Theory of mind and Verstehen (understanding) methodology

Introduction

Theory of mind (ToM) is one of the main topics in psychology, psychiatry, and philosophy of mind. According to Reddy and Morris (2009: 91), ‘Google Scholar yields an incredible 36,000 publications using the term’. In 2014, that same search substantially increased to 130,000 hits. Among them, development of ToM in ordinary children has been studied, and the deficits of ToM with autism and other psychiatric disorders have been reported. Therefore, ToM is an important research topic in psychiatry.

However, studies on ToM are far from monolithic. The divide between theory-theory (TT) and simulation theory (ST) has been noted, and various views exist within TT and ST. Behind these divides, methodological confusion seems to have developed (Obiols and Berrios, 2009). ST is said to be a descendent of the Verstehen (understanding) in the European tradition of human sciences, but ST is not entirely the same as Verstehen methodology.

In the present article, arguments for and against ToM presented in the last three decades will be reviewed. Subsequently, methodological issues with these studies on ToM will be analysed. Lastly, the relationship between studies on ToM and Verstehen methodology will also be examined. Thus, the first section of this article will present a review and history of ToM, and the second part will aid in recapturing the methodologies used in clinical psychiatry.

The term ‘theory of mind’

Many reviews (e.g. Goldman, 2006; Leudar and Costall, 2009a; Sharrock and Coulter, 2009) commonly state that the term ‘theory of mind’ dates back to Premack and Woodruf’s 1978 article entitled ‘Does the chimpanzee have a theory of mind?’. At the beginning of the abstract in their article, Premack and Woodruf (1978: 515) wrote:

An individual has a theory of mind if he imputes mental states to himself and others. A system of inferences of this kind is properly viewed as a theory because such states are not directly observable, and the system can be used to make predictions about the behavior of others.

The authors started their main text with virtually the same sentences, which have been repeatedly quoted by other researchers (Goldman, 2006: 11; Leudar and Costall, 2009b: 28; Sharrock and Coulter, 2009: 59–60). The view presented in these sentences has been seen as the premise of many studies on theory of mind: others’ minds are unobservable and therefore theoretical inference is warranted in order to grasp them.

Premack and Woodruf were well aware of the empathic aspect in understanding others, although that perspective would not be appreciated by later researchers. Premack and Woodruf (1978: 518) stated that ‘it is important to note that empathy and “theory of mind” are not radically different views; they are in part identical’. They also differentiated empathy into (1) ‘if he were in the actor’s position’, and (2) ‘if he were a three-year-old child, a juvenile chimpanzee, a human adult, and so forth’ (p. 518, n. 1). Additionally, they stated that the latter type of empathy seemed equivalent to theory of mind. However, these viewpoints received little subsequent attention.

In 1981, Bretherton, McNew and Beeghly-Smith wrote a book chapter entitled ‘Early person knowledge as expressed in gestural and verbal communication: When do infants acquire a “theory of mind”?’ This work was on the development of infants’ communication. Although they cited Premack and Woodruf’s (1978) work, their chapter suggested that they did not think that ToM alone was sufficient for intentional communication. They added that ‘a framework of meaning’ and ‘an interfacible medium (language or conventional gestures)’ must be shared by an infant and others for communication (Bretherton et al., 1981: 340, original italics). Thus, they reviewed experiments on infants’ acquisition of intentional terms.

In 1985, Baron-Cohen, Leslie and Uta Frith published an article, ‘Does the autistic child have a “theory of mind”?’. They compared the experimental results of false belief tasks conducted on ordinary children with results from children with autism. They interpreted the poor performance of children with autism in the false belief tasks as a deficit in ToM.

Thereafter, Christopher Frith (1992/2008), husband of Uta, interpreted certain symptoms of schizophrenia, such as persecutory delusions, as a deficit in ToM. He preferred the term metarepresentation. Frith (1992/2008: 120) inherited the usage of this term from Leslie (1987), and recognized that metarepresentation was a basic mechanism underlying ToM.

In addition to these works, there were other pioneering studies conducted on theory of mind without using that exact terminology. Bretherton et al. (1981) cited Piaget’s (1954, 1962) observations of infants as a forerunner of their research. They also referred to Heider’s (1958) concept of humans as naive psychologists, although they thought that Heider’s idea was applicable to adults but not to infants. Obiols and Berrios (2009) pursued the origin of the concept of theory of mind in the work of James Mark Baldwin (1861–1934). They quoted passages from Baldwin (1895, 1898) which involved certain fundamental elements of ToM, including imitation, pretend play, and a sense of personality as opposed to things.

