Heredity 65 (1990) 67—74 The Genetical Society of Great Britain Received 2 January 1990

Maintaining a happy face: stable colour in the Theiridion grallator (Araneae, )

Rosemary G. Gillespie and Departments of Zoology and Entomology, Bruce E. Tabashnik University of , Honolulu, HI 96822, U.S.A. Theridiongrallator is a highly polymorphic spider endemic to the wet and mesic forests of the Hawaiian Islands. The frequencies of the two major morph classes, patterned and unpatterned abdomen, did not show significant spatial or temporal variation within undisturbed or disturbed areas on . Similarly, morph frequencies did not vary significantly between undisturbed areas on three different islands. Estimates of spider migration suggest that gene flow is sufficient to account for the similarity in frequency within and between areas on Maui, but it is probably not sufficient to explain the similarity among islands. Fecundity did not differ between morphs. A significant inverse relationship between morph frequency and residence time in certain cases suggests that frequency-dependent selection, perhaps mediated by bird predation, may play a role in maintaining the polymorphism.

INTRODUCTION 1988c). Morph frequencies of this species exhibit temporal stability, but considerable variability Polymorphismhas long fascinated ecological between colonies only short distances apart geneticists, as it offers the opportunity to study (Oxford, 1976; Reillo and Wise, 1988b). Oxford evolution in action. It may be caused by certain and Shaw (1986) found that morph frequencies of types of selection (heterosis, temporally fluctuating E. ovata in their study site in England (to which or frequency-dependent selection), migration, or the species is native) are converging towards a stochastic processes (Ford, 1975; Kimura, 1983). general equilibrium. Some type of weak selection In disturbed environments polymorphisms may is implicated, in addition to migration and genetic reflect historical selection and present neutrality drift. At a local level, Reillo and Wise (1988a, b) and/or novel directional selection. Numerous found no evidence for selection on morph frequen- cases of transient and stable polymorphism of cies in the eastern U.S. (where the spider has been genes affecting pigmentation have now been introduced); they attributed patterns of morph found in diverse groups, including snails (Clarke frequency variation to the effects of migration and et a!., 1978), butterflies (Turner, 1977), moths, genetic drift. beetles (Bishop and Cook, 1980) and The only studies that have identified selective (Gunnarsson, 1985; Oxford, 1976; Oxford and agents responsible for maintaining a polymorph- Shaw, 1986; Reillo and Wise, 1988a, b). ism in spiders are those of Gunnarsson (1985, Colour polymorphism in spiders is a wide- 1987). Gunnarsson examined colour morphs of spread phenomenon, but few studies have Pityohyphantes phrygianus (Araneae: Linyphiidae) attempted to analyse the factors responsible for its and showed that morph frequencies are balanced presence or maintenance. One of the two spiders by opposing selective pressures: melanic forms in which this problem has been addressed is the have an activity advantage over non-melanics at small theridiid, ovata. This spider low temperatures; at the same time, this renders exhibits three distinct morphs (Locket and them more vulnerable to predation. Millidge, 1953; Hippa and Oksala, 1979; Oxford, The present study examined colour variability 1976), which are controlled by three alleles at an in the Hawaiian happy face spider, gral- autosomal locus (Oxford, 1983; Reillo and Wise, lator (Araneae, Theridiidae), a small (<5 mm 68 R. G. GILLESPIE AND B. E. TABASHNIK

