Word learning in the context of semantic prior knowledge: evidence of interference from feature-based neighbours in children and adults

Emma James; M. Gareth Gaskell; Gráinne Murphy; Josie Tulip; Lisa M. Henderson

doi:10.1080/23273798.2022.2102198

Word learning in the context of semantic prior knowledge: evidence of interference from feature-based neighbours in children and adults

James, Emma; Gaskell, M. Gareth; Murphy, Gráinne; Tulip, Josie; Henderson, Lisa M. 2023-02-07 00:00:00 LANGUAGE, COGNITION AND NEUROSCIENCE https://doi.org/10.1080/23273798.2022.2102198 REGULAR ARTICLE Word learning in the context of semantic prior knowledge: evidence of interference from feature-based neighbours in children and adults a b b b b Emma James , M. Gareth Gaskell , Gráinne Murphy , Josie Tulip and Lisa M. Henderson a b Department of Experimental Psychology, University of Oxford, Oxford, UK; Department of Psychology, University of York, York, UK ABSTRACT ARTICLE HISTORY Received 8 April 2022 The presence of phonological neighbours facilitates word-form learning, suggesting that prior Accepted 7 July 2022 phonological knowledge supports vocabulary acquisition. We tested whether prior semantic knowledge similarly beneﬁts word learning by teaching 7-to-10-year-old children (Experiment 1) KEYWORDS and adults (Experiment 2) pseudowords assigned to novel concepts with low or high semantic Vocabulary; word learning; neighbourhood density according to feature norms. Form recall, deﬁnition recall, and semantic prior knowledge; semantic categorisation tasks were administered immediately after training, the next day, and one week neighbourhood density; later. Across sessions, pseudowords assigned to low-density (versus high-density) semantic consolidation; development neighbourhood concepts elicited better word-form recall (for adults) and better meaning recall (for children). Exploratory cross-experiment analyses demonstrated that the neighbourhood inﬂuence was most robust for recalling meanings. Children showed greater gains in form recall than adults across the week, regardless of links to semantic knowledge. While the results suggest that close semantic neighbours interfere with word learning, we consider alternative semantic dimensions that may be beneﬁcial. Our ability to learn new words remains important across The role of prior knowledge in new word acquisition the lifespan, enabling us to communicate about new can be examined both in terms of the initial learning of a things we experience. Encountering an unfamiliar food new word and its long-term retention in vocabulary, as from a diﬀerent cuisine or a rare animal at the zoo, for dissociated within a complementary learning systems instance, requires us to learn its association with a new framework (Davis & Gaskell, 2009; McClelland et al., combination of sounds, encode its key features, and 1995). According to this model, vocabulary knowledge determine how it relates to concepts that we already is stored in a distributed manner in neocortical regions know. For example, we might learn that a pomelo is a of the brain, permitting eﬃcient language processing type of fruit that is similar to a grapefruit, but larger and and communication (Gaskell & Marslen-Wilson, 1997). sweeter. In using what we know about grapefruits to For a new word to become part of this system, it must support new learning, we bring a broader understanding do so in such a way to avoid disrupting existing connec- of citrus fruit features—inferring that a pomelo likely has a tions. The complementary learning systems model pro- waxy peel, pips, and juicy ﬂesh. However, the availability poses that initial lexical representations are formed of this related knowledge will vary across new concepts using the hippocampal memory system, with the new we encounter, and it is not clear the extent to which it form-meaning mapping not yet fully integrated with helps or hinders memory for the new word. We addressed existing neocortical vocabulary (Davis & Gaskell, 2009). this question by examining how children’sand adults’ This system allows knowledge of a new word to be learning of novel concepts is aﬀected by the density of acquired rapidly without disrupting existing connec- associated semantic neighbourhoods. Determining the tions, and the new representation can then support role of prior knowledge has important implications for slower integration of the new word into neocortical- theories of word learning, as well as for understanding based networks. how word learning changes as semantic knowledge The strengthening of neocortical connections is pro- accumulates across development. posed to result from repeated reactivations of the CONTACT Emma James emma.james@psy.ox.ac.uk Department of Experimental Psychology, University of Oxford, Anna Watts Building, Oxford, OX2 6GG, UK Supplemental data for this article can be accessed http://doi.org/10.1080/23273798.2022.2102198. © 2022 The Author(s). Published by Informa UK Limited, trading as Taylor & Francis Group This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. 2 E. JAMES ET AL. hippocampal representation, which can occur “oﬄine” One way of examining the inﬂuence of prior knowl- during sleep. In line with this, increased competition is edge on word learning is to manipulate the psycholin- observed between new and familiar words in lexical pro- guistic properties of the to-be-learned words. By cessing tasks following sleep but not wake in both adults quantifying the similarities between new items and real (Dumay & Gaskell, 2007) and children (Henderson et al., words that would be known to the learner in diﬀerent 2012), suggesting that sleep enhances integration with ways, we can examine whether these relationships neocortical vocabulary. These sleep-associated consoli- predict memory for new words before and/or after oppor- dation processes are further associated with improved tunities for consolidation. In examining word-form simi- accuracy and eﬃciency in recalling the new word- larity, many studies have demonstrated that novel forms (Ashworth et al., 2014; Gais et al., 2006; James words with many phonological neighbours (i.e. words et al., 2020a; Tamminen et al., 2010). Deliberate retrieval that diﬀer by a single sound) are more readily learned practice may also enhance recall of new word knowl- than words with fewer phonological neighbours. This edge (e.g. Hulme & Rodd, 2021), proposed to operate beneﬁt has been found for pre-school children (Hoover via similar principles of hippocampal reactivation et al., 2010;Storkel, 2009;Storkel et al., 2013), school- (Antony et al., 2017). These ﬁndings converge on the aged children (James et al., 2019;van derKleij et al., importance of memory reactivation processes in sup- 2016), and adults(Jameset al., 2019; Storkel et al., porting vocabulary acquisition, and highlight the impor- 2006), suggesting that individuals can access prior knowl- tance of understanding factors that inﬂuence this edge to support learning across the lifespan (although longer-term consolidation of new word knowledge as note that this support may be less apparent in incidental well as those that support initial encoding. word learning; James et al., 2020b). From a complemen- One factor proposed to support lexical consolidation tary learning systems perspective, James et al. (2019) is the learner’s prior linguistic knowledge (James et al., found that this early beneﬁt from phonological neigh- 2017). Recent progress in complementary learning bours was reduced one week later, once words with systems theory has examined how the speed of neocor- fewer neighbours had opportunities to beneﬁt from tical learning may be inﬂuenced by memory schema oﬄine consolidation processes. These ﬁndings support (e.g. Kumaran et al., 2016; McClelland, 2013; McClelland the proposal that items related to prior word-form knowl- et al., 2020), driven by evidence that new information edge may be integrated into neocortical systems more related to existing knowledge becomes integrated into rapidly, and are therefore less dependent on oﬄine reac- neocortical networks very rapidly (Tse et al., 2007). For tivation processes to support long-term memory. example, McClelland (2013) demonstrated that a In this study, we turn our attention towards semantic neural network trained with structured knowledge of neighbours to develop a broader understanding of how birds and their semantic properties (e.g. grow, move, diﬀerent aspects of prior knowledge inﬂuence memory ﬂy) showed more rapid learning of a typical new for new vocabulary. Learning the meaning of a new example that shared these properties (a cardinal) than word is a crucial part of vocabulary acquisition itself, an atypical example (a penguin, which cannot ﬂy but and also supports the long-term retention of new can swim). This model demonstrates how information word-form knowledge (Henderson et al., 2013). that capitalises upon existing knowledge can be