|
Article Excerpt
Research on the effect of lexical ambiguity on eye movements has yielded some interesting results with respect to how such words are processed during reading (see Duffy, Kambe, & Rayner, 2001, for a recent review). This research has made it clear that it is necessary to distinguish between balanced ambiguous words (words with two relatively equal meanings) and biased ambiguous words (words with a highly dominant meaning). Furthermore, it is clear that disambiguating information preceding the ambiguous word differentially affects processing of balanced and biased words (Binder & Morris, 1995; Dopkins, Morris, & Rayner, 1992; Duffy, Morris, & Rayner, 1988; Kambe, Rayner, & Duffy, 2001; Rayner & Frazier, 1989; Rayner, Pacht, & Duffy, 1994; Sereno, O'Donnell, & Rayner, 2006; Sereno, Pacht, & Rayner, 1992).
Most of this research has used nouns that have two meanings as the target word, and the general pattern of results differs when the target words are ambiguous with respect to noun-verb ambiguities (Frazier & Rayner, 1987) or verb-verb ambiguities (Pickering & Frisson, 2001). But, for noun targets, the following pattern of results has emerged. In a neutral context, readers look longer at a balanced ambiguous word than a control word matched on length, frequency and appropriateness in the sentence (Duffy et al., 1988; Rayner & Duffy, 1986). They also look longer at the balanced word than the biased word (Rayner & Frazier, 1989), and there is no difference between the biased ambiguous word and a control word again matched on length, frequency and sentence fit (Duffy et al., 1988). However, when they reach the disambiguating information (which appears after the ambiguous target word) and the subordinate meaning is instantiated, reading is disrupted with long fixations and regressions back to the ambiguous word (Duffy et al., 1988; Kambe et al., 2001).
When the disambiguating information precedes the ambiguous target word, thus making it clear to the reader which meaning of the word should be instantiated, the pattern is somewhat different. In the case of a balanced ambiguous word, there is no difference in how long readers look at the ambiguous word and its matched control word (Duffy et al., 1988; Rayner & Frazier, 1989). Apparently, the contextual information is sufficient for the reader to choose the appropriate meaning, and there is no slowdown in processing that is manifest in the eye movements. However, when the disambiguating information precedes a biased ambiguous target word and the subordinate meaning is instantiated, readers look longer at the ambiguous word than its matched control. This latter finding, which has now been replicated many times (Binder, 2003; Binder & Rayner, 1998; Duffy et al., 1988; Rayner et al., 1994), has been called the subordinate bias effect (Pacht & Rayner, 1993; Rayner et al., 1994).
The subordinate bias effect is theoretically important because it suggests that strong contextual information does not override automatic access processes. That is, although contextual information is consistent with the subordinate meaning, readers look longer at the ambiguous word. According to selective access accounts, appropriate contextual information should lead to the immediate access of the intended meaning. In actuality, there is some controversy concerning the nature of the subordinate bias effect (see Binder & Rayner, 1998; Martin, Vu, Kellas, & Metcalfe, 1999; Rayner, Binder, & Duffy, 1999; Vu & Kellas, 1999), and the extent to which strong preceding context can override the effect (see Wiley & Rayner, 2000). Furthermore, there are questions related to what the appropriate control word should be when the subordinate meaning is instantiated by the context (Sereno et al., 1992, 2006): should the control word be a word matched on overall frequency (as is typically the case) or one matched to the frequency of the subordinate meaning? Sereno et al. (2006) recently presented evidence consistent with the view that the most appropriate control is one matched on the frequency of the subordinate meaning. Interestingly, this methodological issue has important theoretical implications in that how one might model the results is influenced by the answer to the methodological question. Thus, Duffy et al. (2001) modelled the results via a constraint-satisfaction approach in which the subordinate bias effect was due to competition between the alternative meanings, while Reichle, Rayner, and Pollatsek (2003; Reichle, Pollatsek, & Rayner, in press) modelled the results in the context of the E-Z Reader model by assuming that the subordinate meaning was a low frequency word.
In the present article, we sidestep these questions regarding the basis of the subordinate bias effect in lexical ambiguity resolution and address a much simpler question: can an adjective or modifier that immediately precedes a biased ambiguous word influence the immediate interpretation of the intended meaning? Thus, if the ambiguous word table is preceded by the word kitchen (consistent with the dominant meaning) or statistical (consistent with the subordinate meaning), will readers immediately access the appropriate meaning? In our experiment, the issue of appropriate control word is not relevant because we used the ambiguous word as its own control (Rayner & Frazier, 1989). That is, we compared processing time on table when there was a modifier with when there was no modifier, and we instantiated either the dominant or the subordinate meaning of the ambiguous word. If the subordinate bias effect is real, not an artifact of the choice of control word, readers should fixate longer on the ambiguous word when disambiguating information precedes it.
It is also interesting to note that one might think that an immediately preceding adjective that specifies the appropriate meaning of a word might make the subordinate bias effect less evident. Thus, intuitively, it might seem that statistical table should be quite easy to process and not yield any type of increased fixation on table (even though the subordinate meaning is instantiated). Actually, the presence of the subordinate bias effect in prior research led us to suspect that fixation times would be longer on table when an adjective consistent with the subordinate meaning was present in comparison to when an adjective consistent with the dominant meaning was present. The present study is therefore another attempt to determine if strong contextual information (in the form of an adjective highly consistent with one or the other meanings of the ambiguous word) might lead to data more consistent with selective access accounts.
While it might seem evident that there should be an immediate effect from the modifier (given the widely accepted view that language processing is incremental), there are reasons to think that such an effect might not show up directly on the target word. First, in a conceptually similar study to ours (except that unambiguous target words were used so that alternative meanings being instantiated was not at issue), McConkie and Zola (1981) presented readers with a modifier word that was highly constraining of the following word, such as buttered popcorn where buttered should be highly constraining of popcorn in the context of going to the movies. McConkie and Zola (1981) reported that readers did not look at popcorn for significantly less time (1) when preceded by buttered than when preceded by the control word adequate. So, in this case, an immediately preceding constraining adjective apparently did not influence how long it took to read the target word. However, Ehrlich and Rayner (1981) argued that for contextual constraint to have an effect, perhaps it has to build up over larger stretches of text than the immediately preceding word (see also Morris, 1994). Ehrlich and Rayner found clear evidence for the effect of contextual constraint when implemented in this fashion. Given McConkie and Zola's results, perhaps it is the case that contextually constraining variables need more build-up time and one would not find an immediate disambiguation effect in our study. If this were the case, we might expect to see an effect on the spillover fixation (the first fixation after leaving the target word) or even later in the sentence.
Second, a number of studies dealing with various types of ambiguity resolution and interpretation have obtained delayed effects in which the effect of...
|
