R2.1. The need for unambiguous expressions

The most frequent source of ownership intuitions consists of people's statements about particular things and individuals. That is why it is important to understand how human minds process such utterances. I argued that we cannot properly describe the psychological processes that underpin ownership intuitions using common terms like “ownership,” “property,” “possession,” etc., because they carry unexamined assumptions. It would be greatly comforting if such worries were unfounded, and we could share DeScioli's commendable trust in the clarity of everyday English, including words like “possess” or “be kind.”

Reading the commentaries, however, seems to justify one's worries, as these (perfectly relevant) contributions include no fewer than 10 statements to the effect that “ownership is” this, “ownership is founded on” that, or “the origin of ownership lies in” some other thing – a close reading of which suggests that the authors are using the same term in vastly different senses. So it is still worthwhile to use an abstract notation despite DeScioli's forceful objections to the horror of “computer code.”

There is also a substantive question here. We should not assume that mental representations are always translatable as words of the natural language (Jackendoff, Reference Jackendoff1983, Reference Jackendoff1995). For example, the conceptual semantics of possession and ownership verbs, sketched by Jackendoff, requires a description of the sentence “Beth owns the doll” as

$$[ {{\rm State\ B}{\rm E}_{{\rm Poss}}( {[ {{\rm DOLL}} ] } ) , \;\,{\rm Place\ A}{\rm T}_{{\rm Poss}}( {[ {{\rm BETH}} ] } ) } ]$$

which may be cumbersome, inelegant, and difficult to learn, but conveys the precise semantic relations involved and nothing else (Jackendoff, Reference Jackendoff1983, pp. 191–192).

R2.2. Does language reveal implicit concepts?

It is tempting to assume that commonalities in linguistic forms may provide us with information about the mental concepts. Kemmerer rightly emphasizes the recurrence in many languages of a distinction between (what could be glossed as) “alienable” versus “inalienable” features of a person, metaphorically extended to different kinds of close or more distant possession. These are remarkable facts, but they raise the question of their connection to other mental concepts. The alienability feature expressed in some languages might be a foundational mental concept, but might also be a vague prototypical representation with little inferential potential.Footnote ¹

If the latter is the case, one may expect that verbal communication can be parsimonious and efficient, despite expressing categories (like alienable/inalienable) that are vague and often arbitrarily applied, because non-verbal mental concepts related to ownership are precise and provide rich inferences. That is, conceptual knowledge about the connections between people and their shadows, people and their thoughts, people and their shoes, etc., is perhaps rich and precise enough that some extremely vague “alienability” distinctions can be used appropriately in most contexts. This would be the null hypothesis in a Whorfian project to measure the effect of such distinctions. That is of course speculative – as Kemmerer rightly points out, there is very little research in this direction.

R2.3. A unified ownership concept? The evolutionary perspective

Concerning the (non-linguistic) mental concepts activated, the model is minimalist in the sense that we do not assume that a mentally represented concept underlies all the uses of possessive terms, verbs like “own” and “possess,” and all motivations or emotions connected with these terms. Rather, we start from a functional description of what an ownership psychology is for. In humans as in other animals, some cognitive process supports the competitive acquisition and use of rival resources; in humans in particular, rival resources are also crucial to the deployment of cooperative relations. The goal is to describe how the psychological processes involved in these two domains of selective pressure would account for many aspects of ownership intuitions.

As a consequence, the model is relevant to many aspects of these intuitions, for example, it does include the valuation of some possessed objects as more important than others (contra Kemmerer), but leaves aside adjacent phenomena that are not relevant to this evolutionary background. That is why Morewedge is right that the model does not include some phenomenologically important phenomena, like one's attachment to a deceased relative's belongings, a private diary, etc. In the same way, Hood mentions people's motivation to guard or cherish “otherwise worthless sentimental objects, sacred artefacts or memorabilia.” These representations are not produced by the systems that regulate the use of rival things among con-specifics, and therefore lie beyond the scope of the model.Footnote ²

R3.1. Definition

The model specifies that various cues (such as contiguity, usage, etc.) called P-cues, result in a represented P(agent, thing) tag. Frustrated by what seems “enigmatic” to him, DeScioli asks “Does the P allude to possession, property or something else?” The P() tag does not mean any of these things, because none of these common words is at all precise. What matters here is that the term P() can be computationally described, in terms of its inputs, inference rules, and outputs. That is also why, as Belk & Atasoy lament, the term “possession” is not defined in the target article, as it is not part of the technical terms in the model.

