Comparative cognition is the study of cognition across the tree oflife. Nonhuman animals (hereafter animals) display a broad and oftensurprising range of cognitive abilities. Kea (a species of largeparrot endemic to New Zealand) solve problems using domain-generalstatistical inference (Bastos & Taylor 2020), Eurasian jays aredeceived by magic tricks (Garcia-Pelegrin et al. 2021), and bumblebeesmay be conscious (Gibbons et al. 2022). Researchers in the field ofcomparative cognition seek to understand the cognitive mechanismsgiving rise to animal behavior, including their evolutionary,developmental, and socioecological history. Comparative cognition isan interdisciplinary field bringing together tools from ecology,ethology, cognitive science, developmental psychology, evolutionarybiology, and neuroscience, among others (Shettleworth 2009). There isalso a robust philosophical literature on the methods of comparativecognition. Philosophers aim to address questions such as, “whatis evidentially required to show that animals have causalreasoning?” and “how can we avoid anthropocentricism inthe study of nonhuman minds?” How best to define“cognition” is an area of debate (Allen 2017). For ourpurposes, we can follow Sara Shettleworth’s broadcharacterization of cognition as “the mechanisms by whichanimals acquire, process, store and act upon information from theenvironment” (2010a: 4; see also entry onanimal cognition).

• 1. Evolution, Development, and Culture
  • 1.1 Convergent Evolution
  • 1.2 Homology and Development
  • 1.3 Culture
  • 1.4 Cognition in the Wild and Captivity
• 2. Eliminating Alternative Explanations
  • 2.1 Null Hypotheses in Comparative Cognition
  • 2.2 Cognitive and Associative Explanations
  • 2.3 The Role of Epistemic Values
  • 2.4 The Role of Non-Epistemic Values
• 3. Methodological Challenges
  • 3.1 Anthropomorphism
  • 3.2 Anthropocentrism
  • 3.3 Underdetermination
  • 3.4 Signatures and Dimensions
  • 3.5 Reproducibility
• 4. Comparative Cognition and AI
• Bibliography
• Academic Tools
• Other Internet Resources
• Related Entries

1. Evolution, Development, and Culture

Research in comparative cognition takes place against a backdrop ofevolutionary, developmental, and socioecological considerations.Natural selection leads to the convergence and divergence of cognitivetraits over evolutionary time. Cognitive traits might converge becausetwo species face similar socioecological problems, such as navigatinga similar foraging environment. Alternatively, two species might sharecognitive abilities because they have inherited them from a sharedancestor or have developmental modules in common. Finally, cognitiveabilities interact with evolution in ways that not only alter aspecies trajectory through phenotypic space, but their evolvability(Brown 2013). Considerations such as these help researchers predictand explain cognitive and behavioral abilities across taxa:

Only by investigating proximate (mechanism and ontogeny) and ultimate(phylogeny and function) causes can we fully understand the nature andorigin of behavior. (Krupenye & Call 2019: 16)

This section presents several key ways in which evolutionary,developmental, and socioecological considerations shape methods incomparative cognition.

1.1 Convergent Evolution

The evolution of wings is a paradigm case of convergent evolution.Wings have the function of sustaining flight and have evolved in manydifferent organisms, including birds, bats, and insects. Thestructures comprising wings differ across these groups—bat wingsconsist of extended digits, bird wings evolved from an extendedforelimb, and insect wings are outgrowths of exoskeleton. Althoughwings in these groups differ in their underlying structures, theyshare features like rigidity, which allow them to perform the samefunction—producing lift and enabling flight. Similarities intraits such as these can be explained in part by appealing to sharedselection pressures.

Adaptive considerations such at these guide theorizing in comparativecognition. In this case, a “psychological trait is treated as adesign problem” (Ereshefsky 2007: 670). One determines thefunction that a psychological trait serves and asks what selectionpressures might have led to that evolutionary outcome. For example,research on western scrub jays suggests that they have episodic-likememory or the ability to remember the “what, where, andwhen” features of specific events (Clayton et al. 2001). Scrubjays are “scatter-hoarders”, hiding food in multipleplaces for later recovery. Caching behavior such as this enablesindividuals to survive in environments with fluctuating food supplies.When recovering their caches, scrub jays remember where and when theyhave cached a specific food item. For example, they avoid searchingfor perishable food (like a worm) if its window of freshness haselapsed (Clayton & Dickinson 1998). Recent work on cuttlefishsuggests that they face similar foraging challenges toscatter-hoarding birds. They live in environments with variable foodsupplies and appear to keep track of what they have eaten, as well aswhere and when they ate (Jozet-Alves, Bertin, & Clayton 2013;Schnell, Clayton, et al. 2021a). Cuttlefish are cephalopod mollusksand diverged from vertebrates like scrub jays over 550 million yearsago (Figure 1). Given this phylogenetic distance, if cuttlefish and western scrubjays share the capacity for episodic-like memory, this cognitive traitmay have evolved independently in these two taxa as a common solutionto similar ecological problems.

Figure 1. Distantly related organismsmay share similar cognitive abilities because they face similarsocioecological selection pressures over evolutionary time. Imageadapted from Schnell, Amodio, Boeckle, and Clayton (2021b: 4). [Anextended description of figure 1 is in the supplement.]

Alexandra Schnell and colleagues argue that studying phylogeneticallydistant species like chimpanzees, corvids, and cephalopods, allowsresearchers to better understand the selection pressures that giverise to “complex cognition” like episodic-like memory,future planning, causal reasoning, and imagination (Schnell et al.2021b; see also van Horik et al. 2012, Powell et al. 2017, Amodio etal. 2019). For example, the Social Intelligence Hypothesis holds thatcomplex cognition is a product of selection pressures that arise fromliving in complex social groups (see Byrne & Whiten 1988).However, many cephalopods live short and solitary lives. If thesecephalopods do have cognitive capacities such as causal reasoning,these may have evolved in response to selective pressures emergingfrom something other than complex social environments. Focusing onsocially complex vertebrates like chimpanzees and corvids alone makesit “difficult to uncouple the effects of ecological and socialpressures on cognitive evolution” (Schnell et al. 2021a: 172).If it can be shown that organisms such as corvids and cephalopods(e.g., the octopus) have the same cognitive capacities (e.g., causalreasoning), then researchers might be better placed to infer theenvironmental demands that led to the emergence of adaptations inthese distantly related taxa.

1.2 Homology and Development

There are several approaches to determining what counts as ahomologous trait. The “phylogenetic approach” holds thattwo traits are homologous when they are derived from a common ancestor(Brigandt 2007). Humans and chimpanzees have similarphysical-cognitive abilities like tool use (Herrmann et al. 2007).Under the phylogenetic approach, these abilities are homologousinsofar as they were inherited from the evolutionary ancestor thathumans and chimpanzees shared 6–8 million years ago. The“developmental approach” to homology holds that two traitsare homologous when they’re produced by the same developmentalmodule (Ereshefsky 2012). Mammalian vertebrae, arthropod limbs, andbird feathers are often viewed as homologues in this sense. Some holdthat homology is best understood as having both phylogenetic anddevelopmental components (Ereshefsky 2007, 2012). Under this view, adevelopmental module imposes proximate constraints on how a homologoustrait is constructed, while evolution explains why that developmentalmodule is found across taxa (namely, through common descent).

Some philosophers argue that psychologists should focus on identifyinghomologous rather than convergent traits, given their interest in themechanisms responsible for behavior (see Matthen 2007; Clark 2010).For example, Paul Griffiths (1997, 2007a, 2007b) argues thatidentifying cognitive traits as convergent is epistemically moredemanding than identifying cognitive traits as homologous.Explanations that appeal to convergent evolution often adopt alock-and-key model of adaptation where a trait is viewed as a solutionto an existing socioecological problem (e.g., episodic-like memory asa solution to the foraging problem that western scrub jays facetoday). However, Griffiths argues that it is unlikely that mindsevolved in response to contemporary socioecological problems. Instead,the structure and organization of an organism’s phenotype andniche coevolve. Moreover, what counts as a “problem”depends on many factors, such as resource constraints and thedevelopmental plasticity of an organism. Griffiths writes:

Problems whose solutions cannot be developmentally dissociated must besolved as a single problem and so are not separate problems from thestandpoint of adaptive evolution. (2007b: 203)

To provide a compelling case of convergent evolution, one must takesuch constraints under account. In the case of complex cognitivetraits, researchers often know little about the resource anddevelopmental constraints at play in an organism’s evolutionaryhistory. In this way, identifying cognitive traits as convergent maybe epistemically demanding.

1.3 Culture

In addition to evolutionary and developmental factors, comparativecognition researchers aim to understand how cultural factors affectcognitive capacities. Philosophers and scientists have also arguedthat understanding animal cultures has important implications forconservation efforts (Brakes et al. 2019) and animal welfare(Fitzpatrick & Andrews 2022). Broadly, culture can be definedas

information transmitted between individuals or groups, where thisinformation flows through and brings about the reproduction of, and alasting change in, the behavioral trait. (Ramsey 2017: 348; see alsoRamsey 2013; Andrews 2015 [2020]: chapter 8; and entry oncultural evolution)

For example, the skills required to use a tool effectively might betransmitted through social learning (such as copying), creating alasting change in a recipient’s subsequent behavioral abilitiesand reproductive fitness. Cultural transmission has been described asa form of “soft inheritance” or the idea that organismscan inherit phenotypic variation that is the result of non-geneticeffects (Mayr 1982). Eva Jablonka and colleagues argue that manyanimal traditions are transmitted in this way (Avital & Jablonka2000; Jablonka & Lamb 2008). For example, the behavioralinnovation of opening milk bottles spread quickly through Eurasianblue tit populations in the early twentieth century. Studies suggestthat this innovation spread through a mixture of local enhancement(naïve birds learning that milk bottles are a source of food byspending time near birds capable of exploiting this food source) andobservational learning (naïve birds copying the behavior ofexperienced innovators) (Aplin, Sheldon, & Morand-Ferron2013).

One advantage of understanding culture in terms of soft inheritance isthat some of the methods and models employed in the study ofbiological evolution can be applied to the study of culture (Mesoudi,Whiten, & Laland 2006). Rachael Brown (2017a) also argues thatanimal traditions are important for understanding genetic evolution.For example, recent studies suggest that the diversification of beakmorphologies in Galápagos finches is the result of behavioralforaging innovations spread through cultural transmission. Thepresence and spread of these behavioral innovations (e.g., usingone’s beak to puncture the skin and drink the blood of seabirds, as in the case of “vampire finches”) may have beenrequired to allow sufficient selective pressure for morphologicaladaptation to occur. Despite the links between culture and geneticevolution, there remains disagreement regarding whether the modernsynthesis in evolutionary biology should be extended to includeprocesses such as soft inheritance (see Pigliucci 2007; Laland et al.2015). Nevertheless, research on animal culture is flourishing withobservational and experimental studies revealing both its cognitiveunderpinnings and effects (see Whiten 2022 for a recent review).

1.4 Cognition in the Wild and Captivity

As noted, comparative cognition is a field that draws on methods froma variety of disciplines, including ethology, experimental psychology,and neuroscience. Philosophers and scientists have debated therelative merits of these methods for understanding animal minds. Acentral method in ethology is fieldwork on wild animals, for instance,while experimental psychology relies primarily on laboratory studieson captive animals. Wild and captive animals often lead differentkinds of lives. For example, captive chimpanzees tend to have regularaccess to food, extensive experience with human objects (includingobjects designed for enrichment), and live in a relatively smallspace. Wild chimpanzees typically live in territories that are manysquare kilometers in size, spend much time foraging, and must overcomelife-threatening problems not typically found in captivity, such ascompetition with neighboring groups and predators (Boesch 2022). Incomparative cognition, many researchers stress the importance ofobserving animals in their natural environment. From an ethologicalperspective, this is crucial for understanding how animal mindsevolved because “it is in this situation that natural selectionacts” (Healy & Hurly 2003: 326).

