R ESEARCH D ESIGN AND M ETHODOLOGY
3.5 Phase One: Pilot Study
3.5.2 Data Analysis
included the following: each trial began with a forward mask consisting of ten hash marks (##########) which appeared on the screen for 500 ms. It was immediately replaced by a prime word, appearing for a duration of 50 ms, which was followed by a target word (or nonword) and it remained on the screen until the participants responded or for a maximum of 2,000 ms. The presentation order of the word pairs was randomized for each participant. The participant initiated the trial by pressing the space bar key. Their task was to respond as quickly and as accurately as possible to a target word, by pressing the following keys on the keyboard: the “Y” key with their right index finger for a real word and “N” key with their left index finger for a nonword. Before the experiment proper, each participant went through a trial with fifteen numbers of items which resembled the main task. The reaction time data and the error data were measured and saved after each experiment and further analyzed.
= 20.078, p = .000]. The main effect of Bilingual Group did not approach significance [F< 1]. However, the interaction between Prime Type and Bilingual Group approached significance [F(2,852) = 6.724, p = .001]. The mean reaction times and percentage of errors are presented in Table 3.3.
A series of planned comparisons were conducted to investigate the effects of Bilingual Group on reaction times. The Early High Proficient group had a very significant priming effect of 140 ms. The Late High Proficient and Late Low Proficient groups also had priming effects of 40 ms and 26 ms respectively. The pattern of results shows that the AoA may be a critical factor in the conflicting results. Further analysis showed that nonwords were reacted slowly than words.
Table 3.3 Mean RTs (ms) and Percentage of Errors in the two priming conditions for all three Bilingual Groups in Experiment 1
Overall Early Late High Late Low
Priming Condition RT (Error %) RT (Error %) RT (Error %) RT (Error %)
Translation 394 (4.5) 303 (7.6) 381 (9.2) 498 (6.1)
Control 463 (4.5) 443 (7.7) 421 (9.1) 524 (6.2)
Priming 69 140 40 26
The analysis conducted on the error data did not reveal a main effect of Prime Type [F< 1]. Percentage of error was similar when targets were preceded by translation primes and also when they were primed by control primes. However, the main effect of Bilingual Group tended towards significance in the error analysis [F(2,36)
=3.027, p = .032]. Errors were numerous for the Late High Proficient bilinguals than the Early High Proficient and Late Low Proficient bilinguals. A robust translation priming effect from L1−L2 is evident from the results of Experiment 1, which is consistent with earlier bilingual word recognition studies, (e.g., Gollan, et al., 1997;
Jiang, 1999; Jiang & Forster, 2001; Kim & Davis, 2003). In the next experiment, we tried to explore as to what happens when the direction of priming is from L2−L1, with only the languages of primes and targets reversed.
3.5.2.2 Results of Unmasked Translation Priming from L2−L1 (Experiment 2)
As in the previous experiment, separate analyses were run on the reaction time data and the error data. The mean reaction times and percentage of errors are presented in Table 3.4. Analyses conducted on the reaction time data revealed that the main effect of Prime Type did was marginal [F(1,113) = 2.231, p = .018]. This finding suggests that although Bodo−Assamese bilinguals were faster to respond to translation word pairs (324 ms) than to control ones (346 ms), the priming effect of 22 ms was much less compared to the effects observed in Experiment 1. Further investigations also revealed that the main effect of Bilingual Group was not significant [F(2,36) = 2.413, p = .213]. Consistent with our expectations, the Early High Proficient group produced robust priming effects of 79 ms. However, the Late bilinguals showed no translation priming whatsoever (−6 ms and −7 ms). The Late High Proficient group showed no facilitation, despite their fluency. Therefore, they provide empirical support that processing of second language words depends on the age of acquisition of L2, when proficiency is matched. This also proves that not only AoA and proficiency, but language direction also plays an important role in determining the priming effect.
Table 3.4 Mean RTs (ms) and Percentage of Errors in the two priming conditions for all three Bilingual Groups in Experiment 2
Overall Early Late High Late Low
Priming Condition RT (Error %) RT (Error %) RT (Error %) RT (Error %)
Translation 324 (1.2) 283 (2.4) 365 (2.0) 325 (1.4)
Control 346 (1.1) 362 (2.1) 359 (1.9) 318 (1.3)
Priming 22 79 −6 −7
In Experiment 2, we observed that when the priming direction was from L2−L1, translation priming effect was evident but relatively smaller compared to the
priming effect observed in Experiment 1. However, the presence of priming effect in the L2−L1 direction appears to contradict a number of previous studies in which significant translation priming effects in the L1−L2 direction have been observed but null translation priming effects have been evident when the priming direction was from L2−L1 (e.g., De Groot & Nas, 1991, Experiment 4; Grainger & Frenck- Mestre, 1998, Experiment 1; Kim & Davis, 2003, Experiment 1; Voga & Grainger, 2007, Experiment 2).
