Notwithstanding this approach, the first species of Polyplectron which I happened to examine almost made me give up the search; for I found not only that the true tail-feathers, which in the peacock are quite plain, were ornamented with ocelli, but that the ocelli on all the feathers differed fundamentally from those of the peacock, in there being two on the same feather, (fig. 54), one on each side of the shaft. Hence I concluded that the early progenitors of the peacock could not have resembled in any degree a Polyplectron. But on continuing my search, I observed that in some of the species the two ocelli stood very near each other; that in the tail-feathers of _P. Hardwickii_ they touched each other; and, finally, that in the tail-coverts of this same species as well as of _P. malaccense_ (fig. 55) they were actually confluent. As the central part alone is confluent, an indentation is left at both the upper and lower ends; and the surrounding coloured zones are likewise indented. [Illustration: Fig. 54. Part of a tail-covert of Polyplectron chinquis, with two oval ocelli of nat. size.] [Illustration: Fig. 55. Part of a tail-covert of Polyplectron malaccense, with the two oval ocelli, partially confluent, of nat. size.] A single ocellus is thus formed on each tail-covert, though still plainly betraying its double origin. These confluent ocelli differ from the single ocelli of the peacock in having an indentation at both ends, instead of at the lower or basal end alone. The explanation, however, of this difference is not difficult; in some species of Polyplectron the two oval ocelli on the same feather stand parallel to each other; in other species (as in _P. chinquis_) they converge towards one end; now the partial confluence of two convergent ocelli would manifestly leave a much deeper indentation at the divergent than at the convergent end. It is also manifest that if the convergence were strongly pronounced and the confluence complete, the indentation at the convergent end would tend to be quite obliterated.