To what extent can we say that a speaker knows rule R of his or her language rather than rule R', given that both rules produce the same grammatical outcome? If rule R' provides a more general, technically precise formulation of the same conditions formulated by rule R, do we ascribe knowledge of R' to the speaker -- even if the speaker admits only to knowing R? Assuming that a third-person report drawing on the best available theory should take precedence over the speaker's own first-person report, Chomsky claims we can and should ascribe knowledge of the more precise rule to the speaker. I argue that while the third-person reports offered by observers drawing on the best available theories provide standards by which a given behavior may be evaluated, corresponding first-person accounts must be taken into consideration as criteria of assertibility constraining what we may conclude about the person's actual knowledge.
[1] Given the following two choices:
Which is a native English speaker -- call him or her S -- most likely to produce? It should be fairly obvious that he or she would likely produce the grammatically correct sentence A. What may not be so obvious is his or her reason for choosing A over B.
[2] Chomsky explains the choice by citing the speaker's knowledge of the appropriate rule. In rejecting the grammatically incorrect sentence B, Chomsky claims, speaker S shows that he or she "knows that verbs cannot be separated from their objects by adverbs" (KL 266). Call this "rule R." But because he holds that the prohibition of such adverbial intervention is a consequence of the more general rule of strict adjacency, Chomsky goes further and claims that what S really knows is that "the value for the case assignment parameter in E is strict adjacency" (KL 266, emphasis in the original). Call this "rule R'." Both rule R and rule R' describe S's behavior. But are we justified in claiming that S in fact knows rule R'?
[3] It would be helpful first of all to clarify what Chomsky means by knowing a rule. Extrapolating from behavioral evidence, Chomsky claims (with some more or less weak provisos, e.g., KL 265) that if a speaker's utterances conform to the conditions specified by a language rule, then that speaker knows the rule. In short, a speaker who observes a rule can be said to know that rule (KL 268). In addition, Chomsky claims that knowing a rule of language is an instance of knowing-that, and therefore involves propositional knowledge (KL 265). Thus according to Chomsky, if S acts in accordance with rule R of his or her language, then S knows rule R and therefore knows that R (KL 266).
[4] Chomsky's claim that knowledge of language is knowing-that has an important corollary. That is that the person who knows rule R not only knows that R, but believes that R (e.g., RR 93). Ascription of knowledge of language to a person therefore entails a corresponding ascription of belief to that person. When we state that someone knows a language rule, we are in effect making a statement about his or her attitude (belief) toward the propositional content embodying the language rule.
[5] It seems to me that in cases in which knowledge of language is ascribed, we are justified in recasting talk about knowledge into talk about beliefs. That is because what interests us is not whether or not a given rule is true, i.e., whether or not it accurately describes the appropriate language behavior, but rather, whether or not it is considered true of his or her language by the speaker. Our ascription of knowledge of the given language rule thus involves a statement about S, specifically, about how things are with him or her as demonstrated by his or her attitude toward the relevant proposition(s). For that reason, framing the question in terms of S's beliefs is perfectly legitimate, and shows exactly what is at stake when we ascribe knowledge of language to a person.
[6] On the basis of the foregoing, I would suggest that for any language rule R, knowing that R means the following:
We can say that S knows R if S believes the propositions comprising R. If R can be stated as p, then S knows R if S believes that p. Further, for S to believe that p is for S to be disposed normally to feel/hold/agree that p.
[7] Applying this to the example introduced at the beginning of this paper, we would say that S's knowing R means that
[8] By the same token, S's knowing R' means that
[9] Further, if we claim that S produces sentence A and not sentence B because S knows that R, we are in effect asserting that S's producing A comes about by virtue of S's cognitive/doxastic state having a certain content. Or:
[10] The general claim here is that language behavior takes the form it does by virtue of the content of the cognitive/doxastic state that enters into/supports/underlies that behavior. Conversely, the actual content of that cognitive/doxastic state would represent the speaker's knowledge of language and would explain why he or she produces the appropriate utterances.
[11] Given the above definition of what it is to know, and the connection between the content of a cognitive/doxastic state and the role it plays in the explanation of behavior, the question becomes: Which formulation of a rule describes the speaker's actual belief(s), and which formulation simply describes a set of conditions to which the speaker's behavior (unknowingly) conforms?
