Acquisition of a Joystick-Operated Video Task by Pigs (Sus scrofa). Candace C. Croney1 and Sarah T. Boysen. Front. Psychol., February 11 2021. https://doi.org/10.3389/fpsyg.2021.631755
Abstract: The ability of two Panepinto micro pigs and two Yorkshire pigs (Sus scrofa) to acquire a joystick-operated video-game task was investigated. Subjects were trained to manipulate a joystick that controlled movement of a cursor displayed on a computer monitor. The pigs were required to move the cursor to make contact with three-, two-, or one-walled targets randomly allocated for position on the monitor, and a reward was provided if the cursor collided with a target. The video-task acquisition required conceptual understanding of the task, as well as skilled motor performance. Terminal performance revealed that all pigs were significantly above chance on first attempts to contact one-walled targets (p < 0.05). These results indicate that despite dexterity and visual constraints, pigs have the capacity to acquire a joystick-operated video-game task. Limitations in the joystick methodology suggest that future studies of the cognitive capacities of pigs and other domestic species may benefit from the use of touchscreens or other advanced computer-interfaced technology.
Discussion
Overall, all pigs performed significantly above chance on one-walled targets, which indicates that, to some extent, all acquired the association between the joystick and cursor movement. That the pigs achieved the level of success they did on a task that was significantly outside their normal frame of reference in itself remarkable, and indicative of their behavioral and cognitive flexibility. Their high level of social motivation to perform the task was also noteworthy. Although food rewards associated with the task were likely a motivating factor, the social contact the pigs experienced with their trainer also appeared to be very important. Occasionally, during some sessions, equipment failures resulted in non-reward following correct responses. On these occasions, the pigs continued to make correct responses when rewarded only with verbal and tactile reinforcement from the experimenter, who was also their primary caretaker. Additionally, during times when the task demands seemed most challenging for the pigs, and resulted in reluctance to perform, only verbal encouragement by the experimenter was effective in resuming training. This may have been due to the strong bond the pigs developed with the experimenter during training, which would support the assertion of Boysen (1992) that the human-animal bond is a crucial element in the success of animals used in studies of comparative cognition.
It should be noted that despite performing above chance on the SIDE task, even the pig that performed best did not approach the level attained by non–human primates that acquired the task after a comparable number of trials (see Hopkins et al., 1996). Indeed, none of the pigs was able to meet the criteria of Hopkins et al. (1996) for demonstrating motoric or conceptual acquisition of the SIDE task. There are several possible explanations for the pigs’ failure to meet they criteria. First, they were established for dexterous primates (rhesus monkeys and chimpanzees); although no clear rationale was provided for their adoption. Thus, it was difficult to know how to adapt those criteria for pigs, taking into account their more limited perceptual and motor capabilities, which clearly differ from primates. For example, the visual demands of the task may have been particularly problematic for the pigs, since we had previously established that all four subjects were far-sighted. As sufficient visual capability is a prerequisite for successful completion of a joystick-operated-video game task, and despite attempts to position the computer monitor appropriately, it is impossible to know how well the pigs were able to see, and subsequently correctly discriminate between targets. Furthermore, because of the positioning of the pigs’ eyes relative to their snouts, they were often forced to watch the screen prior to moving the joystick, and then check their progress after cursor movement was initiated. This artifact of the pigs’ anatomy likely contributed to some of their errors because in order to succeed, they not only needed dexterity and conceptual understanding of the task, but perhaps also short-term or working memory (which is not well understood in pigs) of the target position locations.
In addition, the pigs’ limited dexterity no doubt constrained their performance. Because the joystick-operated video-task paradigm was initially designed for use by non-human primates with great manual dexterity, modifications to the equipment were necessary so that the pigs could use their snouts to manipulate the joystick. However, the pigs’ ability for such manipulation was restricted to their normal range of head and neck movements. This limitation appeared particularly troublesome for the Yorkshire pigs whose larger size also constrained their ability to reposition themselves as needed to contact targets located in the horizontal plane. Thus, it was not surprising that the Yorkshire pigs performed better on vertical plane movements, which are more frequently seen in their normal behavioral repertoire during routine activities such as rooting. In fact, when faced with left or right targets, the Yorkshire subjects were often observed to alter their stance so that they were parallel to the computer screen. This way, they could approach horizontal targets in the same way they did for those in the vertical plane. Because of their small size, the micro pigs were better able to reorient themselves as needed to view the computer monitor and complete horizontal plane movements. This flexibility likely resulted in better performance in both planes and may have contributed to their superior performance compared to the Yorkshire subjects. Ebony and Ivory’s smaller size also enabled them to be maintained in the laboratory for a much longer period for training and testing (15 months) than the Yorkshire pigs. Thus, they were afforded the opportunity to continue training, thereby contributing to their improved performance on the SIDE task. Consequently, their terminal performance was much better than the Yorkshire pigs that were trained for only 10 weeks on the same task.
Additional problems that may have been attributable to dexterity limitations were observed when the pigs were unable to completely move the cursor toward a target wall and finish the trial, simply because of the angle at which the cursor approached the target. On these occasions, the pigs often nosed the joystick to move the cursor back out of the target wall and then altered the angle at which they approached the target. However, in doing so, they sometimes contacted an incorrect wall, resulting in reduced accuracy on their first cursor attempts. Further, when the pigs were unable to make contact with a horizontal target, they often resorted to strategies that allowed them to move the cursor upward, then down into the correct left or right wall. These responses were consistently observed, particularly for Hamlet and Omelet, who systematically responded with a series of movements that resembled an “inverted v” when faced with right or left targets. The resultant asymmetry in the pigs’ performance relative to target position is similar to that observed in rhesus monkeys (Hopkins et al., 1996). In comparing the performance of rhesus monkeys to chimpanzees on the SIDE task, Hopkins et al. (1996) observed that the monkeys had more difficulty responding to horizontal targets, suggesting that their manipulative behavior was less diverse than chimpanzees. This problem may, in part, explain the pigs’ poor performance relative to primates, as their ability to manipulate objects is significantly less dexterous and flexible.
