1 A Psychologistic Conception of Reasons for Belief

The term ‘reasons for belief’ is sometimes used normatively, to indicate bases of belief that are good (relative to some epistemic standard), whether possessed by subjects or not.Footnote 1 The term also has a nonnormative usage, wherein it simply indicates possessed bases of belief, good or bad. And although one can easily imagine an expanded use of the term according to which it applies to mental events that are not essentially truth-presentational with respect to their contents (e.g. desires, preferences) as well as to ones that are (e.g. perceptual appearances, other beliefs), it is most commonly used in the epistemological literature to talk solely of the latter sort of mental event, which I’ll call “cognitive.”Footnote 2

Here I follow Turri (2009) and adopt a nonnormative, cognitive use of ‘reasons for belief’, thus taking reasons for belief to be analogues of what are called “motivating reasons” in the metaethical literature (p. 507, n. 8). If under this general conception we follow Turri further (as I do) and accept psychologism about reasons for belief—the view, that is, that reasons for belief are cognitive events in the subjects who have them—we still face the question of which such cognitive events in particular they are to be identified with. Assuming psychologism about reasons for belief in general, in other words, which specific form of it should we adopt?

With a view to its application to the reasons on which our testimonial beliefs are directly based, I want to propose an answer that recalls Dretske’s distinction between triggering and structuring causes. A triggering cause of an event e is an event that initiates a discrete causal process terminating in e. The causal process may be simple (i.e. such that it includes no other processes as elements and its initiating event therefore brings about e directly) or it may be complex (i.e. such that it includes one or more other processes as elements and its initiating event therefore brings about e indirectly, by the intervention of one or more other events that initiate included processes). If the process is complex, moreover, it may comprehend more than one triggering case of e (viz. both its initiating event and the initiating event of any included discrete process that also terminates in e). A structuring cause of an event e, by contrast, is not an event that initiates a discrete causal process terminating in e, but rather an event or condition in virtue of which such a process occurs.

Thus, to draw from one of Dretske’s well-known illustrations (Dretske 1988, p. 40ff.), suppose it’s the middle of winter and I’ve set my thermostat at 68 °F. As the temperature in the house drops below 68°, the thermostat kicks in and sends a signal to my furnace, which ignites as a result. The thermostat’s signal may be regarded as a triggering cause of the furnace’s coming on: it initiates a discrete (if simple) causal process terminating in the furnace’s ignition. The drop in temperature may also be regarded as a triggering cause of the furnace’s ignition: it initiates a discrete, complex causal process terminating in the furnace’s ignition. My previously having set the thermostat at 68°, however, like the thermostat’s current connection to the furnace, is merely a structuring cause of the furnace’s ignition: it initiates no discrete causal process terminating in the furnace’s ignition, but it is an event or condition in virtue of which such a process occurs.

There are triggering causes for every formation of belief; so much will be granted all around. Moreover, I take it to be uncontroversial that cognitive events in subjects very often serve as triggering causes of their belief formations. With these points in mind, the specific version of psychologism about reasons for belief (in the relevant nonnormative, cognitive sense) that I want to propose can be simply stated: a reason for S’s belief that p is a cognitive event in S that serves as a triggering cause of S’s forming the belief that p. The belief formed—the belief for which S has a reason—may be either occurrent or dispositional.Footnote 3 Further, not only can S form an occurrent belief that p in circumstances where she already has a dispositional belief that p, S’s formation of an occurrent belief that p can have as a triggering cause a cognitive event distinct from any cognitive event that served as a triggering cause of her forming the already-possessed dispositional belief that p. Where S’s formation of an occurrent belief that p does have as a triggering cause a cognitive event distinct from any cognitive event that served as a triggering cause of her forming an already-possessed dispositional belief that p, we may say that S has a new reason for her existing belief that p.

2 Features of the Conception

The psychologistic conception of reasons for belief I have just articulated has a number of advantages. It allows us, for example, to be appropriately conservative about the number of reasons for belief we attribute to subjects (cognitive structuring causes of belief, at least, don’t count). It allows us to admit that beliefs needn’t always be based on reasons (although there are triggering causes for every formation of belief, not every such cause is cognitive). It doesn’t require that reasons for belief are transparent to their possessors (cognitive or not, triggering causes of belief-formations often run beneath the surface of consciousness). And because this psychologistic conception is consistent with a wide variety of substantive normative epistemological theories, it allows for a wide variety of potential explanations as to why a given reason for belief counts as good, or bad, as the case may be.

