2.1. Literalism and contextualism

The labels literalism and contextualism demarcate the two poles between which there are other possible positions (for a discussion of the full range, see Borg, Reference Borg2012; Cappelen & Lepore, Reference Cappelen and Lepore2005; and Recanati, Reference Recanati2004). Literalists hold that the truth-conditions of sentences are determined by the meanings of their parts and their syntax, independently of the different contexts of use in which one might utter such a sentence. Emma Borg defines literalism, with a common conventionalist slant (statement iv.),Footnote ⁷ as the conjunction of the following four commitments:

1. i. Semantic content for well-formed declarative sentences is truth-evaluable content.

2. ii. Semantic content for a sentence is fully determined by its syntactic structure and lexical content: the meaning of a sentence is exhausted by the meaning of its parts and their mode of composition.

3. iii. There are only a limited number of context-sensitive expressions in natural language.

4. iv. Recovery of semantic content is possible without access to current speaker intentions. (Borg, Reference Borg2012, 4–5)Footnote ⁸

The literalist does not deny that there are context-sensitive terms in natural languages; she is free to acknowledge that the sentence ‘I am hungry’ has different truth-conditions depending on who utters it. She simply maintains that there is a limited number of such context-sensitive terms in natural languages (for one such list, see Cappelen & Lepore, Reference Cappelen and Lepore2005, 1) and that, in most cases, words have the same conditions of correct application across contexts.

Contextualism, on the other hand, is the view that every term (or nearly every term) is context-sensitive.Footnote ⁹ The contextualist thus believes that even well-formed declarative sentences without any paradigmatically context-sensitive terms (e.g., ‘Perrin ordered coffee’ or ‘Lucy opened her store’) do not have determinate truth-conditions independently of specific contexts of use. ‘Open’ can be used in such a variety of cases (I can open a door, open a box, open my mouth) that for any sentence which includes ‘open’, we can only get at the proposition expressed by it by getting at the specific proposition that the speaker intended to communicate in that context. The same context-sensitivity extends even to such mundane terms as ‘coffee’ and ‘tea’. So, e.g., when Matt tells Rand “Perrin ordered coffee” in Starbucks, Matt’s utterance is true if and only if Perrin ordered a cup of hot coffee in a to-go cup, while the same sentence takes on different truth conditions in the context of a restaurant manager talking about the bulk order Perrin just placed with their food supplier. The most radical contextualists will deny that words have meanings at all; they suggest instead that words express something like a partial or schematic concept which has “semantic potential” instead of determinate conditions of correct application (Recanati, Reference Recanati2004, 97).Footnote ¹⁰

In what follows, I will not argue against contextualism; I assume that Davidson does not intend to be a contextualist and that the interest of “Derangement” comes from its purporting to present an anti-conventional literalist position. So, Lepore and Stone (Reference Lepore and Stone2017) and Camp (Reference Camp2016) will have succeeded if they can show that the anti-conventionalism of “Derangement” commits one to contextualism.Footnote ¹¹

2.2. Conventionalism and anti-conventionalism

Separately from the literalist–contextualist debates, philosophers have also argued about the degree to which language is conventional. With Convention (Reference Lewis1969), David Lewis revived the idea that the meanings of a speaker’s words are fixed by the linguistic conventions of her community.Footnote ¹² To be philosophically controversial, conventionalism cannot just be the descriptive thesis that speakers living in proximity often use many of their words in the same ways, nor the suggestion that sometimes speakers can intend to mean by a word whatever the relevant experts mean by that word. These are theses that the anti-conventionalist can take on board. Conventionalism, rather, is a constitutive claim about what fixes the meanings of a speaker’s words. The conventionalist argues that the meanings of a speaker’s words are fixed by the meanings assigned to those words in the lexicon of her linguistic community:

There are different ways one could develop conventionalism, but Lepore and Stone (Reference Lepore and Stone2017) and Camp (Reference Camp2016) largely work with the classic Lewisian and Burgean models.

The contradictory of conventionalism is anti-conventionalism, the position that the meanings of a speaker’s words are not fixed by conventions. Again, there are multiple ways one could develop anti-conventionalism, two of which I will touch on in Section 6. The main point is that for the anti-conventionalist, what a speaker means by her words can come apart from the meanings her community normally assigns to those words. Thus, to understand what someone means by her words, we must grasp what she takes herself to mean by those words rather than what those words are typically used to mean (by her or by others).

