Through the descriptive study of learner language, some SLA researchers seek to better understand language learning without

recourse to factors outside learner language. Researchers may adopt an interlanguage perspective, exploring learner language as a linguistic system, or they may study how learner language compares to the target language. Research is centered on the question: What are the unique characteristics of learner language? Much of the research has focused on the English language as the L2, because of the huge number of people around the world learning and teaching it.

Error analysis
The field of error analysis in SLA was established in the 1970s by S. P. Corder and colleagues. A widelyavailable survey can be found in chapter 8 of Brown, 2000. Error analysis was an alternative to contrastive analysis, an approach influenced by behaviorism through which applied linguists sought to use the formal distinctions between the learners' first and second languages to predict errors. Error analysis showed that contrastive analysis was unable to predict a great majority of errors, although its more valuable aspects have been incorporated into the study of language transfer. A key finding of error analysis has been that many learner errors are produced by learners making faulty inferences about the rules of the new language.

Error analysts distinguish between errors, which are systematic, and mistakes, which are not. They often seek to develop a typology of errors. Error can be classified according to basic type: omissive, additive, substitutive or related to word order. They can be classified by how apparent they are: overt errors such as "I angry" are obvious even out of context, whereas covert errors are evident only in context. Closely related to this is the classification according to domain, the breadth of context which the analyst must examine, and extent, the breadth of the utterance which must be changed in order to fix the error. Errors may also be classified according to the level of language: phonological errors, vocabulary or lexical errors, syntactic errors, and so on. They may be assessed according to the degree to which they interfere with communication: global errors make an utterance difficult to understand, while local errors do not. In the above example, "I angry" would be a local error, since the meaning is apparent.

From the beginning, error analysis was beset with methodological problems. In particular, the above typologies are problematic: from linguistic data alone, it is often impossible to reliably determine what kind of error a learner is making. Also, error analysis can deal effectively only with learner production (speaking and writing) and not with learner reception (listening and reading). Furthermore, it cannot account for learner use of communicative strategies such as avoidance, in which learners simply do not use a form with which they are uncomfortable. For these reasons, although error analysis is still used to investigate specific questions in SLA, the quest for an overarching theory of learner errors has largely been abandoned. In the mid-1970s, Corder and others moved on to a more wide-ranging approach to learner language, known as interlanguage.

Error analysis is closely related to the study of error treatment in language teaching. Today, the study of errors is particularly relevant for focus on form teaching methodology.

Interlanguage
Interlanguage scholarship seeks to understand learner language on its own terms, as a natural language with its own consistent set of rules. Interlanguage scholars reject, at least for heuristic purposes, the view of learner language as merely an imperfect version of the target language. Interlanguage is perhaps best viewed as an attitude toward language acquisition, and not a distinct discipline. By the same token, interlanguage work is a vibrant microcosm of linguistics. It is possible to apply an interlanguage perspective to learners' knowledge of L2 sound systems (interlanguage phonology), and language-use norms found among learners (interlanguage pragmatics).

By describing the ways in which learner language conforms to universal linguistic norms, interlanguage research has contributed greatly to our understanding of linguistic universals in SLA. See below, under "linguistic universals".

Developmental patterns

Ellis (1994) distinguished between "order" to refer to the pattern in which different language features are acquired and "sequence" to denote the pattern by which a specific language feature is acquired.

Order of acquisition
Researchers have found a very consistent order in the acquisition of first language structures by children, and this has drawn a great deal of interest from SLA scholars. Considerable effort has been devoted to testing the "identity hypothesis", which asserts that first-language and second-language acquisition conform to the same patterns. This has not been confirmed, probably because second-language learners' cognitive and affective states are so much more advanced. However, orders of acquisition in SLA do often resemble those found in first language acquisition, and may have common neurological causes.

Most learners begin their acquisition process with a "silent period", in which they speak very little if at all. For some this is a period of language shock, in which the learner actively rejects the incomprehensible input of the new language. However, research has shown that many "silent" learners are engaging in private speech (sometimes called "self-talk"). While appearing silent, they are rehearsing important survival phrases and lexical chunks. These memorized phrases are then employed in the subsequent period of formulaic speech. Whether by choice or compulsion, other learners have no silent period and pass directly to formulaic speech. This speech, in which a handful of routines is used to accomplish basic purposes, often shows few departures from L2 morphosyntax. It eventually gives way to a more experimental phase of acquisition, in which the semantics and grammar of the target language are simplified and the learners begin to construct a true interlanguage.

