0039 0.037 42 0.223 0.0048 0.048 6 0.204 0.0029 0.025 43 0.165 0.0031 0.028 7 0.229 0.0069 0.071 44 0.318 0.0035 0.033 8 0.056 0.0021 0.019 45 0.170 0.0025 0.026 9 o.o6o 0.0016 0.016 46 0.544 0.0050 0.046 10 0.067 0.0013 0.012 47 0.597 0.0063 0.063 11 0.244 0.0030 0.030 48 o.666 0.0088 0.097 12 1.774 0.0231 0.250 49 0.711 0.0073 0.082 13 1.077 0.0110 0.110 50 0.661 0.0090 0.097 14 1.186 0.0117 0.120 51 0.641 0.0077 0.077 15 3.038 0.0293 0.340 52 0.649 0.0080 0.093 16 3.717 0.0478 0.450 53 0.411 0.0071 0.074 17 2.267 0.0419 0.400 54 0.294 0.0049 0.058 18 2.189 0.0492 0.460 55 0.348 0.0044 0.044 19 1.354 0.0128 0.140 56 3.182 0.0341 0.330 20 0.996 0.0115 0.116 57 0.350 0.0091 0.087 21 0.433 0.0080 0.097 58 0.404 0.006o 0.056 22 0.486 00091 0.079 59 0.404 0.0057 0.058 23 0.248 0.0037 0.036 6o 0.675 0.0119 0.127 24 0.185 0.0024 0.025 61 0.752 0.0099 0.103 25 0.164 0.0022 0.022 62 0.708 0.0105 0.110 26 0.311 0.0043 0.043 63 0.673 0.0114 0.124 27 0.173 0.0029 0.030 64 0.646 0.0091 0.097 28 1.684 0.0223 0.230 65 0.567 0.0112 0.115 29 1.537 0.0219 0.240 66 0.746 0.0130 0.128 3o 0.509 0.0080 0.097 67 3.356 0.0346 0.340 31 0.701 0.0120 0.116 68 2.426 0.0232 0.230 32 0.280 0.0041 0.037 69 0.109 0.0094 0.098 33 0.200 0.0031 0.035 70 0.229 0.0036 0.032 34 0.168 0.0025 0.025 71 1.080 0.0222 0.200 35 0.310 0.0038 0.045 72 1.308 0.0181 0.210 36 0.166 0.0043 0.044 73 0.110 0.0046 0.049 37 2.033 0.0273 0.300 74 0.131 0.0037 0.037

...... J -383-

Summary of Growth Data.

E. lapponicus. 5th. Instar Female.

Vari- Vari- able Mean S.E. Range able Mean S.E. Range

1 0.599 0.0036 0.032 38 2.218 0.0200 0.180 2 0.627 0.0032 0.032 39 0.604 0.0059 0.061 3 0.333 0.0021 0.021 4o 0.943 0.0095 0.093 4 0.501 0.0040 0.040 41 0.326 0.0050 0.053 5 0.358 0.0032 0.031 42 0.208 0.0032 0.00 6 0.187 0.0019 0.017 43 0.181 0.0015 0.015 7 0.204 0.0101 0.085 44 0.313 0.0039 0.035 8 0.047 0.0020 0.019 45 0.170 0.0030 0.031 9 0.055 0.0019 0.018 46 0.510 0.0045 0.052 10 0.060 0.0024 0.024 47 0.574 0.0036 0.039 11 0.207 0.0025 0.027 48 0.639 0.0031 0.032 12 1.762 0.0193 0.170 49 o.684 o.004o 0.048 13 1.061 0.0080 0.070 50 0.642 0.0035 0.037 14 1.108 0.0102 0.100 51 0.606 0.0047 0.048 15 3.035 0.0119 0.090 52 0.524 0.0049 0.046 16 3.352 0.0249 0.290 53 0.398 0.0032 0.031 17 1.688 0.0125 0.110 54 0.294 0.0039 0.034 18 1.787 0.0149 0.130 55 0.344 0.0034 0.032 19 1.304 0.0091 0.10o 56 3.180 0.1054 1.150 20 0.929 0.0063 0.075 57 0.445 0.0064 0.069 21 0.435 0.0058 0.058 58 0.382 o.0046 0.034 22 0.454 0.0048 0.049 59 0.401 0.0044 0.038 23 0.226 0.0030 0.026 60 0.652 0.0050 0.053 24 0.178 0.0022 0.023 61 0.729 0.0068 0.060 25 0.161 0.000 0.028 62 0.681 0.0051 0.053 26 0.302 0.0028 0.028 63 0.667 0.0042 0.039 27 0.162 0.0026 0.027 64 1.180 0.0104 0.100 28 1.624 0.0125 0.120 65 0.193 0.0030 0.027 29 1.444 0.0110 0.120 66 0.333 0.0089 0.093 3o 0.524 0.0031 0.027 67 3.199 0.3188 3.402 31 0.658 0.0060 0.054 68 2.313 0.2307 2.388 32 0.264 0.0032 0.029 69 0.048 0.0042 0.040 33 0.187 0.0030 0.029 7o 0.042 0.0043 0.037 34 0.164 0.0020 0.018 71 0.687 0.0104 0.119 35 0.306 0.0014 0.014 72 0.663 0.0085 0.103 36 0.167 0.0022 0.020 73 0.137 0.0035 0.038 37 1.946 0.0169 0.160 74 0.157 0.0064 0.057 -384-

Summary of Growth Data.