long) resident of wet (annual rainfall 200 to METHODS 350 cm) and mesic (annual rainfall 100 to 200 cm) forests of the Hawaiian Islands (Gon, 1985). It has Study organism long, slender legs and a smooth translucent yellow Tgrallator inhabits the underside of leaves of a body. The base colour of the abdomen is pale variety of , especially the native translucent yellow, but a variety of abdominal arguta (Saxfragaceae) and Clermonlia arborescens colour patterns may be superimposed on this back- (Campanulaceae). It is also found on the intro- ground (Gon, 1985). The polymorphism is control- duced ginger, coronarium (Zing- led by one locus, with alleles for patterned morphs iberaceae). The webs are inconspicuous, scantily dominant over alleles for the unpatterned (Gilles- covering the underside of a leaf. Feeding is pie and Tabashnik, 1989). Patterns may be red, primarily nocturnal, and the most common prey maroon, black or white, and vary considerably in items are dipterans of the families form and extent. For the purpose of this paper, and (Gon, 1985). During daylight the morphs were categorised as either unpatterned hours spiders are usually found flat against the (no superimposed pigment) or patterned (some underside of a leaf. Generally, except for the case superimposed pigment). of maternal females with offspring, a single In this study we examined temporal and spatial individual only is found under a leaf. T grallator patterns of morph frequencies in T grallator, and exhibits an annual life cycle. Egg sacs are mostly possible factors responsible for the presence and laid between May and July, with young dispersing maintenance of the polymorphism. (1) Temporal in the later summer (see Gon, 1985; Gillespie and and spatial variation in morph frequencies at Tabashnik, 1989 for additional information). different subsites within one site that included native and disturbed forest. Similar morph frequencies here could suggest selection, but Fieldsites migration between subsites could not be ruled out. Undisturbedsites were located in Maui (the We therefore also estimated migration between Nature Conservancy of Hawaii's Waikamoi Pre- subsites, and morph frequencies between isolated serve, elevation 1360 m), (the Nature sites. (2) Migration within a site, between subsites. Conservancy of Hawaii's Kamakou Preserve, ele- If this was sufficiently high, it could explain vation 1110 m) and Hawaii (Thurston, in Vol- similarities in morph frequencies between subsites. canoes National Park, elevation 1190m). These (3) Variation in morph frequency between islands, were all sites of native Hawaiian wet-to-mesic in undisturbed sites. Because gene flow among forest, with a canopy dominated by Metrosideros islands is expected to be low, similar frequencies polymorpha (Myrtaceae) and Acacia koa on different islands would imply that selection was (Leguminosae); the sub-canopy by Broussaisia important and relatively uniform across islands. arguta (Saxifragaceae), Clermontia arborescens Variation among islands would suggest genetic (Campanulaceae), Cheirodendron trigynum drift, differential selection pressures, or a combina- (Araliaceae), Coprosma sp. (Rubiaceae), hex tion of both. We also measured: (4) Fecundity of anomolum (Aquifoliaceae), Myrsine spp. (Myr- unpatterned versus patterned females. We sinaceae) and Pelea spp. (Rutaceae). examined the hypothesis that differences in A disturbed site was located in an area of Maui offspring numbers play a role in maintaining the (Makawao State Forest, near Waikamoi Preserve, polymorphism. This could occur if females of one elevation 1350 m) which had many introduced morph type lay more eggs than another. In plants (grasses, blackberry and ginger in par- the final part of the study we tested the possible ticular), mammals (especially pigs, mongoose role of frequency-dependent selection in main- and mice) and birds (including Leiothrix lutea, taining the polymorphism: (5) Relationship Lonchura punctulata and Zosterops japonica). between morph frequency and residence time. T grallator was found here on the introduced We used natural low-level fluctuations in morph ginger, H. coronarium. frequency to test the stability of the polymor- phism, using residence time (deviations from the average) as an indicator of survival. Shifts from Temporaland spatial variation in the equilibrium morph frequency should result morph frequencies in lowered survival for the morph that was in Thesite on Maui (with undisturbed and disturbed excess, if frequency-dependent selection were areas) was divided into subsites of 100—200m2 operating. separated by 300-900m, which reflected the COLOUR POLYMORPHISM IN THE SPIDER THER/DION GRALLATOR 69 clumped distribution of the spider's preferred host Variation in morph frequency between islands plants, B. arguta and C. arborescens in undisturbed Morphfrequencies of T grallator (patterned ver- forest, and H. coronarium in the disturbed area. sus unpatterned) were determined at undisturbed In the three undisturbed subsites (labelled 1—3) sites on three islands. This allowed us to determine and the four disturbed subsites (labelled 4—7), the ubiquity and stability of the polymorphism in morph frequencies were determined from a 1-day sampling period of 50 individuals at each subsite largely isolated areas. T grallator were located by between March and April 1988. Two subsites thorough scrutiny of the leaves of all of the larger- (undisturbed subsite 1 and disturbed subsite 4) leaved plants in a site. Spiders were censused by going through the area systematically, turning over were monitored from September 1987 until April each leaf in turn and one time only. Data were 1988. The proportion of unpatterned to patterned collected from a single day's sample at each site. morphs (n =50-70individuals) was determined in The colour and size of individuals (both mature the middle of each month. Chi-square tests were and immature) were recorded for different used to determine differences in morph frequencies species to estimate the frequency of the different between different sites and subsites. morphs. Chi-square tests were used to test for heterogeneity in morph frequencies among sites. We also calculated allele frequencies for each Migration island, and used the frequencies to estimate FST Migrationinto subsites on Maui was estimated in and Nm as described above. Here Nm represents both undisturbed and disturbed areas. In the the number of migrants per generation between undisturbed area, a clump of B. arguta was located islands that would be needed to account for the at approximately 300 m from subsite 1, T grallator observed morph frequencies on the different inhabitants were removed, and the undersides of islands, assuming that the polymorphism is six leaves were coated with StickemTM, a non- neutral. drying sticky substance. Similarly, in the disturbed area, a clump of H. coronarium was located approximately 600 m from subsite 4, any T. gral- Fecundityof unpatterned versus lator were removed, and the underside of six leaves patterned females coated with Stickem. Trapped spiders were picked off the leaves every second day from September Weexamined the possibility of differences in 1987 to August 1988. When sticky leaves senesced, fecundity between patterned and unpatterned the leaf closest to them was coated with Stickem, morphs. All females with egg sacs (in both distur- so that six coated leaves were always present at bed and undisturbed areas) were marked and each sampling area. monitored. The number of offspring produced by Based on previous evidence that morph type unpatterned and patterned females were counted is controlled by a single locus, with alleles for the immediately after the eggs hatched. patterned morphs dominant to unpatterned (Gillespie and Tabashnik, 1989), and the assump- tion that the alleles are in Hardy—Weinberg equili- Relationshipbetween morph frequency and brium, we calculated allele frequencies for each residence time subsite. We used these allele frequency estimates to calculate FST, the fixation index, which is the Usingnatural fluctuations in morph frequencies, heterozygosity of a subpopulation due to random we examined the possibility of frequency-depen- dent selection by testing whether the frequency of genetic drift (Harti, 1988). We then estimated Nm, a given morph in a population affected its residence the number of migrants per generation, from FST time at a subsite. Because we could not separate as follows: emigration from mortality, we use the term "resi- Nm=(1/FST—1)/4 (Hartl, 1988). dence" rather than "survival". At any given subsite morph frequencies fluctu- These calculations assume that selection does not ated by approximately per cent. We examined influence the polymorphism; i.e., gene flow is the the relationship between morph frequency and only factor restricting differentiation among sub- morph residence time. A significant negative populations. Nm is the number of migrants that relationship would suggest that frequency-depen- would be needed to account for the observed dent or temporally varying selection plays a role morph frequencies at the different subsites. in maintaining the polymorphism in T grallator. 70 R. G. GILLESPIE AND B. E. TABASHNIK

