Discussion and Q/A

10 Testimony and Epistemic Autonomy

Elizabeth Fricker

1 . D I V I S I O N O F E P I S T E M I C L A B O U R V E R S U S T H E I D E A L O F I N D I V I D U A L E P I S T E M I C A U TO N O M Y

A reference point in philosophical investigation of knowledge from testimony is the ideal of the ‘autonomous knower’. This ideal type relies on no one else for any of her knowledge. Thus she takes no one else’s word for anything, but accepts only what she has found out for herself, relying only on her own cognitive faculties and investigative and inferential powers. Descartes explicitly espoused this ideal, and method, in his Meditations (Descartes 1641). Locke equally rejected ‘other men’s opinions floating in one’s brain’ as not constituting knowledge (Locke 1690). The wholly autonomous knower will not accept any proposition, unless she herself possesses the evidence establishing it. Thus she will not accept anything on the basis of another’s word for it, even when she has evidence of their trustworthiness on the topic in question.

Such extreme purism restricts how much one can come to know very severely. We humans are essentially social creatures, and it is not clear that we do or could possess any knowledge at all which is not in some way, perhaps obliquely, dependent on testimony. How exactly does the system of empirical belief — hopefully knowledge — of each of us depend on others’ testimony? There is cer- tainly massive causal reliance on testimony in the process by which each of us develops into a language-user and thinker, ‘grows into possession of a world’.! The initial stages of language acquisition by a child inevitably occur through a

Earlier versions of this paper were given at a workshop on ‘Testimony, Trust and Action’ in King’s College Cambridge in September 2003, at a conference on ‘Moral Testimony’ in the Philosophy Department at Birmingham University in March 2004, and at a conference at the Inter-University Centre in Dbrovnik, Croatia, in May 2005. I received very useful comments from audiences at these events, in light of which I corrected various errors. I am also very grateful to both John Hawthorne and Stephen Schiffer for valuable comments and discussion on an earlier draft. The research for this paper was done between January and June 2002, during a period of leave funded by my employers, Magdalen College and Oxford University, and by a Fellowship from the Mind Association. My thanks for their support.

226 Testimony and Epistemic Autonomy

process of simple trust" in its teachers — parents and other carers. In this cog- nitive developmental process learning meanings is not separable from coming to grasp and accept our shared basic world picture, the common-sense theory which structures and frames our empirical thought. There is, for instance, no distinction to be drawn between learning the meanings of ‘chair’, and ‘horse’, and ‘jump’, and ‘cook’, and learning about chairs, and horses, and jumping, and cooking.#

The fact that each of us is causally reliant on others’ testimony in the historical process by which she acquires her system of concepts and beliefs does not entail that, once adult, each of us remains epistemically dependent on testimony for her empirical knowledge. Perhaps each of us can afterwards push away the ladder of trust in others, up which she has climbed into possession of a world. Beliefs which were first acquired through a process involving simple trust in testimony, and were initially epistemically based on testimony (as we may say once core normative epistemic concepts become applicable to the developing child, viz. when she becomes a thinker capable of epistemic self-criticism), may later acquire an alternative basis. It may be that beliefs from the epistemic source of percep- tion, linked by memory and extended by inference, can take over, together with support from inference to the best explanation and broader coherence. Suppose one could, once epistemically matured, thus push away the ladder of testimony, retaining only the portion of one’s beliefs which remain epistemically supported without reliance on it. In maintaining the ideal, one would then be restricted to what one learns from one’s own senses and preserves in memory, plus whatever one can get to by use of one’s own inferential powers from that base — with a ban on even reasoned, empirically backed trust in the word of others!

There is reason to doubt that one can in that way eliminate all epistemic dependence on testimony in one’s mature system of empirical belief, even if pre- pared drastically to prune it. For one to do so, her original epistemic dependence on testimony would have, everywhere, to be replaced by adequate support from other epistemic sources, or the belief in question dropped. Now of course it often happens in particular cases that one first learns of something through another’s testimony, and then is later able to confirm it for oneself through perception, perhaps combined with memory and inference. My daughter tells me her new teacher wears glasses; later I see the teacher for myself. The weather forecaster on Tuesday predicts that it will rain on Wednesday; Wednesday proves wet. Facts about a foreign country known to one at first only through travel literature and friends’ reports are confirmed by perception, when one travels there oneself. In these and countless similar cases one later gets first-hand perceptual evidence of what one first believed on testimony. In such cases contrary perceptual evidence would decisively falsify the testimony.

There are other ways, less direct but no less powerful, in which alternative grounds for belief can grow strong enough to take over the support of a belief originally acquired from and based on testimony. Inference to the best explanation and explanatory coherence more broadly can take over the support

Elizabeth Fricker 227

of many beliefs originally based solely on others’ testimony. It is plausible, for instance, that one’s implicit beliefs about what the words of one’s language mean no longer rest on the past trusted testimony through which they were learned. One’s linguistic interactions with others would not run as smoothly as they do, if one’s first teachers had deceived one! (See Adler 1994; Lyons 1997.)

Nonetheless the role of past trusted testimony in the system of empirical belief of each of us is fundamental, because it has shaped the conceptual frame with- in which current individual perceptions are made — how the sensory given is conceptualized to yield perceptual experience and belief.$ Thus, in our three examples above, while it is true that subsequent contrary perception would refute the earlier testimony, these perceptions are themselves subtly dependent on a framework of concepts shaped in part by earlier trust in testimony. I see the teacher’s reading glasses, but that she is my daughter’s teacher I know only through a set of background beliefs in which testimony is inextricably involved. When it rains on Wednesday I see and hear the rain; but my knowledge that it is Wednesday is testimony-infected, and that very concept and its application is one constituted by human consensus involving testimony. Similarly, when I vis- it Australia for the first time, in one way I gain personal confirmation of what I had previously known of only through testimony; but my knowledge that I am in Australia at all depends on testimony in multifarious and hard to pin down ways: initially I knew where my flight landed only through trusting the testimony of travel agent and airline personnel, and though the evidence of roadsigns and so forth may take over, these are all put there by human agency, and constitute a kind of testimony. Moreover, the controlling idea in terms of which I con- ceptualize and slot in all my own personal experiences — of the spherical planet earth with its land masses and seas, its countries, nations, and other geopolitical institutions, its history and prehistory — was acquired from testimony.

This brief sketch has shown how the epistemic dependence on testimony in the beliefs of each of us socially embedded twenty-first-century individuals is subtle and widespread, if not all-pervading. A more sustained enquiry is needed to see whether it could in principle be eliminated by such an individual while leaving her any beliefs at all. Worse yet for the would-be epistemic autonome: it may not be sharply determinable at all, whether and when freedom from oblique epi- stemic dependence on testimony is achieved, since isolating the contribution of testimony from that of other sources of support, in a system of belief with rich explanatory coherence, is not a clear-cut matter. It is at any rate certain that, in order to live up to the ideal of individual epistemic autonomy, a very great deal of what is believed by a normal member of a modern society, with its exten- ded division of epistemic labour, would have to be bracketed, given up — most of geography, history, the natural and social sciences including medicine, and so forth.

Giving all that up is no more a serious practical possibility than living out the life of a more thoroughgoing sceptic — one who doubts even the evidence of the

228 Testimony and Epistemic Autonomy

senses as indicators of a perceptible external world. Who would really give up the fruits of the sciences including all technology, medicine, dentistry, foreign travel, as well as historical understanding and knowledge — and so on? The epistemic- ally autonomous individual could not trust an electrician to wire her (self-built!) house for her, since she would not accept his testimony about what he was going to do, and that it would work safely; nor her doctor to prescribe medicines; nor would she try ski-ing because her friends (she could not have many!) told her it was fun.

We have found that testimony, for each of us in our modern social and epi- stemic predicament in which division of epistemic labour along with other sorts is the rule, is an essential source of empirical grounding for her beliefs about the world she finds herself in, and her own place in it. This system of empirically based belief is richly coherent, including its ability to explain its own sources.