E. lapponicus. Adult Male.

Vari- Vari- able Mean S.E. Range able Mean S.E. Range

1 0.602 0.0038 0.033 38 3.389 0.0322 0.300 2 0.651 0.0055 0.044 39 0.639 0.0077 0.078 3 0.377 0.0024 0.026 4o 1.347 0.0146 0.170 4 0.668 0.0084 0.10o 41 0.419 0.0055 0.062 5 0.550 0.0119 0.130 42 0.257 0.0029 0.027 6 0.230 0.0015 0.016 43 0.158 0.0014 0.013 7 0.205 0.0072 0.078 44 0.355 0.0022 0.020 8 0.096 0.0018 0.018 45 0.178 0.0027 0.027 9 0.096 0.0016 0.015 46 0.679 0.0043 o.o46 10 0.10o 0.0013 0.020 47 0.742 0.0065 0.056 11 0.164 0.0040 0.038 48 0.839 0.0112 0.110 12 1.957 0.0200 0.210 49 0.870 0.0116 0.100 13 1.102 0.0095 0.090 50 o.845 0.0095 0.088 14 1.339 0.0119 0.130 51 0.842 0.0136 0.117 15 2.950 0.0296 0.320 52 1.015 0.0127 0.127 16 2.217 0.0238 0.290 53 0.773 0.0059 0.051 17 8.835 0.0712 0.600 .54 0.608 0.0073 0.065 18 8.879 0.1245 1.200 55 0.510 0.0069 0.062 19 1.664 0.0106 0.100 56 3.066 0.0272 0.300 20 1.325 0.0079 o.o8o 57 o.6o5 0.0066 0.066 21 0.488 0.0032 0.040 58 0.470 o.0056 0.056 22 0.720 0.0072 0.079 59 0.505 0.0072 0.071 23 0.303 0.0038 0.039 6o 0.862 0.0114 0.100 24 0.216 0.0032 0.029 61 0.942 0.0119 0.115 25 0.174 0.0021 0.025 62 0.912 0.0117 0.126 26 0.362 0.0037 0.034 63 0.927 0.0121 0.141 27 0.192 0.0022 0.025 64 1.044 0.0133 0.132 28 2.321 0.0132 0.110 65 1.055 0.0123 0.125 29 2.332 0.0211 0.220 66 1.655 0.0155 0.140 3o 0.741 0.0057 0.061 67 3.233 0.0265 0.250 31 0.947 0.0075 0.087 68 2.456 0.0191 0.190 32 0.337 0.0029 0.031 69 0.000 0.0000 0.000 33 0.234 0.0021 0.022 70 0.265 0.0025 0.025 34 0.170 0.0019 0.020 71 2.937 0.0425 0.460 35 0.354 0.0026 0.027 72 3.836 0.0256 0.240 36 0.180 0.0020 0.022 73 0.070 0.0044 0.038 37 2.555 0.0182 0.200 74 0.097 0.0071 0.061

1..._ -385-

Summary of Growth Data.

E. lapponicus. Adult Female.

Vari- Vari- able Mean S.E. Range able Mean S.E. Range

1 0.650 0.0119 0.112 38 3.142 0.0336 0.260 2 0.730 0.0075 0.083 39 0.717 0.0083 0.085 3 0.386 0.0043 0.045 40 1.327 0.0186 0.160 4 0.618 0.0065 0.073 41 0.434 0.0068 0.068 5 0.447 0.0044 0.042 42 0.264 0.0038 0.039 6 0.202 0.0014 0.014 43 0.180 0.0032 0.028 7 0.208 0.0055 0.062 44 0.360 0.0027 0.031 8 0.080 0.0013 0.014 45 0.184 0.0018 0.022 9 0.082 0.0017 0.019 46 0.626 0.0092 0.078 10 0.082 0.0017 0.019 47 0.673 0.0081 0.066 11 0.140 0.0029 0.030 48 0.784 0.0088 0.078 12 2.164 0.0240 0.210 49 0.809 0.0084 o.085 13 1.124 0.0130 0.120 50 0.800 0.0064 0.060 14 1.275 0.0185 0.160 51 0.782 0.0081 0.066 15 3.321 0.0334 0.340 52 0.706 0.0055 0.056 16 2.247 0.0256 0.230 53 0.598 0.0071 0.066 17 4.765 0.0762 0.770 54 0.458 0.0046 0.043 18 6.205 0.1072 0.890 55 0.353 0.0050 0.049 19 1.624 0.0175 0.140 56 4.125 0.0391 0.350 20 1.223 0.0122 0.100 57 Q.564 0.0064 0.069 21 0.518 0.0056 0.066 58 0.466 0.0056 0.063 22 0.653 0.0068 0.067 59 0.504 0.0074 0.077 23 0.291 0.0038 0.037 6o 0.827 0.0081 0.065 24 0.217 0.0030 0.032 61 0.916 0.0108 0.110 25 0.188 0.0037 0.040 62 0.881 0.0095 0.073 26 0.357 0.0041 0.040 63 0.939 0.0091 0.095 27 0.187 0.0031 0.034 64 1.898 0.0158 0.150 28 2.055 0.0228 0.220 65 0.663 0.0076 0.088 29 1.965 0.0180 0.170 66 0.355 0.0022 0.026 30 0.607 0.0058 0.049 67 4.475 0.0355 0.330 31 0.912 0.0099 0.104 68 3.387 0.0279 0.250 32 0.337 0.0039 0.038 69 0.000 0.0000 0.000 33 0.233 0.0030 0.036 7o 0.000 0.0000 0.000 34 0.180 0.0016 0.017 71 1.108 0.0307 0.310 35 0.359 0.0038 0.043 72 1.099 0.0315 0.320 36 0.190 0.0033 0.036 73 0.115 0.0050 0.052 37 2.459 0.0256 0.230 74 0.179 0.0057 0.064 -386-

Summary of Growth Data.

