Introduction

Contrafreeloading was first reported by Jensen (1963). This is a phenomenon in animals’ feeding behavior according to which animals will work (e.g. press a lever) for “earned” food even though identical “free” food can easily be obtained from a nearby dish (see review in Inglis et al. 1997). For example, rhesus macaques worked for biscuits from a food puzzle, from which skillful manipulation with the fingers was required to retrieve the biscuits, although they were also available from an ordinary food box (Reinhardt 1994). This phenomenon has also been found in laboratory pigeons (Neuringer 1969), crows (Powell 1974), laboratory rats (Jensen 1963), grizzly bears (McGowan et al. 2010), and chimpanzees (Menzel 1991). Thus, contrafreeloading exists commonly in a wide variety of taxa in animals.

Contrafreeloading appears to contradict a basic tenet of most learning, motivation, and optimal foraging theories, namely that animals strive to maximize the ratio of reward, or benefit, to effort, or cost (Inglis et al. 1997). Inglis et al. (1997) discussed the motivational systems that might exist behind contrafreeloading. They suggested that one possible mechanism to develop contrafreeloading might be an information primacy model. According to this model, animals’ work for earned food is motivated partly by the food itself, and partly by the information about the food resource. The behavior updating their estimate of the profitability of an uncertain food resource is adaptive because some unpredictable environmental change could turn it into the optimal place to feed. Contrafreeloading might be explained by the information primacy model and the motivation to gather information about the food resource.

The information primacy model was established based on the fact that contrafreeloading should occur under conditions of using food as rewards. Also, some sensory stimuli can work as incentives for behavior in the sensory reinforcement paradigm (Matsuzawa 1981). Primates can recognize movies’ contents (Morimura 2006; Morimura and Matsuzawa 2001). So far, contrafreeloading has been investigated using only food as rewards in nonhuman animals. Here, using a sensory stimulus as a reward, I studied a contrafreeloading-like phenomenon under sensory reinforcement, which enabled me to approach the motivational system of contrafreeloading and might suggest another explanation for contrafreeloading. If contrafreeloading for sensory rewards occurs, animals will work for such stimuli even though identical stimuli can be obtained without such work.

Methods

Subjects

The subjects were three male Japanese macaques (Macaca fuscata) named Romio, Tim, and Sabu. They were living at the Primate Research Institute of Kyoto University. They were raised by human caretakers because of their mother’s death or rejection. Romio (9 years and 3 months old when the experiment was started) was hand-reared from birth. Tim (9 years and 7 months old) was reared by his biological mother at first; however, he was hand-reared by his caretakers after the age of 3 months. Sabu (approximately 2 years old) was born in the wild. Starting a few months after his birth, he was protected and reared by human caretakers because he was found to be alone and emaciated. The subjects had experienced some psychological studies before this experiment (Murai and Tomonaga 2009; Murai et al. 2004; Ogura and Matsuzawa, unpublished data). At the beginning of this study, all of the subjects had lived in individual cages (175 cm high × 85 cm wide × 80 cm long) for more than 1 year. Therefore, the visual environments of the subjects were restricted. In this study, touch responses of the subjects made on a display to obtain the presentation of movies could be maintained by providing movie rewards (Ogura and Matsuzawa, unpublished data). During this experiment, the monkeys could see other monkeys in the same room. The monkeys were fed monkey pellets and sweet potatoes daily at about 10:00 a.m. and 5:00 p.m. They could drink water ad libitum. Routine care of the monkeys and experiments were performed in accordance with the guidelines of the Primate Research Institute, Kyoto University.

Apparatus

Two computer-controlled touch-sensitive displays were used in this study. During the experiment, each monkey’s home cage was divided into an upper and a lower compartment by inserting a metal board horizontally 50 cm into the cage (Fig. 1). The monkey could move freely between these two compartments. A touch-sensitive display (30.4 cm × 22.8 cm) (model L352T-C-BK, Eizo Nanao, Ishilawa, Japan and model CV516PJ, Totoku electric, Tokyo, Japan) was attached to each compartment. The monkeys were allowed to touch the displays through the bars of their cages. The apparatus was attached only during the experiment and was removed at other times.

