PhilosophicalPsychology Vol.25,No.2,April2012,187–202 Mind-brain correlations, identity, and neuroscience Brandon N. Towl One of the positive arguments for the type-identity theory of mental states is an inference-to-the-best-explanation (IBE) argument, which purports to show that type- identity theory is likely true since it is the best explanation for the correlations between mentalstatesandbrainstatesthatwefindintheneurosciences.Butgiventhemethodsof neuroscience,thereareotherrelationsbesidesidentitythatcanexplainsuchcorrelations. I illustrate some of these relations by examining the literature on the function of the hypothalamusanditscorrelationwithsensationsofthirst.Giventhattherearerelations besidesidentitythatcanexplainsuchcorrelations,thetype-identitytheoristisleftwitha dilemma: either the correlations we consider are weak, in which case we do not have an IBE to an identity claim, or else the correlations we look at are maximally strong, in which case there are too few cases for the inductive part of the strategy to work. Keywords: Hypothalamus; Identity Theory; Inference to the Best Explanation; Kind Splitting; Mind-Brain Correlations; Psycho-Physical Correlations; Thirst 1. Introduction The identity theory, the theory that mental states are type-identical to brain states, hasbecome popularagain (for examples, seeKim, 1993,1998; Mandik, unpublished manuscript; Polger, 2005; Shapiro, 2004). Though many authors have written defensesofthetheory,thepositiveargumentsforidentitytheoryhavenotreceivedas much scrutiny. Here I would like to examine the argument that starts with (what some call) the psycho-physical correlation (PPC) thesis. The argument tries to establishthatthePPCthesisprovidesgoodevidencefortype-identitytheorybecause type-identitygivesusthebestexplanationofwhythePPCthesisholds.Thuswehave an inference-to-the-best-explanation (IBE) argument for the type-identity theory, given the PPC thesis. This strategy can be found explicitly in the earlier works of Christopher Hill (1991, 1984) as well as more recent work by Brian McLaughlin (1997, 2001; Hill & McLaughlin, 1999).1 BrandonN.TowlisalectureratWashingtonUniversityinSt.Louis. Correspondenceto:BrandonN.Towl,WashingtonUniversity,St.Louis,MO,USA.Email:[email protected] ISSN0951-5089(print)/ISSN1465-394X(online)/12/020187-16(cid:2)2012Taylor&Francis http://dx.doi.org/10.1080/09515089.2011.569920 188 B. N. Towl Aftersettingouttheargument, Ishallarguethatitfacesadilemma thaterodesits initial appeal as justification for type-identity theories. Specifically, correlations betweenmentalstatesandbrainstatesareeitherrelativelyweakorstrong.Ontheone hand,weakcorrelationscouldbeexplainedbyidentifyingamentalstatewithabrain state—but there are a number of other relations that can explain such correlations as well. On the other hand, if the identity theory is meant to explain only strong correlations, it turns out that there are very few of these—too few to justify the induction thatallmental states,as wecurrently describethem,areinfact brainstates. 2. The Psycho-Physical Correlation Thesis and IBE The psycho-physical correlation thesis is usually stated thus: Psycho-physicalcorrelationthesis(PPC):foreverytypeofmentalstate,M,thereis atypeofphysical(orfunctional)stateP,suchthatitisnomicallynecessarythatfor any being x, x is in M if and only if x is in P.2,3 ForthePPCthesistoserveasapremiseinanIBEargumentforanidentitytheory, it must represent a summary of empirical findings thus far. The PPC thesis is more happily read, then, as a summary of certain observations made in the course of neuroscience. Such correlations populate the neuroscience literature: fear is correlated with activity in the amygdala, pain is correlated with activity in the somatosensory cortex, visual sensations are correlated with activity in V1, etc. This means the PPC thesis is not to be read as the conclusion of an a priori metaphysical argument.Rather,theideaisthatthetruthofthePPCthesisisestablishedbyfinding a sufficient number of mind-brain correlations, while the explanation of the PPC thesis,iftrue,isthatidentitiesholdbetweenthementalstatesandthephysicalstates being referred to. Manyauthorsdobelievethat(1)therearecorrelationsbetweenmentalstatetypes and physical state types,4 