Tests were done in (a) an undisturbed area, nor among four subsites in a disturbed area (Chi- and (b) a disturbed area: Square=44, df=3, P =0.22) on Maui (table 1). (a) The undisturbed area was a 5 m2 area in Further, the mean frequency of unpatterned subsite in the Waikamoi Preserve in Maui. Three morphs did not differ significantly between undis- plants of B. arguta, occupied by a large number turbed (mean 722±45 per cent) and disturbed of spiders (40—50 per cent of the leaves occupied) (mean 667±75 per cent) areas (t=0.57,df=5, were selected in October 1987, and the spiders P =059).Variation in morph frequencies during (n =44)were marked with a small dot of paint on a 7-month period was not significant at a single one leg to allow identification of individuals. undisturbed (Chi-Square =263,df= 6, P =0.85) Spiders were repainted following moulting. When or disturbed (Chi-Square=243, df=6, P=0.88) the apparent disappearance of a painted individual subsite (fig. 1). from a subsite was coupled with replacement on its leaf by a similar unmarked individual, we assumed that the former had moulted into the Table 1 Morph frequencies indisturbed and undisturbed latter. Spider densities remained relatively con- areas on Maui stant through January 1988, but began to decline Genefreq. to very low levels in Spring (5-8 per cent of the UnpatternedPatterned leaves occupied). From February to April 1988 Subsite n q2 p2+2pq q p populations were composed almost entirely of mature . Undisturbed Subsite 1 15 080 020 089011 (b) The highly disturbed area was a 10 m2 area Subsite 2 31065 035 080020 in subsite 5 in the Makawao State Forest, elevation Subsite 3 430'72 028 085015 1350 m, where T grallator was found at high Mean 072 028 085015 density (40-45 per cent of the leaves occupied) on Disturbed the introduced ginger, H. coronarium. Spiders were Subsite 1 22073 027 085015 marked and monitored from the beginning of Subsite 2 60045 055 067033 November 1987 until late January 1988. Subsite 3 20070 030 084016 Subsite 4 24079 021 089011 Tests were divided into one week intervals. Mean 067 033 081019 At the outset of each week, the frequency of un- patterned and patterned morphs was determined. Individuals were then monitored every 2-3 days in the subsequent 7-day period to determine how long they remained in the subsite. Tests were con- Migration ducted on alternate weeks from November 1987 The data from the sticky leaves revealed some to March 1988. immigration into subsites from which T grallator We used regression analysis (Sokal and Rohif, had been removed. On B. arguta, a total of 14 1981) to test for a significant relationship between T. grallator (comprising 026 per cent of all morph frequency and residence time at each sub- caught) were trapped between site. At the undisturbed subsite, separate analyses September1987 and August 1988 (1.8 were conducted for periods when spider density immigrants/leaf/year). None of these was marked. was high (40-50 per cent leaves occupied; October Except for one mature female, and one immature, 1987 through January 1988) and when it was low all were mature males, 50 per cent being trapped (5—8 per cent of the leaves occupied; February to between March and May, the peak of sexual April 1988). activity. On H. coronarium, a total of six T. grallator (0.06 per cent of all arthropods caught) were trap- ped between December and August (1.4 RESULTS immigrants/leaf/year). Again, none were marked. Temporal and spatial variation in Five were mature males. We calculated FST, heterozygosity due to ran- morph frequencies dom genetic drift, and estimated Nm, the number Morph frequencies did not show significant spatial of migrants that would be needed to account for or temporal variation at undisturbed or disturbed the observed morph frequencies at the different areas on Maui. Morph frequencies did not vary subsites. The estimate for Nm at undisturbed sub- significantly among three subsites in an undis- sites was 23 (F7- =0.0108);at disturbed subsites turbed area (Chi-Square =04,df= 2, P =082) the Nm value was 5 (FaT =OO464). COLOUR POLYMORPHISM IN THE SPIDER 71

