• E-mail
• Print
• Comment
• Font Size
• Digg
• del.icio.us
• Discuss article

Patterns of Convergence in General Shell Form Among Paleozoic Gastropods

Posted on: Thursday, 18 May 2006, 12:03 CDT

By Wagner, Peter J; Erwin, Douglas H

Abstract.-Recent phylogenetic studies of Paleozoic gastropods show the classification provided by the Treatise on Invertebrate Paleontology to include numerous highly polyphyletic taxa. Here we test whether this classification reflects limits on the range of possible designs, general architectural constraints, or common functional solutions for Paleozoic gastropods. Our test evaluates whether superfamilial/subordinal level archetypes as defined by the Treatise evolved more frequently than expected among 626 Late Cambrian-Middle Devonian species. Using a previously established phylogeny and five general shell features (spire angle, exhalent current position, base angle, umbilical width, and apertural inclination) we show that there are fewer gastropod morphotypes than expected given the frequency of change in these features. This is true even after accounting for architectural constraints implied by the data. Moreover, the most common morphotypes include significantly more species and evolved significantly more times than expected. These results imply a set of architectural attractors for lower-middle Paleozoic gastropods, consistent with ecomorphological theory that particular morphotypes are best suited for particular lifestyles. Thus, the Treatise classifications likely reflect these ecomorphological patterns within subclades rather than phylogenetic patterns.

Introduction

If certain morphologies represent optimal solutions to common adaptive problems, then we expect these morphologies to be particularly common and/or to evolve relatively frequently. However, we expect similar patterns if architectural constraints make many theoretically possible morphotypes improbable or even unobtainable. Indeed, phylogenetic autocorrelation alone predicts both nonuniform distributions of taxa across morphospace and even nonuniform derivations of morphotypes because taxa typically will inherit many (if not most) of their ancestors' traits (e.g., Raup and Gould 1974). Thus, we need to control for the effects of phylogenetic autocorrelation and architectural constraints in order to test and/ or corroborate hypotheses about adaptive morphological types.

Paleozoic gastropods are a model taxon with which to delimit the expectations of phylogenetic autocorrelation and then to separate the expectations of architectural constraints and ecomorphology. First, broad phylogenetic analyses of early snails exist (Wagner 1999a,b; Ntzel et al. 2000). Second, the nature of gastropod coiling limits the possible range of gross shell forms (Raup 1966), which puts obvious architectural constraints on those forms. Third, gross features such as spire angles, umbilicus size, and apertural orientation affect the position of the shell's center of gravity, how the shell interacts with a fluid environment, and how water might have flowed within shells. Consequently, functional and ecomorphological interpretations have typically concentrated on these features (e.g., Yochelson 1971; Cain 1977; Linsley 1977; Linsley et al. 1978; Linsley and Kier 1984; Morris 1991; Gubanov et al. 1995). Finally, many studies suggest high rates of convergent evolution in gross shell form among gastropods (e.g., Batten 1984; Kool 1993; Ponder and Lindberg 1997; Wagner 1999b; Collin and Cipriani 2003).

Traditional suprageneric classifications such as that in the Treatise of Invertebrate Paleontology (Knight et al. 1954, 1960; see also Wenz 1938) rely heavily on gross shell features. Although these gross shell features are meaningful in functional analyses or discussions of architectural limitations, workers now use much more particular characters to infer phylogenetic relationships (e.g., Batten 1984; Wagner 1999a,b; Ntzel and Bandel 2000; Ntzel et al. 2000). Those phylogenetic analyses suggest that suprageneric taxa in the Treatise are polyphyletic assemblages. This raises the possibility that traditional classifications reflect functional archetypes and/or the limitations of architectural constraints rather than portions of gastropod phylogeny. We can use the phylogenetic relationships inferred from the specific shell characters as models to estimate both the frequencies of changes in gross shell form and the expected effects of phylogenetic autocorrelation (see, e.g., Raup and Gould 1974; Felsenstein 1985). If we assume that unobserved combinations of gross shell forms (e.g., high spires plus large umbilici) represent architectural constraints, then we can determine the expected distribution of morphologies given phylogeny and possible architectural constraints. This allows us to evaluate whether general rules for gastropod ecomorphology accurately predict which morphotypes evolve more frequently than expected and/or are more speciose than expected. This also allows us to evaluate whether the Treatise classification of gastropods reflects ecomorphotypes or architectural limitations rather than phylogeny.

Background

Initial Consensus

J. Brooks Knight (e.g., Knight et al. 1954, 1960) established the basic systematic framework for Paleozoic gastropods during the 1950s (Fig. 1) (see also Yochelson 1984; Wagner 2001b). Knight's framework considered archetypes (i.e., basic shell designs) to represent generally paraphyletic or monophyletic groups. The model considers bellerophontoids with bilaterally symmetrical shells (isostrophic) to be the most primitive gastropods. Bellerophontoids give rise independently to two groups with asymmetrical shells (anisostrophic): macluritoids and pleurotomarioids. Macluritoids later gave rise to the euomphaloids. The pleurotomarioids were the ancestral stock for the subsequent lower Paleozoic diversification into patelloids (= patellogastropods or "true" limpets), trochoids, neritoids, and murchisonioids. Murchisonioids included the ancestors of loxonematoids, which in turn were the stem group of the Caenogastropoda, the subulitoids. Subulitoids then gave rise to neogastropods, opisthobranchs + pulmonates (= heterogastropods), and cerithioids, with all other caenogastropods arising through the cerithioids.

Inferring relationships among unusual Paleozoic morphologies to each other and to modern groups presents many challenges. Knight met these by modeling his systematic framework on Theile's (1929) views on the relationships among modern gastropod groups. Knight also was heavily influenced by Yonge's (1947) analysis of the influence of gills and water flow currents (see Lindberg & Ponder 2001 for a discussion of Yonge's influence on perceptions of gastropod evolution). Untorted tergomyans have relatively straightforward, flow-through water currents. Torsion (an twisting of the embryo that places the anus directly above the mouth) later produced a convoluted flow, with currents entering on the right and left side of the shell, passing over the relevant gill, and exiting the shell along the midline (Fig. 1). Knight and colleagues viewed the loss of symmetry as a common solution to the problems created by combining torsion with paired gills (aspidobranchy; see Yonge 1947). Several solutions exist to the problems identified by Yonge and many affect the basic shell archetype. These include (1) increasing apertural width within isostrophic or nearly isostrophic taxa; (2) a deep sinus and/or slit in some anisostrophic pleurotomarioids; and (3) the loss of the right gill and production of a single water current in other anisostrophic gastropods. Yonge's model assumed that each modification happened in several times in parallel: for example, Yonge's model held that aspidobranchy was lost at least four times among the ancestors of extant gastropods. Knight et al. (1960) suggested these all losses of aspidobranchy occurred in the Paleozoic, with each loss establishing a major clade (one, the neritopsines, is not shown in Fig. 1). Indeed, Knight et al. (1960) tacitly implied that paired gills were lost even more often, as they themselves suggested that taxa such as the Trochonematoidea and the Craspedostomatoidea were polyphyletic. However, following publication of the Treatise, the systematics hardened into an evolutionary scenario.

