Discussion

Subdivision of the bryozoan genus Fenestella into several genera during the past half century raises the question of whether these genera 1) represent biological entities (individual or paraphyletic clades) or 2) are concepts of convenience that divide a continuum of morphologies involving independent repetitive evolution of artificially defined, plastic character states. With few exceptions, discussed below, the recently defined genera were initially differentiated on the basis of zooecial endozonal (interior) morphologies, geometric relationships among zooecia in the endozone, or consistent presence of unique heterozooecia, which are not among the nine exterior characters used in this analysis.

In this study, we have addressed the question of whether fenestellid genera as currently used have overall morphological coherence independent, or largely independent, of the initially defining characters, or whether instead species assigned to a genus plot in morphospace independently of one another. To do this we have used nine quantitatively variable characters that can be observed on the colony surface (Figure 2) and that generally were not part of the basis for naming the genera.

The relatively recently named genera, and a small number of genera named earlier and used as conceptual controls, each occur within only a portion of the occupied morphospace as defined by principal components of the nine quantifiable surface characters. Some genera in the study are represented by a single species, and each of these species is limited to a small portion of the occupied morphospace. This pattern is to be expected, inasmuch as elements of size and proportions are typical for discrimination of fenestrate bryozoan species within a genus. In addition, each genus that is represented by two or more species in the study also has a limited range of sizes of externally observable characters that – with one exception – differentiates it from other genera in the study.

Genera Originally Diagnosed in Part on One or More of the Nine Characters Used in This Study

Five of the included genera (Alternifenestella, Cubifenestella, Laxifenestella, Minilya, and Spinofenestella, Table 9) each had one or more elements of size of the meshwork or features seen on branch surfaces embodied in the original diagnosis. Elements of size were an important part of the characterization of Cubifenestella when it was named, but size was less emphasized in defining the other four genera. Nevertheless, it is worth examining the apparent effect of the initially mentioned size element(s) in positioning each of these genera in morphospace.

Alternifenestella. The externally observable size characteristics included in the first diagnosis (Morozova 1974) of Alternifenestella Termier and Termier, 1971 were narrow branches and narrow dissepiments. Branch width (BW) does not appear in Table 9 as an important PCA loading coefficient on any of the axes for Alternifenestella. Dissepiment width (DW) does plot small on PCA-2 and PCA-5 for Alternifenestella (to the top on Figure 13.5 and to the right on Figure 16.2, and Table 9), but the magnitude of the loading coefficient is not dominant relative to other important characters (Table 4).

Cubifenestella. The original diagnosis of this genus noted intermediate to robust zoaria, with open mesh spacing based on intermediate to large fenestrules, with narrow to intermediate branch width and intermediate dissepiment width (Snyder 1991). Indeed Cubifenestella does plot entirely in the positive area of PCA-1 (Figure 14.5), confirming Snyder's concept of its general robustness (large size) based on the three species that he recognized. However, fenestrule length and width are not reflected on PCA-2 through PCA-5 (Table 9), which suggests that fenestrule size has not differentially affected the placement of Cubifenestella in the morphospace defined here. Small dissepiment and branch width, however, do play a role on PCA-4 (to the right on Figure 13.4, Table 9).

Laxifenestella. Morozova (1974) included moderately broad dissepiments in the original diagnosis of Laxifenestella. In addition to overall large features of PCA-1, dissepiment width is large on PCA-5 for Laxifenestella (to the bottom on Figure 13.1, Table 9). However, the Laxifenestella cloud is not differentiated on PCA-2 or PCA-4, which also contain dissepiment width as an important character. Therefore, breadth of dissepiments apparently has a negligible effect among the external characters of Laxifenestella.

Minilya. The original diagnosis of this genus included small nodes (Crockford 1944). The Minilya cloud is generally positive on PCA-3 (to the upper region on Figure 16.1, Table 9) but is roughly centered on both PCA–4 and PCA-5 (Figure 13.5). Node diameter may have had a small, but certainly not preeminent effect in determining the position of the Minilya cloud.

Spinofenestella. The original diagnosis of Spinofenestella Termier and Termier (1971) included short fenestrules. The Spinofenestella cloud is negative on PCA-3 (to the bottom on Figure 16.2, Table 9), which precludes especially short fenestrules.

