The results from the current cross-modal lexical decision task could clarify the
controversial issue of lexical ambiguity resolution (modular/interactive hypothesis) as
well as the internal organization of mental lexicon (single/separate representation) for
polysemy. First of all, it was found that homonymy and polysemy behaved differently
on meaning activation during the sentence resolution of lexical ambiguity. Only the
contextually appropriate meaning was activated in the case of homonymous words
(words with unrelated meanings), whereas in the case of polysemous words (words
with interrelated senses), both contextually relevant and irrelevant meanings were
activated. In other words, while the resolution of homonymous meanings in sentence
processing depended on the contextual information, polysemous meanings were
resolved and processed regardless of the context effect. These results suggest that two
types of ambiguity are resolved differently in sentences and that the distinction
between different types of ambiguity (in terms of meaning relatedness) is of
importance in the issue of lexical ambiguity resolution. As we have discussed in
Chapter 2, previous research has not come to an agreement regarding whether or not
128
the context has an immediate effect on the resolution of lexical ambiguity. When
examining the ambiguity during sentence comprehension, most of the previous
studies in fact centered on only homonyms, or they used polysemous words
interchangeably with homographs or homophones in their experiments. However,
finding that homonymy and polysemy have different patterns during sentence
resolution, our study proposes that the distinction between different types of lexical
ambiguity which different on their meaning relatedness should not be ignored and the
effect of sense relatedness of ambiguity would be a potential factor to influence the
experimental results in lexical ambiguity resolution. To further find out how previous
research, which did not make a distinction between homonymy and polysemy, is
different from ours, a more specific comparison was made. We compared our study
with some of the previous studies on lexical ambiguity resolution whose experimental
designs were similar to the present study, i.e., by using cross-modal lexical decision
tasks with the settings adjusted to detect the immediate stage of processing. These
selected studies consisted of Swinney (1979), Onifer & Swinney (1981), and Ahrens
(1998, 2001, 2002, 2006), in all of which the results supported the modularity
(context-independent) view or the multiple access model (see Table 6.1).
Table 6.1. Comparison of the previous and present cross-modal lexical decision tasks
(English) Not distinguished (H: 58%, P:42%)
As shown in Table 6.1, first of all, the major difference between previous studies
and the current research is the distinction between homonymous words and
polysemous words. For example, we collected the available materials from Onifer and
Swinney (1981) and Ahrens (2001, 2006) and calculated their distribution of different
types of ambiguity, for which the homonymy/polysemy distinction was depended on
whether two meanings of ambiguity were listed as different entries or the same entry
in the dictionary (based on The American Heritage Dictionary of the English
130
distribution of homonymy/polysemy in these studies was not consistent and not
specifically controlled; for instance, Onifer and Swinney (1981) included both
homophones such as flower/flour and polysemous words such as cold (which means
“low temperature” or “flu”) in their experiments but did not make a distinction, and
Ahrens (2001) also included both homonymous word such as bei4shu1 (背書) (which
means “to memorize” or “to endorse”) and polysemous words with metaphorical
extension such as zou3hou4men2 (走後門) (which means “to take illegal advantage
of” or “to get in by the back door”) in her experiment. Different from the results the
above mentioned studies have found, our study here (which distinguishes between
different types of ambiguity) does not provide strong evidence for the
context-independent view (multiple access model), as the resolution of homonymous
words is found sensitive to the context. Therefore, our study suggests that meaning
relatedness of lexical ambiguity would be a crucial factor to influence the results on
lexical ambiguity resolution.
Second, the effect of meaning dominance (frequency distribution) was not
examined in Swinney (1979) and Ahrens (1998, 2001, 2002, 2006), in which
equi-biased ambiguity (i.e., two meanings of ambiguity were equally frequent) was
used. In Onifer & Swinney (1981), on the other hand, unbalanced ambiguity (i.e., two
meanings of ambiguity were unequally frequent) was employed and their results
found multiple access of meanings and thus argued against frequency effect, for both
the dominant and subordinate meaning were activated. If there was a strong effect of
meaning frequency, on the other hand, the less dominant meaning should not be
activated as the more dominant meaning was. Therefore, they suggested that
differences on meaning dominance did not influence the activation of multiple
meanings on lexical ambiguity resolution. However, in the present study, unbalanced
ambiguity was also used and the results found a potential frequency effect for
homonymy, in which only the dominant meaning (also the contextually appropriate
meaning) was activated, but not for the case of polysemy, in which both the dominant
and subordinate meanings were accessed. In other words, unbalance on meaning
distribution of ambiguity only had influence on homonymous meanings but not on
polysemous meanings. The different patterns again not only lay emphasis on the
difference between homonymy and polysemy but also indicate the influence of sense
relatedness. For homonymy, its unrelated meanings might be separately and distinctly
represented, and when there are two incompatible meanings to be resolved and
processed in context, meaning frequency may come into play to help pick out one
meaning for further processing. Therefore, the more dominant one would have greater
possibility to be accessed, and the less dominant meaning may not be able to
overcome the effect of meaning dominance and hence would be suppressed. For
132
polysemy, on the other hand, multiple meanings are interrelated, and there may be no
immediate need for the meaning frequency effect to act in order to choose one of
those compatible meanings. Consequently, both the more dominant meaning and the
less dominant meaning of polysemy could be initially accessed. Although so far the
experiment in the present study cannot make a clear distinction as to whether the
context itself or both context and frequency have effects on the first meaning
activation of homonymy, since its contextually appropriate meaning is also the more
dominant meaning in our experiment, the current research however does illustrate that
homonymy and polysemy have different processing patterns on lexical access and
therefore points out the important effect of meaning relatedness.
Overall, concerning the controversial issue in lexical ambiguity resolution, our
study suggests that whether the context-independent model or the context-dependent
model holds right should depend on which type of ambiguity is considered. We would
take the position that absolute modularity and absolute interaction in lexical
processing can be considered as the two extremes along a continuum, which is close
to the view taken by Tanenhaus, Dell, and Carlson (1987) reconciling modular and
interactive models (see also Simpson & Krueger, 1991; Tabossi, 1988). Different
factors, such as meaning relatedness of lexical ambiguity, may influence the degree of
activation along this continuum. For the issue of lexical ambiguity resolution, as we
have outlined in Chapter 2, two major effects—context and meaning frequency—have
been pointed out in previous research and four models in relation to these two factors
have been proposed. For ease of reference, we reduplicate Table 2.3 in Chapter 2 as
Table 6.2.
Table 6.2. Different types of models of lexical ambiguity resolution (reduplicated) Processing
hypotheses
Language processing
models Types of models
Context effect
(at an early stage)consideration when dealing with the issue of lexical ambiguity resolution— the sense
relatedness of ambiguity itself. Therefore, as Figure 6.1 shows, the two factors,
context and frequency, in relation to sense relatedness, would influence the degree of
meaning activation in lexical ambiguity resolution. On the one hand, whether the
context effect occurs or not is relevant to the effect of sense relatedness, since it was
found that ambiguous meanings with low relatedness showed context effect in
sentence processing but not for ambiguous meanings with high relatedness. On the
134
other hand, whether the frequency effect occurs or not is also relevant to the effect of
sense relatedness, for meaning dominance only had influence on ambiguous words
with unrelated meanings, but not on words with closely related meanings. Therefore,
it is suggested that the examination of the context and frequency effect in lexical
ambiguity resolution should not be separate from the consideration of the sense
relatedness effect.