Date of Award


Degree Type


Degree Name

Doctor of Philosophy (PhD)




Professor A.B. Kristofferson


Accounts of the psychophysics of duration have attributed variance on the temporal decision axis either to variability in a psychological timekeeper or to variability in afferent delays between the times of stimulus presentation and registration at the timekeeper. A similar distinction could be made when accounting for variance in the timing of discrete responses either in terms of timekeeper variability or in terms of variability in efferent delays in responding. Thus in each case, an ambiguity exists for the identification of variance estimates based on behavioural data and strong assumptions have been made about the variance in one process when attempting to characterise the variance attributable to the other. It is demonstrated in this thesis that this need not be the case in tasks where subjects are required to maintain some interval (T) between sequences of identical discrete responses. Two classes of model are recognized as possible accounts of performance in such tasks.

In the tandem process model it is assumed that a timekeeper initiates responses at intervals with mean approximating T but that there is variable time lag in observing each overt response as a result of efferent delays. In the associative element chaining model it is assumed that no timekeeping as such takes place but that a reflex chain of events triggered by one overt response produces, as the last element in the chain, the next overt response. A Morse telegraph key tapping task with T in the range 170 through 350 msec reveals estimates of interresponse interval (I) lag one serial correlation to be consistently negative. This finding rejects the associative element chaining model but confirms predictions of the tandem process model.

Under the tandem process model the variances of the timekeeper intervals and the response delays can be separately determined using estimates of the variance and lag one serial covariance of I. In two experiments which employ different Sa and response equipment it is found that, whereas response delay variance changes little with I, timekeeper interval variance is a linear increasing function of I. The last mentioned finding is consistent with a set of stochastic models of the timekeeper. These postulate that intervals are generated by the time taken to attain a given count of events where interevent times are independently and identically distributed. If it is assumed that the events originate in a Poisson source, estimates of the rate parameter are of the order of 10³ events per second and well within the range of estimates obtained from psychophysical procedures. Implications of findings of non-zero response delay variance for studies of reaction time are also considered. Some of the response delay variance is shown to depend on the type of movement used in making the response. Further analysis of the data of Experiment 2 suggests that successive response delays are negatively correlated. However conclusions with respect to either of the process variances are little changed.

Files over 3MB may be slow to open. For best results, right-click and select "save as..."

Included in

Psychology Commons