Brain and Language Lab

Second Language and Bilingualism

This broad topic is one of our major areas of investigation. We are examining a number of issues and questions in this area.

First, in our theoretical work, we have proposed and are further developing the hypothesis that the learning, knowledge and use of second and additional languages depend on declarative and procedural memory (see especially Ullman, 2001, 2005, 2013, 2015, 2020). On this view - the declarative/procedural (DP) model of second language (L2) - later language learning should depend on the two memory systems, just as first language does. However, given that the memory systems change over the course of childhood, later learned languages should be learned somewhat differently than native languages. In particular, evidence suggests that learning and consolidation in declarative memory improves during childhood (and plateaus during adolescence and early adulthood), whereas learning and consolidation in procedural memory may be well-established early in childhood, and then might attenuate (though the evidence is less clear for developmental changes in procedural memory than in declarative memory). For these and other reasons (e.g., pedagogical instruction of later languages that might encourage explicit learning in declarative memory), later learned languages should depend more on declarative memory than native languages. However, this increased dependence on declarative memory should hold only for rule-governed aspects of grammar, which can depend on either memory system. In contrast, most lexical (idiosyncratic) knowledge can apparently be learned only in declarative memory, and thus should depend on this system in all language learners. Moreover, procedural memory is not posited to be defunct in adults, but simply less available for learning than during childhood (as compared to declarative memory). Therefore, contrary to strong versions of the critical period hypothesis, we predict that rule-governed aspects of grammar will eventually be learned in procedural memory, and thus should become increasingly native-like (first language-like) in their neurocognition over time and with practice/exposure. Finally, various learner, input, and item level factors should affect the extent to which grammar depends more or less on each of the two memory systems, and how this changes over time. For example, learners with a relative advantage in declarative memory (e.g., due to factors such as sex or genotype) should depend earlier and more on this system for grammar. Along the same lines, learning conditions (e.g., types of language training) that enhance learning in either declarative memory or procedural memory should lead to different patterns of dependence of grammar on the two memory systems (Ullman & Lovelett, 2018). Indeed, techniques that enhance learning in either system, such as spaced repetition, retrieval practice, exercise, or diet, should also enhance language learning in either system (Ullman & Lovelett, 2018). We found that Turkish learners of English show a striking benefit of combining spacing and testing, which yielded an 18 percentage-point increase from pretest to two post-tests (Karatas et al., 2021).

Second, we perform empirical studies to examine these predictions, as well as related issues. Our studies use a variety of techniques, including various behavioral methods (e.g., correlational approaches and the examination of frequency effects), event-related potentials (ERPs), and functional neuroimaging (fMRI). We investigate the learning, processing, representation, and retention of language, in various experimental linguistic paradigms, including with natural languages, mini-languages, artificial languages, and artificial grammars. Within natural languages we work mainly with English, but also examine other languages (e.g., Spanish). One exciting line of research has examined the neurocognitive effects (using behavioral and ERP measures) of instructed (explicit, classroom-like) versus uninstructed (implicit, immersion-like) training on the neurocognition of syntax, using an artificial language (Morgan-Short et al., JoCN, 2012). We found that only uninstructed/implicit training led to fully native-like patterns of brain processing. In a follow-up study we tested the learners 5 months later, and found that their brain patterns became more native-like for syntactic processing, and that this held for both the instructed/explicitly and uninstructed/implicitly trained groups, though again only the uninstructed group showed full native-like brain processing (Morgan-Short et al., PLoS ONE, 2012). We hypothesize that this increase in native-like brain processing over time was due to consolidation of syntactic knowledge, possibly in procedural memory. Another study found that university language learners of Spanish with classroom and immersion experience were able to attain native-like processing of syntax (Bowden et al., 2013). We have examined the effects of multiple factors, including sex (male vs. female), age of arrival, and length of residence on the computation (storage vs. composition) of inflected forms in L2 (Babcock et al., 2012). And in an fMRI study we investigated the learning and generalization of affixal morphology in an artificial language paradigm (Nevat et al., 2017).

We are also increasingly performing behavioral and neuroanatomical meta-analyses of second/additional language learning. In one recent paper, meta-analyses of correlational studies showed that grammatical processing abilities, across syntax and morphology, correlated with learning abilities in declarative memory (but not procedural memory) at early stages of L2 learning, whereas at later stages of L2 learning grammatical processing abilities correlated with learning abilities in procedural memory (but not declarative memory) (Hamrick et al., 2018). This paper convincingly links grammar to declarative memory at lower levels of L2 experience, but to procedural memory at higher levels, as predicted by the DP model. In another recent paper, we performed multiple neuroanatomical meta-analyses (using Activation Likelihood Estimation) to synthesize the functional neuroimaging literature of language learning. The analyses showed that grammar learning activates the anterior striatum (anterior caudate nucleus and anterior putamen), implicating procedural memory, while word learning activates occipito-temporal structures linked to the ventral stream and declarative memory. Moreover, grammar learning predicted to rely especially on declarative memory (e.g., with instructed explicit training) shows hippocampal involvement, while grammar learning predicted to rely particularly on procedural memory (e.g., with uninstructed implicit training) shows anterior striatal involvement (Tagarelli et al., 2019).

Finally, we investigate the neurocognitive effects of bilingualism. In an ERP study, Grey et al. (2018) found that bilinguals (who had learned both English and Mandarin at an early age) showed more native-like brain processing in learning an additional language than monolinguals (English-speaking). And in a recent structural imaging study, we investigated the developmental patterns of both grey and white matter structures in a cross-sectional study of a large sample (n=711 for grey matter, n=637 for white matter) of bilingual and monolingual participants, aged 3-21 years. Metrics of grey matter (thickness, volume, surface area) and white matter (fractional anisotropy, mean diffusivity) were examined across 41 cortical and subcortical brain structures and 20 tracts, respectively. As compared to monolinguals, bilinguals showed: a) more grey matter (less developmental loss) starting during late childhood and adolescence, mainly in frontal and parietal regions (particularly in the inferior frontal gyrus pars opercularis, superior frontal cortex, inferior and superior parietal cortex, and precuneus); and b) higher white matter integrity (greater developmental increase) starting during mid-late adolescence, specifically in striatal-inferior frontal fibers. The data suggest that there may be a developmental basis to the well-documented structural differences in the brain between bilingual and monolingual adults (Pliatsikas et al., 2020).