Paleobiology, 30(4), 2004, pp. 543–560 Relative abundance of Sepkoski’s evolutionary faunas in Cambrian–Ordovician deep subtidal environments in North America Shanan E. Peters Abstract.—The relative proportions of Sepkoski’s Cambrian, Paleozoic, and Modern evolutionary faunas in Cambrian–Ordovician benthic marine assemblages from mixed carbonate-shale and shale lithofacies deposited below normal wave base (herein, deep subtidal) in North America are strongly positively correlated with global relative genus richness in Sepkoski’s global compendium. The correlation between local and global faunal proportions is robust regardless of how proportions are calculated, including when local proportions are based on number of specimens. Like the global pattern, the transition between the Cambrian and Paleozoic evolutionary faunas appears to occur gradually, in that Lower Arenigian (Ibexian) deep subtidal assemblages contain approximately equal proportions of Cambrian and Paleozoic faunal elements. In agreement with previous work, an onshore-offshore differentiation of faunas is evident both within Ordovician deep subtidal communities and across a larger environmental gradient. Within the deep subtidal assemblages studied here, the Paleozoic fauna tends to have a greater proportion of individuals for a given proportion of genera than the Cambrian fauna, although both tend to accrue genera at similar rates with increasing relative abundance. The Modern evolutionary fauna appears to accrue genera more rapidly with increasing local relative abundance. The extent to which these differences reflect ecological factors such as biomass, metabolic requirements or larval recruitment patterns, taxonomic practices stemming from variable morphospace saturation, or taphonomy-related counting biases remains unclear, but it suggests the possibility that Sepkoski’s evolutionary faunas may share ecological characteristics that influence both local relative abundance and global rates of taxonomic evolution. Shanan E. Peters. Department of the Geophysical Sciences, University of Chicago, Chicago, Illinois 60637. E-mail: shananp@umich.edu Present address: Department of Geological Sciences and Museum of Paleontology, The University of Michigan, Ann Arbor, Michigan 48109 Accepted: 21 April 2004 Introduction One of the most striking patterns to emerge from a global compilation of fossil first and last occurrences is based not on absolute taxonomic counts, which are sensitive to, among other factors, variation in the amount of exposed sedimentary rock (e.g., Raup 1976; Peters and Foote 2001, 2002; Smith 2001), but rather on their relative numbers, which are unbiased by sample size. Sepkoski (1981) identified the Cambrian, Paleozoic, and Modern evolutionary faunas (EFs) as statistically defined groups of taxa that exhibit covarying richness patterns and that sequentially replace one another as the principal faunal elements during the Phanerozoic (Fig. 1). As such, the evolutionary faunas first emerged as little more than a mathematical description of a stage-by-class family richness matrix. Despite this abstraction, Sepkoski’s three great evoluq 2004 The Paleontological Society. All rights reserved. tionary faunas seemed to resonate with many paleobiologists’ intuitive understanding of the large-scale fossil record. Perhaps for this reason, the ecological and evolutionary significance of the three faunas has been explored and discussed extensively at many temporal and spatial scales (e.g., Sepkoski 1991; Ausich and Bottjer 1982; Sepkoski and Sheehan 1983; Sepkoski and Miller 1985; Bambach 1985, 1993; Westrop et al. 1995; Westrop and Adrain 1998; Li and Droser 1999; Droser and Finnegan 2003). Identifying a profoundly interesting temporal pattern is important, but distinguishing possible underlying evolutionary mechanisms is often considerably more difficult. Sepkoski hypothesized that his three faunas formed macroevolutionarily coherent groups, each with different carrying capacities and intrinsic rates of diversification. Using this theoret0094-8373/04/3004-0004/$1.00 544 SHANAN E. PETERS FIGURE 1. Sepkoski’s evolutionary faunas and unassigned genera for stages in the Phanerozoic. A, Total global genus richness, B, relative global genus richness. Data from Sepkoski’s (2002) global compilation of marine animals and animal-like protists. Because stage-to-stage variation in sample size does not influence proportion estimates, total richness, rather than boundary-crosser richness, is shown and was used to calculate proportions. Dark area, Cambrian EF; stippled area, Paleozoic EF; gray area, Modern EF; unfilled area, unassigned genera. Notice that the proportion of unassigned genera has not changed substantially since the Middle Cambrian and that the Cambrian and Ordovician Periods witnessed a rapid transition between the Cambrian and Paleozoic EFs. Data are plotted for stratigraphic stages treated as having unit length. SEPKOSKI’S EVOLUTIONARY FAUNAS ical framework, Sepkoski explained the major features of each fauna’s richness history on the basis of just a few simple parameters and several perturbations (Sepkoski 1979, 1984). Alroy (2004) questioned the extent to which genera (as opposed to the original level of analysis, families) within each evolutionary fauna exhibited the predicted negative (and possibly competitive) interactions specifically predicted by Sepkoski’s three-phased coupled logistic model, but he pointed to ecological factors, such as the Mesozoic marine revolution (Vermeij 1977), as plausible explanations for some of the globally apparent genus-level behavior of the faunas. Indeed, ecological interactions, both within and between evolutionary faunas, are implied by Sepkoski’s carrying capacities. However, quantitatively projecting global patterns into local communities and positing ecological interactions as a causal mechanism for those global patterns can be problematic (e.g., Gould 2000), particularly without at least demonstrating the potential for the ecological interactions in the first place. In some cases, the link between global richness patterns and local biotic interaction has been made quite effectively (e.g., Vermeij 1977; Lidgard et al. 1993; Sepkoski et al. 2000). In others, the commonly inferred link has been criticized (e.g., Gould and Calloway 1980, but see Sepkoski 1996). Sepkoski certainly recognized the potential importance of local, community-level observations in buttressing an argument for ecologically mediated interactions having global macroevolutionary consequences (Sepkoski and Sheehan 1983; Sepkoski and Miller 1985; Sepkoski 1991, 1996; Sepkoski et al. 2000). Nevertheless, relatively little is known about how the global relative genus richness of the three faunas is related, quantitatively, to their relative richness and abundance in local communities. Because global relative genus richness can be manifested in local communities in any number of ways, including those that demand ecological interactions (or lack thereof) completely opposed to what might be predicted on the basis of global data, understanding the local-global relationship among the faunas is important in constraining the range of hypotheses that might unite them evolutionarily. 