A monstrous account of non-deictic readings of complex demonstratives

ABSTRACTComplex demonstratives are often seen as a source of trouble for the idea that demonstratives are directly referential. Several authors have provided an array of counterexamples that preclude us from treating complex demonstratives as devices of direct reference, since they could hardly be considered rigid designators. In this paper I argue that a revision of the classic theories can accommodate all the counterevidence from non-deictic uses of complex demonstratives. Namely, I argue that the two chief objections that have been posed to the traditional picture, the so-called ‘quantification in’ and ‘no demonstration, no speaker reference’ problems, can be given a unified solution by treating them as instances of monstrous quantification into the character of the term.KEYWORDS: Rigid designationdirect referencemonstersindexicalityquantifying in AcknowldgementsI wish to thank Derek Ball, María de Ponte, Manuel García-Carpintero, Stefano Predelli and Jordi Valor for their comments on earlier versions of this paper, and to an anonymous reviewer for a very thorough and constructive report.Disclosure statementNo potential conflict of interest was reported by the author(s).Notes1 It is in this sense that the nominal can be said to contribute to character. In these theories, the character of a complex demonstrative of the form ⟦that P⟧ is not a function F such that, for each context c, f(c) = dc, but rather a proper subfunction of F: one whose domain only includes contexts c such that ⟦P⟧()(dc) = 1. The character of any complex demonstrative is always given by a subfunction of F, and the role of the nominal is to determine exactly which of these subfunctions is to be selected.2 The validity of these arguments is debatable (cf. Braun Citation2008a, 70, footnotes 20–24), but this is not in conflict with the kind of approach advocated by c-theorists.3 Predelli (Citation2001) has posed a challenge to c-theories based on anaphora, but I have argued elsewhere that the counterevidence he points out can be given an alternative explanation (Gimeno-Simó Citation2021).4 King (Citation2001) also provides some syntactic evidence that seems to point towards the idea that complex demonstratives are quantifiers, but it has been subject to severe criticism (Altshuler Citation2007).5 Some defenders of the traditional paradigm have tried to explain away these readings by arguing that complex demonstratives can act as ‘stylistically altered definite descriptions’ (Dever Citation2001, 286; Salmon Citation2002, 522; Citation2006a, 446; Citation2006b, 272, footnote 11; Corazza Citation2003, 272; Georgi Citation2012). These ambiguity theories have been severely criticised by Ethan Nowak (Citation2014; Citation2021a; Citation2022).6 Notice in fact that the problem can be posed without even resorting to the third sentence in (18): (18*) [A student]1 was sitting in the library. Another student was sitting across from her1.Who is the referent of ‘her’ on this occasion? Let us suppose it is Mary. In that case these two sentences are true together in any world in which any student is sitting in the library and Mary is sitting across from another student in, say, a coffee shop. This is wrong for the same reasons as above.7 For simplicity’s sake, this lexical entry sets aside many details not directly relevant to the discussion at hand, such as distality or plurality. I am also passing over questions such as the interpretability of empty terms and the way to handle multiple occurrences of demonstratives.8 Just to make it sure, what (26) is stating is not that for every actually red object x there is a possible world w such that x is shiny in w: rather, we want all the actually red things to be shiny together in a single world. Neither is it stating that there is a possible world w such that for every x such that x is a red thing in w, x is shiny in w. If we only have a world at the evaluation parameter, we are at a loss: we need ‘red’ to be assessed according to the initial world, meaning that it should take wide scope over ‘it is possible’, but at the same time ‘red’ should be under the scope of the quantifier ‘everything’, which in turn should be under the scope of ‘it is possible’ in order for all these things to be collectively shiny at a single world. The scoping is thus incompatible.9 These analogies between the modal and the temporal domain were highlighted by Schaffer (Citation2012). See also Neale (Citation2006).10 Some authors often include overt world variables for adjectives, since this idea provides a nice way to account for so-called ‘Fodor readings’ (scopings that are neither narrow nor wide; see Heim and von Fintel (Citation2011, §8) for an overview). But, as arguments in Musan (Citation1995) and Kusumoto (Citation1999) made clear, systems with explicit variables for adjectives often overgenerate, since they predict that many sentences have readings that they do not actually display. I prefer not to lose the correspondence between the superficial form of a sentence and its LF; therefore I won’t be including variables that are not contributed by some explicit element of the sentence.11 It is widely acknowledged that past tense often works referentially, as in Partee’s (Citation1973) famous example ‘I didn’t turn off the stove’. Here, the anteriority with respect to the