FIGURE 1. "Consensus" phylogeny implied by Knight et al. (1960), with two subclades (macluritoids + euomphaloids, and pleurotomarioids + descendants) derived from paraphyletic bellerophontoids, and at least two "advanced" groups derived from pleurotomarioids. Bars show the number of reductions in symmetry, some of which retain the primitive aspidobranch condition, others of which involve the loss of the right gill. Arrows show the inferred inhalent and exhalent water flows. Gastropods modified from Kinght et al. (1960).

Recent Challenges to the Consensus

Studies of both modern and fossil gastropods cast numerous doubts on the consensus paradigm. Although phylogenetic analyses of modern taxa indicate that most major gastropod clades diverged early in the Paleozoic (Ponder and Lindberg 1996, 1997), they also reveal multiple shifts from internal symmetry or near-symmetry to internal asymmetry. Lindberg and Ponder (2001) emphasize the importance of the gastropod pallial cavity for gastropod evolution: structures are repeatedly reduced, with at least four separate complete losses of palliai structures in the patellogastropods, vetigastropods, neritopsines+cocculinids (although these might be independent losses), and caenogastropods+heterogas\tropods.

Paleontological studies also suggest that the archetypes used to diagnose higher taxa in the Treatise evolved multiple times. Protoconch (larval shell) and shell mineralogy characters are incongruent with basic adult shell (teleoconch) archetypes (e.g., Batten 1984; Bandel 1988; Nutzel and Bandel 2000). Because neontological work suggests that protoconch morphologies and shell mineralogies evolve much more slowly than do general teleoconch features (see, e.g., Ponder and Lindberg 1997), homoplasy among general teleoconch forms is more likely than is homoplasy among protoconch forms or shell mineralogies.

Phylogenetic analyses of teleoconchs also suggest high homoplasy among adult shell archetypes (Wagner 1999a,b). Among lower Paleozoic gastropods, the morphotypes typifying Treatise macluritoids and euomphaloids, i.e., flat-bottom shells with an exhalant canal (typically demarked by a selenizone) near the top of the aperture, evolved five times, including at least once among eotomarioids only distantly to euomphaloids (Fig. 2A). Open-coiled morphotypes also considered typical of euomphaloids appeared four times (Fig. 2B). Nearly flat-topped to turbinate shells with a highly angular, almost siphonate base and an exhalent canal halfway up the whorl or higher typify Treatise definitions of the Raphistomatidae and appear to have arisen independently within the euomphaloids, murchisonioids and eotomarioids (Fig. 2C). High-spired morphotypes with a vague siphonal notch or canal and a relatively shallow sinus typify Treatise definitions of the Plethospiridae and seemingly evolved independently twice within the murchisonioids and once in the eotomarioids (Fig. 2D). Trochoids have a fairly generalized morphology with medium-spired shells, (usually) heavily inclined apertures, little or no sinus, and strongly reduced exhalent channels near the suture. Morris and Cleevely (1981) already have proposed that various families placed in the Trochoidea by the Treatise were a polyphyletic assemblage related to euomphaloids. Wagner's (1999b) analyses suggest that trochoid morphotypes are even more widely distributed, appearing among the euomphaloid, murchisonioid, and trochonematoid clades (Fig. 2E). Finally, Wagner's results suggest that high-spired shells with siphonate bases and exhalent notches at or near the suture and little or no sinus that typify Treatise subulitoids arose at least twice by the end of the Silurian and once more in the Devonian (Wagner unpublished), albeit all within the murchisonioids (Fig. 2F). Nutzel et al.'s (2000) analyses of middle-late Paleozoic subulitoids further corroborates this pattern. Wagner (2001b) summarized other likely examples of similar architectural forms that likely evolved multiple times among Paleozoic gastropods.

TABLE 1. Compound characters and compound character states. See supplementary information for the resultant data matrix.

Analyses

Gastropod Data

Morphotypes and Compound Shell Characters.-We examine evolutionary patterns among five general shell features (Table 1): (1) spire angle, (2) exhalent current position, (3) base angle, (4) umbilicus width, and (5) apertural inclination (Figs. 3, 4). These features are amalgamations of multiple independent (or at least semi- independent) shell characters; therefore, we refer to them as compound characters. One can make exact measurements of all of these compound characters. However, compound characters vary within species and even within individuals over ontogeny. Also, workers have based functional and taxonomic interpretations on general conditions of these compound characters rather than on exact conditions. Therefore, we partition each compound character into qualitative "states" (Table 1). Each unique combination of compound character states represents a unique morphotype.

TABLE 2. Characters affecting gross morphotype features. (See Wagner 1999a,b for discussions of phylogenetic characters and phylogenetic character states.)

We will briefly review each of these compound characters. We caution that the functional interpretations of these characters do not apply to small snails (i.e., <5 mm in maximum height or width) that live in low Reynold's number environments, where viscosity overrides gravity. These conditions greatly alter the association between shell shape and how the animal balances the shell or between shell shape and how water flows through the mantle (Palmer 1980; Chaffee and Lindberg 1986). This study includes only a few small species. However their inclusion introduces a conservative bias: small taxa theoretically might disperse into regions of morphospace unavailable to large taxa, which would discourage rejection of the null hypothesis (see below).

Spire angle obviously reflects coiling parameters, but it also reflects the general shape and orientation of the aperture (Table 2). For example, shells with lenticular apertures (e.g., Fig. 3B,D) will have greater spire angles than otherwise similar shells with round apertures (e.g., Fig. 3C). Spire height also will increase as the long axis of the aperture becomes perpendicular to the coiling axis (e.g., Fig. 3B vs. Fig. 3D). Aperture shape and orientation are themselves compound features that reflect the dimensions and orientations of the left and right ramps as well as the orientation of the inner margin.

Spire angle affects functional interpretations such as ease of mobility (Linsley 1977), orientation on the substrate (Linsley et al. 1978), and resistance to breakage (Vermeij 1977, 1987). Traditionally, most high-spired species have been assigned to murchisonioid, loxonematoid, or subulitoid families; most taxa with very low (including "negative") spire angles are assigned to euomphaloids or macluritoid families, and most taxa with intermediate spire heights are assigned to pleurotomarioid or trochoid families.

Features such as the selenizone, sinus apex, and/or anal notch denote the position of the exhalent current on most shells. This in turn reflects characters affecting apertural orientation (e.g., the angle at which the inner margin projects from the coiling axis; the angles between ramps) and symmetry (e.g., the relative dimensions of the left and right ramps) (Table 2). The position of the exhalent current plays a prominent role in the inferred symmetry of the animal within the shell as well as the inferred life orientation of the shell. Traditional taxonomy typically assigns taxa with exhalent current positions moderately high on the whorl (e.g., Fig. 3B) to particular pleurotomarioid or murchisonioid families such as the Raphistomatidae or the Plethospiridae. Similarly, traditional taxonomy assigns tightly coiled, moderate-to-high-spired taxa with highly reduced sinuses and/or anal notches near the suture to subulitoid families and those with sinuses lower on the whorl to loxonematoid families. Finally, taxa with loose coiling, low spires, and selenizones, and with sinuses at the top of the aperture (i.e., ~"12 o'clock") typically are assigned to euomphaloid or macluritoid families.