Moreover, the conception’s bearing on certain complex, and deviant belief-forming processes serves not only to distinguish it from nearby competitors, but also to avoid some of their problems. Consider a psychologistic conception according to which reasons for belief are the proximate cognitive causes of belief. One of the virtues of this conception, according to Turri (2011, p. 389), is that it gives correct verdicts about certain cases of complex belief-forming process. Here’s an example from Pollock and Cruz (1999):

I might believe that I’m going to be late to my class, and that might cause me to run on a slippery sidewalk, lose my footing, and fall down, whereupon I find myself flat on my back looking up at the birds in the tree above me. My belief that I was going to be late caused me to have the belief that there were birds in that tree, but I do not believe the latter on the basis of the former (p. 36).

On the proximate cognitive cause conception, we get the right verdict that in this situation the belief about lateness is not a reason for the belief about the birds: the lateness belief may be a distal cognitive cause of the birds belief, but it is obviously not a proximate cognitive cause of the birds belief.

My psychologistic conception is not just consistent with this verdict, it yields it, on the plausible assumption that the complex causal process described in the example (the believing that leads to the pace-quickening that leads to the slipping that leads to the falling that leads to the visual appearance that leads to the believing) is not a discrete one.

For a causal process to be discrete, it must have both a certain unity and a certain distinctiveness. In the case of a simple causal process, this would seem at the very least to require that the process’s initiating event and terminating event each be of a distinctive event type, such that these types are regularly associable. (For the initiating event to be of a distinctive event type is for it to be essentially of a certain event type, where the terminating event is not also essentially of this type; for the terminating event to be of a distinctive event type is for it to be essentially of a certain event type, where the initiating event is not also essentially of this type. And to say that one event type is regularly associable with another is to say that tokens of the one can be regularly associated with tokens of the other.Footnote 4) This explains why, in the thermostat example, the included causal process leading from the thermostat’s signal to the furnace’s ignition may be discrete even if it is regarded as simple: the thermostat’s signal is of a distinctive event type (e.g. electro-informational signaling), the furnace’s ignition is of a distinctive event type (e.g. electro-informational signal reception), and the two types are regularly associable. It also explains why a simple causal process leading from the furnace’s ignition to, say, my deciding to go for a brisk walk in the chilly air outside is not discrete: the furnace’s ignition and my deciding to go for a walk may each be of a distinctive event type, but these types do not seem regularly associable.

A complex causal process necessarily includes simple causal processes, and it is very difficult to see how such a process could be discrete if any of the simple causal processes it includes is not also discrete. Discreteness for a complex causal process thus seems minimally to require that the initiating event of each simple causal process it includes be of a distinctive event type, that the terminating event of each simple causal process it includes be of a distinctive event type, and that the initiating event types be regularly associable with the corresponding terminating event types. Hence, a complex causal process leading from a drop in temperature to a thermostat’s signaling a furnace to a furnace’s ignition to someone’s deciding to go for a walk is not discrete because at least one of the simple causal processes it includes (viz. the furnace’s ignition leading to the decision to go for a walk) is not discrete.

Obviously, these remarks come nowhere close to a full explication of the notion of discreteness on which my psychologistic conception relies. But I think they do suffice to make plain the plausibility of the assumption that the complex causal process described in the Pollock and Cruz example is not discrete. Very plausibly, that complex causal process is not discrete because some of the simple causal processes it includes are not discrete; some of these simple processes are not such that their initiating events are of distinctive event types, their terminating events of are of distinctive event types, and the initiating event types are regularly associable with the corresponding terminating event types. When it comes to the falling that leads to the visual appearance of the birds, for example, the falling and the visual appearance may each be of a distinctive event type, but the types do not seem regularly associable.

