Quantitative paleoecology of marine faunas in the lower Hamilton Group (Middle Devonian, central New York): Significance for probing models of long-term community stability

Beautifully fossiliferous strata in the Hamilton Group (Middle Devonian, central New York) constitute a rich “ecological archive” sufficient to probe and test founda-tional concepts in paleontology. The evident community stability of Hamilton faunas over 4–6 m.y.—including two proposed mechanisms for coordinated stasis—has ignited controversy. Resolving community structure and both taxonomic and ecological temporal persistence within the Hamilton Group thus becomes critical to testing whether these Hamilton communities are stable and whether they are ecologically “locked.” Toward this end, we conducted multivariate analyses (cluster and correspondence analysis) of marine faunas in 81 large samples (~300 specimens each) in shallowing-upward sequences of the Cardiff and Pecksport Members (Marcellus Subgroup, Oatka Creek Formation) of the Hamilton Group. Eight statistically and ecologically distinctive benthic communities characterize the vertical gradient, from depauperate, deeper-water dark shales below to species-rich shelf siltstones above. These communities correlate strongly with grain size, bioturbation intensity, bedding thickness, density of fossils, and faunal and ecological diversity. Species richness varies inversely with weight percent organic matter. We characterized taxonomic distributions using multivariate statistics; these statistical analyses were based on percentages of 50 taxa. In order of decreasing depth, the communities are: Cephalopod- Pterochaenia , Pterochaenia-Eumetabolotoechia , Eumetabolotoechia , Emanuella , Eumetabolotoechia-Ambocoelia , Arcuaminetes- Eumetabolotoechia , Arcuaminetes-Ambocoelia , and Mucrospirifer-Ambocoelia . The Cephalopod- Pterochaenia community represents a mixed benthic-pelagic


INTRODUCTION
Abundant and often exquisitely preserved marine faunas in the well-exposed Middle Devonian Hamilton Group of New York have long attracted international attention as a valuable "ecological archive" for testing new hypotheses in paleontology. Studied intensively by paleontologists and stratigraphers for more than 150 yr, these rich and temporally persistent Hamilton faunas have figured prominently in some of the largest evolutionary and paleoenvironmental debates, both past and present. The hypothesis of punctuated equilibria, the oft-debated speciation model of Eldredge and Gould (1972), derived in part from Niles Eldredge's interpretations of morphological stasis in the Hamilton trilobite Phacops rana 1 and related species in his Columbia University doctoral dissertation. Superbly preserved Hamilton Group fossils also afforded resolution of mid-Paleozoic skeletal microstructures, which in turn allowed inquiry into bold questions regarding shell structure and evolution, especially of bivalves (e.g., Carter and Tevesz, 1978). In addition, outstanding Hamilton fossil preservation afforded early analyses of the isotopic signatures of Paleozoic seas (e.g., Popp et al., 1986aPopp et al., , 1986bVeizer et al., 1999, and references therein;Selleck and Koff, 2008).
In more recent years, Hamilton Group faunas have yet again assumed centrality in the lively and prolonged debate over "coordinated stasis," defined as simultaneous taxonomic and, some would claim, ecologic stasis within independent clades that remain apparently stable for millions of years (Brett and Baird, moderate species richness and are dominated by epifaunal, active suspension feeders, especially the small epibyssate bivalve Pterochaenia fragilis, and the pedunculate brachiopods Eumetabolotoechia multicostata and Emanuella subumbona. The Pterochaenia-Eumetabolotoechia community is an opportunistic fauna that developed when the substrate first became favorable for colonization by benthic organisms. To a lesser extent, this probably also holds true for the Eumetabolotoechia assemblage. Communities near the top of the shallowing-upward cycle-Eumetabolotoechia-Ambocoelia, Arcuaminetes-Eumetabolotoechia, Arcuaminetes-Ambocoelia, and Mucrospirifer-Ambocoelia-have higher taxonomic and ecological heterogeneity, with a more diverse array of trophic and locomotory groups than their counterparts in the finer-grained, and inferred deeper, facies.
Cluster significance tests applied to all pairs of communities known from adequate numbers of samples demonstrated that the communities are statistically valid and distinctive. Multivariate means of all communities were significantly different; furthermore, most pairs of communities were drawn from populations that showed no overlap in terms of rectangular distributions. The community sequence and an ordination derived from the first two axes of the correspondence analysis provided relative depth curves. Our communities, with two exceptions, do not have clear counterparts among upper Hamilton Group faunas. The ecological locking model proposed to explain the stability of Hamilton faunas is not supported by our quantitative tests to date.
1985; Brett et al., 1996;Schopf, 1996;Ivany et al., 2009;Brett, 2012). The marine communities in the Hamilton Group have been claimed to be one of the fossil record's best-documented and most temporally persistent examples of coordinated stasis (Brett and Baird, 1995;Ivany et al., 2009;Nagel-Myers et al., 2009). Various scholars have analyzed taxonomic stasis during this 4-6 m.y. interval and explored the possibilities of ecological stasis, which include both stable community structure and even, in the view of some authors, stable rates of predation.
Explanatory proposals for coordinated ecological stasis in the Hamilton Group interval have engendered controversy. Scholarly discussion, intense and sometimes heated, has centered on possible mechanisms, principally "ecological locking" or "environmental tracking" (Morris et al., 1995;Ivany, 1996). The ecological locking (Morris et al., 1995) model views these marine communities as tightly structured and therefore, once established, resistant to large-scale taxonomic change. The environmental tracking model, in contrast, interprets stability as arising when faunas follow environments through time; during an interval of minimal environmental change, faunas will vary only minimally (Ivany, 1996; see also comments by Miller, 1997). Ivany (1996) and several others observed that brief intervals of accelerated environmental change serve to magnify this coordinated pattern. Synchronous regional environmental changes, expressed through sedimentology and geochemistry as well as paleontology, can produce concurrent faunal eliminations that include both regional disappearances and true extinction (lineage termination). In sum, the ecological locking hypothesis proposes that the observed patterns originate in ecological homeostatic mechanisms, whereas the environmental tracking model hypothesizes fidelity of organisms to their environments and posits coordinated extinctions (or regional disappearances) to create the boundaries of intervals of relative stasis.
Coordinated stasis is inherently a quantitative claim. Only through rigorous quantitative studies that sample rare taxa (not just abundant to common taxa) and that analyze the taxonomic and ecological structure of communities over time can coordinated stasis be properly tested (Bennington and Bambach, 1996;McKinney et al., 1996;Ivany et al., 2009). Moreover, demonstrating ecological stasis throughout the Hamilton interval requires such quantitative documentation from both lower and upper intervals. We present here the results of one such quantitative study of lower Hamilton communities, as one of several necessary steps in evaluating coordinated stasis models as they apply to this Middle Devonian interval.
The stratigraphic framework for the Hamilton Group was established by Carlton Brett and Gordon Baird, building in turn upon a series of earlier fundamental works by others; see discussions of this in the comprehensive stratigraphic volumes edited by Brett (1986) and Landing and Brett (1991). This classic Hamilton Group lithostratigraphy must be correlated with coarser sediments occurring eastward, closer to the deltaic source; these eastern strata were studied by Chuck Ver Straeten (Ver Straeten et al., 2011). These marine sediments of deltaic and other marine environments exhibit marked transgressive-regressive cycles (e.g., Brett and Baird, 1985;Brett et al., 1986). These cycles thicken eastward toward the delta and correspondingly thin westward. In western New York, these thinner cycles consist of shale, calcareous shale, and limestone, and they are approximately symmetrical (Miller, 1986;Savarese et al., 1986). Further east, in central and eastern New York, as the source area in the Acadian delta is approached, cycles increase in thickness, grain size, and clastic input (e.g., Baird et al., 2000;Ver Straeten et al., 2011); in addition, they become dominated by the regressive phase, with a corresponding reduction of the transgressive phase to a few thin beds that are commonly associated with a diastem and lag concentrate of phosphatic and other residual materials.

