Our (not so) unique minds: metacognition in non-primates
Metacognition in Non-Primate Animals: A Capacity with an Unexpectedly Deep Phylogenetic Heritage or Simply Homoplasy?
Few academic discussions have evoked such debate as those of what separates our species, Homo sapiens, from other animals and, thus, defines us as human. Differences of mind and cognitive capacity are often of central importance in such discussions with several mental capacities being considered uniquely human, or at least functioning in a uniquely human way if shared with other animals (domain generality versus specificity for example) (Premack, 2007). One such capacity that has traditionally been considered uniquely human is metacognition (Metcalfe & Shimamura, cited in Kornell, 2009, p. 11), the ability to form a second order thought in considering a first order thought, or put simply ‘having knowledge about knowledge’ (Terrace & Son, 2009). Metacognition has also been philosophically tied to concepts of higher-order consciousness and as such has been important in shaping concepts of human sentience (Edelman & Seth, 2009; Smith, 2009).
Surprisingly, however, despite a wide research base in human psychology, relatively little work has been done, until recently, to assess the metacognitive capacity, or lack thereof, of other animals. Is our attribution of this capacity to ourselves as unique justified? A growing body of research would suggest not. The majority of this modest body of work has focused on the identification of metacognition in primates with some, albeit hotly disputed, success (Couchman, Coutinho, Beran, & Smith, 2010; Smith, 2009; Smith, Shields, & Washburn, 2003; Suda-King, 2008). In identifying metacognition in primates, we have, through attempts to identify what makes us unique, found an unexpected and interesting cognitive homology in our nearest phylogenetic relatives. This discovery raises several larger questions with wider ramifications for comparative psychology, animal cognition and our understanding of the evolution of the brain and cognition, questions that have only recently began to be explored in earnest by scientists (Smith, 2009; Terrace & Son, 2009). Is this ability truly unique to we primates? If it is not, how far back along the phylogenetic tree does it extend? Or if it is present in other animals, is it a case of homoplasy rather than true homology?
The following essay shall consider these questions through, firstly, an explanation of the theory of metacognition and how it is applied, followed by an overview of the few works to focus on metacognition in non-primate animals (NPAs). Problems with past work on animal metacognition shall then be explored before going on to consider the implications of metacognition research in more distantly related taxa and offering recommendations of organisms of focus for further metacognitive study, taking cues from both comparative psychology and neuroscience.
Knowledge about Knowledge
As stated in the introduction, metacognition refers to the ability of an individual to have knowledge of knowledge, or as Terrace & Son (2009) describe it, form a mental representation of a mental representation. Human beings routinely do this when we discuss our feelings, in self-ascribing our mental states we build a secondary representation of a primary representation (Arango-Munoz, 2010). Expressions of uncertainty, especially regarding memory or in relation to an anticipated task, are also considered metacognitive activities as the individual is required to assess the accuracy (secondary representation) of a memory (primary representation) and in turn answer with certainty or with uncertainty (Terrace & Son, 2009). Edelman & Seth (2009) have also stated that a link may exist between this process and self-consciousness, an important factor in investigations of sentience in animals.
Kornell (2009), in agreement with the philosophy of Arango-Munoz (2010) (at least at the gross level of division), further stated that metacognition can be divided into two functional categories, the first being monitoring, the second being control. The monitoring category of metacognitive function is where an individual undergoes the process of forming second-order representations, as described prior (Kornell, 2009). The control category, particularly in the case of certainty judgments of memory, sees the individual act on secondary representations that indicate uncertainty, in order to reduce this uncertainty, or withdraw from the task (Kornell, 2009). An example of active uncertainty reduction could be someone deciding to re-check an instruction manual after deciding that they could not accurately recall instructions. This active seeking of clarification has been observed in primates also in the form of hint requests (Kornell, 2009; Terrace & Son, 2009) but has unfortunately not been seen in NPAs.
Despite the lack of evidence for active uncertainty reduction in NPAs, an ability that could be considered a strong indicator of metacognition as we understand it in humans, a small body of evidence has began to accumulate that suggests that some animals have the capacity to withdraw from affirmative answering, by declaring uncertainty, should they consider a given task too difficult (Terrace & Son, 2009). This can be considered a metacognitive response, as it requires the organism to assess its certainty and respond accordingly. How discrimination and metamemory tasks have been used to measure this in NPAs shall be discussed below.
