Animals think, though not always in ways that resemble human thought. Decades of research across hundreds of species have documented problem-solving, self-awareness, planning for the future, deception, cultural learning, and even what looks like metacognition in creatures ranging from great apes and crows to octopuses and bees. The question scientists grapple with today is less “do they think?” and more “what kinds of thinking do they do, and how far does it go?”
Recognizing the Self in a Mirror
One of the oldest tests of animal cognition is the mirror test, first developed in 1970. The idea is straightforward: place a visible mark on an animal’s body that it can only see in a mirror, and watch whether it uses the reflection to investigate the mark on itself. If it does, the reasoning goes, the animal recognizes the reflection as its own body rather than as another individual. A review of more than 50 years of mirror self-recognition research, covering over 30 species, found that only social animals have consistently demonstrated this ability, while solitary species studied so far have not.1PubMed Central. Sociality and self-awareness in animals The list of species that pass includes great apes, elephants, dolphins, and magpies.
What surprised researchers was the addition of a tiny reef fish to that list. Cleaner wrasses, small fish that pick parasites off larger fish, passed a version of the mirror test and then went further. In a 2024 study, cleaner fish that had achieved mirror self-recognition were shown to accurately recall their own body size, choosing to attack photos of unfamiliar fish only when those fish were smaller than themselves. The researchers concluded that these fish had formed a mental image of their own bodies, making them the first non-human animal demonstrated to possess what the authors called “private self-awareness.”2PubMed Central. Cleaner fish with mirror self-recognition capacity precisely realize their body size based on their mental image That a fish with a brain smaller than a pea can do something so sophisticated rattled assumptions about what kind of nervous system self-awareness requires.
Solving Problems and Using Tools
Tool use was once considered a uniquely human trait, then a uniquely primate trait, and then a uniquely mammalian or avian trait. Each of those boundaries has fallen. New Caledonian crows, for instance, not only fashion stick tools to extract insects from bark but show evidence of causal reasoning: in one experiment, crows that learned to solve a two-trap-tube task transferred that knowledge immediately to a visually different trap-table task, suggesting they understood the underlying cause-and-effect relationship rather than simply memorizing a sequence of actions.3PubMed Central. Causal reasoning in New Caledonian crows: Ruling out spatial analogies and sampling error
Great apes go further still, using multiple tools in planned sequences. In controlled experiments, apes spontaneously used up to five tools in sequence to retrieve a food reward, and they adjusted their tool selection as the costs of reaching the best tool changed.4PubMed Central. Sequential tool use in great apes That kind of flexible adjustment implies more than trial-and-error learning; it suggests the animal is mentally weighing options before acting.
Knowing What You Know
One of the more striking cognitive abilities found in animals is metacognition, loosely defined as thinking about your own thinking. In humans, this shows up as the feeling of uncertainty when you are not sure of an answer, or the confidence that you are right. Testing whether animals experience something similar is tricky, but researchers have found consistent behavioral signatures suggesting they do.
In one study, chimpanzees performing a difficult ordering task on a touchscreen showed a telltale behavior: their hands wavered more before making a choice when the task was objectively harder. This hand-wavering pattern mirrors what humans do when facing uncertainty, and it occurred even though the chimps’ overall accuracy remained high.5PubMed Central. Chimpanzees (Pan troglodytes) show subtle signs of uncertainty when choices are more difficult The researchers argued this lends real plausibility to the idea that feelings of uncertainty, like other emotions, can be studied in non-human animals.
Eurasian jays show a different but equally telling pattern. Given the option to “opt out” of a difficult memory test and take a guaranteed small reward instead, jays chose to opt out more often when trials were harder. Crucially, when jays that typically opted out of hard trials chose to engage instead, they were significantly more accurate, suggesting they were only taking on challenges when they felt confident in their knowledge.6PubMed Central. Uncertainty monitoring in Eurasian jays (Garrulus glandarius) That looks a lot like an animal assessing its own certainty and acting strategically on the assessment.
