Review - Stich and His Critics
by Dominic Murphy and Michael Bishop (Editors)
Wiley-Blackwell, 2009
Review by Robin Aldridge-Sutton
Sep 8th 2009 (Volume 13, Issue 37)

Stich and His Critics is a collection of ten papers exposing and criticizing the philosophy of Stephen Stich, along with his replies.  This format is conducive to understanding but the state of the art nature of the content means that it is not accessible for the philosophically uninitiated.  The main themes of this wide-ranging book are the role of folk psychology in science, how words refer to the world, epistemology, everyday mindreading, and innateness in cognitive science.  The following are outlines of the main arguments and replies.

Stich used to be an eliminativist about representational content.  That is he denied that people's mental states like beliefs and desires have subjects in the way that we ordinarily imagine.  In Is There a Role for Representational Content in Scientific Psychology Francis Egan notes that Stich is no longer an eliminativist but has never explained why.  She thinks he is right to have stopped because representational content is required in several areas of psychology.  Computational psychological modeling needs content to link abstract computationally defined states to their psychological context at the sub-personal level.   Attribution theory and developmental psychology use representational content to work at the personal level. 

Stich is not convinced that the content used in computational psychology is really like that of folk psychology, which is what is under discussion.  He says content does seem useful to cognitive social, developmental, and abnormal psychology.  But without a detailed philosophical account of the explanatory structure of these areas we cannot be completely sure.   Thus, given the vagueness of content, we cannot be sure that realism about it is justified.  He also notes that he no longer believes in the method by which he originally worked out the concept of content that he attributed to folk psychology.  The process of consulting the intuitions of philosophers to discover those of the folk has come under a lot of pressure recently.  There is evidence from surveys of folk, under the banner of 'experimental philosophy', that their intuitions are quite different from those of philosophers.   The discussion is thus premature. 

This is a nice illustration of the difficulties for philosophers in compromising with experimental philosophy.  Stich, himself an 'experimental philosopher', could have only said we don't know what content is like in folk psychology so we don't know if it's real.  But instead he both "plays the game" with Egan, in assuming that our intuitions are evidence of the folk's, and also declares that this assumption is unjustified.  It seems Stich is putting a buck each way in this case.

In Representationalism Reconsidered Peter Godfrey-Smith is interested in explaining how representation is being useful in cognitive science given that its physical underpinnings are not properly understood.   The kind of representation he is talking about is the basic sense in which one thing stands for something else.  His idea is that representation is often being used as a model.  The sense of model he is using is from some recent philosophy of science, allowing for flexibility in the ways and degrees to which theoretical posits resemble their objects.  He thinks that the way representation is discussed and employed indicates a model-based approach though this has not been explicitly acknowledged. 

PGS argues that one of the consequences of this view, presented in detail elsewhere, is that representation should only be employed in certain situations.  When the representation model is employed in an explanatory way it requires a real reader mechanism rather than just an abstract one.  There must be something that uses the representativeness.  Stich had criticized this by pointing out the common useful postulation of abstract mechanisms at an intermediate level of a functional analysis.  PGS says here that this can be a useful approach, but it should not be taken literally, as when explanation is intended. 

In his reply Stich agrees that a more complex theory is required to explain the use of representation in cognitive science but holds off on weighing in on PGS's account until it is presented in more detail.  He wonders 'aloud' whether this may explain how representation is being useful despite it's vagueness.  Readers may find PGS's discussion is speculatively interesting but that he hasn't really presented evidence that this is how representation is being used.

In On Determining What There Isn't Michael Devitt agrees with Stich that semantic assent is unhelpful in doing ontology.  We should not base our beliefs about what exists on beliefs about how words refer to the world.  But unlike Stich he thinks that we can discover the correct reference relation, by looking for the common and distinctive properties of paradigm cases of reference.  In response to Stich's view that it is unlikely that there is a theoretically interesting reference relation, Devitt points out the popular truth-referential theories of meaning in linguistics.  These theories say that meaning is determined by propositions being true if the parts of the world that are referred to have the properties referred to.  As the use of the meaning of mental states in social science is common meaning is theoretically useful.  Thus since meaning is defined by reference, reference is theoretically useful.

Stich replies that Devitt has convinced him that it is not clear and accepted that reference is not used in scientific explanation.  But he thinks his plan for discovering its nature is under described.  That is because some recent experimental philosophy has indicated that certain intuitions about reference vary across cultures.  These intuitions define significantly different reference relations so there will not be common and distinctive properties of a set of cases that doesn't distinguish between them.  But how to justify which side to take is not obvious.  He concludes by stating his alliance with deflationists about reference who think there is no objective reference relation. 

This highlights the fact that Devitt gives no reason to think that there is an objective reference relation, but proceeds by assuming we all know its paradigm cases.  On the other hand if there were a good case for an objective reference relation there is no obvious reason why there couldn't be more than one that was theoretically interesting.  Then perhaps in principle Devitt's plan could work.  Different cultures could have different reference relations.

