Set of Pictures Used as Visual Stimuli in the Experiment. The pictures are displayed in HTML form as thumbnails on a series of pages (see Supporting DMTS Pictures). See text for details.

Training Procedures. Four stages of training were used to shape rats to match-to-sample (MTS). In stage 1, animals were introduced to the motivation-and-reward elements of the task (i.e., swimming to a hidden platform to escape from water) and then trained to discriminate precisely between two visual stimuli. Before training commenced, a picture was placed on each of the displays facing into the choice pool. One of the pictures was chosen arbitrarily as the reinforced stimulus (+; platform placed below it), and the other was chosen as the nonreinforced (–; no platform) stimulus. Rats are instinctive swimmers and the visual water task capitalizes on the natural inclination of rats to escape from water to a solid substrate, the location of which is directly paired with a large, salient picture. On their first training trial, animals were removed from a holding cage and released into the pool facing the reinforced picture, a few centimeters from the platform. Upon being released, animals usually swam forward, touched the platform, and then climbed upon it. The animals were allowed to remain on the platform for a few seconds and were then returned to their holding cages. After a delay of 1 min, the locations of the reinforced and the nonreinforced pictures were exchanged and another trial was run. After this alternation routine was repeated a few times, the release distance from the platform was gradually increased over trials until the rats could swim reliably to the platform after being released from the end of the pool opposite the monitors. On the next day, the training of animals to reliably and consistently discriminate the + from the – stimulus was undertaken. Procedures similar to the above were used, except the alternating pattern of the reinforced picture/platform location was substituted with a Left (L), Right (R) LLRLRR sequence; a simple pattern that cannot be learned by rats in this task, but reduces side biases in their responses when a stimulus is displayed three or more times in succession on the same side of the maze. On each trial, the animals were released into the pool and required to swim until they located the platform. If a rat swam to the platform without entering the arm of the maze that displayed the – stimulus, the trial was recorded as correct. If an animal passed the end of the divider and obviously entered the side displaying the – stimulus, the trial was recorded as incorrect. Achieving 9 of 10 correct during the last 10 trials, in a minimum of 30 trials, was set as the criterion for completing stage 1. Typically, 10 trials were completed on the first day, 15 on the second day, 20 on the third day, and so on, in an effort to reach a maximum of 30 trials per day. Once animals reached 9 of 10 correct, trials were stopped until the next day, when stage 2 was initiated.

The procedures in stage 1 may simply have trained rats to swim toward a picture if they recognized it and avoid a picture if they did not. Based on the idea that successful MTS requires a careful assessment of the visual stimuli, in stage 2 we adopted a procedure that encouraged animals to swim to the end of the divider, grasp it, and then examine both pictures before making a choice. We used procedures identical with those in stage 1, including the same reinforced picture, but used a novel picture on each trial presented in random order (100-plus pictures in the set) as the nonreinforced picture. The performance criterion remained at 9 of 10 correct, and a maximum of 30 trials per day were run. Once animals reached the testing criterion, trials were suspended until the next day, when stage 3 began.

In stage 3, animals were trained to resolve a new discrimination in one trial. On the first day, a novel picture was selected as the reinforced stimulus, a series of novel pictures were used as the nonreinforced stimuli, and the animals were allowed a maximum of 30 trials to achieve 9 of 10 correct. On the next day, a new reinforced picture was selected to be discriminated from a series of novel pictures, and the animals were again tested to criterion. This procedure was repeated each succeeding day until the rats achieved 9 of 10 correct in 9 or 10 trials on 2 consecutive days.

In stage 4, a sample phase was introduced to each trial. Throughout a series of daily sessions, the animals were shifted from one-trial discrimination learning to MTS. On the first trial of the first day for each animal, a picture was selected at random from the picture set as the reinforced stimulus and displayed on the sample pool display, and the rat was released into the pool. Because the previous training stages had entrained swimming to the end of the divider to make a choice, the rats usually swam to the faux divider, grasped it, looked at the sample display, and then swam toward it where they found an escape platform. The rat was then removed to its holding cage. The sample picture was then displayed as the reinforced picture in the choice pool and a nonreinforced picture was selected at random from the picture set. After 1 min, the rat was removed from its cage, placed into the choice pool, and given the opportunity to make the discrimination. The same procedure (unchanging reinforced/novel nonreinforced stimulus on each trial) was repeated 11 more times in the same session (12 sample and choice trials to perform one discrimination; 12–1). The next day, two new reinforced pictures were selected, and six trials with one (discriminated from a series of novel nonreinforced pictures) followed by six trials of the other were run sequentially (six trials each with two reinforced pictures; 6–2). On the third day, four trials with each of three new reinforced of pictures (4–3) were run; on the fourth day, three trials with each of four new reinforced pictures (3–4) were run; on the fifth day, two trials with each of six new reinforced pictures were run; and finally on the sixth day, 12 trials, each with a new reinforced picture, were run. At 12–1, rats are successfully performing trial-unique MTS; on each trial an animal is shown a sample picture, and then 1 min later, it can accurately discriminate the sample picture from a novel picture.

In stage 1 (discrimination training) rats learned to reliably perform a simple visual discrimination between two pictures. Three animals reached the criterion in <30 trials, whereas four animals required >150 trials. The average number of trials for all animals in the study to complete discrimination training was 72 (SEM = 15.0).

In stage 2 (novel picture training), animals were trained to discriminate the same reinforced picture used in stage 1, from a novel picture that changed on each trial. They accomplished this task in about half the number of trials (mean = 38, SEM = 3.9) required to complete discrimination training.

In stage 3 (one-trial learning), rats were trained to discriminate between a new sample picture and novel pictures each day until they could learn the discrimination in one trial (9 of 10 correct in 9 or 10 trials). Animals required an average of 5.3 days (SEM = 0.1) to complete stage 3. The minimum number of days for the 18 animals tested to reach the criterion was 4 (n = 5), and the maximum number of days was 7 (n = 3).

In Stage 4 (MTS training), a sample trial was introduced to the procedure, and throughout a series of daily sessions of 12 trials, rats were shifted from one-trial learning of a discrimination to trial-unique, MTS. In general, performance was unaffected by adding a sample phase between one-trial learning and MTS training; Seventy-five percent accuracy or better was maintained by all animals through the sessions of MTS training in which the number of sample pictures presented per day was increased from 1 to 12 (trial-unique MTS) over 6 successive days. On the sample phase of the first trial of MTS training, most animals approached and grasped the faux divider, looked at the display, and then cautiously swam toward the sample picture. Some animals moved around the divider several times before swimming to the computer display, but this behavior quickly decreased during subsequent trials and did not affect their performance.