We computed the sensitivity measure d' as the difference between the z-scores for the probability of hits (for words) and for false alarms (for nonwords), and the bias as c = (z(p(hits))+z(p(FA))/2. Thus, when bias is a negative number, this indicates a bias to respond 'yes', and where bias is a positive number, this indicates a bias to respond 'no'. We used correction computations for probability values of 1 (was changed to 1/(2N)) and 0 (was changed to 1-1/(2N)) based on the suggestions of Macmillan and Creelman (1990) [26
]. These data were analyzed with native language (Arabic, English, or Hebrew) and presentation mode (unilateral or bilateral) as between groups factors, and visual field (LVF or RVF), as a within-subject factor. The cell means for both sensitivity and response bias are illustrated in Figure .
Figure 1 Interhemispheric interaction for all stimuli during word recognition in Arabic, Hebrew, and English. Panel A: Sensitivity scores (d') in the bilateral and unilateral conditions for native readers of Arabic, Hebrew, and English. * indicate that d' is significantly (more ...)
It can be seen that all of the particpants evinced the expected RVF advantage, reflecting LH specialization for the task. The analysis of sensitivity scores (d') revealed a significant interaction between presentation mode and visual field, F(1,111) = 4.19, ηp2 = .04, p < .05, and a significant interaction between native language and presentation mode, F(2,111) = 4.12, ηp2 = .07, p < .05. In addition, the main effects of each of the three factors were significant (visual field, F(1,111) = 101.47, ηp2 = .48, p < .001; presentation mode, F(1,111) = 32.33, ηp2 = .23, p < .001; and language, F(2,111) = 15.3, ηp2 = .22, p < .001. Planned comparisons revealed that for both Hebrew and English speakers, although the effect of presentation mode on LVF performance was very dramatic, (Hebrew speakers: F(1,39) = 40.25, ηp2 = .51, p < .0001; English speakers: F(1,39) = 14.52, ηp2 = .28, p < .001), it was also significant in the RVF (Hebrew speakers: F(1,39) = 8.70, ηp2 = .19, p < .0001; English speakers: F(1,39) = 4.3, ηp2 = .10, p < .05). For Arabic speakers, presentation mode significantly affected LVF performance, F(1,36) = 5.95, ηp2 = .15, p < .05, but not RVF performance, p > .8. One-sample t-tests were done on the d' scores to test if performance was signficantly better than chance. It can be seen in the Figure that all conditions except the LVF bilateral condition for Arabic speakers resulted in sensitivity that is greater than chance.
The analysis of response bias revealed a significant 3-way interaction between language, presentation mode, and visual field, F(2,111) = 4.15, ηp2 = .07, p < .05, a two-way interaction between language and visual field, F(2,111) = 23.95, ηp2 = .30, p < .0001; a two-way interaction between presentation mode and visual field, F(1,111) = 4.33, ηp2 = .03, p < .05; and two main effects: of visual field, F(1,111) = 12.21, ηp2 = .10, p < .0001; and language, F(2,111) = 5.62, ηp2 = .09, 12 p < .05. Planned comparisons revealed that presentation mode had no effect in either visual field for English speakers, with bias not significantly different from 0 in all conditions, except in the RVF in the bilateral condition, where the effect is marginal (p = .08). For Hebrew speakers, bias was not affected by presentation mode in the LVF, but it was in the RVF, F(1,39) = 11.83, ηp2 = .24, p < .001. It can be seen in the Figure that the bilateral presentation mode resulted in a slight and identical positive bias (to say 'yes') in both visual fields, wheres in the unilateral condition, there is a slight negative bias in both visual fields, which is significant in the RVF but not in the LVF. For Arabic speakers, we see a third pattern: response bias is not affected by presentation mode in the RVF (where it is significantly positive), but it is affected significantly in the LVF, F(1,36) = 5.52, ηp2 = .13, p < .05, where it is significantly negative in the bilateral condition.