Chapter 6. Input, interaction, and Second Language Acquisition
Book Name: Understanding Second Language Acquisition
Writer: Rod Ellis
Professor: Dr.Haniye Davatgari
Saeed Mojarradi PhD. Candidate Monday, November 19, 2018
Three views on input in language acquisition
It is axiomatic that in order for SLA to take place, there must be (1) some L2 data made available to the learner as input and (2) a set of internal learner mechanisms to account for how the L2 data are processed. A major issue in the study of SLA, however, has been to decide what weight to allot to (1) and (2). On the one hand it is possible to conceive of the learner as ‘a language-producing machine’ who automatically and effortlessly learns a 1.2, provided he gets the right input data.
On the other hand, the learner can be seen as ‘a grand initiator’; that is, he is equipped with just those abilities that are needed to discover the L2, no matter how impoverished the 1.2 data are. Also, of course, there are intermediate positions in which the learner is seen as actively contributing to SLA, but dependent on the provision of appropriate input.
Behaviorist accounts of SLA view the) earner as ‘a language-producing machine’. The linguistic environment is seen as the crucial determining factor. In this model of learning, input comprises the language made available to the learner in the form of stimuli and also that which occurs as feedback.
Behaviorist theories emphasize the need to regulate the stimuli by grading the input into a series of steps, so that each step constitutes the right level of difficulty for the level that the learner has reached. Feedback serves two purposes. It indicates when the L2 utterances produced by the learner are correct and so reinforces them, and it also indicates when the utterances are ill formed by correcting them. The regulation of the stimuli and the provision of feedback shape the learning that takes place and lead to the formation of habits. Nativist accounts of SLA view the learner as ‘a grand initiator’.
Thus, whereas a behaviorist view of language acquisition seeks to explain progress purely in terms of what happens outside the learner, the nativist view emphasizes learner-internal factors. A third view, however, is tenable. This treats the acquisition of language as the result of an interaction between the learner’s mental abilities and the linguistic environment.
The learner’s processing mechanisms both determine and are determined by the nature of the input. Similarly, the quality of the input affects and is affected by the nature of the internal mechanisms.
The interaction between external and internal factors is manifest in the actual verbal interactions in which the learner and his interlocutor participate. It follows from this interactionist view of language acquisition that the important data are not just the utterances produced by the learner, but the discourse which learner and caretaker jointly construct.
Three different views regarding the role of input in language development have been discussed. The behaviorist view emphasizes the importance of the linguistic environment, which is treated in terms of stimuli and feedback.
The nativist vim minimizes the role of the input and explains language development primarily in terms of the learner’s internal processing mechanisms. The interactionist view sees language development as the result both of input factors and of innate mechanisms. Language acquisition derives from the collaborative efforts of the learner and his interlocutors and involves a dynamic interplay between external and internal factors. The discussion of the role of the linguistic environment in SLA which is the main purpose of this chapter is conducted largely within the interactionist framework. However, many of the early studies of input and interaction concerned the acquisition of a first language rather than a second language. The next section, therefore, looks at the way mothers talk to young children.
Empirical studies were able to show that the mother’s speech was remarkably well formed, containing few ungrammatical utterances or sentence fragments. Furthermore this speech was characterized by a number of formal adjustments in comparison to speech used in adult—adult conversations.
Snow (1976) lists a number of these: a lower mean length of utterance, the use of sentences with a limited range of grammatical relations, few subordinate and co-ordinate constructions, more simple sentences, the occurrence of tutorial questions (i.e. questions to which the mother already knows the answer), and, overall, a high level of redundancy.
There are also adjustments in pronunciation. Sachs (1977) shows that mothers tune the pitch, intonation, and rhythm to the perceptive sensitivity of the child.
Ferguson (1977) suggests that there are three main functions: (1) an aid to communication, (2) a language teaching aid, and (3) a socialization function. It is the former, however, that motivates motherese. Mothers seek to communicate with their children, and this leads them to simplify their speech in order to facilitate the exchange of meanings.
Mothers pay little attention to the formal correctness of their children’s speech, but instead attend to the social appropriate-ness of their utterances. Brown (1977) describes the primary motivation as ‘to communicate, to understand and to be understood, to keep two minds focused on the same topic’. Thus if motherese also serves to teach language and to socialize the child into the culture of the parents, it does so only indirectly as offshoots of the attempt to communicate.
