Grammar and language production Where do function words come from?

Human Cognitive Processing

Volumn 15, Issue , 2006, Pages 139-168

  Schilperoord, Joost ^a   Verhagen, Arie ^b  

^a TILBURG UNIVERSITY   (Netherlands)

^b LEIDEN UNIVERSITY   (Netherlands)

Author keywords

function words; grammatical schemata; language production; storage vs. computation

Indexed keywords

EID: 85086701757     PISSN: 13876724     EISSN: None     Source Type: Book Series    
DOI: 10.1075/hcp.15.11sch     Document Type: Chapter

References (44)

1. 2” = “N-double bar”).
2. Kempen and Hoenkamp’s notation of syntactic structures differs from standard generative tree structures in that they explicitly specify at least some of the grammatical functions. This functional information, even though it may strictly speaking be redundant, must be represented in the structure at some point anyhow in order to function as a trigger for the relevant functorization procedures, in this case for example the ones that ultimately result in the insertion of the preposition of.
3. This question was raised by one of the reviewers of this paper. Although we are not aware of research into the distribution of pauses with respect to function words in spontaneous conversation, incidental observations, including some reported in the literature (e.g., Clark 1996:268) do suggest that similar patterns at least occur in conversation as well. See Schilperoord (2001) for various methodological and empirical aspects of dictation research.
4. Of course, pauses may have various other sources than cognitive ones, and this may endanger the validity of both our data and the conclusions drawn from them. In our research, we consider a pause ‘cognitive’ if it reflects conceptualization processes or lexical retrieval (see Boomer 1965; Schilperoord 1996). But what about other sources of pausing, how can we be sure to have kept pauses from other sources out of the corpus? We should first distinguish between pauses that are involuntary, and pauses that language producers willingly insert into the stream of speech. These latter pauses occur by intent and often serve rhetorical or communicative purposes, i.e. they are oriented towards an addressee. Clearly, such pauses could not be considered cognitive in the above sense. However, the possibility of such pauses being present in our corpus can safely be ruled out because of the strictly monologic nature of the production circumstances. All letters in our corpus were dictated to a machine, not to secretaries taking notes. Hence, pauses cannot even have resulted from a friendly employer pausing for the typist’s convenience. But even if pauses can be considered involuntary, they still can be caused by various factors. In terms of the IPG-model, pauses may be caused by all main components of the model, and therefore they may reflect conceptualization processes (preparing what to say), lexical-grammatical processes (retrieving lexical items), morpho-phonological processes (accessing word forms), monitoring processes (monitoring one’s own production), or they may originate from the workings of the articulator. Let us briefly consider these factors in turn. Obviously, the first two factors do not pose any problem since these are the factors that we are interested in in the first place. Pauses caused by the articulator were excluded from the corpus on grounds of pause duration. Dechert and Raupauch (1980) have calculated that ‘breathing’ pauses last.3 seconds at most, so we simply excluded pauses up to that length from the corpus. One should keep in mind that pauses lasting longer than. 3 seconds may reflect articular activity, but in those cases one can be sure that this is not the only factor causing these pauses. In other words, pauses lasting over. 3 seconds at least also originate from cognitive processing (see also Note 5). Then there may be morpho-phonological factors causing pauses manifesting the so-called tip-of-the-tongue. Clearly, such pauses are not cognitive. However, as we devote a lengthy discussion to this possibility in Section 3 (p. 14), we leave this issue aside here. In addition, pauses may originate from the workings of the monitor. While producing texts, text producers constantly monitor their own production. They attend to various aspects of their actions, such as content, choices of phrasing, and so on. Monitoring becomes apparent from various types of self-repairs that are produced ‘on the fly’, that is, while producing speech, but the monitoring process may also cause pauses itself. Once again, monitoring pauses are not the type of pauses that we are interested in here, so how can we be sure that monitoring does not interfere with conceptualization and lexical retrieval? To be honest, we cannot in any strict sense. However, there is some circumstantial evidence that in dictation, monitoring predominantly occurs at pre-established locations: major text structural locations such as prior to paragraphs and sentences. While it is clear that in spontaneous speech, the orientation of monitoring is mainly backwards, under the far more controlled production circumstances that we are dealing with here, its orientation is mainly forwards. That is, while dictating letters, text producers devote quite some attention to conceptual planning pieces of text in advance. One factor suggesting