Chapter 4 Experiment Two: Lexical ambiguity resolution with syntactic category
4.1.3 Norming studies of ambiguous words and sentential contexts
4.1.3.7 Norming study 7: Sentence plausibility
國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
4.1.3.6 Norming study 6: Disambiguation
The aim of disambiguation rating was to ensure that the disambiguating word we used in the succeeding context was semantically related to the subordinate meaning of the homograph and can indeed help readers disambiguate. For the ambiguous targets, as soon as the participants chose one definition based on the succeeding context, both the disambiguating word and the homograph with the definition of its subordinate meaning were parallel displayed. For the unambiguous targets, as soon as the participants judge the syntactic category of the target word, the succeeding context was presented followed by a parallel display of the target and the disambiguating word. The participants were asked to rate the semantic relatedness between the disambiguating word and the target word (either the homograph with its subordinate meaning or the unambiguous word) on a 7-point scale (where 1 = not related; 7 = much related). The two-way analysis of variance (ANOVA) with word type and
ambiguity as independent variables was conducted for strength of disambiguation.
The main effects and the interaction were non-significant (Fs < 3, ps > .10), indicating there were no significant differences among word types and between homographs and unambiguous words.
4.1.3.7 Norming study 7: Sentence plausibility
The procedure and the criteria were identical to Norming study 6 in Experiment
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
1. The two-way analysis of variance (ANOVA) with word type and ambiguity as independent variables was conducted for sentence plausibility. There were no significant differences among word types, F(3, 152) = .06, p = .98. However, the plausibility for homograph-embedded sentences was significantly lower than that for unambiguous word-embedded sentences (F(1, 152) = 4.90, p = .03). The interaction was non-significant (p > .08).
4.1.4 Apparatus
Apparatus was identical to Experiment 1.
4.1.5 Procedure
Procedure was identical to Experiment 1.
4.1.6 Data analysis
In Experiment 2, three regions of interest (ROIs) were subject to analyses of eye movements: target word (the homograph or unambiguous word), post-target word (the two-character word following the target word), and disambiguating word (the two-character word providing disambiguating information in the succeeding context).
The eye-movement measures we analyzed and the criteria we used to eliminate data
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
were identical to Experiment 1. Data of 2 participants were excluded from the analyses because they had first-pass blinks in the target region on more than a quarter of all trials; data of another 3 participants were also excluded from the analyses because most of fixations fell out of the range of the sentences. These data were replaced with new qualified data. Overall, for the data of first-pass reading, the removed trials accounted for 3.41% (target), 2.84% (post-target), and 2.56%
(disambiguating). On the other hand, for the data of second-pass reading, the removed trials accounted for 5.66% (target), 4.97% (post-target), and 4.22% (disambiguating).
The average accuracy of comprehension test was 96.33%. The statistical analyses were identical to Experiment 1, except that the effect of word frequency in each ROI was partialled out since we had not controlled the word frequency in the post-target and the disambiguating region.
4.2 Results
4.2.1 Target words 4.2.1.1 Duration measures
Means and standard errors of both the first-pass and second-pass duration measures for each condition on target words are shown in Table 17. The statistic results of all analyzed effects for the duration measures on target words are listed in
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
Table 18. For the first-pass duration measures, both the main effect of syntactic category and that of ambiguity were non-significant in any measure. However, the ambiguity-by-dominant syntactic category interaction was marginally significant in FFD (b = .02, SE = .01, t = 1.81), indicating larger ambiguity effect for dominant verb
meaning. FFD and SFD revealed significant ambiguity effect when the homographs’
dominant meaning was a verb (FFD: b = .04, SE = .02, t = 2.23; SFD: b = .04, SE
= .02, t = 1.90). FFD, SFD, and GD revealed significant ambiguity effect for VN (ts >
2).
For the second-pass duration measures, the main effects of syntactic category were non-significant in any measure. However, RRT revealed significant subordinate-by-dominant syntactic category interaction (b = - .1, SE = .05, t = -2.12).
