[Special Papers in Palaeontology, 73, 2005, pp. 7–25] BETWEEN DEATH AND DATA: BIASES IN INTERPRETATION OF THE FOSSIL RECORD OF CONODONTS by MARK A. PURNELL* and PHILIP C. J. DONOGHUE *Department of Geology, University of Leicester, University Road, Leicester LE1 7RH, UK; e-mail: [email protected] Department of Earth Sciences, University of Bristol, Wills Memorial Building, Queens Road, Bristol BS8 1RJ, UK; e-mail: [email protected] Abstract: The fossil record of conodonts may be among and standing generic diversity. Analysis of epoch ⁄ stage-level the best of any group of organisms, but it is biased nonethe- data for the Ordovician–Devonian interval suggests that less. Pre- and syndepositional biases, including predation there is generally no correspondence between research effort and scavenging of carcasses, current activity, reworking and and generic diversity, and more research is required to bioturbation, cause loss, redistribution and breakage of ele- determine whether this indicates that sampling of the cono- ments. These biases may be exacerbated by the way in which dont record has reached a level of maturity where few genera rocks are collected and treated in the laboratory to extract remain to be discovered. One area of long-standing interest elements. As is the case for all fossils, intervals for which in potential biases and the conodont record concerns the there is no rock record cause inevitable gaps in the strati- pattern of recovery of different components of the skeleton graphic distribution of conodonts, and unpreserved environ- through time. We have found no evidence that the increas- ments lead to further impoverishment of the recorded ing abundance of P elements relative to S and M elements spatial and temporal distributions of taxa. On the other in later parts of the conodont record reflects evolutionary hand, because they are resistant to abrasion and can with- changes in the composition of the apparatus. stand considerable metamorphism conodonts can preserve Ignoring the biases and incompleteness of the conodont evidence of otherwise lost sequences or environments fossil record will inevitably lead to unnecessary errors and through reworking. misleading or erroneous conclusions. Taking biases into We have conducted a preliminary investigation into how account has the potential to enhance our understanding of the various forms of gross collecting bias arising from period conodonts and their application to geological and biological to period variation in intensity of research effort and in pre- questions of broad interest. served outcrop area have affected the conodont fossil record. At the period level, we are unable to reject the hypothesis Key words: completeness, gaps, microfossil, preservation, that sampling, in terms of research effort, is biased. We have taphonomy, vertebrate skeletons. also found evidence of a relationship between outcrop area Our purpose with this contribution is to introduce and of bias. The quality of the conodont fossil record is gener- provide an overview of an issue that underlies all palae- ally held to be among the best of any group of organisms ontological study and provides the common theme of this (Foote and Sepkoski 1999; Sweet and Donoghue 2001), collection of papers: how we interpret the fossil record. and because of their near ubiquity and ease of recovery To what extent do perceived changes in morphology, from marine strata of Late Cambrian to latest Triassic age skeletal composition, abundance and diversity through conodonts have attained an almost unrivalled reputation time reflect changes in biology and evolutionary history for biostratigraphic utility. This in turn has fuelled a and how has this primary signal been biased by post- widespread tacit assumption that because conodont bios- mortem processes? How do biases affect the ways in tratigraphy works, biases in their fossil record cannot which we use the record for evolutionary, biological and be significant (Donoghue 2001a,b; Wickstro¨m and biostratigraphic purposes? Donoghue, 2005). The record must be ‘close enough’ to Conodonts provide a particularly interesting window the original signal. through which to view the sometimes uneasy relationship Yet few would argue that post-mortem factors have not between interpretations of the fossil record and hypotheses played some role in shaping what we see, and it must ª The Palaeontological Association 7 8 SPECIAL PAPERS IN PALAEONTOLOGY, 73 therefore be true that if the record as we perceive it biases at one stage can make elements more or less suscept- reflects both biological-evolutionary patterns and post- ible to the effects of bias during subsequent stages. We mortem biases, failure to take both into account will have attempted to highlight these