Printer Friendly

The usefulness of behavior for phylogeny estimation: levels of homoplasy in behavioral and morphological characters.

Abstract. - It is widely believed that behavior is more evolutionarily labile and/or more difficult to characterize than morphology, and thus that behavioral characters are not as useful as morphological characters for estimating phylogenetic relationships. To examine the relative utility of behavior and morphology for estimating phylogeny, we compared levels of homoplasy for morphological and behavioral characters that have been used in systematic studies. In an analysis of 22 data sets that contained both morphological and behavioral characters we found no significant difference between mean consistency indices (CIs, which measure homoplasy) within data sets for the two types of characters. In a second analysis we compared overall CIs for 8 data sets comprised entirely of behavioral characters with overall CIs for 32 morphological data sets and found no significant difference between the two types of data sets. For both analyses, 95% confidence limits on the difference between the two types of characters indicate that, even if given the benefit of the doubt, morphological characters could not have substantially higher mean CIs than behavioral characters. These results do not support the idea that behavioral characters are less useful than morphological characters for the estimation of phylogeny.

Keywords. - Behavior, character evolution, consistency index, homoplasy, morphology, phylogeny.

One of the basic premises of ethology is that behavior evolves in essentially the same fashion as morphology. This idea both arose from and promoted the use of behavioral characters in studies of phylogeny. Although Darwin (1859) clearly realized that instincts evolve, in modern ethology the idea that behavior and structure should be treated equivalently traces most clearly to Whitman (1899) and Heinroth (1911), whose views stemmed largely from the application of behavior to the systematics of birds. With the rise of ethology the use of behavioral characters in studies of phylogeny became relatively commonplace (Hinde and Tinbergen, 1958; Bekoff, 1977; McLennan et al., 1988). Tinbergen spoke for many ethologists when he claimed "Behaviour characters are in principle neither more nor less useful [for systematics] than morphological or other characters; they merely add characters to the total by which overall likeness is judged" (1959, p. 328).

Despite the ethological tradition of using behavior in systematics, among biologists in general it is widely believed that behavioral characters/are inferior to morphological characters as indicators of phylogenetic relatedness. Two main arguments have been forwarded to support this notion. First, it is claimed that preliminary (i.e., nonphylogenetic) criteria for homologizing characters - for example, the position of the character in relation to other features - are difficult and perhaps even impossible to apply to behavioral traits (Atz, 1970; Hodos, 1976; Aronson, 1981). Second, it is held that behavior is so evolutionarily labile that it is suspect as an indicator of relationships (Atz, 1970; Urbani, 1989).

These two arguments are closely related, thus it is worthwhile to clarify the distinction between them from the outset. Both arguments imply that, following a phylogenetic analysis, behavioral characters will be found to exhibit more homoplasy - i.e., character convergence or reversal - than will morphological characters. The first argument implies that such homoplasy - or mistaken homology as some would call it - occurs because behavioral characters lack some of the properties that allow us to make reasonably good preliminary homology assessments for morphological characters. Put another way, the behavioral characters that we would a priori call homologues are not as easily compared as morphological characters, thus we are more likely to make mistakes in assessing behavioral homology. The second argument, on the other hand, implies that, even if we could somehow equalize the a priori criteria for homology applied to behavior and morphology, we would still find that behavioral characters show more homoplasy than morphological characters. In other words, a given degree of similarity is more likely to arise or be lost multiple times for behavior than for morphology.

The fact that behavioral characters can give apparently good estimates of phylogeny falsifies the notion that such characters absolutely cannot be homologized (e.g., McLennan et al., 1988; Arntzen and Sparreboom, 1989; Prum, 1990). It is also clearly true that some behavioral characters can be quite conservative in their evolution. For example, chemosensory tongue protrusion apparently arose in a common ancestor of squamate reptiles no later than the Jurassic and has been retained in all lineages of squamates since then (Schwenk, 1988). Nonetheless, it is possible that a priori criteria of homology are usually more difficult to apply to behavior than morphology and/or that behavioral characters are usually more intrinsically labile than morphological characters.

Studies of sticklebacks (McLennan et al., 1988) and manakins (Prum, 1990) have shown that behavioral homoplasy is no more prevalent than morphological homoplasy for characters used to estimate phylogeny in these taxa. However, a more inclusive test is required to assess in more general terms the relative usefulness of these two kinds of characters for the estimation of phylogeny.

In this study we use cladistic methods to compare relative amounts of behavioral and morphological homoplasy in a wide variety of taxa and characters. Our study complements an earlier study by Sanderson and Donoghue (1989) comparing levels of homoplasy in morphological and molecular data sets. We discuss the results with respect to both the usefulness of behavior in systematics and the question of whether behavior in general is more evolutionarily labile than morphology.

Materials and Methods


We performed two comparisons of levels of homoplasy in behavioral and morphological characters. In the first, we used data sets that contained both behavioral and morphological characters and compared levels of homoplasy for the two kinds of characters within each data set. In the second, we compared overall amounts of homoplasy for data sets that contained only behavioral characters with those for data sets that contained only morphological characters. The first analysis controls for peculiarities of particular taxa by pairing behavioral and morphological data points by study group. The second analysis lacks this control but provides a larger sample of taxa and a more powerful statistical test. The data sets that include behavioral characters are all the published and unpublished systematic data sets we could find that allowed calculations of homoplasy as detailed below. The purely morphological data sets in the second analysis were taken from Sanderson and Donoghue (1989). This sampling scheme resulted in a bias toward studies of certain taxonomic groups and taxonomic ranks. All of the data sets containing behavioral characters are for either arthropods or vertebrates, and within these two groups hymenopterans and birds are particularly overrepresented. Most of the data sets are from studies of interspecific or intergeneric relationships.

