Learning and consolidation of new vocabulary/1 Learning and consolidation of new vocabulary in autism spectrum disorder Henderson

Autism spectrum disorder (ASD) is characterized by rich heterogeneity in vocabulary knowledge and word knowledge that is not well accounted for by current theories. We take a novel approach and examine whether individual differences in vocabulary knowledge might be partly explained by a consolidation and/or integration impairment. Verbally able children with ASD and typical peers showed similar improvements in recognition and recall of novel words (e.g., biscal) 24-hours after training. Typical children showed competition for exiting words (e.g., biscuit) after 24-hours (suggesting that the new words had been integrated with existing knowledge) whereas children with ASD showed immediate competition effects that diminished after 24-hours. Thus, children with ASD showed strengths in the consolidation of explicit memory for new spoken word forms but weaknesses with the integration of new and existing word knowledge over time. These results are considered from the perspective of a dual-memory systems framework. Learning and consolidation of new vocabulary/4 ASD is characterised by impairments in social interactions and social communication and repetitive and stereotyped patterns of behaviour (Lord & Jones, 2012). Many children with autism spectrum disorders (ASD) have pronounced and protracted delays in language acquisition and for a substantial proportion of these children problems with language and communication are life-long. Yet, there is striking heterogeneity in relation to vocabulary development and word knowledge, with some individuals achieving typical and even above average vocabulary scores (Kjelgaard & TagerFlusberg, 2001; Mottron, 2004; Luyster, Lopez & Lord, 2007) and others showing clear vocabulary impairments despite age-appropriate cognitive skills (Loucas et al., 2008). Language impairments in ASD have been causally attributed to aspects of autistic pathology; for instance, a failure to follow ゲヮW;ニWヴ ェ;┣W I┌Wゲが ラヴ ┌ミSWヴゲデ;ミS デエW ゲヮW;ニWヴげゲ intention may derail early word learning (BaronCohen, Baldwin & Crowson, 1997). “キマキノ;ヴノ┞が ; け┘W;ニげ Sヴキ┗W デラ キミデWェヴ;デW キミaラヴマ;デキラミ マ;┞ Sキゲヴ┌ヮデ デエW ability to learn new information from context (Happe, 1999). However, both theories anticipate pronounced language impairments across the autism spectrum; neither can explain the rich variation of language phenotypes that characterise ASD. While individual differences in social cognition or central coherence may inform our understanding language variation, it is likely that some variance is explained by other aspects of development. It is therefore imperative that we reveal the factors that contribute to language learning in ASD, in order to develop well-tailored intervention programmes. This study takes a novel approach to investigating language learning in ASD. We investigate the time course of new word learning, specifically examining the role of off-line consolidation in learning the phonological form of a new word and the extent to which it has been integrated with existing lexical knowledge. Previous studies of vocabulary acquisition in ASD have largely relied on paradigms in which children are briefly exposed to a novel word and then assessed immediately after learning (e,g., the fast mapping paradigm). Traditionally, research has focused on the social deficits associated with impaired language acquisition (Baron-Cohen, Baldwin & Crowson, 1997; Preissler & Carey, 2005; McDuffie et al., 2006; Parish-Morris et al., 2007; Luyster & Lord, 2009). This research has Learning and consolidation of new vocabulary/5 demonstrated that although many children with ASD are impaired at interpreting social cues, they can learn new words when social cues are salient (Parish-Morris et al., 2007; Luyster & Lord, 2009) perhaps by relying on associative learning mechanisms (Parish-Morris et al., 2007; Preisseler, 2008). In non-social tasks, children with ASD can use mutual exclusivity to fast-map novel words to novel objects over objects they already know (de Marchena, Eigsti, Worek, Ono & Snedeker, 2011; Preissler & Carey, 2005). They also demonstrate the noun bias during word learning, in which a novel word is taken to represent an object rather than an action (Swensen, Kelley, Fein & Naigles, 2007). Reported differences in speech perception in ASD are also relevant when considering how new vocabulary might be acquired (Kuhl, 2007). A number of studies