RESEARCH DOSSIER
Tool Use and Physical Causal Cognition Across Species
Abstract. Tool use and physical causal cognition across species investigate how animals recruit external objects to transform physical relations, acquire material techniques, and generalize effective action to novel causal structures. This fundamental research monograph offers a critical narrative synthesis of field observations, controlled experimental studies, methodological analyses, and contemporary comparative research through 2026. It reviews the conceptual definition of tools, the history of comparative technology, ecological affordances, primate nut cracking and sequential actions, crow hook manufacture and water-displacement experiments, parrot tool-set transport, woodpecker finch acquisition, dolphin sponge carrying, octopus shelter construction, and the evolutionary interpretation of convergent technology. The evidence supports distinct explanatory levels for observed tool performance, perceptual-motor skill, causal discrimination, novel intervention, development, and cultural transmission. Critical evaluation of opacity manipulations, trap problems, Aesop's fable paradigms, and novel-action experiments demonstrates why success in one physical task does not establish a general causal theory. The synthesis identifies problems in species-appropriate task design, sampling, learning-history controls, phylogenetic inference, and independent replication. A final interdisciplinary discussion situates operational tests of physical causal generalization within the Artificial Era while keeping empirical claims about biological cognition separate from hypotheses about artificial systems. This is a narrative scientific monograph, not an original experiment, preregistered systematic review, or meta-analysis.
Keywords: animal tool use; physical causal cognition; comparative psychology; animal cognition; tool manufacture; causal reasoning; physical affordances; New Caledonian crows; chimpanzee technology; capuchin nut cracking; Goffin’s cockatoos; dolphin sponging; octopus tools; social learning; causal intervention; cross-species methods; evolutionary convergence
The Research Object: Tool Use as Situated Physical Problem Solving
Tool use and physical causal cognition across species concern how an organism recruits an external object to transform a physical relation, achieve a goal, or obtain information, and how experimental evidence supports explanations of that activity. This research object spans selection, manipulation, manufacture, transport, and sequential combination of tools, together with perception of mechanical properties and the capacity to generalize beyond practiced routines. The behavioral act of using an object is directly observable; the cognitive processes inferred from that act require discriminating tests. A crow shaping a hook, a chimpanzee selecting a stone, and a dolphin carrying a sponge can all be described as external-object technology, yet the functional problems and acquisition pathways differ considerably St Amant & Horton (2008) Boesch & Boesch (1990) Krützen et al. (2005).
The explanatory unit employed here is an organism acting on an ecologically meaningful physical arrangement under specified perceptual, motivational, developmental, and learning conditions. Tool behavior depends on anatomy, motor control, accessible materials, and an opportunity to discover useful object relations. Evidence of sophisticated performance can coexist with uncertainty about representation, and a failure may arise because the test requires an unfamiliar action or inaccessible perception. An experiment establishes its strongest conclusion when the materials, control conditions, and predictions identify which account survives alternative explanations. Comparative inference therefore requires behavioral description, task analysis, causal manipulation, and justified cross-species measurement rather than ranking animals along an assumed universal ladder of intelligence Seed et al. (2009) Penn et al. (2008).
This monograph presents a critical narrative synthesis of peer-reviewed original experiments, field observations, comparative theoretical work, and recent evidence through 2026. It evaluates how physical problem-solving capacities are defined, acquired, measured, and explained; how the same animal may succeed in one causal task and fail in another; and how technological traditions arise in different lineages. The study does not report new trials, a preregistered systematic review, or a meta-analysis. Its scientific contribution is the explicit integration of levels of explanation and the disciplined distinction between observed action, hypothesized cognition, and evolutionary inference Miller et al. (2026) Bastos et al. (2025).
Definitions and the Boundary of Animal Tools
A comparative definition must distinguish a tool from an environmental feature that simply affords action. Traditional formulations typically describe an external object controlled by an animal to alter another object or organism, or the animal's own state, through an intentionally directed movement. St Amant and Horton proposed a broader explanatory treatment that includes mechanical transformation and information mediation. These approaches classify some behaviors differently: an object used as a probe can change access to sensory information, while stones positioned as support or surfaces may blur the line between portable tools and fixed environmental structures. The boundary is analytically consequential because inclusion rules alter apparent species distributions St Amant & Horton (2008).
Operational coding should record whether an object was detached from its substrate, moved or modified, controlled during use, and functionally coupled to a target. A stone hammer repeatedly directed onto a nut is an especially clear mechanical intervention; collecting a coconut shell and later assembling a shelter presents a different temporal structure. The question of whether external objects qualify as tools must be answered before the separate question of which mental mechanism produced the action. A tool classification is a description of the behavior's organization, and no single tool-use observation proves foresight, causal theory, or cultural learning Boesch & Boesch (1990) Finn et al. (2009).
Comparative datasets also require distinctions among tool possession, habitual deployment, manufacture, flexible selection, and combinations of tools. An individual may spontaneously use a novel object once without the behavior becoming characteristic of its population. A population may repeatedly exploit a material through long apprenticeship without each individual innovating the practice. Manufacture can mean removing leaves, adjusting length, shaping a hook, or joining separate elements; these transformations impose different coordination and material demands. Reported richness of tool repertoires therefore depends on classification, observation effort, ecology, and availability of manipulable materials Hunt (1996) Weir et al. (2002) Bastos et al. (2025).
Historical Formation of Comparative Tool Research
Scientific inquiry into animal tools developed from natural-history accounts and field ethology into experimental comparisons of mechanisms. The extensive longitudinal study of chimpanzees established that tool use was embedded in foraging ecology and in population-specific repertoires. Boesch and Boesch compared tool manufacture and use in chimpanzees at Mahale, Gombe, and Taï, showing meaningful variation in the kinds of actions recorded and in modifications made to materials. This evidence challenged assumptions that complex tool behavior depended upon a single laboratory-trained species or an exclusively human technological tradition. At the same time, regional contrasts posed new questions about local resource conditions, experience, and historical transmission Boesch & Boesch (1990).
The discovery of hook manufacture by New Caledonian crows shifted comparative attention toward independently evolved avian technology. Hunt described naturally occurring hook manufacture in 1996; Weir, Chappell, and Kacelnik subsequently documented a captive crow shaping wire into a hook in an experimental context. Together these findings established the importance of separating a population's natural technical repertoire from an individual's flexible response to an unfamiliar apparatus. Later experiments examined compound tools, material preferences, and causal discrimination, extending observation into explicit tests of generalization Hunt (1996) Weir et al. (2002) von Bayern et al. (2018).
The history of the field also reflects a methodological correction: complex action sequences can be produced by different learning systems. Research on apes and children contrasted opaque with transparent apparatuses to test whether visibility of mechanisms changes the copying of seemingly unnecessary actions. Studies of trap problems removed tool-manipulation confounds and found better task performance in altered conditions. Current comparative research consequently treats representational explanations as empirically testable proposals with observable consequences, and evaluates environmental, perceptual, and procedural alternatives before assigning strong causal understanding Horner & Whiten (2005) Seed et al. (2009) Miller et al. (2026).