3 The basis of adjustments made by mothers
Another question concerns how mothers determine the nature and the extent of the modifications which are needed. Gleason and Weintraub (1978) suggest that parents have a general idea of their children’s linguistic ability, particularly their ability to understand, but they lack an accurate knowledge of what specific linguistic features their children have mastered. Parents may internalize a model of a ‘typical’ child of a given age and then adjust their speech upwards and downwards on the basis of feedback from an individual child. Of crucial importance, therefore, is the extent to which the child comprehends what is said to him and the extent to which he signals his comprehension or tack of comprehension to his caretaker.
4 The effects of rnotherese,
In discussing the ways to which the mother’s speech adjustments affect first language acquisition, a distinction needs to be drawn between the route and the rate of acquisition. Little is known about the relationship between motherese and the route of development.
Newport et at. (1977) suggest that input influences only language-specific features of the language being learnt (e.g. verb auxiliaries in English) and that universal aspects (e.g. basic sentence types) ‘proceed in indifference to the details of varying individual environments’. However, other studies have produced rather different results. Furrow et al. (1979) found a large number of strong relationships between measures of mothers’ input and formal measures of their children’s speech.
Ellis and Wells (1980), and Barnes et al. (1983), among others, to suggest that the way mothers talk to their children influences how rapidly they acquire the language. However, the key features of the input appear to be interactional rather than formal.
How does the input affect first language acquisition? Wells (1981: 109) provides the following account: the general principle involved seems to be one of constructing a linguistic representation on the basis of the speech signal that he child) hears, and comparing that with the conceptual representation of the situation to which he believes the spoken message applies, using any available cues to help him along with the task.
132 Understanding Second Language Acquisition
Thus, listening and being able to relate one sensory modality (the aural) to another (the visual) is of central importance. This process is facilitated by the interactional routines in which the mother and child participate. As Ferrier (1978) explains, the child finds himself in routine interactional contexts in which his mother produces a limited set of predictable utterances.
The regularity and invariance of the caretaker’s utterances, together with the accommodations of the caretaker whenever she observes the child in communicative difficulty create activities of shared attention which serve as the basis for the process of modality matching. In contrast, language heard but not addressed directly at the child does not appear to help.
Mapping lessons occur when the adult provides utterances which the child can decode with the help of context, or when the adult expands on what the child has said by using words for different aspects of the context. Segmentation lessons occur when the adult gives incidental clues as to how utterances can be divided up into words, phrases or clauses (e.g. by putting new words in familiar frames like ‘There’s a.
These lessons are, of course, the result of attempts to communicate with the child, net to teach him. Research into motherese, as reflected in the five issues discussed above, led to a reappraisal of the role of the linguistic environment in L 1 acquisition. It came to be seen to serve as far more than just a trigger to activate innate processing mechanisms and led to an interactionist interpretation of development.
Foreigner talk studies
Studies of foreigner talk were stimulated by Ferguson’s (1971) account of simplified registers. This pinpointed the linguistic similarities among motherese, foreigner talk, and also fossilized forms of interlanguage. As in the case of motherese, research and discussion have centered on describing and explaining foreigner talk and, more recently, speculating what role it plays in SLA. In this section I shall concern myself with the description and explanation of foreigner talk, delaying a consideration of its role in SLA until later.
Tire description of foreigner talk in order to describe foreigner talk, it is necessary to collect and analyses samples of speech addressed by native speakers to non-native speakers. Long (1981a) points out that many of the studies have failed to obtain baseline data (i.e. speech between native speakers) to serve as the basis for comparison.
Also some studies where baseline data have been collected have not taken care to ensure that both sets of data were derived from identical tasks. As with any other register, foreigner talk is likely to he influenced by a whole host of variables such as the topic of conversation, the age of the participants (i.e. whether they are children, adolescents, or adults), and, in particular, the proficiency of the learners. Therefore, foreigner talk is not to be thought of as a static, fixed set of features, but as dynamic, changing in accordance with various situational factors. Foreigner talk has both formal and functional characteristics. Long (1981 a) labels these input and interactional features respectively.
