December 3, 2002

Dear Dr Harnad,

Please find attached a resubmission of our paper ‘Constructing an Understanding of Mind,’ which we first submitted to you on June 7, 2001 and which you returned to us for revisions on April 8, 2002. We have taken some time to respond to the detailed analyses of the paper, in part because the referees set us some interesting challenges that required some thought and exploration and also because the field has moved on somewhat since our first submission and we needed to ensure that this version is as up-to-date as possible. Irrespective of the decision about the outcome of this submission we wish to thank you and the many referees for their insightful comments and requests. We feel that the reviews have forced us to be more clear, to address some issues which we had shied away from, and the tasks set us also provided some very concrete support for the model of cognitive development presented in the article.

Although keeping to the original structure we have made many changes. As a result we will spell out at length just how we have responded to the comments made by each referee. Please note that in order to carry out the tasks you set we have had to increase the word length, but please also note that the original was reported by at least one referee to be too brief in its development and almost all the points highlighted by the reviews refer to additions to the text, not cuts. We have responded to all the underlined comments, and in addition we have tried to take other points into account.

[1] SIGINFICANCE: We took this point very seriously and have taken pains to identify what our argument is and how our position relates to others’ – both friend and foe. Given that we have received a number of offers of commentaries from people who have heard about the paper but not reviewed it (e.g. Michael Siegal, Nancy Budwig and others), we have no concerns that the paper would fall on deaf ears.

[2] PRESENTATION: We are grateful for the pointers to the weaknesses here and hope that we have made the argument more clear. As you requested, sections 1 and 2 contain a more focused exposition of our case - see specific comments below.

[3] SCHOLARSHIP: The straw man issue is well taken and we have perhaps belaboured our attempts to identify the fact that straw people are constructed by our detractors much more readily than ourselves. We hope you now agree that we have inserted sufficient numbers of quotations from our detractors’ work for everyone to identify both who they are and why we disagree with them. In some cases [e.g. the work of Tomasello and Moore] we have contrasted the points upon which we agree and disagree.

[4] EVIDENCE: We have tried to do two things here. First we have inserted more examples of the research evidence with accompanying analyses of why our theoretical framework fits better. The reference section is much larger and more up-to-date as a result. Secondly, we have turned the final section into a ‘Conclusions and future directions’ statement, as we feel that this will enable the reader to see just how different are the implications of our perspective from the dominant positions we criticize.

[5] REASONING: We were very pleased that you asked for clarity on the philosophical basis of the paper and we have spelled out why we feel that a Wittgensteinian analysis is important. Although we completely take referee #10’s point that our analysis does not contribute to Wittgensteinian scholarship we feel that it continues to be important to raise the issues that emanate from such scholarship within developmental science and related disciplines.

[6] THEORY: We hope that we have clarified our use of terms. We have a slight concern that we might not placate one or two detractors about our treatment of "theory of mind" – we have to describe the various perspectives in order to describe the issues to the reader with no background in the area, but realize that there might not be enough here for the expert, particularly a devotee of the theory or simulation approaches. Please note that referee #1 felt that there was sufficient information on traditional accounts for us not to get too bogged down in the tired debates between the ‘traditional’ accounts. We have attempted to define concepts like ‘triadic interaction’ and ‘criteria’ more fully.

Referee #1

We agree that an attempt to reconcile the approaches is timely (however successful we have been!). As with many other reviewers we have been very happy to expand the examples that have been requested. We cite more of Meins’s articles and many other compatible studies/theoretical perspectives. It has been reassuring to find that this work is compatible with ours and we feel that our theory is able to incorporate all these theoretical accounts and interesting data.

As suggested by Referee #1, we have clarified what we mean by triadic interaction in footnote #5.

We agree entirely that there is a rich tradition of authors who use interaction as a basis of social understanding. Indeed we cited in our original submission most of the authors who were raised by this referee. We have added to these citations. However, we are convinced that an analysis that is based upon Piaget and Chapman is vital for an account that takes the child’s construction seriously, plus it is the case that Piaget spent much of his early career and part of his later works dwelling upon social phenomena. We discuss this in section 2.1.

This referee suggests that we over-rely on Chapman and Chandler, although referee #10 states that we should cite more of Chapman’s articles. We have addressed both criticisms referring to a broader range of Chapman’s writings, but also pointing out that Chapman’s work draws on rich theoretical traditions including Piaget, Vygotsky, Wittgenstein and others. In fact, responding to many of the issues raised by the referees has required citing several more articles by Chapman as well as some by Chandler.