Simulation theory

In the mid-1980s, a view in opposition to ToM (in the narrow sense as described above) was presented. This view would subsequently be referred to as simulation theory (ST). Two articles published in 1986 – Gordon’s ‘Folk psychology as simulation’ and Heal’s ‘Replication and functionalism’ – are regarded to be the origins of ST (Goldman, 1989, 2006; Nichols and Stich, 2003; Perner and Kühberger, 2005). These articles commonly countered the view of previous studies on ToM that others’ minds were unobservable and could only be theoretically inferred. According to Gordon’s (1986) article, people predict another’s action by ‘putting themselves in the other’s shoes’ (p. 162) and asking ‘[w]hat would I do in that person’s situation?’ (p. 161, original italics). Heal (1986/2003: 14) also stated, ‘I place myself in what I take to be his initial state by imagining the world as it would appear from his point of view and I then deliberate, reason and reflect to see what decision emerges’.

The background or underlying component of ST was ‘Verstehen, roughly defined as “empathetic understanding”’ (Gordon, 1986: 162). In that statement, Gordon referred to Collingwood, Schütz and von Wright. ¹ Heal (1998a/2003: 29) also referred to the ‘Verstehen’ approach of Vico, Dilthey and Collingwood as the origins of her ‘simulation approach’.

In contrast to ST, ToM in the narrower sense was labelled theory-theory (TT). Around 1990, the terms simulation theory and theory-theory were commonly used by researchers. ² For example, Goldman (1989) used the terms ‘folk-theory theory’ and ‘simulation theory’. ³ Gopnik and Wellman (1992) also used the terms ‘theory theory’ and ‘simulation theory’. Thus, the ST/TT debate became a prominent issue that has been addressed in studies conducted on ToM (see also Zahavi, 2005).

ST and mirror neurons

It should be noted that ST was not necessarily a faithful reproduction of the German tradition of Verstehen. Gordon (1986: 170) stated that ‘[o]ne interesting possibility is that the readiness for practical simulation is a prepackaged “module” called upon automatically in the perception of other human beings’. Goldman (1989, 1992) also pursued the possibility that simulation of others’ mental states might be driven by a cognitive process, although he did not mention a specific conclusion at that time (1992). He argued:

[I]f one person simulates a sequence of mental states of another, they will wind up in the same (or isomorphic) final states as long as (A) they began in the same (or isomorphic) initial states, and (B) both sequences were driven by the same cognitive process or routine. . . . [S]uccessful simulation can be process-driven. (Goldman, 1989: 173, original italics)

The discovery of the mirror neurons was a suitable neural process that Goldman sought. The mirror neuron process was eventually presented by Gallese and colleagues in 1996. ⁴ They conducted single-cell recordings of neurons in the F5 brain area in macaque monkeys, and found neurons that ‘became active both when the monkey performed a given action and when it observed a similar action performed by the experimenter’ (Gallese et al., 1996: 593).

Two years later, Gallese and Goldman published ‘Mirror neurons and the simulation theory of mind-reading’ as an Opinion article in Trends in Cognitive Sciences ⁵ (Gallese and Goldman, 1998). They cited two types of studies as the basis of their argument: a study on macaques’ mirror neurons, and studies on the human brain using transcranial magnetic stimulation (TMS) or functional imaging such as positron emission tomography (PET). Gallese and Goldman (1998) cited the study by Fadiga and colleagues (1995) which demonstrated enhancement of TMS-evoked potentials in muscles by observing another’s action. According to Fadiga and colleagues the pattern of the enhancement of the evoked potentials corresponded to ‘the pattern of muscle activity recorded when the subjects executed the observed actions’ (p. 2608). ⁶ They concluded that humans have ‘a system matching action observation and execution’ (p. 2608). Gallese and Goldman (1998) also cited two studies utilizing PET (Grafton et al., 1996; Rizzolatti et al., 1996) which demonstrated that specific brain regions were activated during observation of action. From these experiments, Gallese and Goldman (1998) posited that humans also have a mirror system akin to that of macaques. They maintained that the supposed mirror system serves as a neural basis for ‘mind-reading’, and that simulation theory corresponds to this model but theory-theory does not. Thus, the explanation relying on the biological mechanisms of the brain was presented as the basis of simulation theory.

Division within TT

TT is not monolithic. The distinction between Theory-theory of mind (TToM) and Theory of mind module/mechanism (ToMM) has been noted and discussed by many researchers (Gallese and Goldman, 1998; Goldman, 2006; Gordon, 1999; Hutto, 2009; Langdon, 2005; Leudar and Costall, 2009a; Nichols and Stich, 2003; Zahavi, 2005).