The input features are of two types; (1) those that involve simplifications within the grammatical rule structure of the language, and (2) those that involve simplifications leading to ungrammatical speech. Interactional features consist of the specific discourse functions performed by native speakers. These do not differ in kind from those observed in conversations involving just native speakers, but there are differences in the frequency with which specific functions are used.
2 foreigner talk consisting of interactional and grammatical input adjustments (i.e. there are no ungrammatical simplifications);
3 foreigner talk consisting of interactional adjustments as well as both grammatical and ungrammatical input adjustments.
Which type of foreigner talk occurs is the result of various factors concerned with the proficiency of the learner and the role relationships between the participants.
In general (1) appears to be more common than (2), which in turn is more common than (3). A number of studies have considered the factors that influence the degree to which input and interactional adjustments are present in foreigner talk.
Scarcella and Higa (1981) have compared the foreigner talk addressed to child non-native speakers with that addressed to adolescents. They conclude that child non-native speakers receive a simpler input in a more supportive atmosphere.
Long (1983b) suggests that the use of ungrammatical foreigner talk depends on four conditions: (1) the non-native speaker has very low proficiency in the L2; (2) the native speaker thinks he is of a higher status; (3) the native speaker has considerable prior experience of foreigner talk; (4) the conversation occurs spontaneously (i.e., is not part of a laboratory experiment).
Explanations of how native speakers are able to adjust their speech include (1) regression (i.e. the native speaker unconsciously moves back through the stages of development that characterized his own acquisition of the language until he reaches an appropriate level for the person he is addressing) in much the same way as Corder (1981) suggests the L2 learner regresses in the early stages of development; (2) matching (i.e. the native speaker assesses the learner’s language system and then imitates the language forms he identifies in it); and (3) negotiation (i.e. the native speaker simplifies and clarifies in accordance (1983 ) with the feedback that he obtains from the learner)
L2 data are made available to the learner in the input he receives. However, this input is not determined solely by the native speaker. It is also determined by the learner himself. The feedback he provides affects the nature of the subsequent input from the native speaker. Also, as Sharwood-Srnith (1981) notes, the learner’s output serves as input to his own language processing mechanisms. it makes little sense, therefore, to consider the contribution of the native speaker independently of that of the learner. Because one affects the other, it is more fitting to consider the joint work done by native speaker and learner by looking at the discourse they construct together.
The method for undertaking this is known as discourse analysis. There is a further reason for investigating discourse, rather than just foreigner talk. It may shed light on how L2 learners learn. As Hatch (1978c: 403) argues: It is not enough to look at input and to look at frequency; the important thing is to look at the corpus as a whole and examine the interactions that take place within conversations to see how interaction itself determines frequency of forms and how it shows language functions evolving. In other words, Hatch proposes that we need to look at discourse in order to study how language learning evolves out of the strategies used to carry on conversations. Harder (1980: 168) makes a similar point.
Conversations involving child learners
Hatch shows that in child-learner discourse, conversations typically commence with attempts by the child to ‘open the channel’ by calling for the adult’s attention (e.g. ‘oh oh’; ‘look it’). The adult then responds by identifying the object that appears to have attracted the child’s attention and the child repeats the name of the object.
The sequence may end there and a similar ‘nominating’ sequence be embarked upon. Alternatively the conversation might move into the development stage, stimulated by the adult demanding some comment on the nominated topic from the child. This may result in an attempt at elaboration by the child. Further developments can occur if the adult calls for further comments, or requests clarification. The entire discourse pattern is summarized in Figure 6.1. It is very similar to the pattern which has been observed in conversations between children and their mothers in Li acquisition research (see, for instance, Clark and Clark 1977; Wells el al. 1979). This suggests that where child language learners are concerned, there may be general ways of going about building a conversanon,
Conversations involving adult learners
Adult–adult conversations are described by Hatch (1978c). Whereas child–adult sequences follow the ‘here-and-now’ principle (i.e. they refer to objects that are physically present and to ongoing activity), adult conversations are more likely to be rooted in displaced activity. As a result, the adult learner has difficulty in identifying the topic. He resorts to the use of requests for clarification (e.g. ‘huh?’), or to echoing part of the native speaker’s question in order to establish the field of reference. Repair strategies are also common on the native speaker’s part. These involve moving the topic to the beginning or end of the sentence where it is more salient (e.g. ‘Holidays this summer—where are you going?’), simplifying lexis, adding gestures, and translating or switching to foreigner talk. Repair strategies are also found in adult–adult conversations where both of the speakers are L2 users (Schwartz 1980). In these cases conversation is organized in favor of self-repair.
