• Tidak ada hasil yang ditemukan

Masked Semantic Priming from L2−L1 (Experiment 2D)

W ORD R ECOGNITION

4.3 Visual Lexical Decision

4.3.2 Semantic Priming

4.3.2.5 Masked Semantic Priming from L2−L1 (Experiment 2D)

mean reaction times to nonwords showed that the participants took significantly longer to respond to nonwords (1057 ms) than to words (770 ms).

Table 4.21 Mean RTs (ms) and Percentage of Errors for Cognate and Non-Cognate Targets Primed by Semantically Related and by Unrelated Control Primes for All Three Bilingual Groups in Experiment 2C

Cognate Status/Bilingual Group

Cognate Non-cognate

Early Late high Late low Early Late high Late low Prime Type RT

(Error %)

RT (Error %)

RT (Error %)

RT (Error %)

RT (Error %)

RT (Error %)

Semantic 942

(15.8)

910 (17.1)

782 (21.1)

904 (12.8)

946 (15.3)

775 (22.7)

Control 878

(14.9)

849 (13.5)

695 (18.5)

831 (9.6)

852 (15.6)

666 (14.3)

Priming −64 −61 −87 −73 −94 −109

4.3.2.4.3 Discussion

The results of Experiment 2C showed that unlike Experiment 2A, L1 primes failed to facilitate L2 targets when a masked paradigm was used. The results of the study is consistent with the findings of Basnight-Brown & Altarriba (2007), who did not find significant semantic priming for semantically related words with a forward mask design.

4.3.2.5.1 Method

Participants.

The same participants, who took part in the previous semantic priming experiments, participated in this experiment as well.

Stimuli.

The eighty Assamese word targets of Experiment 2C and their respective Bodo translation primes were used again, but now, respectively, as Assamese (L2) translation primes and corresponding Bodo (L1) word targets.

Procedure.

The design and procedure of the present experiment were identical to those of Experiment 2C.

4.3.2.5.2 Results

Similar to the previous experiments, a mixed-effects analysis was run on the reaction time data and error data after removing outliers from the data. Table 4.22 shows the mean reaction times and percentage of errors as a function of Prime Type and Cognate Status. In the reaction time analysis, the main effect of Prime Type was observed to be significant [F(1,146) = 20.269, p = .000]. However, similar to Experiment 2C, semantically related words (752 ms) were responded to slowly as compared to unrelated control words (685 ms).

The main effect of Cognate Status approached significance [F(1,146) = 15.312, p = .000]. The main effect of Bilingual Group was not significant [F < 1]. However, the interaction between Prime Type and Bilingual Group approached significance [F(2,6757) = 24.633, p = .000]. The Cognate Status and Bilingual Group interaction was also significant [F(2,6757) = 19.858, p = .000]. Finally, a significant three-way interaction between Prime Type, Cognate Status, and Bilingual Group was also observed [F(2,6757) = 28.028, p = .000].

Table 4.22 Mean RTs (ms) and Percentage of Errors for Cognate and Non-Cognate Targets Primed by Semantically Related and by Unrelated Control Primes in Experiment 2D

Cognateness

Overall Cognate Non-cognate

Prime Type RT (Error %) RT (Error %) RT (Error %) Cognate effect

Semantic 752 (15.7) 785 (21.8) 719 (10.2) −66

Control 685 (9.8) 710 (12.1) 659 (7.7) −51

Priming −67 −75 −60

Planned comparisons performed on the individual priming effects for the three types of bilinguals revealed that the reaction time differences between all three Bilingual Groups was not significant (Early High Proficient = 733 ms, Late High Proficient = 736 ms, Late Low Proficient = 703 ms). Moreover, no cognate facilitation effect was observed with any of the bilingual groups, replicating the pattern of Experiment 2A, 2B, and 2C. The mean reaction times and percentage of errors as a function of Prime Type, Cognate Status, and Bilingual Group are presented in Table 4.23.

Table 4.23 Mean RTs (ms) and Percentage of Errors for Cognate and Non-Cognate Targets Primed by Semantically Related and by Unrelated Control Primes for All Three Bilingual Groups in Experiment 2D

Cognate Status/Bilingual Group

Cognate Non-cognate

Early Late high Late low Early Late high Late low Prime Type RT

(Error %)

RT (Error %)

RT (Error %)

RT (Error %)

RT (Error %)

RT (Error %)

Semantic 806

(21.7)

824 (22.7)

763 (22.1)

744 (7.8)

752 (10.2)

695 (13.6)

Control 733

(11.8)

729 (11.9)

688 (12.4)

650 (2.4)

640 (5.8)

668 (11.2)

Priming −73 −95 −75 −94 −112 −27

In the mixed-effects analysis on the error data, the main effect of Prime Type tended towards significance [F(1,146) = 12.409, p = .000]. Participants recognized Bodo

targets preceded by Assamese unrelated control primes more accurately than those preceded by Assamese semantically related primes. The main effect of Cognate Status was significant, [F(1,146) = 13.153, p = .000]. Errors were numerous for cognate targets than for non-cognate targets. The main effect of Bilingual Group did not approach significance in the error analysis [F < 1]. However, the Cognate Status and Bilingual Group interaction was found to be significant [F(2,6757) = 9.239, p = .000]. The percentage of errors for all three Bilingual Groups was numerous for the cognate targets than for the non-cognate targets. Lastly, comparisons of the nonword data were conducted. Analyses of the mean reaction times to nonwords showed that the participants took significantly longer to respond to nonwords (882 ms) than to words (709 ms).

4.3.2.5.3 Discussion

The results of this experiment demonstrated the lack of facilitation in the L2−L1 direction, when primes were semantically related to the targets. This finding is consistent with Basnight-Brown and Altarriba’s (2007) results where no semantic facilitation was observed in the L2−L1 direction. The finding also conforms to the predictions of the RHM which predicts null effects in this direction.

4.3.2.5.4 Combined Analysis of Experiment 2C and 2D

To test for differences between translation priming in both directions, we analyzed the data from Experiment 2C and 2D in one design using a mixed-effects analysis.

The combined analysis revealed a significant main effect of Target Language [F(1,13271) = 562.029, p = .000]. The interaction between Prime Type and Target Language approached significance [F(1,13266) = 7.170, p = .007]. The Cognate Status and Target Language interaction was also found to be significant [F(1,13265)

= 16.818, p = .000]. Of crucial interest, was the interaction between Bilingual Group and Target Language which was found to be very significant [F(2,13261) = 307.603, p = .000]. This interaction is shown in Figure 4.7. Finally, the three-way interaction between Relatedness, Bilingual Group and Target Language approached significance [F(2,13226) = 22.394, p = .000].

Figure 4.6 Mean RTs (ms) as a function of Bilingual Group and Target Language in Experiment 2C and 2D.