4.1. Malapropisms and antecedently established lexicons

Considering malapropisms, Lepore and Stone offer the example of someone who calls a fancy bowl a “soup latrine.” We certainly do not want to assign the typical meaning to the word ‘latrine’ here; however, once we have given up the typical meaning of ‘latrine’, we are faced with an incredible set of possible interpretations. Does the speaker mean bowl, vessel, urn, or any other of numerous possibilities? Lepore and Stone allow that pragmatic considerations can help us narrow down the possible meanings of ‘latrine’ from just anything (e.g., tiger, elm) to liquid-containing vessels, but they argue that pragmatic considerations cannot take us any further. That is, there is not enough evidence available from purely contextual facts to point to one interpretation of ‘latrine’ over the other once we have gotten to the set of liquid-containing vessels. They assume that the concept tureen is the interpretation we want of ‘latrine’. But there is not sufficient evidence available in the context alone to justifiably think that ‘latrine’ means tureen in this instance, rather than, say, bowl. The only way we can make the jump from ‘latrine’ to tureen is if we take the speaker to have attempted to say a different word than ‘latrine’—attributing a performance error to them—and then search for an appropriate word. Despite its phonetic similarities, ‘tureen’ will still only be salient as the intended word if we are assigning it its conventional meaning, tureen. So, Lepore and Stone’s argument goes, if we can successfully interpret malapropisms only by taking them to be performance errors and can consider them as such only by appealing to an antecedently established lexicon the speaker was trying to follow, then it turns out the meaning of a malapropism is fixed by a social lexicon the speaker was intending to follow.Footnote ¹⁸

Lepore and Stone take it that understanding malapropisms as performance errors precludes anti-conventionalism because the anti-conventionalist is not allowed to appeal to an antecedently established lexicon in explaining what has gone wrong in a malapropic utterance. Without an antecedently established lexicon for comparison, there is little sense to the claim that a speaker intended to utter a different word than the one she actually uttered and so no reason to choose the meaning of ‘tureen’ as the interpretation of a malapropic utterance of ‘latrine’. Viewed as an argument against contextualism, this line of thinking might work, as it is not clear that contextualism is consistent with speakers employing context-independent stable lexicons. However, the anti-conventionalist can maintain that there is an antecedent lexicon by which to judge whether someone has made a performance error or used her word in a way that is innovative or non-standard.Footnote ¹⁹ If malaprops are failed attempts to speak as one intended (or even just as one usually does), then we could get to the bottom of what Mrs. Malaprop meant by ‘latrine’ by looking to our prior theory for her, identifying the word we think she intended to say, and then asking something like “Did you mean to say ‘tureen’?.” If she blushes in embarrassment and says “Yes, of course,” then it makes sense to interpret her utterance of ‘latrine’ as contributing to the meaning of the sentence in just the same way that ‘tureen’, as normally uttered by Mrs. Malaprop, would have.

It is critical to note that the lexicon that we hold up as the standard by which to judge if Mrs. Malaprop has made a performance error need not be a social lexicon. That is, even if Mrs. Malaprop intended to utter ‘tureen’, the meaning of her intended utterance is not necessarily fixed by conventions, despite her attempt to use ‘tureen’ in an expected, common way. This is denied by Lepore and Stone, who conclude that “in these cases [the malapropic speaker] has a completely ordinary communicative intention, in which the plan is to use the conventional articulation of the words” (Lepore & Stone, Reference Lepore and Stone2017, 252). However, the idea that ordinary communicative intentions are intentions to mean whatever one’s community means by the uttered words is not defended. In the above story, it could well be that what Mrs. Malaprop intended to utter was “That’s a nice soup tartine” and that by ‘tartine’ she normally means what most of us mean by ‘tureen’. That is, a malaprop can be a performance error wherein the intended word is nevertheless being used non-conventionally. The very possibility of this highlights the fact that in a performance error reading of malaprops, understanding what a speaker means by her malapropic utterance is a matter of understanding the meaning the speaker intends to convey by the word she intended to utter. Such an appeal to communicative intentions and an antecedently established lexicon is perfectly in line with Davidson, who writes: “The presence of intentions is important, since it gives content to an attribution of error by allowing for the possibility of a discrepancy between intention and accomplishment” (Davidson, Reference Davidson1992, 112). Prioritizing the speaker’s communicative intentions in understanding her is tantamount to prioritizing the speaker over the community in an account of what fixes the meaning of the speaker’s words.