Verbs obtained less rereading times than nouns when the dominant syntactic category is verb (i.e., VV < VN, b = - .16, SE = .07, t = -2.22). In addition, in terms of homographs’ dominant syntactic category, verbs also obtained less RRT and TVT than nouns when the target word was a verb (i.e., VV < NV, ts < -1.8). On the other hand, the main effect of ambiguity was significant in RRT and TVT (ts > 2), especially when the subordinate meaning was a noun (i.e., NN & VN, ts > 2). RRT also revealed significant ambiguity effect when the homographs’ dominant meaning was a noun (b
= .13, SE = .07, t = 2.02) and marginal ambiguity effect for NN (b = .17, SE = .10, t =
‧
1.80). GPT and TVT revealed significant ambiguity effect when the homographs’
dominant meaning was a verb (GPT: b = .06, SE = .03, t = 1.86; TVT: b = .14, SE
= .05, t = 2.67), especially for VN (ts > 2).
Table 17
Experiment 2: Means and standard errors of the duration measures for all conditions on target words
Duration measures (ms) Type
Ambiguity Effect size
A UA A－UA Note. A = ambiguous words; UA = unambiguous words
‧
Experiment 2: Statistic results of all effects for the duration measures on target words
Effects (t value)
Means and standard errors of both the first-pass and second-pass probability measures for each condition on target words are shown in Table 19. The statistic results of all analyzed effects for the probability measures on target words are listed in
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
Table 20. For the first-pass probability measures, SKIP revealed that verbs were less skipped than nouns in terms of homographs’ dominant syntactic category, especially when the target word was a verb (i.e., VV < NV, b = - .35, SE = .16, p = .03). RFR showed no effect.
For the second-pass probability measures, the main effects of syntactic category were non-significant in any measure. However, ROR revealed significant subordinate-by-dominant syntactic category interaction (b = .45, SE = .20, p = .03).
Verbs obtained less probability of regression-out than nouns when the dominant syntactic category was noun (i.e., NV < NN, b = - .66, SE = .29, p = .02). In addition, in terms of homographs’ dominant syntactic category, verbs also obtained less probability of regression-out than nouns when the target word was a noun (i.e., VN <
NN, b = - .53, SE = .28, p = .06). RRR and RIR revealed that verbs obtained less probability of rereading and regression-in than nouns in terms of homographs’
dominant syntactic category, especially when the target word was a verb (i.e., VV <
NV, RRR: b = - .56, SE = .25, p = .03; RIR: b = - .47, SE = .25, p = .06). On the other hand, both RRR and RIR revealed significant main effect of ambiguity (ps < .001), significant ambiguity effect whenever the target word was a noun or a verb (ps < .05), significant ambiguity effect when the homographs’ dominant meaning was a verb (ps
< .01), and marginal ambiguity effect when the homographs’ dominant meaning was a
‧
noun (ps = .07). In addition, both RRR and RIR revealed significant ambiguity effect for VN (ps < .01); RIR also revealed significant ambiguity effect for VV (b = .75, SE
= .36, p = .03).
Table 19
Experiment 2: Means and standard errors of the probability measures for all conditions on target words
Probability measures (%) Type
Ambiguity Effect size
A UA A－UA Note. A = ambiguous words; UA = unambiguous words
‧
Experiment 2: Statistic results of all effects for the probability measures on target words
4.2.2 Post-target words 4.2.2.1 Duration measures
Means and standard errors of both the first-pass and second-pass duration measures for each condition on post-target words are shown in Table 21. The statistic
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
results of all analyzed effects for the duration measures on post-target words are listed in Table 22. For the first-pass duration measures, both the main effect of syntactic category and that of ambiguity were non-significant in any measure. FFD and SFD revealed a reverse syntactic category effect of homographs’ dominant meaning when the target word was a verb (ts < -1.8), indicating less reading time for the post-target words following VV compared to those following NV. The ambiguity-by-dominant syntactic category interaction was significant in FFD, SFD, and GD (FFD: b = .03, SE = .01, t = 2.21; SFD: b = .03, SE = .01, t = 2.04; GD: b = .03, SE = .02, t = 1.98).
In addition, the interaction effect was also marginally significant in FFD when the target word was a verb (b = .03, SE = .02, t = 1.85), indicating larger ambiguity effect for the post-target words following VV than those following NV. FFD, SFD, and GD revealed significant ambiguity effect when the homographs’ dominant meaning was a verb (ts > 2). In addition, FFD also revealed marginal ambiguity effect when the target word was a noun (b = .03, SE = .02, t = 1.8). The ambiguity effect for VN was significant in FFD, SFD, and GD (FFD: b = .06, SE = .03, t = 2. 12; SFD: b = .05, SE
= .03, t = 1.8; GD: b = .06, SE = .03, t = 1.89).
