Vol. 14, No. 5, May 1999 Review The geography of behaviour: an evolutionary perspective Susan A. Foster sfoster@clarku.edu Trends in Ecology & Evolution 1999, 14:190-195 Dept of Biology, Clark University, Worcester, MA 01610-1477, USA

Abstract
Traditional approaches to the study of behaviour have typically assumed that behavioural patterns, especially elements of reproductive behaviour, are invariant within species. Recent research on a diversity of behavioural traits in a wide array of taxa provides evidence that genetically based geographic variation in behaviour is common. Comparisons of populations that display geographic variation in behaviour can offer substantial insight into mechanisms of adaptive divergence and constraining or generative roles of gene flow, initial stages of speciation, and the roles of phenotypic plasticity and ontogeny in determining patterns of behavioural evolution.

In ethology, there exists a rich tradition of comparative research designed to explore patterns of adaptation and evolutionary change. Almost exclusively, these research programmes have employed interspecific comparisons in which the behavioural characterization of the species was based on single populations and the behavioural patterns were assumed to be 'species-typical'; that is, invariant within the species ¹. Although population and quantitative geneticists have demonstrated extensive geographic variation in nonbehavioural traits ^1,2 , evolutionary studies of geographic variation in behaviour have entered the literature slowly. Only now are we coming to recognize the ex-tent of such variation and its value for elucidating evolutionary mechanisms ¹.

When geographical variation in behaviour has been explored, the goal has most often been to infer the selective causes of apparently adaptive differences among populations. Population contrasts can be especially valuable for this purpose, because divergent populations might have been separated for less time than higher taxonomic units, and they might more often reside in the habitats where the differences evolved ². There might also be fewer traits that differ between populations than between species, with the consequence that there is also less opportunity for covarying traits to confound interpretation of adaptive value ³. Comparison of appropriately selected populations can also increase our understanding of the initial stages of speciation ^4–6 and can offer unique insights into the interplay between genes and environment in the ontogeny of population differences in behaviour ⁷.


Divergence in behaviour among populations

Characteristics distinguishing gene pools considered to be subsets of the same species are likely to differ from those distinguishing species because populations tend to have had shorter periods of independent evolution. In particular, novelties (derived behavioural patterns not detected elsewhere within the clade) should appear in population clades less often than in higher-order clades. This is because genetically based behavioural novelties probably appear rarely and, when they do, can often be disadvantageous ⁸. In contrast, loss of ancestral behavioural traits might be frequent because they can be disrupted by mutations at multiple loci; a suggestion that has been made for other complex traits ⁹. However, loss could be restricted to those character states that have evolved recently or that have not become deeply embedded in neural circuitry (which would negatively affect other behavioural patterns if disrupted) ^10,11 .

Comparisons of behaviour in a large number of postglacial freshwater populations of threespine stickleback (Gasterosteus aculeatus) offer evidence of such a pattern ⁸. Not one novel motor pattern (relative to the ancestral repertoire) has been detected, whereas strong patterns of adaptive differentiation reflecting loss of ancestral behavioural patterns or shifts in the frequency of expression of the traits have been demonstrated ( Box 1). Research has revealed similar patterns in a wide array of taxa involving a diversity of behavioural phenotypes ( Table 1). Indeed, I am aware of only one case in which a derived, complex behavioural pattern has been detected in population comparisons: the recent appearance of a novel migratory pattern in one population of the blackcap warbler Sylvia atricapilla ¹². Even this example might be considered equivocal given that changes in the timing and direction of migratory orientations need not be considered novel motor patterns, but instead simply changes in timing and orientation. Despite this, when comparisons are made among populations, there are clearly many more cases in which investigators have shown that ancestral behavioural patterns have decayed than there are cases in which they consider novel behavioural patterns to have been acquired. This pattern contrasts with that observed in higher-order comparisons ^1,8,10,11 .

