Interlanguage¶
Interlanguage treats a language learner's evolving production as evidence of a patterned system of its own, not merely target-language errors.
Core Idea¶
Interlanguage is the developing linguistic system inferred from a person learning and using an additional language. Learner speech or writing is not just a list of departures from a target standard: recurring forms can reveal provisional regularities in sound, grammar, vocabulary, meaning and use. This system differs from both previously known languages and the target variety, while drawing on both and changing through interaction, instruction and use.[1][2]
“System” does not mean perfect internal uniformity. Forms can vary by task, interlocutor, context or developmental moment, and a temporary plateau is not proof of permanent fossilization. Selinker's original adult-learner hypothesis and later, broader uses should be distinguished when making claims about ages or outcomes.[2][3]
Structural Signature¶
- Learner and prior repertoire: a person brings one or more existing languages and communicative practices.
- Additional-language project: a target variety and situations of use supply input and possible comparison points.
- Observable learner forms: repeated speech or writing provides evidence for patterns rather than a direct readout of hidden competence.
- Developing organization: phonological, lexical, grammatical or pragmatic tendencies form an analyzable learner system.
- Shaping processes: transfer, overgeneralization, instruction, learning and communication strategies may contribute; no single cause is mandatory for each form.
- Variation and change: patterns can be reorganized, context-sensitive or relatively stable; permanence requires stronger evidence than persistence alone.[2][3]
Condensed: prior repertoire + additional-language experience → patterned learner production → provisional system hypothesis → test across contexts and time.
Sig role-phrases: learner repertoire; additional-language setting; observed forms; provisional rule; contextual and longitudinal test.
What It Is Not¶
- Not a random error collection. Recurrent learner forms may be patterned even when they depart from a particular target norm.[2]
- Not a literal third language at a fixed midpoint. It is a construct for a learner's changing repertoire, not a measured position halfway between two complete languages.
- Not necessarily internally invariant. Systematicity and contextual variability can coexist; longitudinal study may reveal changing distribution of forms.[3]
- Not identical to first-language transfer. Prior-language influence is one possible shaping process among several.[2]
- Not inevitably fossilized. Selinker's strong claim about universal fossilization is controversial; a stable form may later change, and stabilization can differ among subsystems.[2]
- Not second-language acquisition itself. Acquisition is the wider developmental process; interlanguage names a learner-system construct within its analysis.
Scope of Application¶
In instructed language learning, repeated past-tense forms may show how a learner extends a productive rule, then reorganizes it as exceptions become available. A form like goed can illustrate overgeneralization, but a single instance alone cannot establish a learner's grammar or identify their prior language as its cause. Compare uses across tasks and dates rather than interpreting every non-target form as an unrelated mistake.[2]
In naturalistic additional-language use, an adult may use a new language differently at work and in community conversation. A researcher can ask whether variation follows interlocutor, communicative task or time, instead of presuming one uniform grammar or permanent arrest. The target is an analyzable system of production in meaningful use, not merely classroom test performance.[2][3]
The original hypothesis focused on adults, while later research extended the construct to other learner populations. The expansion does not license treating all ages, settings or target varieties as empirically identical.[2][4]
Clarity¶
The construct changes the question from “How many errors are there?” to “What pattern would make these uses intelligible, and when does the pattern hold?” Target-language comparison can still identify a difference, but it cannot alone explain why the learner repeatedly uses one form and not another. A learner may also know a form yet use it differently under time pressure or unfamiliar social demands; performance is evidence requiring interpretation, not an unfiltered transcript of a single mental rule.[2]
The term fossilization names a proposed relative stability of some non-target patterns. It should not be inferred just because a form appears in two observations, nor treated as the inevitable fate of every interlanguage.
Manages Complexity¶
Language learning produces many changing forms across phonology, morphology, syntax, vocabulary and pragmatics. Interlanguage groups them as evidence of an evolving organization while retaining the contexts that shape use. This allows a longitudinal record to track regularities, exceptions and restructuring without demanding that every variation be random noise or that every learner match one idealized target variety.[3]
The compression has a cost: a neat hypothesized grammar can conceal task effects, multilingual resources or social variation. Testing across settings and time is part of using the abstraction responsibly, not an optional afterthought.
