I I39 1

THE NATURAL CONTROL OF POPULATION BALANCE IN THE KNAPWEED GALL- ( YACEANA)

BYG. C. VARLEY,King's College,Newcastle upon Tyne

(With i i Figures in the Text)

CONTENTS PAGE PART I ...... 140

I. INTRODUCTION I...... 40 2. THE CENSUS I...... 14 3. THE LIFE HISTORY OF THE KNAPWEED GALL-FLY 2...... I4

PART 2. THE FACTORS WHICH AFFECT THE ADULT GALL- AND THEIR FECUNDITY 146

i. THE FECUNDITY OF THE GALL-FLIES IN THE FIELD...... I46

2. THE EXPERIMENTAL MEASUREMENT OF FECUNDITY. I. . . . 48 (a) The effect of mating on fecundity ...... 149 (b) The effect of feeding on fecundity ...... - . . . . 149 (c) The effect of combinations of temperature and humidity on fecundity . . . . 50 3. FIELD OBSERVATIONS ON THE ADULT GALL-FLIES ...... I52 (a) The population density of the gall-flies and its bearing on their fecundity. . . . . I52 (b) Experiment on the dispersal of adult gall-flies ...... I53 (c) The effect of weather on the behaviour of the gall-flies ...... I54 (d) The rate of oviposition in the field ...... 155 (e) The effect of weather on fecundity ...... 157 PART 3. THE FACTORS WHICH AFFECT THE SURVIVAL OF THE EGGS, LARVAE AND PUPAE OF THE GALL-FLY ...... I57 i. THE MORTALITY UP TO THE FORMATION OF THE GALL ...... 58 (a) The egg mortality in 1935 ...... 158 (b) The mortality of the larvae up to the formation of the gall in 1935 ...... 58 (c) The egg mortality in 1936 ...... i6o (d) The mortality of the larvae up to the formation of the gall in 1936 ...... i6o 2. THE MORTALITY AFTER THE FORMATION OF THE GALL ...... i6i (a) Winter disappearance ...... I6I (b) Mortality due to mice ...... I63 (c) Mortality due to unknown causes ...... I63 (d) Mortality due to chalcid parasites ...... I64 (I) Eurytoma curta ...... I64 (2) Eurytoma robustar...... I67 (3) Habrocytus trypetae...... 68 (4) Torymuscyanimus . 71 (5) Macroneuravesicularis .1 . . . .. I7 I (6) Tetrastichussp. B 172 (e) Mortality due to caterpillars ...... 173 (i) hohenwartiana ...... 173 (2) metzneriella ...... I174 (3) Euxanthisstraminea . .1 ...... I74 PART 4. DISCUSSION AND CONCLUSIONS ...... I74 SUMMARY ...... I82 ACKNOWLEDGEMENTS ...... I 82 REFERENCES ...... I86

This content downloaded on Fri, 8 Mar 2013 13:36:49 PM All use subject to JSTOR Terms and Conditions I40 Natural control of population balancein the knapweedgall-fly PART I The knapweed gall-fly, Urophorajaceana(Hering)* (Diptera, Trypetidae), is a member of a large and I.INTRODUCTION complex community which lives in the In this contribution to insect ecology the theory of flower-heads of the black knapweed Centaurea balance of populations, formulated by nemoralis-Jordan (Compositae). Owing to a happy Nicholson (I933) and Nicholson & Bailey (1935), is series of peculiarities in its life history, the gall-fly used for the first time in the interpretation of the provides particularly suitable material for the study results of a field survey. The conclusions are of population problems in the field. sufficiently striking to claim the attention both of The problem considered here is formulated thus: ecologists and economic zoologists, and their im- What factors control the population density of Uro- portance goes beyond that of the insect material on phora jaceana in nature, and how do they operate? which they are based. Nicholson (I933, p. 135) states that 'a controlling It is now more than twenty years since Lotka factor should act more severely against an average published his mathematical studies on the inter- individual when the density of is high, and action between predators and prey which were less severely when the density is low. In other words, applied by Gause (1934) to the oscillations in the the action of the controlling factor must be governed population densities of protozoan predators and by the density of the population controlled.' Control- prey under constant environmental conditions in ling factors, with or without the help of other factors, vitro. Nicholson & Bailey's formulation of the can therefore maintain a species in a state of balance simpler situation which arises when a sp-ecific insect at such an average population density that over a parasite and its host have synchronized generations period of years these factors kill the surplus popula- was first shown to apply under idealized laboratory tion. Where other factors permit its survival it is the conditions by the neat experiment of de Bach & controlling factors which mainly determine whether Smith (I94i), where the oscillations in the population a species shall be rare or common. density of parasite and host agreed excellently with Two groups of controlling factors can be distin- the theory over a period of eight generations. guished. The first have been termed density dependent The present work provides the first attempted factors by Smith (I935). They may be recognized by confirmation from field data of the basic assumptions the fact that at any time the severity of their action of the theory of Nicholson & Bailey. The theory is increases as the population density rises. Intra- subsequently used to interpret the interaction be- specific competition for limited food supply or tween the various factors destroying the knapweed limited space operates in this way, and the sigmoid gall-fly. The quantitative effect of each factor can population curves obtained by Pearl (I925) for be examined separately. The clarification of a com- Drosophila cultures, and by many subsequent plex situation achieved in this way may provide workers for other species, are explicable on this view the economic entomologist with a new and powerful (see Crombie, 1945). According to Nicholson's technique. And the rather paradoxical nature of theory, limitation of host population density acts in the conclusions reached may well revolutionize the the same way on the increase of parasites and pre- methods of assessing the probable value of different dators. However, the parasites and predators also projectedcontrolmeasures to be applied to insectpests. exercise a reciprocal influence on the numbers of the If the cause operating to produce balance in the species on which they feed. population density of a species is known to be a This reciprocal reaction provides a second type of parasite, workers seeking either to reduce, conserve, controlling factor, to which it is proposed to apply or increase the population density of the species can the new term delayed density dependent factor. A use Nicholson & Bailey's theory to investigate the parasite acts as a delayed density dependent factor long-term effects on the balance which may be if its fecundity or its effective rate of increase is expected from any alteration of conditions. strongly correlated with host density. Nicholson

* Until I937 the knapweed gall-fly was known in this the genus Euribia Latreille i802 as valid. This is closely country as Urophora solstitialis (L.), but it had long been bound up with the very vexed question of the validity of known that continental U. solstitialis was usually a gall-fly Meigen's I8oo names, of which Euribia is one (see of thistles. However, gall-flies bred from the continental Collin, 1946). knapweed were found to differ from the Both Seguy (I934) and Kloet & Hincks (I945) accept thistle species, and were described as new by Hering the genus Urophora of Robineau-Desvoidy I830, and (I935) under the name jaceana. Collin (1937) has found they are followed here. So it comes about that the knap- that the British specimens from knapweed are in fact weed gall-fly was called Urophora solstitialis (L.) by jaceana, and not solstitialis. Varley & Butler (I933), Euribia jaceana Hering by The generic name of the knapweed gall-fly is a point of Varley (I937a, b, I94I), and Urophora jaceana (Hering) dispute. Collin (I937) follows Hendel (I927) in accepting in this present paper !

This content downloaded on Fri, 8 Mar 2013 13:36:49 PM All use subject to JSTOR Terms and Conditions G. C. VARLEY 14I assumes that if the host density rises above the dry up, and finally their remains fall off in a lump, density of the steady state in which host and parasite leaving the ripening fruits behind. The fruits, when are in equilibrium, the percentage of hosts destroyed ripe, get squeezed out as the bracts dry and con- by the first parasite generation will not increase, but tract. Towards the end of the summer some of the remain unchanged. The number of hosts killed, and flower-heads fall to the ground. As the winter therefore the number of parasites emerging in the advances more and more flower-heads fall, until in next generation, will be proportionately greater. the following June less than a third of them remain Only after this delay of one generation will the on the dry and bleached stems. increased parasite population begin to destroy a greater proportion of hosts. Eventually after two or more generations the host density will be reduced. This fall in host density will in turn be followed by a fall in parasite density, which will allow the hosts to increase once more. These oscillations are essentially similar to those predicted independently by Lotka (I925) and Volterra (I926, I93 I) (for more complete references see Thompson, 1939). They have been observed experimentally by de Bach & Smith (I94I). In order to find how the mortality factors control ,..,#w:;:~~~~X W l, OOD LAND the population density of the knapweed gall-fly the following programme of work has been carried out. The natural rate of increase of the gall-fly has been measured under field conditions, and the factors which influence this have been studied. The mortality due to all causes has been assessed, and an HAY examination made to determine which are density dependent factors, and which delayed density de- pendent factors. The interaction between these 10 metrers factors and the other agencies which cause mortality has been considered in the light of Nicholson's theory of balance of animal populations.

2. THE CENSUS A site near Madingley, at the edge of the University Farm some 3 miles west-north-west of Cambridge, was chosen for the census work. Knapweed grew in profusion on either side of a grassy cart-track with wide uncultivated verges. The plant community was Fig. i. Sketch map of the census area, showing the distribution of the and not stable, as the ground was being colonized by knapweed (stippled area) the position of the squaremetre plots. The bushes of hawthorn stippled square (Crataegus). Poplar (Populus) metre was sampled twice in different years, and the suckers, rose (Rosa), and bramble (Rubus) grew cross-hatchedsquare metres were sampledthree timnes. thickly in places. The bushes were cut back in I932 and again in I937. In the census a total Of 92 sq.m. samples were Selected specimens of knapweed from the census taken from a striP 30 m. long to the west and 70 m. area were all identified by Dr W. B. Turrill as long to the east of the cart track (Fig. i). In Centaurea nemoralis Jordan, which was formerly February I935, i0 sq.m. were collected. More were included under C. nigra L. The shoots of the knap- taken in early June, and from the end of June until weed appear above the ground in April and May, the end of October samples were collected at weekly and the flower buds appear from amongst the intervals, and a total Of 46 sq.m. were cut in the ensheathing leaves in late June and July. They are year. In 1936 the weekly routine was begun in early then 3 mm. in diameter, and increase to between May, and continued to the beginning of October. 8 and i 2 mm. before the bracts part and the purple The sample squares were not selected at random. florets come into bloom. The growth changes in the Fig. i shows that the knapweed was patchy in its flower-heads during the early summer are shown in distribution, and random samples would frequently the diagrammatic sections in Fig. 3. There are about have contained little or no knapweed. The sample 8o (20-I00) florets in each flower-head. Within a squares were selected so that all had a fair quantity few days of coming into bloom the florets shrivel and of knapweed in them. This had two effects: it

This content downloaded on Fri, 8 Mar 2013 13:36:49 PM All use subject to JSTOR Terms and Conditions 142 Natural control of population balancein the knapweedgall-fly reduced the variation between the individual before they in their turn produced adult flies in samples, and increased the total quantity of knap- July 1936. The removal of the next generation of weed examined, and hence increased the accuracy of galls from 35 sq.m. would have no effect before the the observations. Had the object of the work been to census had finished. Fig. i shows that the site on obtain a valid mean population density per unit area which the census was made included about 500 sq.m. this would not have been admissible, but what was on which knapweed was abundant, so that the re- required was a series of comparable samples of the moval of zi and 46 sq.m. respectively in the two greatest possible homogeneity, and containing the generations amounts to only about 4 and 9 % of the greatest possible amount of material. total. This effect would have been reduced if the Those flower-heads on the standing stems could samples had been collected from a larger area. But all be collected without any difficulty. As very few another error would then have increased, since the flower-heads fell to the ground by October, the percentage of galled flower-heads varied locally census of the fresh flower-heads up to this time is within wide limits, and was considerably less only complete. During the winter a large proportion of a few hundred yards from the site of the census. the flower-heads fall to the ground, where they soon Even within the census area the number of gall-fly decay and disintegrate, and cannot be accurately larvae in each sample area varied so greatly that the counted. However, though the flower-heads may mean number per square metre had a standard error fall to pieces, many galls remain, and these may be equal to at least I 5 % of the mean. It is improbable discovered in some numbers by thorough search. that the systematic errors arising from the census The search for fallen flower-heads and galls usually method employed are as large as this, so that their occupied between I -5 and z hr. for each square effect can be neglected over the small number of metre. All the vegetation was cut down and generations studied. examined, and the ground was teased over with Examination of material. Each of the I7,492 forceps, and the decayed grass and leaves were flower-heads of the knapweed collected on the removed until the ground was bare. Few galls on 92 sq.m. samples was split open and the contents the surface could have escaped detection. Some, were examined. Special attention was paid to the however, were found partly covered in worm castings, knapweed gall-fly, Urophora jaceana, and to those and others must have been buried in this way. The other species in the community which were known ground was also tunnelled both by moles and mice, to affect its numbers (Table i, Fig. 2). All stages of and a few galls must have been buried by their spoil these were counted as accurately as possible. Certain heaps. Thus it is certain that some of these galls other species, such as U. quadrifasciata, the various escaped discovery, and the census is correspondingly Cecidomyiids, and mites, were not counted accu- incomplete. This is discussed below under the rately, since they had little or no direct effect on the heading 'winter disappearance'. It amounted to numbers of U. jaceana. 6o0% of the galls in the winter of I935-6. Treatment of census data. Since the census has Another sampling procedure was used in February been restricted to samples of the whole population, 1935, and again in September 1935 and October the data are subject to sampling errors. Throughout 1936, when the same line of ten adjacent sample this work numerical data have been treated statisti- squares was taken, without any search for fallen cally. Whenever mean values have been used the flower-heads. The variation in the amount of knap- standard errors of the means have been calculated weed in these samples was rather greater than in by the methods of Fisher (1934) or Bond (I935). those selected according to the first-mentioned sampling procedure. 3. THE LIFE HISTORY OF THE All the material collected in the census was removed to the laboratory for examination. This introduces KNAPWEED GALL-FLY an error into the work, since the density of the The life history of this species was first studied in population was thereby reduced, thus affecting some- detail by Wadsworth (1914), and the early stages what the course of events under observation. But have been redescribed by Varley (I937b). Here only the effect was probably small with respect to the the salient features of the life history need to be inaccuracies arising from sampling errors, since the mentioned. area sampled in each generation of the flies was a The adult gall-flies (Fig. 3 A, B) qmerge from the small fraction of the total area inhabited by the flies. flower-heads of the previous summer in July, and Samples were taken from three generations of galls. are to be seen in the field for about a month. The 2I sq.m. were sampled before the emergence of the liberation of marked flies showed the mean life-span adult gall-flies in I935, butinthefirstthirteenof these of a female fly to be about a week, but both in the samples no fallen flower-heads were taken. The next field and in the laboratory certain individuals lived generation of galls was removed from 46 sq.m. much longer than this.

This content downloaded on Fri, 8 Mar 2013 13:36:49 PM All use subject to JSTOR Terms and Conditions G. C. VARLEY I43 Table I. List of the most important forming Oviposition usually starts on the third day after the community in the flower-heads of the black emergence and continues until the fly dies. When a knapweed (Centaurea nemoralis) in the census area female fly finds an unopened flower-head of knap- at Madingley weed which is between 3 and 5 mm. in diameter it The order of magnitude of the larval population walks on to it, and turns around a few times. density of the species is indicated by the symbols: Eventually the fly pushes its ovipositor down at the A=abundant, more than ioo per sq.m. side of the flower-head and inserts the tip between C=common, IO-IOO per sq.m. the bracts. Often the ovipositor is removed after S=scarce, i-io per sq.m. a few seconds and replaced in a slightly different R= rare, less than i per sq.m. position, but finally the fly remains motionless for Synonyms are put in brackets about 2 min., and during this time a few eggs are laid. The slender end-piece of the ovipositor is Plant-feeding species driven through the soft tissue at the base of the Diptera, Trypetidae (gall-flies) flower-head and turns upwards so that the eggs are Urophora jaceana (Hering) A (U. solstitialis Auctt., nec Lin.) laid in the space between the florets and the over- U. quadrifasciata (Meigen) C lapping bracts. The track of the ovipositor is faintly Chaetorellia jaceae (Rob. Desv.) S indicated in Fig. 3 C. Chaetostomella cylindrica (Rob. Desv.) R The eggs are easily seen if a flower-head is split (C. onotrophes (L.)) open. They are usually in groups of two or more. Diptera, Cecidomyiidae (gall-midges) A few days after they are laid their discovery is aided Dasyneura miki (Kieffer) A by the shrivelling, or retardation in growth, of the Clinodiplosis cilicrus (Kieffer) C florets in their immediate neighbourhood (Fig. 3 D). , Tinaeoidea The first larval moult takes place in the egg and the Metzneria (Parasia) metzneriella (Stainton) S eggs hatch as second instar larvae about iz days Lepidoptera, Tortricoidea after being laid. The time of hatching depends partly Euxanthis straminea (Haworth) R on the temperature, and there may be a difference of Eucosma hohenwartiana (Schiff.) C 2 or 3 days between the hatching of the first and the (E. scopoliana (Haworth)) last egg of a single batch. Parasitic species The second instar larva when first hatched creeps Hymenoptera, Chalcidoidea over the florets and eats its way into one of them, Eurytoma curta Walker C leaving a small hole with a brown edge, and slowly E. robusta* Mayr S burrows down the axis of the floret to the ovary. Habrocytus trypetae (Thoms.) C Then almost at once the plant tissue surrounding Torymus cyanimus* Boh. S the ovule swells and elongates, becoming eventually Macroneura (Eupelmella) vesicularis (Retz.) S a pear-shaped fleshy mass about 7 by 3 mm., in Tetrastichus Nees brevicornist R which the larva lies (Fig. 3 E). If two or more Tetrastichus sp. B S dairat (Walker) R adjacent florets contain larvae they fuse together to form a multilocular gall with each larva in a Hymenoptera, Braconidae separate Neochelonella (Chelonus) sulcata (Jurine) S cell (Fig. 3 F). In time the outer wall of the gall cell Macrocentrus nidulator (Nees) S hardens and becomes woody, while the inner tissues Hymenoptera, Ichneumonidae remain fleshy and are eaten by the larva. The passage Omorga ensator (Grav.) R by which the larva entered the ovary remains open, Ephialtes buolianae Hartig R so that the cell is finally flask-shaped with a rather (Scambus depositor var. Roman) wide opening at the top (Fig. 3 G). The details of Glypta longicauda (Hartig) S gall formation have not been studied, since they have (G. nigrotrochanterata Strobl.) S no bearing on the problem in hand. G. vulnerator Grav. S The third instar larva appears some 3 weeks after Predatory species oviposition, and a fortnight after this the hind-end Diptera, Cecidomyiidae of the body becomes pigmented and sclerotized, and Lestodiplosis miki Barnes C forms the perispiracular plate. As the larva feeds head downwards, this hard black plate forms a plug * Not listed by Kloet & Hincks (I945). Recorded which usually fits tightly into the neck of the flask- from Urophora cardui in Britain-see Blair (I ). 93 shaped gall-cell. This is important in connexion with t Not listed by Kloet & Hincks (1945): apparently this is the first British Record. the attacks of parasites, described later. The larva is t Put in the genus Tetrastichus by Kloet & Hincks; fully grown soon afterwards, and remains inactive but according to Mr J. F. Perkins daira Walker is an in its cell during the winter. Aprostocetus. Pupation begins in May when the larva reverses its

This content downloaded on Fri, 8 Mar 2013 13:36:49 PM All use subject to JSTOR Terms and Conditions I44 Natural control of population balancein the knapweedgall-fly position in the gall-cell so that its head faces the (i) All but the brief adult life is spent within a exit, and the larval cuticle becomes a hard brown single flower-head of the knapweed. puparium. Within this skin there develops first a (2) The young larvae cause the formation of hard fourth instar larva, or prepupa, and then the true durable galls. Each gall-cell is isolated from the pupa, as described for the related genus Rhagoletis by others in the same flower-head. From examination Snodgrass (I924). From the puparium the adult fly of the galls, the number of larvae which caused thelr emerges in about a month, in the early part of July. formation can be found. The number surviving can The following important features in this life be counted, and the cause of death of the others can history have made the Urophorajaceana particularly usually be inferred from the contents of the gall- suitable for detailed ecological study: cells.

Glypta Omorga Ephialtes spp. ensator Macrocentrus spp. nidulator

Apanteles Neochelontella sicarius sulcata

Euxanthis Eucosma Metzneria straminea hoheniartiana metzneriella

Mice and UJrophora -Ai +. Winter jalceana } . disappearance

Lestodiplosis Aprostocetus miki daira

Torymus -Tetrastichus cyanimus _ urytoma Eurytom brevicornis / ~~~curta l robusta \ Tetrastichus

Macroneura 1/ r9--- s , - s sp.B. vesicularis Habrocytaxs-X ttry petae

Fig. 2. Food chain of the species which affect the numbers of the knapweed gall-fly, Urophorajaceana.

Explanation of Fig. 3

Fig. 3. The knapweed gall-fly and its life history. A. Knapweed gall-fly, male (x I2). B. Knapweed gall-fly, female ( x I2). C. Knapweed flower-head,5 mm. in diameter,with small florets. It contains four gall-fly eggs. Note the faint track of the gall-fly's ovipositor (x 6). D. Knapweed flower-head,6 mm. in diameter,showing florets half grown. Four eggs have alreadyhatched, and two larvae are shown inside separateflorets. Two eggs have failed to hatch. Note that the floretsnear the trackof the fly's ovipositorare stunted in growth. E. Knapweed flower-head8 mm. in diameter. Most of the florets are alnost ready to bloom. A gall, surmountedby the remains of the pappus of the fruit from which it has been formed, contains a second instar larva. The dark woody layer of the gall is beginning to form. F. Knapweedflower-head in bloom. The gall contains third-instarlarvae which have already consumed a large part of the fleshy gall-tissue. G. Knapweed flower-head after flowering, in September. The fruits have dropped out, leaving only the paraphyses. The gall is hard and woody, and contains (left) a fully fed gall-fly larva, (centre) a larvaof the chalcidparasite Torymuscyanimus. On the remainsof the host can be seen three egg shells of Torymus. On the right is a brown gall-fly pupariumcontaining a larvaof the chalcid parasite Eurytomacurta. To the extreme left is a slightly swollen fruit containing a larva of the small gall-fly Urophoraquadrifasciata.