E. sanzeri. 1st. Instar Male.

Vari- Vari- able Mean S.E. Range able Mean S.E. Range

1 0.294 0.0036 0.032 38 0.674 0.0044 0.045 2 0.297 0.0029 0.029 39 0.214 0.0012 0.014 3 0.153 0.0014 0.013 4o 0.288 0.0024 0.021 4 0.195 0.0013 0.014 41 0.134 0.0019 0.018 5 0.154 0.0014 0.016 42 0.096 0.0010 0.009 6 0.102 0.0011 0.011 43 0.087 0.0007 0.007 7 0.135 0.0017 0.019 44 0.146 0.0024 0.025 8 0.044 0.0015 0.014 45 0.074 0.0012 0.012 9 0.050 0.0017 0.017 46 0.171 0.0023 0.023 10 0.059 0.0019 0.019 47 0.208 0.0029 0.035 11 0.122 0.0011 0.010 48 0.225 0.0028 0.028 12 0.569 0.0050 0.048 49 0.236 0.0035 0.036 13 0.356 0.0022 0.022 50 0.228 0.0028 0.024 14 0.313 0.0020 0.020 51 0.215 0.0029 0.027 15 0.893 0.0042 0.046 52 0.189 0.0026 0.026 16 1.003 0.0061 0.063 53 0.130 0.0013 0.014 17 0.346 0.0040 0.041 54 0.099 0.0009 0.010 18 0.338 0.0032 0.033 55 0.122 0.0021 0.023 19 0.445 0.0028 0.024 56 1.030 0.0119 0.106 20 0.332 0.0015 0.011 57 0.131 0.0021 0.020 21 0.183 0.0023 0.022 58 0.107 0.0008 0.008 22 0.136 0.0015 0.018 59 0.152 0.0027 0.028 23 0.076 0.0011 0.009 60 0.210 0.0035 0.037 24 0.068 0.0008 0.007 61 0.233 0.0034 0.033 25 0.081 0.0010 0.010 62 0.229 0.0028 0.030 26 0.130 0.0021 0.019 63 0.216 0.0024 0.029 27 0.073 0.0013 0.011 64 0.193 0.0021 0.019 28 0.523 0.0044 0.038 65 0.136 0.0022 0.021 29 0.469 0.0037 0.035 66 0.133 0.0013 0.011 3o 0.188 0.0013 0.012 67 1.046 0.0110 0.141 31 0.203 0.0021 0.022 68 0.786 0.0088 0.081 32 0.096 0.0005 0.004 69 0.050 0.0016 0.015 33 0.074 0.0008 o.008 7o 0.048 0.0015 0.016 34 0.082 0.0011 0.010 71 0.162 0.0023 0.022 35 0.134 0.0021 0.019 72 0.159 0.0019 0.019 36 0.074 0.0016 0.017 73 0.045 0.0014 0.013 37 0.626 0.0061 0.057 74 0.049 0.0019 0.017 -387-

Summary of Growth Data.

E. panzeri. 1st. Instar Female.

Vari- Vari- able Mean S.E. Range able Mean S.E. Range

1 0.301 0.0027 0.029 38 o.686 0.0048 0.041 2 0.301 0.0033 0.029 39 0.218 0.0026 0.028 3 0.150 0.0009 0.009 40 0.295 0.0038 0.037 4 0.197 0.0027 0.025 41 0.139 0.0026 0.022 5 0.159 0.0011 0.011 42 0.101 0.0016 0.013 6 0.104 0.0009 0.008 43 0.090 o.0008 o.008 7 0.134 0.0015 0.012 44 0.147 0.0017 0.016 8 0.047 0.0014 0.014 45 0.075 0.0009 0.007 9 0.055 0.0015 0.015 46 0.174 0.0017 0.018 10 0.064 0.0014 0.016 47 0.211 0.0027 0.032 11 0.121 0.0011 0.010 48 0.225 0.0039 0.042 12 0.576 0.0062 0.067 49 0.232 0.0046 0.051 13 0.357 0.0033 0.033 5o 0.225 0.0048 0.055 14 0.316 0.0030 0.026 51 0.211 0.0051 0.056 15 0.893 0.0099 0.099 52 0.181 0.0035 0.040 16 1.009 0.0106 0.117 53 0.126 0.0015 0.014 17 0.349 0.0041 0.043 54 0.098 0.0011 0.009 18 0.343 0.0036 0.038 55 0.124 0.0019 0.016 19 0.445 0.0030 0.026 56 1.037 0.0146 0.140 20 0.334 0.0026 0.024 57 0.128 0.0013 0.012 21 0.193 0.0013 0.013 58 0.115 0.0019 0.017 22 0.136 0.0022 0.022 59 0.152 0.0020 0.017 23 0.074 0.0009 0.009 Go 0.212 0.0040 0.037 24 0.067 0.0011 0.011 61 0.233 0.0062 0.060 25 0.084 0.0007 0.006 62 0.228 0.0058 0.059 26 0.131 0.0014 0.013 63 0.210 0.0065 0.064 27 0.073 0.0014 0.013 64 0.191 0.0052 0.048 28 0.524 0.0043 0.039 65 0.132 0.0024 0.021 29 0.472 0.0032 0.025 66 0.123 0.0020 0.019 30 0.192 0.0015 0.014 67 1.047 0.0147 0.138 31 0.208 0.0031 0.027 68 0.783 0.0107 0.091 32 0.098 0.0013 0.014 69 0.045 0.0011 0.011 33 0.077 0.0012 0.012 70 0.047 0.0013 0.012 34 0.085 0.0011 0.009 71 0.155 0.0036 0.041 35 0.137 0.0013 0.015 72 0.159 0.0037 0.046 36 0.073 0.0014 0.015 73 0.042 0.0014 0.012 37 0.630 0.0062 0.054 74 0.048 0.0014 0.014 -388-

Summary of Growth Data.