The Search for Metacognition in Non-Primate Animals
Thus far, work on assessing metacognition in NPAs has focused primarily on three vertebrate species; the bottlenosed dolphin, the rat and the pigeon. Each of the works based on these organisms has utilised discrimination tasks, either utilising metamemory or not, of controlled difficulty, in order to elicit responses of uncertainty should the subject be capable of them.
In one of the first attempts to assess NPAs for metacognition, Smith et al. (1995) attempted to assess whether dolphins differed greatly from humans in their capacity to opt-out of difficult trials in order to move on to tasks with which they had a greater chance of success. This was achieved through having the subjects (human or dolphin) listen to a tone that could be classified as either ‘high’ or ‘low’, though the tone itself could range in frequency anywhere between these two extremes (Smith, et al., 1995). A correct response would result in reward, an incorrect response would lead to a time out. A third response, the uncertain response (UR), which triggered a new trial with guaranteed success, was also offered. Smith et al. (1995) found that in this trial, both humans and dolphins readily gave an UR to difficult trials (middle range frequencies) in order to move onto trials in which they were more likely to succeed. This was considered a metacognitive response as the animal was able to assess its certainty before responding (Smith, 2009), thus demonstrating the ability to form a second-order mental construct.
Foote & Crystal (2007) attempted a similar experiment utilising a more common model organism, the rat. In this experiment rats were offered rewards for successfully discriminating between sounds that could be deemed ‘long’ or ‘short’ (Foote & Crystal, 2007). If the animals selected the correct sound length, they were rewarded with six food pellets, while an incorrect answer yielded none. Foote & Crystal (2007) also offered the same third option seen in Smith et al. (1995), that of uncertainty, opting to end the trial for a smaller reward of three pellets. As with the dolphin trial, Foote & Crystal (2007) found that rats were also much more likely to opt for an UR in the case of difficult trials. These findings were further bolstered by the fact that Foote & Crystal (2007) also predicted, and found, that should an UR not be offered, accuracy during forced answers to difficult trials should be much lower than when an UR is offered. This gave weight to the conclusion of Foote & Crystal (2007), that rats were aware of their own cognitive state in a manner consistent with theories of metacognition, as if they were offered a chance to avoid high-risk, high-difficulty trials, they were likely to do so in order to increase their overall success.
Two notable pigeon based metacognition, or more specifically metamemory, experiments have been carried out, both reaching similar conclusions. The first, by Inman & Shettleworth (1999), set out to formulate a useful method of assessment of nonverbal animals for metamemory, using pigeons as a model. In this trial, pigeons were shown visual stimuli before being given the opportunity to identify these among distractors some time after initial exposure for a food reward (Inman & Shettleworth, 1999). While they appeared to behave in a way consistent with metamemory, and thus metacognition, in their selection of the ‘safe’ option (similar to the UR in other trials) in relation to how much time had elapsed since first being shown the stimuli, they only did so after being shown the stimuli and distractors. Interestingly, the pigeons failed to commit to a trial or opt out in a way suggestive of metacognition if given this option before being shown the stimuli and distractors (Inman & Shettleworth, 1999). Inman & Shettleworth (1999) concluded that this inability to assess the accuracy of memory in anticipation of a trial was inconsistent with metacognition (the reasoning behind this shall be discussed below). Sutton & Shettleworth (2008) retested the assumptions of Inman & Shettleworth (1999) in the light of more recent and refined concepts of metacognition in non-humans and reached the same conclusion, that pigeons do not display metacognition.
Metacognition in NPAs: Problems, Perspectives and Prospects
As can be seen, two broadly different approaches have been taken in attempting to identify metacognition in NPAs. Firstly, as is seen in Smith et al. (1995) and Foote & Crystal (2007), discrimination trials that do not rely on memory have been used. Secondly, trials based on metamemory tasks have also been used, as seen in the pigeon based experiments of Inman & Shettleworth (1999) and Sutton & Shuttleworth (2008). This division between discrimination based tasks in the presence of the focal stimuli, termed ‘concurrent’ trials by Terrace & Son (2009), and metamemory trials based on assessment of memory accuracy, corresponds with an important schism in how metacognition is defined and applied to animal behaviour.
