Philosophers throughout the ages posited along the Cartesian hypothesis “I think, therefore I am.” At its core this very basic statement represents awareness and specific knowledge of one’s own internal mental state. In cognitive psychology this self-awareness and knowledge of one’s knowledge is referred to as Metacognition. Metacognition is often described as comprising of two aspects: monitoring and control. Monitoring is understood as the ability to know and make judgments about what one knows. Meanwhile, control is characterized by the organization of information and the ability to identify information one does know as well as planning ways to attain new information (Endel and Stephen, 1970). Yet metacognition is far more than strictly a memory and recall process—it in fact represents a very high level of thinking with cognitive processes being controlled by executive processes. Metacognition allows individuals to make plans about acquiring knowledge, monitor comprehension of information, and evaluate progress made toward achieving a specific goal (Livingston, 1997). Furthermore, metacognition is absolutely vital to learning.
Until fairly recently, these metacognitive abilities were believed to be confined to the realm of human beings. That being said, significant research has been invested in determining whether or not non-human animals also possess certain metacognitive abilities (Call et al. 1997; Smith and Shields 1997). Subsequent studies have proven that the great apes such as chimps and bonobos do in fact enjoy metacognitive skills, which led researchers to believe that a common ancestor between humans and apes must have acquired these abilities (Call and Carpenter 2001).
Yet the question remains: to what extent do other animals possess metacognitive abilities? Are these skills unique among humans and their closest evolutionary relatives or was metacognition present among more ancient ancestry? As metacognitive researchers move further and further away from humans they are coupled with the problem of how to communicate with the animals they are observing. Does the absence of language lead to an inability to convey metacognitive ability or does observation of non-human animals require creative experiments that substitute verbal language with other forms of communication and recognition?
As with all metacognitive experiments the ability to identify these skills is of the utmost importance. It is significant to note that these types of experiments, metacognitive studies among non-ape animals, are still in their infancy and must be seen as pioneers in this field. Understanding metacognition in animals such as monkeys will allow researchers to better understand why these animals act in a certain way and whether or not they are aware of these actions. It also provides researchers with a means to test animal learning, linking the presence of metacognition to a greater capacity for learning. By testing monkeys, researchers hope to expand their understanding of animal behavior, with success in these studies hopefully leading to future studies. Finally, these studies have the potential to make a profound impact on the way scientists and psychologists understand cognitive evolution.
This paper will focus on two contemporary studies conducted within the past five years. Both studies were conducted on rhesus monkeys, with researchers focusing their attention on the presence and extent to which the monkeys displayed metacognitive skills. The first study, Hampton et al. (2004) helped set the standard as to whether or not a basic form of metacognitive ability, meta-awareness is apparent in monkeys. Here, meta-awareness will be understood as the capacity to be aware of the knowledge that one does or does not know. The second study, Terrace et al. (2006) expands on the findings of the first study and investigates whether or not monkeys were able to display more complicated forms of metacognition. Using these two studies as a foundation for further information, this paper will engage the idea of metacognition within rhesus monkeys and discuss its presence or lack thereof. Both Hampton et al. (2004) and Terrace et al. (2006) developed specific tests and apparatuses in order to gauge monkey ability in memory awareness, judgment making, and the creation of plans to gauge their own knowledge.
THEORIES AND HYPOTHESIS
Crucial to the study of metacognition is the debate of whether this skill is uniquely human or to what extent non-human animals possess this skill. Endel Tulving, a prominent psychologist, as well as many others believed that metacognition was simply a skill that was unique to humans. In the 1950s Joseph Hart, among others, began conducting experiments concerning knowledge of one’s own knowledge. They began coining terms such as ease of learning (EOL), judgment of learning (JOL), and feeling of knowing (FOK). Ease of knowledge refers to the ability to determine before exposure how easy it will be to learn a specific task or piece of information. Judgment of learning occurs either during or after learning and is a judgment as to whether or not learning should continue or if the information is sufficiently encoded. Finally, feeling of knowing refers to the rate at which individuals expect and/or have an inclination that they remember a specific task or piece of information (Endel and Stephen, 1970). These three terms reflect the three stages of metacognitive memory, mainly acquisition, or the attaining of the original knowledge, retention, or the maintenance of that knowledge, and retrieval, or the reproduction of that knowledge at a different time (Endel and Stephen, 1970).