E. panzeri. 2nd. Instar Male.

Vari- Vari- able Mean S.E. Range able Mean S.E. Range

1 0.335 0.0029 0.031 38 0.856 0.0159 0.148 2 0.333 0.0035 0.035 39 0.263 0.0031 0.00 3 0.174 0.0016 0.016 4o 0.382 0.0056 0.063 4 0.234 0.0027 0.025 41 0.169 0.0029 0.031 5 0.178 0.0020 0.021 42 0.115 0.0022 0.025 6 0.111 0.0014 0.014 43 0.091 0.0013 0.012 7 0.133 0.0020 0.018 44 0.161 0.0019 0.017 8 0.025 0.0010 0.009 45 0.076 0.0018 0.018 9 0.029 0.0009 0.009 46 0.204 0.0024 0.027 10 0.034 0.0013 0.013 47 0.247 0.0026 0.028 11 0.139 0.0016 0.014 48 0.272 0.0037 0.036 12 0.727 0.0068 0.066 49 0.281 0.0048 0.041 13 0.444 0.0037 0.040 50 0.271 0.0053 0.044 14 0.399 0.0042 0.047 51 0.252 0.0050 0.044 15 1.152 0.0132 0.130 52 0.218 0.0042 0.047 16 1.272 0.0116 0.120 53 0.151 0.0020 0.019 17 0.471 0.0065 0.073 54 0.111 0.0010 0.012 18 0.469 0.0067 0.070 55 0.138 0.0012 0.012 19 0.541 0.0060 0.064 56 1.274 0.0133 0.150 20 0.395 0.0048 0.049 57 0.141 0.0032 0.030 21 0.210 0.0038 0.038 58 0.151 0.0026 0.024 22 0.179 0.0025 0.028 59 0.186 0.0022 0.022 23 0.098 0.0014 0.013 Go 0.262 0.0047 0.038 24 0.078 0.0020 0.019 61 0.286 0.0071 0.058 25 0.088 0.0019 0.018 62 0.278 0.0072 o.o58 26 0.149 0.0033 0.036 63 0.258 0.0062 0.052 27 0.076 0.0015 0.014 64 0.234 0.0047 0.041 28 0.642 0.0070 0.063 65 0.169 0.0022 0.023 29 0.574 0.0073 0.076 66 0.160 0.0022 0.020 30 0.226 0.0027 0.024 67 1.319 0.0153 0.180 31 0.268 0.0029 0.029 68 0.987 0.0103 0.119 32 0.125 0.0014 0.014 69 0.067 0.0026 0.025 33 0.092 0.0011 0.010 70 0.063 0.0030 0.033 34 0.091 0.0011 0.012 71 0.207 0.0051 0.048 35 0.153 0.0023 0.025 72 0.204 0.0050 0.048 36 0.078 0.0014 0.015 73 0.057 0.0019 0.017 37 0.765 0.0101 0.101 74 0.060 0.0020 0.020

.6!....M•11•M -389-

Summary of Growth Data.

E. panzeri. 2nd. Instar Female.

r Vari- Vari- able Mean S.E. Range able Mean S.E. Range

1 0.329 0.0036 0.034 38 0.860 0.0095 0.094 2 0.331 0.0025 0.021 39 0.268 0.0020 0.016 3 0.166 0.0012 0.010 40 0.388 0.0040 0.039 4 0.228 0.0024 0.023 41 0.165 0.0016 0.013 5 0.184 0.0023 0.024 42 0.116 0.0017 0.014 6 0.111 0.0011 0.010 43 0.096 0.0005 0.004 7 0.131 0.0036 0.036 44 0.164 0.0015 0.015 8 0.026 0.0008 0.007 45 0.075 0.0014 0.013 9 0.031 0.0016 0.017 46 0.203 0.0026 0.024 10 0.036 0.0022 0.024 47 0.248 0.0027 0.028 11 0.136 0.0019 0.022 48 0.278 0.0032 0.026 12 0.733 0.0060 0.057 49 0.285 0.0043 0.042 13 0.451 0.0048 0.054 5o 0.278 0.0038 0.037 14 0.404 0.0037 0.040 51 0.257 0.0044 0.041 15 1.170 0.0102 0.120 52 0.218 0.0025 0.023 16 1.177 0.0168 0.123 53 0.153 0.0030 0.030 17 0.475 0.0079 0.090 54 0.108 0.0011 0.012 18 0.468 0.0055 0.066 55 0.142 0.0012 0.011 19 0.545 0.0039 0.039 56 1.291 0.0155 0.180 20 0.404 0.0105 0.116 57 0.148 0.0021 0.020 21 0.212 0.0027 0.028 58 0.153 0.0022 0.023 22 0.175 0.0022 0.022 59 0.188 0.0022 0.023 23 0.096 0.0010 0.009 60 0.270 0.0039 0.038 24 0.083 0.0012 0.011 61 0.295 0.0052 0.051 25 0.089 0.0019 0.017 62 0.285 0.0049 0.042 26 0.146 0.0021 0.024 63 0.263 0.0045 0.040 27 0.077 0.0018 0.018 64 0.263 0.0052 0.045 28 0.639 0.0061 0.053 65 0.138 0.0011 0.011 29 0.579 0.0042 0.042 66 0.122 0.0036 0.036 30 0.229 0.0020 0.021 67 1.329 0.0159 0.140 31 0.267 0.0031 0.031 68 1.010 0.010o 0.083 32 0.123 0.0011 0.011 69 0.047 0.0015 0.016 33 0.090 0.0017 0.014 70 0.052 0.0013 0.014 34 0.092 0.0018 0.017 71 0.201 0.0042 0.037 35 0.151 0.0020 0.018 72 0.208 0.0050 0.046 36 0.076 0.0013 0.013 73 0.055 0.0020 0.021 37 0.779 0.0075 0.075 74 0.063 0.0009 0.008 -390-

Summary of Growth Data.