Table 1. Examples of behavioural traits displaying population differentiation
Taxon	Behaviour	Difference a	Refs b
Insects
Drosophila melanogaster	Territorial success	S	13
Drosophila paulistorum	Courtship elements	FS	32
Callosobruchus maculatus	Migratory frequency	S	33
Ephippiger ephippiger diurnus	Syllabic content of male calls	S	34
	Female preference for male calls		35
Jadera haematoloma	Mate guarding duration (reaction norm)	S	23
Nilaparvata bakeri	Structure of male and female calls	S	36
Oncopeltus fasciatus	Migratory tendency	Q	37
Spiders
Agelenopsis aperta	Diet breadth	S	14
	Intensity of conspecific agonistic behaviour	S	14
	Territory size	S	14
	Antipredator retreat probability	S	14
	Effect of predator on feeding latency	S	14
Fish
Phoxinus phoxinus	Predator inspection	S/L?	38
	Schooling tendency	S/L?	38
	Response to alarm substance	S	38
Poecilia latipinna	Male mating behaviour	F	39
Poecilia reticulata	Predator effect on foraging rate and time	S	15
	Predator effect on male courtship	FS	15
	Male courtship composition	FS	15,16
	Male courtship intensity	S	15,16
	Female preference for male colour	S	15,16
	Intensity of conspecific aggression	S	15,38
	Schooling intensity, cohesion and orientation	S	15,38
	Predator effect on schooling behaviour	S	15,38
	Strength of antipredator behaviour	S	15,38
	Aquatic predator inspection and avoidance	S	15,38
Gasterosteus aculeatus	Intensity of conspecific aggression	S	40
	Migratory tendency	L	41
	Diversionary display behaviour	L	8
Amphibians
Acris crepitans	Advertisement call	S	42
Desmognathus ochrophaeus	Reproductive behaviour	S	28,30
Gastrophryne carolinensis	Advertisement call	S	43
Geocrinia laevis	Advertisement call	S	43
Geocrinia victoriana	Advertisement call	S	43
Hyla chrysoscelis	Advertisement call	S	44
Limnodynastes tasmaniensis	Advertisement call	S	43
Littoria ewingii	Advertisement call	S	43
Littoria verreauxii	Advertisement call	S	43
Pseudacris nigrita nigrita	Advertisement call	S	43
Pseudacris feriarum	Advertisement call	S	43
Reptiles
Thamnophis elegans	Dietary preferences	S	3
Thamnophis melanogaster	Method of prey attack	FS	45
Thamnophis ordinoides	Antipredator behaviour	S	46
Thamnophis sirtalis	Defensive behaviour	F	47
	Dietary preferences	S	47
Birds
Cistothorus paulustris	Song repertoire size and recurrence interval	S	48
Molothrus ater	Song recognition and/or preference of females	S	7
Sylvia atricapilla	Migratory route	S	12
Zonotrichia leucophrys	Accuracy and range of dialect tutor imitation	S	24
	Migratory tendency		24
Mammals
Peromyscus maniculatus	Climbing behaviour	S	19
Saimiri sciureus	Reactivity	S	49
Spermophilus beecheyi	Antipredator behaviour	L	11

Shifts in the frequency of expression of behavioural patterns, or in the threshold levels of stimuli that elicit them, could be common because there often seems ample genetic variation within populations to permit evolution of the range of variation observed among populations ^13,14 . Results from population studies suggest that comparisons at this level are probably most valuable to those interested in understanding the evolution of quantitative characters or the processes by which complex characters decay. In the case of stickleback, population comparisons indicate that both loss and shifts in expression can rapidly produce adaptive differentiation.


Adaptive inference

Population comparisons provide some of our best insights into the causes and mechanisms of adaptive differentiation, especially when the geographical distribution of ecotypes is both complex and unlikely to follow a cline in selective regimes ². However, even with an ideal scenario, in which parallel, independent evolution of similar behavioural ecotypes (homoplasies) can be inferred from molecular or geographical data ¹, methods other than population comparison are necessary to demonstrate the action of selection directly and to establish causal relationships ². Although they cannot demonstrate selection unequivocally, comparative studies of populations allow us to detect probable causes of adaptive differentiation, which can then be tested using other methods.