Gopnik, for instance, is regarded as a major proponent of TToM. Gopnik and Wellman (1992: 145) claimed that ‘the child’s early understanding of mind is an implicit theory analogous to scientific theories, and changes in that understanding may be understood as theory changes’. They argued that a child observes others and then learns and revises theory of others’ minds, similar to the way in which scientists revise and update scientific theories according to the results of experiments. Their view has been called ‘child as little scientist view’ (Ravenscroft, 2010; see also Goldman, 2006: 13).

A representative of ToMM is Leslie, who investigated pretend play as a prerequisite for ToM (Leslie, 1987). In his example, when someone uses a banana as if it were a telephone, that person does not simply use the word ‘telephone’ to replace the meaning of banana. In order to use a banana as if it were a telephone, one must be able to ‘decouple’ a representation of the telephone from the original context and apply that representation in another context. Leslie (1987: 417) insisted that such use of representation is a representation of representation and thus called it ‘metarepresentation’. ⁷ He also argued that understanding another’s belief amounts to decoupling the content of that belief. He claimed that the acquisition of decoupling and ‘metarepresentation’ is a discontinuous developmental process and supposed ‘the specific innate basis’ for ToM (Leslie, 1987: 424).

Years later, Scholl and Leslie (1999) cited Fodor’s concept of modules as a basis of their argument. Fodor (1983: 9) originally focused on cognitive mechanisms, but not on propositional attitudes. According to him, the most important characteristic of a ‘cognitive module’ is ‘informational encapsulation’ from other cognitive systems (p. 37). Fodor also expected a cognitive module to be ‘domain specific, innately specified, hardwired, autonomous, and not assembled’ (p. 37). He hypothesized that the candidates of cognitive modules were the ‘input systems’ (p. 41), namely, ‘the perceptual systems plus language’ (p. 44, original italics). However, Fodor did not think that the whole of perceptual mechanism or language use was modular. He clearly distinguished information processing within the input systems and belief formation at the personal level. When a certain proposition is fixed as a belief, the linguistic expression must be compared with the world as perceived and also with the body of accumulated knowledge (p. 102). Therefore, there must be ‘central systems’ that are not encapsulated, that is to say, ‘nonmodular’ (p. 103).

Scholl and Leslie (1999) claimed that their model of ToM as an innate and domain-specific mechanism was consistent with the concept of module (see also Nichols and Stich, 2003: 117–18). They argued that at least infants’ ‘early ToM’ (Scholl and Leslie, 1999: 134) was modular, although they acknowledged that adults’ full-fledged ToM might not be completely modular.

Comparison between various standpoints on ToM

Up to this point, the present review apparently suggests four distinct standpoints: (1) ST as general reasoning, (2) ST founded on a specific mechanism, (3) TT as general reasoning, and (4) TT founded on a specific mechanism. However, these standpoints are not equivalent.

First, ToMM is not very different from TToM. Scholl and Leslie (1999: 147) wrote:

It [ToMM] interprets situations as involving intentional agents, who are represented as holding attitudes to the truth of propositions. In short, ToMM is the innate metarepresentational basis which imparts the essential character of ToM. It incorporates innate notions/concepts of propositional attitudes such as BELIEF and PRETENCE …. (original capitals)

They claimed to present ‘a specific innate basis’ of ToM (Scholl and Leslie, 1999: 134, original italics), but in effect they presented a model for the treatment of propositional attitudes (see Sharrock and Coulter, 2009). This reveals that their argument was at odds with Fodor’s original argument on modularity. Fodor (1983) excluded analysis of propositional attitudes from the investigation on modules, and he considered that the beliefs at the personal levels were not modular (see also Nichols and Stich, 2003: 121–2). As Sharrock and Coulter (2009: 73) precisely observed, ToMM is in effect ‘based upon … the modularity thesis, namely that our capacity for using a portion of our language (the so-called ‘mentalistic’ part) can be distinguished and studied independently from the rest of our linguistic capacities’ (original italics). ⁸

With regard to ST, Goldman’s ST and Heal’s standpoint were vastly different. Heal is an analytical philosopher, and the focus of her argument was not directed at the experimental studies on ToM. Her main aim was to criticize the philosophical view that ‘explanation of action or mental state … is causal’ (Heal, 1986/2003: 11), which she subsumed under functionalism. Heal’s Wittgensteinian standpoint is opposed to reductionistic thinking of functionalists. Her point was that no one has the perfect causal theory that fully grasps and predicts others’ beliefs or behaviours. Thus, according to Heal, people somehow replicate others’ minds in trying to understand them. In addition, Heal (1986/2003: 20) argued it as ‘truism’ that ‘we render the thought or behaviour of the other intelligible’.

