«Knowledge Ascription by Grammatical Construction Laura A. Michaelis University of Colorado at Boulder 1. Introduction1 While little consensus has emerged ...»
have knowledge of its identity, as in (50), express lack of knowledge of its identity, as in (52), or provide its identity, as in (54) below:3
54. What I lost was my wallet.
a Presupposed: I lost x b Asserted: x = my wallet This analysis, while propositional, is distinct from the Intellectualist one, in that we view the propositional content conveyed by the WH-complement of a knowledge-ascription verb as presupposed rather than asserted. What is asserted is the speaker’s stance toward the value of the variable. The assertion of speaker stance is what makes an utterance containing such a clause informative. In other words, a WH-clause alone conveys nothing more than an open proposition (as, for example, what I lost in (54) conveys ‘I lost x’); it takes an embedding context to make such a clause into an assertion. Such embedding contexts need not be complementation contexts; they may instead be conversational or speech-act contexts. According to Lambrecht and Michaelis (1998), main-clause
questions like (55) assert a speaker stance toward the variable, captured by (56):
55. What did you lose?
56. With regard to your having lost x, I am inquiring about the value of x In exploiting properties shared by WH-complements across a variety of syntactic contexts, the present account resembles that of Stanley and Williamson (2001) and Stanley (2011), who take the formal similarity of know-how ascriptions to ascriptions of knowing-why, knowing-where, etc. as evidence that knowing how to F is in a family of mental states that include knowing where to F, knowing why to F, knowing when to F, etc., states that involve the normal knowing relation, together with an embedded question. (Stanley 2011: 226) The current account, however, provides a semantico-pragmatic representation of WHcomplements that makes sense regardless of the embedding verb; it is unclear whether The analysis given here of (54) is potentially controversial, since (54) is generally taken to exemplify a free relative-clause (i.e., the thing that I lost) rather than an indirect question. While the formal similarity between the two constructions creates ambiguities, as in (a), which has both a free-relative and an indirect-question paraphrase, as indicated in (b-c), respectively, certain syntactic tests distinguish the two patterns.
(a) I asked what she asked.
(b) I asked the question that she had asked. (free relative) (c) I inquired about what she had asked. (indirect question) One such test is described by Zwicky and Sadock (1975): insertion of the modifier the
hell, as in (d), allows for only the indirect-question interpretation:
(d) I asked what the hell she asked.
While such facts suggest that free relative clauses and WH-clauses are indeed distinct construction, I maintain that the former can revealingly be treated as denoting an open proposition in equative predications like (54).
the same could be said of the Intellectualist account. Stanley (2011) asserts, for example,
that (57) has the paraphrase given in (58):
57. John knows how to find coffee in New York City. (Stanley 2011, (1d))
58. For some way w, John knows that he can find coffee in New York City in way w.
(Stanley 2011, (2d)) This seems reasonable, and yet the proposed paraphrase relation appears to be restricted to sentences containing factive verbs like know. It does not appear to hold, for example,
when we replace the verb know with the verb ask: (60) is not a valid paraphrase of (59):
59. John asked how to find coffee in New York City.
60. For some way w, John asked whether w was a way to find coffee in New York City.
When John asks how to find coffee in New York City, he is not inquiring about the efficacy of a coffee-locating method that he already has in mind (say, using an iPhone application). Instead, he is seeking to discover a method. Because the paraphrase in (60) contains a wide-scope existential quantifier over methods, it does not capture what is going on in a context of inquiry, where the person making the inquiry does not yet know of a particular method, but only takes for granted that there is one. If, however, we translate the WH-complement how to find coffee in New York City as a proposition containing an unbound ‘means’ variable (i.e., ‘one finds coffee in x way in New York City’), and analyze the matrix verb ask as an indicator of the speaker’s stance toward that variable, it is easy to describe the meaning of (59): it asserts that John inquired about the value of a ‘means’ variable, just as (50) asserts that John knows the value of a ‘means’ variable. In other words, the current account may come closer to the compositional ideal than the Intellectualist one, in that it gives the same analysis of WH-clauses regardless of embedding context. In the present analysis, the true second argument of the verb know in (50), the verb wonder in (52) or the verb ask in (59) is not an open proposition but the variable contained within that open proposition.
Is an unbound variable the kind of thing that can be an argument? A recent study by Birner, Kaplan and Ward (2007) suggests that the answer is yes. This study examines the family of argument-structure constructions consisting of that-clefts (e.g., That’s John who wrote the book), equative clauses containing the epistemic verb would and a demonstrative subject (e.g., That would be John) and simple equatives with demonstrative subjects (e.g., That’s John). The latter two constructions, they argue, should not be analyzed as truncated clefts (pace Hedberg 2003). That is, they reject the
view that (61) is an elliptical version of (62):
61. That’s John.
62. That’s John who’s knocking on the door.
Instead, they argue, all three constructions inherit formal, semantic and informationstructure properties from an argument-focus construction used for equative assertions.