In Eliminativism and the Theory of Reference Frank Jackson agrees with Stich that there is no objective reference relation but argues that there had better be one that can be standard for doing ontology.  Rather than offering a solution he offers a way to find one.  The right reference relations for the ontological context are those which explain how we report indefinitely many beliefs with sentences in a learnable, public language.  There will be some variation but also much agreement. 

Stich replies that this account cannot be right because Jackson's account does not determine the reference of thoughts, only spoken and written language, and ontology can be done in thought.  As Jackson's view of language defines meaning through the thoughts of the language-user it cannot then define the meaning of those thoughts through the language in which they are used.

In Why Isn't Stich an ElimiNativist? Fiona Cowie considers whether the concept of innateness should be eliminated from our discourse due to its being impossible to analyse.  Innateness is roughly the property of being naturally developing regardless of circumstances, though one main point of this paper is that it is impossible to define.  She first shows how a great many proposals for defining innateness all fail to correctly classify paradigm cases of innateness.  Yet research on innateness is flourishing in cognitive science.  To help us understand what is going on she describes two cases from the history of science where concepts that were apparently hopelessly confused went on to become crucial: elements and genes.  Elements in 1750 consisted of phlegma, spirit, oil, salt, and earth, and of course they were no use.  There are two perspectives on genes, functional and biological, and most of the time they have been hopelessly incompatible.  Yet following up on both has been crucial to scientific progress.  Her point is basically that we have to use the concepts we have until we have something better to put in their place. 

Stich is convinced by Cowie both that innateness is troublesome and that it is as yet irreplaceable.  But it seems as though they are mistaking the knowledge that is gained when a concept is shown to be mistaken for a reason to keep using the concept after the fact.  Clearly concepts should not be discarded until we are sure that they are useless.   But we should still only posit what best explains the available evidence.  Readers may feel that Cowie's paper shows that innateness does not do this.

In A Defence of the Use of Intuitions Ernest Sosa defends analytic epistemology, the process of using intuitions to discover the nature of knowledge.  His main project is to challenge the claim made by Joshua Alexander, Jonathan Weinberg, Shaun Nichols, and Stich that their recent surveys of cross-cultural epistemic intuitions show cultural variation that undermines analytic epistemology.  He points out that the response schema allowed only two responses and there could have been significantly more agreement if a third option was included.  He also claims that different cultures may import different background information to the scenarios employed in the surveys, or different ideas of 'knowledge'.  Stich has replied that if cultures had different ideas of knowledge that would still be problematic because we would not know which is the right or best one.  Sosa thinks this is really obscure because we can just value both. 

Stich claims that finding agreement between the cultural positions the surveys uncovered looks to be a difficult task.  He accepts that cultural epistemological diversity has not been shown beyond philosophical doubt.  But he argues that the stakes are very high for it, as the entire western tradition of philosophy is founded on the method of intuition, so more research should be done.   As for the option of valuing both ideas of knowledge he distinguishes his philosophical project from Sosa's.  He thinks Sosa is interested in finding the nature of his own culture's concept of knowledge, whereas he himself is interested in finding the best method of building knowledge, and the first will not take care of the second.  The question then becomes why does the fact that different cultures build knowledge in different ways change whether our way is the best?  Was the idea that everyone reasons the same way ever really good evidence that this is the best or only way?

In Reflections on Cognitive and Epistemic Diversity Michael Bishop agrees that Stich's original and recent attacks on analytic epistemology work but advocates a different positive epistemology.  Stich advocates pragmatism, adopting those methods of inquiry that maximize the probability of getting what you intrinsically value.  That is because he thinks that different cultures interpret belief brain states as different propositions, so whether beliefs are true will be culturally relative.  Thus true beliefs will not be intrinsically valued by most people.  Also there is evidence from medicine that unrealistically optimistic beliefs are beneficial, so true beliefs are not always instrumentally valuable either.  Bishop accepts these arguments but argues for strategic reliabilism on pragmatic grounds.  This is a complex view advocating truth-reliable methods of reasoning but unconscious avoidance of thinking about things where true beliefs are not pragmatic.  His basic argument is that we are attracted to believing the truth so his method will be much easier to take up than Stich's. 

In his reply Stich denies that strategic reliabilism has been shown to be pragmatically better than pragmatism since no empirical evidence has been given to support the claim that we are strongly attracted to the truth.  He asserts that Bishop's record suggests that if there were such evidence he would have included it.  Also what is required is that we would still be attracted to the truth even after understanding Stich's arguments against its value.  It is hard to understand the idea that people do not tend to believe what they think is true.  The same goes for the argument against the intrinsic value of truth.  It seems to say that holding a belief is holding to a brain state rather than a proposition.