A number of learner strategies that contribute to this end are noted, such as ‘stepping in sideways’ (i.e. allowing the native speaker to introduce a topic and then trying to add some new and relevant information), repetition, changing the topic, and the use of conversation fillers like `ya know’. These strategies were rarely used by the child learners. Krashen (1982) speculates that because older learners are more involved in keeping a conversation going, they may learn more rapidly than younger learners (see Chapter 3).
The negotiation of meaning
A major feature of conversations involving L2 learners is that the learner and native speaker together strive to overcome the communicative difficulties which are always likely to arise as a result of the learner’s limited L2 resources. This has become known as the negotiation of meaning. On the part of the native speaker this involves the use of strategies and tactics (Long 1983a). Strategies are conversational devices used to avoid trouble; examples are relinquishing topic control,
Interaction analysis was initiated in subject classrooms. In the 1960s Flanders (see Flanders 1970) developed a category system for analyzing the communicative uses of the teacher’s and pupils’ language. Later this system was adapted for use in language classrooms (e.g. Moskowitz 1971). More sophisticated systems have been devised by Fanselow (1977) and Allwright 1980). AIlwright, for instance, proposes that classroom interaction be accounted for in terms of three types of analysis:
(1) A turn-taking analysis, which consists of several categories grouped under the general headings of ‘turn-getting’ and `turn-giving’;
(2) A topic analysis, which makes use of such categories as instances of the target language intended primarily as ‘models’ (e.g. the teacher says something for the learner to imitate), and instances of communication concerned primarily with information about the target language;’ and
(3) A task analysis, which does not provide a detailed set of categories, but distinguishes tasks at the levels of turn-taking and topic manage-ment, and also at the cognitive level.
Ailwright’s system requires the researcher to code each utterance in the interaction. He illustrates its use on an extract from a language lesson and shows that it is capable of revealing some interesting information about how learners and teachers use the target language. Long (1980) provides a thorough critique of interaction analysis systems.
The major problems lie in the choice of the variables to be examined (i.e. what features of classroom language to categorize). Long argues that the categories selected arc no more than ‘subjective hunches’, as they are not verified by SLA research.
The study of teacher talk parallels that of foreigner talk, the language that teachers address to L2 learners is treated as a register, with its Specific formal and interactional properties.
The following is’ a summary of the main findings: 1 Formal adjustments occur at all language levels. Gaits found that teachers’ utterances were simpler on a range of measures of syntactic complexity when they addressed pupils than when they were talking among themselves.
Henzl compared the language that teachers used when teaching pupils of different levels of proficiency. He observed adjustments in pronunciation (e.g. with low-level students, the teachers used a more accurate, standard pronunciation), in lexis (e.g. they substituted items with a narrow semantic field like ‘young gal’ by more general words like ‘woman’), and in grammar (e.g. they adjusted the mean length of their utterances).
These modifications mirror those observed in foreigner talk (see Table 6.1). 2 In general, ungrammatical speech modifications do not occur. This is presumably because the conditions that permit deviations from the standard language do not arise in the classroom, However, extreme simplification involving deviant utterances can occur in certain types of classroom interaction such as those found in free discussion (see Hatch, Shapira, and Cough 1978). 3 Interactional adjustments occur.
Gales notes interactional devices in teachers’ speech similar to those observed in motherese (e.g. repetition, prompting, prodding, and expansions). It is likely that many of the interactional adjustments found in other simplified registers will also occur in teacher talk. But there are likely to be differences also.
In summary, teacher talk in language lessons is broadly similar to foreigner talk. However, because special constraints operate in the classroom, there are both formal and interactional differences.
Studies of teacher talk in subject lessons involving 1.2 learners include Chaudron (1983a) and Wcsche and Ready (1983). Both these studies looked at teacher talk in university classrooms. The studies showed a range of speech adaptations similar to those discussed above. For instance, the talk directed at L2 speakers (in comparison to native speaking students) was grammatically simpler (e.g., contained fewer words per clause), was slower, with more and longer pauses, and contained more repetition.