This construction contains a copular verb and a demonstrative subject, and it presupposes an open proposition whose variable is referred to by the demonstrative subject. The focal expression following the verb be provides the value of this variable, as in other argumentfocus predications (e.g., I saw JOHN). Thus, for example, in (61), the demonstrative subject refers to the variable in a presupposed open proposition, ‘x is at the door’.
But if the variable is the true second argument of a verb that takes a WH-complement, where is the propositional content of the WH-complement in our representation? It is in the presupposition, as indicated by the existential clauses in the paraphrases of (50) and (52). For example, (51), the paraphrase provided for (50), John knows how to make good coffee, contains the existential clause ‘there is x means by which one makes good coffee’.
Patterns of ellipsis in WH-complements support the view that the open proposition is presupposed rather than asserted: in a pattern called sluicing by Ross (1969), only the question word is present; the predication in which the question word plays an argument
role is deleted on the supposition that it is recoverable:
63. Sue can make good coffee but I don’t know how [...].
64. I left my keys somewhere, but I don’t know where [...].
The fact that the open proposition is omissible under conditions of contextual recoverability follows from its status as a topic, i.e., the entity or proposition about which the speaker is providing new information (Lambrecht 1994: Chapter 4, Lambrecht and Michaelis 1998). Topical arguments are predictable arguments; speakers omit them because hearers can reconstruct them from context. In the case of knowledge-ascription predications like (57), John knows how to find coffee in New York City, the topical proposition is the open proposition ‘One finds coffee in New York City using x method’.
The topic status of the open proposition is further substantiated by synonymy relations
like that in (65-66):
65. I know how to make good coffee.
66. I know the way to make good coffee.4 In (66), a noun phrase is used in place of a WH-complement. It is significant that this noun phrase contains the definite article, which elsewhere evokes an existence presupposition: both (65) and (66) take for granted that there is some way to make good coffee. Thus, while Stanley (2011) is correct in asserting that, e.g., the sentence John knows how to find coffee in New York City (sentence (57) above) “is naturally read as expressing the proposition ‘For some way w, John knows that he can find coffee in New York City in way w'” (p. 4), his analysis fails to distinguish between asserted and The complementation pattern exemplified by (66) may be unique to English. For
example, French native speakers find its direct translation ungrammatical:
(a) *Je sais la manière de faire du bon café.
Such cross-linguistic differences in verb complementation patterns are not unexpected, even in closely related languages. For example, while the complementation pattern exemplified in (64) is also found in French, as in (b), it is ungrammatical in German (c),
which requires a finite clause instead, as in (d):
(b) Il ne sait pas comment répondre. (‘He doesn’t know how to respond.’) (c) *Er weiss nicht, wie zu antworten (d) Er weiss nicht, wie er antworten soll. (lit. ‘He doesn’t know how one should respond.’) presupposed parts of that proposition. This sentence does not assert that John knows that there are ways to find coffee in New York City. Instead, it presupposes that there are ways to find coffee in New York City. It asserts that John knows one or more of those ways, i.e., one or more values of the ‘means’ variable.
4. Conclusion Neither the Intellectualist nor the Rylean model provides an adequate semantic analysis of the two major complementation patterns attested for verbs of knowledge ascription, namely, the infinitival and WH-complement patterns. The Intellectualist model perhaps comes closer, in that it correctly assesses such verbs as expressing a relationship between a person and a proposition. The problem with the Intellectualist model, as I see it, is that neither the infinitival-complement construction nor the WH-complement construction actually denotes this relationship. I have argued that the infinitival-complement pattern denotes a relation between a person and a procedure, where propositional knowledge represents a precondition for performing the procedure, and that the WH-complement pattern denotes a relation between a person and a ‘means’ variable in a presupposed open proposition—namely, the ability to identify that variable. The moral of this story is that the grammar of knowledge attribution is not monolithic, but is instead a constellation of constructions, each with its own array of semantic roles and use conditions.
But why should the grammar of English (or any other language) offer speakers two different ways of saying essentially the same thing? Put differently, what pragmatic considerations induce a speaker to use the infinitival-complement construction rather than the WH-complement construction when formulating a knowledge-ascription predication?
