DIFERENCIAS ENTRE LOS PROYECTOS Y LOS CENTROS DE INTERÉS
6. Elaboración de un dossier.
King’s claim in (ii) about the proper account of the relation between form and meaning turns on how we understand LF (an abbreviation for logical form). Unsurprisingly, not all theorists understand LF the same way. Dowty (1982), for example, prefiguring King (2003), identifies himself as one of those theorists
striving toward analyses which make the logical form of a sentence as close to the apparent surface syntactic form as possible. By ‘logical form’ I of course don’t mean the translation of an English sentence into intensional logic, much less do I mean ‘logical form’ as is currently understood in the Extended Standard Theory. Rather, ‘logical form’ is simply that syntactic
5.4 Syntactic Structures | 153 analysis of a sentence upon which compositional semantic interpretations is directly based, i.e., the analysis tree of a sentence (p. 28; original emphasis).16 Since a given set of expressions, such as the sentences of English, can be generated by any number of recursive syntactic rules, which are in this sense extensionally equivalent (or share a weak generative capacity in Chomsky’s sense), there is noprivileged syntactic
analysis over which a compositional semantic interpretation can be defined. Rather, any appropriate syntactic analysis will do the trick, though Dowty does note a preference for those that cleave closer to surface structure, which leads him, unlike King, to an intensional treatment of tense. Regardless, no question about the exact means by which natural language encodes tense and modality arises on this conception of LF.
Contrast this with understanding of LF that implicit in Kamp (1971). In s. 5.2.1 I alluded to a footnote accompanying Kamp’s double-indexed intensional analysis of (8).
(8) A child was born that will become ruler of the world That footnote reads as follows.
I of course exclude the possibility of symbolizing the sentence by means of explicit quantification over moments. Such a symbolization of [(8)] would certainly be possible; and it would even make the operators P and F super- fluous. Such symbolizations, however, area considerable departure from the actual form of the original sentences which they represent – which is unsat-
isfactory if we want to gain insight into the semantics of English. Moreover, one can object to symbolizations involving quantification over such abstract objects as moments, if these objects are not explicitly mentioned in the sentences that are to be symbolized (fn. 1; emphasis added).
Kamp (personal communication) confirms that, unlike Dowty, he had the LF of generative grammar in mind when he wrote about “the actual form of the original sentences.” More generally, Kamp’s remarks reveal a method of theorizing that takes the deliverances of contemporary syntactic theory as a constraint on semantic analysis. However, Kamp’s talk of what is “explicitly mentioned in the sentences,” on the one hand, and the “actual form of the original expression,” on the other, leaves it somewhat unclear whether surface structure or some other level of syntactic representation is the appropriate constraint on semantic analysis.
16 Extended Standard Theory was the 1970s iteration of generative grammar, beginning with Chomsky
Although some theorists are primarily concerned with adhering to surface form/structure, it’s not immediately clear why this should be thought of as a significant level of repre- sentation that needs to be respected or preserved. In particular, if “surface structure” just refers to the conventional orthographic representation of linguistic expressions in, e.g., the English alphabet, it has no obvious claim to privileged status vis-á-vis theoretical inquiry into natural language. In the first instance, linguistic competence is acquired reflexively, with little or no explicit instruction whereas the ability to read and write in accord with (local) orthographic conventions requires explicit instruction, but is in no sense necessary for competence with a language. There are many illiterate people in the world, but no one would suggest, on that count, that they lack competence with their language. Theoretical inquiry into linguistic competence, then, is strictly speaking independent of orthography.
Additionally, an alphabet, the orthographic system of a given language, is evidently just a folk linguistic division of the continuous sound wave that accompanies speech into discrete units that are meant to represent distinct sounds therein. This is a rather crude representation, however, given that a single letter can carry the weight of more than one sound.17 As such, conventional orthographic representation is not necessarily revelatory of anything besides the primitive conception of language it reflects. But once we see that the “surface structure” is, in this sense, arbitrary, there is no obvious reason to privilege it above other levels of representation of expressions, each of which is a theoretical posit and needs to be justified accordingly.