Other researchers note that control conditions like those found in thelaboratory are necessary for investigating animal minds. For example,Heyes and Dickinson (1990) argue that identifying the mental statesresponsible for a given action requires determining

what the animal would have done if its circumstances had beendifferent in certain, specifiable respects from those in which theaction actually occurred. (1990: 88; but see Allen & Bekoff 1997:chapter 9)

Observing animals in their natural environment often precludes varyingthe environment in systematic ways, while laboratory studies aredesigned to control for alternative explanations and nuisancevariables (seesection 2). Some researchers also argue that “unnatural” environmentsmight direct development and learning in ways that supports theemergence of new or enhanced cognitive capacities. Captive great apes,for example, develop problem-solving and communicative abilities notfound in their wild conspecifics, such as tool use, pointing, and signlanguage (Tomasello & Call 2008; Bandini & Harrison 2020).Studying animals outside of their natural environment might provideinsight into their cognitive potential when appropriately scaffoldedby the environment.

Overall, most researchers acknowledge that a plurality of methods isneeded for understanding animal minds. For example, Michael Tomaselloand Josep Call write regarding primate cognition research that,

fieldwork is primary. It tells us what animals do; it sets theproblem. But then if we wish to figure out what is the nature of thecognitive skill, if any, underlying some activity in the wild, we needexperiments. (2008: 451)

Kristin Andrews documents different possible sources of bias in thefield and laboratory. For example, in the field, ethograms might leadto confirmation bias, while in the laboratory, the importance of therelationships between human experimenters and study participants isoften underemphasized (see Andrews 2020: 42–61). Andrewsconcludes that comparative psychologists should aim to employ andintegrate a variety of methods for studying animal minds, rather thanseek to eliminate bias. Colin Allen and Marc Bekoff (1997) similarlyemphasize the need for an interdisciplinary approach:

Science is not likely to make complete contact with the nature ofanimal minds at any single point—many methods will be useful,and competing hypotheses should be evaluated. (1997: 180)

2. Eliminating Alternative Explanations

Comparative cognition researchers often test hypotheses about animalminds through behavioral studies or experiments. These studies aredesigned to determine whether a species, group, or individual animalbehaves as one would expect on the assumption that a given hypothesisis true. Often these hypotheses specify to some degree the cognitiveprocesses thought to underlie a given suite of behaviors. AsShettleworth notes, researchers seek

not just to confirm that animals are (or are not) capable of doingsomething “clever” but to discoverhow they dowhat they do. (Shettleworth 2013: 4)

In this section, we consider the role that statistical nulls,alternative explanations, epistemic values, and non-epistemic valuesplay in the evaluation of hypotheses in comparative cognition.

2.1 Null Hypotheses in Comparative Cognition

A central element of experimental work in the sciences involvescontrolling for extraneous or “nuisance” variables.Comparative cognition is like other sciences in employing experimentaland statistical methods to control for such variables (Bausman &Halina 2018; Dacey 2023). The aim of an experiment is generally todetermine whether there is a relationship between the independent anddependent variables. For example, crows drop nuts onto roads andretrieve the cracked nuts (Shettleworth 2010a). Are they doing thisbecause they’ve learned that cars can be used as“nutcrackers”? Or has the behavior of dropping nuts from aheight evolved over many generations (perhaps before the advent ofcars)? As a first step towards answering these questions, one mightexperimentally investigate whether the presence of approaching cars ona road (the independent variable) affects nut-dropping behavior incrows (the dependent variable) (Cristol et al. 1997). To detect thiseffect, however, one must control for the “noise” createdby extraneous variables—i.e., those variables that are alsolikely to affect the dependent variable. In the case of foragingcrows, extraneous variables may include time of day, season, proximityto nut trees, etc. In an experiment, the statistical null hypothesistypically holds that any differences observed in the dependentvariable can be attributed to extraneous variables and thus there isno evidence that the independent variable has had an effect (see Sani& Todman 2006). Whether an observed difference in the dependentvariable can be attributed to the independent variable (i.e., whetherthe difference is statistically significant) depends on one’stest statistic, which reflects the variability of the data due toknown extraneous variables. If proximity to nut trees can explain thedifference observed in nut-dropping behavior, for example, then thereis insufficient evidence to conclude that nut-dropping behavior hasbeen affected by the presence or absence of approaching cars.

The term “null hypothesis” is sometimes used more broadlyto refer to an alternative or competing hypothesis, rather than astatistical null hypothesis. For example, associative learning,behavior reading, and hypotheses that deny animals“special” human-like cognitive abilities are oftenreferred to as “null hypotheses” (see Hanus 2016;Dickinson 2012; Andrews & Huss 2014; Mikhalevich 2015). Thesehypotheses are typically contrasted with those that attribute complexor human-like cognitive abilities to animals like causal reasoning ortheory of mind. One problem with this broad use is that the term“null” suggests that a hypothesis should be epistemicallyprivileged (in the sense that it must be rejected before accepting anyalternative hypothesis) when in many of these cases it is not clearthe purported null should be privileged in this way. As Andrews andHuss (2014) write:

the onus is on the skeptic to explain why the skeptical hypothesis andnot the optimistic hypothesis that animals do have psychologicalproperties is the proper null hypothesis. (2014: 720–721)

Statistical null hypotheses are epistemically privileged in the sensethat they must be rejected before one can conclude that the variableof interest (the independent variable) has had an effect. Thisinference strategy is justified in the context of statisticalhypothesis testing due to the nature of experimental design andinferential statistics (Bausman & Halina 2018). However, thefeatures that justify this strategy in the statistical case are notpresent in hypothesis evaluation more generally (the latter lacks atest statistic, for example). Instead, one must supply compellingreasons for preferring one hypothesis (such as associative learning)over another (such as causal reasoning). Such reasons may betheoretical (e.g., that one hypothesis has epistemic virtues likepredictive power that the other lacks) or empirical (e.g., that onehypothesis already has independent empirical support, while the otherdoes not). The next three sections discuss hypothesis evaluation incomparative cognition in this broader sense. For additional discussionof statistical null hypotheses, null modeling, and default models, seeBausman 2018, Zhang 2020, and Dacey 2023.

2.2 Cognitive and Associative Explanations

As noted above, to provide compelling evidence that an animal hascognitive capacities like episodic memory or theory of mind,researchers aim to eliminate plausible alternative explanations forbehavior. This research strategy is not unique to comparativecognition but is found across the sciences. For example, Julian Reissargues that a hypothesis is warranted insofar as one has eliminatedalternative hypotheses. That is, the strength of the warrant dependson how many alternatives have been eliminated and how compelling orsalient those alternatives are (Reiss 2015). In comparative cognition,associative learning is regularly advanced as a salient alternativethat must be eliminated before concluding that an animal has complexcognition. As Starzak and Gray (2021) write,

Over and over again the familiar refrain is, “do animals havecomplex human-like cognitive abilities or can their behavior beexplained in terms of simpler processes such as associativelearning?”. (2021: 2)

What is associative learning? Broadly, associative learning is a classof learning mechanisms characterized by a change in associationbetween two or more variables. Two common forms of associativelearning are classical (Pavlovian) conditioning and operant(instrumental) conditioning. In classical conditioning, a stimuluscomes to elicit a response in an organism because it has becomeassociated with another stimulus. For example, drawing on the studiesof the Russian psychologist Ivan Pavlov, dogs naturally salivate inresponse to the taste of food (this response does not require trainingor conditioning, so is anunconditioned response to anunconditioned stimulus). However, one can pair anotherstimulus, like the sound of a bell, with the arrival of food. Overtime, dogs will learn to associate the food with the sound of thebell, such that the sound of the bell will become sufficient on itsown to elicit salivation (the sound has become aconditionedstimulus leading to aconditioned response). In operantconditioning, an organism’s behavior changes in response to theconsequences of that behavior. For example, a cat might behave in manyways while interacting with a puzzle box (which, say, contains foodthe cat likes) with some behaviors resulting in the cat successfullyopening the puzzle box by accident. Over time, the successfulbehaviors will be strengthened, and the unsuccessful behaviorsweakened, such that the cat will be able to quickly open the puzzlebox as a result of operant conditioning alone. The tendency forbehaviors to be strengthened or weakened in response to their positiveand negative consequences respectively is known as Thorndike’slaw of effect after the American psychologist Edward Thorndike.

It is often possible to formulate plausible associative learningaccounts that can accommodate data originally believed to supporthypotheses like causal reasoning or theory of mind (for examples, seeTaylor, Medina, et al. 2010; Heyes 2012; Halina 2022). In many cases,this is taken to undermine the original interpretation of the databecause associative learning is taken to be preferred as anexplanation of animal behavior over capacities that require“complex” or human-like cognitive abilities (Buckner 2011;Hanus 2016). One justification for this preference is that associativelearning is “simpler” than alternative cognitiveexplanations and thus should be preferred all else being equal (seesection 2.3). Philosophers have argued, however, that appealing to simplicity alonein this context is not enough. Instead, one must show that theassociative and cognitive explanations in question can in fact beordered by complexity and that there are good reasons for preferringthe simpler explanation in a particular case (e.g., because the morecomplex cognitive ability presupposes or requires the simplerassociative ability) (Heyes 2012; Meketa 2014; Dacey 2016, 2017). Asecond justification for preferring associative learning as anexplanation is that it is phylogenetically widespread. If associativelearning is phylogenetically widespread, then it is reasonable toassume that many animals will use it to solve physical and socialcognition tasks. Irina Mikhalevich (publishing as Meketa 2014) refersto this as the “taxonomic ubiquity argument” (2014: 737).One concern with this approach is that if it is standard practice tobelieve that associative learning trumps complex cognition (in caseswhere both are consistent with the data), then this practice mightitself result in associative learning appearing more taxonomicallyubiquitous. Thus, relying on such results to justify associativelearning as a preferred hypothesis appears to beg the question (Meketa2014, but see Heyes 2012).

The above discussion presumes that associative learning and complexcognition are mutually exclusive—that one or the other, but notboth, are needed to explain a behavior of interest. One explanationfor this assumption is that it is simply part of typical definitionsof complex cognition and associative learning. For example, AmandaSeed and colleagues note that complex cognition is often“defined by exclusion, rather than by some positive assessmentof the mechanisms underpinning it” (Seed, Emery, & Clayton2009: 402). Such an approach identifies a flexible or novel behavioras best explained by complex cognition when it “cannot easily beexplained in terms of simple conditioning, or hardwired actionpatterns” (Seed et al. 2009: 410). Under this view, complexcognition and associative learning are mutually exclusive bydefinition. However, researchers have expressed concerns with theassociative-cognitive distinction (Allen 2006; Buckner 2017). Oneconcern is that associative learning models are now sophisticatedenough to capture paradigm cases of cognitive processes (see Buckner2011, forthcoming; Dickinson 2012). Another concern is that cognitiondoes not lend itself to precise definition or clear categorization.Allen (2017), for example, advances a “relaxed pluralism”about cognition, allowing for multiple incompatible accounts. Thosewho reject the associative-cognitive distinction often urgeresearchers to focus on more specific capacities instead. As DavidPapineau and Cecilia Heyes write,

research should refocus on specific explanations of how animals dospecific things, rather than on the presence or absence of somegeneral or ideal form of rationality that contrasts with associativemechanisms. (2006: 187)

Dacey (2016) also argues that the concept of “association”is best understood as a “highly abstract filler term” thatcan be implemented by a wide range of cognitive mechanisms (2016:3763). Understood this way, associative learning and complex cognitionare not mutually exclusive: both may be needed to explain a givenbehavior.

