Article: Reconstruction, composition and homology of conodont skeletons: a response to Agematsu et al.

Agematsu et al. (2018) commented on our recent paper about testing hypotheses of element loss and apparatus stability in the apparatus composition of complex conodonts (Zhang et al. 2017). They take issue not with our approach, but with our specific hypothesis concerning the skeletal apparatus of Hindeodus parvus. We proposed that, in marked contrast to the remarkable anatomical conservatism exhibited by ozarkodinid conodonts, which seem not to vary their 15 element ‘dental formula’ over a period in excess of 250 million years, H. parvus had only 13 elements, lacking 2 elements from the posterior P domain of the apparatus.

Our paper presents an hypothesis of homology for the skeletal elements of H. parvus that Agematsu et al. (2018) argue is incorrrect, based on three lines of reasoning:

1. Previous hypotheses concerning which of the elements found in collections of disarticulated skeletal material belong together as components of the multielement taxon Hindeodus.
2. Evidence of stability in the inferred homology of those elements, implied by how previous authors have applied conodont anatomical notation.
3. The new material and interpretations of Agematsu et al. (2017).

We take issue with the first two parts of their argument. They discuss apparatus reconstruction and composition as if these things were the same as hypotheses of homology, but there are important distinctions, and we take this opportunity to clarify the differences between reconstructing the multielement composition of a conodont taxon, inferring hypotheses of homology for elements within such a reconstructed taxon, and the evidence required to provide a definitive test of such hypotheses.

Reconstruction and proxies for homology

Conodonts are generally found as individual, morphologically discrete microfossils known as elements, sometimes many thousands in a single sample. For the first few decades of their study, each distinct type of element was treated as if it constituted a taxon. Elements with similar shapes were combined into genera, and a form taxonomy developed that looked superficially like the application of a biological system. From the 1930s, however, evidence began to accumulate that conodont microfossils were in fact components of a skeleton that contained a number of different types of elements. Elements that had been, on the basis of their shape, assigned to different species and genera in fact occurred together within the same skeleton. This clearly had major implications for the single‐element approach to naming and defining taxa, but it took several decades of work, starting in earnest in the 1960s to develop a biologically meaningful taxonomy (for discussion see e.g. Sweet 1981, 1988).

This slow revolution required two distinct things, which are essentially the two arguments put forward by Agematsu et al. (2018): first, evidence and methods by which to recognize and put together the elements in collections of isolated skeletal parts that came from the same taxon, and second, the development of hypotheses of homology. Although multielement reconstruction is possible without hypotheses of homology, the ability to recognize, compare and apply consistent terminology to homologous parts of the skeletons of different taxa is a fundamental prerequisite of developing a biologically meaningful taxonomy. Without hypotheses of homology, comparative anatomy is impossible, and discussion of phylogenetic and evolutionary patterns is meaningless.

In many ways it is desirable, and more intellectually robust, to keep separate the process of reconstructing conodont taxa and the process of proposing hypotheses of homology, but the distinction became blurred. Increasing numbers of reconstructed conodont species revealed that many taxa contain the same basic types of elements, conforming to a small number of broad morphological categories, described either using descriptive terms such as pectiniform, ramiform, bipennate and angulate, or terms derived from the pre‐multielement form taxa to which they would have been assigned, such as ozarkodiniform and spathognathodiform (Agematsu et al. (2018, table 1) provides examples of both). It also became clear that these element types were recognizable in fossils preserving articulated skeletal remains. Perhaps inevitably, because the number of taxa preserved as articulated skeletons was small compared to the number of taxa being reconstructed, these two sets of evidence were mixed to provide a morphological search image of what to look for when reconstructing conodont taxa, with the consequence that as ideas of homology developed, they were closely linked to the morphological criteria for recognizing the elements one would expect to find in a conodont taxon (e.g. Sweet & Schönlaub 1975; Sweet 1981, 1988).