Comparative population studies have suggested adaptive differentiation of many of the traits listed in Table 1. Indeed, adaptive differentiation of behavioural phenotypes has proven so pervasive that the most exciting examples go well beyond demonstration of correlations indicating probable adaptation. For example, research on guppies (Poecilia reticulata) has offered compelling experimental evidence about the causes of adaptive geographic variation in several behavioural phenotypes. Introductions to natural habitats differing in predation regime, and laboratory selection experiments in which visual backgrounds have been systematically varied, have demonstrated that population differences in male colouration reflect the counteracting forces of intrasexual selection favouring bright males and natural selection for crypsis by predators ¹⁵. Female mating preferences are, in turn, correlated with population differences in male colour patterns ^15,16 . An artificial selection experiment involving laboratory-reared fish from one population indicates that selection on colour pattern can produce correlated changes in female preference, an outcome predicted by theory ¹⁶. Similarly, antipredator behaviour in the guppy is correlated with predation intensity across populations ( Table 1), and fish derived from an introduction made in 1957 display genetically based shifts in elements of antipredator behaviour that match theoretical predictions and observed patterns of correlations across populations ¹⁷.

Populations that have differentiated in response to divergent selective regimes also offer promising opportunities for testing models that predict adaptive patterns. Most are optimality models ^14,18 , and many invoke differences in selective regimes that vary across populations. A particularly elegant example involves a comparison of contest and foraging behaviour of riparian and grassland populations of the funnel-web spider Agelenopsis aperta ¹⁴. The behaviour of populations from large habitat patches of desert grassland and other more arid habitats in the southwest USA agreed with predictions of evolutionary game theory and optimal foraging models ( Box 2). This agreement demonstrates the power of these models for predicting adaptive behavioural differentiation when there is sufficient genetic variation for populations to evolve to adaptive optima.

Although populations of A. aperta matched model predictions in these large expanses of arid habitat, the behaviour of spiders in desert riparian (woodland) habitats did not always do so. One population inhabited a riparian habitat that was restricted in extent and was subject to gene flow from adjacent, more arid evergreen woodland and desert scrub habitats. This study and other research on population differences in climbing behaviour of the mouse Peromyscus maniculatus ¹⁹ demonstrate the prominent role gene flow can have in preventing populations from reaching adaptive optima in small habitat patches or at the edges of large patches.

Population contrasts also provide evidence that the evolution of adaptive behavioural phenotypes is constrained. The most common examples involve the failure of complex behavioural phenotypes to decay under conditions of relaxed selection. For example, populations of the California ground squirrel Spermophilus beecheyi that have been isolated from snakes for periods spanning 70 000–300 000 years continue to recognize snakes and respond to them appropriately, although they are no longer as good at discriminating rattlesnakes (Crotalus viridis oreganus) from gopher snakes (Pituophis melanoleucus catenifer). In contrast, Arctic ground squirrels (S. parryii), isolated from snakes since the middle of the Pliocene (i.e. for three million years), have lost the ability to recognize snakes entirely, presumably because of the longer period of isolation ^10,11 . Similar variation is observed in the responses of populations of birds that are moved from areas where avian brood parasitism is common to areas devoid of such parasites. In some instances, the ability to reject mimetic eggs has been retained ²⁰, but in others loss of this behaviour has occurred ^21,22 .


Plasticity, learning and ontogeny

Many different types of behaviour in many taxa are extremely responsive to environmental conditions. In some instances, measurements of behavioural phenotypes under common conditions appropriately reflect population differences, but in others, measures of adaptive phenotypic plasticity (reaction norms) are more appropriate ^19,23 . Essentially, the evaluation of behavioural reaction norms is the evaluation of conditional responses to environmental variation, and these can differ among populations. Where appropriate geographical variation exists, comparative studies can be used to test theoretical predictions about the evolution of conditional strategies ^19,23 . An excellent example involves mate guarding behaviour in the soapberry bug Jadera haematoloma. In a population in which natural variation in sex ratio is relatively large, males readily alter mate guarding behaviour in response to sex ratio, whereas males from a population with a more stable sex ratio are less responsive to sex ratio changes. Although this conforms with theory, a second finding, the co-occurrence of genetic variation and plasticity in the former population, was more surprising and suggests that behavioural plasticity, long thought to influence the selective environments experienced by other traits, need not be associated with reduced evolutionary potential ²³.

Learning is an expression of phenotypic plasticity. As with other aspects of behaviour, the ability to learn can evolve and populations can differ genetically in learning capabilities. Population differences in learning skills contribute to geographical variation in a diverse range of behavioural phenotypes, from the capacity for song memorization in white-crowned sparrows (Zonotrichia leucophrys) ²⁴ to antipredator behaviour in threespine stickleback ²⁵. That the evolution of learning ability can take unexpected pathways is well illustrated by predator-avoidance learning in the stickleback. Population differences in antipredator responses match expectations based on differences in the kinds of predator in freshwater habitats. Stickleback from one lake devoid of piscine predators displayed little antipredator behaviour, apparently because the fry had failed to learn avoidance behaviour in previous encounters with their father when he attempted to capture them in his mouth to return them to the nest ( Box 3).