We measure umbilicus size as a modified version of Raup's D (Raup 1966), i.e., the umbilicus "radius" divided by the sum of the umbilicus radius and the aperture width (Fig. 3). Coiling parameters, especially curvature, strongly affect this feature. Several other features also affect umbilicus size (Table 2). The orientation of the inner margin (= columella) is important, as a shell with its columella parallel to the coiling axis (e.g., Fig. 3E) will have smaller umbilici than will an otherwise identical shell with its columella angling away from the coiling axis (e.g., 3E vs. 3B or 3D). However, whereas workers typically measure D from the inner margin, we measure it from when shell material starts. Therefore, columellar thickness and orientation also affect umbilicus size (e.g., Fig. 3B,D). Some snails use a localized thickening (i.e., a callus; Fig. 3C) to reduce the umbilicus relative to shells with otherwise similar characters. Other shells extend or fold the columella forward in such a way that it fills the umbilicus (e.g., Fig. 3B). Finally, taxa such as Liospira partially or totally fill the umbilicus with the parietal inductura (i.e., a funicle).

Large umbilici likely make shells susceptible to breakage (e.g., Vermeij 1977, 1987). Signor and Brett (1984) document trends towards decreased umbilical size for both bellerophonts and ammonites following the diversification of jawed fishes in the Devonian. Umbilicus size also affects the center of shell gravity and thus inferred ease of locomotion, with large umbilici making shells more difficult to carry than otherwise similar shells with small umbilici (Linsley 1978). Traditionally, workers classify most taxa with large umbilici within the euomphaloids, whereas most taxa with small umbilici are assigned to pleurotomarioid or trochoid families.

The basal angle reflects the shape of the base. The presence of a siphonal notch or canal creates a very low basal angle (e.g., Fig. 3F). The general shape and dimensions of the base of the aperture (e.g., Fig. 3C,D) and/or the angle between the inner margin and the aperture base also affect this feature, especially on non-siphonate taxa. The orientation and shape of the inner margin relative to the coiling axis affect the base shape on many nearly planispiral species. Such taxa use the inner margin to make part or all of the shell's base, (e.g., Fig. 3A), which means that the base is homologous with the columella of other shells.

Inhalent siphons are common on predatory gastropods. Traditionally, workers classify taxa with acute basal angles and selenizones in the Plethospiridae and taxa with moderate basal angles and selenizones in the Raphistomatidae. Workers assign taxa lacking selenizones but with acute basal angles to subulitoid families. Workers typically place moderate-to-high-spired taxa with very wide basal angles and se\lenizones in the Luciellidae and moderate-to-high-spired taxa without selenizones in the Pseudophoridae. A flat or very nearly flat base might encourage sedentary habits on soft substrates (e.g., Yochelson 1971), and such taxa usually are classified as euomphaloids.

Unlike the other compound characters used here, apertural inclination (Fig. 4) is not emphasized in traditional suprageneric classifications. However, this compound character strongly affects how the snail carries the shell (with inclined apertures encouraging balancing of the shell over the gastropods foot) and thus affects the animal's mobility (Linsley 1977, 1978). Several specific shell characters affect apertural inclination, including the projection of the columella, the shape of the base, and the inclinations of the left and right ramps independently (Table 2).

FIGURE 2. Independent evolution of morphologies typifying major Treatise taxa. A, "Classic" euomphaloids. B, Open-coiled euomphaloids. C, Raphistomatids. D, Plethospirids. E, Trochoids. F, Subulitoids. Figures modified from Knight et al. (1960). Clade names reflect taxonomy of Wagner (1999b), with Trochonematoidea replacing Lophospiroidea.

FIGURE 2. Independent evolution of morphologies typifying major Treatise taxa. A, "Classic" euomphaloids. B, Open-coiled euomphaloids. C, Raphistomatids. D, Plethospirids. E, Trochoids. F, Subulitoids. Figures modified from Knight et al. (1960). Clade names reflect taxonomy of Wagner (1999b), with Trochonematoidea replacing Lophospiroidea.

Species Analyzed.-Our analyses are based on morphometric measurements of 626 Late Cambrian-Middle Devonian (Givetian) species (see supplemental information online at http:// dx.doi.org/10.1666/ 04092.s1). We based most measurements on photographs or digital images of museum specimens, supplemented with published figures. Given the coding scheme shown in Table 1, these species represent 162 of the 1920 (6 4 5 4 4) possible morphotypes.

The species analyzed covered the breadth of early gastropod phylogeny. However, we excluded bilaterally symmetrical bellerophontoids and "true" subulitids (i.e., species correctly assigned to Subulites, Fusispira, and Cyrtospira) because they show no variation in two or more of the compound characters. Inclusion of these taxa would have biased results in favor of rejecting the null hypothesis by making some morphotypes overly species-rich. However, our analyses did include species assigned to the genus Porcellia that secondarily reacquired a bellerophontoid form. We also included Siluro-Devonian species assigned to subulitiform taxa such as Macrochilina, Decorochilina, and Nanochilina, as this phylogenetic group displays a broader range of morphotypes (including neritaeform species) than do true subulitids.

We also excluded platyceratoids and oriostomatoids. Although there is substantial variation in morphotypes in these two taxa, we have little understanding of the phylogenetic relationships among platyceratoid or oriostomatoid species because they lack comparable shell characters. Their exclusion likely has a conservative bias on the analyses (see "Discussion").

A neighbor-joining cluster analysis of average pairwise dissimilarities (Sneath and Sokal 1973: p. 124) among the 626 species corroborates our suspicion that the Treatise classification of Knight et al. (1960) reflects gross morphotypes (Fig. 5). Here, dissimilarity for each compound character is based on the absolute difference between states given in Table 2. In contrast to inferred phylogenetic relationships, morphotypes reflect traditional classification well.

FIGURE 3. Measurements used to evaluate patterns of morphologic evolution. Abbreviations for measurements: AW, apertural width; BA, base angle; EP, position of the exhalent current; SA, spire angle; UR, umbilical radius. Abbreviations of other morphological features: bm, basal margin; ca, coiling axis; ex, exhalant; im, inner margin; lr, left ramp, rr, right ramp. A, Lytospira (Treatise euomphaloid). B, Raphistoma (Treatise raphistomatid pleurotomarioid). C, Anomphalus (Treatise anomphalid trochine). D, Rhombella (Treatise pleurotomarioid). E, Hormotoma (Treatise murchisoniid). F, Subulites (Treatise subulitoid). Figures modified from Knight et al. (1960).

Phylogenetic Model.-We use the phylogenies of Wagner (1999a,b) as the backbone of our phylogenetic model for Late Cambrian-Silurian species. These are modified and augmented by the addition of Devonian species (Wagner unpublished data) and methodological advancements (e.g., Wagner 2001a). The resulting tree suggests over 400 transitions among the 162 observed morphotypes (Fig. 5; see supplementary information).

FIGURE 4. Apertural inclination. A, Whole aperture inclined ~40. B, Aperture inclined ~30 with the base and left ramp inclined much more steeply than the right ramp. C, Aperture inclined ~30 with right ramp inclined much more steeply than the base and left ramp. Modified from Knight et al. (1960) and Wagner (1999b).