2.3 The Role of Epistemic Values

Epistemic values can be broadly characterized as those features (of atheory or a theory in relation to evidence) that scientists valuebecause they’re believed to lead to epistemic goods like truthand understanding. Heather Douglas distinguishes between thoseepistemic values that are minimal criteria versus ideal desiderata.Minimal criteria are those values that are epistemically necessary:they are “genuinely truth assuring” and their absenceindicates “something is wrong with our theory” (Douglas2013: 799). These include features like internal consistency andempirical adequacy. In contrast, ideal desiderata are not required,but are useful and often provide assurance that we are on the righttrack or that, if we’re not on the right track, we will find outsooner rather than later. Ideal desiderata include values likesimplicity, unification, and novel prediction. For example, a simplertheory may be easier to use and a theory capable of making novelpredictions might assure researchers that the theory is notoverfitting the available data (Douglas 2009a, 2009b; Douglas &Magnus 2013).

Comparative cognition researchers also appeal to epistemic values whenevaluating theories and hypotheses. Researchers minimally expecttheories to be internally consistent and empirically adequate. Whentwo competing theories are both consistent with the availableempirical data, researchers evaluate them with respect to otherepistemic values. For example, in the context of chimpanzee theory ofmind research, Tomasello and Call (2006) note that twohypotheses—theory of mind and learned behavioralrules—account for the available experimental data. However, theyargue that theory of mind provides a unified explanation of theexisting data, while the claim that chimpanzees rely on learnedbehavioral rules requires positing a unique behavioral rule for eachexperimental result. They also express concern that behavioral rulesare ad hoc. Fletcher & Carruthers (2013) concur, maintainingthat

the behavior-rule account is only capable of “predicting”new findings after they are discovered, postulating a novelbehavior-rule for the purpose. (2013: 88)

Here, we find researchers arguing in favor of one explanatory theoryover another based on epistemic values like predictive power,unification, and coherence.

One epistemic value that has received a lot of attention fromscientists and philosophers working in comparative cognition issimplicity or parsimony (Dacey 2016). As we saw in the previoussection, the epistemic value of simplicity is often attributed toassociative learning, leading to debates about whether associativelearning is truly simple and, if so, what this means for theorychoice. However, another reason why simplicity has received a lot ofattention is its connection to Morgan’s Canon—amethodological principle widely adopted in research on animalcognition. This principle holds that when there are two or moreplausible explanations for an animal’s behavior, psychologistsshould favor the explanation that appeals to “lower”rather than “higher” psychical faculties (Fitzpatrick2008).

Many philosophers have rejected Morgan’s Canon as a usefulmethodological principle. First, they argue that what counts as“sophisticated” is often ambiguous (Sober 2005;Fitzpatrick 2008; Meketa 2014). Simplicity has been used todistinguish between sensory and conceptual reasoning,stimulus-response mechanisms and conscious thought, associative andnon-associative learning. Moreover, in all these cases, the contrastseems problematic (Andrews & Huss 2014). Second, if Morgan’sCanon is interpreted as no more than the dictum that “simpler isbetter or more likely to be true”, then it is a poor researchguide. There are numerous parsimony considerations one can make incomparative research (ontological parsimony, explanatory parsimony,evolutionary parsimony): often these considerations pull in differentdirections and rarely do they favor the conclusion urged byMorgan’s Canon (Sober 2005; Fitzpatrick 2008; Dacey 2016).Finally, philosophers of science have argued that justifyingsimplicity as a virtue is context dependent—that is, it is not avirtue that applies across the board (see Longino 2008). As ElliottSober writes:

When a scientist uses the idea [of parsimony], it has meaning onlybecause it is embedded in a very specific context of inquiry. Onlybecause of a set of background assumptions does parsimony connect withplausibility in a particular research problem. What makes parsimonyreasonable in one context therefore may have nothing in common withwhy it matters in another. (Sober 1990 [1994: 140])

If Sober is correct, then determining whether one should prefer thesimpler explanation will depend on the case at hand. Insofar asMorgan’s Canon is a general methodological principle, meant tohold across a wide range of disparate cases or a “blanket biastowards endorsing lower explanations”, it should be rejected,according to this view (Fitzpatrick 2008: 243). Comparative cognitionresearchers have reached similar conclusions. For example, Tomaselloand Call (2006) write,

we are not strong proponents either of parsimony (unless one clearlydefines the criteria for parsimony) or of Morgan’sCanon—certainly not as substitutes for grappling with data whenthere is plenty of it. (2006: 381)

It is worth noting that not all accounts of Morgan’s Canoncharacterize it as a simplicity principle. For example, SimonFitzpatrick and Grant Goodrich (2017) argue that when one looks atConwy Lloyd Morgan’s own formulation of the canon, it does nottake the form of a simplicity principle. Indeed, Morgan explicitlyrejected simplicity as a criterion for choosing between competingexplanations. Instead, he held that one should choose the explanationthat best coheres with our observations and broader backgroundknowledge. One such piece of background knowledge for Morgan was that“higher” faculties evolve from “lower” ones;thus, the former will be rarer in nature than the latter, and thisshould inform our choice of explanation. Similarly, Adrian Currie(2021) argues that Morgan’s Canon is best understood as holdingthat evolutionary ancient and highly evolvable traits are more likelyto be found across the tree of life. Thus, explanations that appeal tosuch “lower” traits should be preferred over those that donot. Under this view, whether a cognitive trait is simple or complexis not relevant—it is its expected taxonomic distribution andevolvability that matters.

2.4 The Role of Non-Epistemic Values

Philosophers have argued that non-epistemic values, such as practicaland ethical ones, play an important role in science. Such values areneeded to evaluate the risk of uncertainty associated with ahypothesis (see Rudner 1953; Douglas 2009a). Hypotheses are notdeductively entailed by the evidence. Instead, one must determine theappropriate trade-off between false positive and false negatives forany given test. A test with a high bar for hypothesis acceptance willresult in more false negatives, while a test with a low bar forhypothesis acceptance will result in more false positives. Determiningthe right balance between errors often requires taking non-epistemicfactors into account. For example, if a false negative hasconsequences we wish to avoid as a society (e.g., death due to lack oftreatment), while a false positive for the same test does not (e.g.,the treatment is harmless and has negligible economic costs), then weshould err on the side of the false positive. Of course, often thecalculus is not this simple and numerous social and economic factorsmust be considered.

The results of comparative cognition research are used to inform lawsand welfare policies concerning animals. For example, in December2013, a group of lawyers, scientists, and policy experts filed apetition for a writ of habeas corpus in a New York State Supreme Court(Grimm 2013). A writ of habeas corpus is a court order to“produce the body”. It requires any person or institutionholding a prisoner to bring the captive to court and justify herimprisonment and treatment. The writ was filed on behalf of Tommy, amale chimpanzee. At the time of the filing, Tommy was seen livingalone in a dark shed. If recognized by the court, the writ wouldrequire Tommy’s holder to justify Tommy’s captivity andtreatment. This was the first time a habeas corpus had been filed onbehalf of a nonhuman animal and, if successful, would represent thefirst case of an animal being given the right not to be treated asproperty in the United States. The case on behalf of Tommy was made inpart by drawing on cognitive evidence. The plaintiffs argued that thecognitive abilities of chimpanzees are such that solitary confinementcauses harm. Numerous comparative cognition researchers andphilosophers have testified in support of this case (SeeNonhuman Rights Project: Client, Tommy (Chimpanzee); Andrews, Comstock, et al. 2018). Thus, how we treat and think weought to treat animals often depends on our knowledge of theircognitive abilities (Bekoff & Gruen 1993). Such knowledge can beused to prevent negative states of mind, such as loneliness, anxiety,and distress, as well as restore and promote positive mentalstates.

Should the methods of comparative cognition take non-epistemic valuesinto account? Jonathan Birch (2018) argues that when there are“clear policy applications in view” comparative cognitionresearchers should adjust their standards to reflect the moralconsequences of error (2018: 1028). Birch advances a criterion to helpdetermine when an animal welfare scientist (X) should acceptthe hypothesis that some species (S) has a mental state(M), given a particular policy context (P) (Birch2018). Under this view, a welfare scientist should accept that somespecies has a mental state in a particular policy contextPif and only if the expected sum of the possible welfare outcomes giventhe scientist’s background knowledge and the decision to affirmthe hypothesis thatS hasM inP isgreater than the expected sum of the possible welfareoutcomes given the scientist’s background knowledge and thedecision not to affirm the hypothesis thatS hasMinP. One concern with precautionary approaches such as thisone is that it is challenging to agree on where to set the burden ofproof in any given case. For example, Birch (2017) proposes to set theevidential bar for animal sentience at “at least one credibleindicator of sentience in at least one species of that order”(2017: 5). Other researchers, however, have objected that thisevidential bar is either too weak or too strong. For example, MichaelWoodruff (2017) argues that this evidential bar is too weak and shouldbe raised to include many more independent indicators of sentience todecrease uncertainty regarding the principle’s level ofscientific support. In contrast, Rachael Brown (2017b) argues thatBirch’s evidential bar is too strong for those situations inwhich there is no statistically significant evidence of a singlecredible indicator of sentience, but instead “multiple, weak,but convergent, lines of evidence that a species is sentient”(2017b: 2). Despite these differences, there is general agreement thatexpected welfare consequences should affect evidential standards inthose areas of comparative cognition that have clear policyimplications, and that the right evidential standard must bedetermined on a case-by-case basis (see also Benz-Schwarzburg,Monsó, & Huber. 2020; Crump et al. 2022).

3. Methodological Challenges

3.1 Anthropomorphism

A perennial concern in animal cognition research is whetherresearchers are being “anthropomorphic”. Shettleworthdefines anthropomorphism as

the attribution of human qualities to other animals, usually with theimplication it is done without sound justification. (Shettleworth2010b: 477)

The term “human qualities” refers to those properties thatwe readily accept as characteristic of humans. These may includecognitive abilities such as future planning, empathizing, insightfulproblem solving, and reliving past experiences. In comparativecognition, researchers may attribute such states to animals based onthe available empirical evidence. In our everyday lives, we also oftenattribute human-like states to animals (Serpell 2005). In both cases,there’s a question whether one’s attributions are corrector not.

Concerns surrounding anthropomorphism have dramatically influenced themethods and conclusions drawn from animal studies throughout thetwentieth and twenty-first centuries. Behaviorism within animalcognition research, for example, can be understood in part as aresponse to concerns about anthropomorphism (Wynne 2007). For thoseworried that we are misattributing human-like cognitive states tononhuman animals, one solution is to focus on describing observedrelationships between environmental cues and behavior instead.Although traditional behaviorism is widely rejected today, there arecontemporary scholars who hold that behaviorism had something right inits unwillingness to anthropomorphize. As Clive Wynne (2004)warns:

Old-time behaviourism may have imposed excessive constraints on animalpsychology. But the reintroduction of anthropomorphism risks bringingback the dirty bathwater as we rescue the baby. (2004: 606)

Should we be concerned about anthropomorphism in comparativecognition? One reason to be concerned is that there are empiricalstudies showing that humans have the tendency to over-attribute mentalstates to objects in their environment. For example, in one classicstudy, the psychologists Fritz Heider and Marianne Simmel (1944)showed human participants a video of three shapes moving in variousdirections and speeds. Despite the objects being two-dimensionalshapes, almost all participants described the scene in anthropomorphicor human-like terms (e.g., as the shapes “fighting” or“chasing” one another). Contemporary studies confirm thathumans are quick to attribute mental states to objects and agentsbased on behavior and other cues (like the presence of eyes) (Fiala,Arico, & Nichols 2011, 2014; Arico et al. 2011). As Dacey (2017)argues, the human tendency to anthropomorphize may serve as a fast andfrugal heuristic allowing one to quickly anticipate the behavior ofother agents. However, if this heuristic activates even when nopsychological agent is present, then this may lead to numerous falsepositive in animal cognition research. Dacey argues that the bestmethodological approach is not a general prohibition against theattribution of human-like mental states to animals, however, butrather the application of methods known to effectively combat implicitbias, like making counter-stereotypical information salient. Forexample, selecting targets that are relatively unlikely to beanthropomorphized (such as insects) and asking researchers to imaginesuch targets behaving intelligently may help counter intuitiveanthropomorphism (Dacey 2017: 1158–1161).

