TO understand the selective forces that have shaped the remarkable diversity of feeding-related morphological traits in snakes, we need several kinds of information. Detailed anatomical studies can provide insight into the ways in which species have diverged through evolutionary time (e.g., McDowell, 1986), functional analyses can clarify the consequences of that morphological diversity for tasks such as prey capture and ingestion (e.g., Deufel and Cundall, 2003), and phylogenetic information can illuminate the sequence and correlates of change (Greene, 1997). However, we also need information about the field biology of the taxa in question to suggest hypotheses about the fitness benefits that have driven particular evolutionary trajectories. Unfortunately, data even on basic topics such as dietary composition are woefully inadequate for many major lineages of snakes. Historical biases in the geographic location of research-intensive institutions (mostly in relatively cool-climate habitats of North America and Europe) mean that taxa from other parts of the world have attracted comparatively little scientific scrutiny, especially in organisms (such as reptiles) that achieve their greatest diversity in areas far distant from the traditional research centers (Seigel and Collins, 1993; Shine and Bonnet, 2000). The current paper provides morphological and ecological data on such a group.

Given the extent of our ignorance about most living species of snakes, the most effective approach may be to concentrate our efforts on taxa that are of particular interest either because they occupy a critical position within snake phylogeny or because they display a suite of morphological features that can best be interpreted in an ecological context (Akani et al., 2001). The atractaspidid snakes of Africa and the Middle East fulfill both of these criteria. Recent studies suggest that atractaspidids represent an ancient radiation and may be sister taxa to the Elapidae (Underwood and Kochva, 1993; Cadle, 1994; Fry and Wuster, 2004; Nagy et al., 2005).

As currently defined, the Actractaspididae comprise about 65 species of snakes distributed broadly through Africa, with a limited penetration into the adjacent Middle East (Branch, 1988, 1998; Underwood and Kochva, 1993). Notoriously confusing taxonomically (Böhme, 1975; Underwood and Kochva, 1993), the atractaspidids share several features (e.g., they are all fossorial and nocturnal and have smooth shiny scales, slender bodies, relatively small heads with indistinct necks, small eyes, and short tails) but also diverge in many aspects. For example, the morphology of the fangs, venom glands, and rictal glands varies substantially among taxa, as does overall mean body size and head shape (Underwood and Kochva, 1993). Long considered atypical viperids, and previously called mole vipers or burrowing adders, extensive recent work unequivocally allies them with the elapids based on morphological attributes (Bates, 1991; Underwood and Kochva, 1993; Wollberg et al., 1998) as well as molecular data (Heise et al., 1995; Gravlund, 2001; Nagy et al., 2005) and venom sequences (Fry, 1999, 2005; Fry and Wuster, 2004). Although the phylogenetic position of atractaspidids thus has attracted considerable attention, much less is known of their general biology (Sweeney, 1971; Spawls et al., 2002). Nonetheless, several authors have commented on their ecology (Branch and Patterson, 1976; Greene, 1977; Boycott, 1995), diets (Douglas, 1982; Akani et al., 2001; Gower et al., 2004), defensive behavior (Golani and Kochva, 1988) and feeding mechanics (Deufel and Cundall, 2003).

Because putative basal taxa within the Elapidae, e.g., African Elapsoidea (Kelly et al., 2003; Nagy et al., 2005), show many similarities in body form and diet to atractaspidids (Broadley, 1971a; Branch, 1998), character states for ecological variables (such as sexual dimorphism, reproductive output, and feeding habits) in atractaspidids may help to clarify the ancestral conditions that prevailed in early elapids. Also, atractaspidids display several morphological novelties, including the possession (by some, but not all species) of a unique “side-stabbing” mode of envenomation that allows the snake to strike sideways and backwards, without opening its mouth (Greene, 1977, 1997). The dietary composition of such animals may clarify selective pressures for the evolution of side-stabbing dentition (Deufel and Cundall, 2003) as well as other unique atractaspidid morphological traits such as the remarkably elongate “quill-snouted” head shape of some taxa.

Materials and Methods

We measured and dissected preserved specimens of six atractaspidid species from southern Africa in the collections of the Transvaal Museum and the Port Elizabeth Museum in the Republic of South Africa and the State Museum of Namibia and the Directorate of Wildlife Conservation (both in Windhoek, Namibia). Species were chosen based on availability of material; a list of specimens examined is provided at the end of this paper. For each animal we recorded snout-vent length (henceforth, SVL), tail length, head length (from the posterior margin of the lower jaw to the tip of the snout), head width (at the widest point), eye diameter, body width (diameter at midbody), and body mass. The specimen was opened with a midventral incision, and any prey items in the alimentary canal (including the hindgut) were removed for later identification. Sex and reproductive status of the snakes were determined by visual inspection of the gonads. Males were considered mature if they had enlarged, turgid testes and/or white, thickened efferent ducts (indicating the presence of sperm). Females were classed as mature if they had thick muscular oviducts, vitellogenic ovarian follicles, and/or oviductal eggs.

