1974; Honig, 1978; Olton, 1978).
By far the most frequently used procedures in the investigation of working memory in animals are variations of the delayed matching-to-sample (DMTS) task. Usually a limited number of stimuli make up a set of "samples." On each trial, one sample is presented to the animal and later, after a delay (retention interval), that sample and another member of the set (the "comparison stimuli") are displayed. Choice of the sample is rewarded. Many (20 - 250) such trials occur within an experimental session separated by intertrial intervals (ITIs) normally longer than delays. To the extent that performance remains accurate in spite of a delay between the sample and comparison stimuli, one can conclude that the animal remembers the sample (or an associate of it; cf. Roitblat, 1980).
The delayed matching-to-sample task is a special case of the more general delayed conditional discrimination task. The delayed conditional discrimination and delayed matching-to-sample tasks are the same except for the "symbolic" nature of the conditional task, in which the sample and comparison sets are physically different but associatively related by the reinforcement contingencies chosen by the experimenter. For example, one might choose two forms (triangle and circle) as samples and two colors (red and green) as comparisons and define choice of red (green) as being correct after a sample of the triangle (circle).
Successive matching-to-sample is another task closely related to delayed matching-to-sample. In this task too, a trial is begun with a sample stimulus, but the trial ends with only one comparison -- either the sample or another member of the set responses to the comparison that is identical to the sample are rewarded while responses to the nonmatching comparison are not. If the identity of the sample is remembered to the end of the delay then there should be a substantial difference between rates of responding to matching and nonmatching comparison stimuli.
There are probably numerous other kinds of tasks that could be mentioned as appropriate to the study of working memory in animals. Two will be of later concern in this chapter; both have been used to investigate spatial memory. The simpler of the two is the delayed alternation task, most often done with a T-maze. Both right and left alleys of the maze end in goal boxes, both of which are baited with rewards at the start of the trial. Trials then consist of two parts. The animal is first "forced" to choose either the right or left alley in which a reward is consumed. After a delay, the animal is returned to the maze and offered a "free" choice between alleys. Correct performance consists of choosing the, alley opposite to the forced: choice (in which an unconsumed reward remains). The other task makes use of the radial-arm maze (e.g., Olton & Samuelson, 1976). Here: there are more (often eight) alternative alleys radiating from the central platform. Rewards are placed at the end of each alley and the animal continues to choose until all rewards are consumed. Repetitions of previously chosen alleys result in nonreward and are thus scored as errors. In both instances' it is presumed that the animal must remember where it has recently been in order to avoid repeating