this is the fact that in dictation self-repairs are almost totally absent. Another point is the fact that pausing between paragraphs or sentences last considerably longer than pausing within sentences and clauses (see Schilperoord 1996, 2001), suggesting that at these locations preplanning the content of text parts takes place. Since the pauses that we are interested in are all located around function words, there seem to be good reasons for assuming that such pauses reflect the processes of refining conceptualization or retrieving lexical items. To conclude, our considerations thus far suggest the pauses in our corpus to be mainly caused by cognitive factors (conceptualization, lexical retrieval). Admittedly, other factors can never be ruled out completely, but in the absence of any compelling evidence that such factors correlate structurally with the relevant location types that we consider in this chapter, we may safely assume that these other factors are randomly distributed, and hence do not jeopardize the validity of the data. Finally, we would like to stress the fact that pausing in language production is an empirical phenomenon, and that pausing parameters, such as pause locations, can be analyzed independently from any pre-established theoretical point of view, be it computational psycholinguistics, or cognitive linguistics. What matters, in our view, is how to arrive at a proper account of this issue.
5. In psycholinguistics. 3 seconds is the generally accepted ‘cut off’ value for a pause to be taken as reflecting some cognitive activity, rather than as resulting from muscular activities of the vocal tract. See for example Dechert and Raupauch (1980).
6. This may have something to do with the somewhat ambivalent status of prepositions with regard to their category status: lexical or functional. See Section 4.2 for further discussion, and also Schilperoord (1996), Schilperoord and Verhagen (1997).
7. In case a reader wonders what is ‘fixed’ about these expressions, compare them with the phrases this cup of coffee and a bathroom (e.g. in Would you like this cup of coffee? or Where can I find a bathroom, please?).
8. In Jackendoff’s (2002) theory, lexical items are viewed as correspondence rules between semantic, syntactic and phonological information. Moreover, a lexical entry may be both larger and smaller than an individual word. Idioms are a case in point, but a plural suffix, as an item licensing the formation of plural forms, is also a lexical entry. These assumptions are shared by all present construction based approaches to grammar. Croft (2001) may be seen as arguing against a separate level SS for syntactic information, essentially because there is no way to define the necessary global syntactic notions (“noun”, etc.) in a non-circular fashion, independently of (language) specific constructions. In his view, the information usually considered syntactic reduces to schematic aspects of form and to the symbolic relation between form and meaning. On the other hand, Croft’s view seems to allow for language specific distributional classes to be included in the specification of the form of a construction. As this issue is not directly relevant for the present discussion, we use the more conservative notation here. To avoid misunderstanding: we use Jackendoff’s formalism only for reasons of convenience. As he has repeatedly and rightly pointed out himself, the formalism does not assume any particular theoretical point of view.
9. This is the term used in Jackendoff (2002); another term used for essentially the same concept is ‘unification’. See Goldberg (1995), among others, for discussion of the way this notion fits into the theory of Construction Grammar.
10. While the possibility of a direct relationship between a function word and conceptual structure is a necessary condition for language production as we see it, a reviewer suggested that it might be a sufficient condition. For example, the production of a determiner such as the could be motivated by the presence of the feature +accessible in the conceptual structure, but it need not activate the structure “det–N”, which might still come from the head noun. Being lexically driven or not and being structure building or not are in principle separate characteristics of a production model. On logical grounds, such a possibility cannot be foreclosed, obviously. However, it is first of all not a part of IPG, and second, we have explicitly based our proposal on the constructional approach. The analyses of function words that we are aware of, all share the view that precisely what makes these elements “grammatical”, is the fact that they do not function independently (they are “bound forms”), and are necessarily associated with other, variable linguistic material. We thus continue to assume that activation of a function word by a feature of the conceptual structure also activates the associated schema.
11. For ease of exposition, we conflated the two formal representational levels S[yntactic] S[tructure] and P[honetic] S[structure]. But see also Note 8.
12. 2 (1) = 3.31, p >. 10).
13. Confusingly labelled ‘clitics’; they are not pronominal and they are also phonologically independent.