SCOPE AND OBJECTIVES
We present here the findings of our quantitative marine paleoecological research on faunas of portions of the lower Hamilton Group-specifically, the Cardiff and Pecksport Members of the Middle Devonian Marcellus Subgroup in central New York (Figs. 1 and 2). The lithologies of the Cardiff and Pecksport Members are well summarized and their faunas illustrated in Linsley (1994). These two members occur in the Oatka Creek Formation, which is best expressed in the Finger Lakes region further west; as this formation was established, it became necessary to examine the detailed correlations of these classic Hamilton Group localities in central New York with the more recently established lithostratigraphic units in western New York and in eastern New York, closer to the clastic source. How the memberlevel lithographic of the two new formations-one in the west, another in the east-correlates through central New York's classic Hamilton Group remains to be resolved by the Brett and Ver Straeten lithostratigraphic teams. We are in a critical transition zone. Following Selleck (2010), we emphasize the memberlevel stratigraphy and its scientific replicability within the central region, with an understanding that the correlations to the east and west will continue to be studied by these other groups.
The stratigraphic interval we studied (upper Cardiff and Pecksport Members of the Oatka Creek Formation, Marcellus Subgroup) records the transition between basinal black shales and limestones with pelagic and mixed pelagic-benthic assemblages into highly oxygenated benthic communities of siltstones that formed in shallower areas of the depositional basin. Brower  and Nye (1991), one of our Syracuse group's earlier contributions, also discussed the pelagic and pelagic-benthic communities from the lower Hamilton Group.
Medium gray, non-calcareous silty shale. Poorly bedded with beds 0.5 to 2 cm thick. Fossils common.
Medium gray, non-calcareous chippy shale. Several horizons of 3 -10 cm diameter concretions, with sparse small concretions on 5 to 12 cm intervals throughout section. Fossils common.   Tables 1 and 2 for community codes. Brower) and our students obtained and analyzed large and closely spaced samples from the shallowing-upward sequences in central New York. Syracuse University faculty actively involved in this long-term program of Devonian paleontological research, and in working on this with graduate students, include James C. Brower, Cathryn R. Newton, Linda C. Ivany, Osborne B. Nye Jr., and now Christopher Junium. In conducting our quantitative paleoecological analyses over this span, we and our students have generally collected large samples, even from sparsely fossiliferous horizons such as those of the dark shales. In addition, several of our students in this larger program have conducted morphological and paleo-autecological studies of individual Hamilton Group taxa during these years. Some Syracuse graduate alumni supported in this group continue to publish on Hamilton Group faunas (e.g., Christopher McRoberts on trilobites and bivalves: McRoberts et al., 2013;Nagel-Myers et al., 2018). This multidecadal research program at Syracuse University has consistently probed Devonian marine community structure and composition using systematically collected samples and analyzed with multivariate quantitative methods. During the 35 yr in which Brower andNewton worked together (1983-2018), we had the further goal of using these large data sets to test some fundamental principles of paleoecology, such as exploring the extent to which these paleocommunities were tightly structured. Our current study fits with both of these long-term aims and forms a key addition to the Hamilton Group record in allowing for tests of both community structure and its temporal persistence.
The long lineage of Syracuse University studies on the paleoecology of Devonian siltstones and shales in upstate New York includes Brower andChute (1964a, 1964b), Nye et al. (1975), Brower et al. (1978), Brower (1987), Thompson and Newton (1987), , Thompson and Newton (1988), , Thompson (1989), Brower and Nye (1991), Bonuso (2001), Bonuso et al. (2002aBonuso et al. ( , 2002b, Brett et al. (2009), Ver Straeten et al. (2019, and diverse other studies cited therein. Nicole Bonuso and Joel Thompson were graduate students who worked with Newton. Another graduate student was Willis Newman, a fine field geologist who collected data for this paper, but who did not complete the doctorate and who died far too young. This study critically extends, and in some ways completes, our earlier work by further analyzing community structures in the Marcellus Subgroup-that basal part of the Hamilton Group now famed for its hydrocarbon potential and production (Nyahay et al., 2007;Kargbo et al., 2010;Soeder, 2010). Our focus is not on the lowermost Marcellus Subgroup, which is the site of most hydrocarbon exploration, but on the intriguing interplay of communities higher, in the Cardiff and Pecksport Memberscommunities that find some, but not all, ecological counterparts further upward in the Hamilton Group. This strengthening of lower Hamilton Group quantitative paleoecology facilitates greater rigor in testing inferences about possible ecological stability and, in particular, ecological locking of marine communities over 4-6 m.y. of Hamilton Group deposition.

QUANTITATIVE PALEOECOLOGICAL METHODS
Biological and sedimentological data were analyzed with multivariate statistics. Lucid ecological references for the methods applied herein include works by Clifford and Stephenson (1975), Orloci (1978), Whittaker (1978aWhittaker ( , 1978b, Gauch (1982), Greig-Smith (1983), Legendre (1983, 2012), Pielou (1984), Birks and Gordon (1985), Ludwig and Reynolds (1988), and Dray et al. (2012). Here, we deployed a variety of methods comparable to those used both in community ecology and in our earlier works on Hamilton Group paleoecology; this parallel approach best facilitates comparisons. The basic analytical strategy follows that of Brower (1987). Initially, clustering and ordination are used to identify and characterize the communities and gradients present in the data. These are tested subsequently to ascertain their statistical significance. In the final step, plotting the communities and gradients against geographic and stratigraphic position leads to the construction of a paleoecological scenario.
We define communities herein as consistent, statistically recurrent groups of fossils. Both ordination and clustering methods are necessary for the recognition of faunal communities and gradients, because these methods are subject to different types of distortion. Ordination methods preserve the main patterns in the data, but the relationships between similar species and samples are often shown incorrectly. Clustering retains the distances between similar items, but at the cost of large-scale deformation. In both approaches, we used percentages of species in samples. Computational details of the exact algorithms are available in the textbooks listed above and Sneath and Sokal (1973). Clusters were calculated by the unweighted-pair-group method (UPGM) on a matrix of correlation coefficients. The ordination algorithm consists of correspondence analysis axis ordination or reciprocal averaging. The correspondence analysis was not detrended because its axes retain the dual space properties of the variable-and specimen-space eigenvectors specified by the Eckart-Young theorem (see any textbook on multivariate analysis). As outlined later, the first two axes of the correspondence analysis are highly arched. Consequently, these were reduced to a single axis by calculating orthogonal projections of the data points onto a polynomial as suggested by Ludwig and Reynolds (1988) and Phillips (1978). This is termed the "correspondence analysis axis ordination" in later parts of the paper. Clustering and ordination lead to similar and complementary conclusions.
The groups of samples recognized in the clusters and ordinations should be tested for statistical significance. We elected to employ significance tests for clusters rather than discriminant analysis because they are largely independent of the number of variables (taxa in this case), and they are less sensitive to statistical assumptions.
Inasmuch as cluster significance tests are probably not familiar, the statistical methods will be outlined briefly (for details, see Sneath, 1977Sneath, , 1979. Visualize an axis connecting the centroids of two clusters of samples in the space defined by the species. The samples are projected onto this axis to form q (or Q) scores or distributions that illustrate the differences between the two clusters. These q scores are analogous to the scores of discriminant functions, except that they are more conservative, because the differences between the two clusters may not be maximized. Various tests can be applied to the q distributions. The conventional hypothesis for univariate or multivariate means ascertains whether or not the differences between them are statistically significant subject to a predetermined risk level. This is obviously trivial for groups that have been previously identified by cluster analysis; one almost invariably rejects the null hypotheses, even if the two groups overlap greatly. It is much more informative to query whether the samples were drawn from two populations that overlap more or less than some specified amount. The degree of overlap of the q distributions is measured by the W statistic of disjunction.
We regard communities as consistent faunal associations that may intergrade. The communities studied here changed in response to a paleoecological shallowing gradient caused by a marine regression. Consequently, a rectangular distribution of the q scores provides a reasonable null hypothesis (Sneath, 1977). In a rectangular distribution, the observed number of samples along equal intervals of the q scores follows a Poisson distribution, so the q distributions lack extended tails and behave like censored data. The null hypothesis specifies that the q scores of the two clusters are drawn from overlapping rectangular distributions. The alternative hypothesis is that the q distributions of the two clusters are disjunct. The critical test values were obtained from Sneath (1979, fig. 2).

