Rachel Wilson, University of Arizona
A central question in research into language learning is to determine what causes learners' performance problems. Language teachers often observe that while their students may "know" a particular construction, they exhibit difficulties in either producing it or understanding it or both. Learners also observe this phenomenon themselves, wondering why their performance does not reflect what they "know."
It has long been believed that linguistic competence is quite different from linguistic performance (Chomsky, 1965), and that, even though one of our goals is to describe a learner's competence, we can only do so through observations of learner performance. However, despite the widespread acceptance of this distinction, serious gaps remain in teasing apart the causes of learner error-- which errors are due to lack of linguistic knowledge and which errors are due to inefficient processing?
A crucial issue that has not been adequately examined so far is the role that the performance systems themselves may play in affecting performance-that is, language processing systems which put linguistic competence into actual use. Research has shown that many linguistic processing strategies are actually language-specific (Abramson & Lisker, 1967; Cuetos & Mitchell, 1988; Cutler, Mehler, Norris, & Segui, 1986), thus the natural implication for second language learners is that they will experience difficulties when using their native processing systems to operate in a language which requires different ones.
This paper will focus on those processing strategies and examine how language-specific processing can affect learner performance. The hypothesis here is that many performance problems are not the result of competence deficits, but rather processing inefficiencies. An eventual goal of this research program is to describe in detail the processing inefficiencies and propose methods for eradicating them. However, first it is necessary to establish that the processing mechanisms themselves are responsible for some learner error; this is the aim of the current paper.
A simple way to test this hypothesis is to compare learner performance on two tasks which differ in the amount of processing load, but require the same amount of linguistic knowledge. Thus, if performance on the two tasks is different, we can conclude that processing load affects learner performance. These results would presumably be most clear in a case where language-specific processing strategies are known to exist and are different in the L2 from the L1.
According to research in language processing, word order plays a significant role in the way that sentences are processed in some languages, but may play a lesser role in other languages. For example, the fact that English speakers are slower to process sentences that are not SVO--for example, object relative clauses (Forster & Olbrei, 1973; Nicol, Forster, & Veres, 1997) shows that in English, word order is a very important factor in language processing. On the other hand, for Serbo-Croatian speakers, word order is not the most important factor in processing: case-marking is (Urosevic, et al. 1988). This is despite the fact that Serbo-Croatian speakers also exhibit a strong preference for SVO word order (Urosevic, et. al. 1986; Slobin and Bever, 1982). No research has been done on the processing of Russian sentences, but here it is assumed that native Russian speakers behave similarly to Serbo-Croatian speakers.
Bearing in mind this crucial difference, it is easy to see the conflicts that must arise for a native speaker of English learning Russian. The speaker, accustomed to only noticing word order, would have to a) learn to pay attention to case-marking endings and b) learn to rely less on word order. This combination of learning to do something new and at the same time "unlearning" old conventions may indeed be very difficult.
This notion leads to the following experiment. Recall that the original hypothesis is that some performance problems are due to the interaction of the language-specific processing protocols of the L1 and the L2. By having learners perform an experiment in a context where language-specific protocols are very salient, varying the processing load but keeping all other elements of the task constant, it should be clear whether or not the processing protocols themselves are having an effect on learner performance.
The basic details of the experiment are as follows. A list of 72 three-word Russian sentences was constructed. The sentences were carefully designed so that they varied in word order and case-marking. All six possible word orders of Russian were represented equally. Half of the 72 sentences were grammatical and half were ungrammatical, arranged in random order. A native speaker of Russian recorded the sentences on a tape, pausing briefly between each sentence. Next, the native speaker made a second recording in which he repeated the sentence and allowed a long lag time in between each sentence. The duration of the first recording was approximately six minutes; the second was approximately 20 minutes.
A group of native English speakers enrolled in a third-year Russian class was asked to listen to the first recording and report on an answer sheet whether each sentence was grammatical or ungrammatical or if they were unsure. After a brief rest period, the same students listened to the second recording and again reported the grammaticality/ungrammaticality of the sentences. A smaller group of native Russian speakers also performed this task to ensure its validity.
Recall that the recordings are identical except that in the first recording the sentences are read very quickly, whereas in the second recording the sentence is repeated and there is a long break in between sentences. Because of the speed difference, clearly the first recording puts a higher demand on the processing system. If performance is better on the second task, we can then conclude that the processing system itself is responsible for poorer performance on the first task. If processing load has no effect on performance, than performance on the two tasks should be identical.
The results of the experiment are somewhat curious. First, as predicted, performance did significantly improve on the second task, confirming the hypothesis that learners' performance is affected by processing. This was the main goal of the experiment. However, there were some secondary predictions as well, based on the word-order preferences of English speakers and the case-marking preferences of Russian speakers. These secondary predictions were sometimes confirmed and other times counter-indicated.
These results point to the complex interaction between word-order and case-marking and the need for further research. It would be desirable in the future to perform the above experiment on a larger subject group. Also, additional tasks which take into account the effects of unambiguous vs. ambiguous case-markers (such as the lexical-decision task described in Urosevic, et al., 1988) should be performed.
The main results are very clear, though. The second language learners examined in this study possess a high degree of linguistic knowledge, but are unable to put that knowledge to use in real time.
BibliographyAbramson, A., & Lisker, L. (1967, 1970). "Discrimination along the voicing continuum: Cross-language tests." Paper presented at the Sixth International Congress of Phonetic Sciences, Prague.