This content downloaded on Fri, 8 Mar 2013 13:36:49 PM All use subject to JSTOR Terms and Conditions min~~~~~~~~~

Fig. 3.-

This content downloaded on Fri, 8 Mar 2013 13:36:49 PM All use subject to JSTOR Terms and Conditions I46 Natural control of population balancein the knapweedgall-fly

(3) There is only one generation in the year. July I935. The completeness and accuracy of the Although the successive stages of development of census data on which this estimate is based are the gall-fly overlap in time during the summer, discussed below under the heading 'winter dis- equivalent stages of successive generations do not appearance'. overlap. Henc a complete census can be made for The proportion of females was estimated from the each generation. number of male and female gall-flies which emerged in the emergence cages in the laboratory. Out of PART 2 662 gall-flies, 28I were females, or 42-4 % ? 1-9. Hence the number of female which THE FACTORS WHICH AFFECT THE gall-flies emerged in 1935 is estimated to have been ADULT GALL-FLIES AND THEIR 6-9 x 0o424 = 29 ? o07/sq.m. FECUNDITY (b) The number of eggs laid per sq.m. cannot be Provided there is no migration, the population counted directly, as all the eggs are not present at density of adult gall-flies in a particular area will the same time. The oviposition period lasts about change from one generation to the next by a factor 4 weeks, which is much longer than the time taken to equal to the fecundity multiplied by the proportion of hatch (about I2 days). Empty eggshells cannot be females, and by the fraction of the eggs which reach found, and the mortality in the very young larvae the adult stage. These three quantities must be would make the sum of eggs and larvae found smaller estimated, and an assessment made of the factors than the total number of eggs laid. Two indirect which influence them. methods have been employed to estimate the number Fecundity has been studied from three aspects. of eggs laid per sq.m. First, the fecundity of the gall-flies in the field has Method i. It is easy to make a direct count of the been estimated from the census data. Secondly, it late second instar larvae by counting all the gall- has been studied in the laboratory, and its depen- cells after the galls have been fully formed. Then, dence on nutrition and on certain climatic factors knowing the proportion of eggs laid which success- measured. Thirdly, detailed field observations have fully form galls, the total number of eggs laid per indicated how oviposition is affected by weather sq.m. is readily calculated. conditions. All the larvae had formed galls by I3 August, and the number of gall-cells subsequently found in i. THE FECUNDITY OF THE GALL-FLIES 22 sq.m. was 3247. The mean number of gall-cells IN THE FIELD per sq.m. in late summer 1935 was I47-6?2I-5 (Appendix, Table A, col. 3). No direct measurement of the fecundity was practi- On p. the mortality which occurred up to the cable in the field, but the census provided an indirect I58 formation of the galls is given as o0289 ? 0-022. method of estimation. This is based on the relation- Hence the survival up to this time was o07I I, and the ship (which has also been used by Sachtleben (I927) number of eggs laid per sq.m. in I935 in his detailed study of the Panolis flammea) Gall-cells per sq.m. Mean fecundity No. of eggs laid per unit area Proportion of eggs surviving = ? No. of females emerged per unit area' I476/0o7I =208 3I. Method 2. A second method of estimating the which is accurate if migration can be neglected. It total number of eggs laid per sq.m. is to find the is shown later that the gall-flies move so little that mean number of flower-heads per sq.m. which the accuracy of the relationship is not likely to be contain seriously affected. eggs, and multiply this figure by the mean number of eggs in each flower-head. Table 6, col. 2 shows that 447 eggs were found in I48 flower-heads, The fecundity of the gall-flies in I 93 5 which gives a mean value of 3-02 ? 0-I5 for the (a) The number of female gall-flies which emerged number of eggs laid per flower-head. per sq.m. was estimated by counting the number of The total number of flower-heads containing eggs live or empty puparia in the year-old galls in and cannot be counted directly, since some eggs have after July. There was considerable pupal mortality hatched before the last have been laid. However, due to parasites in June, but this had virtually ceased the total number of flower-heads containing eggs or in July, and in io sq.m. collected in July and August larvae can be counted, and approximates to the total totals of 9, 6, 6, o, 6, 5, 3, 17, 13 and 4 puparia were number of flower-heads in which eggs were laid. found, I2 of which were about to produce gall-flies, Although a proportion of the eggs fail to develop, while the other 57 had already done so. This gives a there is only a small chance that none of a group of mean of 6-9 ? i -6 gall-flies emerged per sq.m. in eggs will survive. A correction can be applied for

This content downloaded on Fri, 8 Mar 2013 13:36:49 PM All use subject to JSTOR Terms and Conditions G. C. VARLEY I47 any random egg mortality, and allowance made for 2 September, was 34/63 = o054. The first of these additional mortality of whole egg batches. figures might be expected to be too low, as there The mean number of flower-heads which con- were many puparia in these samples yet to emerge. tained eggs or larvae of the gall-fly was 6o03 ? 7-4 The second figure might be expected to be too high, (Appendix, Table A, col. 2). Of this total a mean of as it might include some emergence after July. The 2-6 contained eggs only, and 57.7 contained larvae combined result is likely to be more accurate than already in galls. A figure of o 29 may be assumed for either alone, and gives a fraction of emergence the random mortality which had occurred before 0-563 ? 0o049. gall-formation (p. I58), and it is seen in Table 6 The mean number of larvae-plus-puparia in the that for every 886 flower-heads with eggs which form 36 sq.m. was 3-6I ? o 6o. Multiplying by the frac- at least one gall-cell there would be expected to be tion of emergence, o0563, we estimate the number of 82-93 flower-heads in which none of the eggs sur- flies emerging per sq.m. to be 2 03 ? O-38 per sq.m. vived to form galls. Hence to the figure of 6o03 Since the proportion of females in the population flower-heads with eggs and larvae must be added a was o0424, the number of female flies which correction of (57 7 x 82 93)/8865 54, making a total emerged per sq.m. is estimated at of 8 flower-heads per sq.m. which contained eggs 65 2 O3 x o-424=o-86 ? O0I7. of the gall-fly. Since the mean number of eggs laid per flower-headwas 3'02 ? 0-I5, the mean number (b) The number of eggs laid per sq.m. in 1936. of eggs laid per sq.m. in I935 Method i. The number of gall-cells found per sq.m. in samples nos. 73-92 was 28 ? 5 (Appendix, Table E==65.8 x 3-02 =199 ? 23. B, col. 3). The total mortality up to gall-formation This second estimate agrees well with that of 208 was 37-5 % ? 3-4 (see p. I6I) so that the proportion two obtained by the first method. The mean of the surviving to form galls was o-625. Hence the estimates is 203 ? 27. The standard error has been number of eggs laid per sq.m. in I936 calculated on the assumption that, since the esti- mates were derived from the same data, the correla- Gall-cells per sq.m. z28 =448 ? 8-5. tion between them is unity. Since the number of Proportion of eggs surviving o-625 female gall-flies which emerged per sq.m. in I935 Method 2. The mean number of eggs laid per was 2-9 ? 0o7, the fecundity in I935 flower-head was 267/88 = 3 04 ? o I8 (Table 7). The Eggs laid per sq.m. 203 mean number of flower-heads containing eggs and f= - 70+ ?In9 Female flies emerged per sq.m. 2-9 larvae in 29 sq.m. was 12-3 ? I 6 flower-heads per sq.m. (Appendix, Table B, col. 2). However, only The fecundity of the gall-flies in I936. I-4 per sq.m. contained eggs, and io-9 per sq.m. The same methods have been used to estimate the already contained larvae in galls. Allowance must fecundity in I936. be made for the mortality of eggs and larvae before (a) The numberof gall-flies which emergedper sq.m. gall-formation. It is shown on p. i6i that this in 1936 can be estimated from the data in Table 2. mortality appeared to comprise 7-7 % infertility of Emergence began late in June as it did in 1935, but whole egg batches, followed by a 32-3 % random cold and wet weather delayed the emergence of some mortality of the remainder. Correction must first be gall-flies until August or September. Further, the applied for the random mortality. Table 7 shows rain storms of July caused flooding, which resulted that with a mortality of 0o32, for every 289 galled in about 46 % mortality in the puparia. The drowned flower-heads there would be 35-5 flower-heads in puparia were not at first easy to distinguish from live which none of the eggs laid eventually produced ones. Some doubtful puparia were isolated, and one galls. Hence the IO-9flower-heads containing larvae male gall-fly emerged as late as I9 August; some represent IO9 (I +35-5/289) =I224 flower-heads apparently living puparia were found in September. which had previously contained fertile egg batches. However, in the census no eggs were found after If 77 % of the egg batches were infertile, this the middle of August, and the last eggs must have number of flower-heads must be divided by the been laid by female flies which emerged towards. the survival, o 923, to give a total of I2 -24/0 923 = I3 3 end of July. The few flies which emerged after this flower-heads per sq.m. with eggs. Adding the I14 apparently laid no eggs, and so may be neglected. flower-heads found containing eggs, this gives a It is not easy to see from Table 2 what proportion total of I4-7 flower-heads per sq.m. in which eggs of the larvae and puparia had emerged by the end of were laid. July. The fraction which had emerged in sq.m. Multiplying this by the mean number of eggs laid nos. 57-66, collected between II and 28 July, was per flower-head, we estimate the number of eggs 24/40=o-6o. The fraction which had emerged in laid per sq.m. sq.m. nos. 67-82, collected between 3 August and E= I4'7 X 304=44.7 ? 5 6. J. Anim. Ecol. I6 10

This content downloaded on Fri, 8 Mar 2013 13:36:49 PM All use subject to JSTOR Terms and Conditions I48 Natural controlof population balancein the knapweedgall-fly The estimates by the two methods agree, and their removed and counted in a .drop of water under a mean is 44'8 ? 7- i. Hence the fecundity binocular microscope. Eggs laid per sq.m. The first experiment was designed to discover the ' sizes of flower-heads which were acceptable to the Female flies emerged per sq.m. gall-flies for oviposition. Six pairs of gall-flies were = 44-8/o086 = 52 ? 9. isolated in hurricane-lamp glasses over flower-pots This estimate of the fecundity in I936 is rather filled with damp sand. Flower-heads of the knap- lower than that for 1935, but the difference is not weed were provided with their stalks in glass tubes significant. containing water. Each pair of gall-flies was given

Table 2. The numbersof gall-fly (Urophora jaceana) larvae, puparia, dead puparia, and puparia from which-flies had emergedin the square metre samples nos. 47-82, in the summerof 1936 Date Sq.m. Live Live Flies Date Sq.m. Live Dead Flies (1936) nos. larvae pupae emerged Total (1936) nos. pupae pupae emerged Total I2 May 47 2 . . 2 28 July 65 . . . 0 I9 May 48 . . . 0 z8July 66 . . . o 26 May 49 I . . I 3 Aug. 67 I . I 2 2 June 50 2 3 * 5 3 Aug. 68 . I I 9 June 51 . . . 0 4 Aug. 69 2 . . 2 I6 June 52 3 3 I I Aug. 70 5 * 5 10 23 June 53 2 . 2 IZAug. 7I . . . 0 30 June 54 . 4 I 5 I2 Aug. 72 3 * 6 9 7 JUly 55 . I 3 4 I7 Aug. 73 * * I I 7 July 56 . 2 3 5 I8 Aug. 74 I . I 2 II July 57 . 3 3 I 8 Aug. 75 . * * 0 13 JUIy 58 . . 2 2 24 Aug. 76 . . I 14 July 59 . 3 * 3 25 Aug. 77 . . . 0 I4 JuIy 6o I . 2 3 26 Aug. 78 . I 2 3 2 I July 6i 1 3 4 8 3 I Aug. 79 . 3 I 4 2 I July 62 . 3 2 5 3 I Aug. 8o . 2 I 3 ZI July 63 I I0 I I 2 Sept. 8i . 4 10 14 28 July 64 * I 4 5 z Sept. 82 2 5 4 II Totals i8 7 29 3I 67 I8 14 15 34 63 No. per sq.m. 0o4 I-6 1-7 3 7 o-8 o 8 I-9 3 5 Grand total Ex= I 30, EX2 922, X =3-6I, 36X12 470, E;(x-W _X= 452, estimated standard error S / o6o. 36 x 35 Emergence in samples 57-82 = 58/1io3 = 0-563.

Estimate of standard error s=-| 4 = 0 049. Combining the total per sq.m. (3-61 + o 6o) and the fraction 1wi03 h I103 which emerged (0.563 ? 0-049) the emergence is estimated at 2-03 ? 0-38.

either three or four flower-heads of different sizes. 2. THE EXPERIMENTAL MEASUREMENT OF The experiment was carried out in a cool green- FECUNDITY house, in which temperature and humidity were un- The gall-flies used in these experiments were reared controlled. The flies were fed on sugar solution. in an outdoor insectary from larvae collected during After 2 or 3 days the flower-heads were removed, the winter. Preliminary work showed that mature and measured, and the eggs in each were counted. gall-flies would oviposit in flower-heads in captivity, In every case the majority of the eggs were laid even if they were confined in very small glass jars. in the smallest available flower-head, whether this The eggs, laid in the space between the developing was 3 mm. in diameter or as large as 5 mm. Out of florets and the overlapping bracts, can easily be a total of 663 eggs laid in this experiment, only 66

This content downloaded on Fri, 8 Mar 2013 13:36:49 PM All use subject to JSTOR Terms and Conditions G. C. VARLEY 149 were laid in flower-heads whose diameter exceeded Female gall-flies, which were less than i day old, 5 mm. Flower-heads smaller than 3 mm. are not were taken from the emergence cage and placed in suitable, as they are still enveloped in the young separate glass bottles as in Fig. 4A, each with a male. leaves of the flowering shoot. This preference for The bottom of each bottle was covered with damp oviposition in flower-heads of 3-4 mm. diameter cotton-wool, and a small glass tube supported on a agrees with the fact that the stage of development of card held a flower-head of about 4 mm. diameter the gall-fly larvae found in the census was closely in a little water. Bottles containing flies were set up related to the stage of growth of the flower-head in every few days as the flies became available, and which they were feeding. each day's emergence was divided into two series. Field observations show that after a gall-fly has In the one the cotton-wool was moistened with a laid eggs in a flower-head it walks away and seeks dilute solution of cane-sugar, and in the other, which another. The second experiment was planned to see served as a control, only tap water was given. whether the gall-flies laid fewer eggs if provided with Temperature and humidity were not controlled. The only a limited number of flower-heads in which to eggs were counted every day, and fresh flower-heads lay. Five pairs of gall-flies were isolated with I, 2, 4, provided. The female gall-flies were dissected when 8 and I6 small flower-heads respectively. Other they died, and the eggs remaining in the ovaries were conditions were as in the first experiment. The counted. experiment continued until the death of all the female Altogether 32 female gall-flies were used in the flies, the flower-heads being changed once during experiment, but the first eighteen gave unsatis- the period of life. factory results, as many of them soon died from The first gall-fly, provided with a single flower- a fungus disease. The last fourteen gave better head, laid 2i6 eggs. The second laid 277, the third results. The unfed flies laid on the average 95 eggs 29 and the fourth 78 eggs, but only lived for 3 days, (maximum 149) and lived for about 8 days (maximum while the fifth laid 125 eggs. This demonstrates io days) while the flies fed on sugar solution laid on clearly enough that the gall-flies do not restrict their the average 22o eggs (maximum 3I6) and lived on output of eggs when there is only a single flower- the average 23 days (maximum 3 i days). These head available and in the experiments which follow, differences are strongly significant, and it is con- each female gall-fly was providedwith a single flower- cluded that both the fecundity and the longevity are head of suitable size. doubled if the gall-flies are provided with sugar. Feeding did not alter the length of time before the (a) The effect of mating on fecundity laying of the first eggs, which was about 4 days, nor did it affect the number of eggs which remained in Six unmated female gall-flies were isolated in the ovaries at death, which was about glass bottles as in Fig. 4A, and the flower-head was 50. The question arises whether the gall-flies feed in removed every 3 or 4 days and examined for eggs, the field. The mouthparts of the gall-fly are similar by which time most of them had started develop- in a general way to those of the blow-fly (Calliphora) ment. As controls, six similar females were kept (Graham-Smith, I9II) and might enable the flies to with males, and their performance was compared. feed on liquid food, perhaps including particles in The unmated females laid nearly as many eggs on suspension. Experiments on thirsty gall-flies showed the average as the mated females (22 as against that when presented with a freshly made suspension 29), but none of the eggs of the unmated gall-flies of yeast in cane-sugar solution, the crop contents developed normally. These results are in agree- were devoid of yeast cells. However, Boyce (1934, ment with those of Glaser (I923) for the flies p. 5Io) found that the flies of the related genus Musca and Stomoxys. The yolks of the eggs laid by Rhagoletis ingested solid matter, such as diatomaceous the unmated gall-flies remained opaque, shrank earth, if it was mixed with sugar solution and away from the egg shell, and became shortened, or sprayed on leaves. constricted in various irregular ways. Eggs of The only types of food likely to be available to the exactly similar appearance were found in the field, gall-flies in the field are the nectar of flowers, and and it is probable that much of the egg mortality in the honey-dew of aphids. The flowers in bloom in the field was due to lack of fertilization. the census area when the gall-flies are present were wild rose (Rosa), various small leguminous species (b) The effect of feeding on fecundity and three kinds of Umbelliferae (wild carrot (Daucus The ovaries of newly emerged female gall-flies are carota L.), wild parsnip (Pastinaca sativa L.) and very small, but in a few days they become greatly hogweed (Heracleum sphondylium L.)) of which only enlarged and full of ripe eggs, and the flies start to the wild carrot was in bloom in the early part of the lay eggs. The effect of feeding gall-flies on cane- period. In I936 all the Umbellifers were in bloom sugar was tested in the following way. when the gall-flies were common. Of these flowers, IQ-2

This content downloaded on Fri, 8 Mar 2013 13:36:49 PM All use subject to JSTOR Terms and Conditions I50 Natural control of population balancein the knapweedgall-fly only the Umbellifers were frequented by Diptera. with thermostats. In I935 the temperature control The flower-tables were examined repeatedly, but was maintained by a toluene-mercury gas thermo- although there were many gall-flies close by on the stat, and temperatures between 2o and 320 C. were leaves and flower-heads of the knapweed, none was studied. In 1936 a cooling system was installed, in seen feeding. However, two males of Urophora which an electric thermostat operated a relay to a quadrifasciata were seen on wild carrot in I935. pump which circulated the water through an ice- Aphids were scarce in the census area in 1935 and box. The temperature could be kept constant to I936, and the gall-flies did not seem to be attracted within a quarter of a degree C., and temperatures to them in any way. In 1938 aphids were abundant down to I5? were used. on the knapweed, and the question of feeding was Humidity was controlled by a flow method. The examined directly by the analysis of crop contents. bottles were closed by well-fitting rubber stoppers, Fourteen wild gall-flies were dissected soon after and connected together by T-pieces so that a flow of capture, and in about half the specimens, both male conditioned air could be sent through them in and female, the crop was distended with a clear parallel (Fig. 4B). The rate of flow in each bottle yellowish fluid. The crop was placed on a waxed was adjustable by a screw clip and was observed in slide and punctured. An equal amount of Fehling's a separate bubbling tube which contained oil of low reagent was added, and some of the mixture was vapour pressure. The rate of flow was such that each sucked into a U-shaped capillary tube, and immersed bottle received its own volume of air every z min. in boiling water. In all cases the test clearly demon- The source of air was a pump worked by tap water strated the presence of reducing sugars in the food. (Cannon & Grove, 1927) and this proved very It seems likely that honey-dew was the source of the reliable and easy to adjust. The air was passed sugar. through a series of three jars filled with broken glass To conclude, since both the population density and strong caustic potash, in which the humidity of of aphids and the time of the flowering of the the air was determined. In I935 these jars were kept Umbellifers are so variable, the food supply is in a separate water-bath (Fig. 4 C) which made inconstant. Fluctuations in such food supply possible the adjustment of humidity by alteration of may alter the fecundity of the gall-flies in the the relative temperature of the two water-baths. field. In I936 the whole apparatus was put in the same water-bath (Fig. 4D). The air delivered by these jars could be tapped off at a T-piece and its (c) The effect of combinationsof temperature humidity could be measured by a dew-point hygro- and humidity on fecundity meter. The humidity of the air after passage through The precise effect of weather on an insect is the experimental bottles could be measured in the difficult to determine, since so many variables have same way. The difference in relative humidity never to be considered. However, some information can exceeded 5 % and was usually about 3 %. be obtained on this effect by comparing the behaviour Light was not controlled, and its intensity was of the gall-flies observed in the field under known much lower than in the room as a whole owing to the weather conditions with the results of laboratory submersion of the bottles in the water tank. How- experiments conducted under conditions of constant ever, bright light is not necessary to ovipositing flies, temperature and humidity. since they were seen ovipositing by weak artificial The laboratory experiments were designed to test light. the effect of constant temperature and humidity on Newly emerged gall-flies were kept for 3 or 4 days the fecundity of gall-flies which were already in milk bottles before the experiment; they were fed sexually mature, and ready to lay eggs if given the on dilute cane-sugar solution. This gave time for right conditions. Diagrams of the apparatus used are maturation and mating. The gall-flies were then put given in Fig. 4. The experimental chambers (Fig. in the experimental bottles, each with a suitable 4A, e) were a series of bottles of the same size as flower-head, and the experiment continued until used in other experiments on oviposition, and one their death. The flower-heads were examined daily, male and one female gall-fly were put in each. A and the eggs laid were counted, and fresh flower- suitable flower-head of knapweed was put with its heads substituted. This occupied about 20 min. each stalk in a narrow glass tube containing a little water. day, during which time control of the conditions Evaporation was reduced as much as possible by ceased. removing all the small leaves from the knapweed In 1935 only four experimental bottles were used, stalk, and by plugging the open end of the glass tube but this was increased to eight in I936. The gall- with cotton-wool. flies were available for study for only about a month The temperature of the bottles was kept constant in each year, and the effect of only fourteen different by sinking them in a large tank of water provided conditions was studied, using in all I 12 female flies.