A common concern when using model phylogenies to analyze character evolution is that there might be an element of circularity if the inferred phylogenies relied on those characters (Harvey and Pagel 1991). This is not strictly true of the compound characters examined here, but the phylogenetic model does rely on the constituent characters those compound characters comprise (see Table 2). A possible concern here is that these constituent characters do not evolve independently of one another because of the gross shell features that they create. Independent evolution is one of the simplifying assumptions of all phylogenetic methods that do not explicitly account for it (Edwards and Cavalli-Sforza 1964) and simulations show that nonindependent change wreak havoc on such methods (e.g., Wagner 1998; Huelsenbeck and Nielsen 1999). Thus, our phylogenetic model might be very inaccurate and grossly underestimate the number of transitions (see, e.g., Bandel 2002). The generally high rates of homoplasy and finite character space for shell characters magnify this concern (e.g., Fox et al. 1999; Wagner 2000a,b). However, underestimating the number of times morphotypes evolved reduces the expected number of morphotypes. We emphasize that this makes our analyses conservative by reducing our chances of rejecting the null hypotheses.

Methods

Monte Carlo Assessments of Null Distributions.-Several tree- based tests exist to test hypotheses about biased character evolution or shifts in extinction/speciation associated with characters (e.g., McShea 1994; Wagner 1996). However, such tests are inadequate for testing the hypotheses that we are considering. Instead of asking whether changes (or speciation) tends to be in one direction along a linear gradient, we are asking whether "cells" in a multivariate volume evolve more frequently than expected and/or are too speciose. Methods that test for bimodal trends exist (Alroy 1998), but extending these to multiple dimensions remains computationally unfeasible. Analytic tree-based tests for trends or correlated change (e.g., Maddison 1990) can reject null hypotheses of random change along particular linear gradients of morphospace or hypotheses of common patterns of character correlations throughout the morphospace. However, our test hypotheses predict that the correlations will differ in different regions of morphospace.

Monte Carlo (MC) simulations represent an extremely useful way to approximate the probabilities of morphospace clustering. These can determine the probabilities of the observed patterns under the null hypothesis of diffusion into a bounded morphospace given empirically derived models of phylogeny and morphological change (see below). Essentially, this extends to multiple dimensions the test for the effects of phylogeny on trends used by Wagner (1996). This generates the expectations for the null hypothesis that a model phylogeny and model set of transitions account for an observed distribution of morphotypes. The test begins with a morphotype that likely represents the ancestral morphotype for anisostrophic gastropods (Strepsodiscus major, a very slightly asymmetrical bellerophont species [see Knight 1948; Wagner 1999b]). Each MC simulation uses a topology matching the model phylogeny and distributes the same number of changes (782) across the same branches (424 total) as posited by the model tree. Changes were drawn from the distribution of changes implied by that model phylogeny (Fig. 6) and then "added" to the ancestral condition. This proceeded from the base of the tree upwards, and thus created Markovian distributions.

FIGURE 5. Cluster analysis of 626 species based on compound characters and compound character states given in Table 1. Typical Treatise classifications are included.

FIGURE 6. Transitions across model phylogeny for the five compound characters. Transitions reflect shifts among "states" listed in Table 1. Thus, a shift from "2" to "1" is -1 whereas a shift from "2" to "3" is +1. Simulations sample from these distributions each time a compound character changes. Note that these distributions reflect only branches on which change occurred.

Each MC simulation randomly samples from the distributions of changes implied by the model tree (Fig. 6), with the probability of change for any one compound character equal to its proportion of the total changes. For example, spire height changes 160 times, so the probability of it changing is 160/764 = 0.207. Note that a compound character was permitted to change only once per branch. Using these distributions has two implications. First, the MC simulations explicitly used any biased directional change in compound shell characters (i.e., a driven trend sensu McShea 1994). second, most changes are a single "step" for any compound character. The combination of these two conditions encouraged parallel evolution of morphotypes from the same common ancesto\r, and thus discourages rejecting the null hypothesis. Also, the simulations do not permit the morphospace to exceed that actually observed. Thus, if an MC simulation initially tries to increase the spire height from an ancestor with the maximum spire height, then the simulation discarded that change and replaced it with another. This was repeated until all derived morphotypes fell within the bounds of the observed morphospace. This also induces passive trends near architectural boundaries and further encourages parallelisms under the null hypothesis (see, e.g., Foote 1994; Wagner 2000a).

One might think that it is more appropriate to evolve the underlying phylogenetic characters over the phylogeny instead of the compound characters. However, we are testing hypotheses about the distribution of the compound characters, not their constituent homologies. Thus, the exact processes generating the morphotypes are not important, only the distributions of those morphotypes across phylogeny. Ultimately, the approach that we use here is appropriate for any heritable trait, regardless of how it is inherited. Thus, one could use this approach to examine the distribution features such as environmental preference. Ultimately, given that the underlying homologies are themselves compound features when viewed from the level of developmental genetics, one could even use this premise that only the most basic elements of characters truly are inherited to dismiss evaluating any morphological traits across phylogeny.

FIGURE 7. Expected numbers of compound character state pairs given 626 species, 782 changes in the five morphotypes, and no architectural constraints. Arrow denotes the observed (174 condition pairs). Runs with 174 or fewer pairs (in gray) represent only 3.9% of all runs.

Finally, note that the MC simulations provide a "multitailed" test. That is, it does not test whether specific regions of morphospace evolve too many times or are too speciose, but whether we expect any regions of morphospace to evolve so frequently or to be so speciose. This is appropriate because the null hypothesis is that we will get some degree of clustering and repletion simply because of the Markovian distribution of the data.

Accommodating Architectural Constraints.-The only architectural constraints implicit to the MC simulations described above are those of a bounded morphospace. These constraints do not include unfeasible regions within that bounded morphospace, e.g., combinations of compound characters that are architecturally unfeasible. For example, Raup (1966) noted that shells that are very high-spired but loosely coiled are not observed in nature even though both very high-spired and loosely coiled shells are observed in nature. Raup attributes this to the combination being architecturally unfeasible.

The data show 174 of the 210 possible compound character state pairs given our coding scheme. However, the simulations described above yield 174 or fewer combinations in only 3.9% of the runs (Fig. 7). This is sufficient to reject the idea that all combinations are equally viable. The number of unobserved combinations (36 total) represents an upper bound on what is architecturally impossible. To examine the maximum effects of architectural constraints + phylogeny, we use only simulations that yield 174 combinations. To maintain a multitailed test, we did not require that the 36 unevolved combinations in the simulations match those of the actual data. Clustering and/or parallelism beyond this necessarily exceeds the expectations of phylogenetic autocorrelation, inferred morphotype change, and possible architectural constraints. Given the logic presented above, this leaves only ecomorphology as a viable explanation.

FIGURE 8. Observed vs. expected exhaustion curves for the accumulation of gastropod morphotypes. Expectations and 95% error bars are based on 15,000 replications that generated only 174 state pairs. See Wagner (2000b) for explanation of exhaustion curves.