545 Here I do not attempt to demonstrate ecological interactions between the evolutionary faunas in local communities, nor do I examine the extent to which their ecological aspects might be driving global patterns of relative richness. Instead, I tackle the rather modest question of whether interactions between the faunas, as might be inferred from their relative richness globally, are consistent with their local relative proportions. That is, this study seeks to test the extent to which Sepkoski’s global observations on the relative richness of the three faunas are quantitatively relevant within individual fossil assemblages. I do this by comparing the relative genus richness of the evolutionary faunas in Sepkoski’s (2002) global genus database to the relative genus richness and relative abundance of the faunas in local Cambrian and Ordovician benthic communities occupying environments located below normal wave base (herein, deep subtidal). I also focus on perhaps the more important and interesting relationship between local relative abundance and local relative genus richness within each of the three faunas. I also briefly visit the question of their onshore-offshore distribution (Sepkoski and Sheehan 1983; Sepkoski and Miller 1985). The Cambrian and Ordovician periods were chosen for this study because a dramatic shift between the trilobite-dominated Cambrian EF and the articulated brachiopod-dominated Paleozoic EF, as well as an initial radiation of the Modern EF, occurred during this time and because the Ordovician represents one of the most important transitions in the history of marine animal life (Sepkoski and Sheehan 1983; Droser et al. 1996; Miller 1997). Awareness of how global patterns in the evolutionary faunas are manifested in local fossil assemblages is important for understanding the Ordovician community reorganization as well as the relationship between synoptic global patterns and local communities in general. Data and Methods Sixty-three samples from well-preserved Cambrian and Ordovician assemblages deposited in storm-dominated, mixed carbonate-shale and shale facies located below normal wave base (above to below maximum 546 FIGURE 2. SHANAN E. PETERS Map of locations used in this study. storm wave base in what would be classified as deep subtidal and offshore environments) were collected from 37 locations and 30 formations in the United States and southern Canada (Fig. 2). All told, 20,088 fossil specimens belonging to more than 190 genera were counted from field collections (median sample size is 159 benthic individuals). To augment these field data with samples collected by other workers using different counting and sampling techniques, 13 samples were derived from published whole-assemblage abundance data from similar facies and depth zones (see ‘‘Literature Sources’’). The literature-derived portion of the database comprises 4990 individuals (median sample size 249 individuals). Results are consistent if literature data are excluded. The total number of specimens used in this study is therefore 25,078 representing more than 230 genera from 38 formations. The purpose of restricting depth zone and lithological sampling was to control as much as possible taphonomic variables and to allow the comparison of relatively consistent environments from the Cambrian to the Ordovician, although precise environmental control through time is difficult. This approach has the advantage of minimizing the potential effects of environment and taphonomy on temporal signals, but it also restricts inference to the chosen environment. Samples from the target environment were assigned a more specific depth zone relative to maximum storm wave base on the basis of physical sedimentary structures, taphonomic conditions, and stratigraphic context (e.g., Brett et al. 1986; Brett and Baird 1986). Although absolute water depth is an important component in determining the distribution of benthic taxa on modern marine shelves (e.g., Hill et al. 1982; Ellingsen 2002), substrate characteristics are likely to be much more important in this regard (Ellingsen 2002). Water depth is also not the most important taphonomic variable (Powell et al. 2002; Best and Kidwell 2000). Thus, even if absolute depth assignments differ between locations and time intervals, maintaining relatively consistent lithofacies may allow more relevant control than does a constant depth zone per se. To increase the geographic scope of this study and prevent documenting basin-specific effects, samples were obtained from as many locations as feasible. However, because of outcrop patterns and geographic restrictions in facies tracts owing to basin evolution and sea level variation, not all time intervals are represented over a wide geographic range in the assemblage data (Fig. 2). For example, Arenigian assemblages in these data are restricted to the Basin and Range Province, and no Tremadocian assemblages were field sampled. Because the Tremadocian represents an important interval in the Cambrian–Ordovician transition, a limited number of North American faunal lists from below normal wave-base environments were derived from the literature (see ‘‘Literature Sources’’). These SEPKOSKI’S EVOLUTIONARY FAUNAS TABLE 1. 547 Taxa assigned to each evolutionary fauna globally and locally. Cambrian EF Trilobita Hyolithida Polychaeta ‘‘Inarticulata’’ Eocrinoidea ‘‘Monoplacophora’’ Paleozoic EF Articulata Anthozoa Stenolaemata Cephalopoda Crinoidea Ostracoda Asteroidea Modern EF Gastropoda Echinoidea Demospongia Chondrichthyes Bivalvia Gymnolaemata Malcostraca Osteichthyes data consist only of faunal lists (lists of taxa found at a location without relative abundance information) and are not included in any of the quantitative analyses. When Tremadocian literature data are illustrated, they are shown with dashed lines to emphasize their distinction from the rest of the assemblage data. To achieve a wider range of environments in a comparative analysis of onshore-offshore patterns in the evolutionary faunas, 18 supplemental Lower (12 lists) and Late Ordovician (6 lists) whole-assemblage faunal lists from North American intertidal and shallow subtidal environments were obtained from the literature (see ‘‘Literature Sources’’). These data were included only for comparative purposes in the onshore-offshore analysis and did not enter into any of the quantitative analyses of the temporal patterns in faunal proportions. Sampling Methods and Binning Protocols. Field-samples were collected predominately by excavating quarries at accessible locations in sections with appropriate fossil-bearing lithofacies and by collecting all macrofossils (.1 mm). A limited number of collections were also made from contextual float when quarrying was not possible or practical. Specimens were recovered by finely splitting bedding planes and/or by examining bedding surfaces. Peters (2004) gives more detailed descriptions of sampling methods. The following identifiable skeletal elements were counted as individuals: trilobite cranidia, brachiopod valves, bivalve valves, gastropod spires, echinoderm calyxes or semi-articulated to disarticulated thecal elements, nau- tiloid phragmacones, coral colonies or solitary individuals, and ostracode valves. Bryozoans were counted as discrete colonies when possible, otherwise 1-cm lengths were treated as individuals. Because bryozoans pose special taxonomic and counting problems, they are almost certainly underrepresented in these data. Omitting bryozoans from the analysis does not, however, substantively change results. Although putatively swimming taxa, such as agnostid trilobites and nautiloids, were collected and counted, they were omitted from the analysis, which was limited to benthic assemblages. Including non-benthic taxa does not substantively change the results because swimming taxa constitute less than 7% of the total number of individuals in these data. Genera were assigned to an evolutionary fauna if they belonged to a group designated as a member of a fauna by Sepkoski (1981). Both Sepkoski’s data (Sepkoski 2002) and the assemblage data were parsed into faunas as shown in Table 1. All other taxa were not designated as belonging to any evolutionary fauna (unassigned genera). Richness for each fauna in each stage in Sepkoski’s data was counted as the total number of taxa known to have existed during that stage (i.e., total richness derived from first and last occurrence and range-through data). Assemblage samples were grouped by stratigraphic stages according to the Cambrian subdivisions of Palmer (1998). The Delamaran (9 samples), Marjuman (11 samples), Steptoean (11 samples), Sunwaptan (2 samples), lower Arenigian (Ibexian, [Webby 1998]; 548 SHANAN E. PETERS 13 samples), Caradocian (16 samples), and Ashgillian (14 samples) stages are represented in the field samples and literature data. Because Sepkoski’s (2002) data use a different stratigraphic subdivision in the Cambrian, assemblage data for the Cedaria zone and later Marjuman as well as the Steptoean were combined and compared with Sepkoski’s Dresbachian data. Marjuman assemblages older than the Cedaria zone and Delamaran assemblages were combined and compared with Sepkoski’s pooled middle Middle and upper Middle Cambrian subdivisions. The meager Sunwaptan samples in this study are from the Eurekia zone and therefore correspond to Sepkoski’s Trempealeuan data. Because not all of Sepkoski’s genus data are resolved to the level of upper or lower Arenigian, only those genera that cross the lower