A second argument in favor of avoiding the attribution of human-likemental states to animals highlights that there are competingexplanations that can account for the behaviors inquestion—explanations that do not appeal to sophisticatedcognitive abilities. For example, Shettleworth (2010b) discusses thecases of animal insight, theory of mind, and mental time travel.Research suggests we find these abilities in animals such aschimpanzees and crows. However, Shettleworth argues that we shouldresist this conclusion because there are alternative,non-anthropomorphic, explanations for the behaviors observed in theseanimals. For example, we can explain the apparent insightfulproblem-solving behavior found in crows as part of their naturalbehavioral repertoire or as being driven by cues in the environment.And we can explain the apparent theory of mind abilities inchimpanzees as instead arising from a set of learned and innate rulesabout what to expect in social situations (seesection 2). Crucially, in addition to arguing that sophisticated mental abilitiesmight not provide the best explanations for animal behavior,Shettleworth argues that such abilities might also not provide thebest explanation for human behavior. She writes that empiricalresearch

increasingly reveals an unexpected role in human behavior for simple,unconscious and sometimes irrational processes shared by otheranimals. Greater appreciation of such mechanisms in nonhuman specieswould contribute to a deeper, more truly comparative psychology.(Shettleworth 2010b: 477)

Buckner (2013) additionally argues that humans have an inflated senseof their own cognitive aptitudes; thus, using this sense as a guidefor evaluating other animals “loads the deck against animalmentality” (2013: 853). He calls this tendency“anthropofabulation” which combines anthropocentrism (nextsection) with an exaggerated sense of human intelligence.

Some researchers argue that anthropomorphism has scientific benefits.For example, “heuristic anthropomorphism” holds that wecan use our intuitive understanding of humans and animals to generateconcrete hypotheses about animal behavior and then test thosehypotheses empirically (de Waal 1999). In other words, we can use ourexperience of being human—that is, of being a humananimal—to help us generate hypotheses about what it might belike to be another animal, such as a crow or dog. Crucially, however,the idea is not to simply impose a human perspective on other animals,but to take what is known about an animal’s behavior,evolutionary history, ecological context, etc. into account. The ideais to do this critically; that is, in a conjectural or provisionalway, which leaves the proposed hypothesis open to testing, revision orrejection. As de Waal (1999) writes:

While we should be reluctant to postulate capacities for which thereis no evidence anywhere in a species’ behavior, charges ofanthropomorphism are meaningless without a precise critique of thehypotheses under consideration. In a Darwinian framework, there is nogood reason to avoid concepts merely because they derive from thebehavior of the species to which we belong. Application of theseconcepts to animals not only enriches the range of hypotheses to beconsidered, but it also changes the view of ourselves: the morehuman-like we permit animals to become the more animal-like we becomein the process. (1999: 272).

Andrews (2016, 2020) also argues for the importance of human folkpsychology in comparative cognition research. Folk psychology can bebroadly understood as our human commonsense understanding ofpsychological phenomena (see entry onfolk psychology as a theory). Folk psychology is a form of anthropomorphism insofar as it involvesapplying human qualities to other animals. Andrews argues that folkpsychology is methodologically important for grouping animal behaviorstogether into types. For example, in a now classic study, Whiten andByrne (1988) collected reports of tactical deception in nonhumanprimates. “Deception” is a folk psychological term used toindicate the human act of deceiving or tricking another individual.However, despite its folk-psychological origin, Andrews argues thatthis term provided a useful starting point for individuating andcategorizing behaviors. Such categories were then analyzed, updated,and refined based on empirical evidence. Identifying robust behavioraltypes in this way is in turn important for investigating the cognitivemechanisms responsible for behavior (Andrews 2016a).

3.2 Anthropocentrism

The term “anthropocentrism” describes the tendency tolocate human beings at the “center”. Anthropocentrism canbroadly be understood as the claim that humans are special orexceptional in some way. Some researchers hold that concerns aboutanthropomorphism (see previous section) arise from a place ofanthropocentrism: “Cries of anthropomorphism are heardparticularly when a ray of light hits species other than ourown” (de Waal 1999: 256). In other words, the belief that humansare special gives rise to a bias against attributing human qualitiesto other animals. Brian Keeley (2004) argues that historically,particularly in the theological context, some traits have been takento belong exclusively or categorically to one group, such as humans orgods. For example, one might hold that only humans have souls. In thiscase, to attribute a soul to a nonhuman animal is to make a categorymistake: nonhuman animals are simply not the types of things that canhave souls. Such an attribution will always be false. Keeley contraststhis “categorical anthropomorphism” with“situational anthropomorphism” (2004: 529, see Fisher1996). While categorical anthropomorphism involves mistakenlyattributing a quality to something that simply cannot possess thatquality, situational anthropomorphism involves mistakenly attributinga quality to an agent or system, but this quality is something thatthe agent or system could possess in principle. If cognitivecapacities like insight and episodic memory are products of evolutionand development, then there is no reason in principle why otheranimals should not have them. If on the other hand, such capacitiesare products of cultural inheritance involving language, we should notexpect to find them in nonlinguistic species. In either case, it is anempirical question, in the same way determining whether an animal isomnivorous or land-dwelling is an empirical question. In the same waythat researchers should be concerned about anthropomorphism and falsepositive attributions of mental states to animals, one should beconcerned about anthropocentrism and false negatives—the failureto attribute human-like qualities to animal when the animal in factpossesses them (Sober 2005, 2012). Such false negatives have beentermed “anthropodenial” and “anthropectomy” inthe literature (de Waal 1999, Andrews & Huss 2014).

3.3 Underdetermination

A major challenge in comparative cognition is that claims aboutcognition are often underdetermined by behavioral evidence. AsTomasello and Call note, “the exact same behavior may beunderlain by very different cognitive mechanisms” (2008: 451).Broadly, a hypothesis is underdetermined when the available evidencefails to indicate what we should believe about that hypothesis (seeentry onunderdetermination of scientific theory). For example, if two hypotheses are equally supported by the availabledata, we might not be able to choose between them (contrastiveunderdetermination). Or if a hypothesis is found to be incompatiblewith an empirical result, we might not know whether to reject thathypothesis or some other background assumption instead (holistunderdetermination). Problems of underdetermination are foundthroughout the sciences. However, this problem is particularly salientin the cognitive sciences, given the opaque and complex nature ofcognitive systems.

We encountered an example of contrastive underdetermination whendiscussing cognitive and associative accounts of behavior (section 2.2). There we saw that some competing hypotheses appear to be equally wellsupported by the empirical data: for example, causal reasoning andassociative learning accounts of problem-solving behaviors in corvids(see Taylor, Medina, et al. 2010). If it is true that,“associative hypotheses can be constructedpost-hoc forevery experimental outcome” (Starzak & Gray 2021: 4), thensimply finding that one’s causal-reasoning hypothesis fits aparticular experimental outcome will not be sufficient to accept itover the available associative hypotheses. Instead, one must appeal toother factors, like epistemic values (section 2.3) to determine hypothesis choice.

Another major source of underdetermination in comparative cognition isthat the structure and function of target cognitive phenomena areoften uncertain and open to revision. For example, since the 1970s, asignificant amount of research has been dedicated to determiningwhether nonhuman great apes like chimpanzees have “theory ofmind” (ToM) or the ability to attribute mental states to otheragents. In their classic paper initiating this research program, DavidPremack and Guy Woodruff write:

we speculate about the possibility that the chimpanzee may have a“theory of mind”, one not markedly different from our own.(Premack & Woodruff 1978: 515)

However, they add that a chimpanzee’s ToM may differ from humanToM in important respects—e.g., in the type of mental statesinferred. Moreover, throughout work on animal ToM, researchers’understanding of the target phenomenon and relevant backgroundassumptions have changed in response to new findings. For example,rather than rejecting the hypothesis that chimpanzees attributeperceptual states to other agents, researchers have interpreted somenegative results as due to other factors (e.g., lack of ecologicalvalidity or poor experimental design) (Hare et al. 2000; Kaminski etal. 2004; Bräuer et al. 2007). This has then led to proposalsregarding what additional experimental controls are needed tosuccessfully detect a relationship between the independent anddependent variables (seesection 2.1). Philosophers have noted that such revised understandings of thetarget phenomenon play an important role in the biological andcognitive sciences (Bechtel 2008; Bechtel & Richardson 1993[2010]). Nevertheless, revising the phenomenon in this way leads tounderdetermination: in the face of conflicting data, it is unclearwhether researchers should reject the target hypothesis or revise it(an instance of holistic underdetermination).

3.4 Signatures and Dimensions

One way to overcome problems of underdetermination is to provideadditional constraints on hypothesis construction and evaluation.There have been several recent proposals in the literature on how todo this. For example, Alex Taylor and colleagues argue thatcomparative cognition researchers are currently too focused on whetheranimals succeed at a particular experimental task (Taylor, Bastos, etal. 2022). In their view, the problem with this approach is that itfails to adequately constrain the hypothesis space—there aresimply too many plausible hypotheses that could account for suchsuccess. Given this, researchers should instead seek to identify

the full range of information processing patterns including errors,limits, and biases (whether neutral, adaptive, or maladaptive) shownby an agent, so as to constrain the cognitive hypothesis spaceeffectively. (Taylor, Bastos, et al. 2022: 3)

Taylor and colleagues refer to this as “signature testing”(in contrast to “success testing”). A“signature” is any pattern of evidence that constrains thehypothesis space with respect to a phenomenon of interest (it can beweakly or strongly diagnostic depending on how much it constrains thehypothesis space). A successful hypothesis should be able to accountfor all signatures of a cognitive process, not just an animal’ssuccessful performance on an experimental task.

Starzak and Gray (2021) similarly urge researchers to develop morefine-grained accounts of how to conceptualize cognitive phenomena. Forexample, with respect to the cognitive phenomenon of causalunderstanding, they write:

in thinking about the nature of causal understanding we should thinkabout the extent to which organisms can differ with respect to thekind of information they can pick up; with respect to the differentsources of causal information they can exploit; with respect to theway they can process this information and integrate different types ofinformation or information stemming from different sources; and withrespect to the flexibility with which they can use this information toguide behavior. (2021: 9)

Like signature testing, this multidimensional approach is more nuancedthan asking whether an animal “has causal understanding”tout court. It also helps address the issue that associativelearning and complex cognition are not necessarily mutually exclusive (section 2.2). Rather than asking whether an experimental result is best explainedby appealing to associative learning or causal understanding, one canappeal to both accounts: associative learning might explain how causalinformation is acquired in some cases and cognitive models mightexplain how causal information is integrated. Researchers haveadvocated for a similar multidimensional approach in the context ofwork on behavioral innovation (Brown 2022) and animal consciousness(Birch, Schnell, & Clayton 2020, see section on consciousness inentry onanimal cognition). The upshot is that a more fine-grained approach to cognition andbehavior may help constrain the hypothesis space in such a way thatavoids major problems of underdetermination.