[12] Chomsky offers an answer that can be called the argument from the best theory. He states that
[W]e are entitled to propose that the rule R is a constituent element of Jones's language (I-language) if the best theory we can construct dealing with all relevant evidence assigns R as a constituent element of the language abstracted from Jones's attained state of knowledge (KL 244).
[13] Chomsky's argument is that if our best theory for explaining a speaker's behavior includes attributing to him or her knowledge of a given rule, then we should conclude that this knowledge does in fact enter into the speaker's behavior (KL 249), and that the speaker therefore does know the rule. Implicit in this argument is the provision that a third person attribution of knowledge of language has authority over a first person report, if the third-person attribution is made on the basis of the best possible theory available. From this point of view, the third-person claim that another person's language behavior implicates a given cognitive/doxastic content is true simply by virtue of the claim's having been derived from the best available theory. Because it is derived from the best available theory, the third-person attribution must take precedence over any relevant first-person account.
[14] But this does not tell us whether or not S knows rule R in the required sense of believing that R. It does not, in other words, tell us whether or not S holds the requisite attitude -- belief -- in relation to the propositions and constituent concepts embodying the rule he or she is said to know. This is what we need to ascertain, but how?
[15] It is informative to note in this regard a point that Searle has raised in general objection to Chomsky's claim that speakers are (actually, in fact) following the rules he and other grammarians have formulated. According to Searle, for any attribution of rule-following, we need to show that the attributed rules are "rules that the agent is actually following, and not mere hypotheses or generalizations that correctly describe his behavior." For Searle, the argument from the best theory does not suffice, since the descriptive or predictive accuracy of the attributed rule does not by itself prove that the rule is in fact being followed. We need, instead, "some independent reason for supposing that the rules are functioning causally" (quoted in KL 247).
[16] There seem to be two points bundled into Searle's objection. The first, which Searle explicitly makes, is that behavior which seems to be in accord with a rule must be shown to be guided by that rule in fact and not simply hypothetically. More generally, if we are to claim that a person is behaving in a certain way on account of his or her given cognitive/motivational content, we must show that this given cognitive/motivational content does in fact enter into the production of the behavior in the specified manner.
[17] The second point, which Searle doesn't make but which I find implicit in the call for obtaining an independent reason for attributing a rule, is that the person to whom such rule following is attributed should (somehow) understand him or herself to be following the rule. This would mean (among other things) that he or she should show evidence of believing that R, for the given attributed rule R. Such evidence could be found in the appropriate first-person avowal of belief or acceptance that R; such a first-person avowal would in fact constitute an independent reason for attributing actual, as opposed to hypothetical, rule-following to that person, if we read "independent" to mean something like "coming from a source other than the person doing the attributing." Like the previous point, this point can be generalized. Given cases in which it is claimed that a speaker knows a given rule of language, we would want independent corroboration of that claim.
[18] A common sense attempt to corroborate a knowledge claim would have us solicit a first person report from the speaker him- or herself. We might, for instance, ask the speaker to describe what, if any, language rule he or she understands him- or herself to be following in producing a given utterance. With this evidence, we would be able to determine whether or not our hypothesized ascription of rule-following (and with it the corresponding ascription of belief) is accurate.
[19] This first approach would require the speaker to be able to convey to us on his or her own why his or her language behavior exhibits the regularity observed of it. But there are prima facie two problems here. First, it seems clear that not all speakers can formulate the rules their language behavior seems to conform to, and second, not all speakers are aware of their reasons for producing utterances in the form that they do.
[20] But neither of these considerations should be taken to mean that S necessarily cannot give us the kind of testimony we would want. In the first place, the inability to state or otherwise express a rule is not necessarily evidence that one does not know (or would not recognize) the rule any more than the inability to describe a concept is evidence that one does not know (would not recognize) the concept. And in the second place, one's not being aware of one's reasons for behaving in a given way is not necessarily evidence that one does not know why one behaved in the given way. Many actions do not ordinarily require a high degree of attentiveness to the conditions of their production in order to succeed. This is obviously true of performances based on, e.g., physical skills, which often require little or no attentive monitoring for their success. But it is also true of more formalized behaviors, language performances among them. For example, a speaker may concentrate attention on what he or she is saying and apparently not think about the syntactic conditions his or her utterance must meet. And yet afterward the speaker may acknowledge that he or she did indeed mean to conform to the appropriate syntactic rule. In any case, it seems reasonable to suppose that a speaker initially inattentive to the reasons behind his or her syntactic behavior can at least in principle become aware of them and may impart that awareness to others. The question is how.