Notice that the trust in testimony of which I am stressing the ubiquity need not however be given uncritically, without empirical grounds. I have argued else- where that a mature recipient of testimony need and should not trust another’s word without adequate empirically based warrant to do so. We have seen that it is impractical to live up to the supposed ideal of individual epistemic autonomy. One cannot live in a modern scientifically and technologically sophisticated soci- ety, nor have any social life at all, without trusting others in almost one’s every action. But this is not to say that one’s trust in the vast heritage of knowledge and know-how built up from others’ investigations, expertise, and experience must be blind — uncritical and undiscriminating. Good empirical grounds for taking a fresh instance of testimony to be sincere and reliable — or for being distrustful of it — are often to be had; and inference to the best explanation and rich coherence within one’s accumulated system of belief can support, ex post, one’s reliance on some earlier pieces of testimony, while equally discrediting others. (I here barely touch on issues which need much fuller discussion. See Adler 1994; Coady 1992; Fricker 1994, 2002, 2005.)

Still, one who trusts testimony discriminatingly, only when she has an adequate empirical basis to do so, and whose past trust is now vindicated through support from explanatory coherence, is yet dependent on testimony in her beliefs, and actions based on them. If I take others’ word for things, I extend my knowledge far beyond the range I could achieve on my own, but by this very fact I am not epistemically autonomous. I believe many things for which I personally do not possess the evidence, and my believing is premised on the supposition that some other person or set of persons jointly has, or had, access to that evidence, and evaluated it correctly. (These points are expanded in Section 4, below.)

We have seen that even if one could, by a heroic effort of epistemic recon- struction, push the ladder of past trusted testimony away, the project of attaining and maintaining the ideal of complete individual epistemic autonomy is not an attractive or feasible one — one would forgo too much! Is there reason to regret this? In this paper I shall respond to this question by addressing a closely

Elizabeth Fricker 229

related one: In what way exactly is one’s epistemic self-governance necessarily compromised, by one’s practically inevitable dependence on others’ testimony? Putting the question the other way about: In what way, and to what extent, can one maintain one’s epistemic self-governance despite one’s inevitable reliance on others’ testimony, and the technological fruits of others’ knowledge and expert- ise, in almost every area of one’s life?% I will first prepare the ground to address this question by considering another: In what circumstances, and on what topics, may one properly accept and learn from another’s testimony? In developing an answer to this question we will find material relevant to answering our first one.

2 . T H E C I RC U M S TA N C E S A N D TO P I C S O F P RO P E R A C C E P TA N C E O F T E S T I M O N Y

In what circumstances and on what topics may one person with epistemic pro- priety accept the testimony of another and by so doing learn, acquire know- ledge, from her? Conversely: What are the circumstances in which, and topics on which, one person may tell something to an audience, thereby expressing her knowledge, and reasonably intend and expect to be believed, trusted — to have her word on the matter accepted? These are distinct questions, but the mutuality of the illocutionary act of telling means that their answers will coincide, where the expectation of being trusted is well founded.&

2.i A Precondition for Testimonial Spreading of Knowledge As a preliminary I note a precondition for testimony to be given and received at all. For an act of telling to succeed there must be mutual understanding. A mes- sage must be got across and accepted. So there must be a proposition which the teller intends by her action to present as true, and this must be identical with the one grasped by her audience as so presented, and accepted by her.' This does not, in itself, entail a shared language. Nor does it entail that, when a shared language is employed, the message conveyed is what the speech act literally means — a pro- position which the sentence used is conventionally apt to convey, and is plausibly interpreted as specifying on that occasion.( But communication of a message is most commonly effected by use of sentences of a shared language in accord- ance with the constraints of their literal meaning, to make explicit assertions. The occasions on which other media are feasible vehicles are relatively few and far between. This being so, the spreading of knowledge by means of testimony, or something like it, is possible to any significant extent only when there is a shared language, and mutual understanding of speech acts between speaker and audience.

Our common-sense view — which I shall not question here — is that we do indeed share a language, including its semantics, with our co-speakers to an extent sufficient for mutual understanding and successful communication. But one can

230 Testimony and Epistemic Autonomy

learn from another’s testimony only when one does not already know what she tells one. Hence, if difference of opinion regarding the truth value of some sen- tence S entailed difference of meaning attached to S, there would be no learning from others regarding the proposition expressed by S.) This observation is, I think, enough to discredit extreme ‘holistic’ theories about the fixation of mean- ing (already implausible). But there are certain areas of discourse where disagree- ment might be thought to undermine the supposition that meaning is shared. If this suspicion were confirmed, then learning from testimony in the strict sense (as opposed to changing one’s language to conform more with others) would be shown to be impossible in these areas.

Difference of opinion due to ignorance, where one party simply lacks firm belief either way on the topic, is unproblematic. Equally, disagreement in the strong sense of conflict of opinion is unproblematic, when its origin is traceable to different access to evidence. (In such a case pooling of evidence will produce convergence of opinions.) It is when disagreement in judgement persists despite similar access to evidence that, in certain areas, the supposition of shared mean- ing may be threatened. If there are certain subject matters where disagreement of judgement in response to the same evidence entails difference of meaning, then there can be no learning from testimony in the strong sense of deferring to others’ judgement, letting it override one’s own, on those topics.

It is a point familiar from a certain style of philosophical account of how meanings and beliefs are simultaneously attributed to someone, that a tentat- ive interpretation of an utterance which yields a difference of opinion between interpretee and interpreter not explained by differential access to evidence, is thereby thrown into doubt. Other aspects of the total interpretation being equal, it is more ‘charitable’ hence a priori better warranted, to interpret the other as meaning something else (see Davidson 1984). But other aspects may well not be equal, and so the defeat of the assumption of shared meaning is generally not instantaneous. We all have had futile arguments — ‘‘It’s green’’; ‘‘No it’s not, it’s yellow’’ — where the suspicion lurks that there is not really a substantial matter at issue, rather than a non-concordance of linguistic usage at its vague edges, com- pounded perhaps by a pig-headed refusal of the out-of-line debater to adjust her usage. Equally we all have had arguments where it seems certain that there is a substantial, not merely a semantic matter at stake — ‘‘It’s unfair that you let Juli- an go in the front of the car, but you never let me’’, although progress towards agreement may seem no less hard to achieve.

Colour concepts, and other simple perceptually applied concepts; plus moral and also aesthetic concepts, are ones where sorting out substantive from merely linguistic disagreement on particular occasions is difficult; no less difficult than giving an account of how the precise content of those concepts is fixed. In these and some other cases, there really may be no way to distinguish between defer- ring to others’ judgement about the application of an already shared concept, and adjusting one’s concept. (We remarked earlier that, in one’s initial acquisition of

Elizabeth Fricker 231

one’s language, there is no sharp line between acquiring new information, beliefs, about things one already has a concept of; and acquiring those concepts.)

These considerations will be relevant in a full investigation of the possibilit- ies for learning from testimony about these topics. It will be important to bear them in mind, when considering to what extent one can defer to others’ judge- ment on moral and aesthetic matters. There may prove to be limitations on this grounded in considerations about meaning, for aesthetic judgement in particular, I suspect.!* Having noted this, I will not explore it more fully here.

2.ii A Principle Concerning Deferential Acceptance With these points about the need for shared meaning made, we can proceed with the main positive idea. We can formulate a general principle:

Testimony Deferential Acceptance Principle (TDAP 1): For one properly to accept that P on the basis of trust in another’s testimony that P — her word that P!! — requires that she be epistemically well enough placed with respect to P so that were she to have, or make a judgement to form a conscious belief regarding whether P, her belief would almost certainly be knowledge;!" and that she be better epistemically placed with respect to P than oneself; and that one recognize these things to be so.