Undisturbed Disturbed

f0

U 0 p a t re 0 e d Sep. 15 Oct. 14 Nov. 14 Dec. 16 Jan. 16 Feb. 17 Mar. 17 Sampling Date

Figure 1 Morph frequency fluctuations over a 7-month period. The percentage of unpatterned morphs at an undisturbed subsite (Broussaisia) and a disturbed subsite (Hedychium) from Sep. 1987 to Mar. 1988.

Variation in morph frequency between islands Relationship between morph frequency and residence time Morphfrequencies of T. grallator in undisturbed areas did not vary significantly among the islands Undisturbedarea of Maui, Molokai and Hawaii (Chi-Square = When the population density was high (40-50 per 026,df= 2, P =088,table 2). cent leaves occupied), the proportion of un- patterned morphs ranged from 700 to 905 per Table 2 Morph frequencies at undisturbed areas on different cent. For each morph, residence time was nega- islands tively correlated with relative abundance (fig. 2). Island n Unpatterned Patterned When the population density was low (Site) (February to April 1988; 5—8 per cent of the leaves occupied), there was no significant relationship Maui 41 068 032 between residence time and morph frequency (for (Waikamoi) R2 = F[17] = Molokai 35 066 034 patterned morphs, slope =—28, 014, (Kamakou) 1.10,P=0.33; for unpatterned morphs, slope= Hawaii 22 062 038 —26, R2=0•41, F[17J=184,P=0.21). (Thurston) Disturbed area The estimated value for FST was 00015, which The population density here ranged from 10—25 yielded an estimate of 169 for Nm. These results per cent of leaves occupied. The proportion of imply that a gene flow of 169 migrants per island unpatterned morphs ranged from 430 to 650 per per generation would be required to maintain the cent. Residence time showed no correlation with degree of similarity in morph frequency observed relative abundance for either morph (fig. 3). among islands, assuming that the polymorphism was not influenced by selection. DISCUSSION of unpatterned versus Fecundity Bothpatterned and unpatterned colour morphs patterned females have been found in all populations of T grallator Therewas no significant difference in the number studied to date (Gillespie and Tabashnik, 1989). of offspring between unpatterned (mean 1600, The current study found no significant spatial vari- SD: 10.93) and patterned females (mean 2030, ation between or within islands in the frequencies SD: 12.55) (t-test: t =060,df= 16, P =0.56). of these morphs. Further, no significant temporal 72 R. G. GILLESPIE AND B. E. TABASHNIK