My psychologistic conception of reasons for belief may thus be said to share the proximate cognitive cause conception’s virtue of appropriately handling cases of complex belief-forming process like the one described in the Pollock and Cruz example. Yet my conception, unlike the proximate cognitive cause conception, allows for “transitivity in the basing relation” (Turri 2011, p. 389). Unlike the proximate cognitive cause conception, in other words, my conception allows not just that reasons for belief can themselves have reasons for belief, but that if a reason for belief has another reason for belief, then the latter is a reason for whatever belief the former is a reason for. From her occurrent belief that p, suppose, S infers the occurrent belief that q, from which she in turn infers the occurrent belief that r. Suppose further that this complex inferential belief-forming process is discrete. On my conception of reasons for belief, S’s belief that p is a reason for her belief that q—it is a cognitive event that initiates a discrete, simple causal process terminating in her forming the belief that q—just as her belief that q is a reason for her belief that r. But on my conception, S’s belief that p is also a reason for her belief that r: her belief that p is a cognitive event that initiates a discrete, complex causal process terminating in her formation of the belief that r. On the proximate cognitive cause conception, however, S’s belief that p cannot count as a reason for her belief that r, because her belief that p is not a proximate cognitive cause of her belief that r.

On Turri’s own psychologistic conception of reasons for belief they are nondeviant cognitive causes of belief, where nondeviance is cashed out in terms of the manifestation of cognitive traits of the believer (2011, § 4). One difference between our respective conceptions is that mine clearly rules out cognitive structuring causes of belief as reasons for belief, whereas his does not (not clearly, anyway). The conceptions also differ in that, although there is some overlap between the notion of nondeviance (or of cognitive-trait manifestation) on which Turri’s account relies and the notion of discreteness on which mine relies, the two notions are distinct; hence the two conceptions will presumably render different verdicts about at least some cases in which the nondeviance and the discreteness of a belief-forming process come apart.

Consider, for example, the case of Wilt (Turri 2011, p. 389): Wilt takes a peek into the crisper, and forms the belief as a result that, alas, the lettuce has wilted. In virtue of some neurological quirk, however, his occurrent belief about the lettuce causes him straight away to form the belief that his favorite football team will win twelve games in the upcoming season. The neurological quirkiness of the simpleFootnote 5 belief-forming process (from the belief about the lettuce to the belief about the football team) makes it a deviant one, and hence on Turri’s conception of reasons for belief the lettuce belief does not count as a reason for the football team belief. On my conception, however, the process may nonetheless be discrete—the process at least seems to be such that its initiating event is of a distinctive event type (e.g. occurrent belief about food), its terminating event is of a distinctive event type (e.g. occurrent belief about athletic teams), and the two are regularly associable—so it may well be that on my conception the lettuce belief counts as a reason for the football team belief. I do not see this as a problem for my conception, however, provided we keep in mind the nonnormative sense of ‘reasons for belief’ here at play. It is not obviously counterintuitive to say that the lettuce belief is a reason for the football team belief. Indeed, it seems quite natural to say that the lettuce belief is just a very bad reason for the football team belief.

3 Testimonial Reasons

Consider now testimonial belief, or what Audi (1997) has called “testimonially based belief” (p. 407) and Pritchard (2004, pp. 326–7) “testimony-based belief”. Testimonial belief is not just testimony-related belief—not just any belief bearing a causal connection to the reports of others. It is rather belief with certain content directly based on reception of another’s report of that content, i.e. belief that p whose formation is the terminating event of a simple belief-forming process initiated by reception of another’s report that p. (Where the content of another’s report includes the content in question but more besides, the report counts as a report of that content: another’s report that p and q and r, etc. thus counts as a report that p, as well as a report that q, etc.)Footnote 6

A belief with a certain content that is directly based on reception of another’s report of a different content (i.e. a content that neither is nor includes the content of the belief) is not plausibly a testimonial belief. If my formation of the belief that Burlington is the capital of Vermont is the terminating event of a simple belief-forming process initiated by my reception of your report that Montpelier is the capital of Vermont (due, say, to my structuring cause disposition to take all of your claims about Montpelier to be false but to have correlates true of Burlington), my testimony-related belief that Burlington is the capital of Vermont is not plausibly a testimonial belief.

Nor is a belief with a certain content that is only indirectly based on reception of another’s report of that content plausibly a testimonial belief. If my formation of the belief that Montpelier is the capital of Vermont is the terminating event of an inferential belief-forming process initiated by my occurrent belief that you really think that Burlington is the capital of Vermont, where my formation of this occurrent belief about what you really think is in turn the terminating event of a belief-forming process initiated by my reception of your report that Montpelier is the capital of Vermont, my belief that Montpelier is the capital of Vermont, though obviously a testimony-related belief, is not plausibly a testimonial belief.

