2. Vallée’s pluri-propositional framework

All of these papers except chapter 1 are crafted within a semantic framework influenced by Perry (Reference Perry2001). (Chapter 2, published in 2009, is largely an updated, pluri-propositionalist defense of chapter 1’s [1996] approach to first-person plural pronouns.) As such, it makes sense to begin there. We try to steer clear of questions as to whether or how well Vallée gets Perry correct, focusing instead on an exposition of Vallée’s broad point of view.

There are two fundamental tenets therein:

We may rephrase the approach in terms of the authors Vallée draws most heavily on. From Russell (Reference Russell1905) and Kaplan (Reference Kaplan, Almog, Perry and Wettstein1977), he takes a commitment to directly referential propositional content. A proper name, e.g., is simply a “tag” for an object. For reasons stressed by Frege (Reference Frege1892), however, one needs more than that: something must capture the difference in “cognitive significance” between, say, (1a) and (1b):

1. 1.

   1. (a) I am Richard Vallée.

   2. (b) Richard Vallée is Richard Vallée.

Vallée would have us supplement the directly referential, singular content, specifically with reflexive contents. Here, he draws on Reichenbach (Reference Reichenbach1947). Finally, the framework is Austinian (Reference Austin, Urmson and Sbisa1962) thrice over: it focuses on concrete utterances; it allows indefinitely many contributions by any given utterance; and a crucial aspect of linguistic meaning is a conventional instruction for doing something, a schema for acting. Vallée writes:

Spelling out things out, Vallée distinguishes the following three levels of meaning (51–52):

Vallée says that [1: M] “is not propositional” (76), and that “… the conventional linguistic meaning of a lexical item understood as a type is a rule of use.” Take as an example:

1. 2. I am a philosopher.

[1: M] instructs the hearer: find the speaker S of the utterance in question; find the time t of the utterance; take the content of this utterance to be that speaker S of u is a philosopher at time t of u. Getting from [1: M] to [2: C] requires identifying not just potential utterances, but the actual utterance u ₁. [2: C] is “saturated” with that specific one. Still, it does not require a grasp of the usual and familiar facts about the context of utterance (Who? When? Where? To whom? Etc.) Establishing those facts will yield [3: D]. As will become clear below, [1: M] and [2: C] are reflexive contents in that they are about (among other things) utterances. [3: D] is not reflexive, but rather familiar old referential content.

It is worth pausing to explicitly distinguish a couple of things that Vallée’s level [1: M] is not. It is not a universally quantified token-reflexive proposition such as “For every utterance u of ‘I am a philosopher,’ u is true iff the speaker is a philosopher.” That is a possible approach, but it is clearly not Vallée’s, for that would be truth-evaluable proposition and [1: M] is not. It provides a “how-to” for a particular linguistic tool. It is also not a Kaplanesque (Reference Kaplan, Almog, Perry and Wettstein1977/1989) “character” (understood as a function from sets of contextual parameters to contents). Vallée’s [1: M] is an imperative. Granted, there are similarities between [1: M] and the formal semanticist’s characters. Both are properties of expression-types, and both are crucial elements for determining “contents.” However, Vallée several times stresses differences between an approach to the context-sensitivity of lexical items which is focused on sentences as types in context versus Perry’s post-2001 approach, which is focused on utterances (x, 23, 50, 156).

These meanings-levels are most starkly distinguished in a simple case of indexicality. Take (2) again. Meaning [1: M] will apply in common to any utterance of that sentence; it is what is grasped by those who have mastered the meanings of its constituent expressions. A native speaker can glean the meaning of [1: M] of ‘I am a philosopher’ knowing nothing whatever about whether it was even uttered. Once we take the meaning-rule for (2) and identify a specific utterance, we get [2: C]: The speaker S of this very utterance u₁ is a philosopher at time t of u₁. One can get [2: C], the purported cognitive significance, from [1: M] extremely straightforwardly: simply identify which utterance is in play. More work is required to move forward in the cycle of interpretation. One needs the usual information about that utterance: Who? When? Suppose it is Richard Vallée (RV) who says (2). Fill in who that is, and we get the singular proposition [3: D]:

1. 3. <RV, philosopher>.