Levels of Explanation: Mechanism, Development, Ecology, and Evolution
The functional explanation of a tool action is distinct from the mechanism generating the action. Mechanistic analysis asks which sensory cues, motor routines, reinforcement histories, and representations guide a choice. Developmental analysis asks when the animal can coordinate the relevant movements and how practice or exposure alters behavior. Ecological analysis asks which physical problems make external-object techniques useful, and historical analysis asks whether tool behavior reflects ancestral retention, convergent evolution, or population-specific innovation. These levels form complementary research questions and require different kinds of evidence Boesch & Boesch (1990) Emery & Clayton (2004).
A chimpanzee's successful cracking of a nut can be examined through the geometry of hammer impact, the stability of the anvil, the hardness of the shell, and the choice of stone. The developmental question concerns the prolonged acquisition of these coordinated actions, including opportunities to watch competent individuals. The ecological question concerns the nutritional payoff relative to alternative food sources and the local availability of stones. Evidence of intergroup distributional differences invites investigation of transmission without eliminating ecological and demographic explanations. The same observed outcome supports several linked but methodologically separate explanatory projects Visalberghi et al. (2007) Gruber et al. (2009).
At the evolutionary level, similarities between apes and corvids can arise through convergent functional solutions rather than identical anatomical machinery. The comparison is most informative when the same psychological construct is assessed through ecologically calibrated tasks and when shared performance is distinguished from shared representation. An animal with grasping hands and a bird with a dexterous bill face different mechanical possibilities, even when both can retrieve inaccessible rewards through tools. The inference from outcome similarity to common cognitive architecture must therefore be tested against anatomical constraints, learning opportunities, and divergent evolutionary histories Emery & Clayton (2004) Penn et al. (2008).
Affordances, Materials, and Mechanically Effective Action
An affordance is an action possibility arising from the relation between an organism's capacities and material or environmental structure. A stone can serve as a hammer because its mass, rigidity, and manipulability support a useful impact; a flexible strip can function as a probe because it reaches into an opening while remaining under control. The effectiveness of an object is relational. Tool choice consequently requires measurement of the task's physics and the animal's available movements. A useful comparative experiment changes one material property at a time while holding reward, access, and training as constant as feasible Visalberghi et al. (2007) Fragaszy et al. (2013).
Wild capuchin studies illustrate how apparently simple nut cracking requires coordinated understanding or learning of material relations. Researchers documented selection and transport of stones and characterization of anvils, then investigated how individuals orient nuts for greater stability. Stable placement changes the geometry of impact and the probability of cracking. Such evidence supports finely adjusted physical competence; it does not, by itself, identify whether the adjustment emerged through explicit relational reasoning, perceptual feedback, accumulated motor practice, or a combination. Testing mechanism requires controlled material substitutions and transfer to novel combinations Visalberghi et al. (2007) Fragaszy et al. (2013).
Novel apparatuses permit rigorous contrasts between relevant physical variables and superficial visual cues. In water-displacement tasks, for example, a sinking solid object should elevate the water level more effectively than a floating or hollow alternative of suitable dimensions. The investigator must establish which distinctions were visible to the animal, how preference was assessed before testing, whether reward feedback guided learning, and whether a physical property transfers to previously unseen objects. Tool choice provides stronger evidence about a mechanism when the animal generalizes between materially different examples that preserve the same causal relation Jelbert et al. (2014) Logan et al. (2014).
Methodological Architecture and Evidence Classification
A rigorous cross-species program requires complementary naturalistic and controlled data. Field observation establishes behavioral occurrence, ecological context, variation between populations, and sequences that may be difficult to reproduce in captivity. Laboratory work makes it possible to assign participants to conditions, counterbalance stimulus positions, equalize reinforcement histories, and isolate mechanical variables. Comparative modeling then asks whether multiple populations or lineages show patterns compatible with evolutionary hypotheses. None of these approaches replaces the others: a contrived laboratory success need not be ecologically frequent, and an observed wild tradition may resist immediate mechanistic interpretation Boesch & Boesch (1990) Bastos et al. (2025).
This review classifies claims by the strength of the supported inference. Behavioral evidence establishes that an animal performed a specified sequence under documented conditions. Mechanistic evidence tests a process such as feedback-guided learning, physical discrimination, or anticipatory transport through competing predictions. Generalization evidence demonstrates that information survives changes in materials, contexts, or outcomes. Developmental evidence links acquisition to age, practice, and social opportunity; evolutionary evidence requires species sampling and historically informed inference. A claim becomes more robust when independent methods converge while controlling for shared task-specific artifacts Seed et al. (2009) Taylor et al. (2014).
Assessment of individual studies should include how animals were selected, how many performed each condition, opportunities for prior experience, within-subject dependence, statistical uncertainty, the interpretation of null results, and the ecological fit of the apparatus. Comparative work is particularly vulnerable to making population-wide claims from a small pool of captive individuals or to interpreting naturally occurring behavior without representative observation effort. The strongest scientific synthesis maintains separate records for experimental results, proposed cognitive interpretations, and unresolved alternatives. A cumulative picture emerges from calibrated evidence rather than a count of spectacular performances Penn et al. (2008) Miller et al. (2026) Bastos et al. (2025).
Chimpanzee Technologies in Wild Foraging Systems
Wild chimpanzees provide a rich account of tool behavior integrated into routine foraging rather than isolated test trials. Comparative field records identified diverse techniques for acquiring insects, extracting concealed food, and using stones or wooden hammers. Boesch and Boesch's three-population comparison reported substantial contrasts in tool categories and degrees of material modification, particularly at Taï. The relevant unit is not simply a count of tools: gathering food with a probe, cracking a hard object against an anvil, and combining implements in sequence recruit different physical structures and sensorimotor demands. Descriptive surveys also raise questions about visibility, site coverage, and classification consistency Boesch & Boesch (1990).
The ecological setting determines what a chimpanzee can learn. Nut-cracking requires suitable nuts, durable striking objects, available anvils, and repeated opportunities for practice. A behavior may be absent from a population because useful materials are unavailable, the relevant task is rare, individuals have insufficient opportunity to acquire a technique, or historical transmission has not established a tradition. Field comparisons therefore require documented resource distributions and developmental observations. Inferences about mental sophistication drawn from the mere presence of tool use ignore how local challenges and material opportunities shape action Boesch & Boesch (1990) Gruber et al. (2009).
A causal analysis of chimpanzee technology must ask whether the animal represents relations among implement, target, and outcome, recognizes useful affordances through experience, or depends on a practiced stimulus-response sequence. These processes can cooperate. Skilled impact control is compatible with procedural learning and with representations of mechanical consequences. Experiments using new tools, reversed constraints, and altered reward locations are needed to determine what generalizes. Field observations supply ecological relevance; they cannot alone discriminate every candidate representation Seed et al. (2009) Penn et al. (2008).
Nut Cracking, Hammer Selection, and Mechanical Calibration
Nut cracking is an especially informative comparative task because its mechanics can be independently described. The result depends on hammer mass and material, striking velocity, impact angle, nut stability, and anvil geometry. Bearded capuchin studies document consistent choice of potential hammers and anvils, demonstrating selective engagement with the material environment. Characterization of the stones found around cracking sites constrains competing hypotheses about chance use, indiscriminate collection, and adapted tool selection. Transporting a useful hammer from another location further separates immediate opportunism from organized multi-step activity Visalberghi et al. (2007).