We agree with referee #1 that there are many theorists whom it would be possible to draw on to reach an approach similar to ours. In our revised section 2.1 we outline the families of theories consistent with our approach. We do this in section 3.1 on infancy as well. Also, in section 4.2.1 we discuss attachment theory.

In expanding our review of current research we have cited the authors suggested by this referee, particularly Baldwin and Moses, Fonagy, Steele, Dunn and Meins, as well as others.

Referee #2

We take the point that many would disagree that a flaw exists in the contemporary work and hope that our analysis of the flaw is clearer now.

Referee #3

We have attempted to respond to the ‘straw person’ argument with a clearer account of what we are up against – the idea that the theory-theory and simulation accounts can remain insulated against the social data. In our revision, we acknowledge that essentially everyone realizes that social interaction is important but attempt to stress that to be agnostic about the role of social interaction in accounts of social development becomes increasingly difficult. In section 2.1 we outline some of the different ways in which developmental theories view the role of social interaction. In general it is possible to categorize developmental theories as falling into frameworks emphasizing either the individual, the social collective, or relations between individuals. In this article we propose an approach to social cognitive development based on a relational approach.

Referee #3 is right that there probably isn’t a social constructionist or enculturation account in the "theory of mind" literature. We clarify this in footnote #1. However, we note that several authors mention an enculturation position as a possibility to be considered (e.g., Astington & Olson, and Gopnik & Astington) and sometimes attribute this position to Bruner or Wittgenstein.

Referee #3 very insightfully recognizes that Wittgenstein’s arguments are also relevant for theories of infant development. This would be an additional argument against the analogical argument we discuss in section 3.1. There are also other arguments we could detail but to do so would result in making the section on infancy an article in itself. However, we hope now to have included better research examples to illustrate our case

In line with the suggestion from this referee and referees #4 and #10 we have added a discussion of future directions for research to our concluding section 5. In this section as well as sections 3.1, 3.2, 4.1 and 4.2 we have clarified how our approach differs from other approaches.

We would like to have had the space to develop our ideas on the issue of atypical development. However, we have had to restrict mention of this topic to a few points at the end of section 4.2.

The referee mentioned several relevant authors to consider in our revision: Fonagy; Hughes & Dunn; Baresi & Moore; Ruffman, Perner & Clements. We have discussed all these authors as well as others in our revision: Barresi & Moore, 1996 in section 3.1 on infancy; Ruffman, Perner & Clements, 1994, in section 5; Fonagy in section 4.1, and Hughes & Dunn, 1998 in section 1.

Referee #4

Referee #4 is not alone is being unaware of Piaget’s writings on social factors in development. In our revised section 2.1 we briefly view Piaget’s position and we also cite references for further depth on this point. We feel that this is important because a straw person of enculturation is often constructed by the theoreticians we are criticizing (we cite examples of this in the paper). Also, Piaget’s relational approach enables us to offer an alternative, particularly when combined with Chapman and Wittgenstein .

This referee suggests more specific, concrete examples. In our revision of section 3.2 we provide more examples of how children may learn criteria for mental state terms.

S/he suggests referring to dynamic systems theorists such as Thelen and Smith. We discuss such theories in footnote #3. We note that although there is much we agree with in such theories there are also differences. It is for this reason that we discuss this issue in a footnote rather than break the flow of the article.

On the issue of language and thought we have much more discussion on our position, primarily in section 3.2.

Referee #5

1. This referee and others asks us to clarify our differences with other theorists. We have done this throughout the paper but referee #5 focuses on infancy. On this topic we have revised section 3.1 and we now more clearly describe how we differ from the theorists mentioned by this referee as well as other theorists, e.g., Barresi & Moore; Gopnik; Meltzoff; Tomasello. (Footnote #6 deals with how we differ from Trevarthen.) We have responded to this major criticism by attempting to show that our account goes beyond the accounts s/he mentions, in particular by referring to the post infancy/secondary intersubjectivity period.

This referee states that we waver between Piagetian and Vygotskian accounts. Our account involves aspects of both; they are not mutually exclusive. We hope that it is now more clear that [following Chapman] we believe that there is a strong case to suggest that Piaget’s and Vygotsky’s accounts can [or indeed should] be merged.

The referee also states that we are unclear regarding the role of conversation. We have extensively revised section 3.2, and elsewhere to clarify this issue.

