Language Learning and Teaching

     Researchers and teachers of second languages came to realize that the mistakes a person made in this process of constructing a new system of language needed to be analyzed carefully, for they possibly held in them some of the keys to the understanding of the process of second language acquisition. As Corder (1967:167) noted: "A learner's errors ... are significant in [that] they provide to the researcher evidence of how language is learned or acquired, what strategies or procedures the learner is employing in the discovery of the language."

Mistakes and Errors

In order to analyze learner language in an appropriate perspective, it is crucial to make a distinction between mistakes and errors, technically two very different phenomena. A mistake refers to a performance error that is either a random guess or a "slip," in that it is a failure to utilize a known system correctly. All people make mistakes, in both native and second language situations. Native speakers are normally capable of recognizing and correcting such "lapses" or mistakes, which are not the result of a deficiency in competence but the result of some sort of temporary breakdown or imperfection in the process of producing speech. These hesitations, slips of the tongue, random ungrammaticalities, and other performance lapses in native-speaker production also occur in second language speech. Mistakes, when attention is called to them, can be self-corrected.

     Mistakes must be carefully distinguished from errors of a second language learner, idiosyncrasies in the language of the learner that are direct manifestations of a system within which a learner is operating at the time. An error, a noticeable deviation from the adult grammar of a native speaker, reflects the competence of the learner. Learners of English who ask, "Does John can sing?" are in all likelihood reflecting a competence level in which all verbs require a pre-posed do auxiliary for question formation. As such, it is an error, most likely not a mistake, and an error that reveals a portion of the learner's competence in the target language.

     Can you tell the difference between an error and a mistake? Not always. An error cannot be self-corrected, according to James (1998: 83), while mistakes can be self-corrected if the deviation is pointed out to the speaker. But the learner's capacity for self-correction is objectively observable only if the learner actually self-corrects; therefore, if no such self-correction occurs, we are still left with no means to identify error vs. mistake. So, can we turn to frequency of a deviant form as a criterion? Sometimes. If, on one or two occasions, an English learner says "John cans sing," but on other occasions says "John can sing," it is difficult to determine whether "cans" is a mistake or an error. If, however, further examination of the learner's speech consistently reveals such utterances as "John wills go; "John mays come," and so forth, with very few instances of correct third-person singular usage of modal auxiliaries, you might safely conclude that "cans," "mays," and other such forms are errors indicating that the learner has not distinguished modals from other verbs. But it is possible, because of the few correct instances of production of this form, that the learner is on the verge of making the necessary differentiation between the two types of verbs. You can thus appreciate the subjectivity of determining the difference between a mistake and an error in learner speech. That undertaking always bears with it the chance of a faulty assumption on the part of a teacher or researcher.

     The fact that learners do make errors, and that these errors can be observed, analyzed, and classified to reveal something of the system operating within the learner, led to a surge of study of learners' errors, called error analysis. Error analysis became distinguished from contrastive analysis by its examination of errors attributable to all possible sources, not just those resulting from negative transfer of the native language. Error analysis easily superseded contrastive analysis, as we discovered that only some of the errors a learner makes are attributable to the mother tongue, that learners do not actually make all the errors that contrastive analysis predicted they should, and that learners from disparate language backgrounds tend to make similar errors in learning one target language. Errors—overt manifestations of learners' systems—arise from several possible general sources: interlingual errors of interference from the native language, intralingual errors within the target language, the sociolinguistic context of communication, psycholinguistic or cognitive strategies, and no doubt countless affective variables.

Errors in Error Analysis

There is a danger in too much attention to learners' errors. While errors indeed reveal a system at work, the classroom language teacher can become so preoccupied with noticing errors that the correct utterances in the second language go unnoticed. In our observation and analysis of errors—for all that they do reveal about the learner—we must beware of placing too much attention on errors and not lose sight of the value of positive reinforcement of clear, free communication. While the diminishing of errors is an important criterion for increasing language proficiency, the ultimate goal of second language learning is the attainment of communicative fluency.

   Another shortcoming in error analysis is an overemphasis on production data. Language is speaking and listening, writing and reading. The comprehension of language is as important as production. It so happens that production lends itself to analysis and thus becomes the prey of researchers, but comprehension data is equally important in developing an understanding of the process of SLA.

   Over the years, many studies (James 1998;Tarone 1981; Kleinmann 1977; Schachter 1974) have shown that error analysis fails to account for the strategy of avoidance. A learner who for one reason or another avoids a particular sound, word, structure, or discourse category may be assumed incorrectly to have no difficulty therewith. Schachter (1974) found, for example, that it was misleading to draw conclusions about relative clause errors among certain English learners; native Japanese speakers were largely avoiding that structure and thus not manifesting nearly as many errors as some native Persian speakers. The absence of error therefore does not necessarily reflect nativelike competence because learners may be avoiding the very structures that pose difficulty for them.