The ability to position a nut stably is another experimentally relevant component. Fragaszy and colleagues reported that wild capuchins could orient palm nuts so that their placement improved stability before hammering. This adjustment is an example of physically effective action produced through intricate contact with material feedback. Its importance lies in showing that a complete tool routine includes preparation of the target, not merely movement of an implement. Questions remain about the relative contribution of previous experience, body coordination, and anticipatory sensitivity to forces. Trials that alter anvil shape or nut asymmetry offer ways to distinguish these processes Fragaszy et al. (2013).
Comparing chimpanzees and capuchins requires attention to body size, grip, strength, risk, and ecological niche. A hammer that is mechanically optimal for one species could be prohibitively heavy for another. Laboratory choice tests can normalize tools by functional effectiveness rather than by absolute mass, while field measurements provide ecological constraints on what is readily available. Similar choices under calibrated conditions would support convergent sensitivity to useful properties, whereas different preferences would not automatically demonstrate different causal capacities. The study of tool selection is strongest when material mechanics and anatomical cost are specified before cognitive labels are assigned Boesch & Boesch (1990) Visalberghi et al. (2007).
Sequential Tools and Anticipation of Future Operations
Some physical problems require more than a single intervention. A tool may first expose a concealed cavity, another may retrieve a target, and a third may alter access to the working area. The central scientific question is whether an animal chooses and transports several implements before needing each one, or whether it selects the next object only after encountering the immediate obstacle. The first pattern suggests an integrated organization of action across steps, although it can also be learned through repeated experience. Sequence length alone does not identify the representational process; the crucial measures are conditional selection, advance preparation, and transfer to changed tasks Boesch & Boesch (1990) Osuna-Mascaró et al. (2023).
The comparative relevance of chimpanzee technology became especially clear when experimental tool-set transport was studied in Goffin's cockatoos. Osuna-Mascaró and colleagues presented situations requiring either one implement or a sequence and measured whether birds transported the appropriate combination to a future working location. Some individuals acquired the relevant behavior and adjusted transport according to the demands of the apparatus. This evidence extends the distribution of conditional tool-set transport beyond primates and supports comparative hypotheses about organizing future material operations. It also leaves open whether individuals encode functional relations explicitly or acquire structured strategies through experience Osuna-Mascaró et al. (2023).
A rigorous sequential-tool experiment must avoid treating high success after prolonged training as spontaneous planning. Investigators should document training stages, the temporal distance between selection and use, whether both tools are immediately visible, and whether a new goal changes the optimal sequence. A representation of the anticipated task predicts selective preparation in unfamiliar combinations. A learned rule based on familiar apparatus cues predicts more limited transfer. These alternatives may yield identical success in a repeated condition; testing the mechanism requires carefully structured departures from that condition Osuna-Mascaró et al. (2023) von Bayern et al. (2018).
Physical Causation Beyond Tool Manipulation
Physical causal cognition can be studied without requiring an animal to wield an external object. In one line of work, chimpanzees used a physical effect of hidden weight to infer which location contained a reward. Such tasks distinguish causal perception or inference from the motor skill of using a probe or hammer. They also test whether an animal uses the structure of an event, rather than a memorized object identity, to guide choice. The result can provide evidence for discriminating a particular causal relation while leaving the breadth and explicitness of that relation to further experiments Hanus & Call (2008).
Trap problems illustrate how motor requirements can obscure the cognitive target. A device may require pushing food with a stick while avoiding a trap, thereby combining perceptual discrimination, direction control, and tool-mediated movement. Seed and colleagues reported that chimpanzees solved a version of the trap problem when the requirement to use a tool was removed. This design demonstrates why a poor result in a compound task cannot be interpreted as an isolated deficit in physical reasoning. Comparisons should determine whether successful avoidance transfers between apparatuses that share a causal relation but differ in surface layout Seed et al. (2009).
The inference made from these experiments is constrained by how a physical mechanism is displayed. Visible motion can supply an informative cue without the animal representing an unseen force, and the rewarded choice may be learned by trial-and-error during testing. Experiments should contrast causal and arbitrary predictive cues, include unreinforced transfer trials where ethical and practical, and measure performance before extensive feedback accumulates. The strength of the conclusion lies in the specificity of the alternative ruled out, not simply in whether the animal retrieves food Hanus & Call (2008) Seed et al. (2009) Penn et al. (2008).
Individual Acquisition and the Long Development of Skilled Action
Physical skill develops through repeated interactions between morphology, objects, and opportunities for exploration. An infant may observe a skilled adult long before it can grip and strike efficiently. The acquisition process can involve component abilities that mature at different rates: selecting an object, moving it to a location, orienting a target, coordinating impacts, and evaluating the result. A mature repertoire is therefore an endpoint of individual history, and the presence of competent adults does not show that each component is directly copied. Detailed longitudinal studies remain necessary to establish acquisition sequences and the role of supportive social environments Boesch & Boesch (1990) Miller et al. (2026).
The same methodological principle applies to birds whose bill-based manipulation differs from primate hand coordination. A young crow may become proficient through explorative handling and gradually restricted attention to suitable material. The rate and direction of acquisition can be affected by access to twigs, prey, and adult tool products. Captive-reared subjects may exhibit manipulation under unfamiliar conditions, yet the developmental ecology of wild behavior can remain different. Researchers should report age, rearing background, previous problem-solving, and access to naturalistic objects before comparing developmental milestones across species Hunt (1996) Miller et al. (2026).
Developmental claims become stronger when competing sources of improvement are measured separately. If performance increases with age, one explanation is maturation of motor coordination; another is greater cumulative exposure to the physical contingency. If behavior changes after observing another individual, learning may depend on attention to the location, material, or outcome without copying the other's exact movement. Separating these pathways requires controlled exposure histories and component-level assessment. A well-designed developmental program treats skill acquisition as a dynamic combination of opportunity, movement, reinforcement, and potentially structured causal knowledge Tebbich et al. (2001) Gruber et al. (2009).
New Caledonian Crow Tool Manufacture and Material Selection
New Caledonian crows constitute a major comparative case because their natural foraging includes the manufacture and use of slender implements to extract otherwise inaccessible prey. Hunt's field account described hook-making and the preparation of tools from vegetation, showing that manufactured implements belong to the wild behavioral repertoire. The analysis of these artifacts made it possible to ask how structural features such as curvature, reach, and stiffness contribute to effectiveness. Wild manufacture should be distinguished from a laboratory animal's response to an unfamiliar construction problem. Both are important, but they test different mixtures of established skill and novelty Hunt (1996).
Weir, Chappell, and Kacelnik observed a captive New Caledonian crow shape a metal wire into a hook for retrieving an inaccessible food item. The episode has become influential because the material and method were unfamiliar in the relevant experimental setting. Nevertheless, a single innovative performance does not identify a complete causal model or determine how readily other individuals of the same species would behave. Later studies sought to move beyond celebrated individual cases toward repeatable tasks involving alternative materials, manipulated contingencies, and multiple subjects. The field's shift from discovery to mechanistic experimentation is essential for evaluating the scope of flexibility Weir et al. (2002) Jelbert et al. (2014).
The functional relationship between a tool and its target can be investigated by changing shape, rigidity, or length while preserving reward location. A preference that survives changes in color or identity is more informative about physical structure than a preference for an object repeatedly rewarded during training. Such tests must also measure manipulation costs and prior object preferences. New Caledonian crows show sophisticated tool behaviors, but the most defensible conclusions remain tied to the specific causal dimensions and transfer tests examined rather than a global attribution of human-like physical reasoning Hunt (1996) Jelbert et al. (2018) Penn et al. (2008).
