Studying the evolution of vocal behaviour with the comparative method requires the identification of homologous behavioural characters. In the Paridae, an avian family of some 60 species, repertoires are particularly complex in Poecile (the chickadees), including combinatorial chick-a-dee and gargle calls, as well as narrow-band songs. Note-types are attractive candidates for characters in intra- and intergeneric comparisons, but an objective form of classification is needed to replace conventional subjective techniques. Spectrographic cross-correlation is unsuitable for this task because of the presence of numerous simultaneous frequencies, some of which are not harmonics. I tested the ability of Nowicki & Capranica's heterodyne AM 'model to predict the complex spectral structure of naturally occurring chick-a-dee notes. Synthetic notes produced by addition and subtraction of two fundamentals or their harmonics, representing nonlinear coupling of two syringeal oscillators, agreed well with the real notes from which fundamental contours were traced. Frequency and time parameters measured from the two fundamental contours were submitted to cluster analysis to classify note- types. "Phonological distance'' measured in principal component space, and "syntactical distance,'' based on the rank order in which note-types appear, were highly correlated in a sample of Black-capped Chickadee Poecile atricapillus calls. This relationship will be examined for other species of Poecile, and the phylogenetic extent of the multiplicative AM phenomenon will be explored.