On the Rationality of our Response to Testimony

On the Rationality of Our Response to Testimony ABSTRACT The assumption that we largely lack reasons for accepting testimony has dominated its epistemology. Given the further assumption that whatever reasons we do have are insufficient to justify our testimonial beliefs, many conclude that any account of testimonial knowledge must allow credulity to be justified. In this paper I argue that both of these assumptions are false. Our responses to testimony are guided by our background beliefs as to the testimony as a type, the testimonial situation, the testifiers’ character and the truth of the proposition testified to. These beliefs provide reasons for our responses. Thus, we usually do have reasons, in the sense of propositions believed, for accepting testimony and these reasons can provide evidence for the testimonial beliefs we form. _____________________________ It is evident that, in the matter of testimony, the balance of human judgement is by nature inclined to the side of belief; and turns to that side of itself, when there is nothing put into the opposite scale. If it was not so, no proposition that is uttered in discourse would be believed, until it was examined and tried by reason; and most men would be unable to find reasons for believing the thousandth part of what is told them. (Reid 1764, 197). In this quote Thomas Reid starts with the descriptive epistemological claim that unless those propositions we accept supply reasons for disbelief, we are inclined to accept testimony. Reid presents this ‘natural’ attitude as the epistemologically correct Synthese, volume 131, no. 3, June 2002, 353 – 370. attitude: we are justified in accepting testimony credulously. He then supports this epistemological principle with the argument that if things were otherwise, if we needed to justify our acceptance of testimony, then only a small fraction of our testimonial beliefs would be justified. Michael Dummett articulates an almost identical argument. Dummett starts with the descriptive epistemological claim that in ordinary cases of testimony: I go through no process of reasoning, however swift, to arrive at the conclusion that he has spoken aright: my understanding of his utterance and my acceptance of his assertion are one. (Dummett 1993, 419). Dummett then raises the question of whether this feature of our receipt of testimony is an “epistemological principle” or a “mere psychological phenomenon”; he raises the following doubt. Maybe it is simply in our nature to accept the … assertions of others without, usually, any scrutiny or reflection; but may it not also be that, if we are to possess knowledge acquired by [testimony], we must be able to supply as backing an argument corresponding to the inference we omitted to draw? (Dummett 1993, 420). This question Dummett answers negatively. If the suggestion were adopted — if we must be able to supply reasons to justify our acceptance — then, he claims, “we should have to confess to knowing pitifully little.”1 Thus, Dummett is not merely claiming that we go through no process of reasoning, rather his claim seems to be either that we lack reasons for our acceptance of testimony, or that the reasons we have for acceptance are not sufficient to justify our testimonial beliefs. Our credulous acceptance of testimony, therefore, must be sanctioned by an “epistemological principle”, that is, it must be justified. 2 Synthese, volume 131, no. 3, June 2002, 353 – 370. This argument to the conclusion that credulity is justified may be set out somewhat as follows. (1) We largely lack reasons — in the sense of propositions believed — for accepting testimony. (2) In those cases where we do have reasons, these reasons are usually insufficient to justify our testimonial beliefs. (3) Therefore, if our practice of accepting testimony without reason were not justified, few of our testimonial beliefs would be justified. (4) Many of our testimonial beliefs are justified. (5) Therefore, we must be justified, other things being equal, in accepting testimony without reason, that is, credulously. Variations of this argument have been extremely influential. In arguing against reductionism — the view that it is our reasons for accepting testimony that justify our testimonial beliefs — Coady claims that acceptance without reasons must be justified because we just could not individually justify our testimonial beliefs.2 Similarly, Hardwig argues epistemic dependence must be justified because we cannot possess reasons for the truth of many our testimonial beliefs.3 Both of these arguments start by presuming the truth of premise (1). Furthermore, this presumption tends to be made, independently of its use in any argument, whenever the conclusion that credulity is justified is accepted.4 Contrary to (1) I will claim that we usually have reasons — in the sense of propositions believed — for accepting testimony. Let me label this descriptive epistemological claim (R), (for reasons). If (R) is true, then (1) is false and any argument that employs (1) as a premise is unsound. The particular Reidian argument of the last paragraph may also be contended at premise (2). In disagreement with (2), I 3 Synthese, volume 131, no. 3, June 2002, 353 – 370. will claim that our reasons for accepting testimony can provide evidence for the testimonial beliefs we form. It follows that our reasons for acceptance could justify our testimonial beliefs. Let me label this italicised claim (E), (for evidence). If (E) is true, then (2) may be contended.5 The claims (R) and (E) — we usually have reasons for accepting testimony that can be evidential — need to be framed by three important clarifications. First, the descriptive claim (R) neither states nor implies that acceptance is always inferential. Second, claims (R) and (E) involve an important distinction between the notions of reasons and evidence. Third, neither (R) nor (E) is a claim about what justifies our testimonial beliefs. Let me consider each of these points in turn. Our acceptance of testimony is frequently immediate and non-inferential. If (R) was equivalent to the claim that our acceptance is usually mediated by an inference to the truth of what is expressed, then (R) would be false. However, (R) is not equivalent to this psychologically implausible claim. The beliefs that constitute our reasons for acceptance tend to be background beliefs. To give a background role to our beliefs is simply to acknowledge that acceptance rarely involves explicitly articulating a justifying argument. Similarly, if Grice’s analysis is accepted, then our understanding of implicature is reasoned but it may, nonetheless, be immediate.6 As a way of forming belief, testimony is largely immediate. Second, claim (R) concerns our reasons for belief. Our reasons for belief are those propositions we take to be evidence for what we believe. Claim (E) concerns the evidence for our beliefs. Evidence for a proposition either entails or makes probable 4 Synthese, volume 131, no. 3, June 2002, 353 – 370. that proposition. To be evidence a proposition, at least, must be true. By contrast, a false proposition may be taken to be evidence and a subject’s reasons for belief need not determine that what is believed is true or even probably true. This allows for the possibility that our reasons for acceptance are such that they do not provide evidence for the testimonial beliefs formed through acceptance. That is, the truth of (R) is compatible with the falsity of (E). Third, according to some theories of testimonial justification, it is our reasons for accepting testimony that justify our testimonial beliefs. According to other theories, we are justified in accepting testimony without reason and it is, paradigmatically, the speaker’s reasons that justify our testimonial beliefs.7 Neither (R) nor (E) makes a claim about what justifies our beliefs. The former states that we rarely accept testimony without reason. The latter states that these reasons can provide evidence for the testimonial beliefs we form — our reasons could be justificatory. This is not to claim that such evidence justifies our testimonial beliefs; it is not to claim that our reasons are justificatory. Thus both (R) and (E) are neutral with respect to competing theories of testimonial justification. On any occasion where we comprehend an utterance — testimony — we are disposed to believe, disbelieve or suspend judgement in, the proposition expressed. Whether we accept the testimony, reject it, or suspend judgement our doxastic response is rarely a product of inference. Nonetheless, I hope to make (R) persuasive through showing how our doxastic response to testimony is determined by what we believe, background or otherwise. I shall start by considering the general beliefs we have about the credibility of testimony and use empirical psychology to try and model how 5 Synthese, volume 131, no. 3, June 2002, 353 – 370. our observation of the communicative context allows us to form credibility judgements. This shall occupy the next, and longest, section of this paper. In the second section I offer a rather rough epistemic evaluation of the classes of reason thus presented and thereby argue for (E). 1. Responding to Testimony 1.1. Prior Presumptions We distinguish between different types of testimony. Certain types we are disposed to accept, whilst others we are disposed to reject. Given the comprehension of testimony, we can recognise the testimony to be a certain type and are thereby disposed to accept it or otherwise. Thus, an initial characterisation of the reasons explaining our responses to testimony requires an account of our typing of testimony. The doxastic dispositions thereby explained are essentially synchronic: our beliefs about testimonial types are formed prior to our encounter with any given testimony. In the next section, I shall consider how our doxastic dispositions are diachronically informed by our engagement with the contextual circumstances of testimony. We receive testimony from many sources and testimony to different topics. We distinguish between these sources and topics. We are more likely to believe friends, family and lovers than strangers, doctors than apothecaries, and vicars than lawyers. We are likely to believe speakers talking about everyday events, giving us directions, or telling us the football scores. However, we tend to be sceptical of speakers talking about politics, the greatness of their exploits, or the statistics that favour their opinion. On some quite specific topics we may be utterly sceptical, or conversely credulous.8 6 Synthese, volume 131, no. 3, June 2002, 353 – 370. Allowing that content may be identified as a particular topic, and named individuals as particular sources, six generic types of testimony could be identified.9 These are (1) general topic, (2) general source, (3) general source and general topic, (4) particular source, (5) particular topic, and (6) particular source and general topic.10 It is our distinguishing between different types of testimony that explains our presumptions. According to Burge, “We make use of a presumption of credibility when we read books, signs, or newspapers, or talk to strangers on unloaded topics”.11 We presume credibility, as Burge notes, but not across the board. We presume that certain books and newspapers are credible, but not that all books and newspapers are credible. On the contrary, most would presume that much of what they read in the National Enquirer or The Daily Sport is false. Moreover, if we presume that the testimony of strangers on unloaded topics is credible that is because we have the general belief that the testimony of strangers on unloaded topics is credible. Thus, our background general beliefs — beliefs about what types of testimony tend to be credible — explain our presumptions.12 Our standing disposition to accept a given testimony could then be explained as follows. We have a long history of dealing with testimony and on this basis we have learnt to distinguish testimonies into types some of which we take to be credible, some non-credible and others as credible as not. With this background of belief we are in a position to form the judgement, of any given testimony, that it belongs to a certain type, say type x. We believe that type x testimony is credible; we would assent to the statistical generalisation ‘Testimony of type x is probably true’. We then directly infer that this testimony