In metacognitive studies a further element is added, where the confidence in one’s retrieval is also typically assessed. One effect that metacognition has proven is that human memory must be understood as partially a result of an inference that is based on some aspect of remembering, rather than total recall (Hampton et al. 2004). In many metacognitive experiments, the issue of confidence allows experimenters to interpret whether or not specific information is being remembered and why or why not patients are able to be aware of that information. Tied to this idea of confidence is the idea of uncertainty. Metacognitive theory is based around the idea of being able to discuss an individual’s awareness and the degree to which they are able to remember or not remember specific tasks or information.
Yet because metacognition was originally tied to a verbal awareness of knowledge, it was thought that language was indeed a pre-requisite for metacognition. Newer, less traditional tests brought a sense of creativity to the laboratory and allowed experiments to observe younger patients as well as animals (Hampton 2001). These experiments hypothesize that metacognition was not limited to humans and that other animals would be able to display metacognitive skills, if given the opportunity to. Many of the ensuing studies, such as Gallup’s “rouge test” focused primarily on self-awareness, or the idea of simply knowing about oneself. As self awareness in apes became evident, researchers have pushed the idea of metacognition further, in hopes of expanding their hypothesis about the topic.
Essential to the two studies described above, are further specific concepts and theories found within cognitive psychology. Regarding the Hampton et al. (2004) study the concepts of memory awareness and cost/benefit analysis play a large role. Similarly in the Terrace et al. (2006) study, the particulars of ‘monitoring’ and ‘control’ require a fuller understanding. From a psychological standpoint, awareness is understood as the ability to cognitively react to a given stimuli. Regarding the study, awareness is essential in understanding whether monkeys are able to discriminate between knowing and not knowing and whether they gather information selectively. This awareness is linked to working memory; as the article explains these “studies specify what an animal with memory awareness should be able to do, not what that animal should experience subjectively (Terrace et al. 2006).” Again, this awareness had to be inferred through behavior as verbal communication was obviously impossible. Similarly, as is the case with most experiments involving animals, researchers had to carefully balance the cost/benefit paradigm. This model emphasizes that in order for a specific behavior to occur the benefit that will be achieved far outweighs the risk or cost of committing such a behavior. In the cases of these two studies, experimenters had to hypothesize and formulate ways that will allow them to assess and reward the rhesus monkeys without shifting the cost/benefit paradigm too far in a single direction. Furthermore, in regards to monitoring and control, researchers hoped to prove that language was not a requisite to such a skill (Terrace et al. 2006). They anticipated ways for the monkeys to acquire learning methods, much like humans do. The researchers’ understanding of these concepts in concordance with one another helped them shape the hypothesis that rhesus monkeys will display metacognitive abilities and that they will be able to transfer those abilities to further, unpracticed tasks.
METHOD AND RESULTS:
As these two studies can and should be understood as being an extension of the other, the methods and results of the first study will be discussed followed by the methods and results of the second study:
Hampton et al. (2004)
In this study seven male and two female monkeys with previous training in experiments were taught to choose among four opaque tubes on the basis of which one contained a food reward. A special apparatus was constructed that positioned these four tubes in front of the monkeys. The monkeys were separated from the apparatus by a set of screens (a clear one and an opaque one) as well as a visual inhibitor that ensured that the monkeys could not clearly see into all of the tubes at all times. At any time a monkey reached for a specific tube, the others were automatically locked into position. The experiment contained three training phases as well as the main task. During phase one, familiarization, the monkeys were presented with clear tubes and taught to reach for the tube with the reward. They were also familiarized to the system of clear and opaque screen dividers. Phase two introduced opaque tubes as well as the visual inhibitor. This phase required the monkeys to attune to experimenters baiting, or placing the reward in a given tube. Phase three instructed the monkeys to actively “look” inside the tube. A look was defined as any time a monkey deliberately moved its line of sight toward the tubes. The main task incorporated all of the learning phases and tested monkey ability to both know the placement of food or to actively seek out more information if they did not know. In a set of trials the tube containing the reward varied randomly. Half of the trials monkeys were able to observe an experimenter baiting the tube (seen trials) while the other half the monkeys’ view was blocked from baiting (unseen trials). Monkeys were allowed one chance to select the reward tube; however they were allowed to inspect the tubes to see if they had food. As explained, looking involved considerable effort on the monkeys’ part, and as such the apparatus was moved toward or away from the monkey’s line of sight to balance the benefit/cost paradigm.