The differences in the time of processing in Experiment 1 and 2 were compared to evaluate the asymmetry in bilingual processing. The analysis revealed a significant main effect of Target Language [F(1,1827) = 39.263, p = .000] which indicates that the participants were faster when responding to Bodo target words (335 ms) than when responding to Assamese target words (428 ms). The numerical difference in the reaction time between the two directions was 47 ms which confirms that this difference was significant, thereby supporting the asymmetry in translation priming.
The interaction between Prime Type and Target Language was reliable [F(1,1222) = 3.673, p = .056]. Moreover, the Bilingual Group and Target Language interaction was reliable [F(2,619) = 3.469, p = .032].
3.5.2.3 Results of Masked Translation Priming from L1−L2 (Experiment 3)
Calculations of mean reaction times for correct responses were done using identical procedure as that used in the first experiment. The mean reaction times and percentage of errors are presented in Table 3.5. Analysis on the reaction time data did not reveal a main effect of Prime Type [F < 1). A very insignificant priming effect of 5 ms was observed. Further analysis showed that nonword targets were reacted significantly more slowly than word targets. Comparisons were conducted to determine if there was a significant effect of Bilingual Group. The results showed that the main effect of Bilingual Group did not approach significance [F < 1]. This
pattern was different than the pattern observed in Experiment 1 which followed the unmasked paradigm in the L1−L2 direction. In this experiment, the Early High Proficient group showed null effects (–29 ms). This null effect might indicate that the L1 primes were not able to facilitate the L2 targets because of the mask which means that the processing of words not only depends on AoA, proficiency and language direction, but also on different paradigms. Another interesting observation in this experiment was that contrary to the Early High Proficient group, the Late High and Low Proficient groups were able to produce priming effects of +11 ms and +34 ms respectively. What was noteworthy here was that the Late Low Proficient group had larger priming effect compared to the others which is indicative of using different processing mechanisms.
Table 3.5 Mean RTs (ms) and Percentage of Errors in the priming conditions for all three Bilingual Groups in Experiment 3
Overall Early Late High Late Low
Priming Condition RT (Error %) RT (Error %) RT (Error %) RT (Error %)
Translation 945 (4.6) 906 (7.6) 906 (9.0) 1023 (6.8)
Control 950 (4.6) 877 (7.6) 917 (9.0) 1057 6.8)
Priming 5 −29 11 34
3.5.2.4 Results of Masked Translation Priming from L2−L1 (Experiment 4)
Analysis on the reaction time data did not reveal a main effect of Prime Type [F <
1]. Translation pairs (701 ms) and control pairs (708 ms) were responded to almost similarly, showing a very insignificant priming effect of 7 ms. Moreover, mean reaction times of nonwords were not significantly slower than words. Planned comparisons showed no significant effect of Bilingual Group [F < 1]. Contrary to Experiment 3, the Early High Proficient group did produce an effect 13 ms. The
Late High Proficient group and Late Low Proficient group also produced priming effects of 8 ms and 2 ms respectively, but it was not significant. Contrary to some of the previous findings, the present experiment did produce a priming effect in the L2−L1, although it was lesser in magnitude and not significant. The mean reaction times and percentage of errors are presented in Table 3.6.
Table 3.6 Mean RTs (ms) and Percentage of Errors in the two priming conditions for all five Bilingual Groups in Experiment 4
Overall Early Late High Late Low
Priming Condition RT (Error %) RT (Error %) RT (Error %) RT (Error %)
Translation 701 659 699 743
Control 708 672 707 745
Priming 7 13 8 2
We conducted a combined analysis of Experiment 3 and 4 and the results revealed a significant main effect of Target Language [F(1,1827) = 85.845, p = .000]. This suggests that the participants were significantly faster to Bodo targets than to Assamese targets. A significant interaction was also observed between Target Language and Bilingual Group [F(2,1256) = 4.744, p = .009], indicating the role of second language age of acquisition and proficiency. This finding of our study is consistent with the assumptions of the RHM that mapping of L2 words into the lexical and conceptual representations is dependent on the proficiency of the bilinguals. In other words, high proficient bilinguals can directly access the conceptual representation, whereas, low proficient bilinguals access the conceptual representation indirectly via the L1 lexical representation.