FACIES ANALYSIS
In these Hamilton Group regressive intervals in central New York, lithofacies vary consistently in ways that are traceable laterally but exhibit a narrow range of lithologic variation. The cycles at our two study sites are typical: They begin in basal strata of dark-gray, noncalcareous shales (Figs. 3 and 4). These are overlain by progressively siltier, concretion-rich shales, and each regressive cycle culminates in noncalcareous silty shales or siltstones. Dark-gray shales at the base of these cycles preserve bioturbated fabrics at one site ( Fig. 3) and are only sparsely bioturbated at the other (Fig. 4).
In addition to faunal data, we analyzed grain size (scaled meristically from 1.0 for shale to 5.0 for siltstone) and weight percent of total organics (measured by loss on ignition). These lithologic variables were then analyzed relative to the percentages of common species in the assemblages. Statistics considered included correlation coefficients, multiple correlation coefficients, and regression weights for polynomials, where the correspondence analysis axis ordination of species composition was written as a function of the lithologic code and/or organic content. Inasmuch as the organics were measured continuously, whereas the grain size was coded meristically, it is difficult to compare regression coefficients based on both parameters. Visual inspection and significance tests suggested that third-order polynomials provide reasonable fits to the data. These polynomials were calculated for the lithology, organic matter, and lithology and organic matter considered together. The conclusions were largely based on comparing the squared multiple correlation coefficients. Interactions between lithology and organic matter were not treated, because significance tests indicated that the interactions did not explain significant components of the variance for the dependent variable.
The squared multiple correlations for the correspondence axis ordination versus the lithology and organics were 0.758 and 0.802, respectively, whereas the same value for both lithology and organics was 0.855. (Detailed results of the correspondence axis ordination, including relation of lithologic data to communities, appear in the "Sequence of Communities" section.) Thus, the correspondence axis ordination seems to represent an approximately equal compound of organic matter and grain size if all eight communities are considered. This is evidenced by the fact that the variances associated with the two polynomial regressions for the organics and lithology do not differ significantly. The regression for both lithology and organic matter explains a significantly larger amount of the variance in the ordination than those of lithology or organic matter alone. The F ratio for the added component of variance is 6.91 with 3 and 65 degrees of freedom, which is significant at the 0.05 and 0.01 probability levels. The seven exclusively benthic communities yielded similar results, except that grain size was slightly more important than organic matter. Figure 5 illustrates the dendrogram for 81 samples based on 50 species and the unweighted-pair-group method clustering of a matrix of correlation coefficients. Eight main clusters or communities can be recognized at correlations of 0.7 and higher. The dendrogram is tightly structured, and virtually all of the links within the clusters take place at levels of 0.85 and higher. Communities will be discussed in order of increasing grain size and decreasing organic content. As seen in Figures 3 and 4, the lithologies vary, including relatively dark and fine-grained shales and siltstones with 3% total organic matter; gray silty shales with lower organic content; and siltstones in with approximately 2 wt% of total organic matter.

CLUSTER ANALYSIS
Samples 1-9 on the dendrogram constitute the Cephalopod-Pterochaenia community (Fig. 5). Diversity is low, with only three to eight species per sample. Cephalopods dominate the community; these include both large orthocones and Agoniatites vanuxemi, which represent from 43% to 81% of the individuals, respectively (Table 2). Pterochaenia fragilis, also common (11%-30% of the community), is enigmatic in terms of living habits; it is clearly a byssate suspension feeder (e.g., Brower and Nye, 1991; see illustrations in Clarke, 1904), but the site of byssal attachment (i.e., benthic vs. epipelagic) is uncertain. We interpret it here as a nonendemic benthic opportunist, but we recognize that others might also reasonably infer it was an epiplanktonically attached taxon; this can likely be resolved with further studies. Eumetabolotoechia multicostata also occurs in many samples. This community characterizes the finest-grained rocks observed in this study-namely, thinly bedded dark-gray shales with small burrows and relatively high organic contents (averaging 3.0 wt%).
The Eumetabolotoechia community contains samples 11-24 on the cluster diagram (Fig. 5). The pedunculate rhynchonellid Eumetabolotoechia multicostata invariably dominates; in one sample, it represents 81% of the individuals present. Other taxa encountered in all or most samples are (in decreasing order of abundance): Emanuella subumbona, Pterochaenia fragilis, Nuculoidea sp., and Arcuaminetes scitulus (Table 2). Samples 11-45 join the cluster at high correlation coefficients of 0.955-0.997. However, samples 31 and 24 group at somewhat lower values of 0.903 and 0.885; these two samples also contrast in having a slightly higher species richness and a somewhat lower proportion of Eumetabolotoechia.
Fourteen samples, listed from 35 to 55 on the dendrogram, represent the Eumetabolotoechia-Ambocoelia community. Eumetabolotoechia multicostata and, to a lesser extent, the pedicleattached brachiopod Ambocoelia umbonata and the recliner Arcuaminetes scitulus form the core of this cluster. Eumetabolotoechia and Ambocoelia individuals occur in all samples, whereas Arcuaminetes is present in only 13 samples. Although generally present, Nuculoidea sp., Nuculites oblongatus, and large orthocones are less abundant; they typically constitute 1% to 5% of any one sample. Accessory species that occur in approximately half of the samples are Mucrospirifer mucronatus, Pterochaenia fragilis, Lingula sp., and Palaeozygopleura hamiltoniae (Table 2).
Samples 25 and 26 differ greatly from all others; more than 50% of the fauna is Emanuella subumbona, a small, pedicle-attached spiriferid brachiopod. We thus designate this the Emanuella community. Eumetabolotoechia multicostata and Arcuaminetes scitulus each typically account for 10%-15% of the fauna. A few shells of Mucrospirifer mucronatus and Nuculoidea sp. are associated (Table 2). Although the geographic distribution of this assemblage is uncertain, it seems locally to be restricted to a small stratigraphic interval within the Cardiff Member along Gulf Road.
The Arcuaminetes-Eumetabolotoechia community, represented by samples 27-44 on the clusters in Figure 5, is dominated by Arcuaminetes scitulus and Eumetabolotoechia multicostata, which together generally comprise slightly over 50% of each sample. In most samples, these are followed by Ambocoelia umbonata, Nuculoidea sp., Mucrospirifer mucronatus, Lingula sp., and Nuculites oblongatus (in decreasing order of abundance). A few specimens of large orthocones and Bembexia sulcomarginata occur in most samples ( Table 2).
The Arcuaminetes-Ambocoelia community characterizes the sequence of samples 28-73 on the dendrogram. Approximately half of the individuals consist of Arcuaminetes scitulus, in conjunction with highly variable numbers of Ambocoelia umbonata. Other common constituents in most samples include Mucrospirifer mucronatus, Nuculoidea sp., and Emanuella subumbona. Eumetabolotoechia multicostata is conspicuously absent from most samples (Table 2).
The Mucrospirifer-Ambocoelia and Arcuaminetes-Ambocoelia communities occupy the coarsest sediments at both the Gulf Road and Swamp Road localities. These sediments are gray silty shales and siltstones that are relatively thick and poorly bedded. Average organic matter content is 2.17%-2.3%.
The previous discussion indicates that the communities are mainly defined by the most common or dominant species. This was expected for clustering of a matrix of correlation coefficients between the original percentages of taxa within samples.
Consequently, we further explored community composition by calculating constancy (proportion or percent of samples within a community in which a species, say species i, is present) and fidelity (proportion or percent of samples within a community, with species i divided by total number of samples over all communities containing that species). The basic pattern is clear (Table 2).
Species with high constancies generally have low fidelities for any one community, and vice versa. This reflects the ecological strategies of Hamilton Group taxa. Common forms are broadly adapted and inhabit a wide variety of communities and environments. Such organisms are ecological generalists. Conversely, forms with a high fidelity for a certain community are General characteristics Lithologic code       Bembexia sulcomarginata environmentally restricted, specialized, and usually rare. Also, such taxa are usually restricted to one or a small number of samples within that community. Figure 6 illustrates the distribution of species in samples, showing that only a few taxa are found in a large number of samples, and vice versa. Clusters were also generated for the 50 species (Fig. 7). As is usual for paleoecological data, this dendrogram is much more loosely structured than that for the samples. For example, only 6% and 14% of the links are above correlation coefficients of 0.70 and 0.60, respectively. Despite low levels of similarities, the clusters do identify coherent groups of taxa with common distribution patterns, as shown below in telegraphic form. The species are listed as given on the dendrogram.
Cluster 1: Eumetabolotoechia to Orbiculoidea on the dendrogram of Figure 7. Common and rare taxa found in fine sediments; these taxa are most abundant in the Eumetabolotoechia, Eumetabolotoechia-Ambocoelia, Emanuella, and Arcuaminetes-Ambocoelia communities.
Cluster 2: Pterochaenia to Hallotheca. Common and rare species found in the finest sediments, mostly from the Cephalopod-Pterochaenia and Pterochaenia-Eumetabolotoechia communities.
Cluster 3: Arcuaminetes to Eldredgeops. Widely tolerant, common to rare forms, generally from the Eumetabolotoechia to Mucrospirifer-Ambocoelia communities; all or most species occur in the Arcuaminetes-Ambocoelia and Mucrospirifer-Ambocoelia communities.
Cluster 5: Rhipidomella and Modiomorpha. Rare species in a variety of assemblages, ranging from the Eumetabolotoechia to the Mucrospirifer-Ambocoelia communities.
Cluster 6: Leiopteria to Greenops. Rare species, mostly centered in the Arcuaminetes-Ambocoelia and adjacent communities. Cluster 7: Lingula to Echinocaris. Broadly adapted, common to rare organisms ranging from the Eumetabolotoechia to the Mucrospirifer-Ambocoelia communities, with all or most being recorded in the four communities from the coarsest sediments.
Cluster 10: Schuchertella to Grammysia. Wide-ranging, rare species found in communities varying from Pterochaenia-Eumetabolotoechia to Mucrospirifer-Ambocoelia communities; all of these taxa are known from the Arcuaminetes-Eumetabolotoechia and Arcuaminetes-Ambocoelia communities.