‧
Experiment 2: Means and standard errors of the duration measures for all conditions on post-target words
Duration measures (ms) Type
Ambiguity Effect size
A UA A－UA Note. A = ambiguous words; UA = unambiguous words
For the second-pass duration measures, RRT and TVT revealed a reverse main effect of homographs’ dominant syntactic category (ts ≦ -1.8), indicating that the
post-target words obtained more rereading and total viewing times when the syntactic
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
category of homographs’ dominant meaning was noun compared to when being verb.
GPT and RRT also revealed a reverse syntactic category effect of homographs’
dominant meaning when the target word was a verb (ts < -2), indicating more go-past and rereading times for the post-target words following NV than those following VV.
On the other hand, GPT, RRT, and TVT revealed significant main effect of ambiguity (ts > 2), significant ambiguity effect when the target word was a noun (ts > 2), significant ambiguity effect when the homographs’ dominant meaning was a verb (ts
> 2), and significant ambiguity effect for VN (ts > 2). In addition, TVT also revealed significant ambiguity effect when the target word was a verb (b = .08, SE = .03, t = 2.77), significant ambiguity effect when the homographs’ dominant meaning was a noun (b = .08, SE = .03, t = 2.81), and significant ambiguity effect for both NN and VV (ts > 3).
‧
Experiment 2: Statistic results of all effects for the duration measures on post-target words
Means and standard errors of both the first-pass and second-pass probability measures for each condition on post-target words are shown in Table 23. The statistic results of all analyzed effects for the probability measures on post-target words are
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
listed in Table 24. For the first-pass probability measures, SKIP showed no effect.
RFR revealed marginally significant main effect of ambiguity (b = .26, SE = .14, p
= .06), significant ambiguity effect when the target word was a noun (b = .38, SE
= .20, p = .05), and marginal ambiguity effect for NN (b = .47, SE = .26, p = .06). For the second-pass probability measures, RIR showed no effect. Both RRR and ROR revealed significant main effect of ambiguity (ps < .001), significant ambiguity effect whenever the target word was a noun or a verb and whenever the homographs’
dominant meaning was a noun or a verb (ps < .05). In addition, RRR showed significant ambiguity effect for the post-target words following all types of homographs (ps < .05); ROR showed significant ambiguity effect for the post-target words following VN and NV (ps < .05).
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
Table 23
Experiment 2: Means and standard errors of the probability measures for all conditions on post-target words
Probability measures (%) Type
Ambiguity Effect size
A UA A－UA
First-pass SKIP NN 24.48 (2.20) 23.38 (2.16) 1.10 VV 32.47 (2.38) 32.48 (2.37) -0.01 VN 25.84 (2.23) 30.28 (2.32) -4.44 NV 23.92 (2.15) 27.06 (2.26) -3.14 RFR NN 16.38 (2.19) 11.22 (1.84) 5.16 VV 13.51 (2.13) 11.03 (1.94) 2.48 VN 12.72 (1.98) 10.29 (1.85) 2.43 NV 14.14 (2.03) 12.72 (1.98) 1.42 Second-pass RRR NN 28.00 (2.32) 20.90 (2.09) 7.10 VV 23.68 (2.18) 18.09 (1.96) 5.59 VN 28.00 (2.32) 20.10 (2.04) 7.90 NV 27.49 (2.29) 22.61 (2.16) 4.88 RIR NN 12.00 (1.68) 9.26 (1.49) 2.74 VV 12.11 (1.68) 10.08 (1.53) 2.03 VN 11.47 (1.65) 12.37 (1.67) -0.90 NV 11.52 (1.64) 10.37 (1.57) 1.15 ROR NN 13.62 (2.06) 8.93 (1.68) 4.69 VV 13.62 (2.14) 11.03 (1.94) 2.59 VN 16.19 (2.21) 7.75 (1.63) 8.44 NV 17.01 (2.19) 11.11 (1.88) 5.90 Note. A = ambiguous words; UA = unambiguous words
‧
Experiment 2: Statistic results of all effects for the probability measures on post-target words
4.2.3 Disambiguating words