In Simulation Theory and Cognitive Neuroscience Alvin Goldman defends a simulation theory of mindreading with neuroscientific evidence, from Stich and Nichols' claim that simulation theory is not a natural or theoretically interesting category.  The specific theory he is defending is that we discover what other people think and feel by our neurological states resembling theirs.  He presents evidence that neurological resemblance occurs in non-mindreading activities.  The same areas of people's brains are activated when they perform, imagine, and observe the same actions.  Then evidence that it also happens in various kinds of low-level mindreading activities (those concerning people's sensations rather than their beliefs or desires).  The same brain areas are active in feeling and observing disgust, pain, or touch.  Also people whose capacities for disgust, fear, or anger are removed cannot recognize the corresponding emotion in the faces of other people as well as control groups can. 

In response Stich says that this account of mindreading is not natural or interesting because it says that when someone fails to read someone else's mind correctly they have not engaged in simulation at all.  That is because simulation is neurological resemblance and they failed to achieve it.  Stich points out that, in his book, Goldman avoids this problem in the case of high-level mindreading.  He allows for simulation to fail if it is directed towards mindreading as a goal, whether consciously or otherwise.  This allows for mistaken simulation based mindreading where the goal of resemblance is there even if the resemblance is not.  But Stich notes there is no evidence that such unconscious aims exist but there is evidence of lots of unconscious mindreading.

Goldman could respond that unconscious aims have been posited in dual processing research, where aims that have been taken on consciously have been found to influence automatic processing preconsciously.  This gives some reason to think unconscious aims could figure into the mindreading picture, but it is far from evidence that they do.  It is surely theoretically interesting that when we observe the sensations or actions of others we use the same part of our brain as when we feel or imagine them ourselves.  What is lacking is any reason to think that this is related to mindreading. 

Stich and Nichols' account of mindreading has a good chance of being right according to Kim Sterelny in The Triumph of a Reasonable Man: Stich, Mindreading, and Nativism, and if it is it strengthens his arguments against mindreading being innate.  It says that quasi-perceptual mechanisms play an important role in desire-reading through reading of emotional expression and direction of attention, requiring less of innate belief and desire concepts.  It does not require that we represent other agents' planning processes, instead allowing us to use our own planner module with estimations of their goals.  It requires little innate information because it breaks down belief and desire understanding into learnable chunks. 

Stich replies that both the terms innate and concept are controversial and because Sterelny does not clarify his use of them (and in fact acknowledges their opacity) he does not make much headway on whether belief and desire concepts are innate.  As for the learnable chunks he thinks no one has said they were all learned at once, even Leslie, a nativist about mindreading and the only major mindreading figure quoted in Sterelny's article.  Sterelny cites a conference as evidence of the appeal of the main argument for nativism that he considers.  This may be a signal that the view he is opposing is more often espoused in conversation than in print.

In the anthropologically informative Against Moral Nativism Jesse Prinz says that morality is universal in being found in every known culture but we do not have mechanisms specifically evolved for it.  To show this he defines three degrees of moral universality that may be claimed for morality and proceeds to present empirical evidence against their showing moral nativism.  The most demanding degree of universality he considers is that specific moral norms are universal.  He considers the norms against harming the innocent, disobeying authorities, and incest, and shows that all are applied so differently that no specific variant is in fact universal.  Next he considers whether having moral norms about the same things is innate.  Against this he says that it's plausible that the three moral schemas he has considered were culturally rather than biologically evolved.  Further he shows that for each there is variation in whether they are in fact moralized, taking moralization to be the attachment of the emotions anger and guilt.  Finally he accepts that we universally form moral norms in general, but denies that we have innate mechanisms to do so given that neither anger or guilt are domain specific or hard to learn.   

Stich's main response is to question the sense of morality Prinz intends.  He separates morality as a natural kind of thing from morality as a folk theoretical thing and says that Prinz's claims do not make sense from either point of view.  If he intends to discuss a folk theory of morality he fails because his theory includes norms about food, burial, religion, and etiquette that evoke anger and guilt but do not seem moral.  If he intends a natural kind of thing then why is it better than the account of norms that Stich and Sripada set out that differs from folk theory in just the same ways.  Stich thinks this problem is typical of empirically minded moral psychology at the moment.  One way that this seems unfair is that Stich offers no evidence that the norms he claims seem not to be moral seem that way to most folk.  However it is certainly fair to say that Prinz does not make it clear which he means.

The tone of the book is almost universally one of mutual respect and generosity in interpretation, with the exchange between Rutgers colleagues Goldman and Stich the only one where tempers may have flared a little.  One anomaly is that the replies are all at the end, out of order, and not indexed so that if you want to read them after each chapter, as seems most natural, you have to leaf through to find them.  Maybe Stich intended the replies as a piece but it is not obvious that the order has any advantage.  The contributors are all experts in their fields and the contributions all well-informed and thought-provoking.  In the introduction Murphy and Bishop display editorial commitment in providing relevant background information to many of the issues.

The book as a whole may be less useful in being so wide-ranging, but it is connected in representing a strongly empirical approach.  Overall it is an excellent collection. 

© 2009 Robin Aldridge-Sutton

Robin Aldridge-Sutton is a philosophy MA student and tutor at Victoria University of Wellington.  His research focus is the role of intuitions in philosophy, particularly the free will debate.