A Brief Survey on Animal Behavior Studies

This blog post offers a brief survey of two questions:

What. What behaviors — and more importantly, what cognitive skills — are observed in which species?
How. How are these cognitive skills conceptualized and tested, and what do the criticisms say?

My intuition of the comparative psychology method:

Animals can’t talk. So you test their cognitive skills only through their behavior, but you don’t know how animals emit their behavior because they can’t explain and they just act. There’ve been rounds of criticism and rebuttal over alternative hypotheses in which a different (often lower-level) cognitive skill can predict the same behavior.
Ecological validity matters because animals don’t show cognitive skills even if they have them without the right ecological validity.
Action matters. Attention (gaze) matters as well.
Inclination can be tested through accepting negative reward - give up food, pay efforts, etc.

Social Skills

Understand Others Are Perceptual

References

Five primate species follow the visual gaze of conspecifics. Animal Behaviour (1998)

Chimpanzees know what conspecifics do and do not see. Animal Behaviour (2000)

Do chimpanzees know what conspecifics know? Animal Behaviour (2001)

Chimpanzees understand psychological states – the question is which ones and to what extent. Trends in Cognitive Sciences (2003)

Chimpanzee minds: suspiciously human? Trends in Cognitive Sciences (2003)

Chimpanzees versus humans: it’s not that simple. Trends in Cognitive Sciences (2003)

The goggles experiment: can chimpanzees use self-experience to infer what a competitor can see? Animal Behaviour (2015)

Ravens attribute visual access to unseen competitors. Nature Communications (2016)

Chimpanzees gesture to humans in mirrors: using reflection to dissociate seeing from line of gaze. Animal Behaviour (2018)

Cognitive skill. Can animal understand others act according to their perceptual state?

The interesting points. Alternative hypotheses are interesting. (1) Following others’ gaze. Chimps may just learn a behavior-rule: someone faces there, there’s something interesting. (2) Pass the goggle test (stated later). Chimps may just encode a heuristic: when there’s an unobstructed line (barrier affords gaze line) between eyes and the object, behavior emits. I agree with Povinelli that these are different levels of “behavior abstraction” and with Tomasello that the key is understanding others act according to some conditions. Human (ToM) can attribute behaviors to their fundamental cause (psychological state). But even if a chimp only learns associative knowledge like “you face somewhere + something is there and your behavior follows that”, it is just learning a coarser (and wrong but useful) behavior abstraction. In the other words, what is the difference between the physical state of light shooting in eyes and the psychological state of seeing? Does it really have difference between understanding others behave according to that physical state and to that psychological state.

Test with competitive environment. Subordinate reacts to dominant and compete for food. Three interesting settings: (1) Dominant’s door is fully closed so can’t read its behavior. (2) Subordinate observes dominants’ gazing for food, so it is given chance to read behavior. (3) Does subordinate figure out which individual dominant sees food. In (1), subordinate prefers hidden food anyway. In (2), subordinate can’t locate food. In (3), subordinate ignores ignorant dominant. My takeaways: (1) Ecological validity matters. Use competitive environment because apes compete for food naturally. Subordinate does not need to pay attention to dominant when they don’t have food in the wild, so they don’t read behavior even given the chance. (2) Animals model others’ mental state because they react even when there’s no behavior to read.

Test with goggles. Two steps: (1) Learn a barrier is transparent or vague. (2) Predicting others’ behavior when they’re faced with the barrier. Everything is the same except for their personal experience with the barrier.

Test with mirror. Chimps react when human can see them through the mirror (solve the straight gaze hypothesis)?

Conclusion. (1) Ecological validity matters. (2) The behavior abstraction level is at least very close to understanding others are seeing and act according to what they see.

Understand Goal

References

Chimpanzees’ flexible targeted helping based on an understanding of conspecifics’ goals. PNAS (2012)

Great apes infer others’ goals based on context. Animal Cognition (2012)

Cognitive skill. Not only attend to others’ behavior, but also others’ goal, and predict others’ behavior accordingly.

Test with help. Chimps choose the tool from a tray of tools to help a human when requested. Chimp will choose the correct tool that best helps the human. When chimp and human are split by a opaque barrier, one chimp even stand up to look human’s behavior, and choose the right tool. Chimps have to understand the goal state of human to help correctly.