Arenigian stage boundary (Xbl 1 Xbt [Foote 2000]) or that are explicitly assigned a lower Arenigian first occurrence were included in the analysis. All Arenigian first occurrences unresolved to substage in Sepkoski’s data were excluded from the global tabulation. Each field sample was treated as a paleontological ‘‘grab sample’’ and proportions were calculated separately for each sample. The assemblage data for each stage were not pooled to determine a single average community composition. Rather, proportions for each fauna in each sample were calculated separately and mean values for stages were calculated by pooling these individual sample proportions. Faunal proportions were calculated in four ways at the assemblage level. The first two ways are based on the number of genera (relative genus richness) and were calculated by tallying genus richness in each evolutionary fauna and then dividing either by the total number of genera in the sample (all taxa) or by the total number of genera that could be assigned to an evolutionary fauna (faunas only). The second two ways of calculating faunal proportions in assemblages are based on the number of individuals (relative abundance). In these cases, the number of individuals representing each fauna was counted and then divided either by the total number of individuals in the sample (all taxa) or by the total num- ber of individuals that could be assigned to an evolutionary fauna (faunas only). Local Evolutionary Faunal Proportions The transition between the trilobite-dominated Cambrian EF and the brachiopod-dominated Paleozoic EF appears at the temporal scale of this study to occur gradually and uniformly in North American fossil assemblages from mixed carbonate-shale and shale environments located below normal wave base (Fig. 3). It is of course possible that a more rapid transition actually occurs in the upper Arenigian (Whiterockian [Li and Droser 1999; Droser and Finnegan 2003]) or in the Llanvirnian and Llandeilian, which are all unrepresented in this study. There is no significant change in relative genus richness in the three faunas between the Caradocian and the Ashgillian, implying that most of the transition between the Cambrian and Paleozoic EFs, at least in terms of relative genus richness in offshore environments, was completed by the Caradocian. The Modern EF, which is represented by approximately three times as many gastropods as bivalves in these assemblage data, exhibits little increase in relative genus richness during the Ordovician and remained only a small fraction of the total community in the environments studied here. In the Cambrian, genera from the Cambrian EF dominate benthic assemblages almost to the exclusion of all other elements. In the Late Ordovician, Paleozoic genera are dominant, but to a lesser degree (Fig. 3). The Early Ordovician is an important transitional interval with communities of intermediate composition that are unlike those of the Cambrian and later Ordovician. The average intermediate composition of assemblages in the Early Ordovician is not the result of pooling disparate samples that are individually Cambrian-dominated (Cambrian-like) or Paleozoic-dominated (Ordovician-like); instead, most of the Early Ordovician assemblages in the data set are composed of 40–60% Cambrian EF genera (gray points in Fig. 3) and are therefore generally distinct from older and younger assemblages. It is important to realize that in Figure 3, and in all other figures, the mean assemblage SEPKOSKI’S EVOLUTIONARY FAUNAS FIGURE 3. Relative genus richness of each evolutionary fauna in local assemblages. Gray points are values for single samples. Stippled datapoint is for literature-derived estimate in Tremadocian. Spacing of points along abscissa is proportional to geologic time. Mean 6 one standard deviation shown. proportion is obtained by taking the unweighted average of proportions observed in each sample, not by pooling all of the samples from a time interval and then calculating a single proportion for each of the three faunas. Thus, an average value of 0.58 for the Cambrian fauna in the lower Arenigian (Fig. 3) in- 549 dicates that a given sample is, on average, composed of 58% Cambrian EF genera. The intermediate composition of lower-Arenigian assemblages documented here is at least broadly consistent with the results of Li and Droser (1999: Figs. 6, 9), who also reported approximately equal abundance of shellbeds dominated by Cambrian and Paleozoic fauna elements in the lower Arenigian of the North American Basin and Range province. Li and Droser (1997) also reported data for Cambrian shell beds that are consistent with these assemblage-level data. This consistency emerges despite the fact that assemblage data presented herein were collected primarily from a very different taphofacies. Most of the lower Arenigian samples in this study were collected from the muds and thin carbonates that are in many cases directly interbedded with the skeletal tempestites and other shell accumulations tabulated by Li and Droser (1997, 1999). This suggests that the relative abundance of taxa in shell accumulations agrees broadly with presumably less time-averaged and less taphonomically altered intervening assemblages (Kidwell 1986), at least at the taxonomic level relevant to the identification of the evolutionary faunas. Additional section-by-section data from skeletal concentrations and intervening sediments are required in order to test this hypothesis, but the agreement between these two estimates of the relative richness of the evolutionary faunas is noteworthy. This is particularly true because the reliability of shellbed data has been justifiably questioned (e.g., Westrop and Adrain 1998). Another ecologically important way of tabulating faunal proportions in local assemblages is to count the total number of individuals that belong to each fauna rather than the total number of genera, as in Figure 3. Because specimens need only be identified to class level to determine their faunal affinity, this approach has the additional advantage of circumventing any potential problems, inconsistencies, or errors in genus-level taxonomy that may exist between the evolutionary faunas in this tabulation. Figure 4 presents the same assemblage data shown in Figure 3, but with evolutionary faunal proportions calculated on 550 SHANAN E. PETERS FIGURE 4. Mean relative abundance of each evolutionary fauna in local assemblages. Gray points are values for single samples. Spacing of points along abscissa is proportional to geologic time. Mean 6 one standard deviation shown. the basis of number of individuals. Although the overall temporal pattern in mean assemblage relative abundance is similar to that for mean relative genus richness, there are some notable differences between the genus- and individual-based tabulations. First, the variance in many time intervals is greater for proportions calculated on the basis of individuals (Fig. 4) than it is for relative genus richness (Fig. 3). The most notable example of this occurs in the lower Arenigian, where variance in the proportion of individuals in the Cambrian and Paleozoic EFs far exceeds variance in the proportion of genera in each fauna. In general, high variance in the lower Arenigian is consistent with the intermediate average composition of these assemblages because the standard error of sample proportions is greatest when true proportions are near 0.5 (Hayek and Buzas 1997: Fig. 8.3). However, the difference in variance between sample genus proportions and individual proportions cannot be explained by this fact alone. To evaluate potential sources of variance in the individual-based tabulation for the lower Arenigian, the influences of lithology and depth zone (which are not entirely independent in these data) were examined. Figure 5 shows the effects of these environmental parameters on the proportion of Cambrian faunal elements calculated on the basis of genera and individuals, as in Figures 3 and 4. The proportion of individuals belonging to the Cambrian fauna in the lower Arenigian is significantly different between lithologies (p 5 0.05) and depth zones (p 5 0.01) according to a nonparametric Kruskal-Wallis test. However, there is no significant variation in relative genus richness across these same environmental variables (Fig. 5; p . 