We define a behavioral character broadly as one representing movement (or the lack of movement - e.g., "freezing" in response to a predator) of the organism as a whole or any of its external parts. The behavioral data include characters representing a wide assortment of functional categories, including courtship, nesting, territorial and other social behavior, antipredator responses, and feeding behavior. A wide range of organizational levels of behavior are also represented, from simple, stereotyped movements to more inclusive behaviors such as nesting dispersion (e.g., solitary or colonial) and habitat type. Physical manifestations of behavior, principally characters of nest architecture, are also included. Notwithstanding this wide range of behavioral character types, mating display and nest architecture characters make up the majority of the behavioral data.

The morphological characters are also varied and include color patterns and other gross external characteristics, many osteological features, and characteristics of the soft anatomy. We excluded karyological and biochemical characters from the analyses of levels of homoplasy. These latter characters might be considered morphological; however, the ideas regarding behavioral and morphological characters that we are trying to address were not formulated with these kinds of characters in mind. This is not to suggest that we believe such characters are more or less useful than classical morphological features. Sanderson and Donoghue (1989; see also Wyss et al., 1987) found no difference in levels of homoplasy between classical morphological characters and molecular characters.

When entire data sets were taken from a single systematic study we used the tree(s) obtained by the original author(s). If two or more data sets from the literature were combined, or if only part of an original data set was amenable to cladistic analysis, we obtained trees using the computer programs PAUP version 3.0 (Swofford, 1989) or Hennig 86 (Farris, 1988). In these cases we retained the ordering scheme of the original authors; if no obvious ordering scheme had been used we treated the characters as unordered. A data set was included only if the number of phylogenetically informative characters was large enough to at least potentially give a fully resolved tree; thus the number of informative binary-character equivalents had to be at least equal to the number of taxa minus two. (The number of binary-character equivalents is the number of characters that a data set would contain if all the characters were converted to binary characters.)

We used the consistency index (CI; Kluge and Farris, 1969) as our measure of homoplasy. The CI is defined as the minimum number of character state changes required by a given data set divided by the number of character state changes required for the same data given the tree in question. This can be restated as the number of state changes required (given the number of characters and character states but without reference to a tree) if no character exhibits any homoplasy divided by the number of state changes, including homoplasious changes, required by the tree in question. CIs can be calculated both for individual characters and for all characters combined (Farris, 1989b). A CI of 1.0 indicates no homoplasy in the character(s) in question for the given tree. As homoplasy increases the CI decreases asymptotically toward a theoretical minimum value that is determined by the number of characters, the number of character states, and the number of taxa (Archie, 1989). As an illustration of the calculation of the CI, consider a character that, within the group of taxa under consideration, has two character states, a primitive state A and a derived state A'. The numerator of the CI for this character is always I because the minimum number of character state changes is 1 (i.e., a single change from A to A'). If, for the given tree, A' is required to arise twice independently from A, then the denominator is 2 and the CI is 0.5. If, in addition to arising twice, A' in one lineage must revert back to A, then the denominator is 3 and the CI is 0.333. Insofar as homoplasy is a form of "noise" in phylogenetic analysis, higher CIs indicate greater usefulness for the estimation of phylogeny.

The consistency index has been criticized as a comparative measure of homoplasy on the grounds that it varies with number of characters and number of taxa and is sensitive to the inclusion of unique derived (autapomorphic) characters, which by definition are phylogenetically uninformative (Archie, 1989). We have used the CI because it is easily calculated, its meaning is generally understood by systematists, and its inherent biases can be accounted for in comparative analyses (see below). Furthermore, the alternative measures of homoplasy whose properties are well known - the homoplasy excess ratio, homoplasy excess ratio maximum, and character retention index (Archie, 1989, 1990; Farris, 1989a) - all take into account not only the number of states for each character but the number of taxa that have each state. Part of our aim was to measure the evolutionary lability of characters and this ability can be associated with the number of taxa that have each character state. For example, the number of taxa in which a derived character state occurs is also the maximum number of independent origins of that state that can be inferred from the data. Thus we did not want to factor out the number of taxa that have each state in our measure of homoplasy. The CI does not account for this variable and thus is more appropriate than the alternative measures as an indicator of evolutionary lability.

For calculations of CIs we used only data coded as discrete states; CIs can be calculated for continuous characters (Kluge and Farris, 1969), but it is unclear how one should compare such CIs with those obtained for discretely coded characters. Phylogenetically uninformative characters - i.e., autapomorphies and characters invariant within the ingroup - were excluded for the calculations of CIs. Inclusion of such characters in calculations of CIs gives an inflated indication of the phylogenetic usefulness of the data. In addition to assessing the usefulness of characters in phylogenetic studies a goal of our study was to measure the evolutionary lability of characters. Autapomorphies and shared derived characters (synapomorphies) common to all members of the ingroup do indicate evolutionary change. However, because some of our sources presented only informative characters, we did not perform a separate analysis with uninformative characters included.

Statistical analyses were performed using Systat version 5.1 (Wilkinson, 1987).

Comparison of Character CIs within

Data Sets

For each of the 22 data sets listed in Table 1 we obtained a most parsimonious tree or set of trees (the most trees for any one study being six) as outlined above. Because our goal was to measure actual levels of homoplasy we excluded data sets (not referenced) that gave highly unresolved trees. We did not use a specific algorithm for rejecting data sets on these grounds. However, decisions to reject data sets were made without knowledge of Cls for these data, thus their exclusion should not have biased the results. To obtain the best estimate of the true tree, we included any available karyological and molecular data that could be coded as characters with discrete states. (As noted above, levels of homoplasy were not calculated for such characters.) In estimating the phylogeny we included all types of data in a single parsimony analysis rather than performing separate analyses and obtaining a consensus tree (for the rationale for this approach see Barrett et al., 1991).


