On Kudo’s (2011) pragmatic model of mental representation

In spite of these differences, Kudo (2011) analyses the BPOC as a pattern whose interpretation is parallel to that of (other) resultatives (e.g. He cried his eyes red; She ate herself sick). That is, Kudo suggests that sentences such as (6a-b, 7a-c, 8b) are made sense of by hearers as referring to intense events only because their literal interpretation (on which they refer to events in which a body part actually comes off as a result of the action) would not be feasible in the real world. The BPOC, according to Kudo, is thus interpreted by means of a general interpretive mechanism, one which we also use to interpret an utterance such as The joggers ran the pavement thin (Goldberg 1995:184), whose form is that of a causative property resultative, as having an intensifying postverbal sequence.

It would seem, then, that we are dealing here with what Croft (1998) calls the “pragmatic model of mental representation”, which he defines as follows: “There is one independently represented unit in the mind with a general meaning U, [a/U]; (a/U1) and (a/U2) are derived from the general meaning U and general cognitive principles relevant to the specific context of use” (Croft 1998: 154). In the case of Kudo’s proposal, [a/U] is then a general resultative construction, from which a literal reading and an excessive-event reading can be derived pragmatically, both for the caused property resultative and for the BPOC.

Croft (1998) proposes a fairly straightforward means of excluding this pragmatic model for any given linguistic unit. A linguist should check whether the translation equivalent of a supposedly pragmatically accountable use of that unit in another language is (mutatis mutandis, i.e. taking into account more general structural differences as well as trivial lexical differences) formally equivalent; if there is no formal equivalence, then the use of that unit must be considered to be stored as such in the first language, rather than being computed “on-line”. As an example, Croft gives the noun-noun (NN) compound water tower, which in essence is indeed a tall structure supporting a tank of water. In French, however, such structures are referred to as “water castles” (chateaux d’eau). Croft (1998:159) concludes: “This cross-linguistic difference is strong evidence that the pragmatic model is not appropriate for the derivation of this particular noun compound; otherwise speakers of both languages would come to the same solution to the naming of this object”. Of course, one might object that the French term is in fact a rather special and unexpected combination while water tower is a fully transparent NN combination, so that it is only the French term, not the English one, which requires storage in the mental lexicon. One might also point out, as Croft is in fact aware one might, that castles may have been culturally more salient (i.e. frequently occurring) structures in France when water towers made their appearance, so that when they were first dubbed chateaux d’eau, this made perfect sense, given the fortress-like appearance of the earliest water towers. The point is, though, that for present-day speakers of French and English, one cannot plausibly assume that water tower and chateau d’eau are coined anew on each occasion a term referring to the object in the world is called for; more likely, speakers in a given speech community, French or English, just re-use the conventional expression available to them to denote this object.

Crucially, Croft (1998:159) further suggests that “[t]he same argumentation can be applied to more general grammatical patterns.” In the next section (section 3.3) I will indeed apply Croft’s “test” to the BPOC. For one anonymous reviewer, Croft’s translatability test fails to convince as an argument against a purely pragmatic approach: “Why would ‘pragmatic’ constructions necessarily be the same in all languages?” I nevertheless believe that the test can be usefully applied to languages which otherwise provide their speakers with comparable encoding possibilities. So, even though water tower may look as though it is a purely “pragmatic” choice, there is nothing, except the blocking effect triggered by water tower itself, which in principle prevents speakers of English from producing alternative terms for this object. Among other imaginable possibilities, we could have had terms similar to French chateau d’eau, such as water castle, water fortress or water keep, given that there was also a Gothic architecture revival in England and the United States. The crucial idea here is that the existence of a rather different encoding choice in another language may alert one to the not-so-obviousness of a given lexical choice in the language under study. Note, moreover, that speakers of English could just as well have modeled the term for the object in question on the familiar word light house, and thus have coined water house, which indeed is actually a now obsolete term for water tower. Still other terms could have been water tank tower, hydrotower, water building, water pillar, water silo or cistern tower. English does in fact have an alternative: water towers are sometimes referred to as standpipes. In short, alternative encodings in other languages - or in (older stages of) the same language - may help us realize that a complex expression which seems transparently compositional is actually conventionalized and thus listed as a learned language item. And the further claim made here is that if a combination is conventionalized, it is no longer assembled from scratch each time it is used, nor is it decoded inferentially (“it definitely can’t be a tower made of water, so it must be a tower that contains water”. . .) each time it is used.

For now, let me point out that from the early days of construction grammar, it has been suggested that constructions can be stored along with specific “pragmatic” properties beyond their so-called “literal” meaning. For example, in their classic paper on the let alone construction, Fillmore, Kay and O’Connor (1988) explicitly link up pragmatic instructions on the use of this construction with the construction itself. Likewise, Fillmore and Kay (1999) argue that the What’s X doing Y? construction comes with rich semantic information, which includes instructions for its conventional pragmatic usage. As a result, an utterance such as Waiter, what’s that fly doing in my soup? is - or should be - understood as expressing that the situation described is judged to be incongruous rather than as being a straightforward question for information. The well-known joke (“I believe that’s the backstroke, Sir”) hinges on the waiter blatantly ignoring this pragmatic part of the construction. Conventionalized indirect speech acts of the type exemplified in Can you pass me the salt? have also been discussed from a construction grammar perspective by Stefanowitsch (2003).^[1]

In the present paper, too, the position defended is that the “non-literal” reading of the BPOC construction has become conventionalized as part of the semantic component of a separate construction. Even though the literal reading clearly keeps on lingering in the background, it is not via this reading and Gricean or other general rules of communication that the intended, “non-literal” reading is arrived at. This position does not blur the traditional distinction between semantics (or “literal meaning”) and pragmatics (“typical usage”). It merely requires that for some complex language items, the stored semantic information be backgrounded and the stored pragmatic information be foregrounded.

• [1] “The problem a hearer is faced with when hearing strings like [Can you close the window?;Would you mind telling me the time?; I would like a cheeseburger] is not to process its “literal”meaning, determine that the utterance does not lend itself to a literal interpretation, and theninfer possible “non-literal” meanings. [...] Instead, the problem faced by the hearer is to realizethat each of these expressions is ambiguous in terms of the conventional meanings attached tothe constructions they instantiate, and to use context in order to determine which of their conventionalized meanings is the one intended by the speaker. [...]” (Stefanowitsch 2003:122-123)