TDAP1 specifies a condition necessary for epistemically proper trusting acceptance of another’s testimony on some topic. It is not sufficient, because while the hearer’s cognizance of the testifier’s strong epistemic position vis-à- vis the topic makes it rational for her, other things being equal, confidently to expect the testifier’s judgement about the matter in question to be correct — to deem her competent about the matter in question — TDAP1 does not speak to the question of the testifier’s sincerity. As I have argued elsewhere, the overall trustworthiness of a speaker’s testimony breaks down into these two quite separate components. In this investigation I concentrate on the circumstances in which deferential acceptance of another’s judgement, as expressed in her sincere testimony, is epistemically proper. Thus, having noted the need for adequate warrant to believe the speaker sincere, I put further consideration of sincerity aside, assuming in what follows that insincerity is not an issue in whether to trust the other’s testimony.!#

The matter of sincerity is one reason why TDAP1 specifies only a necessary, not a sufficient condition, for epistemically proper deference to another’s testimony. The only other reason I can think of why other things would not be equal, regarding the hearer’s expectation of correctness of the testifier’s expressed judgement, is if she were also aware of significant contrary testimony. Contrary testimony will be epistemically significant if it either comes from another equally well-qualified expert; or, in some cases, if it is from many mutually independent sources, albeit not especially expert ones.!$ A more refined

232 Testimony and Epistemic Autonomy

condition incorporating these two factors, which is normatively both necessary and sufficient for deferential acceptance is:

TDAP 2: One properly accepts that P on the basis of trust in another’s testimony that P — her word that P — just if she speaks sincerely, and she is epistemically well enough placed with respect to P so that were she to have, or make a judgement to form, a conscious belief regarding whether P, her belief would almost certainly be knowledge; and she is better epistemically placed with respect to P than oneself; and one recognizes all these things to be so; and one is not aware of significant contrary testimony regarding P.

TDAP 2 specifies when it is proper to accept another’s testimony that P out- right. There will also be situations where neither party is in a position to make a knowledgeable judgement as to whether P, but one is better epistemically placed than the other. In these cases it will be epistemically rational for the worse-placed person to defer to the other’s opinion, while falling short of taking her utterance as an expression of knowledge; hence forming only a tentative belief regarding P. This is required when, for instance, an informed decision about how to act is urgently needed. There may be some topics for which this situation is the rule — that is, where knowledge as opposed to more or less well-grounded spec- ulation is very hard to come by.!% It remains true that one should not accept outright another’s testimony that P, unless one reasonably believes her to be so placed as to (almost certainly) form knowledgeable belief regarding P. Hence TDAP2 is the correct general principle governing the outright acceptance of another’s testimony.

An explanatory comment is needed on the role of the complementary ‘intern- al’ and ‘external’ components of epistemic propriety in TDAP1 and 2. I have formulated TDAP1 and 2 incorporating both internal and external components, because I am concerned to describe what happens when things go right, and thus how knowledge is spread by means of testimony. Externally, things are going right when the testifier speaks from her expertise-generated knowledge, and is sincere. But epistemic rationality has a key internal component: it is not rational to accept unquestioningly the testimony of an expert who, so far as one knows, is no such thing. And, though not all-in epistemically proper, it is subjectively rational and epistemically blame-free to accept another’s testimony, when one falsely but justifiedly believes her to be an expert about the topic, being deceived about this through no fault of one’s own. I am against purely exernalist accounts of when acceptance of testimony is epistemically proper. These fail to incorporate the requirement that the subject maintain epistemic responsibility for her own beliefs. In Section 3 I spell out the implications, and means of satisfying, this requirement.!&

We may distinguish between a weak and a strong form of deference to anoth- er’s testimony:

Elizabeth Fricker 233

Weak Deferential Acceptance occurs when I form belief that P on the basis of trust in another’s testimony that P, when I myself have no firm pre- existing belief regarding P; nor would I form any firm belief regarding P, were I to consider the question whether P using only my current epistemic resources, apart from the current testimony to P. Strong Deferential Acceptance occurs when I let another’s trusted testimony regarding P override my own previous firm belief, or disposition to form a firm belief, regarding P.

The distinction between strong and weak deferential acceptance may or may not turn out to be important. First off, it seems that there could be subject matters where strong deferential acceptance is never epistemically appropriate, although weak deference can be. This fact may illuminate the nature of that sub- ject matter.

Whether for weak or strong deferential acceptance, it seems that TDAP2 is the correct normative principle: her sincerity not being in question, and my being aware of no significant contrary testimony, it is epistemically proper that I defer to another’s testimony in forming belief regarding P, or in overriding my own previous belief regarding P, just if I recognize that she is better epistemically placed than I am to determine whether P; and it is epistemically proper that I accept her testimony outright just if I recognize this, and also that she is so placed as to form (almost certainly) knowledgeable belief regarding P. We may intro- duce a thin and inclusive sense of ‘expert’ capturing this core normative necessary condition for deferential acceptance expressed in TDAP1 (which is also norm- atively sufficient, apart from the matters of sincerity and absence of significant contrary testimony):

S is an expert about P relative to H at t just if at t, S is epistemically well enough placed with respect to P so that were she to have, or make a judge- ment to form a conscious belief regarding whether P, her belief would almost certainly be knowledge; and she is better epistemically placed than H to determine whether P.

2.iii Bases of Expertise We can now explore the different possible bases of such relative epistemic expert- ness of S over H regarding some collection of propositions P comprising a subject matter W. In so doing we will be developing a description of the various cir- cumstances in which it is epistemically proper, when she knows them to obtain, for one person deferentially to accept another’s testimony regarding some sub- ject matter. The idea of someone’s being epistemically ‘well placed’ regarding P is used so far in a broad catch-all sense. We will now see how a variety of specific cir- cumstances may contribute to this. One is literally the spatio-temporal location of the person; another is particular skills and perceptual and cognitive equipment she possesses.

234 Testimony and Epistemic Autonomy

There are various kinds of situation in which it is obvious and unproblematic that S will at that time be epistemically expert relative to H regarding some sub- ject matter W.

Case One: W is an observable event or state of affairs, and S is or was at the time of its occurrence so positioned as to be able to observe it, whereas H was not.

Suppose, for instance, that Natalie went to the RadioHead concert in South Park, while I did not. Then — assuming she has normal observational and memory capacities — she knows quite a bit of what happened during it, and I can learn from her telling me about it. Without access to some such eyewitness account (written or spoken), I cannot know very much about what happened. Of course Natalie is not my only possible informant. And there may be a limited amount I can infer from other sources of evidence — walking there the next day I observe huge numbers of empty drinks cans scattered around, and see the stage being dismantled. But I cannot learn any detail except from testimony; and only hers is conveniently available.!'

Natalie is an expert on what happened at the concert, relative to me. Hence — if she is truthful — I can learn from her. But her expert status on the topic relative to me is highly accidental. It is based in a mere happenstance about our locations on one particular day, not on any more stable and intrinsic epistemic talent, skill, or base of knowledge that she possesses whereas I lack. Had I possessed a ticket and gone to the concert, while Natalie stayed at home, I would instead have been the expert vis-à-vis her. Such merely accidental and extrinsic expert status is often brief and transient. Emily is momentarily an expert relative to me about what is in the fridge; but only until I take a look for myself.

Case Two: Superior perceptual skill of S over H.

Now suppose Natalie and I are both at the concert, and are trying to make out what is happening on the stage from some distance. I am shortsighted, whereas Natalie has excellent distance vision. She reports ‘the supporting band is coming on’ and I accept her report, not being able to see anything specific for myself. Or I may think I can see something different, but I allow my visually based judgement to be overruled by hers, in the knowledge that she has better distance vision than me.!( This case is more interesting. Natalie is an expert relative to me about what is happening in the distance not because of an accidental difference in our loca- tions, but due to a superior epistemic skill she has relative to me, which is (in a relaxed sense) intrinsic, and fairly stable. She is not just accidentally better placed, that is spatio-temporally located, than me, regarding the topic; she is better epi- stemically equipped than me to make judgements of a certain kind — namely, judgements about events occurring in the visual perceptual distance. Let us say that she is not only currently an expert relative to me about the happenings on the stage, in the thin sense defined above; but that this is due to an epistemic expertise

Elizabeth Fricker 235

she possesses relative to me, regarding a range of matters-in-circumstances. Spe- cifically:

S has an expertise relative to H on some subject matter W at a time t just if S has a superior ability at t to determine the truth of propositions in W which is based in superior perceptual and/or cognitive skills and know- ledge, and is hence (in a relaxed sense) intrinsic, or has a crucial intrinsic component.