   Finally, error analysis can keep us too closely focused on specific languages rather than viewing universal aspects of language. Gass (1989) recommended that researchers pay more attention to linguistic elements that are common to all languages. The language systems of learners may have elements that reflect neither the target language nor the native language, but rather a universal feature of some kind. Such assertions are in keeping with the bioprogramming theories referred to in Chapter 2. But there are problems, of course, with the search for universal properties of learner's errors. "It is not at all clear in any precise way when the influence of the universal will appear in the interlanguage of learners rather than a violation of it based on influence from either the source or target language" (Celce-Murcia & Hawkins 1985:66).

   We do well, therefore, in the analysis of learners' errors, to engage in "performance analysis" or "interlanguage analysis" (Celce-Murcia & Hawkins 1985: 64), a less restrictive concept that places a healthy investigation of errors within the larger perspective of the learner's total language performance. While a significant portion of this chapter deals with error analysis, let us nevertheless remember that production errors are only a subset of the overall performance of the learner.

Identifying and Describing Errors

     One of the common difficulties in understanding the linguistic systems of both first and second language learners is the fact that such systems cannot be directly observed. They must be inferred by means of analyzing production and comprehension data. What makes the task even thornier is the instability of learners' systems. Systems are in a constant state of flux as new information flows in and, through the process of subsumption, causes existing structures to be revised. Repeated observations of a learner will often reveal apparently unpredictable or even contradictory data. In undertaking the task of performance analysis, the teacher and researcher are called upon to infer order and logic in this unstable and variable system.

     The first step in the process of analysis is the identification and description of errors. Corder (1971) provided a model for identifying erroneous or idiosyncratic utterances in a second language. That model is schematized in Figure 8.1. According to Corder's model, any sentence uttered by the learner and subsequently transcribed can be analyzed for idiosyncrasies. A major distinction is made at the outset between overt and covert errors. Overtly erroneous utterances are unquestionably ungrammatical at the sentence level. Covertly erroneous utterances are grammatically well-formed at the sentence level but are not interpretable within the context of communication. Covert errors, in other words, are not really covert at all if you attend to surrounding discourse (before or after the utterance). "I'm fine, thank you" is grammatically correct at the sentence level, but as a response to "Who are you?" it is obviously an error. A simpler and more straightforward set of terms, then, would be "sentence level" and "discourse level" errors.

     Corder's model in Figure 8.1 indicates that, in the case of both overt and covert errors, if a plausible interpretation can be made of the sentence, then one should form a reconstruction of the sentence in the target language, compare the reconstruction with the original idiosyncratic sentence, and then describe the differences. If the native language of the learner is known, the model indicates using translation as a possible indicator of native language interference as the source of error. In some cases, of course, no plausible interpretation is possible at all, and the researcher is left with no analysis of the error (OUT^).

     Consider the following examples of idiosyncratic utterances of learners, and let us allow them to be fed through Corder's procedure for error analysis:

1. Does John can sing?

A. NO С YES

D. Can John sing?

E. Original sentence contained pre-posed do auxiliary applicable to most verbs, but not to verbs with modal auxiliaries. OUT₂

2. I saw their department.

A. YES

B. NO (context was in a conversation about living quarters in Mexico)

С NO

E YES, Spanish.

G. Yo vi su departamento.YES

H. I saw their apartment.

E.Departamento was translated to false cognate department. OUT₂

3. The different city is another one in the another two.

A. NO

С NO

F.  YES, Spanish.

G. No plausible translation or interpretation. 
I. No analysis. OUT3

     It can be seen that the model is not complicated and represents a procedure that teachers and researchers might intuitively follow. Of course, once an error is identified, the next step is to describe it adequately, something the above procedure has only begun to accomplish.

     A number of different categories for description of errors have been identified in research on learner language (for an overview, see Lennon 1991). 

1. The most generalized breakdown can be made by identifying errors of addition, omission, substitution, and ordering, following standard mathematical categories. In English a do auxiliary might be added (Does can be sing?), a definite article omitted (I went to movie), an item substituted (I lost my road), or a word order confused (I to the store went). But such categories are clearly very generalized.