If the conventionalist is to allow communicative intentions a role in determining what a speaker means by her words, then he must explain how conventions circumscribe what a speaker can intend to mean by her words (as a constitutive matter of fact). The most plausible way to do this is by maintaining that what a speaker can rationally intend to communicate is restricted by how she believes she will be understood by her interlocutor; therefore, ordinary communicative intentions are sensitive to how we believe other speakers use their words. Davidson himself endorses something like this principle when he follows Donnellan, Reference Donnellan1968 in maintaining that a speaker must have a “reasonable belief” that she could be understood as intended in order to mean what she does by her words:Footnote ²⁰

However, this principle does not entail that speakers are bound to speak according to a social lexicon. Different speakers will have different beliefs about what those around them mean by their words, so it is not a social lexicon which guides a speaker’s use of words but her beliefs and expectations about how she will be understood.Footnote ²¹ Furthermore, beliefs about how an interlocutor uses his words do not entail that we must use those words in the same ways; what a hearer is prepared to understand and how he is disposed to use his words in conversation can come apart. For example, a speaker could believe that her current interlocutor does not use ‘hamburger’ to refer to packaged ground beef and yet genuinely use ‘hamburger’ to refer to ground beef based on a reasonable expectation that she will be so understood by her interlocutor (as my wife does). So even if beliefs about how one will be understood constrain how one uses one’s words, it is still facts about the speaker (e.g., her lexicon and her history of interpreting others) and not her community or some social lexicon which determine what she means by her words.

To return to the issue of malapropisms, we are now in a position to see that there should be no issue with the anti-conventionalist appealing to an antecedently established lexicon to explain not just what has gone wrong in the case of a malaprop but how an interlocutor could come to give a malaprop its intended interpretation. Now, on the anti-conventionalist picture I have sketched so far, we might not always be able to track down exactly what our interlocutors mean (perhaps we do not have time to ask them or observe their behavior), so we fall back on tentatively assigning a “conventional” (i.e., what we believe to be typical) interpretation to their words.Footnote ²² Such is fine, for conventions are still not playing an essential role in an account of meaning, only a practical, contingent role in certain communicative circumstances. What plays an essential role in a performance error reading of malaprops is the positing of an antecedently established lexicon that the malaprop-utterer is trying to follow. This, I suspect, is an account of malaprops that the contextualist cannot easily accommodate, as it relies on an appeal to an established lexicon with entries that have determinate meanings. Thus, Lepore and Stone’s arguments concerning the interpretation of malapropisms are best aimed at contextualism, not anti-conventionalism.

4.2. Neologisms and externalism

Similar concerns arise with Lepore and Stone’s handling of neologisms. They argue that neologisms can have their meanings fixed only once there is a regular use of the term in the community: “[the meaning of a neologism] is passed from one speaker to another like links in a chain, and it is the whole network of speakers that work together to lock onto its meaning” (emphasis added; Lepore & Stone, Reference Lepore and Stone2017, 258). A successful neologism will pick something out in the world but the precise extension, and hence meaning, of the neologism cannot be known to the coiner of the term: “The original speaker, who understood and named the phenomenon, just offers the starting point” (Lepore & Stone, Reference Lepore and Stone2017, 258). It is only later, in the history of the use of the neologism, that a linguistic community narrows in on the meaning of the term because it takes time for the community to triangulate on the nature of the picked-out thing/relation and thus it takes time for a clear extension to come into view. Thus, Lepore and Stone argue, anti-conventionalism cannot account for neologisms because a newly coined term cannot be meaningful until there is a convention governing its use.

This view of neologisms does not hold up. Consider Lepore and Stone’s example of ‘bromance’, coined by Dave Carnie in the 1990s. There are three options for describing Carnie’s coining of ‘bromance’: (1) Carnie meant bromance by ‘bromance’, (2) Carnie expressed an indeterminate content by ‘bromance’, (3) Carnie did not mean anything by ‘bromance’. Lepore and Stone seem committed to (2). This position is thrust upon them because (1) concedes that Carnie coined a meaningful neologism prior to any convention governing its useFootnote ²³ and (3) is both implausible and ill-fitting with their idea that later speakers have modulated the meaning of Carnie’s term.

