2 CAPÍTULO II MARCO TEÓRICO
2.21 VIGAS
2.21.2 SECCIONES RECTANGULARES SIMPLEMENTE REFORZADAS
Returning to how these predispositions relate to the current research, perspective maintenance is hypothesized to play a role in multiple clauses appearing in sequence (both in sequential monoclausal utterances and even more so in multiclausal utterances), in which a perspective established in the first clause is maintained into following ones. The phenomenon of perspective maintenance may aid in the interpretation of relatives, complements, and conjoined clauses, in which elements like pronominals play a role (e.g.,Louise told Suki that [her] coffee was decaf; “her” = Louise/Suki). Furthermore, and more relevant to the experiments described here, perspective also appears to help children in the interpretation of passives, by establishing a semantic focus.33 By hypothesis, once
32As a side note,
Behaghel’s Law of Increasing Terms (1932) resembles MacWhinney’s (1982) predisposition of grammatical complexity; the former suggests that whenever possible, a shorter term will precede a longer one. On other matters, though, MacWhinney and Behaghel disagree. For instance, Behaghel’s Second Law claims that old information will be placed before new; this stands in direct opposition to MacWhinney’s notion ofinformativeness. Behaghel’s statement might instead correspond more closely with linguists’ notions of the “theme-rheme” (i.e., topic-focus, or topic-comment) structure inherent in language (seeLi and Thompson,1976, for extension discussion on this matter).
Indeed, MacWhinney himself notes that, although informativeness is the predisposition which has received the most attention in the child language literature, this predisposition runs completely counter to the adult tendency to order the old before the new. Both strategies are explainable, given the distinct (but related) cognitive effects of primacy and recency (Deese and Kaufman,1957;Murdock,1962; Glanzer and Cunitz,1966;Atkinson and Shiffrin,1968; see also the discussion of Experiment 7 in Chapter7below). In general, however, it is worth noting explicitly here that “unmarked” (default, or first acquired) in child language simply does notalwaysequate to “unmarked” (most common) in adult language; the childhood phenomenon of informativeness and the patterns leading to the formulation of Behaghel’s Second Law may be a perfect example supporting this notion.
33For instance, Dewart (1975; cited in
MacWhinney,1982) found that, given the sequencebad duck
# the cat was bitten by the duck, children who “did not know the passive” correctly chose the duck as the actor in the event; in contrast, children who heard poor cat # the cat was bitten by the duckcould correctly choose the duck as the actor, but could not correctly reenact the passive. MacWhinney does not spell the following out, but it seems that without the perspective cue, the second group perhaps parsed the sentence less fully (or even more “shallowly”) than the first group did.
a perspective is established, it will be maintained until it is overtly canceled by the establishment of a new perspective.
MacWhinney (1982) notes that the first major area in which children’s per- spective maintenance has been explored is in the interpretation of relative clauses. He discusses data from a number of studies in which children have been asked to enact sen- tences which vary on both the placement of the relative clause with regard to the matrix clause (i.e., whether it modified the matrix subject or object), and the syntactic role of the relative pronoun within its clause (i.e., subject- or object-extracted). Given these two dichotomous possibilities, the range of possible sentence types includes the following:
(75) Relative clause types (S-/O-modifying, S-/O-extracted)
SS: The dog [thati ti chased the cat] kicked the horse
SO: The dog [(thati) the cat chased ti] kicked the horse
OO: The dog chased the cat [(thati) the horse kicked ti]
OS: The dog chased the cat [thati ti kicked the horse]
Given the perspective maintenance hypothesis, which claims that a perspective established in the matrix clause should be held through the relative, we can predict the relative levels of difficulty that children should have in the comprehension of each of these sentences.