inte- However, while psycholinguistic measures that capture grated into neocortical networks without requiring semantic neighbourhoods have well-documented inﬂu- repeated hippocampal reactivations (Tse et al., 2007). ences on the processing of known words (see Pexman, From a language learning perspective then, it follows 2020, for a recent review), they have received relatively that individuals with rich lexical networks will likely little attention in experimental studies of word learning. have more relevant structures to draw upon to Network approaches to modelling language acquisition support new word learning, and thus consolidation support that semantic neighbours broadly predict voca- into neocortical networks is proposed to proceed more bulary growth, ﬁnding that new words are more likely to rapidly. Capitalising upon existing knowledge may be be acquired when they are highly connected to known one means by which individuals with good vocabulary ones (Engelthaler & Hills, 2017; Hills et al., 2009), knowledge acquire new vocabulary at faster pace than although semantically distinct words may be acquired those with weak vocabulary, thereby increasing the earlier (Engelthaler & Hills, 2017). Yet only two studies “vocabulary gap” in performance across development to our knowledge have examined the learning and (James et al., 2017). However, the nature of this lexical retention processes that might underlie these inﬂu- support is not well-speciﬁed, and it is not clear how ences: Tamminen et al. (2013) in a study of sleep-associ- word learning might be inﬂuenced by diﬀerent aspects ated consolidation in adults, and Storkel and Adlof of linguistic knowledge. (2009) in a study of preschool children. In contrast to LANGUAGE, COGNITION AND NEUROSCIENCE 3 the beneﬁts seen for word-form neighbours, the ﬁndings processes enhance engagement of the new words of both studies point towards interference from related with existing semantic knowledge. semantic knowledge when learning new words. Thus, unlike for word-form neighbours, it appears from Tamminen et al. (2013) taught adults pseudowords as the existing research that connections to semantic knowl- names for novel concepts with either sparse or dense edge may interfere with new word acquisition. However, semantic connections. These novel concepts were semantic relationships can be conceptualised along created from existing concepts with either low- or high- several diﬀerent dimensions (Hameau et al., 2019), and density semantic neighbourhoods, as quantiﬁed by the diﬀerent measures have been shown to contribute number of associates provided in free association norms unique variance in predicting performance in speeded (Nelson et al., 2004). To make each concept novel, they word recognition tasks (Pexman et al., 2008). Tamminen added a novel feature (e.g. bee whose sting feels pleasant, et al. (2013) derived their measures of semantic neigh- crab that has a beak). Participants were tested on their bourhood from free-association norms, which were knowledge of the new pseudowords immediately after created by asking participants to produce the ﬁrst word learning, the next day, and one week later—allowing for that came to mind when presented with the target an assessment of the inﬂuence of prior knowledge (Nelson et al., 2004). Free-association norms are con- before and after opportunities for consolidation. Across sidered a language-based measure that typically reﬂects all test sessions, participants had poorer performance for the co-occurrence of concepts in spoken and/or written high- versus low-density concepts in a synonym judge- language, and can be broadly distinguished from object- ment task, which required them to identify which of based metrics that draw upon the content of the concepts three familiar words was associated with the trained themselves (Buchanan et al., 2001). For example, feature word. This perhaps indicates that participants experi- production norms are collected by asking participants enced interference from existing knowledge in learning to list features of target concepts (McRae et al., 2005), pro- the new concepts. However, there was no inﬂuence of ducing measures of semantic richness (i.e. the number of the density of prior semantic knowledge on explicit features produced) and common features between con- recall of the new word-forms or their meanings, cepts. Semantic neighbourhoods conceptualised in this suggesting that this interference might arise from way may arguably be more beneﬁcial to a new learner lexical processing during the synonym judgement than linguistic co-occurrence. For example, the target task rather than during learning per se. In line with “bird” leads to “cat” as a more frequent lexical associate this, responses only slowed to high-density novel than “robin” (Nelson et al., 2004), yet knowledge of items in a speeded semantic categorisation task one birds is intuitively more useful when learning about a week later, suggesting that a longer period of consoli- robin than a cat. In this study, we test the hypothesis dation enhanced competition between new and that semantic prior knowledge deﬁned by object-based known concepts as they became better integrated in similarities will facilitate word learning and consolidation. memory (similar to the way in which lexical compe- This proposal garners support from studies of early tition was observed following sleep in the studies language learning that capture the variability in knowl- described above). Thus, dense existing semantic knowl- edge that preschool children bring to the task. For edge elicited interference when processing the novel example, Borovsky et al. (2016)demonstrated that concepts in the context of familiar words, but there infants were more able to learn and recognise new was no evidence that memory for the new words them- words from categories that they had more knowledge selves was aﬀected by links to semantic knowledge. about compared to categories for which they had lower However, evidence of semantic interference in estab- levels of existing knowledge. Similarly, Perry et al. (2016) lishing new word representations comes from a study found that preschool children with larger shape-based with children. Storkel and Adlof (2009) quantiﬁed noun vocabularies were more likely to remember object semantic set sizes of non-objects by collecting free shapes during word learning. These studies support the associations from presented line drawings. In a sub- proposal that semantic knowledge relevant to a new con- sequent word-learning task, preschool children were cept’s properties may aid acquisition, although they more accurate in identifying nonwords associated with cannot address questions of longer-term memory as chil- objects from small semantic set sizes, suggesting that dren were only tested on the same day as learning. —at least in young children—connections to existing semantic knowledge can interfere in learning and/or The present study remembering new information. In this study, the eﬀect only emerged at a delayed test one week later, We used a similar design to Tamminen et al. (2013), but oﬀering further support to the hypothesis that oﬄine instead drew upon shared features as an object-based 4 E. JAMES ET AL. measure of semantic relationships to test whether example, Davies et al. (2017) showed that eﬀects of psy- semantic neighbourhood density can beneﬁt word cholinguistic variables on lexical processing decline learning (rather than the associative measure used pre- across the lifespan as the lexical system accumulates viously). We taught participants pseudowords and experience and maximises learning eﬃciency. We associated deﬁnitions, formed by adding a novel present two experiments to explore these possibilities feature to known concepts from low and high feature across children aged 7–10 years (Experiment 1) and density neighbourhoods. We tested explicit recall of adults (Experiment 2). While the developmental studies the pseudowords and deﬁnitions immediately after described above typically assessed semantic prior learning, the next day, and one week later, to examine knowledge inﬂuences in pre-school children, selecting the inﬂuence of existing knowledge on word learning a school-aged group allowed us to use the same exper- before and after opportunities for consolidation. We imental tasks for children and adults, thus facilitating also used a speeded semantic categorisation task to developmental comparisons. Further, these age groups index integration of the new words into neocortical overlap with those that have been examined in studies vocabulary, capitalising on previous ﬁndings that of phonological neighbours in word learning (James known words with high semantic neighbourhood et al., 2019; James et al., 2020b), allowing comparisons density are responded to more quickly in this task than with diﬀerent types of prior knowledge inﬂuence. words with low semantic density (Mirman & Magnuson, Our main research questions were as follows: First, is 2008). Tamminen et al. (2013) found that implicit seman- explicit memory for novel words helped or hindered by tic neighbourhood eﬀects for trained pseudowords links to existing semantic knowledge during word learn- emerged only after a period of consolidation, consistent ing, as deﬁned by object-based norms? Second, can with other studies that have found implicit semantic newly trained concepts acquire the lexical properties activation to emerge after one/more periods of sleep of their neighbours, beneﬁting from rich semantic con- (Clay et al., 2007; Tham et al., 2015). nections in speeded reaction time tasks? Third, what is One motivation for understanding the role of seman- the time course of this engagement with semantic tic prior knowledge is to understand how word learning knowledge? By using both explicit and implicit measures might change across development. The studies of new word knowledge, we aimed to capture initial described above span a broad age range from preschool encoding processes and those that indicate integration children to adults, yet there is a lack of direct develop- with existing knowledge, proposed to require periods mental comparisons. A study of known words taken of oﬄine consolidation. Finally, across experiments we from linguistic corpora suggested that young infants also explored whether children and adults are diﬀerently start by learning words from sparse semantic neigh- inﬂuenced by semantic knowledge during word learn- bours but increasingly beneﬁt from dense neighbours ing, reﬂecting diﬀerences in the amount of prior knowl- across the preschool years (Storkel, 2009). However, edge or their sensitivity to it during learning. In the ﬁrst these developmental diﬀerences have not been tested experiment with children, we thus tested three exper- experimentally, and no studies to our knowledge have imental hypotheses: 1) A large number of shared fea- considered the inﬂuence of semantic neighbours in tures will facilitate explicit aspects of word learning, as school-aged children. There are two possibilities here: demonstrated by superior performance in recall and rec- ﬁrst, school-aged children may show smaller semantic ognition tasks; 2) Novel concepts that share lots of fea- density eﬀects than adults given that our selected tures with existing concepts should show a reaction measure is based on adult norms, and children may time advantage when compared to novel concepts not have yet acquired rich enough knowledge about that share fewer features, in a speeded semantic categ- concepts to have such extreme diﬀerences in low- orisation task; and 3) Across tasks, eﬀects of neighbour- versus high-density items. Speaking to this proposal, hood density (i.e. better recall/recognition for high Pexman and Yap (2018) found that adults with better density items; a density eﬀect in speeded semantic cat- existing vocabulary knowledge showed more sensitivity egorisation) will emerge only after a night’s sleep (24- to semantic neighbourhood density in speeded hour test) or longer period of consolidation (week responses during a semantic categorisation task com- follow-up test). pared to adults with lower vocabulary knowledge, suggesting that the knowledge learners bring to the Experiment 1 task is highly relevant. Alternatively, children might show larger semantic density eﬀects under the possi- Experiment 1 can be considered exploratory in the sense bility that an underdeveloped system may be more sen- that it was not pre-registered, and that the sample size sitive to the inﬂuence of existing knowledge. For was determined opportunistically (school availability LANGUAGE, COGNITION AND NEUROSCIENCE 5 within the timeframe of the study). However, our analy- pronounceable (Appendix 1), and these were split into sis plan and key hypothesis tests were consistent with a two lists matched on length, number of orthographic pre-registered adult study being conducted in parallel neighbours, and bigram frequency (Marian et al., (Experiment S1, http://osf.io/3vnsg; detailed below). 2012). A subset of 16 pseudowords were selected for Experiment 1 with children. Novel concepts were created by taking an existing Experiment 1 methods base concept and adding an additional feature (further details below). For example, a gorilla (base concept) Participants that has green skin (added feature). Half of the base con- Two whole classes of children took part in the study, cepts were animals, for purposes of the semantic categ- recruited via two schools in North Yorkshire. The result- orisation task. Critically, the base concepts were selected ing sample included 51 children (25 male) aged 7–10 for having high (n = 12) or low (n = 12) semantic neigh- years (M = 8.67 years). One additional child was excluded bourhood density according to the McRae et al. (2005) from analyses due to hearing diﬃculties. Two of the 2 feature norms. In this ﬁrst step towards examining included children were absent on the second day of feature-based semantic neighbourhoods in word learn- testing, and thus only contributed data for two out of ing, we took a broad approach to deﬁning semantic the three follow-up tests. neighbourhood density. First, we selected for the The study was approved by the Research Ethics Com- number of features of each item (also termed semantic mittee of the Department of Psychology, University of richness), as has consistently been shown to facilitate York. Consent was obtained from the school head tea- lexical processing (Pexman, 2020). Second, we con- chers. Parents were fully informed about the study and sidered the density of the semantic neighbourhood by were given the opportunity to opt their child out of selecting for low versus high intercorrelational feature taking part. density—the extent to which the listed features co- occurred in other normed concepts. This metric is described in detail within the database documentation Design and procedure (McRae et al., 2005): pairs of features are considered sig- Children completed a single training session in a whole- niﬁcantly correlated if they share ≥ 6.5% of their variance class setting, which lasted approximately 45 min. Test within the database (i.e. they often co-occur together in sessions were then conducted individually in a quiet the 541 normed concepts), and the proportion of signiﬁ- setting outside the classroom at three time points: the cantly correlated feature pairs is calculated for each same day (T1), the next day (T2), and one week later concept. We predicted that this co-occurrence would (T3). Standardised assessments of vocabulary and non- also support learning, being indicative of many shared verbal ability (matrix reasoning) from the Wechsler features that could support processing (Grondin et al., Abbreviated Scale of Intelligence II (Wechsler, 2011) 2009) and representing many existing connections were also collected during these sessions for descriptive between concepts. purposes. The mean t-scores were within the average Our selected low-density base concepts had fewer range for both matrix reasoning (M = 46.53, SD = 9.73) features listed in the norms (≤ 16), and fewer of these and vocabulary (M = 59.71, SD = 11.92). listed features (≤ 14%) co-occurred in other normed concepts. High-density base concepts had more features overall (≥ 18) and more of these (≥ 25%) also co- Stimuli occurred in other concepts. The two groups of stimuli Pseudowords were initially selected using the English were otherwise well matched on measures of frequency, Lexicon Project (Balota et al., 2007) according to the fol- age of acquisition, imageability, concreteness and word lowing criteria: 5–6 letters long, no orthographic neigh- length (Table 1). A pilot study of these base concepts bours, and a nonword rejection Z-score of −0.45 to 0.45 with adults supported a reaction time beneﬁt for high- (i.e. an average range response time for rejection in a density concepts in a semantic categorisation task lexical decision task). These criteria were used to (mean diﬀerence = 14 ms; t(70) = 2.56, p = .01). ensure that the word-forms were well-matched across The added features that made each concept novel conditions, and to minimise alternative sources of varia- were also selected from the McRae et al. (2005) norms, bility in the speeded semantic categorisation task. and each occurred only once in the norms to minimise Twenty-four bisyllabic pseudowords were selected in the inﬂuence of additional semantic neighbourhoods. total such that each began with diﬀerent vowels or con- The features were drawn from a range of perceptual, sonant clusters and were judged to be easily behavioural and functional categories, which were 6 E. JAMES ET AL. Table 1. Properties of stimuli in the low and high semantic neighbourhood density conditions. a a b c c d e No. of features % features correlated AoA Frequency Log10 freq Imageability Concreteness No. of phonemes Low 12.75 5.58 5.14 16.41 1.02 607.56 4.89 4.33 High 18.92 40.00 5.17 16.38 1.13 616.00 4.89 4.50 p <.001* <.001* .96 1 .56 .61 .90 .79 a b c d e McRae et al. (2005). Kuperman et al. (2012). CELEX English linguistic database (Baayen et al., 1995). MRC Psycholinguistic database (Coltheart, 1981). Brys- baert et al. (2014). *signiﬁcant diﬀerence between low vs high semantic density items at p < .05. matched in type across low- and high-density base con- Form repetition cepts (Appendix 1). To ensure that these combinations Children heard each new word-form spoken by the of base concepts and features did not diﬀer in plausi- experimenter, with its orthographic form projected on bility across low- and high-density conditions, 58 the PowerPoint at the front of the classroom. They adults completed online ratings of how plausible they repeated the pseudoword aloud twice, and sub- would ﬁnd each item in a children’s storybook. High- sequently copied it into their workbooks. and low-density items did not diﬀer in plausibility (ps >.2). A subset of 16 items (8 per density condition) Deﬁnition repetition were selected for Experiment 1. A single ﬁxed set of Children were introduced to the deﬁnition of each pseu- pseudoword-concept pairings was used for Experiment doword, and again repeated it aloud twice. 