CORRESPONDENCE ANALYSIS
The first three axes of the correspondence analysis for 81 samples and 50 species account for 30.4%, 17.9%, and 10.2% of the information in the data set, respectively. The third axis mainly polarizes the unique Emanuella subumbona community from the rest, and this axis will not be discussed further. Correspondence axes I and II display the overall relations between the eight communities (Fig. 8A). The communities are arranged in a horseshoe-shaped figure in a systematic sequence: (1) the Cephalopod-Pterochaenia community; (2) the Pterochaenia-Eumetabolotoechia community; (3) the Eumetabolotoechia community; (4) the Emanuella community and (5) the Eumetabolotoechia-Ambocoelia community; (6) the Arcuaminetes-Eumetabolotoechia community; (7) the Arcuaminetes-Ambocoelia community; and (8) the Mucrospirifer-Ambocoelia community. The Emanuella community is most closely linked to that of Eumetabolotoechia. Interestingly, these relations resemble those that could be inferred from the dendrogram of Figure 7, except that the position of the Emanuella community is greatly distorted by the clusters. From left to right on the graph of the correspondence analysis (Fig. 8A), the communities reflect a trend of increasing sedimentary grain size, generally decreasing amounts of organic matter, and increasing sedimentary bed thickness. This sequence presumably reflects an overall tendency of decreasing depth, increasing agitation, and higher amounts of oxygenation. Furthermore, the sequence also replicates the most frequent vertical changes between and within communities, as annotated below.
Other information can also be gleaned from the correspondence analysis plot. The wide gap between the Cephalopod-Pterochaenia and the Pterochaenia-Eumetabolotoechia communities is notable. This gap indicates considerable cleavage between the two groups and suggests much variance between the two clusters. However, the Cephalopod-Pterochaenia and Pterochaenia-Eumetabolotoechia communities sometimes occur in adjacent samples. Perhaps a critical or threshold change in one or more environmental parameters dictated the assemblage that is found. Likewise, the Pterochaenia-Eumetabolotoechia community is discontinuously separated from the more diverse and shallower-water communities that are closely packed on the right side of the graph. Although they tend to occupy different regions in the correspondence axis plot, overlap can be seen between the Eumetabolotoechia and the Eumetabolotoechia-Ambocoelia communities, the Eumetabolotoechia-Ambocoelia and the Arcuaminetes-Eumetabolotoechia communities, the Arcuaminetes-Eumetabolotoechia and the Arcuaminetes-Ambocoelia communities, and the Arcuaminetes-Ambocoelia and the Mucrospirifer-Ambocoelia communities.
To study the patterns for the more shallow-water communities, we performed another correspondence analysis, in which the nine samples of the Cephalopod-Pterochaenia community were deleted in order to decrease the total dispersion of the data set. Axes I and II extracted 27.7% and 14.8% of the variance of the 72 samples and 50 species, respectively. The graph of the first two axes for the samples groups the communities from Pterochaenia-Eumetabolotoechia to Mucrospirifer-Ambocoelia, but the intergradations between the Eumetabolotoechia, Eumetabolotoechia-Ambocoelia, Arcuaminetes-Eumetabolotoechia, Arcuaminetes-Ambocoelia, and Mucrospirifer-Ambocoelia communities are illustrated more clearly (Fig. 8B). Although the second axis isolates the Emanuella community from the others, this assemblage remains closest to that of the Eumetabolotoechia community. The third correspondence axis, which is associated with 9.07% of the variance, separates the Pterochaenia-Eumetabolotoechia community from the rest of the data; it thus does not provide any new information. Figure 9 contains the first correspondence axes for the 50 species in all 81 samples. The horseshoe-shaped pattern for the taxa duplicates that of the samples and communities. The organ-   isms that are restricted to the finest-grained and darkest shales lie on the left side of the graph. Proceeding from left to right along the horseshoe, taxa represent associations found in progressively coarser rocks and shallower lithofacies.

CLUSTER SIGNIFICANCE TESTS
Cluster significance tests were applied to the communities known from adequate numbers of samples, namely, Cephalopod-Pterochaenia, Eumetabolotoechia, Eumetabolotoechia-Ambocoelia, Arcuaminetes-Eumetabolotoechia, and Arcuaminetes-Ambocoelia. The Pterochaenia-Eumetabolotoechia, Emanuella, and Mucrospirifer-Ambocoelia communities were deleted because of the small numbers of samples (which varied from two to four, in the communities deleted). Allowing for the missing communities, all pairs of communities that occurred adjacent stratigraphically were tested; the results are given in Table 3.
Student's t values demonstrate that mean species compositions of all pairs of communities differ significantly at the 0.01 probability level. The multivariate means of the different communities were clearly derived from different populations. However, this hypothesis tells little about the amount of overlap between the parent populations. This information is provided by  Tables 1 and 2. the W statistics of disjunction, which were tested using the data of Sneath (1979). The W figures denote that most pairs of communities were drawn from populations with rectangular distributions that do not overlap, again subject to a risk of 0.01. However, the Eumetabolotoechia-Ambocoelia community overlaps with the Eumetabolotoechia and Arcuaminetes-Eumetabolotoechia communities in terms of rectangular distributions. The Eumetabolotoechia and Arcuaminetes-Eumetabolotoechia communities are completely disjunct. The statistics suggest that one assemblage, the Eumetabolotoechia-Ambocoelia community, may not have occupied a completely separate region in the ecological gradient seen in the Cardiff Member. Figures 3 and 4 show the stratigraphic sections with the communities superimposed. The stratigraphic order of the communities is summarized in Figure 10 for the two sections and their composite. Each arrow denotes a vertical change from older to younger rocks. In general, the composite sequence of communities, listed from base to top, is Cephalopod-Pterochaenia, Pterochaenia-Eumetabolotoechia, Eumetabolotoechia, Emanuella or Eumetabolotoechia-Ambocoelia, Arcuaminetes-Eumetabolotoechia, Arcuaminetes-Ambocoelia, and Mucrospirifer-Ambocoelia; this sequence agrees with that given by  Table 1. the correspondence analysis. Several lines of evidence suggest that this sequence reflects decreasing depth. The basic sequence, ranging from the Cephalopod-Pterochaenia to Mucrospirifer-Ambocoelia communities, correlates strongly with increasing sedimentary grain size, increasing bed thickness, and progressively more intense bioturbation; it also relates to decreasing amounts of organics, increased species richness, and ecological diversity. Thus, the main trend is one of decreasing depth, with changes in associated parameters, such as increasing agitation, and we acknowledge that additional ecological factors may also have been operative. The two collections with the Emanuella community are only known at the Gulf Road section; these samples follow a large stratigraphic interval dominated by the Eumetabolotoechia community and are in turn overlain by the Arcuaminetes-Eumetabolotoechia community (Figs. 3 and 10). The lithology and organic content of the Emanuella samples resemble those of the Eumetabolotoechia community. Both Eumetabolotoechia and Emanuella subumbona are pedicle-attached brachiopods, although they exhibit few other common features. For example, the former is a rhynchonellid, whereas the latter is a spirifer, and Emanuella subumbona is much smaller than Eumetabolotoechia multicostata. The paleoecological changes involved here are not entirely certain and are discussed later.