factors in Text-figure 1 decrease the reliability and accuracy of any interpreta- and in the discussion below. We also present a more tions. Every fossil sample lies somewhere in a spectrum detailed discussion of the potential effects of biology and that ranges from complete preservation to complete loss, bias on the relative abundance of different components of and the papers in this volume explore the fertile ground the conodont skeleton in collections of isolated elements. of interactions between bias and biology. Biases, for the purpose of this paper, are taken as fac- tors that distort or selectively filter the patterns of spatial PRE- AND SYNDEPOSTIONAL NON- and temporal distribution of fossils, as revealed through PRESERVATION AND SELECTIVE LOSS analysis of collections, causing them to deviate from a perfect record of ‘true’ biological and evolutionary his- Predation and scavenging tory. This is more than taphonomy, as we include other biasing factors such as sampling, collecting and processing Numerous examples of elements or apparatuses pre- methods, and consider how assumptions and methods, served within predators and scavengers (Scott 1969, 1973; especially phylogenetic methods, can bias interpretations. Melton and Scott 1973; Nicoll 1977; Conway Morris Any simple classification will inevitably underempha- 1990; Purnell 1993; Purnell and Donoghue 1998) or in size the complex interactions and feedbacks that occur coprolites (Higgins 1981) demonstrate that conodonts between biases, but in order to provide a framework for were food for other animals. It is possible that many ele- discussion and to be consistent with the overall structure ments in conodont collections have passed through the of this collection of papers we consider biases primarily guts of other animals, but there has been no systematic in terms of when they exert their influence, as summar- study of how this may have affected what is preserved. ized in Text-figure 1. It is important to note that not all Given the well-known effects on the enamel of gnathos- conodonts are equally susceptible to different biases, with tome teeth of passage through a gut (Fisher 1981), it is various aspects of conodont biology and evolution mak- likely that conodont elements, composed primarily of ing some species, or some types of elements within the enamel-like tissues (Donoghue 1998, 2001c), could be apparatus more likely to be lost. Similarly, different ele- partially or completely dissolved in the process of diges- ments or species may be more susceptible to bias at dif- tion by some conodont eaters. Fragmentation is also ferent stages in the transition from death to data, and possible. Species with small elements, or the more gracile pre- and syndepositional post-depositional sampling, collecting and interpretative bias bias laboratory bias bias predatation/scavenging: compaction and stratigraphic and element identifiability dissolution diagenesis/metamorphism: geographic coverage fragmentation dissolution fragmentation of rock and exposure preservation phylogenetic methods incorporation into sediment sampling strategy: current activity: lithological preferences abrasion of fossils recovery loss of host rock through sample thickness fragmentation assumptions regarding bias erosion or tectonic sample spacing winnowing & transport and completeness recycling sample size reworking, bioturbation and sample processing: time averaging rock dissolution and disaggregation sieving density and/or magnetic sediment type and rate of separation accumulation non-preservation of host sediment TEXT-FIG. 1. Biases that act to distort recovery of conodont elements. The diagram summarizes when different biases exert their influence and indicates how they interact. PURNELL AND DONOGHUE: BIASES IN INTERPRETATION OF CONODONT RECORD 9 elements in an apparatus are more likely to be lost or implications for temporal and spatial resolution of fragmented during digestion. Compaction of a coprolitic interpretations that draw directly on stratigraphic order- mass may also result in higher levels of fragmentation ing of fossils (see Barrick and Ma¨nnik 2005; Dzik 2005; because of the close juxtaposition of elements. On the Roopnarine 2005). Reworking is also likely to result in other hand, incorporation of elements into a coprolite significant element size bias. may enhance their chances and quality of preservation if Transport of elements, either all the elements of a spe- it is mineralized or lithified before significant sedimentary cies or just the more easily entrained components of the compaction. skeleton, may result in their removal to different deposi- tional settings. This can ultimately result in their com- plete loss from the record if those environments are less Current activity