is credible. Our disposition to acceptance is thereby 7 Synthese, volume 131, no. 3, June 2002, 353 – 370. explained by our believing the premises of the statistical syllogism: testimony of type x is credible, this testimony is of type x, and therefore this testimony is credible. Our standing disposition to reject or suspend judgement in a given testimony is equally explained by our possessing a general belief about the credibility of these testimonies, only in these cases the testimony will be believed non-credible or as credible as not respectively. This is not to claim that audiences explicitly articulate the statistical syllogism when accepting, rejecting or suspending judgement in testimony. It is to claim that an audience’s recognition of a testimony as a certain type in conjunction with the belief that such a type is credible provides an explanation of their doxastic disposition. The statistical syllogism thereby represents our reasons for belief, or otherwise, even if it is rarely explicitly articulated.13 Let me provide an illustration. Suppose that I suffer certain pains and visit a doctor who diagnoses no more than malaise. I may go through no process of inference in accepting the doctor’s diagnosis, my understanding and acceptance may be one, but this is not to say that I lack reasons for acceptance. I accept because I have the background belief that general practitioners are credible on medical matters and the surgery setting supports my presumption that I have communicated with a general practitioner. Call the doctor’s testimony d, my credibility belief c and my type belief t. I accept the doctor’s testimony because I take the conjunction of c and t to be evidence for d. 8 Synthese, volume 131, no. 3, June 2002, 353 – 370. Support for this claim is provided by a consideration of subsequent suspension of judgement. If we recognise that p was our reason for believing q, then if we come to believe not-p, we will, other things being equal, suspend our belief that q. Thus if I were to subsequently come to believe either not-c or not-t, then I should cease to believe d. That I should do so shows the conjunction of c and t to have originally been my reason for acceptance. It is doubtful that I would abandon my credibility belief, but I might learn that the doctor was a charlatan and if I were to learn that the doctor has fabricated his qualifications, I should suspend my belief that my pains were no more than malaise. In seeking a second, authoritative, diagnosis I show my belief that the doctor is a qualified general practitioner to have been, at least part of, my reason for accepting his testimony. We find that in the process of communicating, dispositions to accept, reject or suspend judgement are engendered. A testimony might fit no type for which we have a general belief concerning its credibility yet our perception of its context can dispose us to accept it or not, and a testimony that we might be inclined to reject initially can become more plausible as the communicative context supplies further information, and conversely.14 I shall now consider how to model such contextual reasons. 1.2. Contextual Reasons Suppose that you do not presume to trust strangers even on unloaded topics; maybe this is wartime and you are in an occupied land. Nonetheless, needing to find a certain landmark you approach a passer-by to ask for directions. It seems intuitive, if you understand what is said, that the passer-by could communicate her knowledge of the whereabouts of this landmark. Despite an absence of background general beliefs 9 Synthese, volume 131, no. 3, June 2002, 353 – 370. about the credibility of this type of testimony, the communicative context, I would argue, can supply reasons for acceptance or otherwise. How this could be so requires an empirically plausible explanation. I think that the first thing to note is that we treat certain, but not all, contextual particulars as relevant to the credibility of testimony. In the present example that the sun was overhead is a fact that does not seem relevant. However, that the passer-by seemed well disposed, smiled, repeated her directions and did not seem in a hurry, do seem relevant in that attention to them would, in all probability, dispose one towards acceptance. Other circumstantial features, attention to which could dispose one towards acceptance or otherwise, include: 1. The speaker seemed suspicious. 2. The speaker paid careful attention to what was said. 3. The speaker contradicted himself. 4. The speaker’s utterances were thoroughly coherent. 5. The speaker seemed worryingly eager to be believed. 6. What the speaker said cast him in a bad light. Attention to half of these contextual details — namely 1, 3 and 5 — might dispose one to reject the accompanying testimony and conversely the other details. This list could be obviously extended almost endlessly. On the assumption that this is correct and that we do treat such contextual features as relevant to the question of whether or not to accept testimony, then why do we treat these features, and not others, as relevant? An adequate answer to this question requires an appeal to empirical psychology.15 The following answer could be thus supported. 10 Synthese, volume 131, no. 3, June 2002, 353 – 370. As audiences, we have a history of engaging with testimony. A good deal of this history has been successful: much of what we know, we know through testimony. For any audience, past testimonies could be divided into three classes: accepted testimony, rejected testimony, and testimony on which judgement was suspended. Call these classes A, R, and S respectively. Given our past success, we will be inclined to accept testimony similar to those in class A, reject testimony similar to those in class R, and suspend judgement on testimony similar to those in class S. Suppose that a testimony falls into class S if it is equally similar to those in A and those in R. The identification of relevant contextual details could then be based on recognition of the similarities and dissimilarities amongst the testimonies that are members of A, R. For example, suppose two rejected testimonies are similar in point of feature φ. Other testimonies in R could also be similar in point of φ. Let Φ stand for the set of φ features extracted from class R. From set Φ a proto-typical φ can be judged and in this manner, a set of proto-typical features can be extracted from classes R and A. Given that this involves no more than rudimentary similarity judgements it seems reasonable to suggest that in recognising details 1 to 6 as relevant, we recognise such proto-typical features. 