Learning processes
Psychology provides us with three major frameworks for the focus on learning processes: Information Processing (IP), Connectionism, and Complexity Theory. IP has had more influence on the study of SLA than any other psychological perspective, following an approach developed by John Anderson (e.g. 1976, 1983). All three make the claim that learning language is essentially like learning other domains of knowledge: that whether people are learning mathematics, or learning to drive a car, or learning Japanese, they are not engaging in any essentially different kind of mental activity. Learning is learning. We take a general look at the information- processing framework and then discuss three approaches based on it, the Multidimensional Model, Processability, and the Competition Model, respectively. The Connectionism framework also claims that “learning is learning,” but considers learning processes as a matter of increasing strength of associations rather than as the abstraction of rules or principles. Complexity Theory focuses on processes and states of change in a wide variety of domains. As it has been applied to language development, it differs from other psychological approaches in the importance it gives to (1) social and contextual as well as cognitive factors and (2) the role of variability.
Information Processing (IP)
Approaches based on IP are concerned with the mental processes involved in language learning and use. These include perception and the input of new information; the formation, organization, and regulation of internal (mental) representations; and retrieval and output strategies.
The information processing approach makes a number of assumptions (McLaughlin 1987):
(1) Second language learning is the acquisition of a complex cognitive skill. In this respect language learning is like the acquisition of other complex skills.
(2) Complex skills can be reduced to sets of simpler component skills, which are hierarchically organized. Lower-order component skills are prerequisite to learning of higher-order skills.
(3) Learning of a skill initially demands learners’ attention, and thus involves controlled processing.
(4) Controlled processing requires considerable mental “space,” or attentional effort.
(5) Humans are limited-capacity processors. They can attend to a limited number of controlled processing demands at one time.
(6) Learners go from controlled to automatic processing with practice. Automatic processing requires less mental “space” and attentional effort.
(7) Learning essentially involves development from controlled to automatic processing of component skills, freeing learners’-controlled processing capacity for new information and higher-order skills.
(8) Along with development from controlled to automatic processing, learning also essentially involves restructuring or reorganization of mental representations.
(9) Reorganizing mental representations as part of learning makes structures more coordinated, integrated, and efficient, including a faster response time when they are activated.
(10) In SLA, restructuring of internal L2 representations, along with larger stores in memory, accounts for increasing levels of L2 proficiency.
Our mental capacity requirements for controlled processing are obvious when we are beginning to learn a second language, as we need to concentrate our attention to comprehend or produce basic vocabulary and syntactic structures. It is only after these have been automatized that we can attend to more complex, higher-order features and content. We encounter similar capacity limitations (we easily experience “information overload”) in learning a new “language” for computerized word processing: we must initially use controlled processing to select appropriate symbols and apply the right rules, and it is difficult or impossible to simultaneously pay attention to higher-order content or creative processing. It is only after we have automatized the lower-level skills that our processing capacity is freed for higher-order thought. Writers usually cannot compose “online” effectively until lower-level word-processing skills such as typing, saving documents, and changing fonts have become automatized. Further examples can readily be drawn from learning other complex nonverbal skills, such as driving or skiing, where tasks that initially require attentional control become automatized with practice; they then generally remain out of conscious awareness unless some unusual occurrence returns them to controlled processing. Behaviors under attentional control are permeable, i.e. they are changeable; but once automatized, they are both more efficient and more difficult to change. In fact, one explanation for L2 fossilization (or apparent cessation of learning) from an IP perspective is that aspects of L2 may become automatized before they have developed to target levels, and positive input no longer suffices to lead to their improvement.
Information Processing has three stages, as shown in Table 4.2 (adapted from Skehan 1998).
Input for SLA is whatever sample of L2 that learners are exposed to, but it is not available for processing unless learners actually notice it: i.e. pay attention to it. Then it can become intake. It is at this point of perception of input where priorities are largely determined, and where attentional resources are channeled. Richard Schmidt (1990) lists the following features as likely contributors to the degree of noticing or awareness which will occur:
• Frequency of encounter with items
• Perceptual saliency of items
• Instructional strategies that can structure learner attention
• Individuals’ processing ability (a component of aptitude)
• Readiness to notice particular items (related to hierarchies of complexity)
• Task demands, or the nature of activity the learner is engaged in
In line with this IP approach to learning, developing, and testing strategies to heighten learner awareness of input and to structure attention has been a major thrust in foreign language instructional design and pedagogy, so that successful intake can occur.