Water Displacement and the Aesop's Fable Paradigm
The Aesop's fable paradigm translates a familiar physical principle into an experimental choice: dropping an appropriate object into water raises the level and can bring floating food into reach. Jelbert and colleagues used this paradigm with New Caledonian crows to test differences between water and sand, sinking and floating objects, solid and hollow objects, and apparatuses with different starting water levels. In several comparisons the birds preferentially selected effective options. These performances indicate sensitivity to properties relevant to successful displacement under the conditions tested Jelbert et al. (2014).
The same experiments exposed boundaries of inference. Birds did not succeed consistently in all variants, including tasks requiring attention to tube width and a counterintuitive U-shaped connection. Consequently the appropriate scientific description is a differentiated profile: accurate choices across several functional discriminations and difficulty when success depended on different or hidden relations. A uniform label such as complete causal understanding would erase this structure. Specific successes must be linked to the visible cues and learning opportunities available in each problem Jelbert et al. (2014) Taylor et al. (2014).
Subsequent modifications to the paradigm showed that seemingly similar tasks can yield altered performance when apparatus details change. Logan and colleagues reported that New Caledonian crows could solve a U-tube variant and discriminate between water-filled tubes of different volumes, while their results did not support the hypothesis of an inferred hidden causal mechanism. This comparison is instructive because earlier failure and later success were not equivalent measures under identical conditions. Replication should examine task configuration, sample differences, training histories, and perceptual demands before interpreting a change as a direct correction of the animals' underlying capacity Logan et al. (2014) Jelbert et al. (2014).
Causal Intervention, Prediction, and Novel Action
A major theoretical distinction separates recognizing that one event predicts another from inventing an action that reproduces an observed causal relation. An animal might prefer heavy objects after seeing them raise water, yet remain unable to generate a previously unpracticed intervention required by an unfamiliar device. Taylor and colleagues tested a causal intervention problem involving two-year-old children and New Caledonian crows. The results showed a contrast between the children's production of the target action and the tested crows' failure, despite the latter group's existing tool-related capacities Taylor et al. (2014).
This result bears on a specific proposed route from observation to intervention. It does not negate crow tool manufacture, material discrimination, or learning in other physical tasks. The original experiment itself identified the limited crow sample and need for further experiments as constraints on generalization. It also controlled for the ability to perform the relevant motor action in a separate condition, strengthening the interpretation that the immediate difficulty could not be reduced to mere inability to move the object. The evidence supports separating distinct dimensions of causal behavior instead of translating one result into an overall ranking of species Taylor et al. (2014).
Future intervention tests should contrast a merely correlated cue with a manipulated causal link, include unfamiliar action structures, and measure whether the subject spontaneously acts in a way predicted by a newly observed dependency. An adequate design also includes demonstrations that the animal perceives the relevant events and can physically perform the response. Repeated rewarded attempts may transform an insight test into an acquisition test, so the temporal structure of learning matters. Conclusions should specify whether the result concerns causal detection, action invention, intervention selection, or explanation of an unobserved mechanism Taylor et al. (2014) Penn et al. (2008).
Compound Tools, Construction, and Component Relations
Compound tool construction introduces a demand that can be absent from ordinary tool choice. The animal must recognize that two individually inadequate components can be combined into an effective implement, and the resulting assembly must satisfy the physical geometry of the target. Von Bayern and colleagues demonstrated that some New Caledonian crows constructed compound tools from shorter components in an experimental setting. The result expanded the empirical range of avian constructive behavior and invited new tests of whether individuals anticipate the properties of the completed object or discover successful configurations through exploration von Bayern et al. (2018).
Construction tasks should distinguish assembling objects, choosing a working configuration, and understanding the causal contribution of a connection. An animal may learn that inserting one element into another permits retrieval without representing all aspects of tensile strength or joint stability. Conversely, successful combination in a novel task can support abstraction about reach or continuity when object identities and training history are appropriately controlled. Comparisons with manufacture in the wild require attention to ecological opportunities: manipulating natural hooks and linking unfamiliar laboratory components have different developmental histories Hunt (1996) von Bayern et al. (2018).
Research on material templates offers another route into representation. Jelbert and colleagues investigated whether New Caledonian crows could reproduce aspects of tool size or shape using an experimentally defined target, suggesting that mental template matching could contribute to the reproduction of tool forms. A laboratory demonstration of matching is a candidate mechanism for cultural continuity, but it is not, by itself, a demonstration that wild tool traditions are transmitted through that mechanism. Demonstrating transmission would require linking the candidate process to observed learning across individuals and generations Jelbert et al. (2018).
Parrot Innovation, Tool Sets, and Flexible Transport
Parrots provide particularly informative contrasts with corvids and apes because species without a famous natural history of habitual tool use can sometimes innovate object-based solutions. Auersperg and colleagues described spontaneous manufacture and use of an implement by a Goffin's cockatoo in captivity, providing a case where a novel material problem elicited an unusual technique. Such cases help test whether physical flexibility is confined to specialist tool-using ecologies. Their interpretation must nevertheless preserve the distinction between individual innovation, repeated experimental success, and a stable population repertoire Auersperg et al. (2012).
The tool-set experiments of Osuna-Mascaró and colleagues examined whether Goffin's cockatoos could acquire and conditionally transport two implements to a task location. Some birds carried both tools when the apparatus required the pair and adjusted behavior when one was sufficient. This is a more stringent test than simply using the second tool after the first has exposed a new problem: selection precedes direct contact with the later operation. The results support flexible multi-step organization in some individuals, while raising questions about experience-dependent policies, planning demands, and the range of physical relations to which the behavior can generalize Osuna-Mascaró et al. (2023).
The strongest comparisons among crow, cockatoo, chimpanzee, and child performance use matched functional requirements with species-appropriate handling. A device that requires simultaneous carrying can impose very different physical costs on a flying bird and a walking primate. Experiments should therefore distinguish conceptual selection of a tool set from the logistics of transport. This distinction is especially important when a highly skilled tool-user fails a task constructed around an unfamiliar motor configuration. Similar outcomes across species can indicate functional convergence while leaving the underlying learning algorithms and representations open to inquiry Osuna-Mascaró et al. (2023) von Bayern et al. (2018) Penn et al. (2008).
Woodpecker Finches, Independent Acquisition, and Ecological Opportunity
Woodpecker finches demonstrate how the comparative study of tools can address questions about the acquisition and ecological distribution of technique in a distantly related bird lineage. Finches use suitable objects to probe for prey in crevices, a functional context that differs from both crow hook-making and primate hammering. Tebbich and colleagues experimentally examined whether tool use is acquired by social learning and provided evidence relevant to the possibility that individual learning can support acquisition under appropriate ecological circumstances. The research question concerns acquisition mechanisms rather than a binary contrast between instinct and intelligence Tebbich et al. (2001).
A species-typical opportunity to exploit a crevice does not imply that every individual will acquire an efficient technique at the same age or by the same route. Developmental experience, opportunities to manipulate small objects, seasonal prey access, and the distribution of materials can affect both frequency and form. If socially isolated individuals can learn relevant procedures, this supports an individual acquisition route under that experimental condition; it does not exclude effects of parental presence, social attention, or local traditions in natural environments. Multiple developmental pathways can produce similar ecological outcomes Tebbich et al. (2001) Miller et al. (2026).