Results

Sexual Dimorphism

Adult female Atractaspis bibronii averaged significantly larger than conspecific adult males, but the sex difference in body sizes was small (Table 1; females averaged 11.5% longer). Indeed, the largest specimen examined was a male. Adult males and females were thus similar in the mean values of other morphological variables (Table 1). However, significant sex differences were apparent in body proportions: at the same SVL, males had longer tails and larger heads than did females, and were more heavy-bodied (Table 2). Relative to head length, males also had larger eyes (Table 2). Female Amblyodipsas polylepis attained much greater body sizes than did males; indeed, the size ranges of adult males and females did not overlap in our sample (Table 1; females 75.9% longer). Thus, adult females significantly exceeded adult males for mean values of all the morphological traits that we measured. Over the size range shared by both sexes (<50 cm SVL), males had longer tails than same-sized females, but did not differ in relative head sizes, eye diameters, or body masses (Table 2).

Table 1 Sample Sizes and Morphological Characteristics of Adult Specimens of the Atractaspidid Snake Species Examined in our Study. Criteria for assessment of snakes as adult are given in the text. Some variables could not be measured reliably for some specimens, so sample sizes differ among traits. Data on body mass were log-transformed prior to analysis (to achieve equality of variances) but raw (untransformed) values are reported in the Table. Asterisk refers to a significant (P < 0.05) difference between mean values for adult males vs. females, based on one-factor ANOVA

View Fullscreen

Table 2 Sex Differences in Body Proportions in the Atractaspidid Snake Species Examined in our Study. This Table shows the results of tests for heterogeneity of variances, followed by single-factor analysis of covariance if slopes did not differ significantly. Some variables could not be measured reliably for some specimens, so sample sizes differ among traits. Data on body mass were log-transformed prior to analysis (to achieve equality of variances). Asterisk refers to a significant (P < 0.05) difference between adult males vs. females. Sex was used as the factor in all analyses

View Fullscreen

Unlike their larger congeners, Amblyodipsas ventrimaculata showed little sexual size dimorphism, and mean values for males and females did not differ significantly for most traits (Table 1; females 18.8% longer). However, males had longer tails and larger heads than did females at the same body length, and were more heavy-bodied (analysis restricted to snakes <30 cm SVL, to ensure overlap of sexes: Table 2).

Macrelaps microlepidotus displayed significant sexual size dimorphism in most of the traits that we measured, with females growing much larger than males (Table 1; females 42.1% longer). The sole exception was tail length, reflecting the fact that males have much longer tails than same-sized females (analysis restricted to snakes <70 cm SVL; Table 2). The heads of males were also significantly wider (but not longer) than those of females at the equivalent body length (Table 2).

Adult female Xenocalamus bicolor significantly exceeded adult males for mean values of all the morphological traits that we measured (e.g., 26.1% longer in mean SVL), except for tail length (which showed the reverse pattern; Table 1). Further analysis confirmed the sex difference in relative tail length (Table 2; Fig. 1) and revealed that males are heavier than females of the same body length. The heads of males are larger (relative to snout-vent length) than are those of females, but are narrower in shape (Table 2; Fig. 1).

View Figure

• Download as Powerpoint
• Download Figure

Female Xenocalamus mechowii grew larger than males (mean SVL 34.7% longer) and averaged larger for all morphological traits except tail length (Table 1). Males had longer tails and were more heavy-bodied than females at the same body length, but otherwise showed no significant sex differences in body proportions (analysis on all snakes <60 cm SVL; Table 2).

Reproductive Output

Based on counts of vitellogenic follicles and oviductal eggs, we recorded clutch sizes from 2–19 eggs (Fig. 2). Analysis of covariance with maternal SVL as the covariate, clutch size as the dependent variable, and species as the factor, shows that larger females did not produce larger clutches overall (F_1,34 = 1.02, P = 0.32), but that species differed significantly in their mean clutch sizes relative to maternal body size (F_4,30 = 3.48, P < 0.02), with no significant interspecific divergence in the relationship between fecundity and maternal body size (F_4,30 = 2.38, P = 0.08; see Fig. 2). Removing maternal SVL as a covariate did not modify the conclusion that mean clutch sizes differed significantly among species (F_4,35 = 15.32, P < 0.0001); Fisher's PLSD post hoc tests showed that two species (Amblyodipsas polylepis and Macrelaps microlepidotus) had significantly higher mean clutch sizes than did any of the other taxa.