present moment is presupposed, rather than asserted. However, this does not entail that some sentences may fail to provide an explicit representation of present tense. On the contrary, representing present tense is necessary for ensuring that the presuppositional material is correctly interpreted. For example, in a system like von Stechow’s (Citation2003) any sentence p containing a referentially used past tense receives an LF like (a), containing an explicit present tense variable: (a) t0 [t0 < t1pres] (p)12 Co-indexing is necessary in this system, for otherwise we could end up attributing to a sentence like (26) a reading it does not have. There are alternatives, however. For example, the same results can be achieved by fronting every sentence with two variables ‘’ which always pick out the parameters of the circumstance, and by letting any variable marked with the indicative have the following presuppositional semantics: ⟦wnind⟧ = λc: gc(wn) = gc(w0). gc(wn), where n is any natural number.⟦tnpres⟧ = λc: gc(tn) = gc(t0). gc(tn), where n is any natural number.I will resort to co-indexing for simplicity’s sake.13 Admittedly, the latter reading is harder to achieve, but it can be made perfectly felicitous, and even preferred to the former one, if some context is provided. Imagine you and me have been classmates to Julia, and we are having a conversation remembering past times. If I utter (31) in this context, the second reading is more salient.14 This is, indeed, the way I have been dealing with this expression throughout this paper. Other authors prefer to treat it as a semantically vacuous expression and to add the lambda as a covert element within the relative clause itself. I see no real difference between these two approaches, and my proposal is compatible with either of them.15 Higher-type traces are a relatively common device for solving several puzzles in semantics, and they often go hand-in-hand with the introduction of new combination principles (for a handbook treatment, see Heim and von Fintel Citation2011, 94–95). Admittedly, introducing this kind of machinery is not cost-free, but it is often necessary in order to achieve the right results without resorting to more doubtful resources, such as an unorthodox syntax. Importantly, this device is not being introduced just for the sake of accounting for problems related to complex demonstratives: the very same problem we are dealing with here can be posed to any NP containing a relative clause, such as the one in (31), and it can be solved analogously (see footnote 17).16 Some philosophers may be reluctant to adopting yet another composition principle. Notice, however, CIRCUMSTANCE SPECIFICATION is complementary to INDIVIDUAL IDENTIFICATION above, meaning that it does not come at a greater conceptual cost, much less considering that the only reason why it has been introduced is in order to preserve the overall simplicity derived from adopting type > for predicates – which is, after all, a far less common type in intensional settings.17 This system predicts that (31) may receive two different logical forms: (31a) λ. [A fellow student who2 [Ǝt3: t3 < t1] ( r2 wears glasses)] λ4. Mary is married to r4(31b) λ. [A fellow student who2 [Ǝt3: t3 < t1] ( R2 wears glasses)] λ4. Mary is married to r4The only difference between these two is the type of the movement trace. In (31a), the property ‘fellow student’ combines with ‘who wears glasses’ by INDIVIDUAL IDENTIFICATION and it is passed up and saturated with the tense of the main clause, just like ‘married’. In (31b), ‘fellow student’ is taken as argument by ‘who’ and saturated with the tense of the embedded clause, thus making it insensitive to the tense of the main clause and simultaneous with the wearing of the glasses.18 Sentence (33a) may be made felicitous under some assumptions. If we supposed, for example, that Waverley had been co-authored, (33a) would sound much better. Nowak (Citation2021a) himself provides an example of a felicitous utterance of a sentence much like (33a): (b) That author of Principia looks friendly, but I wouldn’t try to get an autograph from that one.Thus, the appropriateness of this kind of constructions seems to be highly context-dependent.19 In a recent paper, Nowak (Citation2022) poses a dilemma having to do precisely with the contrast between complex demonstratives with and without a relative clause. He argues that the only way to account for the aforementioned data is to either give up on standard syntax or to reject the idea that semantic composition proceeds locally. In a footnote (Nowak Citation2022, 14, footnote 23), however, he also acknowledges that there is room for a way out of the dilemma in case predicates like ‘author of Waverley’ and ‘author who wrote Waverley’ happen to differ hyperintensionally. This is exactly what my solution amounts to, and the verbal morphology is responsible for it.Notice, however, an important difference between his system and mine. When the verb in the relative clause is tenseless, my approach predicts that the complex demonstrative will only display deictic uses. In light of examples like (12b) and (13b), this is as it should be.20 For a critique of Nowak’s own way to account for this contrast, see Gimeno-Simó (Citation2021).21 Here is a complication to the circumstance-binding solution: the complex demonstrative in (16), unlike the one in (15), contains no