100 .0

90 'V 80 0

4, 70 E

04, 60 C 4,

4, 50

40 1 .0

Morph Frequency

Figure 2 Effect of morph frequency on residence time of unpatterned and patterned morphs in populations in an undisturbed area at high spider density. The percentage of the 7-day period that a spider remained at the subsite is shown for the separate morphs, as a function of the morph frequency at the beginning of the time period (forpatterned morphs, for unpatterned morphs). Residence times demonstrated a significant inverse relationship to their frequency for both morph types. For patterned morphs, slope=—2156, R2=076 (ANOVA: Fj191=25.69, P<0.001). For unpatterned morphs, slope=—788, R2=045 (ANOVA: F[1,7]— 560, P<005).

100 0 L * 4, . a. 90 .0 * 'V * N 80 *0 0 0 •**. 4, 70 0 E * 4, 60 C(5 4,

4' 50

40 0.0 0.2 0.4 0.6 0.8 1 .0

Morph Frequency Figure 3 Effect of morph frequency on residence time of unpatterned and patterned morphs in populations in a disturbed area. • for patterned morphs, for unpatterned morphs. Regression analysis: For patterned morphs, slope =—601, R =019, ANOVA: I1Io]=2•43,P>0•10;for unpatterned morphs, siope=—192, R2=003, ANOVA: F[IloI=031, P>010. COLOUR POLYMORPHISM IN THE SPIDER THERIDION CRALLATOR 73 variation was detected in 7-month studies of two in an undisturbed area where spider density was populations. How can we explain both the high. If we assume that emigration was indepen- existence of the polymorphism and its stability? dent of morph type and frequency, this might Repeated founder events and genetic drift suggest frequency-dependent selection. have been used to explain the presence of the One of the most likely potential agents that polymorphism in the spider Enoplognatha ovata, could induce fluctuating selection pressure on T. and its considerable spatial variability on a local grallator is bird predation. The development of a scale (Reillo and Wise 1988a). E. ovata bears a search image is a common phenomenon in birds striking resemblance to T grallator in its colour (Clarke, 1962; Allen, 1974, 1976; Murdoch and polymorphism (Gillespie and Tabashnik, 1989). In Oaten, 1975; Atkinson and Warwick, 1983; Green- T grallator, the finding of similar morph frequen- wood, 1984; but see Guilford and Dawkins, 1987). cies in different subsites on Maui could be As a result, their foraging intensity in a given area explained by extensive gene flow between these may vary with the frequency and/or abundance of populations. Estimates of Nm indicate that 23 their prey. Drepanidine birds (Fringillidae) coexist immigrants per population per generation would with T. grallator in Hawaii. They are obligate or be needed to account for the observed morph facultative gleaners (Scott eta!., 1986; Pimm frequencies at the undisturbed subsites, five at the and Pimm, 1982) and have long been recognized disturbed subsites. Our estimates from the sticky as important predators of spiders (Perkins, 1913). leaves indicate that immigration from all sources They have been observed searching the undersides into undisturbed and disturbed subsites was exten- of B. arguta leaves on several occasions (personal sive (1.8 immigrants/leaf/year on B. arguta. 14 observation). It may be that when patterned immigrants/leaf/year on H. coronarium), which morphs of T. grallator are rare, they are favoured; suggests that gene flow between subsites is prob- as they become more numerous, birds develop a ably sufficient to account for the similarity in search image towards them, and they suffer higher morph frequencies between subsites. predation. Further, populations of T gral!ator It is highly unlikely, however, that gene flow exist in discrete patches. In such situations, avian among islands is sufficient to maintain the degree predators tend to confine their foraging effort to a of similarity in morph frequencies observed among single patch before moving to another (Oaten, islands. Assuming that gene flow is the only factor 1977). Therefore, the search image of a bird could restricting differentiation in morph frequency be