17

initially (test). The critical comparison was of the duration of responses following the initial baseline presentation with those evoked by the final test presentation.

The experiments exploring the vervets' responses to calls produced by their companions (Cheney and Seyfarth, 1988) involved systematic manipulations of the relationship between the baseline/test stimulus and the sounds used in the intervening habituation series, varying acoustic structure, individual identity of the caller, and the putative referent. There was a significant decrement in the response to a 'chutter' call (characteristic of intergroup interactions) when the baseline and test presentations were separated by an repeated presentations of the same individual's 'wrr' (another call type characteristic of interactions with neighboring groups). No such decrement was obtained when the habituation series was a 'wrr' recorded from another individual. Analogous experiments assessed response to leopard or eagle alarms, with a habituation series made up of the other type of alarm call recorded either from the same individual as the baseline/test stimulus, or from a different individual. There were no statistically significant differences between the responses to baseline and test presentations in these trials.

Cheney and Seyfarth suggest that the results from their habituation/dishabituation studies demonstrate that the principal determinant of the vervets' responses was the meaning of the calls presented, rather than their physical structure, and argue that this provides the strongest evidence available for representational  signalling (Cheney and Seyfarth, 1988, 1990, 1992; Seyfarth and Cheney, 1992, 1993). Specifically, it is claimed that vervet monkeys, "appear to process information at a semantic level, and not just according to acoustic similarity" (Cheney and Seyfarth, 1988), and that "monkeys have some representation of the objects and events denoted by different call types and that they compare and respond to vocalizations on the basis of these representations" (Seyfarth and Cheney, 1993). If true, these claims represent an important advance in our understanding of the cognitive capabilities of non-human primates, extending substantially the parallels between referential signalling and human speech. The case for representations has additional theoretical importance because it is part of the foundation upon which subsequent arguments are constructed. These include consideration of more complex cognitive processes, such as the possibility that vervets attribute mental states (e.g., knowledge or belief) to their companions (Cheney and Seyfarth, 1990, 1992), and can thus be said to have a 'theory of mind'.

The conclusion that the responses of vervet monkeys to alarm calls are mediated by mental representations is based upon two assertions: (i) that the results obtained cannot be accounted for simply in terms of physical differences between the stimuli, as in traditional habituation/dishabituation studies (e.g., Nelson and Marler, 1989), and (ii) that differences in the duration of response between baseline and test presentations are attributable to the intervening habituation experience. I shall discuss each of these issues in turn.

Cheney and Seyfarth provide descriptive statistics for a selection of call characteristics such as total duration, dominant frequency, fundamental frequency, and voicing. They conclude that the structural properties of the intergroup calls (wrr and chutter) were not more similar than those of the leopard and eagle alarms, and suggest that the acoustic characteristics of the playback stimuli cannot therefore account for the results obtained (Cheney and Seyfarth, 1988). Quantitative analyses of this kind are an essential first step in exploring the recognition and categorization of animal signals, although the data provided are not sufficient to provide a comprehensive mapping of the acoustic space occupied by the calls in the vervet repertoire, as has been possible for analogous descriptions of other signals (e.g., Marler and Pickert, 1984; Nelson and Marler, 1990). A more serious concern, however, is that there is no way of assessing the perceptual significance of the acoustic features selected. Systematic playback experiments involving either titration along acoustic continuua (e.g., Gerhardt, 1978;Zoloth et