verbal morphology: (16) That senator with the most seniority on each committee is to be consulted.Thus, my circumstance-binding solution is unable to handle the problem of intensional mismatch for (16): in this sentence the descriptive material within the complex demonstrative will always have to be interpreted according to the circumstance of the context. Something similar occurs with a simple QI case like (c): (c) Every professor cherishes that first publication of hers.Nowak (Citation2021a) acknowledges that sentences like these may be problematic for his account too. I don’t think, however, that this is a knockdown argument against my circumstance-binding solution. In the first place, an important thing to have in mind is that, unlike the other sentences we’ve been discussing, neither (16) or (c) enjoy crosslinguistic support: their translations into other languages are rather infelicitous. For instance, the literal translation of (c) into French, Italian or Spanish is not entirely ungrammatical, but a bound reading is impossible (i.e., it can only be employed for referring to a particular object; see the discussion in Corazza (Citation2003, 272–273)): (d) #[Chaque professeur]1 estime celle sa1 première publication.(e) #[Ogni professore]1 valorizza quella sua1 prima pubblicazione.(f) #[Todo profesor]1 aprecia aquella primera publicación suya1.This suggests that pragmatic arguments such as those presented in Braun (Citation2008a) are more likely to hold for sentences containing no relative clause, like (16) and (c), i.e. these do seem to be an indiosyncrasy of English – in fact, even some English speakers judge examples like (c) to be ungrammatical, or at least highly unnatural (see the discussions on analogous examples in Higginbotham (Citation1988), Neale (Citation1993, 107) and Georgi (Citation2012, 383, footnote 18)).But, even if further research should disconfirm the idea that felicitous uses of these examples are indiosyncratic, the c-theorist willing to preserve intensions still has many other options at hand. Let me sketch a possible answer. Arguably, both (16) and (c) contain quantifiers that can take wide scope over the demonstrative (‘hers’ may be thought of as a quantifier, and indeed its translation into other languages suggests it probably is). And quantifiers are usually thought to contain hidden variables for implicitly restricting their domains (von Fintel Citation1994). These variables, represented as ‘P’, ‘Q’, ‘R’ and ‘S’ below, may fall outside the scope of the demonstrative in case the quantifier takes wide scope: (16a) λ. [Each committee in P] λ2. [The most seniority among Q in r2] λ3. that senator with r3 is to be consulted(g) λ. [Every professor in R] λ2. [An x such that x is of r2 and x is S] λ3. r2 cherishes that publication of r3All that is required is to fill the variables with the corresponding properties: ‘Q’ should pick out ‘being a senator’ and ‘S’ should be ‘being a publication’ (the other two are irrelevant). Thus, these two properties will be assessed according to the world and time of the circumstance, for they take wide scope over the demonstrative. This is one among many possible solutions.22 There is an additional assumption to be made, one that the theories by Heim (Citation2005) and von Stechow (Citation2003; Citation2004) helped establish as standard: a mechanism of feature transmission under binding. The world variable within the relative clause (‘w2’ in (19d)) is phonologically realised as indicative, and this is so because it is bound by a verb in indicative ‘believes’. It is, however, not marked with this feature at the semantic level.23 Imagine, for example, that I know Mary pretty well, and that Bob and me are talking about Mary’s outstanding performance on the exam. For some reason, I am unable to remember her name, so I utter (19a) for reporting Greg’s belief. Our intuitions in this situation are similar to the ones with proper names: whether my utterance is true depends to a great extent on our interests, expectances and the required degree of accuracy.24 Here is an example of a previously grammatical use of subjunctive in English: ‘I will not let thee go, unless thou bless (as opposed to ‘blessest’) me’ (King James Bible, Genesis 32: 26). There are still some remnants of these archaic uses: ‘come what may’, ‘God save the Queen’. Subjunctive in Romance languages works just like in these examples.25 For some reason, bound uses are available only for complex demonstratives. Achieving a bound reading of a bare demonstrative is much harder, if available at all (Nowak Citation2021b): (h) OK [Every IKEA kit]1 comes with instructions for assembling it1/[that kit1].(i) # [Every IKEA kit]1 comes with instructions for assembling that1.Interestingly, some donkey readings are also often unavailable for bare demonstratives but perfectly acceptable for complex ones (Borg Citation2000, 248, footnote 14): (j) OK I bought [a donkey]1 and had it1/[that donkey]1 vaccinated.(k) # I bought [a donkey]1 and had [that]1 vaccinated.These contrasts are further proof that complex demonstratives pattern together with pronouns. While the mechanism that explains these facts is unclear, a plausible hypothesis to be explored is that it has to do with Φ-featural agreement. This would conform with the idea that the role of the nominal attached to a complex demonstrative is analogous to that of Φ-features.