During the main task seven of the monkeys looked significantly more at unseen trials. Similarly, no monkey looked significantly more during seen trials rather than unseen trials. Aborted trials (trials that the monkey did not choose within a given timeframe) were rare but happened far more for unseen trials than seen trials. Monkeys chose accurately but not perfectly on seen trials (Fig. 2) and unseen accuracy was far greater when they looked. Because looking patterns changed between knowing and not knowing depending on whether or not the trials were seen or unseen researchers claim that this shows that these monkeys are aware of what they know (Table 2). These results further indicate that monkeys do in fact possess a certain characteristic of metacognition, mainly the awareness of what knowledge they do or do not know.
Terrace et al. (2006)
In this experiment monkeys were trained to make retrospective (looking back on knowledge/ behavior) judgments of their accuracy on perceptual tasks. Based on these judgments they were differentially rewarded for either high or low confidence decisions. Reward was greatest when monkeys chose high confidence and were correct. A lesser reward was given during low confidence trials.
This study contained a set of experiments that first taught rhesus monkeys to learn basic perceptual tasks and to then reflect on these tasks by making metacognitive decisions about their choices. The experiment was conducted on two male rhesus monkeys who had previously conducted experiments. The training apparatus included chambers with both visual and auditory responses that helped attune the monkey toward specific choices they made. The first two tasks they were asked to complete was a line discrimination test (which line is longest) and a numerical discrimination test (which shape has the most features). After these basic tasks were completed monkeys were asked to retrospectively judge how confident they were in their decisions. Two icons appeared on a screen, one that represented high confidence and one that represented low confidence. If a monkey chose the high confidence option he was rewarded or punished by the addition or removal of three tokens. Choice of low confidence included an addition of one token. Once a specific amount of tokens were collected, the monkey was rewarded with a preferred reward. This experiment alleges to be one of the first in which monkeys in a token society faced the option of losing tokens. After these more basic tasks were completed, monkeys were then presented with a series of photographs, which is referred to as a serial-working memory task. After being presented with the photographs, the monkeys used a touch screen to showcase the order in which they were originally represented and were given the same high versus low confidence icons to choose from. Monkeys excelled at both line and numerical discrimination but had certain difficulties remembering during the serial working memory task. To judge whether or not monkeys would be able to make judgments on how well they remembered the serial working memory task, experimenters added an element of hint giving to the monkeys. Rather than using trial and error monkeys could request hints from a monitor that pointed to the correct picture that they were to choose. Monkeys were only rewarded after a correct, complete trial; all failed attempts ended at any selection of a wrong sequence with no reward given. Additionally, the requesting of hints brought a lesser reward, while a completed sequence without the assistance of hints provided an optimal reward.
Results indicate that hint requesting was far less requested on a familiar list and than on unfamiliar list (figure 5). These results seem to imply that the monkeys were able to use their recently learned skills and apply them to novel situations. Monkeys began performing reasonably well during the serial working memory task with the amount of requests of hints decreasing declining significantly from trial to trial. Researchers point to the decreased reaction time and the increase of accuracy during these trials to support these findings.
The results from the first study (Hampton et al. 2004) seem to indicate both a presence and more importantly an awareness of short-term memory among rhesus monkeys. The observed monkeys discriminated between seen and unseen trials; however, there were errors among the results, with monkeys opting to occasionally look during seen trials. Researchers explain this trend by pointing to a weak memory among the monkeys or a lack of attention during baiting. Yet even among these seen trials that monkeys chose to look, the monkeys very often looked first at the tube that indeed contained the reward, signaling a feeling of knowing (FOK) (Hampton et al. 2004). Nevertheless researchers claim that these results reflect many possible outcomes. Further research is required to determine whether monkeys have weak memory systems, they have a natural conservative decision making systems, or that they do demonstrate characteristics of an implicit memory (Hampton et al 2004).
The important findings of Hampton et al. (2004) do in fact conclude that humans and the great apes are not the sole possessors of metacognition. This study paves the way in determining which ancestor occupied these skills and at what point evolution caused these different species to retain this ability. However, this study was careful in limiting the extent of metacognition displayed within the rhesus monkeys. Researchers were able to identify memory awareness through the specific tasks required of the monkeys but point to other studies such as Gallup (1994) rouge test which old world monkeys such as the rhesus monkeys subsequently failed. This has led the Hampton et al. (2004) researchers to conclude that the memory awareness showcased in their study serves as one of the many elements that constitute complete self-awareness. Whether or not these elements can exist independently of one another remains unresolved, and further research should be aimed at pinpointing how aspects of metacognition developed as well as focusing on whether monkeys display other metacognitive abilities. These future studies could help researchers better understand how humans and non-human animals are both similar and dissimilar, particularly concerning learning patterns and judgments regarding their knowledge.
