Since its original description as a chordate, the Late Devonian Scaumenella mesacanthi has been interpreted alternately as a prochordate, a larval ostracoderm and an immature acanthodian. For the past 30 years, these minute specimens were generally considered as decayed acanthodians, most of them belonging to Triazeugacanthus affinis. Among the abundant material of ‘Scaumenella’, we identified a size series of 188 specimens of Triazeugacanthus based on otolith characteristics. Despite taphonomic alteration, we describe proportional growth and progressive appearance of skeletal elements through size increase. Three ontogenetic stages are identified based on squamation extent, ossification completion and allometric growth. We demonstrate that what has been interpreted previously as various degrees of decomposition corresponds to ontogenetic changes.
In 1935, Graham-Smith  described Scaumenella mesacanthi based on 560 specimens from the Upper Devonian (380 Ma) Escuminac Formation (Miguasha, Canada) as a ‘chordate, and probably a vertebrate’ because of the presence of a head, an abdomen with branchial arches, a notochordal or vertebral region and a hypocercal tail. Subsequently, Scaumenella was frequently considered when dealing with vertebrate origin [2–4]. Lehman  and Piveteau et al.  reinterpreted Scaumenella as a prochordate closer to cephalochordates than ascidians, whereas Tarlo  proposed that Scaumenella was most similar to an ammocoete (his ‘larval ostracoderm’).
In the early 1980s, the Quebec ichthyologist Vianney Legendre made detailed observations on more than 900 specimens of Scaumenella from the Musée d'Histoire Naturelle de Miguasha (MHNM). Among other things, Legendre identified nasal bones on some of the Scaumenella diagnostic to those of the acanthodian Triazeugacanthus affinis (figure 1), which he used to associate these two taxa. Therefore, Béland & Arsenault  upended Graham-Smith's  reconstruction of Scaumenella and proposed that these specimens correspond to partially decomposed Triazeugacanthus (fig. 11 in ). They referred to ‘scaumenellization’, the taphonomical alteration leading to an ultimate state of degradation being the ‘scaumenelle’. This interpretation was based on the so-called progressive disappearance of fin spines, scales, cranial and girdle bony elements  and used subsequently as a classical example of decaying effect on the anatomical and taxonomic interpretation of a vertebrate . Since then, various fossil fish specimens were said to be scaumenellized [7–9].
Besides scaumenellization, Béland & Arsenault  mentioned that some of the small Scaumenella could correspond to immature Triazeugacanthus. Recently, Cloutier et al.  and Cloutier  agreed with this interpretation based on shape morphometry and squamation pattern. However, this ontogenetic interpretation has to be quantitatively tested on a large number of specimens.
Our main objective was to describe a size series of Scaumenella–Triazeugacanthus in order to recognize if their primary source of variation corresponds to ontogenetic changes or taphonomic alteration. In favour of the ontogenetic changes, we expect to observe, despite taphonomic alterations, (i) linear relationships among the size of individual skeletal structures and body size, and (ii) progressive appearance of skeletal elements correlated with body size. Concerning taphonomic alteration, we expect (i) non-proportional relationships among the size of individual skeletal structures and body size, and (ii) progressive disappearance of skeletal structures not correlated to body size. The large number of available fossils provided the opportunity to quantify the variation along the size series.
2. Material and methods
Triazeugacanthus specimens come from the Middle Frasnian Escuminac Formation (Miguasha, Canada), which is a UNESCO World Heritage site . Triazeugacanthus and the so-called ‘scaumenellized’ Triazeugacanthus are housed in the MHNM (1620 specimens) and NMS (National Museums of Scotland) (2015 specimens) collections. A subsample of 188 specimens was analysed based on (i) the presence of diagnostic Triazeugacanthus features (electronic supplementary material, figure S1), (ii) the integrity of the specimens (e.g. specimens without preparation artefacts) and (iii) the presence of representatives along an optimized size range. Specimens were observed under water immersion (Leica MZ9.5), drawn using a camera lucida and photographed (Nikon D300).