Output for SLA is the language that learners produce, in speech/sign or in writing. The importance of output for successful L2 learning has been most fully expounded by Merrill Swain (e.g. Swain and Lapkin 1995). Meaningful production practice helps learners by:
• Enhancing fluency by furthering development of automaticity through practice
• Noticing gaps in their own knowledge as they are forced to move from semantic to syntactic processing, which may lead learners to give more attention to relevant information
• Testing hypotheses based on developing interlanguage, allowing for monitoring and revision
• Talking about language, including eliciting relevant input and (collaboratively) solving problems
Fluency is achieved in production both through use of automatized rule-based systems and through memory-based chunks which serve as exemplars or templates and are “retrieved and used as wholes” (Skehan 1998 :60).
Central processing is the heart of this model, where learning occurs. It is here that learners go from controlled to automatic processing, and where restructuring of knowledge takes place. It is possible to test for degree of automatization because controlled processing requires more time. Research that measures the amount of time it takes multilinguals to recall words and grammatical structures shows that the L2 of even fluent speakers of both languages is generally less automatized than their L1, and less proficient L2 is less automatized than more fluent L2.
In the model of learning that was proposed by Anderson (1976), development from declarative to procedural stages of knowledge is parallel to development from controlled to automatic processing in many respects. The declarative stage involves acquisition of isolated facts and rules; processing is relatively slow and often under attentional control. Development to the procedural stage involves processing of longer associated units and increasing automatization, which frees attentional resources for higher level skills. Proceduralization requires practice.
As noted in the assumptions about IP listed above, the restructuring that takes place during central processing makes mental representations more coordinated, integrated, and efficient. It involves qualitative changes, meaning that L2 development cannot be characterized as a seamless continuum along which new forms are added to old, but as a partially discontinuous plane along which there is regular systemic reorganization and reformulation. Two types of evidence from learners’ speech and writing are often cited. One is the sequence of acquisition which learners exhibit when they produce unanalyzed chunks of L2 correctly and then make errors as they restructure the elements they have processed in accord with newly formulated patterns and rules: i.e. an onset or increase of ungrammaticality in utterances is often an indicator of “progress” in SLA. A related type of evidence is found in U-shaped development: i.e. learners’ use of an initially correct form such as plural feet in English, followed by incorrect foots, eventually again appearing as feet. In this case, feet is first learned as an unanalyzed word, without recognition that it is a combination of foot plus plural. The later production of foots is evidence of systemic restructuring that takes place when the regular plural -s is added to the learner’s grammar. Feet reappears when the learner begins to acquire exceptions to the plural inflection rule.
Theories regarding order of acquisition
Psychological approaches to SLA have made significant contributions to understanding why certain elements are acquired in a fixed sequence. One of the best known of these approaches is the Multidimensional Model, developed by researchers who initially studied the German L2 learning of adult L1 speakers of Italian, Spanish, and Portuguese in the ZISA project (see Clahsen, Meisel, and Pienemann 1983). This model includes the following claims:
• Learners acquire certain grammatical structures in a developmental sequence.
• Developmental sequences reflect how learners overcome processing limitations.
• Language instruction which targets developmental features will be successful only if learners have already mastered the processing operations which are associated with the previous stage of acquisition.
The processing strategies which account for developmental sequences in perception and production are explained by Clahsen (1984) in relation to the IP constraint of limited capacity: “linguistic structures which require a high degree of processing capacity will be acquired late” (p. 221). Which syntactic structures require more processing capacity (i.e. are more com plex) is determined by the extent to which their underlying relations are preserved in output, and by the perceptual salience of any reordering that does occur. Clahsen (1984:23) infers the following hierarchy:
Canonical Order Strategy: There is no reordering from “basic” word order. Structures which can be processed with this strategy will be acquired first.
Initialization/Finalization Strategy: Reordering which moves underlying elements into the first or last position in a grammatical string are perceptually more salient, and thus easier to process than permutations to internal positions.
Subordinate Clause Strategy: Reordering in subordinate clauses is not allowed. This accounts for why “learners initially use certain reorderings only in main clauses and [. . .] thus the order of the elements in subordinate clauses is less varied.”
A reorientation of the Multidimensional Model is known as Processability Theory (Pienemann 1998; Pienemann and Kessler 2011); it also has the aim of determining and explaining the sequences in which processing skills develop in relation to language learning. The following acquisitional hierarchy of processing skills is proposed (from Pienemann and Håkansson 1999):
Lemma/word access: Words (or lemmas) are processed, but they do not yet carry any grammatical information, nor are they yet associated with any ordering rules.
Category procedure: Lexical items are categorized, and grammatical information may be added (e.g. number and gender to nouns, tense to verbs).
Phrasal procedure: Operations within the phrase level occur, such as agreement for number or gender between adjective and noun within the noun phrase.