The simple belief-forming processes whose terminating events are formations of testimonial belief and whose initiating events are receptions of others’ reports of the relevant contents are of course causal processes. They may also be essentially discrete; every imaginable example seems at least to satisfy the relevant minimal condition for discreteness suggested in the previous section. (They seem all to be simple causal processes such that their initiating events are of distinctive event types—e.g. reception of a report that p, their terminating events are of distinctive events types—e.g. belief that p formed on the basis of reception of a report that p, and the initiating event types are regularly associable with the corresponding terminating event types.) In any case, where they are discrete I will call such belief-forming processes testimonial belief-forming processes.

Because the report receptions that initiate testimonial belief-forming processes are cognitive events (they are not mere occurrences, in foro externo, of reports—not merely the assertions or claims of others—but rather subjects’ receptions of these reports), they will always be triggering causes of testimonial beliefs. It follows, then, that testimonial beliefs formed by testimonial belief-forming processes will always be directly based on reasons, viz. the report receptions that initiate the testimonial belief-forming processes. I’ll call these reasons testimonial reasons.Footnote 7

4 The Question of Basicality

Are testimonial reasons, in the sense just specified, basic? The question has the virtue of being less vague than that of whether testimony is a basic source of belief. Talk of the sources of belief, like talk of the sources of the normative status of belief (cf. Audi 2002; Lackey 2008, ch. 1) leaves unclear whether the main concern is with features of belief-forming processes, of the initiating events of belief-forming processes, of the structuring causes of the processes’ terminating events, of distinct processes somehow causally linked to the processes, or of some or all of the above. The question of the basicality of testimonial reasons is also an important one because, as I will explain below, its answer bears on the plausibility of an important kind of inferentialism about testimonial belief. But understanding the question requires us first to get clear about how, in the light of the psychologistic conception of reasons for belief I have given, we are to understand the notion of basic reasons for belief.

Traditionally, epistemological debates of relevance have tied the basicality of reasons for belief to nondoxastic considerations: basic reasons for belief have been construed as cognitive (i.e. truth-presentational) events other than beliefs. When foundationalists, for example, have asserted the existence of epistemically basic beliefs, they have meant to assert the existence of a stock of beliefs whose good reasons (i.e. reasons yielding a sufficient measure of justification, evidence, warrant, etc. for the beliefs) are nondoxastic cognitive events, for example, perceptual appearances. And when coherentists have denied the existence of epistemically basic beliefs, they have thereby asserted that all good reasons for belief are themselves beliefs.

I take the usual considerations of appearance-belief divergence in cases like the Müller–Lyer illusion to provide compelling support for the assumption that there are nondoxastic cognitive events—truth-presentational events distinct from occurrent beliefs.Footnote 8 Although both types of cognitive event are essentially truth-presentational to a subject, occurrent beliefs are also essentially truth-presentational by the subject (for a subject to have an occurrent belief requires that she present a content as true, if only in foro interno to herself) whereas nondoxastic cognitive events are essentially not: nondoxastic cognitive events are essentially truth-presentational only by something other than the subject. Alternatively put, whereas occurrent beliefs consist of occurrent assentings to content truth by the subject, nondoxastic cognitive events consist merely of occcurrent “attractions to assent” to content truth (Sosa 2007, p. 48ff).

On the assumption that there are nondoxastic cognitive events so distinct from occurrent beliefs, it makes good sense in the light of my psychologistic conception of reasons for belief to follow the traditional epistemological lead and tie the notion of basic reasons for belief to nondoxastic cognitive events. Nondoxastic cognitive events can be reasons for belief but—unlike occurrent beliefs—not have them (because, trivially, only beliefs can have reasons for belief);Footnote 9 hence it is to nondoxastic cognitive events that we should turn for basic reasons for belief—reasons for belief beyond which no further reasons for belief can be found. Consonantly with my psychologistic conception of reasons for belief, then, we can say that reasons for belief are basic just in case they are (reasons for belief and) nondoxastic cognitive events.