We tested the null hypothesis that behavioral characters are no more homoplasious than morphological characters by comparing mean CIs for the two kinds of characters within each of the data sets. In compiling the sets of character CIs we excluded characters that were polymorphic within any terminal taxon on the grounds that CIs for such characters may be heavily influenced by the unknown cladistic structure within these terminal taxa. In addition, we excluded behavioral characters that obviously depend on unique morphological features. We also left out characters that had unknown states for more than 20% of the taxa. This 20% rule allowed us to retain the great majority of characters without including what seemed an unreasonable number of unknown states, but this threshold is obviously somewhat arbitrary. Including characters with unknown states for certain taxa tends to raise CIs because parsimony algorithms assign the most parsimonious state to such unknowns. A greater percentage of the behavioral than the morphological characters had unknown states, thus the 20% threshold (rather than exclusion of all characters with any unknown states) may have biased the results toward higher CIs for behavior. However, the great majority of behavioral characters were known for all taxa so this effect is likely to be slight. For cases in which more than one most parsimonious tree was found, we used the mean CI averaged over all the characters in all the trees.

For phylogenetically informative data sets in general the distribution of character CIs tends to be heavily skewed toward 1.0 (perfect consistency). In addition, because of the limited set of numerators and denominators only certain values for character CIs are possible and in reality only very few values are at all frequent. A typical distribution of character CIs has a large peak at 1.0 with several smaller peaks at other likely outcomes (e.g., 0.5 for a binary character with one extra step; 0.333 for a binary character with two extra steps; see Fig. 1). This distribution is obviously highly nonnormal. The distributions of the mean character CIs for the behavioral and morphological samples are also nonnormal, although they are more normal than the CI distribution. However, the distribution of differences between mean morphological and behavioral CIs within data sets does not differ significantly from a normal distribution (Lilliefors test, N = 22, maximum distance = 0.126, P > 0.4). This justified the use of a parametric test; we used a paired t-test to assess differences between morphological and behavioral CIs, with each data set providing a paired comparison in the analysis.

Because of the small sample size (i.e., number of data sets) and the substantial variance in the difference between mean morphological and behavioral CIs, a type II error (i.e., failing to reject a false null hypothesis) is a distinct possibility. To address this problem we calculated 95% confidence limits for the difference between morphological and behavioral mean CIs. If the significance test on the difference failed to reject the null hypothesis, these 95% confidence limits might yet give an indication that a substantial difference was a reasonable possibility.

Comparison of Overall CIs among

Data Sets

In the second analysis we compared overall CIs for data sets (listed in Table 2) consisting either entirely of behavioral characters or entirely of morphological characters. Although we have no "behavioral" data on plants, we included morphological studies of plants because CIs for plant and animal data sets apparently do not differ (Sanderson and Donoghue, 1989). Because of the very limited number of usable behavioral data sets, studies used in the first analysis were included among the behavioral studies in this second analysis. However, to increase the independence of the two analyses, none of the studies from the first analysis were used as morphological data sets in the second analysis. Trees were constructed as outlined above. For unordered multistate characters, derived states that occur in only one taxon are uninformative. Following Sanderson and Donoghue (1989) we removed the effects of such states by subtracting 1 from both the numerator and the denominator of the overall CI for each such state for the data set in question.


We performed analyses of covariance to compare CIs of the behavioral and morphological data sets using number of taxa and number of binary-character equivalents as covariates. Previous studies have shown significant effects of one or both of these variables on CIs (Archie, 1989; Sanderson and Donoghue, 1989). Parametric tests are appropriate because CIs for entire data sets do not exhibit the same highly skewed distribution shown by individual character CIs within a data set. We transformed the overall CIs to their natural logarithms for the analyses of covariance to make the variances independent of the means and because Cl is curvilinearly related to number of taxa (Archie, 1989; Sanderson and Donoghue, 1989).

In the analyses of covariance we tested for differences between morphological and behavioral log CIs at the mean number of taxa and mean number of characters of the behavioral data sets. We used the mean for the behavioral data sets rather than the mean for all the data sets because the latter would have resulted in a comparison outside the actual range of number of characters for the behavioral data sets. To obtain comparisons at the means of the behavioral data sets we used variables denoted adjusted number of taxa and adjusted number of characters. These variables were created by subtracting the mean number of taxa of the behavioral data sets from the actual number of taxa, and the mean number of characters of the behavioral data sets from the actual number of characters. For analysis of covariance, Systat makes the comparison at the Y-intercept; using the adjusted values shifts the Y-intercept from zero to the behavioral mean for number of taxa and characters.

As for the comparison within data sets, we were concerned about the possibility of a type II error. This may be a particular problem for analysis of covariance when the values taken by the covariates differ substantially between levels of the "treatment" (here, character type) because this difference can inflate the variance of the difference between treatment levels (Snedecor and Cochran, 1980). We therefore calculated 95% confidence limits for the difference between morphological and behavioral log CIs at the Y-intercept (i.e., at the mean number of taxa for the behavioral data sets). The potentially inflated variance can only make the 95% confidence interval conservative (i.e., wider) with respect to the conclusions drawn below.


In both of our analyses the data for which levels of homoplasy are being assessed are also used to estimate phylogenies. This introduces a problem of circularity because the phylogenies are used to estimate homoplasy. Several authors have discussed this problem with respect to the reconstruction of character evolution, and two "solutions" have been advocated. Some authors (e.g., Olmstead, 1989) suggest that the character(s) of interest should be left out of the data set used to estimate the phylogeny. Others (e.g., Donoghue and Sanderson, 1992) suggest that the character(s) of interest should be included in the data set to increase the likelihood of obtaining the true tree.