SEQUENCE OF COMMUNITIES
Both stratigraphic sections analyzed here have thick intervals characterized by frequent changes in community composition. Examples are: Pterochaenia-Eumetabolotoechia and Different ranges of communities and environments occur in the two stratigraphic sequences. The oldest and youngest communities at the Swamp Road section consist of Eumetabolotoechia and Mucrospirifer-Ambocoelia, but those at Gulf Road vary from Cephalopod-Pterochaenia to Arcuaminetes-Eumetabolotoechia and Arcuaminetes-Ambocoelia. Thus, the Swamp Road area began and ended in shallower habitats than did the Gulf Road section.
The relations between the samples are shown by the first two axes of the correspondence analysis, as has been discussed above (Fig. 8A). Inasmuch as the overall pattern involves both axes, the data were combined into a single sequence, which will be termed the correspondence ordination. First, a fourth-order polynomial was fitted to the first two axes of the correspondence analysis, yielding a squared correlation coefficient of 0.890 (Fig. 11). Next, orthogonal projections of the 81 samples onto the polynomial were determined. Finally, the polynomial was "unfolded" and scaled so that the points on the left and right of the curve equaled 0 and 95, respectively. This technique is commonly employed by ecologists to reduce curvilinear ordinations of various types (e.g., Ludwig and Reynolds, 1988;Phillips, 1978), although typically with lower-order polynomials.
The correspondence ordination is interpreted as a generalized measure of relative depth and position along an ecological gradient, as has been suggested in other studies, including Cisne and Rabe (1978), , and . Several facts indicate that this ordination does provide Downloaded from http://pubs.geoscienceworld.org/books/book/chapter-pdf/5177698/spe545-09e.pdf by guest such information. Scores for the samples on the ordination are highly correlated with sedimentary grain size and community composition, and they vary inversely with the organic matter content. For example, the Spearman's rank correlations between the correspondence ordination versus sedimentary grain size, organic matter content, and community composition are 0.828, -0.732, and 0.964, respectively. In addition, there are associations between the correspondence axis ordination and increasing amounts of sedimentary bed thickness, bioturbation, density of fossils, and taxonomic and ecological diversity, although the exact correlations have not been calculated. High ordination scores in this analysis were linked with comparatively shallow water, high agitation, coarse substrates, and high oxygen contents, and vice versa. The correlation between the correspon-dence axis ordination and depth is both generalized and blurred; the independent effects of depth, such as agitation and substrate texture, cannot be completely disentangled, although some suggestions are presented below.
The scores for the samples on the correspondence axis ordination are plotted against stratigraphic position in Figure 12. For the purposes of interpretation, we will assume that depths and scores exhibit a linear relationship. The first-order pattern in both stratigraphic sections is regression or relative fall of sea level. The rapid increase in the scores at the base of the Gulf Road section, especially involving the Cephalopod-Pterochaenia, Pterochaenia-Eumetabolotoechia, and perhaps the Eumetabolotoechia communities, suggests rapid shoaling and augmented oxygenation. Later rates of falling sea level at Gulf Road are  Figure 11. Fourth-order polynomial fitted to axes I and II of the correspondence analysis of 81 samples shown in Figure 8A. The species codes are listed in Table 1.  Table 1. more gradual and exhibit several smaller-scale transgressions and regressions superimposed on the overall trend. The base of the Swamp Road outcrop began in deeper water than that seen in the youngest beds at Gulf Road. Consequently, the range of depths spanned by the two sections overlaps considerably. The Swamp Road area is characterized by a decrease in relative sea level that appears almost linear. The superimposed fluctuations of scores for the correspondence axis ordination are minor, and they may or may not be statistically significant. The Mucrospirifer-Ambocoelia community caps the Swamp Road section, which ended in shallower habitats than did the older section at Gulf Road.

COMMUNITY COMPOSITION AND ECOLOGICAL STRUCTURE
Ecological groups within the eight Cardiff Member communities will be discussed in order of decreasing depth. The living habits of the shelled organisms are divided into ecological categories that represent a combination of feeding habits and mode of locomotion or attachment. The nine categories are: (1) pedicle-attached suspension feeders; (2) reclining suspension feeders in which the shell may lie at, or slightly above, the substrate; (3) epibyssate or very shallow endobyssate suspension feeders; (4) endobyssate suspension feeders; (5) burrowing suspension feeders that are probably mostly sedentary (rather than active) burrowers; (6) burrowing deposit feeders; (7) grazers that crawl along the surface; (8) a highly variable suite of "collectors," which includes mainly bottom plowers, crawlers, or walkers, some of which may have been capable of swimming or burrowing; the organisms denoted as "collectors" may have been deposit feeders, scavengers, or predators on small animals and plants; and (9) swimming predators and scavengers. Reconstruction of living habits for most species is straightforward, and it was based mainly on references listed previously. Several forms do require additional explanation. The shape of the small and thin-shelled bivalve Pterochaenia fragilis definitely denotes the presence of a byssus (Stanley, 1972), but the site of byssal attachment is uncertain. This bivalve spans a wide variety of facies, from laminated black shales and limestones to siltstones. The species may be extremely abundant in dark-gray shales, where some specimens are preserved in inferred living position; such occurrences are interpreted as representing organisms attached to the seafloor. Pterochaenia fragilis is a common accessory in various level-bottom Hamilton Group communities, a distribution consistent with a benthic habit. A few shells were obtained from laminated black shales, where they are associated with pelagic faunas. These could represent epipelagic organisms, as suggested by Thayer (1974) for Upper Devonian material (see also discussion in Boyer and Droser [2009] on such organisms). 2 Formerly known as Leiorhynchus, this taxonomy was revised by Paul Sartenaer.
Although much less common, the distribution and some morphological features of Buchiola retrostriata parallel those of Pterochaenia fragilis, and their living habits are considered to be the same. "Praecardium" robusta, a rare element in many Cardiff Member communities, is a large (up to 60 mm long) bivalve that was probably epifaunal and may have been epibyssate. Although "Praecardium" often occurs in deeper-water, fine-grained facies, its large size is not typical of most epipelagic taxa. The asymmetric valves suggest that this bivalve did not have the swimming capacity characteristic of some modem pectinaceans in comparable habitats (cf. Stanley, 1972). We therefore conclude that most individuals of Praecardium likely dwelled epifaunally, but some specimens may have been epipelagic.
Shell shapes and apertural orientations of the gastropods and bellerophontids in the Cardiff Member demonstrate that these animals dwelled epifaunally on the sedimentary substrate or on plants (e.g., Linsley, 1977). Their food supplies are uncertain. Recent shallow-water counterparts graze on benthic algae, whereas those dwelling in aphotic regions are deposit feeders (Rex, 1973). Inasmuch as stratigraphic position and lithology indicate that these sediments were deposited at moderate depths, the Cardiff Member gastropods are tentatively considered to have been herbivores. The bellerophontids-extinct molluscs thought by some to have been gastropods and by others possibly to have been monoplacophorans (this latter grouping is, however, itself considered polyphyletic)-are here classified as herbivores as well. Following Bouchet et al. (2005Bouchet et al. ( , 2017, we consider bellerophontids as incertae sedis within Class Gastropoda for now.

Cephalopod-Pterochaenia Community
This community occurs only in the organic-rich, deep-water shales of the Gulf Road section (Fig. 3). Both species richness and ecological diversity are low, with means of 5.22 species and 3.56 ecological categories, respectively (Table 2). Surprisingly, equitabilities are comparatively high, with a more even distribution of species frequencies and ecological categories than in any other Cardiff Member community. The assemblage is dominated by pelagic cephalopods. Among the most common bivalves, Pterochaenia fragilis is tentatively interpreted as either epipelagic or more likely epibyssate. Eumetabolotoechia multicostata, a pedunculate rhynchonellid that also occurs in samples from the upper stratigraphic range of this community, 2 commonly forms dense accumulations in dark-gray shales stratigraphically higher in the sections, and it is likely that individuals of this species were moored to the substrate. The presence of pedicle borings on valves further demonstrates that some individuals were attached to other individuals of multicostata (see also discussions of the ecology of this brachiopod genus in Thompson and Newton [1987] and Boyer and Droser [2009]). Epipelagic specimens of Eumetabolotoechia multicostata are not known. We thus infer that Eumetabolotoechia multicostata and Pterochaenia fragilis were epibenthic rather than epipelagic. If our interpretation is correct, the Cephalopod-Pterochaenia community reflects a mixed pelagic-benthic fauna.