We use these simulations to generate expected distributions of (1) numbers of morphotypes, (2) species richness within morphotypes, and (3) frequencies of morphotype evolution given the constraint of 174 compound character combinations. Again, note that we do not compare particular simulated morphotypes to observed data. Instead, we determine how frequently we can expect any morphotype to have 10, 11, 12, etc. taxa, and how frequently we can expect any morphotype to evolve 10, 11, 12, etc., times.

FIGURE 9. Observed versus expected richness for morphotypes ranked by richness. Error bars reflect 95% confidence intervals from 15,000 Monte Carlo simulations.

Results

Number of Morphotypes.-We expect far more than the observed 162 morphotypes even when simulations evolve only 174 compound character state pairs: in 15,000 such simulations, only one evolved as few as 203 morphotypes. Indeed, the observed pattern deviates significantly from expectations after only 50 morphotype transitions even given the restriction of only compound character combinations (Fig. 8). This indicates both that morphotypes were rapidly exhausted (see Wagner 2000b) and that morphotypes were recycled far beyond the expectations of simple architectural constraints.

Observed versus Expected Morphotype Species Richness.-Species richness is concentrated in fewer morphotypes than expected given the models of phylogeny and compound character change and a limit of 174 state pairs (Fig. 9). Observed deviations from expectations include both morphotypes with too many species and morphotypes with too few. Morphotypes with seven or more species (i.e., the 27 richest morphotypes) all have significantly more species than expected. Moreover, there are more morphotypes with six or more species (32) than expected even with 95% error bars (28). In most cases, the difference is appreciable: 19 of the top 22 exceed the 95% error bars by two or more species. Conversely, we observe only 60 morphotypes with three or more species despite expecting (given 95% error bars) at least 68 such morphotypes. Similarly, we observe only 100 morphotypes with 2+ species despite expecting at least 117. Thus, the pattern underlying the exhaustion of morphotypes (Fig. 8) reflects not simply a few morphotypes being overly rich, but numerous morphotypes being both species poor and species rich.

Inferred versus Expected Morphotype Originations.-The most commonly derived morphotypes evolve far more frequently on our model phylogeny than expected given the MC simulations (Fig. 10). Here, 23 of the 25 and 31 of the 40 most commonly derived morphotypes evolve significantly more frequently on the reconstructed phylogeny than they do in the MC simulations based on the same parameters. Unlike the observed richness pattern, however, none of the morphotypes evolve less frequently than expected given the model phylogeny and frequencies of change and 174 state pair limit.

Discussion

We can reject the null hypothesis that the evolution of any particular morphotype is equally probable. We can further reject the idea that this is due to simple architectural limitations. Here we discuss some possible explanations for these patterns.

Abiotic Explanations

Model Error.-A universal concern is whether model error might bias results. For tree-based studies, inaccurate phylogeny is an obvious source of error. Of course, it is highly improbable that the model phylogeny used here is entirely correct. Phylogenetic methods (including those using stratigraphic data and/or likelihood approaches) are strongly biased in favor of trees that underestimate change (Wagner 1998). However, for our results to be an artifact of phylogenetic error, the phylogeny must grossly overestimate morphotype transitions: the observed number of morphotypes is expected only if 65% of reconstructed transitions occur (Fig. 8). Preserving the null hypotheses would require trees giving much more weight to general morphotypes than to the constituent shell characters. We consider such alternatives very unlikely given studies showing that gross shell shapes are not congruent with soft anatomy (e.g., Kool 1993) and other studies suggesting that substantial structure consistent with phylogenetic autocorrelation exists among constituent shell characters underlying those gross morphotypes (e.g., Wagner 1995, 2001b; Vermeij and Carlson 2000; Papadopoulos et al. 2004).

FIGURE 10. Observed versus expected numbers of derivations ranked by the number of times evolved. Error bars as in Figure 9.

Preservation Biases.-Another possible explanation is that some morphotypes are easier to sample than others. If so, then overly species-rich morphotypes might represent "taphonomic optima" rather than ecomorphological ones. This explanation also is extremely unlikely. The only obvious taphonomic bias affiliated with gross morphotypes known to us is a tendency for very high-spired species to be poorly preserved in localities where other gastropod species are well preserved (Wagner personal observation). Another bias is the tendency for calcitic shells to preserve more easily than aragonitic shells (e.g., Cherns and Wright 2000). However, calcitic shells typically occur among medium- and high-spired shells rather than high-spired shells (e.g., Yochelson et al. 1967; Batten 1984). Moreover, two prominent taxa with calcitic shells (the platyceratids and the oriostomatids) are excluded from these analyses owing to difficulties in placing them phylogenetically. Given that high- spired morphotypes are prominent among the morphotypes that are species rich and frequently derived, these biases should be dampening signal rather than creating it.

Shell characters such as elaborate shell sculpture and spiral ornament make it easier to distinguish species with otherwise similar shell morphologies (see, e.g., Schopf et al. 1975). Such features therefore increase the number of identifiable species wi\thin a morphotype. However, these traits vary independently of the gross morphotypes documented here. Inornate species abound among the richest morphotypes and various types of ornament occur among species in "rare" morphotypes.

Biotic Explanations

We can think of no compelling arguments suggesting that abiotic factors could produce the patterns documented here. Therefore, biological processes are most likely responsible for the patterns of gross convergence.

Ecomorphological Archetypes.-The most likely explanation for the patterns documented here is that several morphotypes represent architectural attractors during the Early Paleozoic. That is, these forms likely facilitated basic functions better than did the many other forms that evolved much less frequently or not at all. Moreover, the richest and most frequently evolving morphotypes group species assigned to familiar families and genera that epitomize the common archetypes of the Paleozoic. We recognize the following primary archetypes among these morphotypes (Table 3): euomphaliform, pleurotomariiform, murchisoniiform, loxonematiform (= turritelliform), trochiform, and subulitiform. Here, we will summarize the implications of each group for trends in basic ecology and/or soft anatomy.

Euomphaliforms encompass several of the significantly overly species-rich and excessively derived morphotypes. This strongly suggests that the archetype represents an attractor in general morphospace. However, the euomphaliform archetype becomes markedly less diverse and less abundant in the Mesozoic and it is practically nonexistent in the modern oceans, suggesting that this archetype ceased to be an attractor. It follows from this that euomphaliform morphotypes likely were associated with a general ecological strategy that was feasible in the Paleozoic, but which became less viable over time. Sessile epifaunal habits represent a general ecological strategy with such a history. The diversification and increased abundance of carnivores (e.g., Vermeij 1987) and increased bioturbation (e.g., Thayer 1979) are likely causes for the reduced viability of the general strategy. This interpretation also conforms to functional analyses suggesting that euomphaliforms were sessile epifaunal suspension feeders (Yochelson 1971; Linsley 1977, 1978; Linsley et al. 1978; Morris and Cleevely 1981; Morris 1991). There are no other general ecologic strategies for which euomphaliforms seem suited that also have declined over time. Thus, the conclusion that they were sessile, epifaunal organisms is difficult to reject. Euomphaliforms frequently evolved from pleurotomariiform and trochiform morphotypes that are associated with active mobile grazing, suggesting that the transition from active to sessile lifestyles was not difficult.