Test with behavior context. A human repeatedly fed chimpanzees grapes from bucket A, then stood up, walked to bucket B, and fed them there, so chimps learned to anticipate bucket B when she stood up. In test trials, her measured behavior was identical: she stopped feeding, looked toward bucket B, looked back, stood up, turned, crossed barriers, and paused. Only the prior context differed: no event, self-dropped clipboard, or external event such as walkie-talkie call, thrown clipboard, or another human calling her. Chimps, especially females, waited longer after external events, suggesting they inferred her goal had changed.

Difference between tests. In test 1, experimenter’s action is different and goal is inferred accordingly. In test 2, action sequences are identical, and goal is inferred according to behavior context.

Understand Belief

References

Great apes anticipate that other individuals will act according to false beliefs. Science (2016)

A test of the submentalizing hypothesis: apes’ performance in a false belief task with inanimate controls. Communicative & Integrative Biology (2017)

Great apes use self-experience to anticipate an agent’s action in a false-belief test. PNAS (2019)

Great apes distinguish true from false beliefs in an interactive helping task. PLOS ONE (2017)

How children come to understand false beliefs: a shared intentionality account. PNAS (2018)

Understand belief is understanding agents behave not with respect to reality but with respect to their belief about reality. The way to test is false belief understanding: when agents’ belief about the reality is false, can the subject who knows better about the reality tell how the agent will behave correctly (according to the false belief rather than to reality):

Implicit task. Anticipatory-looking / gaze task. A subject watches an agent hide a toy in Box A. While the agent is absent, the toy is moved to Box B. When the agent returns and starts to search, the subject’s gaze anticipates that the agent will look in Box A, where the agent falsely believes the toy is.
Classical task. Sally puts a marble in a basket and leaves. Anne moves the marble to a box. Sally returns. The child is asked: “Where will Sally look for the marble?” Passing means answering “in the basket,” because Sally falsely believes it is still there. Failing children usually answer “in the box,” where the marble really is.

Implicit false belief is not true false belief. Evidence is infants solve implicit one but can’t solve classical one until 4 years old. Cognitive skill of implicit task is simple epistemic tracking (what others perceive, know, want, etc) and predict others’ behavior accordingly; cognitive skill of classical task is perspective coordination between the agent’s subjective perspective, one’s own perspective, and the objective situation. A theory of children’s failure on the classical task at 3-y-old is they don’t understand two conflicting perspectives can be “true” about a single reality - they can’t coordinate perspectives, and are therefore pulled toward reality (their own perspective, actually).

Great apes, while watching videos, anticipatorily looked where an actor falsely believed a hidden object to be.

Alternative explanations are ruled out.

Submentalizing. For instance, apes simply memorize the spatial relationship between the green shirt and the object. However, when the green shirt is replaced by a green object, apes gaze randomly.

Behavior-rule account. For instance, apes simply predict “agents search where they last looked”. The experiment setting for rebuttal: object hidden in box 1 → actor goes behind barrier → object moved to box 2 → object removed entirely from the scene. Apes first experience the barrier by themselves, one group can see through the transparent barrier and another group can’t. The first group gazes randomly, and the second group gazes according to false belief. Therefore, apes at least reason about “belief-like states” — “where did the agent last encounter the object?”

Ignorance-based rule. Apes don’t just look in the right direction — they actively act on their understanding of another’s beliefs. Apes help agents to retrieve the hidden object. Apes do this not because they think the experimenter is ignorant (doesn’t know where the object is). Indeed, when the experimenter never saw the object placed in any box and thus he has no belief about its location, apes gaze randomly.

Vicarious Emotion Learning

References

Empathy and pro-social behavior in rats. Science (2011)

Pro-social behavior in rats is modulated by social experience. eLife (2014)

The roots of empathy: through the lens of rodent models. Neuroscience & Biobehavioral Reviews (2017)

What behavior. Rats free trapped cage mates and choose to help even when chocolate is available. Their helping is governed by strain familiarity rather than genetic relatedness: a rat will help strangers from the strain it grew up with, rather than a strain that is more closely related genetically.

Mechanism. Rats perceive a conspecific’s distress → this triggers affective arousal (shared emotional state) in the observer → that shared distress is aversive → which motivates prosocial action to end it. They don’t do so out of an intrinsic motivation to interact with others, because even when the freed cage mate is then sent into a separate room, rats will still go and free it anyway.