We have chosen the second course for two reasons in addition to the one just given. First, if the CI for a character must be obtained with that character removed from the data used to construct the tree, then for each data set as many tree-finding analyses must be run as there are characters in the data set. In contrast, if characters are not excluded then only one tree-finding analysis must be run for each data set. For large data sets, computer time becomes a significant bottleneck for such analyses. Second, and more importantly, because we are interested in relative rather than absolute levels of homoplasy, circularity is a problem only if it systematically underestimates or overestimates behavioral homoplasy relative to morphological homoplasy. This problem is addressed in the Discussion section with reference to the specific results of this study.


The means ([+ or -] SD) of the mean CIs for morphological and behavioral characters within data sets are 0.839 [+ or -] 0.120 and 0.841 [+ or -] 0. 140, respectively. A paired t-test shows no statistically significant difference between the two types of characters ([X.sub.morph-behav] = 0.002 [+ or -] .171SD; t = 0.052, df = 21, P > 0.9). The 95% confidence limit for the difference extends from the behavioral CI being 0.074 worse than the morphological CI to the behavioral CI being 0.078 better than the morphological CI.

The analyses of covariance on overall log CIs of behavioral and morphological data sets showed no statistically significant difference in intercept between these character types when adjusted number of taxa alone, adjusted number of characters alone, or both were entered as covariates (Table 3). A plot of log CI against number of taxa (Fig. 2) indicates graphically the absence of evidence that the morphological data sets have higher log CIs than do the behavioral data sets.


Both covariates were significantly correlated with log CI when used by themselves. Only adjusted number of taxa was significantly correlated with log CI when both covariates were used. Interactions of the covariates and character type were insignificant for all analyses. For the analyses reported here, we omitted morphological data sets with exceptionally many taxa and/or characters because these data sets had an unduly large influence on the slope and rendered the regression nonlinear. Because of the large effect of number of taxa on log CI and the potential for nonlinearity, we were especially concerned that the range of number of taxa for the two character types be comparable. Therefore, we omitted any morphological data set for which the number of taxa was even moderately higher (i.e., 10 or more higher) than the greatest number of taxa for a behavioral data set. The essential result - the lack of a significant effect of character type on log CI - holds when all of the data sets are included. However, in this latter analysis the assumption of linear relationships between log CI and the covariates is violated, thus the conclusions regarding data sets with numbers of taxa greater than 40 and numbers of characters greater than 130 should be treated with caution.

We used the results from the analysis of covariance with adjusted number of taxa as the lone covariate (because this was the only significant covariate in the analysis with both covariates) to obtain a 95% confidence interval for the difference between behavioral and morphological log CIs. This gave estimates for the behavioral and morphological log CIs (at the mean number of taxa for the behavioral data sets) of -0.266 and -0.378, respectively. The 95% confidence interval extends from the behavioral log CI being 0.038 worse than to 0.262 better than the morphological log CI. Transformed back to CIs, the point estimates for behavior and morphology are 0.766 and 0.685, respectively. To transform the 95% confidence interval from log CIs to CIs one must assume a particular point on the CI scale because log CI and CI are curvilinearly related. Assuming that the estimate of the morphological CI is correct, the 95% confidence interval extends from the behavioral CI being 0.025 worse than to 0.205 better than the morphological CI.


Behavioral Characters and the

Estimation of Phylogeny

The estimation of phylogenetic relatedness relies on the recognition of character states shared through descent from a common ancestor. Homoplasious characters confound such analysis, thus levels of homoplasy are in general inversely related to the usefulness of characters in phylogenetic studies. For features that are at least potentially useful for estimating phylogeny, our comparisons of consistency indices do not support the idea that behavior is more homoplasious than morphology. Both the comparison of mean morphological and behavioral character CIs within data sets and the analysis of overall CIs among purely morphological and purely behavioral data sets (Table 3, Fig. 2) show no statistically significant difference between the two types of characters. The 95% confidence limits of the difference indicate that morphological characters, at best, might on average have CIs 0.03-0.07 higher than those of behavioral characters. Even this difference, which was arrived at by giving morphology all the benefit of the doubt, is not very striking. In short, the data give no justification for discriminating against behavioral characters as indicators of phylogenetic relationships.

Because the characters of interest were used to construct the phylogenies, the comparison of overall CIs, strictly speaking, only indicates that characters within behavioral data sets evolve as congruently as those within morphological data sets. If behavioral characters evolve as congruently as morphological characters but are more homoplasious this comparison would underestimate behavioral homoplasy relative to morphological homoplasy (see the comment on circularity in the Materials and Methods section). However, we see no reason to believe that behavior evolves in a fashion that produces congruence despite high levels of homoplasy. In any case, the analysis of mean CIs within data sets is not subject to this criticism. Most of these data sets are comprised primarily of morphological characters. Mere congruence among the relatively few behavioral characters in these data sets is unlikely to result in high consistency estimates for these characters because the tree upon which the estimates are based results primarily from morphological characters.

It might be argued that only particularly useful behavioral characters are normally included in systematic studies and thus that our sample indicates nothing about behavior in general. With regard to this argument one point should be clarified from the outset. Characters used in estimating phylogeny are clearly not a random sample of all features of living things; typically the characters chosen for analysis are invariant (or nearly so) within the taxa to be studied but vary among those taxa. Our claim is that within this subset of characters the evidence does not indicate that morphology is more phylogenetically informative than behavior. The argument can still be made that, even within this subset, systematists have been especially selective in their choice of behavioral characters. However, this argument is weakened by the fact that overall CIs for purely behavioral data sets are as high as those for morphological data sets (Table 3, Fig. 2). In this case many behavioral characters were included in each data set, thus it is unlikely that the investigators chose only those characters that seemed especially consistent. The few behavioral characters in many of the data sets in our first analysis (Table 1) is more likely due to the difficulty of collecting behavioral information than to heightened selectivity on the part of the investigators.