Response biases can often be inevitable when testing animals, and they emerged during testing with the pigs as well. For example, while Ebony, like all of the subjects, showed some level of side bias (left), he did correctly move the cursor to the right numerous times on all but the one-wall task. As previously noted, these trials created the smallest targets for the pigs. Side bias training was instituted for all pigs manually upon observation of biases because although the software titrated to an easier level of task difficulty if a subject made errors consistently, the program’s random generation of target locations did not facilitate training to overcome bias. This intervention was not successful, however. Learning on manual side-bias training with objects or with the joystick with the computer turned off (necessary given the previously mentioned software limitations) did not appear to generalize to the joystick-operated task. A few explanations for this observation are plausible. First, Ebony may simply have been limited in either or both dexterity and the paw/snout/eye-coordination needed to hit right-sided, one-walled targets. It is also possible that because the video-task apparatus was not centered in the pen due to constraints of the testing space, Ebony’s body positioning to complete such tasks may have further constrained his performance given that additional training did not correct the side-bias problem with the joystick, although it was effective on bias correction using objects (Croney, 1999). It is also possible that some degree of instinctive drift may have impacted his and the other pigs’ performance, especially as the tasks became more challenging and rewards for behaviors performed were reduced due to errors.
An alternative explanation for the difference between the pigs’ and primates’ performance that must be considered is that the pigs may have been unable to fully comprehend the concepts required to perform well on the SIDE task. Difficulties with the conceptual component of the task may have been due, in part, to the spatial discontiguity of the stimulus and response. Meyer et al. (1965) suggested that a primate’s learning efficiency might be impaired when the hand used to execute a response was placed in an area distant from the location of the discriminative stimuli. A similar rationale may have been a factor for the pigs, since the movement of their snouts was some distance from the images displayed on the monitor, and the lateral placement of their eyes may have contributed to a cognitive disconnect between their movements and the resulting changes appearing on the screen.
In addition to the difficulties posed by limited dexterity and vision, several methodological factors may also have impeded the pigs’ performance on the SIDE task. First, because a protocol for testing pigs using the joystick-operated video-game task paradigm had not previously been established, the methods used in the current experiment were exploratory. As such, some changes in procedures and equipment were necessary during the experiment to correct concerns as they emerged. For instance, early design flaws in the joystick apparatus were detected and required correction. Initially, the protective welded plastic area surrounding the joystick was too high and impeded movement of the joystick in all directions. In addition, positioning of the feed delivery tube attached to the automatic dispenser sometimes resulted in failure to deliver rewards to the pigs after correct responses early in training and required correcting. This delay in reinforcement following a correct response may have impeded the animals’ initial learning. Finally, the test pen was designed so that the joystick apparatus was positioned approximately 0.04 m away from the right side of the pen. This initial positioning proved to be significant in that it restricted the pigs’ abilities to stand or move to the right of the joystick.
Initial training procedures also proved to be problematic. One problem in the training process was that the pigs were allowed to work at their own pace, which resulted in a large set of data consisting primarily of four- and three-sided tasks. After the protocol was amended to require performance of a minimum number of two- and one-walled targets during each session, improved performance on these conditions was observed. However, the Yorkshire pigs had been terminated from testing by the time procedures were revised, and thus did not benefit from the revision. Moreover, this change in training made it extremely difficult for the micro pigs to achieve stringent criteria of Hopkins et al. (1996) for all facets of task acquisition.
Taken together, the failure of all subjects to meet the criteria for SIDE task acquisition may reflect the limitations first imposed by procedural methodology issues, and visual and motor skill limitations, rather than learning deficits. Although their performance was limited compared to primates tested, that they were able to perform as successfully as they did on one-walled targets suggests they acquired some important aspects of the task demands. However, it is impossible to determine to what extent their ability to demonstrate conceptual understanding of the SIDE task may have been constrained by their perceptual and motor capacities. Nonetheless, evaluation of their terminal test results showed that all pigs improved their performance with respect to the various target positions. This improvement was particularly noteworthy for the Yorkshire pigs (Hamlet and Omelet), who completed only a few 100 trials in their 10 weeks of training on the task. Furthermore, the high level of performance attained by one of the micro pigs (Ivory), regardless of target position or number of walls, strongly suggests some level of conceptual acquisition of the task.
In summary, the results of the present study underscore the importance of understanding the basic perceptual and motor capabilities of a species prior to developing appropriate methods of testing their cognitive abilities. While the joystick-operated video-game paradigm has proven suitable for testing several species, including monkeys, pigeons, and chimpanzees (Rumbaugh et al., 1989; Washburn et al., 1990; Spetch et al., 1992; Hopkins et al., 1996), it is not optimal for testing the cognitive abilities of pigs, as their performance was clearly hindered by dexterity limitations and visual constraints. Thorough investigations of the pig’s visual and motor capabilities are necessary before their cognitive abilities can be adequately assessed using this or any type of technology. Use of a computer touch screen may better address the problem of limited dexterity and would likely provide a more viable alternative in future computer-interfaced studies of the cognitive abilities of pigs.