This content downloaded on Fri, 8 Mar 2013 13:36:49 PM All use subject to JSTOR Terms and Conditions G. C. VARLEY 151 This small number, coupled with the great vari- The values for the fecundity in this experiment ability in the performance of individual gall-flies, are all very low, and seem to differ by a factor of three makes the statistical error in the results large. It was from the results in similar bottles in which the impossible to perform the experiments on a larger temperature and humidity were not controlled. The scale without much more expensive apparatus>, difference here might be due to low light intensity, or owing to the limited time during which gall-flies and to some toxic emanation from the rubber (Mellanby flower-heads were available, and also because of the & Buxton, I935), or simply to the constancy of the pressure of census work which had to be done at the conditions (see Uvarov, I931). same period. The fall of the fecundity at high temperatures is Although the accuracy of this physiological study partly due to a reduced length of life. Fig. 6 indi-

\ e ,t,t c ; ge

e c ~ ~ ~ ~ ~~bA

C D Fig. 4. Apparatus for controlling temperature and humidity. A. Experimental bottle e containing a pair of flies and a flower-head of the knapweed, connected to the bubbling tube b which contains oil. B. Diagram showing the method of connecting eight experimental bottles in parallel. C. Apparatus as used in 1935; air was led through the potash bottles p in the first water-bath, passed through a metal coil in the second water-bath, and then past a tap t into the experimental bottles. D. Apparatus as used in 1936. is low, the results given in Table 3 and Fig. 5 show cates the effect of temperature on the length of life. sufficient consistency for the drawing of certain The data for the lower temperatures come from important conclusions. The highest fecundity (77 Table 3. The effects of subjecting the gall-flies to eggs per female) was recorded at 30? C. and 8z % higher temperatures for periods of i hr., or 24 hr., R.H., and the graph indicates that there is an were observed separately in a much simpler appa- optimal region for oviposition under constant condi- ratus. The effect of a temperature of 440 was severe. tions which lies between 22 and 32'. The effect of The flies appeared to be dead after i hr., but at room humidity was too small to be statistically significant, temperature they all revived, although recovery was but the results suggest that at any one temperature often incomplete. At 460 all were dead within an the fecundity is highest at a humidity of over 6o % hour. At 370 the flies were active and apparently saturation. normal in their behaviour for some hours, but

This content downloaded on Fri, 8 Mar 2013 13:36:49 PM All use subject to JSTOR Terms and Conditions I52 Natural controlof population balancein the knapweedgall-fly invariably they died within a day. Similar results above and below which various factors reduce the were observed both at IOO % R.H. and under drier number of eggs laid. There is evidence that for at conditions. The fecundity experiment showed that least part of the season the population density of the at 350 the mean length of life was about a day, but gall-flies was below this optimum. the gall-flies laid a number of eggs in that time. The The population density was estimated by walking highest temperature recorded in the field was 27'. very slowly along the side of the cart track, and counting all the gall-flies seen. The area covered by this search in I936 was about ioo sq.m., but only 3. FIELD OBSERVATIONS ON THE ADULT about 50 of these contained much knapweed. As the GALL-FLIES flies spend almost the whole of their time on the The gall-flies were abundant in the field only in knapweed and seldom stay on any other plant for July, and at this time of the year there was much more than a few seconds at a time, it is best to census work and laboratory experimentation to be express the density of gall-flies in terms of the total done. However, in 1935 i day a week was spent in number of flies observed divided by the area con- the field on the site of the census, and 3 days a week taining knapweed. The number of gall-flies seen in

Table 3. The effect of constant experimental conditions of temperatureand humidity on the fecundity and survival of adult gall-flies (Urophora jaceana) Relative No. of No. of eggs laid by Mean no. of Mean no.* Temp. humidity female each female fly Total no. eggs and of days (0 C.) (%) flies (zeros omitted) of eggs standarderror alive 20-4 88 4 2I, 96 II7 29 33 3 22-4 43 4 4, I33 I37 34? 33 2.5 2413 34 4 27, 98 I25 3I223 32 24 3 85 8t ii6,I 122, 204 442 55? 21 3 5 27'3 40 4 9 94?i 27 3 40 8t 46, 120, 137, 203 56 43?6 27 3 I00 4 43, 47, I20 2I0 52 25 I 31 4 30 4 77, 90 I67 42 26 2 15 40 12 4, 6, 87 97 8 7 6 15 82 I0 I6,1 I7, 21, 40, 110 204 20+ II 9 27 53 8 31, I24, 128 283 35 ? 20 3 30 82 I5 3, 58, 63, 75, 83, 1148 77?2I I.7 Io8, I58, I97, I98, 205 35 43 8 97 97 12 ? 12 I*6 35 82 I4 13, 47, 87, 95, II6 358 26+ ii I 35 100 5 28, 42 70 '4? 9 I.5 * This does not include the 3-4 days at room temperaturebefore the experimentbegan. t In these two instanceseach jar containedtwo female flies, and the total numberof eggs laid by the two is recorded. in 1936. Part of this time was spent in collecting I936 was greatest on 7 July, when twenty males and material for the census, and the rest in observing the six females were seen. On 27 July, ten males and adult gall-flies. four females were seen. Later observations showed A field study of the activities of the gall-flies in two males on 4 August, a male and a female on relation to the varying conditions of the environ- IO August, and a single male on I5 August. The ment cannot give a quantitative estimate of the maximum density recorded was therefore one fly in effect of environmental conditions on the fecundity. each 2 sq.m., and the density fell to one fly in But observations of activity, when considered in the 25 sq.m. by the beginning of August. light of the results of the laboratory experiments and This low population density apparently reduces the census data, lead to certain important conclusions. the chance of finding mates. In 1936 the gall-flies were seen in normal coitus some twenty times, while on four occasions interspecific pairing was seen. On (a) The population density of the gall-flies and 7 July a male of the gall-fly Urophora jaceana was its bearing on their fecundity seen paired with a female U. stylata. This is an Chapman (i928) and MacLagan (I 932) have shown easily distinguished species which forms galls in the that the population density of an insect may affect its flower-heads of the spear thistle, vulgare fecundity; there is an optimum population density, (Savi) Ten. Males of Urophora jaceana were seen

This content downloaded on Fri, 8 Mar 2013 13:36:49 PM All use subject to JSTOR Terms and Conditions G. C. VARLEY '53 on 30 June and io August paired with females of the movements within a small area. On 7 July I936 smaller and differently marked species U. quadri- seven males remained for Io hr. within a foot of fasciata, which is common in knapweed, but forms where they were first seen; they were observed at no gall. On I0 August a female of U. jaceana was least eight times during this period. The males are found paired with a male of U. quadrifasciata. pugnacious, and when two males meet they buffet Only one of these cases was noted in July, and on one another with head and vibrating wings, or even that day the males of U. jaceana greatly outnumbered grapple together, until one eventually retreats. the females of the species. On the other 3 days the Boyce (1934, p. 45O) describes similar behaviour in population density of the gall-flies U. jaceana was the related genus Rhagoletis. It is almost as if each very low. No interspecific mating was seen in I935 male maintains a territory, as do the males of certain when the maximum population density was more birds! The female gall-flies, on the other hand, than twice as high as in I936, and the emergence was partly by short flights, but mostly by walking, move more concentrated into the month of July. Inter- distances of a few feet an hour in their most active specific pairing is therefore regarded as a sign that periods. If a female meets a male during this mates of the same species are difficult to find. wandering, courtship and mating may follow. But

35"C 1 ( 14 40?PC-

42 Over 4-0 e _ 300C fe\( per 00 _gatt'- fY 00C 0

0 2012icrc p . ft .

0 0 20?C 0 2 4 6 8 10 MEAN SURVIVALIN DAYs S5C 0 0 Fig. 6. The effect of temperature on the longevity of adult gall-flies. White circles-data from Table 3; 0Z 20Z 40% 60% 80Z 100;% black circles-data from experimentson thermaldeath RELATIVE HUmIDITY point. (The .effectsof humidity are neglected.) Fig. 5. The effect of constant conditions of temperature and humidity on the fecundity of the gall-fly. The pairing does not always result from such encounters. figures in the circles show the mean values for the A female which was seen walking from flower-head differentconditions. The curves show fecundity under to flower-head met a male. She at once flew a few the approximatelimits of regions in which the fecun- inches, and went under a leaf where she remained dity is below 2o, between 2o and 40, and above 40 eggs per female gall-fly. (Data in Table 3.) hidden for 2 min.

Interspecific pairing between related species of (b) Experiment on the dispersal of adult tsetse flies (Glossina spp.) has been observed in the gall-flies field by Vanderplank (I947), who recorded no mating In 1938 four liberations of marked gall-flies were preferences. made between 24 June and I July, all in exactly the No attempt was made to get eggs from the gall- same spot within z yd. of the cart-track in the former flies found cross-mated. But these observations may census area. Each batch of flies was distinguishable be correlated with the egg mortality discussed below. by a different mark in 'Robialine' enamel, either on Egg mortality was greater in I936 than in I935, and the wing, or more usually on the thorax or abdomen. in I936 7 % of the egg batches were infertile, and the Altogether Io8 flies were liberated, of which 70 eggs resembled those laid by unmated females. were males and 38 were females. They were The meeting between the sexes in the gall-flies is sought on seven occasions, the last being I3 July. not a rapid process, because the movements of the The distance travelled and the type of mark flies are so slow; they seldom fly, and when they do was noted in each case, without the fly being so they rarely fly more than 2 ft., and more usually captured. Marked flies were seen on I47 occasions; only 2 or 3 in. Males are usually solitary, and even I07 of the recoveries were of males and 40 of when they are active they tend to confine their females, so that, allowing for the differences in the

This content downloaded on Fri, 8 Mar 2013 13:36:49 PM All use subject to JSTOR Terms and Conditions I54 Natural controlof population balancein the knapweedgall-fly numbers liberated, males were seen about one and temperature and relative humidity measured in an a half times as frequently as females. out-door insectary at the Entomological Field Counts of the flies only a few minutes after they Station, 2 miles away. had been liberated showed that not more than 75 % The highest temperature recorded in the out-door could usually be found. However, in the case of two insectary was 32? with a humidity of 38 % in July liberations more than half the flies were still found I935, and the lowest maximum day temperature in after an interval of z or 3 days, while only just under July was I5? with 8o0 humidity. The minimum half were found after 4 days had elapsed. This high temperature recorded was 7? with a saturated proportion of recoveries shows how different the atmosphere. All these conditions are tolerated by behaviour of the gall-fly must be from that of the the gall-flies in the laboratory, and the daily maximum related Rhagoletis (Phipps & Dirks, 1932), in which temperature was always within the range of condi- 12% recoveries were made, at distances of 38 to tions in which the gall-flies laid eggs in the laboratory 156 yd. from the point of liberation. under constant conditions. This is in contrast to the Early in the experiment the weather was cold and results of Buxton & Lewis (1934, p. 225) on the windy, and dispersal was slow, no flies being found tsetse flies. These authors found that the maximum more than 3 yd. -from the point of liberation even recorded temperature reached the upper fatal limit after 5 days. There followed some sunny days, and of the flies, and that conditions in the wet season of the second liberation of 37 flies i8 individuals were such that the fecundity of the tsetse flies was were found after 4 days, the most distant male reduced to zero. having by then reached a point 9 yd. from the Although the weather conditions in the census point of liberation. Six days after this liberation area were always within the range of tolerance of the I of the flies were still found, the furthest gall-flies, nevertheless, changes in the weather having got I 5 yd. away. After 14 days one female alteredthe behaviourof the flies. On I5 July I935, was found 7 yd. away, and six males were found at 9.30 p.m. G.M.T., a short search was made for flies between 3 and 22 yd. from the point of liberation. by lamp-light, and six males and two females were After i6 days only two males were found, these seen resting on the unopened flower-heads of the being 3 and 20 yd. away respectively. Every day knapweed. Next morning at 4 a.m. an hourly the search was continued far outside the area in routine began; a strip of ground to the east of the which flies were discovered. Curiously enough not cart-track was examined carefully and all the gall- a single fly was found to have crossed the grassy cart- flies seen were noted, and their positions were marked track, which was only about 6 yd. wide, although by gummed labels stuck to the plants an inch or so knapweed was abundant on the other side. The away from the flies. If this was done carefully the marking did not appear to impair the flight of the gall-flies seldom flew away, although they turned gall-flies, and they flew readily if disturbed. and watched the operation. The area searched in The rate of disappearance of the flies is consistent I935 was about 6o sq.m. and it took nearly an hour with the hypothesis that about 75 % of the flies to cover it. At first the gall-flies moved very little, present were discovered, and that mortality rather but, as the morning advanced and the temperature than migration accounted for the slow fall in rose, their activity increased, and many flies ap- numbers. The figures suggest that after i week the peared which had certainly not been in plain view number of flies present had fallen to a half, and after before. By 8 a.m. most of the flies had moved some 2 weeks to a quarter of the number originally present, distance from their labels, and it was not easy to which gives an average life of 8-io days. It is guess which was which, so the labelling was dis- concluded that migration is far too small a factor to continued. The relevant observations are summarized invalidate the use of the formulae used to estimate in Table 4. Of the time periods observed, that the fecundity of the gall-flies from the census data. between 7.43 and 8.43 a.m. was the one in which most gall-flies were seen, and after this they soon began to disappear. The weather on the (c) effect of behaviour A more complete series of observations was made of the gall-flies on 7 July I936, over an area of about IOOsq.m. On all the days in 1935 and 1936 when flies were which included the area searched in 1935. Table 4 observed in the field, the weather conditions were shows that the results were rather similar, but that noted every hour. The temperature and the humidity there was a much greater excess of males. The were measured with a whirling hygrometer at a general activity of the gall-flies was low. No ovi- height of 4 ft. from the ground, and sometimes also position was observed (though it probably occurred) at i ft. from the ground. The wind velocity and the and many of the gall-flies remained close to their amount of cloud were also estimated. These results labels for long periods. Two males confined their were compared with continuous records of the movements within a radius of only 6 in. of their

This content downloaded on Fri, 8 Mar 2013 13:36:49 PM All use subject to JSTOR Terms and Conditions G. C. VARLEY I55 labels, and were seen on each round from 4.50 a.m. were resting in the shade beneath leaves or flower- to 4.40 p.m. On the average each male was seen heads. Presumably all the gall-flies which had close enough to its label for its identity to be certain disappeared were hidden in the dense herbage close on four consecutive visits, but each female on only to the ground. two visits, which shows that the females move about Another series of observations was made on single more than do the males. The peak of activity was less female gall-flies, whose activities were noted con- clear than in I935, and activity lasted considerably tinuously, and compared with changes in the weather. longer, but was less intense. Fig. 7 shows how the temperature and humidity The difference between the results in the 2 years changed during these days, and the black circles is probably due to the difference in weather between mark the times at which the flies were observed to the 2 days. On I5 July 1935 the temperature rose to lay eggs. Oviposition was seen over almost the whole a maximum of 27.50, and the gall-flies had mostly range of conditions met, except that the lack of vanished before i I a.m., when the temperature oviposition below I6? is probably significant. It (measured at 4 ft. above the ground) had risen to has already been seen that in the field general 24', and the humidity was down to 5I %. In I936 activity was greatest at 20. the temperature only just reached 230 with an 8o % The data obtained on I4 July I936 are particularly

Table 4 Total Total No. of Relative females males females Temperature humidity Period (G.M.T.) seen seen Pairs laying eggs (C.)(C.) A. The number of gall-flies (Urophora jaceana) seen in an area of 6o sq.m. on 15 July I935 4-5 a.m. I 0 0 0 I4 83 5-6a.m. 3 4 0 0 14 75 6-7 a.m. 3 I5 0 2 15 72 7.43-8.43 a.m. 20 24 4 8 I9 67 IO.I0-I0.45 a.m. 6 6 I I 24 5I 2.30 p.m. 0 ?2 0 0 24'5 50

B. The number of gall-flies seen in an area of ioo sq.m. on 7 July I936

4-5 a.m. I 4 0 0 I4 100 5-6 a.m. I 8 0 0 I5 100 6-7 a.m. I 9 0 0 17.5 92 7-8 a.m. 5 15 0 0 I9 8i 8-9a.m. 6 2I 3 0 21 79 9-Io a.m. 6 17 3 0 21 8o io-i i a.m. 6 20 3 0 2I 77 i i a.m.-i p.m. 3 2I 0 0 23 8o 4-5 p.m. 3 i8 0 0 I9 93 humidity. The peaks of activity in both years were instructive, as the weather was changing rapidly. at almost the same temperature, near 200, and Temperature varied between i6 and 20?. The wind activity had become less at 23 or 24?. This seems not was gusty,.and often made it difficult to keep the to be in agreement with the experimental results at gall-flies in view, owing to the movement of the constant temperature, which showed maximum plants on which they were standing. The sun shone fecundity at about 30?. It may of course be that the fitfully and there were occasional showers of rain. weather conditions in which the gall-flies are most The observations showed that neither rain nor lack active is not near the optimum for constant condi- of sunshine prevented oviposition, but that gall- tions. Uvarov (193 i) notes that the temperature flies were often stimulated to activity by the arrival preferred by an insect is greatly altered by its of sunshine, and would stop moving when a cloud previous treatment; thus the ant, Formica rufa, passed by; but this was not invariable. prefers 23' if it has previously been at 50, but prefers 32' if it has been at 27'. A change of similar magni- tude in the gall-fly would account for the difference (d) The rate of oviposition in the field between the optimum in the field and the optimum From these continuous observations of the gall- under constant conditions in the laboratory. flies, an estimate can be made of the rate at which The few gall-flies which remained in view when eggs can be deposited. The data are shown in the temperature was at its maximum of 270 in I935 Table 5. The eggs laid by female no. I were not

This content downloaded on Fri, 8 Mar 2013 13:36:49 PM All use subject to JSTOR Terms and Conditions 156 Natural control of population balancein the knapweedgall-fly counted, but since the mean number of eggs laid at If the flies maintained their maximum rate of one time is three (p. I46) it may be surmised that oviposition, they would require something like 4 hr. about I 5 eggs were laid. The rate of egg laying to lay the 70 eggs which was the estimate for the during the active period of this gall-fly was i6 eggs fecundity in 1935, and 3 hr. to lay the 52 eggs in per hour. The rate for female no. 6 was io eggs per 1936. This is a very small proportion of the lifetime hour. The other flies observed laid far less rapidly, or of a gall-fly, and indicates why oviposition is ob- not at all. The mean rate for all the flies on I4 July served only rarely. 1936 was 3 eggs per hour per female gall-fly over the Although oviposition can be so rapid, and can take whole period of observation. place over so wide a range of conditions of tempera-

IZNoo JIJLY 1935

JtLY 23 1935 20tC

84A

JULY 41936 150C

6A.M.

JLULY 15 1935

40% 60% 8070 100% RELATIvE HuMIDITY Fig. 7. Oviposition of gall-flies in relation to conditions of temperatureand humidity in the field. The lines representthe changes in temperatureand humidity on days when gall-flies were observed ovipositing. Black circles representthe times (B.S.T.) at which oviposition took place.

Table 5. Observationson the behaviour of individual female gall-flies (Urophora jaceana) in the field No. of flower- Period of observation heads examined No. of No. of Times at which Fly no. Date (G.M.T.) by fly ovipositions eggs laid oviposition occurred

I 23 July 1935 5.50-8.25 a.m. 15 5 ? 7.30, 7.40, 7.48, 8.oo, 8.15 a.m. 2 14 July 1936 7.12-7.52 a.m. 3 0 0 3 ,, ,, 7.39-8.25 a.m. 3 1 3 8.07 a.m. 4 ,, ,, 8.50-II.29 a.m. i8 I 3 8.55 a.m. 5 ,, ,, I0.45 a.m.-I.25 p.m. 6 2 5 11.23 a.m., I.I9 p.m.

6 ,, ,, l2.I4-I.I6 p.m. I7 4 I0 I2.I4, I2.30, I2.49 I.00 p.m. 7 I5 July 1936 2.59-4.30 p-m. 3 0 0

This content downloaded on Fri, 8 Mar 2013 13:36:49 PM All use subject to JSTOR Terms and Conditions G. C. VARLEY I57 ture and humidity, the flies in captivity contained an and humidity (data in Table 3), and thirdly the average, of 50 apparently mature eggs when they expectation of life at different temperatures (Fig. 6) died. Why gall-flies containing mature eggs fail to it was calculated that the weather might be respon- lay them remains an unanswered question. sible for a change in the fecundity in the ratio of I-4:I. This expected difference is not large, but it agrees with the ratio of the estimates of the fecundity (e) The effect of weather on fecundity in the z years concerned,since 70:52= I 3:I. How- Oviposition was restricted in the field to the ever, the statistical errors in the field observations month of July, and the weather in this month was were so large that little reliance can be placed on the much cooler in I936 than in I935. This is seen in the ratio between these figures. 2-hourly means of temperature and humidity To conclude, the available data are insufficient to

250C 16

12~~~~~~~~~~~~~~~~~~~1

20 JUY2950tJL 1936 8244

150C200C-m 8 24 JuLY 1935 JuLY 19,364

6Wf/o 70% 80% 90% 100% 70% 80% 90% 100% RELATIVE HUMIDITY

Fig. 8. Mean weather conditions for July in I935 and 1936. Two-hourly means of temperature and humidity for the month of July in 1935 and I936, obtained from continuous records in an out-door insectary at the Entomological Field Station, Cambridge. Numbers indicate the time of day (B.S.T.). measurements made in an out-door insectary for the assess the effect of weather on the fecundity of the month of July in 1935 and 1936, which are plotted gall-flies in the field. against each other in Fig. 8. The difference in the mean maximum temperature in the 2 years is 30, but PART 3 the humidity at corresponding temperatures was THE FACTORS WHICH AFFECT THE SUR- almost the same. An attempt was made to see how VIVAL far this difference in mean temperature might be OF THE EGGS, LARVAE AND expected to alter the fecundity. PUPAE OF THE GALL-FLY No rigorous method is available, and the method The mortality can conveniently be divided into two used is too complex and too uncertain to warrant periods, the first up to the formation of the gall, and detailed description. Using first the 2-hourly means the second after this event. As the causes of the of temperature and humidity in July I935 and 1936 mortality and the methods of estimating it are (Fig. 8), and secondly curves showing the rate of different in the two periods, they can best be treated oviposition under different conditions of temperature separately.