and (2) that these correlations do provide evidence for identity based on an IBEargument. Theargument takes roughly thefollowing form: First Premise: if a theory provides a good explanation of a set of facts, and the explanationisbetterthananyexplanationprovidedbyacompetingtheory,then one has a good and sufficient reason to believe that the theory is true. (IBE Principle) Second Premise: type materialism [a.k.a., type identity theory] provides a good explanationofthepsycho-physicalcorrelationsthatareclaimedtoexistbythe PPC. Third Premise: moreover, the explanation that it provides is superior to the explanations provided by all competing theories. Conclusion: provided that the psycho-physical correlation thesis is true, we have goodandsufficientreasontobelievethattypematerialismistrue.(Hill,1991,p. 23) Hill’sargumenthereisaclearstatementofthestrategy,andsomakesagoodtarget for the ensuing discussion. Indeed, on its face, this argument seems sound (it is Philosophical Psychology 189 certainlyvalid).Aproblemarises,however,dependingonwhatwetakeourevidence for the second premise to be. Consider: the PPC thesis, as stated, implies that the correlations to be found betweenmentalstatetypesandbrainstatetypeswillberelativelystrong,meaningthat there will be few to no exceptions. It also implies that there will be a strong correlation for every mental state type of interest. (More will be said about correlations andtheirrelativestrengthbelow.)Thesetwoimplicationsareintension with each other: if the correlations to be explained are strong, then they are unlikely to apply to all cases of mental states of interest. Likewise, if we wish to capture all mental states of interest, the correlations we wish to explain might be relatively weak—and then the identity theory is a less viable explanation for them (or so I shall argue). Put another way: if the identity theorist seeks to explain (relatively) weak correlations, thenthethirdpremiseisfalseand theIBE argument fails. Ontheother hand,if onlymaximally strong correlations are allowed, thenthePPCthesis willnot hold true for many pairings of mental states and physical states that are of interest, and the identity theory loses its generality as a theory of mind. Inthenextsection,Ioutlinethewaysthatmind-brain correlations areestablished empirically to provide background for the discussion to come. In sections 4 and 5, I will look at the first horn of the dilemma, that identity might not be the best explanation for weak correlations. Section 6 will look at the other horn: focusing on maximally strong correlations erodes the appeal of an identity theory as a general theory of mind. I end with some common objections to my argument in section 7. 3. Mind-Brain Correlations Tosaythat‘‘Mis(type)identicaltoB’’istomakeakindofmetaphysicalclaimabout the kinds of things named by ‘M’ and ‘B’.5 A mind-brain correlation, on the other hand,isapatternofco-occurrencebetweenagivenMandagivenB.ThePPCthesis claims that there will be a mind-brain correlation for every mental state of interest, while the IBE argument provides evidence for the identity theory based on these correlations. So what exactly does it mean to say that a mind-brain correlation is a ‘‘pattern of co-occurrence between a given M and a given B’’? Roughly, when a type of mental stateispresent,atypeofbrainstatealsoseemstobepresent(andviceversa).Amore rigorousdefinitionofmind-braincorrelationcanbehadintermsofgoodevidencefor the presence or absence of a given state: Mind-brain correlation: a mental state type M is said to be correlated with brain state type B iff, forall (ormost) tokens ofM and B, thefollowing criteria obtain: (1) WhenwehavegoodevidencethatamentalstatetokenofMispresent,wealso have (or could obtain) evidence that a brain state token of B is present (M!B); 190 B. N. Towl (2) When we have good evidence that a brain state token of B is present, we also have (or could obtain) evidence that a mental state token of M is present (B!M); (3) When we have good evidence that M is absent, B also appears to be absent ((cid:2)M!(cid:2)B); and (4) When we have good evidence that B is absent, M also appears to be absent ((cid:2)B!