As discussed, both Smith et al. (1995) and Foote & Crystal (2007) have argued that the results of their concurrent discrimination trials are indicative of animal metacognition and awareness. Terrace & Son (2009) opposed this by arguing that the results of these experiments were at best ambiguous for the reason that the task was performed in the presence of the stimuli, allowing for simpler non-metacognitive explanations. This concern was also expressed by Inman & Shettleworth (1999) in reference to the dolphin experiments of Smith et al. (1995).
The prime non-metacognitive process put forward as a plausible explanation for the results seen in the concurrent trials, was explained by Kornell (2009) in what was termed the ‘third response problem’. Kornell (2009) stated that it is possible that the animals under trial learned to associate middle-range stimuli with the third response, rendering it simply a conditioned choice made in order to obtain a reward rather than a true UR indicative of metacognition and thus high-order consciousness. If this was the case, the behaviour observed neither proved nor disproved the presence of metacognition. In order to be more certain, Terrace & Son (2009) advised that such experiments should always attempt to assess retrospective or prospective metamemory (prospective metamemory being the task on which the pigeons failed), as they are better indicators of true metacognition than concurrent URs.
In more recent work, Smith (2009), contra Kornell (2009), Terrace & Son (2009) and the earlier opinions of Inman & Shettleworth (1999) asserts that the making of URs, is per se, evidence of metacognition. While Smith (2009) does not dispute that conditioning of the sort proposed above could occur, he does assert that the observed results, as a function of parsimony, are most likely true URs indicative of consciousness without necessitating the kinds of metamemory trials utilised by others.
Regardless of the technicalities of definition, it is evident that, whether partially or in full, some cognitive faculties we would have traditionally attributed only to ourselves, or to our nearest primate relatives, are present in some form in other animals of relatively distant relatedness (mammals with millions of years of divergence from our line). Future trials could, perhaps, attempt both concurrent and metamemory based trials in order to determine whether metacognition is present in a more comprehensive fashion and in doing so allow us to form more concrete assertions about its evolutionary past.
Future Directions in NPA Focused Metacognition Research
The evidence available for metacognition in NPAs raises interesting questions about the evolutionary history of this ability that could perhaps be answered through further exploration of the capabilities of animals other than those already discussed. These could be other mammal species in order to discover whether this is a ubiquitous mammalian trait, or more distantly related organisms, to either further push back the ‘metacognitive family tree’, or identify where distantly related organisms have developed this capacity as a function of convergent evolution.
Despite the fact that the only positive evidence for metacognition has been found in mammals (Terrace & Son, 2009), recent research on similarities between the brain structure of such vastly different creatures as the marine ragworm and the mouse has suggested that many organisms share more neural commonalities than we would have traditionally admitted (Strausfeld, 2010; Tomer, Denes, Tessmar-Raible, & Arendt, 2010). Does this suggest that cognitive functions like metacognition may also be more deeply ingrained phylogenetically than we had suspected? Additionally, convergence is a realistic scenario here also, as metacognition is considered an adaptive tool for maximising efficiency of behaviour (Sutton & Shettleworth, 2008), rendering it entirely possible that it arose separately on more than one occasion. Two animal groups of use to future research focused on the answering of these questions are the Corvids, the family of birds to include the ravens and crows, and molluscs of the order Octopoda.
The Corvids, in particular the ravens and crows, are well known for feats of memory and problem solving in lab settings (Edelman & Seth, 2009). Some groups of crows have even been observed to actively engage in tool construction in the wild (Edelman & Seth, 2009), another behaviour that has been considered uniquely human in the past. Tool production implies the ability to plan actions, a trait that along with metacognition is associated with high-order consciousness and sentience (Edelman & Seth, 2009). Despite the failure of the pigeon, another avian, to perform successfully in metacognitive tasks as defined by Inman & Shettleworth (1999) (though Smith, 2009, may disagree), Corvids may perform better. If the metamemory based definition of metacognition is to be followed this would indicate a positive case of metacognition occurring as a result of convergence (conversely, if the per se argument of Smith, 2009, were taken it could indicate a deep phylogenetic heritage). Should Corvids fail also, it would go some of the way to suggesting an absence of metacognitive capacity in birds.
