A Method to Train Marmosets in Visual Working Memory Task and Their Performance

Abstract

Learning and memory processes are similarly organized in humans and monkeys; therefore, monkeys can be ideal models for analyzing human aging processes and neurodegenerative diseases such as Alzheimer's disease. With the development of novel gene modification methods, common marmosets (Callithrix jacchus) have been suggested as an animal model for neurodegenerative diseases. Furthermore, the common marmoset's lifespan is relatively short, which makes it a practical animal model for aging. Working memory deficits are a prominent symptom of both dementia and aging, but no data are currently available for visual working memory in common marmosets. The delayed matching-to-sample task is a powerful tool for evaluating visual working memory in humans and monkeys; therefore, we developed a novel procedure for training common marmosets in such a task. Using visual discrimination and reversal tasks to direct the marmosets' attention to the physical properties of visual stimuli, we successfully trained 11 out of 13 marmosets in the initial stage of the delayed matching-to-sample task and provided the first available data on visual working memory in common marmosets. We found that the marmosets required many trials to initially learn the task (median: 1316 trials), but once the task was learned, the animals needed fewer trials to learn the task with novel stimuli (476 trials or fewer, with the exception of one marmoset). The marmosets could retain visual information for up to 16 s. Our novel training procedure could enable us to use the common marmoset as a useful non-human primate model for studying visual working memory deficits in neurodegenerative diseases and aging.

Keywords: aging; delayed matching-to-sample task; marmoset; neurodegenerative diseases; non-human primate; visual working memory.

FIGURE 1

Apparatus. Front view of the apparatus (top left). The apparatus was 30 cm (W) × 20 cm (D) × 25 cm (H) in size, and the screen was 154 mm (W) and 91 mm (H) in size (1024 pixels × 600 pixels). A food tray was located underneath the screen. Back view of the apparatus (bottom left). A black USB-powered feeder was attached to the acrylic panel. A marmoset touching the screen (right). The marmoset typically crouched in the testing cage to touch the screen or pick up a reward. The distance between the screen of the apparatus and the front panel of the cage was approximately 45 mm.

FIGURE 2

Steps for touch training. In Step 1, the marmosets touched any part of a large red square (500 pixels × 500 pixels) on the screen to receive a reward. In Step 2, the color of the square was randomly set to red, yellow, or blue. In Step 3, the size of the square gradually decreased to the final size (200 pixels × 200 pixels). In Step 4, the location of the square varied from trial to trial.

FIGURE 3

Delayed matching-to-sample task and visual stimuli. (A) After the marmosets touched the sample stimulus, a matching stimulus was presented at the same center position, and a non-matching stimulus was presented 250 pixels right or left of the center. (B) A sample stimulus was presented twice to ensure that the marmosets looked at it. The non-matching stimulus was always located 250 pixels right or left of the center, but the position of matching stimulus gradually shifted 0, 125, 150, 175, 200, 225, and 250 pixels from the center. (C) Task events in a trial of the delayed matching-to-sample task. The same sample stimulus was presented thrice. (D) Visual stimuli used in this study, including four graphic patterns (G1, G2, G3, and G4) and two flower photographs (F1 and F2).

FIGURE 4

Trials needed to meet the proficiency criteria in the delayed matching-to-sample task. The numbers of trials needed to meet the proficiency criteria in the third (left), fourth (center), and fifth phases (right) are summarized. G1, G2, G3, and G4 represent the graphic patterns used as stimuli. The thick shaded line indicates median values.

FIGURE 5

Learning curve. Learning curves of all 13 subjects (S1–S13) in the third phase are presented. Two subjects failed to meet the proficiency criteria in 5000 trials.

FIGURE 6

Delayed matching-to-sample task performance as a function of delay length. (A) Mean performance of the marmosets in the short (red line), middle (green), and long (blue) delay conditions. The error bars represent standard errors. (B) Estimated performance curve based on the parameters described by Lind et al. (2015). The data for individual marmosets are indicated by gray dots.