Exercise of an expertise will almost certainly require that the environment be normal in various respects — as with perceptual skills — and so is intrinsic only in a sense which is relaxed, though surely intuitive. Exercise of specialized cognitive skills may require access to equipment, even laboratories; but has a crucial intrins- ic component. An expertise is, in this lenient sense, a superior epistemic power possessed by a person due to her specific differentiating characteristics, such as superior perceptual skills, or specialized field of training and knowledge. Her expertises are relatively stable properties of a person, since they are not owed to mere accidents of spatio-temporal location, but are more deep-seated properties of that person; some owed to genetic endowment, but many acquired through special training or education.!)

S’s possessing a superior perceptual ability to H is one kind of expertise which S may have relative to H. This may be due to native differences in perceptual equipment, as with acute versus poor distance vision. But differential percep- tual ability may also be due to training and background knowledge. An expert at cricket can see and describe what is taking place during the game — ‘‘It was a fast ball that moved in from outside the off stump, and the batsman caught an edge on it, and was caught behind by the wicket-keeper’’ — when a novice will have discerned almost nothing specific at all. The same goes for aural per- ceptual abilities, for instance to discern and describe the harmonic progressions in a complex piece of music; or to catch and understand the words of speech in a particular language. Because background knowledge and skills inform and shape perception in this way, there is no sharp distinction between perceptu- al versus knowledge-based expertise. Many bases of expertise involve both in inextricable combination. The complex perceptual-cum-knowledge-based skill provides a superior ability to determine the truth of a range of propositions in certain circumstances. Other bases of expertise are more heavily grounded in spe- cialized knowledge and training, with a lesser role for associated perceptual skills. This includes scientific knowledge and skills in experimental procedures and the evaluation of data; and technological knowledge and know-how, such as that of the garage mechanic, builder, or computer technician. Most purely cognitive is superior ability at reasoning in a particular abstract domain such as mathematics.

When another has expertise relative to me in a certain, perhaps esoteric, field of knowledge, it is clear that I can and should — assuming I trust in her sincer- ity — defer to her in forming beliefs about the domain in question. Where the

236 Testimony and Epistemic Autonomy

field is sufficiently advanced and complex, I may not even be able to evaluate the arguments, nor the significance of the evidence, myself; and I may lack the nat- ive talents to acquire the skills to do so, even if I had both time and inclination (see Hardwig 1991). We will draw out the significance of these facts in the next section.

We can make some observations about the relations between weak and strong deference, and expertise. Deference to another is appropriate (assuming that she has, like me, a normal endowment of perceptual and cognitive skills) when she but not I has had access to the relevant evidence — for instance, when she but not I has had opportunity to exercise normal perceptual judgemental abilities, as in Case One. Since I have no basis for firm belief in such cases, this will be weak deference. I learn from the other about something of which I would oth- erwise be ignorant. Her report informs me, rather than overriding my own prior firm belief. Thus we can conclude that: Weak deference is often appropriate, even when the other has no superior expertise to me regarding the topic, she is merely contingently more expert than me, at this moment.

In contrast, when I and the other both have access to relevant evidence,"* deference to her will be appropriate only if I accept that she has a relevant epi- stemic power, an expertise, which is superior to mine. Since ex hypothesi we each have access to relevant evidence, I also have a basis for firm belief myself; so this will generally be strong deference. In this type of situation a stronger kind of deferring to another’s epistemic power, her superior authority, is involved. I accept the other’s judgement as overruling my own, in light of my acknowledge- ment of her superior epistemic power regarding the matter in question. Our Case Two above instances such strong deference. Natalie has an expertise relative to me on events going on in the perceptible distance. Thus she is better epistem- ically equipped than me to make judgements about what is taking place on the stage in the park, even though we are standing next to each other, and each able to look towards the stage. If my vision is as good as Natalie’s, then I will ration- ally defer to her testimony regarding what went on at the concert only if I was not there myself. But if I know her sight is better than mine, I may and should rationally allow her reports of her perceptual judgements to overrule my own perceptually based judgements, when we are similarly spatially located.

We may thus conjecture that: Strong Deference is appropriate only when the other has a superior expertise — an intrinsic epistemic power — to me."! This is largely true, although there are two counter-cases. First, it can be that I have a basis adequate for firm belief, but S, while having no relevant greater epistemic powers than me, has a stronger one which trumps it. For instance, I believe that Tom is away on holiday on the basis of my memory of his testi- mony of three weeks ago; but Chloe testifies to having seen Tom in town today. It is not determined whether I should accept Chloe’s testimony in these circum- stance, without further details. But there surely will be some cases of this kind, where it is right to accept another’s testimony overriding one’s own previous firm

Elizabeth Fricker 237

belief, because she has had access to fresh evidence, though her relevant epistem- ic powers are no greater than mine. One factor which would do the trick is if Chloe’s testimony was independently corroborated by many others. This is our second type of counter-case: I should bow to others’ testimony about some mat- ter, even if their skills and evidential position severally are not superior to mine, if weight of numbers is massively on their side. Notwithstanding these counter- cases, our conjecture captures a general tendency.

3 . D E F E R E N C E O N M O R A L A N D A E S T H E T I C M AT T E R S ?

We started with the idea that it is rational to defer to another’s apparently sincere testimony on some topic P just if I recognise that she is better placed than me to judge whether P (and I am aware of no significant contrary testimony regarding P). This being so, learning from testimony is possible only in domains where it makes sense to think that one person can be better placed than another to make judgements. This in turn requires some notion of objective standards of evid- ence and correct judgement for the domain in question. If any basis whatever on which a judgement is made is as good as any other, then the idea of anoth- er’s being better placed than me does not apply. In fact this restriction imposes little more than the very idea of judgement imposes in the first place. There is a determinate content to judge only if there are standards for correct judgement, independent of what seems to any particular individual to be correct. However, it could perhaps be that this minimal notion of objectivity applies in some domain, but for some reason someone else can never be better placed than me to make judgements about it, or at least it could never be rational for me to believe this. Some accounts of self-ascriptions of certain conscious mental states would place them in this category. Exploring the possibility or otherwise of rational deference to testimony may give fresh insights onto this topic, as well as others, though I cannot pursue this thought further here.""

We are investigating the circumstances in which, and topics on which, it can be rationally permissible, indeed mandatory, deferentially to accept another’s testimony. Whether and if so in what circumstances deference to others’ testimony on moral and aesthetic matters is ever rational — epistemically and morally proper — is a large topic, an adequate discussion of which would require a separate paper. But I shall make a key preliminary point. The kind of objectivity in standards of judgement which we have just seen to be required — the idea that there are better and worse ways of arriving at judgements in the domain — is relatively unproblematic, and no more than common sense, for both moral and aesthetic judgements. Thus it is only a very moderate thesis to hold that there is such a thing as superior expertise on moral and aesthetic matters. This view can be held without commitment to any metaphysically outlandish and epistemically problematic form of moral or aesthetic realism. The notion of objectivity which

238 Testimony and Epistemic Autonomy

must be invoked need not be understood in terms of correspondence with wholly mind-independent facts, and more or less accurate means of homing in on those. (There may or may not be any domains in which we wish so to understand it!) The required notion of objectivity of judgements in the moral and aesthetic domains can be explained in terms of better and worse ways of arriving at judgements of the class in question, with better and worse defined otherwise than in terms of homing in on mind-independent truth."# This being so, there is nothing immediately incoherent, or metaphysically neck- out-sticking, about the idea of moral and aesthetic experts. The idea accords with common sense, and normal practices of deference. Aesthetic experts are those who are specially trained, experienced, and knowledgeable in a certain area — say Baroque music, or Renaissance painting. One would defer to them about the qualities, including aesthetic ones, of items in their field of expertise. The idea of moral experts is equally valid. Some people are especially trained, experienced, and knowledgeable in the kinds of considerations involved in making moral judgements. Such expertise may be primarily in a specific field, involving a particular kind of empirical matter."$ A quasi-realist about the moral, or aesthetic, can make perfect sense of expertise in these domains. I may defer to an expert in an aesthetic field because I know my own opinion-forming processes are crude and uninformed, untrained, in the relevant aesthetic — which amongst Caravaggio’s paintings are the greatest masterpieces; or what is a specially good example of an early nineteenth-century English transfer-printed cup and saucer. (Learning through deference, I may come in time to be a bit of an expert myself!)