1 (with counterbalancing of pseudowords across density conditions introduced in Experiment 2). Drawing task Children were given 30 s per item to draw a picture of the new concept, designed to help them to engage with its diﬀerent features. These were not scored or ana- Training tasks lysed further. The training tasks were conducted with the class as a whole. Children were given workbooks to support their Meaning matching learning, and were guided through a number of tasks After the workbooks had been collected, further learning using a PowerPoint presentation projected at the front and feedback took place via a multiple choice quiz. In the of the classroom (see Figure 1). The ﬁrst three tasks ﬁrst round, a pseudoword and three possible options for were completed for each item in turn (form and its deﬁnition were presented on screen, and children had deﬁnition repetition, drawing), followed by the to show their answer by raising one, two or three ﬁngers. meaning matching task for all items. In total, children In the second round, the deﬁnition was presented and the heard each new word-form nine times, and each children had to choose the correct word-form to match. deﬁnition six times. Each item was presented once in each round, with the correct answer provided after each one. Test tasks Children completed the test tasks individually with the experimenter. There were four test tasks to assess diﬀerent aspects of word knowledge. All test tasks were presented using DMDX software v5.1.3.4 (Forster & Forster, 2003), with item order randomised. The tasks were presented in the following ﬁxed order. Cued form recall Children were presented with the ﬁrst consonant(s) and vowel of the word (both aurally and visually), and were asked to speak the remainder of the word. Children Figure 1. Schematic of the training tasks used in the learning phase. were encouraged to attempt partial responses even if Note: For each item, children ﬁrst repeated the new word aloud twice and they were not sure of the answer, and the experimenter wrote it down, before repeating the deﬁnition twice. They were given 30 transcribed the responses for scoring on the basis of s to draw a picture of the new concept. After completing these learning tasks for all items, children completed two rounds of multiple choice quizzes. whole word accuracy (0, 1). LANGUAGE, COGNITION AND NEUROSCIENCE 7 Form recognition contributing to model ﬁt(p <.2). We then used a Children were presented with auditory and orthographic forward “best-path” approach to test for the inclusion of presentations of the pseudoword alongside a corre- appropriate random slopes (Barr et al., 2013). The results sponding foil in which the ﬁnal vowel was changed presented are from the most complex model supported (see Appendix 1). Both of the written stimuli remained by the data. The data and analysis scripts are available on screen for up to 7 s, or until the child had selected on the OSF (https://osf.io/35ftn). Figures were made their answer with a key press response. using ggplot2 (Wickham, 2009). Speeded semantic categorisation Experiment 1 results Children were presented with each word-form visually Cued form recall and auditorily, and were asked to make speeded judge- ments about whether or not the concept was an animal Children recalled a mean proportion of .20 (SD = .40) of the using a key press response. They were asked to respond word-forms at T1, and performance improved substantially as quickly and accurately as possible, and each trial ter- over tests (Figure 2A; delay1: β = 0.95, SE = 0.05, Z = 21.10, p minated after a response or 7 s. To allow for adjustment <.001). Recall continued to improve between T2 (M = .51, to the task and response format, the experimental task SD = .50) and T3 (M = .80, SD = .40; delay2: β = 0.91, SE = was preceded by 24 practice trials using existing 0.07, Z = 13.35, p <.001).There wasnoinﬂuence of seman- English words, providing feedback for erroneous tic neighbourhood density in recall of word-forms, alone or responses. We analysed both the accuracy (0, 1) and in interaction with test session (ps>.6; Table 2). the response time (ms) for correct trials. Form recognition Cued meaning recall Children were given an auditory and visual presentation Children could successfully recognise the new word-forms of each word-form, and asked to provide as much of the at above chance levels at T1 (M= .83, SD = .38), and deﬁnition as they could remember. Verbal responses improved at subsequent tests (T2: M = .92, SD = .28; T3: were transcribed by the experimenter. A total of two M = .94, SD = .24). This eﬀect of test session was statistically points could be awarded per item for correctly recalling signiﬁcant across both contrasts (delay1: β = 0.39, SE = 0.05, the base concept and the added feature. Z= 8.07,p < .001; delay2: β = 0.21, SE = 0.10, Z = 2.08, p = .037), again demonstrating signiﬁcant improvements in form knowledge across the week. As with the recall of Analyses word-forms,there wasnoinﬂuence of semantic neigh- Data were analysed in R (R Core Team, 2015), using lme4 bourhood density on their recognition (ps > .18; Table 2). (Bates et al., 2015b) and ordinal (Christensen, 2015)to ﬁt mixed eﬀects models. For each dependent variable, we Cued meaning recall initially ﬁtted a model with ﬁxed eﬀects of test session, semantic density, and their interaction. Fixed eﬀects Children scored an average of .36 out of a maximum of 2 were deviance coded to enable interpretation of each points for each item at T1 (SD = .76). There were no sig- predictor in relation to the overall mean. Test session niﬁcant changes in performance across test sessions (ps is a three-level factor, and we set two orthogonal con- > .36; Table 3), but there was a signiﬁcant diﬀerence in trasts to interpret the data: delay1 tested for diﬀerences memory for words from diﬀerent semantic neighbour in memory performance without versus with opportu- conditions (β = −0.48, SE = 0.18, Z = −2.62, p = .009). nities for consolidation (T1 vs. T2&T3); delay2 tested for Children were better at recalling deﬁnitions with low continued changes across the week (T2 vs. T3). For semantic neighbourhood density (M = .47, SD = .84) models with discrete dependent variables, Wald’sZ than high semantic neighbourhood density (M = .26, was used to determine statistical signiﬁcance. For reac- SD = .67; Figure 2C). There was no evidence of an inter- tion times, we report signiﬁcance computed using the action between test session and semantic neighbour- lmerTest package (Kuznetsova et al., 2017). hood density (pruned from the ﬁnal model; p = .687). In light of earlier convergence issues in attempting to ﬁt maximal models (Barr et al., 2013), we adopted a parsimo- Semantic categorisation nious modelling approach for these experiments (Bates et al., 2015a). We ﬁrst ﬁtted a model with our ﬁxed Accuracy eﬀects of interest and random intercepts for participants Performance was very low on the semantic categoris- and items, and then pruned away the interaction if not ation task (M = .59, SD = .49). Neither test session nor 8 E. JAMES ET AL. Figure 2. Explicit recall performance by semantic neighbourhood condition and test session. Note: Proportion correct for (A) Form recall in Experiment 1; (B) Form recall in Experiment 2; (C) Meaning recall in Experiment 1; and (D) Meaning recallin Experiment 2. Individual points mark average participant recall for each condition for each test session. Error bars denote 95% conﬁdence intervals. semantic neighbourhood density inﬂuenced accuracy The data were log-transformed to remediate issues of on this task (all ps > .4; Table A2-1). skewness