[21] On the face of it, at least, introspection would appear to be the most obvious way for the speaker to gain such awareness, since what we are concerned with here are psychological facts which, one would think, could be discovered through one's focusing attention on one's own inner states. But Chomsky rejects introspection, claiming that it can tell the introspector neither that the given rule holds nor that the rule enters into the appropriate "mental computations" involved in language production (KL 269).
[22] But introspection does not exhaust all possible avenues for securing the kind of first-person evidence we would want to obtain. We could state the rule we think S is following, and ask S whether or not he or she would accept this as the correct description of the reason he or she produced the given utterance. We could likewise present S with different formulas presenting under different descriptions the same linguistic regularity observed of S, and ask him or her to choose which one correctly describes his or her understanding of why he or she conformed to that regularity. We might, to return to our previous example, show S rules R and R', and ask him or her which one, if any, describes his or her understanding of why sentence A is preferable to sentence B. No matter which specific approach would be taken, the crucial criterion would be that ascription to S of knowledge of a rule be contingent upon S's recognizing and agreeing to the propositions contained in the rule.
[23] Thomas Nagel has in fact suggested something like this. As he puts it,
So long as it would be possible with effort to bring the speaker to a genuine recognition of a grammatical rule as an expression of his understanding of the language, rather than to a mere belief, based on the observation of cases, that the rule in fact describes his competence, it is acceptable, I think, to ascribe knowledge of that rule to the speaker. ( OM 60, emphasis in the original).
I will call this the acceptance condition. In essence, the acceptance condition holds that
An ascription of knowledge to a person should be contingent upon the acceptance by that person of the appropriate propositions and/or concepts as accurately articulating what he or she believes.
[24] Generally, we can ascribe belief B to S if S, when B is brought to his or her attention, feels/holds/agrees that B. Without S's feeling/holding/agreeing that B, we could not confidently ascribe B to S. In addition, S's feeling/holding/agreeing that B can consist in the recognition that B or the acquisition of the attitude that B.
[25] Thus for S to accept that rule R correctly reflects what he or she knows about the appropriate aspect of language, S must either recognize that R or acquire the belief that R. If S were to recognize that R, then S would simply be exercising an already-existing disposition to normally hold/feel/agree that R, given the appropriate circumstances. If S were to acquire the belief that R, then S would, on the basis of, e.g., evidence presented, become disposed to hold/feel/agree that R is the case, given the appropriate circumstances. In other words, when we recognize that R, we are exercising or expressing a belief we already have, though perhaps we never had the need or opportunity to do so before. When we are brought to accept that R, we are acquiring, and consequently expressing, the belief that R. Note that in either case, the acceptance condition involves a first-person avowal of belief.
[26] Note also that it is not necessary that the speaker come to this avowal through reflection or "introspection" or otherwise on his or her own. If a rule is described to the speaker, and the speaker agrees that he or she believes (or is brought to believe) that the rule holds in the appropriate circumstance, then it is reasonable to attribute to him or her knowledge of that rule. But it is also true that if the speaker does not recognize or accept the rule as articulating something he or she believes or has come to believe, then the plausible attribution to him or her of knowledge of that rule would be difficult to maintain.
[27] Would Chomsky agree to make the ascription of knowledge of language rules contingent on the acceptance condition? On the one hand, he seems to accept a scenario in which a speaker comes to know the rules of grammar "from the outside" -- that is, by having them taught or otherwise brought to his or her attention by another party (KL 269). On the other hand, his position on the usefulness of the first-person perspective generally is that it isn't. His view seems to be based not only on his own belief that much knowledge of language is tacit, but on the widely recognized observation that first-person accounts are inherently unreliable. If we examine these two points, however, we will find that, rather than invalidating the first-person perspective altogether, they serve only to qualify the claims that can be made for it.
[28] If, as Chomsky claims, knowledge of language is largely tacit, then claims regarding a speaker's knowledge of a given rule may be a difficult matter to decide from the speaker's point of view. Given Chomsky's understanding of tacit knowledge as knowledge that is "generally inaccessible to consciousness" (KL 260) and therefore presumably opaque to the knower, it is easy to see how it would be difficult to make knowledge ascription contingent on the appropriate first-person avowal. But this difficulty may be more apparent than real.