TABLE 3. The general archetypes of the morphotypes with four or more derivations. In all cases, both the number of derivations and the richness are higher than expected given Monte Carlo simulations. See Table 2 for morphotype state values.

Subulitiforms present a scenario nearly opposite that of euomphaliforms. Workers typically interpret shells with moderate-to- high-spired shells, highly angular (siphonate) bases, and anal notches at the top of the aperture as belonging to predatory gastropods (e.g., Linsley 1977). For example, most neogastropods would fall in the subulitiform archetype. We also include in this archetype the "plethospirid" morphotypes (e.g., morphotype 11342), which still retain sinuses and selenizones despite siphonate bases and highly asymmetrical apertures. The subulitiform archetype is extremely common among modern gastropods and typically is considered "advanced" (e.g., Knight et al. 1960). Although this archetype was a less strong attractor during the early Paleozoic than were many other archetypes, it nevertheless evolved more frequently than expected given our null hypotheses. This in turn suggests that predatory habits have long been important among gastropods.

Classic "archaeogastropod" archetypes (e.g., pleurotomariiform and trochiform) evolve frequently and typically are quite speciose. The combination of tangential apertures, intermediate spire height, and reduced umbilicus size lowers the shells' centers of gravity relative to the spire. All else being equal, shells with a lower center of gravity should be easier to carry than those with a higher center of gravity (Linsley 1978). The pleurotomariiform morphotypes, with the anal notch in the middle of the shell indicating (near) bilateral symmetry of the soft parts, are no longer common. However, trochiform morphotypes, with greatly reduced anal notches near the suture indicating high internal asymmetry, remain both diverse and abundant in the modern world.

Finally, the murchisoniiform and loxonematiform morphotypes evolve frequently and possess very high species richness in the Paleozoic. The two archetypes are distinguished primarily by the placement of the anal notch/ sinus/selenizone, with the murchisoniiform archetype possessing greater bilateral symmetry than the loxonematiform archetype. Modern snails with such shells typically are slow-moving shell draggers and frequently engage in suspension feeding (e.g., Linsley 1978; Hughes 1986). The two archetypes likely reflect a similar ecomorphological type (or range of types), with the differences due primarily to the soft anatomy of the snail.

Implications for Trends in Internal Anatomy.-Inferred exhalent currents set "high" on the whorl frequently accompany left-right apertural asymmetry and diagnoses trochiform, loxonematiform, and (especially) subulitiform morphotypes. This also occurs among some murchisoniiform and euomphaliform morphotypes. The functional biology of the internal anatomy, especially the size and symmetry of the pallial cavity, offers plausible explanations for the repeated evolution of these archetypes. Our results suggest that the general "Yongeian" trend toward internal asymmetry (see above) was more pervasive than Yonge (1947, 1960; Morton and Yonge 1964) envisioned. This corroborates Lindberg and Ponder's (2001) conclusions based on extant gastropods.

Explanations invoking ecomorphology and internal anatomy converge in subulitiforms. Reduction of the right side accompanied the development of siphons to facilitate laminar water flow within the mantle cavity (Lindberg and Ponder 2001) and precise location of potential prey. As noted above, this general archetype is extremely common in the post-Paleozoic world, owing to the diversification of neogastropods. If predation was less common in the Paleozoic world, as some studies suggest, (e.g., Vermeij 1977, 1987) and if the functional interpretation of the subulitiform archetype is correct, then it is unsurprising that it appears less frequently than typical Paleozoic archetypes. However, this study suggests that this archetype already represented an ecomorphological attractor by the Devonian.

Evolution among Archetypes.-Although the basic archetypes are distributed broadly across gastropod phylogeny (Fig. 2), evolution from one archetype to another is nonrandom. For example, although the most probable descendant of a euomphaliform morphotype is another euomphaliform morphotype, there is a much higher probability of its giving rise to a trochiform morphotype than to a morphotype of any other archetype (Table 4). Similarly, the trochiform archetype is largely a dead-end for yielding morphotypes belonging to any archetype other than trochiform or euomphaliform. Despite the low sample sizes, these numbers are significant. For example, the one-parameter archetype transition hypothesis in which euomphaliforms might give rise to any archetype if there is a change in archetype (i.e., P = 1/5) has a log-likelihood of -17.1. A two- parameter hypothesis in which P[trochiform descendant | change from euomphaliform ancestor] = 0.6 and P[other archetype | change from euomphaliform ancestor] = 0.1 yields a log-likelihood of -9.3, which is significantly better given a standard log-likelihood ratio test (see, e.g., Sokal and Rohlf 1981: p. 695). Moreover, these analyses exclude oriostomatoids, which include both trochiform and euomphaliform morphotypes. Thus, our results probably underestimate the strength of this relationship.

TABLE 4. Archetype derivations. Transition matrix showing numbers of morphotype transitions that result in a particular archetype transition. The matrix includes only morphotypes that we could easily assign to a particular archetype.

Other archetypes appear to be more plastic, but still suggest "preferred" archetype transitions. It seems that it is easy for murchisoniiform morphotypes to have loxonematiform morphotypes descendants and vice versa. However, the loxonematiform morphotypes have a much higher probability of yielding subulitiform morphotypes than do murchisoniiform morphotypes. Finally, pleurotomariiform morphotypes apparently could give rise to morphotypes of any archetype, although there seems to be a tendency for pleurotomariiform gastropods to yield archetypes allowing bilateral symmetry if there is a major change in spire height (i.e., euomphaliform or murchisoniiform) or maintaining spire height if there is a notable change in bilateral symmetry (i.e., trochiform).

In summary, the archetype transition matrix hints at a "decision tree" that is reminiscent of the consensus phytogeny shown in Figure 1, but with the archetypes replacing clades. Thus instead of representing the phylogeny of gastropods, it instead represents a common phylogeny among gastropods. Pleurotomariiform gastropods were apt to give rise to euomphaliform, murchisoniiform, and trochiform gastropods; murchisoniiform gastropods were apt to give rise to loxonematiform gastropods, which were in turn apt to give rise to subulitiform gastropods. Notably, the one major modification, the tendency for euomphaliform and trochiform gastropods to \give rise to one another, is consistent with several revisions to the Treatise consensus (e.g., Golikov and Starobogatov 1975; Morris and Cleevely 1981).

There are two related implication of these results. First, one must take care when evaluating general phylogenetic statements. For example, Morris (1990) suggested that opisthobranchs evolved from subulitoids. This idea was difficult to reconcile with the fossil record because most phylogenetic inferences consider opisthobranchs highly derived whereas subulitoids diverged very early in gastropod evolution. However, Morris's scheme of morphologic change is perfectly consistent with the fossil record and neontological schemes after it is recognized that traditional definitions of the Subulitoidea are polyphyletic. In other words, it is very plausible that opisthobranchs evolved from a subulitiform group, but not that one that includes Subulites.

Second, this study again emphasizes the need to examine trends in multiple dimensions and with a regard to local optima (Cheetham 1987; Alroy 1998). Attempts to summarize general trends along any one of these morphological variables would miss the fact that probabilities of change depend not simply on the condition for that variable, but also on the conditions on one or more additional morphological variables.