in model ﬁtting. We also removed responses < 200 ms or that were ≥ 2.5 standard deviations above each participant’s condition mean. We analysed RTs to Reaction time correct responses only, leaving 49.05% of original trials. We were cautious in analysing the RT data considering Responses were slowest at T1 (M = 2154 ms, SD = that performance accuracy was so low in this task, and 1211 ms) compared to later test points (β = −0.07, SE = removed participants who were at/below chance perform- 0.01, t = −7.66, p < .001), but the decrease in response ance (n = 11). This left 40 participants in the analysis, who times between the T2 (M = 1833ms, SD = 1150 ms) and ranged from .52-.83 in categorisation accuracy (M =.63). T3 (M = 1696 ms, SD = 977 ms) tests were not statistically Table 2. Final analysis models for Experiment 1: Form tasks. a b Cued form recall Form recognition β SE Z p β SE Z p (Intercept) 0.02 0.28 0.07 .946 2.89 0.25 11.38 <.001 delay1 0.95 0.05 21.10 <.001 0.39 0.05 8.07 <.001 delay2 0.91 0.07 13.35 <.001 0.21 0.10 2.08 .037 density 0.07 0.25 0.28 .780 −0.29 0.22 −1.34 .181 Note: (a) Analysis based on from 2416 observations across 51 participants and 16 items. The ﬁnal model included by-participant and by-item intercepts. The two-way interaction between time and density was pruned from the model with no reduction in model ﬁt χ = 0.88, p = .64; (b) Analysis based on from 2416 observations across 51 participants and 16 items. The ﬁnal model included by-participant and by-item intercepts, as well as by-participant random slopes for the eﬀect of density. The two-way interaction between time and density was pruned from the model with no reduction in model ﬁt χ = 1.76, p = .41. LANGUAGE, COGNITION AND NEUROSCIENCE 9 Table 3. Final analysis model for Experiment 1: Cued meaning concepts. This hindrance was observed at the immediate recall. test and did not change at the delayed tests. High- rela- β SE Z p tive to low-density semantic knowledge activated 0|1 2.11 0.27 7.69 <.010 during encoding may thus interfere in forming the new 1|2 2.19 0.27 7.99 <.001 representation, and/or could make the novel concept delay1 0.01 0.04 0.22 .827 delay2 −0.07 0.08 −0.9 .368 harder to retrieve amongst its competitors at test. density −0.48 0.18 −2.62 .009 In Experiment 2, we examined the contribution of Note: Analysis based on from 2416 observations across 51 participants and semantic neighbourhood density to word learning in 16 items. The ﬁnal model included by-participant and by-item intercepts. The two-way interaction between time and density was pruned from the adults. In an adult study carried out in parallel to Exper- model with no reduction in model ﬁt χ = 0.75, p = .69. iment 1, we found no inﬂuence of semantic neighbour- hood density in a comparable word learning experiment signiﬁcant (β = −0.03, SE = 0.02, t = −1.84, p = .066). (Experiment S1, available on the OSF at https://osf.io/ There was no inﬂuence of semantic neighbourhood ksdfu/). Speciﬁcally, adults did not show interference from dense semantic neighbourhoods in recalling the density on reaction times (ps > .14; Table A2-2). word meanings, as we observed for children. However, adults showed much higher levels of recall performance than children, and diﬀerent training tasks were used Experiment 1 discussion across the two experiments. To facilitate developmental Experiment 1 examined how children learn and remem- comparisons, we repeated the experiment with adults ber pseudowords paired with novel concepts. Children using the same training procedures as Experiment 1, recalled 20% of the pseudowords on the same day as but we reduced the number of exposures during train- learning, but showed substantial improvements across ing to ensure comparable levels of performance the week: averaging 51% and 80% at the day and between age groups. week follow-up tests, respectively. However, they were much poorer at learning the word meanings: they Experiment 2 showed low accuracy in both the meaning recall (18%) and semantic categorisation (59%) tasks, which neither Three hypotheses were pre-registered on the Open improved nor declined with repeated tests. This increase Science Framework (http://osf.io/yk3d5): 1) Cued recall in recall for word-forms is consistent with previous for word-forms will improve over time, consistent with ﬁndings of an oﬄine consolidation and/or retrieval prac- Experiment 1 (and Experiment S1), and with extant evi- tice beneﬁt for this aspect of word knowledge, and adds dence supporting strengthening of novel word-forms to growing evidence that deﬁnition recall does not by delayed tests; 2) Where a neighbourhood density beneﬁt from the same opportunities for reactivation eﬀect emerges, we predict that low-density items will (James et al., 2020a; Tamminen et al., 2012; Tamminen be better learned than high-density items—consistent & Gaskell, 2013). However, participants may also have with our ﬁndings from Experiment 1 (and non-signiﬁcant beneﬁted from opportunities to re-encode the word- numerical diﬀerences in Experiment S1); and 3) If the forms (but not meanings) during repeated tests in this absence of a density eﬀect in the deﬁnitions task for study, facilitating improvements in this aspect of word adults in Experiment S1 was driven by their higher per- knowledge across the week. formance, then we would expect a neighbourhood Our primary research questions related to the new density eﬀect to emerge at lower performance levels words’ engagement with existing semantic knowledge, in this task. However, if the absence of the density as indicated by performance diﬀerences related to eﬀect is driven by adults’ learning eﬃciency (relative to semantic neighbourhood density. We found that existing the enhanced sensitivity of developing learners to semantic knowledge can inﬂuence new vocabulary acqui- semantic competitors), we would expect no eﬀect of sition in school-aged children: they were better at recal- density in the deﬁnitions task for adults regardless of ling novel semantic concepts from low- versus high- performance levels. density semantic neighbourhoods. However, recall of word-forms appeared unaﬀected by these semantic manipulations. Thus, in line with the processing interfer- Experiment 2 methods ence observed for language-based neighbourhoods pre- Participants viously in adults (Tamminen et al., 2013), dense feature- based semantic neighbourhoods also appear to elicit 70 participants were recruited via the University of York interference observable in children’s learning of new Psychology Department participant pool according to 10 E. JAMES ET AL. the following criteria: native monolingual English speak- with children. Participants circled their meaning match- ers, aged 18-35, with normal or corrected-to-normal ing answers (1, 2, or 3) in an additional training booklet. hearing and vision, and no reading or language dis- orders. Three participants did not complete more than Test tasks one of the three follow-up sessions, and were excluded from analyses. Thus, the ﬁnal sample consisted of 67 par- The four test tasks were programmed for participants ticipants (14 male), with a mean age of 20.33 years (SD = to complete online from home, in the same ﬁxed 2.54). Nine participants contributed only partial data (2/3 order described above. Adults were provided with sessions) having missed the ﬁnal session. only the written cues, and gave typed responses for Participants received either £10 or course credit for the recall task. Given that we were most interested in their time. The study was approved by the Department vocabulary learning (rather than orthographic learning of Psychology Research Ethics Committee at the Univer- speciﬁcally), answers were scored according to whether sity of York. they read as phonologically correct (e.g. attee or atty instead of attie; chiypod instead of chipod). The form and deﬁnition recall tasks were hosted online using Design and procedure Qualtrics (Qualtrics, 2014). A link within the survey took participants to the form recognition and semantic To make Experiment 2 as comparable as possible to categorisation tasks, which were programmed using Experiment 1, we conducted training in a group setting Testable (Rezlescu, 2015) to enable response time lasting approximately 45 min. Three test sessions were recordings. then completed online according to the same schedule: the same day (T1), next day (T2), and one week later (T3). Participants were asked to complete the ﬁrst test Analyses session within 2 h of training, and complete each sub- Analyses were conducted as in Experiment 1. sequent session at a similar time (by 6pm at the latest). All sessions completed on the correct day were included in the analyses. Although we did not implement speciﬁc Experiment 2 results attention checks in the online tests, inspection of task per- formance conﬁrms that participants were engaged with Cued form recall the activities (i.e. recognition task performance