S-procedure: Grammatical information may be exchanged across phrase boundaries, such as number agreement between subject and verb.
Clause boundary: Main and subordinate clause structures may be handled differently.
This is an implicational hierarchy in the sense that processing skill at level 1 is a prerequisite for processing skill at level 2, level 2 is prerequisite for level 3, and so forth. The sequence of strategies describes the developing learner grammar in terms of processing prerequisites needed to acquire grammatical (syntactic and morphological) rules at successive stages.
The universality of this sequence in SLA is being tested by researchers, with generally supportive results. In addition to Pienemann’s analysis of German L2 (1998) and reanalysis of data from prior research on Swedish L2 (Pienemann and Håkansson 1999), the most extensive studies thus far have been on Danish, Norwegian, and Swedish (Glahn et al. 2001).
Claims that language instruction will be effective only if it targets the next stage in an L2 learner’s developmental sequence (rather than more advanced levels) have been tested on many languages since the 1970s (reviewed in Spada and Lightbown 1999). Results are mixed concerning the interaction of developmental order and instructional level, with indication that at least for some structures, and for some learners, instruction at a more advanced level can be more efficient. Complexities include the type of instruction (e.g. whether explicit contrastive L1–L2 information on the structure is presented), and the degree to which L1 knowledge may be applicable. However, these complexities do not appear to invalidate claims about order of acquisition; even when learners profit from more advanced levels of instruction, they progress through the same developmental sequence.
Competition Model
Another psychological approach that has addressed the general question of how languages are learned is the Competition Model (Bates and MacWhinney 1981; MacWhinney 2001). This is a functional approach which assumes that all linguistic performance involves “mapping” between external form and internal function. The form of a lexical item is represented by its auditory properties, and its function by its semantic properties; the forms of strings of lexical items are word-order patterns and morphological inflections, and their functions are grammatical. For example, for the word horse the form is represented by the sounds [hors]; the function is the meaning of a four-legged, hay-eating animal. In the sentence Horses eat hay, the word orders of horses before and hay after the verb are forms; the functions are to convey that horses is the subject and hay is the object. The inflection -s on horses is also a form; its function is to convey that more than one horse is being referred to.
This approach considers that learning the system of Form–function mapping is basic for L1 acquisition. SLA involves adjusting the internalized system of mapping that exists in the learner’s L1 to one that is appropriate for the target language. This is accomplished by detecting cues in language input which are associated with a particular function, and by recognizing what weight to assign each possible cue (the cue strength). The cue in English that horses is the subject in the sentence Horses eat hay is word order – horses comes in front of the verb. If the sentence were in Japanese, the cue would be a case marker, the inflection -ga that is attached to the end of a word which means it is the subject (i.e. that it has nominative case).
Multiple cues are available simultaneously in input; language processing essentially involves “competition” among the various cues. For example, for the grammatical function of subject, possible cues are word order, agreement, case marking, and animacy (i.e. capacity for volitional action). All of these possible cues are illustrated in the following sentences (some are not grammatical or grammatically felicitous):
The relative strength of word order as a cue in English over the other possibilities can be tested by presenting native speakers with sentences such as these and asking them to identify the subject or agent in each (i.e. who/ what does the “kicking”).
In spite of the ungrammaticality of (b–c), or in the case of (d) its anomalous character, native English speakers are most likely to identify the first noun phrase in each of these sentences as subject, even though in (b) the verb agrees with the second noun phrase rather than the first, in (c) him is case-marked as object (the receiver of the action) rather than subject, and in (d) fence is inanimate and cannot be interpreted literally as a “doer” of the verb kick . If these sentences were translated into other languages, different identifications of subject would likely be made depending on whether agreement, case marking, or animacy carried more weight. In Japanese, for instance, the case marker -ga attached to a noun phrase (if no other -ga occurred) would generally carry more weight in identifying that NP as the subject, no matter where in the word order it occurred. An English L1 speaker learning Japanese as L2 might inappropriately transfer the strong word-order cue to initial form–function mapping (and identify the wrong noun phrase as subject if it occurred first), whereas native speakers of Japanese might transfer their L1 cue weights to English L2 and also provide nonnative interpretations.
Acquisition of appropriate form–function mappings is driven primarily by the probability that a particular functional interpretation should be chosen in the presence of a particular cue. If the probability is high, the cue is reliable. The following determinants of cue strength are also discussed by MacWhinney (2001 :74–75; see Ellis 2008 :473–79):
• Task frequency: how often the form–function mapping occurs. The vast majority of English sentences have a subject before the verb, so the mapping of word-order form to subject function is very frequent.