Predictions for Other Systems

Other Archetypes.-This study does not cover additional gastropod archetypes that are common in the Paleozoic. Prominent among these are the limpets, bellerophontiform (i.e., closely coiled planispiral shells), and naticiform gastropods. Limpets clearly evolved many times among extant gastropods (Ponder and Lindberg 1997). By the Devonian, limpets appeared among platyceratoid gastropods. Indeed, the limpet archetype likely represents a molluscan archetype rather than simply a gastropod archetype as all extant tergomyans are limpets (e.g., Wingstrand 1985). It also is a matter of debate as to whether any of the limpetiform molluscs that existed before platyceratoids are gastropods (see, e.g., Wahlman 1992; Peel and Horn 1999), and it is considered nearly certain that many Cambrian and Ordovician limpets were non-gastropod molluscs (e.g., Knight and Yochelson 1960). Indeed, the best cases that we have for pre- Devonian gastropod limpets come from the Bellerophontoidea, which include the Crepidulalike Carinaropsidae and the Pterothecidae.

The bellerophontoid archetype likely is primitive for gastropods, but it arises at least once more by the Devonian with the appearance of the Porcellidae. Workers have suggested independent derivations of the bellerophontiform archetype among Carboniferous gastropods (e.g., Peel 2001). Moreover, shell mineralogy synapomorphies link late Paleozoic bellerophontiform taxa such as Euphemites to extant fissurelloids (MacClintock 1968; McLean 1984). Given that extant phylogenies necessitate a normally coiled vetigastropod ancestor for fissurelloids, this suggests that Euphemites and relatives represent a parallel derivation of bellerophontiform shells. Still other genera such as Bucanospira, Craspedostoma, and Spirina (assigned to the Craspedostomatoidea in the Treatise) frequently converge upon a bellerophontiform morphotype, especially during the final stages of ontogeny (Erwin personal observation). Finally, the bellerophontiform archetype might also be a molluscan archetype rather than a gastropod archetype, as it is likely that the bellerophontiform cyrtonelloids are tergomyans rather than gastropods (Wahlman 1992).

Naticiform shells typifying members of the Neritoidea begin to appear in the Devonian. No morphotypes that one would label naticiform deviate significantly from our null expectations. However, this might simply be due to the late appearance of this possible archetype, and it is possible that the addition of Carboniferous taxa will reveal it to be an attractor archetype.

Protoconchs.-Ntzel and Frda (2003) document active trends in larval shell (protoconch) morphotype evolution through the Paleozoic, with small and tightly-coiled protoconchs becoming relatively more diverse than large and open-coiled protoconchs. Nutzel and Frda suggest that the trend reflected differential extinction of clades with large, opencoiled protoconchs. However, simulation studies indicate that patterns such as this are much more likely to reflect a driven trend (i.e., iterative evolution) than a sorting trend (i.e., differential speciation and/or extinction) (Simpson 2004). The driven trend also is more likely given the model phylogeny used here.

Post-Devonian Patterns.-The patterns detailed here coupled with ideas about the evolution of the general ecosphere through the Paleozoic permit predictions about expected patterns among Carboniferous and Permian gastropods. We noted above that we expect that the euomphaliform archetype became a weaker attractor or possibly even a repulsor after the apparent diversification of predators. However, euomphaliform morphotypes likely remained attractors into the Paleozoic, as classic form genera typifying the morphotype such as Straparollus and Euomphalus likely are polyphyletic (Wagner personal observation), and other euomphaloids such as Cydoscapha and Planotectus share shell features with taxa of other archetypes (Erwin personal observation). Moreover, Batten (1984) observed that Triassic-Jurassic euomphaliform genera (e.g. Discohelix, Anisostoma, Woehrmannia, and Weeksia) shared shell mineralogies with the Trochoidea rather than with the Euomphalidae. This implies both that euomphaliform morphotypes were still attractors in the Mesozoic, and that the evolutionary distance between trochiform and euomphaliform morphotypes continued to be short during the Mesozoic.

Other analyses hint that other archetypes continued to evolve iteratively. For example, the Microdomatoidea appear to represent a polyphyletic assemblage of trochiform morphotypes (Erwin personal observation). Similarly, protoconch evidence suggests that additional murchisoniiform groups evolved during the Carboniferous (Ntzel and Bandel 2000; Ntzel and Mapes 2001).

Analogous Patterns in Other Taxa.-Several studies suggest that architectural attractors exist among other taxa. For example, Stanley (1972) outlined ecomorphologic types for bivalves that likely are polyphyletic. Phylogenetic and functional analyses of brachiopods suggests high degrees of polyphyly among Permian and Triassic members of the same functional groups (Carlson 1991). Finally, general tiering patterns and filtration systems might be polyphyletic not simply within crinoid echinoderms, but also across blastozoan and crinozoan echinoderms (see, e.g., Cowen 1981). These all represent cases where workers might be able to establish ranges of possible morphotypes and examine whether morphotypes associated with particular ecological and/or functional strategies evolve more frequently than expected given apparent rates of change over model phylogenetic topologies.

Conclusions

Particular gastropod morphotypes arose repeatedly during the early Paleozoic. These morphotypes can be grouped into general archetypes that match traditionally defined taxonomic units. These results strongly corroborate ideas that general shell form reflected gastropod ecology and functional biology, and that specific shell morphotypes facilitated particular functions better than did others. Indeed, the morphotypes that evolved most frequently match the expected "ideals" given views of shell balance and water-flow patterns through the mantle cavity. Further, the suggested trends in mantle cavity evolution independently match those implied by phylogenetic analyses of extant gastropods. We conclude that through the Devonian at least, general rules associated with ecological and functional factors channeled gastropod evolution toward a restricted set of architectural attractors. Finally, our analyses indicate that some archetypes were predisposed toward giving rise to others, which suggests limitations to morphologic transitions as well as to morphospace occupation.

Traditional classification likely identified ecomorphological groups rather than phylogenetic clusters. This raises the question of what had the greater effect on evolutionary dynamics of early gastropods: clade membership or ecomorphology? This is a question of general concern that applies to numerous other taxa and should be a focus of future research.

Acknowledgments

Over a decade ago, PJ.W.'s doctoral committee forgot that analyses like these were supposed to be part of his dissertation. He thanks them for this oversight. PJ.W.'s Devonian contributions were funded by National Science Foundation grant EAR-990328. DH.E. acknowledges support from the NASA Astrobiology Institute and the Santa Fe Institute. For comments on an earlier draft of the manuscript, we thank D. Lindberg and T. Baumiller. Taxonomic and locality data for species used in this analysis are available from the P.J.W. upon request. For access to specimens used in this study, we thank K. S. W. Campbell (Australia University), A.G. Cook (Queensland Museum, Brisbane), R. J. Horn (Narodn Museum, Prague), P. Jones (Australian Museum, Sydney), D. Korn and W. Kiessling (Humboldt Museum, Berlin), and J. Todd (Natural History Museum, London).

Literature Cited

Alroy, J. 1998. Cope's rule and the dynamics of body mass evolution in North American fossil mammals. Science 280:731-734.

Bandel, K. 1988. Reprsentieren die Euomphaloidea eine natrliche Einheit der Gastropoden? Mitteilungen Geologie-Palontologie Institut der Universitt Hamburg 67:1-33.