0.83; results similar for ANOVA). Thus, the relative genus richness of the Cambrian EF in lower Arenigian samples tends to be consistent, regardless of depth zone or lithology, but relative abundance tends to be more variable across depth zones and lithologies. The lower Arenigian Paleozoic EF exhibits a similar pattern, but individuals tend to be more abundant in shallower water and carbonate-dominated environments. The effect in the Paleozoic EF is also not as strong (p 5 0.03 for depth effect in individuals; p 5 0.14 for lithology effect in individuals; p . 0.74 for both effects in proportion of genera). The lower Arenigian Modern EF, which is dominated by gastropods in the assemblages studied here, does tend to have a SEPKOSKI’S EVOLUTIONARY FAUNAS 551 FIGURE 5. Effects of lithology and depth on relative taxonomic richness (top) and relative abundance (bottom) of the Cambrian EF in the lower Arenigian. Gray points are values for single samples; dark lines represent mean of samples 6 one standard error of the mean. There is no significant difference in relative genus richness between depths and lithologies (p . 0.10; Kruskal-Wallis test; ANOVA), but relative abundance differs significantly (p # 0.05); wb refers to wave base; N, individuals; S, genera. See text for discussion. greater proportion of individuals in shallower-water settings and in carbonate-dominated samples, but the effect is not significant (p . 0.1). However, the possibility that gastropods (which dominate the Modern EF in these assemblage data) are more prevalent in carbonate-dominated settings than in siliciclastic settings is consistent with the lithofacies preferences reported by Novack-Gottshall and Miller (2003). Unfortunately, the sample sizes used to test for significant environmental variation in lower Arenigian faunal proportions are quite small. Nevertheless, greater between-sample variance in the proportion of individuals is the expected consequence of sampling an environmental gradient occupied by taxa that tolerate a range of environmental conditions, but that achieve peak abundance in specific settings. Provided that taphonomic differences between the evolutionary faunas have not caused the selective destruction of individuals from different faunas in different depth zones and lithologies (which is a distinct possibility), this result suggests that there are significant differences in the preferred environment of the evolutionary faunas even within the relatively restricted lithofacies and depth zone covered here. Broader-scale patterns in the environmental distribution of the evolutionary faunas are discussed below. Another difference between the genusbased and individual-based tabulations shown in Figures 3 and 4 is the strong domi- 552 SHANAN E. PETERS FIGURE 6. Relative genus richness plotted against relative abundance (proportion of specimens) for each evolutionary fauna in each sample. Lines are least-squares linear regression lines calculated after excluding assemblages with zero values for a fauna. Including zero values increases all slopes, but relative relationships remain qualitatively similar. Regressions are significant (p , 0.001); r2 values are 0.73 (Cambrian), 0.72 (Paleozoic), 0.67 (Modern). Slopes for the Cambrian and Paleozoic fauna are 0.61. Slope for Modern fauna is 0.93. See text for discussion. nance of the Paleozoic EF in the Late Ordovician when assemblage proportions are calculated on the basis of specimens. Most Late Ordovician assemblages are dominated by the Paleozoic EF in terms of specimens, but with respect to the number of genera, the Cambrian and Modern EFs tend to be better represented. The comparatively large proportion of individuals belonging to the Paleozoic EF for a given proportion of genera is a general feature of the assemblage data. Figure 6 shows local relative genus richness plotted as a function of local relative abundance for each of the three faunas. Each point in Figure 6 represents a single sample and each sample has three points (one for each of the three faunas). Least-squares linear regression lines are shown for each fauna. To prevent undue weighting of assemblages that are missing a fauna altogether, as commonly occurs in the Cambrian for example, regression lines are calculated with zero values omitted, but including zero values does not change the qualitative relationships. The slopes of the Cambrian and Paleozoic EFs shown in Figure 6 are indistinguishable, but the intercept of the Paleozoic EF is sub- stantially less than the intercept of the Cambrian EF. This indicates that an increase in the proportion of individuals tends to correspond to an equivalent increase in the proportion of genera in both the Cambrian and Paleozoic EFs, but that for a given proportion of genera, the Paleozoic EF tends to have more individuals. The slope of the Modern EF is substantially larger than the slope for the Cambrian and Paleozoic EFs, indicating that an increase in the proportion of individuals in the Modern EF results in a larger increase in relative genus richness. However, the range of values for the Modern EF is not large in these data so the nature of the richness versus abundance relationship is difficult to evaluate. The difference between the intercepts of the lines for the Cambrian and Paleozoic EFs shown in Figure 6 could be the result of variation in the way individuals are counted (see ‘‘Data and Methods’’) or systematic variation in the reliability and consistency of genus-level taxonomy between faunas. Alternatively, if the Paleozoic EF really does tend to have a greater proportion of individuals for a given proportion of genera, then this result may have many interesting biological implications relating to, for example, the ability of the Paleozoic EF to accommodate and recruit a greater number of individuals per genus per unit area of seafloor. This may be possible because of the overall low metabolic requirements (Bambach 1993) and larvae settlement preferences of brachiopods and other Paleozoic EF taxa relative to those of the Cambrian and Modern EFs. The pattern may also reflect variation in the degree to which the evolutionary faunas are differentiated morphologically. If the Cambrian EF tends to have a larger suite of distinguishing characters and a more dispersed distribution of taxa within morphospace, then many genera may be recognized for a given number of individuals. Likewise, if the Paleozoic EF tends to be comparatively depauperate in morphological characters and to have a rather densely occupied morphospace, then relatively few genera may be recognized for a given sample size. However, adequately testing of these and other hypotheses and establishing the generality of the patterns SEPKOSKI’S EVOLUTIONARY FAUNAS 553 shown in Figure 6 require additional field data. Local versus Global Evolutionary Faunal Proportions The relationship between local, assemblagelevel relative genus richness of the three faunas and their relative genus richness in Sepkoski’s global compilation is shown in Figure 7. There is remarkably good agreement between the two data sets, despite the very different scales at which they were compiled. Although the overall correlation is quite high, it is important to note that not all of the points fall perfectly on the 1:1 line. For example, the lower Arenigian (Ibexian) assemblages studied here have, on average, more genera belonging to the Cambrian fauna than would be predicted on the basis of Sepkoski’s global compilation. Although residual variation around the 1:1 line is present, evaluating the meaning of this rather small amount of variation is difficult. Subtle changes in the average sampled environment from stage to stage, for example, could easily drive residual variation (see following discussion on onshore-offshore patterns), but quantifying this effect is impossible without high-resolution environmental data. It is also possible that the residuals reflect the fact that only North American assemblages have been sampled in this study. In some cases, biogeographic differences in the apparent timing of clade diversification, such as the rather late arrival of bivalves to Laurentia (Babin 2000; Sánchez and Babin 2003; Novack-Gottshall and Miller 2003), could explain residual variation. Interestingly, local relative genus richness of the Modern EF in the lower Arenigian is far less than predicted by Sepkoski’s global compilation, which is precisely what is expected on the basis of a late arrival of bivalves to North America (Novack-Gottshall and Miller 2003: Fig. 3). If bivalves are excluded from Sepkoski’s compilation in the Early Ordovician, then the Arenigian point shown in Figure 7 falls near the one-to-one line. Another plausible explanation for residual variation is imperfect stage assignment in Sepkoski’s data. For example, many of the Arenigian genera in Sepkoski’s compilation are not resolved to substage. In the analyses pre- FIGURE 7. Mean relative genus richness in local assemblages plotted against relative genus richness in Sepkoski’s global compilation. Unassigned genera are excluded from the calculation but results are similar if they are included (Table 2). Stippled datapoint shows literature-derived Tremadocian estimate. The y-error bars are 6 one standard error of the mean; x-error bars are 95% binomial confidence limits for global proportion. Dashed lines are 1:1 lines. Points are labeled according to Sepkoski’s stages: MidC (Middle Cambrian), Dres (Dresbachian), Trep (Trempealeauan), Aren (Arenigian), Cara (Caradocian), Ashg (Ashgillian). 