Consider the intuitive thought that whereas perceptual belief-forming processes are, in Goldman’s (1979) terms, “belief-independent” with respect to their inputs, inferential belief-forming processes are “belief-dependent” with respect to their inputs (p. 13). One plausible way to render this thought is to say that whereas the initiating events of perceptual belief-forming processes are thoroughly perceptual (i.e. if the processes are simple, their initiating events are perceptual appearances; and if they are complex, both their initiating events and the initiating events of any processes they include as elements are perceptual appearances), the initiating events of inferential belief-forming processes are thoroughly doxastic (i.e. if the processes are simple, their initiating events are occurrent beliefs; and if they are complex, both their initiating events and the initiating events of any processes they include are occurrent beliefs).Footnote 10 Given this intuitive thought, so rendered, the conception of basic reasons for belief I have just offered fits well with the suggestion that whereas the reasons for belief yielded by perceptual belief-forming processes are basic, those yielded by inferential belief-forming processes are not.

Notice further that one can admit the existence of basic reasons for belief in the sense I have specified without holding that such reasons are ever epistemically good, or at any rate good enough alone to render the beliefs for which they serve as reasons justified (evidenced, warranted, etc.). Accepting the existence of basic reasons in the sense here articulated, in other words, does not an epistemological foundationalist make; it is consistent with the coherentist’s hallmark rejection of the foundationalist’s epistemically basic beliefs.

Thus, the question of whether testimonial reasons are basic may be understood as the question of whether the report receptions that initiate testimonial belief-forming processes are all nondoxastic cognitive events. If the report receptions are all nondoxastic cognitive events, testimonial reasons are basic; if even some of the report receptions are occurrent beliefs, testimonial reasons are not basic.

5 Basicality and Inferentialism about Testimonial Belief

I said above that the question of whether testimonial reasons are basic bears on an important kind of inferentialism about testimonial belief. What kind do I have in mind, and what’s the bearing?

Following a helpful suggestion from Kenyon (2013, p. 74), we may distinguish normative and nonnormative inferentialism about beliefs in a given class: normative inferentialism holds that beliefs in the class are justified (evidenced, warranted, etc.) only by inferential belief-forming processes, whereas nonnormative inferentialism holds that beliefs in the class are formed only by inferential belief-forming processes. With respect to testimonial belief, the kind of inferentialism I have in mind is a nonnormative one: it maintains that testimonial beliefs are formed only by inferential belief-forming processes, and this does not alone entail the normative inferentialism about testimonial belief favored by “reductionists” in the epistemology of testimony, according to which testimonial beliefs are justified (warranted, etc.) only by inferential belief-forming processes.

One reason nonnormative inferentialism about testimonial belief is important is that, despite not strictly entailing normative inferentialism about testimonial belief, it nevertheless coheres better with the latter than with an opposing normative noninferentialism. At least this is so on a naturalistic approach to epistemology that sensibly seeks to constrain epistemic prescriptions by psycho-causal realities. On the naturalist approach nonnormative inferentialism about testimonial belief also seems to sit very ill at ease with typical “antireductionist” claims about testimonial evidence constituting a “fundamental” (Coady 1973, p. 154) or “sui generis” (Welbourne 2002, p. 409) category of evidence.

At any rate, the question of whether nonnormative inferentialism about testimonial belief (hereafter, simply “inferentialism”) is true has been garnering increasing attention in the contemporary epistemological literature (see, e.g. Audi 1997, pp. 406–8; 2011, p. 151ff; 2013, pp. 510–12; Kusch 2002, p. 20ff; Kusch and Lipton 2002, p. 213; Fricker 2004, pp. 115–7; Kenyon 2013), and it’s therefore worth noting how the present question of whether testimonial reasons are basic bears on it.

Simply put (at least on what I take to be a standard construal of inferential belief-forming processes as belief-dependent in the Goldmanian sense, where such belief-independence is rendered along the lines I suggested in the previous section), if testimonial reasons are basic, inferentialism is false. For if testimonial reasons are basic, they are all nondoxastic cognitive events, and testimonial belief-forming processes are all initiated by nondoxastic cognitive events. But if testimonial belief-forming process are all initiated by nondoxastic events, then testimonial beliefs are not—contrary to inferentialism—formed only by inferential belief-forming processes (which are belief-dependent, or thoroughly doxastic in terms of their initiating events): they are formed by testimonial belief-forming processes and testimonial belief-forming processes are not (because never) inferential.