_____. 2002. Reevaluation and classification of Carboniferous and Permian Gastropoda belonging to the Caenogastropoda and their relation. Mitteilungen Geologie-Palontologie Institut der Universitt Hamburg 86:81-188.

Batten, R. L. 1984. The calcitic wall in the Paleozoic families Euomphalidae and Platyceratidae (Archaeogastropo\da). Journal of Paleontology 58:1186-1192.

Cain, A. J. 1977. Variation in the spire index of some coiled gastropod shells, and its evolutionary significance. Proceedings of the Royal Society of London B 277:377-428.

Carlson, S. J. 1991. A phylogenetic perspective on articulate brachiopod diversity and the Permo-Triassic extinctions. Pp. 119- 142 in E. C. Dudley, ed. The unity of evolutionary biology, Vol. 1. Dioscorides Press, Portland, Ore.

Chaffee, C., and D. R. Lindberg. 1986. Larval biology of Early Cambrian molluscs: the implications of small body size. Bulletin of Marine Science 39:556-549.

Cheetham, A. H. 1987. Tempo of evolution in a Neogene bryozoan: are trends in single morphologic characters misleading? Paleobiology 13:286-296.

Cherns, L., and V. P. Wright. 2000. Missing molluscs as evidence of large-scale, early skeletal aragonite dissolution in a Silurian sea. Geology 28:791-794.

Collin, R., and R. Cipriani, 2003. Dollo's law and the re- evolution of shell coiling. Proceedings of the Royal Society of London B 270:2551-2555.

Cowen, R. 1981. Crinoid arms and banana plantations: an economic harvesting analogy. Paleobiology 7:332-343.

Edwards, A. W. F., and L. L. Cavalli-Sforza. 1964. Reconstruction of evolutionary trees. Pp. 67-76 in J. H. Heywood and J. McNeil, eds. Phenetic and phylogenetic classification. Systematics Association, London.

Felsenstein, J. 1985. Phylogenies and the comparative method. American Naturalist 125:1-15.

Foote, M. 1994. Morphological disparity in Ordovician-Devonian crinoids and the early saturation of morphological space. Paleobiology 20:320-344.

Fox, D. L., D. C. Fisher, and L. R. Leighton. 1999. Reconstructing phylogeny with and without temporal data. Science 284:1816-1819.

Golikov, A. N., and Y. I. Starobogatov. 1975. Systematics of prosobranch gastropods. Malacologia 15:185-232.

Gubanov, A. P., J. S. Peel, and I. A. Pianovskaya. 1995. Soft- sediment adaptations in a Silurian gastropod from Central Asia. Palaeontology 38:831-842.

Harvey, P. H., and M. D. Pagel. 1991. The comparative method in evolutionary biology. Oxford University Press, Oxford.

Huelsenbeck, J. P., and R. Nielsen. 1999. Effects of nonindependent substitution on phylogenetic accuracy. Systematic Biology 48:317-328.

Hughes, R. N. 1986. A functional biology of marine gastropods. Johns Hopkins University Press, Baltimore.

Knight, J. B. 1948. Further new Cambrian bellerophont gastropods. Smithsonian Miscellaneous Collections 111:1-6.

Knight, J. B., and E. L. Yochelson. 1960. Monoplacophora. Pp. 177- 184 in Moore et al. 1960.

Knight, J. B., R. L. Batten, and E. L. Yochelson. 1954. Mollusca: Gastropoda. Bulletin of the Museum of Comparative Zoology 112:173- 179.

Knight, J. B., L. R. Cox, R. L. Batten, and E. L. Yochelson. 1960. Systematic descriptions. Pp. 1169-1324 in Moore et al. 1960.

Kool, S. P. 1993. Phylogenetic analysis of the Rapaninae (Neogastropoda: Muricidae). Malacologia 35:155-259.

Lindberg, D. R., and W. F. Ponder. 2001. The influence of classification of the evolutionary interpretation of structure-a reevaluation of the evolution of the palliai cavity of gastropod molluscs. Organisms, Diversity and Evolution 1:273-299.

Linsley, R. M. 1977. Some laws of gastropod shell form. Paleobiology 3:196-206.

_____. 1978. Locomotion rates and shell form in the Gastropoda. Malacologia 17:193-206.

Linsley, R. M., and W. M. Kier. 1984. The Paragastropoda: a proposal for a new class of Paleozoic Mollusca. Malacologia 25: 241- 254.

Linsley, R. M., E. L. Yochelson, and D. M. Rohr. 1978. A reinterpretation of the mode of life of some Paleozoic frilled gastropods. Lethaia 11:105-112.

MacClintock, C. 1968. Shell structure of patelloid and bellerophontoid gastropods (Mollusca). Peabody Museum of Natural History Bulletin 22:1-140.

Maddison, W. P. 1990. A method for testing the correlated evolution of two binary characters: are gains or losses concentrated on certain branches of a phylogenetic tree? Evolution 44:539-557.

McLean, J. H. 1984. A case for derivation of the Fissurellidae from the Bellerophontacea. Malacologia 25:3-20.

McShea, D. W. 1994. Mechanisms of large-scale evolutionary trends. Evolution 48:1747-1763.

Moore, R. C., et al. 1960. Mollusca 1. Part I of R. C. Moore, ed. Treatise on invertebrate paleontology. Geological Society of America, New York, and University of Kansas Press, Lawrence.

Morris, N. J. 1990. Early radiation of the Mollusca. Pp. 73-90 in P. D. Taylor and G. P. Larwood, eds. Major evolutionary radiations. Clarendon, Oxford.

Morris, N. J., and R. J. Cleevely. 1981. Phanerotinus cristatus (Phillips) and the nature of euomphalacean gastropods, Molluscans. Bulletin of the British Museum of Natural History (Geology) 35:195- 212.

Morris, P. J. 1991. Functional morphology and phylogeny: an assessment of monophyly in the Kingdom Animalia and Paleozoic nearly- planispiral snail-like mollusks. Ph.D. dissertation. Harvard University, Cambridge.

Morton, J. E., and C. M. Yonge. 1964. Classification and structure of the Mollusca. Pp. 1-58 in K. M. Wilbur and C. M. Yonge, eds. Physiology of the Mollusca. Academic Press, New York.

Ntzel, A., and K. Bandel. 2000. Goniasmidae and Orthonemidae: two new families of the Palaeozoic Caenogastropoda (Mollusca, Gastropoda). Neues Jahrbuch fr Geologie und Palontologie Mitthandelung 2000:557-569.

Ntzel, A., and J. Frda. 2003. Paleozoic plankton revolution: evidence from early gastropod ontogeny. Geology 31:829-831.

Ntzel, A-, and R. H. Mapes. 2001. Larval and juvenile gastropods from a Carboniferous black shale: palaeoecology and implications for the evolution of the Gastropoda. Lethaia 34: 143-162.

Ntzel, A., D. H. Erwin, and R. H. Mapes. 2000. Identity and phylogeny of the Late Paleozoic Subulitoidea (Gastropoda). Journal of Paleontology 74:575-598.

Source: Paleobiology

More News in this Category