was The proportion of word-forms recalled on the same day always well above chance, ≥ 67%). No standardised of learning (M = .21, SD = .40) was highly comparable to assessments were collected for Experiment 2. Experiment 1 (M = .20, SD = .40), suggesting a similar level of diﬃculty for children and adults. Recall improved signiﬁcantly at the delayed tests (delay1: β = 0.39, SE = Stimuli 0.03, Z = 13.32, p < .001; Figure 2B), but continued The full set of 24 items were used for the adults. We improvements between T2 (M = .36, SD = .48) and T3 additionally incorporated two elements of counterba- (M = .38, SD = .49) were not statistically signiﬁcant (p lancing for this experiment to ensure that idiosyncratic = .122; Table 4). diﬀerences in the stimuli were not responsible for the There was a small but statistically signiﬁcant eﬀect of neighbourhood density eﬀects. The two versions of the density (β = −0.11, SE = 0.05, Z = −2.254, p = .025): word- stimuli altered the set of pseudowords and novel fea- forms associated with low neighbourhood density con- tures assigned to each density condition. cepts were better recalled (M = .33, SD = .47) than those associated with high-density concepts (M = .29, SD = .46). This density eﬀect did not change over time, Training tasks and the interaction was pruned from the ﬁnal model (p = .383). The training tasks were identical to Experiment 1, except with form and deﬁnition repetitions reduced to one per item. Only one round of meaning matching was admi- Form recognition nistered, presenting each deﬁnition once with three options for its word-form on each occasion. This meant A technical issue meant that T1 form recognition and that participants had ﬁve exposures to the new word- semantic categorisation data from the ﬁrst set of partici- forms in total, and only two exposures to the deﬁnitions, pants was not saved from Testable (n = 9), and this issue intended to reduce adults’ performance levels in line also aﬀected a later session for two participants. LANGUAGE, COGNITION AND NEUROSCIENCE 11 Table 4. Final analysis models for Experiment 2: Form tasks. a b Cued form recall Form recognition β SE Z p β SE Z p (Intercept) −1.11 0.24 −4.62 <.001 3.38 0.26 12.89 <.001 delay1 0.39 0.03 13.32 <.001 0.11 0.04 2.59 .010 delay2 0.07 0.05 1.55 .122 −0.04 0.08 −0.56 .578 density −0.11 0.05 −2.24 .025 0.21 0.11 1.97 .049 Note: (a) Analysis based on from 4608 observations across 67 participants and 24 items. The ﬁnal model included by-participant and by-item random intercepts, and by-participant random slopes for the eﬀect of density. The two-way interaction between time and density was pruned from the model with no reduction in model ﬁt χ = 1.92, p = .38; (b) Analysis based on from 4200 observations across 65 participants and 24 items. The ﬁnal model included by-participant and by-item random intercepts, and by-item random slopes for the eﬀect of density. The two-way interaction between time and density was pruned from the model with no reduction in model ﬁt χ = 0.10, p = .95. Unfortunately it was not possible to replace these par- Semantic categorisation ticipants due to timing constraints, and our main Accuracy hypotheses related to the explicit recall measures for Accuracy was generally very low (M = .57, SD = .50), and this experiment. We removed any participants who did did not change across the course of the week (ps > .35). not have data from at least two of the three sessions, There was also no signiﬁcant eﬀect of neighbourhood leaving 65 participants for these analyses. density (p = .508; Table A2-3). Recognition of the new word-forms was much higher than participants’ ability to recall them. Performance was lowest at the ﬁrst test point (M = 0.91, SD = 0.29; delay1: Reaction time β = 0.11, SE = 0.04, Z = 2.59, p = .010), but there were no At this low level of performance, 16 participants were further changes in performance between the day (M excluded from RT analyses on the basis of chance-level = .94, SD = .24) and week (M = .93, SD = .26; p = .578) performance (note that this exclusion was not tests. There was a small but signiﬁcant eﬀect of neigh- speciﬁed in the pre-registration due to an oversight). bourhood density (β = 0.21, SE = 0.11, Z = 1.97, p This left 49 participants in the analysis, who ranged = .049): performance was slightly higher for high- from .51-.74 in categorisation accuracy (M = .60). Only density items (M = .93, SD = .26) than low-density (M 44.98% of the data was retained after data trimming = .92, SD = .27). However, there was no evidence of an (as above), and so caution is needed in interpreting interaction with test session (pruned from ﬁnal model; these data. Modelling was carried out on the log-trans- p = .951; Table 4). formed data, and showed only a decrease in reaction time across test sessions: participants were slowest at the ﬁrst test (M = 1202 ms, SD = 495 ms; delay1: β = Cued meaning recall −0.08, SE = 0.01, t = −6.45, p < .001), and continued to Participants scored an average of 0.39 (SD = 0.78) points improve between the day (M = 1017 ms, SD = 430 ms) per item at the ﬁrst test, which did not change over time and week (M = 911 ms, SD = 403 ms) memory tests (ps > .7; Figure 2D; Table 5). Whilst this level of perform- (delay2: β = −0.05, SE = 0.02, t = −2.71, p = .010). There ance was highly comparable to Experiment 1 (M= .36, was no eﬀect of neighbourhood density (p = .344; SD = .76), recall of meanings was not aﬀected by the Table A2-4). semantic neighbourhood density of the concepts in adult participants (p = .704). Experiment 2 discussion In Experiment 2, we examined whether semantic neigh- Table 5. Final analysis model for Experiment 2: Cued meaning bourhood density inﬂuences adults’ word learning. To recall. draw comparisons with children in Experiment 1, we β coeﬃcient SE Z p used more items and fewer exposures to the new Fixed eﬀects stimuli to create a similar level of task diﬃculty 0|1 1.86 0.22 8.63 <.001 1|2 1.98 0.22 9.18 <.001 between the two groups. Adults recalled a comparable delay1 −0.01 0.03 −0.33 .738 proportion of the stimuli across the diﬀerent tasks to delay2 −0.02 0.05 −0.29 .773 density −0.07 0.19 −0.38 .704 children, but note that the overall information learned Note: Analysis based on from 4536 observations across 66 participants and was still higher for adults as they were provided with 24 items. The ﬁnal model included by-participant and by-item intercepts. more items (24 vs. 16). Consistent with the results of The two-way interaction between time and density was pruned from the model with no reduction in model ﬁt χ = 0.01, p = .995. Experiment 1, memory for new word-forms improved 12 E. JAMES ET AL. over repeated tests distributed across the week, whereas Z = −2.22, p = .027), showing a beneﬁt for recalling deﬁnition knowledge remained stable. meanings associated with low-density semantic neigh- Like children, adults were inﬂuenced by semantic bourhoods. There was no eﬀect of age group, alone or neighbourhood density in recalling the new items. in interaction with any other variable. However, while children had been inﬂuenced by the semantic manipulation in recalling the word meanings, General discussion adults showed this eﬀect in recalling the word-forms— despite no explicit demands on accessing semantic The two experiments showed that semantic prior knowl- knowledge in these tasks. For word-form recall, the edge inﬂuences new word learning. We used an object- eﬀects of semantic neighbourhood density were based metric to test the hypothesis that more shared similar in direction to those observed Experiment 1, features could facilitate learning, in contrast to previous demonstrating a disadvantage for high-density items. studies that found semantic interference from language- However, this was also accompanied by a small beneﬁt based measures of relatedness. However, this was not for recognising high-density items. In contrast to chil- the case: both children and adults showed interference dren, adults were not inﬂuenced by semantic density from dense feature-based neighbourhoods in recalling in recalling the novel meanings. the new items. This semantic interference emerged for the task drawing upon meaning knowledge for children and form knowledge in adults, although cross-exper- Additional exploratory analyses iment analyses indicated that these task diﬀerences The results suggest that both children and adults experi- may not be robust. In the following discussion, we enced interference from high-density semantic neigh- focus ﬁrst on the nature of semantic inﬂuences during bourhoods, but there were group diﬀerences in how word learning, before considering possible developmen- this interference manifested in the diﬀerent measures tal diﬀerences and implications for word learning more of word learning. We conducted additional exploratory broadly. analyses to assess whether these patterns of perform- ance for explicit recall of word-forms and meanings