Topological homology in conodonts and its application to Hindeodus parvus

As more fossils preserving articulated conodont remains were discovered, and more effort was expended in interpreting them (e.g. Norby 1976; Aldridge et al. 1987; Purnell 1993; Purnell & Donoghue 1997), the confusion created by not separating the meaning and definition of homology from the proxy criteria used to infer it started to create issues of communication, and the need (anticipated by Sweet 1981) for a more strictly defined concept of topological homology and means of differentiating this from less secure morphologically inferred hypotheses, became clear (Purnell 1993; Purnell et al. 2000). The widely used anatomical notation of Purnell et al. (2000) proposed an operational concept of homology as an hypothesis of similarity based on topological relations and which contains potential phylogenetic information (see Rieppel 1994; Purnell et al. 2000 for discussion). Topology refers to the numbers of, and the relative spatial relationships between, recognizable anatomical units (i.e. conodont elements in a skeletal apparatus). Purnell et al. (2000) were explicit that ‘similarity’ did not mean morphological similarity of elements. This is the anatomical notation and underpinning homology concept employed by Agematsu et al. (2014, 2017, 2018) and by Zhang et al. (2017). And because this system is based on topology, the evidence for numbers of elements and the relative spatial relationships between them in articulated skeletons provides the ultimate test of hypotheses of homology that have been inferred on the basis of element morphology and proposed evolutionary relationships.

It also follows from this that these inferred hypotheses cannot test or falsify the evidence of homology provided by articulated skeletons, and this is what we take issue with in Agematsu et al. (2018). In this context, it does not matter how many authors over multiple decades have consistently adhered to a concept of Hindeodus in which elements of a particular morphology are considered to be the ‘same’ element, and Agematsu et al. (2018) are wrong to state that this ‘strongly suggests that the angulate elements identified in the bedding plane assemblages of Hindeodus parvus by Agematsu et al. (2014) are P₂ elements’. It does not matter because, before the discovery of articulated skeletons (Agematsu et al. 2014), all hypotheses of homology in Hindeodus were inferred on the basis of morphology and are falsifiable by the direct evidence of topology contained in the articulated skeletons. The principal difference between Agematsu et al. (2017, 2018) and Zhang et al. (2017) concerns the presence of a P₂ element, but the arguments of Agematsu et al. (2018) that draw on previous reconstructions of the multielement taxon Hindeodus and stability in the inferred homology of those elements (their table 1, fig. 1) are simply not relevant because they do not provide a test of the hypothesis.

Agematsu et al. (2018, table 1) confuses the issue on multiple levels. Because the table is organized into columns labelled with the anatomical notation of Purnell et al. (2000), Agematsu et al. (2018) imply that the authors of the reconstructions listed in rows each expressed an hypothesis of homology consistent with the topological concepts of Purnell et al. (2000), but this is incorrect. In many cases, the authors applied morphologically based descriptive terms to their reconstructions, not terms designed to imply homology between taxa; the apparent stability in these reconstructions simply reflects the fact that different authors were consistently able to find the ‘same’ element with angulate/ozarkodiniform/ozarkodinan morphology (see Sweet 1981 for discussion). Inferring that within Hindeodus these elements were homologous with one another, because of their close morphological similarity, is a reasonable hypothesis, and in many cases, the authors’ intentions probably extended no further than this. It is a different hypothesis to suggest that these elements were homologous with all other elements identified as occupants of P₂ positions in all other taxa. Yet, this is what table 1 implies, without recognizing the distinction between the relatively strong evidence that morphology can provide for homology within a genus compared to its weaker power and lesser reliability as evidence of homology between taxa above genus level. As we note above, prior to the discovery of articulated skeletons of Hindeodus, hypotheses of homology between elements of Hindeodus and those of other taxa were inferred on the basis of morphological criteria, and this applies to all the reconstructions in Agematsu et al. (2018, table 1) except Agematsu et al. (2014) and Zhang et al. (2017). Irrespective of how reliable these morphological criteria seem, they carry less weight than topological evidence: the direct evidence can test and overturn hypotheses inferred from morphology, but inferred hypotheses cannot overturn the evidence of topology.