2. Regarding the mechanisms of social development we are now more explicit in presenting our views. We engage with the views of Astington throughout the paper (especially sections 2.1, 2.2, 3.2), with Dunn in sections 1, 4.2 and elsewhere, with Fonagy in section 4.1 and with Meins throughout the paper but especially sections 1, 3.2, 4.1 and 5.

3. In our revision we refer to Tomasello et al., 1993 and Barresi & Moore, 1996 and we describe our differences, but also how our position is similar in ways to theirs.

In regard to pretend play, we disagree that it is "the best candidate for a causal role in theory of mind development." However, we do agree that it is one likely context that could facilitate social development. We address this issue in footnote #9. We cite Lillard (1993a, 1993b) and Youngblade & Dunn (1995), but we believe that Meins et al. (1998) is not especially relevant for this topic. Although this is an important issue we discuss it in a footnote because it did not fit with the flow of our argument and we felt that we could not devote the space to a whole section on this topic.

4. This referee states that any theory of social development must account for atypically developing children and primates. Clearly we agree. However, we are now stretching the word limit and we have chosen to devote our attention to presenting our theory. We do mention the topics of atypically developing children and primates but only briefly in footnotes (footnotes #11 and #12). We regret having to deal with so many important topics in footnotes but in fact many of these topics are important enough to devote an entire article to and fitting them into the body of our paper would require restructuring the paper and changing the focus, or adding another section.

Referee #6

This referee states that our paper is not sufficiently original because published work already discusses the role of social interaction and gradualism. We hope that both the first and this version sufficiently acknowledge that others have directed our attention there. However, we feel that we have done enough to establish differences between ourselves and others. Plus we contend that the majority of accounts simply treat social interaction as an influence that has to be bolted on to their account.

This referee asks what kind of processing is involved in making inferences about minds. The two alternatives given are mental simulation or rules. In our revision (end of section 3.2) we are more clear that our answer is: Neither. Rather children understand talk about human action (e.g., words like, think, know, remember, happy, sad, etc. but also words like lost, found, trick, hide, look, see, tell, say, fair, not fair, not nice, good, etc.) in terms of the circumstances of their correct usage. That is the pattern of interaction for which we use these words. Reasoning or thinking about psychological matters (e.g., solving the false belief task) involves understanding the question in terms of the circumstances for the use of the words. We agree with the referee that we did not sufficiently argue our case before and have now done this by making more clear what our objections to other positions are, what our position is and how it is supported by the available data.

The new version can be read as a specific analysis of the nature of social cognitive development rather than a description of the social experiences in development. Indeed it could be read as a statement about the tensions in contemporary developmental theory, although we have avoided making such a claim for fear of stretching word limits and our referees’ credulity.

Referee #7

This referee (Chris Moore) asks "How the infant is able to move from participating in triadic interactions without an individualistic understanding of intentionality to having such an individualistic understanding"? Moore also phrases the question as how to "characterize the psychological processes that allow diversity in perspectives to be coordinated into higher forms of understanding." We have extensively revised section 3.1 on infancy, but we still acknowledge that we have not provided a complete answer to this very difficult and important problem. However, we have pointed out problems with other approaches and provided a firm foundation on which to build an adequate theory of infant social development, as well as the conceptual tools with which to build it (i.e., a description of the psychological processes drawing on ideas from Baldwin, Chapman, Piaget and Vygotsky). More fully describing such a theory would require a complete article in itself and that is not the focus of our current paper. Instead, we need the discussion of infancy, and especially the infant’s capacity to grasp others’ referential intent, in order to lead into our theory of the role of language in social development.

This referee notes that we rely on the "standard Tomasello/Baldwin line that understanding reference is foundational for acquiring language". We take it that something like this position necessarily follows from what we mean by language so our approach is clearly consistent with Tomasello and D. Baldwin in many ways. However, there is an important distinction that may not always be clear in other theorists. That is, we follow Chapman in emphasizing that infant’s ability to grasp others’ referential intent is not recognizing a correspondence between words and reality but understanding what others intend you to pay attention to in that situation. We believe that this position is consistent with Tomasello’s approach to language development, although it may not be explicit in his work. Dare Baldwin’s experimental work is excellent and very important in this area but we believe that the distinction drawn from Chapman is not explicit.

We thank Chris Moore for drawing our attention to some of his other articles that are relevant for our paper. Although we did know these articles we hadn’t cited them in our original manuscript. Addressing the reviews in our revision has made these articles by Moore quite relevant and we now discuss them in section 3.1, to identify areas of overlap and difference in order to respond to his claim that the paper "does not provide much further insight".

