C. elaphus: Late Pleistocene fossil material from Britain and Germany (Lister, 1981, 1984, 1986), and modern material including representatives of a range of modern subspecies (Natural History Museum, London).
D. dama: Modern and Late Pleistocene fossil material from Britain (Lister, 1981, 1984, 1986; Natural History Museum, London).
D. mesopotamica: Modern material (Natural History Museum, London & Hebrew University, Jerusalem); Late Pleistocene material from Tabûn, Israel (Natural History Museum, London).
M. giganteus: Fossil material from the Irish late-glacial, and from Middle to Late Pleistocene interglacials of Britain and Germany (Lister, 1994).
Cervus eldi, C. canadensis, C.nippon, Axis axis and A. porcinus, living (Natural History Museum, London; American Museum of Natural History, New York).
Muntiacus: representatives of the range of living species and subspecies (Natural History Museum, London).
Details of specimens in text Fig. 2:
B: left, Cervus canadensis, modern, AMNH 123050; right, M. giganteus, Allerød, Ireland, NHM M18078.
C: left, Cervus elaphus, modern, NHM 18220.127.116.11; right, Megaloceros giganteus, Allerod, Ireland, NHM 14171.
D: left, Cervus elaphus, Middle Pleistocene, West Runton, NHM M17508; right, Dama dama, Middle Pleistocene, Swanscombe, Harrison Museum (photo reversed).
E: left, Cervus elaphus, Middle Pleistocene, Grays, NHM M22034; right, M. giganteus, Late Pleistocene, Kent’s Cavern, NHM M14242.
It is typical of cervid dental and skeletal morphology that characters show considerable intra-specific variability (Lister 1996, Pfeiffer 1999). To take account of this, every effort was made to score as many individuals of each species as possible on each character (Table S1). Variation was quantified as described by Lister (1996). Almost all characters were defined with two alternative character states, but in the scoring of individual specimens, dental and skeletal characters were scored on a five-point scale from showing the full expression of one state, through intermediate expressions, to the full expression of the alternative state. These were then coded as 100, 75, 50, 25 and 0% of one of the two states chosen arbitrarily. Within each sample (i.e. one character scored on a sample of individuals of one species), a mean value was calculated. This gives some estimate of the degree of consistency of expression of the character states in the species.
Provided sample size was ≥4, a mean expression of ≥75% (¾) for a character state was required for the sample to be coded with that character state for the purpose of analysis. For sample sizes of ≤3, the required figure was increased to 87.5% (7/8). Lower figures (e.g. 70% one state, 30% the other) were coded as polymorphic. This is a form of ‘confidence coding’ (Wiens, 1999).
Characters were discarded only if (i) they were found to be too difficult to define or score consistently, or (ii) available sample sizes were too low.
A character for which samples showed polymorphism, instead of being coded ‘0’, ‘0&1’, and ‘1’, was instead coded as a three-state ordered character ‘0’, ‘1’ and ‘2’ respectively (cf. Tables S1 & S2). This appropriately treats the polymorphic condition as though an intermediate stage in the replacement of one morphology by another, and corresponds to the ‘ordered scaling’ method described by Wiens (1999). The method is particularly advantageous in the present project in that it gives equal weight to (a) a sample showing ‘intermediacy’ due to intermediate character expression in individual specimens; (b) a sample showing ‘intermediacy’ due to a mixture of specimens showing fully expressed, but different, character states; or (c) any combination of the two. All of these situations are commonly encountered, and cannot in practice be readily coded in such a way as to distinguish among them. In addition, the scoring of polymorphism as an intermediate in an ordered series allows PAUP to utlilise the information in tree-building, which it does not for characters entered as polymorphisms per se. Where only a single specimen was available and its morphology was intermediate between two character states, it was coded as ‘?’ for the purposes of analysis.
The re-coding of a two-state character into a three-state ordered character does, however, potentially double the weight of the character in cladistic analysis, as two steps are required to change from ‘0’ to ‘2’. For this reason, all ordered characters were weighted by 1/(n-1), where n is the number of character states, using the ‘scaling’ option in PAUP (Swofford 2002).
In Table S1 the character types (ordered/unordered), character states, and their percentage expressions, are shown before these transformations. Table S2 shows the transformed matrix ready for phylogenetic analysis, and indicates the weightings given to each character. In Table S2 also, character states are colour-coded to give an impression of the distribution of derived character states. Shades of yellow indicate degrees of derivation from the outgroup (assumed ancestral) condition in an ordered character. In the many cases where polymorphic two-state characters have been re-coded as ‘0, 1, 2’, pale yellow indicates the polymorphic condition (code 1), bright yellow the fixed derived state (code 2). For characters where the outgroup is itself polymorphic, taxa fixed for one of the alternative states are indicated in yellow, those fixed for the other state are indicated in blue.
Comparison with previous studies
In a preliminary study including a wider range of cervid taxa (Nock 2001, data not shown), some interesting convergences between the M. giganteus-Dama clade and Rangifer tarandus (reindeer) were evident, including palmated antlers and double-ridged axis (main text, Fig. 2 C). However, the position of the reindeer outside the Cervini is incontrovertible (Groves & Grubb 1987, Randi et al. 1998, Pitra et al. 2004).
In an important study, Pfeiffer (1999, 2002, 2005) has undertaken a major analysis of cervid phylogeny based on morphological characters. This work broke new ground in the application of cladistic methodology to Quaternary mammal palaeontology, and in the extremely detailed observation and recording of characters. In several of her conclusions, Pfeiffer’s work and the present study are in agreement. However, in the key question of the relationship of M. giganteus, the two studies differ dramatically, as Pfeiffer (1999, 2002) places it in a clade with red deer, Cervus elaphus - albeit with bootstrap values of 60% or less (Pfeiffer 2005) - , whereas our molecular and morphological data reject that topology in favour of a relationship with Dama. Comparison of datasets and methodology suggests the following reasons for this difference:
1. Most of the characters here interpreted as synapomorphies of Dama and M. giganteus were not found in Pfeiffer’s study (Table S1).
We mostly restricted individual characters to dichotomies between two or three states of a single morphological entity. Pfeiffer more frequently coded characters as up to 6 ordered ‘grades’, or in other cases gave as (unordered) character states a considerable range of morphologies. An example of the effect of this is that whereas we find the presence of a posterior antler tine (the ‘back tine’, character 1, Table S1) to be shared between D. dama and M. giganteus against its total absence in our other taxa, Pfeiffer’s (2002) character 97 defines four varieties of posterior tine, with different species of Dama and M. giganteus scored as different character states which are not treated as homologous by the analysis.
3. We allowed PAUP4.0b10 to determine character polarity using an outgroup (Muntiacus spp.), whereas Pfeiffer determined polarities on the basis of the distribution and frequency of the character states among living and fossil cervids, the more widespread character state being treated as plesiomorphic.