E. panzeri. 3rd. Instar Male.

Vari- i Vari- able Mean S.E. Range able Mean S.E. Range

1 0.380 0.0057 0.064 38 1.148 0.0117 0.120 2 0.385 0.0047 0.046 39 0.342 0.0037 0.033 3 0.202 0.0023 0.025 4o 0.509 o.0086 0.078 4 0.290 0.0032 0.032 41 0.217 0.0026 0.030 5 0.219 0.0030 0.024 42 0.143 0.0017 0.021 6 0.133 0.0016 0.014 43 0.105 0.0014 0.016 7 0.156 0.0024 0.026 44 0.194 0.0021 0.019 8 0.024 0.0006 0.006 45 0.089 0.0012 0.012 9 0.030 o.0008 0.008 46 0.251 0.0032 0.031 10 0.033 0.0013 0.012 47 0.309 0.0035 0.033 11 0.170 0.0021 0.021 48 0.343 0.0059 o.058 12 0.945 0.0132 0.144 49 0.356 0.0063 0.062 13 0.568 0.0064 0.063 50 0.345 0.0070 0.070 14 0.522 0.0065 0.059 51 0.317 0.0068 0.066 15 1.542 0.0118 0.130 52 0.274 0.0070 0.072 16 1.682 0.0103 0.090 53 0.184 0.0045 0.052 17 0.675 0.0081 0.071 54 0.131 0.0022 0.022 18 0.687 0.0081 0.081 55 0.170 0.0028 0.025 19 0.706 0.0073 0.077 56 1.630 0.0163 0.170 20 0.507 0.0059 0.058 57 0.189 0.0037 0.038 21 0.275 0.0043 0.041 58 0.200 0.0027 0.025 22 0.244 0.0034 0.031 59 0.233 0.0032 0.036 23 0.132 0.0023 0.025 6o 0.340 0.0071 0.074 24 0.105 0.0013 0.015 61 0.365 0.0081 0.086 25 0.106 0.0022 0.021 62 0.343 0.0121 0.128 26 0.174 0.0027 0.025 63 0.327 0.0075 0.079 27 0.084 0.0024 0.025 64 0.297 0.0067 0.073 28 0.844 0.0090 0.082 65 0.221 0.0054 0.056 29 0.767 0.0076 0.070 66 0.231 0.0039 0.042 3o 0.297 0.0026 0.023 67 1.700 0.0174 0.140 31 0.360 0.0044 0.040 68 1.250 0.0150 0.150 32 0.156 0.0017 0.017 69 0.083 0.0020 0.018 33 0.113 0.0014 0.015 7o 0.076 0.0025 0.025 34 0.100 0.0015 0.016 71 0.296 0.0087 0.093 35 0.176 0.0021 0.021 72 0.287 0.0084 0.084 36 0.086 0.0018 0.017 73 0.073 0.0019 0.020 37 1.020 0.0112 0.103 74 0.088 0.0025 0.030 -391-

Sum..,ry of Growth Data.

E. panzeri. 3rd. Instar Female.

VariL Vari- able Mean S.E. Range able Mean S.E. Range

1 0.385 0.0051 0.045 38 1.145 0.0150 0.140 2 0.389 0.0033 0.029 39 0.344 0.0050 0.050 3 0.204 0.0037 0.031 4o 0.501 0.0072 0.067 4 0.286 0.0058 0.057 41 0.207 0.0043 0.038 5 0.215 0.0018 0.015 42 0.139 0.0027 0.026 6 0.127 0.0018 0.021 43 0.100 0.0021 0.020 7 0.159 0.0036 0.039 44 0.185 0.0020 0.017 8 0.024 0.0011 0.011 45 0.089 0.0016 0.013 9 0.029 0.0015 0.015 46 0.249 0.0027 0.028 10 0.035 0.0016 0.017 47 0.304 0.0020 0.020 11 0.161 0.0018 0.015 48 0.336 0.0026 0.028 12 0.945 0.0098 0.097 49 0.354 0.0051 0.051 13 0.572 0.0068 0.061 50 0.344 0.0039 0.044 14 0.516 0.0064 0.064 51 0.319 0.0045 0.042 15 1.539 0.0146 0.140 52 0.268 0.0047 0.057 16 1.655 0.0127 0.120 53 0.191 0.0026 0.032 17 0.643 o.0064 0.068 54 0.138 0.0018 0.020 18 0.650 0.0080 0.075 55 0.174 0.0024 0.022 19 0.700 0.0078 0.060 56 1.639 0.0126 0.120 20 0.503 0.0072 0.067 57 0.187 0.0026 0.022 21 0.271 0.0026 0.024 58 0.200 0.0024 0.028 22 0.238 0.0040 0.037 59 0.231 0.0030 0.028 23 0.128 0.0020 0.020 60 0.331 0.0037 0.036 24 0.102 0.0015 0.017 61 0.364 0.0063 0.064 25 0.100 0.0012 0.013 62 0.353 0.0058 0.059 26 0.174 0.0022 0.022 63 0.332 0.0051 0.046 2? o.o86 0.0011 0.011 64 0.412 0.0059 0.063 28 0.832 0.0110 0.099 65 0.136 0.0036 0.031 29 0.758 0.0112 0.100 66 0.130 0.0031 0.028 30 0.289 0.0041 0.039 67 1.705 0.0223 0.230 31 0.352 0.0058 0.056 68 1.233 0.0156 0.150 32 0.155 0.0032 0.033 69 0.045 0.0028 0.031 33 0.109 0.0015 0.016 7o 0.050 0.0023 0.022 34- 0.098 0.0012 0.013 71 0.259 0.0030 0.033 35 0.180 0.0015 0.017 72 0.265 0.0031 0.033 36 0.085 0.0012 0.013 73 0.066 0.0027 0.024 37 1.005 0.0117 0.100 74 o.088 0.0035 0.030

____ -392-

Summary of Growth Data.

