Referring to theories which hypothesize that L2 in late learners initially relies on different substrates than does L1 but later gradually comes to rely on L1 neurocognition, Babcock et al. (Reference Babcock, Stowe, Maloof, Brovetto and Ullman2012, p. 821) write: “Paradis (Reference Paradis and Ellis1994, Reference Paradis2004, Reference Paradis2009) argues that this reliance increases across both lexical and grammatical functions”. This declaration might appear surprising to readers familiar with Paradis’ work since, in the very publications cited by the authors, one can find repeated statements to the contrary. Particularly puzzling is the immediately following assertion that “In contrast, both Clahsen and Ullman and their colleagues posit that lexical functions are always similar in L1 and L2, and that only aspects of grammar change over time regarding the neurocognitive mechanisms they depend on” (my emphasis), when this is precisely what Paradis says – and hence there is no contrast. In fact, this has been the basis for Paradis’ main contention over the past 20 years, namely that explicitly known forms (including irregular verb forms) are always of the same nature in L1 and L2 (i.e., declarative). This is why bilingual experimental studies using lexical items as stimuli – including neuroimaging (Paradis, Reference Paradis2004, pp. 173–180; Reference Paradis and Arabski2007) and laterality studies (Paradis, Reference Paradis2008a, Reference Paradis2009) – cannot claim to address the representation of the language system, because lexical items in isolation are stored declaratively (like English irregular verb forms) and thus do not reveal anything about the grammatical procedures (like the implicit composition of the regular past tense) that underlie implicit language competence. In L2, only the neurocognitive mechanisms on which aspects of the grammar depend change over time, in that, unlike in L1, much of the grammar is processed declaratively, at least in the early stages; with extensive exposure, though, it may eventually rely on the same neurocognitive processes as L1 (Paradis, Reference Paradis2009).
In spite of Paradis’ repeated pleas to heed the specific technical meanings of terms associated with a specific theory in order to eliminate ambiguities as much as possible, and thereby avoid barren controversies and distracting red herrings (Paradis, Reference Paradis2004, p. 233), we seem to be witnessing an instance of attributing to an expression a meaning other than the one intended by its author. This would constitute a misrepresentation (see Paradis, Reference Paradis2006), especially when the term has been defined by the author in question. Now, one must admit that the matter can sometimes be confusing, particularly when common terms (e.g., lexicon and lexical item) are used with a technical meaning (see Ryan, Reference Ryan1985), and Paradis may not always have been sufficiently unambiguous. So, let us straighten the record by clarifying whatever may have led to misinterpretation.
To begin with, Paradis has never used the term lexical function. It is therefore difficult to determine what this expression might refer to in his framework. To my knowledge, lexical functions do not correspond to any construct in current generative or any other type of grammar, except Meaning-Text Theory, in which the term lexical function is used to describe semantic relationships, such as in verb–noun collocations (e.g., Mel'čuk, Reference Mel'čuk and Wanner1996, Reference Mel'čuk and Cowie1998, Reference Mel'čuk, Burger, Dobrovol'skij, Kühn and Norrick2007).Footnote 1 In Babcock et al.'s context, given that lexical functions are said to be stored and, unlike grammatical functions, to be of the same nature in both L1 and L2, one may infer that they refer to what Paradis calls lexical items (or words). Previously, Ullman (Reference Ullman2006, p. 100) had already stated that “Paradis discusses the increased reliance on procedural memory, in both L1 and high-proficiency L2, largely in terms of greater automatization and implicitness across various domains of language, including both lexicon and grammar” (italics in the original). The issue was then addressed in Paradis (Reference Paradis2009, p. 12), apparently to no avail. The properties attributed by Babcock et al. to lexical functions are indeed those of what Paradis (Reference Paradis and Ellis1994, Reference Paradis2004, Reference Paradis2009) calls words (lexical items).