This content downloaded on Fri, 8 Mar 2013 13:36:49 PM All use subject to JSTOR Terms and Conditions 158 Natural control of population balancein the knapweedgall-fly

i. THE MORTALITY UP TO THE FORMATION frequency distribution of the gall-cells in col. 3 should have been the same as that of the in OF THE GALL eggs col. 4, apart from sampling errors. The frequency (a) The egg mortality in I935 distribution of gall-cells shows that groups i and In July I935 many eggs were found in various stages 2 are larger, and groups 3-I2 are smaller than in of development. In normal development the yolk the frequency distribution of the eggs. first shrinks away from the ends of the egg shell, and This difference is strongly significant, as is shown becomes sausage-shaped. Soon the blastoderm by the x2 test. The value of x2 is shown at the foot shows as a transparent outside layer, but as differen- of the column, groups 7-I2 having been lumped tiation proceeds to the formation of a definite larva, together for the purposes of its calculation, leaving the blastoderm becomes less clearly distinct from seven groups. As the totals of the groups have been the yolk within. equalized, this leaves five degrees of freedom; In I935 49 abnormal eggs were found, which Fisher'sTable 3 (I934) shows that a value exceeding would certainly not have hatched. Of these, I 2 were a tenth of this would be expected only once in a filled with yolk which contained large oil globules hundred trials. The strongly significant change in and appeared to be decaying. In I3 the yolk had the frequency distribution of the gall-cells must be shrunk from the ends of the eggs, but was still due to mortality in the eggs and young larvae. opaque, and no embryo was forming. These two The amount of this mortality can be estimated, types of egg were not always clearly distinct from assuming that its incidence is random, and that the normal eggs, but could at once be recognized if death of one egg or larva in no way alters the other eggs in the same batch had developed normally. expectation of life for other larvae in the same flower- In 2z eggs the yolk was irregularly shaped, twisted, head. Then, if there is a chance m that any one larva or even broken into two fragments. Lastly, 2 eggs will die before it forms a gall, the chance that two contained dead second instar larvae. All the dead larvae in the same flower-head will die is M2n; that eggs except the last two resembled eggs laid by only one of two will die is 2m (I - m), and the chance unmated females; it is therefore likely that absence that neither will die is (i-rm)2. This reasoning can of fertilization prevented their development. be applied to egg groups of all sizes. By multiplying Few of the eggs which were alive at the time of the chance of mortality by the frequencies of the egg examination would have died before hatching, and groups in col. z of Table 6 a new larval frequency a fairly accurate estimate of the total mortality is distribution can be built up for any assumed value of obtainable by taking the ratio of dead eggs to the random mortality. This calculation has been made total found. The mortality was calculated only from for a number of different values of mortality, and data of egg batches in which no hatching had yet for comparison with the totals found in the field the taken place. There were 148 such egg batches with resultant frequency distributions have been multi- a total of 447 eggs (Appendix, Table F). Of these plied by a factor to bring the sum of groups i-iz up 40/447 = 8-9 % were dead. The standard error of to 886. this mortality is estimated to be o0oI3. In only two These calculated frequency distributions are given of the egg batches which contained more than one in Table 6, and their goodness of fit with the ob- egg were all the eggs dead. served frequency distribution of the gall-cells in col. 3 can be estimated as before from the value of x2, after lumping groups 7-I2 together. (b) The mortality of the larvae up to the The values of x2 at the bottom of Table 6 have formation of the gall in 1935 a minimum of I4'77, when the mortality is 0-29. The mortality in this period cannot be obtained The number of degrees of freedom is five, and the directly, but the fact that the eggs are normally laid corresponding value of P, the probability of such a in groups in the flower-heads provides an indirect difference occurring;by chance, is between o-oi and method by which the total mortality up to gall- o0o2. Such a high value of x2 would be expected formation can be estimated. The egg mortality only once in eighty trials, if the difference between being already known, the larval mortality can readily the frequency distributions were due solely to be found. random errors. This suggests strongly that the The frequency distribution of eggs laid is shown assumption of random mortality is not correct. in Table 6, col. 2, and the same frequency distribu- Comparing the sums of the items in Table 6, the tion is given again in col. 4, but with the figures mortality appears to be rather higher than o03z for multiplied up so that the total is 886, thus making eggs in groups of four or more, and less than o-28 the figures comparable with the frequency distribu- for eggs laid singly or in pairs. However, the mean tion of gall-cells in col. 3. Had there been no mortality must be near the minimum value of x2, mortality amongst the eggs and young larvae the which occurs when ni=0 289 (Fig. 9). This figure

This content downloaded on Fri, 8 Mar 2013 13:36:49 PM All use subject to JSTOR Terms and Conditions Table 6. Comparison of the frequency distributions of the eggs and the gall-cells of the gall-fly (Urophora jaceana) in I93 5 Frequency of flower-heads containing each number of eggs or gall-cells

Observed No. of frequency Calculated frequency of flower-heads with gall-cells if random mortality=m eggs or , _ A_-__ _A gall-cells Eggs Gall-cells m= o m = 0-2 m=o-25 m=o-z6 m=0-27 m=o-28 m = 0-29 m=o-30 m = 0-31 m=0-32 m - 0-35 0 - - 48-28 66-i8 701I5 74-26 78-52 82-93 87-50 92 23 971I6 I13-09 I 29 287 I73-6I 2491 3 272-o6 276 85 28i170 286-62 29I-6I 296-66 30IP79 306-98 322-98 2 38 272 227-48 266-32 27246 273 48 274 43 275'30 276-I0 276 83 277 49 278-o6 279 33 3 36 I96 215-51 I89-79 i8o-i6 I78-I4 I76-og 174-01 I7I-9I I69-78 i67,64 I65-47 158-8I 4 23 79 I37-69 92-22 83-66 82-04 80o43 78-84 77 27 75 75 741I9 72.66 68-I6 5 8 29 47 89 39 94 37.I8 36-56 35 93 35-27 34-60 33'90 33 20 32-47 30-21 6 5 20 29-93 24 77 21-51 20-84 20-17 19-50 I8-84 I8-17 17-5I I6-87 14'97 7 5 2 29'93 I3-I0 IO33 9.84 9.38 8.93 8-5I 8-II 7-72 7-35 6-34 8 2 0 II-97 4.86 3-98 3 84 3.69 3-56 3-43 3-30 3.I8 30o6 2-7I 9 I I 5 99 2.34 21I4 2-09 2.04 I9-8 I-gI I-84 177 170 146 10 0 0 0 I79 I.49 I-42 I.34 I-26 i-i8 I-I0 102 0o95 0-73 II 0 0 0 I.30 o-82 o073 o-66 0o59 0?53 0?47 0-42 0?37 0-25 I2 I 0 5-99 0?43 0-20 O-I7 0-15 O0I3 O-II 0-09 o-o8 o-o6 0-04 Total 1-I2 148 886 886-o 886-o 886-o 886 o 886-o 886-o 886-o 886-o 886-o 886-o 886-o x2, lumping groups 7-12 168 35 30-22 17-84 16-47 15'52 14-91 1477 14'99 15 63 1673 2265

Table 7. Comparison of the frequency distributions of the eggs and the gall-cells of the gall-fly (Urophora jaceana) in I936 Frequency of flower-heads containing each number of eggs or gall-cells

Observed No. of frequency Calculated frequency of flower-heads with gall-cells if random mortality = m eggs or __ gall-cells Eggs Gall-cells m=o m=o02 m = 0-25 m = 0-27 m = 0-29 m=0'30 m=0-3I m=0-32 m?==033 m=0o34 m=0o35

0 i8- - I8-47 24-82 27.64 30 63 32-20 33 82 35-50 37 23 39-02 40 87 I 22 90 72-25 88-88 94-36 96-72 99-I8 100-45 10-75 103-08 104'43 105-79 I07-22 2 i8 96 59-II 72-28 75-I2 76-I9 77-23 77 73 78-22 78-70 791I6 79-64 80-o0 3 i8 57 59-II 56-4I 5556 55-I4 54-66 54-40 54-12 53 82 53-50 53-I6 52-8i 4 II 26 36 12 35'03 33 44 32-66 31.79 3I.32 3o084 30'34 -29-82 29-29 28-73 5 9 I0 29-56 21-66 I8-98 I7-89 i6-8o I6-25 I5-7I 151-7 14-63 14-08 I3-57 6 6 4 19-70 9-96 7-97 7-25 6-58 6-26 5 94 5-64 5 35 5-o6 4 79 7 3 5 9-85 3 25 2-50 2-24 199 i-88 1.76 i-66 I-56 145 I.36 8 0 0 0 i-o6 o-8o 0-71 0-6i 0 57 0 53 0?49 0?45 0-41 0-38 9 I I 3-28 0-47 0-27 021I 0-17 OI5 0-I3 O-II O-I0 0-09 o-o8 Total I-9 88 289 289-o 289-o 289-o 289-o 289-o 289-o 289-o 289-o 289-o 289-o 289-o x', lumping groups 6-9 59.II 17-93 12-I5 I0-53 9-36 8-96 8-71 8-59 8-63 8-82 9-21

This content downloaded on Fri, 8 Mar 2013 13:36:49 PM All use subject to JSTOR Terms and Conditions i6o Natural controlof population balancein the knapweedgall-fly has been used in calculations in other parts of this 0'20, expressed as a fraction of the eggs laid. paper. Expressed in terms of the larvae which hatched it It is necessary to estimate the standard error of becomes 22%. the x2minimum to find the accuracy of the estimate For the most part this larval mortality was not of the mortality. I am indebted to Prof. R. A. directly observed, and much of it may have been due Fisher for the method employed. The variance of a to intrinsic causes which killed the larvae before they formed galls in the florets. However, in a single instance two larvae were found in the same floret, and one of them was already dead. This observation suggests that some of the larval mortality was due to competition, and provides an explanation for the fact that the mortality seemed to be higher in the large egg batches than in the small. A B1 The early larval mortality is therefore to be regarded as density dependent, and its operation will be considered in more detail in the discussion. 15 ] 12 (c) The egg mortality in 1936 Eggs were found in I936 from the beginning of 15~~~~13 July to the middle of August. In the 88 egg batches found in which no hatching had yet taken place, there were in all 267 eggs of which 41 were dead x2 C (Appendix, Table G). The egg mortality was there- fore I5 3 %, with an estimated standard error of 1935 \ 1936 0'022. Altogether 49 dead eggs were examined in I936. Of these zi were filled with decaying yolk, 7 had the 14\ I 10 yolk contorted, the yolks of I7 were opaque and stumpy, and 4 larvae had died in the first instar while still within the egg. The proportions of the different types of dead eggs were rather different than in I935. Moreover, in 1936 about half of the dead eggs were found in egg batches in which every egg was undeveloped and dead. These dead eggs resembled those laid by unmated females. It may be inferred that the low population density of gall- flies in I936, which led to interspecific mating in a number of instances, also led to the laying of eggs by 0.25 0.30 0.35 unmated or cross-mated females. ASSUMED VALUEOF RANDOM1ORTAUTY (d) The mortality of the larvae up to the Fig. 9. Estimation of randommortality from x2minimum. formation of the gall in I936 The values of x2 in Tables 6 and 7 are plotted against the values assumed for the random mortality. The Table 7 shows the frequency distribution of the construction on. the curve indicates the method used eggs laid and of the gall-cells discovered in I936, and for estimating the standarderror of the x2 minimum, in cols. 4-I0 are shown the frequency distributions see text. derived from the egg distribution assuming different values for random mortality. These expected distri- x2 minimum equals twice the radius of curvature of butions are compared with the frequency distribu- the x2 curve at the minimum. The curve is shown in tions of gall-cells in col. 3 by the x2 test, having Fig. 9, and the radius is measured by AB2/2AC, lumped groups 6-9 together, leaving four degrees of since the curve is parabolic. The standard error so freedom. The values of x2 are plotted in Fig. 9. The estimated was found to be O-022. minimum value of P is between o-i and 0o05. This To conclude this section, the larval mortality value is not unreasonably high, and the figures are prior to gall-formation is obtained by subtracting the not inconsistent with the assumption that the mortality in the egg stage (o0o89) from the total difference between the frequency distributions of mortality (o-289). This leaves a larval mortality of eggs and gall-cells was due to random mortality. As

This content downloaded on Fri, 8 Mar 2013 13:36:49 PM All use subject to JSTOR Terms and Conditions G. C. VARLEY already noted, in I935 the data suggested that the The following method was used. The totals for mortality was not random. The difference between ten or more successively collected square-metre the results in the z years cannot be attributed to a samples were added together, and the mean numbers greater degree of competition in I935, since the or percentages of gall-fly larvae killed by various mean number of larvae in the occupied flower-heads, agencies were found. The census data also gave the and hence the competition, was very slightly greater period of time over which each agency operated, and in I936 than in 1935. Possibly the smaller amount the results were built up into Table 8, which shows of data available for I936 is insufficient to demon- both the time of operation and the numerical effect strate the effect observed in 1935, owing to the con- of every important factor. comitant reduction in accuracy. For those factors which operated simultaneously The random mortality occurring between the separate percentages are not given. Similar tables laying of the eggs and the formation of the galls is are given by Schwerdtfeger (1936) for the mortality estimated from the x2 minimum to be 0o323. The of the moth Dendrolimuspini, which is a serious pest standard error of the estimate, determined from the of conifers in north . graph, is o0o35. This mortality includes only a part The census of galls on the s,tanding stems of of the egg mortality. Half of the I5.3 % egg mor- knapweed began in February i935, but the first tality was of entire egg batches, which would not complete census including fallen galls was not made alter the frequency distribution of the gall-cells, so until May. However, although the fallen galls were long as small and large batches failed to hatch with omitted, the census of io sq.m. in February 1935 equal frequency. Only the remaining 7-6 % of the gave an idea of the factors which had caused egg mortality was probably random, and would be mortality in the previous summer. But they included in the 32-3 % total random mortality up to provided a low estimate of the total numbers. gall-formation. This leaves In the sections which follow, each of the major 0-323 (IOO-7'7)-7-6 = 22-2 % causes of mortality listed in Table 8 will be given as the random mortality of the larvae or separate consideration. 22-2 -=26-2 % (IOO- I5.3) (a) Winter disappearance when expressed in terms of the larvae which hatched. The total number of newly formed gall-cells found The over-all mortality from the egg stage up to in the late summer of 1935 was estimated from the formation of the gall consisted of 7.7?2-2% twenty samples to be I476?2I 5 per sq.m. (Ap- mortality of whole egg batches followed by penclix, Table A). By the following spring and early 32-3 ? 3-5 % random mortality of the remainder, summer the mean for 36 sq.m. had fallen to 56-8 ? 7-o which taken together give an over-all mortality of gall-cells per sq.m. (Appendix, Table C). This 37.5 ? 3'4%. difference is strongly significant, and the winter The relation between these various figures is made disappearancewas estimatedto be 6I +?7-3 %. clearer by the following schematic representation: That this loss was due in part to the observer's

Mortality given Nos. left from each as successive IOO eggs laid percentages % % whole batches ...... dead Dead eggs I5-3 f77 ...... 7-7 eggs 77 t7 6 % random ...... 7-6 dead eggs 8-2 4Total random death of 32-3 % 22-2 dead larvae 26-2 Live eggs 84-7% .. eggs and larvae I. Survival 67.7% 62-5 live larvae IOO*O% IOO-O% IOO-O% failure to notice the galls is certain, but in some at 2. THE MORTALITY AFTER THE FORMATION least of the sq.m. samples examined this explanation OF THE GALL is quite inadequate. Some of the sq.m. were crossed The examination of square-metre samples of knap- by a maze of mouse or vole runs, and when the weed throughout the summer gave information vegetation was cleared away the soil surface was about the contents of each gall-cell. From these almost smooth. It is inconceivable that under these data, with the addition of those already given in the conditions more than 5 % of the fallen galls could previous sections, it has been possible to build up have been overlooked. Yet it was in these places that a picture of the course of events over the period of fewest galls were found. Frequently small piles of 2 years. partly destroyed galls were found in mouse runs, and

This content downloaded on Fri, 8 Mar 2013 13:36:49 PM All use subject to JSTOR Terms and Conditions i62 Natural control of population balancein the knapweedgall-fly

Table 8. The effect of successive mortality factors on the numbers of the knapweed gall-fly (Urophora jaceana) found per sq.m. at Madingley Non-specific mortality (i.e. that which affects the gall-fly and its parasites indiscriminately) is marked with an asterisk. No. killed No. alive per % killed per sq.m. sq.m. 1934 July No. of larvae in gall-cells 43 larvae Died due to unknown causes 5 2 41 Parasitized by Eurytoma curta I5 6 35 Aug. Miscellaneous parasitism: Habrocytus trypetae 4 Torymus cyanimus 3 Tetrastichus sp. f 2 - *Destroyed by caterpillarsJ 4 6 21-4 larvae 1935 Winter *Winter disappearance not estimated ? ? 21-4 *Destroyed by mice i8-5 4 174 May-June Miscellaneous parasitism: Habrocytustrypetae JO - *Macroneura vesicularis 6o 0-4 Tetrastichus sp. J oli 6-9 flies

July 6-9 flies emerged per sq.m., 421 % were females Mean number of eggs laid: 70 per female 203 eggs Infertile eggs 9 I8.3 I84.7 ,, Larvae died before forming galls 20 37.I 147-6 larvae Larvae died in galls due to unknown cause 2 3 144.6 Parasitized successfully by Eurytoma curta 45.5 65 8 78.8 Aug.-Sept. Miscellaneous parasitism: Habrocytus trypetae j 5-6 *Eurytomarobusta 4-I *Torymus cyanimus 37 3-7 *Tetrastichus sp. o-6 *Destroyed by caterpillars I4.8 50 larvae 1936 Winter *Winter disappearance 6I.5 30o8 I9-2 *Destroyed by mice 64 12-2 7o0 Larvae died due to unknown causes 26 I-8 5-2 Miscellaneous causes: *Birds l 0'4 May-June Habrocytus trypetae 0-7 *Macroneura vesicularis 3 0 25 *Tetrastichus sp. 0-25 3-6 larvae July *Drowned in floods 44 1.57 2-03 flies

July 2 03 flies emerged per sq.m., 42 % were females Mean number of eggs laid: 52 per female - - 44-8 eggs Infertile eggs I5.3 6-9 37-9 " Larvae died before forming galls 26-2 9 9 28-o larvae Larvae died in galls due to unknown causes 4-3 I2 26-8 Parasitized by Eurytoma curta 27 7.2 I9-6 Aug.-Sept. Miscellaneous parasitism: Habrocytus trypetae 0O2 *Eurytoma robusta i-6 * Torymuscyanimus 36 I .5 *Tetrastichus sp. 0-2 Killed by Lestodiplosis 0-2 *Destroyed by caterpillars 3-3 I2'6 larvae

This content downloaded on Fri, 8 Mar 2013 13:36:49 PM All use subject to JSTOR Terms and Conditions G. C. VARLEY I63 much of the winter disappearance was probably due As already noted, mice attacked only the fallen to mice carrying galls underground. galls. In June and July I935 the total number of For this reason the 6 i '5 % winter disappearance gall-cells per sq.m. was 38, of which 23 had fallen. has been counted as a form of mortality in the Of the latter 6-5, or 28 %, were destroyed by mice. calculations which follow. In I936 the total number of gall-cells was 57 per As a summer census was not made in I934 it is not sq.m., and 53 of these had fallen, of which 36, or possible to estimate the winter disappearance for 68 %, had been destroyed by mice. I934-5. In the first census, made in FebrutaryI935, the mean number of gall-cells found on the standing stems was 43 ? 9, by which time, of course, an un- (c) Mortality due to unknown causes known number of galls had already fallen. By July A proportion of the larvae and pupae of the gall- 1935 a mean number of only 15 gall-cells remained fly were found dead or destroyed without the cause on the standing stems in each sq.m., but a mean of being apparent. Into this class will fall all the 23 per sq.m. was found fallen to the ground. This mortality which is not due to parasitic and pre- gives a mean total of 38 ? IO gall-cells per sq.m., daceous insects, or to mice. It will therefore include which is not significantly less than the 43 ? 9 found any mortality due to parasitic disease, roving pre- in February. Probably winter disappearance was dators (e.g. mites), fungi, and climatic factors, as well much less than in the following winter. As the as any intrinsic functional failure. The appearance winter disappearance is believed to be due to mice, of the dead larvae varied greatly. In a few cases the it is also significant that mouse damage was far less gall-cell was empty, and there was no sign that the also (Table 8). soft tissues of the gall had been eaten; these larvae must have died at a very early stage. Some dead larvae were brown and flabby, while others were dry (b) Mortality due to mice and hard, and covered with fungal hyphae. Dr Petch Many of those galls which fell to the ground during very kindly named the fungi as Aegerita sp., the winter became quite free from the flower-heads Fusarium sp., Cladosporiumsp. and Cephalosporium in which they were formed. At the base of some of muscarium. None of these is a parasitic species, and them were large gaping holes, each opening into a it is very likely that they attacked the fly larvae only separate gall-cell. Sometimes the whole wall of the after death. gall was destroyed, leaving perhaps only one partly There is no evidence that climatic factors prove intact gall-cell. Such gall-cells never contained any fatal in winter. Nor is high temperature in summer live insect larva or pupa, and their previous history ever likely to be lethal to the larvae under natural could not be inferred unless perhaps fragments of a conditions in England. Experiment showed that for puparium of the gall-fly indicated previous attack by i hr. exposure the upper fatal limit of the larvae was the chalcid parasite Eurytoma curta. about 430 C. The temperature of the inside of a The evidence that mice destroyed these galls is as flower-head in bright sunshine did not exceed that follows: Only the fallen galls were affected. They of the air by more than 50, as was shown by measure- were often found in small heaps in mouse runs. The ments made with a small thermocouple. The larvae holes in the galls had been made from the outside, would be subjected to a temperature of 430 only if and were not at all like the neat circular holes made the air temperature reached or exceeded 380, which by emerging parasitic Hymenoptera. The galls were is IO1 higher than any reading taken during the attacked with great thoroughness, and usually had period of the census. a hole into each gall-cell. It was only in the largest The gall-fly larvae are very resistant to dry condi- galls, with six or more cells, that the central cell was tions, and when fully grown they can be kept for sometimes intact. Whatever did the damage must months in dry gelatihe capsules, in which they will have turned the gall over to deal with it from all complete their development, and emerge as flies. sides. It is extremely unlikely that any insect would Gelatine capsules are in general rapidly fatal to most do this, and the only other likely animals are mice, fly larvae, unless the humidity is kept very high. voles or shrews. There is no direct evidence avail- Equally thewet conditions of the winter seem to cause able, but in view of the facts put forward it is little or no mortality to the gall-fly larvae in the field. concluded that mice or voles were responsible. The only evidence that climatic factors caused the The percentage of gall-cells found destroyed by death of any of the inhabitants of the galls was seen mice was i8-5% early in I935 and 64% early in in the very wet period in July I936, when the ground 1936. As mice were probably responsible for most was waterlogged and covered with-puddles for some of the 6i-5 % winter disappearance in I935-6, mice days. Such conditions were fatal to a large propor- may have caused a mortality of 86 % of the gall-fly tion of the larvae and pupae of the gall-fly and its larvae in the gall-cells of the I935-6 generation. parasites which were submerged in the fallen galls.