16 Consequently our disposition to accept testimony or otherwise because of the communicative context could be represented in the following manner. First, the relevant circumstantial features of the context are recognised. This may be a matter of consideration or it may be immediate. Each of these features will be similar to a proto-typical feature extracted from A and R. The testimony will thereby be judged similar to those in class A, to those in class R, or to those in both classes. Thus, we would be disposed to accept, reject or suspend judgement in the testimony. In the 11 Synthese, volume 131, no. 3, June 2002, 353 – 370. former case, for instance, we could be represented as reasoning, ‘this testimony is similar to those previously accepted, and therefore it is credible’. Certain weaknesses and strengths of this model are transparent. On the positive side the model accounts for how our reasons for accepting testimony, or otherwise, can be (1) contextual, (2) intuitive and (3) learnt. On the negative side, the model seems too oversimplified and, crucially, it seems a hostage to fortune. Let me take the positive points first. (1) Our reasons for accepting testimony, or otherwise, can be contextual because their ground can be the identification of certain contextual details as evidence for the acceptability of the testimony, or otherwise. (2) This judgement is intuitive because the audience need not occurrently remember any of the past occasions of testimony where similar contextual features were identified. It is enough that these occasions have defined the proto-type by means of which the present identification is effected. (3) Our ability to assess the credibility of testimony improves with experience: each new success alters the set of relevant features, such as φ, and thereby the proto-typical feature associated with each such set.17 On the negative side, this model is too oversimplified. Relations of similarity exist not simply amongst the testimonies in A and R but also between the testimonies in these classes. Thus, the definition of proto-typical features on the basis of experience seems more difficult than suggested. This complication might reasonably be put aside insofar as it is not clear that insurmountable processing problems are posed. However 12 Synthese, volume 131, no. 3, June 2002, 353 – 370. to do this may seem cavalier in the light of the second weakness of this model, namely that it is a hostage to fortune. It has been argued on many fronts that empirical psychology requires a richer account of similarity. In particular, accounts of similarity must take into consideration the subject’s ‘theory’ about those objects taken to be similar.18 Of particular concern to the model just outlined is the fact that people seem to be psychological essentialists. That is, [P]eople act as if their concepts contain essence placeholders that are filled with ‘theories’ about what the corresponding entities are … [and which] provide or embody causal linkages to more superficial properties. (Medin and Ortony 1989, 186). Thus it seems that whilst kinds may be represented as prototypical configurations of observable features, kinds are identified through the possession of a shared ‘essence’. It is because of this ‘essence’ or ‘nature’ that a member of a given kind is taken to possess the propensity to manifest the features by which the kind is represented and members of the kind are recognised. Insofar as recent psychology requires a richer notion of similarity that incorporates a subject’s theories about those objects that are taken to be similar, then it seems that an alternative account of our credibility judgements is needed. To assess a testimony as credible is, paradigmatically, to explain the speaker’s utterance in terms of his possessing the intention to share his knowledge. Judging the credibility of testimony therefore involves some explanation of a speaker’s behaviour. Recent psychology suggests that our explanations of behaviour tend to proceed as follows. First, as it is an individual’s behaving as he did that is the explanandum, this behaviour is salient rather than the situational background.19 Second, the individual’s behaviour is taken to 13 Synthese, volume 131, no. 3, June 2002, 353 – 370. be representative: we take a subject to have acted as he did because of a stable disposition to act in this way.20 Third, in revealing these dispositions the subject is judged to possess a certain character or nature and the subject’s behaviour is then explained in terms of the theory possessed concerning persons with this character. Thus it seems that our observation of contextual details allows a judgement of the speaker’s character. Whether we take the testimony to be credible will then hinge on the theory we possess concerning persons of this nature. For instance, to return to the original example, our observation that the passer-by seemed well disposed, smiled, repeated her directions and did not seem in a hurry, seem relevant because these details seem to allow a judgement of the passer-by’s character. Insofar as these observations suggest an honest helpful character, we should be disposed to acceptance. Call the first account of our contextual reasons the situational model; call the second account of our contextual reasons the character model. Like the situational model, the character model equally shows how our reasons for accepting testimony, or otherwise, can be (1) contextual, (2) intuitive and (3) learnt. As with the situational model, our reasons can be contextual because their ground can be the identification of certain contextual details as evidence for the acceptability of the testimony, or otherwise. A judgement of credibility is intuitive because the audience need not occurrently remember past encounters with similar characters. These occasions