Ch.+11+by+Michael+Harrington


 * Chapter 11 **

By Michael Harrington
 * Cognitive Perspectives on Second Language Acquisition **

A cognitive theory of Second Language Acquisition (SLA) tries to clarify the existing psychological mechanisms which allow the learner to develop the competence for comprehension and production. Harrington (1995) discusses when and how the learner uses different kinds of sources of information, syntax, and context and so on in real-time processing outcomes in comprehension in the sentence processing. There are two approaches which discuss real-time sentence processing. The //syntax-based// approach shows the comprehension process as the application of independent syntactic principles. It is assumed that the semantics, frequency, and contextual information do not play any role in initial parsing decisions. In contrary, according to the //constraint-based// models the comprehension is a result of the interaction of multiple sources of knowledge i.e. linguistic, pragmatic, contextual, and real-world.


 * SLA as a Cognitive Science **

Cognitive theory in SLA has been identified with the information-processing paradigm and was dominant metaphor for cognition a few decades ago. An information process is the device by using which a system gives systematic answers (which are mainly goal oriented) to certain environmental conditions. This approach implies that the mind is a general purpose symbol processing system which is ground to capacity constraints. Information processing approach differs from sentence processing one in three ways. //First// it says that the mind is a general purpose symbol processor something which is the main issue of controversy in sentence processing research. The syntax-based and constraint-based approaches present, respectively, classical and connectionist views of cognition and language. The classical view has a symbolic level of representation in contrast with the connectionist view which characterizes language knowledge. //Second// in the information processing approach it is unclear about the nature of the linguistic knowledge whereas in sentence processing approach there are models of knowledge development. //The third// difference is in terms of temporal dimensions. For sentence processing research it is vital to find out how comprehension processes are carried out in real time. In comparison with this information processing research has to do with identifying the conditions that affect learners’ outcomes. Though there are some differences between these two approaches, they are complementary.


 * A Syntax-Based Approach to Sentence Processing **


 * Language processing as Symbol Computation **

In the syntax-based approach, cognition and language are characterized as a symbol manipulation process (Newell, Rosenbloom, and Laird 1989) and knowledge is presented in the mind in symbols and that computations. In natural language computation of these symbols include phonemes, morphemes, grammar, etc. The symbolic view has been the main approach to cognition. Sentence processing has seen as a process of symbolic computation. The role of //syntactic structure// is vital in terms of how the sentence parser (or processor) builds it that eventually leads to an interpretation of the sentence.


 * Syntax-Based Processing **

According to this approach syntactic knowledge plays a primary role in the sentence interpretation process and that knowledge consists of an independent competence grammar. There is also a distinction between the mechanisms responsible for lexical processing (e.g., word recognition and lexical access) and syntactic processing.

Two categories can be considered to classify the syntax-based approaches: //principle-based// (Crocker 1994, Frazier 1989, Pritchett 1992) and //referential// (Altmann and Steedman 1998, Crain and Steedman 1985).


 * <span style="font-family: 'Times New Roman',serif; font-size: 12pt;">A Syntax-Based Second Language Processing **

<span style="font-family: 'Times New Roman',serif; font-size: 12pt;"> Juffs and Harrington (1995) tried to investigate the asymmetry evident in L2 learner performance on grammatical judgments for //wh//-structures. According to earlier research ESL learners experience more difficulty judging the acceptability of subject extraction sentence (ex. Who did Ann believe –likes her friend?) than object extraction one (ex. Who did Ann believe her friend likes?). Structures containing constraints on //wh//-movement are interesting because they give an opportunity to test the supposed effects of UG (Universal Grammar) in adult SLA. The difficulties which arise in SLA might be due to the lack of processing rather than lack of competence.