View Figure

• Download as Powerpoint
• Download Figure

Dietary Habits

The composition of the diet differed significantly among our study species (Table 3). Restricting analysis to the three species with largest sample sizes and dividing prey into major types (frogs, amphisbaenians, other lizards, snakes, mammals), a contingency-table test confirmed significant interspecific divergence (χ² = 30.49, df = 8, P = 0.0002). This difference was due almost entirely to the diet of Xenocalamus (specialization on amphisbaenians) differing from those of the other two species (more generalized; see Table 3). Despite this greater dietary breadth, however, fossorial squamates clearly predominated in atractaspidid diets overall. From Table 3, we recorded 32 amphisbaenians and nine fossorial skinks; the 11 non-fossorial skinks and 11 snakes (including five scolecophidians) were almost certainly taken in burrows at night as well (Broadley, 1971b). In contrast, only a single snake contained anuran prey and only five contained mammals (Table 3).

Table 3 Prey Items Identified from Guts of Atractaspidid Snakes in the Current Study. The Table shows the number of snakes of each species containing each kind of identifiable prey item

View Fullscreen

Detailed inspection of data revealed no clear sex difference in dietary composition in the three species for which large samples were available (Amblyodipsas polylepis: three lizards and three snakes in females vs. five lizards and three snakes in males, χ² = 0.00, df = 1, P > 0.99; Atractaspis bibronii, 13 reptiles and no mammals in males vs. 14 reptiles and three mammals in females, χ² = 0.97, df = 1, P = 0.33; Xenocalamus bicolor, amphisbaenians in eight of ten males vs. nine of nine females, χ² = 0.51, df = 1, P = 0.48). Despite a small sample size, however, we detected a significant ontogenetic shift in diet within X. bicolor: two of three prey items in juvenile snakes were other snakes (one conspecific, one leptotyphlopid), whereas adult animals contained only amphisbaenians (χ² = 6.16, df = 1, P < 0.015).

Discussion

Our results are consistent with previously published information on the general biology of atractaspidid snakes, but differ in some respects and suggest novel functional underpinnings to the high morphological diversity within this ancient lineage. Many previous authors have commented on this diversity, both within and among atractaspidid genera. For example, this lineage encompasses species with a wide variety of dentition types, including the only snake taxon with a unique “side-stabbing” mode of venom delivery (Underwood and Kochva, 1993; Branch, 1998). Additionally, the atractaspidids comprise a diverse array in terms of mean adult body sizes (e.g., adult males averaged <7 g in Amblyodipsas ventrimaculata vs. >100 g in Macrelaps microlepidotus; Table 1) and relative head sizes and shapes (Table 2). The range from relatively large blunt-snouted heads through to narrow quill-shaped heads is seen in intrageneric (Amblyodipsas, Xenocalamus) as well as intergeneric comparisons (Broadley, 1971b). Other features associated with prey ingestion and transport are similarly diverse, including rictal glands and venom glands (in some species, extending for 20% of SVL; Underwood and Kochva, 1993).

The interspecific variation within some of these traits agrees with earlier reports and with general patterns among living snake species. For example, our data on reproductive output (small numbers of relatively large eggs) are similar to the figures summarized by Branch (1988, 1998). He reported clutches of 3–7 eggs for Atractaspis bibronii (vs. 4–6 in our study), 3 eggs in Amblyodipsas ventrimaculata (vs. 2–6), 3–4 eggs in Xenocalamus bicolor (vs. 2–4), and 3–10 eggs in Macrelaps (vs. 5–19). Although we did not find a statistically significant correlation between clutch size and maternal body size (as seen among most snakes; Fitch, 1970; Seigel and Ford, 1987), our data show a strong trend in that direction (Fig. 2).

The interspecific diversity in the degree of sexual size dimorphism in mean adult body size (SSD) among atractaspidids is more surprising, because comparisons among closely-related taxa generally reveal a strong conservatism in this trait (Shine, 1994). Indeed, the degree of SSD in one atractaspidid (Amblyodipsas polylepis, in which adult females averaged 76% longer than conspecific males) ranks as the greatest yet recorded for snakes, far above the previous maximum (58%, for Nerodia rhombifera, in Shine's [1994] review of SSD in 375 species of snakes). In striking contrast, adult females averaged <20% longer than males in the congeneric A. ventrimaculata (Table 1). However, two broad patterns in these data support earlier conclusions. First, adult female snakes generally attain larger body sizes than do conspecific males, especially in taxa (such as atractaspidids) in which males do not engage in physical combat bouts over mating opportunities (Shine, 1994). The only exception may involve Atractaspis bibronii, where our largest specimen was a male, and Broadley (1991a) recorded an even larger male (650 mm SVL). Second, the degree of female-biased SSD tends to be greater in species with larger mean adult body sizes in snakes (Shine, 1994) as well as in other types of animals (e.g., frogs—Shine, 1979; turtles—Berry and Shine, 1980; primates—Clutton-Brock et al., 1977). Atractaspidids show both of these patterns. Thus, the primary surprise is simply the magnitude of interspecific divergence in SSD, which perhaps simply reflects the long timescale of adaptive radiation within this lineage (Cadle, 1994) and, hence, the long duration of opportunities to accumulate interspecific divergence.