Content (3) is what utterances of ‘I am a philosopher,’ ‘Richard Vallée is a philosopher,’ and ‘That man is a philosopher’ have in common. It is the course-grained truth conditions. <RV, philosopher> is what is literally stated using those sentences. It is not reflexive but is rather “directly referential”: i.e., of the general form <object o having property P>, or <objects o ₁ and o ₂ standing in relation R>, or…

3. A case study: Vallée on complex demonstratives

Vallée’s overarching framework promises explanatory riches. He urges, in particular, that its pluri-propositional and reflexive dimensions can help with otherwise puzzling features of complex demonstratives, of gradable adjectives, of certain of the phenomena which Grice (Reference Grice1975) classifies as ‘implicature,’ and more besides. As noted, this aspect of Vallée’s work is groundbreaking; various novel, interesting complexities arise and are explored.

Let us take a focused look at complex demonstratives. Consider (4):Footnote ³

1. 4. [_S [_NP [_DET That] car][is yellow]]

From the perspective of old-fashioned philosophy of language, complex demonstratives stick out like sore thumbs, and have thus attracted much attention. (See the citations on 68 for a fairly comprehensive sampling of this literature.) Grammatically, they seem odd because they combine syntactic features of pure demonstratives (‘that,’ ‘this’) with aspects of definite descriptions (‘the F’). They are like definite descriptions in containing a determiner and a nominal restriction, but they are unlike them in that said determiner belongs to a special category, namely context sensitive terms. Relatedly, an expression like ‘That car’ looks semantically odd. On the one hand, such phrases involve compositional machinery over descriptive material, and that would seem to put them in Russell’s (Reference Russell1905) category of a “denoting phrase” (i.e., a quantificational NP). On the other hand, they seem to yield genuine reference to an object within de re singular propositions (e.g., <car_i, yellow>, where the first element of the pair is the object under discussion). Such seemingly quantificational-referential hybrids are at best unique and at worst impossible. Either way, their status cries out for treatment by philosophers of language.

Vallée rejects two broad approaches as promising but imperfect. There are, to coin some labels, reference-over-satisfaction views. These emphasize the context sensitivity of the determiner ‘that’ and reference for the complex ‘that car.’ ‘Car,’ goes the idea, merely helps the hearer find the right referent. It renders a certain object more salient, albeit via linguistic means; but, like manifestly staring at something and then saying ‘that,’ the contextual prop drops out of the proposition expressed. As Vallée nicely phrases it, on this view it is ‘that’ which genuinely refers, not ‘that car’ (56). Then there are satisfaction-over-reference views. They emphasize the nominal restriction ‘car’ and its descriptive content. An object must satisfy the condition specified by the word/phrase; and said condition thereby factors into the official truth-conditions. Phrased using jargon that we will later call into question, in reference-over-satisfaction the descriptive material ‘car’ is “merely pragmatic”; in satisfaction-over-reference, it is “genuinely semantic.”

His highly innovative deployment of the pluri-propositional, reflexive-referential framework falls neatly into neither of the above. He argues that his meaning-levels provide a better overall account because each of the above misses something. We may summarize Vallée’s stance on complex demonstratives with (5):

In brief, the story is platypuslike: the meaning is simultaneously satisfactional because of [2: C] and referential because of [3: D]:

This proposal has considerable advantages. Four spring immediately to mind. Imagine an exchange like (6):

1. 6.

   1. (a) Who is that king?

   2. (b) [That king] is [that man we saw wearing a funny hat yesterday].

At least on a very simple reference-over-satisfaction approach, there is only a singular proposition in play, whose constituents are the identity relation and an ordered pair containing the same person twice over: e.g., <=, <George IV, George IV>>. Hence, very implausibly, (6b) is predicted to function just like ‘That is that’ as far as semantics is concerned. The reference-over-satisfaction approach elides cognitive significance. Vallée offers a solution: said difference lies in content [2: C]. Continuing with the example, Vallée’s content [2: C] for ‘that king’ would refer to the utterance of the first NP (call that u_k), whereas the content [2: C] for ‘that man we saw wearing the funny hat yesterday’ would refer to the utterance of the second NP (call that u_m). Crucially, u_k and u_m are utterances of different sentence types. So, an utterance of (6b) is different from both ‘that is that’ and from ‘that king is that king.’ A second advantage: rule (5) isn’t ad hoc, with an otherwise unattested mechanism introduced to deal with complex demonstratives. As we stressed at the outset, Vallée’s reflexive pluri-propositional perspective applies widely. Moreover, as Vallée points out, the restriction does its usual job of imposing a satisfaction condition that is met or not depending upon how the world happens to be. Advantage number three. Vallée’s story also accounts for what happens with a “misdemonstration.” A use of ‘that king’ or ‘that car’ seems to go awry when the object referred to is not a king/car in a way that isn’t parallel to staring or gesturing in a misleading direction while using a bare demonstrative. Such an error is correctly classed as “properly semantic.” Fourth and finally, let us underscore a feature of Vallée’s proposal which is easily missed. In complex demonstratives, there is material in [1: M] that does not make its way into the corresponding [3: D]. Therein lies what he calls “overarticulation” of the type-meaning vis-à-vis the official truth-conditions. Precisely because what is “overarticulated” falls out of [3: D], Vallée’s rule rightly treats the official truth-conditions as genuinely singular—as if the contribution to “what is said” had been directly referential, say via a proper name or pure indexical.