Now, Paradis (Reference Paradis and Ellis1994, Reference Paradis2004, Reference Paradis2009) refers to lexical items as meaning–form relationships, excluding the grammatical properties that take effect in the context of their use. Lexical items (meaning–form relationships), which are explicit, are differentiated from their grammatical properties, which are implicit. In 1994, lexicon and vocabulary were used as synonyms that referred to the storage of lexical items, or words, and contrasted with morphosyntax. Based on the evidence available at the time, morphosyntax, but not vocabulary, was considered to be affected by maturational constraints imposed on procedural memory. Later, the term lexicon received a technical meaning in the context of a neurolinguistic theory of bilingualism (Paradis, Reference Paradis2004) in which the linguistic competence for each language is hypothesized to constitute a subset (or subsystem) of the neurofunctional system that sustains language. Each subsystem contains a parametric instantiation of phonology, morphology, syntax and the implicit grammatical properties and implicit semantic constraints of lexical items. The lexicon of each subsystem is an abstract construct that contains that subsystem's lexical items together with their grammatical properties; the vocabulary contains lexical items (meaning–form correspondences without their implicit grammatical properties), most often called words precisely to avoid possible confusion with items in the lexicon and to make the vocabulary/lexicon distinction clear. Lexical items (i.e., words) constitute the vocabulary (i.e., the set of explicit words stored by a speaker) and are sustained by declarative memory.
Paradis (Reference Paradis2004, Reference Paradis2009, and elsewhere) assumes that implicit properties of L1 lexical items, like all other implicit grammatical functions, are acquired incidentally and used automatically, whereas, in L2, like other grammatical processes, they are either ignored or explicitly learned and stored in declarative memory. There is no reason to believe that implicit grammatical properties of lexical items should behave differently than other grammatical functions. For the sake of logical coherence, what is implicit is not explicitly known and hence not declarative, and should be expected to be neurofunctionally processed accordingly. Explicit knowledge of the meaning–form correspondence (words, lexical items), is independent of the implicit knowledge of their morphosyntactic properties: L2 learners often learn words without acquiring their morphosyntactic properties (as demonstrated in the production, by English speakers of French L2, of sentences such as *je l'ai téléphoné – unaware that téléphoner requires an indirect object; and in French speakers’ English L2 of sentences such as *he provided three informations – unaware that information is a mass noun in English). In other words, the grammatical properties of lexical items behave in every way like other aspects of the grammar. Implicit competence leads one to add the plural marker to count nouns (N + s), but not to mass nouns, just as one adds the past tense marker to regular verbs (V + ed), but not to irregular ones.
Processes that are used implicitly in L1 may be replaced by the use of the corresponding explicit rules in L2 until such time as, through frequent and prolonged use, they are acquired and hence processed by the same neurocognitive mechanisms as in L1. Both age of appropriation and length of exposure may impact on the transition from using stored knowledge to using implicit competence (including composition), which represents a shift from relying on different neurocognitive processes to relying on the same ones that L1 relies on, namely implicit competence subserved by procedural memory.
According to Paradis (Reference Paradis2004, Reference Paradis2009), L2 learners initially depend largely on different neural substrates than L1 speakers do (i.e., they store both words and grammatical functions, including grammatical properties of lexical items in declarative memory) but, with increasing exposure or proficiency, they may gradually come to rely on procedural memory for these implicit grammatical functions, as L1 speakers do from the beginning. Thus, irregular past tense forms continue to be stored, as in L1 (as is the case with kanji vs. kana or irregularly vs. regularly spelled Portuguese words – Paradis, Reference Paradis2008b); and some – potentially most – implicit grammatical functions, which were initially stored, may eventually become implicitly processed (e.g., the regular past tense may gradually be composed). The nature of the underlying neurocognition may differ between Paradis’ and Ullman's accounts in that for Paradis implicit grammatical properties of lexical items rely, like other implicit grammatical properties, on procedural memory, whereas for Ullman it would appear that they rely, together with words, on declarative memory (personal communication).
This difference, however, has no bearing on the interpretation of the data in the Babcock et al. article, because, in both theories, the items of interest – namely irregular verb forms – are considered to be stored (in L1 and L2) while regular past tense forms are considered to be composed in L1. According to Paradis, as well as Ullman, complex forms that depend largely on rule-governed compositional mechanisms in L1 rely instead in L2, at least at low levels of proficiency, on different mechanisms, and in particular declarative memory, which also includes lexical memory (Babcock et al., Reference Babcock, Stowe, Maloof, Brovetto and Ullman2012; Paradis, Reference Paradis2009) – I assume that lexical memory refers to memory for lexical items. Therefore, in both frameworks, lexical items are stored, English irregular past tense forms are stored, in L1 and L2 alike. However, in L2, grammatical functions (i.e., implicit processes of which the speaker is not necessarily aware), including regular past tense composition (unlike in L1), may be stored at first; with sustained practice over time, however, they may come to depend on the same neurocognitive mechanisms as in L1, namely implicit procedures (such as V + -ed composition). In a nutshell, declarative memory underlies lexical memory (the set of vocabulary words) in L1 and L2. But late-onset L2 speakers, especially at lower levels of exposure, rely instead on declarative memory for complex forms that depend on rule-governed mechanisms in L 1 (Babcock et al., Reference Babcock, Stowe, Maloof, Brovetto and Ullman2012; Paradis, Reference Paradis2004, Reference Paradis2009).