Pterochaenia-Eumetabolotoechia Community
Although many taxa in this community occur in the lower Cephalopod-Pterochaenia assemblage discussed above, the frequency of cephalopods is much lower (~2.16%, or an order of magnitude lower than that of the underlying Cephalopod-Pterochaenia community; Table 2). Many benthic species in this community also occur in the deeper-water Cephalopod-Pterochaenia community; the most common of these are pedunculate brachiopods (Eumetabolotoechia multicostata and a few specimens of Emanuella subumbona) and epibyssate bivalves, most notably Pterochaenia fragilis. These two ecological categories taken together account for over 90% of the individuals in this community. Minor frequencies of infaunal deposit-feeding bivalves (Nuculoidea sp.) and reclining brachiopods represent the lowest stratigraphic occurrences of these groups in our study. Most of the sediments associated with this community show some traces of bioturbation (usually small, well-defined, horizontal burrows) of 1.5-2.0 mm diameter, implying that the bottom waters were at least dysaerobic (and probably aerobic). Such burrows do not occur in the Cephalopod-Pterochaenia community. The introduction of new species and ecological categories in the Pterochaenia-Eumetabolotoechia fauna augments species richness and ecological diversity considerably, with means of 8.33 species and 5.0 ecological types per sample. This greater diversity suggests a shift in environmental conditions relative to those of the mixed pelagic-benthic Cephalopod-Pterochaenia community. Because of the dominance of pedicle-attached and epibyssate suspension feeders, the equitability of the Pterochaenia-Eumetabolotoechia community is less than that of most of the other assemblages.
The Pterochaenia-Eumetabolotoechia community likely reflects a pioneer fauna developed as the substrate initially became favorable for benthic settlement. The grain size of sediments associated with the Cephalopod-Pterochaenia and the Pterochaenia-Eumetabolotoechia communities appears constant, but the mean organic content declines slightly, from 3.0% to 2.8%. The contrast between the Cephalopod-Pterochaenia and Pterochaenia-Eumetabolotoechia communities is consistent with a threshold or trigger effect, in which small environmental changes may have catalyzed large ecological and taxonomic shifts in faunal composition. One such environmental change between these two communities may have been an increase in oxygenation.

Eumetabolotoechia Community
The pedunculate brachiopod Eumetabolotoechia multicostata attains peak abundance in this community, where it constitutes nearly two thirds of the individuals (Table 2). Some beds of dark-gray shale are densely covered with this assemblage, whereas other bedding surfaces are almost barren, a pattern we interpret as reflecting opportunistic settlement behavior. Pedicleattached brachiopods, especially Eumetabolotoechia multicostata and small numbers of Emanuella subumbona, form 77% of the fauna. Other accessory taxa, at levels of 6% to 7% each, include epibyssate bivalves (mainly Pterochaenia fragilis), reclining brachiopods such as Arcuaminetes scitulus (formerly Devonochonetes scitulus) and Spinulicosta spinulicosta, and deposit-feeding nuculoids. Nuculoidea sp., the most common burrowing deposit feeder, was a shallow-burrowing, asiphonate bivalve. Other ecological categories are either sparsely represented or unrepresented. Species richness and ecological diversity continue to be comparatively low, with means of 8.79 species and 5.21 ecological categories, respectively. The dominance of a few pedicle-attached suspension feeders results in an equitability that is the lowest of all eight communities in the data set.
The Eumetabolotoechia community is usually underlain by, and often succeeded by, the Pterochaenia-Eumetabolotoechia community. In fact, most taxa in the Eumetabolotoechia community were already present in the Pterochaenia-Eumetabolotoechia fauna; the community transition mostly involves changes in the frequencies of these fossils (e.g., Figs. 5 and 8).
The Eumetabolotoechia community occurs in distinctly coarser rocks with thicker beds, more bioturbation, and slightly lower organic content (Figs. 3, 4, and 10). These interrelated changes suggest a pattern of decreasing depth, with an associated increase in agitation and a shift toward firmer substrates.

Eumetabolotoechia-Ambocoelia Community
This assemblage represents a higher species richness community closely related to the Eumetabolotoechia community (Table 2), as is documented by the multivariate analyses discussed above (Figs. 5 and 8) and by recurring stratigraphic juxtapositions of these two assemblages. The Eumetabolotoechia-Ambocoelia community invariably succeeds the Eumetabolotoechia community stratigraphically (Figs. 3, 4, and 10). Pedunculate brachiopods, typically Eumetabolotoechia multicostata and Ambocoelia umbonata, account for nearly 60% of the fauna. This represents a significant decline from the 77% in the Eumetabolotoechia community. Reclining suspension feeders-Arcuaminetes scitulus and a few individuals of Mucrospirifer mucronatus-constitute 21% of the fauna, which is also significantly higher than in the Eumetabolotoechia community. Burrowers, both deposit and suspension feeding, are also present. Deposit-feeding burrowers are represented by forms that lived at various depths. Nuculoidea sp. was a shallow burrower, whereas Nuculites oblongatus, which becomes common in this community for the first time, has a shell shape signaling the presence of siphons and a deeper burrow. Epibyssate bivalves decline to about 3%, a downshift caused by decreasing Pterochaenia fragilis.
High species richness values and ecological diversities appear for the first time in this community, averaging 11.9 and 6.21 species and ecological categories, respectively. Likewise, the equitabilities increase, as the Eumetabolotoechia-Ambocoelia community consists of a greater number of taxa and ecological categories than do the communities occurring stratigraphically lower in the sections. Sediments associated with this community resemble those of the Eumetabolotoechia suite, except that organics are sparser, with an average of 2.4 wt%, as compared with 2.7 wt%. We infer that the Eumetabolotoechia-Ambocoelia assemblage reflects a more agitated, and possibly slightly more oxygenated, habitat than the underlying Eumetabolotoechia and other communities.

Emanuella Community
Multivariate analyses demonstrate that this peculiar community relates most closely to the Eumetabolotoechia fauna. These communities are associated, and they share similar ecological roles, species, diversity, and equitability (Figs. 3, 4, and 10; Table  2). Pedicle-attached suspension feeders dominate, but Emanuella subumbona is much more abundant than Eumetabolotoechia multicostata. Compared to the Eumetabolotoechia community, reclining brachiopods-mostly Arcuaminetes scitulus and Mucrospirifer mucronatus-are more frequent, but burrowing deposit feeders and epibyssate bivalves are rarer. The Emanuella community is associated with silty shales, in contrast to the Eumetabolotoechia community, which occurs in shales. Organic content is slightly less abundant than in the rocks of the Eumetabolotoechia community. These lithologic differences suggest that the Emanuella community occupied a slightly shallower and more agitated habitat than the Eumetabolotoechia community.
Taxonomic and ecological diversity exceed those of the Eumetabolotoechia-Ambocoelia assemblage. The Arcuaminetes-Eumetabolotoechia fauna occurs in gray shale and silty shales that are generally coarser than those of the associated Eumetabolotoechia and Eumetabolotoechia-Ambocoelia assemblages. Organic contents are lower than in the sediments of the Eumetabolotoechia community, but they are comparable to those of the Eumetabolotoechia-Ambocoelia community. The inferred environmental changes include decreasing depth and increasing agitation.
The abundance of the small reclining brachiopod Arcuaminetes scitulus deserves special note. Small recliners such as Arcuaminetes scitulus are often considered characteristic of soft soupy substrates in quiet water. In fact, suspension-feeding brachiopods require some water movement to provide fresh food and oxygen and to remove wastes. The chonetid shell must have lain close to the substrate and the boundary layer between the substrate and seawater. Current velocity decreases exponentially as the seafloor is approached. Consequently, a certain minimal amount of agitation is required to enable these small reclining animals to survive. This is wholly consistent with the distribution of these small reclining taxa in the Cardiff Member communities. Arcuaminetes scitulus is either completely absent or rare in the deepest and quietest water benthic communities (Pterochaenia-Eumetabolotoechia and Eumetabolotoechia). The species appears only after a critical amount of agitation is reached.