significantly altered by changes of morph among islands, the number of migrants needed to frequency of T. grallator even within a single account for the observed morph frequencies patch. appears to be unreasonably high (Nm =169 We should give some consideration to the spiders per island per generation). These results absence of any significant relationship between the imply, therefore, that the similarity in morph frequency of a morph and its residence in disturbed frequency among islands is maintained, at least in areas and in areas where spider densities were low. part, by selection. These results further suggest that One possible explanation would be reduced preda- similar selection occurs on each island. tion. In disturbed areas spiders are found on H. Types of selection that could maintain the poly- coronarium, a plant with large, slippery leaves that morphism in T grallator include balancing selec- may well restrict predation. Further, when a prey tion, heterosis, frequency-dependent selection and species is very scarce, predator search images tend temporally varying selection. In E. ovata, sig- to be lost (Murdoch and Oaten, 1975). Thus, when nificant fecundity differences have been detected spider density is low, frequency-dependent selec- between morphs in certain populations, and may tion may be weak or absent. On the other hand, play a role in maintaining the polymorphism of there was no significant difference in morph this species by balancing selection on other frequency between undisturbed and disturbed sub- features of a morph (Reillo and Wise, 1988b). sites. This could be explained by migration However, no differences in fecundity were detected between the undisturbed and disturbed subsites. in T grallator. We have shown here that the polymorphism of We examined the possible role of selection in T. gral!ator is ubiquitous and morph frequencies the polymorphism of T. grallator through the effect appear to be balanced. Frequency-dependent of naturally varying morph frequencies on resi- selection is implicated in both the exhibition and dence period. The results indicated a significant stability of the polymorphism. However, alterna- negative relationship between the frequency of a tive explanations cannot be ruled out. Further morph at any one time and its residence, but only study is needed to definitively pinpoint the 74 R. G. GILLESPIE AND B. E. TABASHNIK mechanisms responsible for maintaining the poiy- GUNNARSSON, B. 1985. Melanism in the spider Pityohyphantes phrygianus C. L. Koch: The genetics and the occurrence morphism. of different color phenotypes in a natural population. Heredity, 59(1), 55—62. Acknowledgements The study was inspired and instigated GUNNARSSON, B. 1987. Phenotypic variation in dark coloration through the work of Sam Gon. It would have been impossible in Pityohyphantes phrygianus (C. L. Koch) (Araneae: without an enormous amount of help from the Nature Conser- Linyphiidae). Bull Brit. ArachnoL Soc., 6(9), 369-374. vancy of Hawaii: Rob Rydell encouraged the initial efforts; HARTL, D. L. 1988. A Primer of Population Genetics, 2nd edn. Alan Holt, Paul Higashino and Mark White allowed progress Sinauer Assoc., Inc., Sunderland, Mass. to continue throughout the year; the work on Molokai would HIPPA, H. AND OKSALA, i. 1979. Colour polymorphism of have been impossible without the help of Ed Misaki. Dr Lloyd Enoplognatha ovata (Clerck) (Araneae: Theridiidae) in Loope and Art Medeiros (Haleakala National Park) gave western Europe. Hereditas, 90(2), 203-212. immense support and encouragement throughout the study. KIMURA, M. 1983. The Neutral Theory of Molecular Evolution. Thanks also to Bill Mull for invaluable advice and discussion, Cambridge University Press, Cambridge. to Geoff Oxford and Jay Rosenheim for useful comments on LOCKET, G. H. AND MILLIDGE, A. F. 1953. British Spiders, Vol. the first draft, and to John Bordley and the University of the 1. Ray Soc. London. South for use of computer equipment. Chris Parrish helped in MURDOCH, W. W. AND OATEN, w. 1975. Predation and popula- completion of the study. The work was funded by Whitehall tion stability. Adv. Ecol. Res., 9, 1—131. Foundation grant J86-47, with additional support from the OATEN, A. 1977. Optimal foraging in patches: a case for stochas- Nature Conservancy of Hawaii, the National Park Service, and ticity. Theor. Pop. Biol., 12, 263-285. a Fujio Matsuda Scholar Award from the University of Hawaii OXFORD, G. s. 1976. The colour polymorphism in Enoplognatha Foundation. ovatum (Clerck) (Araneae: Theridiidae). Heredity, 36, 369—38 1. OXFORD, G. 5. 