The woodpecker finch case also highlights the importance of matching causal questions to apparatus construction. A probe chosen for a narrow opening tests different physical relations from a hammer chosen for a hard shell. A comparative measure called tool proficiency would collapse these differences unless it specifies contact geometry, material demands, prior experience, and independently measurable task success. The scientifically useful comparison concerns whether organisms can learn and transfer relations between object properties and physical outcomes within tasks suited to their bodies and ecological niches Tebbich et al. (2001) St Amant & Horton (2008).
Dolphin Sponge Carrying and the Transformation of Foraging Interfaces
Bottlenose dolphins in Shark Bay provide an aquatic case of external-object technology with mechanical and ecological demands unlike those of birds or primates. Some individuals carry marine sponges on their rostra during benthic foraging. The sponge appears to protect the sensitive surface while the animal searches along the seafloor. In this setting the tool mediates physical contact between body and substrate rather than being held to strike or retrieve a target. Understanding the behavior requires integrating object placement, underwater sensory ecology, foraging opportunity, and the costs of transporting an object during swimming Krützen et al. (2005) Mann et al. (2008).
Krützen and colleagues investigated the distribution of sponge carrying in relation to maternal lines and argued for cultural transmission of this technology. The inference depends on the association between social relationships and behavioral distribution, alongside plausible ecological and genetic alternatives. Later ecological analyses explored why the practice occurs in particular foraging niches. These results show why a population-specific technique can be shaped jointly by social history and functional opportunity. They do not determine whether individual dolphins consciously understand every mechanical effect of the sponge or learn its efficacy through repeated sensory and foraging feedback Krützen et al. (2005) Mann et al. (2008).
Comparisons with terrestrial technologies must be calibrated to the medium and functional goal. A dolphin cannot manipulate a stone hammer in the same way as a capuchin, and a sponge's role may be protective rather than mechanically transformative in the usual sense. Nevertheless, the core comparative questions remain recognizable: how an external item changes the organism-environment relation, whether its use is conditional on context, what individuals learn, and how behavior persists in a population. The dolphin case guards against equating tool sophistication with the visible complexity of hand-like manipulation St Amant & Horton (2008) Mann et al. (2008).
Octopus Coconut Shells and Delayed Protective Use
A coconut-carrying octopus offers another distinct configuration of animal technology. Finn and colleagues described veined octopuses collecting and transporting coconut-shell halves and later assembling them into protective shelters. The behavior is notable because carrying shells can be awkward and may provide little immediate protective benefit until the pieces are reassembled. The sequence therefore draws attention to delayed utility, transport costs, and environmental transformation. The observation establishes a tool-use candidate under definitions that recognize protective external-object functions, while particular claims about future-directed mental representation require additional discriminating evidence Finn et al. (2009) St Amant & Horton (2008).
This case demonstrates the interpretive difference between an organized temporal sequence and an explicit plan. An animal may repeatedly carry useful material because a history of reinforcement and ecological regularities supports that routine; alternatively, it may prepare for a predicted later need. Naturalistic observation can document the temporal gap and physical work involved but usually cannot compare these mechanisms experimentally. Future protocols could vary future refuge availability, shell cost, current shelter opportunities, and location of threats to determine how flexibly transport tracks anticipated need Finn et al. (2009).
Cephalopod anatomy and nervous-system organization also make direct comparisons with vertebrate tool users difficult. Flexible arms provide manipulation and locomotion capabilities absent from birds, yet the ecological and perceptual conditions of underwater shelter building are distinct. Similar words such as construction, planning, and instrument use can hide different operational demands. The principal evolutionary lesson is that external-object behavior can emerge in markedly different body plans, and evidence for convergent function must be distinguished from evidence for identical cognition Finn et al. (2009) Emery & Clayton (2004).
Hawaiian Crows and the Problem of Species-Level Inference
Research on Hawaiian crows, or ʻalalā, expanded the recognized distribution of corvid tool use. Rutz and colleagues reported widespread use of tools among captive surviving individuals and examined features of tool behavior relevant to the species' natural history. Such evidence matters because it shows that a well-developed capacity may be overlooked when few individuals remain accessible for study or when field opportunities are constrained. The finding broadens comparative hypotheses about tool-use evolution in crows, but captive performance and historical prevalence in extinct or inaccessible wild populations are different inferential targets Rutz et al. (2016).
A species-level claim requires care when sampling is determined by conservation status, institutional housing, and previous exposure. Captive populations can provide extraordinary experimental access, while also displaying idiosyncratic environments and restricted ecological variation. The fraction of individuals demonstrating a technique under a particular experimental protocol is an observable sample statistic; it is not automatically an estimate of historical wild frequency. Comparative records should specify whether behavior was spontaneous, elicited, trained, or observed in ordinary foraging contexts, along with the number and developmental background of participating animals Rutz et al. (2016) Miller et al. (2026).
The Hawaiian crow findings invite evolutionary comparison with New Caledonian crows, yet similar tool use can arise through ancestral predispositions, independent innovations, or behavior facilitated by ecological opportunities. To choose among these histories, researchers need phylogenetic context, precise trait coding, and descriptions of the relevant habitat and manipulating structures. The contrast between the two crow lineages is most informative when it motivates competing tests rather than relying on species labels as explanations for mental abilities Rutz et al. (2016) Emery & Clayton (2004).
Evolutionary Convergence, Ecology, and Uneven Observation
The taxonomic distribution of tool behavior reflects both evolutionary history and the uneven intensity of observation. Apes, corvids, parrots, dolphins, and octopuses are represented by very different research traditions, sample sizes, and definitions of tool use. Absence of a published report may indicate lack of behavior, unsuitable opportunity, insufficient observation, or exclusion under a narrow coding rule. A phylogenetic comparison that records an unobserved species as biologically incapable will distort estimates of how often tool behavior emerged or was lost. The sampling process is therefore part of the scientific object St Amant & Horton (2008) Bastos et al. (2025).
Bastos and colleagues used publicly available videos and phylogenetic modeling to examine tool use in parrots, identifying examples of self-care tooling in species not previously documented in their compilation. The study illustrates how digitally mediated observations can expand behavioral discovery and challenge traditional assumptions of rarity. At the same time, opportunistic uploads are influenced by human contact, visibility, audience interest, and unequal video coverage across species. Model-based estimates of possible tool-use prevalence should not be conflated with directly observed prevalence, and claims about evolutionary association remain contingent on trait definitions and phylogenetic assumptions Bastos et al. (2025).
Convergence is a hypothesis about historical emergence and functional resemblance, not a guarantee of equivalent representation. A crow can shape a hook and a primate can fashion a probe, but shared outcomes need not depend on the same motor architecture or developmental program. Researchers can investigate convergence through common causal contrasts, functional performance calibrated for each body, and analysis of opportunities for learning. Evolutionary claims gain explanatory value when they identify the conditions favoring a capacity and the phylogenetic structure of its distribution rather than celebrating surprising resemblance in isolation Emery & Clayton (2004) Bastos et al. (2025).
Tool Behavior and the Development of Comparative Cognitive Constructs
Comparative psychology regularly uses constructs such as flexibility, planning, physical knowledge, and innovation, but each must be defined through testable indicators. Flexibility might mean transfer after a tool's material properties change; planning might mean transporting a future-use tool before confronting the obstacle; innovation might mean creating a new effective action without seeing a model. These measures do not automatically correlate within an individual or across species. A broad construct is justified when multiple tasks show a coherent pattern and when alternative contributors have been measured Osuna-Mascaró et al. (2023) Taylor et al. (2014).