Arcuaminetes-Ambocoelia Community
This fauna is closely linked to, and is interbedded with, the Arcuaminetes-Eumetabolotoechia and Mucrospirifer-Ambocoelia communities 8,and 10). Reclining suspension feeders, mostly Arcuaminetes scitulus, along with some shells of Mucrospirifer mucronatus and Spinulicosta spinulicosta, lead the roster of ecological categories and represent 58% of the individuals (Table 2). Next are pedunculate brachiopods, largely Ambocoelia umbonata with rare Emanuella subumbona, which comprise 21% of the assemblage. Burrowing deposit feeders are diverse, but they are slightly less abundant than in the Arcuaminetes-Eumetabolotoechia community. The small cohort of burrowing suspension feeders includes forms that occur also in the deeper-water communities. Species richness values and ecological diversities of the Arcuaminetes-Ambocoelia community slightly exceed those of Arcuaminetes-Eumetabolotoechia, but the relatively high abundance of a few reclining suspension feeders leads to lower equitabilities in this assemblage. Lithologies associated with the Arcuaminetes-Ambocoelia community are shaly siltstones and silty shales that are coarser, thicker bedded, and more highly bioturbated than those of the Arcuaminetes-Eumetabolotoechia fauna. Organic contents of the three relatively shallow-water communities in the Cardiff Member are quite similar. Thus, the Arcuaminetes-Ambocoelia assemblage almost certainly occupied shallower, higher-energy habitats than the Arcuaminetes-Eumetabolotoechia community.

Mucrospirifer-Ambocoelia Community
This group of organisms occurs in the coarsest sediments in the study-silty shales to siltstones-and represents the most shallow-water fauna present (Figs. 3 and 4). However, the lithologies and sedimentary structures indicate gentle to moderate (rather than high) amounts of wave and current agitation below normal wave base.
Reclining brachiopods (45%) and pedunculate brachiopods (37%) form the trophic nucleus of this community ( Table 2). The most common taxa in this assemblage, e.g., Ambocoelia umbonata, Mucrospirifer mucronatus, and Arcuaminetes scitulus (in decreasing order of abundance), are also known from the closely associated Arcuaminetes-Eumetabolotoechia and Arcuaminetes-Ambocoelia communities (Figs. 5, 8, and 10). The large spiriferid Spinocyrtia granulosa is a common accessory in this fauna; this is the lowermost stratigraphic occurrence of significant numbers of Spinocyrtia granulosa. Burrowing deposit feeders continue to decline, a trend probably correlated with the higher agitation and lower amounts of organic matter. Large epibyssate bivalves, such as Actinopteria boydi, Cornellites fasciculata, Gosseletia triquetra, and Leiopteria sp., and endobyssate clams such as Goniophora hamiltonensis, Grammysia bisulcata, and Modiomorpha spp., show increased abundance. Maximum diversitiesboth taxonomic and ecologic-are attained in some samples of the Arcuaminetes-Ambocoelia and Mucrospirifer-Ambocoelia faunas. However, taxonomic and ecological evenness are higher in the Mucrospirifer-Ambocoelia community.

Cluster Analysis of Trophic Structures
Taxonomic and ecological structures of the communities are highly correlated, as is strongly indicated in the cluster analysis of the 81 samples for the nine ecological categories (Fig. 13). This dendrogram derives from the unweighted-pair-group method on a matrix of correlation coefficients. The communities can be divided into four major types with respect to the ecological data: (1) Samples 1-9 belong to the Cephalopod-Pterochaenia community, characterized by a very low species richness and mixed pelagic-benthic assemblage.
(2) The Pterochaenia-Eumetabolotoechia community includes samples 10, 12, and 13. This is a low-species-richness pioneer fauna that mostly includes pedicle-attached brachiopods and small epibyssate bivalves. These organisms seem to appear as soon as the seafloor becomes generally suitable for occupation by bottom dwellers. (3) The samples ranging from numbers 11 to 55 on the dendrogram are mostly from the Ambocoelia and the Emanuella communities. All of these samples are dominated by pedicle-attached brachiopods. Species richness values and equitabilities range from low to moderate. Within this large group, the samples within the Eumetabolotoechia community tend to separate from those of the Eumetabolotoechia-Ambocoelia and Emanuella communities, which are more diverse, both ecologically and taxonomically. (4) A large cluster extends from samples 27 to 29 on the dendrogram. These samples are grouped in the Arcua-minetes-Eumetabolotoechia, Arcuaminetes-Ambocoelia, and Mucrospirifer-Ambocoelia communities. Pedicleattached and reclining suspension feeders form the trophic core for all of these assemblages. Maximum taxonomic and ecological diversities are observed in these faunas. A matrix of correlation coefficients among the nine ecological roles was calculated for 72 samples from the seven benthic communities (Table 4). The magnitudes for the correlations that differ significantly from a population value of nil, i.e., no correlation at all, equal 0.232 and 0.302 at the 0.05 and 0.01 probability levels. Most (53%) of the observed correlations are not statistically significant. The correlation with the highest magnitude is -0.906 for the pedicle-attached versus reclining suspension feeders. Thus, these two categories are inversely related, as would be expected from the composition of the communities. Many Cardiff Member recliners-particularly common groups such as chonetids-had relatively low profiles with respect to the sediment. Therefore, these animals could not become well established until critical amounts of gentle agitation were attained, in order to supply them with food and oxygen and to remove waste materials from near the substrate. Most pedicle-attached brachiopods filtered water higher above the substrate, where more water movement would have been available in the quieter habitats. In addition, the fine sediments of the deeper communities could easily have fouled the shells of small recliners such as Arcuaminetes scitulus. The correlation coefficient with the next highest magnitude, a value of 0.469 between reclining filter feeders and surface grazers, reflects mainly the rarity of both groups in the finest sediments of the Cardiff Member. Here, one category only explains 22% of the variance of the other (i.e., correlation coefficient squared). Therefore, except for the pedicle-attached and reclining brachiopods, the various ecological categories seem to be basically independent of each other, as would be predicted from diversity theory.
Correlations between the taxonomic and ecological diversities and between the taxonomic and ecological equitability are all high; these range from 0.808 to 0.858 (Fig. 14). Thus, these pairs of parameters consistently track one another, as can be inferred from the annotations on the communities. Equitabilities in our samples are largely independent of the diversities, and vice versa. Contrary to popular belief, communities with low species richness may have taxa that are evenly distributed; they are not necessarily dominated by a single species. To some extent, the converse is also true for the Cardiff Member data. For example, the Arcuaminetes-Ambocoelia community is rich in taxa, yet Arcuaminetes scitulus makes up almost half of the individuals.

Species Richness and Equitability
As expected from the previous discussion of the communities, both taxonomic richness and ecological diversity increase with progressively younger beds in both stratigraphic sections (Fig. 15). The stratigraphically highest samples at Swamp Road (including portions of the Mucrospirifer-Ambocoelia and Arcuaminetes-Ambocoelia communities) exhibit slightly lower species richness values, representing a slight reversal of the overall trend of increasing species richness in the sections. For both sections, species richness and ecological diversity are highly correlated. The overall environmental changes of decreased depth   Table 1. and increased sedimentary grain size, agitation, and oxygenation favor greater diversities.
Equitabilities are more independent of stratigraphic position. The most evenly distributed communities are those of Cephalopod-Pterochaenia, with low species richness and ecological diversity, and Arcuaminetes-Eumetabolotoechia, with higher species richness and higher ecological diversity. The Eumetabolotoechia fauna, highly dominated by the pedicle-attached brachiopod E. multicostata, yields the lowest equitabilities. Low equitabilities also characterize the Pterochaenia-Eumetabolotoechia and Emanuella communities. This reflects the fact that these deep-water communities are basically pioneer epifaunal assemblages that developed when conditions first became favorable for the establishment of benthic organisms. As depth decreased in both stratigraphic sections, the increased agitation and oxygenation favored more numerous taxa and ecological categories that exploited more resources.

Taxonomic Composition
Vertical distributions of some of the dominant taxa appear in Figure 16. Several generalizations can be made. Virtually all common Hamilton Group species are protean organisms with broad environmental tolerances, and they span many types of communities. Some organisms are concentrated in one or several assemblages-for example, orthocones, Emanuella subumbona, and Pterochaenia fragilis-but others range more widely, such as Eumetabolotoechia multicostata, Arcuaminetes scitulus, Mucrospirifer mucronatus, and Ambocoelia umbonata. In some instances, species composition and abundance can fluctuate widely between adjacent samples within a stratigraphic section, especially at transitions between different communities; however, in most cases, less dramatic variation occurs during transitions. On average, changes between communities explain more of the variation in the data than changes within assemblages.
Among abundant species, Ambocoelia umbonata, Arcuaminetes scitulus, Eumetabolotoechia multicostata, Mucrospirifer mucronatus, and Pterochaenia fragilis seem controlled more by factors related to grain size than by organic content, even though both parameters contribute significantly. Conversely, Nuculoidea sp. and large orthocones are more closely correlated with organic content. This was expected for Nuculoidea because of its depositfeeding habits. Likewise, the orthocone frequency declines drastically from the Cephalopod-Pterochaenia community to the Pterochaenia-Eumetabolotoechia community. The distribution of Emanuella must be influenced by parameters other than grain size and organic content, based upon the regressions for these variables, which are not significant.

