The performance of rat subjects was investigated using a two-lever, mixed fixed-interval schedule. This schedule selected, at the beginning of each interval, either a 'short' fixed- interval duration of n seconds with reinforcement available on one lever, or a 'long' interval of 5n seconds, making reinforcement available on the second lever. The selection of interval duration was random.
This schedule produced bell-shaped response rate functions with the peak occurring at the shorter of the two fixed-interval durations. Normal and lognormal curves were fitted to these data, producing for each distribution three
parameter estimates representing the height (H), the spread(S), and the central tendency(M) .
In Experiment 1, the values of the fixed- interval durations were varied; H was a decreasing function of schedule value, while M and S remained approximately
constant proportions of the shorter interval duration.
In Experiment 2, the probability of reinforcement on the 'short' lever was varied. H decreased with probability while S increased; M varied slightly but unsystematically.
In Experiment 3, the magnitude of reinforcement was varied on a within-session basis; 'short ' and 'long' intervals were equally probable. H and S varied
unsystematically, while M was an increasing function of reinforcement magnitude.
In Experiments 4 and 5, the development of stable performance was followed with and without the imposition of a limited-hold condition on the shorter fixed-interval component. When the limited hold was in force, stability was achieved sooner and response rate functions were more symmetrical than when the limited hold contingency was absent.
The results obtained in these experiments were further investigated using computer modelling techniques.
An attempt was made to integrate the results with existing knowledge of fixed-interval performance, and thereby to shed some new light upon unresolved problems in the field.