Cognitive development can separate procedural expertise from causal generalization. An older individual may outperform juveniles on a familiar manufacturing routine while showing little advantage in a new physical inference task. Conversely, an individual may immediately discriminate a physical cue but struggle to perform a complex manipulation. Reviews of corvid development emphasize the interaction of experience, ecological conditions, and physical maturation. Comparative studies should therefore avoid treating ontogenetic timing of one observable skill as a fixed schedule for an entire psychological faculty Miller et al. (2026).
The formal distinction between capacity and performance can improve interpretation. A test outcome depends on latent competence, sensory access, opportunity, motivation, training, and task mechanics. Cross-species comparisons become stronger when each factor is manipulated or constrained and when several tasks are used to evaluate the same construct. The objective is to build a network of evidence with clear operational definitions, rather than infer a single hidden capacity from one striking demonstration or one unsuccessful attempt Seed et al. (2009) Penn et al. (2008) St Amant & Horton (2008).
Cultural Ecology and the Boundaries of Social Explanations
The ecological distribution of tools can be conditioned by learned local traditions. In chimpanzees, comparisons among populations and experimental exposure to honey-acquisition problems demonstrate why social history is part of a full explanatory account. Dolphin sponge carrying likewise combines a specialized benthic task with evidence relevant to transmission through maternal relationships. These cases help distinguish group-level stability of a behavior from immediate individual acquisition. They also underscore the need to identify what is socially transmitted: knowledge of a location, attention to an object, a movement pattern, or a relation between tool and goal Gruber et al. (2009) Krützen et al. (2005).
The distinction matters for causal cognition. An animal may learn a useful technique by copying a familiar sequence yet remain unable to explain its effectiveness or transfer it to a novel mechanism. Another may independently discover the physical relation and subsequently influence others through its products or actions. Both pathways can support behavioral traditions, and a single task may recruit them together. Evidence of stable population variation cannot, by itself, establish high-fidelity imitation, cumulative improvement, or representation of hidden mechanical causes Horner & Whiten (2005) Jelbert et al. (2018).
An integrated research design follows individual developmental histories, manipulates observational opportunity, and measures functional transfer after training. For example, an apprentice exposed only to finished tools can be compared with one exposed to tool construction and with an individual given equivalent access to raw materials alone. Such designs are constrained by animal welfare and by the rarity of some natural behaviors, but they permit more discriminating inferences about how material traditions persist. The scientific goal is to explain distributed acquisition without collapsing social, mechanical, and developmental processes into one category Tebbich et al. (2001) Jelbert et al. (2018) Gruber et al. (2009).
Competing Models of Physical Causal Cognition
The central explanatory alternatives concern how an animal connects observed conditions with effective action. Model-free reinforcement learning can gradually favor a tool choice when its consequences are rewarded. Perceptual-motor learning can calibrate action to material resistance and movement feedback. Feature-based generalization can extend a learned association to new stimuli with shared properties. Structured causal models posit relations among interventions, physical variables, and outcomes that support predictions in configurations not encountered during training. These classes can contribute jointly to behavior, and the scientific task is to identify observations on which their predictions diverge Penn et al. (2008) Seed et al. (2009).
A well-controlled experiment can vary one causal property while making superficial cues unreliable. If an animal chooses a heavier object after familiar training, both feature learning and knowledge of a mechanical effect may predict success. If an apparatus changes how weight operates while preserving its appearance, a theory sensitive to the new relation makes a more distinctive prediction. Conversely, flexible transfer may sometimes arise from sufficiently rich learning histories rather than an explicit mental representation of a causal graph. Inferences must therefore specify the learning model being challenged and the degree of novelty offered by the test Jelbert et al. (2014) Logan et al. (2014).
The debate becomes more productive when it shifts from asking whether an animal 'understands physics' to specifying what representation, if any, is necessary to explain a result. A sequence of successes can support sensitivity to material properties without providing evidence for reasoning about invisible mechanisms. A failure to invent an unfamiliar intervention can coexist with competence in familiar mechanical routines. Studies of crows and chimpanzees demonstrate that these abilities can dissociate across tasks, and a model adequate for one paradigm need not explain all performance across species Taylor et al. (2014) Hanus & Call (2008) Seed et al. (2009).
Cross-Species Experimental Validity and Mechanism Transfer
A valid comparison aligns scientific questions while respecting the body and ecology of each test species. Identical apparatuses may create unequal demands for animals with hands, bills, rostra, or flexible arms. Likewise, food rewards vary in attractiveness, visibility, and natural relevance. Researchers should establish that each species can perceive the relevant cue, operate the mechanism, and maintain adequate motivation before interpreting failures. Calibration experiments can independently assess grasping, lifting, and force application. These controls are essential because every tool test is simultaneously a cognitive and sensorimotor task Seed et al. (2009) St Amant & Horton (2008).
Mechanism transfer offers a powerful diagnostic. After establishing performance with one set of objects, an experiment introduces new shapes or materials that preserve the same relation, then changes the causal relation while keeping appearance familiar. A subject guided by functional relations should track the new dependency more reliably than one guided only by familiar objects, although experience-rich associative models may also generalize. The investigator should preregister the contrasts that discriminate candidate mechanisms and retain records of every training episode. Reported transfer should include both successful and unsuccessful novel conditions to reveal the boundaries of generalization Jelbert et al. (2014) Logan et al. (2014) Horner & Whiten (2005).
Statistical interpretation requires attention to repeated measures and individual differences. Trials within one bird or ape are not independent samples of the species, and selective reporting of unusually capable individuals can inflate the perceived prevalence of an effect. Hierarchical models and transparent individual-level summaries can help quantify uncertainty without treating a small captive sample as representative of a whole evolutionary lineage. Independent replications should reproduce the causal structure of an experiment while accommodating species-appropriate interfaces. The objective is construct-level replication rather than superficial duplication of hardware Taylor et al. (2014) Osuna-Mascaró et al. (2023) Miller et al. (2026).
Evidence Quality, Contradictions, and Reproducibility
The apparent contradictions within comparative tool research often arise because studies ask different questions under different constraints. One crow experiment may demonstrate reliable preference for sinking objects, while another documents failure to infer a hidden relation or produce a novel action. These results can jointly describe a cognitive profile rather than form mutually exclusive verdicts. Such differentiation is visible in the water-displacement literature and the causal-intervention comparisons. The productive question is which component of a hypothesized capacity each task successfully isolates Jelbert et al. (2014) Logan et al. (2014) Taylor et al. (2014).
Naturalistic evidence has high ecological relevance but often limited experimental control; captive experiments offer manipulable contingencies but may involve unusual environments and small samples. Historical case reports establish possibility, not necessarily frequency or prevalence. Field traditions implicate acquisition and local ecology but do not specify the exact internal learning process. Phylogenetic modeling can illuminate historical associations while remaining sensitive to missing data and trait definitions. These evidence classes should be combined through explicit assumptions rather than assigned a single global strength independent of the research question Rutz et al. (2016) Bastos et al. (2025) Boesch & Boesch (1990).