As described in section 3, the P() cues are species-specific, being relevant to the kinds of competitive interactions that humans can establish. Although Arnhart claims that the model provides “no evolutionary explanation” for P() intuitions of associations between agents and things, it would seem that the existence of external, localized, rivalrous, and partly excludable fitness-relevant resources provides the selective pressure for such concepts, as observed in many different species (Eswaran & Neary, Reference Eswaran and Neary2014; Strassmann & Queller, Reference Strassmann and Queller2014). This explains why humans and other animals sometimes use similar P() cues, as described by Kangiesser, who also notes that the likelihood of prevailing in a conflict is crucial in many primates. This was probably true of early hominins as well. In many apes and humans, as noted by Merker, one crucial domain of P() cues would concern territories (DeScioli & Wilson, Reference DeScioli and Wilson2011).Footnote ³

Beyond territories, humans in prehistory probably extended P() representations to tangible objects like tools. It may be the case, as Merker proposes, that incipient property in tools was a watershed in ownership representations, as the strength of P() tags about tools would reflect the amount of past investment, not just immediate utility. Also, tools illustrate another aspect of ownership, the possibility of priority in future access to the thing, which becomes central in human ownership intuitions (Tatone). Indeed, many of the P() cues that are specific to human interactions concern access to future utility, and the exclusion of others from that future access (Pesowski & Powell, Reference Pesowski and Powell2023).Footnote ⁴

R3.2. What processes create P() tags?

The ownership psychology model describes generic P() cues. In actual situations, domain-specific information makes these cues more precise. What counts as contiguity for ideas is not the same as for objects. As far as tangible, physical objects are concerned, it is clear that some low-level perceptual processes provide P() cues that remain implicit, as emphasized by Tummolini. In particular, the spatial contiguity between people and material objects triggers a specific form of processing, which in a way primes the expectation of special associations between person and thing, leading to the P() tags described in the model. For instance, it is quite striking that, as Tummolini reports, when a tag becomes activated, observers perceive the thing as further away from themselves.

The developmental emergence of P() and L() intuitions may also be linked to control, as Tummolini suggests, although as he points out, there is no systematic study of that process. He proposes that some form of conceptual learning process may lead from maps of controllable objects, to a more abstract concept that would be a precursor of P() tags. This however would not be the starting point of P() concepts, because as Tatone demonstrated, elements of both P() tags and even the GIVING relation appear very early in infancy, before children have much experience of which parts of the world are under control. The point is that control provides gradual enrichment of the database of things that become the t in P() tags. That is why Wispinski et al. are right that people's representations of things that they control certainly take into account possible actions, the limitations of one's body, etc. This is very much assumed by the description of P() cues in the minimalist model. For instance, contiguity between agent and thing is a parameter that depends on species-specific cues, which themselves certainly integrate expectations about agents' mobility – that is why the extension of a territory for instance may vary a lot between species.

P() tags are based on cues that may be processed very differently by two agents in interaction, which is one major reason why it is important to focus on disagreements, as Echelbarger & Tully point out, and as developmental psychologists have done for some time (Blake & Harris, Reference Blake and Harris2009; Noles & Gelman, Reference Noles and Gelman2014). Indeed, ownership disputes (outside the legal system) are almost never about the explicit principles, but about the relevant facts. The study of disagreement can also reveal that people's intuitions about their own connection to things are of course different from other people's perception of that connection. As Morewedge illustrates, Yasser and Yitzhak may have misperceptions of each other's attachment to a piece of land. A third party will not share or represent those feelings.

R3.3. Are P() tags related to body representation?

It is tempting to speculate that P() cues result from some more “primitive” processes in the mind. For instance, Arnhart writes that the “best explanation” for how agents could have representations of possessing objects (which he calls “ownership,” illustrating the problem of using such terms) lies in the neurobiology of “self-ownership.” Also, Wispinski, Inns, & Chapman state that the implicit rules of ownership for things “very much follow the implicit rules for possessing one's body parts.” Many intellectual traditions assume that ownership has its “origins” in ownership of one's own body. This can also be seen in developmental arguments, see for instance (Rochat, Reference Rochat2014).

There may well be connections between the domains of self-perception and ownership intuitions. But it is difficult to state that the latter are “grounded in” or “founded upon” the former, because those terms do not by themselves denote any precise, tractable computational process. They may imply that, ontogenetically, humans develop their ownership intuitions as an inferential outcome of representations self and other. That is tempting, and psychologists like Rochat have indeed followed that path (Rochat, Reference Rochat2011). But that cannot be the whole story. As mentioned before, infants do have intuitions about transfers of possessions, which do not seem to be extracted from their intuitions about their own bodies (Tatone, Geraci, & Csibra, Reference Tatone, Geraci and Csibra2015).

R3.4. Are P() tags represented as extensions of the self?