Lepore and Stone take the fact that others in Carnie’s linguistic community advocated for various expansions of the extension of ‘bromance’ to be evidence that the meaning of ‘bromance’ was indeterminate until there was a convention governing its use. However, it is dubious that there is a single governing convention for ‘bromance’ even now: the term is still “litigated” by speakers, as emphasized by Lepore and Stone (Reference Lepore and Stone2017, 257–258).Footnote ²⁴ Moreover, if Lepore and Stone are committed to the stability of use across time and the community as a requirement for the fixing of a word’s meaning, then relatively few of our words currently have determinate meaning because so many of the words used by English speakers across the world are still undergoing litigation or modulation and will continue to vary in their application across time and place.

Another issue with appealing to communal meaning litigation in this way is that such modulations of meaning assume a meaning to be modulated. That is, if one wants to expand Carnie’s concept of bromance to include friendships between male dogs, then one must presume that Carnie’s word ‘bromance’ meant something in the first place (i.e., that it had an extension to expand). Fuzziness on the full extension of a term should not be confused with a term not having content. Getting clear on this distinction shows that even if the ways most people now apply the term ‘bromance’ are more varied than how Carnie first used the term, there was still something Carnie meant by ‘bromance’ and that that act of meaning was indeed a convention-independent neologism.

While the admission that Carnie meant something determinate by ‘bromance’ should be incisive against conventionalism, Lepore and Stone try to block this conclusion by appealing to semantic externalism (citing Kripke, Reference Kripke, Harman and Davidson1972, Putnam, Reference Putnam1975, and Burge, Reference Burge1979). They suggest that neologisms resist anti-conventionalist explanations because successfully interpreting a neologism requires a speaker and hearer to jointly focus on the external regularity that fixes the meaning of the term:

By “regularity,” I take it that Lepore and Stone mean a persisting phenomenon or object. Their main idea is that a neologism, to pick something out in the world, must lock onto a regularity as such and not merely, e.g., a discrete time-slice of something. Someone hearing a neologism for the first time can only understand it if they lock onto the same external regularity the speaker picks out. From this plausible conception of what it takes to grasp a neologism, Lepore and Stone want to extract two problems for anti-conventionalism.

First, they suggest that an individual cannot, on her own, lock onto an external regularity. Since Carnie’s knowledge of the bromance relationship “turned out to be neither exhaustive nor authoritative” when he coined the term ‘bromance’, Lepore and Stone want to draw the conclusion that Carnie could only mean something indeterminate by his neologism (Lepore & Stone, Reference Lepore and Stone2017, 258). Instead, the collective knowledge of the community is needed to lock onto a determinate meaning for ‘bromance’. Clearly, though, this epistemological principle does not apply to every introduction of a new term. One can successfully christen a baby ‘Annie’ (given the right circumstances) without knowing anything about the baby’s blood type, personality, future career choice, and so forth. It must be argued, then, and not simply assumed, that one must first have exhaustive or authoritative knowledge about the phenomenon itself before coining a word like ‘bromance’. There are surely some epistemic constraints a speaker must meet in order to successfully coin a new word (we might want the coiner to be able to make explicit some rules for applying his term and/or be able to reliably gesture to an external regularity that someone else could pick up on), but such constraints do not entail that an individual cannot name a regularity all on his own.Footnote ²⁵ A linguistically competent agent with sufficient background knowledge is perfectly capable of identifying a yet-unnamed regularity in the world and giving it a name. To suggest that we must have exhaustive knowledge of some phenomena before we can name it merely backfires on the conventionalist, for our linguistic communities can go wrong in understanding some external phenomenon just as an individual can. Perfectly exhaustive knowledge of a regularity is illusory, a never-reached state (by an individual or a community), and there is no non-question-begging way of defining ‘authoritative knowledge’ here that would privilege the community over the speaker in every case.