Babcock et al. also mention that the finding that adult age of arrival (AoA) affected the storage vs. composition of regular past tenses was not predicted by either Clahsen or Ullman but is not inconsistent with either view. Nor is it inconsistent with Paradis’ framework (Reference Paradis2009, Chapter 4). Both age of first encounter with, and length of exposure to, L2 are expected to influence the eventual shift to acquiring (i.e., using composition for) those aspects of grammar that were initially stored (for Paradis, including the implicit grammatical properties of lexical items). There is therefore no basis for assuming that “the finding that LoR [length of residence] influences the mechanisms underlying regulars, but not irregulars, does not seem to be expected by Paradis” (Ullman et al. 2012, p. 837). It is perfectly compatible with his proposed neurolinguistic theory of bilingualism (Paradis, Reference Paradis2004), further developed in Paradis (Reference Paradis2009), given that the processing of regular forms – at least at a low level of proficiency – differs between L1 and L2 (composed in L1 but stored in L2), but the processing of irregulars does not, and that longer exposure allows for implicit internalization of L2 grammatical processes.
In plain language, then, irrespective of what anyone else might claim that Paradis says, this is what Paradis actually maintains: words are stored in declarative memory in both L1 and L2. Therefore the nature of their neural substrate does not change over time. Implicit grammar (including implicit grammatical properties of lexical items) is subserved by procedural memory in L1. At least in the early stages of a late-learned L2, it is stored in declarative memory in the form of metalinguistic knowledge that is used to construct sentences. Over time, and with increasing exposure, it may eventually be acquired, at least in part, and thus processed in the same way as in L1. The time it takes an individual to shift from using declarative memory to using implicit procedures will depend on a number of factors, including age of first exposure, length and amount of exposure, and motivation – as well as estrogen levels and BDNF gene expression, as identified by Ullman (Reference Ullman2008).
Referring to theories which hypothesize that L2 in late learners initially relies on different substrates than does L1 but later gradually comes to rely on L1 neurocognition, Babcock et al. (Reference Babcock, Stowe, Maloof, Brovetto and Ullman2012, p. 821) write: “Paradis (Reference Paradis and Ellis1994, Reference Paradis2004, Reference Paradis2009) argues that this reliance increases across both lexical and grammatical functions”. This declaration might appear surprising to readers familiar with Paradis’ work since, in the very publications cited by the authors, one can find repeated statements to the contrary. Particularly puzzling is the immediately following assertion that “In contrast, both Clahsen and Ullman and their colleagues posit that lexical functions are always similar in L1 and L2, and that only aspects of grammar change over time regarding the neurocognitive mechanisms they depend on” (my emphasis), when this is precisely what Paradis says – and hence there is no contrast. In fact, this has been the basis for Paradis’ main contention over the past 20 years, namely that explicitly known forms (including irregular verb forms) are always of the same nature in L1 and L2 (i.e., declarative). This is why bilingual experimental studies using lexical items as stimuli – including neuroimaging (Paradis, Reference Paradis2004, pp. 173–180; Reference Paradis and Arabski2007) and laterality studies (Paradis, Reference Paradis2008a, Reference Paradis2009) – cannot claim to address the representation of the language system, because lexical items in isolation are stored declaratively (like English irregular verb forms) and thus do not reveal anything about the grammatical procedures (like the implicit composition of the regular past tense) that underlie implicit language competence. In L2, only the neurocognitive mechanisms on which aspects of the grammar depend change over time, in that, unlike in L1, much of the grammar is processed declaratively, at least in the early stages; with extensive exposure, though, it may eventually rely on the same neurocognitive processes as L1 (Paradis, Reference Paradis2009).
In spite of Paradis’ repeated pleas to heed the specific technical meanings of terms associated with a specific theory in order to eliminate ambiguities as much as possible, and thereby avoid barren controversies and distracting red herrings (Paradis, Reference Paradis2004, p. 233), we seem to be witnessing an instance of attributing to an expression a meaning other than the one intended by its author. This would constitute a misrepresentation (see Paradis, Reference Paradis2006), especially when the term has been defined by the author in question. Now, one must admit that the matter can sometimes be confusing, particularly when common terms (e.g., lexicon and lexical item) are used with a technical meaning (see Ryan, Reference Ryan1985), and Paradis may not always have been sufficiently unambiguous. So, let us straighten the record by clarifying whatever may have led to misinterpretation.