The apparatus of H. parvus, and the implications for stability and constraint

The important differences between the interpretations of Agematsu et al. (2014), Zhang et al. (2017) and Agematsu et al. (2017) are differences in hypotheses of homology (Fig. 1). As Agematsu et al. (2017) noted, the articulated skeletal material described by Agematsu et al. (2014) was insufficient to determine whether the S₁ locations of the apparatus were occupied. Their specimens provided clear evidence for only 13 elements (rather than the typical 15), and in finding evidence in new fused material for a complete S array, Zhang et al. (2017) proposed that it was the P₂ locations, and not S₁, that were unoccupied. The new material and interpretations of Agematsu et al. (2017) provide a definitive test of these alternative hypotheses of element homology for H. parvus. They clearly support the hypothesis of Zhang et al. (2017) for a full suite of S₁–S₄ elements, but also find evidence for an additional pair of elements which, in the best preserved of the specimens figured by Agematsu et al. (2014, 2017; i.e. those exhibiting the least disruption of original element juxtaposition), are located near to the S array. The most parsimonious interpretation is that put forward by Agematsu et al. (2017): these are P₂ elements (the alternative would require duplication of an S element pair and loss of the P₂ elements). However, the apparent location of P₂ elements, closer to the S array than they are in other ozarkodinid taxa known from articulated skeletons, raises some interesting questions.

Figure 1 Open in figure viewerPowerPoint The different hypotheses of topological element homology proposed by Agematsu et al. (2014), Zhang et al. (2017) and Agematsu et al. (2017): diagrammatic representations of how these authors interpreted the articulated skeletal material preserved on bedding planes, not plans of apparatus structure (rostral is towards top of page). Element morphology is simplified. Both Agematsu et al. (2014) and Zhang et al. (2017) proposed hypotheses of homology for a 13 element apparatus, but they differed in the interpretation of the elements located on what is the medial side of the S array as drawn. Agematsu et al. (2017) found evidence for 15 elements; this is typical for ozarkodinids, and their figured specimens support their hypotheses of homology, but the apparent location of the P₂ elements is unusual. CaptionThe different hypotheses of topological element homology proposed by Agematsu et al. (), Zhang et al. () and Agematsu et al. (): diagrammatic representations of how these authors interpreted the articulated skeletal material preserved on bedding planes, not plans of apparatus structure (rostral is towards top of page). Element morphology is simplified. Both Agematsu et al. () and Zhang et al. () proposed hypotheses of homology for a 13 element apparatus, but they differed in the interpretation of the elements located on what is the medial side of the S array as drawn. Agematsu et al. () found evidence for 15 elements; this is typical for ozarkodinids, and their figured specimens support their hypotheses of homology, but the apparent location of the P2 elements is unusual.

The evidence presented by Agematsu et al. (2017) is not conclusive (well‐preserved specimens exhibiting lateral collapse orientations would provide a more stringent test) but it suggests that in H. parvus, the P₂ elements are more closely associated with the S array than in other conodont taxa for which we have direct evidence of the 3D architecture of the skeleton. This implies a shift in function to one more associated with prey prehension or perhaps moving food from the rostral S–M array towards the P₁ elements for processing. Interestingly, the P₂ elements are morphologically somewhat different from more typical examples of ozarkodinid P₂ elements. They have relatively larger cusps, with the flexure and denticulation of the process (e.g. Agematsu et al. 2018, fig. 1B) to a degree reminiscent of elements that occupy S positions (albeit shorter). The location of the P₂ elements in H. parvus, and their morphology, supports our functional hypothesis (Zhang et al. 2017) for a shift in H. parvus towards foods that can be ingested with less slicing or crushing. The evidence is less clear, but the specimens of Agematsu et al. (2014, 2017) also suggest that the caudal processes of the M elements were perhaps more parallel to the S array that in other ozarkodinid taxa. Although the new evidence highlights subtle differences between H. parvus and other ozarkodinid taxa, the broader point we make (Zhang et al. 2017) remains true: the 15 element ‘dental formula’ of ozarkodinid conodonts remained stable for more than 250 million years, and this signal of remarkable functional and/or developmental constraint is worth further investigation.

Acknowledgements

MAP's views on homology in conodonts were largely developed during the tenure of Natural Environment Research Council Fellowships (GT5/F/GS/95/6 and GT5/98/4/ES) and in discussion with Phil Donoghue and Dick Aldridge. Thanks to Steve Leslie for comments on an earlier version of this paper.