Learning of predator avoidance behaviour by one stickleback population early in life but not by another is a consequence of an evolutionary change in behavioural ontogeny between populations. Although genetically based differences in learning ability can be involved in many such population differences, this is not invariably the case. Development of male song dialects in brown-headed cowbirds (Molothrus ater) is a consequence of complex interactions between male learning and population differences in female perceptual preferences ⁷. In this instance, a primary cause of dialect variation between populations lies in differences in female preferences and responsiveness, rather than in genetically based differences in learning by males. In both stickleback and cowbirds, population comparisons have offered unique insights into the ways that evolutionary modification of behavioural phenotypes can come about, and at the same time have afforded better understanding of the proximate causes of shifts in ontogenetic trajectories.


Speciation

If, as is so often argued, allopatric differentiation is the most common means by which new species arise, speciation is implicitly geographical in nature. Given this observation, the paucity of instances in which population comparisons have been incorporated into research programmes exploring speciation is stunning ^5,6,26 . Equally surprising is the tendency to view behavioural phenotypes involved in courtship and mate choice as invariant across populations when these could be the very traits whose differentiation is most likely to contribute to speciation. Although uncommon, several research programmes have revealed the value of geographical comparisons for understanding the mechanisms by which speciation is achieved ^5,6 .

One approach is to study a taxon in which two or more populations exhibit ecotypic differences that parallel differences between sibling species within the group. When the divergent population characteristics reflect differences in selective regimes between sites, variation in the relationship between fitness and phenotype across populations can provide insight into the ecological causes of speciation in the group. This method can reveal the role of adaptive differentiation in the speciation process ^5,6 , as demonstrated by research on the threespine stickle-back ^4,5 . In this taxon, the widespread ecotypic population differentiation (described in Box 1) parallels differences between ecotypes in Benka Lake (Alaska, USA) ²⁷ and between species pairs in six small lakes in the Strait of Georgia (British Columbia, Canada) ^4,26 . The divergence in all cases is thought to have been driven by similar ecological pressures.

At present, both sympatric and allopatric origins are plausible for the species pairs ^4,26 . Research on the polytypic population in Benka Lake can provide insights into the processes that would have been involved in sympatric speciation, whereas population contrasts can reveal more about speciation in allopatry. In the allopatric model, McPhail ⁴ envisions the marine ancestor colonizing small, eutrophic lakes and evolving into the benthic forms typical of such habitat. A second marine incursion then gave rise to the limnetic species, and the differences between the two were subsequently enhanced by character displacement ^4,26 . This explanation leaves us with a behavioural puzzle in that benthic lacustrine populations are extremely similar to marine populations in aspects of behaviour plausibly expected to contribute to mate choice and species discrimination ⁵ ( Box 1). Whether speciation in this case proves to have arisen in allopatry or sympatry, this hierarchical comparative approach clearly offers windows onto speciation mechanisms and the role of behaviour in speciation that would not be available to us otherwise.

Insights into behavioural aspects of the speciation process are not limited to taxa that display ecotypic differentiation. Plethodontid salamanders within the Desmognathus ocrophaeus complex (a mixture of populations that probably includes several closely related species) display a complete range of sexual incompatibilities that do not appear to covary with environmental characteristics. Thus, there is little to suggest a role for natural selection in this divergence ^28,29 , although sexual selection could have been an important influence ³⁰. Comparisons across populations of grey tree frogs suggest that the diploid Hyla chrysoscelis has given rise to the tetraploid H. versicolor three or more times, in each instance producing parallel changes in advertisement call pattern ³¹.


Conclusions

Behavioural traits, like other phenotypic characteristics, often exhibit geographic variation, and the variation often proves to be the expression of underlying genetic differentiation. Consequently, judicious comparisons across populations can help us understand the patterns and causes of both microevolutionary change and speciation. This approach is most likely to be of value when the taxa under consideration have relatively low dispersal potential or are separated by effective barriers to dispersal.