To begin with, Paradis has never used the term lexical function. It is therefore difficult to determine what this expression might refer to in his framework. To my knowledge, lexical functions do not correspond to any construct in current generative or any other type of grammar, except Meaning-Text Theory, in which the term lexical function is used to describe semantic relationships, such as in verb–noun collocations (e.g., Mel'čuk, Reference Mel'čuk and Wanner1996, Reference Mel'čuk and Cowie1998, Reference Mel'čuk, Burger, Dobrovol'skij, Kühn and Norrick2007).Footnote 1 In Babcock et al.'s context, given that lexical functions are said to be stored and, unlike grammatical functions, to be of the same nature in both L1 and L2, one may infer that they refer to what Paradis calls lexical items (or words). Previously, Ullman (Reference Ullman2006, p. 100) had already stated that “Paradis discusses the increased reliance on procedural memory, in both L1 and high-proficiency L2, largely in terms of greater automatization and implicitness across various domains of language, including both lexicon and grammar” (italics in the original). The issue was then addressed in Paradis (Reference Paradis2009, p. 12), apparently to no avail. The properties attributed by Babcock et al. to lexical functions are indeed those of what Paradis (Reference Paradis and Ellis1994, Reference Paradis2004, Reference Paradis2009) calls words (lexical items).
Now, Paradis (Reference Paradis and Ellis1994, Reference Paradis2004, Reference Paradis2009) refers to lexical items as meaning–form relationships, excluding the grammatical properties that take effect in the context of their use. Lexical items (meaning–form relationships), which are explicit, are differentiated from their grammatical properties, which are implicit. In 1994, lexicon and vocabulary were used as synonyms that referred to the storage of lexical items, or words, and contrasted with morphosyntax. Based on the evidence available at the time, morphosyntax, but not vocabulary, was considered to be affected by maturational constraints imposed on procedural memory. Later, the term lexicon received a technical meaning in the context of a neurolinguistic theory of bilingualism (Paradis, Reference Paradis2004) in which the linguistic competence for each language is hypothesized to constitute a subset (or subsystem) of the neurofunctional system that sustains language. Each subsystem contains a parametric instantiation of phonology, morphology, syntax and the implicit grammatical properties and implicit semantic constraints of lexical items. The lexicon of each subsystem is an abstract construct that contains that subsystem's lexical items together with their grammatical properties; the vocabulary contains lexical items (meaning–form correspondences without their implicit grammatical properties), most often called words precisely to avoid possible confusion with items in the lexicon and to make the vocabulary/lexicon distinction clear. Lexical items (i.e., words) constitute the vocabulary (i.e., the set of explicit words stored by a speaker) and are sustained by declarative memory.
Paradis (Reference Paradis2004, Reference Paradis2009, and elsewhere) assumes that implicit properties of L1 lexical items, like all other implicit grammatical functions, are acquired incidentally and used automatically, whereas, in L2, like other grammatical processes, they are either ignored or explicitly learned and stored in declarative memory. There is no reason to believe that implicit grammatical properties of lexical items should behave differently than other grammatical functions. For the sake of logical coherence, what is implicit is not explicitly known and hence not declarative, and should be expected to be neurofunctionally processed accordingly. Explicit knowledge of the meaning–form correspondence (words, lexical items), is independent of the implicit knowledge of their morphosyntactic properties: L2 learners often learn words without acquiring their morphosyntactic properties (as demonstrated in the production, by English speakers of French L2, of sentences such as *je l'ai téléphoné – unaware that téléphoner requires an indirect object; and in French speakers’ English L2 of sentences such as *he provided three informations – unaware that information is a mass noun in English). In other words, the grammatical properties of lexical items behave in every way like other aspects of the grammar. Implicit competence leads one to add the plural marker to count nouns (N + s), but not to mass nouns, just as one adds the past tense marker to regular verbs (V + ed), but not to irregular ones.
Processes that are used implicitly in L1 may be replaced by the use of the corresponding explicit rules in L2 until such time as, through frequent and prolonged use, they are acquired and hence processed by the same neurocognitive mechanisms as in L1. Both age of appropriation and length of exposure may impact on the transition from using stored knowledge to using implicit competence (including composition), which represents a shift from relying on different neurocognitive processes to relying on the same ones that L1 relies on, namely implicit competence subserved by procedural memory.