Heal (1996a/2003, 1996b/2003) supported the ST side of the ST/TT debate, but only for a short time. Her approval of ST seems to have been partly due to her definitions of ST and TT. She defined TT as an ‘imperialistic’ standpoint that ‘[e]verything comes from theory’, but ST as a modest view that ‘[a]t least some important things do not come from theory but come from simulation’ (Heal, 1996b/2003: 67). However, it did not mean that she fully approved of ST. She did not think that simulation can vividly reproduce other’s experience of intentional attitudes (Heal, 1996a/2003) or that simulation alone can always faithfully reproduce what another thinks or believes (Heal, 1996b/2003; 1998a/2003: 32). In addition, Heal distinguished her ‘a priori claim about the relations of certain personal-level cognitive abilities’ from ‘an a posteriori hypothesis about the workings of sub-personal cognitive machinery’ (Heal, 1998b/2003: 91). She coined the new term ‘co-cognition’ for her standpoint (Heal, 1998b/2003), and withdrew from the ST/TT debate. Therefore, Heal’s standpoint was different from the ST of other researchers who sought empirical evidence for their views.

The above analysis indicates that the primary distinction at the level of empirical studies is focused on distinctions between ST and TT viewpoints. ST insists that others’ minds are understood by simulation, but at the same time researchers of ST endeavour to explain their arguments on a neural basis. In contrast, TT insists that humans infer others’ beliefs and actions theoretically. TT seems to be the view that does not distinguish understanding others’ minds and grasping non-human phenomena. However, TT presents the arguments as psychologically understandable.

This paradox is usually overlooked, apparently because researchers’ methodologies are not adequately scrutinized. As Obiols and Berrios (2009: 379) noted, studies on ToM have ‘two epistemological levels’: an observer has a certain belief about others’ minds, and a researcher investigates that observer’s belief about others. If the two levels are considered equivalent, this structure potentially engenders an infinite regress; if the distinctive authority (though not fully justified) is granted to the researcher, the possibility of that regress is concealed (Obiols and Berrios, 2009). ⁹

Hybrid views

The ST/TT debate does not seem to have ended in a victory for either side. Hutto (2009: 223) stated that ‘[a]fter years of marshalling philosophical considerations and evidence from psychology, few TT/ST purists remain’. Goldman, who was a major proponent of ST, proposed ‘an ST-TT hybrid’ (Goldman, 2006: 21). Prior to that, Nichols and Stich (2003), who had criticized Goldman, also supported a hybrid view. They wrote that the ST/TT debate ‘proved to be a very unfortunate way of characterizing the theoretical landscape, since it ignored the possibility that the correct account of mindreading might be provided by a hybrid theory’ (Nichols and Stich, 2003: 132, original italics).

Stich and Nichols (1992) were once highly critical of ST, but later switched to ‘an integrated hybrid theory’ (Nichols and Stich, 2003: 148). ¹⁰ They drew a hypothetical flow chart of the relationship between functional components as a model for the ToM mechanism, although they did not commit to the functional localization within the brain (Nichols and Stich, 2003: 11). First, they drew a model for the basic cognitive functions found in humans (p. 14), and subsequently added the ‘Possible World Box’ (p. 28), ¹¹ and some subsidiary mechanisms supposed to be characteristic of (though not exclusive to) ToM. Possible World Box is supposed to ‘represent what the world would be like given some set of assumptions that we may neither believe to be true nor want to be true’ (p. 28). We humans tend to attribute ‘one’s own beliefs to the target’, which Nichols and Stich called ‘default belief attribution’ (p. 66), but we overcome that tendency. Beliefs about another person’s perspective and that person’s statements are taken into account in understanding the mind of others. It is accomplished, as Nichols and Stich (2003: 93) argued, by ‘updating’ the representations in the Possible World Box: if another person seems to have a certain representation p, p is input to the Possible World Box and the representations inconsistent with p are removed.

Furthermore, Nichols and Stich (2003) claimed that their view was different from those of ST, TToM and ToMM. Their standpoint was not a pure TT view, because they claimed (p. 135) that their view was similar to that of ST in assuming that a person uses one’s own inference mechanism to understand and predict others’ mental states and behaviours. However, Nichols and Stich (pp. 135–40) argued that we cannot simulate another’s perceptual states or desires. In addition, default belief attribution cannot be explained by any of these theories. It is not mediated by either simulation or theory; thus, neither ST nor TT can adequately explain it (pp. 106–7, 140). With regard to ToMM, using one’s unspecific representations and beliefs is contrary to the concept of a module which is supposed to be encapsulated (pp. 120–1). Nichols and Stich (pp. 118–19) focused attention on Scholl and Leslie’s (1999: 147) statement that the selection of the specific content of another’s alleged belief need not be modular. These arguments revealed that ToMM cannot explain either practical reasoning or its attribution to others.