We can gain some insight into this question by contrasting reports of mundane abilities,
like those in (67-68), with reports of refined abilities, like those in (69-70):
67. Sue learned to swim.
68. Sue learned how to swim.
69. Sue learned to change lanes quickly.
70. Sue learned how to change lanes quickly.
While (67), (68) and (70) all assert that Sue attained knowledge required to perform a procedure, (69) seems instead to assert that Sue attained knowledge of when to use an
already mastered ability. In this respect, (69) appears similar to (71):
71. Sue knows to swim.
If (71) means anything, it means ‘Sue knows that one ought to swim (under some conditions)’. By the same token, (69) means something like ‘Sue learned that one ought to change lanes under some conditions’. It makes sense that (71) should have a deontic reading, since, as observed above in connection with example (42), the English verb know otherwise fails to license the infinitival-complement pattern. However, the English verb learn is clearly compatible with the infinitival-complement pattern. What then accounts for the deontic flavor of (69)? Comparison with French gives some clue. Native
speakers report the pattern of grammaticality in (72-73):
72. Elle sait (??comment) nager.
lit. ‘She knows (??how) to swim.’
73. Elle sait *(comment) changer de voie rapidement.
lit. ‘She knows *(how) to change lanes rapidly.’ As (72) shows, while the bare-infinitive form is preferable to the WH-infinitive form for nager (‘swim’), the reverse pattern holds for the more complex verb phrase changer de voie rapidement, as shown in (73). If we assume that the infinitival complement denotes a
mundane skill while the WH-complement denotes a rarer skill, this pattern makes sense:
the opposition seen in (69-70), and (72-73), may be a case of linguistic iconicity (Haiman 1980, 1983). According to Haiman’s quantity principle, linguistic complexity reflects conceptual complexity, and it therefore stands to reason that the less elaborate complementation pattern (the bare infinitive) should denote a more conventional skill than the more elaborate complementation pattern (the WH-complement).
5. References Birner, Betty, Jeffrey Kaplan and Gregory Ward 2007. Functional Compositionality and the Interaction of Discourse Constraints. Language 83: 317–343.
Fillmore, Charles J. 1986. Pragmatically Controlled Zero Anaphora. In V. Nikiforidou, M. Van Clay, M. Niepokuj and D. Feder, (eds.), Proceedings of the Twelfth Annual Meeting of the Berkeley Liguistics Society. Berkeley: Berkeley Linguistics Society.: 95-107.
Goldberg, Adele. 1995. Constructions: A Construction Grammar Approach to Argument Structure. Chicago: University of Chicago Press.
Haiman, John. 1980. The Iconicity of Grammar: Isomorphism and Motivation. Language 56: 515-540.
Haiman, John. 1983. Iconic and Economic Motivation. Language 59: 781-819.
Hedberg, Nancy 2000. The Referential Status of Clefts. Language 76: 891–920.
Jackendoff, Ray. 1997. The Architecture of the Language Faculty. Cambridge, MA: MIT Press.
Kay, Paul and Charles J. Fillmore. 1999. Grammatical Constructions and Linguistic Generalizations: The ‘what’s X doing Y’ Construction. Language 75: 1-33.
Kay, Paul. 2002. English Subjectless Tag Sentences. Language 78: 453-81 Lakoff, George. 1993. The Contemporary Theory of Metaphor. In A. Ortony, (ed.), Metaphor and Thought. Cambridge: Cambridge University Press. 202-251.
Lambrecht, Knud. 1994. Information Structure and Sentence Form. Cambridge:
Cambridge University Press.
Lambrecht, Knud and Laura A. Michaelis. 1998. Sentence Accent in Information.
Questions: Default and Projection. Linguistics and Philosophy 21: 477-544.
Moffett, Marc. 2005. Constructing Attitudes. Protosociology (Compositionality, Concepts and Representations I: New Problems in Cognitive Science) 21: 105-128 Oshima, David. 2009. Between being Wise and Acting Wise: A Hidden Conditional in Some Constructions with Propensity Adjectives. Journal of Linguistics 45: 363Ross, John. R. 1969. Guess who? In R. Binnick, A. Davison, G. Green and J. Morgan, (eds.), Papers from the Fifth Regional Meeting of the Chicago Linguistic Society.
Chicago: Chicago Linguistics Society. 252-286.
Michaelis, Laura A. 2004. Type Shifting in Construction Grammar: An Integrated Approach to Aspectual Coercion. Cognitive Linguistics 15: 1-67.
Sag, Ivan A. 2010. English Filler-Gap Constructions. Language 86: 486–545.
Stanley, Jason. 2011. Knowing How. Nous 45: 207-238.
Stanley, Jason and Timothy Williamson. 2001. Knowing How. The Journal of Philosophy 98: 411-444.
Zwicky, Arnold & Jerrold Sadock. 1975. Ambiguity Tests and how to Fail them. In P.
Kimball, (ed.), Syntax and Semantics 4. NY: Academic Press, Inc. 1-36.