E. panzeri. 4th. Instar Hale.

VariL Vari- able Mean. S.E. Range able Mean S.E. Range

1 0.471 0.0055 0.061 38 1.511 0.0114 0.100 2 0.450 0.0052 0.048 39 0.423 0.0042 0.051 3 0.247 0.0014 0.015 4o 0.655 o.0046 0.047 4 0.356 0.0043 0.041 41 0.271 0.0026 0.029 5 0.264 0.0015 0.012 42 0.182 0.0020 0.022 6 0.147 0.0012 0.012 43 0.118 0.0011 0.011 7 0.179 0.0031 0.033 44 0.224 0.0038 0.036 8 0.028 0.0012 0.013 45 0.102 0.0015 0.015 9 0.033 0.0013 0.015 46 0.310 0.0039 0.040 10 0.038 0.0009 0.010 47 0.379 0.0026 0.028 11 0.190 0.0025 0.027 48 0.425 0.0052 0.045 12 1.179 0.0077 0.080 49 0.444 0.0053 0.048 13 0.704 0.0036 0.042 5o 0.432 0.0056 0.047 14 0.660 0.0034 0.034 51 0.405 0.0047 0.047 15 1.888 0.0179 0.210 52 0.372 0.0028 0.030 16 2.057 0.0150 0.170 53 0.240 0.0036 0.027 17 0.959 0.0093 0.112 54 0.175 0.0018 0.016 18 0.977 0.0135 0.157 55 0.206 0.0037 0.042 19 0.886 0.0037 0.036 56 1.970 0.0122 0.120 20 0.648 0.0031 0.028 57 0.219 0.0030 0.032 21 0.331 0.0051 0.045 58 0.253 0.0019 0.021 22 0.320 0.0027 0.027 59 0.283 0.0039 0.041 23 0.172 0.0024 0.022 6o 0.430 0.0054 0.047 24 0.133 0.0019 0.022 61 0.471 0.0088 0.082 25 0.118 0.0019 0.019 62 0.454 0.0077 0.072 26 0.207 0.0026 0.022 63 0.421 0.0058 0.050 27 0.096 0.0015 0.014 64 0.385 0.0059 0.058 28 1.071 0.0046 0.050 65 0.306 0.0043 0.042 29 1.002 0.0065 0.065 66 0.385 0.0151 0.179 3o 0.358 0.0065 0.070 67 2.049 0.0179 0.180 31 0.464 0.0027 0.025 68 1.506 0.0119 0.120 32 0.207 0.0038 0.030 69 0.102 0.0026 0.030 33 0.148 0.0026 0.026 7o 0.065 0.0033 0.031 34 0.115 0.0015 0.015 71 0.533 0.0248 0.251 35 0.212 0.0032 0.031 72 0.485 0.0194 0.198 36 0.101 0.0017 0.015 73 0.087 0.0024 0.025 37 1.289 0.0075 0.070 74 0.112 0.0024 0.025 -393-

Summary of Growth Data.

E. panzeri. 4th. Instar Female.

Vari- Vari- able Mean S.E. Range able Mean S.E. Range

1 0.467 0.0049 0.052 38 1.507 0.0087 0.080 2 0.467 0.0037 0.044 39 0.429 0.0047 0.047 3 0.248 0.0019 0.018 4o 0.661 0.0041 0.049 4 0.350 0.0061 0.062 41 0.265 0.0040 0.045 5 0.255 0.0036 0.043 42 0.173 0.0030 0.029 6 0.141 0.0015 0.014 43 0.121 0.0019 0.021 7 0.170 0.0032 0.026 44 0.223 0.0025 0.025 8 0.029 0.0007 o.006 45 0.105 0.0023 0.020 9 0.034 0.0010 0.011 46 0.310 0.0033 0.034 10 0.038 0.0009 0.008 47 0.379 0.0054 0.060 11 0.182 0.0022 0.023 48 0.421 0.0074 0.062 12 1.199 0.0124 0.120 49 0.445 0.0098 0.083 13 0.713 0.0073 0.078 50 0.438 0.0098 0.095 14 0.658 o.0046 o.o46 51 0.412 0.0100 0.086 15 1.966 0.0147 0.130 52 0.357 0.0074 0.070 16 2.083 0.0181 0.190 53 0.242 0.0056 0.059 17 0.834 0.0084 0.083 54 0.111 0.0016 0.018 18 0.886 0.0106 0.100 55 0.134 0.0017 0.018 19 0.882 0.0063 0.057 56 2.048 0.0227 0.220 20 0.625 0.0036 0.028 57 0.241 0.0042 0.038 21 0.331 0.0062 0.073 58 0.260 0.0031 0.027 22 0.315 0.0021 0.019 59 0.282 0.0036 0.034 23 0.161 0.0022 0.021 6o 0.430 0.0081 0.078 24 0.129 0.0022 0.022 61 0.467 0.0097 0.079 25 0.119 0.0016 0.015 62 0.459 0.0098 0.083 26 0.205 0.0018 0.019 63 0.437 0.0067 0.062 27 0.103 0.0018 0.018 64 0.647 0.0064 0.065 28 1.061 0.0071 o.o6o 65 0.236 0.0052 0.052 29 0.977 0.0069 0.071 66 0.274 0.0058 0.061 30 0.368 0.0034 0.029 67 2.135 0.0186 0.170 31 0.459 0.0028 0.030 68 1.582 0.0140 0.170 32 0.200 0.0030 0.025 69 0.034 0.0039 0.045 33 0.144 0.0010 0.011 7o 0.043 0.0042 0.044 34 0.121 0.0013 0.013 71 0.382 0.0078 0.070 35 0.209 0.0025 0.024 72 0.390 0.0088 0.079 36 0.103 0.0019 0.020 73 0.092 0.0027 0.022 37 1.297 0.0090 0.090 74 0.119 0.0027 0.028 -394-

Summary of Growth Data.