Not only are there clear cases where an individual could successfully coin a term themselves, but it seems that in most cases the introducer of the neologism must mean something by his word before his community can collect more knowledge about the targeted regularity. Lepore and Stone assume that a community can fix the meaning of a new term when an individual cannot because the community achieves exhaustive (or at least authoritative) knowledge of the regularity in question. But how are others in the community to lock onto that same regularity that Carnie picked out with ‘bromance’ and add to the communal body of knowledge about it, if Carnie had not already picked out the phenomenon of bromance with his coining of the term ‘bromance’? If we think about the meaning of a neologism being passed around like “links in a chain” then we must posit something that links the various individuals who take up use of the neologism. The second (or n^th) person to use ‘bromance’ can only add to the knowledge Carnie initially had of bromance if Carnie himself meant bromance by ‘bromance’. The chain of people who mean bromance by ‘bromance’ must start with someone and thus the communal use of the word (i.e., the regular/expected use of it in the community) must come after an individual has successfully meant bromance by ‘bromance’.

The second lesson Lepore and Stone draw from externalism is that to understand someone else’s neologism, a hearer must appeal to more than her “nonce insights” into the speaker’s mental states (Lepore & Stone, Reference Lepore and Stone2017, 247). They explain: “[if] we want to figure out what bromance is, then, we do not just look inside the head of the first speaker; rather, we engage with the world and our own understanding” (Lepore & Stone, Reference Lepore and Stone2017, 258). Note the lack of any appeal to linguistic conventions in this explanation of how we “figure out what bromance is”—this is a telling omission. It is true that in interpreting another’s unfamiliar use of words, we cannot rely merely on intuitions about the speaker’s mental state; we usually need to look to the shared environment, consider salient regularities, and interact in some way to triangulate on the regularity they intend to pick out. However, there is nothing in this appeal to an externalist, triangulatory process that requires us to invoke conventions. Indeed, Lepore and Stone’s assertion that in interpreting others we must “engage with the world and our own understanding” is just the sort of externalism that Davidson supports: “The triangular relationship between agents and an environment to which they mutually react is, I have argued, necessary to thought.” (Davidson, Reference Davidson1997b, 130).Footnote ²⁶ Davidson even flags his basic agreement with Burge and Putnam:

However, Davidson will go on to clarify that “it doesn’t follow, simply from the fact that meanings are identified in part by relations to objects outside the head, that meanings aren’t in the head” (Davidson, Reference Davidson1987, 31). So, even if we agree with Burge and Putnam (and Lepore and Stone) that causal interactions with the world partially determine the meanings of our thoughts and utterances, we are not forced to say that what we mean by our words is fixed by features of the world of which we are entirely ignorant or by the ways a community of speakers use their words. The externalist thesis that Lepore and Stone sketch is perfectly compatible with the view that what speakers mean by their words is fixed by facts about the speaker—her history of language learning and causal interactions with the world as well as her present communicative intentions—and not her linguistic community.Footnote ²⁷

Regimenting externalist commitments against Davidson conflates his anti-conventionalism with contextualism. It is uncertain whether contextualism is compatible with externalism, as there are strong grounds for linking contextualism with internalism about meaning.Footnote ²⁸ So the following concluding remarks by Lepore and Stone are best directed at contextualists:

As long as we agree that a speaker’s mental states (her perceptual states and intentions) are relevant to interpreting her speech acts—which they must be on pain of utterly alienating a speaker from what she means by her words—then this emphasis on the need to look to the world in interpreting someone, to be sensitive to features of one’s environment and to harness one’s knowledge about the world and language, is completely consistent with anti-conventionalism.Footnote ²⁹

5.1. The stability of prior theories

Camp’s first error is to assume that a speaker’s prior theories need not have any “robust stability” over time for Davidson (Camp, Reference Camp2016, 117). The pressure for relative stability in one’s prior theories is brought out most clearly when we begin to think of the prior–passing theory distinction from the perspective of the speaker (rather than just the interpreter). Prior theories are not just regimented in interpreting others, they are regimented by the speaker in the production of her own speech acts. In clarifying his position in “Derangement,” Davidson explicitly commits himself to the idea that speakers possess a theory of meaning that they are disposed to use in producing speech acts, all other things being equal:

A person’s standing prior theory is not static—what one is generally disposed to mean by a term can change throughout her life—but Davidson certainly assumes that an individual’s standing prior theory will enjoy stability over time. There is no reason not to grant Davidson this commitment to the stability of prior theories alongside his commitment to the need for local passing theories, for general stability in one’s standing prior theory does not entail that what one means by any given utterance is necessarily going to be a function of that theory. Of course, we are free to posit that often a speaker’s passing theory—what she means in a context—is going to mirror a fragment of her prior theory, what she generally means by her words. The anti-conventionalist just wants two things to be clear: (1) for any given term or expression we could mean something different by it in an appropriate context, so no individual word is guaranteed sameness of meaning across every context, and (2) the prior theory that one’s passing theory is based on is not fixed by linguistic conventions or a community-wide lexicon, but by facts about that speaker’s use of her words and history of language acquisition.

