An Introduction to the Cognitive Neuroscience of Second and Artificial Language Learning

Besides acquiring one's first language(s) in infancy, learning an additional language at a later point in life is one of the most fascinating accomplishments of the human mind. It is so fascinating that it verges on the miraculous. Scientists are not usually fond of the term, however, and many a scholar has striven to identify, characterize, and quantify, if not resolve, the cognitive and neural mechanisms underlying this uniquely human ability. The study of language learning is rendered particularly challenging by the fact that the process is driven by the complex interplay of essential learner-intrinsic and learner-extrinsic variables. Moreover, a great deal of language learning happens outside conscious awareness, that is, without the learner relying on explicit processes involving declarative memory or metacognitive awareness. It follows that subjective and questionnaire-based assessments provide limited insights into the cognitive adventure that represents the learning of a language. In order to gain theoretical insights, we must thus turn to methods that go beyond the description of overt manifestations of language knowledge and “surface” indicators of performance (i.e., external manifestations of comprehension and production). This explains the spectacular expansion of the cognitive neuroscience of language over the past three decades, involving increasingly sophisticated and diversified methods, such as neuropsychological testing, reaction time experiments, computational modeling, eye-tracking, electrodermal conductivity, electrophysiology, magnetoencephalography, positron emission tomography, functional magnetic resonance imaging, near-infrared spectroscopy, transcranial magnetic stimulation, and transcranial direct current stimulation. The risks attached to using such sophisticated methods should not be underestimated, however, because the more sophisticated the method, the more abstract the data obtained, and the less intuitive the interpretation. This special issue on the cognitive neuroscience of second and artificial language learning (CoNSALL) covers a range of key topics in the study of how we learn and use novel languages. Far from being monolithic, the collection of contributions covers a spectrum of questions, approaches, methods, and considerations. The seven empirical studies and three conceptual reviews deal with a variety of target populations, including child and adult learners; monolinguals, bilinguals, and multilinguals. They cover a wide range of languages, from natural ones such as Basque, Dutch, English, German, Italian, and Spanish to artificial miniature systems, and they focus on different language areas, from vocabulary to morphology and syntax. The studies employ pictures, speech, words, sentences, animations, and videos to study lexical development, the learning of syntactic features such as grammatical gender or word order, and conceptual mapping, from the early stage of learning a second or an artificial language, to full proficiency, and even attrition of the native language. The data are based on a wide range of methodologies, from questionnaires and behavioral (psycholinguistic) measures to event-related potentials, brain stimulation, as well as structural and functional neuroimaging. In order to help readers choose a path through the CoNSALL special issue, or perhaps identify those contributions that they will find particularly relevant, we have compiled Table 1, which conveys the diversity of approaches and methods whilst highlighting overlaps among the papers. Accuracy RTs Learning of L2 word meanings, plasticity, L1 Attrition Liu and van Hell Behavioral ERPs Definition learning Semantic priming Reading and button press Accuracy LPC, N400 Bultena et al. Bilingual and multilingual learners Dutch- German Behavioral ERPs Pictures Written words Exposure + feedback 2AFC decision Confidence rating Reading and button press Accuracy RTs, Ratings ERN, CRN L1 interference during L2 grammatical gender learning Recall 4AFC MCQ Translation Listening or reading and writing Word error rate Accuracy Semantic similarity Pliatsikas et al. Children Adults Gray matter volume Brain metabolism Behavioral ERPs Pictures Definitions Accuracy RTs P300 German-English Italian- English Behavioral ERPs Accuracy RTs LAN, P600 Vocabulary learning exposure, matching, and production Syntactic priming Viewing, speaking, matching, and reading Accuracy Scores Priming effect English- Varied Artificial Animations Spoken sentences 2AFC task Vocabulary test Grammatical judgement Viewing, listening and button press Accuracy Procedural memory Declarative memory Neuroimaging Neurostimulation Patient studies Activation levels Accuracy, RTs The diversity that characterizes this special issue is representative of research in second language acquisition and artificial language learning. It should not be seen as a weakness; on the contrary, we believe that the wide range of dimensions covered offer excellent opportunities for transfer of knowledge across research domains and cross-disciplinary applications of methods. And this is not just a fuzzy-warm, well-intended, diplomatic statement. Truly interdisciplinary science does not suddenly happen just because a group of open-minded academics have been invited to exchange ideas and interact in a given venue for a day or two. A process of discovery and gradual adaptation is required, which entails not only