Experiment 2: Legal Nonwords

16 undergraduate students from National Taiwan University (mean age=19.81 years, SD=1.33 years; eight males and eight females) participated in experiment 2, online cross-modal priming task for words in isolation (using legal nonwords, 0 ISI, target presentation duration=1000 ms). All participants were native speakers of Mandarin Chinese, right-handed, and had normal or corrected-to-normal eyesight or hearing, and none dwelt abroad for more than one year before the age of seven. They were also screened for brain injury and auditory or speech impairment. All participants were paid NT$100 for participation.

Materials & Design

The materials and design were similar to those in pretest 4,⁷⁰ while the nonwords were changed to legal nonwords (the proportion of related/unrelated trials hence was balanced in experimental trials only) prepared in 3.6.2 and target presentation duration to 1000 ms. The experimental auditory stimuli were more accurately cut off to a zero crossing point at the ending of formants by using Praat v.5.0.46 (see 3.8 for further details).

Procedure

Participants sat in the same sound-attenuated room in front of the experimental computer and a button box. An introductory video was prepared before the experiment began, which was similar to the one used before, except that this time the example for pseudohomophone was changed to a legal nonword. Since in earlier experiments or

70 There were four lists created in the beginning. After the first three subjects were run for list 1a, however, I further modified the script and created another new four lists. The remaining 13 participants were run using these new lists. The major design was the same, that legal nonwords are used and target presentation duration is 1000 ms at prime offset. Minor differences (e.g., PreRelease settings in E-Prime objects, sound loading factor, or ITI settings (2000 ms or 3000 ms), not recorded in detail) may exist but were considered to have only minute influences, if any.

informal pretests some participants thought memory tests were the most important parts in the experiment or mentioned they developed a strategy to memorize the auditory primes during lexical decisions, it was further emphasized that reaction times to lexical decisions were the most important to the study and that they did not need to consciously memorize anything during lexical decisions because it was not hard to pass the memory tests. E-Prime measured the time from the appearance of target words to the time responses were made in milliseconds until the time limit for response passed, and recorded time stamps for necessary information such as target presentation time or target onset delays. An inter-trial interval was also added.⁷¹ The remaining procedure was similar to that in the previous experiment. The whole experiment lasted for roughly 35 minutes.

Apparatus

For experiment 2, scripts were compiled using E-Prime 2.0.8.22 and the experiment was run on a Intel Pentium 4 (1.70GHz) desktop computer (with 1 GB RAM, NVIDIA GeForce FX 5200 graphics card, and Intel ICH 4 sound card) with a PST serial response box, and a ViewSonic E773 CRT monitor. The experiment was configured as such that prime words were played while a fixation point appeared on the center of screen. At the offset of the auditory prime word a visual target appeared on the position of fixation point for 1000 ms or until a response was made. E-Prime measured the response time in milliseconds. Positive responses were always placed on the right since the participants were all right-handed. The sounds were played out through a closed-ear headphone. The participants sat in front of the computer and the response box at a distance of about 60 cm from the computer screen to their eyes.

71 An ITI of 2000 ms was generally adopted. As mentioned earlier, however, I modified list 1a after the first three participants finished the experiment and created new lists based on it. The new lists (which were run for the remaining 13 participants) all had their ITI set as 2000 ms, though it was not documented whether list 1a adopted a 2000 ms or 3000 ms ITI. However the influences seemed minute for the current purpose.

Results & Discussion

All participants’ lexical decision accuracy rates were above or equal to .82, and memory test accuracy rates were above .79, and no data were removed for this reason.

Incorrect responses, outliers (RTs more than 1000 ms) and data for visual targets whose presentation delayed for more than 30 ms were removed, which meant 12.50% of the experimental trials (Mean RT= 550.64 ms, SD=124.07 ms).

A four-way ANOVA was run on individual participants’ raw RT data with Lists as a random factor and Prime Group (highly literal group and highly metaphorical group), Target Type (literal target and metaphorical target), and Prime Relatedness (related and unrelated) as fixed factors. There was no significant main effect found for Prime Group [F(1, 885)=.06, p>.80] or Target Type [F(1, 885)=1.11, p>.29], but a main effect for Lists [F(3, 885)=12.39, p<.001]. There was a near-significant effect found for Prime Relatedness [F(1, 885)=2.72, p=.10]. There were no interactions.⁷² Further examination of the data revealed that the near-significant effect for Prime Relatedness may be due to the low power (.38) of the analysis (given relatively small number of participants), and due to the fact that although in the highly literal word group, literal targets in the related condition were faster responded to than metaphorical targets, in the highly metaphorical word group, there was virtually no difference between related and unrelated conditions, thus slightly weakening the effect.

Given the near-significant effect and the fact that the main purpose of study was to compare the same group of target words under different conditions (see 2.5.2 for details), further separate independent t-tests were done for RTs of each group of visual targets under the related and unrelated conditions. The results showed a significant difference between the related and unrelated conditions for highly literal words’ literal targets

72 Equal variances not assumed.

[t(212)=-2.04, p<.05]. There was no significant effect found for highly literal words’

metaphorical targets [t(219)=-0.20, p=.84], for highly metaphorical words’ literal targets [t(222)=-0.54, p>.58], or for highly metaphorical words’ metaphorical targets [t(235)=-0.49, p>.62]. The data are summarized in Table 14 and Figure 9:

Table 14. Mean RTs for Highly Literal/Metaphorical Words in Isolation (Legal

Literal Meaning Metaphorical Meaning

RT Priming RT Priming

RTs for Highly Literal/Metaphorical Words in Isolation

500

Figure 9. Mean RTs for Highly Literal/Metaphorical Words in Isolation (Legal Nonword).

When literal visual targets were presented for 1000 ms immediately after isolated words that had highly frequent literal senses in corpus, a significant priming effect was found. However, no effect was found for infrequent metaphorical senses of highly literal words, or for literal or metaphorical senses of highly metaphorical words. This indicated

*

that, based on the categorization of corpus sense frequency, highly literal words’ literal meanings may be regarded as salient meanings based on the current online priming measure, which corresponded to the results in the earlier word association task.

It was of interest whether frequent metaphorical meanings could still be primed when frequency and association consistently suggested that these meanings were salient, since words with highly frequent metaphorical meanings did not always obtain first associates related to such meanings. If they were primed, then based on the criterion proposed in the beginning of this chapter, these meanings could be defined as salient meanings in the study.

Further second-order analyses were done on nine groups of items selected from the previously defined highly metaphorical word groups. The experimental primes in these groups all had frequent metaphorical meanings in corpus as well as 70%⁷³ or more first associates related to their metaphorical meanings in the association task. A two-way ANOVA was run on the items’ raw RT data with Target Type (literal target and metaphorical target), and Prime Relatedness (related and unrelated) as fixed factors.

There was no significant main effect found for Prime Relatedness [F(1, 254)=.006, p=.94] or Target Type [F(1, 254)=.003, p>.95]. There was no interaction effect. Hence, even when the frequency measure and the association measure consistently suggested the words’ metaphorical meanings were salient, these meanings were not primed.

Although the results could also be due to the limited number of items and participants, in all, there was not adequate support for the current study to define these metaphorical meanings as salient, based on the previously established criterion.

73 Due to the current criteria for frequency threshold (65%) and concern for the limited number of items, this threshold was considered appropriate for the association measure.