How grouping improves the categorisation of frequency in song birds and humans and why song birds do it better

Abstract

There is evidence that song bird species produce, recognize, and discriminate song notes on the basis of frequency within a range. How song birds do this is unknown. One hypothesis is that song birds represent individual frequencies separately, somehow knowing which identify conspecifics. This hypothesis suggests that song birds memorise individual song frequencies as a list without underlying rules (see Herrnstein, 1990). Because of natural continuous variability in song frequency within and among individuals (Borror, 1961; Weisman et al. 1990) song birds might need to memorize a number of individual frequencies to discriminate conspecifics from heterospecifics. A second hypothesis is that song birds categorise song notes into frequency ranges, forming large open-ended categories (see Herrnstein, 1990), and use knowledge about these ranges in song. Here, song birds treat exemplars of a frequency range collectively as suggested by theories of category learning (Keller & Schoenfeld, 1950). To decide between these hypotheses we trained zebra finches and, for comparison, humans in a (distributed S+) discrimination that required memorisation of individual frequencies and in a (compact S+) discrimination that could be acquired by classifying frequencies into ranges