Referee #8

This referee states that there is "nothing new or radical about the proposition that social cognition develops through social interaction". We agree but there are still very large differences in the ways that different theories account for the role of social interaction in development.

This referee notes that we do not mention Rogoff. There are many theories that have something in common with our approach but also differences. In our original manuscript we had chosen not to devote a lot of space to reviewing different theories and instead to focus on presenting our theory. In section 2.1 of our revision we now review more approaches, but this is still far from exhaustive. In our revision we now cite Rogoff, Cole and others. We also describe in footnote #2 how we differ from Rogoff’s participation theory.

The article by Sabbagh and Callanan (1998) is relevant and we discuss it in our revision in sections 1 and 3.2. We also agree that Lillard’s (1998) discussion of cultural differences is relevant and we discuss her work in our revised manuscript in section 2.1, 3.2, and 5.

To make a general point here, in response to the many suggestions from the referees to review more theoretical perspectives and more empirical research we have added over 100 new references. In the interest of space, we have also deleted some references from the original manuscript that were not absolutely necessary.

We have edited the manuscript extensively and we hope that this referee will now agree that the writing is tighter and more concise, although we note that Referee #9 liked the writing style of the original manuscript. In doing so we have both articulated our ideas in more detail and taken them further. In fact, we acknowledge that this referee has a good point that we did not take our ideas far enough in the original manuscript. We thank the referees for the encouragement and the difficult questions to address that made us develop our ideas more thoroughly.

We take the points concerning the quotation taken from Raver and Leadbeater that a social interactionist account does not rule out a focus on the individual as a unit of analysis and that it still allows us to make a distinction between the nature and origins of competence. We hope that we have made it clear that we agree with the first point and would respond to the second by agreeing since it acknowledges the place of a social constructivist analysis.

In section 1 of our revision we clarify how we use the phrase "theory of mind."

This referee thought that we were suggesting a conflict between children’s understanding of emotion and false belief. In fact, we were talking about the difference between children’s performance on tests compared to naturalistic observation. We have tried to clarify this point and we discuss it in section 4.1. The analysis of the epistemic triangle and its implications [e.g. gradualism] will hopefully provide an answer to the question posed to us about the nature of the conflict and acknowledge this referee’s agreement with us about the possibility of inconsistencies in development. We also agree that there is a problem in moving from individual differences data into a developmental account. Again we have not addressed this issue in the paper in as much detail as it would need [it would require a whole paper], but hope to have provided sufficient concordant experimental data to make our case for a developmental account which also takes into account the individual differences data.

This referee suggests that we refer to Linda Smith’s dynamic view of concepts. We do not think that this is appropriate for the concepts we are concerned with, although we acknowledge that there are interesting parallels to draw upon. We have responded by referring to Smith and Thelen’s dynamic systems theory in footnote #3.

Referee #9

Vasu Reddy makes two criticisms of our original paper. We have responded to the first by devoting section 3.1 to a critical analysis of the "theory of mind" analyses of infancy. In many ways it has been difficult to accommodate the depth of her point here, simply because our view has so much in common with her own – we suspect that she would prefer us to have addressed areas of overlap and potential difference between us when our agenda was to attack a position which is antithetical to all of us. We have now made clearer [albeit brief] reference to the non-verbal interactions in infancy that are the starting point for the establishment of criteria. However, we would hope and expect that Dr Reddy would write an interesting commentary on the paper if it were published.

We are clearer in our revision (section 3.2) that learning to talk about the psychological world is built onto prior prelinguistic shared practices or everyday activities or joint routines. It is this form of activity that Dr. Reddy refers to as narrative in non-verbal interactions. We agree that this is an area in which our theory could be elaborated and applied in exciting research on the types of shared activities that result in the development of joint attention.

Dr. Reddy suggests that we imply that social understanding is late in appearing and language-based. Our position is that we can describe two forms of social understanding, although there can be no sharp demarcation between them. During infancy infants gradually develop a practical or lived form of social understanding or knowledge of other people. This can be thought of as expectations or the beginning ability to anticipate what people may do (sections 3.1 and 4.2.1). And this early understanding or skill begins to develop during the first year, as Reddy suggests, becoming more obvious with joint attention. Adding words to such interaction rituals allow children to begin to develop a reflective understanding of social situations. We think that we are fundamentally in agreement with Reddy and that there are ways in which Reddy’s suggestions can extend the position we are arguing for.