A transparent evidence record should state whether the full original study or its abstract was consulted, what variables were manipulated, which primary outcomes were measured, and how authors framed their limitations. Especially careful treatment is required for novelty, since prior experience may be undocumented, and for spontaneous behavior, which can be influenced by unobserved exposure. Negative results need sufficient power and demonstrated task comprehension before strong absence claims are justified. Reproducibility improves when materials, coding criteria, trial sequences, and individual outcomes are available for independent scrutiny Taylor et al. (2014) Seed et al. (2009) Miller et al. (2026).
Technology, Environmental Scaffolding, and Artificial Systems
External tools alter the effective organization of an animal's action by incorporating environmental objects into a functional sequence. This observation provides a conceptual bridge to embodied and distributed accounts of cognition, where performance depends on interactions among organism, environment, and manipulable artifacts. Such a description can be operationalized without assuming that every successful tool user maintains an abstract theory of physics. The demonstrable relation is between an action system and a material arrangement; claims about specific representations remain separate hypotheses assessed through transfer and intervention tests St Amant & Horton (2008) Visalberghi et al. (2007).
In the Artificial Era, engineered systems that select robotic implements or learn manipulation policies can be compared with biological tool users using common experimental principles. A system's ability to succeed with a familiar gripper is different from its ability to transfer a causal relation to unfamiliar materials or adapt when a tool's affordance is reversed. The useful analogy is methodological: both domains require operational definitions, tests of causal generalization, and controls for training distribution. Existing animal experiments cannot directly establish properties of artificial cognition, and engineered task performance cannot determine the phenomenology of a bird or primate Penn et al. (2008) Taylor et al. (2014).
This comparison can motivate future cross-system research in which behavioral flexibility is measured through intervention sequences, material substitution, and prospective tool choice. Experimental histories must be documented for each system, whether they involve individual experience, social exposure, or computational training. Evaluation should distinguish functionality, predictive robustness, and causal transfer as separate outcomes. The Artificial Era broadens the range of possible cognitive architectures under study while preserving the evidentiary discipline required to compare different bodies, learning systems, and technological environments Osuna-Mascaró et al. (2023) Bastos et al. (2025).
Open Research Problems and an Integrated Comparative Program
The first open problem is identifying the structure of causal knowledge across tasks. A species may display effective tool manufacture and yet fail to create a novel intervention from passive observation; another may master functional material choices without solving hidden-mechanism problems. Research should map these dimensions across independently designed paradigms rather than infer a general capacity from a familiar demonstration. Comparative test batteries can vary perceptual access, temporal separation, transport demand, novelty, and physical dependency while keeping track of shared motor and motivational requirements Taylor et al. (2014) Jelbert et al. (2014) Osuna-Mascaró et al. (2023).
A second priority is reconstructing development and transmission. Longitudinal studies of individuals in ecologically appropriate settings could identify how exploration, maturation, access to materials, and exposure to experts jointly organize tools and causal judgments. Controlled social-learning conditions should determine whether observers acquire a physical relation, an action sequence, or simply attention to a resource. Strong evidence of cultural transmission requires a plausible pathway between individuals and a careful assessment of environmental and genealogical alternatives. Dolphin sponging, chimpanzee foraging techniques, and crow manufacture offer distinct tests of this agenda Krützen et al. (2005) Gruber et al. (2009) Jelbert et al. (2018).
A third priority is correcting the taxonomic and methodological sampling imbalance. Observational effort is disproportionately concentrated in a few charismatic lineages, and ecological differences make absence claims unreliable. Opportunistic digital evidence can reveal overlooked behavior but introduces selection bias, while phylogenetic comparisons are sensitive to how tool use is defined. A coordinated research program should publish species-level inclusion rules, sampling effort, material affordances, and negative observations, and should report both individual and population variation. Improved evidence would permit more credible explanations of convergent evolution and the ecological conditions favoring technology Bastos et al. (2025) Rutz et al. (2016) St Amant & Horton (2008).
Conclusion
Tool use and physical causal cognition across species form a multidimensional domain of comparative psychological science. Field research demonstrates that animals in several distant lineages recruit external objects to modify access to resources, control physical contact, and build protective structures. Experimental work establishes capacities for selecting effective materials, manufacturing novel implements, combining components, adjusting sequences, and conditionally transporting tool sets. These achievements are supported by different combinations of ecological opportunity, anatomical organization, motor learning, social exposure, and information processing. The study of tools is therefore an inquiry into the organization of effective physical action and the mechanisms that make it flexible Boesch & Boesch (1990) Hunt (1996) Osuna-Mascaró et al. (2023) Finn et al. (2009).
The strongest evidence also defines the boundaries of interpretation. Crows succeed on several material discriminations while encountering difficulties with some hidden connections and novel interventions; chimpanzees' trap-task performance changes when motor-tool demands are removed; population traditions coexist with unresolved questions about what is transmitted socially. These contrasts show why causal sensitivity, action invention, planning, manufacture, and culture should be examined as distinguishable constructs. An explanation becomes compelling when it predicts transfer to unfamiliar configurations and survives controls for perceptual cues, reinforcement history, and motor opportunity Jelbert et al. (2014) Logan et al. (2014) Taylor et al. (2014) Seed et al. (2009) Horner & Whiten (2005).
The cumulative scientific synthesis supports a comparative program centered on operationally defined capacities, controlled physical interventions, developmental evidence, ecological field validation, and phylogenetically responsible inference. Tool behavior is well established beyond humans, and its distribution is broader than traditional sampling suggested. The open questions concern the organization, generality, acquisition, and historical emergence of the mechanisms supporting this behavior. Research in the Artificial Era can extend the same evaluative principles to engineered action systems, maintaining clear distinctions among observable performance, explanatory model, and empirically verified causal generalization Bastos et al. (2025) Miller et al. (2026) Penn et al. (2008).