In summary, for the five genera in which size of externally visible characteristics were mentioned when the genera were established, the mentioned characteristics have a small role (if any role at all) in determining placement of the genera in the nine-dimensional morphospace generated in this study.

Position Distribution of Each genus Within Defined Morphospace

Density and Size of Multispecies Coherent Clouds (PCA-1 vs. PCA-2 vs. PCA-3). Observations of Rectifenestella (seven nominal species), Hemitrypa (nine nominal species), Archimedes (five nominal species including the type species, A. wortheni Hall, 1857), Cubifenestella (three nominal species including the type species, C. rudis (Ulrich)), Minilya (two nominal species), and Spinofenestella (four nominal species) each formed their own single, coherent cloud.

The clouds of observations for each genus, while coherent, have varying size and density. The most observation-rich cloud, Rectifenestella (243), is dense and occupies a relatively small portion of morphospace located on the near zero negative portion of the PCA-1 axis (Figure 11.1-11.2). In contrast, the Cubifenestella cloud is represented by only 72 observations, which are very broadly distributed across the positive region of PCA-1 and equally broadly distributed though approximately centered on zero on both PCA-2 and PCA-3 (Figure 14.5-14.6).

The density and extent of the clouds for other multispecies coherent clouds (Archimedes, Hemitrypa, Minilya, and Spinofenestella) each fall between the end members. The cloud of OTUs for Archimedes largely overlaps that of Fenestella (Figure 14.3-14.4 and Figure 13.3). The cloud of OTUs for Hemitrypa is similar to that of Rectifenestella (Figure 14.1-14.2 and Figure 13.1-13.2), however, the Hemitrypa cloud has an asymmetrical distribution as the overall size of characters become larger. That is, larger PCA-1 values of Hemitrypa (toward the right on PCA-1, Figure 14.1) expand into the positive region of PCA-3 and into the negative region of PCA-2 (upper left quadrant, Figure 14.2). Hemitrypa is one of the few examples among the genera analyzed that does not correspond to a single geometric shape such as a conical cloud or truncated ellipsoid, i.e., apparently at least two somewhat independent controls affect the distribution of Hemitrypa OTUs in this morphospace.

The OTUs for Minilya represent only two nominal species, but they form a coherent cloud, expanding into larger values for PCA-3 and PCA-2 with larger overall size (Figure 16.1-16.2). Additional specimens and/or species of Minilya would provide an excellent test to establish whether morphospace would be filled for the existing distribution or whether the cloud would expand (toward more negative PCA-3) to mimic the conical shape of the clouds of several other genera or confirm an elongated cloud.

The cloud of OTUs for Spinofenestella (and Alternifenestella in part) overlaps parts of the clouds for Fenestella and Archimedes, but forms a much smaller cloud, largely restricted to negative PCA-3 values and centered on more positive values for PCA-2 (c.f. Figure 16.1-16.2 and Figure 14.1-14.4).

Regardless of the density of packing of observations for these genera represented by two or more species, each occupies a well-delimited portion of the morphospace that partially overlaps the morphospace of other genera but which has its own unique pattern and centroid. In other words, these genera have morphologies that can be used to discriminate them more or less confidently from at least a subset of the other genera. Although they were defined exclusively or largely on characteristics other than size, they also can be described by a separate set of external, size-determined characteristics.

Multispecies Discontinuous Clouds. Observations of Laxifenestella (six nominal species), Fenestella (five nominal species), Apertostella (three nominal species including the type species A. crassata Snyder 1991), and Alternifenestella (five nominal species) are organized into two separate clouds for each of the genera. In each case one of the clouds is formed by a single nominal species, and all the other nominal species (ranging from two to five) comprise the other cloud. There are three possible reasons for the discontinuity in distribution of observations assigned to a single genus: 1) there are other species within the genus that if included in the study would have bridged the gap, 2) size characteristics of a species are not phylogenetically constrained and do not necessarily form a continuum, and 3) the single isolated species was incorrectly assigned in the study from which the data were taken for the present study.

The isolated species in Laxifenestella was originally named Fenestella serratula Ulrich, 1890, reassigned to Laxifenestella, and measured in Snyder (1991). F. serratula has strong morphological affinity with Laxifenestella sarytshevae (Shul'ga-Nesterenko 1951), the type species of Laxifenestella, having all the qualitative characters included in the original generic diagnosis given by Morozova (1974) and visible in the type specimens of F. sarytshevae (illustrated in Morozova 2001). Other than the isolated position in morphospace, there is no apparent reason to question determination of F. serratula as a species of Laxifenestella. Excluding the discrete subcluster noted above, the majority of Laxifenestella observations form a truncated ellipsoid (positive of 0.0 on PCA-1) near a boundary with Rectifenestella (Figure 11.3), reflecting an overall larger size for measured characters.