rather supply those configurations of character-attributes needed in order to form a theory of this character and to determine the prototype by means of which such characters are recognised. Moreover, an audience’s ability to assess the credibility of testimony 14 Synthese, volume 131, no. 3, June 2002, 353 – 370. could be learnt as our experience of testimony could ground our knowledge human wiles.21 My purpose is to provide an account of our reasons for responding to testimony as we do. It is plausible to suppose that, in this regard, our observations and beliefs about the testimonial situation and the character of the testifier are both relevant. Thus, I shall not attempt to arbitrate between these models, but assume without further ado that there is some descriptive truth to both. As classes of reason, situational and characterbased reasons are similar to and differ from our prior reasons in the following respects. They are similar in that a given testimony is characterised as a certain type and a judgement of credibility is a consequence of this judgement of type. They differ in that for situational and character-based reasons testimonial types are broadly individuated as credible, non-credible and as credible as not. The given testimony is then identified as belonging to a type on the basis of contextual features, either of the testimonial situation or the testifier. By contrast in the case of our prior reasons, types of testimony may be quite narrowly characterised and a given testimony may identified as belonging to a type prior to its encounter. There remains one more important class of reason governing our response to testimony. 1.3. Existing Beliefs The general beliefs we have about testimonial types, our observation of the communicative situation and our observation of the character of the speaker can provide our reasons for accepting, rejecting or suspending judgement in testimony. So too can our beliefs (1) as to the truth of the proposition expressed and (2) as to the cost of our accepting a falsehood. 15 Synthese, volume 131, no. 3, June 2002, 353 – 370. Our disposition to accept testimony or otherwise can depend on our beliefs as to the truth of the proposition expressed. If, for instance, we already know what the speaker says, then we will be disposed to accept his testimony even though this will not alter our knowledge. Conversely, if we believe that what the speaker says is very unlikely, then we will be disposed to reject his testimony. To give Hume’s example, given that he had no experience of sub-zero temperatures, it required very strong testimony to get the Indian Prince to believe that water freezes.22 In general, our beliefs pre-dispose us towards accepting, rejecting or suspending judgement in any proposition. We then weigh this predisposition to acceptance or otherwise against the testimony we have received.23 Our dispositions to accept testimony or otherwise can also depend on our beliefs as to the cost of accepting a falsehood. Where the cost to the speaker is high, we should be disposed towards acceptance. Within institutional scientific contexts the cost to speakers of fabricating results is such that there is a powerful constraint on speakers to be trustworthy. Thus we judge that the scientific testimony of scientists, as a type, tends to be credible.24 Where the cost to ourselves of accepting a falsehood is high, we should be disposed towards rejection or suspension of judgement. Our disposition to accept, reject or suspend judgement in testimony can be explained in terms of our background beliefs. A complete explanation might have to refer to (1) our general beliefs about testimonial types; (2) our judgement of the testimonial situation; (3) our assessment of the speaker’s character and (4) our beliefs as to the truth of the proposition expressed and the cost of our accepting a falsehood. These 16 Synthese, volume 131, no. 3, June 2002, 353 – 370. background beliefs articulate our implicit reasons for believing or disbelieving a proposition expressed by intelligible testimony. Our response to testimony is rationally sophisticated: when we accept testimony, we usually have reasons for doing so. The claim that we usually have reasons for accepting testimony — claim (R) — embodies little more than recognition of this rational sophistication. 2. Evaluating Our Reasons for Acceptance When we accept testimony we usually have reasons for doing so. However, a subject’s reasons for belief need not determine that the proposition believed is probably true. What we take to be evidence need not in fact be evidence. Thus, to evaluate our reasons for acceptance is to determine whether what we take to be evidence tends to be so; it is to determine whether our reasons for belief are truthconducive. To evaluate our reasons for acceptance each class of reason identified above must be considered independently. If the dispositions engendered by a class lead, on average, to true belief, then the practice of accepting testimony, or otherwise, on the basis of such reasons is truth conducive. 2.1. Prior Presumptions Our disposition to accept a given testimony can be partially explained by our background reasoning, ‘this testimony is a certain type, testimonies of this type are credible therefore this testimony is credible’. Our dispositions to reject or suspend judgement in a given testimony may be similarly explained. These statistical syllogisms clearly articulate justifying arguments. In the limiting case, the stated inference is deductive. The premises of a statistical syllogism provide evidence for the 17 Synthese, volume 131, no. 3, June 2002, 353 – 370. conclusion when true. The major premise in our acceptance, or otherwise, of testimony concerns the credibility of a certain type of testimony. The premise ‘type x testimony is credible’ is true only if type x testimony is probably true. This raises the question of whether or not we are in a position to know such statistical generalisations and how we come to form type credibility beliefs. One possibility is that our type credibility beliefs are in fact beliefs as to the probable truth of types of testimony, formed because of our observing the truth of testimony. However, if this is the case there