Goldman (2006: 21) also presented a hybrid view, though his version was one ‘with emphasis on the simulation component’. As an advocate for ST, he considered that one ‘makes special uses of her own mind in assigning mental states to others’ (p. 40). Goldman focused on ‘the act of assigning a state of one’s own to someone else’ and labelled it as projection (p. 40). He spelled out his view as ‘simulation-plus-projection’ (p. 40). In his account, simulation-plus-projection is almost equal to default attribution. Of course, simulation-plus-projection alone cannot appropriately reproduce another’s mental state. Goldman wrote that ‘a requirement for successful simulation is quarantining one’s own genuine states that don’t correspond to states of the target, that is, keeping such states from intruding into the simulation’ (p 41, original italics). He claimed that failure in quarantining leads to egocentrism which cannot be explained by TT.

Goldman (2006) acknowledged that ST cannot explain how people speculate others’ mental states from their behaviours, since simulation does not seem to function backwards. Goldman postulated a ‘generate-and-test strategy’ (p. 45, original italics): one hypothesizes another’s mental state and simulates whether that state leads to the observed behaviour. A problem with this strategy, as Goldman noted, is how to generate plausible hypotheses out of a great many possible mental states. Theory is necessary to generate such plausible hypotheses (p. 184).

Nichols and Stich’s Possible World Box and Goldman’s quarantining are similar in their role of overcoming egocentric bias. Goldman (2006: 181–2) argued that the Possible World Box is just like a pretend state within ST, although he did not directly compare the Possible World Box with quarantining. On the contrary, Nichols and Stich (2003: 135) did not say whether their Possible World Box had any counterpart within the ST framework.

Nichols and Stich (2003) and Goldman (2006) have different views of the role of simulation. This is largely due to their definitions of simulation. Nichols and Stich (2003: 132–4) listed 10 possible definitions for simulation and they limited the meaning of simulation to the use of one’s own inference mechanism or practical reasoning mechanism in understanding others. In their account, attribution of one’s own preferences, perceptual states and default beliefs are not included in simulation, since these are not brought about by an inference mechanism. According to Goldman (2006: 37), if one’s mental process imitates another’s mental process, then it is simulation. This definition leads to the view that default attribution is included in simulation.

Anti-‘ToMism’

Apart from ST and TT, there are yet other views on the understanding of others’ minds. Leudar and Costall (2009a) and Sharrock and Coulter (2009) labelled both ST and TT as ‘ToMism’ and criticized these views. Their definitions of ToMism were slightly different. Sharrock and Coulter (2009: 63) characterized ToMism as indirectness in grasping others’ minds. Additionally, Leudar and Costall (2009a: 4) noted that ToMism relied on inference in grasping others’ minds. Therefore, Sharrock and Coulter more explicitly included ST into ToMism.

The reasons for criticizing ToMism are diverse. Sharrock and Coulter’s account of others’ minds and language is based on Wittgenstein’s later philosophy (Wittgenstein, 1953/2001). Sharrock and Coulter (2009) address the problems common to ST and TT, and the specific problems with ST and TT. They counter the general assumption of ToM that ‘the minds of other individuals are not directly available to us’ (p. 58, original italics) as follows:

One can directly discern someone’s ‘state of mind’ on many occasions; as when one can see that someone is grief-stricken, depressed, distracted, … and only when some relevant contextual details are lacking and where observational details are unclear might one need to infer such ‘states’. (p. 66, original italics)

Mental states are not hidden things, but are expressed as utterances and behaviours. Behaviours are then described with language. Understanding others is not separate from understanding language.

Children learn language not by establishing a theory, but by using it with others within a certain cultural setting.

Instead of needing to construct a theory of mind, children learn to speak the language and to apply the criteria that are part of it, though learning the language is not some single, unified or independent task, but something that is done in and through learning to participate in a range of diverse practices that make up the lives of whatever collectivities we belong to. (Sharrock and Coulter, 2009: 78)

In other words, ‘there is no independent mentalistic vocabulary as such to be segregated from the language’ (p. 76). Sharrock and Coulter (2009: 76) remark that ‘the capacity to attribute beliefs, intentions, understandings and the like’ cannot work ‘purely on the basis of theories about putative inner states and without any embedding in a culturally accumulated mountain of understandings of the ways in which practices are organized’.