J. Anim. Ecol. i6 II

This content downloaded on Fri, 8 Mar 2013 13:36:49 PM All use subject to JSTOR Terms and Conditions i64 Natural controlof populationbalance in the knapweedgall-fly Some of the mortality included in unknown females. The details of the search for hosts and the causes may be due to the feeding habits of the chalcid spatial distribution of parasitism have been described parasite, Habrocytus trypetae. The females of this elsewhere (Varley, I94I). The females discover parasite sometimes fed on larvae in the gall without flower-heads of the knapweed during flight, and laying eggs, and, unless the delicate feeding tube often alight and walk over them, tapping them with was found, the cause of destruction would not be the antennae. They may eventually insert the ovi- apparent. positor into a flower-head, even though, as shown by subsequent dissection of the flower-head, it may be (d) Mortality due to chalcid parasites devoid of hosts. The gall-fly larvae are suitable for attack soon after hatching from the egg (in the second The gall-fly larvae and pupae in the puparia were instar), but no eggs have been found in third instar attacked by parasites at three main periods in the gall-fly larvae. year. Some of the common species are illustrated An individual gall-fly larva spends about 2 weeks in Fig. iO. The common Eurytoma curta laid its eggs in the second instar. In I935 second instar larvae in the gall-fly larvae soon after they had hatched, and were found between 9 July and 25 August, and the parasites destroyed their hosts soon after the females of Eurytoma curta were seen in the field latter had completed their growth in August. between 9 July and 6 August. In I936 the hosts From the time when the gall-fly larvae approached were available from I7 July to I September, and the full size until the end of the summer, they were females of E. curta were seen in the field between attacked by various other chalcids, chief amongst 7 July and 8 August. The census showed that some which were Torymus cyanimus, Eurytoma robusta, E. curta must have emerged later than this, for some Habrocytus trypetae, and Tetrastichus sp. B, all pupae of E. curta collected from fallen in except the last of which are ectophagous. galls August did not emerge in captivity until early The third period of attack began in the early September, when there were no suitable hosts in the summer of the following year, when other genera- knapweed. The retarded emergence of these indi- tions of Habrocytus trypetae and Tetrastichus sp. B, viduals was probably one of the effects of and Macroneura vesicularis attacked any larvae or flooding. Probably in normal years the emergence of pupae of the gall-fly or of any other parasite which E. curta corresponds fairly closely with the period was in the galls. The competition between the various during which gall-fly larvae are suitable for parasiti- species was severe. The method of attack of each zation. species is described below, and the hosts selected for The egg and larval stages have been described attack and the success of the attacks are considered. elsewhere (Varley, I937a). The egg is like a short In one or two instances gall-fly larvae were found sausage with a long 'tail', and its volume may be as which had apparently been killed by the very gmall much as a tenth of that of the host in which it is larvae of the predacious gall-midge Lestodiplosis laid. The egg hatches in a few days, but the endo- miki. These larvae normally attacked other gall- phagous larva grows very slowly, and is in midge larvae present in the flower-heads. They have usually the third instar when the gall-fly larva is fully grown been describedby Otter (I938). in August. At this time the parasite exerts a peculiar influence on the host, which, instead of passing into (i) Eurytoma curta (Fig. ioA) a diapause and hibernating in the larval state, turns Eurytoma curta parasitizes only the larvae of the to face the exit of the gall-cell, and forms its gall-fly Urophorajaceana in the knapweed, but it has puparium(Varley & Butler, I933). Inside the brown been recorded as a parasite of various gall-forming puparium of the host the parasite larva begins to insects on other plants, such as the gall-flies Urophora grow rapidly, and consumes its host completely eriolepidis (Lw.), U. stylata (Fabr.), U. cardui (L.), within a few days. It passes the winter as a fifth Tephritis vespertina(Lw.) and the gall-wasp Aulacidia instar larva in the otherwise empty puparium of the hieracii (Bouche). No other trypetids were common host. This early pupation makes it very easy to on the site of the census breeding in plants other recognize parasitized hosts, until the normal time than the knapweed, although a few specimens of for pupation comes in May (Fig. 3 G). Urophora stylata, Xyphosia miliaria (Schr.) and The number of E. curta per sq.m. during the Icterica westermanni(Meig.) were seen. No evidence period of the census is indicated in Table 9. The was obtained to indicate whether or not these were factors which caused mortality were mostly the same attacked, but they were far too uncommon to be as for its host, the gall-fly, and the percentages killed important alternative hosts. by mice, and winter disappearance, have been given The adults of Eurytoma curta emerge mostly in the the same values as for the gall-fly in Table 8, since first half of July. Out of I03 adults reared fifty-four these factors destroyed the contents of the gall more were females, which gives a proportion of 0o52 ? 0?05 or less indiscriminately.

This content downloaded on Fri, 8 Mar 2013 13:36:49 PM All use subject to JSTOR Terms and Conditions Scate: mm Fig. io. Chalcid parasites of the knapweed gall-fly. A. Eurytoma curta: -colour black except for brown fore-tibiae, brown apices of the femora, and brown tarsi. (E. robusta differs mainly in the shape of the abdomen.) B. Habro- cytus trypetae: colour of head thorax and abdomen dark metallic green; coxae and all but apices of femora metallic green; tibiae and apices of femora brown, tibiae centrally infuscate; tarsi yellow, last joint dark brown. C. Torymus cyanimus: colour brilliant metallic green with blue and violet reflexions; legs mainly bright yellow, but femora centrally and coxae wholly metallic green, and hind tibiae centrally infuscate. D. Tetrastichus sp. B: colour dull metallic green, legs brown. E. Macroneura vesicularis: colour very dull green with dull coppery reflexions. Legs of female pale yellow with darkened femora. Legs of male with dark apices to hind and mid tibiae.

I 1-2

This content downloaded on Fri, 8 Mar 2013 13:36:49 PM All use subject to JSTOR Terms and Conditions i66 Natural control of population balancein the knapweedgall-fly The mortality due to parasitism in August and other parasites had hatched, then the newly hatched September is inevitably rather too low in Table 9. ectophagous larvae died of starvation, being unable As will be seen later, the ectoparasites do not (with to bite through the hard puparium; and the E. curta the exception of Habrocytus trypetae) choose between survived. The acceleration of the host's pupation by healthy larvae of the gall-fly, and those which already E. curta is thus advantageous to the parasite, as it contain larvae of Eurytoma curta. Cases of multi- protects it from the attacks of some of its enemies. parasitism were observable only before the host Those ectophagous chalcids which attacked after was completely consumed by an ectoparasite. Many E. curta had completed the destruction of the gall- found later than this doubtless escaped notice, and fly within its puparium sometimes laid their eggs

Table 9. The effect of successive mortality factors on the numbersof the chalcid parasite Eurytoma curta found per sq.m. at Madingley Non-specific mortality (i.e. that which affects the gall-fly and its parasitesindiscriminately) is markedwith an asterisk. No. killed No. alive per % killed per sq.m. sq.m.

I935 Feb. Larvae per sq.m. 2'7 larvae May-June Miscellaneousparasitism: Habrocytustrypetae 26 0?4 *Other parasites 030 2-o adults July 2zo adults emerged per sq.m., 52 %were females Mean number of eggs laid: 63 per female - - 66 eggs Egg or larva died-host survived 0o3 0-2 65-8 larvae Aug.-Sept. Miscellaneousparasitism: *Torymuscyanimus 0-2 Habrocytustrypetae j o-6 *Tetrastichussp. 25 0?4 Died due to unknown causes- o05 *Destroyed by caterpillars 14.I 5? laryae I936 Winter *Winter disappearance 6i 5 30X8 192' *Destroyed by mice 64 I2-3 6-9 Missing 33 2-3 4-6 *Destroyed by birds 0 -4 June Miscellaneousparasitism: Habrocytustrypetae 26 0o3 Tetrastichussp. 0o3 *Macroneuravesicularis JOI 3-5 larvae July *Drowned in floods 53 i.84 I-66 adults

July i 66 adults emerged per sq.m., 52 % were females Mean number of eggs laid: 8-4 per female - 7'2 eggs Aug.-Sept. *Miscellaneousparasitism 003 Died due to unknown causes 25 O-I *Destroyed by caterpillars II4 5-4 larvae

the death of the gall-fly would be credited solely uselessly outside the puparium, and sometimes laid to the successful ectoparasite. The incidental death them within the puparium on the body of the larva of any small Eurytoma larva would be unrecorded. of E. curta, which then served as a host. Only those Although the ectophagous larvae of the chalcids E. curta remaining in the standing flower-heads were Torymus cyanimus, Habrocytus trypetae and Macro- subject to this form of attack, which was prevalent in neura vesicularis might destroy the larva of Eurytoma the spring. curta with that of its host, the outcome of such Three instances were noted in which a healthy competition depended very much on the circum- gall-fly larva contained a dead egg of E. curta. Two stances, and in particular. on the timing of the other gall-fly larvae contained either a dead egg or attack. If the E. curta larva had caused the pupation larva of E. curta, but both had also been attacked by of the gall-fly larva to begin before the eggs of the E. robusta, which may perhaps have been responsible

This content downloaded on Fri, 8 Mar 2013 13:36:49 PM All use subject to JSTOR Terms and Conditions G. C. VARLEY I67 for the death of the E. curta larva. These instances This reduction in the fecundity of E. curta in I936 have been included in Table 9 in the o03 % mortality must be partly due to the cold weather. Now the in July I935. gall-fly's oviposition period overlapped with that of Six cases were observed in August and September E. curta, and the two species might be expected to in which either a dead gall-fly larva or dead gall-fly have been similarly affected. But whereas the puparium contained a dead larva of E. curta. These fecundity of the gall-fly fell from 70 tO 52 eggs per have been included in the o.5 E. curta found per female, that of E. curta fell from 63 to 8-4. The sq.m. to have died from unknown causes. change is far greater than can be accounted for by the One case of superparasitism was observed. Two weather. larvae of E. curta, one of them already dead, were The only other factor likely to have affected the found in a single live host. fecundity of E. curta is the availability of hosts. The mortality of E. curta in the winter of I935-6 Table A in the Appendix shows that in I935 there requires further explanation. Starting from the were I47'6 gall-fly larvae distributed amongst 240 calculated number of eggs per sq.m., the expected flower-heads per sq.m., giving a mean of o-6i number of live E. curta at the end of 1935 was suitable hosts per flower-head. For I936 Table B in 50 per sq.m. (Table 9). But in the census in the Appendix shows that there were only 28 gall-fly September a mean of only 46 was found. In the larvae in I40 flower-heads, giving a mean of 0-2 hosts following spring the number found alive, recently per flower-head. Thus the host population density killed, or parasitized was only 4-6 per sq.m., instead expressed as gall-fly larvae per flower-head is only of the expected figure 6-9. This latter figure is based one-third as great in I936 as in the previous year. on the values of destruction by mice and winter E. curta seeks its hosts by probing with the ovi- disappearance, assuming these factors to have acted positor, and probes flower-heads either with or with- equally on both E. curta and gall-fly larvae. The out any contained hosts. Hence, other conditions difference (2 3) between these figures is recorded as being equal, the rate of discovery of hosts would be missing' in Table 9. It may easily be due to expected to be about a third as great in 1936 as in random sampling errors, as the two figures are based I935. on different samples. The estimated fall in the fecundity of E. curta is Of the remaining 4 6 E. curta larvae found per not significantly bigger than the combined effect of sq.m. in June, 0o4 in the galls in the standing flower- these two factors, of which the most important is the heads had been destroyed from the outside, pre- fall in host density. This result supports the view sumably by birds (tits, Parus spp. have been seen that the difficulty in finding hosts was the main feeding on knapweed galls) and 0o7 per sq.m. factor limiting the mean number of eggs laid per suffered parasitism by various chalcids, leaving female of E. curta. The great reduction in the success 3.5 per sq.m. Of this remainder 53 % were drowned of searching in I936 apparently caused no substantial by the July floods, leaving a mean number of only increase in the number of flower-heads examined by is66 adult E. curta to emerge per sq.m. each female. The fecundity of E. curta is shown in Table 9 for Since the reproductive rate of the parasite E. curta the two generations studied. Details of the com- appears to be controlled by host density, this putations are given in the Appendix, Tables D and parasite is presumed to act on the gall-fly as a E. The values are subject to large sampling errors delayed density dependent factor. This relationship because of the small number of adults which emerged is further considered in the discussion. per sq.m. In July I 935 2o00 ? o 65 adults emerged per sq.m., and the number of eggs laid per female (2) Eurytoma robusta was estimated to be 63 ?23. In I936 i 66?0-38 The adults of this species resemble those of adults emerged per sq.m., and the number of E. curta very closely, but the female has a rather eggs laid was estimated to be only 8-3 ? 26 per longer abdomen. However, the larvae of E. robusta female. Although the standard errors of the esti- are ectophagous, and both eggs and larvae are easily mates are large, the seven-fold difference between distinguished froni those of the other species. The the mean number of eggs laid in the 2 years is dark brown eggs are laid on the third instar larvae significant (P

This content downloaded on Fri, 8 Mar 2013 13:36:49 PM All use subject to JSTOR Terms and Conditions i68 Natural controlof population balancein the knapweedgall-fly of the gall-fly in August, and the parasite larvae feed 6-7 % of the mortality in I935 and 28 % in I936. A singly and rapidly destroy the hosts. They then feed smaller percentage of larvae failed to kill the gall-fly a little on the gall-tissue, and leave a number of larvae and died themselves. In the summer of I935 small woody chips in the gall-cell (larvae of the a few larvae of E. robusta were found parasitized by related genus Harmolita are entirely photophagous). the chalcids Habrocytus trypetae and Tetrastichus. The winter is passed in the larval stage and the adults In August I936 25 % of the larvae of Eurytoma emerge in the following July and August. This robusta hatched from eggs laid on such tiny gall-fly species has been recorded as a parasite of the thistle larvae that they were only half grown when the host gall-fly Urophora cardui L., and of the gall-wasp was completely consumed. Their food supply ex- Aylax papaveris (Perris), and has been reared from. hausted, they died of starvation. the flower-heads of various species of Centaurea and There is some information about host selection. from crispus L. The female parasites which emerged in August 1936 The distribution of this species in space was very sought their hosts at a time when many were still irregular (Varley, 194 I). No trace of it was discovered very small. As noted above, these small gall-fly in the preliminary census work in thirty-six different larvae were accepted as hosts, although they were

Table io. The effect of successive mortality factors on the numbersof the chalcid parasite Eurytoma robusta found per sq.m. at Madingley No. killed No. alive per % killed per sq.m. sq.m.

I935 July No. of eggs 4'5 eggs Died due to superparasitism 6-7 0-3 4-2 larvae Failed to attack host 3 012 4-o8 Aug. Miscellaneous causes: Parasitism A 0 I2 Destroyed by caterpillars - 7 o-o6 Died due to unknown causes) O-I2 3-78 larvae

I936 Winter Winter disappearance 6I.5 2-33 I-45 , Destroyed by mice 64 0-93 0-52 , Died due to unknown causes) July Drowned in floods J 8o 042 O-o adults

July O-I adults emerged per sq.m., 50 % were females Mean number of eggs laid: 50 per female 2-5 eggs Died due to superparasitism 28 07 i -8 larvae Failed to attack host II 0-2 I-6

Aug. Died of starvation on tiny host 25 0-4 12 localities in England and Wales. In the census area unsuitable for the parasite larvae. In seven out of it was first found in the fresh flower-heads in July twenty-one instances it was possible to see that the 1935, when a few eggs and young larvae were seen. gall-fly larva had previously been parasitized by Its distribution in the various sq.m. samples was E. curta. This indicates that parasitized and un- also very patchy. 23 sq.m. were examined during parasitized hosts were accepted by E. robusta with the time when this species was available for dis- approximately equal readiness. The endophagous covery, and over half of the 83 hosts found larva of E. curta was always killed when its host was attacked were discovered in 3 adjacent sq.m. In parasitized and killed by the ectophagous larva of I936 its localization was even greater, and 32 out E. robusta. of 38 parasitized hosts were in a single one of the 20 sq.m. samples examined. (3) Habrocytus trypetae (Fig. ioB) Table io shows the changes in the numbers of live The early stages of this chalcid parasite have been Eurytoma robusta found per sq.m. described elsewhere (Varley, 1937a). The eggs are In a number of instances two or more eggs of laid in gall-cells containing larvae or puparia of the E. robusta were found in a single gall-cell of the gall-fly, or in gall-cells already containing other host, although only one parasite at most could mature parasites. When the host attacked is either a gall-fly on the one host. Such superparasitism has been pupa, or a larva of Eurytoma curta inside a puparium, discussedelsewhere (Varley, I94I). It accountedfor some of the eggs may be laid outside the puparium,

This content downloaded on Fri, 8 Mar 2013 13:36:49 PM All use subject to JSTOR Terms and Conditions G. C. VARLEY I69 and the parasite larvae die of starvation, being these will have been included in the mortality of the unable to penetrate the hard puparium. gall-fly due to unknown causes. Many eggs may be laid on a single host, but only Table i i shows the numbers of H. trypetae found one larva ever matures, because newly hatched at different times of the year. The figures for the larvae destroy any other eggs or larvae which they numbers of eggs laid per female have been calculated find. The degree of superparasitism is as great as, on the assumption that this parasite did not utilize or greater than, would be expected if the egg distri- alternative hosts outside the knapweed. The figures bution were random(Varley, I94I). are quite consistent with this view, but, as H. trypetae There may be two or three generations in the has been recorded as a parasite of various species of year. In 1935 adults emerged mainly in May, July insect,* the assumption requires justification. and September. In the colder year 1936 they In the region of the census the only gall-flies from emerged in June and September. other plants were a few Urophora stylata (Fab.) from

Table i i. The effect of successivemortality factors on the numbersof the chalcid parasite Habrocytus trypetae found per sq.m. at Madingley No. killed No. alive per % killed per sq.m. sq.m. 1935 Feb. No. of larvae found - 3-4 larvae May Observed mortality o 0 3-4 adults May 3 4 adults emerged per sq.m., 50 % were females Mean number of eggs laid: 20 per female -- 33 eggs Died due to superparasitism 64 21 12 larvae Failed to attack host 12-5 15 10-5 Parasitized 2 Died due to unknown causesL 21 0-2 8 3 adults July 8-3 adults emerged per sq.m., 50 % were females Mean number of eggs laid: 0o24 per female - I egg Died due to superparasitism 30 0o3 0o7 adults Sept. 0o7 adults emerged per sq.m., 50 % were females Mean number of eggs laid: 29 per female - I0 eggs Died due to superparasitism I5 I.5 8-5 larvae Failed to attack host 23 2 6-5 Destroyed by caterpillars 04 ?'9 5-6 1936 Winter Winter disappearance 6I-s 3-4 2-2 Destroyed by mice 64 I -4 o-8 adults June o-8 adults emerged per sq.m., 50 % were females Mean number of eggs laid: i -9 per female - - 0-76 eggs After this the mean number did not rise above i in I0 sq.m., and further analysis is superfluous.

The newly emerged female is not sexually mature, the spear thistle (Cirsium vulgare (Savi) Ten.), and it first feeds on the host without laying eggs. Xyphosia miliaria (Schr.) from the field thistle It pushes the ovipositor down the neck of the flask- (Cirsium arvense (L.) Scop.), and Icterica westermann shaped gall until it stabs a host, which is then stung (Meig.) from the ragwort (Senecio jacobaea L.). In and paralysed. The female remains motionless for the knapweed the normal host was the gall-fly some time while a secretion hardens round the ovi- Urophorajaceana, while other occupants of the gall, positor to form a tube, from which the ovipositor is such as Eurytoma curta were also attacked. However, withdrawn. Through this tube the blood of the host Habrocytus trypetae was not found attacking the exudes, and the parasite drinks it up. This method * of feeding was first described by Lichtenstein (I92l) Habrocytustrypetae has been recorded as a parasite for two other species of Habrocytus. It is also known of the gall-flies Terelliaserratulae (L.), Urophoracardui from at least three other chalcid genera (Eurytoma, (L.), Noieta pupillata (Fall.) and from the moth Sparganothis(Oenophthira) pilleriana Schiff. and it has Pteromalusand Spintherus:see Clausen (1940) for been reared from the flower-heads of species of the references). Some of the hosts of Habrocytus Composite genera Centaurea,Carduus, Cirsium, Arctium trypetae in the knapweed were killed by this treat- and Hieracium.The great difficultyin naming species of ment. The flimsy broken feeding-tube is inconspic- the genus Habrocytusperhaps makes some of the records uous, and was probably overlooked in some cases; doubtful.

This content downloaded on Fri, 8 Mar 2013 13:36:49 PM All use subject to JSTOR Terms and Conditions 170 Natural control of population balancein the knapweedgall-fly non-gall-forming trypetids in the knapweed, such 89 larvae attacked by Habrocytus trypetae, only 12, as Chaetostomella cylindrica, Chaetorellia jaceae and or I3-5 %, were also parasitized by Eurytoma curta, Urophora quadrifasciata. The first two of these whereas the percentage of gall-fly larvae containing species were indeed attacked by another quite E. curta was 45 %. Both species were found together distinct species of Habrocytus, H. albipennis (Walk.), only one-third as frequently as would be expected and Urophora quadrifasciata was parasitized by if the Habrocytus trypetae females laid eggs on another species to which Dr FerriZerewas unable to parasitized and healthy gall-fly larvae with equal give a name. This negative evidence is consistent readiness. with the assumption that this race of Habrocytus In May I935, when there were only 23 hosts trypetae is restricted to hosts in the knapweed galls available per sq.m. in the standing flower-heads which remain in the standing flower-heads, and (Table I2), H. trypetae had a choice between larvae searches for its hosts neither in other parts of the and pupae of both the gall-fly and Eurytoma curta. knapweed, nor in other plants. In this instance 82 % of the gall-flies present, and The census did not indicate whether the very 68 % of the E. curta were attacked. The preference small number of eggs laid in the knapweed in July for the gall-fly is revealed by the difference in the I935 was due to eggs being laid elsewhere. However, percentages of survivors-i8 % of the gall-flies if females reared elsewhere had laid eggs in the compared to 32 % of the E. curta.