I might defer for similar reasons in a moral matter. Or I might defer, or seek advice here, not because of a general lack of expertise, but because my consultant is better placed than me — a relative expert — regarding the current matter. She may know more of the relevant background facts about a difficult case regarding custody of children in divorce proceedings, or she may unlike me be impartial, not being involved in the situation as I am. Or I may just want a second opinion, or to talk the matter over with someone else, as part of the process of forming my own judgement. Would it be just, or cruel, to carry out my threat and deprive my son of his Beano,"% for getting into trouble at school for fooling about in class again? This is the kind of situation where one may want to confer, and maybe defer to another’s judgement.

A full investigation of the possibilities for, and constraints on, rational deference on moral and aesthetic matters must canvas more considerations than those raised here. I suggested earlier that there may be constraints deriv- ing from the meaning of aesthetic predicates — their tie to a specific non- judgemental cognitive-cum-affective response in the subject — on the extent to which deference on aesthetic matters is possible. (One possibility is that only weak, never strong deference, is rationally possible.) For deference on moral judgements, there are important ties with the idea of individual autonomy and responsibility which may place limits. In the present paper I merely wish to

Elizabeth Fricker 239

point out that the idea of expertise on these topics is an everyday and apparen- tly sensible one, and thus that deferential acceptance of testimony on these matters is prima facie rationally possible, as well as being a common occurrence (see Jones 1999). We have briefly reviewed the various bases on which another person may some- times be far better placed than oneself to make judgements about a certain subject matter. We saw that such superior epistemic status is sometimes based in acci- dents of location, and may be short-lived; but is sometimes based in intrinsic and relatively stable differences in epistemic powers between two individuals. When I appreciate that another person is thus expert relative to me, it is not merely rationally permissible, but rationally mandatory, to defer to her judgement over my own conclusions, regarding the subject matter in question. This being so, one may question whether the supposed ideal figure of the autonomous knower, who refuses ever to trustingly accept another’s testimony, a fortiori will never allow her own judgement to be corrected by another’s, is really such an ideal after all. I will return to this question in my final section. First I address my earlier question: To what extent can one maintain one’s epistemic self-governance despite one’s inevitable reliance on others’ testimony, and the technological fruits of others’ knowledge and expertise, in almost every area of one’s life?

4 . R E L I A N C E O N O T H E R S ’ W O R D A N D E P I S T E M I C S E L F - G O V E R N A N C E

We have seen, as encapsulated in TDAP2, how it is rational to accept another’s word on a topic, and even to allow her expressed judgement to override one’s own prior opinion, when one knows that she is strongly placed epistemically, and better placed than oneself, regarding the matter in question. For each of us, her appreciation of her own circumscribed and feeble epistemic powers and small position in the larger scheme of things, together with her grasp of folk psycho- logy, including where applicable appreciation of others’ superior expertise and epistemically more advantageous position, entails that deference to others’ opin- ions is rational, in these circumstances. Lack of such appreciation of one’s limited powers and others’ superior ones, and an accompanying refusal to bow to others’ judgement or advice even when they are clearly relatively expert, is pig-headed irrationality, not epistemic virtue or strength.

Does this mean, then, that there is after all no loss of epistemic autonomy incurred by the way in which, in our modern condition, we rely on others’ know- ledge and its technological fruits for whole swathes of our fabric of knowledge and in our daily lives (as sketched in my introduction)? It does not. It is cru- cial for the maintenance of epistemic self-governance that our trust in the word of others is given not blindly and universally, but discriminatingly. By trust- ing only cannily, and with good grounds, we can do much to retain epistemic

240 Testimony and Epistemic Autonomy

self-governance. I shall return to this theme shortly. But there is still an important loss of autonomy, as I will now explain.

I mentioned our awareness of our own cognitive limitations, our feeble powers. We can only see what is here and now, and that only to a limited extent. Our memories even of this are less than total and often corrupt, and our inferential powers are feeble. A superior being, one who lacked our cognitive limitations, and could do all the work herself, in finding out about the universe, could be epi- stemically autonomous in a way that no one of us, with our limited research time and processing capacities, is able to be. She would not need to take anything on trust from another’s word, because she would have the epistemic power to check up, to find out for herself about everything she wanted to know, without reliance on others. We are not such beings, and so we can extend our knowledge beyond a small base only through rational trust in the spoken or written word of oth- ers. My trust in another’s word is rational when I have good grounds to believe her competent about her topic and sincere, and by this means I can know about all kinds of matters which I lack the time or talents to find out for myself. But this knowledge from trust in testimony is knowledge at second hand (or third, or fourth . . .), and as such my epistemic position vis-à-vis what I know is in at least one respect inferior to when I know at first hand.

When I form belief that P through my trust in a speaker’s word given to me that P, her testimony that P, I take her to speak from knowledge. That is, this is a normative commitment of my accepting her utterance at face value, as an expression of knowledge. If I come to know she does not speak from know- ledge, this is a normative defeater for my belief. Additionally, in my own view of knowledge as requiring adequate grounds, I must be disposed upon reflec- tion to form the belief that she speaks from knowledge. This belief is an essential justifying ground for my belief in what I am told and trustingly accept, and so must itself be knowledge. In short: my reason for believing P true is because I believe, or am disposed to form belief upon reflection, that my informant is telling me what she knows. This being so, I know only because someone else’s knowledge has been passed on, spread to me by the mechanism of telling, of testimony."&

Knowledge can be passed on in this manner through many links in a chain of trusted testimony. But the regress must stop eventually with someone who knows that P not from trust in testimony. The following axiom holds:

T: If H knows that P through being told that P and trusting the teller, there is or was someone who knows that P in some other way — not in virtue of having been told that P and trusting the teller.

It is a consequence of T that if someone knows that P through trust in testi- mony, there must be some other way in which P is or once was known. Hence T has the corollary:

Elizabeth Fricker 241

T corollary: For any proposition P that can be known, there must be some way other than trust in testimony through which P can or once could be known."'

Why cannot a chain of trusted testimony go in a circle, falsifying T? The regress must end with someone who knows that P in some other way, because knowledge requires evidence or grounds. When I know that P from someone’s testimony, my personal ground for my belief that P, the warrant in virtue of which I am entitled to it, is my knowledge that my informant knows that P. But in taking P to be known I am rationally committed to an existential supposition: that there is, that it is to say that some individual or group of persons between them possesses, evidence or warrant for P, which is not just that someone they trust has told them that P. As T expresses, knowledgeable belief based on trusted testimony implicitly refers back to the existence of a non-testimonial ground or warrant for what is testified to: the ground or warrant in virtue of whose posses- sion the original teller(s) spoke from knowledge."( Hence there cannot be a state of affairs that is known of only through trust in testimony. A chain of testimo- nially spread belief which went in a circle would lack any empirical grounding, and what is believed would not be true unless by luck.") Consonant with this fact, there is a sense of ‘the evidence for P’, used in scientific-style discourse, when it is asked: ‘‘What is the evidence for P?’’, in which someone’s testimony that P is not evidence for P at all. For instance the question: ‘‘What is the evidence that smoking causes lung cancer?’’ is not answered by responding: a lot of distin- guished scientists have asserted that it does. The question asks for an account of the real evidence, the evidence on which the experts’ conclusion is based. The well-groundedness of belief spread about through testimony depends on the existence of such non-testimonial evidence for P — that is, on its possession, per- haps distributedly, upstream in the chain of informants.