554 SHANAN E. PETERS sented here, all unresolved first occurrences were omitted because many of these may in fact represent upper Arenigian occurrences. Regardless of the reasons for the departures from the 1:1 line, the magnitude of the residual variation is small; the largest deviation in Figure 7 is 12%. Figure 8 shows the relationship between local assemblages and global relative genus richness when assemblage proportions are calculated on the basis of number of specimens. In general, results are consistent with the assemblage-level relative genus richness tabulation. As discussed above, one of the main differences between the genus-based and specimen-based tabulation is that Late Ordovician assemblages are more strongly dominated by the Paleozoic EF when proportions are calculated from the number of individuals. This is shown by positive residuals on the 1:1 line for the Ashgillian and Caradocian. Another difference is the low slope of the Modern EF in Figure 8. This indicates that the proportion of individuals representing the Modern fauna in local assemblages tends to be low in comparison to relative genus richness in the global compilation. This is perhaps not surprising given the known environmental preferences of bivalves (Miller 1989; NovackGottshall and Miller 2003), which are greatly outnumbered by gastropods in these assemblage data. Nevertheless, the strength of the correlation, which measures the association between global richness and local abundance, remains high for the Modern EF. Figures 7 and 8 compare faunal proportions for genera from the three faunas only. In these cases, unassigned genera (Fig. 1) were omitted from the calculations. However, including unassigned genera does not substantively affect the results. Table 2 shows the linear productmoment correlation coefficients between global and assemblage faunal proportions calculated by using all genera, including those that are unassigned to an evolutionary fauna (all taxa). Also shown in Table 2 are correlation coefficients for comparisons in which only genera belonging to an evolutionary fauna are included (faunas only), as in Figures 7 and 8. Because random time-series are often spuriously correlated, the significance of the correlations FIGURE 8. Mean relative abundance in local assemblages plotted against relative genus richness in Sepkoski’s global compilation. Unassigned genera are excluded from the calculation but results are similar if they are included (Table 2). The y-error bars are 6 one standard error of the mean; x-error bars are 95% confidence limits for global proportion. Dashed lines are 1:1 lines. Points are labeled by stage as in Figure 7. SEPKOSKI’S EVOLUTIONARY FAUNAS TABLE 2. Linear product-moment correlation coefficients between the mean proportion of an evolutionary fauna in local assemblages and the equivalent proportion in Sepkoski’s global genus compilation for six stratigraphic intervals in the Cambrian and Ordovician as in Figure 7. ‘‘Faunas only’’ refers to proportions calculated with unassigned genera omitted. ‘‘All taxa’’ refers to proportions calculated with unassigned genera included. ‘‘Genera’’ refers to assemblage proportions calculated by using the number of genera in each fauna. ‘‘Individuals’’ refers to assemblage proportions calculated by using the number of individuals. Faunas only All taxa Genera Cambrian Paleozoic Modern 0.986 0.997 0.764 0.976 0.996 0.828 Individuals Cambrian Paleozoic Modern 0.997 0.996 0.937 0.984 0.989 0.960 in Table 2 was not tested. Although correlations are consistent between tabulations, the slopes of the relationships shown in Figures 7 and 8 increase when all taxa are counted (total fauna; plots not shown). This is because Sepkoski’s global compilation includes taxa, such as radiolarians, that do not belong to an evolutionary fauna and that were not counted in local assemblages. The effect of including these taxa is to decrease uniformly the global proportions of the evolutionary faunas relative to the assemblage data. However, the strengths of the assemblage-global correlations are not markedly influenced by unassigned genera (Table 2). Environmental Distribution of the Evolutionary Faunas The large-scale patterns of faunal change during the Ordovician, including onshore-offshore expansion of the evolutionary faunas, have been well documented (e.g., Sepkoski and Sheehan 1983; Sepkoski and Miller 1985; Miller 1989; Patzkowsky 1995; but see Westrop et al. 1995; Westrop and Adrain 1998). Trilobite-dominated assemblages of the Cambrian EF occupied all environments in the Cambrian, but these were superseded during the Ordovician radiation by brachiopod-dominated assemblages of the Paleozoic EF. Sepkoski and Sheehan (1983) and Sepkoski and Miller 555 (1985) found that the Paleozoic EF first became prevalent in shallow-water environments and then rapidly expanded to become dominant over most parts of the shelf during the Ordovician radiation. They also found that the mollusk-dominated Modern EF exhibits a similar pattern, in that it first achieved dominance in nearshore environments and then slowly expanded offshore, albeit in an irregular fashion (Sepkoski 1991). Although there is widespread consensus regarding the overall pattern, the underlying evolutionary and ecological processes responsible for the shifting environmental distribution of the evolutionary faunas are still under discussion (Droser et al. 1996; Westrop and Adrain 1998). For example, Adrain et al. (1998) found that some trilobite groups depart from the stereotypical Cambrian EF syndrome and were active participants in the Ordovician radiation, suggesting that the apparent replacement of the Cambrian EF by the Paleozoic EF was more a matter of dilution than of displacement (Westrop et al. 1995; Westrop and Adrain 1998). The results of this study (Table 2, Figs. 7, 8) may at first seem to contradict an onshore-offshore gradient in the composition of communities because the gradient hypothesis predicts that faunal proportions in a given environmental zone should not reflect global relative genus richness. Instead, under the onshore-offshore hypothesis, local assemblages should be disproportionately dominated by genera from the evolutionary fauna preferentially occupying that depth zone. At the very least, environmental patterns in the distribution of the evolutionary faunas predict environmental heterogeneity in their relative richness. To evaluate the onshore-offshore signal in these data, Early Ordovician (Tremadocian and Arenigian) and Late Ordovician (Caradocian and Ashgillian) assemblages were separately assigned to three depth zones relative to maximum storm wave base (as in Fig. 5). This was possible because although the total environmental gradient encompassed in this study is relatively narrow when compared with the complete spectrum of shelf environments, sufficient variability exists to make approximate depth divisions (see discussion of 556 SHANAN E. PETERS FIGURE 9. Depth effects on the relative genus richness of Sepkoski’s evolutionary faunas in Early and Late Ordovician assemblages. Data on left represent assemblages located below normal wave base (deep subtidal). Within deep subtidal environments, assemblages were distributed according to their