A large cloud, comprised of observations derived from four Fenestella species is elongated on PCA-3, centered on PCA-1, and located on the negative portion of the PCA-2 axis (Figure 11.5-11.6). Fenestella sp. 1 of Ernst and Schroeder (2007) plots separately, forming an isolated, small cloud between -1.0 and -2.0 on PCA-1 (Figure 11.5).

Fenestella sp. 1 (Ernst and Schroeder 2007) appears to have greater affinity with Rectifenestella than with Fenestella s.s. It has elongate pentagonal zooecial chamber cross-sectional shape in the endozone and moderately large keel nodes (Ernst and Schroeder 2007, p. 218, figure 6E-H), which are characteristic of Rectifenestella but different from the slightly elongate rectangular zooecial cross-sectional chamber shape in the endozone and small or absent keel nodes characteristic of Fenestella subantiqua (d'Orbigny 1850), the type species of Fenestella, as described by Snell (2004). If Fenestella sp. 1 (Ernst and Schroeder 2007) were reassigned to Rectifenestella on the basis of its zooecial chamber shape and the presence of conspicuous nodes, it would form a contiguous part of the Rectifenestella observations on PCA-1, PCA-2, and PCA-3 (Figure 11.5-11.6).

Apertostella forms two disjoint but independently coherent clouds. One subcluster consists of two species, including the type species, A. crassata Snyder, 1991. This subcluster occupies morphospace shared by Fenestella (PCA-1 values from -1.0 to +1.0 and PCA-3 values from -1.0 to +1.5) and is restricted to negative values on PCA-2 (Figure 14.5-14.6). This group of observations bridges the gap on PCA-1 between Rectifenestella and Cubifenestella (Figure 14.5), is centered on PCA-3, and is centered farther left on PCA-2 than either Rectifenestella or Cubifenestella (Figure 14.6).

The second subcluster of Apertostella, composed solely of the species Apertostella venusta Snyder, 1991, occupies morphospace independent of any other genus in this study. It is adjacent to Cubifenestella, overlapping it but extending even farther into positive values on PCA-1 (Figure 14.5). However, it has more negative PCA-3 values than the Cubifenestella observations that it overlaps on PCA-1.

Morozova (2001) considered Cubifenestella and Apertostella to be synonyms of Rectifenestella. She argued that the proportions of length:width:height ratios of endozonal chambers that Snyder used in part to characterize the genera were encompassed in the concept of Rectifenestella. She also mistakenly interpreted a small, strategically placed and oriented dog-tooth crystal of calcite as a superior hemiseptum in an illustration (Snyder 1991, Pl. 30, figure 5) of the type species of Cubifenestella; superior hemisepta are characteristic for Rectifenestella but according to Snyder (1991) are absent in Cubifenestella and Apertostella. The distribution of observations of the three putative genera on PCA-1, PCA-2, and PCA-3 do form a continuum. One subcluster of Apertostella, comprised in part by its type species, bridges the gap between Rectifenestella and Cubifenestella, and the other subcluster of Apertostella, together with the Cubifenestella observations, extends the cone formed in negative PCA-1 space well into positive PCA-1 space and expands it in PCA-2 and PCA-3 space (Figure 14.5-14.6).

The distributions in morphospace represented in Figure 14.5-14.6 suggest a closer affinity of Apertostella venusta with species that comprise Cubifenestella than with the other two species originally included in Apertostella. If the absence of a superior hemiseptum is considered sufficient to exclude the subcluster of Apertostella observations centered at the crossing of the PCA-1 and PCA-3 axes from Rectifenestella, then the two species that comprise that subcluster need to be closely examined to determine if they warrant discrimination from Cubifenestella. The discrimination of these three genera from one another requires further study.

Alternifenestella observations group into two subclusters, one of which is comprised of observations from four species and corresponds with the placement of Spinofenestella, i.e., slightly negative on PCA-1 and PCA-3, and slightly positive on PCA-2 (Figure 16.1-16.2).