Abstract Reasoning¶
Collect multiple samples of meaningful learner production. Identify candidate regularities, then compare how they appear across forms, contexts and time. Consider possible contributions from prior languages, overgeneralization of new patterns, instruction or communication strategies, without deducing a cause from one surface form. Ask what observation would refute a proposed rule: a contextual contrast, a later reorganization, or a different pattern in comprehension. If a form persists, describe its stability and sampling limits before calling it fossilized.[2][3]
This reasoning infers a system from evidence; it does not imply every learner consciously knows or can state the rules they use.[2]
Knowledge Transfer¶
The learner-system lens transfers across target languages, classroom and naturalistic settings, and linguistic subsystems because each presents patterned use under development. It does not transfer as a universal fixed sequence of stages, a guaranteed first-language-transfer account, or a verdict that all learners stop below native-speaker norms. Broader developmental fields may similarly infer changing systems from behavior, but that analogy is a possible higher-order abstraction, not a reason to relabel every learning process as an interlanguage.
Examples¶
French immersion: three forms for one meaning¶
Selinker's author-published collection reproduces the French-immersion study and reports one child's three ways of expressing “I like” within a two-minute taped segment. Its searchable text gives J'ai aime and J'aime and begins a third numbered form with Je, but drops the remainder; the exact third spelling is therefore not verified here. The first two transcriptions alone show a non-target and a target-like realization of the same intended meaning in the same short communicative setting, while the source reports a third variant. Merely counting an error would miss the learner's alternation of one construction. Conversely, this short record does not establish which grammatical rule, discourse context or monitoring state selected each form. A defensible interlanguage hypothesis here is variable realization of the first-person “like” construction, to be tested by more tokens and surrounding contexts; an invariant single form is already contradicted. This is an attested child-immersion observation, not an author-made transcript or proof of an adult-universal developmental path.[5][2]
Mapped back: English-speaking French-immersion learner → meaningful French production → three printed variants in two minutes → provisional variable-construction system → insufficient tokens for cause or fossilization.
Adult picture descriptions: two syllable-repair distributions¶
Tarone's original study recorded six adults learning English as they described a picture sequence: two each with Cantonese, Portuguese and Korean first languages. Rather than reducing the speech to “pronunciation errors,” she coded syllable-structure deviations as consonant deletion, vowel epenthesis or glottal-stop insertion. About one fifth of attempted syllables per participant had a structural deviation. Her Table B reports that the two Korean speakers used deletion in roughly 70–80% of their deletion/epenthesis repairs, while the two Portuguese speakers favored epenthesis in roughly 80–90%; Cantonese speakers also favored deletion. Thus a strategy distribution, not one isolated mispronounced word, is the learner-system evidence. The cross-background difference supports a qualified prior-language influence, while both repair strategies simplify difficult syllables toward easier forms; Tarone also warns that individual error origins are often uncertain. The percentages describe this small picture-description sample, not every speaker of those languages, and they do not show a universal or permanent learner grammar.[6]
Mapped back: six adult English learners with three prior-language backgrounds → picture-description task → coded syllable repairs and Table B contrast → provisional group-pattern inference → sample-size and causal-attribution limits.