4. Twin difficulties facing the framework

As befits the halfway mark, let us summarize. We began by overviewing the contents of Words and Contents. We then briefly presented the background framework. It is pluri-propositional in the sense that utterances express at least two propositional contents. It is reflexive twice over. There are the reflexive contents of the specific utterance. And the type meaning, a procedural rule of use, is metalinguistic as well. Recall [1: M] for ‘that car.’ It tells the hearer to find the utterance u of ‘that car’; it tells her that one object is most salient to the utterer of ‘that car’ at the time ‘that car’ is spoken. There are nonreflexive propositions expressed too, as “official truth conditions.” Summarizing in terms of predecessors, there is the influence of both Austin (Reference Austin, Urmson and Sbisa1962) and Reichenbach (Reference Reichenbach1947) in that [1: M] is a reflexive rule of use. [2: C] does Fregean (Reference Frege1892) work. And [3: D] captures the singular contents of Russell/Kaplan et al. To illustrate the framework’s virtues beyond indexical pronouns, we applied it to complex demonstratives like (4):

1. 4. [_S [_NP [_DET That] car] [is yellow]]

Vallée proposed a clever amalgam of reference-over-satisfaction with satisfaction-over-reference approaches. As such, it promised the benefits of each more familiar alternative, with the failings of neither.

The volume abounds with other applications.

In what remains, we raise two sorts of difficulties corresponding to the “pluri” aspect of the view and to the “reflexive” aspect. We thereafter diagnose a potential source of these flaws in an equivocation on ‘semantic.’ We end with a suggestion which takes Vallée’s laudable pluralism a step further than he does, thereby avoiding these pitfalls.

The first difficulty: on the present framework, all contents of utterances must be propositional. Yes, the types have nonpropositional meanings, namely, rules of use. But tokens/“speakings” have only “contents,” where the latter are by definition claimlike. It is part and parcel of Vallée’s whole approach that for every challenge, additional propositions are the solution. He posits them for troubling (lexical) parts: there are additional propositional constituents for the first-person plural pronouns, contextuals (‘enemy,’ ‘foreign’), and colour terms. He posits them for problematic grammatical combinations: there are additional truth-conditions corresponding to complex demonstratives, graded adjective phrases (‘tall man’), and differentially informative sentences.

Vallée is inclined to describe this as a “feature” rather than a “bug.” Uniformity is simplicity, he maintains, and hence theoretically desirable (xiii). Not in this case. The exclusive reliance of propositions has, as a consequence, that Vallée always faces a pair of questions:

1. 7.

   1. (a) Which proposition is it exactly?

   2. (b) How exactly is it conveyed?

Each question introduces a subdifficulty.

We will illustrate with slurring words. Vallée’s view (162) is that (8a) carries the conventional implicature (8b):

1. 8.

   1. (a) Hans is a Boche.

   2. (b) Hans is German and despicable because of being German.

This seems to us a partially correct answer to (7b) because it is not the case that is despicable because of being German is part of the “official” truth conditions (163). We agree that no such thing is literally asserted/claimed/stated. However, and this is the first subdifficulty, Vallée doesn’t provide enough evidence against various alternatives with respect to the “How conveyed?” question. Once one is committed to the “extra content” being a proposition, the means of conveyance could be via generalized conversational implicature, via a special kind of manner implicature, or via presupposition.