We can also see Davidson’s commitment to the stability of prior theories in his discussion of how we interpret others. He acknowledges that most of the time we interpret others “effortlessly, even automatically” (Davidson, Reference Davidson1994, 112),Footnote ³² chalking this up to the fact that we have learned to talk in similar sorts of ways as others around us do. We often interpret others effortlessly and are so interpreted because we are working with fairly stable prior theories in the interpretation and production of utterances. I write “prior theories” (plural) because when Davidson says that we often interpret others effortlessly because we have “learnt to talk as others do,” he is not saying that we have learnt to talk in one particular way—the way in which all others in our community do—but that we have acquired various prior theories covering various sorts of interactions with others and each of these various prior theories usually allows for effortless, automatic communication in their proper contexts. The important bit is that even while he is committed to the view that linguistically mature humans possess multiple prior theories, Davidson can and does maintain that the prior theories we regiment in the interpretation of others are going to enjoy a robust stability over time. This general stability is necessary if the lexical entries in our prior theories are to have their meanings fixed holistically and thereby have determinate, cross-contextual conditions of correct application.

5.2. The connection between prior and passing theories

Nevertheless, the fact that prior theories can enjoy relative stability over time does not, on its own, respond to Camp’s worries about how Davidson is going to maintain his holist commitments (and thus his literalism) while being an anti-conventionalist. This is because of the relation Camp envisions between the holistic prior theories that speakers possess and the local passing theories we construct. Her stated concern is that Davidson is committed to a merely causal relationship between a speaker’s prior and passing theories and that, therefore, a speaker’s prior theory cannot contribute to fixing the (cross-contextual) meanings that her passing theory assigns to her words in a particular context.

It is not clear what Camp means by “a merely causal relation,” but it seems to entail at least the idea that there are no rules or norms connecting one’s prior theory to one’s passing theory: “Nor are there any rules for getting from either prior theories or conventional meaning to a passing theory, just ‘rough maxims and methodological generalities’” (Camp, Reference Camp2016, 117). In one sense, this is accurate: the anti-conventionalist is committed to the idea that one’s prior theory (for oneself or another speaker) does not provide any categorical norm for how to use or interpret any given word in a context. With this commitment, Davidson is rejecting a simple, mechanical connection between the semantics of one’s prior and passing theories. But “rough maxims” and “methodological generalities” are still normative connections: there are better and worse ways to move from a prior theory for a person to a passing theory for them. And, importantly, Davidson is clear that in speech and interpretation we always move from prior to passing theories; that is, we always work from a more stable base of linguistic knowledge.

Davidson does not provide an explicit example of any such maxim or generality in “Derangement,” but some possibilities are forthcoming. Speakers will often use their words in familiar ways, so until she receives mitigating contextual or verbal cues, the interpreter has reason to use the semantic values from the prior theory she has for the speaker. So one should consider all of the T-sentences in one’s prior theory for a person as generalities: each hypothesis of the form “for agent A (or agents of type X), ‘S’ is true iff S” is, as it were, a rough maxim, one which we can and should consider in constructing a passing theory for A, but given the possibility of innovative, malapropic, and simply different uses of words, our antecedently acquired T-sentences for A are defeasible. When potentially deviant uses of words appear, interpreters should interpret the speaker so as to make her maximally rational given the beliefs we ascribe to her and the available facts (so we ought to interpret malaprops and non-standard uses of words charitably). Most importantly, perhaps, is Davidson’s repeated insistence that interpretation is most often a dynamic process; understanding another comes not by considering her words in isolation, nor by looking to features of the world in isolation, but by engaging with the other speaker about the world one shares with her.Footnote ³³ This sort of engagement with another about the world can itself be done better or worse.