A distinct but related claim is that owned things are in some way represented as parts of or extensions of the self. Hood for instance emphasizes that “owned items can represent part of own's identity,” a sentiment echoed by Nancekivell & Pesowski, who emphasizes similarities between the way we represent (some) things possessed and our representations of ourselves. Rochat also links the development of possession or ownership concepts to other important conceptual changes after infancy, also connecting ownership to control and the body.

The notion of possession as extending the self raises difficult questions. First, we would need to complete it with a description of the processes that are specific to non-really-self objects, since humans do not represent their connection to their coats or money exactly like their connection to other aspects of their selves. Second, if we claim that possessions are extensions of the self, we must also explain how these self-extensions ensure the efficient management of competitive acquisition or the cooperative interactions around property. What cues would make people understand that a thing is an extension of a particular individual's self, as opposed to others agents'?

Nancekivell & Pesowski describe the extended-self perspective as more minimalist than the minimalist model. Without wishing to escalate this competition in parsimony, one should point out that the model is minimal in the following sense – that we do not need to postulate any psychological processes, beyond two mechanisms (competitive acquisition and mutualistic cooperation) that are already independently documented in a vast literature. By contrast, the extended-body or extended-self metaphors are additional mechanisms postulated specifically in order to explain ownership phenomena.Footnote ⁵

R4.1. Beyond Bourgeois equilibria

Humans differ from other apes in (among many other things) two relevant features: (a) Having a special attitude to things used by other agents, and (b) maintaining extensive cooperation with non-relatives. In the minimalist model, that is not a coincidence, as attitudes to others' possessions are driven by cooperation expectations. Human ownership intuitions and motivations do not reduce to the representation of P() cues – they are vastly different from the representations that underpin the Bourgeois equilibria of evolutionary game theory.

This argument goes against a possible “cynical” interpretation of ownership, following which agents just follow the Bourgeois strategy, with an added veneer or cover of respectability provided in the human case by moral justifications. This is what Shechter, Gilead, & Bereby-Meyer (Shechter et al.) seem to suggest, arguing that ownership principles are in effect formulated by the haves against the have-nots, as a justification for material dominance. Even allowing for some exaggeration, is this perspective justified? The evidence from psychology suggests the opposite. Children from an early age and adults in all known cultures make a sharp distinction between mere possession and accepted, “legitimate” possession, between P() and L() tags. As André, Fitouchi, & Baumard (André et al.) point out, in this domain, people just do not assume that might be right, and the inferences from L() tags are often exactly the opposite of what competitive acquisition would recommend – which is why people generally do not condone that a thief appropriate an old lady's things (see sect. 8.2.2). It is also difficult to follow Shechter et al. when they state that “legitimate ownership of one object can influence perceptions of legitimate ownership of another object,” so that P(A, t₁) would justify a representation L(A, t₂). That generally does not happen, as L(A, t₂) requires its own cues.

Comparative psychology too suggests that human ownership psychology does not reduce to P() cues. Interactions around the use of rival resources among apes can be described entirely in terms of competitive acquisition, with P() cues (including rank, relative formidability, etc.) providing the required information. But as Kangiesser reports, non-human apes do not seem to engage in any of the behaviors associated with “respect” to legitimate possession.

The addition of L() tags to competitive acquisition may be one of those features that allowed human communities to become gradually larger and more complex. The study of social evolution provides many descriptive and explanatory models of complexification processes (Dubreuil, Reference Dubreuil2010; Johnson & Earle, Reference Johnson and Earle2000; Marcus, Reference Marcus2008; Sanderson, Reference Sanderson2014; Service, Reference Service1962). Dale connects this increase in group size to ownership, pointing out that “at a certain point, the costs of intragroup competition become so high that they begin to outweigh the benefits of group living.” That would describe groups in which interactions are ruled by people's P() cues about possession and competitive acquisition. By contrast, a group in which ownership is “respected” can grow much larger, as this vastly reduces the costs of competition, and indeed increases the opportunities for mutualistic cooperation.

It may be that Dale's argument is on the right track, albeit once formulated in a broader frame. Recall that, in the minimalist model, ownership intuitions are explained by the broader set of cognitive mechanisms that support mutualistic cooperation via partner choice and reputation (André & Baumard, Reference André and Baumard2011). In other words, a parsimonious description of the complexification process would be that the emergence of partner–choice cooperation ushered in both the possibility of large-scale communities and a generalized “respect” for property.