Deception and Social Scheming
If you can think about what another individual knows, you can potentially manipulate what they know. Ravens provide one of the clearest examples. In food-caching experiments, ravens that had hidden food adjusted their behavior depending on whether they had been watched. They would delay returning to a cache if a rival was nearby, or create false caches to mislead observers. Researchers described this as “tactical deception,” noting that both cachers and raiders appeared capable of withholding information about their intentions and manipulating the other bird’s attention.7Animal Behaviour. Observational learning and the raiding of food caches in ravens, Corvus corax: is it ‘tactical’ deception? This kind of behavior suggests that ravens hold some model of what another raven can see and know, a capacity sometimes called “theory of mind” in the human literature.
Similar behaviors appear in primates, dolphins, and even some insects. The common thread seems to be social pressure: species that live in complex social groups, where competition for food or mates is intense, tend to develop more sophisticated abilities to predict and influence the behavior of others.
Symbols, Words, and Meaning
The most famous attempts to teach human-like language to animals have involved great apes, and the results are genuinely impressive even if they fall short of full human language. Bonobos Kanzi and Panbanisha learned to use a keyboard of over 200 visual symbols called lexigrams, and published studies showed they could comprehend spoken English at roughly the level of a two-and-a-half-year-old human child. When researchers analyzed more than a decade of vocabulary test errors, they found the mistakes followed patterns strikingly similar to those of humans: the bonobos confused symbols that sounded alike, looked alike, or were frequently associated in time or space, revealing that they had organized their vocabulary into meaningful categories rather than memorized isolated associations.8PubMed. Mental representation of symbols as revealed by vocabulary errors in two bonobos (Pan paniscus)
Kanzi’s abilities were further tested with degraded speech, the kind of muffled or distorted audio that is hard for even human listeners to parse. He was presented with both natural human speech and computer-generated speech that had been artificially degraded. Kanzi recognized both natural and computer-generated degraded voices at rates significantly above chance, suggesting his understanding of spoken words was robust enough to survive substantial distortion.9PubMed Central. Degraded and computer-generated speech processing in a bonobo No one claims this is language in the full human sense, with recursive grammar and open-ended creativity. But it demonstrates a mental architecture capable of symbolic representation and flexible comprehension.
Memory, Planning, and Mental Time Travel
Can animals think about the past and the future? The evidence, especially from food-caching birds, is surprisingly strong. Western scrub-jays remember not just where they hid food, but what type of food they hid and how long ago they hid it. If a perishable item was cached too long ago to still be fresh, they skip it and go to a cache with non-perishable food instead. This “what-where-when” memory is considered an animal analog of human episodic memory.
Research on future planning has been even more contested. The long-standing assumption, sometimes called the Bischof-Köhler hypothesis, was that acting to satisfy a future motivational state was beyond the capacity of non-human animals. Disproving this hypothesis became something of a holy grail for the field, sparking a wave of studies on whether animals can plan for needs they do not currently feel.10Wiley Interdisciplinary Reviews: Cognitive Science. Mental Time Travel in Animals Scrub-jays have come closest to meeting these criteria, though the debate continues about whether what looks like planning could be explained by simpler learning rules.
Counting and Numbers
The ability to judge quantity turns out to be remarkably widespread. Animals as varied as monkeys, crows, fish, and honeybees can assess the number of items in a group, a capacity researchers call a “sense of number.”11Journal of Experimental Biology. Neuroethology of number sense across the animal kingdom This is not just noticing “more versus less.” Controlled studies in honeybees trained to discriminate the quantity of visual elements found effects matching those seen in vertebrate number systems: the bees found it harder to distinguish quantities that were close together (four versus five) than quantities that were far apart (two versus seven), the same pattern humans show.12Trends in Cognitive Sciences. A Sense of Number Emerging from Phylogenetically Diverse Brains
The fact that this ability appears in insects, cephalopods, fish, birds, and mammals suggests it evolved very early or arose independently multiple times. Either way, it tells us that some form of abstract numerical processing does not require a large brain. It may be a fundamental feature of nervous systems sophisticated enough to interact flexibly with the world.