6 Audi’s Challenge to the Prima Facie Case Against Basicality

In one obvious respect, denial of the claim that testimonial reasons are basic is considerably weaker than the claim itself: denial of the basicality claim does not require that all testimonial reasons are occurrent beliefs, whereas the basicality claim requires that all testimonial reasons are nondoxastic cognitive events. As such, denial of the basicality claim appears to have a prima facie plausibility that the claim itself lacks. It is prima facie plausible that at least some testimonial reasons are occurrent beliefs, and the claim that at least some testimonial reasons are occurrent beliefs entails denial of the basicality claim. Even if it is also prima facie plausible that at least some testimonial reasons are nondoxastic cognitive events, this doesn’t similarly afford the basicality claim prima facie plausibility, for the claim that some testimonial reasons are nondoxastic cognitive events doesn’t entail the basicality claim.

Is it really prima facie plausible that at least some testimonial reasons are occurrent beliefs? Consider occurrent beliefs about what others have reported (or are reporting)—occurrent beliefs to the effect that a “speaker has made an assertion with a particular content,” as Fricker (1987, pp. 69–70) puts it, or occurrent beliefs “of the form ‘Person R reports X’,” to use Goldman’s (1999, p. 129) description. It is prima facie plausible that testimonial belief-forming processes are at least sometimes initiated by occurrent beliefs of this sort, for not only do occurrent beliefs of this sort have all that is required to be the initiating events of testimonial belief-forming processes (they are cognitive events, and they are report-receptive in the relevant sense that they present as true contents about others’ reports), it would be an at least somewhat surprising discovery to learn that occurrent beliefs of this sort never initiate testimonial belief-forming processes. And if it is prima facie plausible that occurrent beliefs about what others have reported at least sometimes initiate testimonial belief-forming processes, then it is prima facie plausible that at least some testimonial reasons are occurrent beliefs, for the initiation of such processes by occurrent beliefs of this sort would make those beliefs testimonial reasons.

Call the claim that testimonial belief-forming processes are at least sometimes initiated by occurrent beliefs about what others have reported the prima facie case against basicality. Despite a wealth of arguments in the epistemological literature in favor of antireductionist views and “default” (to justification) theses about testimonial belief, and despite the growing interest in the inferentialist–noninferentialist divide, challenges to this prima facie case are very difficult to find. To his credit, and as far as I’m aware, Audi (1997) presents the only such challenge in the literature. He writes:

To be sure, in order to acquire, on the basis of testimony, a belief that [p], I may have to be disposed to believe that someone said [p]. But that seems to be only because I must perceive this being said, not because I must form the belief (or otherwise believe) that it was said, just as perception of a sentence in a convincing editorial can produce belief of what it says without one’s forming the belief that the sentence says that. It is my perception of what is said, typically my hearing or reading it, that is required for formation of a testimonially based belief of the proposition attested to. Understanding and believing testimony that p when we hear that testimony may require that in some sense we presuppose the attester said that p—so that if (e.g.) we disbelieve the attester said that p, we will not believe p from the testimony. […] But I doubt that believing p on the basis of testimony requires believing that the attester said that p, anymore than understanding a sentence which says that p requires believing that the sentence says that p. Surely the testimonial acquisition of beliefs does not require the mind to keep double semantic books (1997, pp. 408–9).

In Audi’s view, then, perceptual appearances of others’ reports (as contrasted with occcurrent beliefs about what others have reported) may well be the only report-receptive cognitive events that ever initiate testimonial belief-forming processes—even if the testimonial beliefs formed by these processes require (by way of structuring causes) dispositions to form beliefs about, or presuppositions concerning, what others have reported. And with his comments about both the necessity of perceptual appearances for testimonial belief formation and the keeping of “double semantic books,” we have a challenge to the prima facie case against basicality. The challenge seems to go along the following lines:Footnote 11

Audi’s challenge (to the prima facie case against basicality)

P1:

Testimonial belief-forming processes require perceptual appearances of others’ reports.

P2:

If testimonial belief-forming processes require perceptual appearances of others’ reports, perceptual appearances of others’ reports always initiate testimonial belief-forming processes.

C1:

Perceptual appearances of others’ reports always initiate testimonial belief-forming processes.

P3:

If perceptual appearances of others’ reports always initiate testimonial belief-forming processes, then occurrent beliefs about what others have reported never initiate testimonial belief-forming processes (for this would involve double bookkeeping of the contents of reports received).