The limitation of our samples to vertebrates and arthropods - necessitated by the lack of appropriate data sets for other taxa - is clearly a shortcoming of the study. It should be noted, however, that the taxa that are particularly overrepresented - namely hymenopterans and birds - are groups that have played a major role in shaping biologists' notions about behavior. In other words, the groups used are in large part the same ones that helped give rise to the ideas we are countering in this paper. In any case, we see no reason to think that the included taxa exhibit particularly low (or high) levels of behavioral homoplasy; these taxa are, for the most part, simply groups for which behavioral information can be relatively easily obtained. Nonetheless comparisons should be extended to taxa other than the vertebrates and arthropods.

Our samples may also be biased toward studies at low taxonomic levels (e.g., among species or genera), although this is not clear because we do not know the overall distribution of systematic studies among taxonomic levels. Assuming that the bias is real it might be taken as evidence that behavior is in fact more homoplasious than morphology. The argument is that levels of behavioral homoplasy are reasonable for low level studies but are intolerably high for the estimation of higher level relationships with the result that few high level studies include behavioral characters. The idea that behavior is only useful at low taxonomic levels has been articulated by several authors (Blair, 1953; Tinbergen, 1963; Atz, 1970). However, there is another plausible explanation for the lack of high level studies that incorporate behavioral characters. The ability to obtain behavioral information for a group depends on key characteristics of the taxon in question. For example, the courtship behavior of ducks has been well characterized because these animals frequently display on open water and, in addition, are often bred in captivity for their aesthetic appeal. As increasingly distant groups are included in an investigation it becomes less and less likely that all of them will share the characteristics that facilitate study of their behavior. (It should be noted that the facilitating characteristics themselves may or may not be behavioral.) In a sense, the problem is an inability to adequately detect potential homologies. However, this is not to say that it is intrinsically difficult to detect homology among behavioral traits, only that it is difficult to obtain behavioral data that one might even think of homologizing. This problem is not nearly as pronounced for morphology. As a result, few high level studies will rely on behavioral characters. In short, we do not believe that the lack of high level studies in our samples constitutes evidence that behavior is more homoplasious than morphology. However, attempts should be made to incorporate behavioral characters in high level studies to more fully address this issue.

Because of the difficulty of collecting behavioral data, morphology will no doubt continue to play a larger role than behavior in phylogeny estimation. Nonetheless our study indicates that behavioral characters should be used whenever they are available and that phylogenies based on behavior should engender as much confidence as those based on morphology. In concert with Sanderson and Donoghue's (1989) demonstration of a lack of support for the claim that molecular data are more consistent than morphological data, our study suggests that intuitive notions about the relative systematic worth of broad categories of characters may be ill-founded.

The Evolutionary Lability of


As discussed in the introduction, behavioral characters might be expected to show more homoplasy than morphological characters either because a priori criteria for recognizing homology are especially difficult to apply to behavior or because behavior is intrinsically more evolutionarily labile than morphology. For the set of characters used in phylogeny estimation the similarity in levels of homoplasy for behavior and morphology does not support either of these notions. It is tempting to generalize these results to all behavioral and morphological features. However, as mentioned above, the characters included in our analysis are obviously not a random sample of all possible features, thus such a generalization is not warranted.

Nonetheless the fact that our results fail to support claims about the peculiarity of behavioral characters in a case where support might have been expected should be food for thought. It seems to us that arguments for the uniqueness of behavioral character evolution are unconvincing at best. An exhaustive critique of such arguments is beyond the scope of this paper; however, we would like to briefly consider two widespread claims, both of which have been used to argue that behavior is more evolutionarily labile than morphology.

The first points to the great inter- and intraindividual variation in behavior and claims that this variation must be associated with the evolutionary lability of such features. Although we have not found this argument stated explicitly in the literature we include it here because it is perhaps the most common argument we have heard in conversations on this subject with other biologists. It should be noted that this variation is not the standard heritable variation of population genetics, but rather is the result of proximate responses of organisms to environmental stimuli. Examples include the tendency for some species of birds to breed cooperatively (i.e., to have nonbreeding "helpers-at-the-nest") in some parts of their range but not others (Stacey and Bock, 1978), switches from wide dispersal to philopatry between years in individuals of some mammalian species (Elton, 1942), and the varying responses of many animals to different feeding stimuli (Curio, 1976). However, although the behavior of an individual may change more rapidly than its morphology, it is not clear that phenotypic plasticity in general is more characteristic of behavior than morphology. Countless examples exist of such plasticity in morphological characters. Specific examples include the development of alternative jaw and body shape morphs induced by differences in diet in cichlid fishes (Meyer, 1987; Wimberger, 1991, 1992), the inducible morphological defenses of many aquatic invertebrates (Havel, 1986), and the many organ and tissue level effects of exercise (e.g., Lanyon and Rubin, 1985) and of life at high altitude (e.g., Hock, 1974). Furthermore, although there is some relationship between such plasticity and evolution (West-Eberhard, 1989) the nature of that relationship is complex (Via and Lande, 1985).

The second argument is that behavior is more prone to convergence than morphology because behavior is particularly susceptible to natural selection (Mayr, 1958; Atz, 1970; see below). Behavioral traits apparently do tend to have lower heritabilities than morphological traits (Mousseau and Roff, 1987), suggesting more intense selection on the former than the latter (or that behavioral traits are more often at evolutionary equilibrium than morphological traits). However, one could equally well argue that features subject to selection should show less convergence if selection is usually stabilizing. This, for example, is the generally accepted explanation for the observation that first and second positions in nucleotide codons often exhibit less change than third positions (Li et al., 1985).