R4.2. Discrepancies between inferences from P() and L() tags

The discrepancy between P() and L() tags should be emphasized as it may lead to misunderstandings. For instance, Friedman points out that we should not directly equate the perceived legitimacy of an ownership claim (“It really is her car”) with an argent's willingness to fight to keep a thing. Willingness to fight is also influenced by both the likelihood of prevailing, and the intrinsic value of the thing to the current user. That is true and should have been emphasized, as André et al. also make that point. True, the model does stipulate that L() tags inherit information from P() tags, notably as concerns the strength of association. But that does not mean that L() tags reduce to the information in P() tags. Indeed, in many cases, L() tags support inferences that counter those from P() tags. That is clearly true when cooperation leads to “respecting” property, as described in section 8.2. In such cases, the motivation to preserve Min() interactions over-rides the advantages of acquisition.

That is also why people can be seen as “owners” of things they don't particularly like, as Friedman points out. The fact that I have Min() expectations toward Melanie means that I cannot leave the party with the coat she was wearing on her way in. That is independent from the value of that coat to Melanie, because cooperative interactions (made possible by Min() expectations) require that we do not impose any uncompensated costs on others, even if those costs were minimal (e.g., if she did not like that coat at all). Indeed, when the value of partners is extremely high (e.g., with one's spouse), people can even engage in what would seem to be violations of L() tags, as noted by Fonn, Zahl, Kristensen, Margoni, & Thomsen (Fonn et al.) One can for instance borrow their things without asking. As mutualistic cooperation models predict, people may accept interactions in which they willingly incur costs to benefit others' fitness – see models of friendship (Tooby & Cosmides, Reference Tooby, Cosmides, Runciman and Smith1996) and fitness interdependence (Barclay, Reference Barclay2016).

On a related point, Friedman also points out, quite rightly, that L() tags are generally not based on explicit computations about the cues concerning willingness to compete. That is, we routinely go about interacting with people without explicitly representing how strongly they would resist if we tried to seize their coats or their shoes. Rather, by activating Min() expectations toward them, we assume by default that all the things they use, for which we have P() cues, are also described by default L() tags (see sect. 7.2.3). The person who is wearing the shoes is by default assumed to be in an L() relation with those shoes – a default assumption that is of course defeasible.

R4.3. Cooperation and moral psychology

The activation of cooperation expectations (Min() in the model) is the reason why many aspects of ownership intuitions are moralized. People do not just “respect” property but feel that they ought to do so, and that others should too. This is not surprising, as moral intuitions and emotions are grounded in natural selection for cooperation (Baumard, André, & Sperber, Reference Baumard, André and Sperber2013; Curry, Reference Curry2016).

In the minimalist model, there is no moral intuition about ownership intuitions that is not in fact derived from a moral intuition about cooperation. There is no special morality of ownership. That of course runs against Atari & Haidt's proposal that ownership should be seen, alongside care/harm, fairness, purity, loyalty, authority, and liberty, as one of the separate domains or “foundations” of morality (Haidt & Joseph, Reference Haidt and Joseph2007). It is an empirical question, whether the psychology of morality, in its different domains, requires distinct cognitive processes (Curry, Alfano, Brandt, & Pelican, Reference Curry, Alfano, Brandt and Pelican2022; Haidt & Joseph, Reference Haidt and Joseph2007) or can be derived from a common set of cooperative assumptions (André, Debove, Fitouchi, & Baumard, Reference André, Debove, Fitouchi and Baumard2022). The point of the model was not to adjudicate these general debates. However, in the domain of ownership, it does seem that all the intuitions concerning interactions and their moral aspects could be derived from cooperative fairness computations.

In the minimalist model, P() tags lead to L() tags but only within the range of Min() cooperation expectations. This suggests, as mentioned in section 10.3, that communities with different assumptions about the range of Min() would have different levels of “respect” for property. But the shift from cooperation to competitive acquisition may also result from transient aspects of a situation. Mitkidis & Elbaeck point out that extreme scarcity or cues of competition may lead to ignoring Min() expectations, with the consequence that people are back in competitive acquisition, and that L() tags concerning people and things are not activated anymore. So changes in P() cues can have a deep effect on people's activation of moral inferences concerning the taking and sharing of resources, activating what Mitkidis & Elbaeck call a “maximizing mindset” (Elbaek, Mitkidis, Aarøe, & Otterbring, Reference Elbaek, Mitkidis, Aarøe and Otterbring2021). They also predict that particular, temporary cues may get people to shift in and out of that mindset, in contrast with the literature that generally focuses on stable, general traits of a community like generalized social trust.

R4.4. Is cooperation sufficient?

The minimalist model specified that cooperation is necessary for L() tags. But one may want to argue that cooperation is actually sufficient, as André et al. suggest. In their view, the intuitions delivered by cooperation systems are very different from those of competitive acquisition, whose input is not actually relevant to ownership intuitions as such, or L() tags. In that proposed modification of the minimalist model, once two agents entertain cooperation expectations, what I called Min(), the input from P() cues would be entirely shut off. All inferential processes concerning who owns what, and what can be done, would be derived from the cooperation systems only.

