8. RESULTADOS Y ANÁLISIS
8.2 Tratamientos aplicados a la vinaza
8.2.1 Tratamiento electrodisolución-heterocoagulación
L1 vs. Advanced L2. What should be carefully considered is the interaction effect of article and speaker group. Table 4.5 is the output summary of a glmer model that fits the
interaction between article and speaker group in L1 and advanced L2 speakers. The grand mean of unique referent prediction is about 56%, which roughly matches the average of the grand mean of the two groups: 61.8% in L1 speakers and 50.3% in advanced L2 speakers.
Table 4.5. The glmer output for the interaction effect of article and group in L1 and adv L2.
(a) Fixed effects Subject (Intercept) 𝛾&& 0.4681 0.6841
article 𝛾* 0.0010 0.0316 1 item (Intercept) 𝛿&& 0.5438 0.7374
article 𝛿* 0.0003 0.0165 -1 Number of obs: 1514, groups: Subject, 76; item, 20
The coefficient estimate for the article is .200, which is also the same as that in L1 and advanced L2 speakers. Because both L1 and advanced L2 speakers showed a significant increase of unique referent prediction in the definite condition, they show the same significant main effect
of article; thus, no main effect of speaker group was observed. The two groups behaved alike and no significant interaction effect of article and speaker group were observed.
L1 vs. intermediate L2. Table 4.6 shows the output of a glmer model that includes only L1 speakers and intermediate L2 speakers. This model shows a marginal main effect of group and a significant interaction effect of article and group. The main effect of article observed in L1 speakers and advanced L2 speakers cannot be seen here. This is due to the opposite patterns of behavior in the two groups. That is, the unique referent prediction rate increases in L1 but
decreases in intermediate L2 at the cue of a definite article; thus, a main effect observed in the L1 group analysis is canceled out by the opposite behavior of the intermediate L2 speakers.
Table 4.6. The glmer output for the interaction effect of article and group in L1 and inter L2.
(a) Fixed effects
Groups Name Variance Std.Dev.
Subject (Intercept) 𝛾&& 0.4515 0.6719 item (Intercept) 𝛿&& 0.7230 0.8503 Number of obs: 1213, groups: Subject, 61; item, 20
Figure 4.4 clearly visualizes why the main effect of article cannot be observed here. The dotted line hiding behind the solid line in Figure 4.4 is the L1 group. And the dashed line in the opposite direction is the intermediate group. Because the two groups make different
predictions for each article, the main effect of article cannot be observed. Instead, the marginal main effect of group indicates that the overall rate at which L1 speakers predict unique referents was higher than intermediate L2 learners would do so.
Figure 4.4. The interaction of speaker group and article.
Finally, the cross-over interaction effect is significant at the .013 alpha level. This means that the opposite tendency observed between the two groups is not a mere coincidence. It is for this reason that the numerical difference between definite and indefinite conditions observed in the intermediate L2 group analysis cannot be disregarded. Although the pairwise comparison between definite and indefinite conditions within the intermediate L2 group was not statistically significant, the interaction effect clearly shows how the linguistic information (definite vs.
indefinite articles) takes a different effect by speaker group.
Advanced L2 vs. intermediate L2. The comparison between advanced L2 speakers and intermediate L2 speakers showed the same result. Table 4.7 summarizes the output for the glmer model with only L2 speakers. There was no main effect of article for the same reason as in the
comparison between L1 and intermediate L2 speakers. Because advanced L2 speakers predicted a unique referent more at the cue of a definite article and intermediate L2 speakers did so at the cue of an indefinite article, no main effect of article was observed across the two groups. No main effect of group was observed, which means that, unlike in the comparison between L1 and intermediate L2, the grand mean difference between advanced and intermediate L2 was not statistically significant. The cross-over interaction of article and speaker group was significant.
This means that the opposite behavior of the two speaker groups was due to chance.
Table 4.7. The glmer output for the interaction of article and group between adv. and inter. L2.
(a) Fixed effects
Groups Name Variance Std.Dev.
Subject (Intercept) 𝛾&& 0.3703 0.6085 item (Intercept) 𝛿&& 0.6239 0.7899 Number of obs: 1379, groups: Subject, 69; item, 20
Intermediate L2 vs. the rest. Since L1 and advanced L2 speakers showed an identical behavior, another glmer model was fit where all three speaker groups were included. In this model, L1 and advanced L2 speakers were collapsed into one group. The interaction effect of article and speaker group was significant in this model as well. The output summary is provided
cannot be dismissed just because the main effect of article within the intermediate group was not significant.
Table 4.8. The glmer output for the interaction of article and group b/w inter. L2 vs. the rest (a) Fixed effects Subject (Intercept) 𝛾&& 0.4310 0.6565 item (Intercept) 𝛿&& 0.7178 0.8473
article 𝛿* 0.0059 0.0770 0.10 group 𝛿- 0.0552 0.2350 0.57 0.88 Article x group 𝛿/ 0.0005 0.0220 0.39 0.96 0.98 Number of obs: 2053, groups: Subject, 103; item, 20
Their pattern of behavior, which is in the opposite direction from the other groups, is noteworthy enough due to this interaction effect. This interaction indicates how the same linguistic information takes a completely different effect in the speaker groups and the different patterns of behavior might signify the developmental path of English article acquisition in L1 Korean L2 English speakers.
One possible interpretation of the intermediate L2 speakers’ behavior is that they might interpret the indefinite article as a marker of singularity. Most English-Korean dictionaries list
‘one’ as the first entry for the meaning of ‘a(n),’ and sample sentences also have direct
translations including the word hana in Korean meaning ‘one.’ Due to the influence of their L1 or their English instruction in Korea, Korean learners of English might think of the indefinite article as a singularity marker. If this is the case, the effect of article in the intermediate group might have been significant if the study had included a larger sample size; this question remains for further research.