References
- Auersperg, A. M. I., Szabo, B., von Bayern, A. M. P., & Kacelnik, A. (2012). Spontaneous innovation in tool manufacture and use in a Goffin’s cockatoo. Current Biology, 22(21), R903–R904. https://doi.org/10.1016/j.cub.2012.09.002
- Bastos, A. P. M., Claessens, S., Nelson, X. J., Welch, D., Atkinson, Q. D., & Taylor, A. H. (2025). Evidence of self-care tooling and phylogenetic modeling reveal parrot tool use is not rare. iScience, 28(4), 112156. https://doi.org/10.1016/j.isci.2025.112156
- Boesch, C., & Boesch, H. (1990). Tool use and tool making in wild chimpanzees. Folia Primatologica, 54(1–2), 86–99. https://doi.org/10.1159/000156428
- Emery, N. J., & Clayton, N. S. (2004). The mentality of crows: Convergent evolution of intelligence in corvids and apes. Science, 306(5703), 1903–1907. https://doi.org/10.1126/science.1098410
- Finn, J. K., Tregenza, T., & Norman, M. D. (2009). Defensive tool use in a coconut-carrying octopus. Current Biology, 19(23), R1069–R1070. https://doi.org/10.1016/j.cub.2009.10.052
- Fragaszy, D. M., Liu, Q., Wright, B. W., Allen, A., Brown, C. W., & Visalberghi, E. (2013). Wild bearded capuchin monkeys (Sapajus libidinosus) strategically place nuts in a stable position during nut-cracking. PLOS ONE, 8(2), e56182. https://doi.org/10.1371/journal.pone.0056182
- Gruber, T., Muller, M. N., Strimling, P., Wrangham, R., & Zuberbühler, K. (2009). Wild chimpanzees rely on cultural knowledge to solve an experimental honey acquisition task. Current Biology, 19(21), 1806–1810. https://doi.org/10.1016/j.cub.2009.08.060
- Hanus, D., & Call, J. (2008). Chimpanzees infer the location of a reward on the basis of the effect of its weight. Current Biology, 18(9), R370–R372. https://doi.org/10.1016/j.cub.2008.02.039
- Horner, V., & Whiten, A. (2005). Causal knowledge and imitation/emulation switching in chimpanzees (Pan troglodytes) and children (Homo sapiens). Animal Cognition, 8(3), 164–181. https://doi.org/10.1007/s10071-004-0239-6
- Hunt, G. R. (1996). Manufacture and use of hook-tools by New Caledonian crows. Nature, 379, 249–251. https://doi.org/10.1038/379249a0
- Jelbert, S. A., Taylor, A. H., Cheke, L. G., Clayton, N. S., & Gray, R. D. (2014). Using the Aesop’s fable paradigm to investigate causal understanding of water displacement by New Caledonian crows. PLOS ONE, 9(3), e92895. https://doi.org/10.1371/journal.pone.0092895
- Jelbert, S. A., Hosking, R. J., Taylor, A. H., & Gray, R. D. (2018). Mental template matching is a potential cultural transmission mechanism for New Caledonian crow tool manufacturing traditions. Scientific Reports, 8, 8956. https://doi.org/10.1038/s41598-018-27405-1
- Krützen, M., Mann, J., Heithaus, M. R., Connor, R. C., Bejder, L., & Sherwin, W. B. (2005). Cultural transmission of tool use in bottlenose dolphins. Proceedings of the National Academy of Sciences, 102(25), 8939–8943. https://doi.org/10.1073/pnas.0500232102
- Logan, C. J., Jelbert, S. A., Breen, A. J., Gray, R. D., & Taylor, A. H. (2014). Modifications to the Aesop’s fable paradigm change New Caledonian crow performances. PLOS ONE, 9(7), e103049. https://doi.org/10.1371/journal.pone.0103049
- Mann, J., Sargeant, B. L., Watson-Capps, J. J., Gibson, Q. A., Heithaus, M. R., Connor, R. C., & Patterson, E. (2008). Why do dolphins carry sponges? PLOS ONE, 3(12), e3868. https://doi.org/10.1371/journal.pone.0003868
- Miller, R., Claisse, E., Timulak, A., & Clayton, N. S. (2026). Development of cognition in corvids. Animal Cognition, 29, Article 76. https://doi.org/10.1007/s10071-026-02088-8
- Osuna-Mascaró, A. J., O’Hara, M., Folkertsma, R., Tebbich, S., Beck, S. R., & Auersperg, A. M. I. (2023). Flexible tool set transport in Goffin’s cockatoos. Current Biology, 33(5), 849–857.e4. https://doi.org/10.1016/j.cub.2023.01.023
- Penn, D. C., Holyoak, K. J., & Povinelli, D. J. (2008). Darwin’s mistake: Explaining the discontinuity between human and nonhuman minds. Behavioral and Brain Sciences, 31(2), 109–130. https://doi.org/10.1017/S0140525X08003543
- Rutz, C., Klump, B. C., Komarczyk, L., Leighton, R., Kramer, J., Wischnewski, S., Sugasawa, S., Morrissey, M. B., James, R., St Clair, J. J. H., Switzer, R. A., & Masuda, B. M. (2016). Discovery of species-wide tool use in the Hawaiian crow. Nature, 537, 403–407. https://doi.org/10.1038/nature19103
- Seed, A. M., Call, J., Emery, N. J., & Clayton, N. S. (2009). Chimpanzees solve the trap problem when the confound of tool-use is removed. Journal of Experimental Psychology: Animal Behavior Processes, 35(1), 23–34. https://doi.org/10.1037/a0012925
- St Amant, R., & Horton, T. E. (2008). Revisiting the definition of animal tool use. Animal Behaviour, 75(4), 1199–1208. https://doi.org/10.1016/j.anbehav.2007.09.028
- Taylor, A. H., Cheke, L. G., Waismeyer, A., Meltzoff, A. N., Miller, R., Gopnik, A., Clayton, N. S., & Gray, R. D. (2014). Of babies and birds: Complex tool behaviours are not sufficient for the evolution of the ability to create a novel causal intervention. Proceedings of the Royal Society B: Biological Sciences, 281(1787), 20140837. https://doi.org/10.1098/rspb.2014.0837
- Tebbich, S., Taborsky, M., Fessl, B., & Blomqvist, D. (2001). Do woodpecker finches acquire tool-use by social learning? Proceedings of the Royal Society B: Biological Sciences, 268(1482), 2189–2193. https://doi.org/10.1098/rspb.2001.1738
- Visalberghi, E., Fragaszy, D., Ottoni, E., Izar, P., de Oliveira, M. G., & Andrade, F. R. D. (2007). Characteristics of hammer stones and anvils used by wild bearded capuchin monkeys (Cebus libidinosus) to crack open palm nuts. American Journal of Physical Anthropology, 132(3), 426–444. https://doi.org/10.1002/ajpa.20546
- von Bayern, A. M. P., Danel, S., Auersperg, A. M. I., Mioduszewska, B., & Kacelnik, A. (2018). Compound tool construction by New Caledonian crows. Scientific Reports, 8, 15676. https://doi.org/10.1038/s41598-018-33458-z
- Weir, A. A. S., Chappell, J., & Kacelnik, A. (2002). Shaping of hooks in New Caledonian crows. Science, 297(5583), 981. https://doi.org/10.1126/science.1073433
Social Exposure, Cultural Traditions, and Causal Discovery
Tool traditions can persist when individuals encounter distinctive techniques through a population's social environment. Research on chimpanzees has long treated geographically varying repertoires as evidence prompting cultural explanation, while also requiring ecological controls. Gruber and colleagues used an experimental honey-acquisition task with wild chimpanzees to probe how established population knowledge influenced solutions. The results illustrate how prior local practice can channel responses to a novel problem. Such evidence concerns the distribution and transmission of techniques, and the physical knowledge supporting a technique must still be evaluated through independent tests Gruber et al. (2009) Boesch & Boesch (1990).
Horner and Whiten's opaque-versus-transparent experimental conditions separated copying of demonstrated actions from sensitivity to visibly unnecessary steps. Children and chimpanzees differed in their treatment of apparently irrelevant movements, with the visibility of a mechanism altering the chimpanzees' copying strategy. The finding is important because observational learning can support either action imitation or outcome-focused reproduction, and the same successful final result may arise through different social processes. Transparent conditions also provide visual information unavailable in opaque apparatuses; interpreting group contrasts therefore requires careful attention to causal access and task demands Horner & Whiten (2005).
Social transmission is closely connected to material cognition but remains an independently analyzable phenomenon. The presence of a distinctive tool pattern may reflect social copying of motions, exposure to finished artifacts, enhancement of attention toward a resource, or independent discoveries within a shared ecology. The existence of cultural variation thus does not establish a particular cognitive theory of tool function. Combining field distribution, experimental social-learning conditions, and tests of functional transfer would help identify how a tradition can transmit both a practical solution and information about its causal properties Gruber et al. (2009) Jelbert et al. (2018) Tebbich et al. (2001).