is little reason for thinking our type credibility beliefs are true because, to quote Coady, “it seems absurd to suggest that, individually, we have done anything like the amount of fieldwork … require[d].”25 A psychologically more plausible possibility is that type credibility beliefs are based upon our successful history of dealing with testimony. After all, our history of acceptance has largely been a history of knowledge acquisition. Insofar as we have acquired knowledge rather than false belief and our doxastic responses are at least a partial consequence of our type credibility beliefs, then this past success should ensure that our type credibility beliefs are largely true. This is not because we have confirmed through observation that our history of dealing with testimony has been successful, but simply because as a matter of fact it has been so. If our type credibility beliefs were largely false, then past success would not be what in fact it is. Thus our type credibility beliefs must be largely true and if this is the case, then as the major premise of a statistical syllogism, these beliefs provide evidence for the truth or falsity of testimony.26 18 Synthese, volume 131, no. 3, June 2002, 353 – 370. 2.2. Contextual Reasons — the situational model Our disposition to accept a given testimony can be partially explained by our implicitly judging this testimony similar, in its relevant contextual details, to those hitherto accepted. Being relevantly similar to the class of accepted testimony is evidence for the truth of a given testimony if and only if two conditions are satisfied. First, the class of accepted testimony is largely true. Second, it is not purely accidental that the contextual details defining the similarity relation define this relation to a class of testimony that is largely true. In like manner, our disposition to reject a given testimony can be explained by our implicitly judging it similar to testimonies hitherto rejected. And the existence of this similarity relation is evidence for the falsity of the testimony if and only if most rejected testimonies have been false and it is not accidental that this testimony is similar to this class of mainly false testimony. Our disposition to suspend judgement is to be explained in terms of these two cases. Given the present descriptive, non-sceptical epistemological context I assume that the first conditions are satisfied. I assume that we usually acquire knowledge rather than false belief through testimony. This implies that that those testimonies we accept are true more often than not. I assume that the corollary holds and those testimonies we reject are false more often than not. The difficulty lies with the second two conditions. I focus on acceptance. To form a judgement of credibility is simultaneously to explain and evaluate a speaker’s testimony. Thus, a judgement of credibility involves an explanation of behaviour. Two ways of explaining the behaviour may be distinguished. Dispositional explanations treat the enduring dispositions of the subject — his ‘character’ or 19 Synthese, volume 131, no. 3, June 2002, 353 – 370. ‘nature’ — as the cause of behaviour. Situational explanations treat the situational pressures — where ‘situations’ are characterised at all levels, physical, social and personal — as the cause of behaviour. Accurate explanations tend to be situational, (see below). This is because most people tend to respond similarly to comparable situational pressures. If it is then assumed that the contextual details that underpin our acceptance provide an adequate characterisation of the testimonial situation, then it does not seem purely accidental that an accepted testimony is probably true. It has been assumed that the class of accepted testimony is largely true and it may be assumed that in similar situations speakers will manifest similar behaviour. The truthfulness of their utterances will be an aspect of this. 2.3. Contextual Reasons — the character model Our disposition to accept a given testimony can be partially explained by our implicit assessment of the speaker’s character. This judgement in conjunction with the ‘theory’ we possess concerning persons of such nature determines whether we take the testimony to be credible or otherwise. A judgement of character can be evidence for the truth of a speaker’s testimony; it is taken to be so by the law courts. However, one may presume that a judgement of character is evidence only if such judgements form part of behavioural explanations that tend to be accurate. This condition is not satisfied. Explanations of behaviour in terms of an assessment of character both under-estimate situational factors and over-estimate the stability of an individual’s behaviour across situations. This conjunction Ross and Anderson label the fundamental attribution error.27 The error is ‘fundamental’ because accurate explanations should exhibit the converse bias. 20 Synthese, volume 131, no. 3, June 2002, 353 – 370. 2.4. Existing Beliefs Our disposition to accept, reject or suspend judgement in a given testimony can be partially explained as a consequence of our beliefs as to the truth of the proposition expressed and beliefs as to the cost of accepting a falsehood. Our beliefs as to the truth of the proposition expressed and the cost to the speaker of our accepting a falsehood would be evidence for the truth of a testimony if they were knowledge.28 Given the assumption that most of our beliefs are not merely true but constitute knowledge, I assume that this condition is largely satisfied. Our beliefs as to the cost to ourselves of our accepting a falsehood are not evidence for the truth of a testimony; rather they determine the level of reasons required for acceptance to be doxastically responsible. A complete explanation of our doxastic responses to intelligible testimony could have to refer to (1) our general beliefs about testimonial types; (2) our judgement of the testimonial situation; (3) our assessment of the speaker’s character and (4) our other relevant beliefs. Insofar as (1), (2) and (4) articulate the reasons for our doxastic response to testimony, our reasons for acceptance should provide evidence for the truth of the proposition believed and our reasons for rejection should provide evidence for the falsity of the proposition disbelieved. We should suspend belief when there is equal evidence for and against the proposition expressed. Thus, if our reasons for acceptance were only based on (1), (2) and (4), one could unequivocally conclude that our reasons for acceptance provide evidence for our testimonial beliefs. However, where acceptance is based on (3), our assessment of the speaker’s character, our reasons might explain the cause of our belief but should not, ordinarily, provide evidence for this belief. As such the strongest conclusion that can be drawn is that our 21 Synthese, volume 131, no. 3, June 2002, 353 – 370. reasons for acceptance can provide evidence for the testimonial beliefs we form. This is all that (E) states. I should note again that I remain neutral as to what justifies the propositions we come to believe through acceptance. 