BARYONYX WALKERI 
phylogenetic significance, and it would not be against the Inter- 
national Code of Zoological Nomenclature to call Baryonyx a 
spinosaurid; nevertheless it could be very misleading. 
3. The similarities between the two genera, as known at present, are 
not (in our opinion) sufficiently close to justify their treatment, for 
cladistic analysis, as asingle O.T.U. (operational taxonomic unit). 
: Also of importance in this connexion are two recently discovered 
theropods from the Romualdo Member of the Santana Formation 
: (Albian) of the Araripe Basin of north-eastern Brazil. Kellner & 
_ Campos (1996) described a fragmentary snout as a new member of 
_the Spinosauridae and named it Angaturama. Angaturama is espe- 
cially useful in that it shares certain apomorphous characters with 
_ Spinosaurus, yet, at the same time, possesses other apomorphous 
characters that it shares with Baryonyx but which could not be 
demonstrated in Spinosaurus itself because of the latter’s incom- 
_pleteness. We therefore believe that, using Angaturama as a link, we 
can justify a closer relationship between the Spinosauridae 
| (Spinosaurus, Angaturama and the Moroccan maxilla) and the 
| Baryonychidae (Baryonyx) than between either of those two fami- 
| lies and any other. We further believe that this relationship might be 
reflected in the classification by placing those two families together 
in the same superfamily Spinosauroidea' Stromer, 1915, excluding 
the other genera previously placed there by Sereno ef al. in 1994 
(Torvosaurus, Eustreptospondylus). 
The Spinosauroidea as defined here is characterized by the fol- 
_ lowing apomorphous characters, unique among the Theropoda: 
} 
1. The elongation of the jaws, especially in the prenarial region. 
2. A greater or lesser tendency to develop a terminal rosette (greater 
in the upper jaw than in the lower). 
}3. The shape of the front of the lower jaw, turned upwards at the 
| extreme anterior end and constricted transversely just behind that. 
| 4. An increase in the number of premaxillary teeth to seven. 
5. Teeth that show a reduction in: 
, (a) the compression of the whole tooth in a labio-lingual direc- 
| tion; 
| (b) the recurvature of the crown (only slight in Baryonyx, practi- 
: cally non-existent in Spinosaurus); 
(c) the size of the denticles on the anterior and posterior carinae 
of the crown (very fine in Baryonyx, absent altogether in 
Spinosaurus and Angaturama). 
| The otherAlbian theropod from Brazil, /rritator Martill, Cruickshank, 
\Frey, Small & Clarke 1996, is a partial skull lacking the end of the 
‘snout; this makes it difficult to compare with Angaturama, except in 
')so far as both genera obviously had elongated jaws, external nares 
set far back, and teeth that share all the unique tooth characters of 
)Spinosauridae (5. above). It seems very likely that /rritator too is a 
| spinosaurid, and even possible that it is a senior synonym (by one 
-month!) of Angaturama. Its authors, however, assigned it to the 
/Bullatosauria Holtz, 1994a, a new taxon of manuraptorans that 
. includes the Troodontidae and the Ornithomimosauria. 
| These characters could, of course, represent independent adapta- 
tions to similar diets. On the other hand, it is more parsimonious to 
: ‘Some confusion surrounds the superfamilial name Spinosauroidea (which, according 
0 the Principle of Coordination laid down in Article 36(a) of the International Code of 
ological Nomenclature, must retain the same author and date as existing family- 
roup names based upon the same type-genus). Sereno er al. (1994) proposed a taxon 
including Baryonyx, Spinosaurus, Torvosaurus and Eustreptospondylus, naming it 
orvosauroidea. However, Sereno et al. (1996) changed the name of that nominal taxon 
(© Spinosauroidea, presumably because the family name Spinosauridae Stromer, 1915 
is Senior to Torvosauridae. 
Si) 
interpret the unique distribution of all seven in the Spinosauridae and 
the Baryonychidae as indicating a sister-group relationship between 
those two families. This latter interpretation is consistent with the 
fact that the degree of reduction shown in the tooth characters (b) and 
(c) above is significantly less in Baryonychidae than in the much 
later Spinosauridae, i.e. the characters reflect a trend which seems to 
have increased greatly during the time interval between the Barremian 
and the Cenomanian. 
The systematic position of Baryonyx and its allies 
within the Theropoda 
By far the most detailed analysis of theropod relationships published 
to date is Holtz’s work of 1994a. His computer analysis of the data, 
based ona matrix of 19 O.T.U.s and 126 characters, produced a single 
most-parsimonious cladogram (his fig. 4), and a strict consensus 
cladogram of the six equally parsimonious ‘runners-up’ (his fig. 5). 
Holtz’s cladogram agrees in many respects with the cladograms 
of earlier workers, notably Gauthier’s. In some details, however, it 
differs greatly. The most important difference lies in the positions 
assigned to the Tyrannosauridae and the Troodontidae, among the 
most highly derived Coelurosauria (within the Manuraptora and 
close to the Ornithomimosauria). 
We assessed the phylogenetic relationships of Baryonyx by incor- 
porating it as an additional, twentieth taxonomic unit into Holtz’s 
data-matrix. It seemed to us, however, that a few of the characters 
that Holtz had used in his analysis were unsatisfactory and that 
others were scored wrongly. This opinion has been confirmed 
independently through the recent publication of criticisms by Clark, 
Perle & Norell (1994) of several of Holtz’s characters (see Appen- 
dix C) leading to scepticism of his main conclusion, that the 
Tyrannosauridae should be placed within the Manuraptora. Our 
relatively minor modifications of Holtz’s analysis have not affected 
that conclusion. 
We deleted 8 characters from Holtz’s data-matrix and re-scored a 
few others, thereby producing a modified data-matrix (Appendix C). 
All available information on Baryonyx (on 57 characters, informa- 
tion on the 61 others being unavailable or too uncertain to be of any 
value) was added and the analysis run with both the same computer 
programme (PAUP Version 3.0, Swofford 1990) and with Hennig86 
(Farris 1988) and the results compared. PAUP was run under Heuris- 
tic and Branch & Bound algorithms and produced six equally 
parsimonious trees of 228 steps, with a consistency index (C.I.) of 
51% and a retention index (R.I.) of 70%. Hennig86, using mh* and 
Branch & Bound options, gave similar results: six equally parsimo- 
nious trees of 234 steps, witha C.I. of 50% and an R.I. of 70%. These 
figures resemble Holtz’s (51% and 71% respectively). Holtz’s best 
cladogram had 244 steps, a few more, but it should be remembered 
that he used 126 characters as against our 118. All trees were rooted 
to a hypothetical ancestor on unordered characters. 
All these trees (and the corresponding consensus trees) produced 
by both programmes display topologies that are generally similar to 
each other and which, in their broad outlines, resemble Holtz’s 
single most-parsimonious cladogram. In all of them there is a 
monophyletic Ceratosauria arising from the basal node, and, much 
farther away from the root, a monophyletic Neotetanurae. Between 
them are three other nodes that give rise to a polyphyletic assem- 
blage that might be referred to collectively and informally as ‘basal 
Tetanurae’ (among which Baryonyx is placed). The more detailed 
arrangement of the O.T.U.s was constant within the Ceratosauria but 
varied from tree to tree within the Neotetanurae; those clades, 
however, do not concern us here, and discussion of such problems is 
beyond the scope of the present work. 