Reddy’s second point is also a valid criticism of the previous version of the paper. We hope that the closer attention to the detail of Chapman’s position and the application of the epistemic triangle as a unit of analysis will answer the claims that we were not specific enough about the social nature of the child’s construction of knowledge and mind-behaviour dualism. We have attempted to spell out why the theory metaphor won’t work in our analysis of the nature of mental representations.

Referee #10

We have followed this referee’s suggestion to do something more generic at the beginning of the paper regarding the debate between individual and social constructivism. In our revision we do this in section 2.1.

In response to this referee and the others we have extensively revised our paper so that we think our position is far more explicit.

p2-3ff: In revising the paper we have been more careful in not implying causality from correlations.

p3 We have responded to the comment that we draw largely upon experimental work, by suggesting in section 5 that we need a richer set of data, as well as acknowledging more throughout that we draw upon a wider set of studies that the, largely experimental, theory of mind data. We have avoided the temptation to get immersed in a discussion about comparisons with the neo-Vygotskian literature simply because [1] this literature does not really focus upon issues which overlap with this domain of research; [2] the social interaction and mental states literature covered here is large enough [and growing] to merit the focus of this paper; [3] a complete comparison would have taken up many journal pages, with perhaps little addition to our thesis.

p4 We agree that Piaget’s sociological ideas merit explanation and clarification. In section 2.1 we have elaborated on his ideas about relationism, to avoid the charge of lacunary by-passing. However, we do not draw on his model of social exchange and this article is not the place to present that model. One of us (JC) is editing a book, in which most of the authors of most of the chapters explain and critique Piaget’s Sociological Studies from a number of theoretical perspectives. We know Chapman’s (1986) article on Piaget’s sociological studies, and Mays (2000) critique (one of us, JC, has an article in that special issue of New Ideas in Psychology). Yes, Piaget did have a model of subject-subject interaction. Chapman’s (1991) point was that although Piaget worked on both subject-object interaction and subject-subject interaction, he did not integrate them as a single interacting system. This is the point of the epistemic triangle.

p4 ‘internalize’ and ‘participate’: Referee #10 points out that some sociocultural theorists (Matusov, Rogoff) criticize internalization and argue instead for "transformation of participation." However, they critique a particular view of the nature of internalization. We agree with their critique but that is not the view of internalization we endorse, i.e., we do not reject internalization, only one view of internalization. In footnote #2 we discuss Rogoff and Matusov’s views and how we differ from them. Our rejection of introspection is a particular form of introspection, that is, the idea that we could learn the meaning of psychological terms through introspection. As we discuss in section 3.2, introspection is possible once children have learned the meanings of mental state terms. However, this is not introspection on beliefs and intentions etc. as mental objects.

We have found it challenging to respond to some of the suggestions, particularly from referee #10, that are of a more general nature. Although we fully agree that these issues arise and are relevant, and we have published on these topics elsewhere, the challenge has been that these are large issues but we do not want to change or dilute the topic of the paper. The challenge has been to say something brief to avoid disrupting the flow of the paper, yet not to be too dangerously synoptic. It is dangerous to say something too brief because our comments may be misunderstood. One of the ways we have dealt with this problem is to use footnotes, although we recognize that this is less than ideal and that important issues are addressed in footnotes we have done so in order to maintain the flow of the paper.

p5 In our revision we spend more time presenting Chapman’s account, both from this 1991 chapter and his 1999 article. We hope that this will provide the "full Chapman exegesis" that this referee has asked for. This referee also suggests drawing on other papers by Chapman from the late 1980s. (Although referee #1 suggested that we over rely on Chapman.) Responding to the many reviews has, in fact, required drawing on some of Chapman’s other papers.

We believe that our use of Chapman’s ideas is not a simplified version. If our version is augmented it is because it applies the epistemic triangle to a new area. Indeed the application to children’s reasoning about social matters rather than concrete operational reasoning requires us more thoroughly to integrate Chapman’s analysis of Piagetian ideas with Wittgensteinian ideas.

Regarding what difference the use of the ideas we present in this article would make for research, we argue that our framework provided a way of integrating and making sense of a great deal of current research. It also leads to a new research agenda. Several of the referees asked that we provide suggestions for future research based on our theory. We do this in section 5.

p7 This referee refers to the following statement in the original manuscript and asks is it "must" or just "likely":

We argue that learning the meaning of metal state terms must be social and cannot be a matter or individual introspection because there are no inner objects such as beliefs or intentions that can be introspected upon. We have extensively revised section 3.2 and we believe it is clearer.

