Time perception: discussion paper.

Perhaps the most obvious contrast in quality of timing is that between human and animal performance. In both cases considerable evidence is provided that mean times are quite accurate and that, to a first approximation anyhow, all distributions of normalized responses are the same. However, substantial differences exist between the animal and human data in the magnitude of the relative variability: the Weber fraction for the human data runs at about 5% and for the animals nearer to 50%.'-' However, in at least two respects these two classes of data are not comparable. First, the ranges over which they have been studied do not overlap, being between tens of milliseconds and a few seconds for the humans and from seconds to tens of seconds for the animals. Second, the pressure in the human experiments has been for precision of performance and it is far from clear that the animal studies have been designed with that in mind. The consquences for an animal who does not exhibit exact timing are really not very severe, being nothing worse than some unrewarded responding. Perhaps it would be useful for someone doing animal studies to attempt to establish the limits of their performance, which we have no reason to expect to be worse by an order of magnitude than that of people. And in the other camp, perhaps it would be useful to determine whether the 5% figure continues to hold into the region of tens of seconds. I do not underestimate the difficulties and effort required in each case, but both questions seem important. All of us have been astonished by the precision of timing that Kristofferson and his associates have managed to achieve, and until he demonstrated it 15 years ago few of us would have anticipated that the variability in timing would remain constant or nearly so over any substantial region. What is new in the present data, and even more surprising, is the series of plateaus in estimates of the period of the clock (see below), which are spaced by factors of two over intervals that increase by factors of two. This, it seems to me, has important implications for modeling, to which I now turn.