The Brain Does Not Have to Look the Same
A persistent misconception is that complex thinking requires a brain structured like ours, with the layered neocortex that defines the mammalian brain. Bird brains lack this layered architecture entirely, yet corvids (crows, ravens, jays) rival great apes on many cognitive benchmarks. The key structure in the crow brain, called the nidopallium caudolaterale, functions as an analog to the mammalian prefrontal cortex and plays a crucial role in higher cognitive and executive functions.13eNeuro. Input and Output Connections of the Crow Nidopallium Caudolaterale Research comparing the two structures found that while the layered organization differs dramatically, virtually every other aspect of the neural architecture is “extremely comparable,” leading to the conclusion that cortical layering is not required for higher cognitive functions.14PubMed. The avian ‘prefrontal cortex’ and cognition
This finding has important implications. It means evolution has arrived at prefrontal-like cognitive abilities through at least two independent architectural designs. The functional outcome, flexible decision-making and executive control, matters more than the wiring diagram used to achieve it.
Octopuses, Cuttlefish, and the Invertebrate Surprise
If bird brains challenged the neocortex assumption, invertebrates demolished it entirely. Octopuses, with nervous systems organized in a radically different way from any vertebrate, routinely solve problems in the lab. In a classic study, common octopuses learned to open sealed glass jars containing live crabs, and their performance improved significantly over repeated trials.15Behavioral and Neural Biology. Problem solving ability of Octopus vulgaris lamarck (Mollusca, Cephalopoda) In a more complex puzzle task requiring alternating push-and-pull actions across multiple difficulty levels, all tested octopuses reached the success criterion at every level and adapted quickly when the rules changed, demonstrating behavioral flexibility that would be noteworthy in any vertebrate.16PLOS ONE. Pull or Push? Octopuses Solve a Puzzle Problem
Cuttlefish, close relatives of octopuses, have shown a different kind of cognitive sophistication: self-control. In a delay-of-gratification task similar to the famous marshmallow test used with children, cuttlefish were presented with a choice between an immediately available but less preferred food and a better food that required waiting. They maintained delays of 50 to 130 seconds to hold out for the better option, tolerances comparable to those of some large-brained vertebrates.17Proceedings of the Royal Society B: Biological Sciences. Cuttlefish exert self-control in a delay of gratification task In corvids, this same ability to wait for a better reward is correlated with performance on other cognitive tasks, suggesting that self-control may be a marker of general intelligence.18PubMed Central. Waiting for a better possibility: delay of gratification in corvids and its relationship to other cognitive capacities
Empathy and Emotional Contagion
Whether animals have emotions is an old question that has gradually shifted from philosophical speculation to empirical investigation. The strongest evidence comes from social species. Asian elephants, for instance, have been observed approaching distressed companions and offering physical contact with their trunks and vocalizations. In a study of a semi-captive herd, bystander elephants affiliated with each other and matched the behavioral and emotional state of the first distressed individual. The researchers classified this as consolation behavior best compared to similar responses observed in apes, possibly arising through independent evolutionary paths.19PubMed Central. Asian elephants (Elephas maximus) reassure others in distress
Emotional contagion, where one individual’s distress spreads to others, has also been documented in rats, mice, and prairie voles. These are not interpretive stretches; the animals measurably change their behavior and physiology in response to another’s state. Whether this constitutes empathy in the rich sense humans experience remains debated, but the functional building blocks are there.
Culture Passed Down Through Generations
Social learning, where an animal picks up a skill by watching another, has been identified in vertebrates and invertebrates alike. But sustained traditions that pass from one generation to the next are rarer, and the multiple traditions that researchers call “culture” are rarer still.20PubMed Central. The evolution of animal ‘cultures’ and social intelligence The clearest examples come from chimpanzees, where neighboring groups maintain distinct toolkits and foraging methods despite living in similar environments.
Wild capuchin monkeys have provided some of the most detailed recent evidence. In an open diffusion experiment, researchers introduced a novel extractive foraging task to two wild primate populations and tracked how information spread. The capuchins learned primarily through direct observation, and naïve individuals showed a bias toward watching successful males. Social tolerance, how willing established group members were to let newcomers watch them work, predicted which pathways the knowledge traveled.21PubMed Central. Social tolerance and success-biased social learning underlie the cultural transmission of an induced extractive foraging tradition in a wild tool-using primate The study’s framing was explicit: this supports claims of genuine cultural transmission in non-human primates.