This referee refers to Simon (2001). There is much that we would disagree with Simon about concerning his view of the mind as a physical symbol system, but his claim that the referee points out that "learning takes place inside the learner and only inside the learner" (Simon, 2001, p. 210), seems, at least in a sense, to be uncontroversial. In one way this has to be true, but where disagreements arise concerns the nature of development of knowledge. We argue that it is not possible to learn the meaning of psychological words through an individual process of introspection because there are no such inner objects as beliefs, desires and intentions. Beliefs, etc. are not separable from activity. Learning the use of such psychological terms depends on a shared practice, an understanding of everyday activities.

p7 We are more clear about the three dominant accounts.

p8 Knowledge is based on action. We have attempted to clarify that this assumption that knowledge originates in activity is the basic assumption for constructivism. We have not taken up the bait to analyze the differences between cognitive behaviourism or social activity theory, although we were tempted!

p10 We have changed undifferentiation to non-differentiation.

p12 We have put quotation marks around primitive.

p12 Circumstances-criteria of psychological words: We agree with Putnam’s (1973) argument that ""meanings" just ain’t in the head!". That is, "the extension of a term is not fixed by a concept that the individual speaker has in his head." The referee mentions one of Putnam’s arguments, i.e., the social division of labour, and Putnam also argues that meaning is partly determined indexically.

Putnam’s position that meaning is not attached to mental representations in the head is similar to/derives from Wittgensteinian arguments that meaning cannot be attached to representations (Goldberg, 1991). The referee suggests that this leads to relativism. (In our revision we now deal explicitly with relativism in section 2.1.) This is the classic dichotomy; if meaning is not from the individual then it must be from the collective and that implies relativism. We reject both ‘alternatives’. Instead, we argue that people use symbols to convey meaning to others, and the meaning will depend on the location in the ongoing sequence of interaction. That is, symbols are triadic in structure—they are used to direct another person’s attention to specific aspects of the world—rather than dyadic, i.e., connected to aspects of the world.

Regarding circumstances and criteria, we believe that there may be a misunderstanding here. The referee states that some concepts can be defined independently of circumstances, e.g., arithmetical concepts. However, the criteria for such concepts are based on the activities, e.g., of joining together and taking apart. These are the circumstances in which children learn the criteria for the use of words such as addition and subtraction.

This referee’s third point is that "the motivation for Wittgenstein’s PL is to eliminate a major source of subjectivity." Here we are not sure that we understand the referee and if we do then we disagree. In any case, we have extensively revised our paper with regard to Wittgenstein and we believe that is it now much clearer (section 3.2). Wittgenstein’s private language argument, from our perspective, is just one aspect of his more general rejection of a causal psychological view of the mind.

The referee next raises the issue of reference. This is a complex and important issue. (In fact, the paper could have been structured around this issue.) We have not surveyed views of reference (except to contrast our position with a more common view of reference as a dyadic relation between a sign and a thing referred to). Our view of reference is not represented in the list provided by this referee but rather by Chapman (1999), Putnam (1988) and Sinha (1999). What we have done is to be clear about our position on reference. We do this in yet another footnote (footnote #7). We also more clearly write about grasping others’ referential intent, rather than the development of reference, to be clear that we are concerned with shared meaning. But meaning is not fixed by signs having a dyadic connection to reality, thus ensuring their meaning. Rather signs are used to direct others’ attention and shared meaning is achieved in the context of ongoing social interaction. Our view follows from the epistemic triangle and we believe it is more explicit in our revision. We use the following quotation from Chapman in section 5 of our revision: "The child’s understanding is not just a matter of recognizing a correspondence between the mother’s words and reality, but of grasping her referential intent in that situation – knowing what one is meant to attend to in response to her words and gestures" (Chapman, 1999, p 34, emphasis original).

p12 Our intention was not to contribute to Wittgenstein exegesis but rather to use Wittgensteinian arguments that are highly relevant for the topic of the paper, but still relatively unknown in developmental psychology. Many assume that Wittgenstein did not propose a theory, but in his critique of views of the mind he does provide a foundation on which to build a theory. However in the process of applying these arguments and elaborating a positive account of development we may have contributed to a better understanding of Wittgenstein’s views and their implications.

This referee is of course completely correct that our position relies on some interpretations of Wittgenstein and in the original paper we did not mention others. (But note that the original paper was right at the word limit. In responding to all the reviews we are now beyond it.) In our revision we follow this referee’s suggestion to discuss Kripke’s interpretation of Wittgenstein. Although this is a very important point discussing it in the text would break the flow of our argument so we discuss it in a footnote (footnote #4). We disagree with Kripke (1982) and we cite McDowell (1984), Russell (1987) and ter Hark (1990) in showing how Kripke went wrong in his understanding of Wittgenstein.