(76) Predicted difficulty on comprehension (perspective maintenance assumed): SS>OO/OS >SO
Namely, the easiest sentences to comprehend should be SS utterances, which can be interpreted with no perspective shift at all (since the perspective of the subject, “the dog,” which is established in the matrix clause, can be maintained as the perspectival subject of the subject-modifying relative clause). Meanwhile, OO and OS sentences require one perspective shift; in the former case, this occurs when the NN sequence “cat–horse” is encountered, since the second N establishes a new perspective; in the
latter case, the shift occurs when the object noun “the cat” is followed by a relative pronoun. Finally, SO sentences require a double shift of perspective: once when the NN sequence “dog–cat” is encountered, and again when “chased” requires a return to the perspective of the dog (which is the semantic object of that verb). In short, in contrast with semantic scaffolding, which predicts a uniform benefit for certain configurations of thematic roles and syntactic positions, the perspective maintenance theory predicts an alignment interaction between extraction type and modifier position (Gibson et al., 2005).
MacWhinney reports that data from a number of enactment studies with chil- dren supports the predictions of the perspective maintenance hypothesis quite strongly34 (though many of them have only tested a subset of the predicted difficulty pairs; e.g., OO
> SO, Sheldon, 1974—but see MacWhinney, 1982, for many, many more).
Returning to the issue of adult language, similar effects—and specifically the object/subject asymmetry—are evident in adults’ processing of subject- and object- extracted relative clauses. In an eye-tracking study by Traxler et al. (2002), adult participants read sentences containing relative clauses which modified the matrix subject; the relatives differed in whether the relative pronoun was subject- or object-extracted. As predicted by the perspective maintenance account, subjects experienced consider- ably more difficulty on object-extracted relatives (i.e., SO, in MacWhinney’s terminol- ogy; e.g.,The banker that the lawyer irritated played tennis every Sunday) than subject- extracted relatives (SS;The banker that irritated the lawyer played tennis every Sunday), since the latter maintain the perspective established in the subject of the matrix clause.35
34
MacWhinney(1982) notes that the only exception to the predictions of the prediction maintenance hypothesis has been failure to find significant differences—that is, there is no data supporting significant reversal (e.g., SS>OS/OO). He hypothesizes that this effect may be a result of a ceiling effect.
35
Traxler et al.(2002) discuss a number of possible alternative mechanisms for why this pattern might obtain in adult language processing. Apart from MacWhinney’s perspective-shifting account, the authors also mention a number of memory-based accounts (theintegration costaccount, the similarity- based interference account) and syntax-based accounts (theparallel function account, the active filler
strategy). MacWhinney himself cites these accounts, as well as several others; for a list of these hy- potheses with relevant citations, seeMacWhinney(1982); for details, seeTraxler et al.(2002).
On the other hand, in contrast to the predictions made by the perspective maintenance theory,36
Gibson et al. (2005) conducted research with adults which systematically varied modifier position and extraction type and found that object-extracted RCs were more difficult (i.e., they were read more slowly) than subject-extracted RCs, but that object-modifying RCs were also more difficult than subject-modifying RCs. The authors concluded that there was no evidence in processing for the predicted interaction between extraction type and modifier position. They explained the observed effect with theinfor- mation flow hypothesis, that is, that “[b]ackground information (like that in restrictive RCs) is processed more quickly earlier in a sentence rather than later in a sentence” (p. 337). Thus, subject-modifying RCs are read more quickly than object-modifying RCs because the former both include background information and tend to occur ear- lier in the sentence.37 Furthermore, the authors hypothesize that there is a continuum relating “expectedness” of input words (or syntactic categories) to processing difficulty, such that less-expected input words (as determined by linguistic experience) will cause a higher processing load; object-modifying RCs are less usual and therefore less expected and harder to process than subject-modifying RCs. Gibson et al. suggest that this ten- dency to order old information (such as the background information in RCs) before new arises from the relative cognitive ease of starting with information that is already known; this resonates quite strongly with the hypothesis in Li and Thompson(1976) that the ordering of topic before comment relates to the serialization of the information to be communicated (see footnote above).