The original diagnoses of Alternifenestella and Spinofenestella included two similarities (presence of a keel, single row of keel nodes). Termier and Termier (1971) indicated regular, short fenestrules and occurrence of an aperture at the base of each dissepiment for Spinofenestella, features not mentioned in Morozovas characterization of the genus three years later (1974). The only differences that Morozova (1974) indicated for the two genera were branch width (wide in Spinofenestella, thin in Alternifenestella) and basal shape of endozonal zooecial chambers (triangular in Spinofenestella, triangular-trapeziform in Alternifenestella). The co-occurrence of Spinofenestella and the second subcluster of Alternifenestella in morphospace as defined by PCA-1, PCA-2, and PCA-3 invite reevaluation of these genera.

Comparison of original type specimens of Alternifenestella and Spinofenestella reveals no differences in qualitative characters and even minimal differences in quantitative characteristics (Figure 17). In type specimens of both species, branch width is essentially equal and zooecial cross-sectional shape in deep tangential sections is triangular. The apparent foreshortening of endozonal chambers of Fenestella donaica minor (Figure 17.1) relative to those of F. spinulosa (Figure 17.2) is due at least in part to the greater distal downward inclination of the thin section of F. donaica minor relative to that of F. spinulosa. The contention by Morozova (1974) that branches are relatively broader in Spinofenestella and that deep cross sections of zooecial chambers of Alternifenestella encompass trapezoidal shapes lacking in Spinofenestella must have been based on other species included in her concept of the two genera.

Based on such close correspondence in morphology as seen in thin sections of the type species, supported by correspondence of externally determined morphospace (Figure 16.1-16.2) as expressed in several non-type species assigned to the two genera analyzed in the current study, we consider Alternifenestella and Spinofenestella to be subjective synonyms. Termier and Termier (1971, p. 42) named both genera on the same page. There is no objective criterion for determining that one of the genera should have priority over the other, but we prefer to give Spinofenestella priority because although both genera were given a legal basis for naming (i.e., designation of a type species) only Spinofenestella was provided with a diagnosis by Termier and Termier (1971).

The second subcluster comprised of A. bifida (Eichwald) in Nakrem (1994) is positive on PCA-1 and extends into the positive region of PCA-3 (Figure16.1). These observations are from a single specimen assigned to Alternifenestella bifida (Eichwald 1860) in Nakrem (1994). The positive distribution of this subcluster of Alternifenestella indicates overall large sizes, with a positive range on PCA-1 (Figure 16.1) similar to that of Minilya, which is characterized by a double row of keel nodes. Observations of bifida are scattered but centered on PCA-3, and their slightly positive placement on PCA-2 indicates some combination of small apertures, narrow dissepiments, and/or distantly spaced apertures along branches. Nakrem's specimen has a single row of keel nodes, and zooecial chambers have triangular cross sections in deep tangential sections. Despite its more robust features (Figure 16.1) as measured externally, it therefore appears to belong within the generic concept Spinofenestella based on chamber shape and the single row of keel nodes rather than to Minilya.

Genera Represented by Single Species. The central position of observations in the single species of Exfenestella and Wjatkella (centroids of their respective clouds plotting near zero on PCA-1, PCA-2, and PCA-3; Figure 16.3-16.4, Table 10) gives no indication of distinctive size of their exterior features. Although additional species would undoubtedly give some idea of whether these species trend out into some oriented direction within the morphospace, the position of the two species indicates that at least their respective genera include species that have no distinctive size characteristics of any kind. In contrast, the distinctly positive PCA-1 position of observations from the single species of Flexifenestella (Figure 16.3) indicates overall large size, and their negative position on PCA-3 (Figure 16.3-16.4) is notable. Whatever other characters may be involved in the negative placement on PCA-3, the negative PCA-3 loading of large fenestrules is sure to be involved, because the species included here and Flexifenestella overall is characterized by broad branches that are highly sinuous (Morozova 1974), the sinuosity generating unusually large fenestrules for a fenestellid taxon.