Morphotypes were recognized based on squamation: morphotype 1 (morph-1) shows no body scales, morphotype 2 (morph-2) displays a partial body squamation and morphotype 3 (morph-3) shows complete body squamation (figure 1). Continuous [length of skeletal elements and distances among them (electronic supplementary material, figure S2)] and discrete data [presence/absence of skeletal structures (electronic supplementary material, figure S3)] were collected. Linear regressions between log10-transformed measurements and total length (TL) were calculated for individual morphotypes and for combined morphotypes. The cumulative number of 23 skeletal elements (electronic supplementary material, figures S3 and S4) was calculated in relation to TL. The squamation extent (maximum length of scaled area/TL × 100; the maximum length of scaled area is used as a proxy for the scaled area) is given in relation to TL. A von Bertalanffy growth model is used on squamation data.
Three principal component analyses (PCA) on variance–covariance matrices of five log10-transformed measurements were performed for morph-2, morph-3 and the combined dataset (electronic supplementary material, table S1). Morph-1 was excluded from the PCA because most landmarks are absent. Multivariate normality is accepted (Mardia's test: p < 0.05). Multivariate coefficients of allometry were calculated from PC1 loadings  for morph-2 and morph-3. PCA, multivariate normality and allometric coefficients were calculated with PAST 2.17, whereas all remaining statistics were performed with R v. 3.0.2.
The size series includes 188 specimens of Triazeugacanthus ranging from 4.51 to 52.72 mm in TL. There is a significant linear relationship between the length and height of the eye lens (n = 142, R² = 0.79; F = 110.4, p < 2.2 × 10−16; figure 2a), suggesting a proportional size change. The length of the pectoral spine (figure 2b) shows a significant linear relationship with TL (n = 126, R² = 0.77; F = 396.1, p < 2.2 × 10−16).
In the smallest specimen (4.51 mm TL; figure 1a,b), eye lenses, otoliths and traces of the notochord are present. Pectoral, pelvic, anal and dorsal spines are present in slightly larger specimens (13.58 mm TL; figure 1c,d). Squamation is recorded in larger specimens (19.2 mm TL; figure 1e,f) at the level of the dorsal fin and extends towards the skull and the caudal fin in larger specimens. The squamation is total at around 25 mm TL (figures 1d and 2c) and palatoquadrates, Meckel's cartilages, circumorbital bones, branchiostegal rays, head scales and the hypochordal lobe of caudal fin are recorded. The cumulative number of skeletal elements and the squamation extent are significantly correlated with TL (rs = 0.99, p < 2.2 × 10−16 and rs = 0.67, p < 2.2 × 10−16, respectively; figure 2c). All skeletal structures are present at around 40 mm TL (figure 2c).
In our sample, 81.5% of the specimens show signs of taphonomic alteration. The most recurrent signs are: a dorsal curvature of the body (owing to post-mortem tetany ; 73%), a rupture of the abdominal cavity (owing to bacterial activity or fermentation ; 33%) and a scattering of scales (owing to decay; 33%). Most detached scales are localized ventrally in the mid-trunk region and are associated with the rupture of the abdominal cavity; these two conditions are not observable in morph-1 because scales are absent. Therefore, loss of scales does not affect specimen TL. Specimens showing the greatest deviation from the means in terms of squamation and cumulative number of skeletal elements display evidence of abdominal rupture (electronic supplementary material, figure S4).