Now we see the respect in which knowledge from trust in testimony is in one way inferior owing to its being at second hand. When I know that P solely from trust in testimony, I do not possess the evidence for P. Instead, my knowledge is premissed on the existential supposition that there is non-testimonial evidence for P, although I myself do not possess it. I am rationally committed to the pro- position that a person or persons upstream in the chain of informants between them possess that evidence — the grounds for believing P true. Where the propos- ition is an empirical one that is part of a theory, I am also rationally committed to the proposition that these others have evaluated the evidence and drawn conclu- sions from it correctly (often, this ability is a large part of a special expertise the others possess vis-à-vis me).#* Is it a weakness in my epistemic position regarding P that my ultimate ground for believing P is this derivative second-order one, the proposition — which I must be disposed to form belief in — that there is empir- ical warrant,#! though unknown to me, for believing P? Where my informant’s expert status vis-à-vis me is accidental, it does not seem a worry. — My son tells

242 Testimony and Epistemic Autonomy

me there is still some milk left in the fridge, and I believe him. But if it mattered a lot I could easily check up for myself, and if what he told me were false I would quickly find out. I can get to the first-hand evidence, if need be, and I can eval- uate it correctly. But where my reliance on others depends on an expertise they possess relative to me which is more deep-seated, and I lack the ability to check up for myself if it seems worth it, the existential supposition and dependence on others’ epistemic skills and truthfulness is more troubling.#"

Epistemic dependence on others is troubling first because it is risky — there are many motives for deceit, and causes of honest error, on the part of each of us; and while each can try to trust only where there is ground to expect sincerity and competence, as elaborated below, each link in a chain of testimonial transmission incurs its own risk of error. It is troubling second, because along with the epi- stemic dependence on others comes a no less risky practical dependence on them, in many areas — for instance, for maintenance of all the technological devices on which one depends every day, from electric lighting to computer to driving one’s car, and so forth. Third, epistemic dependence on others, while it extends one’s knowledge base so enormously, also lessens one’s ability rationally to police one’s belief system for falsity. There are many things a layperson believes for which she would not know how to assess the scientific evidence which supports them, even if presented with it. This being so, these beliefs of one will lack the characteristic sensitivity to defeating evidence, should it come along, which is usually taken to be a hallmark of belief which amounts to knowledge.##

I have spelled out the bad news for epistemic self-governance entailed by our dependence on the word of others. The good news is that — as I already emphas- ized — our trust in others need and should not be given blindly, but cannily, only where it is due. Although cognitively limited beings as we are, we must perforce rely on others if we want to enjoy the epistemic and technological riches of mod- ern society, we can take care only to trust those we have good reason to hold worthy of our trust.#$ Fortunately we all have some basic cognitive equipment to help us assess both the sincerity and competence of others in many, though by no means all circumstances. This is because we are all experts (though of varying degrees of skill) in one special topic, namely that of folk psychology. Thus, where we do not have access to or cannot evaluate the evidence for propositions in some domain ourselves, we move one level up, and instead evaluate the experts, our human sources of knowledge about this domain.#% But assessing an informant’s trustworthiness is not always easy, and sometimes there are not sufficient epi- stemic resources available to the layperson to enable a firmly based evaluation to be made at all. The risks involved in trusting others are considerable, especially where there are motivations for deception at work. As I have been arguing, there is often good empirical ground for trusting others, and where so it is consistent with our maintenance of our epistemic self-governance, our responsibility for our own beliefs, that we believe on trust in the word of others, relying on their report for the truth of something where we do not possess for ourselves the evidence,

Elizabeth Fricker 243

and may not even be capable of appreciating its significance. Moreover, as we saw in Section 2, where I know another to be epistemically expert relative to me on a topic, it is not just rationally permissible, but rationally mandatory for me to accept her judgement in preference to my own, just so long as I have good ground to trust her sincerity. Where there is not good ground to believe an inform- ant trustworthy, however, epistemic self-governance entails that we should not accept the reports of others. Caution and canniness should govern our response to others’ testimony. Unless we exercise it, we fail to maintain responsibility for our own beliefs.

5 . T H E I D E A L R E V I S I T E D

I return now to the figure with which I began, the autonomous knower, who trusts no one else’s word on any matter, hence believes only where she herself pos- sesses sufficient evidence, non-testimonial grounds, for what is believed. In the light of the material of the last section we can clarify the autonomous knower in this way: she never believes on the basis of a second-order warrant for belief, the belief that someone else knows, someone else possesses evidence showing the truth of the proposition believed. Is this figure really an ideal? We observed that a superior being, with all the epistemic powers to find out everything she wanted to know for herself, could live up to this ideal of complete epistemic autonomy without thereby circumscribing the extent of her knowledge. Given the risks involved in epistemic dependence on others we saw in the last part of the previous section, this superior being is, I suppose, epistemically better placed than humans are. That is, if she knew at first hand just as much as I myself know in large part through trust in others’ testimony, she would be epistemically more secure, hence both practically more independent, and — in some abstract sense — more autonomous than I am. In the same way that I might regret that I cannot fly, or live to be 300 years old, I might regret that I am not such a being.#&

But what of a human, with no more than human perceptual, physical, and cognitive powers, who attempted to maintain a regime of complete epistemic autonomy — that is to say, who never took anyone’s word for anything, and nev- er deferred to another’s judgement on any matter? We have seen that rational prudence dictates that one should bestow trust only where it is due; where one has good grounds to believe one’s informant competent and sincere. But equally, as encapsulated in TDAP2, where there is good ground to believe another expert relative to oneself, it is not just rationally permissible, but mandatory, deferen- tially to accept the other’s judgement. So what would this individual’s beliefs about others have to be like, for her refusal ever to believe on anyone else’s say-so to accord with maintenance of a rationally coherent system of beliefs? If ration- al at all, she would be not an ideal, but rather a paranoid sceptic about others’ intentions and capacities. Or perhaps she would be severely cognitively lacking,

244 Testimony and Epistemic Autonomy

simply lacking any adequate grasp of what other people are, their capacities and positions in the world — not a master of folk psychology, but an individual sol- ipsist. She cannot ever admit that anyone else knows anything which she does not independently know herself since — as we saw — to admit this is to provide oneself with a second-order warrant to believe that thing oneself (‘A knows that P’ entails that P, and this entailment is a priori and obvious). One might won- der also whether she trusts the recorded beliefs of her own past self, as written down in her personal diaries and other records. The human would-be epistemic autonome on closer investigation is not an ideal, but either paranoid or severely cognitively lacking, or deeply rationally incoherent. We all can remember occa- sions on which someone we know has irrationally refused to change her opinion in response to testimony from someone evidently better placed to judge of the matter than she is. The individual autonome carries this irrational tendency to its irrational extreme.

For each one of us the extent and occasions on which she should accept and rely on others’ testimony is a delicate matter, decisions about which require careful assessment on particular occasions. But that there are some occasions on which it is rational deferentially to accept another’s testimony, and irrational to refuse to do so, is entailed by her background knowledge of her own cognitive and physical nature and limitations, together with her appreciation of how other people are both like and in other respects unlike herself, hence on some occasions better epistemically placed regarding some matter than she is herself. I may rationally regret that I cannot fly, or go for a week without sleep without any loss of performance, or find out for myself everything which I would like to know. But given my cognitive and physical limitations as parametric, there is no room for rational regret about my extended but canny trust in the word of others, and enormous epistemic and consequent other riches to be gained from it.

R E F E R E N C E S

Adler, Jonathan (1994), ‘Testimony, Trust and Knowing’, Journal of Philosophy, 91: 264 – 75.

