Chimpanzee Gaze-Following: A Lesson in the Effects of Experimental Design

Blog || Politics || Philosophy || Science || Fiction || Quotes



Previous studies investigating the ability of primates such as chimpanzees to follow each others' gazes or those of humans have had mixed results. Some show an ability to intelligently follow the gaze of another, while other studies have failed to demonstrate this ability. Thus in 2005 Barth, et al.1 published a study they ran that sought to tease out an explanation for those mixed results and clear up the apparent contradiction.

Most studies of this sort of behavior have involved experimental manipulation where a human experimenter gives cues to a watching chimpanzee (e.g., gazing directly at an object with head aimed at it, glancing at an object with just the eyes, or pointing at an object with the index finger of an outstretched hand). The most common method involves what is known as an object-choice task: hide a food treat in one of two locations such that the subject cannot see where it was hidden (say, under one of two cups on a table), and then provide a cue to where it is hidden and observe the chimpanzee's choice between the two locations.

While most studies investigating this subject are similar in that they use an object-choice task with the cues given by humans, differential procedures have been used for training and testing at the different institutions carrying out the studies. One salient difference in the basic experimental setup for the studies was whether the chimpanzees were given the task as they approached the human giving the cue to an already hidden treat, or whether instead the subject sat in the room and watched the experimenter go through the hiding procedure (blocking the result from the animal's eyes, of course) and then was cued.

These two basic approaches are labeled in the Barth, et al. article as LEAVE and STAY conditions. In the LEAVE condition, the chimpanzee is let in and out of the experiment room between every trial (thus each trial is a new approach to a human who is already cueing toward one location or the other). In the STAY condition, the subject stays in the room through an entire block of trials.

This study hypothesized that the different setup - LEAVE versus STAY - might explain the divergent results in previous studies. The current experiment involved a within-subjects setup, meaning that the same group of chimpanzees was tested in the different conditions (LEAVE-style testing and STAY-style testing) rather than using a different group for each procedure (which is called a between-subjects design). The order of the different conditions each subject took part in was counterbalanced to avoid possible effects from always testing one condition first, e.g.

Seven subjects began in the experiment, receiving orientation training to the procedure. Five of the subjects went on to the test phase after mastering the training criterion on both conditions: for LEAVE, they had to make the right choice in five out of six trials where the experimenter was pointing very close (5 cm) to the cup hiding the food; for STAY, they had to make the right choice in five out of six trials where the experimenter hid the food in full sight of the chimpanzee (allowing the animal to watch which cup was put over the food) and then after a delay prompted the chimpanzee to choose without any cue at all. The two subjects that did not master the training phase were not tested in the experimental conditions.

The actual testing done in the study split up each condition with three different types of cues (gaze with head and eyes, glance with just eyes, and pointing with the finger from further away). Thus every subject was eventually tested for six different situations (with multiple sessions of each) - each of the three types of cues in both LEAVE and STAY designs.

Analyzing the data showed successful choice in the gaze-LEAVE condition as significantly above chance (in fact, they did perfect). In all of the other five conditions, however, the chimpanzees failed to make the correct choice at significantly above chance levels.

The authors conclude that the basic experimental setup of LEAVE or STAY influences whether chimpanzees can successfully follow gaze cues from a human. They tentatively suggest that the fact of approaching the experimenter is the main factor at work here.

Perhaps in the wild chimpanzees have more need to follow the gaze of conspecifics that they just approached (to assess the new situation), and it is less important to follow the attention of a chimpanzee that has been nearby for a while. Or, as the authors propose, perhaps the way the STAY condition is set up in the lab (where the subject has to sit and watch the experimenter go through many preparatory motions before starting the trials and cueing) confuses the animal (e.g., too much movement to pay attention to, or extraneous movement of the experimenter during setup). Either way, the results here at least seem to suggest that in the approaching sort of situation, chimpanzees can successfully use gaze-following.

As a side analysis, the study investigated whether the subjects used different hands preferentially to make their choice in the two different basic setups. It was found that some subjects' hand-preference differed significantly between LEAVE and STAY procedures. This further supports the conclusion that the experimental setup is a factor in their behavior and that the results are not due to something other than the difference in experimental setup between LEAVE and STAY.

One possible flaw in the study which the authors address is that the chimpanzee subjects used in this experiment were already trained on procedures similar to the LEAVE tasks (especially gazing and glancing) from previous experiments. However, the gaze task results were immediately and consistently above chance performance from the very start of the earliest experiment, so while there was some improvement (from about 80% to 100%), familiarity with the setup cannot explain the results of this current study. Likewise, the glance version of the LEAVE procedure started out around chance and was still around chance in this latest study, so that further supports the idea that learning effects were minimal (assuming glancing is analogous to gazing in that one respect). Thus previous experience with the procedure does not seem to explain the gaze-following ability in the LEAVE condition.

However, another explanation is possible for the difference between LEAVE and STAY setups aside from the one given by the authors. It might be something else at work besides the chimpanzee approaching versus staying and watching. For example, perhaps the different orientation procedures for each of those conditions had an effect. One established that the chimps could follow close proximal pointing cues to the right cup while the other established that the chimps could choose - without cueing - the cup they had seen the food hidden in. This difference might have carried over in some unknown way to the results of the testing.

As the authors point out, further study should be done with na´ve subjects (to replicate, and to confirm their argument that practice from previous experiments was not the sole factor involved), and applied to other species. The important lesson to take from the study, though, is that it appears the overall method in which an object-choice procedure is set up could have a significant effect on whether or not the subjects' ability to respond to gaze cues - if such an ability exists - will show up.

This lesson can be applied generally. It is easy to set up experiments to test a hypothesis about what other creatures can and cannot do, but sometimes small details in the design of an experiment can make all the difference in the results. It is nearly impossible to catch all such details when setting up experimental conditions, so really these problems only come out after similar studies have been done by different researchers using similar but varied conditions leading to divergent results.

It is when these seemingly contradictory results come up that we are forced to question our previous conclusions about the results of a study and look for deeper explanations. It is there that we are more likely to find a better understanding of behavior (e.g., the importance chimpanzees place on others' gazes when approaching a new situation) as well as an explanation that can take into account and explain all of the divergent results of different studies.

Thus not only is it important to replicate studies, but sometimes it can be of great help to vary the designs in small or significant ways in order to separate the variables involved and get a deeper picture of the subject being investigated.


1Barth, J., Reaux, J.E., & Povinelli, D.J. (2005). Chimpanzees' (Pan troglodytes) use of gaze cues in object-choice tasks: different methods yield different results. Animal Cognition, 8, 84-92.

Originally Written: 08-09-05
Last Updated: 08-09-05