Nor are we sanguine about the possibility of responding by simply laying out the evidence against these other options, because of the second subdifficulty. To our minds, (7a) is a pseudoquestion when posed about a slur’s alleged present-but-unasserted proposition. ‘What is the extra proposition?’ has a false presupposition. There is no such thing because the meaning difference between a slurring word and its neutral counterpart is descriptively ineffable. This is now widely recognized in the literature, so we will merely provide a hint of a defense here and direct the reader to Diaz-Legaspe, Liu, and Stainton (Reference Diaz-Legaspe, Lui and Stainton2020), which pursues the details. We’ll also revisit “slurring content” at the very end. We think it’s clear on the face of it that is despicable is too precise. Similarly, for is contemptible and is appalling. More deeply, no exact paraphrase is fully satisfactory. A potential reply: (7a) isn’t really a pseudoquestion. It merely calls for an answer in terms of borderless concepts or a penumbral cluster. Granted, that would render Vallée’s actual gloss incorrect. But it would leave his pluri-propositional strategy intact. We’re unmoved. First, it’s not clear to us whether Vallée’s framework allows such contents. More deeply, the meaning of slurring words is more descriptively ineffable than that. It isn’t a matter of open-endedness. The problem can’t be overcome by massaging the details. The contrast in meaning between (pardon the word) ‘chink’ and its neutral counterpart, let alone the N-word versus ‘Black person,’ cannot be captured by any proposition. We would echo what Davidson (Reference Davidson1978) says about metaphor: to look for a state of affairs which distinguishes Hans being a Boche from Hans being German is akin to looking for the clichéd “1000 words” that captures the exact worth of, say, Picasso’s Guernica.

Another problem with Vallée’s answer to (7a) is that anyone who understands ‘chink’ or ‘spic’ would need to possess, on his view, not just the concept despicable but is despicable because of.… Yet young children can (unfortunately) use and understand such slurring words. That takes us to the second difficulty.

Vallée’s framework seemingly attributes to ordinary speakers a great deal of hyperintellectualized knowledge of linguistic contents. We may highlight the point by considering something even more straightforward than the [1: M]-rule for ‘That car is yellow.’ Consider ‘Are you hungry?’:

Given what a corresponding [2: C] content will look like, what concepts must one have at their disposal to competently use and understand ‘Are you hungry?’ on Vallée’s approach? At least SPEAKER, ADDRESSEE and UTTERANCE. We find it doubtful that all language users can deploy even these: does every three-year-old know what an addressee is, let alone an utterance? Consider how hard it is to explain to undergraduates the type/token/utterance contrast! What about adults suffering from dementia? Worse, even run-of-the-mill folk may not grasp THE TIME OF THIS VERY UTTERANCE. If this already sounds demanding, revisit rule (5) itself for ‘That car is yellow’: it is even more hyperintellectualized. Now consider a sentence which contains a series of the expressions dealt with in the book:

1. 10. We even saw the local Boche who owns that big red car.

This includes, in addition to a complex demonstrative, a first-person plural pronoun (‘we’), a conventional implicature (‘even’), a contextual (‘local’), a slurring word (‘Boche’), a graded adjective (‘big car’), and a colour term (‘red’). According to Vallée, each will add its own subvariety of reflexive content. So, an utterance of (10) will have many reflexive contents, of many sorts. Now, it must be conceded, (10) is a complex expression. Some young children would have difficulty using and understanding it. Nonetheless, we and our academic peers do not; yet it’s implausible that even our linguistic knowledge and online processing encompass so very much.

There is an obvious potential reply. We said, guardedly, that Vallée “seemingly” attributes a great deal of hyperintellectualized knowledge. One might be tempted to respond in either of two ways: (a) it is only a “seeming” because Vallée is not telling a psychological story; or (b) his is a psychological story but it’s all subpersonal. Let us consider each.

The first potential line of defense doesn’t pan out because it jars with the Perry-style demands placed upon reflexive contents. The focus of these theorists is cognitive—not an abstract, Platonic, logico-linguistic formal system. In particular, a large measure of the motivation for going pluri-propositional is to explain things which stump referentialists such as Mill: e.g., one not reacting in the same way to coreferential terms (e.g., ‘My pants are on fire’ versus ‘That man’s pants are on fire’ or ‘I am Richard Vallée’ versus ‘Richard Vallée is Richard Vallée’).

One might try to save Vallée’s rules by suggesting that a theorist can use them to rationally reconstruct the observed behaviors. Relatedly, in a broadly Davidsonian (Reference Davidson, Evans and McDowell1976) vein, one might hold that a causal role for the contents is not genuinely necessary for an explanation; all that’s required is that the posited rules, were they known, would suffice to yield the observed behavior. Hence, claims of psychological reality aren’t required after all. We demur because that sort of explanation is not acceptable in general. No one should be satisfied, for instance, with: ‘Though the morbillivirus is not present, it would suffice to yield the patient’s symptoms if it were; so, what explains her symptoms is the measles.’ Given that such gambits yield terrible explanations in general, a move of that sort by Vallée would be equally undesirable. Returning to the example: by our lights, what’s genuinely required to capture the impact of ‘My pants’ as opposed to ‘That man’s pants’ are different etiologies.