1983. Genetics of colour and its regulation REFERENCES during development in the spider Enoplognatha ovata (Clerck) (Araneae: Theridiidae). Heredity, 51, 621—634. ALLEN,.1. A. 1974. Further evidence for apostatic selection by OXFORD, G. S. AND SHAW, M. W. 1986. Long-term variation in wild passerine birds: training experiments. Heredity, 33, colour-morph frequencies in the spider Enoplognatha ovata 361—372. (Clerck) (Araneae: Theridiidae): natural selection, migra- ALLEN,J. A. 1976. Further evidence for apostatic selection by tion and intermittent drift. Biol. J. Linn. Soc., 27, 225-249. wild passerine birds: 9-1 experiments. Heredity, 36, 173— PERKINS,R.C. L. 1913. Introduction (to Fauna Hawaiiensis). 180. In Sharp, D. (ed.) Fauna Hawaiiensis Vol. 1: xv-cxxviii, ATKINSON, W. U. AND WARWICK, T. 1983. The role of selection Cambridge Univ. Press, Cambridge. in the colour polymorphism of Liltorina rudis Maton and P1MM, S. L. AND P1MM, .i. w. 1982. Resource use, competition Littorina arcana Hannaford-Ellis (Prosobranchia: Lit- and resource availability in Hawaiian honeycreepers. torinidae). BioL J. Linn. Soc., 20, 137-151. Ecology, 63, 1468-1480. BISHOP, J. A. AND COOK, L. M. 1980. Industrial melanism and REILLO, P. R. AND WISE, U. H. 1988a. An experimental evalu- the urban environment. Adv. Ecol. Res., 11, 373-404. ation of selection on color morphs of the polymorphic CLARKE, B. 1962. Balanced polymorphism and the diversity of Enoplognatha ovata (Araneae: Theridiidae). Evolution, 42, sympatric species. In Nichols, D. (ed.) and 1172— 1189. Geography, Systematics Association, Oxford, pp. 47-70. REILLO, P. R. AND WISE, D. H. 1988b. Temporal and spatial CLARKE, B., ARTHUR, W., HORSLEY, U. T. AND PARKIN. U. T. patterns of morph frequency variation among coastal 1978. Genetic variation and natural selection in pulmonate Maine populations of the polymorphic spider Enoplog- molluscs. In Fretter, J. and Peake, J. (eds) Pulmonates, natha ovata (Araneae: Theridiidae). Am. Mid. Nat., 119, Vol. 2A, Systematics, Evolution and Ecology, Academic 337—3 54. Press, London. REILLO, P. R. AND WISE, D. H. 1988c. Genetics of color FORD, E. B. 1975. Ecological Genetics, 4th edn. Chapman and expression in the spider Enoplognatha ovata (Araneae: Hall, London. Theridiidae) from coastal Maine. Am. Mid. Nat., 120, GILLESPIE, R. G. AND TABASHNIK, B. E. 1989. What makes a 3 18-326. happy face? Determinants of colour pattern in the SCOrr, J.M.MOUNTAINSPRING,S., RAMSEY,F. L. AND Hawaiian happy face spider Theridion grallator. Heredity, KEPLER, C. B. 1986. Forest bird communities of the 62, 355—363. Hawaiian islands: Their Dynamics, Ecology and Con- GON III, S. M. 1985. Comparative Behavioral Ecology of the servation. Studies in Avian Biology, 9. Cooper Ornithologi- Spider, Theridion grallator Simon, (Araneae: Theridiidae) cal Society, Los Angeles. in the Hawaiian Archipelago. Ph.D. Dissertation, U.C. SOKAL, R. R. AND ROHLF, F. J. 1981. Biometry: The Principles Davis. and Practice of Statistics in Biological Research. Freeman GREENWOOD, J. .1. D. 1984. The functional basis of frequency and Co., San Francisco. dependent food selection. Biol. J. Linn. Soc., 23, 177-199. TURNER, j.R.G. 1977. Butterfly mimicry: the genetical evol- GUILFORD, T. AND DAWKINS, M. s. 1987. Search images not ution of an adaptation. In: Hecht, K., Steere, W. C. and proven: a reappraisal of recent evidence. Anim. Behav., Wallace, B. (eds) Evolutionary Biology, Vol. 10, Plenum 35, 1838—1845. Press, New York.