The referee’s third point in this paragraph concerns whether Wittgenstein’s PL argument is a denial of the facts. It might appear that Wittgenstein is denying mental states, and therefore he must be a behaviorist. If Wittgenstein is denying that a sensation is a mental object it may appear that he is saying there is no sensation, in which case the child’s understanding of mind amounts to no more than learning to talk about mental states through enculturation. But that is not what he intended. "And now it looks as if we had denied mental processes. And naturally we don’t want to deny them" (Wittgenstein, PI, para. 308). "To deny the mental process would mean to deny the remembering; to deny that anyone ever remembers anything" (PI para. 306). What Wittgenstein did reject was labeling mental processes as private objects in the same way we label physical objects. His denial of pain as an inner object has to do with the grammar of sensation words. It is an argument about how our language makes us think about psychological processes not a denial of psychological processes (see section 2.2).

With respect to introspection, Wittgenstein was arguing against the idea that introspection could be the way that we learn the meaning of psychological terms. Here again we are clear in our revision that Wittgenstein’s arguments are not against all forms of introspection but rather against the possibility that introspection could be the way we learn the meanings of terms for inner processes. He does not rule out introspection once someone has learned how to talk about the psychological world. Wittgenstein writes: "It makes sense to ask: ‘Do I really love her, or am I only pretending to myself?’ and the process of introspection is the calling up of memories; of imagined possible situations, and of the feelings that one would have if…" (PI, 587). The Ferrari et al. (2001) article on consciousness mentioned by this referee is not vulnerable to Wittgenstein’s private language arguments.

This referee’s fourth point concerns internalization for a constructivist perspective. In our revision our main task was to clarify our argument and this is a central aspect of it. Wittgenstein’s arguments about how children learn to talk about sensations also applies to talk about the psychological world. Such talk is build upon the child’s ability to engage in shared practices or everyday activities, i.e., knowledge-in-action, or practical, lived knowledge. Children learn to talk about people’s activities in psychological terms and can then think about the psychological world. This argument is elaborated throughout the paper but primarily in section 3.2.

p13. In our revision we believe that we have been much more explicit about our position on gradual development and critical of magic bullet accounts.

p15 We have defined SAT in section 4.2.1 ("the Separation Anxiety Test, a projective measure of attachment security").

p17 We do not address this point directly in the paper but do so indirectly in footnote #8.

Referee #11

1. We have hopefully fixed the stylistic problems this referee has pointed out and we have gone carefully through the paper to avoid other similar problems.

2. Astington & Gopnik’s view of Wittgenstein is relativistic. The examples of dressing properly and eating politely show that these forms are specific to particular cultures and would be different in other cultures. This implies that social understanding would similarly vary across cultures. "Cultural practices" "make themselves the case." This is one common view of Wittgenstein that could be derived from some things he wrote, but we don’t think that this view is consistent with many other things he wrote and his general approach. Language games may vary somewhat cross-culturally, but they are not free to vary without limit, which is implied in the A & G quote. Wittgenstein includes aspects of the world, like color samples, in language games. Also, many of the forms of interaction that we are concerned about here, on which mental state talk is based, involve forms of social interaction that must be universal, e.g., sharing attention, caring for babies, etc. That is, language is built onto simple practices or everyday activities that are rooted in basic human forms of life and therefore will not vary across cultures. Some of these practices are constrained by the form of our bodies and the nature of the environment. We discuss this in sections 2.1 and 3.2.

3. This referee makes many good points in these three paragraphs. Some of the points are distributed across the three paragraphs. One point concerns how the infant develops the capacity to share attention with others. Several other referees also point out this weakness in our original manuscript. In defense, we acknowledge that this is one of the most difficult and important problems in developmental psychology. We had not intended to provide a complete theory to account for this development; this would be an ambitious article in itself. Rather, we meant to describe the development of the capacity for joint attention as a foundation on which we built our view of the role of language in social development. In our revision we have gone much further in the section on infancy (3.1) but we acknowledge that it is still not a complete theory. However, we have pointed out problems with other approaches, described the starting point in development from a constructivist perspective, and suggested how to develop a constructivist theory of infant social development.