One such study, Terrace et al. (2006) aimed at investigating these questions and largely built off of the findings found in Hampton et al. (2004). Part of the Terrace et al. (2006) study targeted whether language was a prerequisite of metacognitive ability. Because metacognition has been hard to determine among humans under the age of four, researchers had speculated that these abilities were intrinsically linked with the ability to use language as an expression of what one does and does not know (Tulving and Madigan 1970). However, this experiment has shown how direct observation of behavior can suggest the presence of metacognitive skills (Terrace et al. 2006). The rhesus monkeys’ decisions were based primarily on previous memory and not external stimuli, as some objections to the work have claimed. This assertion is underscored by monkey ability to transfer a skill (making high and low confidence judgments) toward a novel task (serial working memory task). Rather than relying on external memory the monkeys found ways to correct their memory deficiencies by requesting hints (Terrace et al. 2006) enabling them to make more accurate judgments. In metacognitive terms, this displays a level of control that was previously thought to be beyond the capability of old world monkeys.
In this regard, Terrace and colleagues concluded that these metacognitive abilities are practically analogous to human ability; however, they also concede that it is harder to prove that these animals are aware of their abilities. Rather than possessing the full range of metacognitive skills, it is now believed that these monkeys retain the ability to unconsciously make metacognitive judgments (Terrace et al. 2006). Even more impressive is that these animals are able to think and to think about their thinking without the luxury of language. Yet the explanation for this ability remains unresolved; further research is imperative in understanding how this occurs.
As researchers focus more of their efforts in this sphere, the evolution of metacognition will become clearer (Terrace et al. 2006). One must understand these two studies as a progression of the understanding of metacognitive abilities within more primitive types of non-human animals. Hampton et al. (2004) allowed researchers to validate their hypothesis of the existence of some semblance of metacognition within rhesus monkeys. Meanwhile Terrace et al. (2006) extended this hypothesis showing that these same animals displayed judgment systems that allowed them to determine how to acquire more information. It becomes necessary to continue challenging the results of these studies—broadening them to see the full extent of their findings. Are these metacognitive abilities only observable in laboratory settings? Will other animals display similar skills? Is it possible for different species to show different elements of metacognition? The Hampton et al. (2004) and Terrace et al. (2006) findings have paved the road for further research. Subsequent work will hopefully continue to improve this understanding of metacognition among researchers. Similarly, these two studies as well as future ones will enable researchers to better understand how and why animals both act and choose to learn new information in the ways that they do.
Figures and Graphs
Call, Joseph and Carpenter, Malinda 2001. Do Apes and Children Know What They Have Seen? Animal Cognition 4:207-220
Call, Joseph, Tomasello, Michael, and Gluckman, Andrea. 1997 Comprehension of Novel Communicative Signs by Apes and Human Children. Blackwell Publishing. Vol. 68, No. 6 (Dec., 1997), pp. 1067-1080
Hampton, Robert R. 2001. Rhesus Monkeys Know When They Remember. Proceedings of the National Academy of Sciences of the United States of America. 98:5359-5362
Hampton, Robert R, Zivin, Aaron, and Murray, Elisabeth A. 2004. Rhesus Monkeys (Macaca Mulatta) Discriminate between knowing and not knowing and collect information as needed before acting. Animal Cognition (2004) 7:239-246. Springer-Verlag 2004
Gallup, Gordon G. 1994. Self recognition: Research Strategies and Experimental Design. In: Parker ST, Mitchell RW, Boccia ML (eds) Self-Awareness in Animals and Humans. Cambridge University Press, New York, pp 35-50
Livingston, Jennifer A. 1997. Metacognition: An Overview. http://www.gse.buffalo.edu/fas/shuell/CEP564/Metacog.htm
Smith, J. David and Shields, Wendy E. 1997. The uncertain response in humans and animals. Cognition Volume 62, Issue 1, 2 February 1997, Pages 75-97
Terrace, Herbert S., Kornell, Nate, and Son, Lisa K. 2006. Transfer of Metacognitive Skills and Hint Seeking in Monkeys. Psychological Science: Volume 18 Number 1
Tulving, Endel and Madigan, Stephen A. 1970. Memory and Verbal Learning. Annual Review of Psychology Vol. 21: 437-484