Structural Tensions¶
The cited learner-system work does not establish an intrinsic two-sided cost tension between systematicity and variability. A context-conditioned model can preserve both: recurrent forms may be rule-governed while their distribution changes with task and development. An overly uniform model and a purely random account are two inference errors, not ends of a single tradeoff. Diagnostic: does a proposed rule predict the observed distribution across contexts and dates?[3]
Target comparison and learner-system analysis are likewise complementary lenses. Target-language norms identify contrasts; describing the learner's recurring forms on their own terms can reveal organization that an error-only inventory misses. Neither lens requires rejecting the other. Diagnostic: what recurs when forms are described before being graded correct or incorrect, and how does that recurrence relate to the target and task?[2]
Structural–Framed Character¶
Interlanguage mixes a structural claim—the existence of analyzable patterned production—with a framed research choice about the learner, target variety, task, and relevant notion of competence. The claim is not a moral ranking of speakers, but the chosen comparison norm and data-collection practice matter. It arose in a named applied-linguistics research program; the historical origin marks its theoretical scope rather than guaranteeing that every version of the hypothesis is true. Its vocabulary travels legitimately across additional languages and learner populations when the patterned-system inference is preserved; calling any intermediate skill “an interlanguage” is only a metaphorical import. Its character: a theory-bound, domain-specific construct of learner linguistic organization.
Structural Core vs. Domain Accent¶
The general skeleton is infer a revisable internal organization from repeated variable behavior during learning. The domain accent makes that behavior specifically linguistic: prior repertoires, an additional-language target, utterance patterns, and context-sensitive language use. Remove those roles and the interlanguage identity no longer applies. The live Second-Language Acquisition entry describes the broader developmental process and mentions this learner-system role, but the reviewed identities do not establish a necessary typed dependency; “an interlanguage is a kind of acquisition process” would be a category error. A broader learning-system prime is a future question, not established here.
Instantiates / Related Primes¶
This identity is an approved provisional unparented root. Live Second-Language Acquisition is a process in which the learner system may develop, not a strict genus of the inferred linguistic system itself. An error inventory is likewise evidence about the system rather than its genus. No necessary typed dependency or parent edge is asserted here.
Neighborhood in Abstraction Space¶
Interlanguage sits in a sparse region of the domain-specific corpus (92nd percentile for distinctiveness): few abstractions share its structure, so a faithful description tends to retrieve it precisely.
Family — Unclustered & Miscellaneous (2551 abstractions)
Nearest neighbors
- Second-language acquisition — 0.83
- Processability theory — 0.80
- Diglossia — 0.79
- Generative Linguistics — 0.79
- Natural approach — 0.78
Computed from structural-signature embeddings · 2026-10-08
Not to Be Confused With¶
Language transfer is possible influence from a previously known language, not the whole learner system. Fossilization is a disputed stability claim about some patterns, not the definition of interlanguage. Error analysis can count target deviations but need not infer a developing organization. Second-language acquisition is the encompassing process that can produce and reshape interlanguage patterns.
References¶
[1] Larry Selinker, “Interlanguage”, International Review of Applied Linguistics in Language Teaching 10 (1972), 209–231. Original publisher record checked; full text unavailable in this pass. registry ↩
[2] Elaine Tarone, “Interlanguage,” specialist reference account, on original adult scope, hypothesized processes, variation and contested universal fossilization. registry ↩a ↩b ↩c ↩d ↩e ↩f ↩g ↩h ↩i ↩j ↩k ↩l ↩m ↩n
[3] Richard Towell, Roger Hawkins and Nives Bazergui, “Systematic and Nonsystematic Variability in Advanced Language Learning”, Studies in Second Language Acquisition 15 (1993), 439–460. Original longitudinal-study abstract checked. registry ↩a ↩b ↩c ↩d ↩e ↩f ↩g
[4] Beverly Olson Flanigan, “Variable competence and performance in child second language acquisition”, Second Language Research 7 (1991). Original study abstract checked. registry ↩
[5] Larry Selinker, Papers in Interlanguage, 1988 author-published revised collection, chapter “Systematicity/Variability and Stability/Instability in Interlanguage Systems,” p. 46 (PDF p. 50), two-minute French-immersion tape example. This collection is available in full; the 1972 original article remains uninspected. registry ↩
[6] Elaine E. Tarone, “Some Influences on Interlanguage Phonology”, original 1976 study, Method/Results pp. 9–15 and Tables A–B (PDF pp. 12–14, 24). The scan is marginally legible; group directions are corroborated by the author's prose. registry ↩