C2:

Occurrent beliefs about what others have reported never initiate testimonial belief-forming processes.

Notice that ‘perceptual appearances of others’ reports’ (or ‘perception of what is said’, in the quote from Audi) is ambiguous: it may mean nonsemantic perceptual appearances of others’ reports—that is, perceptual appearances merely of the speech acts or phrase tokens (particular verbal utterances, inscribed sentences, etc.) that encode the contents of others’ reports—or it may mean semantic perceptual appearances of others’ reports—perceptual appearances of the encoding linguistic items as expressing the contents they encode [cf. the sort of semantically loaded perceptual events that Siegel (2006) defends]. P1 of Audi’s challenge is apt to seem unassailable where this ambiguity is unnoticed or glossed over: virtually everyone on every side of the reductionist–antireductionist–inferentialist–noninferentialist debates would agree, at any rate, that testimonial belief-forming processes cannot occur in the complete absence of any sort of perceptual contact with others’ reports—without either nonsemantic perceptual appearances or semantic perceptual appearances of these reports. Where disambiguation occurs explicitly in favor of one or the other reading of ‘perceptual appearances of others’ reports’, P1 is likely to have more detractors; but perhaps the premise would still have enough plausibility to make the argument in which it appears worthy of attention.

If ‘perceptual appearances of others’ reports’ in disambiguated in favor of the nonsemantic reading, however, it seems clear that P2 of Audi’s challenge must be rejected, because mere nonsemantic perceptual appearances of others’ reports cannot have the report-receptive nature required of the cognitive events that initiate testimonial belief-forming processes. Mere nonsemantic perceptual appearances of others’ reports cannot present as true contents about what others have reported; they can only present as true contents about the “low level” sensory qualities—the shapes, sounds, colors, etc.—of the speech acts or phrase tokens whereby others report what they do. Hence even if such nonsemantic perceptual appearances are (as per P1) always required for testimonial belief-forming processes, they cannot ever (let alone always) be the events that initiate testimonial belief-forming processes.

To rescue P2 of Audi’s challenge, then, the semantic reading must be adopted, for at least on the semantic reading the perceptual appearances of others’ reports can have the requisite report-receptive nature and hence serve as the initiating events of testimonial belief-forming processes. Adopting the semantic reading, moreover, boosts the plausibility of P3. It certainly seems right that if testimonial belief-forming processes are always initiated by semantic perceptual appearances of others’ reports and sometimes initiated by occurrent beliefs about what others have reported, then we would in some cases be landed with an unparsimonious bookkeeping of report contents by two distinct sorts of report-receptive cognitive event: the semantic perceptual appearances would keep track of report contents, and the occurrent beliefs would (needlessly) do so as well. The suggestion of such double bookkeeping smacks of ontological extravagance. The occurrent beliefs would do no real work if the semantic perceptual appearances were there anyway, and the occurrent beliefs would thus be prime candidates for excision by Ockham’s razor.

But even with P2 so rescued, a serious problem remains for it: testimonial belief-forming processes may require semantic perceptual appearances of others’ reports only because they require the semantic perceptual appearances as triggering causes of occurrent beliefs about what others have reported, or as structuring causes of the terminating testimonial beliefs, where occurrent beliefs about what others have reported alone stand as the initiating events of the processes. On either option, testimonial belief-forming processes would require semantic perceptual appearances of others’ reports, yet the semantic perceptual appearances would never (let alone always) initiate the processes; hence on either option, P2 of Audi’s challenge would be false.

The first option is represented by the following figure (with the rightward arrows representing discrete, simple belief-forming processes) (Fig. 1). And the second option is represented by the next figure (with the upward arrow representing the structuring cause relation) (Fig. 2).

Fig. 1
figure 1

First Option

Full size image
Fig. 2
figure 2

Second Option

Full size image

On the first option, semantic perceptual appearances of others’ reports might well be reasons for testimonial beliefs by virtue of being perceptual reasons for beliefs that are testimonial reasons (given the transitivity of the basing relation considered in Sect. 2, and assuming that the complex belief-forming processes they initiate are discrete), but they would not themselves be testimonial reasons, for they would not be the initiating events of the simple processes that are testimonial belief-forming processes; only occurrent beliefs about what others have reported would on the first option count as testimonial reasons. On the second option, semantic perceptual appearances of others’ reports would not even count as reasons for testimonial beliefs, let alone testimonial reasons, for they would merely be structuring causes of testimonial belief formations. As on the first option, semantic perceptual appearances of others’ reports, though required for testimonial belief-forming processes, wouldn’t (ever) initiate those processes.