What’s more, Vallée seems to agree. For instance, in the particular case of grasping the meaning of ‘we,’ Vallée (chapter 2) makes multiple (seemingly not-merely-reconstructive) demands on ordinary speakers:

Similar causal requirements are echoed at various points later on in the volume, such as in chapter 3 on complex demonstratives (51), chapter 4 on implicit comparison classes (78–79), chapter 6 on contextuals (138), and in chapter 7 on slurs (157). So, these reflexive contents are, for Vallée, charged with playing psychologically real roles. Rightly so. Absent causal efficacy, no genuine explanation is forthcoming.

The second potential response is that his is a psychological story positing a series of very sophisticated inner representations; however, the knowledge is tacit. It is comparable to knowledge of language in generative grammar. Put otherwise, the theorist uses very abstract representations to describe the mechanisms. So, understanding her tools is very cognitively demanding. That makes things sound hyperintellectualized: e.g., something outside the ken of a young child. But, in fact, though the theorist may, the language user doesn’t “know” these rules in the sense of ‘know’ from commonsense. They are merely psychophysical causes, which mechanically produce behaviors.

This is unpromising on two fronts. Unlike the case of generative grammar, there is no independent evidence of rules like (5) and (9)—not from acquisition, clinical deficits, online errors, etc., nor even from unexpected (un)grammaticality. There’s no reason to believe that there is anything in our overall linguistic competence, conscious or otherwise, to be described in this logicomathematical way. More deeply, even if there were tacit rules, such subpersonal processes do not rationalize; they don’t, taken alone, explain the actions of the person. In Chomsky-style explanations, the personal-level agent doesn’t have the relevant concepts; nor, then, would a person using Vallée’s (5) and (9). Yet cognitive significance is supposed to make sense of what agents do. That is, the suggestion has been that people intend others to believe these contents; and that such beliefs may interact with desires and other propositional attitudes to motivate action. Remember that Fregean aspirations are a core part of Perry’s framework, and Vallée approvingly cites them, at multiple places. (See, for example, 50 and 79.) Besides, once again, this simply is not Vallée’s approach. Recall the quotations immediately above. His “I exploit” looks personal-level and rationalizing, as does his “the reasoning we do.” His choice of terms is apt, no slip of the pen, given the contents’ functions.

Neither of the potential responses rescue Vallée from attributing hyperintellectualized linguistic knowledge to persons. He really must, as an empirical hypothesis, saddle ordinary utterances and our knowledge of their meaning with many, many reflexive contents—more than is plausible given human psychology.

To summarize this section, we spied trouble with the pluri-propositional aspect of Vallée’s framework in that, for every explanatory challenge, he posits more propositions. But then, for each one posited, two questions arise: What exactly is it? And how exactly is it conveyed? Answers go begging because the whole issue is ill posed. We spied trouble with the second fundamental tenet, the reflexive one, as well. Those additional propositions in particular are implausibly demanding, psychologically speaking. Looking ahead, our suggested fix will rest on this: if one didn’t need to recourse to so many propositions, one could avoid much of the hyperintellectualizing.Footnote ⁴

5. A potential diagnosis: Equivocation on senses of ‘semantic’

The book’s contents and background framework have been explained. We’ve looked at two applications and noted twin difficulties. In the present section, we hazard a diagnosis both in general and specifically regarding the application to complex demonstratives: there is an equivocation on a complex, protean term ‘semantics,’ which infects much of the field. We will end by pressing Vallée’s promising insights even farther than he himself did, thereby making progress on the difficulties from section 4.

There are multiple senses of ‘semantics.’ Two are especially important here:Footnote ⁵

These are clearly not the same notion intensionally speaking. Nor are they coextensive in the case of the vernacular.

Meanings, in some pretheoretical sense, come in subvarieties. There’s the merely conveyed versus the full-on stated, the decoded versus the inferred, names which refer versus sentences which are true/false. There are, we’d add, subvarieties within semantics_Truth which do not belong in semantics_Type. Conversational implicatures provide a ready example. They involve reference and truth conditions; the things suggested/implied/intended stand in entailment relations. Yet they are paradigm cases of unencoded meaning. An example. Asked ‘Would you buy a Mercedes?,’ a person who responds with sentence (11), notwithstanding potential cancelability, deniability, etc., in some sense puts forward something bearing semantic_Truth properties, namely the propositions I would not buy a Mercedes and A Mercedes is an expensive European car.

1. 11. I don’t like expensive European cars.