Cognitive skills. Vicarious emotion learning and affective arousal. Learning about danger by watching someone else get hurt is far safer than learning by getting hurt yourself. It points to a biologically ancient mechanism for the social transmission of survival-relevant information.

Inequity Aversion (IA)

References

Monkeys reject unequal pay. Nature (2003)

Inequity responses of monkeys modified by effort. PNAS (2007)

Evolution of responses to (un)fairness. Science (2014)

Social disappointment explains chimpanzees’ behaviour in the inequity aversion task. Proceedings of the Royal Society B (2017)

Young children, but not chimpanzees, are averse to disadvantageous and advantageous inequities. Journal of Experimental Child Psychology (2017)

No evidence for inequity aversion in non-human animals: a meta-analysis of accept/reject paradigms. Proceedings of the Royal Society B (2024)

The question is whether an animal will be frustrated by getting a higher (guilty) or lower (envy) reward than a con-species. Advantageous IA almost certainly exists only in humans. There’s no clear conclusion on whether disadvantageous IA exists in more species.

The key is whether animals in these scenarios will be increasingly frustrated: (1) they put in effort and get a low reward; (2) in addition, a high reward is visible; (3) in addition, a near-by and visible con-species get higher reward in the same situation. In the last scenario, two monkeys both have grapes and cucumbers in front of them, but the experimenter gives one grapes and the other cucumbers. A meta-analysis shows that from (1) - (2), apes are more frustrated, but from (2) - (3), no evidence. The downside of meta-analysis is ignorance of experiment context by pulling data together for statistical robustness.

In experiments with more control but lower statistical robustness, generally, three things are essential for IA behavior: (1) a partner is visible and near (2) the animal puts in effort for the reward and (3) the affinity between partners may increase tolerance towards inequity. (2) and (3) are interesting, because they may be the psychological foundation of cooperative behavior - I put in effort, so I want an equal reward, but if you’re my long-term partner, getting mad at you is costly.

Reputation And Mutualism

References

Biological markets: supply and demand determine the effect of partner choice in cooperation, mutualism and mating. Behavioral Ecology and Sociobiology (1994)

Biological markets. Trends in Ecology & Evolution (1995)

Image scoring and cooperation in a cleaner fish mutualism. Nature (2006)

Cleaner wrasses Labroides dimidiatus are more cooperative in the presence of an audience. Current Biology (2011)

Biological market. A is male and B is female. In an environment, only A and B exist, so B evolves to be ugly (energy-saving) because A chooses B anyway. In another environment, (A,B,B’) exist, so B and B’ both evolve to be beautiful (energy-costing) because A chooses from them and they have to compete as if in a market. Concepts like bargaining power and friction cost apply in the biological market.

Behavior of cleaner fish. Bystanders avoid cleaners whose clients jolt (indicating cheating), and cleaners spontaneously reduce cheating when a bystander is visible. Together they establish that clients evaluate cleaners by observation and cleaners strategically increase cooperation when watched — a non-human audience effect. The learning mechanism of “reputation” can be simple: bystander exists + I'm serving -> I cheat / don't cheat -> bystander leave / don't leave -> I get reward from bystander / not get.

Cooperation

References

Chimpanzees recruit the best collaborators. Science (2006)

Engineering cooperation in chimpanzees: tolerance constraints on cooperation. Animal Behaviour (2006)

Tolerance allows bonobos to outperform chimpanzees on a cooperative task. Current Biology (2007)

Elephants know when they need a helping trunk in a cooperative task. PNAS (2011)

No third-party punishment in chimpanzees. PNAS (2012)

Preschool children and chimpanzees incur costs to watch punishment of antisocial others. Nature Human Behaviour (2018)

Third-party punishment by preverbal infants. Nature Human Behaviour (2022)

Cooperation in a human-like sense. It involves a free choice or preference to work with others, not a forced situation, and a recognition that all collaborators — but not free riders — deserve their fair share of the spoils even if that means sacrificing one’s own resources. And when there is enforcement against noncooperators or freeloaders, it is done not for selfish motives—such as obtaining the food for oneself or maintaining dominance—but rather for the good of the cooperative group, ultimately preserving shared group norms.