A central problem with Audi’s challenge to the prima facie case against basicality, then, is that nothing in the argument provides us with a good reason to discount either of these options,Footnote 12 and in the absence of any such reason, P2 has nothing significant to be said in its favor.

A further problem with Audi’s challenge can be seen by considering what C1 and the double bookkeeping rationale for P3 jointly commit us to (keeping with the semantic reading of ‘perceptual appearances of others’ reports’ required to rescue P2). Together, C1 and the double bookkeeping rationale do not merely commit us to C2; together, it seems, they commit us to the claim that all testimonial reasons are perceptual appearances of a certain sort (viz. semantic perceptual appearances of others’ reports), which is even stronger than the basicality claim according to which they are all nondoxastic cognitive events. Given C1, the double bookkeeping rationale seems to preclude report-receptive cognitive events of any type other than report-receptive perceptual appearances from serving as the initiating events of testimonial belief-forming processes. Suppose, for example, one were to take nondoxastic nonperceptual “experiences of understanding” (cf. Hunter 1998; Pitt 2004) as report-receptive cognitive events; given C1, the double bookkeeping rationale for excluding occurrent beliefs as initiating events of testimonial belief-forming processes would seem to apply to these putatively report-receptive cognitive events as well: if testimonial belief-forming processes were sometimes initiated by both report-receptive semantic perceptions of others’ reports and report-receptive experiences of understanding, we would have an unparsimonious bookkeeping of report contents that is best avoided.

Otherwise put, C1 and the double bookkeeping rationale for P3 of Audi’s challenge seem jointly to commit us to the view that the option represented by the following figure (whether or not upward arrows representing structuring causation are added) is the only possible option as far as testimonial belief-forming processes go (Fig. 3).

Fig. 3
figure 3

Third Option

Full size image

The problem is that if this is the only possible option, testimonial belief-forming processes seem to end up as mere species of perceptual belief-forming processes and testimonial belief a mere a species of perceptual belief. For if this is the only possible option, those processes appear to count as mere species of perceptual belief-forming processes (they are simple belief-forming processes initiated only by perceptual appearances of some sort, and hence their initiating events are thoroughly perceptual); and if this is the only possible option, the belief-formations in which the processes terminate—formations of testimonial belief—appear to count as mere species of perceptual belief formations.

I take this to be a very serious problem for Audi’s challenge because I take the suggestions that testimonial belief-forming processes are mere species of perceptual belief-forming processes and that testimonial belief is a mere species of perceptual belief to be very implausible. Although (as the debate about inferentialism shows) it is a live issue in the epistemological literature whether testimonial belief-forming processes are mere species of inferential belief-forming processes, and hence whether testimonial belief is a mere species of inferential belief, it is not a live issue whether testimonial belief-forming processes are mere species of perceptual belief-forming processes, or whether testimonial belief is a mere species of perceptual belief. Nobody in the literature, including Audi himself (outside of his challenge to the prima facie case against basicality, that is) seems prepared to accept such “mere species” views about the relationship between testimonial belief-forming processes and perceptual belief-forming processes and about the relationship between testimonial belief and perceptual belief. Antireductionists’ and noninferentialists’ fondness for pointing out analogies between the testimonial and the perceptual is always hedged by acknowledgments of disanalogies between the two (e.g. Coady 1973, pp. 154–5; 1993, pp. 146–8; Audi 1997, pp. 408–9; 2013, pp. 512–3), and no antireductionist or noninferentialist I’m aware of goes so far as to claim that the analogies are underwritten be a mere species relation.

7 Conclusion

Audi’s challenge to the prima facie case against the basicality of testimonial reasons thus fails in at least two ways: it assumes for no good reason that if testimonial belief-forming processes require perceptual appearances of others’ reports, they are always initiated by perceptual appearances of others’ reports; and it seems to yield the deeply counterintuitive consequences that testimonial belief-forming processes are mere species of perceptual belief-forming processes and that testimonial belief is a mere species of perceptual belief. In the absence of any other challenge in the literature to the prima facie case against basicality, therefore, that case stands and the viability of inferentialism is preserved.