Failure of coextension in one direction, then. There are also subvarieties within semantics_Type which don’t pertain to semantics_Truth. Consider the difference between otherwise synonymous expressions with contrasting sociolinguistic registers. The variance in meaning between French’s ‘vous’ (singular) and ‘tu’ is real and important, but not a matter of the kind of people they are true of. Again, the variance among the system of words which all designate feces—‘poop,’ ‘shit,’ ‘excrement,’ ‘stool,’ ‘turd,’ ‘night soil’—is likewise a matter of what discourse situations each is suited for. (We’ll expand on this in section 6.) The first, ‘poop,’ is fit, as a matter of linguistic convention, to be used in child-oriented talk; the second is vulgar; the third is designed, in terms of its linguistic function, for academic/clinical situations, etc. Nor, to anticipate an objection, is a speaker of (12) inevitably stating, presupposing or suggesting that their addressee is a child:

1. 12. There is poop on Richard’s lawn.

Even more starkly, phatic words such as ‘bye’ and ‘gesundheit’ simply lack semantics_Truth. But for the grip of an ideology, one wouldn’t think to ask under what circumstances these were true or false. Still less would one inquire about their “cognitive significance” as opposed to their “official truth conditions.” QED.

Should one simply dismiss these nontruth-conditional differences as not really part of meaning? Might one simply declare that they have no place in semantics_Type? That would certainly bring the meaning varieties in its extension into closer correspondence with those in semantics_Truth. However, the differences between ‘poop’ versus ‘turd’ and ‘congrats’ versus ‘bye’ are clearly not a mere matter of form. Like meanings of the truth-theoretic sort, they are instead closely tied to proper linguistic usage in conversational exchanges. What’s more, despite the shared semantics_Truth (whether because of the same reference or the joint absence of reference), a translator who rendered Spanish ‘excremento’ as ‘poop,’ or ‘bye’ as ‘hola,’ would have mixed up their meanings; and we very appropriately tutor young children and foreigners in the meaning differences among ‘engaging in sexual intercourse,’ ‘screwing’ and ‘fucking’ and among ‘gesundheit,’ ‘cheers’ and ‘amen.’

Continuing with our potential diagnosis, consider how neglecting all this might lead to mischief. The following implicit argument becomes very tempting:

C1 obviously encourages, as a general matter, the overpopulation of language with propositional contents. As a specific case in point, recall the puzzle about complex demonstratives like (4) and (6b), repeated below:

1. 4. That car is yellow.

   6b. [That king] is [that man we saw wearing a funny hat yesterday].

Vallée asked: Is the role of the nominal restriction (i.e., ‘car’/‘king’) semantic or not? Satisfaction-over-reference theorists emphasized conventional encoding, what knowledge of linguistic meaning is involved, etc. Doing so, the temptation is to infer:

The reference-over-satisfaction contingent point out, among other things, that one doesn’t say ‘That’s false’ to ‘That king is from Belgium,’ just on the grounds that he isn’t a genuine king but is merely dressed like one. (That response would seem appropriate if its meaning really were conjunctive, including he is a king as a “semantic” conjunct.) The contrary conclusion beckons:

Vallée, we saw, ingeniously split the difference to overcome the “puzzle.” He agrees that C2 is correct. However, it is correct only about a reflexive content: the nominal’s restriction is (merely) part of the level [2: C] truth-condition. In particular, the descriptive material provided by the nominal in (4) is:

Meanwhile, Vallée agrees that C3 is also correct. Applied to our example, the descriptive content λx. car(x) is not part of the “official” truth condition for (4). Instead, the level [3: D] result is the de re:

Herein lies the mischief of the equivocation. To overcome the impasse, Vallée posits two propositions. Both sides are right, but about different contents. This may represent a significant improvement (assuming for the nonce that reconciliation is the right goal). It certainly exemplifies the originality and insight we underscored at the very outset. However, it lands Vallée with the last section’s twin difficulties, here specifically with respect to complex demonstratives.

What’s the alternative for reconciling satisfaction-over-reference with reference-over-satisfaction without additional propositions? Disambiguate. Both sides can be right because they are using ‘semantic’ in different ways. C2 looks plausible with respect to semantics_Type; C3 looks plausible with respect to semantics_Truth.

6. From reform to revolution

We began by praising Vallée’ work as introducing not just novel issues but innovative methods. It’s a paradigm to be built upon. In this final section, we try to do just that. We’ll frame our discussion in terms of his reforms, and how we’d press beyond them to something more revolutionary.