In short, the two above arguments for the evolutionary lability of behavior are based on unsubstantiated empirical claims and/or questionable theoretical justification. The weakness of these arguments, coupled with the results of our study, gives no reason to believe that behavioral evolution is particularly rapid or characterized by high levels of convergence.

Divergent Views on the Use of

Behavior in Systematics

Curiously, both comparative psychologists, who have been characterized as resisting the integration of behavior and modern evolutionary thought (Lockard, 1971; but see Dewsbury, 1984), and behavioral ecologists, who have championed the applicability of evolutionary theory to behavior (Krebs and Davies, 1987), have been skeptical of the use of behavior as an indicator of phylogenetic relationship (Aronson, 1981; Burghardt and Gittleman, 1990). On the other hand, classical ethologists have typically embraced the use of behavior in systematics. We suspect that a two-step process accounts for this pattern. First, the ephemeral nature of behavioral acts and the apparent ease with which animals can alter their behavior could lead - partly by confusing proximate and ultimate causation - to the view that behavioral features are very evolutionarily labile and difficult to characterize. Second, this natural tendency could be erased by careful observation of the behavior of many, and especially of closely related, species of organisms. Of the three groups of scientists mentioned, only the ethologists have a strong tradition of comparative studies of the behavioral repertoires of closely related species. Comparative psychologists have emphasized studies of learning, often of very distantly related organisms, while comparative studies in modern| behavorial ecology usually focus on a few behavioral patterns of particular theoretical interest (e.g., mating systems, territoriality). Given these different approaches it is perhaps not surprising that the ethologists have been much more inclined than the other two groups to accept the use of behavior in studies of phylogeny.

Behavioral ecology may have undermined the use of behavior in systematics in another way as well. Both comparative and single species studies in behavioral ecology have emphasized the apparent fit of behavioral characteristics to current environments, leading to the belief that behavior is strongly molded by natural selection (Burghardt and Gittleman, 1990). This in turn may have promoted the view that behavior is too evolutionarily labile to be useful in systematics. However, as argued above, even if behavior is particularly susceptible to selection (a debatable point), it does not necessarily follow that behavioral evolution must be extremely labile. The predominance of selection as a process does not necessarily imply a particular pattern of change. It seems appropriate to end a paper espousing the usefulness of behavior in systematics by quoting Tinbergen (1959, p. 321), himself a great believer in the survival value of behavior (and, as such, the father of behavioral ecology), but imbued with a phylogenetic perspective as well: "It is of course only after completion of a taxonomic study that one can distinguish between more conservative and more changeable characters, and a priori, each character, however trivial it may seem to be, has to be given the benefit of the doubt."

