ANALYSIS OF COMPLEX VARIATION: DICHOTOMOUS SORTING OF PREDATOR-ELICITED CALLS OF THE FLORIDA SCRUB JAY

ABSTRACT

The acoustically complex predator-elicited calls of the Florida scrub jay Aphelocoma c. coerulescens were classified operationally by a new procedure, dichotomous sorting. Vocalizations were tape-recorded in the field during natural and experimental encounters between scrub jays and several types of live and mounted predators. Six continuous, independent variables of frequency and duration were measured in 539 randomly selected calls. Principal Components Analysis (PCA) was used to identify the variables that contributed the most variation to a given data set. Univariate frequency-of-occurrence histograms and bivariate scatter plots were used to locate distributional breaks or separations that established dichotomous subdivisions of the data set. The analytical steps were repeated on each subsequent subset of calls. Some variables that appeared graded by visual inspection were distinctly discontinuous and defined call types. Hierarchical Cluster Analysis (HCA) was used to confirm the PCA-based approach and to explore unresolved variation in heterogeneous categories. These steps produced specific criteria to classify 12 call types and subtypes that included several Inflected types that rose and fell in frequency, Steep and Low calls that only rose in frequency over time, and Flat call types that did not change in frequency over time. The criteria can be applied much like a dichotomous key to classify other scrub jay call data sets, and the methodological strategy is widely applicable to acoustical communication systems in general.