(cid:2)M).6 And so there are actually four kinds of evidence we can look for in order to establish a mind-brain correlation. If we have all four kinds of evidence, and the evidence is good, we can say that we have a strong correlation. If we have fewer than these four, or if some of the evidence is contested, we can talk about a weaker correlation.7 It should not be surprising, given the above, that the methods neuroscientists employ to look for such correlations mirror these criteria.8 For example, neuroscientists do experiments that: (1) Activateagivenbrainstate,andthenmeasurewhetherornotagivenmentalstateoccurs. Examples of such ‘‘activation’’ studies include direct stimulation experiments using electrodes or chemical intervention, as well as indirect stimulation studies using administered drugs; (2) Prevent,destroy,orhandicapabrainstate,andmeasurewhetheragivenmentalstatesis absent. Most lesion studies fall under this category, as do experiments that pharmaco- logically block the activity of a mechanism and ‘‘knock out’’ studies that prevent the expression ofgenes responsible for generating certain kinds ofstructures; (3) Present stimuli constructed to activate a given mental state, and then look for correspondingbrainstates.Mostimagingexperimentsfallunderthiscategory,asdoEEG experiments; and (4) Presentstimulithatpreventordeactivateamentalstate,andthenseewhichbrainstates are absent. Though much more rare, such experiments do exist. For example, there are fMRIexperimentsthatrevealdifferencesinbrainactivitywhenasubjectinanexperiment hasfailedtoencodeapieceofinformation,asshownbysubsequentforgettinginarecall task. In other words, the various methods that neuroscientists actually employ mirror the criteria from the definition above. If all four types of experiment are done for a givenMandB,andtheseexperimentsshowthatMdoesinfactcorrelatewithB,then we have a strong correlation between the two. There could be many or few mind-brain correlations (a correlation is between a given M and a given B, after all). And some mind-brain correlations might be weak whileothersarestrong.ThisfactsuggeststhattheargumentgivenbyHillneedstobe broken intotwosteps.First, ifamind-brain correlation isfound, thereisapotential IBE argument for a single identity claim—that is, that a given M is type-identical to agiven B.Second, there needs tobe aninductive argumentto establish amind-brain identity theory that covers all (or most) mental states.9 The inductive argument is needed for the identity theory to be a general theory of mind: finding a correlation between one mental state and one brain state, though interesting, would not be adequateevidencetobuildanargumentfortheidentitytheory,ageneraltheoryabout Philosophical Psychology 191 mental states. After all, if we found that pain, say, correlated strongly with C-fiber firing,butthatnoothermentalstateseemedtocorrelatewithanyknownbrainstate, no one would claim that we have identity theory of mind (though we might have a case for a single identity claim). It would have to be the case, then, that neuroscientists findmanymind-brain correlations—enough to support aninductive argument for identity theory. Now we can bring the initial tension into better focus. As mentioned previously, correlations can be weaker or stronger. If we allow weak correlations to provide evidenceforthePPCthesis,thenitisunclearthatarelationofidentitywillalwaysbe thebestexplanationforthosecorrelations.Ontheotherhand,ifweallowonlystrong correlations to provide evidence for the PPC thesis, then there are vanishingly few such correlations—so much so that the inductive step becomes suspect. In summary, I am suggesting that it is misleading to ask whether type identity theory is the best explanation for the PPC thesis (as stated). Rather, we need to ask: (1) ‘‘for individual cases, would an identity claim be the best explanation for a given mind-brain correlation?’’ and (2) ‘‘are there enough such identity claims to merit a general identity theory?’’10 4. Identity Claims and Correlation Let us examine the former question first—whether identity claims are a sufficient explanation for a given mind-brain correlation. Kim (2005) has recently argued that identities are not actually explanatory, and thus that IBE arguments for identity theory will fail (thus denying Hill’s second and third premises). Briefly, Kim argues that identity claims simply afford re-write rules that allow us to rephrase statements about mental states into statements about neural states. Any actual explanation that occurs, however, is done outside of the identity statement. Kim seems to consider two types of correlation through his given examples: correlations between phenomena that are diachronic, and correlations between phenomena that are synchronic. Diachronic phenomena are perhaps best explained intermsofcausalchains;theyarenotthesortsofcorrelationsbeingconsideredhere. The sorts of correlations that Hill, McLaughlin, and others are concerned with are synchronic—that is, where tokens of the types involved co-occur. So let us focus on Kim’s synchronic example. Kimprovidesonesuggestedderivationofacorrelationfromanidentitystatement: Pain occurs$pain occurs. Pain¼Cfs [C-fiber firings]. Therefore, pain occurs$Cfs occur. (Kim, 2005, p. 135) In this example, the correlation is derived from the identity statement plus a tautology. But, says Kim, the only explanatory premise is a factually empty logical truth (the tautology); the identity claim here simply allows us to rewrite this truth using substitution. This sort of explanation, for Kim, ‘‘is a far cry from genuine, 192 B. N. Towl factually substantial and empirically falsifiable, explanations of correlations found in science’’ (2005, p. 136). Kim’s assessment is a bit unfair. First, even if identities only provide ‘‘re-write’’ rules, identities of the form ‘a¼b’ can still be informative. And if they are informative, we should not pass them off as just rewrite rules. Second, and more importantly, identitiescanhaveexplanatorypowerwhen:(1)thecorrelations weare faced with are, by themselves, mysterious; and (2) an identity would be a better explanation of the correlations than competing explanations. Granted, it would take some work to show that (1) and (2) obtained for any given example—but their possibility shows that identity claims might still be ‘‘in the running’’ as explanations of particular mind-brain correlations. The main point of this paper is not to argue against Kim, however. I am taking it as an assumption that it is at least plausible for an identity claim to be part of an explanation as to why some M and some B are correlated. Such correlations are themselves summaries of patterns of observation. Thus identity claims must be able to explain the systematic pattern of observations we obtain, and do so better than rival explanations. My main argument is that, in doing so, identity theory faces a dilemma. Although an identity claim could be a good explanation of a mind-brain correlation (and thus also explain the patterns of observation we obtain), there are many other ways in which a correlation can occur, ways that are consistent with the results from experiments. To suggest some alternatives, consider that we would find a correlation between a given mental state and some brain state because: (1) The mentalstate caused thebrainstate; (2) The brainstate caused thementalstate; or (3) Both themental state andthebrain state arisefrom a commoncause. These relations would still yield mind-brain correlations, and yet the mental state and brain state would not be type-identical.11 In what follows, then, I will lay out a particular example where one would expect an identity claim to hold: that of the correlation between thirst and hypothalamic activity. What a detailed examination of this case will show is that, even here, the correlation between thirst and hypothalamic activity is not as strong as needed to fullysupportanidentityclaim—otherrelationscouldexplainthecorrelationaswell. 