• Contrastive availability: when the cue is present, whether or not it has any contrastive effect. In example (a) above, for instance (The cow kicks the horse), the third person singular - s on the verb agrees with both noun phrases and so the agreement cue tells nothing about which is the subject. An available cue must occur contrastively if it is to be useful.
• Conflict reliability: how often the cue leads to a correct interpretation when it is used in comparison to other potential cues.
Transfer of L1 cue strengths to L2 is the most likely outcome in early stages of SLA when the systems differ, but research has shown that some learners ultimately abandon L1 cue strengths in favor of L2, while some compromise and merge the two systems, and some differentiate between the languages in this aspect of processing.
Connectionist approaches
Connectionist approaches to learning have much in common with IP perspectives, but they focus on the increasing strength of associations between stimuli and responses rather than on the inferred abstraction of “rules” or on restructuring. Indeed, from a connectionist perspective learning essentially is change in the strength of these connections. Some version of this idea has been present in psychology at least since the 1940s and 1950s (see McClelland, Rumelhart, and Hinton 1986 for an overview of historical developments), but Connectionism has received widespread attention as a model for first and second language acquisition only since the 1980s.
The best-known connectionist approach within SLA is Parallel Distributed Processing, or PDP. According to this viewpoint, processing takes place in a network of nodes (or “units”) in the brain that are connected by pathways. As learners are exposed to repeated patterns of units in input, they extract regularities in the patterns; probabilistic associations are formed and strengthened. These associations between nodes are called connection strengths or patterns of activation. The strength of the associations changes with the frequency of input and nature of feedback. The claim that such learning is not dependent on either a store of innate knowledge (such as Universal Grammar) or rule-formation is supported by computer simulations. For example, Rumelhart and McClelland (1986) demonstrated that a computer that is programmed with a “pattern associator network” can learn to associate English verb bases with their appropriate past tense forms without any a-priori “rules,” and that it does so with much the same learning curve as that exhibited by children learning English L1. The model provides an account for both regular and irregular tense inflections, including transfer to unfamiliar verbs, and for the “U-shaped” developmental curve (discussed in the previous section on order of acquisition) which is often cited in linguistic models and in other cognitive approaches as evidence for rule-based learning.
Assumptions about processing from a connectionist/PDP viewpoint differ from traditional IP accounts in other important ways. For example (McClelland, Rumelhart, and Hinton 1986; Robinson 1995):
(1) Attention is not viewed as a central mechanism that directs information between separate memory stores, which IP claims are available for controlled processing versus automatic processing. Rather, attention is a mechanism that is distributed throughout the processing system in local patterns.
(2) Information processing is not serial in nature: i.e. it is not a “pipeline . . . in which information is conveyed in a fixed serial order from one storage structure to the next” (Robinson 1995 :288). Instead, processing is parallel: many connections are activated at the same time.
(3) Knowledge is not stored in memory or retrieved as patterns, but as “connection strengths” between units which account for the patterns being recreated.
It is obvious that parallel processing is being applied when tasks simultaneously tap entirely different resources such as talking on a cell phone while riding a bicycle, but it also less obviously occurs within integrated tasks such as simply talking or reading, when encoding/decoding of phonology, syntactic structure, meaning, and pragmatic intent occur simultaneously. Many connections in the brain must be activated all at once to account for successful production and interpretation of language, and not processed in sequence (i.e. one after the other).
Little research based on this approach has been conducted in SLA, but the assumption is that transfer from L1 to L2 occurs because strong associations already established in L1 interfere with establishment of the L2 network. Because frequency is the primary determinant of connection strength, it might be predicted that the most common patterns in L1 would be the most likely to cause interference in L2, but research on trans fer from linguistic perspectives does not support this conclusion in any strong sense; L1–L2 relationships are not that simple. Proponents of connectionist approaches to language acquisition note that while frequency is “an all-pervasive causal factor” (Ellis 2002 :179), it interacts with other determinants, including how noticeable the language patterns are in the input learners receive, and whether the patterns are regular or occur with many variations and exceptions.
Many linguists and psychologists would argue against a strong deterministic role for frequency of input in language learning. One counter argument is that some of the most frequent words in English (including the most frequent, the) are relatively late to appear, and among the last (if ever) to be mastered. Still, whatever one’s theoretical perspective, the effects of frequency on SLA clearly merit more attention than they have typically received since repetition drills went out of fashion in language teaching. Researchers from several approaches to SLA which focus on learning processes are taking a renewed look at how frequency influences learning.
الاكثر قراءة في Linguistics fields
اخر الاخبار
اخبار العتبة العباسية المقدسة