Section 3.2: How are mental state terms learned? How do we steer a middle course through individual and collective approaches? Children learn words to talk about the activities of people, this involves intention (what someone is trying to do, etc.) and why (involving beliefs desires, etc.). Is this just a passive collective approach in which children a cultural way of talking? No, in two senses. First, such words are built onto shared practices. Children develop competence or skill in these shared practices. For example, in situations of joint attention, children have developed competence in this form of triadic interaction through first differentiating their directedness toward objects from others (through experience of resistance from others, that is others don’t always do what the infants wants or expects), and second coordinating their attention with others’ attention. This involves a development from situations in which the joint engagement is supported by adults and the environment to situations in which the infant has become more flexible in achieving that joint attention. This is development in social understanding but it is still practical and unreflective. That is, with this practical knowledge infants can share attention in the moment with others, but cannot reflect on it and think about it. They cannot use their imagination and reason about possibilities. This requires a way of representing situations that are not present. Language serves this function. Words refer to patterns of interaction. Children’s understanding of such words is in terms of their understanding of such patterns of human interaction.

But, still, how do children learn the words? Is it passive? In joint attention when children are involving in pointing as a declarative adults use the word ‘look’. Children learn the circumstances in which it is appropriate to use this word. It is grafted onto their ability to direct adults’ attention with a pointing gesture. But does the child have a full adult understanding of the meaning of the word ‘look’ at this point (i.e., the range of ways it is used in talk to convey meanings)? No. The child starts off with a partial understanding or subset of criteria and there is a protracted development (Nelson). This involves the epistemic triangle because children learn more about situations involving joint attention through other people’s reactions, and when they can communication with language this communicative interaction allows them realize their partial and incomplete understanding of how the words ‘look’ or ‘see’ can be used.

Referee #11 asks if an argument for a constructive process of development really counters cultural relativism. There is something to cultural relativism and enculturation, but this is because these are parts of a relational view. That is, if development occurs in relations between individuals, then culture comes into play. We recognize that there could be cultural differences, for example, in some forms of complex emotions. However, a focus on enculturation gives the impression that these concepts about the social world are free to vary without limit. In contrast, at the level of basic social understanding in infancy and young childhood that we are concerned with we would expect very little if any cultural variation. Infants are directed toward the world and have desires and begin having clear intentions that are obvious to their parents. And their parents begin talking about the infant’s desires and intentions etc. Infants need caring for, they have an extended period of dependency, and so interaction with adults to do with feeding etc. must be a cross cultural universal. Granted, cultures may have different ways of talking about intention, but it would seem that some way of talking about intention would be universal since this seems to be a common factor in human forms of life (Canfield, 1993; also Lillard 1998).

The difference between our view and Astington & Gopnik’s interpretation of Wittgenstein is that ways to eat politely and dress properly vary across cultures, whereas having likes and dislikes and attempting to do things in the world are part of our human and animal nature (emotional reactions, the way our senses work etc.) and we are talking about how the child learns to talk about his or her own desires and intentions, etc. and also how to talk about others and in the past and future. These aspects of human activity are not completely free to vary without limit across cultures, although clearly there can be some variation. The point is that for Wittgenstein language is an activity; it is tied into the world, and it is not free floating, which is the impression we get from some interpretations of Wittgenstein.

4. Are children learning concepts first and then the meaning of words or vice versa? We have clarified our position on this in section 3.2. The default assumption might that children learn concepts first and then words. We acknowledge that there is a Piagetian view of sensorimotor concepts that are based on action schemes and combinations of such schemes. These are practical concepts, i.e., what the infant can do with things, e.g., push a ball, or suck it etc. This can also apply at the social level with people in joint attention situations, e.g., give and take, direct attention and follow gaze. These could be thought of as practical or sensorimotor concepts. In answer to this question, however, the referee is right that we would say both. In section 3.2 we use a quote from Wittgenstein where he asks if children learn the concepts of multiplication before of after they learn the words. In this example it is fairly clear that they learn about both at the same time. The concepts are the operations that are criterial for the correct use of the word. In the case of complex talk about the psychological world development is more protracted and so children may learn something about the concept and how the word is used but through more interaction with others using the word they will learn more about other criteria and other uses, and so learn more about the concept. The meaning of words is not simply given ready made by the teaching of adults; it must be reconstructed.

We hope that we have done enough to convince these referees of the validity of our case. We look forward to the next stage in the review process. We also thank you for inviting us to submit this version for Psycoloquy. We have mixed feelings only because if the paper is rejected by BBS we would of course attempt to resubmit it elsewhere. We could be persuaded on this matter but at the moment would prefer to keep our powder dry.

Yours sincerely,

Jeremy Carpendale and Charlie Lewis