The nature of the transition between formulaic and simplified speech is disputed. Some, including Krashen, have argued that there is no cognitive relationship between the two, and that the transition is abrupt. Thinkers influenced by recent theories of the lexicon have preferred to view even native speaker speech as heavily formulaic, and interpret the transition as a process of gradually developing a broader repertoire of

chunks and a deeper understanding of the rules which govern them. Some studies have supported both views, and it is likely that the relationship depends in great part on the learning styles of individual learners.

A flurry of studies took place in the 1970s, examining whether a consistent order of morpheme acquisition could be shown. Most of these studies did show fairly consistent orders of acquisition for selected morphemes. For example, among learners of English the cluster of features including the suffix "-ing", the plural, and the copula were found to consistently precede others such as the article, auxiliary, and third person singular. However, these studies were widely criticized as not paying sufficient attention to overuse of the features (idiosyncratic uses outside what are obligatory contexts in the L2), and sporadic but inconsistent use of the features. More recent scholarship prefers to view the acquisition of each linguistic feature as a gradual and complex process. For that reason most scholarship since the 1980s has focused on the sequence, rather than the order, of feature acquisition.

Sequence of acquisition
A number of studies have looked into the sequence of acquisition of pronouns by learners of various IndoEuropean languages. These are reviewed by Ellis (1994), pp. 96-99. They show that learners begin by omitting pronouns or using them indiscriminately: for example, using "I" to refer to all agents. Learners then acquire a single pronoun feature, often person, followed by number and eventually by gender. Little evidence of interference from the learner's first language has been found; it appears that learners use pronouns based entirely on their inferences about target language structure.

Studies on the acquisition of word order in German have shown that most learners begin with a word order based on their native language. This indicates that certain aspects of interlanguage syntax are influenced by the learners' first language, although others are not.

Research on the sequence of acquisition of words is exhaustively reviewed by Nation (2001). Kasper and Rose (2002) have thoroughly researched the sequence of acquisition of pragmatic features. In both fields, consistent patterns have emerged and have been the object of considerable theorizing.

Variability
Valid though the interlanguage perspective may be, which views learner language as a language in its own right, this language varies much more than native-speaker language, in an apparently chaotic way. A learner may exhibit very smooth, grammatical language in one context and uninterpretable gibberish in another. Scholars from different traditions have taken opposing views on the importance of this phenomenon. Those who bring a Chomskyan perspective to SLA typically regard variability as nothing more than "performance errors", and not worthy of systematic inquiry. On the other hand, those who approach it from a sociolinguistic or psycholinguistic orientation view variability as a key indicator of how the situation affects learners' language use. Naturally, most research on variability has been done by those who presume it to be meaningful.

Research on variability in learner language distinguishes between "free variation", which takes place even within the same situation, and "systematic variation", which correlates with situational changes. Of course, the line between the two is often subject to dispute.

Free variation, variation without any determinable pattern, is itself highly variable from one learner to another. To some extent it may indicate different learning styles and communicative strategies. Learners that favor high-risk communicative strategies and have an other-directed cognitive style are more likely to show substantial free variation, as they experiment freely with different forms.

Free variation in the use of a language feature is usually taken as a sign that it has not been fully acquired. The learner is still trying to figure out what rules govern the use of alternate forms. This type of variability seems to be most common among beginning learners, and may be entirely absent among the more advanced.

Systematic variation is brought about by changes in the linguistic, psychological, social context. Linguistic factors are usually extremely local. For instance, the pronunciation of a difficult phoneme may depend on whether it is to be found at the beginning or end of a syllable.

Social factors may include a change in register or the familiarity of interlocutors. In accordance with communication accommodation theory, learners may adapt their speech to either converge with, or diverge from, their interlocutor's usage.

The most important psychological factor is usually taken to be planning time. As numerous studies have shown, the more time that learners have to plan, the more regular and complex their production is likely to be. Thus, learners may produce much more target-like forms in a writing task for which they have 30 minutes to plan, than in conversation where they must produce language with almost no planning at all.

Affective factors also play an important role in systematic variation. For example, learners in a stressful situation (such as a formal exam) may exhibit much less target-like forms than they would in a comfortable setting. This clearly interacts with social factors, and attitudes toward the interlocutor and topic also play important roles.