3.5 Reproducibility

In psychology, reproducibility and replication typically refer toredoing an experiment to assess its reliability. If a study isreliable, then running the study again should produce the sameresults. If a replication fails, then this may mean that the originalresult was a product of measurement error, sampling error, imprecisemanipulation, questionable research practices, or other factors (foran overview, see Romero (2019) and the entry onreproducibility of scientific results). Efforts to replicate studies have increased over the past two decadesacross a wide range of fields, including medicine, computer science,and psychology (Bohannon 2014). This has led to what some describe asa “replication crisis” because a surprising number ofstudies have failed to replicate. For example, an effort to replicate100 studies by 270 psychologists as part of the Open ScienceCollaboration found that only 38% of the original results werereproduced unambiguously, with some attempted replications finding aneffect opposite to that of the original study (Bohannon 2015).Findings such as these have led researchers to ask whether comparativecognition also suffers from a replication crisis, and if so, what canbe done to improve the field (Brecht et al. 2021). In a recent survey,for example, comparative cognition researchers were asked their viewson replications (Farrar, Ostojić, & Clayton 2021). Out of 210respondents, the majority agreed (34.8%) or strongly agreed (54.8%)that replications are important to perform in animal cognitionresearch. The majority also disagreed (55.5%) or strongly disagreed(23.6%) that enough replications were already performed in thefield.

Despite the above survey results, comparative cognition has beendescribed as “an absolute beacon for replication efforts”(Beran 2018: 2). Beran (2018) notes that a standard in the field hasbeen to conduct a series of experiments with the first experimentconsisting of a replication of the study that inspired the work andsubsequent experiments dedicated to extending that experiment (2018:2). Halina (2021a) also argues that replications in comparativecognition are common. She adopts Edouard Machery’s resamplingaccount of replication (Machery 2020), showing how under this viewsuccessful replications occur frequently in areas such as chimpanzeetheory of mind research. If replication success were a clear indicatorof reliability, then this would be good news for comparativecognition. The picture is however complicated by several factors.First, failed replications are difficult to interpret. Often areplication study will differ from the original study in severalrespects: in such cases, it is possible to attribute failure to thesedifferences (what Colaço, Bickle, and Walters (2022) refer toas “mismatch explanations”). In chimpanzee theory of mindresearch, for example, failed replications have regularly beenexplained by appealing to changes in the experimental setting (Halina2021a). As we saw insection 3.3, revisions to background assumptions are an important part of science.However, such revisions mean that failed replications are often notinterpreted as undermining reliability (Nosek, Spies, & Motyl2012). As Alexandria Boyle (2021) argues, for this and other reasons,replications in comparative cognition are

poorly placed to deliver clear judgments about the reliability ofcomparative cognition’s methods or its scientific bona fides.(2021: 296, see also Feest 2019)

Another factor complicating the picture is the “file-drawerproblem”, which refers to the phenomenon that studies that failto find statistically significant results may be relegated to the filedrawer (i.e., not published or communicated to the larger scientificcommunity). In the survey cited above, comparative cognitionresearchers were also asked “What percent of the studies thatyou have performed have been published and/or you think will bepublished?” (Farrar, Ostojić, & Clayton 2021). Themedian response was 80% with a large spread (out of 210 responses, 23said that they published 50% or fewer of their studies, while 17reported publishing all their studies). Twenty-nine respondents citednegative or uninteresting results as the reason for not publishing(2021: 18). The file-drawer problem combined with mismatchexplanations of failed replications may create an impression ofreliability via the reporting of many successful replications when infact the record is mixed.

If replications on their own are not a route to more robust research,what else could help? Beran (2018) emphasizes the importance ofpre-registration. Pre-registration involves publishing the methods andstatistical analyses that one will use in a study before datacollection. This prevents one from adjusting or“massaging” elements of the experimental design and dataanalyses with the aim of getting a positive result. Farrar, Voudouris,and Clayton (2021) also advance several methods that comparativecognition researchers could use to assess the reliability of smallsample research, such as statistically modeling variation. Morebroadly, Felipe Romero (2020) argues that major changes to scientificincentive structures are needed: in particular, a shift away fromrewarding novelty and towards rewarding replication and confirmationwork. Along these lines, Brecht et al. (2021) note that while thereare still many disincentives to conducting replication studies incomparative cognition, the field is working towards improving this.For example, the journalAnimal Behavior and Cognition hascommitted to publishing pre-registered studies (includingreplications) regardless of the results. In addition, several newglobal consortiums (such as ManyPrimates, ManyBirds, and ManyDogs)have formed with the explicit aim of assessing reliability,encouraging transparency, supporting large collaborations, andfostering other open science practices (seeOther Internet Resources).

4. Comparative Cognition and AI

There has been a large amount of collaborative work betweencomparative cognition and artificial intelligence (AI) research overthe past decade. On the one hand, formal models developed in AI can beused to predict and explain animal behavior in ways that move beyondfolk psychological accounts. On the other hand, the methods ofcomparative cognition are well placed to inform AI research, given thediversity of behavior and information-processing abilities foundacross the tree of life. We’ll consider both aspects here, whilealso highlighting why animal-AI inferences should be handled withcare.

Colin Allen (2014) argues that comparative cognition would benefitgreatly from developing mathematically rigorous formal models. As wehave seen throughout this entry, comparative cognition researchersoften rely on intuitive, natural language concepts for constructingaccounts of animal minds like causal reasoning, mental time travel,imagination, and self-recognition (see Schnell et al. 2021b).Regarding this approach Allen (2014) writes:

The conceptual framework guiding most work in comparative animalcognition (whether by ethologists or psychologists) is insufficientlyformalized to support rigorous science in the long run. (2014: 82)

One concern with more formal approaches, however, is that they willfail to “scale up” to predict and explain the behavioralphenomenon that are of interest to many comparative cognitionresearchers, such as natural social and physical interactions. Work inAI is starting to show that there are ways of closing this gap,however. For example, Peter Battaglia and colleagues advance aformalized mental model in the form of an “intuitive physicsengine” (IPE) analogous to the machine physics engines used ininteractive video games (Battaglia, Hamrick, & Tenenbaum. 2013,see also Ullman et al. 2017). The IPE is probabilistic andoversimplifies the nature of objects, such as their geometry and massdensity distribution, but has been developed to explain how peoplequickly make inferences about physical scenes in a dynamic and noisyworld. Battaglia and colleagues found that this formal model doesindeed capture people’s intuitions and prediction about physicalscenes (such as “will this tower of blocks fall?”) acrossa wide range of complex and novel scenarios. Moreover, the modelcaptures not just successful predictions, but also other signatures ofhuman judgment like illusions and biases (seesection 3.4).

The above work uses AI to explain human cognition and behavior.However, such methods are also being applied to nonhuman animals. Forexample, much of contemporary AI depends on the tools of reinforcementlearning (RL). These techniques involve linking states of theenvironment and an agent’s actions in such a way that allows theagent to maximize future rewards. The tools of RL were originallyinspired by animal learning research (Hassabis et al. 2017); however,they have since been developed by AI researchers in ways that allowfresh insights into animal behavior. For example, thetemporal-difference (TD) model has been used to explain a wide rangeof results from animal classical conditioning studies (see Sutton& Barto 1998: chapter 14). Similarly, Buckner (2018) draws on workon Deep Convolutional Neural Networks (DCNNs) to explain therelationship between sensory experience and representations ofabstract categories in humans and animals—a longstandingquestion in philosophy of mind. Halina (2021b) also shows how one canunderstand aspects of animal insightful problem solving (namely,mental scenario building) through Monte Carlo tree search. Finally,Bohn et al. (2022) advance a Bayesian computational model of great apecommunication, which they show accurately predicts the communicativeinteractions of real-world chimpanzees living semi-wild in theChimfunshi Wildlife Orphanage in Zambia. These are a few examples ofresearchers drawing on AI for formal models that can be successfullyapplied to capture animal minds and behavior (see also van der Vaartet al. 2012).

A major aim of AI research has been to build machines that“learn and think like people” (Lake et al. 2017).Currently, there are many AI systems that match or exceed humans onvarious tasks, like the ability to play Chess and Go, and even makescientific discoveries (Shevlin et al. 2019). However, how to build amachine with domain-general intelligence or “common sense”remains elusive. This challenge has led some researchers to argue thatanimal cognition research offers the best path towards buildingthinking machines. Animals exhibit many of the “buildingblocks” of human common sense, such as an understanding ofobjects and their affordances, space, and causality (Shanahan et al.2020). Through the application of RL techniques, it is also possibleto train artificial agents in 3D virtual environments analogous to thereal world. Using this approach, one can combine different RLarchitectures and training environments with the aim of encouragingthe development of domain-general abilities. As Murray Shanahan andcolleagues write,

animal cognition supplies a compendium of well understood,nonlinguistic, intelligent behaviour; it suggests experimental methodsfor evaluation and benchmarking; and it can guide environment and taskdesign. (2020: 863; see also Crosby 2020)

Work such as this is already underway. For example, the Animal-AITestbed applies experimental protocols developed in comparativecognition to test AI (Crosby, Beyret, & Halina 2019; Crosby,Beyret et al. 2020). In 2019, the testbed included 300 tasks groupedinto 12 categories such as spatial elimination, delayed gratification,numerosity, and tool use tasks. Konstantinos Voudouris and colleagues(2022) compared AI performance on this testbed with children aged6–10. They found that children and AIs performed similarly onbasic navigational tasks, but that children outperformed AIs on morecomplex tasks like object permanence and detour tasks.

There is much to be gained from animal-AI comparisons. Such work mayrepresent the beginning of a transformation in the field ofcomparative cognition—one that brings artificial systems intothe fold. It may also lead to the development of cognitive models thatbridge the gap between intuitive accounts of capacities likeimagination and causal reasoning on the one hand, and mathematical andcomputational models on the other. However, when applying “richpsychological concepts” like awareness, episodic memory, andtheory of mind to AI, one should also proceed with caution (Shevlin& Halina 2019). Some such concepts have normative dimensions, forexample, indicating a potential moral agent or moral patient. Thesenormative dimensions are important to acknowledge before adopting aterm in the context of AI, particularly if that term is useddifferently in the two contexts. AI findings may also dramaticallyalter our understanding of some cognitive capacities. Rather thanadopting existing cognitive concepts, it may be fruitful in some casesto develop a radically new approach in the context of AI. Ifsuccessful, such an approach could then be imported into comparativecognition and potentially revolutionize our understanding of animalminds.