It is possible, however, that fewer lexical adjustments occur, perhaps because the choice of vocabulary is determined by the subject content of the lessons. It should be noted that the adjustments in subject lessons are not motivated by attempts to teach the U (which is a possibility in language lessons), but by the attempt to share information.
An interesting issue in teacher talk is how the teacher determines what level of adjustment to make. Foreigner talk normally occurs in one-to-one interactions where there is plenty of feedback from the learner. Teacher talk occurs in one-to-many interactions, where the learners may vary in their level of proficiency and where there is likely to be only limited feedback from a few students. However, teachers do succeed in varying their adjustments to suit the linguistic competence of the class they are teaching, as Henzl showed. Adjustments are more frequent with beginners than with advanced students. Teachers must gauge the general level of proficiency of a class and then determine the nature and extent of the modifications to make.
Like interaction analysis (but unlike teacher talk studies), discourse analysis considers both the teacher’s and the learner’s contribution. It differs from interaction analysis in that it aims to describe not just the function of individual utterances, but how these utterances combine to form larger discourse units. Also it seeks to account for all the data, avoiding a ‘rag bag’ category for coding awkward utterances which do not fit any of the other categories.
(4) Real communication, which consists of spontaneous natural speech. Communicative breakdown can occur when the teacher and the pupils are in conflict about which type of language use is in operation. For example, the teacher may ask a question designed to practice a specific formal structure, and the pupil may respond as if it were a genuine question.
He proposes a framework for analyzing the various possibilities.
This is based on distinguishing three basic kinds of pedagogic goal:
(1) Core goals, which relate to the explicit pedagogic purpose of the lesson (e.g. to teach specific aspects of the L2, to impart specific subject content, to help the pupils make something);
(2) Framework goals, which relate to the organization requirements of the lesson (e.g. giving out materials, managing pupil behavior); and
(3) Social goals, involving the use of language for more personal purposes (e.g. imparting private information; quarrelling).
In addition Ellis distinguishes types of address (i.e. who functions as speaker, listener, hearer). Classroom discourse can be described in terms of the types of goal and address which occur. Ellis examines the discourse which occurs in elementary ESL classrooms, where English is both the pedagogic target and the medium of instruction.
He demonstrates that a wide variety of interactions rake place. Some of the interactions have a very different pattern from that discussed by Gremmo, Holec and Riley. For example, interactions with framework goals, where the teacher is engaged in organizing classroom.
Input and interaction in classrooms have been investigated by means of interaction analysis, the study of teacher talk, and discourse analysis. Interaction analysis has spawned numerous category systems, some specifically designed for use in language classrooms.
The role of input and interaction in SLA So far the discussion has focused on describing and explaining the different types of input and interaction. The key question, however, is whether SLA is significantly affected by the quality and quantity of the input and interaction and, if so, how. This section will consider this question. It will begin with an account of the different ways in which the question can be tackled. It will then consider how input and interaction affect the route of SLA and, following this, the rate of SLA.
1 Formulaic speech
One way in which interaction can aid SLA is by providing the learner with ready-made chunks of speech which can be memorized as `unanalyzed wholes’. This is likely to occur when the learner participates in routinized interactions involving the use of invariant utterance types by the native speaker. Hatch (1983b) has referred to this type of input as ‘canned speech’. She suggests that adult learners may be able to keep control of topics (i.e. avoid the risk of the conversation taking off into fields the learner is not linguistically competent to handle) by the adroit use of formulae. In this way they serve an immediate communicative purpose.
2 Vertical structures
Vertical structures are learner utterances which are constructed by borrowing chunks of speech from the preceding discourse. Consider the following example from Ellis (1984a): Teacher: Take a look at the next picture. Pupil: Box. Teacher: A box, yes. Pupil: A box bananas. The L2 pupil’s final utterance can be seen to consist of a repetition of the teacher’s preceding utterance. ‘A box’) plus an extra noun. It has been constructed ‘vertically% Scolion (1976) has observed large numbers of similar constructions in first language acquisition. He points out that they are one of the principal means by which the child overcomes the constraints on the length of his utterances imposed by processing limitations.
Another way in which the structure of conversations can influence the route of SLA is by modelling specific grammatical forms which are then subconsciously acquired by the learner. In other words, the first structures the learner acquires arc those to which he is exposed most frequently. Hatch and Wagner-Gough (1975) report a positive correlation between the frequency of various Wh-questions in the input and the order in which they are acquired.