5. A Historical Case: The Hypothalamus and Thirst ThecaseIwanttoexploreisoftenconsidereda‘‘classictextbookcase’’ofmind-brain correlation: the correlation between the activity of the hypothalamus and the sensation of thirst. What Iwant to showwith this case isthat, once one explores the details,evenaseeminglystrongcorrelation(and,hence,agoodcaseofamind-brain claim) turns out to be a relatively weak correlation—and so the third premise of Hill’s argument is false for many cases of mind-brain correlation. Philosophical Psychology 193 The hypothalamus is a cluster of structures located near the brain stem between the pituitary gland and the thalamus, surrounding the third ventricle of the brain. It is densely connected with a number of areas, including prefrontal cortex, hippocampus, amygdala, pituitary, and the brainstem. Since at least the 1950s, experimenters have suspected that the hypothalamus plays a pivotal role in the regulation of thirst. Because part of the hypothalamus lies outside the blood-brain barrier, it is ideally situated to receive blood-borne messengers. Its dense intercon- nections with the pituitary also make it well situated to control the internal environment. Indeed, the received view is that the hypothalamus, or parts thereof, is ‘‘the center’’ of our thirst sensations.12 Why is the hypothalamus a good test case for the claim that a mind-brain correlationisevidenceforanidentityclaim?Therearethreethingsthatmakethisan appropriate case to study. First, the hypothalamus is an evolutionarily old structure. Itisfoundinseveralspeciesandseemstohavechangedlittleoverevolutionarytime, making it comparable across species (compared to, say, parts of neocortex). Second,thehypothalamusseemstobeinvolvedinanumberof‘‘drives’’whichare relatively easy to study: hunger, thirst, sexual response, fear, etc. Whereas the behaviorswewouldexpectfrom, say,avisualsensation may varyfromindividualto individual, the behaviors that arise from drives are especially stereotyped. Thus we havegoodwaysofdetectingthirst:wecanmeasuresuchthingsasmovementtowards water, amount of water drunk, number of drinking episodes, etc. Third, we know a great deal about the chemistry and physiology of the hypothalamus. We know what neurotransmitters it traffics in, and we know what chemicalchangesinthebodytendtoaffectit(forexample,seeRolls&Wood,1977). This gives us the ability to both manipulate and measure this structure. Sothe hypothalamusseems arather goodcandidate forthekind of IBEargument I have been considering. We have methods to influence the activity of the hypothalamus,andwealsohavegoodmethodsforinducingorslakingthirst.Wealso have good ways of operationalizing thirst—i.e., there are a number of behaviors we can measure that depend directly on thirst (such as movement towards water, amount of water drunk, number of drinking episodes, etc.). And given that the hypothalamus exhibits little variation across individuals and across species, we have a number of animal models we can use to test hypotheses. The first critical experiments that correlated the hypothalamus with thirst were done in the 1950s by Andersson and colleagues (Andersson, 1953; Andersson, Dallman, & Olsson, 1969; Andersson & McCann, 1955; Greer, 1955). These early experimentsshowedthatinjectionsofhypertonicsalineinthehypothalamusofgoats eliciteddrinking,evenintheabsenceofotherfactors(lowbloodosmolalityandhigh body temperature, for example). Greer (1955) and Mogensen and Stevenson (1967) went on to show that electrical stimulation of the lateral part of the hypothalamus alsoinduceddrinking.Thesestudiesshowthatactivatingaparticularbrainstatelead totheexpressionofbehaviortypicalofbeinginacertainmentalstate—beingthirsty. In the symbolization from above, these studies establish that B!M. 194 B. N. Towl Further studies showed that lesioning parts of the hypothalamus in rats and cats often extinguished drinking behavior (Bellinger, 1987; Cambiasso & Chiaraviglio, 1992; Clark et al., 1991; Saad, Luiz, De Arruda Camargo, Renzi, & Manani, 1996). OnestudydonebyCareyandGoodall(1975)showedthatlesionstomedialpreoptic areaofthehypothalamusinratsloweredtheir‘‘tolerance’’forwateradulteratedwith quinine, suggesting that their thirst drive was having less of an effect on behavior. These studies are consistent with (cid:2)B!