That is of course possible but may not be parsimonious, because (a) there is evidence that human minds automatically activate P() tags from P-cues (see discussion of the psychophysics in sect. R3.2 above), and (b) the information contained in these P-cues is in many cases precisely the same as the information used by cooperation systems. So it would be surprising if one cognitive system (managing cooperation) ignored the output of another system (managing competitive acquisition) but also happened to re-create the information already provided by that other system.

For instance, as described in section 7.4, the fact that Melanie worked a lot on her garden would explain why (in contexts of competitive acquisition) she will defend it with greater energy than if she had not done much work, and (in contexts of cooperation) why people will be more motivated to leave her flowers alone than if she had not, why they will be more motivated to help her defend her garden against intruders, etc. So the same information (work, contiguity, etc.) with the same inferential outcome (Melanie's association with the garden is stronger than if she had done no work) is activated in both contexts. It would be strange if the cooperation system ignored that information and then created it anew.

To take another example, people standing in line occupy a position whose value is a direct function of its proximity to the head of the queue. In a situation of competitive acquisition, we can expect that people near the head of the line will defend their position against intruders with more vigor than the ones near the end, and that is actually the case (Fagundes, Reference Fagundes2017; Mann, Reference Mann1969). In situations of cooperative trade, when people sell their place in line, we can expect that places near the head of the queue will command a higher price than those at the tail, and that is actually the case (El Haji & Onderstal, Reference El Haji and Onderstal2019; Humphrey, Reference Humphrey1991; Zhou & Soman, Reference Zhou and Soman2008). Again, one might say that cooperative trade re-creates some information that was already available from competitive interaction, but parsimony would suggest otherwise.

It is of course an empirical question, whether this coincidence – the L() tags inherit information already contained in P() tags – is the general case as described in the minimalist model. I would argue that it is very general.

Evolutionary considerations would support this inheritance interpretation. Our environments of evolution included many situations in which (a) one might interact with both Min() cooperators and non-cooperators (e.g., tribal outsiders), (b) one might be unsure whether the interaction with a particular agent A falls under Min() or not, and (c) that situation might change, so that one extends Min() expectations to previous non-cooperators, or vice-versa. In all these situations, an agent may, first, need to consider both perspectives simultaneously (the two of us may be in competition or perhaps in cooperation) with their respective costs and benefits, and second, be able to shift easily between these alternative readings if the situation changes. Given the likely frequency and fitness relevance of such situations, it would be odd to expect that the computations of one system ignore the output of the other.

R5.1. Why development matters

Developmental psychology is a major contributor to our understating of ownership, thanks in particular to Ori Friedman and his collaborators and students (Friedman, Reference Friedman2010; Friedman & Neary, Reference Friedman and Neary2008; Friedman & Ross, Reference Friedman and Ross2011; Friedman, Neary, Defeyter, & Malcolm, Reference Friedman, Neary, Defeyter, Malcolm, Ross and Friedman2011; Nancekivell, Friedman, & Gelman, Reference Nancekivell, Friedman and Gelman2019), several of whom commented on the target article. More generally, the models of ownership psychology proposed by developmentalists (Friedman et al., Reference Friedman, Neary, Defeyter, Malcolm, Ross and Friedman2011; Gelman, Manczak, & Noles, Reference Gelman, Manczak and Noles2012; Rochat, Reference Rochat2011) provide a rich source of hypotheses in a field often neglected by non-developmental cognitive psychology.

The developmental process may be complex, as Rochat points out, and it may not be entirely clear how it maps to the adult concept. For instance, Rochat suggests that there is a connection between the development of a sense of control over things and the emergence of explicit P() tags. This is confirmed by the low-level psychophysics of possession, as commented on by Tummolini. But we must also be cautious of reading our adult intuitions in infant behavior. For instance, is the infant's familiar game of gaining and losing and regaining control over objects necessarily related to ownership? One might as well consider it as, precisely, a control game, that is, prompted by a motivation to develop motor capacities that allow control and release of objects, independent of any inferences concerning ownership or possession.

It is also possible that Noles is correct in proposing that there is a trade-off between the generality of the model, which applies to many domains of adult intuitions, and its explanatory power in the domain of development. That is also the substance of Blake's detailed comments on possible discrepancies between the minimalist model and some aspects of development, discussed below.

R5.2. Does the developmental sequence match the model?