In any event, setting aside the theoretical import of “surface structure”, the pertinent level of representation in (ii) is LF, which is also the interface level between syntax and semantics posited in generative grammar. The simplest way of articulating the concept of LF, as it is understood in generative grammar, is the following.
Logical Form
LF is the level of linguistic representation at which all grammatical structure relevant to semantic interpretation is provided.18
In other words, whatever interpretable constituents are present in an expression of natural language are found at LF.
Within the minimalist program, the current iteration of generative grammar, two conditions hold that are important for how to understand LF, full interpretation
17 For example, the ‘t’ in North American English corresponds to the so-called plain ‘t’, as in ‘sty’,
the aspirated ‘t’, as in ‘tie’, as well as a flap, as in ‘butter’, and a glottal stop, as in ‘cat’. See Isac and Reiss (2008) s. 6.3 for discussion.
5.4 Syntactic Structures | 155
and inclusiveness. Full interpretation is a general condition of all levels of linguistic
representation that “there are no superfluous symbols in representations” (Chomsky (1995), p. 24). When applied to LF, full interpretation “entails that every element
of the representation have a (language-independent) interpretation” (ibid., p. 24). So LF provides all structural information relevant to semantic interpretation, and every bit of that structural information is given a semantic interpretation, there are no redundant constituents.19 Inclusiveness is a condition applying to the computational procedure that gives rise to complex expressions: “no new objects are added in the course of computation apart from rearrangements of lexical properties (in particular, no indices. . . )” (p. 209). In terms of LF, inclusiveness amounts to the constraint that
all constituents of LF derive from the lexicon.
The burden of the extensional treatment of tense (and modality), then, is to demonstrate that the structures underwriting its analysis cohere with the syntactic structures posited by syntactic theory. In particular, to be a serious proposal regarding natural language syntax, the extensional theorist will have to motivate the presence of the constituents it posits, such as variables, at LF. As Collins (2007) notes, “an item of syntax must be such that it is (i) merged and (ii) carries un/interpretable features to which the derivation is sensitive” (p. 840). In other words, each posited syntactic constituent needs to be specified with respect to its category and features, the latter of which determine which other lexical items it can merge with and which interpretive dependencies it can instantiate (since movement is driven by morphological feature checking).20 Additionally, each posited item must be an element of the lexicon (the only way to enter into a syntactic derivation). Extensional treatments of tense that posit covert syntactic items in the service of semantic analyses without specifying these syntactic details are really only halfway proposals, not yet serious syntactic proposals, not yet complete proposals for how to tense (and modality) is encoded in natural language.
Particularly difficult, in this regard, will be the status of variables, since it’s not clear, on the traditional logical conception of variables as entities that are inherently free or bound, that variables are to be found in syntax. Chomsky (1981), for example, provides the following definition of a variable in his syntactic theory: “α is a variable if and only if α = [N P e] in S bound by an operator [. . . ] [we] understand operators to
19 Chomsky adds, “[t]here can, for example, be no true expletives, or vacuous quantifiers, at the LF
level” (p. 24).
20 Cf. Chomsky (1995) p. 210: “A core property of C
HL [the computational component of the
language faculty] is feature checking, the operation that drives movement under the Last Resort condition.”
include quantifiers, wh-phrases, or a binding NP in COMP; or definite or indefinite operators as in relative clauses” (p. 102). More recently, in terms of minimalism, Chomsky (1995) writes, “FI [full interpretation] at LF includes the property of strong binding: every variable must have its range fixed by a restricted quantifier, or have its
value determined by an antecedent” (p. 51). This conception of variables is consistent with the extensional approach at least insofar as the variables are bound by an operator of some sort, but it leaves no room for “free” variables, as a referential analysis would contend for so-called deictic cases of tense (see ss. 5.2.2 and 5.3.1 above). In other words, only bound variables appear to be present in syntax, marking interpretive dependencies between syntactic constituents (i.e., operators and variables), not the logician’s variables.21 Additionally, operators move during the course of syntactic derivations of complex expressions, and as such they are subject to constraints on movement more generally (e.g., no extraction our of subjects, out of relative clauses, etc.), leave a trace (i.e., a bound variable), and have a determinate category.