Bibliography

• Allen, Colin, 2006, “Transitive Inference in Animals:Reasoning or Conditioned Associations?”, in Hurley and Nudds2006: 175–186.doi:10.1093/acprof:oso/9780198528272.003.0007
• –––, 2014, “Models, Mechanisms, and AnimalMinds”,The Southern Journal of Philosophy, 52(S1):75–97. doi:10.1111/sjp.12072
• –––, 2017, “On (Not) DefiningCognition”,Synthese, 194(11): 4233–4249.doi:10.1007/s11229-017-1454-4
• Allen, Colin and Marc Bekoff, 1997,Species of Mind: ThePhilosophy and Biology of Cognitive Ethology, Cambridge, MA: MITPress. doi:10.7551/mitpress/6395.001.0001
• Aplin, Lucy M., Ben C. Sheldon, and Julie Morand-Ferron, 2013,“Milk Bottles Revisited: Social Learning and IndividualVariation in the Blue Tit,Cyanistes caeruleus”,Animal Behaviour, 85(6): 1225–1232.doi:10.1016/j.anbehav.2013.03.009
• Amodio, Piero, Markus Boeckle, Alexandra K. Schnell, LjerkaOstojíc, Graziano Fiorito, and Nicola S. Clayton, 2019,“Grow Smart and Die Young: Why Did Cephalopods EvolveIntelligence?”,Trends in Ecology & Evolution,34(1): 45–56. doi:10.1016/j.tree.2018.10.010
• Andrews, Kristin, 2015 [2020],The Animal Mind: AnIntroduction to the Philosophy of Animal Cognition, London:Routledge. Second edition, New York: Routledge, 2020.doi:10.4324/9780203712511.
• –––, 2016, “Pluralistic Folk Psychology inHumans and Other Apes”, inThe Routledge Handbook ofPhilosophy of the Social Mind, Julian Kiverstein (ed.), London:Routledge, 133–154 (chapter 7).
• –––, 2020,How to Study Animal Minds,Cambridge/New York: Cambridge University Press.doi:10.1017/9781108616522
• Andrews, Kristin and Jacob Beck (eds.), 2017,The RoutledgeHandbook of Philosophy of Animal Minds, New York: Routledge.doi:10.4324/9781315742250
• Andrews, Kristin, Gary Comstock, G.K.D. Crozier, Sue Donaldson,Andrew Fenton, Tyler M. John, L. Syd M. Johnson, Robert C. Jones, WillKymlicka, Letitia Meynell, Nathan Nobis, DavidPeña-Guzmán, and Jeff Sebo, 2018,Chimpanzee Rights:The Philosophers’ Brief, New York: Routledge.doi:10.4324/9780429461071
• Andrews, Kristin and Brian Huss, 2014, “Anthropomorphism,Anthropectomy, and the Null Hypothesis”,Biology &Philosophy, 29(5): 711–729.doi:10.1007/s10539-014-9442-2
• Arico, Adam, Brian Fiala, Robert F. Goldberg, and Shaun Nichols,2011, “The Folk Psychology of Consciousness”,Mind& Language, 26(3): 327–352.doi:10.1111/j.1468-0017.2011.01420.x
• Avital, Eytan and Eva Jablonka, 2000,Animal Traditions:Behavioural Inheritance in Evolution, Cambridge/New York:Cambridge University Press. doi:10.1017/CBO9780511542251
• Bandini, Elisa and Rachel A. Harrison, 2020, “Innovation inChimpanzees”,Biological Reviews, 95(5):1167–1197. doi:10.1111/brv.12604
• Bastos, Amalia P. M. and Alex H. Taylor, 2020, “Kea ShowThree Signatures of Domain-General Statistical Inference”,Nature Communications, 11: article 828.doi:10.1038/s41467-020-14695-1
• Battaglia, Peter W., Jessica B. Hamrick, and Joshua B. Tenenbaum,2013, “Simulation as an Engine of Physical SceneUnderstanding”,Proceedings of the National Academy ofSciences, 110(45): 18327–18332.doi:10.1073/pnas.1306572110
• Bausman, William C., 2018, “Modeling: Neutral, Null, andBaseline”,Philosophy of Science, 85(4): 594–616.doi:10.1086/699021
• Bausman, William and Marta Halina, 2018, “Not Null Enough:Pseudo-Null Hypotheses in Community Ecology and ComparativePsychology”,Biology & Philosophy, 33(3–4):article 30. doi:10.1007/s10539-018-9640-4
• Bechtel, William, 2008,Mental Mechanisms: PhilosophicalPerspectives on Cognitive Neuroscience, New York: Routledge.doi:10.4324/9780203810095
• Bechtel, William and Robert C. Richardson, 1993 [2010],Discovering Complexity: Decomposition and Localization asStrategies in Scientific Research, Princeton, NJ: PrincetonUniversity Press. New edition, Cambridge, MA: The MIT Press,2010.
• Bekoff, Marc and Lofe Gruen, 1993, “Animal Welfare andIndividual Characteristics: A Conversation Against Speciesism”,Ethics & Behavior, 3(2): 163–175.doi:10.1207/s15327019eb0302_2
• Benz-Schwarzburg, Judith, Susana Monsó, and Ludwig Huber,2020, “How Dogs Perceive Humans and How Humans Should TreatTheir Pet Dogs: Linking Cognition With Ethics”,Frontiers inPsychology, 11: 584037. doi:10.3389/fpsyg.2020.584037
• Beran, Michael J., 2018, “Replication and Pre-Registrationin Comparative Psychology”,International Journal ofComparative Psychology, 31. doi:10.46867/ijcp.2018.31.01.09
• Birch, Jonathan, 2017, “Animal Sentience and thePrecautionary Principle”,Animal Sentience, 2(16):article 1. doi:10.51291/2377-7478.1200
• –––, 2018, “Animal Cognition and HumanValues”,Philosophy of Science, 85(5): 1026–1037.doi:10.1086/699744
• Birch, Jonathan, Alexandra K. Schnell, and Nicola S. Clayton,2020, “Dimensions of Animal Consciousness”,Trends inCognitive Sciences, 24(10): 789–801.doi:10.1016/j.tics.2020.07.007
• Boesch, Christophe, 2022, “What Does a Notion of Weight Meanto Chimpanzees?”,Animal Behavior and Cognition, 9(2):222–238. doi:10.26451/abc.09.02.07.2022
• Bohannon, John, 2014, “Replication Effort ProvokesPraise—and ‘Bullying’ Charges”,Science, 344(6186): 788–789.doi:10.1126/science.344.6186.788
• –––, 2015, “Many Psychology Papers FailReplication Test”,Science, 349(6251): 910–911.doi:10.1126/science.349.6251.910
• Bohn, Manuel, Katja Liebal, Linda Oña, and Michael HenryTessler, 2022, “Great Ape Communication as Contextual SocialInference: A Computational Modelling Perspective”,Philosophical Transactions of the Royal Society B: BiologicalSciences, 377(1859): 20210096. doi:10.1098/rstb.2021.0096
• Boyle, Alexandria, 2021, “Replication, Uncertainty andProgress in Comparative Cognition”,Animal Behavior andCognition, 8(2): 296–304.doi:10.26451/abc.08.02.15.2021
• Brakes, Philippa, Sasha R. X. Dall, Lucy M. Aplin, Stuart Bearhop,Emma L. Carroll, Paolo Ciucci, Vicki Fishlock, John K. B. Ford, EllenC. Garland, Sally A. Keith, Peter K. McGregor, Sarah L. Mesnick,Michael J. Noad, Giuseppe Notarbartolo Di Sciara, Martha M. Robbins,Mark P. Simmonds, Fernando Spina, Alex Thornton, Paul R. Wade, MartinJ. Whiting, et al., 2019, “Animal Cultures Matter forConservation”,Science, 363(6431): 1032–1034.doi:10.1126/science.aaw3557
• Bräuer, Juliane, Josep Call, and Michael Tomasello, 2007,“Chimpanzees Really Know What Others Can See in a CompetitiveSituation”,Animal Cognition, 10(4): 439–448.doi:10.1007/s10071-007-0088-1
• Brecht, Katharina F., Edward W. Legg, Christian Nawroth, HannahFraser, and Ljerka Ostojić, 2021, “The Status and Value ofReplications in Animal Behavior Science”,Animal Behaviorand Cognition, 8(2): 97–106.doi:10.26451/abc.08.02.01.2021
• Brigandt, Ingo, 2007, “Typology Now: Homology andDevelopmental Constraints Explain Evolvability”,Biology& Philosophy, 22(5): 709–725.doi:10.1007/s10539-007-9089-3
• Brown, Rachael L., 2013, “Learning, Evolvability andExploratory Behaviour: Extending the Evolutionary Reach ofLearning”,Biology & Philosophy, 28(6):933–955. doi:10.1007/s10539-013-9396-9
• –––, 2017a, “Animal Traditions: What TheyAre, and Why They Matter”, in Andrews and Beck 2017:362–371 (chapter 34).
• –––, 2017b, “Not StatisticallySignificant, but Still Scientific”,Animal Sentience,2(16): article 14. doi:10.51291/2377-7478.1247
• –––, 2022, “Mapping Out the Landscape: AMulti-Dimensional Approach to Behavioral Innovation”,Philosophy of Science, 89(5): 1176–1185.doi:10.1017/psa.2022.42
• Buckner, Cameron, 2011, “Two Approaches to the Distinctionbetween Cognition and ‘Mere Association’”,International Journal of Comparative Psychology, 24(4):314–348. doi:10.46867/IJCP.2011.24.04.06
• –––, 2013, “Morgan’s Canon, MeetHume’s Dictum: Avoiding Anthropofabulation in Cross-SpeciesComparisons”,Biology & Philosophy, 28(5):853–871. doi:10.1007/s10539-013-9376-0
• –––, 2017, “Understanding Associative andCognitive Explanations in Comparative Psychology”, in Andrewsand Beck 2017: 409–418 (chapter 39).
• –––, 2018, “Empiricism without Magic:Transformational Abstraction in Deep Convolutional NeuralNetworks”,Synthese, 195(12): 5339–5372.doi:10.1007/s11229-018-01949-1
• –––, forthcoming,From Deep Learning toRational Machines: What the History of Philosophy Can Teach Us aboutthe Future of Artificial Intelligence, Oxford: Oxford UniversityPress.
• Byrne, Richard W. and Andrew Whiten (eds.), 1988,Machiavellian Intelligence: Social Expertise and the Evolution ofIntellect in Monkeys, Apes, and Humans, Oxford: ClarendonPress.
• Clark, Jason A., 2010, “Relations of Homology between HigherCognitive Emotions and Basic Emotions”,Biology &Philosophy, 25(1): 75–94.doi:10.1007/s10539-009-9170-1
• Clayton, Nicola S. and Anthony Dickinson, 1998,“Episodic-like Memory during Cache Recovery by ScrubJays”,Nature, 395(6699): 272–274.doi:10.1038/26216
• Clayton, N. S., D. P. Griffiths, N. J. Emery, and A. Dickinson,2001, “Elements of Episodic–like Memory in Animals”,Philosophical Transactions of the Royal Society of London. SeriesB: Biological Sciences, 356(1413): 1483–1491.doi:10.1098/rstb.2001.0947
• Colaço, David, John Bickle, and Bradley Walters, 2022,“When Should Researchers Cite Study Differences in Response to aFailure to Replicate?”,Biology & Philosophy,37(5): article 39. doi:10.1007/s10539-022-09873-y
• Cristol, Daniel A., Paul V. Switzer, Kara L. Johnson, and Leah S.Walke, 1997, “Crows Do Not Use Automobiles as Nutcrackers:Putting an Anecdote to the Test”,The Auk, 114(2):296–298. doi:10.2307/4089172
• Crosby, Matthew, 2020, “Building Thinking Machines bySolving Animal Cognition Tasks”,Minds and Machines,30(4): 589–615. doi:10.1007/s11023-020-09535-6
• Crosby, Matthew, Benjamin Beyret, and Marta Halina, 2019,“The Animal-AI Olympics”,Nature MachineIntelligence, 1(5): 257–257.doi:10.1038/s42256-019-0050-3
• Crosby, Matthew, Benjamin Beyret, Murray Shanahan, JoséHernández-Orallo, Lucy Cheke, and Marta Halina, 2020,“The Animal-AI Testbed and Competition”, inProceedings of the NeurIPS 2019 Competition and DemonstrationTrack, PMLR, 164–176. [Crosby et al. 2020 available online]
• Crump, Andrew, Heather Browning, Alex Schnell, Charlotte Burn, andJonathan Birch, 2022, “Sentience in Decapod Crustaceans: AGeneral Framework and Review of the Evidence”,AnimalSentience, 7(32): article 1. doi:10.51291/2377-7478.1691
• Currie, Adrian, 2021,Comparative Thinking in Biology,(Cambridge Elements. Elements in the Philosophy of Biology),Cambridge: Cambridge University Press. doi:10.1017/9781108616683
• Dacey, Mike, 2016, “The Varieties of Parsimony inPsychology”,Mind & Language, 31(4): 414–437.doi:10.1111/mila.12113
• –––, 2017, “Anthropomorphism as CognitiveBias”,Philosophy of Science, 84(5): 1152–1164.doi:10.1086/694039
• –––, 2023, “Evidence in Default: RejectingDefault Models of Animal Minds”,The British Journal for thePhilosophy of Science, 74(2): 291–312.doi:10.1086/714799
• De Waal, Frans B. M., 1999, “Anthropomorphism andAnthropodenial: Consistency in Our Thinking about Humans and OtherAnimals”,Philosophical Topics, 27(1): 255–280.doi:10.5840/philtopics199927122
• Dickinson, Anthony, 2012, “Associative Learning and AnimalCognition”,Philosophical Transactions of the Royal SocietyB: Biological Sciences, 367(1603): 2733–2742.doi:10.1098/rstb.2012.0220
• Douglas, Heather E., 2009a,Science, Policy, and theValue-Free Ideal, Pittsburgh, PA: University of PittsburghPress.
• –––, 2009b, “Reintroducing Prediction toExplanation”,Philosophy of Science, 76(4):444–463. doi:10.1086/648111
• –––, 2013, “The Value of CognitiveValues”,Philosophy of Science, 80(5): 796–806.doi:10.1086/673716
• Douglas, Heather and P.D. Magnus, 2013, “State of the Field:Why Novel Prediction Matters”,Studies in History andPhilosophy of Science Part A, 44(4): 580–589.doi:10.1016/j.shpsa.2013.04.001
• Ereshefsky, Marc, 2007, “Psychological Categories asHomologies: Lessons from Ethology”,Biology &Philosophy, 22(5): 659–674.doi:10.1007/s10539-007-9091-9
• –––, 2012, “Homology Thinking”,Biology & Philosophy, 27(3): 381–400.doi:10.1007/s10539-012-9313-7
• Farrar, Benjamin G., Ljerka Ostojić, and Nicola S. Clayton,2021, “The Hidden Side of Animal Cognition Research:Scientists’ Attitudes toward Bias, Replicability and ScientificPractice”,PLoS ONE, 16(8): e0256607.doi:10.1371/journal.pone.0256607
• Farrar, Benjamin G., Konstantinos Voudouris, and Nicola S.Clayton, 2021, “Replications, Comparisons, Sampling and theProblem of Representativeness in Animal Cognition Research”,Animal Behavior and Cognition, 8(2): 273–295.doi:10.26451/abc.08.02.14.2021
• Feest, Uljana, 2019, “Why Replication Is Overrated”,Philosophy of Science, 86(5): 895–905.doi:10.1086/705451
• Fiala, Brian, Adam Arico, and Shaun Nichols, 2011, “On thePsychological Origins of Dualism: Dual-Process Cognition and theExplanatory Gap”, inCreating Consilience: Integrating theSciences and the Humanities, Edward Slingerland and Mark Collard(eds.), New York: Oxford University Press, 88–110 (chapter 4).doi:10.1093/acprof:oso/9780199794393.003.0005
• –––, 2014, “You, Robot”, inCurrent Controversies in Experimental Philosophy, EdouardMachery and Elizabeth O’Neill (eds.), New York: Routledge,31–47.
• Fisher, John Andrew, 1996, “The Myth of AnthropomorphismJohn Andrew Fisher”, inReadings in Animal Cognition,Marc Bekoff and Dale Jamison (eds.), Cambridge, MA: MIT Press,3–16.
• Fitzpatrick, Simon, 2008, “Doing Away with Morgan’sCanon”,Mind & Language, 23(2): 224–246.doi:10.1111/j.1468-0017.2007.00338.x
• Fitzpatrick, Simon and Kristin Andrews, 2022, “AnimalCulture and Animal Welfare”,Philosophy of Science,89(5): 1104–1113. doi:10.1017/psa.2022.34
• Fitzpatrick, Simon and Grant Goodrich, 2017, “Building aScience of Animal Minds: Lloyd Morgan, Experimentation, andMorgan’s Canon”,Journal of the History ofBiology, 50(3): 525–569. doi:10.1007/s10739-016-9451-x
• Fletcher, Logan and Peter Carruthers, 2013,“Behavior-Reading versus Mentalizing in Animals”, inAgency and Joint Attention, Janet Metcalfe and Herbert S.Terrace (eds.), New York: Oxford University Press, 82–99.doi:10.1093/acprof:oso/9780199988341.003.0005
