The strong resemblance of the scales of the Garnett specimens to those of _Rhabdoderma elegans_ (Newberry) caused Moy-Thomas (1937:399) to add Hibbard's two species to the synonymy of _R. elegans_. But at that time only the scales could be adequately described. If the shape of the scale and the number and pattern of ridges can vary with age, size and shape of the scale, it follows that assignment of isolated scales to a species should not be attempted. Assignment to genus should be made only with caution.
_Discussion._--The relationship of _Synaptotylus_ to other coelacanths is obscure at present. The knoblike antotic processes on the basisphenoid are unlike those of any other known coelacanth. The palatoquadrate complex is shaped like that of _Rhabdoderma elegans_ but consists of fewer bones, probably because of fusion. The scales resemble those of _Rhabdoderma_. With regard to general shape of fin girdles, the pectoral girdle resembles that of _Eusthenopteron_ more than that of _Rhabdoderma_, but the cleithrum is more nearly like the cleithrum of _Rhabdoderma_. The pelvic girdle appears to be midway between those of _Rhabdoderma_ and _Coelacanthus_ in general appearance. Regarding the basal plates of the remaining fins, those of _Synaptotylus_ appear to resemble basal plates of both _Rhabdoderma_ and _Coelacanthus_. Considering the structure of the sphenethmoid region of the braincase, _Synaptotylus_ is probably more closely related to _Rhabdoderma_ than to other known coelacanth genera.
Classification of Carboniferous coelacanths has been difficult, partly because the remains are commonly fragmentary, and significant changes in anatomy did not become apparent in early studies. In general, coelacanths have been remarkably stable in most characters, and it has been difficult to divide the group into families. As Schaeffer (1952:56) pointed out, definition of coelacanth genera and species has previously been made on non-meristic characters, and the range of variation within a species has received little attention. For example, Reis (1888:71) established the genus _Rhabdoderma_, using the strong striation of the scales, gular plates and posterior mandible as the main characters of this Carboniferous genus. Moy-Thomas (1937:399-411) referred all Carboniferous species to _Rhabdoderma_, redescribed the genus and compared it to _Coelacanthus_, the Permian genus. He cited as specific characters the ornamentation of the angulars, operculars and gular plates (Moy-Thomas, 1935:39; 1937:385). Individual variation in some species has rendered ornamentation a poor criterion. This variation is apparent in _Synaptotylus newelli_ (Hibbard), some specimens having little or no ornamentation; others having much more. The number of ridges and pattern of ridges on the scales also varies. Schaeffer (1952:56) has found this to be true of _Diplurus_ also. Moy-Thomas (1935:40; 1937:385) realized that the type of scale is a poor criterion for specific differentiation. In the search for features useful in distinguishing genera of coelacanths, Schaeffer and Gregory (1961:3, 7) found the structure of the basisphenoid to be distinctive in known genera, and thought it had taxonomic significance at this level. Higher categories should have as their basis characters that display evolutionary sequences. A recent classification (Berg, 1940), followed in this paper, reflects two evolutionary trends in endocranial structure of coelacanths: reduction of endocranial ossification and loss of the basipterygoid processes. Because there has been little change in other structures in coelacanths, Berg's classification is the most useful. Berg (1940:390) includes _Rhabdoderma_ in the suborder Diplocercidoidei because of the presence of the basipterygoid processes, and in the single family, Diplocercidae, but remarks that because of the reduced amount of endocranial ossification the Carboniferous Diplocercidae "probably constitute a distinct family." In considering this concept of classification, the subfamilies Diplocercinae and Rhabdodermatinae of the family Diplocercidae are proposed above. The subfamily Rhabdodermatinae includes at present _Rhabdoderma_ and _Synaptotylus_. The principal characters of the subfamily Rhabdodermatinae, named for the first known genus, are the retention of the basipterygoid processes and the reduction of endocranial ossification. Application of this classification based upon endocranial structure would probably change existing groupings of species of Carboniferous coelacanths; the entire complex of Carboniferous genera should be redescribed and redefined. It will be necessary to consider endocranial structure in any future classification.