the consideration of questions slightly different from our own, but also what the issues are that critically matter to researchers in neighboring fields, as well as an understanding of how other methods can shed light on the questions we have. Toward the end of this introduction we propose ways in which the different studies, reviews, and opinions presented in the special issue can be related to one another to further our understanding of language learning and use in the future, and how we see them offering interesting opportunities for novel research endeavors and cross-fertilization. This special issue of the Cognitive Neuroscience Series of Language Learning is based on an international conference held at Bangor University, Wales, in September 2018, with the same title as this special issue (conference URL: http://consall.bangor.ac.uk). The conference brought together over 60 scholars from 17 countries and from a wide range of disciplines with a strong interest in how languages are learned, from linguistics and computer science to psychology and neuroscience. Papers and communications were invited that deal with the cognitive and neural bases of the acquisition and use of a second language or of an artificial language. In the call for papers, the following questions were particularly emphasized: How are the novel representations of a second language neuroanatomically organized or neurophysiologically implemented in relation to representations of the first? How do they dynamically interact? How do the characteristics of the first language influence second and artificial language processing, and, in turn, how do the characteristics of a second or artificial language influence first language processing? How much functional autonomy do language representations enjoy from generic cognitive control systems during and after learning? Having attended the conference ourselves, and having perused the feedback received from delegates, CoNSALL certainly appears to have delivered passionate moments of communication and interactions between scholars with highly different perspectives on language learning. Keynotes were given by Janet van Hell, Gary Oppenheim, and Patrick Rebuschat, and invited presentations were delivered by Laura Batterink, Jon Andoni Duñabeitia, Rob Hartsuiker, Kara Morgan-Short, Karsten Steinhauer, and John Williams, all of whom were invited to contribute a paper to this special issue. In addition, two papers were invited on the basis of ratings and comments received in the process of abstract evaluation for the selection of oral presentations. Two authors were selected to make a contribution at this stage: Sybrine Bultena and Christos Pliatsikas. Thus, at least one author of each of the articles presented in this special issue participated in the conference, whether by invitation or through the selection process. The only exception was Morten Christiansen who, despite having a keen interest in participating in the conference, was unable to do so as he was delivering the University of Alberta's 41st Annual Distinguished Scholar lecture series in the same week as CoNSALL was held. Morten thus received an independent invitation to join in the special issue venture at a later time. The first three articles of the special issue deal with the early learning of specific language features: novel (made-up) words, second language words, and grammatical gender in individuals who have little or no knowledge of gender in their native language. Oppenheim, Griffin, Peña, and Bedore (2020) adopt a skill-based view of language, according to which we learn language (and other skills) incrementally, through extensive practice, and possibly also lose language as competing linguistic systems are developed. This view has clear implications for the study of bilingual language acquisition and use. It assumes that, in bilingual populations, the ability to speak two languages should depend on how much each language is practiced and not just on the order of acquisition. It also raises the possibility that extensive use of a second language could cause gradual “unlearning” of the first language, a matter of concern for learners, parents, and educators, as well as communities and society as a whole. To explore these issues, Oppenheim and colleagues observed the lexical development of 139 Spanish-English bilingual children for several years as they transitioned from mostly-Spanish homes to mostly-English schools. During this period, the children's English proficiency predictably overtook that of their Spanish. However, the children's Spanish skills continued to improve, thus demonstrating that losing one's first language (here, Spanish) is not an inevitable consequence of becoming dominant in one's second language (English). On the contrary, continued practice of both languages permits the consistent improvement of both languages. Liu and van Hell (2020) investigated how monolingual adults learn novel words from definitions when they have had no significant experience of new word learning as adults. The authors used well established indices of word learning, the late positive component (LPC, indexing lexicalization) and the N400 (indexing consolidation) to show that, although inexperienced learners manifested signs of having lexicalized novel words after 24 hours of offline consolidation, the meaning of learned words had not become consolidated enough to modulate the amplitude of the N400 component. This pattern of results resembles that seen in experienced