• Garcia-Pelegrin, Elias, Alexandra K. Schnell, Clive Wilkins, andNicola S. Clayton, 2021, “Exploring the Perceptual Inabilitiesof Eurasian Jays (Garrulus glandarius ) Using MagicEffects”,Proceedings of the National Academy ofSciences, 118(24): e2026106118. doi:10.1073/pnas.2026106118
• Gibbons, Matilda, Elisabetta Versace, Andrew Crump, Bartosz Baran,and Lars Chittka, 2022, “Motivational Trade-Offs and Modulationof Nociception in Bumblebees”,Proceedings of the NationalAcademy of Sciences, 119(31): e2205821119.doi:10.1073/pnas.2205821119
• Griffiths, Paul E., 1997,What Emotions Really Are: TheProblem of Psychological Categories, (Science and Its ConceptualFoundations), Chicago, IL: University of Chicago Press.
• –––, 2007a, “The Phenomena ofHomology”,Biology & Philosophy, 22(5):643–658. doi:10.1007/s10539-007-9090-x
• –––, 2007b, “Evo-Devo Meets theMind”, inIntegrating Evolution and Development: From Theoryto Practice, Roger Sansom and Robert N. Brandon (eds.),Cambridge, MA: The MIT Press, 195–225.doi:10.7551/mitpress/7465.003.0008
• Grimm, David, 2013, “Lawsuits Seek ‘Personhood’for Chimpanzees”,Science, 342(6163): 1154–1155.doi:10.1126/science.342.6163.1154
• Halina, Marta, 2021a, “Replications in ComparativePsychology”,Animal Behavior and Cognition, 8(2):263–272. doi:10.26451/abc.08.02.13.2021
• –––, 2021b, “Insightful ArtificialIntelligence”,Mind & Language, 36(2):315–329. doi:10.1111/mila.12321
• –––, 2022, “Unlimited Associative Learningas a Null Hypothesis”,Philosophy of Science, 89(5):1186–1195. doi:10.1017/psa.2022.66
• Hanus, Daniel, 2016, “Causal Reasoning versus AssociativeLearning: A Useful Dichotomy or a Strawman Battle in ComparativePsychology?”,Journal of Comparative Psychology,130(3): 241–248. doi:10.1037/a0040235
• Hare, Brian, Josep Call, Bryan Agnetta, and Michael Tomasello,2000, “Chimpanzees Know What Conspecifics Do and Do NotSee”,Animal Behaviour, 59(4): 771–785.doi:10.1006/anbe.1999.1377
• Hassabis, Demis, Dharshan Kumaran, Christopher Summerfield, andMatthew Botvinick, 2017, “Neuroscience-Inspired ArtificialIntelligence”,Neuron, 95(2): 245–258.doi:10.1016/j.neuron.2017.06.011
• Healy, Susan D. and T. Andrew Hurly, 2003, “CognitiveEcology: Foraging in Hummingbirds as a Model System”, inAdvances in the Study of Behavior, Volume 32, New York:Elsevier, 325–359. doi:10.1016/S0065-3454(03)01007-6
• Heider, Fritz and Marianne Simmel, 1944, “An ExperimentalStudy of Apparent Behavior”,The American Journal ofPsychology, 57(2): 243–259. doi:10.2307/1416950
• Herrmann, Esther, Josep Call, María VictoriaHernàndez-Lloreda, Brian Hare, and Michael Tomasello, 2007,“Humans Have Evolved Specialized Skills of Social Cognition: TheCultural Intelligence Hypothesis”,Science, 317(5843):1360–1366. doi:10.1126/science.1146282
• Heyes, Cecilia, 2012, “Simple Minds: A Qualified Defence ofAssociative Learning”,Philosophical Transactions of theRoyal Society B: Biological Sciences, 367(1603): 2695–2703.doi:10.1098/rstb.2012.0217
• Heyes, Cecilia and Anthony Dickinson, 1990, “TheIntentionality of Animal Action”,Mind & Language,5(1): 87–103. doi:10.1111/j.1468-0017.1990.tb00154.x
• Hurley, Susan and Matthew Nudds (eds.), 2006,RationalAnimals? Oxford: Oxford University Press.doi:10.1093/acprof:oso/9780198528272.001.0001
• Jablonka, Eva and Marion J. Lamb, 2008, “Soft Inheritance:Challenging the Modern Synthesis”,Genetics and MolecularBiology, 31(2): 389–395.doi:10.1590/S1415-47572008000300001
• Jozet-Alves, Christelle, Marion Bertin, and Nicola S. Clayton,2013, “Evidence of Episodic-like Memory in Cuttlefish”,Current Biology, 23(23): R1033–R1035.doi:10.1016/j.cub.2013.10.021
• Kaminski, Juliane, Josep Call, and Michael Tomasello, 2004,“Body Orientation and Face Orientation: Two Factors ControllingApes? Begging Behavior from Humans”,Animal Cognition,7(4): 216–223. doi:10.1007/s10071-004-0214-2
• Keeley, Brian L., 2004, “Anthropomorphism, Primatomorphism,Mammalomorphism: Understanding Cross-Species Comparisons”,Biology & Philosophy, 19(4): 521–540.doi:10.1007/sBIPH-004-0540-4
• Krupenye, Christopher and Josep Call, 2019, “Theory of Mindin Animals: Current and Future Directions”,WIREs CognitiveScience, 10(6): e1503. doi:10.1002/wcs.1503
• Lake, Brenden M., Tomer D. Ullman, Joshua B. Tenenbaum, and SamuelJ. Gershman, 2017, “Building Machines That Learn and Think likePeople”,Behavioral and Brain Sciences, 40: e253.doi:10.1017/S0140525X16001837
• Laland, Kevin N., Tobias Uller, Marcus W. Feldman, Kim Sterelny,Gerd B. Müller, Armin Moczek, Eva Jablonka, and John Odling-Smee,2015, “The Extended Evolutionary Synthesis: Its Structure,Assumptions and Predictions”,Proceedings of the RoyalSociety B: Biological Sciences, 282(1813): 20151019.doi:10.1098/rspb.2015.1019
• Longino, Helen E., 2008, “Values, Heuristics, and thePolitics of Knowledge”, inThe Challenge of the Social andthe Pressure of Practice: Science and Values Revisited, MartinCarrier, Don Howard, and Janet A. Kourany (eds.), Pittsburgh, PA:University of Pittsburgh Press, 68–86.
• Mayr, Ernst, 1982,The Growth of Biological Thought:Diversity, Evolution, and Inheritance, Cambridge, MA: BelknapPress.
• Machery, Edouard, 2020, “What Is a Replication?”,Philosophy of Science, 87(4): 545–567.doi:10.1086/709701
• Matthen, Mohan, 2007, “Defining Vision: What HomologyThinking Contributes”,Biology & Philosophy, 22(5):675–689. doi:10.1007/s10539-007-9088-4
• Meketa, Irina, 2014, “A Critique of the Principle ofCognitive Simplicity in Comparative Cognition”,Biology& Philosophy, 29(5): 731–745.doi:10.1007/s10539-014-9429-z
• Mesoudi, Alex, Andrew Whiten, and Kevin N. Laland, 2006,“Towards a Unified Science of Cultural Evolution”,Behavioral and Brain Sciences, 29(4): 329–347.doi:10.1017/S0140525X06009083
• Mikhalevich, Irina, 2015, “Experiment and Animal Minds: Whythe Choice of the Null Hypothesis Matters”,Philosophy ofScience, 82(5): 1059–1069. doi:10.1086/683440
• Nosek, Brian A., Jeffrey R. Spies, and Matt Motyl, 2012,“Scientific Utopia: II. Restructuring Incentives and Practicesto Promote Truth Over Publishability”,Perspectives onPsychological Science, 7(6): 615–631.doi:10.1177/1745691612459058
• Papineau, David and Cecilia Heyes, 2006, “Rational orAssociative? Imitation in Japanese Quail”, in Hurley and Nudds2006: 187–196.doi:10.1093/acprof:oso/9780198528272.003.0008
• Pigliucci, Massimo, 2007, “Do We Need an ExtendedEvolutionary Synthesis?”,Evolution, 61(12):2743–2749. doi:10.1111/j.1558-5646.2007.00246.x
• Powell, Russell, Irina Mikhalevich, Corina Logan, and Nicola S.Clayton, 2017, “Convergent Minds: The Evolution of CognitiveComplexity in Nature”,Interface Focus, 7(3): 20170029.doi:10.1098/rsfs.2017.0029
• Premack, David and Guy Woodruff, 1978, “Does the ChimpanzeeHave a Theory of Mind?”,Behavioral and Brain Sciences,1(4): 515–526. doi:10.1017/S0140525X00076512
• Ramsey, Grant, 2013, “Culture in Humans and OtherAnimals”,Biology & Philosophy, 28(3):457–479. doi:10.1007/s10539-012-9347-x
• –––, 2017, “What Is AnimalCulture?”, inThe Routledge Handbook of Philosophy of AnimalMinds, Kristin Andrews and Jacob Beck (eds.), New York: Routledge,345–353 (chapter 32).
• Reiss, Julian, 2015, “A Pragmatist Theory ofEvidence”,Philosophy of Science, 82(3): 341–362.doi:10.1086/681643
• Romero, Felipe, 2019, “Philosophy of Science and theReplicability Crisis”,Philosophy Compass, 14(11):article e12633. doi:10.1111/phc3.12633
• –––, 2020, “The Division of ReplicationLabor”,Philosophy of Science, 87(5): 1014–1025.doi:10.1086/710625
• Rudner, Richard, 1953, “The ScientistQua ScientistMakes Value Judgments”,Philosophy of Science, 20(1):1–6. doi:10.1086/287231
• Sani, Fabio and John Todman, 2006,Experimental Design andStatistics for Psychology: A First Course, Malden, MA: Blackwell.doi:10.1002/9780470776124
• Schnell, Alexandra K., Nicola S. Clayton, Roger T. Hanlon, andChristelle Jozet-Alves, 2021, “Episodic-like Memory Is Preservedwith Age in Cuttlefish”,Proceedings of the Royal Society B:Biological Sciences, 288(1957): 20211052.doi:10.1098/rspb.2021.1052
• Schnell, Alexandra K., Piero Amodio, Markus Boeckle, and Nicola S.Clayton, 2021b, “How Intelligent Is a Cephalopod? Lessons fromComparative Cognition”,Biological Reviews, 96(1):162–178. doi:10.1111/brv.12651
• Seed, Amanda, Nathan Emery, and Nicola Clayton, 2009,“Intelligence in Corvids and Apes: A Case of ConvergentEvolution?”,Ethology, 115(5): 401–420.doi:10.1111/j.1439-0310.2009.01644.x
• Serpell, James A., 2005, “People in Disguise:Anthropomorphism and the Human-Pet Relationship”, inThinking with Animals: New Perspectives on Anthropomorphism,Lorraine Daston and Gregg Mitman (eds.), New York: Columbia UniversityPress, 121–136 (chapter 6).
• Shanahan, Murray, Matthew Crosby, Benjamin Beyret, and Lucy Cheke,2020, “Artificial Intelligence and the Common Sense ofAnimals”,Trends in Cognitive Sciences, 24(11):862–872. doi:10.1016/j.tics.2020.09.002
• Shettleworth, Sara J., 2009, “The Evolution of ComparativeCognition: Is the Snark Still a Boojum?”,BehaviouralProcesses, 80(3): 210–217.doi:10.1016/j.beproc.2008.09.001
• –––, 2010a,Cognition, Evolution, andBehavior, second edition, Oxford/New York: Oxford UniversityPress.
• –––, 2010b, “Clever Animals and KilljoyExplanations in Comparative Psychology”,Trends in CognitiveSciences, 14(11): 477–481.doi:10.1016/j.tics.2010.07.002
• –––, 2013,Fundamentals of ComparativeCognition, (The Fundamentals of Cognition Series 2), New York:Oxford University Press.
• Shevlin, Henry and Marta Halina, 2019, “Apply RichPsychological Terms in AI with Care”,Nature MachineIntelligence, 1(4): 165–167.doi:10.1038/s42256-019-0039-y
• Shevlin, Henry, Karina Vold, Matthew Crosby, and Marta Halina,2019, “The Limits of Machine Intelligence: Despite Progress inMachine Intelligence, Artificial General Intelligence Is Still a MajorChallenge”,EMBO Reports, 20(10): e49177.doi:10.15252/embr.201949177
• Sober, Elliott, 1990 [1994], “Explanation in Biology:Let’s Razor Ockham’s Razor”,Royal Institute ofPhilosophy Supplement, 27: 73–93. Reprinted as“Let’s Razor Ockham’s Razor”, in hisFroma Biological Point of View: Essays in Evolutionary Philosophy,(Cambridge Studies in Philosophy and Biology), Cambridge: CambridgeUniversity Press, 1994, 136–157. doi:10.1017/S1358246100005051doi:10.1017/CBO9780511624940.008
• –––, 2005, “Comparative Psychology MeetsEvolutionary Biology: Morgan’s Canon and CladisticParsimony”, inThinking with Animals: New Perspectives onAnthropomorphism, Lorraine Daston and Gregg Mitman (eds.), NewYork: Columbia University Press, 85–99.
• –––, 2012, “Anthropomorphism, Parsimony,and Common Ancestry”,Mind & Language, 27(3):229–238. doi:10.1111/j.1468-0017.2012.01442.x
• Starzak, Tobias Benjamin and Russell David Gray, 2021,“Towards Ending the Animal Cognition War: A Three-DimensionalModel of Causal Cognition”,Biology & Philosophy,36(2): article 9. doi:10.1007/s10539-021-09779-1
• Sutton, Richard S. and Andrew G. Barto, 1998,ReinforcementLearning: An Introduction, (Adaptive Computation and MachineLearning), Cambridge, MA: MIT Press.
• Taylor, Alex H., Amalia P.M. Bastos, Rachael L. Brown, and ColinAllen, 2022, “The Signature-Testing Approach to MappingBiological and Artificial Intelligences”,Trends inCognitive Sciences, 26(9): 738–750.doi:10.1016/j.tics.2022.06.002
• Taylor, Alex H., Felipe S. Medina, Jennifer C. Holzhaider, LindsayJ. Hearne, Gavin R. Hunt, and Russell D. Gray, 2010, “AnInvestigation into the Cognition Behind Spontaneous String Pulling inNew Caledonian Crows”,PLoS ONE, 5(2): e9345.doi:10.1371/journal.pone.0009345
• Tomasello, Michael and Josep Call, 2006, “Do ChimpanzeesKnow What Others See—or Only What They Are Looking At?”,in Hurley and Nudds 2006: 371–384.doi:10.1093/acprof:oso/9780198528272.003.0017
• –––, 2008, “Assessing the Validity ofApe-Human Comparisons: A Reply to Boesch (2007).”,Journalof Comparative Psychology, 122(4): 449–452.doi:10.1037/0735-7036.122.4.449
• Ullman, Tomer D., Elizabeth Spelke, Peter Battaglia, and Joshua B.Tenenbaum, 2017, “Mind Games: Game Engines as an Architecturefor Intuitive Physics”,Trends in Cognitive Sciences,21(9): 649–665. doi:10.1016/j.tics.2017.05.012
• Van Der Vaart, Elske, Rineke Verbrugge, and Charlotte K.Hemelrijk, 2012, “Corvid Re-Caching without ‘Theory ofMind’: A Model”,PLoS ONE, 7(3): e32904.doi:10.1371/journal.pone.0032904
• Van Horik, Jayden O., Nicola S. Clayton, and Nathan J. Emery,2012, “Convergent Evolution of Cognition in Corvids, Apes andOther Animals”, inThe Oxford Handbook of ComparativeEvolutionary Psychology, Todd K. Shackelford and Jennifer Vonk(eds.), Oxford: Oxford University Press, 80–101.doi:10.1093/oxfordhb/9780199738182.013.0005
• Voudouris, Konstantinos, Matthew Crosby, Benjamin Beyret,José Hernández-Orallo, Murray Shanahan, Marta Halina,and Lucy G. Cheke, 2022, “Direct Human-AI Comparison in theAnimal-AI Environment”,Frontiers in Psychology, 13:711821. doi:10.3389/fpsyg.2022.711821
• Whiten, Andrew, 2022, “Blind Alleys and Fruitful Pathways inthe Comparative Study of Cultural Cognition”,Physics ofLife Reviews, 43: 211–238.doi:10.1016/j.plrev.2022.10.003
• Whiten, Andrew and Richard W. Byrne, 1988, “TacticalDeception in Primates”,Behavioral and Brain Sciences,11(2): 233–244. doi:10.1017/S0140525X00049682
• Woodruff, Michael L., 2017, “Scientific Uncertainty and theAnimal Sentience Precautionary Principle”,AnimalSentience, 2(16): article 11. doi:10.51291/2377-7478.1237
• Wynne, Clive D. L., 2004, “The Perils ofAnthropomorphism”,Nature, 428(6983): 606–606.doi:10.1038/428606a
• –––, 2007, “What Are Animals? WhyAnthropomorphism Is Still Not a Scientific Approach toBehavior”,Comparative Cognition & BehaviorReviews, 2: 125–135. doi:10.3819/ccbr.2008.20008
• Zhang, Mingjun, 2020, “The Use and Limitations ofNull-Model-Based Hypothesis Testing”,Biology &Philosophy, 35(2): article 31.doi:10.1007/s10539-020-09748-0