This remark applies to both children and adults alike. Adults’ continual use of language is also embedded within such a cultural background. Therefore, learning and using language is not an implementation of a single theory. Sharrock and Coulter (2009) argue against the TT’s assumption that people grasp others’ minds using theoretical terms. They criticize the concept of theoretical terms and note that they were outdated and that language cannot be reduced to some explicit theories. Since language is not constituted theoretically, Sharrock and Coulter argue that understanding others’ minds is not theoretical.

These authors are also critical of ST, although not completely refuting simulation. They write that ST ‘overstates their significance amongst the diversity of ways in which individuals are enabled to understand others’ (Sharrock and Coulter, 2009: 88, n. 19, original italics). Citing Gallagher (2007), Sharrock and Coulter (2009: 83) criticize ST as a circular argument that is ‘wholly dependent upon what we already understand about other people and what they do’. If one can say something about another person by asking ‘what would I feel in that person’s situation’, it entails already knowing about another’s mental state. Sharrock and Coulter criticize ST as ‘rather more a case of putting other people in our shoes’ (p. 84).

Leudar and Costall (2009a) emphasize that people understand each other by interacting in various ways. They state that ‘relationship with another person involves knowing her or him as an individual in that relationship’ (Costall and Leudar, 2007: 293, original italics). They argue that such knowledge is not just verbal or intellectual, but also involves other modalities like feelings and emotions (Costall and Leudar, 2007; Leudar and Costall, 2009a).

In later years, Costall and Leudar’s view is subsumed under ‘interactionism’ (Bohl and van den Bos, 2012). Interactionism is defined by Michael (2011: 559–60) as ‘the claim that social understanding and interaction do not require mindreading because various embodied and/or extended capabilities sustain social understanding and interaction in the absence of mindreading’ (see also Bohl and van den Bos, 2012: 3). By mindreading, Michael (2011) refers to both explicit (conscious) and implicit (unconscious) modes of ToM.

Proponents of interactionism acknowledge that it is diverse and not a unified movement (Bohl and van den Bos, 2012; Michael, 2011). Gallagher (2004), who coins a similar term ‘interaction theory’ (p. 199), addresses everyday human interactions. He stresses that such everyday interactions are second-person and ‘based on environmental and contextual factors’ (p. 202). He also emphasizes that ‘we have a direct, perception-based understanding of another person’s intentions because their intentions are explicitly expressed in their embodied actions’ (p. 205). Gallagher argues such understanding is based on innate ‘pretheoretical (nonconceptual) sensory-motor capabilities’ (p. 205).

There are other views subsumed under interactionism. Bohl and van den Bos (2012) list three characteristics of interactionism. First, ‘social interaction in many real-life pragmatic contexts’ is considered to be ‘more adequately characterized as participants making sense of the situation together’ (p. 3). Such interactions ‘cannot be fully explained by referring to processes evolving in the minds (or brains) of individuals’ (p. 3). Second, mental states such as emotions and intentions are considered to be directly perceivable (p. 4). Third, interactionism stresses ‘the importance of affect and engagement in interpersonal interactions’ (p. 4). Thus, there are various versions of interactionism since each researcher emphasizes these three features to varying degrees.

Furthermore, the relationship between interactionism and other versions of ToM is diverse among researchers. Michael (2011) observes that some researchers completely reject ordinary versions of ToM. However, as Michael (2011) acknowledges, interactionism cannot make a definitive argument against ST or TT regarding how person A understands person B’s emotions from observing B’s countenance. According to theory-theorists, A infers B’s emotion from B’s countenance using psychological knowledge. According to simulationists, A experiences B’s emotion by putting him/herself into B’s shoes. Interactionists cannot exclude the possibility that ‘sub-personal information processing’ within individuals (Michael, 2011: 564) may be working in understanding others. Therefore, other interactionists, including Michael, endeavoured to integrate interactionism with ToM; for instance, Bohl and van den Bos (2012) tried to reconcile interactionism with TT.

There are two ways to support the interactionists’ claim that participants understand each other in that relationship at the ‘supra-individual’ level. The one is simply declaring ‘we can understand each other automatically without need of theory, so there is nothing difficult’. The other is to find experimental evidence for the claim. In order to do this, interactionists tend to focus on the sub-individual level, at which participants are supposed to interact automatically. Therefore, interactionism tends to skip the personal level by directly connecting the social level and the sub-personal level. However, as Michael’s analysis above indicated, it is difficult to overcome theory-ladenness in interpreting experimental results on the ways of understanding others. ¹²

Sharrock and Coulter’s argument and interactionism are common in emphasizing the role of social interaction, but these views are different in other respects. They focus on the personal and verbal level of understanding within a social context. This is different from interactionism, but similar to Heal’s standpoint. The arguments of Sharrock and Coulter and of Heal have the following in common: (1) consisting of a Wittgensteinian stance, (2) arguing that understanding others’ minds is not an implementation of single theory, and (3) focusing on the personal and linguistic level of understanding others.