learners (bilingual individuals), who also did not show N400 modulations after 24 hours of offline consolidation, but who, in contrast to inexperienced learners, appeared to have lexicalized novel words immediately after exposure to definitions (Bakker et al., 2015). Bultena, Danielmeier, Bekkering, and Lemhöfer (2020) relied on an electrophysiological index of error monitoring, namely the error-related negativity (ERN), to study the early stages of learning Dutch grammatical gender in native speakers of German. While participants initially made errors based on intuitions deriving from misguided generalizations of their native German grammar, they improved considerably over three learning sessions with corrective feedback. Interestingly, in the case of cognates with mismatching gender across the two languages, a stronger ERN effect was elicited by answers that were correct in the new language and a weaker ERN was elicited by errors (which aligned with the correct gender in the native language). The author interpreted these findings as a sign of overreliance on intuitions derived from native language knowledge. But, as participants became more accurate and knowledgeable regarding grammatical gender in Dutch, the ERN pattern gradually inverted to match the traditional response to errors. The article opens an interesting new avenue for monitoring learning progress at the interface between implicit indices (electrophysiological) and explicit measures (performance). The next four articles move on from the case of learners of completely new words or properties of a second language to the case of individuals who already have achieved a certain level of proficiency in their second languages: bilinguals. In their contribution, Culbertson, Andersen, and Christiansen (2020) validate a faster and simpler method to assess second language proficiency. Determining a speaker's language proficiency is important for a variety of well-known reasons, from employment, university admissions, and placement in language courses to experimentation and research. However, proficiency testing is often time-consuming, and many tests may not provide reliable indications of how the learner would fare in real-world conversations. Culbertson and colleagues explored whether our ability to listen to and recall an utterance in the second language (here, Spanish) could function as a reliable test of second language proficiency. Recall taps into our ability to process language input in real-time, as needed, for example, in a conversation in the real world. The authors found that their recall measure was a better predictor of a learner's ability to translate heard sentences than a shortened version of a standardized listening multiple-choice comprehension test. These findings suggest that recall of naturalistic utterances may provide an accurate and efficient alternative to traditional tests for predicting foreign language proficiency. Pliatsikas, DeLuca, and (2020) a to a that is, in a highly abstract and that bilingual and performance are first and by exposure and by many beyond those of of acquisition and of learning. The by the authors an interesting between experienced bilinguals, who show in brain related to language control and who are in the process of or have limited experience of their second language, and who to show in related to vocabulary acquisition. Pliatsikas and colleagues make a case for more consideration to be given to in research on They also to the of approaches and they the of the between and cognitive in and (2020) provide new this that of new conceptual is by the language or in which definitions are participants were artificial and learned to these with two definitions that the novel to But, it did not to matter whether the two definitions learned were presented in the native language this or if one was presented in the native and the other in the second language This is interesting because it at with the that language during should be which is in implemented in bilingual The study is particularly because of on implicit measures of derived from brain using a of event-related well to index of a target and (2020) of language They and the first from that the native language can of in individuals who have become or even on a later second language. speakers of who have using their native language for a of time can grammatical sentences in their native language as such effect is only seen in individuals who have become dominant in their second language and have using their native language with the in the article by Pliatsikas and colleagues the level of in is that a native language is one's native language for The presented in this article is particularly because it indices of and syntactic by the the N400 and the The three articles the first three of the special issue by questions of learning new words and grammatical but this time in relation to an artificial than a second language. and (2020) investigated the of in bilingual It is often that syntactic representations of two languages are in are across the two languages the in Dutch and but it is not clear how syntactic representations to be for to and colleagues by syntactic priming between Dutch and three artificial languages in of word order and case They found for syntactic between Dutch and the artificial languages, this was limited to They also observed across the three of the artificial language, that word order the of syntax. and (2020) data from an artificial language learning involving and