Academic Tools

	How to cite this entry.
	Preview the PDF version of this entry at theFriends of the SEP Society.
	Look up topics and thinkers related to this entry at the Internet Philosophy Ontology Project (InPhO).
	Enhanced bibliography for this entryatPhilPapers, with links to its database.

Other Internet Resources

• Beyret, Benjamin, José Hernández-Orallo, Lucy Cheke,Marta Halina, Murray Shanahan, and Matthew Crosby, 2019, “TheAnimal-AI Environment: Training and Testing Animal-Like ArtificialCognition”, arXiv:1909.07483. doi:10.48550/ARXIV.1909.07483 [Beyret et al. 2019 available online]
• ManyPrimates
• ManyBirds
• ManyDogs
• “Psychlab: A Psychology Laboratory for Deep Reinforcement Learning Agents”, Joel Z. Leibo, Cyprien de Masson d’Autume, Daniel Zoran, DavidAmos, Charles Beattie, Keith Anderson, Antonio GarcíaCastañeda, Manuel Sanchez, Simon Green, Audrunas Gruslys, ShaneLegg, Demis Hassabis, and Matthew M. Botvinick, 2018, arXiv.doi:10.48550/ARXIV.1801.08116
• Nonhuman Rights Project

Related Entries

abduction |animal: cognition |animal: consciousness |animals, moral status of |artificial intelligence |associationist theories of thought |behaviorism |bias, implicit |cognitive science |connectionism |evidence |functionalism |models in science |other minds |science: theory and observation in |scientific explanation |simplicity |underdetermination, of scientific theories

Acknowledgments

Many thanks to Colin Allen and Kristin Andrews for helpful comments onprevious drafts of this entry.