The inﬂuence of semantic neighbours during were statistically diﬀerent between the two word learning experiments. For each of the two recall measures, we ﬁtted a mixed The results are consistent with the few previous studies eﬀects model with ﬁxed eﬀects of session, density, and that have manipulated the availability of semantic group (children versus adults), with all interaction neighbours during word learning, ﬁnding that memory terms. Random eﬀects were speciﬁed as above, and for new words can be hindered by links to denser the full model tables can be found in Appendix 2 semantic neighbourhoods (Storkel & Adlof, 2009; Tam- (Tables A2-5 and A2-6). For the word-form recall task, minen et al., 2013). Our ﬁndings add two key contri- there were main eﬀects of test session (delay1: β = butions to the literature here: ﬁrst, that school-aged 0.68, SE = 0.03, Z = 25.07, p < .001; delay2: β = 0.50, SE = children are similarly aﬀected by semantic neighbours 0.04, Z = 12.10, p < .001), reﬂecting the improvements as preschool children and adults; and second, that seen across the week in each experiment. There was feature-based conceptualisations of semantic neigh- also a main eﬀect of group (β = −0.58, SE = 0.16, Z = bourhoods inﬂuence learning as well as associative −3.55, p < .001), with adults performing worse than chil- (language-based) metrics. Counter to predictions that dren, but this was in the context of signiﬁcant inter- semantic inﬂuences would emerge at delayed tests— actions with test session (group*delay1: β = −0.29, SE = following increased opportunities for the new represen- 0.03, Z = −10.98, p < .001; group*delay2: β = −0.43, SE = tations to integrate with existing knowledge (Davis & 0.04, Z = −10.44, p < .001). Pairwise contrasts for each Gaskell, 2009; McClelland et al., 1995)—the eﬀects test session showed that the two groups did not diﬀer emerged immediately after training for tasks assessing at the ﬁrst test point (p = .942), but that children increas- explicit knowledge of the new forms and/or meanings, ingly outperformed adults in their likelihood of recalling and did not change across the week. This early the word-forms at T2 (β = −0.88, SE = 0.33, Z = −2.62, p inﬂuence of existing knowledge may be due to the = .009) and T3 (β = −2.60, SE = 0.34, Z = −7.60, p < .001). nature of the training task, given that related concepts There was no eﬀect of density, alone or in interaction were explicitly incorporated during encoding (i.e. the with any other variable. base concepts were named in the deﬁnitions, and par- For the meaning recall task, only the main eﬀect of ticipants used their prior knowledge of these concepts density was statistically signiﬁcant (β = −0.27, SE = 0.12, to draw the items). A key question for future studies is LANGUAGE, COGNITION AND NEUROSCIENCE 13 thus whether the time course of semantic inﬂuence derived, and the extent to which they capture relevant would diﬀer if these similarities were not made explicit semantic relationships for supporting learning. Feature during encoding, requiring learners to infer similarity norms are created by asking participants to list features with known concepts using images or feature descrip- of diﬀerent concepts, but these reports are biased tions alone. towards salient and distinctive features; participants Why then do dense semantic networks lead to poorer are less likely to report the ordinary features that they memory performance for new concepts in this context? share with many other concepts. Thus, the metrics we One possible explanation is that the co-activation of used to deﬁne semantic neighbourhood density may multiple related concepts during encoding leads to not capture the vast array of highly familiar features competition or interference in processing. Recent known and shared for certain concepts, which may be ﬁndings from the lexical processing literature indicate beneﬁcial when learning new related concepts. To that the high-density disadvantage may relate to the explore this possibility further, we conducted some speciﬁc metrics we chose when designing our stimuli. additional analyses (Appendix 3) using an alternative We selected base concepts that varied in both semantic metric from the McRae et al. (2005) feature norms as a richness (the number of reported features) and density predictor of performance across experiments: the pro- (the co-occurrence of those features in other concepts), portion of distinctive features (i.e. the proportion of based on early evidence that shared features facilitate the base concept’s listed features that were not listed lexical processing (Grondin et al., 2009). Since then, for other concepts). This metric was not signiﬁcantly cor- several studies have demonstrated that our selected related to either the number of features reported or the measures might have opposing inﬂuences, particularly percentage of correlated features used in initial selection in studies of word production (Hameau et al., 2019; of the stimuli, suggesting that it captures a diﬀerent Lampe et al., 2022; Rabovsky et al., 2016). According to semantic dimension. The results showed that items Lampe et al. (2022), an abundance of semantic features with high feature distinctiveness were slightly harder leads to stronger lexical activation that supports faster to recall (66%) than items with fewer distinctive features and more accurate responses during picture naming (71%), suggesting that atypicality may hinder concept tasks, whereas high intercorrelational density more memory. Thus, there may be a beneﬁt for semantic strongly activates related concepts that cause interfer- prior knowledge in word learning that was not well-cap- ence. Applied to the present ﬁndings, the co-activation tured by our design. of a large number of related concepts when learning It is not possible to dissociate between these possible the high-density items may have led to interference in theoretical explanations with the present results, but establishing the new semantic representation and/or they highlight two important aspects to consider for when performing the recall tasks, with further research future studies: ﬁrst, that multiple semantic dimensions required to pinpoint the locus of this eﬀect. should be examined simultaneously to understand An alternative (not mutually exclusive) possibility is their inﬂuences on learning (similar to recent analyses that a dense network of co-occurring features makes it for word recognition, e.g. Lampe et al., 2022); and more challenging to integrate the highly distinctive second, the need to distinguish between the availability feature that made each concept novel (i.e. the new of existing knowledge (here, the base concept) and the concept is relatively more atypical of existing knowl- ease at which new information can be incorporated edge). Framed in this way, our stimuli perhaps more into existing networks (the novel feature). Speaking to closely align with computational models of learning aty- this distinction, studies that do not require the inte- pical category exemplars (McClelland et al., 2020): across gration of new semantic information ﬁnd the opposite both conditions, learners had the same amount of new pattern of results to those presented here: pseudowords information—a single feature—to integrate with exist- are more readily remembered when paired with existing ing knowledge. However, when the known concept concepts from higher density semantic networks (Mak & comes from a dense neighbourhood, this additional Twitchell, 2020). Thus, both the availability of semantic feature can be considered more atypical of existing con- knowledge and the need to integrate new information cepts. Thus, it becomes more challenging to integrate are key considerations in understanding how prior this novel information than when there are fewer knowledge can inﬂuence new word learning. related concepts, requiring more extensive opportu- Finally, it is important to consider that the initial chal- nities for learning and reactivation than were oﬀered lenge of learning concepts from high-density neigh- by the present study. bourhoods may yet translate to longer-term A third possibility for the density disadvantage relates processing beneﬁts over time and further exposures. to the way in which the feature norms themselves were Although we originally set out to examine semantic 14 E. JAMES ET AL. integration with the semantic categorisation tasks, we aﬀects the resources available to encode or retrieve later reduced the number of stimulus exposures in the associated word-forms, given that this activation is adults’ learning phase to aid in interpreting diﬀerences experience-dependent (Pexman, 2020). On the converse, in semantic inﬂuences between children and adults strengths in explicit learning may mean that adults can when performing at a similar level of diﬃculty. Thus, overcome semantic competition in the deﬁnitions task. perhaps unsurprisingly, we