Blackburn, Simon (1984), Spreading the Word (New York: Oxford University Press). Burge, Tyler (1993), ‘Content Preservation’, Philosophical Review, 102: 457 – 88 Clement, Fabrice, Koenig, Melissa, and Harris, Paul (2004), ‘The Ontogenesis of Trust’,

Mind and Language, 19/4: 360. Coady, C. A. J. (1992), Testimony: A Philosophical Study (Oxford: Clarendon Press). Davidson, Donald (1984), Enquiries into Truth and Interpretation (Oxford: Oxford Uni-

versity Press). Descartes, René (1641), Meditations on First Philosophy, in The Philosophical Works of

Descartes, ed. Haldane and Ross vol. i. (Cambridge: Cambridge University Press, 1967). Fricker, Elizabeth (1994), ‘Against Gullibility’, in B. K. Matilal and A. Chakrabarti

(eds.), Knowing from Words, (Amsterdam: Kluwer Academic Publishers), 125 – 61.

Elizabeth Fricker 245

(1998) ‘Self Knowledge: Special Access versus Artefact of Grammar — A Dicho- tomy Rejected’, in C. Wright, B. Smith, and C. MacDonald (eds.), Knowing Our Own Minds (Oxford: Clarendon Press), 155 – 206.

(2002), ‘Trusting Others in the Sciences: A Priori or Empirical Warrant?’, Studies in History and Philosophy of Science, 33: 373 – 83.

(2005), ‘Testimony: Knowing through being Told’, in I. Niiniluoto, M. Sintonen, and J. Wolenski (eds.), The Handbook of Epistemology (Netherlands: Kluwer Academic Publishers).

(forthcoming), ‘Second-Hand Knowledge’, Philosophy and Phenomenological Research

Goldberg, Sanford (2001), ‘Testimonially Based Knowledge from False Testimony’, Philosophical Quarterly, 51: 512 – 26.

Goldman, Alvin (2002), ‘Experts: Which Ones Should You Trust?’, Philosophy and Phe- nomenological Research, 63/1: 85 – 110.

Grice, H. (1957), ‘Meaning’, Philosophical Review, 66: 377 – 88. Hardwig, John (1985), ‘Epistemic Dependence’, Journal of Philosophy, 82: 335 – 49.

(1991), ‘The Role of Trust in Knowledge’, Journal of Philosophy, 88: 693 – 708. Jones, Karen (1999), ‘Second-Hand Moral Knowledge’, Journal of Philosophy, 96: 55 – 78. Locke, John (1690), An Essay Concerning Human Understanding, abridged and ed. John

(London: Everyman, 1993). Lyons, Jack (1997), ‘Testimony, Induction and Folk Psychology’, Australasian Journal of

Philosophy, 75: 163 – 78. McDowell, John (1994), Mind and World (Cambridge, Mass.: Harvard University Press). Quine, W. V. O. (1953), ‘Two Dogmas of Empiricism’, in his From a Logical Point of

View (Cambridge, Mass.: Harvard University Press). Schiffer, Stephen (1972), Meaning (Oxford: Oxford University Press). Welbourne, Michael (1994), ‘Testimony, Knowledge and Belief ’, in B. K. Matilal

and A. Chakrabarti (eds.) Knowing from Words (Amsterdam: Kluwer Academic Publishers), 297 – 313.

N O T E S

1. See McDowell (1994). I am talking here about humans, and how they are psycho- developmentally able to acquire language-and-thought; not about other logically possible intelligences, nor the philosophical fiction who springs instantaneously into existence, a functional replica of a human. ‘Testimony’ here is to be taken broadly, to include verbal teaching and coaching by others. It would be a mistake to obscure our dependence on trust in others’ sincerity and competence, in this developmental process, through a definitional stop.

2. By ‘simple trust’ I mean: trusting response to what others tell or teach us, by one who as yet lacks the conceptual resources to entertain doubts about the reliabil- ity of others’ teaching. This is the inevitable initial condition of the infant learn- ing its first words through interaction with its carers. (However many writers on testimony exaggerate how long this initial condition persists — don’t underestimate children — they get wise pretty soon! See Clement, Koenig, and Harris (2004).)

246 Testimony and Epistemic Autonomy

3. A distinction cannot be drawn between analytic versus synthetic, amongst the famil- iar platitudes involving cluster concepts like these, and so many other of our con- cepts. See Quine (1953).

4. Mature visual experience is basically the result of the visual system’s best guess as to what is out there, given the proximal stimulus to the retina. Some of this is hard- wired and hence culturally invariant, notably the perception of 3-D shaped solid objects. But perception is also soaked by thicker, culture-specific concepts, so that its perceptual deliverances to consciousness are much richer: it’s a mobile phone; a tomato; my daughter — these and their like are typical parts of the content of percep- tual experience, not inferences from it.

5. Complete epistemic autonomy, as described here, by definition requires not relying on anyone else’s testimony for any of one’s knowledge. I shall explore whether a weaker, but crucial, notion of epistemic self-governance — epistemic responsibility for one’s own beliefs — is consistent with accepting things on other people’s word for them.

6. That is to say, there is — as will be developed below — a set of conditions regarding speaker’s and hearer’s circumstances such that both the offering, and the accept- ance, of testimony on a topic is objectively epistemically appropriate just when they obtain; so that a speaker gives testimony epistemically properly, and a hear- er epistemically properly accepts it, when each knows these to obtain. (Where the speaker or hearer believes justifiedly, but falsely, that they do so, her act or response is subjectively but not all-in epistemically proper.) See Fricker (forthcoming) for a supporting account of the speech act of telling. I there show how the nature of the communicative speech act of telling is crucial to the question when, and on what basis, the teller may properly be believed — to the epistemology of telling, and testi- mony more broadly. The qualification ‘epistemic’ to the type of propriety here is not idle — a telling could be epistemically appropriate, but grossly inappropriate in some other dimension, e.g. irrelevance, or rudeness.

7. When going for detail some qualifications are needed here. First, for statements made with sentences containing indexicals, understanding may require grasping an appropriately related content or proposition, rather than the very same one — same referent but different senses: ‘‘I’m hungry’’; ‘‘It’s hot here’’, uttered in a telephone conversation. Second, there can be cases where something is correctly conveyed by testimony, although the utterance is partly misunderstood; it may be that only the correctly understood part is believed. See Goldberg (2001).

8. There clearly can be Gricean (Grice 1957; Schiffer 1972) acts of communication which do not employ language as their medium. There are also non-literal message- conveying linguistic utterances such as ironic or sarcastic ones. And a speaker may succeed in getting her message across, be correctly interpreted and believed, despite using words wrongly in some respect — not in accordance with the constraints and permissions of the literal meaning of the sentence she mistakenly employs. Even where communication of what is literally asserted in the speech act occurs, presup- positions and conventional implicatures may be conveyed too. These acts all share with paradigm tellings the successful getting across of a message. I shall not investig- ate here the respects in which they differ; except to say that where what is conveyed

Elizabeth Fricker 247

is not explicitly asserted there is, I believe, a diminution in the responsibility for the truth of what is got across incurred by the utterer. This is one reason to reserve the term ‘telling’ for acts of communication via explicit assertion exploiting literal meaning, as is done in ordinary parlance.

9. I am being careless here about distinguishing sentence types from particular utter- ances of them effecting speech acts, and the role of context in fixing what precise pro- position a particular utterance of a sentence expresses. This matter, though crucial and pervasive in natural language, is tangential to the current point. The fastidious reader may imagine the necessary complicating adjustments.

10. ‘‘There are very beautiful pictures in the Uffizi in Florence, though I have never seen them.’’ — this sounds deviant to my ear. As opposed to ‘‘There are said to be very beautiful pictures in the Uffizi, though I have never seen them’’. On the other hand ‘‘There are famous paintings by Botticelli in the Uffizi, though I have never seen them’’ sounds fine.

11. I form belief that P on the basis of trust in another’s testimony that P, when I do so because I take her utterance at face value, as an expression of her knowledge that P. In so doing I take her word for it that P. There is a variety of other cases where a hearer forms belief that P in response to observing testimony that P, which are not cases of trust in that testimony. Fricker (forthcoming) contrasts these cases with the case of trust in the testimony, and argues that the latter relatively narrow category is the key epistemic kind to discern, in theorizing how knowledge can be spread by means of testimony. The condition proposed in TDAP for forming belief in what is stated would not be correct, for a broader category. Rather, it further characterizes the narrow category.