Fig. 1
figure 1

A monkey’s home cage was divided into two compartments, and a touch-sensitive display was attached to each compartment. During the experiment, the monkeys were allowed to move between the compartments and to touch the displays

Full size image

Stimuli

The stimuli were 21.1 cm × 14.2 cm digitized color movies (720 × 480 pixels, MPEG 1—Layer 2 files) (Fig. 2). They did not include sound. The duration of each movie was 9.5 s. Each stimulus set was composed of 10 movies showing humans and 10 movies showing animation characters, because the monkeys showed preferences for these movies in our previous experiments (Ogura and Matsuzawa, unpublished data). The humans in the movies were novel persons for the monkeys. The behaviors of the humans in the movies were walking, working, and conversation with another human. Computer-generated humans were used as the animation movies. The backgrounds of the characters were plain vivid colors. I used four sets of stimuli, and therefore, the total number of stimuli was 80.

Fig. 2
figure 2

Examples of the stimulus movies. These photographs show the images at 0, 3, 6, and 9 s from samples of a human movie and an animation movie

Full size image

Procedures

In the experiment, both displays showed movies that could be viewed according to different behaviors of the monkeys, namely “free” movies and “earned” movies. As the free movies, one of the displays showed movies in a random order within a stimulus set continuously irrespective of the monkey’s response. A session began with a blank screen for 0.5 s with a beep sound, and then, the movie was played for 9.5 s. This routine was continued until the end of the session. As the earned movies, the other display showed movies under a sensory reinforcement paradigm with a conjugate reinforcement condition (Fujita and Matsuzawa 1986; Matsuzawa 1981; Rovee-Collier and Gekoski 1978). In this paradigm, the amount of the reinforcer varies depending on the subject’s response. A session began with the appearance of a starting stimulus (a red, blue, or green square, 3.7 cm × 3.7 cm) at the center of the display. After the subject touched the starting stimulus, a beep sound was played and the starting stimulus disappeared. One second after that, the same movie as the free movie being shown at that same moment was presented in the center of the display as the earned movie. The earned movie was presented as long as the subject touched the movie. If the subject had not touched the movie for 3 s, the beep sound was played and the movie disappeared. One second after that, the start stimulus was presented at the center of the display again. The positions of the two movies (the upper display or the lower display) were randomly changed among the sessions.

In this procedure, subjects’ touch responses to the displays were recorded by the computer that controlled the presentation of the movies. Also, the subject’s position in the cage was video recorded.

The sessions started between 2:00 p.m. and 4:30 p.m. and continued for 30 min. Each subject experienced 20 sessions of the experiment, with one session per day.

Statistical analyses

The duration that the subject stayed in each compartment was measured using instantaneous sampling with a 1-s interval (Altmann 1974) from the video record. The effects of the presentation procedure on the duration that the subject stayed in each compartment were analyzed using a Generalized Linear Mixed Model (GLMM) (lmer, lme4 library, the freeware package R, Version 2.9.2; R Development Core Team 2009); the model was constructed using a Poisson distribution because the number of the sampling points at which the subject stayed in each compartment was non-negative count data (Dobson 2002). The presentation procedure (free/earned) and the position of the display (upper/lower) were the fixed factors. Individual was included as a random factor. Within the selected model, the numbers of the sampling points at which the subject stayed in each compartment were compared in each subject by using the Wilcoxon rank sum test because the normalities of these data were not confirmed by the Kolmogorov–Smirnov one-sample test (P < 0.10). This statistical test provided only an informal test of significance because the data points for a single individual were not independent. Each data point represented one session in each presentation procedure.

