CHAPTER 2 LITERATURE REVIEW
2.5 M ETHODOLOGICAL & M INOR I SSUES
2.5.1 Review of Ahrens (2006)
Ahrens (2006) reviewed previous studies that supported the modular or interactive accounts of lexical ambiguity resolution, and found that studies supporting the modular view tended to adopt a target presentation duration less than 1000 ms, while studies supporting the interactive view usually adopted a target presentation duration more than 1500 ms. She then suggested that long target presentation duration only allowed researchers to examine effects after the immediate access of meanings, and thus facilitation was only observed for contextually appropriate meanings. She examined meaning activation in unbalanced ambiguous nouns in a cross-modal lexical priming task. The primary/secondary meanings were determined based on results in meaning generation tasks, and the sentences were biased to the nouns’ primary meanings. Her results showed that, when visual targets (either related to the primary meanings or the secondary meanings) were presented for 300 ms at offset of ambiguous words, both meanings were facilitated. Results were similar when visual targets were presented for
750 ms. However, when visual target words were presented for 1500 ms, only the contextually appropriate meanings were facilitated. The results supported her hypothesis and showed that there was an immediate access stage where all meanings were activated, and after a short while only contextually appropriate meanings remained.
Closely examining Ahrens’ (2006) work however reveals four important problems, which need to be dealt with in a cross-modal lexical priming study and will be discussed in following sections:
Firstly, this study did not distinguish different types of meanings based on clear criteria, as discussed in 2.2.
Secondly, this study did not pay adequate attention to polysemy issues in the selection of visual target words (see later discussion for criteria in the study). Since polysemy is a permeating phenomenon in language, visual targets gathered from production tasks such as meaning generation tasks, like ambiguous primes, are likely to be polysemous too. The primary issue that follows is that meaning frequencies for polysemous meanings in visual target words should be considered, or else it is likely that the effects will be confounded with other meanings of the visual target words. Also, and more importantly, visual target words related to the auditory primes tend to have a similar direction of semantic extension or semantic contents. Hence it is very likely that the two visual targets words (assigned to the two meanings of ambiguous primes) actually overlap in their semantic contents, and the secondary-meaning target in fact relates to the primary meaning, thus leading to an inaccurate observation. For example, Ahrens’ (2006) visual targets for the two meanings of the ambiguous prime word 湯
tang1 ‘soup
1; hot spring2’ were 飲料 yin3 liao4 ‘drink’ and 熱水 re4 shui3 ‘hot water’ respectively. However, most soup is a drink composed of hot water (and other stuff), and therefore facilitation of the second meaning target ‘hot water’ may bepartially attributed to first meaning activation. Similarly, the target words for the prime 茅房 mao2 fang2 ‘lavatory1; straw room2’, namely 廁所 ce4 suo3 ‘lavatory’ and 處 所 chu4 suo3 ‘location’, were also clearly related. Therefore, her report for secondary meaning activation was confounded due to the fact that her secondary meaning targets were related to the primary meanings. The semantic overlappings in polysemous primes and target words hence rendered her results less convincing (see discussions below; this study attempts to eliminate this by using monosemous targets).
Thirdly, repeated characters in ambiguous prime words and visual target words also led to a potential bias. Two out of the 16 experimental items had this problem. For example, the prime 河北 he2 bei3 ‘Hebei Province (in China)’ was paired with the secondary meaning target 北邊 bei3 bian1 ‘north’. This may bias the results because the priming effects obtained for secondary meanings again may be attributed to phonological and semantic (and even orthographic) priming caused by the same character, since Chinese characters may carry meanings by themselves.
Fourthly, lexical meanings and phrasal meanings were not distinguished. The secondary meanings may be used idiomatically in a phrase and the meanings are not totally encoded in the specific lexical form (Tyler & Evans, 2001). For example, in the case of 架子 jia4 zi5 ‘shelf1;arrogant manner2’, the secondary meaning (idiomatically) co-occurs with 擺 bai3 ‘set’, as it does in the sentential stimuli. Similarly, in the case of 飯碗 fan4 wan3 ‘bowl1; job2’, the secondary meaning idiomatically co-occurs with 砸 za2 ‘break’, as in the sentential stimuli. Placing prime words in such idiomatic expressions however may make activation of secondary meanings easier than otherwise, thus biasing the results.
The problems above pertained to the selection of primes and targets. In what follows, I continue to present crucial issues that are not problematic but should be considered in a cross-modal lexical priming experiment, and will be specifically discussed in following sections:
(1) In terms of design of the control condition, Ahrens’s (2006) study adopted a matched-targets design, whose control condition was a set of visual target words unrelated to the primes and matched with the experimental related primes for reaction times out of context. This design differs from other types of design, e.g., switched-targets design and matched-primes design, and will be discussed in a later section.
(2) In terms of the methods for collecting visual target words and accompanying timing issues, different prime-target relationships (due to different collecting methods) may affect the timing of priming effects. Ahrens (2006) used frequently occurring words in meaning generation tasks as visual targets, which was similar to a type of association tasks. And whether associatively-related targets can be distinguished from purely semantically-related targets constitutes an important issue in literature, partly because this affects the timing of priming effects. This will be discussed later.
(3) Target presentation duration and inter-stimulus interval (ISI), the timing variables that can be manipulated, may also affect the observed results, as argued in Ahrens (2006).
(4) Types of nonwords can also affect lexical decision results. Ahrens (2006) used legal nonwords, while pseudohomophones are also used in literature (see discussions below).
(5) Lexical category is an issue in literature too. Ahrens’ (2006) study focused on nouns, while other studies may employ adjectives (Williams, 1992), etc. Although this is not the primary focus in the study, the issue will be discussed in a later section.
(6) The creation of sentential stimuli is important. Ahrens (2006) claimed that her stimuli correctly biased the intended meanings based on sentence completion tasks, which will be discussed in a following section.
(7) The design of memory tests should be noted. Ahrens (2006) conducted memory tests after the whole lexical decision experiment (with a 0.8 accuracy threshold), while memory tests can also be done during the experiment. The interval between memory tests and the accuracy threshold can affect the results (see discussions below).
(8) Finally, a technical issue worth considering is how auditory stimuli are actually processed, which may affect participants’ perception.
The above issues will be considered for the current study in the following sections:32