3. Conclusion Credulity, Hume claims, is one of man’s greatest weaknesses. Many would counter that our acceptance of testimony is rarely other than credulous. Given this assumption, our background beliefs are given a largely negative filtering role: we are disposed to accept testimony unless our beliefs provide us with a reason for rejection. This is the psychological account of testimony that Audi finds in Reid. [Our background beliefs] play a mainly filtering role: they prevent our believing testimony that does not ‘pass’, for instance because it seems insincere; but if no such difficulty strikes us, we ‘just believe’ (non-inferentially) what is attested. These filtering beliefs are like a trap door that shuts only if triggered; its normal position is open, but it stays in readiness to block what should not enter. (Audi 1997), 406). This psychological account, I have argued, is premised on the false descriptive claim that we largely lack reasons — in the sense of propositions believed — for acceptance. However, if those propositions we believe — our background beliefs — provide the reasons for our responses to testimony, then a different psychological account is suggested. This is the account I have presented in this paper. When we accept testimony, it is usually because we accept other propositions that we take to be evidence for the testimony. The contrary assumption that our acceptance of testimony is usually credulous is an assumption has dominated the epistemology of 22 Synthese, volume 131, no. 3, June 2002, 353 – 370. testimony and figured prominently in arguments for anti-reductive theories of testimonial justification. I hope that this account of the reasons underlying our response to testimony at once shows this assumption to be false and undermines its plausibility. Credulity is a weakness not a psychological default.29 Paul Faulkner The Department of Philosophy, University College London, 1 (Dummett 1993, 420). 2 (Coady 1973) and (Coady 1992), Ch.3. 3 (Hardwig 1985). 4 For example, Webb endorses (5) and claims, “In only the rarest instances do we reason at all before forming a belief based on what someone has told me.” (Webb 1993, 262). 5 In addition, premise (3) of the Reidian argument is arguably false. One could assert that whilst we are not justified in accepting testimony without reason, it is not our reasons for acceptance that ordinarily justify our testimonial beliefs. See (Faulkner 2000). 6 This analogy is owed to (Adler 1994, 272). 7 These theories are commonly known as ‘reductionism’ and ‘anti-reductionism’ respectively, after (Coady 1973). See (Fricker 1995). 8 Similarly, Adler notes, “we draw all manner of distinctions among classes of potential informants. Car salesman are not as trustworthy (within their role) as librarians; professional journals are more reliable than the popular press” (Adler 1994, 270). 9 It is plausible to identify content as a particular topic because for any proposition that we are disposed to accept one can imagine many fractionally different propositions that we should be equally disposed to accept. 23 Synthese, volume 131, no. 3, June 2002, 353 – 370. 10 I have not included the combinations general source and particular topic and particular source and particular topic because I take these to be irrelevant subsets of (5) and (6) respectively. 11 (Burge 1993, 469). 12 Similar points are made by (Adler 1994). 13 Continuing the analogy with understanding implicature, Adler claims, “So long as our beliefs relevant to evaluating testimony are no harder to process than other contextual information readily exploited in conversation, their use cannot be objected to.” (Adler 1994, 273). 14 For an example, see (Audi 1997, 407). 15 “The principle way that cognitive science can contribute to epistemology is to identify basic belief forming processes.” (Goldman 1986, 181). 16 The proto-typicality of an object with respect to a class Tversky defines as a linear combination of the measures of the features of that object that are shared with the other members of that class and the features that are not shared with the other members of that class. A proto-type maximises prototypicality. The state of empirical psychology in 1977 was such that Tversky could claim, “Research has demonstrated the importance of prototypicality or representativeness to perceptual learning, inductive inference, semantic memory, and the formation of categories.” (Tversky 1977, 347). 17 An analogy suggests itself. On a Davidsonian theory of meaning, the meaning of a sentence depends on the meaning of the words that compose it and the meaning of these words depends on what they contribute to the truth conditions (meaning) of all the sentences that contain them. Similarly, the credibility of a testimony depends on the circumstantial features that individuate it and the relevance of each of these features depends on all the observed credible testimonies that contain them. On this analogy judging the credibility of testimony from its circumstances is analogous to understanding a sentence. 18 Thus it seems that even as young children our inductive inferences are influenced more by causal notions than brute measures of similarity. See (Carey 1985) and (Gelman and Markman 1986). And recent work in cognitive and developmental psychology has employed the idea that theory change is a model of cognitive development. See (Gopnik and Meltzoff 1997). 