Computer simulations based on the Scalar Expectancy Theory (SET) and t
he connectionist model of Church and Broadbent (1990) were run to matc
h data sets from the peak procedure. On the peak procedure, a light or
tone usually signals a reward for a response after a fixed interval (
FI), but occasionally, the signal is left on for a long time and rewar
d is withheld. On such a test, a period of high rate of responding (ru
n) is sandwiched between periods of low rates of responding. Models we
re run to match the means and standard deviations of the start, the en
d, the middle, and the duration of the run, as well as the correlation
s among them. On a trial, the models based on SET determined the start
and the end of the run according to a memory of expected time of rewa
rd and one or two thresholds. Models sampling two thresholds, with bot
h difference and ratio comparison rules, fit the data well. In the con
nectionist models the memory was a matrix of vector autocorrelations,
with a vector representing a clock reading on a set of oscillators. Th
e thresholds were each an angle between the clock vector and a compari
son vector derived from memory. These models did not fare well.