Another trait showing significant interspecific divergence is diet, notably the specialization on amphisbaenians by Xenocalamus compared to the more generalized diets of the other taxa (Table 3). Broadley (1971b) considered Xenocalamus to feed “entirely on amphisbaenians”, a dietary specialization that is basically confirmed in this study as well as by our review of previously published information on atractaspidid diets (Tables 4, 5). Ontogenetic shifts in prey type, as shown by our data for X. bicolor, are widespread among snakes (Greene, 1997). Fossorial elongate squamates constitute a high proportion of all prey items even for the more generalized atractaspidid species (Tables 3, 4, 5). However, the small sample of prey from Macrelaps (one diurnal lizard, three mammals) hints at a very different diet (Tables 3, 4, 5).

Table 4 Literature Records of Prey Items Identified from Guts of Atractaspidid Snakes. The Table shows the number of snakes of each species containing each kind of identifiable prey item in parentheses (if absent, means that only a single specimen recorded) followed by [Reference]. Records from the present study are not included

View Fullscreen

Table 5 Summary of Dietary Records (Literature and Present Study) for Atractaspidid Snakes

View Fullscreen

Our data agree with the limited amount of previously published information on these topics (compare Tables 3 and 4). For example, Akani et al. (2001) reported a total of 18 reptiles and four mammals from guts of Nigerian Atractaspis. However, dissections of preserved specimens by Greene (1977, 1997) revealed a slightly higher proportion of mammalian prey (rather than fossorial squamates, as in our study) in the alimentary tracts of Atractaspis (mammals in 11 snakes vs. reptiles in 31 snakes). Greene (1977) also reported a single anuran in one snake and 12 reptile eggs in another, consistent with our own data for A. bibronii (Table 3).

This putatively greater reliance on mammalian prey in Atractaspis than in related taxa has been used to support the hypothesis that foraging for nestling mammals played a critical role in the evolution of “side-stabbing” dentition (Deufel and Cundall, 2003). We suggest a different scenario, oriented around predation on fossorial elongate squamates rather than mammals (see Table 5; overall, mammals constitute <25% of prey items recorded from Atractaspis species). The phylogenetic reconstructions of Underwood and Kochva (1993) suggest a Macrelaps-like ancestor for living atractaspidids, presumably foraging above-ground or in large burrows for a diverse prey base. Specialization on burrow-dwelling elongate squamates then took multiple routes within the atractaspidids and involved a series of morphological changes that facilitated capture of such animals in confined spaces. Reduction in body size must have been one of the first steps (Lee, 1998), but has been followed by complex shifts in feeding-related structures such as side-stabbing envenomation and head-shape modification. These changes might represent alternative adaptations to overcome the same problem: how can the snake seize an escaping prey item in a burrow barely large enough to contain both predator and prey?

One obstacle involves the fact that the snake has immediate access only to the rear end of the escaping prey (i.e., the tail). Tails of amphisbaenians and burrowing skinks typically are relatively thick (similar to the body in diameter), tapering only at the very end (Greer, 1991; Greer et al., 1998; Shine and Wall, unpubl. data). That shape makes it difficult for the snake to move past the lizard's tail to seize the body. If the snake bites the tail, the lizard can autotomize that portion (an ability widespread in scincids and amphisbaenians: Pianka and Vitt, 2003), thus providing a physical barrier likely to stop the predator from further pursuit until it has ingested the tail. Indeed, the ability to block burrows in this way and to discard a tail before injected venom can reach the lizard's body, might be heretofore unappreciated benefits of the capacity to autotomize. From the predator's perspective, such behavior favors an ability to push past the tail and envenomate or seize the lizard's body. This aim can be accomplished by side-stabbing (push forward, then inject venom as the prey moves forward past the snake's head, as in Atractaspis) or by the evolution of a quill-like head with an undercut jaw (as in Xenocalamus), allowing the snake's head to slide in past the prey item and then turn to seize it midbody. Thus, the superficially divergent trophic morphologies within the atractaspidid radiation may represent alternative evolutionary solutions to the same challenge: how to capture a prey item in a narrow burrow when the only immediately accessible part of that prey item is autotomizable and, hence, does not offer an optimal site for seizure. Laboratory trials of prey-capture techniques in restricted spaces could provide an empirical test of such speculations.

Material Examined

List of specimens examined, excluding unregistered animals (note that many animals in the Directorate of Nature Conservation were unregistered).