12. There is scope for further refinement here: it could be that an informant is very unlikely to form a belief that P which is not knowledge; but is more prone to error, or careless judgement, than not-P. This kind of one-sided reliability is quite plausible in some cases — e.g. someone who is slow to make a judgement of guilt of anoth- er — and a hearer could be aware of this epistemic disposition of an informant. But more usually, someone will be in this way reliable regarding P only if she is also similarly reliable regarding not-P. TDAP as formulated specifies this stronger con- dition. Perhaps someone could be self-deceived, so that she in some sense ‘really knows’ that P, while kidding herself, and telling others, that not-P. TDAP concerns knowledge expressed in conscious judgement, and so excludes repressed knowledge, if such is possible.

13. In contrast with her competence, or expertness as I am here calling it, I think that one is entitled to presume a speaker sincere, unless there are specific cues or other evidence calling this into question. This fact is not an epistemic principle special to testimony, but is fall-out from correct general principles governing the ascription of mental states to other persons. See Fricker (1994).

14. The issues here are delicate. Mere weight of numbers of concurring testifiers does not per se increase the probability of correctness; it depends on the details regarding the likely explanation of how they have come to hold their expressed beliefs. See Gold- man (2002) for an excellent discussion of what epistemic resources a layperson may have, to decide which to trust out of two experts giving contrary testimony.

248 Testimony and Epistemic Autonomy

15. For instance regarding various future matters: the weather, currency and interest rate movements, etc. Here one should defer to and act on the basis of the best advice; while being aware that it is not knowledge — hence hedging one’s bets accordingly.

16. An account such as Wellbourne (1994), which holds it sufficient for the recipient of testimony to come to know, that the hearer speaks from knowledge, is purely external and as such violates my requirement. However accounts like Burge (1993) or Coady (1992), which maintain an entitlement to trust testimony as such which is however defeasible, can be seen instead as proposing a specific thesis regarding how internal rationality is satisfied, in this case — albeit one with which I disagree. (In the case of perception, it is plausible that epistemic responsibility permits one to take one’s senses on trust, unless aware of defeaters.)

17. This is why one knows so little about what goes on during one’s children’s days at school!

18. Can I know this, without begging the question — ungroundedly trusting her testi- mony over mine? Certainly: I have found on many previous occasions that what she has judged from a distance proves correct, as we get nearer. The fact that expert- ise is time and circumstance relative, often transient, means that another’s epistemic expertise relative to oneself can often be conclusively established by oneself, despite one’s own inferior epistemic power.

19. Expertise of S has been defined as relative to another person, H. But we can easily extract a more general concept of expertise, which is a superior epistemic power regarding some topic relative to all those without the specialist training or skills in question — the layperson or non-specialist.

20. The notion of ‘access to relevant evidence’, and certainly of two persons having equal access to relevant evidence, is fraught with difficulty, given the theory-dependence of one’s observational powers — as my cricketing example above illustrates. It does not bear much theoretical weight in the present argument, and all I require is that there be some cases where it clearly applies, and others where it clearly does not. I intend that it holds of Case Two, and similar situations.

21. Is superior expertise also normatively sufficient for strong deference? No, since two people both with superior expertise to me may supply contrary testimony. Apart from this, I cannot think why else it should fail to be.

22. Fricker (1998) argues, on precisely this point, that accepting the possibility of cor- rection of one’s self-ascriptions of mental states made through avowal, by other evidence from one’s behaviour, which might be pointed out to one by others, is a condition for one to be ascribing a genuine concept in these self-ascriptions.

23. I here make a large, but unoriginal claim, which requires at least a fat book for adequate defence. I have in mind positions like the ‘quasi-realism’ of Blackburn (1984).

24. The inextricable interweaving of fact and value in the considerations relevant to a final conclusion on a complex matter reinforce this point. Consider, for instance, the members of a panel appointed to draw up proposed legislation controlling research using human embryos. Both scientific and moral expertise are required, and intel- ligent conclusions rest on inextricable understanding of both. Another example of a specific partly moral expertise is making decisions about when children should be taken away from their parents and into care.

Elizabeth Fricker 249

25. The Beano is a popular comic-strip magazine for children, in the UK. 26. Knowledge requires grounds, and if I trust a speaker who tells me something true but

does not herself know it, my own belief will be based on a false premiss and so not be knowledge. This is the general conception of knowledge I favour, and my account of knowledge from testimony is shaped by it. Even if a different view of necessary conditions for knowledge is taken, that the speaker knows what she tells is clearly a rational commitment of a belief based on trust in testimony.

27. The ‘other ways’ may however include deduction, induction, or inference to the best explanation from premisses some of which were supplied by diverse bits of testi- mony. See Fricker (forthcoming). The tense qualification is important here — the original informant may have since died, or simply forgotten what she once knew and told to others.

28. T and its corollary do not imply the stronger claim: For any P which is known, there is someone who knows it in a way which has no epistemic dependence on testimony. This stronger claim is false, as is explored in Fricker (forthcoming). The source of testimonially spread knowledge that P may have learned some of the facts from which she inferred P from others’ testimony. Thus the ultimate, non- testimonial evidence for any complex theoretical proposition may be possessed only distributedly, by the members of a group. See Hardwig (1991).

29. This remark remains true, but needs careful explanation, when we are dealing with facts constituted by human practices — the boundary between two countries, what something’s name is, and so forth. The testimony itself would not make the belief true, but enough people acting on belief in it would do so.

30. My belief is premissed on these suppositions not in the strong sense that I must occurrently believe them; rather, they are normative commitments of my form- ing belief on trust in testimony. As such, I must come to acknowledge them if talked through it — and my trust is normatively defeated if I come to believe any of them false.

31. I say ‘warrant’ here rather than ‘grounds’, since there are some types of belief — e.g some beliefs regarding one’s own mental states, and perhaps basic perceptual beliefs — which are empirically warranted, but not by grounds for belief.

32. Epistemic dependence of this sort is explored in a series of seminal articles by Hard- wig (1985, 1991). Hardwig suggests the schema: ‘H has reason to believe that S has reason to believe that P ! H has reason to believe that P.’ The schema only holds of prima facie reason, however — I could know that S has reason to believe that P, while myself being aware of defeaters for those reasons. Our present point is that the reasons in question are different. As I have been emphasizing, the ground for belief supplied by trust in testimony is a second-order one. My reason to believe is that I believe that my informant knows that P, hence that she or someone upstream of her has a non-testimonial warrant to believe that P. My original source’s reason to believe is this non-testimonial warrant, the evidence for P.

33. A further point is that once the original source of a testimonially spread belief is no longer available, the original warrant for the belief is no longer retrievable. However this feature characterizes most of our beliefs. Cognitively limited beings that we are, we generally form a belief from the evidence, then store the fact in memory and

250 Testimony and Epistemic Autonomy

jettison the evidence. The lack of sensitivity to potentially refuting new evidence is, in contrast, a risk of testimonial belief only.

34. It should be abundantly clear by now that I am against all accounts of how know- ledge may be gained through testimony which do not require that the recipient trusts only where she has good grounds to do so. They are inconsistent with the requirement of individual rationality, that epistemic self-governance in the sense of responsibility for policing one’s beliefs for truth, is maintained by the individual, the thinking, believing, and acting subject. A rational individual cannot delegate this responsibility to others, although as I am elaborating here, the requirement can be discharged by moving up a level: evaluating the reporters, when we are unable to test their reports for truth directly.

35. In Fricker (1994, 2002) I discuss how non-question-begging evaluation of the dual components of a speaker’s trustworthiness, her sincerity and competence, is often possible. See also the excellent discussion in Goldman (2002).

36. No heavy commitment to the coherence of the conception of this superior being is intended or incurred. I use her merely as a heuristic device in the development of my argument.