Chapter 5. The integration of linguistic and non-linguistic information in L2 processing
6.1 Summary
This dissertation investigated how two different sources of information are integrated in L2 sentence processing. The first two experiments tested whether L1 and L2 speakers have the same grammatical knowledge of English articles with respect to definiteness, which is
operationalized here as unique identifiability. The third experiment examined whether the two groups share the same type of real-world knowledge (that associates people of certain
occupations with certain objects). The final experiment investigated the integration of the two sources of information. The results are summarized below.
Experiment 1. A self-paced reading task probed L1 and L2 speakers’ sensitivity to the mapping of unique and non-unique referents to definite and indefinite articles, respectively. L1 speakers showed a reading time increase when the (non-)uniqueness of referents was not aligned with the (in)definiteness of the corresponding referring NPs. In L1 participants, the effects were observed in the spill-over region or one region after the critical NPs were shown. However, advanced L2 speakers displayed the same pattern of behavior one region later than their L1 counterparts. The results suggest that advanced L2 speakers know that the (in)definiteness of English articles indicates the unique identifiability of a referent, although they are slower in applying such knowledge in processing L2 input.
Experiment 2. A referent prediction task inversed the order of information presentation.
Unlike Experiment 1, which provided information on the (non-)uniqueness of a referent first and recorded participants’ reaction to either matching or mismatching articles, Experiment 2 tested
whether definite or indefinite articles would lead to the prediction of unique or non-unique referents. Incomplete sentences whose final word was a definite or indefinite article were presented along with visual stimuli that had one unique and three non-unique referents. The experiment was designed to see whether either article influences participants’ prediction of unique or non-unique referents. Both L1 and advanced L2 speakers predicted unique referents at the cue of a definite article significantly more than at the cue of an indefinite article. Intermediate learners showed the opposite tendency of predicting a unique referent at the cue of an indefinite article numerically more than at a definite article. Although statistically significant effects were not observed in the intermediate group perhaps due to the small sample size, a significant interaction effect of article and speaker group indicated that intermediate learners predict a unique referent significantly more at the cue of an indefinite article than L1 and advanced L2 speakers do at the same cue.
The two takeaway messages from Experiment 2 is that (1) advanced L2 speakers use definite and indefinite articles in a native-like manner to predict unique and non-unique referents and (2) intermediate L2 speakers show the tendency of using indefinite articles for singularity marking instead of as markers of definiteness. There appears to be a native-language effect at work here. Korean, the native language of the L2 participants in this dissertation has no articles and is forced to translate a/an as hana, the Korean equivalent of one in English. A larger sample size in the intermediate group may help confirm this suggestion.
Experiment 3a. A web-based norming survey confirmed that L1 and L2 speakers shared real-world knowledge when it comes to certain associations between people and objects. Some examples of these associations include the relationship between a doctor and a stethoscope, or that between a scientist and a microscope. Both L1 and L2 speakers agreed that doctors are more
strongly associated with stethoscopes than with laptops and that scientists are more closely related to microscopes than to sweaters. Proficiency levels in L2 did not have any effect on the type of world knowledge.
Experiment 3b. As Experiments 1 and 2 confirmed that L1 and advanced L2 speakers shared the same linguistic information with regard to English articles and Experiment 3a
demonstrated that non-linguistic information, operationalized as real-world knowledge, was the same for both L1 and L2 groups, it was established that both L1 and L2 speakers have access to the same linguistic and non-linguistic information. With this confirmation, the last experiment looked into the integration of linguistic and non-linguistic information via a referent
identification task. The results showed that L1 speakers incrementally integrate both linguistic and non-linguistic information. In particular, they confirm or revise their earlier predictions as they integrate information that becomes available later.
Advanced L2 speakers, who displayed the same capacity for using linguistic and non-linguistic information independently, did not make active use of non-linguistic information when the sentence corresponded to what they had initially predicted based on their real-world knowledge.
When the sentence referred to an object their real-world knowledge hadn’t predicted, however, advanced L2 speakers employed linguistic information to revise their initial expectations. This was due to the timing difference of processing linguistic and non-linguistic information. In general, non-linguistic information that matched their real-world knowledge was processed faster than information that is misaligned with their real-world knowledge, and processing linguistic information was as slow as processing non-linguistic information that did not align with real-world knowledge.
The results dovetail with claims that sentence processing is guided not only by linguistic constraints but also by situational, visual, and contextual constraints in L1 processing (Altmann
& Kamide, 1999; 2007; Tanenhaus, Spivey-Knowlton, Eberhard, & Sedivy, 1995). What was different in L2 sentence processing is that one source of information was processed faster than the other. That is, processing of real-world knowledge was not different in L1 and L2 speakers but L2 processing of the (in)definiteness semantics of English articles was not as automatized as L1 processing of the same information. This difference in the extent to which processing of each information source is automatized was discussed as the potential cause of the fossilization of Interlanguage L2 grammar.