E. panzeri. 5th. Instnr Male.

Vari- Vari- able Mean S.E. Range able Mean S.E. Range

1 0.532 0.0052 0.047 38 2.070 0.0182 0.160 2 0.536 0.0049 0.047 39 0.513 0.0068 0.076 3 0.296 0.0030 0.030 40 0.880 0.0112 0.099 4 0.455 0.0029 0.032 41 0.353 0.0050 0.043 5 0.335 0.0023 0.025 42 0.230 0.0027 0.024 6 0.174 0.0026 0.023 43 0.140 0.0014 0.014 7 0.194 0.0033 0.030 44 0.274 0.0033 0.039 8 0.048 0.0011 0.011 45 0.122 0.0013 0.012 9 0.050 0.0012 0.011 46 0.413 0.0026 0.022 10 0.055 0.0018 0.018 47 0.498 0.0058 0.055 11 0.218 0.0021 0.016 48 0.568 0.0059 0.054 12 1.477 0.0130 0.130 49 0.595 0.0079 0.085 13 0.872 0.0060 0.073 5o 0.582 0.0082 0.091 14 0.854 0.0063 0.073 51 0.558 0.0062 0.066 15 2.367 0.0202 0.210 52 0.574 0.0054 0.051 16 2.697 0.0349 0.420 53 0.360 0.0044 0.044 17 1.597 0.0120 0.120 54 0.253 0.0032 0.033 18 1.641 0.0129 0.110 55 0.267 0.0035 0.037 19 1.152 0.0111 0.100 56 2.415 0.0218 0.220 20 0.858 0.0067 0.062 57 0.275 0.0038 0.033 21 0.406 o.0054 0.063 58 0.326 0.0030 0.030 22 0.439 0.0048 0.047 59 0.364 0.0026 0.024 23 0.231 0.0023 0.020 60 0.578 0.0084 0.095 24 0.178 0.0024 0.020 61 0.613 0.0102 0.113 25 0.140 0.0013 0.013 62 0.585 0.0087 0.096 26 0.263 0.0028 0.029 63 0.557 0.0070 0.086 27 0.118 0.0024 0.019 64 0.531 0.0050 0.056 28 1.404 0.0124 0.120 65 0.487 0.0072 0.065 29 1.343 0.0110 0.090 66 0.625 0.0061 0.062 3o 0.436 0.0076 0.092 67 2.498 0.0147 0.130 31 0.615 0.0045 0.048 68 1.858 0.0152 0.16o 32 0.268 0.0037 0.034 69 0.129 0.0037 0.035 33 0.189 0.0016 0.014 70 0.029 0.0025 0.027 34 0.139 0.0016 0.016 71 0.876 0.0098 0.117 35 0.264 0.0028 0.00 72 0.769 0.0091 0.091 36 0.120 0.0016 0.019 73 0.092 0.0024 0.029 37 1.705 0.0145 0.140 74 0.128 0.0053 0.050 -395-

Summary_ of Growth Data.

E. panzeri. 5th. Instar Female.

Vari- Vari- able Mean S.E. Range able Mean S.F. Range

1 0.531 0.0070 0.070 38 1.988 0.0165 0.150 2 0.548 0.0049 0.054 39 0.533 0.0053 0.048 3 0.305 0.0032 0.030 40 0.886 0.0077 0.069 4 0.434 0.0057 0.057 41 0.3,:8 0.0036 0.041 5 0.307 0.0021 0.021 42 0.225 0.0018 0.017 6 o.161 0.0017 0.015 43 0.140 0.0018 0.018 7 0.175 0.0055 0.061 44 0.272 0.0029 0.034 8 0.041 0.0013 0.013 45 0.123 0.0026 0.022 9 0.050 0.0017 0.019 46 0.385 0.0035 0.036 10 0.055 0.0020 0.023 47 0.485 0.0040 0.040 11 0.189 0.0017 0.019 48 0.542 0.0050 0.045 12 1.513 0.0105 0.120 49 0.579 0.0067 0.056 13 o.884 0.0065 0.063 5o 0.573 0.0073 0.075 14 0.835 0.0066 0.056 51 0.546 0.0076 0.074 15 2.470 0.0183 0.200 52 0.478 0.0054 0.062 16 2.566 0.0151 0.180 53 0.330 0.0043 0.043 17 1.100 0.0080 o.o8o 54 0.246 0.0026 0.029 18 1.193 0.0100 0.090 55 0.278 0.0041 0.044 19 1.117 0.0072 0.070 56 2.545 0.0193 0.180 20 0.807 0.0052 0.054 57 0.330 0.0030 0.028 21 0.405 0.0055 0.054 58 0.323 0.0030 0.028 22 0.422 0.0040 0.036 59 0.351 0.0031 0.027 23 0.220 0.0038 0.043 6o 0.551 0.0075 0.067 24 0.168 0.0020 0.023 61 0.611 0.0096 0.099 25 0.145 0.0019 0.022 62 0.591 0.0081 0.080 26 0.260 0.0025 0.022 63 0.575 0.0076 0.062 27 0.127 0.0031 0.026 64 0.990 0.0076 0.071 28 1.364 0.0099 0.090 65 0.204 0.0068 0.065 29 1.265 0.0097 0.090 66 0.245 0.0025 0.023 30 0.455 0.0035 0.038 67 2.684 0.0181 0.200 31 0.612 0.0069 0.075 68 1.948 0.0183 0.190 32 0.256 0.0031 0.029 69 0.032 0.0157 0.151 33 0.179 0.0031 0.034 70 0.149 0.0653 0.686 34 0.139 0.0012 0.012 71 0.512 0.0054 0.051 35 0.254 0.0038 0.038 72 0.522 0.0057 0.045 36 0.124 0.0025 0.022 73 0.102 0.0036 0.038 37 1.653 0.0105 0.110 74 0.149 0.0041 0.039 -396-

Summary of Growth Data.