Table I 2. The effect of host density (expressedas the numberof available hosts per sq.m.) on the host preference and the fecundity of Habrocytus trypetae When the number of hosts or the number of eggs laid is one per sq.m. or less the error in the estimates is likely to be large, and the figure must be taken as indicative only of the order of magnitude. Available hosts

- A Gall-fly larvae or pupae, Approximate Urophorajaceana The chalcid, Eurytomacurta no. of eggs laid A r- & -AI, per female of No. available No. available Habrocytus per sq.m. % attacked per sq.m. % attacked trypetae 1935 May 20 82 3 4 68 20 July o I 0I - 0-24 Sept. 64 I2 48 I.4 29 I936 June I - 0-5 - I9 knapweed in September I935 or September I936 The number of eggs laid per female was very when hosts were abundantly available, the apparent different in the different generations, and was clearly oviposition rate of the few females known to have correlated with the changes in the density of avail- emerged from the knapweed might have been very able hosts (Table I2). The estimates are less accurate large. Since it is quite reasonable to suppose that a than those for the fecundity of the gall-fly, partly single female Habrocytus could deposit 30 eggs in its because of the smaller numbers involved, and also ifetime, this provides no support for the idea that it because in I935 the three generations followed each attacks alternative hosts elsewhere. other very quickly. The July and September genera- Within the galls there are definite host preferences, tions so nearly overlapped that it was difficult in and some of the data are given in Table I2. In some cases to decide the generation to which empty September 1935 there were in each sq.m. 64 fully egg-shells were to be counted. However, although grown gall-fly larvae and 48 larvae of Eurytoma the accuracy of the estimates is low, the changes curta suitable for parasitization. Gall-fly larvae observed are so large that they cannot be attributed were preferred, and I2% of them were attacked, entirely to errors in measurement. compared with only I-4% of the Eurytoma larvae; In May and September I935 the fecundity was which is a preference of nearly ten to one. estimated to be 20 and 29, when there were about Unlike any other of the ectophagous parasites 20 and ioo hosts available per sq.m. In July I935 studied, Habrocytus trypetae even avoided laying eggs and June 1936, when the density of hosts was less on gall-fly larvae which contained small larvae of than two per sq.m., the fecundity was less than Eurytoma curta. In August, I935, eggs and larvae 2 eggs per female. It is evident that the very of Habrocytus trypetae were found on 89 gall-fly low densities of available hosts greatly reduced larvae in which the presence or absence of larvae the rate of increase of Habrocytus trypetae. This of Eurytoma curta could be established. Of the species, like Eurytoma curta, is therefore to be

This content downloaded on Fri, 8 Mar 2013 13:36:49 PM All use subject to JSTOR Terms and Conditions G. C. VARLEY I7I regarded as a potential controlling factor of the as those of Tephritis truncata (Loew) and Urophora population density of the gall-fly. cardui (L.), but no other gall-flies were common in The action of Habrocytus trypetae upon the gall- the area except U. quadrifasciata. Alternatively, the fly is complicated by three facts. First, it accepts adult Torymusmay wait from May until August, and other hosts, such as the parasite Eurytoma curta, then mature its eggs and attack the hosts. This has besides the gall-fly larvae. Secondly, it does not been shown by Flanders (935) to happen with attack those hosts in fallen galls. Thirdly, the certain other chalcids, which wait for many months emergence of this species is not synchronized with in the adult stage until hosts are available. the life history of the host. This last factor caused the In the present work Torymus cyanimus has been virtual disappearance of Habrocytus trypetae from found only in the knapweed gall-cells, and eggs were the census area during the period of observation. In laid on healthy gall-fly larvae, and on larvae parasi- 1935 the second generation of H. trypetae emerged tized by Eurytoma curta quite indiscriminately. very early, and was seeking hosts in July, when most Occasionally eggs were also laid in or on puparia gall-flies were in a stage unsuitable for parasitization containing E. curta larvae. -as adults, eggs, or very small larvae. The result The number of eggs, larvae and adults of Torymus was a reduction in the population density of H. cyanimus found in the census are shown in Table I 3. trypetae from 8-3 adults per sq.m. in July to less Owing to the possibility of an alternation of hosts, than one adult per sq.m. in September of 1935. The the fecundity of this species has not been estimated. following generation of this species was retarded in The mortality from superparasitism was very great emergence until June by the cold spring of I936, and in this species, and in addition a number of larvae only o 8 adults emerged per sq.m. Almost all the failed in their attack on the host, which survived. The hosts were then in the fallen galls and H. trypetae surviving hosts numbered sixteen in 1935 (six gall- found so few to attack in the standing flower-heads fly larvae, nine larvae of Eurytoma curta, and one of that the number of parasites never exceeded one in E. robusta), and two gall-fly larvae and two larvae of io sq.m. during the rest of the summer of I936. This E. curta in I936. catastrophic fall in the population density of this parasite was perhaps exceptional. In normal years (5) Macroneura vesicularis (Fig. io E) H. trypetae probably emerges in May and August, The life history and the larval stages of this and is able to find plenty of gall-fly larvae at a stage common and polyphagous chalcid have been de- suitable for parasitization. scribed by Morris (1938). The species was found only in the early summer in the old knapweed io (4) Torymus cyanimus (Fig. C) flower-heads, and the eggs were laid in the galls in The larva of the chalcid Torymus cyanimus is an May and June. The eggs were laid in small groups, ectoparasite of the gall-fly larva. The early stages and superparasitismwas common(Varley, I 94 1), al- have been described elsewhere (Varley, I937a). The though only a single parasite could come to maturity eggs are usually laid in August, and though some- on a single host. Superparasitism resulted in 63 % times laid singly they are commonly laid upon the mortalityin 1935. gall-fly larva in small groups (Varley, I94I). The Each egg was found under a small silken pad. The first larva to hatch usually destroys the other eggs, larvae hatched and grew quickly, and became adults but, although two larvae may feed upon the same in July and August of the same year. These adults host for a short time, only one larva comes to did not attack gall-fly larvae, but probably found maturity. Development is rapid, and in 1935 some alternative hosts in the knapweed stems, where of the eggs laid in August gave rise to adults in larvae of the gall-wasp Phanacis centaureae were September, while others passed the winter in the commonly parasitized by this species in the middle larval stage. The few small larvae found late in of the summer. September may perhaps have been the progeny of In May and June I935 Macroneura vesicularis these adults. was fairly common in the census area. Eggs or The emergence of adults from the hibernating larvae were found in 53 gall-cells, but in four other larvae takes place in May of the following year. instances M. vesicularis was found to have attacked However, no fresh Torymus eggs or larvae were larvae or puparia of the small gall-fly Urophora found on the available hosts in June or July, and no quadrifasciata. In two other instances the host was further fresh eggs or larvae of this species were seen a puparium of the related gall-fly Chaetorelliajaceae. until the next generation of gall-fly larvae became The cells formed by the knapweed gall-fly were well grown in the following August. There are two apparently its chief habitat in the knapweed at this possible explanations for this. First, Torymus may season. find some alternative host. The species has been In seven of the 53 cases the gall-fly larva recorded as a parasite of other gall-fly larvae, such was the host, and in six of these the attack was

This content downloaded on Fri, 8 Mar 2013 13:36:49 PM All use subject to JSTOR Terms and Conditions I72 Natural control of population balancein the knapweedgall-fly successful. But only one out of another seven was a pupa of the moth , attacks on gall-fly puparia was successful. In one whose larva had fed in the gall. case this species competed unsuccessfully with Torymnus cyanimus for the same gall-fly larva, (6) Tetrastichus sp. B (Fig. ioD) and in 38 cases it competed with Habrocytus Specimens of this unnamed species have been trypetae for the same gall-fly larva or pupa. Macro- deposited in the British Museum Collections. neura was the victor in ten of these cases, and Tetrastichus sp. B has at least two generations in Habrocytus in 22 cases, but in three of the latter the year, and the main emergences seem to take the Habrocytus larva was subsequently destroyed place in June and August. The adults which emerge by Tetrastichus sp. B, and in six cases both species in June attack the hosts in the old-standing flower- were unsuccessful and failed to mature. Thus heads of the knapweed, and various species are only 17 of the 5 3 attacks on hosts in the gall parasitized. The larvae are gregarious endoparasites,

Table 13. The effect of successive mortality factors on the numbersof the chalcid parasite Torymus cyanimus found per sq.m. at Madingley No. killed No. alive per %killed per sq.m. sq.m. 1934 Sept. Eggs laid per sq.m. - 9 eggs Died due to superparasitism 62 5.6 3-4 larvae Failed to attack host I5 0-5 2-9 Destroyed by caterpillars 3 OI 2-8

1935 Winter Winter disappearancenot estimated - ?28 larvae April Pupae or adults died I0 0?3 2-5 adults May 2-5 adults emerged per sq.m.

Aug. Eggs laid per sq.m. - - 7-7 eggs Died due to superparasitism 43 3-3 4-4 larvae Failed to attack host II 0-5 3 9 Sept. Died due to unknown causes 6-5 0-25 3-65 o-8 adults emerged per sq.m. - 2 85, I936 Winter Winter disappearance 6I5 1 75 I-Io Destroyed by mice 64 0?7 0o4 pupae May 0o4 adults emerged per sq.m.

Aug. Eggs laid per sq.m. - 5-6 eggs Died due to superparasitism 65 3-65 1-95 larvae Failed to attack host I0 0'2 I75 Sept. Died due to unknown causes O'I5 Killed by Habrocytustrypetael II II-55 larvae were successful, and the mortality was 68 %. The and any number between three and twenty may feed relative time of attack seems to be the chief factor in the same host. The adults emerging in August determining which of the competing parasites will usually attack gall-fly larvae in the fresh galls, and be successful, and in general the last parasite to their progeny hibermate as larvae in the dry skin of attack was the victor. In some cases larvae of the host. Macroneura were found feeding on fifth instar larvae In February I935 22 hosts were found attacked of Habrocytus trypetae, and one was found feeding by this parasite; i9 hosts were gall-fly larvae, on a Habrocytus adult which had only recently two were Eurytoma curta, and in one instance the emerged from its pupal skin, and was still in the larva of the ichneumon fly, Ephialtes buolianae, gall-cell. Habrocytus was never found feeding on a was the host. These parasites became adult in larva of Macroneura. June, and of I 4 hosts found attacked by them in In the summer of 1936 only seven gall-cells were the standing flower-heads six were gall-fly larvae, found containing Macroneura. Of the seven hosts and eight were larvae or pupae of the chalcid two were larvae and three were puparia of the gall- Habrocytus trypetae. In the next generation of fly, and one was a puparium containing Eurytoma flower-heads eight gall-fly larvae, four Eurytoma curta, which survived the attack. The seventh host curta, one E. robusta and one Habrocytus trypetae

This content downloaded on Fri, 8 Mar 2013 13:36:49 PM All use subject to JSTOR Terms and Conditions G. C. VARLEY 173 were found attacked by October. Few or no further pupates in the soil, from which it emerges as a moth instances of attack by this species were found until in the following July. October I936, when three gall-fly larvae were found Apparently the caterpillar completes its growth attacked. Perhaps the fall in numbers of this species in a single flower-head, and does not move into a was due to lack of available hosts, as was the similar second. Occasionally two caterpillars were found in fall in Habrocytus trypetae, but the parasite was not the same flower-head. common enough for much information to be available on this point. Two other tetrastichine chalcids, Aprostocetus daira Walk. and Tetrastichusbrevicornis,* were found as occasional parasites of the gall-fly larvae. The larvae of Aprostocetus daira were more commonly found as gregarious endoparasites of the related gall- fly Chaetorellia jaceae, but its larvae were not distinguishable from those of Tetrastichus B. T. brevicornis larvae were recognizable because they were not found in the skin of the dead host. They were probably ectophagous. A A

(e) Mortality due to caterpillars The caterpillars of three species of , Eucosma hohenwartiana, Metzneria metzneriella and Euxanthis straminea (Fig. i i), live in the knapweed flower-heads, and if they encounter a gall in the flower-head they almost invariably enter it. They feed on the succulent gall tissue, tunnel from one cell to another, and destroy the contents. Gall-fly larvae were occasionally found soon after they had been killed in this way, and the wounds made by the caterpillars were clearly seen. When a caterpillar has finished feeding on a gall-cell, it is left empty save for a mass of dry faecal matter. <;:~~~~~~~Sae5 m In 1935 three-fifths of the destruction of the galls by caterpillars was due to Eucosma hohenwartiana, and the remainder to Metzneria metzneriella. In I936 Eucosma was responsible for three-quarters of the caterpillar damage. Euxanthis stramineawas very rare in the census area; a few galls were found which had been damaged by it in the summer of I934 but none was found either in 1935 or I936, althoughthe species was fairly common near by. Fig. I I. Moths from knapweedflower-heads. (i) Eucosmahohenwartiana (Fig. i I A) A. Eucosmahohenwvartiana: forewings pale greyish brown, with darker markings; hindwings greyish The moths are on the wing in the last half of brown. July. In daytime they rest on the flower-heads and B. Euxanthis straminea: forewings yellowish buff, leaves of the knapweed, and the eggs are laid on the with brown central shading; hindwings pale grey. bracts of the flower-heads. The young caterpillars C. Metzneria metzneriella:forewings russet brown enter the flower-heads and feed on the contents, with grey-brownmarkings; hindwings pale grey. which they web together with silk in a characteristic If a way. caterpillar meets a gall, it enters and feeds The perce'ntage of flower-heads attacked by largely on the gall tissues, and kills any gall-fly Eucosma caterpillarswas 17-5 % in I935, and 23-5 % larvae it encounters in the cells. When fully fed in inl I936, and these were approximatelythe propor- September the caterpillar leaves the flower-head and tions of galls affected by them, inldicating that the caterpillars and the galls were distributed inde- * Tetrastichusbrevicornis appears to be new to the pendently. In the galls attackedby Eucosma 45 % British list. of the gall-cells were destroyed.

This content downloaded on Fri, 8 Mar 2013 13:36:49 PM All use subject to JSTOR Terms and Conditions I74 Natural controlof population balancein the knapweedgall-fly Very few Eucosma caterpillars were found dead in I935 which contained cocoons of Metzneria, two the flower-heads. In October I935 three dead contained live caterpillars, eleven contained live caterpillars were found in 25 I flower-heads which pupae, and the moths had emerged from nine. had contained this species, and in August and Of the 59% mortality, 23 individuals had dis- September I936 six dead caterpillars were found in appeared due to unknown causes, seven were 657 such flower-heads. parasitized by the braconid Neochelonella sulcata, Some of the caterpillars were parasitized by and one pupa had been killed by the chalcid ichneumonoid larvae. One was found attacked by Macroneura vesicularis. the ectophagous pimpline Ephialtes buolianae Hartig. The mortality was more carefully studied in the All the live caterpillars found in 1936 were dis- next year, late in the summer of I936, when all the sected and examined for internal parasites. Of 243 caterpillars found were examined for internal ichneu- examined 28 contained first instar larvae of ichneu- monoid parasites. Three dead caterpillars were monoids. There were i8 Glypta larvae (Glypta found, and dissection of 107 caterpillars found alive longicauda Hartig and G. vulnerator Gr. were both revealed I 7 first instar larvae of Neochelonella sulcata, found as adults seeking hosts in knapweed flower- two of Macrocentrus nidulator, and one of Omorga heads). Two of the Glypta larvae were dead, one of ensator, making a total of 23, or 2I % parasitized them being in the same caterpillar as a live Glypta caterpillars. Elsewhere this species was parasitized by larva. the Microbracon and Ephialtes species which also One of the 28 Eucosma caterpillars contained attacked Eucosma. the first instar larva of Omorga ensator (Gr.), and five had larvae of Macrocentrus nidulator Nees. (3) Euxanthis straminea (Fig. i i B) Three contained unknown larvae of two different species. In other localities Eucosma caterpillars This double-brooded species passes the winter as were attacked by various ectophagous ichneumonoid a caterpillar in the shoots of knapweed. In June and larvae, such as Ephialtes brevicornis (Gr.), and July caterpillars of the second generation enter the Microbracon marshalli Szepl., but none of these small flower-heads, always leaving a hole in the stem was found in the census, although they were quite just beneath the bracts. common about a quarter of a mile away. The species was present in less than i % of the flower-heads in 1935 and I936, and no galls were (2) Metzneria metzneriella (Fig. i i C) destroyed in this period. However, a small number The moths are to be seen in the daytime on the was damaged by them in the summer of I934. leaves and flower-heads of the knapweed in July, and the eggs are laid on the bracts. The caterpillars feed PART 4 on the young florets, and may also tunnel a few millimetres down into the stem. They become fully DISCUSSION AND CONCLUSIONS fed in September and October, and make a silken This study of the community in the black knapweed gallery or cocoon in which they pass the winter flower-heads has shown the numerical effect of the inside the flower-head, where they eventually pupate various biotic and climatic factors on the population in the following May. density of the knapweed gall-fly over a period of Metzneria caterpillars were found in 5 -8 % of the 2 years. By the analysis of this tangled skein of fresh flower-heads in October 1935, and in 3-8 % in relationships we wish to find an answer to the I936. These percentages are also approximately the question asked in the introduction: What factors percentage of galls attacked. If a Metzneria cater- control the population density of the knapweed gall- pillar is in a galled flower-head, it almost invariably fly in nature, and how do they operate? takes up residence in the gall, and eventually makes It has already been pointed out that controlling its cocoon there. In these flower-heads attacked by factors must be density dependent. They are the Metzneria about 8o % of the gall-cells were de- factors responsible for maintaining a balance in the stroyed, compared with 45 % for Eucosma. population density, by acting more severely when The mortality of the Metzneria caterpillars was the population density is high, and less severely quite small in the flower-heads up till October. Only when the population density is below the average. three dead caterpillars were found out of a total of However, some confusion has arisen in the literature 83 in io sq.m. in October I935. In the summer since the word 'control' has been given different of 1936 this same generation of flower-heads was meanings by different workers. By 'control' many examined, and the number of Metzneria caterpillars economic entomologists mean the maintenance of found had dropped from 8-3 to I5 per sq.m., due the population density of a pest below a level at to winter disappearance and mice. Out of a total which it does economic damage. Some now use it of 54 old flower-heads found in the summer of simply as a synonym for destruction.

This content downloaded on Fri, 8 Mar 2013 13:36:49 PM All use subject to JSTOR Terms and Conditions G. C. VARLEY 175

Bodenheimer (1938, p. 105), in a discussion of the that the mortality was rather higher in those flower- relative values of climatic and biotic factors in the heads containing many larvae than in those con- control of animal populations, claims that 'all factors taining only one or two. Thus the early larval are of destructive value in direct proportion to the mortality is density dependent, and overcrowding percentage per stage destroyed by each'. On this at this stage is potentially able to limit the popula- view it is sufficient in the case of the knapweed gall- tion density. However, the degree of competition fly to examine Table 8, in which the destructive within a single flower-head is not simply proportional value of the various factors can be seen at a glance. to the population density, but will rise only when The most important are: destruction by mice and more than one egg batch is commonly laid in the winter disappearance, which in the winter of I935-6 same flower-head. Although the percentage of destroyed 64% and 6i-5% of the gall-fly larvae flower-heads galled was 440% in 1935 and only 8% respectively; Habrocytus trypetae, which destroyed in 1936, the mean number of eggs found in flower- nearly 6o % of the gall-fly larvae and pupae in May heads was 3-02 in I935 and 3 04 in 1936. Hence the and June of I935; flooding, which caused 44% degree of competition was the same in the 2 years. destruction in July I936; and Eurytoma curta, which The population density of the gall-fly would parasitized 45X5of the young larvae in July 1935. begin to increase the larval competition within the Had any of these destructive factors not operated, flower-heads only when the percentage of flower- the population density of gall-flies emerging from heads attacked rose to well over 50 %. But in none that generation of larvae or pupae would indeed of the many localities where preliminary census work have been proportionately greater. But Boden- was undertaken was this figure reached. Clearly heimer's claim applies only to the immediate effect some other factors prevent the population density in one generation. It is not necessarily true that the rising to such a level that competition between larvae best method of reducing the average population is effective as a controlling factor. density of the knapweed gall-fly over a number of Had this crowding factor alone been responsible years would be the artificial intensification of any for the limitation of the population density, almost one of these factors. Volterra's third law, the law of I000% of the flower-heads might have been galled. disturbance of the averages, states that if a predator In the census as many as ten gall-cells were found and prey are in equilibrium, the equilibrium popula- in a single flower-head, and sixteen were once found tion density of the prey will be increased if equal under insectary conditions. Taking a fairly con- proportions of the predator and its prey are killed servative figure of I50 flower-heads per sq.m. in the in each generation. census area, population densities as high as 2400 Before it can be decided whether the intensifica- larvae per sq.m. would be possible, which is tion of a factor which causes the destruction of a sixteen times greater than the highest larval popula- pest will tend either to reduce or to increase the tion density found in the field, and 350 times greater mean population density of the pest, further facts than the largest adult population. are needed. If in addition to destroying the pest, The two potential controlling factors remaining it destroys any parasitic species which may act as are the delayed density-dependent factors Eurytoma a controlling factor, then its long-term effect may be curta and Habrocytus trypetae. Their fecundity has as Volterra predicts for predator and prey. For the been shown to be greatly influenced by changes in special case of insects which have discrete genera- the population density of gall-fly larvae, which acts tions, the interaction between one factor and another upon them as a density-dependent factor. The can conveniently be investigated by means of estimated number of eggs laid per female Eurytoma Nicholson & Bailey's theory- of balance of animal curta fell from 63 in 1935 to 8-4 in 1936 in response populations, once the necessary fundamental in- to a cold summer and a three-fold fall in the larval formation about the factors is available from suitable population of gall-flies. This is the kind of change census data. expected on the theory of Nicholson and Bailey, but The first step in the analysis is to find which of the it is possible to go further and examine the agree- potential controlling factors is in fact operating. The ment between some of the fundamental assumptions three factors which appear to have the right proper- of the theory and the field data. ties are the density-dependent early larval mortality, Nicholson and Bailey assume that parasites search and the two common chalcid parasites, Eurytoma for their hosts at random. Various meanings have curta and Habrocytus trypetae, which act as delayed been attached to the term 'random', and an attempt density-dependent factors. to clarifythis has been madeby Varley(I94I), where The early larval mortality was due partly to it is shown that parasites which seek their hosts by competition between the larvae in the same flower- random movements may produce a distribution of head before the formation of the gall. Evidence has progeny amongst the available hosts which is non- already been presented which shows at least for 1935 random in space. The data from 20 sq.m. examined

