Conclusions

My intention has been to give some insight into how, by combining the evolutionary approach of optimal foraging theory with the hypotheses about the psychological mechanisms underlying interval timing, we can move toward a fuller understanding of both foraging behavior and timing. In all of the foraging examples described above, the basic strategy has been to take the constraints embodied in the scalar timing model, apply these to different foraging problems, and investigate how they affect optimal foraging decisions.

In all cases, the addition of scalar timing explains details of the birds' behavior that could not be accounted for with an unconstrained evolutionary approach. Three key assumptions of the scalar timing model emerge as being particularly important in producing these results. The first important assumption is the scalar property, whereby the duration of an interval is proportional to the precision with which it is represented in memory. The second assumption is that time intervals are assumed to be represented in memory as distributions of the intervals experienced rather than as a single statistic. Finally, the third crucial assumption is that reference memory is accessed via random sampling. There is extensive evidence for the scalar property; however, the latter two assumptions are more difficult to test empirically, and given the failure of the basic scalar timing model to account for some quantitative features of behavioral data (e.g., Bateson, 1993; Bateson and Kacelnik, 1996; Brunner et al., 1997), it is probably wise to maintain an open mind about the exact details of how memory is represented and accessed at this stage (e.g., Kacelnik and Brito-E-Abreu, 1998).

The examples described also highlight how our understanding of interval timing can benefit by thinking in detail about what animals use timing for. Specifically, the applications above have given insights into how the functional demands of different uses of temporal information may have shaped how this information is both represented and used in decision making. The basic scalar timing model makes no assumptions about the learning process and assumes that animals are equipped with memories that represent all possible experiences. However, the study of Todd and Kacelnik (1993) makes it clear that scalar timing models need to take account of learning, and that it may make sense for the memory to be weighted in favor of recent experience. In terms of decision making, the study of Brunner et al. (1992) shows that the biases assumed in the decision stage of scalar timing should be thought of as reflecting the optimal trade-off between responding early and late, and that the position of this trade-off will differ depending on the decision being made.

The flexibility of the scalar timing model has proved central in applying the model to a range of different foraging problems. The modularity of the model makes it easy to modify the details of how temporal information is represented in memory and how this information is used. For example, if we look first at how information is represented in the various examples, some of the applications described require a single reference memory, whereas others involving choice assume that there are two (Bateson and Kacelnik, 1995b; Reboreda and Kacelnik, 1991) or more (Shet-tleworth et al., 1988) memories. In most applications, experience is weighted equally in reference memory, but in some cases, it is necessary to assume that memory is biased toward recent experience (Todd and Kacelnik, 1993). If we look at how the temporal information represented in memory is subsequently used, again all of the applications make different assumptions. In some cases, a single sample from memory is compared with the current value in the accumulator (Brunner et al., 1992); in other cases, a weighted average of a sample from memory and the value in the accumulator is used in decision making (Todd and Kacelnik, 1993); and in yet other applications, the critical comparison is between two samples drawn from different reference memories (Bateson and Kacelnik, 1995b; Reboreda and Kacelnik, 1991; Shettleworth et al., 1988). An interesting point that emerges from Brunner et al.'s (1992) study is that the same reference memory may be accessed by two or more different decision-making mechanisms.

As a final thought, it is interesting to note that the flexible use of the same basic components described above has a close analogy with current thinking about the evolutionary process. The neural mechanisms responsible for producing new adaptive behavior patterns are not created from scratch, but are shaped by natural selection from small modifications of existing mechanisms. Thus, it is plausible that the tinkering necessary to apply scalar timing to a range of foraging problems is an accurate reflection of how interval timing mechanisms have evolved.

0 0

Post a comment