[29] First, tacit knowledge as Chomsky understands it would appear to differ very little from ordinary knowledge outside of its being tacit (e.g., KL 269). Chomsky does not claim that a speaker's tacit knowledge of language is inferentially isolated from his or her other attitude states, and in fact he has stated that speakers' decisions to use their tacit knowledge are influenced by their "goals, beliefs, expectations, and so forth" (KL 261). Far from existing behind a kind of firewall separating it from ordinary beliefs and other attitude states, tacit knowledge of language would seem to be woven into the speaker's overall network of attitude states, and to exert some variety of influence on -- as well as to be influenced by -- those states.
[30] Second, ordinary beliefs themselves may be largely tacit. As indicated above (1.2.), beliefs are to some extent dispositional. Following Cohen (6), our having consciously thought about or avowed a belief is a contingent rather than a necessary feature of beliefs. This means that, as with tacit knowledge, we may "have" beliefs without necessarily having consciously thought about them. Nevertheless, when a belief of ours is brought to our attention, we do, under ordinary circumstances, tend to recognize it as such. There is no reason this cannot hold for tacit knowledge as well. In fact all that would be necessary for us to say that someone knew (believed) something, whether tacitly or not, is that when confronted with a statement or other formulation of the belief, that person should be disposed normally to feel/hold/agree that it is true.
[31] It may be objected here that the acceptance condition is contingent on the belief's accessibility to consciousness, and that tacit knowledge is, by definition, inaccessible to consciousness and therefore exempt from the acceptance condition. Again, there is no reason to suppose that tacit knowledge cannot behave like ordinary dispositions to believe, and thus to be brought to awareness given the proper circumstances. Certainly, Chomsky's statement that one can come to know initially tacit rules "from the outside" would seem to indicate his acknowledgement that one could at least in principle have conscious access to one's tacit knowledge. If this is so, then there is no reason in principle that tacit knowledge must remain tacit and thus exempt from the acceptance condition. We might say then that tacit knowledge of language is tacit to the extent that it is initially inaccessible to the person to whom it is attributed, but that given the proper conditions, this inaccessibility can be converted to the kind of accessibility enjoyed by our ordinary knowledge and thus can be brought into play in relation to the acceptance condition.
[32] As mentioned above, Chomsky believes that first person reports regarding what one thinks one is doing are not always reliable. As he puts it, "We might ask Jones what rule he is following, but...such evidence is at best very weak because people's judgments as to why they do what they do are rarely informative or trustworthy" (KL 254). There is truth to this assertion, but a closer look is warranted.
[33] What Chomsky seems to be referring to here is the normal indeterminacy that may and often does characterize an agent's first person accounts of his or her reasons for performing in a given way. Such indeterminacy may be a product of any or all of a number of factors, including the relative attentiveness with which one does something, the degree of fine-grainedness or explicitness demanded of the first-person account, and the fact that internal states are not objectively separate from the first-person perspectives form the basis of reports about those states. Absolute certainty here is out of the question -- but that does not in and of itself invalidate first person accounts.
[34] In fact, I would be inclined to understand the indeterminacy of first person reports as analogous to the underdetermination of theory by evidence. Because of the latter, we cannot (and can never) be certain that the evidence pointing to certain theoretical conclusions is absolutely conclusive. But -- and Chomsky has argued this point against Quine (KL 13 n. 5) -- such underdetermination does not in and of itself automatically invalidate any reasonable conclusions we may feel we are warranted in drawing from the evidence. Just because it is possible that our conclusions will be proven wrong by more or better or subsequent evidence does not mean that we are not justified in drawing the most reasonable conclusions we can based on the evidence available to us. A similar case can be made for the value of first-person accounts. They may be far from infallible, but because they represent expressions or manifestations of what one thinks is the case with oneself, they constitute admissible evidence regarding a person's attitude states.
[35] In fact it reasonably can be held that because they tell us how things are with a person from that person's point of view, first-person reports have a certain privileged status in instances where we are trying to determine someone's attitude toward a given proposition or set of propositions. Evidence regarding what someone thinks he or she believes would seem to be especially relevant if we want to determine whether or not that person knows (believes) a given rule of language. It seems to me that in this case a first-person account would make for useful evidence that should not be ruled out on a priori grounds.