Molluscs of the order Octopoda, the octopi, are also potentially good models for studies of metacognition. As with the Corvids, these animals, particularly the species Octopus vulgaris, have demonstrated incredible adaptability, problem solving capacity and memory in lab settings (Edelman & Seth, 2009). Additionally, this species has shown a surprising capacity for observational learning and retention, demonstrating a greater affinity for learning from fellows than through conditioning (Fiorito & Scotto, 1992). Should metacognition be present in Octopus vulgaris, explanations other than convergence would be extremely difficult to propose given the exceptional phylogenetic distance of this species from the vertebrates studied so far.
In our attempts to identify what makes us unique, especially cognitively, we have discovered numerous, surprising functional analogues in animals. While the majority of these have been found in our nearest primate relatives, one cognitive capacity, that of ‘having knowledge about knowledge’, metacognition, has been identified, at least tentatively, in other mammal species of quite distant relation to ourselves. Unfortunately, however, due to differences of definition of behaviour indicative of metacognition, and a sheer lack of data, we are unable to readily explore the evolutionary history of this cognitive capacity. Should researchers come to some agreement on how to define this capacity, important as it is to our notions of sentience as they stand, perhaps future investigations of other, more distantly related, animals will yield some greater understanding of how widespread this faculty is and how, via homology or homoplasy, it came to be so.
Arango-Munoz, S. (2010). Two Levels of Metacognition. Philosophia, 39, 71-82.
Couchman, J. J., Coutinho, M. V. C., Beran, M. J., & Smith, D. J. (2010). Beyond Stimulus Cues and Reinforcement Signals: A New Approach to Animal Metacognition. Journal of Comparative Psychology, 124(4), 356-368.
Edelman, D. B., & Seth, A. K. (2009). Animal consciousness: a synthetic approach. Trends in Neurosciences, 32(9), 476-484.
Fiorito, G., & Scotto, P. (1992). Observational Learning in Octopus vulgaris. Science, 256, 545-547.
Foote, A. L., & Crystal, J. D. (2007). Metacognition in the Rat. Current Biology, 17(6), 551-555.
Inman, A., & Shettleworth, S. J. (1999). Detecting Metamemory in Nonverbal Subjects: A Test With Pigeons. Journal of Experimental Psychology, 25(3), 389-395.
Kornell, N. (2009). Metacognition in Humans and Animals. Current Directions in Psychological Science, 18(1), 11-15.
Leal, M., & Powell, B. (2011). Behavioural flexibility and problem solving in a tropical lizard. Biology Letters.
Metcalfe, J., & Shimamura, A. P. (1994). Metacognition: Knowing about knowing. Cambridge, MA: MIT Press.
Premack, D. (2007). Human and animal cognition: Continuity and discontinuity. Proceedings of the National Academy of Sciences, 104(35), 13861-13867.
Smith, J. D. (2009). The study of animal metacognition. Trends in Cognitive Sciences, 13(9), 389-396.
Smith, J. D., Schull, J., Strote, J., McGee, K., Egnor, R., & Erb, L. (1995). The Uncertain Response in the Bottlenosed Dolphin (Tursiops truncatus). Journal of Experimental Psychology, 124(4), 391-408.
Smith, J. D., Shields, W. E., & Washburn, D. A. (2003). The comparative psychology of uncertainty monitoring and metacognition. Behavioral and Brain Sciences, 26, 317-373.
Strausfeld, N. J. (2010). Brian Homology: Dohrn of a New Era? Brain, Behavior and Evolution, 76, 165-167.
Suda-King, C. (2008). Do orangutans (Pongo pygmaeus) know when they do not remember? Animal Cognition, 11, 21-42.
Sutton, J. E., & Shettleworth, S. J. (2008). Memory Without Awareness: Pigeons Do Not Show Metamemory in Delayed Matching to Sample. Journal of Experimental Psychology, 34(6), 266-282.
Terrace, H. S., & Son, L. K. (2009). Comparative metacognition. Current Opinion in Neurobiology, 19(1), 67-74.
Tomer, R., Denes, A. S., Tessmar-Raible, K., & Arendt, D. (2010). Profiling by Image Registration Reveals Common Origin of Annelid Mushroom Bodies and Vertebrate Pallium. Cell, 142(5), 800-809.
~ by confusedious on October 27, 2011.