Davidson also allows that grammatical features will be subject to rough maxims and methodological generalities, though they too will not yield strict laws.Footnote ³⁴ In fact, Davidson explicitly acknowledges that something like a universal grammar might be biologically endowed: “What we are born with, or what emerge in the normal course of early childhood, are constraints on syntax, not semantics” (Davidson, Reference Davidson1997a, 134). We could thus consider the syntactic parsing of sentences of natural language to be a process which has stricter or more precise maxims and rules than the assigning of meanings to another’s words (“syntax is so much more social than semantics” (Davidson, Reference Davidson1984, 279)). Still, of course, even with regard to grammar a speaker might go against the grain and still be understood (e.g., a speaker might conjugate the past tense of ‘go’ as ‘goed’, she might use ‘they’ as a third-person singular neuter pronoun (which was non-standard before a few years ago but now readily understood)). The grammatical point is less important here, but in both cases, Davidson is not denying the existence of generalities and maxims that guide how we produce and interpret sentences; he is denying that such generalities and maxims are categorical, that is, he is denying that knowledge of them is sufficient for interpreting another speaker correctly in any particular circumstance. By so doing, Davidson is pressing the fact that meaningful utterances are products of intentional creatures, prone to irregularities, creativity, and changes in disposition, and so interpretation of those around us is sometimes going to require more than mere knowledge of how a speaker or community generally uses their words.

5.3. Holism and passing theories

Dispelling the idea that prior theories and passing theories are utterly divorced, connected only causally, we can return to our suggestion that the holism underwriting the meaning of a speaker’s utterance is located in her prior theory. If a speaker is using her words in a particular context according to how she usually uses those words, then there is no problem for the anti-conventionalist. We might show this by explaining that the speaker has a standing, holistic prior theory and, in this particular instance, she has spoken in accordance with her prior theory, without altering or changing it; thus, the holistic connections supporting her prior theory are directly transmitted to her utterance in the context.Footnote ³⁵ This is a baseline for what is happening in cases where no innovation or creativity occurs, which we can extend to make sense of innovative uses of language. Consider Mrs. Malaprop: in accidentally uttering ‘epitaph’ instead of ‘epithet’, Mrs. Malaprop takes her utterance of ‘epitaph’ to have all the inferential connections that her prior uses of ‘epithet’ had. She can only be understood as taking her own use of ‘epitaph’ to have the semantic value that ‘epithet’ normally does for her if we grant that her grasp of the meaning of ‘epitaph’ is derived from how she takes it to fit in her prior theory. That is, Mrs. Malaprop gives ‘epitaph’ the meaning of ‘epithet’ precisely by understanding ‘epitaph’ in relation to her other known words and beliefs. Similarly, when coining a new term, the coiner will, even if she does not coin her term by explicit definition, understand the meaning of her neologism in part by understanding how it would contribute to the truth-conditions of many other possible sentences.Footnote ³⁶

Once we are clear that a speaker creates a passing theory for a context by drawing from her prior theory, then the worry about how interpreters can appeal to holistic networks in understanding speakers should go away. We understand Donnellan’s use of ‘glory’ and Mrs. Malaprop’s use of ‘epitaph’ in part by appealing to our prior theories for those speakers. When he articulates what he means by ‘glory’ we take Donnellan to mean a nice knockdown argument by “a nice knockdown argument,” giving ‘nice’, ‘knockdown’, and ‘argument’ the values that we would normally give for them in interpreting Donnellan. So here, our understanding of ‘glory’, introduced as it is by definition, is scaffolded by our prior theory for Donnellan. If we were given reason to think that by “a nice knockdown argument,” Donnellan meant something strange in 1968, perhaps embarrassing fact, then we would alter our passing theory for ‘glory’ for him and we would also alter at least our passing theory for his use of ‘nice knockdown argument’ as well. Each such change is done against a background of stability; we could not hope to interpret someone who consistently changed the meaning of every one of her words, just as we cannot change the meanings of all of our words at once.Footnote ³⁷ But the anti-conventionalist is nowhere committed to the possibility of such wholesale change; stable, holistic theories of meaning established before a particular conversation play a crucial role in Davidson’s account of how speakers come to understand one another because it is only against the background of such theories that one can make sense of a speaker’s (or one’s own) novel or non-standard uses of words in a particular context.