Heal’s view on the bottom line of understanding

What is the bottom line in understanding others, given the difficulty in interpreting empirical studies on ToM impartially? This question warrants a closer examination of Heal’s view on understanding others, which she claimed as truism or a priori.

In trying to understand another person, ‘we render the thought or behaviour of the other intelligible’ (Heal, 1986/2003: 20). That is to say, ‘we presuppose the rationality of others, presuppose that we share standards of what follows from what and what is relevant to what’ (1998a/2003: 44). Heal meant by rationality broadly ‘what is exercised in the formation of intention and desire as well as belief’ (1986/2003: 20). She did not require perfect obedience to the logical principles as the standard of rationality. Nor did she insist that predictions of others’ thoughts and behaviours are perfect.

Patently other people are often difficult to understand. Often we know that we are ignorant of their thoughts and feelings, or we have little confidence in our conjectures about what they may be. Simulationism is not the promise of some easy answer to these difficulties. (Heal, 1998a/2003: 32)

In attempting to recreate others’ mental contents, including their thoughts, one presupposes that others have cognitive capabilities and recognize norms. According to Heal (1998a/2003: 41), a human is ‘a rational subject with a point of view, having multifaceted abilities to think effectively about many subject matters and so forth’. She also claimed that her argument was ‘the natural theory of the understanding of other minds for someone who conceives of persons as unified rational subjects’ (p. 41) In summary, it is impossible to ‘do without’ positing minimal rationality ‘because it is presupposed by unavoidable kinds of common everyday thinking’ (1998b/2003: 102).

Heal’s concept of minimal rationality was influenced by Cherniak (1986), since she referred to him in other articles (Heal, 1998a/2003: 42; 2003: 233). Cherniak (1986) also stated that rationality is an indispensable requirement to be an agent, while the agent need not be perfectly rational. He defined ‘minimal general rationality’ as: ‘If A has a particular belief-desire set, A would undertake some, but not necessarily all, of those actions that are apparently appropriate’ (p. 9). Cherniak also acknowledged that an agent need not be perfectly consistent.

Such analysis of our inevitable presuppositions leads to the claim that Heal’s (1998b/2003) standpoint is an a priori one. In this context, what she meant by ‘a priori’ was not tautology in logic or mathematics, but what ‘is deeply embedded in our worldview’ and ‘we rely on unhesitatingly in making inferences’ (p. 94). Heal’s argument leads to the view that people grasp others’ thoughts and beliefs by ordinary principles of inference:

It is an a priori truth that thinking about others’ thoughts requires us, in usual and central cases, to think about the states of affairs which are the subject matter of those thoughts, namely to co-cognise with the person whose thoughts we seek to grasp. (p. 99)

Heal did not find a fundamental difference between conjecturing q from p on the one hand and asking oneself ‘What if p’ and obtaining q from ‘If p then q’ on the other hand (p. 113): both were subsumed under Heal’s co-cognition. Heal (2000/2003: 134–5) argued that both types of inferences presuppose normal cognition and rationality. When one person A understands another person B’s belief by an inference from p to q, A presupposes that B is rational enough to infer q from p (and A’s rationality is also presupposed). When A asks oneself ‘What if p?’ and understands B’s belief q, A presupposes that B and A are both rational to a similar degree. So rationality is presupposed in both cases (this is an abbreviated version of Heal’s more formal presentation).

Later, Heal proposed a ‘two-element conception of rationality’ (2003: 237) and distinguished the general and normative level of rationality from the actual inferences exemplified by each person at each time. The first is ‘the high level and general notion’ (p. 237) of rationality and norm common to almost all human beings. In contrast, the second need not to be the perfect judgement. The scope of the second level is not a strict one since each instance of inference varies from context to context. When people disagree and/or seek a better understanding of each other or of their common issues, they conform to the norm and try to find a better answer. The first level of rationality and norm ‘gradually dawns on us’ in the ‘forward-looking commitment’ to seek a better answer (p. 247). Heal argued that such ‘practice of discussing with others issues of better and worse in reasoning’ cannot be falsified by experimental studies (p. 248): experimental studies on ToM are also the instances of that practice.