What behavior. (1) Know that you need a cooperator. (2) Select the best cooperator (similar to biological market). (3) Second-party punishment - punish others when you’re hurt. (4) Third-party punishment - punish others when you are not hurt (norm-preserving). (5) Reward sharing.

Test with string pulling. Two must pull the string together to get food. Settings: (1) Simultaneous launch. Free two animals at the same time for the task. (2) Delayed launch. Free one and free another much later (e.g., 45 seconds). (3) Food dispersed or not. Dispersed food means hard to monopolize food and vice versa. (4) Competitor. Whether someone is pulling from the other side to compete for the food. These settings are motivated by: (1) Rule out accidental coordination where they happen to pursue reward at the same time. (2) How power dominance and tolerance on cofeeding influence cooperation success rate. (3) If competition encourages cooperation.

Test with food theft. Settings: (1) 2P theft. Food stolen and a con-species gets it. (2) 2P loss. Food lost and no one gets it. (3) 3P stolen. Witness one steals another’s food. (4) 3P loss. Witness food lost, but no one steals it. (5) 3P no victim. Witness one gets food from an empty cage - the food belongs to no one. Punishment enforcement: push a button to destroy the food, but the actor gains nothing - cheap punishment enforcement, but no reward. These settings are motivated by: (1) For 2P settings, see whether apes understand the difference or merely react because of frustration of loss. (2) For 3P settings, see if the existence of a norm violator matters. (3) For 3P victim, see if apes react merely because of frustration out of competitive jealousy by seeing others get reward.

Results. (1) Chimps know when they need to involve collaborators. (2) They choose collaborators that have a higher cooperation success rate. (3) Tolerance towards cofeeding increases cooperation rate. (4) Apes punish more in 2P theft than 2P loss, and no third-party punishment in apes.

One caveat: apes may pay to see punishment done, but don’t enact punishment themselves. Children gaze at norm violators at a very early age (8 months).

Self-recognition

References

Chimpanzees: self-recognition. Science (1970)

The nature of visual self-recognition. Trends in Cognitive Sciences (2013)

Self-recognition in animals: where do we stand 50 years later? Lessons from cleaner wrasse and other species. Psychology of Consciousness: Theory, Research, and Practice (2020)

Cleaner fish recognize self in a mirror via self-face recognition like humans. PNAS (2023)

On the mirror test and the evolutionary origin of self-awareness in vertebrates. Philosophical Transactions of the Royal Society B (2025)

Test with mirror. The question is can animals recognize “self”. The experiment step by step: (1) An animal co-live with a mirror for days. (2) The mirror is taken away and some mark is created on animal’s body. (3) The animal with marker is observed for some time as a baseline to see how animal respond to see how animal respond to marker (rule out dishabituation). (4) Mirror is back, and if animal responds to the marker more compared with animal that hasn’t been exposed to a mirror, then “pass the mirror test”.

Test with photograph. Alternative hypothesis is animal recognizes themselves at test-time using kinesthetic visual matching. So test with photograph with settings: (1) Self-photo with mark. (2) Self-photo without mark. (3) Photo of a familiar fish with mark.

Result on cleaner fish, interestingly. (1) Fish passes the mirror test with brown mark. (2) Fish responds specifically to self’s face but not self’s body. (3) Fish does not respond to a familiar fish with mark, so it is recognizing itself rather than a familiar con-species. (4) Passing the photograph task means fish preserve a mental image of self when co-living with the mirror. They don’t just recognize themselves throught test-time kinesthetic visual matching.

Problem with false-negative. Cleaner fish does not pass the mirror test with ecologically irrelevant blue or green marks (same material, same size, same location). Brown marks resembling ectoparasites that they want to remove naturally. Therefore, failures may simply reflect lack of motivation to investigate the mark. In this way, negative results on most species are not convincing.