have reported superior processing of auditory information, including enhanced performance relative to typically developing (TD) peers on neural responses to frequency changes (Kujala, Aho et al., 2007), frequency discrimination and categorisation (Bonnel et al., 2003), and processing of the pitch contours of sentences (Jarvinen-Pasley, Wallace, Ramus, Happe, & Heaton, 2008). Such enhanced performance may be more characteristic of individuals with ASD who go on to have structural language abilities that meet or exceed age expectations, despite earlier delays in language acquisition, suggesting that it may act as a compensatory or protective mechanism for language learning (Jones et al., 2009). Despite strengths in the initial mapping of a new word to a new referent and with enhanced phonological processing of speech, children with ASD generally have smaller vocabularies than expected for their age (Charman, Drew, Baird & Baird, 2003; Hudry et al., 2010; Tager-Flusberg, Paul & Lord, 2005). Even when children with ASD are well matched to controls on measures of verbal ability they show qualitative differences in how they activate vocabulary knowledge in the service of language comprehension (Henderson, Clarke & Snowling, 2011; McCleery et al., 2010). Hence, the mechanisms underlying vocabulary acquisition in ASD remain poorly understood. The predominance of findings from fast-mapping studies may not be revealing about the extent to which a new word is fully acquired into the mental lexicon (Bedford et al., 2013; Horst & Samuelson, 2008). Indeed, few studies have considered that word learning is protracted or that it Learning and consolidation of new vocabulary/6 relies on off-line consolidation. One exception is a word-learning study by Norbury, Griffiths and Nation (2010). Participants were asked to define and name novel objects (testing semantic and phonological knowledge, respectively), immediately after learning and four weeks later. For verbally able participants with ASD, recall of phonological information was impressive at both time points. In fact, they outperformed TD controls at mapping phonological forms to novel referents immediately after learning. Norbury et al. argued that this strength in mapping novel phonological forms to novel objects reflects enhanced phonological and associative (declarative) learning mechanisms. However, while TD children showed evidence of consolidation, with significantly improved performance on both tasks after four weeks, children with ASD did not show any such improvement. These results provide novel evidence of qualitative differences in the consolidation of new vocabulary in ASD. The measurement of word learning in Norbury et al. was restricted to recognition and recall tasks, providing limited insights into the extent to which a new phonological form has become fully word-like. No studies have explicitly tested the time course with which new lexical representations are integrated with existing knowledge in ASD (cf. Gaskell & Dumay, 2003). According to dualmemory systems frameworks (e.g., Hasselmo, 1999; McClelland et al., 1995) new memories are integrated with existing knowledge slowly, to prevent new information from over-writing or distorting existing memories. Episodic memories (for events that occur at a specific place and time) are proposed to be initially encoded in the connections of the hippocampus but then through off-line consolidation a neocortical representation is strengthened and integrated with semantic memory. Davis and Gaskell (2009) advocated a dual-memory systems approach to vocabulary acquisition, a key prediction being that competition during recognition between novel words and similar sounding well-established lexical neighbours occurs only after the new lexical entry has been integrated into the lexicon and has reshaped existing neocortical networks involved in language processing. Gaskell and colleagues have examined how lexical competition changes when adults (Dumay & Gaskell, 2007; Gaskell & Dumay, 2003) and children (Brown, Weighall, Henderson & Gaskell, 2012; Henderson, Weighall, Brown & Gaskell, 2012; Henderson, Weighall, Brown & Gaskell, Learning and consolidation of new vocabulary/7 2013) learn fictitious novel nonwords (e.g., さbiscalざ) that are close neighbours of established words (e.g., biscuit). Participants made speeded decisions about the presence of a pause inserted toward the offset of existing words (e.g., さHキゲIぱ┌キデざ). Pause detection latencies in existing words became slower if