As a preliminary, we need to explain very briefly what Vallée is moving forward from. Within Early Analytic philosophy, there was a very reasonable focus on truth, evidence, and logical relations. Frege, Russell, Tarski, Church et al. brilliantly developed artificial logical languages for modeling those. What we’ll call High Church Formalism is a tradition which grows out of the calculi of these pioneers. It retained, when theorizing about natural languages, the narrow focus. As a result, natural languages such Amharic, Bengali, Cantonese, Dutch, Ewe, etc., were treated, but for some curious epicycles, basically like the Predicate Calculus and Intensional Logical.

Vallée and his fellow travelers advanced matters by (i) switching the focus to “on the hoof” communication in ordinary utterance contexts and (ii) highlighting more contents (in the plural, as per the book’s title). Vallée pushes (ii) forward on dual fronts. Some meanings, namely those of expression types, are rules of use. Nor are utterance meanings exhausted by “official truth-conditions.” And yet, despite repeatedly emphasizing that his concern is communication rather than logic, Vallée ends up staying too close to the Formalist tradition, with its primacy of truth-conditional/referential contents. For reasons we’ll explore shortly, this strikes us as a deep mistake.

First, a digression into an important question left open in section 5. How could Vallée, of all people, run the two notions of “semantics” together? What could have induced a conflation between (what should now appear) two obviously different things—so as, e.g., to implicitly license the fallacious inference from P1/P2 to C1? Vallée’s focus, after all, are constructions whose encoded meanings are not per se truth evaluable. In particular, the manners in which semantics_Truth and semantics_Type part ways are central to understanding indexical pronouns such as ‘we,’ demonstratives (whether simple or complex), and (more controversially) comparison classes, contextuals, and colour terms. More than that, the contrasts among his meaning levels echo the disambiguations of ‘semantic.’ Nonetheless, Vallée seems to embrace (what is by our lights) the fateful slide in so many words. He writes, for instance: “The linguistic meaning of a term is a property of an expression as type and a rule assigning the truth-conditional contribution…” (49) and “If [‘Hans is German’ and ‘Hans is a Boche’] do not share the same linguistic meaning, they cannot share the same truth conditions” (160; see also 97). In his book, there appears no explicit recognition of meanings that have nothing whatever to do with truth or reference. Our ultimate diagnosis is a historical hangover: he adopts as his departure point symbolic systems wherein semantics_Truth and semantics_Type are roughly coextensive. Given that, it may seem innocent enough to equate them in natural language too, worrying only about that “roughly.”

Returning to the main thread, our alternative suggestion is to press Vallée’s insights to their logical end. To take even more seriously the richness of actual conversational interaction. To insist that various kinds of “nonlogical” interactions are fundamental, not curious outliers. Thus, the semanticist’s paradigm shouldn’t be a philosophical debate in the pages of Canadian Journal of Philosophy, let alone a mathematical proof. Rather, if one must identify “core cases,” Austinian ones would be better: bidding farewell with ‘bye,’ comforting with ‘there, there,’ and closing a prayer service with ‘amen.’ Encapsulated in a slogan, pace Montague (Reference Montague and Visentini1970), English, far from theoretically indistinguishable from a formal language, isn’t even in the same species.

Fleshing out the sloganeering, in some cases, to give an expression’s meaning is simply to describe its job in linguistic interactions. Not all the ways that the linguistic device is used, but how it is supposed to be used: its function in our “language games.” Let’s apply this ever so briefly to some examples from Section 5. To give the standing linguistic meaning of a phatic expression such as ‘bye’ or ‘gesundheit’ is simply to say: ‘bye’ is the informal word for parting; ‘gesundheit’ is the word one uses when someone else has sneezed. To specify the contrasting semantics_Type of a formal-register pronoun and its informal sister is to say, e.g., that ‘vous’ is the second person singular which is apt for social superiors of some kind. As for the contrasting “colouring/shading/tone” among words having the same semantics_Truth, recall (12):

1. 12. There is poop on Richard’s lawn.

But for ideology, but for a Formalist “hangover,” it would be obvious that there is no place laden with poop but not excrement/shit; obvious that a user of (12) is not stating, nor even conveying in some other way, that the addressee is a child. Instead, to specify the use-theoretic aspect of ‘poop’ is simply to say that, among the system of co-referring terms, ‘poop’ is lexically marked as [+ child-oriented, – clinical, + slang].

Putting our “out-Vallée even Vallée” reflections another way, philosophical logic has proven a very valuable enterprise. But, as with philosophy of biology—which rightly has no truck with species or methods of reproduction which are merely logically possible—philosophy of language must pay far more attention to mundane worldly facts than a philosophy of the Predicate Calculus, or even a philosophy of modal or tense logic. Doing so, one must expand upon Vallée’s meaning pluralism. Add more genuine varieties of type-meanings, including rules-of-use for words, phrases, and sentences pertaining not just to the resulting propositional contents of utterances, but to when and where to use such-and-such “verbal implement,” and to what sociocultural ends. Add more genuine kinds of utterance-meanings beyond Vallée’s “contents.” Not just more denominations, more currencies.