E. panzeri. Adult Male.

Vari- Vari- able Mean S.E. Range able Mean S.E. Range

1 0.563 0.0099 0.096 38 2.850 0.0401 0.390 2 0.578 0.0039 0.042 39 0.530 0,0040 0.045 3 0.342 0.0032 0.033 4o 1.226 0.0151 0.130 4 0.544 0.0073 0.068 41 0.452 0.0045 0.050 5 0.389 0.0041 0.039 42 0.284 0.0037 0.043 6 0.193 0.0025 0.025 43 0.150 0.0024 0.026 7 0.176 0.0053 0.056 44 0.311 0.0039 0.036 8 0.077 0.0027 0.032 45 0.116 0.0013 0.014 9 0.083 0.0013 0.013 46 0.476 0.0035 0.039 10 0.032 0.0013 0.012 47 0.561 0.0043 0.040 11 0.164 0.0016 0.014 48 0.635 0.0062 0.061 12 1.657 0.0088 0.090 49 0.687 0.0074 0.069 13 0.878 0.0040 0.037 5o 0.680 0.0018 0.031 14 0.891 0.0077 0.070 51 0.687 0„0066 0.059 15 2.409 0.0102 0.100 52 0.912 0.0099 0.097 16 1.591 0.0062 0.070 53 0.746 0.0083 0.089 17 5.354 0.0719 o.65o 54 0.523 0,0066 0.067 18 5.738 0.0641 0.640 55 0.367 0.0069 0.075 19 1.420 0.0167 0.150 56 2.215 0.0165 0.170 20 1.135 0.0162 0.180 57 0.488 0,0067 0.067 21 0.428 0.0029 0.030 58 0.341 0.0041 0.042 22 0.655 0.0091 0.090 59 0.423 0.0049 0.051 23 0.318 0.0034 0.033 6o 0.646 0.0063 0.057 24 0.242 0.0043 0.038 61 0.690 0.0073 0.077 25 0.148 0.0015 0.017 62 0.681 0.0077 0.080 26 0.308 0.0037 0.035 63 0.703 0.0067 0.062 27 0.113 0.0018 0.017 64 0.764 0.0075 0.077 28 1.736 0.0178 0.160 65 0.849 0.0099 0.092 29 1.804 0.0240 0.220 66 1.-326 0.0130 0.140 3o 0.448 0.0040 0.045 67 2.302 0.0171 0.150 31 0.850 0.0083 0.068 68 1.903 0,0118 0.130 32 0.349 0.0041 0.042 69 0.108 0.0048 0.044 33 0.242 0.0039 0.032 70 0.000 0.0000 0.000 34 0.143 0.0021 0.024 71 2.869 0.0480 0.500 35 0.297 0.0031 0.029 72 2.130 0.0363 0.380 36 0.113 0.0016 0.015 73 0.064 0,0026 0.026 37 2.114 0.0213 0.200 74 0.085 0.0046 0.034 -397-

Summary of Growth Data.

E. panzeri. Adult Female.

Vari- Vari- able Mean S.E. Range able Mean S.E. Range

1 0.601 0.0070 0.071 38 2.616 0.0433 0.500 2 0.637 0.0062 0.070 39 0.614 0.0097 0.091 3 0.349 0.0056 0.053 4o 1.203 0.0160 0.16o 4 0.505 0.0069 0.073 41 0.430 0.0041 0.038 5 0.363 0.0054 0.052 42 0.271 0.0041 0.045 6 0.174 0.0029 0.027 43 0.157 0.0027 0.027 7 0.175 0.0050 0.052 44 0.313 0.0032 0.035 8 0.059 0.0012 0.014 45 0.132 0.0026 0.027 9 0.065 0.0018 0.017 46 0.483 0.0071 0.068 10 0.067 0.0015 0.015 47 0.607 0.0068 0.066 11 0.147 0.0023 0.022 48 0.687 0.0099 0.110 12 1.784 0.0232 0.260 49 0.728 0.0095 0.108 13 0.971 0.0101 0.083 50 0.730 0.0099 0.108 14 0.980 0.0106 0.113 51 0.713 0.0086 0.097 15 2.774 0.0328 0.300 52 0.646 0.0095 0.110 16 1.799 0.0165 0.160 53 0.522 0.0082 0.090 17 1.226 0.0159 0.160 54 0.371 0.0071 0.083 18 2.316 0.0659 0.670 55 0.342 0.0068 0.078 19 1.366 0.0185 0.210 56 3.129 0.0337 0.360 20 0.997 0.0113 0.119 57 0.443 0.0054 0.045 21 0.475 0.0074 0.077 58 0.387 0.0043 0.048 22 0.576 0.0099 0.108 59 0.428 0.0046 0.035 23 0.284 0.0045 0.043 6o 0.723 0.0077 0.082 24 0.215 0.0032 0.030 61 0.772 0.0071 0.070 25 0.155 0.0017 0.018 62 0.761 0.0081 0.070 26 0.295 0.0045 0.047 63 0.760 0.0068 0.069 27 0.133 0.0014 0.012 64 1.531 0.0145 0.150 28 1.675 0.0225 0.250 65 0.615 0.0072 0.066 29 1.610 0.0226 '0.250 66 0.197 0.0034 0.00 3o 0.507 0.0076 0.085 67 3.374 0.0429 0.420 31 0.804 0.0090 0.084 68 2.601 0.0313 0.300 32 0.319 0.0034 0.041 69 0.000 0.0000 0.000 33 0.223 0.0036 0.038 70 0.000 0.0000 0.000 34 0.152 0.0017 0.017 71 0.804 0.0101 0.095 35 0.295 0.0036 0.037 72 0.816 0.0120 0.106 36 0.133 0.0011 0.012 73 0.127 0.0033 0.033 37 2.052 0.0275 0.330 74 0.152 0.0070 0.056