Larsen-Freeman (1976) compared the standard morpheme order for SLA with the frequency of the same morphemes in the parental speech examined by Brown in his study (1973) of first language acquisition. She found a significant correlation. Larsen-Freeman comments: Thus, the tentative conclusion is that morpheme frequency of occurrence in native-speaker speech is the principal determinant for the oral production morpheme accuracy order of ESL learners (1976:
4 Comprehensible input
It is likely that simple ‘exposure’ to input data is not enough. Learners need comprehensible input. Krashen (1981a; 1982) and Long 1983b; 1983c) have argued strongly that SLA is dependent on the availability of comprehensible input before the learner’s internal processing mechanism can work. Krashen presents the case for comprehensible input in the form of the input hypothesis. He argues that for SLA to cake place, the learner needs input that contains exemplars of the language forms which according to the natural order are due to be acquired next.
Krashen and Long’s case for comprehensible input is a strong one. There are, however, problems. First, SLA can take place without two-way communication and hence without interactional modifications. As Larsen-Freeman (1983a) notes, there are cases of successful SLA when the only input is that obtained from reading or watching television.
5 Input and intake
Even if input is understood, it may not be processed by the learner’s internal mechanisms. That is what Krashen meant when he stated that comprehensible input is not a sufficient condition for SLA. It is only when input becomes intake that SLA takes place. Input is the L2 data which the learner hears; intake is that portion of the L2 which is assimilated and fed into the interlanguage system. We know very little about how the learner selects from the input data he receives. Is it to do with the way pieces of input data are presented? Is it to do with socio-affective factors such as motivation? Krashen argues that these act as a ‘filter’, controlling how much input is let in and how much is excluded. Is it to do with the nature of the internal processing mechanisms themselves? The crux of a nativist account of SLA is that it is these mechanisms which regulate the intake, It may be that learners do not respond to the available data on an all-or-nothing basis, either assimilating it or rejecting it. They may attend differentially to features of the input, using some to confirm or disconfirm existing hypotheses, others to form new hypotheses, and keeping others as ‘savings’ so that some ‘trace’ remains which can be worked on later (Hatch 1983b). The discussion of how input becomes intake necessarily involves a consideration of the nature of the internal mechanisms.
Ellis (1984a) suggests that the following features are likely to facilitate rapid development:
1 A high quantity of input directed at the learner.
2 The learner’s perceived need to communicate in the L2.
3 Independent control of the propositional content by the learner (e.g. control over topic choice). 4 Adherence to the ‘here-and-now’ principle, at least initially.
5 The performance of a range of speech acts by both native speaker/ teacher and the learner (i.e. the learner needs the opportunity to listen to and to produce language used to perform different language functions).
6 Exposure to a high quantity of directives.
7 Exposure to a high quantity of `extending’ utterances (e.g. requests for clarification and confirmation, paraphrases and expansions). Opportunities for uninhibited ‘practice’ (which may provide opportunities to experiment using ‘new’ forms). There are strong theoretical grounds for believing that a learning setting rich in these features will lead to successful SLA, but as yet there is little empirical proof.
The study of input and interaction in SLA seeks answers to the following questions: 1 what characteristics are displayed by input and interaction involving 1.2 learners? 2 Arc the characteristics of input and interaction related to SLA, and if so in what way? 3 What are the differential contributions of input and interaction on the one hand and internal processing mechanisms to SLA on the other?
There are fairly clear answers to the first question. We know quite a lot about native speaker input and discourse between native speakers/ teachers and L2 learners. We know much less about the second question. Strong claims have been advanced that SLA is aided by two-way communication in which comprehensible input is provided by means of interactional adjustments. However, two-way communication is not a necessary condition for SLA. Nor is it sufficient. Where question (3) is concerned, we are even less sure.
We do not know what proportion of responsibility to allocate to the linguistic environment, as opposed to the internal mechanisms. It may be that this question cannot be answered, because SLA is jointly determined by factors inside and outside the learner. As Hatch (1983a: 180) writes: While social interaction may give the learner the ‘best’ data to work with, the brain in turn must work out a fitting and relevant model of that input This being so, it is an interactionist view of SLA, rather than a behaviorist or nativist view, that may be most acceptable.