approximate position relative to maximum storm wave base (swb); below, near, and above swb from left to right in the deep subtidal zone. Data on right are for published faunal lists from intertidal and shallow nearshore environments. Means in each depth bin 6 one standard error of the mean are shown. Symbols for the faunas as in Figure 1. criteria used in ‘‘Data and Methods’’). A limited number of additional intertidal and shallow-water nearshore data were derived from the literature to supplement the onshore offshore analysis by providing a much wider range of depth zones and environments (see Appendix). Lithologically, the nearshore literature data range from algal-laminated dolomitic carbonates to shaly flat-pebble conglomerates. Figure 9 shows the effect of depth on the local relative genus richness of each evolutionary fauna for the Early and Late Ordovician. In Early Ordovician deep subtidal environments, there is rather little difference in mean relative genus richness within each fauna (Fig. 9). However, in Early Ordovician intertidal and nearshore environments, not only are all three faunas more equally represented in terms of relative genus richness, but their rank-order relative richness is completely different (Fig. 9). In the nearshore environments sampled here, the rarest of the three faunas in deep subtidal environments, the Modern EF, is dominant. The other two faunas have each shifted down by one rank in nearshore settings. A similar pattern is observed in the Late Ordovician, although each fauna tends to be more variable within Late Ordovician deep subtidal environments. Considerably more data on the relative and absolute abundance of the three evolutionary faunas across high-resolution environmental transects (including both absolute water depth and substrate type) are required in order to fully resolve and test the onshore-offshore pattern of faunal succession and to distinguish this pattern from substrate-driven environmental preferences (e.g., Miller 1989; Miller and Connolly 2001; Novack-Gottshall and Miller 2003). Nevertheless, these data support a broad onshore-offshore distribution of evolutionary faunas that is consistent with previous assemblage-based studies. Agreement between the proportions of evolutionary faunas in the assemblages studied here and in Sepkoski’s global genus compilation (Table 2) may be at least partly the result of choosing a depth zone that harbors most of the known Cambrian and Ordovician genera. Other environments may contribute such a small proportion of total global richness in the lower Paleozoic that their unique signals do not contribute significantly to the global compilation. Clearly, as shown in Figure 9, had this field study been conducted in intertidal and shallow-water environments, there would have been marked disagreement between the global compilation and local assemblages. Perhaps Sepkoski’s global compilation provides a good estimate of the average assemblage-level faunal composition in a range of marine environments, but not in more restrict- SEPKOSKI’S EVOLUTIONARY FAUNAS ed or more taphonomically biased settings, such as the intertidal zone. Discussion Sepkoski’s global compilation of marine animals has recently come under sharp criticism, both for its taxonomic and stratigraphic inaccuracies (e.g., Adrain and Westrop 2000; Jeffery 2001; although the former found that errors were randomly distributed and therefore of little consequence to the overall pattern) and for its naivety with respect to large-scale bias by differential fossil preservation in time and space (e.g., Raup 1976; Allison and Briggs 1993; Cherns and Wright 2000; Peters and Foote 2001, 2002; Smith 2001; Westrop and Adrain 2001; Wright et al. 2003). Nevertheless, some of the most interesting and fundamental features of Sepkoski’s global compilation, and other similar global taxonomic databases, are based on relative proportions and are therefore immune to many of the first-order sample size biases that afflict global fossil data. The patterns of relative genus richness among the evolutionary faunas in Sepkoski’s (2002) global genus compilation appear to be very accurately borne out within individual marine assemblages from deep subtidal environments in North America. This is remarkable and unexpected given the dramatically different scales of analysis and fundamentally disparate nature of the data. It is also surprising given that regional factors can influence the structure and composition of local communities. Miller (1997) pointed out that disparate regional effects, summed over the global record, produce the patterns that emerge in global compilations. For example, Patzkowsky and Holland (1993) credited a regional decline in the importance of brachiopods across a stratigraphic boundary in the Middle Ordovician of eastern North America to shifting water mass characteristics and sediment input instigated by the Taconic orogeny. Bivalves and gastropods also appear to have had different histories on different paleocontinents in the Ordovician (Babin 2000; Sánchez and Babin 2003; Novack-Gottshall and Miller 2003). Even if there is not geographic variation in biological patterns, unequal sampling of marine habitats in time and space can influence large- 557 scale paleobiological patterns (e.g., Westrop and Adrain 2001). It is likely that some of the rather small amount of residual variation in the local-global relationships results from disparate regional histories and unequal environmental sampling in Sepkoski’s data, but the results presented here suggest that Sepkoski’s global data can serve as a reasonable proxy for local assemblage composition in some environments and vice versa. The agreement between faunal proportions in local assemblages and in global compilations may also have biological implications for how communities are assembled. For example, the correspondence may indicate that the communities studied here are composed of taxa that are drawn randomly from the available global pool. Alternatively, and perhaps more likely, the agreement may indicate that most of the known Cambrian and Ordovician genera occupy this depth zone so that their signal overwhelmingly dominates the global pattern. This interpretation is supported by several studies that have identified deep subtidal environments as among the most taxonomically rich environments in the Paleozoic fossil record (e.g., Lockley 1983; Sepkoski 1988; Patzkowsky 1995; Adrain et al. 2000). Of course it is also possible that the agreement between local and global proportions reflects disproportionate representation of North America in the published paleontological literature and therefore geographic bias in Sepkoski’s global genus database. Testing the environmental and geographic sampling in Sepkoski’s data requires additional information about the environmental and geographic distributions of the constituent taxa and is therefore difficult at this time. However, at least some of the correspondence between local and global proportions may indicate that the environments studied here harbor much of the world’s known marine biodiversity and that North America is an important component of Sepkoski’s compilation. Even so, this would not necessitate a strong positive correlation between relative richness estimated on the basis of literature-compiled genus first and last occurrences and local community composition. Thus, the possibility that Sepkoski’s compilation is North American biased does not 558 SHANAN E. PETERS detract from the fact that it appears to have captured ecologically relevant parameters in some Cambrian and Ordovician deep subtidal communities. Sepkoski’s evolutionary faunas have been discussed extensively, and this study may be interesting for its simultaneous treatment of global and local patterns, but a perennial question goes something like this: ‘‘What are the evolutionary faunas anyway?’’ When analyzed at the global scale, the answer is quite simple, but somewhat dissatisfying. Globally, the evolutionary faunas are no more than groups of taxa with covarying richness patterns and characteristic rates of turnover that sequentially replace one another as the dominant faunal elements during the Phanerozoic. But why they should exhibit this pattern in the first place is far from obvious. The evolutionary faunas are certainly not phylogenetically coherent, nor are there clear ecological niches or strategies that unify them. Even in Sepkoski’s original factor analysis, the evolutionary faunas are not fully distinct but exhibit fuzzy boundaries that are often overlooked. For example, the Gastropoda score rather well on two factors, the ‘‘Modern’’ and the ‘‘Paleozoic’’ (Sepkoski 1981), making them, statistically speaking, part of both evolutionary faunas. Thus, it is interesting but perhaps not surprising that within an evolutionary fauna, some clades might behave in a way that is distinctly at odds with the expected pattern (Adrain et al. 1998). Nonetheless, the results of this study seem to strengthen the ecological significance of the evolutionary faunas, as defined on the basis of global first and last occurrence data. This is especially true because the taxonomic richness of a group does not necessarily translate into local ecological dominance (e.g., Wing et al. 1993). Sepkoski’s (1979, 1981) specific threephased kinetic model that sought to explain the macroevolutionary history of the evolutionary faunas has recently been questioned by Alroy (2004), and it seems clear that at the genus level, turnover rates do not neatly divide marine animal classes into three distinct groups. Nevertheless, the results of this study indicate that the relative genus richness of Sepkoski’s evolutionary faunas is not just a global phenomenon but is locally manifested in at least some marine assemblages in very ecologically relevant ways. These results do not support Sepkoski’s specific macroevolutionary model for the three faunas, but they do suggest the possibility that some of the ecological interactions that might be inferred on the basis of global relative genus richness could be relevant in local communities. Although this study in no way answers the question of how the evolutionary faunas maintain coherence at such different scales of analysis (measured here in terms of the agreement between global relative richness and local relative richness and abundance), these data do suggest some possibilities that can be tested further, many of which have already been suggested in some form or another by Bambach (1993) and others. For example, the relationship between local relative abundance and local relative genus richness (Fig. 6) as well as the onshore-offshore patterns of faunal dominance, may reflect fundamental differences between the faunas in such parameters as biomass and metabolic requirements. It is not a stretch to imagine phylogenetically diverse and ecologically disparate taxa having similar characteristics in this regard. This was a major theme of Bambach’s (1993) ‘‘seafood through time’’ paper, and it would be interesting if such low-level differences among unrelated taxa could have such far reaching evolutionary consequences. Of course this does not explain mechanistically how these specific characteristics could result in differential global rates of evolution and local relative richness, but other life-history traits have been linked to rates of evolution and large-scale patterns of richness (e.g., Jablonski 1986) so it is not inconceivable that this may be the case. Whatever the ultimate explanation for the coherence of the evolutionary faunas, playing off local and global patterns of relative abundance and richness in conjunction with a comparative ecological approach will undoubtedly help to better understand Sepkoski’s three great evolutionary faunas. Acknowledgments Fieldwork was supported by grants from the Paleontological Society, the Environmental SEPKOSKI’S EVOLUTIONARY FAUNAS Protection Agency STAR Fellowship program, the Society of Sigma Xi, and the University of Chicago Hinds Fund. B. Gaines, S. Finnegan, M. Foote, K. Karns, S. Kidwell, A. McGowan, and A. Ziegler provided helpful discussion. P. Wagner provided a picture of his favorite poorly preserved Ordovician snail. S. Finnegan, M. Foote, H. De Simone, and T. Rothfus provided invaluable field assistance at various stages of this work. K. Karns and M. Behrendt offered excellent specimen preparation skills. M. Foote, D. Jablonski, S. Kidwell, and A. I. Miller helped to improve early drafts of this manuscript. I also thank J. Adrain, P. Sheehan, and M. Patzkowsky for very helpful reviews. Literature Cited Adrain, J. M., and S. R. Westrop. 2000. An empirical assessment of taxic paleobiology. Science 289:110–112. Adrain, J. M., R. A. Fortey, and S. R. Westrop. 1998. Post-Cambrian trilobite diversity and evolutionary faunas. Science 280: 1922–1925. Adrain, J. M., S. R. Westrop, B. D. E. Chatterton, and L. Ramsköld. 2000. Silurian trilobite alpha diversity and the end-Ordovician mass extinction. Paleobiology 26:625–646. Allison, P. A., and D. E. G. Briggs. 1993. Paleolatitudinal sampling bias, Phanerozoic species-diversity, and the End-Permian extinction. Geology 21:65–68. Alroy, J. 2004. Are Sepkoski’s evolutionary faunas dynamically coherent? Evolutionary Ecology Research 6:1–32. Ausich, W. I., and D. J. Bottjer. 1982. Tiering in suspension-feeding communities on soft substrata during the Phanerozoic. Science 216:173–174. Babin, C. 2000. Ordovician to Devonian diversification of the Bivalvia. American Malacological Bulletin 15:167–178. Bambach, R. K. 1985. Classes and adaptive variety: the ecology of diversification in marine faunas through the Phanerozoic. Pp. 191–253 in J. W. Valentine, ed. Phanerozoic diversity patterns: profiles in macroevolution. Princeton University Press, Princeton, NJ. ———. 1993. Seafood through time: changes in biomass, energetics, and productivity in the marine ecosystem. Paleobiology 19:372–397. Best, M. M. R., and S. M. Kidwell. 2000. Bivalve taphonomy in tropical mixed siliciclastic-carbonate settings. I. Environmental variation in shell condition. Paleobiology 26:80–102. Brett, C. E., and G. C. Baird. 1986. Comparative taphonomy: a key to paleoenvironmental interpretation based on fossil preservation. Palaios 1:207–227. Brett, C. E., S. E. Speyer, and G. C. Baird. 1986. Storm-generated sedimentary units: tempestite proximality and event stratification in the Middle Devonian Hamilton Group of New York. New York State Museum Bulletin 457:129–156. Cherns, L., and V. P. Wright. 2000. Missing mollusks as evidence of large-scale, early skeletal aragonite dissolution in a Silurian sea. Geology 28:791–794. Droser, M. L., and S. Finnegan. 2003. The Ordovician radiation: follow-up to the Cambrian explosion. Integrative and Comparative Biology 43:178–184. Droser, M. L., R. A. Fortey, and X. Li. 1996. The Ordovician radiation. American Scientist 84:122–131. Ellingsen, K. E. 2002. Soft-sediment benthic biodiversity on the 559 continental shelf in relation to environmental variability. Marine Ecology Progress Series 232:15–27. Erwin, D. H., and S. L. Wing, eds. 2000. Deep time: Paleobiology’s perspective. Paleobiology 26 (Suppl. To No. 4). Foote, M. 2000. Origination and extinction components of taxonomic diversity: general problems. Pp. 74–102 in Erwin and Wing 2000. Gould, S. J. 2000. Beyond competition. Paleobiology 26:1–6. Gould, S. J., and C. B. Calloway. 1980. Clams and brachiopods —ships that pass in the night. Paleobiology 6:383–396. Hayek, L., and M. A. Buzas. 1997. Surveying natural populations. Columbia University Press, New York. Hill, G. W., K. A. Roberts, J. L. Kindinger, and G. D. Wiley. 1982. Geobiologic study of the south Texas outer continental shelf. U.S. Geological Survey Professional Paper P1238. Jablonski, D. 1986. Larval ecology and macroevolution in marine-invertebrates. Bulletin of Marine Science 39:565–587. Jeffery, C. H. 2001. Heart urchins at the Cretaceous/Tertiary boundary: a tale of two clades. Paleobiology 27:140–158. Kidwell, S. M. 1986. Models for fossil concentrations: paleobiologic implications. Paleobiology 12:6–24. Li, X., and M. L. Droser. 1997. Nature and distribution of Cambrian shell concentrations: evidence from the Basin and Range Province of the western United States (California, Nevada, and Utah). Palaios 12:111–126. ———. 1999. Lower and Middle Ordovician shell beds from the Basin and Range Province of the western United States (California, Nevada, and Utah). Palaios 14:215–233. Lidgard, S., F. K. McKinney, and P. D. Taylor. 