To illustrate the payoffs of revolution over reform, we’ll end by revisiting slurring words. The problem, recall, was that in taking all utterance contents to be propositional, Vallée was burdened with questions (7 a–b):

1. 7.

   1. (a) What exactly is the propositional content?

   2. (b) How exactly is it conveyed?

His answer to “How conveyed?” about slurs was: via conventional implicature. His answer to “What is conveyed?,” applied to example (8a), may be distilled as:

We urged that treating slurring as specifically an implicature of the conventional sort was underevidenced. We added that (7a) has a false presupposition and, hence, Vallée’s answer is despicable because of being German is wide of the mark in a deep and irreparable way. What’s more, his rule is manifestly implausible as an account of what the agent who understands ‘Boche’ must know: it requires, e.g., understanding of the concepts ENGLISH, UTTERANCE, CONVENTION and DESIGNATING.

Our radical way forward is to maintain that a slurring word and its neutral counterpart differ with respect to an entirely different kind of meaning. Here, begging pardon again for offensive terms, is a last set of examples:

1. 14. Slurring words for Jewish Person

   1. (a) abe

   2. (b) hebe

   3. (c) hymie

   4. (d) kike

   5. (e) yid

As far as reference/truth goes, we hold that all of (14a–e) are equivalent. We don’t merely mean that the “official truth conditions” are the same. There is no difference in “contents” in Vallee’s technical sense at all—not at any “level,” no matter how conveyed. The propositional contribution of ‘hebe’ is exactly that of ‘Jewish person’ (except, of course, in hyperintensional contexts, wherein even synonyms cannot be substituted salva veritate). Notwithstanding this, it seems to us perfectly obvious that none of (14a–e) has exactly the same meaning as ‘Jewish person.’ What’s more, they differ meaningwise among themselves: some are more offensive than others with ‘abe’ being the least aggressive and ‘kike’ being the worst. Only someone under the sway of a stipulative terminology would treat them all as synonyms with the neutral designator. How can all this be? Because semantics_Type far outpaces semantics_Truth. Because, more specifically, the difference in meaning among (14a–e) is wholly use-theoretic; more specifically still, there is contrasting sociolinguistic register, comparable to the difference between ‘shit’ versus ‘feces’ or ‘tu’ versus ‘vous.’ All are, to varying degrees, impolite, derogatory slang. The NP ‘Jewish person,’ in contrast, is the socioculturally appropriate and polite wording option for naming that same group.

This is why (7a) was a pseudoquestion: there is no “extra” proposition hereabouts to paraphrase. And what information needs to be stored in the lexical entries of language users—including the young child, the second language learner, and the translator? Something not especially intellectualized, viz, that there are various word choices for referring to Jewish people, and (14a–e) are nasty options which ought (with varying degrees of firmness) to be avoided.

7. Summary

Words and Contents is an extremely valuable contribution to the literature. As Vallée notes at the outset, it explores a different perspective on different natural language expressions (ix). Vallée’s pluri-propositional and reflexive framework develops a novel variant on Perry’s and applies it to a diverse and impressive range of challenging constructions. Multiple meanings systematically attach to types and utterances of them. There is the rule of use for the potential utterances, meaning [1: M]. It, in turn, gives rise in context to (at least) two propositional contents. There’s the familiar content [3: D] so important to direct reference theorists. There are also in-between reflexive contents (our focus was on content [2: C]).

All of this is so original that it’s unsurprising that it leaves open issues to be pursued by future researchers and may call for some rethinks. We drew attention to issues with both the reflexive and pluri-propositional aspects of the general perspective. We conjectured that an equivocation on ‘semantic,’ itself a historical aftereffect, may ultimately underlie these. Although at the [1: M] level of meaning, Vallée is more pluralist and context-aware than some Frege-to-Kaplan formal semantic traditionalists, he remains too monolithic about utterance contents. He squeezes too many different phenomena into just one sort of mold, i.e., claim-type truth-conditional contents. To repeat our suggested escape: having multiple meanings is an excellent idea, but, as per Vallée’s perfectly appropriate focus on actual conversational interaction, one should allow various semantic meanings (in the “deriving from the type” sense) which are genuinely use-theoretic.

We recommend the book in the highest terms.