A complementary systems account of word learning: neural and behavioural evidence.
Davis MH, Gaskell MG. · 2009
Read A complementary systems account of word learning: neural and behavioural evidence. by Davis MH, Gaskell MG. (2009) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/europmc-pmc2846311.txt.
Introduction
Davis and Gaskell set out a theory in 2009 of how adults learn new spoken words, drawing on memory research and brain evidence. Words like blog pass from total novelty to instant familiarity, and the authors argue the journey runs in two stages. A fast recording stage in the hippocampus, the brain's memory hub, captures the new word at first hearing. A slower stage then weaves it into the cortex during rest and sleep.
Meta-analysis of brain scans plus behavioral studies supported the split. Hippocampal responses shifted during first learning, whereas cortical responses to new words only turned word-like after overnight consolidation, meaning sleep- or rest-based strengthening of memories. The account unites word learning with general memory systems.
Catalog record
| CREATOR | Davis MH, Gaskell MG. |
|---|---|
| YEAR | 2009 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 151 KB |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
A complementary systems account of word learning: neural and behavioural evidence By Matthew H. Davis, M. Gareth Gaskell Published in: Philosophical Transactions of the Royal Society B: Biological Sciences DOI: 10.1098/rstb.2009.0111 ABSTRACT In this paper we present a novel theory of the cognitive and neural processes by which adults learn new spoken words. This proposal builds on neurocomputational accounts of lexical processing and spoken word recognition and complementary learning systems (CLS) models of memory. We review evidence from behavioural studies of word learning that, consistent with the CLS account, show two stages of lexical acquisition: rapid initial familiarization followed by slow lexical consolidation. These stages map broadly onto two systems involved in different aspects of word learning: (i) rapid, initial acquisition supported by medial temporal and hippocampal learning, (ii) slower neocortical learning achieved by offline consolidation of previously acquired information. We review behavioural and neuroscientific evidence consistent with this account, including a meta-analysis of PET and functional Magnetic Resonance Imaging (fMRI) studies that contrast responses to spoken words and pseudowords. From this meta-analysis we derive predictions for the location and direction of cortical response changes following familiarization with pseudowords. This allows us to assess evidence for learning-induced changes that convert pseudoword responses into real word responses. Results provide unique support for the CLS account since hippocampal responses change during initial learning, whereas cortical responses to pseudowords only become word-like if overnight consolidation follows initial learning. Recent advances in information technology have led to significant changes in both the activities of daily life and the words that we use to describe those activities. Many of us ‘google’ information on the Internet, looking for articles and commentary from influential ‘blog’ writers or ‘bloggers’; activities and words that have become increasingly common in the last decade. A search of the Google index in January 2001 (via www.google.com/search2001.html) yields 76 000 pages containing the word ‘blog’, whereas the same search in January 2009 reveals 3.2 billion hits. There are more Google hits for ‘blog’ than for ‘tv’ (2.3 billion) and ‘science’ (0.78 billion). These observations illustrate how a contraction of the phrase ‘weblog’ first coined in May 1999 (see Origins of ‘Blog’ and ‘Blogger’, American Dialect Society Mailing List, 20 April 2008) has gone from being an obscure piece of technological jargon to a familiar word for many speakers of English. In this paper we examine the neurocomputational processes by which new words such as ‘blog’ come to achieve their status as familiar and meaningful units stored in the brains of language users. We propose that: (i) Isolated representations of new words are initially encoded like other novel experiences as ‘episodic’ memories of their first occurrences. These representations are supported by medial temporal lobe memory systems, functionally and neurally distinct from (ii) the neo-cortical representations that support long-term retention of words. Stable cortical representations are derived from multiple encounters with new words by consolidation processes that abstract away from the episodic representations that encode specific occurrences of novel words. Thus two distinct stages of learning and representation (initial episodic and subsequent lexical) are associated with the acquisition of new words. This account of word learning directly parallels existing dual-systems accounts of memory processes in other domains. The goal of the current paper is therefore to develop a cognitive and neuroscientific account of word learning that reflects the computational constraints proposed by general accounts of memory processes. By integrating this framework with the specific computational demands of the perception of spoken language we can make detailed behavioural and neural predictions concerning processes involved in word learning. While the majority of this paper is focused on the acquisition of novel spoken rather than written words, our goal in developing this account is to achieve a broad coverage of phenomena that are relevant for word learning in general. The outline of this paper is as follows. We first explore the complementary learning systems (CLS) model (McClelland et al. 1995), which partitions memory into hippocampal and neocortical components. These two systems contribute differentially to initial acquisition of episodic memories, and subsequent long-term retention, with transfer of information between these systems through offline consolidation. In the second section, we consider predictions from CLS accounts concerning the acquisition of new spoken words. These predictions are framed within the computational context of the Distributed Cohort Model of spoken word recognition (Gaskell & Marslen-Wilson 1997), and neuroanatomical constraints suggested by dorsal and ventral pathway accounts of spoken language processing (Scott & Johnsrude 2003; Hickok & Poeppel 2004; Davis & Johnsrude 2007). Unifying these two aspects into a single, neurocomputational account of word learning requires that we: (i) specify the functional contribution of neocortical systems that retain long-term representations of spoken words, and (ii) explain how hippocampal/medial temporal memory systems interface with neocortical systems during acquisition. We will propose specific computational signatures of processes supported by rapid, hippocampal learning as distinct from slower neocortical learning mechanisms. Of particular interest, here, is the proposal that learning involves overnight, sleep-associated consolidation processes to mediate between fast-learning hippocampal and slow-learning neocortical systems. We will also derive specific predictions for behavioural differences between newly learned and consolidated novel words. Next we will review behavioural, neuropsychological and functional imaging evidence relating to this dual-process account. We begin by assessing behavioural evidence for two distinct stages of word learning. A particular focus of recent behavioural research has been to explore cognitive processes such as lexical competition that are unique to familiar words. Consistent with the CLS accounts we review evidence that newly learned words only become effective lexical competitors after offline consolidation. The next two sections review neuropsychological and functional imaging evidence for the involvement of two distinct neural systems in the medial temporal lobe and neocortex in word learning. We examine the evidence of impaired word learning in amnesic individuals with acquired or developmental injuries to the hippocampus, for whom word learning should be severely impaired according to the CLS account. Evidence that the medial temporal lobe contributes to initial word learning is also provided by functional imaging studies in which participants are scanned while learning new words. We then review evidence for cortical involvement in word recognition and learning. A particular focus for the CLS account is evidence from functional brain imaging studies in which neural responses are measured at different stages during the acquisition of novel words. Interpretation of these findings requires background information on the cortical systems that respond differently to familiar and unfamiliar spoken words. We therefore report a meta-analysis of functional imaging studies that compare words and pseudowords and test predictions derived from this meta-analysis for changes in neocortical responses to pseudowords following initial learning and consolidation. Finally we set out some outstanding questions and directions for future research. THE CLS MODEL The CLS model has much in common with other accounts of memory that invoke distinct learning systems specialized for different types of memory (e.g. procedural versus declarative, Squire 1992; semantic versus episodic, Tulving 1972). However, the CLS account goes beyond these traditional descriptions by considering the computational properties of neural networks that learn from experience. An important distinction for network models is whether learned information is stored in independent, sparse codes (in which distinct memories have little or no overlap in