Imitation Learning In General

References

Imitation as behaviour parsing. Philosophical Transactions of the Royal Society B (2003)

Causal knowledge and imitation/emulation switching in chimpanzees and children. Animal Cognition (2005)

Conformity to cultural norms of tool use in chimpanzees. Nature (2005)

Observational learning in chimpanzees and children studied through ‘ghost’ conditions. Proceedings of the Royal Society B (2008)

Ratcheting up the ratchet: on the evolution of cumulative culture. Philosophical Transactions of the Royal Society B (2009)

Associative sequence learning: the role of experience in the development of imitation and the mirror system. Philosophical Transactions of the Royal Society B (2009)

Identification of the social and cognitive processes underlying human cumulative culture. Science (2012)

The importance of witnessed agency in chimpanzee social learning of tool use. Behavioural Processes (2015)

‘Over-imitation’: a review and appraisal of a decade of research. Developmental Review (2019)

Population connectivity shapes the distribution and complexity of chimpanzee cumulative culture. Science (2024)

Forms of imitation learning, the first three indicate low-fidelity copy:

Emulation. Learns the result or affordance, e.g. “this door opens”. Object-centric.
Stimulus enhancement. Draws the observer’s attention to a location or object, making individual discovery more likely.
Response facilitation. Increases the probability of performing an action already in the observer’s repertoire. Frequent substitution of familiar motor patterns.
(Over) imitation learning. Copies the demonstrator’s specific action.

Test with two-action method. Two ways can achieve the same goal but only one way is demonstrated. Related with social learning helps convention formation: a goal can be achieved arbitrarily but is achieved in a certain way regardlessly.

Test with ghost string. No demonstrator exists and the action is conducted using an invisible string. Test emulation.

Over-imitation and transmission fidelity. Children copies demonstrator’s exact actions even when they know some actions are unnecessary through observing the causality between action and reward, while apes don’t. That is, children over-imitate. But over-imitation can be modulated by the degree to which learning goals are activated (e.g., by prior experience with efficient solutions) and the degree of social motivation or pressure to comply. It is relavent with transmission fidelity (ratchet effect or cumulative culture): humans achieve collective intelligence and show social conventions without reinventing the wheels, and one generation’s innovations are faithfully preserved until further improvements ratchet things up again.

Associative Sequence Learning (ASL). Says imitation is learned through ordinary sensorimotor association: repeated correlated experience of seeing an action and doing that action builds visual–motor links. Later, seeing the action activates the corresponding motor program, enabling imitation, including novel sequences built from familiar components. My opinion: ASL explains the capacity/mechanism for imitation, but does not fully explain the motivation or social function of imitation.

Conclusion. Social learning is often analyzed as instrumental and payoff-oriented for non-human primates but sometimes as social task for children. Human children are unusual not because they socially learn, but because they often treat demonstrated actions as socially/normatively meaningful even when those actions are instrumentally useless. Orient the task towards instrumental learning can reduce over-imitation.

Physical Cognitions

Inhibition

References

Development of the ability to use recall to guide action, as indicated by infants’ performance on AB. Child Development (1985)

Object and spatial representations in detour problems by chicks. Animal Behaviour (1995)

The evolution of self-control. PNAS (2014)

Can be tested with three tasks:

Detour Task. The animal is usually first familiarized with getting a reward directly. Then a barrier is introduced, and the animal must learn or infer that it has to move around the barrier rather than persist in the direct path.
Cylinder Task. The animal is first trained with an opaque cylinder, where food can only be obtained by going to the open ends. In the test phase, the cylinder becomes transparent: the animal can see the food through the side but must inhibit the direct reach and use the previously learned side-opening response.
A-not-B Task. The subject is repeatedly trained to find a hidden object at location A. Then the object is visibly moved to location B, and the subject must inhibit the previously reinforced search at A and search at B instead.
Middle-cup task. Three cups; in experimental trials, both outer cups are baited and the middle is empty. After retrieving from one outer cup, the subject must skip the adjacent empty middle cup and go to the other outer cup.

Behavior. Suppress a prepotent or dominant response. In the detour task, the animal must inhibit the tendency to move straight toward a visible reward and instead go around a barrier. In the cylinder task, after being trained to retrieve food from the open ends of an opaque cylinder, the animal is tested with a transparent cylinder and must inhibit the impulse to reach directly through the visible wall. In the A-not-B task, the subject must suppress the previously reinforced tendency to search at location A and instead search at the new location B. In the middle-cup task, it tests inhibition of reaching toward a nearby cup after just finding food. Thus, all four tasks involve inhibiting an automatic or learned response in favor of a less direct but correct action.

