Foreign language learning contrasts with native language development in two key respects: It is unreliable and it is nonconvergent. At the same time, it is clear that foreign languages are languages. The fundamental difference hypothesis (FDH) was introduced as a way to account for the general characteristics of foreign language learning. The FDH was originally formulated in the context of the theory of rich Universal Grammar, and this theory has guided much foreign language acquisition research over the past two decades. However, advances in the understanding of language have undermined much of the supporting framework.

The FDH—indeed all of SLA research—must be rethought in light of these advances. It is proposed here that (a) foreign language grammars make central use of patches, which are also seen as peripheral phenomena in native languages; (b) non-domain-specific processes are used in foreign language acquisition, but that these are also employed—although more effectively because they are integrated into the language system—by native language development; and (c) foreign language online processing relies heavily on the use of shallow parses, but these are also available in native language processing, although less crucially.

This article examines how changes in linguistic theory affect the debate between the fundamental difference hypothesis and the access-to-Universal Grammar (UG) approach to SLA. With a focus on subjacency (Chomsky, 1973), a principle of UG that places constraints on wh-movement and that has frequently been taken as a test case for verifying second language (L2) access to UG, we reanalyze earlier L2 findings in terms of a revised constraint, which effectively prohibits extraction out of subjects and adjuncts. We show that L2 learners indeed observe such a constraint on wh-movement, and, at the same time, we suggest that recent claims for a universal computational system (e.g., Chomsky, 1995; Uriagereka, 1999) make the respective roles of the first language and UG difficult to tease apart.

The fundamental difference hypothesis (FDH), as formulated by Bley-Vroman (1990), claims that SLA tends to be nonconvergent because domain-specific linguistic mechanisms available in early childhood cannot be used for language learning in adulthood: Instead, second language (L2) learners deploy domain-general problem solving skills. I claim that nonconvergence is also true of some cases of unbalanced early bilingualism, when the target language was acquired in childhood, suggesting that nontarget attainment in these cases may have different roots—namely, inefficient learning mechanisms in L2 learners but insufficient input in early bilinguals. The FDH then predicts that early bilinguals should still be more successful at attaining nativelike knowledge than the L2 learners due to their early age of acquisition. This review article examines this prediction in light of recent studies of postpuberty L2 learners and unbalanced early bilinguals with nonconvergent knowledge of their first language. I show that the incidence of nativelike achievement is higher in the early bilingual groups than in the L2 groups and that differential performance by the two populations on different tasks lends some support to the processing claims of the FDH.

This article reexamines Bley-Vroman’s original (1990) and evolved (this issue) fundamental difference hypothesis that argues that differences in path and endstate of first language acquisition and adult foreign language learning result from differences in the acquisition procedure (i.e., language faculty and cognitive strategies, respectively). The evolved assessment of the theoretical and empirical developments of the past 20 years is taken into account with respect to Universal Grammar and parameters in generative theory and with respect to cognition and acquisition in data processing. This article supports the spirit of Bley-Vroman’s proposals in light of the discussion of three topics: pathway of acquisition, endstate age of acquisition effects, and language processing by monolinguals and bilinguals. I argue that the difference between child and adult language acquisition is, above all, quantitative not qualitative, a gradient continuum rather than a precipitous break.

This article presents evidence that supports the claim that second language (L2) grammars arise in a domain-specific, informationally encapsulated module with contents provided by Universal Grammar and enriched by native language knowledge, as entertained by Schwartz (1986, 1987, 1999) contra Bley-Vroman (1990). I consider state-of-the-art evidence representative of a body of research on the poverty of the stimulus (POS) that argues for the domain-specificity of L2 representations, with a main focus on interpretation. Then I examine interpretive evidence relevant to the role of informational encapsulation and compositionality in SLA. I seek to demonstrate that the acquisition of syntax-linked interpretive properties where the POS is severe provides opportunities for a type of fingerprinting of mental organization that can inform a variety of epistemologically relevant questions.

The fundamental difference hypothesis (FDH; Bley-Vroman, 1989, 1990) contends that the nature of language in natives is fundamentally different from the nature of language in adult nonnatives. This study tests the FDH in two ways: (a) via second language (L2) poverty-of-the-stimulus (POS) problems (e.g., Schwartz & Sprouse, 2000) and (b) via a comparison between adult and child L2 learners, whose first language (L1) is the same, in terms of their developmental route (e.g., Schwartz, 1992, 2003). The phenomena under investigation are Korean wh-constructions with negative polarity items (NPIs). Korean has subject (S)-object (O)-verb (V) as its canonical word order and it is also a wh-in-situ language, but scrambling of the object to presubject position (i.e., movement that results in OSV word order) is generally optional; however, in the context of negative questions with a NPI subject (e.g., amwuto “anyone”), (a) object wh-phrases must scramble on the wh-question reading and (b) the nonscrambled order has a yes/no-question reading. These properties of Korean wh-constructions with NPIs constitute POS problems for nonnatives whose L1 is English (as well as for native Korean-acquiring children). L1-English adult L2 learners (n = 15) and L1-English child L2 learners (n = 10), independently assessed for Korean proficiency, as well as L1-Korean children (n = 23) and L1-Korean adults (n = 15) completed an elicited-production task, an acceptability-judgment task, and an interpretation-verification task. The results show that (a) high-proficiency (adult and child) L2 learners performed like the native adult controls on all three tasks, thereby demonstrating L2 POS effects; and (b) adult and child L2 learners follow the same (inferred) route to convergence, a route differing from—yet subsuming—the L1-child route. Both sets of results lead us to conclude that, contra the FDH, the nature of language is fundamentally similar in natives and (adult or child) nonnatives.