Results

Figure 3 shows the numbers of touch responses to the starting stimulus of the earned movies in each session of each subject. For 20 sessions, the subjects kept touching the starting stimulus, and the presentations of the earned movies were maintained. The proportions of the numbers of touch responses to the starting stimulus, the earned movies, and the black area surrounding the starting stimulus and the movie were 92.4, 6.4, and 1.3%, respectively. All touch responses were momentary. Sustained touch responses were rarely observed.

Fig. 3
figure 3

Numbers of subjects’ touch responses to the starting stimulus of the earned movies in each session. S1, S2, S3, and S4 mean Stimulus set 1, 2, 3, and 4, respectively

Full size image

The duration of staying in each compartment was different depending on the presentation procedure. The mean number of sampling points spent staying in each compartment is shown in Table 1. The Akaike Information Criterion (AIC; Akaike 1974, Dobson 2002) can be used to compare models with different numbers of fitted parameters. The model with the lower AIC is preferred. The model including both the presentation procedure and the playing position as the fixed factors showed the smallest AIC, although all models explained the data well (Table 2). The likelihood ratio test showed a significant difference between the model including the presentation procedure and the model without it as the fixed factor (χ 2 = 752.54, P < 0.001). The presentation procedure was a factor affecting the duration of staying in each compartment. Within the selected model, however, only Romio showed a significant difference of the staying duration between the compartments of the earned movies and the free movies (Romio: W = 110, P < 0.05, Tim: W = 160, P = 0.29, Sabu: W = 211, P = 0.78).

Table 1 Mean (± SE) number of sampling points spent staying in each compartment
Full size table
Table 2 Influence of the variables (i) procedure (free/earned), (ii) position (upper/lower), and (iii) procedure and position on the staying duration
Full size table

Discussion

Here, the contrafreeloading phenomenon was tested using movie stimuli in monkeys. In this experiment, the subject touched the starting stimulus to obtain the presentation of earned movies, although the identical movies were being played as free movies. Any deviation from complete preference for the free reward suggests some level of contrafreeloading (Inglis et al. 1997). Primates spontaneously manipulate some novel objects even without any incentive (Ehrlich 1970). In the present study, however, the subjects kept making responses to obtain the presentation of earned movies continuously throughout the series of sessions even though the manipulandum was a visual stimulus, not a real object. This finding demonstrated a contrafreeloading-like behavior for movie rewards in Japanese macaques.

The presentation procedure might have no significant effect on the value of each compartment. Regarding the duration of staying in each compartment, Romio stayed significantly longer in the compartment with the earned movies than in that with the free movies. Tim and Sabu showed no significant difference in the time stayed between the two compartments. None of the subjects stayed preferentially in the compartment with the free movies. Therefore, only for Romio, the compartment of earned movies had higher value than that of free movie. At least, the free movies did not increase the value of the corresponding compartment.

This study showed that contrafreeloading-like behavior occurs in response to visual stimuli rewards in Japanese macaques. This implies that control over the environment may be the motivation behind the earned reinforcers rather than information useful for locating an alternative food source in the event of a change in the environment. Contrafreeloading is related to the value of control, according to the competence theory (White 1959), which posits that behavior is primarily directed toward controlling and modifying the environment and that such behavior is self-reinforcing (Singh 1970; Singh and Query 1971). Some previous studies revealed the empirical evidences of the value of control over environment for monkeys. In Washburn et al. (1991), the performance of rhesus monkeys on computer tasks that were selected by themselves significantly exceeded performance on identical tasks when assigned by the experimenter. In Hanson et al. (1976), the plasma cortisol level, the stress state indicator, of rhesus monkeys that had control over high intensity noise was significantly lower than that of the monkeys that received identical amounts of high intensity noise but which had no control over the noise. These studies showed the value of control over environment for monkeys, which seems to work as an incentive to the contrafreeloading behavior. The findings of this study, however, do not necessarily contradict the information primacy model, because contrafreeloading for movies might have a different mechanism from contrafreeloading for food. This study provides the first evidence that contrafreeloading occurs with movie rewards.