Case study I: the let alone construction
Fillmore et al. are particularly interested in accounting for formal idioms because, while it is at least plausible that speakers might learn substantive idioms item by item rather like learning individual words, it is not plausible that a speaker learns each instance of a formal idiom item by item. In principle, the number of instances of formal idiom constructions is infinitely large. Despite this, such constructions often have a clearly identifiable pragmatic force. For this reason, formal idioms pose a particularly interesting challenge to the ‘words and rules’ model of grammar: they are productive and therefore rule-based (systematic), yet often defy the ‘usual’ rules of grammar. Fillmore et al. therefore took as their case study the idiomatic let alone construction.
According to Fillmore et al., the let alone construction can be described in terms of its structural, semantic and pragmatic properties, some of which are regular and some of which are idiosyncratic. The let alone construction displays regular syntactic properties, and is characterised by the presence of the coordinating conjunction let alone, which coordinates two prosodically prominent (stressed) expressions. This construction is illustrated by example (1a), which is repeated here as (3). In this example, the expressions in bold type, maths and rocket science (labelled as A and B respectively), are prosodically prominent and are coordinated by let alone.
In semantic terms, the construction has the idiosyncratic property that the coordinated expressions are interpreted as contrasted points on a scale, where the second conjunct (rocket science) has greater emphatic force than the first (maths). In the context of knowing that Lily, the famous rocket scientist besot ted with George, often tells George about her work, we might ask whether George in fact understands rocket science. The utterance in (3), as a result of the let alone, conveys they information that because George doesn’t understand maths, he is even less likely to understand rocket science. This rests upon the assumption that ‘understanding maths’ is a prerequisite for ‘understanding rocket science’.
Closely related to this property of the construction is the fact that let alone can be described as a negative polarity item. This means that it can only occur in negative contexts, whether this is determined by a morphosyntactic negation, as it is in example (3), or by a lexical item like doubt, which brings with it a negative interpretation. This is illustrated by example (4).
The let alone construction has pragmatic point. Not only does the construction reject a particular proposition (for example, that George understands rocket science or can drive a car), but it does so by providing additional relevant information. The relevant information relates to the first conjunct (A) and establishes an implicational scale between the expressions conjoined by let alone. If George doesn’t understand maths (A) this implies that he doesn’t understand rocket science (B). The pragmatic impact of this construction is that by first rejecting a weaker proposition, the proposition that our attention is focused upon (e.g. whether George understands rocket science) is more force fully rejected than it would otherwise have been. These idiosyncratic properties of the let alone construction are in fact shared among a ‘family’ of similar constructions. Some examples are provided in (5).
In light of their findings concerning the let alone construction, Fillmore et al. argue against the ‘words and rules’ view (which they call the ‘atomistic’ view) of grammatical operations, where lexical items are assembled by phrase structure rules into complex units that are then assigned compositional meaning and only subsequently subjected to pragmatic processing. In other words, they argue against a modular view of the language system. Instead, Fillmore et al. (1988: 534) argue that speakers have, as part of their linguistic knowledge or competence, ‘clusters of information including, simultaneously, morphosyntactic patterns, semantic interpretation principles to which these are dedicated, and, in many cases, specific pragmatic functions in whose service they exist.’ In other words, speakers have access to constructions.
At this point, we should pause to consider the various senses of the term ‘construction’. In traditional grammar, this term refers to a clause type, such as the ‘passive construction’ or the ‘cleft construction’. These labels apply to the sentence as a whole, which can be classified as construction X or construction Y on the basis of certain morphosyntactic or semantic proper ties. For example, the passive construction (6a) is identified by the fact that the subject is interpreted as the PATIENT, while the (optional) by-phrase expresses the AGENT. In addition, it is identified by the presence of the passive auxiliary be and the past participle form of the content verb. This information can be schematically represented as in (6b). In a similar way, the (subject) cleft construction (7a) can be captured by the schematic representation in (7b).
In the Chomskyan generative model, these constructions have the status of ‘taxonomic epiphenomena’ (Chomsky 1991: 417). In other words, the model of grammar does not need to contain whole constructions because these can be predicted on the basis of the words and rules that the grammar contains. This means that most generative linguists use the term ‘construction’ as a shorthand for describing certain types of syntactic structures that have certain identifiable properties (for example, ‘the passive construction’ or ‘the wh-construction’), but these constructions are not themselves primitives in the model. Instead, they are the output of the ‘words and rules’ model and as such are not of central importance. Instead, the emphasis in this model is upon characterising the rules that give rise to the constructions.
Against this background, it is clear that Fillmore et al.’s proposal reflects a very different view of how language should be modelled. Instead of a model in which syntactic, semantic, phonological and pragmatic knowledge is represented in encapsulated subsystems, the constructional model proposes that all this information is represented in a single unified representation, which is the construction. In the next section, we will look in detail at a representation of an idiomatic construction, and we will discuss in more detail what it means to develop a constructional model of language and in what sense this type of approach can be held to account for both regular and idiomatic properties of language. Indeed, the constructional model proposed by Charles Fillmore, Paul Kay and their colleagues, grounded in their work on idioms, provided the empirical basis for the symbolic thesis which, as we saw in Chapter 14, is central to a cognitive approach to grammar.
However, it is important to emphasise that Fillmore et al.’s discussion of the let alone construction is situated within a broadly generative paradigm rather than a cognitive linguistics paradigm. For example, part of their paper is concerned with the rules that might underlie the let alone construction, and this theoretical context also explains the separation of semantic and pragmatic meaning in their discussion of the construction. Nevertheless, their proposal that speaker knowledge is not ‘compartmentalised’ but ‘clustered’ around individual constructions represented an important shift in terms of how speaker knowledge could be modelled and set the scene for the emergence of constructional models of grammar.