14. Recall that analyzability does not imply compositionality (in the sense of ‘having been composed’). If elements can be distinguished within a linguistic unit (analyzability), it does not follow that the unit has been constructed out of these elements. Even obvious idioms, necessarily stored as units, may exhibit analyzability: in spill the beans, the element spill corresponds to the semantic component divulge and the beans corresponds to information. For a recent discussion, moving in a somewhat different direction, cf. Croft (2001: 180–184).
15. This position resembles the one defended for linguistic theory in general on the basis of methodological, typological and analytic considerations in Croft (2001), and from the perspective of acquisition in Slobin (2001). In a sense, our analysis provides an additional argument from processing for the hypothesis that global structural notions do not really have explanatory power, and are not primitive but rather based on similarities between specific constructions (cf. Verhagen 2002:420/421).
16. Ariel, Mira (1988). Referring and accessibility. Journal of Linguistics, 24, 65–87.
17. Baddeley, Alan (1990). Human memory. Theory and practice. Hove: Lawrence Erlbaum Associates.
18. Boomer, David S. (1965). Hesitation and grammatical encoding. Language and Speech, 8, 148–158.
19. Carroll, David W. (1999). Psychology of language (3rd ed.). Pacific Grove, CA: Brooks/Cole Publishers.
20. Clark, Herbert H. (1996). Using language. Cambridge: Cambridge University Press.
21. Croft, William (2001). Radical construction grammar. Syntactic theory in typological perspective. Oxford: Oxford University Press.
22. Dechert, Herbert W. & Marius Raupauch (Eds.). (1980). Temporal variables in speech. Studies in honour of Frieda Goldman-Eisler. The Hague: Mouton.
23. Erman, Britt & Beatrice Warren (2000). The idiom principle and the open choice principle. Text, 20, 29–62.
24. Goldberg, Adele (1995). Constructions: A construction grammar approach to argument structure. Chicago/London: University of Chicago Press.
25. Haaften, Ton van (1991). De interpretatie van verzwegen subjecten [The interpretation of understood subjects]. Diss. VU Amsterdam. Dordrecht: ICG Printing.
26. Jackendoff, Ray (1990). Semantic structures. Cambridge, MA: MIT Press.
27. Jackendoff, Ray (1995). The boundaries of the lexicon. In M. Everaert et al. (Eds.), Idioms: Structural and psychological perspectives (pp. 133–167). Hillsdale, NJ: Lawrence Erlbaum Associates.
28. Jackendoff, Ray (1997). The architecture of the language faculty. Cambridge, MA: MIT Press.
29. Jackendoff, Ray (2002). Foundations of language; Brain, meaning, grammar, evolution. Oxford: Oxford University Press.
30. Kay, Paul & Charles J. Fillmore (1999). Grammatical constructions and linguistic generalizations: the What’s X doing Y? construction. Language, 75, 1–33.
31. Kempen, Gerard & Edward Hoenkamp (1987). An incremental procedural grammar for sentence formulation. Cognitive Science, 11, 201–257.
32. Langacker, Ronald W. (1990). Concept, image and symbol. The cognitive basis of grammar. Berlin/New York: Mouton de Gruyter.
33. Langacker, Ronald W. (1991). Foundations of cognitive grammar, Volume II, Descriptive application. Stanford: Stanford University Press.
34. Levelt, Willem J. M. (1989). Speaking: From intention to articulation. Cambridge, MA: MIT Press.
35. Levelt, Willem J. M. (1999). Producing Spoken Language: a blueprint of the speaker. In P. Hagoort & C. W. Brown (Eds.), The Neuro-cognition of Language (pp. 94–122). Oxford: Oxford University Press.
36. Pardoen, Justine A. (1998). Interpretatiestructuur. Een onderzoek naar de relatie tussen woordvolgorde en zinsbetekenis in het Nederlands. [Interpretation Structure. A study of the relation between word order and sentence meaning in Dutch.] Amsterdam/Münster: Stichting Neerlandistiek VU/Nodus Publikationen.
37. Sanders, Ted, Wilbert Spooren, & Leo Noordman (1992). Towards a taxonomy of coherence relations. Discourse Processes, 15, 1–35.
38. Schilperoord, Joost (1996). It’s about time. Temporal aspects of cognitive processes in text production. Amsterdam/Atlanta: Rodopi.
39. Schilperoord, Joost (2001). On the cognitive status of pauses in discourse production. In T. Olive & M. C. Levy (Eds.), Contemporary tools and techniques for studying writing (pp. 60–89). Dordrecht, Boston, London: Kluwer Academic Publishers.
40. Schilperoord, Joost & Arie Verhagen (1997). Functionele elementen in een cognitief perspectief. Evidentie uit taalproductie [Functional elements in a cognitive perspective. Evidence from language production]. Nederlandse Taalkunde, 3, 223–248.
41. Schilperoord, Joost & Arie Verhagen (1998). Conceptual dependency and the clausal structure of discourse. In J. Koenig (Ed.), Discourse and cognition; Bridging the gap (pp. 141–165). Stanford, CA: CSLI Publications.
42. Slobin, Dan I. (2001). Form-function relations: How do children find out what they are? In Melissa Bowerman & Stephen C. Levinson (Eds.), Language acquisition and conceptual development (pp. 406–449). Cambridge: CUP.
43. Verhagen, Arie (2001). Subordination and discourse segmentation revisited, or: Why matrix clauses may be more dependent than complements. In Ted Sanders, Joost Schilperoord, & Wilbert Spooren (Eds.), Text representation. Linguistic and psycholinguistic aspects (pp. 337–357). Amsterdam: John Benjamins.
44. Verhagen, Arie (2002). From parts to wholes and back again. Cognitive Linguistics, 13, 403–439.