[36] It is useful in this context to think of first-person reports in terms of the Wittgensteinian notion of criteria. Criteria, briefly, are normative considerations that provide grounds for justifying assertions, and thereby help to set conditions under which assertions are appropriate. By this reading, first-person reports would provide the criteria of assertibility constraining third-person ascriptions. First-person reports would, in other words, set out the normative conditions under which third-person ascriptions would be deemed appropriate or not, and would thus serve to restrict the range of attitudes we can (reasonably) ascribe to others. Since, as Finch has pointed out, such criteria serve to help fix "what we may be wrong about, deceived about, under an illusion about" (62 quoted in McDonald 61), then first-person reports would exert a potentially limiting influence on third-person ascriptions. Specifically, they would show how claims made from the third-person perspective may fall outside the range of possible understandings that reasonably can be attributed to the person in question.
[37] By this light, the acceptance condition would act as the relevant criterion for ascribing knowledge of a given language rule. A third-person ascription that met the acceptance condition would, all things being equal, be considered a justified ascription. Conversely, a third-person ascription that did not meet the acceptance condition would, all things being equal, be difficult to justify. Thus S's accepting rule R as expressing his or her reason for uttering sentence A rather than sentence B would provide justification for ascribing knowledge of R to S. If on the other hand S did not accept R as expressing his or her understanding of the appropriate language behavior, then by the criterion of the acceptance condition we would not be justified in ascribing knowledge of R to S.
[38] In spite of their built-in indeterminacy, then, first person reports and avowals would seem, for better or for worse, to be the relevant criteria by which to check the plausibility of third person ascriptions of knowledge. Consequently, first-person reports and avowals would be useful in cases where we wish to adjudicate apparently conflicting claims regarding what a given speaker knows.
[39] It is easy to see how such conflict could arise. Take, again, the example of S's producing sentence A rather than sentence B. This may been explained alternately as being due to S's knowing that R -- i.e., believing that adverbs are prohibited from intervening between verbs and their objects -- or S's knowing that R' -- i.e., believing that the value for the case assignment parameter in E is strict adjacency. Given the two very different sets of propositions comprising these rules, we would appear to have two competing claims regarding the speaker's object of belief.
[40] For even if we agree that rule R is nothing more than a consequence of the more general rule R' (KL 266), it is not at all certain that S would recognize this. Nor is it certain that S would understand the constituent concepts of which R' is comprised, even if he or she understood the constituent concepts of R. Many ordinary speakers of English (and others) know what verbs, objects, and adverbs are, but do not know what strict adjacency is. We could expect these speakers' first-person accounts of why they produced sentence A rather than sentence B to be put in terms of verbs, objects, and adverbs, and not in terms of strict adjacency. Accordingly, we reasonably could expect that linguists (and perhaps even only a subset of linguists) would be disposed to describe S's language behavior in terms of strict adjacency, but that S, as an ordinary speaker, would not. As Chomsky concedes, many people may be reluctant to attribute knowledge of R' to S on account of the "unfamiliarity of the notions Case assignment and adjacency parameter" (KL 266-267). Chomsky, of course, does not hesitate to claim that S does indeed know strict adjacency. But his willingness to acknowledge others' reluctance to grant this point is interesting.
[41] Still, Chomsky holds that the unfamiliarity of the concepts used to explain S's behavior is "irrelevant to the description of [S's] state of knowledge" (KL 267). What would seem to matter here is only that the concepts belong to the best available theory, in which case we must assume that they accurately reflect S's state of knowledge. But it seems to me that what really is at issue here is not the relative familiarity of the concepts per se, but rather whether or not these (or other) concepts are properly part of S's repertoire of beliefs about the language. It seems reasonable to suppose that S cannot have the requisite attitude toward concepts that he or she cannot be said to possess. If that is the case, then S's familiarity with the given concepts is hardly a matter of indifference. As Davies has pointed out in a similar context, whether or not a person understands the concepts he or she is said to know is indeed a relevant consideration (Davies 135).