Delayed Gratification

References

Delay of gratification in children. Science (1989)

A hot/cool-system analysis of delay of gratification: dynamics of willpower. Psychological Review (1999)

The evolutionary origins of human patience: temporal preferences in chimpanzees, bonobos, and human adults. Current Biology (2007)

Fission-fusion dynamics, behavioral flexibility, and inhibitory control in primates. Current Biology (2008)

Evolutionary pressures on primate intertemporal choice. Proceedings of the Royal Society B (2014)

Cuttlefish exert self-control in a delay of gratification task. Proceedings of the Royal Society B (2021)

Behavior. A smaller immediate food reward versus a larger delayed one. The delay is adjusted across sessions until the subject reaches an indifference point. Tests temporal discounting and impulse control.

Evolution. (1) In non-human primates, fission-fusion (FF) dynamics best correlates with delayed gratification and inhibition tasks. High FF indicates unstable group context, which requires constantly tracking who is present, who is dominant, and adjusting behavior accordingly should demand strong suppression of prepotent but contextually inappropriate responses. (2) Ecological pressure can be the need for effective ambush predation. For instance, cuttlefish, who is short-lived, shows delayed gratification.

Self-distraction. Mainly observed in children delayed gratification study. With uncovered high reward, children should come up with strategies to deley gratification. They distract attention away from the reward - covering their eyes, resting their heads, singing, talking to themselves, inventing games with their hands and feet, even trying to sleep. These children are academically and socially competent later.

Behavior Flexibility

References

Serial reversal learning and the evolution of behavioral flexibility in three species of North American corvids. Journal of Comparative Psychology (2007)

Problem-solving skills are predicted by technical innovations in the wild and brain size in passerines. Nature Ecology & Evolution (2024)

Two types of behavior Flexibility:

Reversal learning. An animal first learns that one option is rewarded, for example “choose yellow, not green.” After it reaches a learning criterion, the reward rule is reversed: now green is rewarded and yellow is not.
Innovation. A puzzle box with several independent ways to get the same reward. After an animal solves one route, that route is blocked, so it must discover another solution.

MAB is closer to innovation/problem-solving under changing affordances, whereas reversal learning is closer to inhibition and associative updating under changed reward contingencies.

Tool-Use

References

Manufacture and use of hook-tools by New Caledonian crows. Nature (1996)

Tool-related cognition in New Caledonian crows. Comparative Cognition & Behavior Reviews (2007)

Insightful problem solving and creative tool modification by captive nontool-using rooks. PNAS (2009)

Animal tool-use. Current Biology (2010)

New Caledonian crows reason about hidden causal agents. PNAS (2012)

Did tool-use evolve with enhanced physical cognitive abilities? Philosophical Transactions of the Royal Society B (2013)

Behavior. New Caledonian crows show extraordinary behavioural competence in tool-related domains—tool selection, tool manufacture, flexible tool modification, and sequential/meta-tool use. The competence is clear; the mechanism is not. The data rule out the simplest accounts—pure imitation, purely random object manipulation, or a narrow reinforced action sequence — but researchers do not identify a specific cognitive skill such as causal reasoning, planning, or theory-like physical understanding.

Test:

Tool selection. Food inside transparent tube / hole. Tools differ in length or diameter. Crow must choose long-enough tool, or thin-enough tool, to reach/dislodge food.
Tool manufacture. Crow gets raw material: twigs, branches, Pandanus leaves, cardboard, etc. Food hidden in hole/crevice/tube. Crow must make a usable tool before extraction.
Flexible tool modification. Food bucket in vertical tube/well. Straight wire or aluminium strip initially useless. Crow must bend into hook; in variants, unbend, flatten, narrow, or lengthen tool depending on task.
Sequential / meta-tool use. Food reachable only with long tool. Long tool itself out of reach. Crow uses short tool to get long tool, then long tool to get food. Longer versions add more intermediate tools.

Numerosity

References

Use of numbers by a chimpanzee. Nature (1985)

Ordering of the numerosities 1 to 9 by monkeys. Science (1998)

From “sense of number” to “sense of magnitude”: the role of continuous magnitudes in numerical cognition. Behavioral and Brain Sciences (2017)

Research question:

Do organisms perceive numerosity directly, like color or size, before language, symbols, education, or training?
Can an organism acquire numerical/symbolic representations through training, language, or abstraction?