2. Within each category, levels of language can be considered: phonology or orthography, lexicon, grammar, and discourse. Often, of course, it is difficult to distinguish different levels of errors. A word with a faulty pronunciation, for example, might hide a syntactic or lexical error. A French learner who says "[zhey] suis alle a l'ecole" might be mispronouncing the grammatically correct "je," or correctly pronouncing a grammatically incorrect "j'ai."

3. Errors may also be viewed as either global or local (Burt & Kiparsky 1972). Global errors hinder communication; they prevent the hearer from comprehending some aspect of the message. For example, "Well, it's a great hurry around," in whatever context, may be difficult or impossible to interpret. Local errors do not prevent the message from being heard, usually because there is only a minor violation of one segment of a sentence, allowing the hearer/reader to make an accurate guess about the intended meaning. "A scissors," for example, is a local error. The global-local distinction is discussed in the vignette at the end of this chapter.
4. Finally, Lennon (1991) suggests that two related dimensions of error, domain and extent should be considered in any error analysis. Domain is the rank of linguistic unit (from phoneme to discourse) that must be taken as context in order for the error to become apparent, and extent is the rank of linguistic unit that would have to be deleted, replaced, supplied, or reordered in order to repair the sentence. Lennon's categories help to operationalize Corder's overt-covert distinction discussed above. So, in the example just cited above, "a scissors," the domain is the phrase, and the extent is the indefinite article.

Sources of Error

Having examined procedures of error analysis used to identify errors in second language learner production data, our final step in the analysis of erroneous learner speech is that of determining the source of error. Why are certain errors made? What cognitive strategies and styles or even personality variables underlie certain errors? While the answers to these questions are somewhat speculative in that sources must be inferred from available data, in such questions lies the ultimate value of learner language analysis in general. By trying to identify sources we can take another step toward understanding how the learner's cognitive and affective processes relate to the linguistic system and to formulate an integrated understanding of the process of second language acquisition.

Interlingual Transfer

As we have already seen, interlingual transfer is a significant source of error for all learners. The beginning stages of learning a second language are especially vulnerable to interlingual transfer from the native language, or interference. In these early stages, before the system of the second language is familiar, the native language is the only previous linguistic system upon which the learner can draw. We have all heard English learners say "sheep" for "ship," or "the book of Jack" instead of "Jack's book"; French learners may say "Je sais Jean" for "Je connais Jean," and so forth. All these errors art attributable to negative interlingual transfer. While it is not always clear that an error is the result of transfer from the native language, many such errors are detectable in learner speech. Fluent knowledge or even familiarity with a learner's native language of course aids the teacher in detecting and analyzing such errors.

     The learning of a third language (and subsequent languages) pro vides an interesting context for research. Depending upon a number of factors, including the linguistic and cultural relatedness of the languages and the context of learning, there are varying degrees of interlingual interference from both the first and second language to the third language, especially if the second and third languages are closely related or the learner is attempting a third language shortly after beginning a second language.

Intralingual Transfer

     One of the major contributions of learner language research has been its recognition of sources of error that extend beyond interlingual errors in learning a second language. It is now clear that intralingual transfer (within the target language itself) is a major factor in second language learning. In Chapter 4 we discussed overgeneralization, which is the negative counterpart of intralingual transfer. Researchers (see Jaszczolt 1995; Taylor 1975) have found that the early stages of language learning are characterized by a predominance of interference (interlingual transfer), but once learners have begun to acquire parts of the new system, more and more intralingual transfer—generalization within the target language—is manifested. This of course follows logically from the tenets of learning theory. As learners progress in the second language, their previous experience and their existing subsumers begin to include structures within the target language itself.

     Negative intralingual transfer, or overgeneralization, has already been illustrated in such utterances as "Does John can sing?" Other examples abound—utterances like "He goed," "I don't know what time is it," and "Il a tombe." Once again, the teacher or researcher cannot always be certain of the source of an apparent intralingual error, but repeated systematic observations of a learner's speech data will often remove the ambiguity of a single observation of an error.

  The analysis of intralingual errors in a corpus of production data can become quite complex. For example, in Barry Taylor's (1975:95) analysis of English sentences produced by ESL learners, just the class of errors in producing the main verb following an auxiliary yielded nine different types of error: 

Table 8.1. Typical English intralingual errors in the use of articles (from Richards 1971: 187) 

1. Past-tense form of verb following a modal
2. Present-tense -s on a verb following a modal
3. -ing on a verb following a modal
4. яге (for be) following will
5. Past-tense form of verb following do
6. Present-tense -s on a verb following do
7. -ing on a verb following й?о

8. Past-tense form of a verb following be (inserted to replace a modal or do)

9. Present-tense -5 on a verb following be (inserted to replace a modal or do)