Literature Cited

Alexander, B. A. 1991. A phylogenetic analysis of the genus Apis (Hymenoptera: Apidae). Ann. Entomol. Soc. Am. 84:137-149. Archibald, G. W. 1977. The unison call of cranes as a useful taxonomic tool. Ph.D. Diss. Cornell University, Ithaca, NY USA. Archie, J. W. 1989. Homoplasy excess ratios: New indices for measuring levels of homoplasy in phylogenetic systematics and a critique of the consistency index. Syst. Zool. 38:253-269. _____. 1990. Homoplasy excess ratios and retention indices: A reply to Farris. Syst. Zool. 39:169-174. Arnold, E. N., and J. A. Burton. 1978. A Field Guide to the Reptiles and Amphibians of Britain and Europe. Collins, London, UK. Arntzen, J. W., and M. Sparreboom. 1989. A phylogeny for the Old World newts, genus Triturus: Biochemical and behavioural data. J. Zool., Lond. 219:645-664. Aronson, L. R. 1981. Evolution of telencephalic function in lower vertebrates, pp. 33-58. In P. R. Laming (ed.), Brain Mechanisms of Behaviour in Lower Vertebrates. Cambridge University Press, Cambridge, UK. Atz, J. W. 1970. The application of the idea of homology to behavior, pp. 53-74. In L. R. Aronson, E. Tobach, D. S. Lehrman, and J. S. Rosenblatt (eds.), Development and Evolution of Behavior: Essays in Honor of T. C. Schneirla. W. H. Freeman, San Francisco, CA USA. Barrett, M., M. J. Donoghue, and E. Sober. 1991. Against consensus. Syst. Zool. 40:486-493. Bekoff, M. 1977. Quantitative studies of three areas of classical ethology: Social dominance, behavioral taxonomy, and behavioral variability, pp. 1-46. In B. A. Hazlett (ed.), Quantitative Methods in the Study of Animal Behavior. Academic Press, N.Y., USA. Blair,W.F. 1953. Nonmorphological data in anuran classification. Syst. Zool. 2:72-84. Burghardt, G. M., and J. L. Gittleman. 1990. Comparative behavior and phylogenetic analyses: New wine, old bottles, pp. 192-225. In M. Bekoff and D. Jamieson (eds.), Interpretation and Explanation in the Study of Animal Behavior, Volume II: Explanation, Evolution, and Adaptation. Westview Press, Boulder, CO USA. Cannatella, D. C., and L. Trueb. 1988. Evolution of pipoid frogs: Intergeneric relationships of the aquatic frog family Pipidae (Anura). Zool. J. Linn. Soc. 94:1-38. Carpenter, J. M. 1982. The phylogenetic relationships and natural classification of the Vespoidea (Hymenoptera). Syst. Entomol. 7:11-38. _____. 1987. Phylogenetic relationships and classification of the Vespinae (Hymenoptera: Vespidae). Syst. Entomol. 12:413-431. _____. 1988. The phylogenetic system of the Stenogastrinae (Hymenoptera: Vespidae). J. N.Y. Entomol. Soc. 96:140-175. _____.1989a. Testing scenarios: Wasp social be havior. Cladistics 5:131-144. _____.1989b. The phylogenetic system of the Gayellini (Hymenoptera: Vespidae; Masarinae). Psyche 95:211-241. Carpenter, J. M., and J. M. Cumming. 1985. A character analysis of the North American potter wasps (Hymenoptera: Vespidae; Eumeninae). J. Nat. Hist. 19:877-916. Cracraft, J. 1985. Monophyly and phylogenetic relationships of the Pelecaniformes: A numerical cladistic analysis. Auk 102:834-853. Curio, E. 1976. The Ethology of Predation. Springer-Verlag, Berlin, Germany. Darwin, C. 1859. The Origin of Species, 1st ed. Penguin Books, Hamondsworth, UK. (Penguin edition published 1968.) de Queiroz, K. 1989. Morphological and biochemical evolution in the sand lizards. Ph.D. Diss. University of California, Berkeley, USA. de Sa, R. O., and D. M. Hillis. 1990. Phylogenetic relationships of the pipid frogs Xenopus and Silurana: An integration of ribosomal DNA and morphology. Mol. Biol. Evol. 7:365-376. Dewsbury, D. A. 1984. Introduction, pp. 1-4. In D. A. Dewsbury (ed.), Foundations of Comparative Psychology. Van Nostrand Reinhold, N.Y., USA. Donoghue, M. J., and M. J. Sanderson. 1992. The suitability of molecular and morphological evidence in reconstructing plant phylogeny, pp. 340-368. In P. S. Soltis, D. E. Soltis, and J. J. Doyle (eds.), Molecular Systematics of Plants. Chapman and Hall, N.Y., USA. Elton, C. 1942. Voles, Mice and Lemmings: Problems in Population Dynamics. Clarendon Press, London, UK. Estes, R., K. de Queiroz, and J. Gauthier. 1988. Phylogenetic relationships within Squamata, pp. 119-281. In R. Estes and G. Pregill (eds.), Phylogenetic Relationships of the Lizard Families: Essays Commemorating Charles L. Camp. Stanford University Press, Stanford, CA USA. Farris, J. S. 1988. Hennig86, version 1.5. Copyright J. S. Farris. _____. 1989a. The retention index and homoplasy excess. Syst. Zool. 38:406-407. _____. 1989b. The retention index and the rescaled consistency index. Cladistics 5:417-419. Gaudin, A. J. 1969. A comparative study of the osteology and evolution of the Holarctic tree frogs: Hyla, Pseudacris, Acris, and Limnaoedus. Ph.D. Diss. University of Southern California, Los Angeles, USA. _____. 1974, An osteological analysis of Holarctic tree frogs, family Hylidae. J. Herpetol. 8:141-152. Grimaldi, D.A. 1987. Phylogenetics and taxonomy of Zygothrica (Diptera: Drosophilidae). Bull. Am. Mus. Nat. Hist. 186:103-268. Havel, J. E. 1986. Predator-induced defenses: A review, pp. 263-278. In W. C. Kerfoot and A. Sih (eds.), Predation: Direct and Indirect Impacts on Aquatic Communities. University Press of New England, Hanover, NH USA. Hedges, S. B. 1986. An electrophoretic analysis of Holarctic hylid frog evolution. Syst. Zool. 35:1-21. Heinroth, O. 1911. Beitrage zur Biologie, namentlich Ethologie und Psychologie der Anatiden. Verh. Ver. Int. Ornithol. Kongr. (Berlin) 1910:589-702. Hellmich, W. 1962. Reptiles and Amphibians of Europe. Translated by C. Winter. Blandford Press, London, UK. Hinde, R. A., and N. Tinbergen. 1958. The comparative study of species-specific behavior, pp. 251-268. In A. Roe and G. G. Simpson (eds.), Behavior and Evolution. Yale University Press, New Haven, CT USA. Hock, R. J. 1974. The physiology of high altitude, pp. 195-204. In N. K. Wessells (compiler), Vertebrate Structures and Functions. W. H. Freeman, San Francisco, CA USA. Hodos, W. 1976. The concept of homology and the evolution of behavior, pp. 153-167. In R. B. Masterton, W. Hodos, and H. Jerison (eds.), Evolution, Brain, and Behavior: Persistent Problems. Lawrence Erlbaum, Hillsdale, NJ USA. Johnsgard, P. A. 1960. Classification and evolutionary relationships of the sea ducks. Condor 62: 426-433. Kluge, A. G., and J. S. Farris. 