Research on non-human primates positively answered the second question. There’s still debate on the first question - number may be a learned abstraction rather than an innate perceptual primitive: organisms first sense continuous magnitudes such as more/less, larger/smaller, denser/sparser; then object individuation lets them treat a scene as containing discrete units; language and symbols stabilize equivalences such as “three apples,” “three dots,” and “three sounds”; and cognitive control lets them ignore misleading cues, such as when three large objects occupy more area than five small objects.

Mean-End Understanding

References

The string-pulling paradigm in comparative psychology. Journal of Comparative Psychology (2015)

Cognitive skill. Use an intermediate object/action as a means to achieve an end.

Test with string-pulling. Simplest form: the animal sees food attached to a string and can get it by pulling.

Lower-level strategies: (1) learn “pull → food moves closer”; (2) pull for play; (3) rely on immediate visual/proprioceptive feedback. Rule out lower-level strategies: (1) crossed or slanted strings test whether the animal follows the functional connection rather than choosing the string closest to the food; (2) contact/no-contact strings test whether it understands that the string must physically connect to the reward; (3) coiled strings test whether it can persist without immediate reward movement; (4) covered or visually ambiguous strings test reliance on visual continuity; (5) too-heavy or directly reachable rewards test flexibility rather than rigid pulling.

Object Permanence

References

Do animals understand invisible displacement? A critical review. Journal of Comparative Psychology (2014)

Reply to Jaakkola (2014): “do animals understand invisible displacement? A critical review”. Journal of Comparative Psychology (2015)

Visible/Invisible displacement task. Visible displacement: animal sees an object covered in a cup, and must choose where it is among cups. Invisible displacement: object moved while still hidden, and the subject must infer where it ended up.

Cognitive skill. (1) Object permanence: internal representation stands for an absent reality. (2) Secondary representation: action begins to be guided by an internally represented, unseen world rather than by the currently perceived scene.

Metacognition

References

Great apes and human children rationally monitor their decisions. Proceedings of the Royal Society B (2022)

Social cognition and metacognition in great apes: a theory. Animal Cognition (2023)

Metacognition in wild Japanese macaques: cost and stakes influencing information-seeking behavior. Animal Cognition (2024)

Metacognition in Japanese macaques: does impulsivity explain unnecessary looks in the tubes task? Animal Cognition (2024)

Metacognition in nonhuman primates: a review of current knowledge. Primates (2025)

Test with information-seeking. To check if the subject represent uncertainty. Settings: (1) Low uncertainty: the subject witnessed a human putting reward in an opaque tube; (2) High uncertainty: the subject can’t know which tube has the reward but knows one of them has. See if the subjects looks more into the tube in high uncertainty cases.

Test belief-revision. Animal first forms a choice/belief, then receives new evidence that conflicts with it. It can either stick with the original choice or re-check before final choice. Key measure: does conflict between old belief and new evidence trigger renewed information seeking?

Conclusion. Great apes and macaques show strong evidence of procedural metacognition: they seek information when ignorant, avoid or hesitate on uncertain trials, make confidence-like judgments, and sometimes re-check when new evidence conflicts with an earlier decision.

Some interesting behavior. (1) Monkeys still looked in obvious trials, and their first look was usually into the baited tube, meaning unnecessary looking was not simply forgetting, but like “double-check”. (2) Monkeys reduced looking when it is costly mainly in obvious trials, but keep looking in ambiguous trails despite high looking cost. (3) In ambiguous trials, monkeys often stopped searching once they had enough information. (4) Rarely chose the last tube after checking three empty tubes; they preferred direct visual confirmation to inference. (5) Large individual differences.

Common Lower-Level Factors

Persistence: how long the animal keeps trying despite failure. In puzzle tasks, a persistent animal may solve the task simply because it tries more.

Motor diversity: how many different actions the animal attempts, such as biting, pulling, pushing, lifting, scratching, or flipping. More motor diversity increases the chance of accidentally finding a solution.

Neophobia: fear or avoidance of novel objects/situations. A neophobic animal may fail not because it lacks cognition, but because it avoids the apparatus.

Stimulus salience: how noticeable or attractive a cue is. For example, a bright colour or strong smell may dominate attention, making learning easier or harder.

Reward history: the animal’s past experience with rewards. If one colour, location, or action was rewarded before, the animal may keep choosing it even after contingencies change.

Associative learning: knowledge that one cue predicts an outcome.