1993. Competition, clade replacement, and a history of cyclostome and cheilostome bryozoan diversity. Paleobiology 19:352–371. Lockley, M. G. 1983. Brachiopod dominated palaeocommunities from the type Ordovician. Palaeontology 26:111–145. Miller, A. I. 1989. Spatio-temporal transitions in Paleozoic Bivalvia: a field comparison of Late Ordovician and upper Paleozoic bivalve-dominated fossil assemblages. Historical Biology 2:227–260. ———. 1997. Dissecting global diversity patterns: examples from the Ordovician radiation. Annual Review of Ecology and Systematics 28:85–104. Miller, A. I., and S. Connolly. 2001. Substrate affinities of higher taxa and the Ordovician Radiation. Paleobiology 27:768–778. Novack-Gottshall, P. M., and A. I. Miller. 2003. Comparative geographic and environmental diversity dynamics of gastropods and bivalves during the Ordovician Radiation. Paleobiology 29:576–604. Palmer, A. R. 1998. A proposed nomenclature for stages and series for the Cambrian of Laurentia. Canadian Journal of Earth Sciences 35:323–328. Patzkowsky, M. E. 1995. Gradient analysis of Middle Ordovician brachiopod biofacies: biostratigraphic, biogeographic, and macroevolutionary implications. Palaios 10:154–179. Patzkowsky, M. E., and S. M. Holland. 1993. Biotic response to a Middle Ordovician paleoceanographic event in eastern North America. Geology 21:619–622. Peters, S. E. 2004. Evenness in Cambrian–Ordovician benthic marine communities in North America. Paleobiology 30:325– 346. Peters, S. E., and M. Foote. 2001. Biodiversity in the Phanerozoic: a reinterpretation. Paleobiology 27:583–601. ———. 2002. Determinants of extinction in the fossil record. Nature 416:420–424. Powell, E. N., H. K. M. Parsons, C. W. Russell, G. M. Staff, G. T. Gilbert, C. E. Brett, S. E. Walker, A. Raymond, D. D. Carlson, S. White, and E. A. Heise. 2002. Taphonomy on the continental shelf and slope: two-year trends—Gulf of Mexico and Bahamas. Palaeogeography, Palaeoclimatology, Palaeoecology 184:1–35. 560 SHANAN E. PETERS Raup, D. M. 1976. Species diversity in the Phanerozoic: an interpretation. Paleobiology 2:289–297. Sánchez, T. M., and C. Babin. 2003. Distribution paléogéographique des mollusques bivalves durant l’Ordovicien. Geodiversitas 25:243–259. Sepkoski, J. J., Jr. 1979. A kinetic model of Phanerozoic taxonomic diversity. II. Early Phanerozoic families and multiple equilibria. Paleobiology 5:222–251. ———. 1981. A factor analytic description of the Phanerozoic marine fossil record. Paleobiology 7:36–53. ———. 1984. A kinetic model of Phanerozoic taxonomic diversity. III. Post-Paleozoic families and mass extinctions. Paleobiology 10:246–267. ———. 1988. Alpha, beta, gamma: where does all the diversity go? Paleobiology 14:221–234. ———. 1991. A model of onshore-offshore change in faunal diversity. Paleobiology 17:68–77. ———. 1996. Competition in macroevolution: the double wedge revisited. Pp. 211–255 in D. Jablonski, D. H. Erwin, and J. H. Lipps, eds. Evolutionary paleobiology. University of Chicago Press, Chicago. ———. 2002. A compendium of fossil marine animal genera. Bulletins of American Paleontology 363, 560 p. Sepkoski, J. J., Jr., and A. I. Miller. 1985. Evolutionary faunas and the distribution of Paleozoic benthic communities in space and time. Pp. 393–396 in J. W. Valentine, ed. Phanerozoic diversity patterns. Princeton University Press, Princeton, NJ. Sepkoski, J. J., Jr., and P. M. Sheehan. 1983. Diversification, faunal change, and community replacement during the Ordovician radiations. Pp. 673–718 in M. J. S. Tevesz and P. L. McCall, eds. Biotic interactions in Recent and fossil benthic communities. Plenum, New York. Sepkoski, J. J., Jr., F. K. McKinney, and S. Lidgard. 2000. Competitive displacement among post-Paleozoic cyclostome and cheilostome bryozoans. Paleobiology 26:7–18. Smith, A. B. 2001. Large-scale heterogeneity of the fossil record: implications for Phanerozoic biodiversity studies. Philosophical Transactions of the Royal Society of London B 356:351– 367. Vermeij, G. 1977. The Mesozoic marine revolution: the evidence from snails, predators, and grazers. Paleobiology 3:245–258. Webby, B. D. 1998. Steps towards a global standard for Ordovician stratigraphy. Newsletters on Stratigraphy 36:1–33. Westrop, S. R., and J. M. Adrain. 1998. Trilobite alpha diversity and the reorganization of Ordovician benthic marine communities. Paleobiology 24:1–16. ———. 2001. Sampling at the species level: impacts of spatial biases on diversity gradients. Geology 29:903–906. Westrop, S. R., J. V. Tremblay, and E. Landing. 1995. Declining importance of trilobites in Ordovician nearshore paleocommunities: dilution or displacement? Palaios 10:75–79. Wing, S. L., L. J. Hickey, and C. C. Swisher. 1993. Implications of an exceptional fossil flora for Late Cretaceous vegetation. Nature 363:342–344. Wright, P., L. Cherns, and P. Hodges. 2003. Missing mollusks: field testing taphonomic loss in the Mesozoic through early large-scale aragonite dissolution. Geology 31:211–214. Appendix Literature Sources Literature-derived data used in this paper are referenced below. A list of formations used in the study follows each citation. If the formations cited preserve shallow-nearshore and intertid- al environments, then the formations are followed by a parenthetical statement indicating that these data represent shallow habitats that were used to generate Figure 9. If a citation does not indicate a shallow-water environment, then the data represent the deeper-water settings described in ‘‘Data and Methods.’’ See http://dx.doi.org/10.1666/02055.s1 for deep subtidal data from both field and literature sources. Steptoean: Shaw, A. B. 1956. A Cambrian Aphelaspis fauna from Steele Butte, near Boulder, Wyoming. Journal of Paleontology 30:48– 52. Dry Creek Shale. Early Ordovician (taxonomic lists only): Cloud, P. E., and V. E. Barnes. 1957. Early Ordovician sea in central Texas. In H. S. Ladd, ed. Treatise on marine ecology and paleoecology. Geological Society of America Memoir 67:163– 214. Tanyard Formation, Honeycut Formation, Gorman Formation (shallow). Hintze, L. F., L. F. Braithwaite, D. L. Clark, R. L. Ethington, and R. F. Flower. 1972. A fossiliferous Early Ordovician reference section from the western United States. Proceedings of the International Paleontological Union 5:385–399. House Limestone. Lochman, C. B., and J. L. Wilson. Stratigraphy of Upper Cambrian-Early Ordovician subsurface sequence in Williston Basin. American Association of Petroleum Geologists Bulletin 51:883–917. Deadwood Formation (shallow). Mazzullo, S. J., and G. M. Friedman. 1977. Competitive algal colonization of peritidal flats in a schizohaline environment: the Early Ordovician of New York. Journal of Sedimentary Petrology 47:398–410. Middle member, Great Meadows Formation, Fort Ann Formation (shallow). Sando, W. J. 1957. Beekmantown Group (Early Ordovician) of Maryland. Geological Society of America Memoir 68. Stonehedge Limestone, Rockdale Run Formation (shallow). Shaw, A. B. 1958. Stratigraphy and structure of the St. Albans Area, northwestern Vermont. Geological Society of America Bulletin 69:519–568. Lower Highgate Formation. Wilson, J. L. 1954. Late Cambrian and early Ordovician trilobites from the marathon Uplift, Texas. Journal of Paleontology 28:249–285. Lower Marathon Formation and Woods Hollow Formation. Late Ordovician: Bayer, T. N. 1967. Repetitive benthonic community in the Maquoketa Formation (Ordovician) of Minnesota. Journal of Paleontology 41:417–422. Maquoketa Formation. Bretsky, P., and J. J. Bermingham. 1970. Ecology of the Paleozoic scaphopod genus Plagioglypta with special reference to the Ordovician of eastern Iowa. Journal of Paleontology 44:908– 924. Maquoketa Formation (shallow). Frey, R. C. 1987. The occurrence of pelecypods in early Paleozoic epeiric-sea environments, Late Ordovician of the Cincinnati, Ohio area. Palaios 2:3–23. Sample A; Kope Formation. Patzkowsky, M., and S. M. Holland. 1999. Biofacies replacement in a sequence stratigraphic framework: Middle and Late Ordovician of the Nashville Dome, Tennessee, USA. PALAIOS 14:301–323. Lebanon Formation, Hermitage Formation, Arnheim Formation, Sequatchie Formation. Springer, D. A. 1982. Community gradients in the Martinsburg Formation (Ordovician), southwestern Virginia. Ph.D. dissertation, Virginia Polytechnic Institute, Blacksburg, Va. Martinsburg Formation (shallow). Titus, R., and B. Cameron. 1976. Fossil communities of the lower Trenton Group (Middle Ordovician) of central and northwestern New York State. Journal of Paleontology 50:1209– 1225. Lower Trenton Group (shallow).