Elephants present an interesting contrast. Cultural knowledge is widely presumed to be important for them, and individuals in all three elephant species tend to cluster around older companions, creating opportunities for social transmission. But direct evidence of sustained cultural traditions in elephants remains surprisingly thin. A recent systematic review focused instead on what happens when elephants lose access to knowledgeable individuals through poaching, culling, or translocation, finding that such social disruption has measurable downstream effects on group behavior.22PubMed Central. Knowledge transmission, culture and the consequences of social disruption in wild elephants The fact that removing older, experienced elephants degrades group functioning is itself evidence that something important is being transmitted socially, even if we cannot yet catalog the specific traditions.
The Pitfalls of Studying Minds Unlike Ours
All of this research comes with a significant methodological caveat. Every experiment designed to test animal cognition involves assumptions about how the animal perceives and interprets the stimuli presented to it. Researchers call this the “umwelt gamble,” referring to the concept that each species lives in its own unique sensory and perceptual world. An animal might fail a test not because it lacks the cognitive ability being tested, but because the artificial stimuli used in the experiment do not make sense within its perceptual reality.23Ethology. The Limits of Artificial Stimuli in Behavioral Research: The Umwelt Gamble Animals can be biased in unforeseen ways in how they perceive and interpret their environments, which means a negative result on a cognition test is much harder to interpret than a positive one.
The opposite problem, anthropomorphism, also looms. Researchers sometimes see human-like thought where simpler explanations suffice. A rat pressing a lever in a particular pattern might be following a learned rule rather than “planning.” An elephant approaching a distressed companion might be responding to a conditioned cue rather than feeling empathy. Good experimental design tries to distinguish these possibilities, but the line between sophisticated cognition and sophisticated learning is often genuinely blurry. The field’s best work acknowledges this uncertainty rather than resolving it prematurely in either direction.
When Animal Minds Decline
One of the more unexpected lines of evidence for complex cognition in animals comes from studying its breakdown. Age-related cognitive decline is not unique to humans. Even without dementia, aging affects specific types of memories and brain structures in both humans and animal models.24PubMed Central. Assessing cognitive decline in the aging brain: lessons from rodent and human studies As veterinary care has extended the lifespans of companion animals and captive wildlife, veterinarians are seeing more age-related cognitive dysfunction. Conditions resembling Alzheimer’s disease, complete with amyloid plaques and behavioral changes, have been documented in non-human primates, dogs, cats, dolphins, whales, and elephants.25Next Research. Ageing in animals: Alzheimer’s like disease
In aging rats, researchers have found a dissociation between two types of cognitive control. Older rats showed clear deficits in inhibitory control, the ability to stop a habitual response, but largely preserved their capacity for self-control when it came to waiting for a delayed but larger reward.26PubMed Central. Healthy aging in rats is associated with a decline in the ability to inhibit maladaptive responses, but not in measures of self-control by delayed gratification The fact that aging selectively impairs specific cognitive functions in animals, just as it does in humans, tells us these functions are genuinely distinct mental processes rather than different manifestations of one general-purpose learning system.
How Artificial Intelligence Compares
The rise of AI has prompted a new kind of comparison. Modern deep neural networks can outperform humans on narrow tasks like image classification, but how do they measure up against animal cognition more broadly? In one large-scale study, researchers compared the visual object-recognition behavior of deep neural networks against that of humans and rhesus macaque monkeys, using over a million behavioral trials. The neural networks accurately predicted the broad patterns of which objects primates confused with each other, but when the comparison moved to individual images within each discrimination task, every tested model was significantly non-predictive of primate performance.27Journal of Neuroscience. Large-Scale, High-Resolution Comparison of the Core Visual Object Recognition Behavior of Humans, Monkeys, and State-of-the-Art Deep Artificial Neural Networks In other words, the AI systems got the big picture right but missed the fine-grained details of how primate brains actually process individual images. Animal minds remain, in important ways, qualitatively different from the pattern-matching engines we have built so far.
Songbirds offer another angle on this comparison. Young birds learn to sing by listening to adult tutors during a critical developmental window, and the neural circuits sculpted during this period affect the bird’s later vocal communication skills.28PubMed. Neuronal mechanisms regulating the critical period of sensory experience-dependent song learning This experience-dependent learning is nothing like how a speech-recognition algorithm is trained. The bird’s brain physically reorganizes in response to what it hears, and the window for that reorganization opens and closes on a developmental schedule. Machine-learning systems have no analog to this kind of embodied, time-sensitive learning, which is one reason animal cognition research continues to offer insights that AI research alone cannot provide.