19 Our inductive explanations and predictions are guided by the availability heuristic. (Tversky and Kahneman 1973). 24 Synthese, volume 131, no. 3, June 2002, 353 – 370. 20 In accord with psychological essentialism, our inductive explanations and predictions, further, are guided by a representativeness heuristic. (Kahneman and Tversky 1973). This seems a happy alliance. “Our conceptual and inferential tendencies jointly conspire, at least roughly, to carve nature at its joints and project the features of a kind which are essential to it. This pre-established harmony between the causal structure of the world and the conceptual structure and inferential structure of our minds produces reliable inductive inference.” (Kornblith 1993, 94). Kornblith offers an evolutionary explanation for this harmony. 21 This character model is close to the presentation of Hume’s theory of testimony found in (Faulkner 1998). 22 (Hume 1777, 113). 23 Owen claims that this balancing of evidence can be represented in terms of the probability calculus. (Owen 1987). 24 At least we should make this judgment if we were scientists. Root suggests that such sanctions do not govern communication with laypersons, see (Root 2001, 26). 25 (Coady 1992, 82). 26 Of course, this argument cannot be happily employed if one hoped to offer a reductive account of testimonial warrant. This is because its conclusion that our type credibility beliefs are largely true depends upon the assumption that it is knowledge that has been largely acquired from testimony. But if it were supposed that our type credibility beliefs underwrite our testimonial knowledge, then this argument should be epistemically circular. However, those who reject (E) should also reject reductive accounts of testimonial justification and, therefore, should find this use of the assumption is acceptable. For the reductionists the challenge is to either deny Coady’s point, through claiming that our observations are sufficient, or show how this epistemic circularity is not debilitating. 27 (Ross and Anderson 1982). A corollary is the tendency to abandon quickly the use of generalisations in explaining behaviour. If an individual’s dispositions are stable, any specific information should be diagnostic. Thus our psychological essentialism is a source of error in social judgements, see (Nisbett and Ross 1980) and contrast note 20. 28 Williamson would claim a belief is evidence if and only if it is knowledge. See (Williamson 1997). The sufficiency claim is less contentious. 25 Synthese, volume 131, no. 3, June 2002, 353 – 370. 29 Particular thanks are owed to Matthew Soteriou, Mike Martin and an anonymous referee. References Adler, J. 1994. 'Testimony, Trust and Knowing', The Journal of Philosophy 91 (5):264-75. Audi, R. 1997. 'The Place of Testimony in the Fabric of Knowledge and Justification', American Philosophical Quarterly 34 (4):405-422. Burge, T. 1993. 'Content Preservation', Philosophical Review 102 (4):457-488. Carey, S. 1985. Conceptual Change in Childhood. Edited by L. G. e. al, Learning, Development and Conceptual Change. Cambridge MA: A Bradford Book The MIT Press. Coady, C.A.J. 1973. 'Testimony and Observation', American Philosophical Quarterly 10 (2):149-155. Coady, C.A.J. 1992. Testimony: A Philosophical Study. First ed. Oxford: Clarendon Press. Dummett, M. 1993. 'Testimony and Memory', In The Seas of Language, edited by M. Dummett. Oxford: Clarendon Press. Faulkner, P. 1998. 'Hume's Reductionist Theory of Testimony', Pacific Philosophical Quarterly 79 (4):302-313. Faulkner, P. 2000. 'The Social Character of Testimonial Warrant', Journal of Philosophy 97 (11):581-601. Fricker, E. 1995. 'Telling and Trusting: Reductionism and Anti-Reductionism in the Epistemology of Testimony', MIND 104 (414):393-411. 26 Synthese, volume 131, no. 3, June 2002, 353 – 370. Gelman, S.A., and E.M. Markman. 1986. 'Categories and induction in young children', Cognition 23:183-209. Goldman, A. 1986. Epistemology and Cognition. Cambridge MA: Harvard University Press. Gopnik, A., and A. N. Meltzoff. 1997. Words, Thoughts, and Theories. Edited by L. Gleitman, S. Carey, E. Newport and E. Spelke, Learning, Development and Conceptual Change. Cambridge, MA.: A Bradford Book - The MIT Press. Hardwig, J. 1985. 'Epistemic Dependence', The Journal of Philosophy 82 (7):335-349. Hume, D. 1777. Enquiries Concerning Human Understanding and Concerning the Principles of Morals. 3rd Ed. revised by P.H.Nidditch ed. Oxford: Clarendon Press. Kahneman, D., and A. Tversky. 1973. 'On the Psychology of Prediction', Psychological Review 80:237-251. Kornblith, H. 1993. Inductive Inference and its Natural Ground: An Essay in Naturalistic Epistemology. Cambridge MA: The MIT Press. Medin, D.L., and A. Ortony. 1989. 'Psychological Essentialism', In Similarity and Analogical Reasoning, edited by S. Vosniadou. Cambridge: Cambridge University Press. Nisbett, R.E., and L. Ross. 1980. Human Inference: Strategies and Shortcomings of Social Judgement. Englewood Cliffs, N.J.: Prentice-Hall. Owen, D. 1987. 'Hume versus Price on Miracles and Prior Probabilities', The Philosophical Quarterly 37:187-202. Reid, T. 1764. 'An Inquiry into the Mind on the Principles of Common Sense', In The Works of Thomas Reid, edited by W. H. Bart. Edinburgh: Maclachlan & Stewart. 27 Synthese, volume 131, no. 3, June 2002, 353 – 370. Root, M. 2001. 'Hume on the virtues of testimony', American Philosophical Quarterly 38 (1):19-35. Ross, L., and C. Anderson. 1982. 'Shortcomings in the attribution process: on the origins and maintenance of erroneous social assessments', In Judgement Under Uncertainty: Heuristics and Biases, edited by D. Kahneman, P. Slovic and A. Tversky. Cambridge: Cambridge University Press. Tversky, A. 1977. 'Features of Similarity', Psychological Review 84 (4):327-352. Tversky, A., and D. Kahneman. 1973. Availability: A heuristic for judging frequency and probability. Judgement Under Uncertainty: Heuristics and Bisases, 163178. Webb, M.O. 1993. 'Why I know about as much as you: a reply to Hardwig.', The Journal of Philosophy 90 (5):260-70. Williamson, T. 1997. 'Knowledge as Evidence', MIND 106:717-741. 28