participants had recently learned an onset competitor. This finding is argued to reflect the increased amount of lexical activity at pause onset once a novel competitor has been learned and a subsequent reduction in processing resources allocated to the task of detecting the pause (Mattys & Clark, 2002). Crucially, lexical competition emerged 12-hours after exposure to the nonword competitors, but only if sleep occurred (Dumay & Gaskell, 2007; Henderson et al., 2012). Children and adults were also able to recall significantly more newly learned nonwords after sleep than after a similar period of time awake. This suggests that off-line consolidation during sleep not only aids the integration of novel words into the mental lexicon 1 but also strengthens explicit memory. Importantly, previous studies with children have shown that improvements in explicit memory and the delayed emergence of lexical competition are not dependent upon re-exposure to the novel stimuli in the tests (Brown et al., 2012; Henderson et al., 2012) and they occur for real as well as fictitious words that are trained with or without meaning (Henderson, Weighall & Gaskell, 2013). This body of research is strengthened considerably by two key findings. First, fMRI data show hippocampal sensitivity to the familiarity of nonwords encountered on the day of scanning, and neocortical consolidation effects of recently learned nonwords (Davis, Di Betta, MacDonald, & Gaskell, 2009). Second, sleep spindle activity (11-15 Hz oscillations lasting up to 3 seconds) is positively associated with overnight increases in lexical competition observed for existing word competitors of taught novel words suggesting that they play an active role in lexical consolidation (Tamminen, Payne, Stickgold, Wamsley & Gaskell, 2010). This is consistent with the view that spindles are implicated in hippocampal-neocortical consolidation (Diekelmann & Born, 2010). Spindles increase in activity during the up-state of slow oscillations (Molle, Marshall, Gais & Born, 2002) and are temporally aligned with hippocampal ripples (Sirota, Czicsvari, Buhl & Buzsaki, 2003). Learning and consolidation of new vocabulary/8 ASDs are characterised by aberrant structural and functional neural connectivity, which could disrupt hippocampal-neocortical interactions (Belmonte et al., 2004; Just et al., 2004; Herbert et al., 2004; Herbert, 2005). Furthermore, children and adults with ASD experience elevated rates of sleep disturbance relative to TD peers (Hoban, 2000; Wiggs & Stores, 2004), including decreased sleep spindle activity (Limoges et al., 2005), and disrupted sleep in ASD has been associated with poor receptive vocabulary (Malow et al., 2009). Many individuals with ASD (42-82%) show epileptiform activity during sleep in the absence of clinical seizures (Giovanardi Rossi, posar & Parmeggiani, 2000; Lewine et al., 1999; Richdale, 1999; Tuchman, Rapin & Shinnar, 1991; Tuchman, 2000). Hence, it is plausible that qualitative differences in vocabulary acquisition in ASD could stem from aberrant sleep-associated consolidation processes (Femia & Hasselmo, 2002). We do not know how individual differences in consolidation may be related to individual differences in language learning within ASD. No previous research has systematically investigated this issue; therefore, this study will have immediate implications for understanding language heterogeneity in children with ASD. Children were trained on novel nonwords ふWくェくが さbiscalざぶ as used in Henderson et al. (2012). For TD children, it was エ┞ヮラデエWゲキ┣WS デエ;デ ノW┝キI;ノ IラマヮWデキデキラミ WaaWIデゲ aラヴ W┝キゲデキミェ ┘ラヴSゲ ふWくェくが さbiscuitざぶ would emerge 24-hours after exposure to the novel competitors but not immediately (Henderson et al., 2012; Henderson et al., 2013). We hypothesized that TD children would show significant improvements in their ability to explicitly recall and recognize the novel words 24-hours after exposure (Brown et al., 2012; Henderson et al., 2012, 2013). Such evidence would lend further support to the dual-systems account of vocabulary acquisition (Davis & Gaskell, 2007). For verbally able children with ASD, we anticipated similar performance to TD peers immediately after training (Norbury et al., 2010). However, based on the view that children with ASD have impairments in offline consolidation, we predicted that children with ASD would show smaller improvements in explicit recall and recognition of the novel words (Norbury et al., 2010) and a reduced lexical competition effect 24 hours after exposure.