(cid:2)M. Studies have also been done altering the mental states of animals and looking for changes of brain state. Patronas, Horowitz, Simon, and Gertsberger (1998), for example, did a series of experiments that induced thirst in rats by dehydration and heat acclimation. They then noticed an increase in the expression of the c-fos protooncogene, which suggests that hypothalamic activity increases during long exposure to thirst inducing stimuli. Other labs have shown that the mere visual presentation of a known water source (a drinking syringe, for example) increases hypothalamic activity (Rolls & Rolls, 1982). Thus manipulations of the mental state cause noticeable changes in the activity of the hypothalamus: M!B. These studies nicely mirror the criteria I set out above in section 3.13 It is because of these and similar studies that many researchers consider the hypothalamus a ‘‘thirst center’’ in the brain, involved in the regulation and control of drinking behavior.Butsubsequentresearchshowsthatthestoryhereisnotsoclear-cut:there are several problems with the ‘‘hypothalamus as thirst center’’ story. First, there seem to be conflicting data about which part (or parts) of the hypothalamuscountas‘‘thecenter.’’Thedorsalmedialnucleusofthehypothalamus seems a likely candidate, since lesioning this area reduces drinking behavior (Bernardis, 1970; Bernardis & Bellinger, 1998; Clark et al., 1991), and electrical stimulation causes drinking (Rolls & Rolls, 1982). But lesioning this area often also reduces eating behavior, and both behaviors tend to recover spontaneously (Rowland, 1976; Zigmond & Stricker, 1973; also see the debate covered in Wayner, 1964).Thusitisnotclearthatthedorsalmedialhypothalamusisresponsiblejustfor thirstbehavior,norisitanecessarycomponentinsomecases.Forotherareasofthe hypothalamus, the story seems to be the same: there doesn’t seem to be one area singly responsible for, or primarily dedicated to, thirst related activity.14 Second, thirst related behavior can be elicited through activation of brain areas outside of hypothalamus. A series of studies have showed that activating the ventrolaterialperiaqueductalgraymatter(orPAG—apartofthebrainstem)ofcats, rats, and monkeys causes many of the same thirst-related behaviors as activation of the hypothalamus (Wyrwicka & Doty, 1966; Sharpe & Swanson, 1974; Swanson & Sharpe, 1973). Likewise, lesions in this area cause the loss of drinking and feeding behavior, just as lesions of the DMH do (Lyon, Halpern, & Mintz, 1968). There is evenevidencethatactivatingpartsoftheneocortexofmacaquemonkeys(specifically area32inmedialprefrontalcortex)elicitsdrinkingbehavioraswell;similarfindings have come out of PET studies of human brains (Denton et al., 1998, 1999). Also playing key roles are mesial cortex and thalamus (Sewards & Sewards, 2003) and medial forebrain (Rundgren & Fyhrquist, 1978). Granted, these other areas are Philosophical Psychology 195 connected to the hypothalamus. But they show their own correlations with drinking behavior in a way that throws doubt on the identity claim. Third,aseriesofexperimentsbyRollsandcolleagues(Rolls,1974;Rolls,Burton,& Mora, 1980) showed that some hypothalamic neurons in monkeys increased their activity with the taste of water in the mouth, as compared to isotonic saline, and decreased activity when a puff of dry air was presented. Thus, manipulations that increase the satiation in monkeys cause increased activity, while manipulations that cause thirst sometimes cause decreased activity. We would expect the opposite pattern of results if activity in the hypothalamus represented thirst. Thesethreeconsiderationscastsomeseriousdoubtontheclaimthatactivityinthe hypothalamus, or parts thereof, are identical with the sensation of thirst. Although the correlation looks quite strong at first, there are several experiments that show that: (1) multiple other areas produce thirst responses; (2) the hypothalamus is responsible for more than just thirst; and (3) lack of hypothalamic activity does not always correlate with lack of thirst. This is not to say that there is no correlation between hypothalamic activity and thirst. The initial findings of these researchers show some correlation. The problem is simply that the correlation is not a maximally strong one. Sowhatdoesthehypothalamusdowithregardstothethirstmechanism,ifitisnot the ‘‘center’’ for our thirst sensations? There are at least two likely theories worth exploring.Oneisthatthefunctionofthehypothalamusmightbetomonitorcellular dehydration via blood osmolality, and ‘‘kick-off’’ various thirst mechanisms in response.Onthistheory,althoughtheactivityofthehypothalamusisnotidenticalto the sensation of thirst, it regularly causes the sensation of thirst. A second theory (Rolls & Rolls, 1982) is that the