[42] In fact, it is difficult to see how one's not understanding a concept one is said to know can be irrelevant to deciding whether or not one knows a rule or proposition in which that concept figures. Consider the following example: I drink water because I am thirsty and I know that water will quench my thirst. But the best theory of why I drink water goes something like this: when I drink water, the water is absorbed into my bloodstream by osmosis as it enters my stomach. This causes both my blood volume and pressure to increase, and the osmotic strength of my blood to be restored to a normal level. Because this is the best theory, does that mean that I drink water because I know what that theory states?
[43] If we take a position analogous to the position Chomsky takes regarding knowledge of the rules of language, it seems to me we would have to answer "yes." Just as Chomsky holds that S's producing sentence A is guided by his or her knowing that the value for the case assignment parameter is strict adjacency, we would hold that my drinking water is guided by my knowing that the intake of water works through osmosis to cause blood volume and pressure to rise, and osmolarity to reach the proper level. In both of these cases, we would be claiming that the person in question knows what the best theory available states about the reasons for his or her behavior, and that this knowledge enters into the relevant mechanisms for producing the behavior.
[44] But do I in fact know this technical explanation for my drinking water? Again, it is drawn from the best theory available, and certainly, my behavior is perfectly in accord with what it would be if I did know what the theory describes. But the fact is that I did not know the theory (nor for that matter had I even heard of the term osmolarity) until I asked an expert. I did not, in other words, have the requisite familiarity with the propositions I would have to have if I could be said to know what the theory states.
[45] My having consulted an expert raises a crucial point. For, according to Chomsky, my knowledge includes what is known to experts within my speech community. Citing Putnam's ( REP 19-41) notion of the division of linguistic labor (KL 18), Chomsky asserts that the meaning of a term may be expressed in terms of the specialized knowledge of others in my speech community (KL 267). By virtue of my being a member of a given speech community (presumably, in this case, speakers of English), in other words, my knowledge of language encompasses the best theories as formulated by the appropriate experts.
[46] But if, as I believe we should, we are to agree that the acceptance condition sets legitimate assertibility criteria constraining knowledge ascriptions, we cannot automatically attribute the experts' knowledge to any given member of a speech community. Recall the definition of knowing introduced above (1.2.): even given that S belongs to a speech community in which R' is accepted as the best available explanation of a particular language behavior, we still would have to show that S him- or herself stands in the proper attitude to the propositions and concepts making up R'. It is not enough that someone from his or her speech community stands in such an attitude; he or she must him- or herself stand in that attitude.
[47] In light of this, I believe we can reconceive the relationship between S and rule R' of the best (yet unfamiliar) theory explaining S's language behavior. Assuming that ascription of knowledge of R' to S is unjustified given the acceptance condition and the corresponding criteria of assertibility set by S's first-person reports, we can say that, because R' is potentially available to S by virtue of its arising from the best theory available to the relevant experts in S's speech community, R' is the standard against which S's knowledge can be measured. This is not to say that S knows R' but rather that S's state of knowledge can be brought to a level such that S will accept R' as the correct explanation of the given language behavior.
[48] Like first-person criteria of assertibility, third-person standards of explanation cast our assertions of knowledge and avowals of belief in a normative light. My first-person report of why I think I behaved in a given way may be an adequate account of my own beliefs on the subject, but it may fail utterly as an adequate explanation of that behavior -- in the context of the most advanced or accepted thinking on the subject. The upshot of this is that we must think of the best third-person ascriptions of knowledge as hypotheses embodying explanatory standards that people may (or perhaps should) meet in the appropriate context.
[49] This last qualifier is crucial, for there is a degree to which the adequacy of a response will be gauged in terms of the analytical or explanatory framework within which it is elicited. There may be circumstances in which "Because I knew it would quench my thirst" would be a sufficient answer to the question "Why did you drink that glass of water?" Similarly, it can be argued that there may be contexts -- the teaching of grammar to children, for instance -- in which the preferability of sentence A to sentence B is better explained in terms of verbs, adverbs, and objects rather than in terms of strict adjacency.
[50] In a general sense, third-person standards and first-person criteria set certain conditions that our assertions and avowals may meet. Explanations drawn from the best theories provide the standards toward which our own state of knowledge and repertoire of beliefs may aspire. Criteria of assertibility derived from first-person reports and avowals provide conditions placing constraints on what third-person ascriptions may hold. Thus even if the best theories for explaining behavior serve as standards to which knowledge of that behavior can aspire, first-person accounts still must be factored in as legitimate constraints on the range of third-person ascriptions.