Blake describes how young children may both defend a third-party's access to a thing, but it is later that they “respect” another child's property in dyadic interactions. This seems to go against the sequence suggested in section 7.2, which describes the dyadic, triadic, and then generalized (default) creation of L() tags. But the order in the model is a logical order, that is, it describes the interactions from most reduced to most extended. The point is that the inferences described in generalized interactions (hypothesis 4) could be represented by an agent using a concatenation of dyadic or triadic interactions (described in hypotheses 2 and 3). That does not imply a chronological sequence, whereby people start by interacting in dyads and then gradually complexify their social world.

Such discrepancies are of interest, because they illustrate how cognitive development certainly does not consist of a coherent piling-up of complex, detailed structures on top of simpler, more general ones. Developmental sequences may reflect constraints other than conceptual coherence, as I discuss below.

Another challenge to the minimalist model stems from infancy research. As Tatone comments, there is no evidence that the sort of intuitions captured by L() tags is available to very young infants. But, by contrast, there is ample evidence that infants have sophisticated expectations about the transfer of control over objects, as described in section 10.1.1 for giving, but also concerning competitive acquisition and the allocation of resources. So, some parameters of P() tags seem to be available at that early stage of development. Tatone also points out that these could be interpreted as incipient L() tags, to the extent that they are connected to social interaction between the agents. As briefly mentioned in that same section 10.1.1, a voluntary transfer of control over a thing t from A to B predicts subsequent social interaction between the agents. Tatone proposes that the early concepts of resource control, voluntary exchange, and elementary debt monitoring, do all the work that is supposed to be done by an ownership system. Why develop an ownership system then? Tatone would suggest that this is because the utility of things around us is complex enough, that it needs to be detached from the rough expectations delivered by the P() tags. For instance, potential and deferred utility may be one situation that makes it necessary to go beyond the P() tags. That is very much in the same spirit that I propose more general (and more speculative) explanations below.

R5.3. A functional speculation

One may be tempted to speculate, on the basis of these commentaries by Blake, Noles, and Tatone, that the order in which different aspects of ownership psychology emerge in development may obey constraints that have little to do with conceptual coherence (one concept appearing after its presuppositions appeared), and much to do with efficiency given ecological constraints.

Specifically, the minimalist model proposes that (adult) ownership intuitions are grounded in competitive acquisition and mutualistic cooperation. Now these evolutionary aspects of human interaction have different consequences at different developmental stages. As a consequence, one may hypothesize that the various conceptual components emerge at times when they favor a child's welfare (in typical ancestral conditions of human evolution). That way of considering development may not be totally eccentric. For instance, it provides a simple account for the developmental emergence of different anxiety targets (Boyer & Bergstrom, Reference Boyer and Bergstrom2011).Footnote ⁶

This may illuminate apparent paradoxes of development. Relevant to Tatone's remarks, one might hypothesize that infants need to figure out the parameters of agency and causation, the interaction of third parties around them, including relations of dominance, the possibility of forced and unforced transfers, the role of transfers in creating social relations – because all these are involved in early interactions. By contrast, the long-lasting aspects of ownership (L() tags) are irrelevant to their social interactions. Blake's point about the discrepancy between norms and own behavior may also stem from ecological conditions. In the model of cooperation assumed here, agents must balance two independent sets of factors in their cooperation decisions – direct costs/benefits, and reputation costs–benefits. Survival in a partner-choice social exchange context requires that one be chosen as a cooperator, so that reputation is crucial (Baumard et al., Reference Baumard, André and Sperber2013). But the welfare of young children depends, not on social exchange, but on provision from parents, motivated by inclusive fitness. So we should expect that a crucial aspect of adult exchange computation (How does my current behavior affect my status as a potential cooperator?) has no impact on their welfare. By contrast, understanding local norms and monitoring their observance are fundamental to organizing interaction with other children and adults at that age, which is why even preschoolers see norms as imperative (even if local) and why they resent violations (Rakoczy & Schmidt, Reference Rakoczy and Schmidt2013; Rakoczy, Warneken, & Tomasello, Reference Rakoczy, Warneken and Tomasello2008).

R6.1. An evolutionary model predicts local variations

The minimalist model is based on the assumption that evolution by natural selection results in very specific cognitive capacities, which make it possible to acquire vast amounts of information from con-specifics and govern inferences from that information (Sperber, Reference Sperber1996; Tooby & Cosmides, Reference Tooby, Cosmides, Barkow and Cosmides1992). That is why, contra Belk & Atasoy, one should not try to separate the “innate” and “learned” aspects of ownership intuitions, as such distinctions are both misguided and misleading (Boyer, Reference Boyer2018, p. 272ff). A related and common misconception is that evolutionary considerations could only explain cultural universals, which is certainly not the case (Boyer & Petersen, Reference Boyer and Petersen2012).

