Abstract: In the field of second language acquisition (SLA), a core theoretical issue concerns how second language (L2) lexical representations develop from a first language (L1)-dependent “parasitic” state into a relatively autonomous and independent representational system. To elucidate the cognitive mechanisms underlying this “parasitic-to-independent” developmental continuum, the present review synthesizes classic empirical phenomena and modulating factors documented in L2 lexical acquisition, and systematically contrasts and evaluates the explanatory scope and inherent limitations of six connectionist models of L2 vocabulary learning in accounting for these phenomena.
Previous research has documented some classic L2 processing difficulties, such as age-of-acquisition effects, cross-language competition, asymmetric priming, persistent first-language (L1) interference, and early reliance on formal rather than semantic information. Existing models explained these phenomena through varied aspects, including bilingual words recognition and competition, lexical development, computational simulations, and neurocognitive accounts. Nevertheless, the shared cognitive mechanisms underlying these empirical phenomena have yet to be systematically investigated and thoroughly elucidated. The main innovation of this study is to align these approaches within a unified connectionist framework and to reinterpret the parasitic-to-independent transition as a process of distributed weight reconfiguration in the bilingual lexical network.
Building on Hall's (2002) Parasitic Hypothesis of Vocabulary Development, the paper proposes that newly encountered L2 forms initially lack sufficiently strong and differentiated connections with semantic representations. Learners therefore rely on established L1 lexical-semantic pathways as representational scaffolds. From a connectionist perspective, this dependence reflects weak L2-specific connection weights rather than the absence of L2 representations. Because activations generated by the L2 forms are unstable and poorly differentiated, accesses to meaning are mediated primarily through entrenched L1 representations. This network perspective can explain several behavioral phenomena that were often treated separately, including weak L2 activation, strong L1 interference, asymmetric L1-L2 priming, and sensitivity to orthographic and phonological forms rather than semantics.
The central theoretical contribution of the paper is to conceptualize lexical development as the gradual strengthening of direct L2 form-meaning mappings, accompanied by a relative reduction of the reliance on L1-mediated routes. Within the framework of connectionist, increased exposure and proficiency play a more important role than merely improve lexical access speed. They reorganize activation patterns, stabilize L2-specific pathways, reduce representational overlap, and increase the efficiency of direct semantic access. The independence of L2 lexical processing, therefore does not require the formation of fully separate L2 lexicons. Rather, it emerges through the self-organization of increasingly stable L2 pathways within an interactive bilingual network. The relationship between parasitic and independent representation is consequently treated as a developmental continuum rather than a categorical distinction.
A further innovation lies in the mechanism-level comparison of six influential approaches to L2 lexical acquisition. The Bilingual Interactive Activation Model is interpreted as an account of the competitive dynamics produced by differences in representational strength across languages. Lexical Entrenchment Account explains how repeated L2 exposure gradually stabilizes and entrenches lexical representations. The Fuzzy Lexical Representations hypothesis is incorporated as an account of fuzzy lexical representations resulting from insufficient or inaccurate encoding. The Ontogenesis Model of the L2 Lexical Representation is situated within a distributed architecture in which orthographic, phonological, and semantic representations develop at different rates. The Developmental Lexicon-II Model is treated as a computational realization of fuzzy representation. Finally, the Complementary Learning Systems model provides a neurocognitive substrate for this trajectory: rapid hippocampal learning supports the initial encoding of novel L2 words, whereas slower neocortical consolidation enables their gradual integration, differentiation, and stabilization.
By aligning these approaches under shared computational principles, the paper argues that they are not competing explanations but complementary descriptions of the same developmental process at different levels. Behavioral theories describe changes in priming, interference, and lexical access; representational accounts characterize changes in encoding quality and lexical organization; computational models formalize the adjustment of connection weights; and neurocognitive frameworks specify the learning systems that support these changes. This cross-level synthesis clarifies conceptual overlap, and identifies common mechanistic denominators across theories that have traditionally been considered separately.
In conclusion, this paper offers a mechanism-driven synthesis of classic theories of L2 lexical acquisition by locating the transition from parasitic to relatively independent representation within a unified connectionist architecture. Its originality lies in demonstrating how existing behavioral, representational, computational, and neurocognitive accounts can be organized around shared principles of distributed representations, competitive activations, experience-dependent weight adjustments, and complementary learning. The resulting framework provides a coherent developmental explanation of L2 lexical independence and a foundation for future longitudinal, computational, neurocognitive, and pedagogical research.

Key words: L2 lexical representation, “parasitic”, representation, independent lexical representation, Connectionist framework