over each as participants to with sentences made of completely new words in an syntax. The study features an in implicit and explicit learning and on the one with judgement regarding word order, and word on the and colleagues were to show that adult learners could novel words and syntactic as a of exposure and without feedback. They observed that and were by and that the after 24 the authors found that declarative memory initially learn and memory initially with and after 24 memory became more The special issue with a by which a highly and of selected and studies in an to between implicit learning and language The is if the processes in implicit learning of an artificial those in language processing, it should that brain by and neuropsychological research into implicit artificial learning at least in observed in research into language The is complex and not without The for Morgan-Short, and the diversity of make it particularly to direct between the two a by a of key and brain in language that also to a in implicit artificial learning, and the same the contrast between domains and The author by highlighting questions that we must be to if we to better the neural of both language acquisition and that of implicit learning in The contributions in this special issue provide an at the of second and artificial language learning and they to how the in the the one the articles an from the CoNSALL conference on which the special issue was progress in the study of language learning to our and with scholars and of other It is to for but it is often perhaps even to see how little across research with scholars in and neuroscience often the also the conference, the special issue together from all of whom have an interest in how we learn language(s) and who are to in with those colleagues with a different in articles that deal with questions, such as vocabulary and learning, but that use different such as performance or electrophysiological index and we the special issue to the of this essential cross-disciplinary this not just the reading of from fields, more understanding how a method different from our can shed light on the same we are to one's into the is the first into and the special issue one to do this when reading related articles that use In this special issue, the of Liu and van Hell with that of Oppenheim and contrast and approaches with those of Pliatsikas and findings from and colleagues with made by beyond the for interdisciplinary discovery by this special issue, the approaches by between studies and for research. studies of artificial learning et al., et al., would from using the electrophysiological indices that to a of performance errors or grammatical in natural language Bultena et al., In the study by and if a or is learned through exposure to an artificial language, of such words when they the conceptual of a or should influence the component to to such in natural language LAN, or see the of using the measure of language by Culbertson et al. (2020) in experiments, which have little time to to the of language proficiency, such an and measure could highly when used as a in or as a in index of error such as the ERN Bultena et al., explicit measures of performance in Culbertson et al., Liu van Hell, Oppenheim et al., would an index of of their errors. In other words, this additional data would shed light on ability to not only errors but errors as such that they to have learned words or grammatical an on our of The research in this special issue on how we learn and use language(s) a first linguistic is least in that is, second language and To the the studies employ natural languages artificial languages Table Artificial languages have been used for to a years as miniature of how language is learned outside the these systems has many example, they excellent control of the input in the that the can control what the participants are being to the same are several which to be example, many artificial systems are of a that is, learners are to with that are or their are highly This is if artificial languages are used to on how languages are learned in the a process that on learning to and utterances with a syntactic that can be more complex than that used in most artificial language this it would be important for studies to the development of artificial systems that natural languages more thus the of artificial language research to the study of natural language acquisition. In addition, it would be important for research to the between artificial and natural language learning. have been made Morgan-Short, and is that the of artificial systems as to study language learning Williams, but a and is This special issue, and the which would not have been without the and contributions of many are to for all over and the and for their with abstract and conference go to and with and to for and in not only the and but most for the that to be the point of the The was and by in of and and It was made by from the for Wales, and we are to for it we would to our and all the for participating in CoNSALL and the conference so and would to particularly all the authors of this special issue for their as well as our who each and feedback. our and Liu for with the task of the special issue, and with our for error that may and that should be our are also to our and and the and passionate of Language our for invitation to on this adventure in the first and we for the CoNSALL conference and of and in relation to

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.