In the domain at hand, as local cooperation parameters (notably the extension of Min() expectations) vary a great deal between times and places, we would expect corresponding differences in ownership intuitions. Indeed, the model results in specific predictions about, for example, differences in ownership that stem from differences in social trust (see sect. 10.2.). Although Atari & Haidt “warn against” ignoring cultural differences, those are precisely what the model allows us to explain.

R6.2. The case of slavery, a widespread institution

A striking case of cultural variation is the acceptance of slavery. The model specifies that the difference between accepting and rejecting the institution of slavery rests not on differences in ownership principles, but on the extension of the Min() cooperation expectations. But Starmans argues that the rejection of slavery stems from other considerations.Footnote ⁷ Specifically, our psychology would stipulate that one cannot own things with autonomy. Indeed, children readily exclude from the domain of ownership things with some “autonomy” (Starmans & Friedman, Reference Starmans and Friedman2016), as they see control as a crucial feature of ownership (Espinosa & Starmans, Reference Espinosa and Starmans2020). It is of course tempting to infer that, if human minds assume that “autonomy” → “cannot-be-owned,” they would also assume that “can-be-owned” → “is-non-autonomous.”

But is that the case? Historical data suggest otherwise. Slavery was practiced in all kinds of societies for as long as historical documents are available. In many of these cases, there is every indication that people construed slaves as both (a) property that can be traded, stolen, guarded, etc., as well as (b) autonomous agents with volition, cognition, emotions, etc. To consider an extreme example, the Arab scholar Ibn Qutaybah, writing in classical Baghdad, celebrated the exceptional talent of several women poets, whom many rich people were very keen to buy from their owners (Qutaybah, Reference Qutaybah2019).

R6.3. Collective action and private ownership

In the model, it was suggested that, when people represent an abstract collection or institution as “owning” something, that may result in (or be caused by) representations of that institution as an agent. This happens often enough, when we consider that, for example, the state “does not want to sell” some asset. But, as Umbres points out, in many socially important situations, people manage collective rights in resources without generating a notion of ownership by a collective agent. That is for instance the case in commons management. Also, as Umbres writes, in many forms of collective action, people have access to the collectively generated benefits but they privatize them upon receipt.

As Umbres also notes, the notion of ownership by a quasi-agent becomes much clearer, when people engage in a collective action to which they apply coalitional psychology (Tooby, Cosmides, & Price, Reference Tooby, Cosmides and Price2006). That is because coalitions are themselves represented as quasi-agents to begin with (Boyer, Reference Boyer2018, pp. 217–225). So, if people consider that coalitions “want this” or “remember that,” it is quite natural to think that they can “own” a thing. Engaging coalitional psychology also creates an intuition of similarity between members of an alliance, which Grüning & Krueger rightly describe as a crucial factor in facilitating coordination. But similarity needs to be unpacked. For one thing, people cannot really engage in efficient competitive acquisition if they do not have some expectation that P() cues are represented in the same way by the different agents they interact with. But also, people could not extend cooperation to others if they did not have some mental representation of what their Min() expectations consist of. Some of this similarity is of course given “for free” by the fact that different agents belong to the same species, therefore have highly similar capacities. But another set of expected similarities stems from exposure to the same cues, for example, by belonging to the same larger community of norms. The fact that commons can be efficient or unravel shows that the Min() expectations may be necessary but not sufficient.

R6.4. Coordination and institutions

Interactions around rival goods require coordination, specifically in the form of coordination games. That is true of competitive acquisition, in the form of Bourgeois equilibria, and also of cooperative interactions, in the form of mutualism. Because coordination games often have multiple equilibria, arbitrary cues can motivate the choice between strategies, for example, between driving on the right or on the left. The cost of subsequent changes of strategies, which escalates as the number of coordinated agents increases, ensures the stability of such choices, which we call path dependence.

The fact that coordination points may be arbitrary leads Blazsek & Heintz to emphasize what they see as the unbounded variety of ownership institutions. But that may be an exaggeration. First, the fact that some cue is arbitrarily chosen does not imply that its use stems from arbitrary processes. For instance, it is arbitrary that men wear neckties and women skirts, but it is certainly not arbitrary that fashion cues emphasize gender distinctions. In the domain of ownership, there are similar situations. The “choice” between matri- and patrilinear inheritance is a good example. Even if the coordination cue is arbitrary, people find the rules intuitive because they rely on evolved intuitions about genetic relatedness.Footnote ⁸ In the same way, when ranchers and farmers coexist, they have to “choose” which of the two groups will have to enclose their resources. But their inferences from that “choice” follow the intuitive path of optimizing mutual benefits (Ellickson, Reference Ellickson1991). Generally, there are striking similarities in the manifestations of the “property instinct” across the species (Stake, Reference Stake2004).