According to Paradis (Reference Paradis2004, Reference Paradis2009), L2 learners initially depend largely on different neural substrates than L1 speakers do (i.e., they store both words and grammatical functions, including grammatical properties of lexical items in declarative memory) but, with increasing exposure or proficiency, they may gradually come to rely on procedural memory for these implicit grammatical functions, as L1 speakers do from the beginning. Thus, irregular past tense forms continue to be stored, as in L1 (as is the case with kanji vs. kana or irregularly vs. regularly spelled Portuguese words – Paradis, Reference Paradis2008b); and some – potentially most – implicit grammatical functions, which were initially stored, may eventually become implicitly processed (e.g., the regular past tense may gradually be composed). The nature of the underlying neurocognition may differ between Paradis’ and Ullman's accounts in that for Paradis implicit grammatical properties of lexical items rely, like other implicit grammatical properties, on procedural memory, whereas for Ullman it would appear that they rely, together with words, on declarative memory (personal communication).
This difference, however, has no bearing on the interpretation of the data in the Babcock et al. article, because, in both theories, the items of interest – namely irregular verb forms – are considered to be stored (in L1 and L2) while regular past tense forms are considered to be composed in L1. According to Paradis, as well as Ullman, complex forms that depend largely on rule-governed compositional mechanisms in L1 rely instead in L2, at least at low levels of proficiency, on different mechanisms, and in particular declarative memory, which also includes lexical memory (Babcock et al., Reference Babcock, Stowe, Maloof, Brovetto and Ullman2012; Paradis, Reference Paradis2009) – I assume that lexical memory refers to memory for lexical items. Therefore, in both frameworks, lexical items are stored, English irregular past tense forms are stored, in L1 and L2 alike. However, in L2, grammatical functions (i.e., implicit processes of which the speaker is not necessarily aware), including regular past tense composition (unlike in L1), may be stored at first; with sustained practice over time, however, they may come to depend on the same neurocognitive mechanisms as in L1, namely implicit procedures (such as V + -ed composition). In a nutshell, declarative memory underlies lexical memory (the set of vocabulary words) in L1 and L2. But late-onset L2 speakers, especially at lower levels of exposure, rely instead on declarative memory for complex forms that depend on rule-governed mechanisms in L 1 (Babcock et al., Reference Babcock, Stowe, Maloof, Brovetto and Ullman2012; Paradis, Reference Paradis2004, Reference Paradis2009).
Babcock et al. also mention that the finding that adult age of arrival (AoA) affected the storage vs. composition of regular past tenses was not predicted by either Clahsen or Ullman but is not inconsistent with either view. Nor is it inconsistent with Paradis’ framework (Reference Paradis2009, Chapter 4). Both age of first encounter with, and length of exposure to, L2 are expected to influence the eventual shift to acquiring (i.e., using composition for) those aspects of grammar that were initially stored (for Paradis, including the implicit grammatical properties of lexical items). There is therefore no basis for assuming that “the finding that LoR [length of residence] influences the mechanisms underlying regulars, but not irregulars, does not seem to be expected by Paradis” (Ullman et al. 2012, p. 837). It is perfectly compatible with his proposed neurolinguistic theory of bilingualism (Paradis, Reference Paradis2004), further developed in Paradis (Reference Paradis2009), given that the processing of regular forms – at least at a low level of proficiency – differs between L1 and L2 (composed in L1 but stored in L2), but the processing of irregulars does not, and that longer exposure allows for implicit internalization of L2 grammatical processes.
In plain language, then, irrespective of what anyone else might claim that Paradis says, this is what Paradis actually maintains: words are stored in declarative memory in both L1 and L2. Therefore the nature of their neural substrate does not change over time. Implicit grammar (including implicit grammatical properties of lexical items) is subserved by procedural memory in L1. At least in the early stages of a late-learned L2, it is stored in declarative memory in the form of metalinguistic knowledge that is used to construct sentences. Over time, and with increasing exposure, it may eventually be acquired, at least in part, and thus processed in the same way as in L1. The time it takes an individual to shift from using declarative memory to using implicit procedures will depend on a number of factors, including age of first exposure, length and amount of exposure, and motivation – as well as estrogen levels and BDNF gene expression, as identified by Ullman (Reference Ullman2008).