Again, the reflection of Cherniak’s idea can be seen in this context. Cherniak (1986) also stated that normative rationality is stricter than descriptive rationality. He added that minimal normative rationality is weaker than ideal rationality and he further divided ideal rationality into several levels.

Heal (2003) introduced these two elements of rationality in order to avoid a dilemma between mysticism and eliminative materialism. If one supposes that each actual neural activity reifying each individual judgement is caused by certain norms, one has to postulate ‘a kind of mysterious, non-natural top-down causation by a particular package of norms’ (pp. 236–7). In contrast, if one rejects the normative level entirely, one has to be an eliminativist. If one distinguishes the normative level from each individual judgement, one can escape from these alternatives.

Heal’s two-element conception of rationality is different from other two-system views on ToM. Many researchers propose a distinction between a fast but fallible mechanism and a slow but flexible mechanism of understanding others’ minds (Adolphs, 2009; Apperly and Buterfill, 2009). However, researchers disagree about how to differentiate the two systems (Apperly and Butterfill, 2009). Both of the alleged systems would be subsumed under a second type of rationality in Heal’s scheme. She acknowledged that the experimental studies can ‘fill in the details of the patterns to be discerned in the candidate intentional states’ with regard to second-level instances of intentional actions (Heal, 2003: 246).

Heal’s argument on understanding was an a priori one, but she was not hostile to experimental studies on ToM. She thought that her a priori argument and experimental studies could be complementary to each other (1998b/2003: 114). She just wanted researchers to be aware of the implicit premises from which all of us cannot be free.

Lessons for clinical psychiatry

Heal’s argument on understanding others does not posit a fundamental difference between how people in general understand others and how researchers (including Heal) understand others. Therefore, certain lessons can be drawn from her argument on how clinicians understand patients.

Heal’s argument indicates that understanding at the personal level is not in conflict with but rather compatible with the progress seen in experimental studies in the neural sciences. This is good news to clinical psychiatry, since clinicians never cease to infer patients’ mental state at the personal level using language. A personal and verbal level of mutual understanding cannot be eliminated, no matter how advanced the experimental studies on subpersonal mechanisms are in inferring others’ mental states.

Such a personal level of understanding was appreciated in traditional psychiatry, though presently it tends to be discarded as unscientific. Methodology in psychiatry was presented by Jaspers (1962/1997) in a well-organized form. He presented both Verstehen and causal explanation as the foundations of psychopathology. Psychiatrists understand normal and neurotic aspects of patients. In contrast, psychotic experiences cannot be understood and they warrant a causal explanation. This is the golden mean between the psychological approach and the biological approach (Ghaemi, 2003). Jaspers (1962/1997) criticized the prejudices in appreciating only one approach (i.e. psychological or biological), although he acknowledged that some presuppositions are necessary to start any argument.

Jaspers’ concept of Verstehen is not without difficulty. It has both empathic and rational aspects (Kumazaki, 2013a, 2013b): rational understanding is indispensable for guaranteeing validity of understanding. ¹³ However, how can these two aspects be reconciled? In addition, what is meant by ‘rational’ also poses a problem. If the meaning of ‘rational’ is taken to be perfectly rational, irrational behaviours of patients cannot be understood by that method.

The Cherniak–Heal line of argument sheds light on these problems, although it is not a direct descendant of Jaspers’ methodology. ¹⁴ According to Heal, empathy is not in conflict with rationality; rather, rationality is a necessary precondition for empathic understanding. Moreover, the concept of minimal rationality helps in the understanding of a patient’s behaviour which may not be perfectly rational but nevertheless goal-oriented.

Conclusion

The present study reviewed the history of research on ToM. There are various views on ToM, and various methodological issues are involved in the relationship among these studies on ToM. ToM tends to be considered as a field of experimental studies, but experimental studies cannot be free from theory-laden observations. No one can discuss other people’s minds without any presuppositions. Therefore, it is crucial to be aware of inevitable presuppositions in order to avoid other prejudices (Jaspers, 1962/1997). Heal’s a priori account of understanding others illuminates such inevitable presuppositions; people replicate each other’s mental states, and people presuppose each other’s minimal rationality in doing this. This theory falls in line with traditional psychopathology.

In clinical practice, clinicians infer patients’ mental state at the personal level from communication involving language. In everyday life, ordinary people understand others on a personal level through communication involving language. Personal and verbal levels of mutual understanding cannot be eliminated, no matter how advanced experimental studies in neural sciences become. Studies on the personal level of mutual understanding have their own value that is neither secondary to neural sciences nor contrary to neural sciences.