Scatter plots of ARDI versus various other possible metrics of acoustic musicality.

Colors are coded the same as in Fig 3 (with warmer colors corresponding to more human music relevant names) and circles enclose each of the three primate songs: Indri indri ’s: “song” (#717), Tarsius spectrum ’s “duet song” (#769), and Hylobates leucogenys ’ “trio song” (#862). ARDI identifies fewer false positives* than both PC1 (e.g. trills and twitters) and syllable count (e.g. #643 twitter & #113 trill) whereas log(SCI) captured songs more efficiently*. SCI tends to reward longer, more repetitive calls such as the loud call (#448) of Macaca silenus , the duet of Lepilemur edwardsi (#815), the chirrup pumping (#46) of Callicebus moloch , or the pant-hoot-drum (#820) of Pan troglodytes . ARDI, alternatively, tends to boost scores of shorter, more transpositionally musical calls such as the gothic chucks (#648) of Saimiri sciureus , u-trills (#113) of Cebus olivaceus , the twitter (#726) of Tarsius syrichta , or the trill (#247) of Hylobates lar (see also Fig 1 ). Note that SCI is arbitrarily scaled, but conveniently happens to align well with the scales that manifest for syllable count and ARDI. (*compared with researcher designations).