hypothalamus is a kind of ‘‘reward center’’ for successful eating and drinking behaviors. Again, this would make the activity of the hypothalamus not identical to the sensation of thirst, but it would be caused by the satiationofthirst.Thesetworelations—bothcausalrelations,bytheway—seemtofit the evidence better than the relation of identity. Given this, the correlation between thirst and hypothalamic activity is not best explained by the thesis that one is identical to the other. Theseresultsarenotspecifictothegivenexample.Wheneverwehavealess-than- strong correlation, it will be possible that another relation besides identity could explainthecorrelationwefind.Thisisnottosaythatnomentalstatewillbefoundto be identical to any brain state. But, for there to be an IBE argument for such an identity claim, there would have to be a maximally strong correlation. 6. The Dilemma For Hill’s argument above to work, then, the PPC must capture not just any correlationsbetweenmentalstatesandbrainstates;rather,itmustcapturemaximally strong correlations. Otherwise, the third premise of the argument would be false for many cases. The hypothalamus case is one example of this. The case illustrates a 196 B. N. Towl general principle, however: the weaker the correlations one considers, the more cheaply they can be had. Identity claims, however, are strong claims. Thus the numberofidentitiesisboundtobemuchsmallerthanthenumberofcorrelationswe uncover. So not all correlations can be evidence for an identity. So suppose we take the set of correlations to include maximally strong correlations. If we insist that a correlation be sufficiently strong to support an identityclaim,wewillfindthatthenumberoftrulystrongcorrelationsisvanishingly small in neuroscience. To cite some anecdotal examples: . Some neuroscientists claim that there are facial recognition cells in inferotemporal cortex, though there is evidence for this, these cells also become active when a toilet brush is moved rapidly through theirreceptive fields (Gross,2008; Richmond &Wurtz,1982). . Fearresponsesandfearconditioningdocorrelatewithamygdaleactivity,buttheamygdalaalso seemstoplayrolesinlong-termmemoryconsolidation,theprocessingoffacialexpressions,and detectionof‘‘personalspace’’violations(Kennedy,Gla¨scher,Tyszka,&Adolphs,2009;Williams etal.,2006).Furthermore,areassuchasthemedialgeniculatenucleus(MGN)arealsouniquely poised to regulatefear responses (Weinberger, forthcoming). . Though anterior cingulate activation seems to correlate well with response selection, error detection, and emotional salience, there are a number of studies contradicting these roles (Fellows &Farah,2005; Nachev, 2006; Rushworth,Behrens, Rudebeck, &Walton, 2007). Space limits me from going into detail about these further cases, but we can explain their occurrence—and the possibility of many such others—straightfor- wardly. Our mental taxonomy is not devised solely on the basis of neuroscientific evidence. Mental categories might well come from folk psychology, older-but-still- popular psychological theories, current theories in psychology, or psychiatry, or anthropology, or sociology, etc. Given the diversity of sources for our mental categories, we would expect many cases where a given mental category does not line up neatly with our neuroscientific categories. Ifthisiscorrect,thereareboundtobemany examplesofweakcorrelations inthe neuroscience literature. This creates a problem for the induction needed for Hill’s argumenttowork.IfonlystrongcorrelationsgiveusanIBEargumenttoanidentity claim, there will be very few identity claims. And so the general thesis, that type- identity theory is true for all or most mental states, will be an induction based on an exceedingly small number of cases. So we still have a very unconvincing positive argument for the type identity theory. 7. Objections and Implications TherearesomereadyobjectionstotheargumentIhavegivenhere;tomakemycase, there are three that I need to address. The first objection is that I have focused only on correlations that have seemed strong but that are, in fact, weak. The second objectionisthatIhavefocusedonneuralareasthataretoosmall.Thethirdobjection is that I have focused on a largish brain area, rather than smaller network or