This content downloaded on Fri, 8 Mar 2013 13:36:49 PM All use subject to JSTOR Terms and Conditions I76 Natural controlof population balancein the knapweedgall-fly in the summer of 1935 show a significant correlation equal to that in about ioo flower-heads. This does between the percentage of parasitism by E. curta not seem an unreasonable figure for the whole life of and the fraction of flower-heads containing galls. a female parasite. However, this spatial discrepancy was of a kind Estimates of the area of discovery of Habrocytus which did not greatly affect the numerical result of trypetae can be made for the generations emerging parasitism. For common parasites in a limited area in May and September 1935. Insufficient data were such as the census area it is concluded that search available at other times. In May I935 the number is random to a first approximation. of adults which emerged per sq.m. was 3-4 and eggs The idea of random search has been attacked by were laid on 8o % of the available hosts, leaving o0z Thompson (I939) whose main argument seems to be as the fraction of hosts surviving. From equation (i) based on a misunderstanding of Nicholson's use of the area of discovery works out at o047 sq.m. At this the term 'random'. Thompson argues that search time the number of flower-heads still available on cannot be random since parasites tend to select the standing stems was izo per sq.m., of which definite host species for attack. But by random I8 were galled. As the proportion of females was search Nicholson & Bailey meant a search for the 0o5, the area of discovery of H. trypetae females is particular host species. Their mathematical formu- equivalent to I 13 old flower-heads. lation requires simply that the rate of discovering In September 1935 a less accurate estimate is new hosts at any instant should be proportional to possible. The number of adult H. trypetae was the product of the population densities of the estimated to be o07 per sq.m., and they attacked searching parasites and the undiscovered hosts. I2z% of the available hosts, leaving a fraction of The second basic assumption of Nicholson & o-88 survivors. The area of discovery works out at Bailey's theory is that parasites in their random o-i8 sq.m. The mean number of flower-heads in the search for hosts search an area, termed the area of ten samples on which these figures are based was discovery, the average size of which is constant and I90 per sq.m., so the area of discovery can be independent of the population densities of hosts and expressed as 8o fresh flower-heads per female. parasites. In this area all the hosts are supposed to These estimates of the area of discovery, when be found and parasitized. The area of discovery of expressed in terms of flower-heads, are not signi- E. curta can be estimated for two generations by ficantly different. In September all the flower-heads substitution of values in the formula given by contained a number of fruits, and a mass of para- Nicholson & Bailey (1935, p. 555) which can be physes between them, often surmounted by the dead written: florets. In May the remaining flower-heads were dry a - log-, (I) and dead; many were indeed dry circlets of bracts u p only, from which the contents had fallen to the where a is the area of discovery, p the population ground. Hence it might be expected that the time density of adult parasites, and ul and u the population taken to seairch flower-heads in May would be less density of host larvae before and after parasitization, than the time required in September, making the so that u/ul is the fraction of hosts which escapes area of discovery a smaller number of flower-heads parasitization. in September than in May. In I935 the population density of E. curta was It cannot be claimed that these data demonstrate estimated to be p= 2o adults per sq.m., and 45 5 0 of the accuracy of the assumptions made by Nicholson the gall-fly larvae were parasitized, leaving the & Bailey. The assumptions made in the mathe- fraction of survivors u/uL,= 0545. In I936 the figures matical theories which are used in biology are made were p = '66i adults per sq.m., and the fraction of rather for their mathematical simplicity than because hosts escaping parasitism was o073. Substituting they are believed to be exact. Indeed, it is incon- these values in the equation the area of discovery of ceivable that the area of discovery of a parasite E. curta works out at o03 I sq.m. in 1935 and o- I9 sq.m. remains absolutely constant over a very wide range in 1936. Possibly the much colder July in 1936 may of host and parasite population densities. At high have been responsible for a reduction in the area of host densities the parasite will be limited by egg discovery, but the difference between these estimates supply. However, over the narrow range of host and is not significant. Clearly the area of discovery of parasite densities observed in the census area, the E. curta is much less affected by the changes in host data for both species are in broad agreement with density than is its fecundity, in which there was a the assumptions of Nicholson & Bailey. They do not significant seven-fold change. The average value for agree nearly so well with the idea that the fecundity the area of discovery will be taken to be o25 sq.m. of a parasite is constant, which is fundamental to As half the adults of E. curta are males the area of the only other mathematical theory which can readily discovery of a female must be o 5 sq.m., so that a be applied to the interaction between such parasites female must be able to parasitize a number of hosts and hosts with synchronized generations (Thomp-

This content downloaded on Fri, 8 Mar 2013 13:36:49 PM All use subject to JSTOR Terms and Conditions G. C. VARLEY '77

son, 1924, etc.). Thompson's theory applies only of increase (F), which equals I8. Hence we can when host density is so high that the parasite's rate write 23 of increase is limited by egg supply. The parasite is p = -log F. (2) then not density dependent. a With caution, therefore, Nicholson & Bailey's Substituting values for a and F we find the steady theory may be used to study the interactions between density of adult E. curta to be I I 5 per sq.m. All the the different factors causing mortality. gall-fly larvae which are parasitized are supposed to To attempt to predict over many generations the give rise to adults of E. curta, hence p = h (F- I). course of interaction between the various species Substituting this in (2) we find the steady density of affecting the gall-flies would be useless, owing to the adult gall-flies to be unpredictable effects of weather and other factors. F h2-3 log (3) Furthermore, any errors in the calculation would be ha (F- I) cumulative. However, the theory of Nicholson & Bailey can be used more simply. It is possible to Substituting values for F and a the steady density of calculate the population density of parasite and the host works out at o68 gall-flies per sq.m. host, assuming them to be in the steady state. This This calculated estimate of the steady density of involves making a prediction only one generation gall-flies is far below the values of the population ahead, so that the effect of any errors is minimized. found in the census in the 2 years, which were 6-9 The steady state is the condition in which the gall-flies per sq.m. in 1935 and 2-0 per sq.m. in host and its parasite are in exact equilibrium. The 1936. At the same time the steady density calculated host density is such that the parasite can only just for E. curta was II-5 adults per sq.m. compared with mairytain its numbers from generation to genera- 2-0 and I-66 adults per sq.m. in the census. In the tion, and the parasite's density is such that it is just census, however, E. curta caused only 45-5 % sufficient to kill off the surplus host progeny. The mortality of the gall-fly larvae in I935, and 27 % in actual population density would be expected to I936, whereas in these calculations it has been oscillate about this value calculated for the steady assumed that E. curta was responsible for 94-5 % state, and the mean population density over a mortality. It remains to be seen what effect the other number of generations would be expected to be close mortality factors might have on the steady state. to this value. The mortality factors can be considered under The.following section of this paper is admittedly three headings: (i) non-parasitic factors acting speculative, but even if the conclusions reached have specifically on the gall-fly; (2) specific parasites little quantitative accuracy and indicate trends only, acting only on the gall-fly; (3) non_specific factors, they are still of considerable interest. acting both on the gall-fly and on its parasites. Using the data supplied by the census, it is (i) The specific non-parasitic factors either alter possible to calculate the steady state which Eurytoma the fecundity of the adult gall-flies, or cause curta could maintain if it were the only mortality mortality at some stage in the life history. The factor operating on the gall-fly larvae after gall- factors affecting the fecundity are: (a) the weather formation. The only attribute of E. curta for which during the oviposition period; (b) the amount of food a value is required is the area of discovery, which available to the adult gall-flies; (c) the population has an average value of o025 sq.m. A value is also density of gall-flies. required for the natural rate of increase of the gall- In I936 the population density of the gall-flies 'fly, which can be found at once from Table 8. In was so low that interspecific mating was common 1935 the number of gall-flies which emerged per during part of the breeding season, and about 8% sq.m. was 6 9, and they produced 147-6 suitable of the egg batches were infertile. Population density hosts per sq.m., which is a 2i-fold increase. In acted here as a density-dependent factor of a negative 1936 two gall-flies produced 28 suitable hosts type, since its destructive effect was inversely pro- per sq.m., giving a rate of increase of 14. This portional to the population density. However, the change may well have been caused by weather total effect was not large at the population densities differences. The natural rate of increase will there- observed. This factor must in fact always operate at fore be given a mean value of i 8. In the steady the beginning and end of the breeding season. state this i8-fold increase must be balanced by a Changes in weather and food supply must cause mortality of I7/i 8 = 94-5 % in each generation. The the fecundity to vary from year to year in an population density p of adult E. curta needed to irregular manner. This will prevent the population discover 94-5 % of the gall-fly larvae can be found density from reaching a steady value, even if it had from equation (i), since the term (ul/u), the ratio of a tendency to do so. the population densities of host larvae before and The mortality factors which come into this after parasitization, must be equal to the natural rate category are the egg mortality, the early larval

This content downloaded on Fri, 8 Mar 2013 13:36:49 PM All use subject to JSTOR Terms and Conditions 178 NVaturalcontrol of populationbalance in the knapweedgall-fly mortality, and any mortality due to functional Eurytoma curta and the gall-fly in proportion to the disease which has been included in 'unknown relative numbers present, their effect will be the same causes'. These factors together produced a total of as that of the other non-specific mortality factors. 30 % mortality in the summer of 1935, and 40 % in The percentages of mortality caused by all the non- I936. They have been allowed for in the estimation specific factors, whether parasitic or otherwise, can of the natural rate of increase of the gall-fly. therefore be summed together. (2) The only specific parasite of the gall-fly is The total effect of the non-specific factors can be E. curta, whose effect is already under discussion. obtained from Tables 8 and 9. In the I934-5 (3) Chief amongst the non-specific mortality generation the non-specific mortality (marked with factors which act indiscriminately on the gall-fly an asterisk) amounted to 5/35=27'4% before the and its parasites are mice, winter disappearance, and winter, followed by i8-5 % killed by mice, and caterpillars. The effects of these factors were identical 05/I7'4=3% killed by parasites. These together so far as could be determined both on the gall-fly amount to 43 %. The true figure would be higher, and on E. curta. None of these factors can be if the winter disappearance had been estimated. If regarded as density dependent, since neither mice the partly specific mortality due to Habrocytus nor caterpillars rely on the contents of the gall for trypetae is included, it brings the total non-specific their main food. mortality to 76-4 %. Another non-specific factor is summer flooding, In the I935-6 generation Table 8 shows that which Table 8 shows to have destroyed 44% of the 28/79 6=35 2 % of the gall-fly larvae were killed gall-fly pupae, while Table 9 gives the destruction by non-specific factors in August and September. of E. curta as 53 %. The difference here may perhaps Winter disappearance removed 6i5 %, mice de- be due to physiological differences between the species, stroyed 64 %, birds, Macroneura and Tetrastichus but it is too small to be statistically significant. together killed o9/52= I7%, and the summer This factor operated only in the summer of 1936. floods a further 44%, which together give a total The non-specific chalcid parasites E. robusta and mortality of 95-8 % for the non-specific factors. The Torymus cyanimus attacked the gall-fly larvae in figure works out at 96 % for the mortality of August when some of them contained small larvae of Eurytoma curta by the same factors in Table 9. Eurytoma curta. Such parasitized gall-fly larvae were In the I936-7 generation 340% non-specific attacked just as readily as healthy ones. The chalcids mortality had occurred by September. Macroneura vesicularis and Tetrastichus sp. B parasi-. The available data do not allow us to assign an tized both Eurytoma curta and the gall-fly indiscri- average value for the non-specific mortality. It is minately. Both had at least two generations in the produced by a series of diverse factors, each of year, but their effect was greatest in the early which varies in severity from year to year. But, summer, when they attacked the contents of the old except for that part caused by parasites, there is no galls on the standing stems. Habrocytus trypetae reason to suppose that the variations in severity will shows preference for the larvae and pupae of the be influenced by the population density of either the gall-fly, and so cannot be regarded as non-specific. gall-fly or E. curta. It will be instructive to find the It will be given separate consideration. effects different average values of non-specific Probably all the non-specific parasites are capable mortality might have on the balance between the of acting as delayed density dependent factors; but gall-fly and E. curta. they were not sufficiently common in the census area In the steady state the gall-fly must have a to destroy a high percentage of the gall-flies. mortality of 94-5 % to balance the i8-fold natural Although there was no evidence that either Eurytoma rate of increase. If after the attack by E. curta curta or Tetrastichus spp. used hosts outside the non-specific mortality factors destroy a fixed per- knapweed, Macroneura and possibly Torymus cyani- centage of gall-fly larvae so that a fraction x survive, mus had alternative hosts elsewhere, and so were not then E. curta will need to destroy only (i- i/i8x) entirely dependent on the gall-fly population. In the of the gall-fly larvae instead of I7/I8. The successive light of these facts, none of the latter parasites is changes in the population densities of gall-flies and considered to be acting as effective controlling agents of E. curta during one generation can be represented in the census area. In so far as they destroy diagrammatically as follows:

Larvae Surviving larvae Adults (ul) (u) Adults Successive gall-fly densities h -?hF - DDx --Dx=h

Successive E. curta densities p -*hF- D -*(hF- D) x ->(hF-D) x=p

This content downloaded on Fri, 8 Mar 2013 13:36:49 PM All use subject to JSTOR Terms and Conditions G. C. VARLEY '79 If the steady density of adult gall-flies is h per from which the values for the population densities sq.m., hF larvae will be found, F being the natural can be obtained by putting in values for a, F and x. rate of increase, i8. These larvae are attacked by Similarly, the steady densities of the gall-fly and E. curta. Assuming that the number remaining un- its parasite E. curta can be calculated for other parasitized is D, then hF-D must be parasitized, conditions in which different types of mortality and contain larvae of E. curta. The fraction x of these operate. Various conditions are compared in Table I4 survive in each case and emerge as adults. In the with the values of the steady density calculated for steady state the number of adults in successive different percentages of non-specific mortality. generations is the same, so we can write Table I4A shows the steady state which E. curta h=Dx, (4) could maintain alone. Mortality acting only on the p=(hF-D) x. (5) gall-fly (B and C) increases slightly the steady

Table I4. The theoretical effect of different types of mortality factors on the balance between the gall-fly Urophora jaceana and its chalcid parasite Eurytoma curta The naturalrate of increase of the gall-fly is taken as I8, and the area of discoveryof Eurytomacurta 025 sq.m. Steady density of the Steady gall-fly density Percentage -_ __A of adult of gall-fly Adults Available Eurytoma larvae per sq.m. larvae curta parasitized

A. Eurytomacurta acting alone o-68 I2-2 11-5 94-5 Mortality acting only on the gall-flies: B. Before Eurytomaattacks (I) 50 % mortality I0 8-8 89 (2) 90% mortality 2-9 5.3 2-4 45 (3) 92% mortality 3'3 4.8 I.5 3I C. After Eurytomaattacks (I) 50% mortality 0-55 I0 8-8 89 (2) 90% mortality 0-29 5-3 2-4 45 (3) 9Z% mortality o-26 4.8 I.5 31 D. Mortality acting only on Eurytoma: (I) 50 % mortality IP35 244 I I 5 94.5 (2) 90% mortality 6-8 I22 II-5 94*5 (3) 92% mortality 8.5 I53 II15 94 5 E. Non-specific mortality acting on gall-flies and Eurytomaafter its attack: (I) 50% mortality 20 8*8 89 (2) 90% mortality 2-9 53 2-4 45 (3) 92% mortality 3'3 6o I.5 3I (4) Over 94'5 % mortality 0 0 0 - F. Census data for comparison: I935 6-9 147 I-9 42 1936 2 28 I.5 27

The change in the population density of the gall- densities of available gall-fly larvae, but decreases fly from hF to D per sq.m. is due to parasitism by the steady density of E. curta. The result (B) is in E. curta. Substituting for ul/u in equation (i) we accordance with conclusion 7 of Nicholson (1933, can write hF p. I50). Mortality acting only on E. curta leaves p= 2'.3 log-. (6) a D its steady density unchanged, but greatly increases that of the gall-fly (D). Compare Nicholson's con- The unknown D can be eliminated from equations clusion 5. (5) and (6) by substituting D = h/x from Hence (4). The most important conclusion is that increasing we have two equations percentages of non-specific mortality (E) acting p 2 3 log Fx, (7) equally on both species, increase the steady popula- a tion density of gall-flies, and decrease the steady population density of E. curta. This is in accordance (8) hF= x with Volterra's law of disturbance of the averages. J. Anim. Ecol. i6

This content downloaded on Fri, 8 Mar 2013 13:36:49 PM All use subject to JSTOR Terms and Conditions i 8o Natural controlof populationbalance in the knapweedgall-fly A comparison between the calculated values and on the gall-fly. A series of formulae similar to those the census data (F) shows a striking agreement. used for Eurytoma curta can easily be derived to fit With a non-specific mortality of 92 % the calculated the life history of Habrocytus trypetaeand the changes steady density of gall-flies is 3-3, compared with 6-9 in host availability. It is enough here to state the and 2zo in I935 and I936. The number of available general conclusions reached, which are: larvae per sq.m. is 6o, compared with I47 and z8 (i) If H. trypetae alone acts on the gall-fly, and in the census. The population density of E. curta is all the gall-fly larvae remain available for parasitiza- I5 per sq.m., compared with z2o and I-66, and the tion, the calculated steady density of the gall-flies is percentage of parasitism by E. curta is 3I % com- independent of the number of generations of H. pared with 45 and Z7 % in the z years. Thus with trypetae per year. Since some gall-fly larvae are not 92 % non-specific mortality acting with E. curta all in fact available in early summer, the emergence of the calculated figures are between the census figures H. trypetae at this time reduces its efficiency as a for the z years. In none of the other conditions controlling agent. Thus, according to this applica- considered in A-D of Table I4 is this so. tion of Nicholson & Bailey's theory, a parasite with This correspondence between the calculated two or more generations to every one of the host steady densities and the census data has been cannot be more efficient than another species with achieved by giving the non-specific mortality a fixed the same area of discovery which has one generation value of 92 %. Doubtless the mean figure for this synchronized with the period of maximum host mortality must be less than 960%, which was the availability. If the parasite with many generations estimate for I935-6, since with an i8-fold natural commonly emerges at times when a proportion of rate of increase this would eventually cause the hosts are not suitable for parasitization, then, extinction of both gall-fly and its parasites. Possibly though it may be able to increase more rapidly in a the absence of the gall-fly from some localities may high host population, it may be very inefficient as a be due to high values of non-specific mortality. controlling agent. The effects of the remainder of the observed (2) The critical factors for the success of H. mortality must now be considered. Table 8 shows trypetae in the field are the proportion of the hosts that after gall-formation the factors killing gall-fly available, and the proportion of the non-available larvae which remain are Habrocytus trypetae, the hosts which survive. predaceous gall-midge Lestodiplosis miki, and un- (3) If H. trypetae and Eurytoma curta both attack known causes. Their percentage values are small the gall-flies, the result of the competition between compared with the total non-specific mortality. them depends on the host selection of Habrocytus Unknown causes resulted in 5 % empty gall-cells trypetae. The areas of discovery of the two species in I934, 2 % in 1935 and 4-3 % in 1936. This factor are about equal. Hence if H. trypetae attacked only would have a very small effect of the kind shown in gall-fly larvae as hosts, it would in time be com- Table 141B-increasing slightly the steady density of pletely displaced by Eurytoma curta. On the other the gall-flies, and decreasing those of the larvae and hand, if Habrocytus trypetae attacked Eurytoma curta of Eurytoma curta. and the gall-fly with equal readiness, it would itself After the winter of I936 unknown causes de- displace that species. The co-existence of both species stroyed z6 % of the gall-fly larvae. Although this in the field is presumably due to the variable host same factor did not operate on E. curta, Table 9 preference of Habrocytus trypetae. The preference shows that at this time 33 % of the Eurytoma were for the gall-fly is partly overcome when suitable hosts recorded as 'missing'. If two separate factors destroy are scarce, as in May I935. the same percentage of host and parasite, the effect This application of Nicholson & Bailey's theory to on the balance is the same as if a single non-specific the study of the steady state has produced a series of factor had been responsible for the destruction of calculated values for the population densities of the both species. This mortality can therefore be added gall-fly and of Eurytoma curta which agree closely to the non-specific mortality, and would increase its with the values found in the census. This result has effect. been achieved by the substitution in the formulae of Habrocytus trypetae has been shown to be poten- values estimated from the census data for both the tially a controlling agent, but in the census area its natural rate of increase of the gall-fly, and the area influence was too small to be effective. During the of discovery of E. curta. A value for the non- course of the census the numbers of this species fell specific mortality has been used which is rather less catastrophically, owing to its lack of synchronization than that found in the only complete year studied. with the period of host availability, and its failure to By thus combining theory and measurement, two attack thc3e hosts in the fallen galls. There is no major conclusions have been reached: First, that in need to publish detailed results of the calculations the census area the chalcid parasite E. curta was the which have been made for the action of this species factor primarily responsible for the control of the