Ecological Roles
Ecological categories exhibit the following major systematic vertical changes in the two stratigraphic sections (Figs. 17 and 18): (1) Swimming predators abound in the Cephalopod-Pterochaenia community at the base of the Gulf Road section, but frequencies of nektonic predators decline sharply in all other communities. (2) Following establishment of benthic faunas, the dominant ecological categories vary among three groups of epifaunal suspension feeders. In the deepest waters, small epibyssate bivalves, notably Pterochaenia, predominate. Pedicle-attached brachiopods, especially Eumetabolotoechia multicostata, become most abundant in the Eumetabolotoechia and associated faunal suites. Communities in shallower water exhibit an inverse relation between reclining and pedicle-attached brachiopods, in which recliners are most frequent in the shallowest sediments in the two sections. Higher agitation is apparently required to provide adequate water circulation close to the seafloor for small recliners such as Arcuaminetes scitulus.
(3) Benthic assemblages of deeper waters are predominantly epifaunal. Burrowing deposit feeders commonly appeared at shallow depths below the substrate in the Eumetabolotoechia community, and greater burrowing depths were penetrated when the Eumetabolotoechia-Ambocoelia and Arcuaminetes-Eumetabolotoechia communities occupied the area. Burrowing deposit feeders began to decline in shallower waters as sediments progressively had decreasing amounts of organic matter and coarser sediment grain sizes. (4) Endobyssate and burrowing suspension feeders are not abundant. Both groups were most frequently collected in the faunas of the shallower lithofacies, ranging from Eumetabolotoechia-Ambocoelia to Arcuaminetes-Ambocoelia for the burrowers and to Mucrospirifer-Ambocoelia for the endobyssates. Peak abundances of burrowing and endobyssate suspension feeders coincided with the decline of burrowing deposit feeders. (5) Surface-dwelling gastropods, believed to have grazed on plants, occurred most commonly in comparatively shallow regions with diverse associated taxa. (6) Surface collectors, such as trilobites and hyolithids, were erratically distributed as minor accessories of varied communities. Admittedly, this ecological category is highly variable, and a range of feeding habits is included-for example, deposit feeding, scavenging, and predation. This trophic role accounts for no more than 1.0% of the individuals in any community.

TESTING PROPOSALS OF COMMUNITY STABILITY AND COORDINATED STASIS
Hamilton Group faunas are frequently cited as paradigmatic of the pattern of long-term community stability termed coordinated stasis (Brett and Baird, 1995;Brett et al., 1996;Schopf, 1996), with several workers referring to these assemblages as the "type" example of coordinated stasis. However, this intriguing proposal-that some marine evolutionary ecological units (EEUs of Boucot [1983] and Sheehan [1996]) exhibit long-term stability on a scale of millions of years-has been discussed more far widely than it has actually been tested, even now. Presenceabsence data and taxonomic ranges of common taxa do not suffice to demonstrate coordinated stasis (McKinney et al., 1996;Schopf and Ivany, 1997). Instead, quantitative analysis of large samples distributed over a significant temporal interval must be undertaken (see also Bennington and Bambach, 1996). In particular, the stability of rare taxa (and their variability) has not previously been explored adequately in general (Boucot, 1996;McKinney et al., 1996;Ivany, 1999), let alone within the faunas of the lower Hamilton Group.
This study provides both documentation and analysis of paleoecological structure and taxonomic composition for the lower Hamilton Group-an interval necessary for rigorous tests of whether the communities themselves, rather than just the taxa, range through this entire stratigraphic unit and the time it encompasses. It provides a first detailed look at community stability in this lower interval, integrating taxa that are moderately rare (>1% of the fauna). Taken together with the few other quantitative Hamilton Group studies (e.g., Savarese et al., 1986;Brower and Nye, 1991;McCollum, 1991;Bonuso,  2001; Bonuso et al., 2002aBonuso et al., , 2002b, we can make some initial inferences about the structure and temporal persistence of Hamilton Group marine communities. Our data reveal a few-but not all-lower Hamilton Group communities with counterparts in upper Hamilton Group faunas. This means that stability is indicated for some associations but not others. For instance, the Arcuaminetes-Ambocoelia community and the Eumetabolotoechia community apparently occur higher in the Hamilton Group, as described by other workers (cf. Savarese et al., 1986;McCollum, 1991) 3 . However, the other six communities we recognized herein do not clearly recur with the same array of taxa in other studies. Abundant and common taxa within these other associations are known from higher in the Hamilton Group, and hence range through (Brett and Baird, 1995), but the internal structure of later associations, as described in Savarese et al. (1986) and McCollum (1991), is incongruent with that of our communities.
In coordinated stasis, and particularly in the ecological locking model, these communities should recur temporally and should prove statistically similar throughout the temporal interval. As Ivany has pointed out, frequencies of taxa may vary, but the associations ought to recur. This pattern would be predicted in both the environmental tracking and ecologic locking models-and it would especially be an integral prediction of ecologic locking. Our information from the lower Hamilton Group suggests that this is not supported by currently available data. In that regard, the Hamilton faunas may resemble those documented by Patzkowsky and Holland (1999, p. 301) from the Middle to Upper Ordovician of the Nashville Dome. Their results indicate that Ordovician communities "were not tightly integrated in their structure, but rather were fluid with species and genera moving in and out of communities in response to local changes in environmental conditions." Further rigorous quantitative studies of upper Hamilton Group faunas should reveal whether, as in the Ordovician study, loosely integrated communities characterize the duration of Hamilton Group deposition.
Our study also yielded several quantitative results for the lower Hamilton Group that, in themselves, pose further questions for the hypothesis of ecological locking. First, these Marcellus Subgroup communities exhibit low levels of integration and correlation among the constituent species, in contrast to the predictions of ecological locking. Second, rapid swings in the abundance of taxa occur in adjacent beds, with no apparent corresponding shifts in lithology or other environmental indicators. Ecologic categories of lower Hamilton Group taxa in our study show low correlations (53% of observed correlations are not statistically significant). In sum, the lower Hamilton Group faunas analyzed here reflect only loosely structured associations, even though the individual communities are statistically discreteresults that seem not to suggest that the faunas are tightly integrated as has been predicted by some (but not all) coordinated stasis proponents.
These conclusions resemble those reached by Tang and Bott jer (1996), who inferred that Jurassic paleocommunities in western North America also fail to show evidence in support of the ecological locking model. However, Tang and Bottjer's conclusions came solely from presence-absence data or data on dominant taxa; Schopf and Ivany (1997) therefore correctly pointed out that Tang and Bottjer had not conducted all the tests necessary to probe the question. Our study, with a far richer abundance database, now reveals that with appropriate tests of Hamilton Group faunas, 4-6 m.y. persistence of ecologically locked and tightly structured communities is still not indicated.

ACKNOWLEDGMENTS
Brower and Newton dedicate this paper to the honor and memory of John M. Dennison, M.S. advisor to Cathryn Newton and colleague of James Brower, and Willis B. Newman, our joint doctoral student, much respected and beloved, who conducted this fieldwork and analyses but did not complete the dissertation.
Acknowledgment is made to the donors of the Petroleum Research Fund of the American Chemical Society for support of this research program. We also gratefully acknowledge a New York State Honorarium to W.B. Newman and ongoing research support to C.R. Newton from the Vice Chancellor and Provost of Academic Affairs at Syracuse University. Thanks are due to G.C. Baird of the State University of New York at Fredonia, C.E. Brett of the University of Cincinnati, T.X. Grasso of Monroe Community College, L.C. Ivany of Syracuse University, D. and R.M. Linsley of Colgate University, H.B. Rollins of the University of Pittsburgh, and B. Selleck of Colgate University for assistance in the field and for helpful discussions and scientific friendship over many years. Cathryn Newton offers deep and lasting gratitude to Jim Brower, Bob Linsley, and Bud Rollins for becoming tremendous mentors who showed her how to invest in others. John Harper and an anonymous reviewer suggested slight revisions that markedly improved the figures and paleontological and stratigraphic descriptions; K. Lee Avary further lent wisdom and knowledge to the revisions process-and friendship always. CRN also thanks her students, from whom she learned much. Special thanks go to G.K. Wis-