Two species in the study are not biserial fenestellids and were intended to serve as reference points. One, Lyroporella, has branches organized like those of Polyporella, with two rows of zooecia distal to bifurcations but proliferating to three rows at an appreciable distance preceding the next bifurcation. The second, Anastomopora anaphora, typically has multiserial branches but in some species – such as the one included here – branch segments begin with only two rows of zooecia after a bifurcation before interpolating additional rows of zooecia. Even though having large sections of branches with sufficient width to accommodate three rows of zooecia rather than two, Lyroporella surprisingly is characterized overall by small exterior measurements, occurring near the edge of the composite cloud on the negative end of PCA-1 (Figure 14.1). The slightly positive position on PCA-3 (Figure 14.1-14.2) and slightly negative position on PCA-2 (Figure 14.2) are almost certainly strongly influenced by small fenestrules and wide dissepiments. Lyre shaped bryozoans in general tend to have closely spaced branches and broad dissepiments, resulting in small fenestrules that comprise an anomalously small ( 15%) proportion of the fenestrate sheet (McKinney et al. 1993). These structural features, which apparently are related to the overall colony morphology and its relation to the highly kinetic environment in which the colonies lived (McKinney 1977) may be largely responsible for the location of the Lyroporella cloud within the densely packed conical (negative PCA-1) end of the composite cloud.

The second control, Anastomopora anaphora, would be predicted to plot anomalously high on PCA-1, because the species typically has three to four rows of zooecia along branches (McColloch et al. 1994), which would be predicted to require greater branch width than fenestellids with two rows. Polyporid fenestrates with three to four rows of zooecia typically have larger branch widths and disproportionately larger branch spacing (i.e., relatively wider fenestrules) than do fenestellids (Starcher and McGhee 2002). Indeed, Anastomopora's positive position on PCA-1 (Figure 14.3) reflects larger overall sizes, but it falls well below the high end of the PCA-1 range of Laxifenestella (Figure 11.3), Cubifenestella (Figure 14.5), Apertostella (Figure 14.5), and Minilya (Figure 16.1). Anastomopora observations are almost centered on zero for PCA-3, but on PCA-2 they extend from the center of the composite cloud of all fenestellid observations to include the highest negative value in the study (Figure 16.4), minimally beyond the most negative values for Laxifenestella (Figure 11.4) and Apertostella (Figure 14.6). Negative values for PCA-2 are due to some combination of large apertures, wide dissepiments, and close spacing of apertures along branches (Figure 4), more than to large branch width and large fenestrules.

Lyroporella and Anastomopora are currently assigned to two separate families differentiated from the Fenestellidae, thus they have appreciable morphological differences from each other and the rest of the genera included. Nevertheless, they plot within the borders of the composite cloud of fenestellid observations. Their inclusion within this morphospace supports the interpretation that the multispecies fenestellid genera represent appreciable differences in gross morphology among genera.

Relationships Among Morphometric Characters

It is not surprising that a strong degree of covariance exists among morphometric characters (Table 8). However, the combination of characters and their relative significance is noteworthy.

These relationships among covarying characters and relationships expressed in PCA loading coefficients (Table 4) invite interpretation of their biological significance, if any. However, the relationships are complex and in places contradictory among methods of analysis. Future work in this area will require partitioning of parameters of both size and shape.

1. The first complex (FL-FW-ALB-NS) involves a relationship between fenestrule size and spacing of apertures laterally across branch (but not so much along branch) (Table 8, red highlight). Fenestrule size (meshwork openness vs. compactness) is typically associated with water currents and feeding. Characters of aperture spacing are associated with lophophore and polypide size. Thus, the inclusion of node spacing in this character complex is somewat puzzling.

2. The second complex of covaring characters (AAB-ALB-ND) involves spacing of apertures (both along and lateral to branch) and node diameter. Again, features of apertural spacing are typically associated with lophophore and polypide size, which makes the inclusion of node spacing in this character complex enigmatic (Table 8, purple).

3. The third complex of covaring characters (BW-DW) is associated with the robustness of the meshwork (branch and dissepiment width) (Table 8, blue).

4. The fourth complex of covaring characters (AD-ND) is associated with the size of both apertures and nodes (Table 8, green). The biological significance of this relationship is unclear.

5. The combination of aperture size and the spacing of apertures along branch (AD-AAB) displays the least (though still marginally significant at p = 0.02) covariance of all character pairs (Table 8, orange). This weak relationship may reflect an independent relationship between lophophore diameter and length of space required to accommodate polypides when retracted, perhaps reflecting independence of lophophore size and whether or not the polypide is doubled onto itself or does or does not include tentacle wrinkling when retracted.