This content downloaded on Fri, 8 Mar 2013 13:36:49 PM All use subject to JSTOR Terms and Conditions G. C. VARLEY i8i population density of the knapweed gall-fly, although any time must be a complex function of the factors it caused only a small proportion of the whole operating during the previous years, and cannot be mortality. Secondly, that most of the remaining related solely to the conditions of the immediate mortality factors, which together destroyed the past. greater proportion of the gall-flies, had the long-term Schwerdtfeger (I935) gives annual estimates of effect of increasing the average population density of the population densities of four moths which are the gall-fly. This was because they killed E. curta pests of the German coniferous forests. Since i88o, and the gall-fly indiscriminately, and hence reduced when the records begin, all four species have inde- the efficiency of E. curta as a controlling agent. pendently shown great peaks of abundance, in which Up till now consideration has been given only to the population density is hundreds or even thousands the steady state. In nature the steady state is not of times as great as that in the minima in between. realized because environmental conditions are not In these maxima the trees are largely defoliated. steady, and affect the balance in various ways. Thus Here intraspecific and interspecific competition for the weather influenced the behaviour of the gall- a limited food supply apparently acts as a check to flies, and probably altered their fecundity. Tem- oscillations which might otherwise increase inde- porary summer flooding killed many gall-fly larvae finitely in magnitude. and some of the parasites. The weather in the early Were there a natural tendency for the oscillations spring and summer greatly altered the times of in the population density of the gall-fly to increase emergence of Habrocytus trypetae in relation to the in amplitude, the only factor known which could availability of gall-fly larvae or pupae suitable for check this increase is the density-dependent early parasitization. The fraction of flower-heads which larval mortality. This, as we have seen, would be fell to the ground differed in different winters, thus effective only at very high population densities. But varying the fraction of larvae subjected to winter high population densities have not been observed in disappearance and destruction by mice. Changes of the field. Preliminary census work in over thirty these kinds at times favour the gall-fly, while at different localities in England and Wales showed no other times one or more of the parasites may find sample with more than 48 % of the flower-heads conditions particularly suitable. This will result in containing galls. Eleven samples gave o %. Eight more or less irregular changes in the population gave percentages between i and i0. Seven were densities of the various species. between io and zo%, four between 2o and 30%, The theories of Nicholson & Bailey, of Lotka, and four between 30 and 40 %, and three localities only of Volterra suggest that any disturbance of the had between 40 and 48 % of the flower-heads steady state will lead to periodic oscillations in the galled. The expected period of oscillation is just population density of host and parasite. Calculation over 4 years. Of the 26 samples in which the shows that in the gall-fly such oscillations should have gall-fly was found, perhaps one-quarter represent a periodicity of a little over 4 years (Nicholson & peaks of population density. This suggests that the Bailey, 1935, p. 585). Opinions differ as to whether, peaks of oscillation lead at most to some 40 % of the under constant conditions, such oscillations are flower-heads being galled, after which the population damped (Lotka), or are of constant amplitude density falls again. (Volterra), or increase in amplitude (Nicholson & The observed change in the percentage of galled Bailey). But unless the primary assumptions of these flower-heads from 44 to 8 % in the census area may theories are known to be exact, such long-term be interpreted as a glimpse of these oscillations. predictions are of little objective value. Besides, it Considerable differences have been noted both in can be shown with Nicholson & Bailey's theory that the population density and in the specific composi- if a proportion of hosts is not available to parasitism, tion of the population of the knapweed flower-heads oscillations will be damped instead of increasing in at places only a few hundred yards apart. Hence the amplitude. Gause (I934) has shown experimentally rate of dispersal of the insects concerned is not that oscillations between Paramecium and its pre- sufficiently rapid either to equalize populations, or to dator Didinium are damped by provision of situations synchronize oscillations over these distances. Very in which Paramecium is partly protected from possibly the changes in the population density in a Didinium. Perhaps the irregular distribution of the large area will be comparable with the changes in gall-fly larvae in space tends to damp the inherent level of a choppy sea. Each point may show more or oscillations between it and Eurytoma curta. In those less regular oscillations in level, but the oscillations square metres in which there were few gall-fly observed at different points may not be in phase with larvae, the percentage of them parasitized by E. each other. curta was below the average(Varley, 194I). How- To obtain information on these points will require ever, in a fluctuating environment such oscillations an uninterrupted study of io years or more. This cannot in any case be regular, and the population at present communication, embracing as it does only

I2-2

This content downloaded on Fri, 8 Mar 2013 13:36:49 PM All use subject to JSTOR Terms and Conditions I82 Natural controlof population balancein the knapweedgall-fly z years of detailed census work, must be regarded as was estimated to be o025 sq.m. Using Nicholson & preliminary and incomplete. However, since the Bailey's theory, the steady density of gall-flies which work was first broken off in 1937 no opportunity has it could maintain if it was the only factor operating arisen to continue it on a sufficient scale. The results after gall-formation was calculated to be o-68 adult are therefore presented as they stand in the hope gall-flies per sq.m. This is well below the observed that the methods employed may have wider applica- figures. tion in the fields of ecology and economic ento- (c) The chalcid parasite Habrocytus trypetae. This mology, and that they may serve to stimulate further species has two or three generations in the year, and intensive studies of this and other animal com- adults emerged at times when few or no suitable munities. hosts were available. This resulted in a great fall in its population density, and prevented it from being SUMMARY an effective controlling factor. i. A detailed study has been made of the insect 8. The effect of the non-specific mortality, killing community living in the flower-heads of the black as it does an equal fraction of the larvae of the gall- knapweed, Centaurea nemoralis (Compositae). The fly and those of Eurytoma curta, is to decrease the problem considered here is: What factors control efficiency of this parasite as a controlling agent. It the population density of the knapweed gall-fly in thus increases the steady density of the gall-flies and nature, and how do they operate? their larvae. Using constants derived from the 2. A solution is found by the application of census data, and using a figure of 92 % for the non- Nicholson & Bailey's theory of 'balance of animal specific mortality, the calculated steady density of populations' to the results of a detailed census of a the adult gall-flies is 3-3 per sq.m., and the per- series of 92 square-metre plots from an area near centage of parasitism is 3 I. These figures are Cambridge over a period of 2 years. intermediate between the observed figures for I935 3. The population density of the gall-flies was and I936. estimated to be 6-9 flies per sq.m. in July I935, and 9. Combining all the major mortality factors, this about 2 0 per sq.m. in I936. The changes in the theory gives calculated steady densities of host and population density per sq.m. are given in Table 8. parasite in good agreement with observed facts. 4. The fecundity of the gall-flies was estimated Some reliance may therefore be placed on its analysis from the census data to have been 70 eggs per of the effects of the different types of factor. All female in I935, and 52 eggs per female in the colder mortality factors do npt necessarily tend to reduce July of I936. These results are considered in relation the mean populations on which they act. to experiments on the effects of mating, nutrition, Io. Regular oscillations about the steady state are temperature and humidity on the fecundity of the inherent in the parasite-host relationship. These flies, and in relation to detailed observations of the oscillations may not be of ever-increasing amplitude, behaviour of the flies in the field. as supposed by Nicholson & Bailey, but may in fact 5. The cause of the mortality in the eggs and young be damped if some hosts are less available to larvae is discussed. The young larvae form galls in parasites than others. the knapweed florets. i i. In a fluctuating environment the oscillations 6. Mortality after gall-formation was partly due cannot be regular, and the census data are inter- to the chalcid parasites Eurytoma curta and Habro- preted as a glimpse of irregular oscillations occurring cytus trypetae. Subsequent mortality was largely about the steady state. non-specific, and killed these parasites and the I2. This use of Nicholson & Bailey's theory remaining gall-fly larvae indiscriminately. These supplies for the first time an analysis of the mutual factors, including caterpillars, mice, winter dis- effect of parasitic and other factors of destruction on appearance, other parasites and summer flooding, the population density of an insect. Its possible killed between them 96 % of the gall-fly larvae in the applications to other problems of insect ecology, and I935-6 generation. to biologic'al and insecticidal control, are obvious. 7. The controlling factors which keep a popula- tion in balance must be affected in their severity of action by the population density on which they act. ACKNOWLEDGEMENTS Three factors were found to be so affected. This work was carried out at the Entomological (a) The early larval mortality. If this competitive Field Station in Cambridge under the supervision of factor had operated alone, a population density of Dr A. D. Imms, F.R.S., to whom I am greatly 2400 flies per sq.m. might have been reached. indebted for much help and advice. The work was (b) The chalcid parasite Eurytoma curta. Its made possible by a Research Fellowship at Sidney fecundity was reduced by a fall in the population Sussex College, and I wish particularly to thank the density of the gall-fly larvae. Its area of discovery Master and Fellows for their encouragement.

This content downloaded on Fri, 8 Mar 2013 13:36:49 PM All use subject to JSTOR Terms and Conditions G. C. .VARLEY I83 Many systematists have given much time to the Newmarket who named the gall-flies. Dr W. B. naming of the species encountered in this work. I Turrill of Kew was kind enough to name the speci- am especially grateful to Dr Ch. Ferriere, Mr J. F. mens of knapweed submitted to him. Last, but not Perkins, and Mr J. E. G. Nixon of-the Natural least, it is a pleasure to thank my friend David Lack History Museum, on whom fell the heavy burden of for his many helpful criticisms of the typescript. naming the Hymenoptera, and to Mr J. E. Collin of

APPENDIX

Table A. Data from sq.m. samples nos. 23-46, collected between 30 July and zz October I935 No. of flower-heads Total no. of containing gall-fly flower-heads eggs or larvae No. of gall-cells 342 76 *(i8i) 375 56 (I07) 430 94 208 390 I12 254 224 63 1I72 236 33 86 306 54 I02 405 84 I9I 200 47 94 270 6i I45 436 42 99 332 135 422 383 I30 300 264 98 267 I33 54 I47 I5 6 13 43 I0 23 207 53 II9 I64 68 I7I 79 27 68 88 13 37 136 31 90 73 '7 36 225 83 203 Sum 5756 1447 3247 Mean ~~~~~~~~5756 I447 3247 Mean _ = 239 83 44=760-29 147-591 24 24 22 Sum of squares of deviations 385,I93 30,030 213,766 No. of degrees of freedom N 23 23 2I Estimated standard error 26-4 7-37 21-5

This content downloaded on Fri, 8 Mar 2013 13:36:49 PM All use subject to JSTOR Terms and Conditions I84 Natural controlof population balancein the knapweedgall-fly

Table B. Data from sq.m. samples nos. 64-92, collected between 28 July and 6 October I936

No. of gall-cells given for samples nos. 73-92. Some gall-fly eggs present in samples up to nio. 72, collected I4 August. Total no. No. of flower-heads of flower-heads infested by gall-flies No. of gall-cells (176) 20 (102) 6 (59) 0 (io6) 5 (132) 7 (20I) 3 (I78) 28 (230) 21 (156) 2I All galls complete 103 7 17 Aug. I9 302 23 53 228 II 33 III 10 22 136 Io I9 '39 I7 3.7 I9I 7 I5 75 7 I5 319 41 I04 130 21 50 I17 13 28 131 II 24 II0 7 9 I04 6 I2 I30 9 20 I33 I0 27 128 I5 40 52 3 5 8I 6 Io 74 12 20 Sum 2794 357 562 Mean ~~~~~~2794 357 562 Mean 2794 = 13977 123I2 28-I 20 29 20

Sum of squares of deviations 93,860-2 2,I68-2I 9,286-2 No. of degrees of freedom N I9 28 I9 Estimated standard error 15.7 I-6 4-9

Table C. Number of gall-cells in the year-old galls found in the summer of I936 in sq.m. samples nos. 47-82

62 69 II 33 I6 43 0 13 62 32 27 29 97 57 22 15 46 76 87 47 23 II9 I84 28 49 100 12 29 77 72 I55 83 74 6i 25 II9 EX 2045. X 56-8055. Squares of deviations 60,836. Estimated S.E. 6-96. Mean no. of gall-cells = 56-8 ? 7-0.

This content downloaded on Fri, 8 Mar 2013 13:36:49 PM All use subject to JSTOR Terms and Conditions G. C. VARLEY I85

Table D. Emergenceand fecundity of Table E. Emergenceand fecundity of Eurytoma curta in I935 Eurytoma curta in I936 The number of E. curta found in sq.m. samples nos. The number of E. curta found in the year-old galls in 14-21, 23-27 gives directly the number to emerge. the summer of 1936 in sq.m. samples nos. 47-82, excluding those killed by factors in Metre no. Larvae Pupae Emerged Total operating i9S5.

I4 2 . . 2 I0 3 0 I I5 . . . 0 I I 0 0 I6 . 7 * 7 2 I 4 5 I7 * I 3 4 3 3 4 0 i8 . . . 0 I 14 II 3 I9 . . . 0 3 I5 8 I 20 . . 2 2 0 8 3 I 2I . . . 0 2 7 I5 9 23 * . 5 5 0 2 2 I6 24 . . . 0 25 * 4 4 I59 26 . . 2 2 Mean 159/36 = 4-42. Sum of squares of deviations 8Iii. 27 . . . 0 No. of degrees of freedom N= 35. Estimated s.E. s = o 8o. 26 Effectively emergence had ceased by the end of July. Sq.m. samples nos. 67-82 examined after July showed Mean 26/1I3=20. Sum of squares of deviations =66. that out of 85 Eurytoma which had survived the No. of degrees of freedomN= I2. EstimatedSE. s = o'65. winter, 32 had emerged. The fraction of emergence The number of eggs laid per sq.m. can be estimated 32/85 = 0o377 + 0o053. from the total number of available hosts per sq.m. Combining these figures, the number of E. curta which = I47-6 ?21-5 minus 3 per sq.m. died of unknown emerged per sq.m. =4-42x o 377 = i66 + 038. causes, leaving I44-6, and the fractionof these parasitized The number of eggs laid per sq.m. can be estimated by Eurytomacurta. from the total number of available hosts per sq.m. The fractionparasitized was estimatedonly from those =28-o0+4 9 minus I 2?+o 4 per sq.m. which died, hosts in which the presence or absence of E. curta could leaving 268 ? 4-9 hosts per sq.m. be ascertained. The figures were: II 3 I E. curtaobserved The fraction of available hosts parasitized w4s in 2479 hosts, giving a fractionparasitized = 0457 00-I0. ? I07/396 = 0-270 ? 0-022. The number of eggs laid per sq.m. The number of eggs laid by E. curta per sq.m. is given = 144-6 X 0-457 = 66 + io. by 26-8 x 0270=7-2 ?+4. The proportionof femaleswas estimatedat o 52 ? 005. Since the proportion of females is 052? o0os, the The number of eggs laid per female number of eggs laid per female =66/(2-o x o52)=63? 23. = 7-2/(I 66 x 0-52) = 84 ? 2-6.

Table F. Frequency distribution of live and dead gall-fly eggs in 1935 No. of eggs No. of live eggs Total no. in flower- , of flower- head 0 I 2 3 4 5 6 7 8 9 Io II 12 heads I 3 26 ...... 29 2 I 5 32 . . , ...... 38 3 0 I 6 29 ...... 36 4 0 0 I 4 i8 23 5 0 0 0 0 I 7 . . . . 8 6 0 0 0 0 I 0 4* 5 7 0 0 0 0 I 0 I 3 * * * * 5 8 0 0 0 0 0 0 0 0 2 . . . . 2 9 I 0 0 0 0 0 0 0 0 0 . . . I 10 0 0 0 0 0 0 0 0 0 0 0 . . 0 II 0 0 0 0 0 0 0 0 0 0 0 0 . 0 12 0 0 0 0 0 0 0 0 0 0 0 0 I I Dead eggs 40/447=oo8g.

This content downloaded on Fri, 8 Mar 2013 13:36:49 PM All use subject to JSTOR Terms and Conditions i86 Natural controlof population balancein the knapweedgall-fly

Table G. Frequency distribution of live and dead gall-fly eggs in I936 No. of eggs No. of live eggs Total no. in flower- A - of flower- head o I 2 3 4 5 6 7 8 9 heads

I 6 i6 ...... 22 2 3 0 I5 . . . . . i8 3 2 3 2 II ...... I8 4 0 I 0 I 9 . . . . . II 5 I 0 0 0 2 6 . . . . 9 6 0 0 0 0 I I 4 . . . 6 7 0 0 0 0 0 0 I 2 . . 3 8 0 0 0 0 0 0 0 0 0 . 0 9 0 0 0 0 0 0 0 0 0 I I Dead eggs 41/267=0- 53.

REFERENCES Blair, K. G. (I93I). [No title], Proc. S. Lond. Ent. Nat. Hendel, F. (1927). 'Trypetidae', in Lindner, E., 'Die Hist. Soc. 193I-32 [recte I930-3]: 78. Fliegen der Palaearktischen Region', Pt. 49: I-22I. Bodenheimer, F. S. (I938). 'Problems of animal Hering, M. (I935). 'Markische Tierwelt.' Berlin. P. ecology.' Oxford. I69. Bond, W. N. (1935). 'Probability and random errors.' Kloet, G. S. & Hincks, W. D. (I945). 'A check list of London. British insects.' Stockport. Boyce, A. M. (I934) 'Bionomics of the walnut husk Lichtenstein, J.-L. (192I). 'Sur la biologie d'un fly, Rhagoletis completa.' Hilgardia, 8, no. II: 363- chalcidien.' C.R. Acad. Sci., Paris, 173: 733-5. 579. Lotka, A. J. (1925). 'Elements of physical biology.' Buxton, P. A. & Lewis, D. J. (I934). 'Climate and Baltimore. tsetse flies: laboratory studies upon Glossina sub- MacLagan, D. S. (1932). 'Effect of population density morsitans and tachinoides.' Philos. Trans. (B), 224: I75- upon rate of reproduction with special reference to 240. insects.' Proc. Roy. Soc. (B), III: 437-54. Cannon, H. G. & Grove, A. J. (I927). 'An aerating and Mellanby, K. & Buxton, P. A. (I935). 'A toxic emana- circulating apparatus for aquaria and general use.' tion from rubber.' Nature, Lond. 136: 873. J. R. Micr. Soc. 47: 3I9-22. Morris, K. R. S. (1938). 'Eupelmella vesicularis Retz. Chapman, R. N. (1928). 'The quantitative analysis of (Chalcididae) as a predator of another Chalcid, environmental factors.' Ecology, 9: 11-22. Microplectron fuscipennis Zett.' Parasitology, 30: Clausen, C. P. (I940). 'Entomophagous insects.' New 20-32. York, etc. Nicholson, A. J. ('933). 'The balance of animal Collin, J. E. (1937). 'Trypeta vectensis sp. n. and other populations.' J. Anim. Ecol. 2: I32-78. new or little known British species of Trypetidae Nicholson, A J. & Bailey, V. A. (I935). 'The balance (p iptera).' Ent. Rec. 49 (suppl.): 1-7. of animal populations. Part I.' Proc. Zool. Soc. Lond.: Collin, J. E. (1946). 'On the attempts to select as geno- 55I-98. types species originally included under Meigen's Otter, G. W. (I938). 'On the morphology of the larvae " i8oo " generic names.' Ent. Mon. Mag. 82: 142-3. of three species of Cecidomyidae (Diptera) from Crombie, A. C. (1945). 'On competition between knapweed (Centaurea) flowers.' Trans. R. Ent. Soc. different species of graminivorous insects.' Proc. Roy. Lond. 87: 39-68. Soc. (B), 132: 362-95. Pearl, R. (1925). 'The biology of population growth.' de Bach, P. & Smith, H. S. (194I). 'Are population New York. oscillations inherent in the host-parasite relation?' Phipps, C. R. & Dirks, C. 0. (1932). 'Dispersal of the Ecology, 22: 363-69. apple maggot.' J. Econ. Ent. 25: 576-82. Fisher, R. A. (1934). 'Statistical methods for research Sachtleben, H. (1927). 'Beitrage zur Naturgeschichte workers.' Edinburgh, etc. sth ed. der Forleule, Panolis flammea Schiff. (Noct., Lep.), Flanders, S. E. (I935). 'An apparent correlation be- und ihrer Parasiten.' Arb. biol. Abt. (Anst.-Reichs- tween the feeding habits of certain Pteromalids and the anst.), Berl., 15: 437-536. condition of their ovarian follicles (Pteromalidae, Schwerdtfeger, F. (1935). 'Studien iuber den Massen- Hymenoptera).' Ann. Ent. Soc. Amer. 27: 438-44. wechsel einiger Forstschaidlinge. II....' Z. Forst- u. Gause, G. F. (1934). 'The struggle for existence.' Jagdw. 47: 5 I3-40. Baltimore. Schwerdtfeger, F. (I936). 'Beitriage zur Kenntnis des Glaser, R. W. (1923). 'The effect of food on longevity Kiefernspinners, pini L., und seiner and reproduction in flies.' J. Exp. Zool. 38: 384-4I2. Bekampfung.' Mitt. Forstwirt. Forstwiss. 7: I69-242.

This content downloaded on Fri, 8 Mar 2013 13:36:49 PM All use subject to JSTOR Terms and Conditions G. C. VARLEY I87

Seguy, E. (I934). 'Dipteres (Brachyceres) (Muscidae parasitic on the knapweed gall-fly.' Proc. R. Ent. Soc. Acalypterae et Scatophagidae).' Faune de , 28: Lond. (B), 6: 122-30. I-832. Varley, G. C. (1937b). 'The life-history of some Smith, G. S. Graham- (I9I1). 'Some observations on Trypetid flies, with descriptions of the early stages the anatomy and function of the oral sucker of the (Diptera).' Proc. R. Ent. Soc. Lond. (A), I2: 109- blow-fly (Calliphora erythrocephala).' J. Hyg., Camb. 22. II: 3go-408. Varley, G. C. (1941). 'On the search for hosts and the Smith, H. S. (I935). 'The role of biotic factors in the egg distribution of some chalcid parasites of the knap- determination of population densities.' J. Econ. Ent. weed gall-fly.' Parasitology, 33: 47-66. 28: 873-98. Varley, G. C. & Butler, C. G. (1933). 'The acceleration Snodgrass, R. E. (1924). 'Anatomy and metamorphosis of development in insects by parasitism.' Parasitology, of the apple maggot, Rhagoletis pomonella Walsh.' 25: 263-8. J. Agric. Res. 28: 1-36. Volterra, V. (I926). 'Variazioni e fluttuazioni del Thompson, W. R. (1924). 'La theorie mathematique numero d' individui in specie animali conviventi.' de l'action des parasites entomophages et le facteur du Mem. Accad. Lincei, 2: 3i-84. (The substance of this hazard.' Ann. Fac. Sci. Marseille (2), 2: 69-89. work is available in English in Volterra, V. (1928), Thompson, W. R. (1939). 'Biological control and the J. Cons. Int. Explor. Mer, 3 (I): I-5I; and in an theories of the interactions of populations.' Parasi- Appendix to Chapman, R. N. (I93I), 'Animal tology, 31: 299-388. ecology, with especial reference to insects.' New Uvarov, B. P. (1931). 'Insects and climate.' Trans. Ent. York, etc. Pp. 409-48.) Soc. Lond. 79: 1-247. Volterra, V. (193I). 'Levons sur la theorie mathe- Vanderplank, F. L. (1947). 'Experiments in the matique de la lutte pour la vie.' Paris. hybridisation of tsetse-flies (Glossina, Diptera) and-the Wadsworth, J. T. (1914). 'Some observations on the possibility of a new method of control.' Trans. R. Ent. life-history of the knapweed gall-fly Urophora solstilialis Soc. Lond. 98: i-i8. Linn.' Ann. Appl. Biol. I: I42-69. Varley, G. C. (1937a). 'Description of the eggs and larvae of four species of Chalcidoid Hymenoptera

This content downloaded on Fri, 8 Mar 2013 13:36:49 PM All use subject to JSTOR Terms and Conditions