<span style="font-family: 'Times New Roman',serif; font-size: 12pt;">Research has revealed that a range of non-syntactical information can neutralize or topple down predicted structural processing effects (Altman & Steedman, 1988). Moreover, the manipulation of prior discourse can bias responses toward the main verb/reduced relative clause readings in a garden path sentence and the sentence interpretations as well. The initial parsing decisions can also be affected by semantic effects. <span style="font-family: 'Times New Roman',serif; font-size: 12pt;">Syntax-based approach gives us a view of how knowledge of L2 learner develops and provides an explicit and testable set of processing claims. However, the contribution of this approach to a cognitive theory of SLA remains unclear.
 * <span style="font-family: 'Times New Roman',serif; font-size: 12pt;">Syntax-Based Approaches in SLA **


 * <span style="font-family: 'Times New Roman',serif; font-size: 16pt;">A Constraint-Based Approach to Sentence Processing **

<span style="color: black; font-family: 'Times New Roman',serif; font-size: 12pt;">In contrary to syntax-based approach, constraint-based models describe sentence interpretation as a highly interactive process.

<span style="font-family: 'Times New Roman',serif; font-size: 12pt;">According to constraint-based models language is represented as distributed (produce rule-like behavior) and graded features (result of experience) and linguistic knowledge –as a complex pattern of associative links between units (//microfeatures//) which are smaller than the traditional units of linguistic analysis. <span style="font-family: 'Times New Roman',serif; font-size: 12pt;">The distributed and graded nature of knowledge representations means that language knowledge is describable in //probabilistic// terms (Seidenberg and McDonald, 1999). Probabilistic models have a benefit over rule-based ones in catching the variable nature of behavior. <span style="font-family: 'Times New Roman',serif; font-size: 12pt;">In constraint-based approach there is a direct relationship between processing and learning where the mechanisms involved in processing input are also responsible for learning new knowledge (Shakey, 1996).
 * <span style="font-family: 'Times New Roman',serif; font-size: 12pt;">Language as a Distributed Probabilistic System **


 * <span style="font-family: 'Times New Roman','serif'; font-size: 16px;">Sentence Processing as Constraint Satisfaction **

<span style="font-family: 'Times New Roman','serif'; font-size: 16px;">Sentence interpretation is done based on a lexicalist model suggested by McDonald at al. (1994). In this model different information types for the same unit are activated in parallel. However a particular unit is activated stronger, reflecting the type, number and strength of the links with others.


 * <span style="font-family: 'Times New Roman','serif'; font-size: 16px;">Constraint-Based Models of Sentence Processing: Connectionist Sentence Processing **

<span style="font-family: 'Times New Roman','serif'; font-size: 16px;">Connectionist approach has been effective in characterizing learning in local domains of language such as phonological, lexical, morphological, and syntactical. This approach focuses on frequency of thematic roles, gender, number, and verb morphology (Ellis & Schmidt, 1998; Gasser, 1990; and Sokolik, 1990). As for tenses, native speakers seem to produce past tense of regular and irregular verbs in different frequencies. High frequency past forms are produced quickly, whereas low frequency ones slower.

<span style="font-family: 'Times New Roman','serif'; font-size: 16px;">There are two different mechanisms to be in charge of this phenomenon (Pinker & Prince, 1994). Irregular verbs that are sensitive to frequency effects appear to be stored as individual items in associative memory. Whereas regular pasts are generated linking the past tense morpheme to the stem at the time of speech production.

<span style="font-family: 'Times New Roman',serif; font-size: 12pt;">Constraint-Based Models are applied to put variation across learners, languages and even across setting, as it is also considered to be an essential part of SLA. These models, however, have certain faults and do not provide good alternatives to the syntax-based models of sentence processing.
 * <span style="font-family: 'Times New Roman',serif; font-size: 14pt;">Constraint-Based Models and SLA Theory **

<span style="font-family: 'Times New Roman',serif; font-size: 12pt;">Sentence processing gives full notion of how speech is processed and also a full understanding of the mind and behavior.
 * <span style="font-family: 'Times New Roman',serif; font-size: 14pt;">Cross-Linguistic Processing **

<span style="font-family: 'Times New Roman',serif; font-size: 12pt;">SLA researchers tend to concentrate on the study of sentence processing and so far research has been carried out only from the point of view of sentence processing. This limitation is because of the lack of technical reasons and methodological expertise.
 * <span style="font-family: 'Times New Roman',serif; font-size: 14pt;">Sentence Processing Research Methodology **