Population comparisons also provide evidence of rapid adaptive differentiation of some behavioural phenotypes, although mismatches between phenotype and environment can occur because of genetic drift and constraints on evolutionary change. The interactions between genes and the environment also vary across populations, offering especially exciting clues about the evolution of behavioural development.

Finally, as our ability to resolve population phylogenies improves, we should be able to take an explicitly phylogenetic approach to understanding the evolution of geographic variation in behaviour that will provide unique insights into the mechanisms of behavioural evolution and speciation.


Unlinked list

⁵⁰



Acknowledgements

I thank R.G. Coss, W.A. Cresko, S.E. Riechert, S.I. Rothstein, P.A. Verrell and M.E. West for comments that improved the quality of this article. R.J. Scott and J.A. Walker drew all or parts of figures. I was partially supported by NSF Presidential Faculty Fellowship DEB-9253718.

References

[1]
Foster S.A. and Endler J.A. (Eds.) (1999) Geographic Variation in Behavior: Perspectives on Evolutionary Mechanisms. : Oxford University Press

[2] Endler J.A. (1986) Natural Selection in the Wild. : Princeton University Press

[3] Arnold S.J. (1992) Behavioural variation in natural populations. VI. Prey responses by two species of garter snakes in three regions of sympatry.
Anim. Behav., 44:705-719. [Cited by]

[4] McPhail J.D. (1994) Speciation and the evolution of reproductive isolation in the sticklebacks (Gasterosteus) of south-western British Columbia.
In: Bell M.A. and Foster S.A. (Eds.) The Evolutionary Biology of the Threespine Stickleback. (pp. 399-437) : Oxford University Press

[5] Foster S.A., Scott R.J. and Cresko W.A. (1998) Nested biological variation and speciation.
Philos. Trans. R. Soc. London Ser. B, 353:207-218. [Cited by]

[6] Verrell P.A. (1999) Geographic variation in sexual behavior: sex, signals and speciation.
In: Foster S.A. and Endler J.A. (Eds.) Geographic Variation in Behavior: Perspectives on Evolutionary Mechanisms. (pp. 262-286) : Oxford University Press

[7] West M.J., King A.P. and Freeberg T.M. (1996) Social malleability in cowbirds: new measures reveal new evidence of plasticity in the eastern subspecies (Molothrus ater ater)
J. Comp. Psychol., 110:15-26. [MEDLINE] [Cited by]

[8] Foster S.A. et al. (1996) Patterns of homoplasy in behavioral evolution.
In: Sanderson M.J. and Hufford L. (Eds.) Homoplasy. The Recurrence of Similarity in Evolution. (pp. 245-269) : Academic Press

[9] Omland K.E. (1997) Examining two standard assumptions of ancestral reconstructions: repeated loss of dichromatism in dabbling ducks (Anatini)
Evolution, 51:1636-1646. [Cited by]

[10] Coss R.G. and Goldthwaite R.O. (1995) The persistence of old designs for perception.
In: Thompson N.S. (Eds.) Perspectives in Ethology 11. (pp. 83-148) : Plenum Press

[11] Coss R.G. (1999) Effects of relaxed natural selection on the evolution of behavior.
In: Foster S.A. and Endler J.A. (Eds.) Geographic Variation in Behavior: Perspectives on Evolutionary Mechanisms. (pp. 180-208) : Oxford University Press

[12] Berthold P. (1999) Geographic variation and the microevolution of avian migratory behavior.
In: Foster S.A. and Endler J.A. (Eds.) Geographic Variation in Behavior. (pp. 164-179) : Oxford University Press

[13] Hoffman A.A. (1994) Genetic analysis of territoriality in Drosophila melanogaster
In: Boake C.R.B. (Eds.) Quantitative Genetic Studies of Behavioural Evolution. (pp. 188-205) : Chicago University Press

[14] Riechert S.E. (1999) The use of behavioral ecotypes in the study of evolutionary processes.
In: Foster S.A. and Endler J.A. (Eds.) Geographic Variation in Behavior: Perspectives on Evolutionary Mechanisms. (pp. 3-32) : Oxford University Press

[15] Endler J.A. (1995) Multiple-trait coevolution and environmental gradients in guppies.
Trends Ecol. Evol., 10:22-29. [Cited by]

[16] Houde A.E. (1997) Sex,