Testing Feldman's theories of Ottoman-Turkish music from a computational ethnomusicological perspective

Abstract

This article aims to test six distinct theories proposed by Feldman concerning Ottoman-Turkish music using the MakamKit computational musicology analysis engine on a sample consisting of 149 peşrevs and saz semâîsis drawn from the open-access SymbTr symbolic score corpus (Karaosmanoğlu, 2012), together with three Mevlevi âyins. Each study derived a priori directional hypotheses from Feldman’s writings and tested them under multiple-comparison correction using methods including Spearman rank correlation, composer-clustered bootstrap, mixed-effects models, blind clustering, composer-grouped cross-validation, and similarity ranking. Across the six studies, the findings exhibit a common pattern. No test produced consistent evidence in the direction predicted by Feldman. Instead, the most consistent and recurrent deviation was the cadential-convergence measure, which moved in the direction opposite to that predicted in two independent tests of the transition from terkib to seyir (Studies 1 and 2). Apart from this reverse signal, the remaining analyses either produced no consistent pattern or could not be tested at all. The period-specific change in the articulation between usul and melody (Study 3) left no detectable trace on the surface of modern notation; the model–nazire relationship (Study 4) produced not a validated discovery but only a priority ranking for expert examination; the four-period periodization (Study 5) could not be recovered as a natural cluster structure, and only limited statistical evidence was found for a coarse early/late distinction; and the form–style relationship in the First Selâm (Study 6) could not be tested because of the absence of source-traceable scores and was therefore presented as a Stage 1 registered protocol. In conclusion, these six studies do not falsify Feldman’s ethnomusicological historical thesis. Rather, the findings should be interpreted within the limitations of the modern symbolic transmission sample and the selected measurement proxies. The principal contribution of this research is to place theories of Ottoman-Turkish music within a testable and reproducible computational ethnomusicological framework while honestly documenting the points at which that framework remains silent.