The size of each morphotype increases progressively although there are some overlaps among morphotypes (figure 2d). Comparison among morphotypes shows significant differences in linear regression slopes for both eye lenses (morph-1: 0.81; morph-2: 0.77; morph-3: 0.91; figure 2a) and pectoral spines (morph-1: 0.95; morph-2: 1.1; morph-3: 0.74; figure 2b). These changes are suggestive of allometry. Multivariate allometric coefficients on PC1 of the separated PCAs (morph-2 PCA and morph-3 PCA) show a trend (allometric coefficients are non-significant, but the sample size is small) towards allometry in the distance between the anal and dorsal fins (morph-2: 4.47; morph-3: 2.16) with respect to the remaining four variables. Morph-2 specimens have a greater propensity to allometric changes. Morph-2 and morph-3 specimens cluster into two groups in Triazeugacanthus combined PCA (electronic supplementary material, figure S5) corroborating their morphotypic assignation.
We have shown clearly that (i) the size of individual anatomical structures, (ii) the number of skeletal elements and (iii) the squamation extent increase with TL. Based on the squamation extent and pattern, ossification completion and allometry, the three morphotypes are best interpreted as three ontogenetic stages making this growth series one of the best documented early gnathostome fossilized ontogenies [10,11]. Morph-1 showing no body squamation corresponds to a larval stage [10,11,16]; morph-2 starts with the initiation of squamation and shows allometry, which is characteristic of a juvenile stage [10,11,17]; and the completion of squamation and ossification characterising morph-3 corresponds to an adult stage . The evidence for ontogeny in Triazeugacanthus that we have presented does not rule out taphonomic alteration but shows that the main trend of variation is explained by ontogenetic changes.
In the smallest Triazeugacanthus, we recorded the presence of eye lenses (‘optic plates’ , ‘orbits’ ), otoliths and notochordal elements (‘rotten scaly skin’ , ‘elements of the vertebral axis’ ). If only taphonomic alterations occurred in Triazeugacanthus, the smallest specimens would represent the most rotten specimens; accordingly some of these observed structures are among the last elements to decay or to be lost during decomposition of a vertebrate in an aquatic environment [18,19]. These elements are also among the first anatomical structures to form in early ontogeny that have a great potential to be fossilized. In numerous living osteichthyans, post-hatching larvae show distinctive features including a head with limited chondrified elements, large eyes, a notochordal axial support, a finfold without differentiated fins, scaleless skin, and simplified muscular and digestive systems. The absence of finfold and yolk sac are most likely owing to taphonomic loss of soft tissues. There is no indication that the length of an individual would reduce slowly during decomposition other than by the losses of the anteriormost and posteriormost elements . What was considered as the progressive disappearance of fin spines, scales, cranial and girdle bony elements corresponds in fact to the sequential appearance of these elements, primarily as a sequence of skeletal formation. The shape of the cumulative curve of elements is similar to ‘maturity curves’ or ‘ontogenetic trajectories’ in living  and extinct fishes . The postero-anterior direction of squamation development in Triazeugacanthus is also congruent with that observed for other acanthodians and some actinopterygians [11,21]; it would be unlikely for scales or patches of scales to be lost in a non-random pattern (detached scales were found in the abdominal region as a result of decay).
Thus, what has been interpreted specifically as evidence for the decomposition of Triazeugacanthus [5–8] is reinterpreted as evidence of ontogenetic changes. Interpretation of qualitative decay in fossils would benefit from being combined with quantitative analysis. Some fossil vertebrate taxa, such as the Middle Devonian Achanarella , should be reanalysed with such an approach because of their anatomical and taphonomic similarities with ‘Scaumenella’.
Data are available as the electronic supplementary material.
R.C. conceived the project. M.C. realized observations and data analyses and produced figures and the electronic supplementary material with input from R.C. and J.-Y.S. M.C. wrote the first version, all authors contributed to the final version.
R.C. was supported by NSERC 238612, by QCBS, and Research Chair in Paleontology and Evolutionary Biology (UQAR)
We thank O. Matton, F. Charest and J. Kerr (MHNM) and S. Walsh (NMS) for access to the collections, and Z. Johanson and two anonymous referees for providing constructive comments.
- Received November 12, 2014.
- Accepted January 20, 2015.
- © 2015 The Author(s) Published by the Royal Society. All rights reserved.