Tests for fit of unimodal, bimodal and trimodal cline models for each variable (A), tests for common position of two vocalization traits with steep clines (B), and tests for common position of the vocalization clines and the microsatellite cline (C).

For clines of single traits (A), unimodal models were never the best fits. A trimodal model fit best for fundamental frequency, a bimodal model fit best for peak frequency and mean FM, bimodal and trimodal models fit equally well for peak frequency pattern and duration. For microsatellite assignment, a bimodal model fit best but a trimodal model with a deficiency of hybrids in the east was better still.For tests of common position of the vocal clines (B), individual models fit better than pooled models. A trimodal model with an assumption of common position was the best joint fit.For tests of common position of vocal clines and the microsatellite cline (C), individual models fit better than joint models.*indicates best fitting models.