In this as in other domains, it may be misleading to think of institutions as fundamentally distinct from evolved intuitions. Institutions require mental representations of the distribution of particular mental representations in people's minds in one's social environment (Heintz, Reference Heintz2004; Sperber, Reference Sperber, Levinson and Pierre2006). That is why many institutions owe their persistence not just to path dependence, which would be vulnerable to random drift, but also to their “fit” with evolved intuitive understandings of the domain at hand (Boyer & Petersen, Reference Boyer and Petersen2012). Intuitions constrain inferences from arbitrary choices, which is why legal regimes of ownership are much less varied than we may expect – as discussed below.

R7.1. Why the law matters

Ownership psychology consists of intuitions, that sometimes prompt explicit, reflexive thoughts about principles, which themselves can influence the production of legal norms (sect. 8.1.1, Fig. 2). One can only agree with Feeney & Hickey, as well as Patrick and Blazsek & Heinz, that there should be greater integration between the psychology of intuitions with the study of legal institutions. Indeed, legal scholars were far ahead of psychologists in putting forth sophisticated models of ownership. First, they accepted deflationary (or, one might say, minimalist) accounts of ownership for a long time. A clear example is Honoré's proposal that ownership is not a single unified right, but rather a “bundle” of distinct, conceptually independent rights, combined in different ways in different domains and different societies (Honoré, Reference Honoré1961; see also Hohfeld, Reference Hohfeld2014[1923]). This view is largely shared by legal scholars to this day (Attas, Reference Attas2006; Glackin, Reference Glackin2014). Second, legal scholars and economists have long recognized that ownership decisions cannot just be construed as a matter of dominion over a thing, but include trade-offs to do with externalities (Coase, Reference Coase1960).

R7.2. The intuitive basis of norms and the law

Legal scholars have documented the influence of intuitive understandings on the law in the domains of criminal justice (Robinson, Kurzban, & Jones, Reference Robinson, Kurzban and Jones2007; Sznycer & Patrick, Reference Sznycer and Patrick2020) and more recently of morality (Lieberman & Patrick, Reference Lieberman and Patrick2018). Patrick argues that the domain of ownership leads to the same conclusion – legal nuances about squatters' rights, for instance, seem to mirror uncertainties in our intuitions (sect. 8.3). In the same way, failure to protect or defend one's things leads to a diminished intuition of ownership, an L(A, t, s) tag with small s parameter, but also to a weaker legal claim on property.

More generally, even though legal systems are somewhat detached from our intuitions, they also require that intuitive background. For instance, Honoré's notion of ownership as a bundle of independent rights does in fact imply some shared intuitions about ownership – otherwise there would be no reason to think that the different rights are related at all, that they are about the same problem (Harris, Reference Harris1996). In the same way, decisions about externalities often require intuitively grounded judgments about what externalities must be internalized (Glackin, Reference Glackin2014, p. 22ff), a point that is at the center of recent accounts of ownership (Feeney & Hickey) and of legal decisions about liability (Patrick).

This raises the question, how intuitive representations and the explicit legal principles coexist in lawyers' minds. Feeney & Hickey suggest that they may be juxtaposed without much interaction, or that they may actually interact, although there are no studies of that process. As empirical studies in the “law and norms” tradition suggest, it may happen that people know the legal norm though they abide by the informal norms, but it also happens that people follow their intuitive norm and wrongly assume that they are following the law (Ellickson, Reference Ellickson1991). More generally, legal knowledge is certainly “meta-represented,” like many aspects of scientific knowledge (Sperber, Reference Sperber1997).Footnote ⁹

R7.3. Possible effects of laws on folk notions and intuitions

Information may also flow in the opposite direction, from institutions to intuitions. The law can serve as the coordination point described by Blazsek & Heintz, as suggested by Kleiman-Lynch & McCullough's study, in which participants seem to have conflicting intuitions about who actually “owns” the money owed for taxes. Note that people's representations of whether the state “owns” the money you pay in taxes or not, belong to those explicit, “reflexive” beliefs described in section 8.1.1, as a response to intuitions delivered by cognitive systems. In this case (and that is not limited to taxes), it is plausible that our intuitive systems deliver no clear intuition at all. As Kleiman-Lynch & McCullough suggest, participants may interpret the situation as one of contested ownership. That indeed seems quite plausible.

Taxation indeed triggers very different intuitions. People in some countries are happy to contribute to the nations' maintenance, for example, in Scandinavia, while in other places they see taxes as illegitimate, for example, in Russia (Rothstein, Reference Rothstein2005). In one case, people consider that they are contributing to a collective action. In the other, people assume that payment of taxes only reflects their lack of bargaining power, as they cannot resist a powerful predatory state. These differences are correlated with general social trust, as suggested by the minimalist model (sect. 10.2).