1969. Quantitative phyletics and the evolution of anurans. Syst. Zool. 18:1-32. Krebs, J. R., and N. B. Davies. 1987. An Introduction to Behavioral Ecology, 2nd ed. Sinauer, Sunderland, MA USA. Lanyon, L. E., and C. T. Rubin. 1985. Functional adaptation in skeletal structures, pp. 1-25. In M. Hildebrand, D. M. Bramble, K. F. Liem, and D. B. Wake (eds.), Functional Vertebrate Morphology. Harvard University Press, Cambridge, MA USA. Lanyon, W. E. 1986. A phylogeny of the thirty-three genera in the Empidonax assemblage of tyrant flycatchers. Am. Mus. Nov. 2846:1-64. _____. 1988a. The phylogenetic affinities of the flycatcher genera Myiobius Darwin and Terenotriccus Ridgway. Am. Mus. Nov. 2915: 1-11. _____. 1988b. A phylogeny of the flatbill and todytyrant assemblage of tyrant flycatchers. Am. Mus. Nov. 2923:1-41. Li,W.-H., C.-C. Luo, and C.-I.Wu. 1985. Evolution of DNA sequences, pp. 1-94. In R. J. Maclntyre (ed.), Molecular Evolutionary Genetics. Plenum Press, N.Y., USA. Livezey, B.C. 1986. A phylogenetic analysis of recent anseriform genera using morphological characters. Auk 103:737-754. Lockard, R.B. 1971. Reflections on the fall of comparative psychology: Is there a message for us all? Am. Psychol. 26:168-179. Mayr, E. 1958. Behavior and systematics, pp. 341-362. In A. Roe and G. G. Simpson (eds.), Behavior and Evolution. Yale University Press, New Haven, CT USA. McLennan, D. A., D. R. Brooks, and J. D. McPhail. 1988. The benefits of communication between comparative ethology and phylogenetic systematics: A case study using gasterosteid fishes. Can. J. Zool. 66:2177-2190. Meyer, A. 1987. Phenotypic plasticity and heterochrony in Cichlasoma managuense and their implications for speciation in cichlid fishes. Evolution 41:1357-1369. Mousseau, T. A., and D. A. Roff. 1987. Natural selection and the heritability of fitness components. Heredity 59:181-197. Nelson, J. S. 1971. Comparison of the pectoral and pelvic skeletons and of some other bones and their phylogenetic implications in the Aulorhynchidae and Gasterosteidae (Pisces). J. Fish. Res. Board Can. 28:427-442. Nikol'skii, A. M. 1962. Fauna of Russia and Adjacent Countries: Amphibians. Israel Program for Scientific Translations, Jerusalem. Obst, F. J., K. Richter, and U. Jacob. 1988. The Completely Illustrated Atlas of Reptiles and Amphibians for the Terrarium. Translated by U. E. Friese. T. F. H. Publications, Neptune City, NJ USA. Olmstead, R. 1989. Phylogeny, phenotypic evolution, and biogeography of the Scutellaria angustifolia complex (Lamiaceae): Inference from morphological and molecular data. Syst. Bot. 14:320-338. Pinto, J. D. 1984. Cladistic and phenetic estimates of relationship among genera of eupomphine blister beetles (Coleoptera: Meloidae). Syst. Entomol. 9:165-182. Presch, W. 1988. Phylogenetic relationships of the Scincomorpha, pp. 471-492. In R. Estes and G. Pregill (eds.), Phylogenetic Relationships of the Lizard Families: Essays Commemorating Charles L. Camp. Stanford University Press, Stanford, CA USA. Prum, R. O. 1990. Phylogenetic analysis of the evolution of display behavior in the neotropical manakins (Aves: Pipridae). Ethology 84:202-231. Sanderson, M. J., and M. J. Donoghue. 1989. Patterns of variation in levels of homoplasy. Evolution 43:1781-1795. Schram, F. R. 1986. Crustacea. Oxford University Press, N.Y., USA. Schultz, J. W. 1990. Evolutionary morphology and phylogeny of Arachnida. Cladistics 6:1-38. Schwenk, K. 1988. Comparative morphology of the Lepidosaur tongue and its relevance to Squamate phylogeny, pp. 569-598. In R. Estes and G. Pregill (eds.), Phylogenetic Relationships of the Lizard Families: Essays Commemorating Charles L. Camp. Stanford University Press, Stanford, CA USA. Snedecor, G. W., and W. G. Cochran. 1980. Statistical Methods, 7th ed. Iowa State University Press, Ames, USA. Stacey, P. B., and C. E. Bock. 1978. Social plasticity in the acorn woodpecker. Science 202:1298-1300. Strauch, J. G., Jr. 1985. The phylogeny of the Alcidae. Auk 102:520-539. Swofford, D. L. 1989. Phylogenetic Analysis Using Parsimony (PAUP) Computer Program. Version 3.0. Illinois Natural History Survey, Champaign, IL USA. Thorn, R. 1968. Les Salamandres d'Europe d'Asia et d'Afrique du Nord. Paul Lechevalier, Paris, France. Tinbergen, N. 1959. Behaviour, systematics, and natural selection. Ibis 101:318-330. _____. 1963. On aims and methods of ethology. Z. Tierpsychol. 20:410-433. Urbani, C.B. 1989. Phylogeny and behavioural evolution in ants, with a discussion of the role of behaviour in evolutionary processes. Ethol. Ecol. Evol. 1:137-168. Via, S., and R. Lande. 1985. Genotype-environment interaction and the evolution of phenotypic plasticity. Evolution 39:505-522. West-Eberhard, M. J. 1989. Phenotypic plasticity and the origins of diversity. Annu. Rev. Ecol. Syst. 20:249-278. Whitman, C. O. 1899. Animal behavior, pp. 285-338. In C. O. Whitman (ed.), Biological Lectures, Wood's Hole. Ginn and Co., Boston, MA USA. Wiley, J. E. 1982. Chromosome banding patterns of tree frogs (Hylidae) of the eastern USA. Herpetologica 38:507-520. Wilkinson, L. 1987. SYSTAT: The System for Statistics. Systat, Evanston, IL USA. Wimberger, P. H. 1991. Plasticity of jaw and skull morphology in the neotropical cichlids Geophagus brasiliensis and G. steindachneri. Evolution 45: 1545-1563. _____. 1992. Plasticity of fish body shape: The effects of diet, development, family and age in two species of Geophagus (Pisces: Cichlidae). Biol. J. Linn. Soc. 45:197-218. Wyss, A. R., M. J. Novacek, and M. C. McKenna. 1987. Amino acid sequence versus morphological data and the interordinal relationships of mammals. Mol. Biol. Evol. 4:99-116.
COPYRIGHT 1993 Society for the Study of Evolution
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 1993 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Queiroz, Alan de; Wimberger, Peter H.
Date:Feb 1, 1993
Previous Article:The effects of host-plant genotype, hybridization, and environment on gall-aphid attack and survival in cottonwood: the importance of genetic studies...
Next Article:Plastic relationships between reproductive and vegetative mass in Solidago altissima.

Terms of use | Privacy policy | Copyright © 2020 Farlex, Inc. | Feedback | For webmasters