did not see any inﬂuence These possibilities warrant further investigation in of semantic neighbourhood density on speeded proces- studies designed and powered to examine developmen- sing as we initially had predicted (and as was observed tal diﬀerences. by Tamminen et al., 2013). We consider that the present results thus reﬂect a relatively early stage of Encoding and consolidation processes in word new word knowledge, and that the new word meanings learning were not well-consolidated into vocabulary during the course of the experiment. With additional learning Moving beyond the inﬂuence of semantic neighbours, opportunities, dense semantic neighbourhoods may the results are consistent with previous studies demon- yet provide a beneﬁcial role in new word knowledge. strating improvements in word-form knowledge with In line with this possibility, Mak et al. (2021) demon- repeated tests across a week period (e.g. Henderson strated that words encountered across semantically et al., 2013; James et al., 2019; Storkel, 2001; Tamminen diverse texts are more poorly learned than those et al., 2010). These improvements are consistent with the encountered in a single semantic context in the ﬁrst hypothesis that new word knowledge is strengthened instance, but that diversity comes to beneﬁt word “oﬄine”, via opportunities for hippocampal reactivation knowledge after an initial period of stabilisation has during sleep (Davis & Gaskell, 2009; McClelland et al., occurred. Thus, with more opportunities for training 1995), and/or that performance improves with repeated over a longer period of time, the disadvantages seen retrieval practice (Goossens et al., 2014). A ﬁrst notable for learning high-density concepts in the present study ﬁnding here is that participants demonstrated clear may translate to a longer-term processing beneﬁt for gains in word-form knowledge but not meaning knowl- the new words, in line with the lexical processing advan- edge, despite comparably low initial performance in tage observed for the base concepts themselves. these tasks. A likely key factor in this diﬀerence in gains is that participants were re-exposed to the word- forms at each test point (i.e. in the form recognition Developmental diﬀerences in semantic inﬂuences task, and as a cue for the deﬁnition recall task), We tested children and adults using the same exper- whereas there was no further re-exposure to the novel imental paradigm, providing an insight into the meanings. With further opportunities for (re-)encoding inﬂuence of semantic knowledge on word learning the word-forms, it is perhaps no surprise that such across development. It was clear that both groups impressive gains were seen across the course of the accessed related semantic knowledge during the exper- week. iment, marked by superior recall of items associated with However, we also consider the possibility that diﬀer- low- versus high-density semantic neighbourhoods. Yet ences in gains are additionally inﬂuenced by the there was some indication of a developmental diﬀerence extent to which diﬀerent aspects of new word knowl- in how these neighbourhood eﬀects manifest in task edge can build on existing representations, in line with performance. In Experiment 1, children were inﬂuenced a complementary learning systems perspective. That is, by neighbourhood density only in their recall of mean- this discrepancy is similarly observed in studies that ings and not word-forms. For adults in Experiment 2 equate opportunities to re-encode form and meaning however (and non-signiﬁcantly in Experiment S1), neigh- aspects of new word knowledge, suggesting that bourhood density eﬀects were most apparent in the repeated exposures may not be the only contributing form recall measure—despite no requirement for factor. For example, the repeated tests in James et al. semantic knowledge to be retrieved for task success. (2020a) incorporated a single re-exposure of the word- These task diﬀerences were not anticipated and may form (as a cue for meaning recall) and training image be spurious—indeed, it is important to stress that the (as a cue for a picture naming task) at three test points cross-experiment analyses did not ﬁnd clear evidence over 24 h. While word-form recall improved after sleep, of developmental diﬀerences in neighbourhood deﬁnition recall remained stable across periods of eﬀects. However, we can consider that perhaps only wake and sleep despite repeated opportunities to re- the mature lexical-semantic system activates semantic encode semantic features. In the context of the knowledge so automatically during learning that it present experiment, the word-forms represented a LANGUAGE, COGNITION AND NEUROSCIENCE 15 relatively arbitrary combination of sounds, proposed to that multiple mechanisms may contribute to enhanced be dependent on the hippocampal system at encoding vocabulary consolidation during this period. Indeed, it and thus most reliant on reactivation to support neocor- could be that children beneﬁt more than adults from tical consolidation (James et al., 2019). On the converse, the retrieval practice or re-encoding opportunities at the word meanings created for this experiment were each test point. A valuable next step here will be to novel variants of known concepts, directly building on examine whether developmental diﬀerences remain existing representations in both semantic neighbour- when the test delay is a between-subjects manipulation, hood conditions. Recent computational studies have with diﬀerent groups completing only a single test conceptualised the neocortical system as being prior either the same day, the next day, and one week later. knowledge-dependent (Kumaran et al., 2016; McClelland, Understanding these developmental diﬀerences in 2013; McClelland et al., 2020), indicating that neocortical longer-term memory processes, and whether we can learning can occur rapidly in the context of existing capitalise upon them to support vocabulary develop- knowledge without requiring hippocampal reactivation. ment, presents an exciting avenue for future research. Thus, while semantic aspects of new word knowledge may sometimes beneﬁt from oﬄine consolidation (e.g. McGregor et al., 2013), the scope for capitalising upon Conclusions existing semantic connections may render this eﬀect It is well-established that prior knowledge aﬀects new less robust across studies than the beneﬁts observed learning. This study built upon previous studies of pho- for word-form memory. Whether this explanation nological knowledge in vocabulary learning to show that would hold beyond the confound of repeated word- semantic neighbours also aﬀect the acquisition of new form tests within the present design remains an open words. Further, these semantic inﬂuences can be cap- question, which could be better addressed by using a tured by object-based metrics, as well as the associative between-subjects manipulation of test delay. semantic dimensions used in previous studies (Storkel & The second notable ﬁnding is that children showed Adlof, 2009; Tamminen et al., 2013). We found that by greater improvements in word-form knowledge across training pseudowords and associated novel concepts the week than adults. We emphasise caution in inter- with close semantic neighbours (i.e. diﬀering in a preting this result: ﬁrst because it was the result of an single feature), children and adults found it harder to exploratory analysis, and second because of some meth- learn and/or remember new words associated with odological diﬀerences between the two experiments (i.e. dense feature neighbourhoods. Given that these the spoken versus written modality of the recall test). ﬁndings somewhat contradict the semantic density However, superior long-term retention of word-forms beneﬁts observed in studies of known words, we in children has been demonstrated across several pre- propose that time and/or experience, semantic dimen- vious studies (James et al., 2019; Smalle et al., 2018), sion, and semantic distance should each be thoroughly including those using identical training and test para- examined to understand the role that prior semantic digms across the two groups (James et al., 2020b). A knowledge plays in vocabulary acquisition. valuable contribution of the present study is that interpretation of these group diﬀerences is often con- founded by adults’ relative strength in initial encoding: Notes do children beneﬁt more by delayed tests because of developmental diﬀerences in memory processes, or is 1. Experiment S1 used original items selected from the it simply that there is more scope for improvement English Lexicon Project. Due to experimenter error, when initial learning is weak? We found here that chil- slight variants of four of the items were used in Exper- iment 1 (i.e., attie, bryat, shamal, vorgol instead of dren continue to show delayed beneﬁts relative to attay, bryet, shimal, vorgal). However, we recomputed adults even when matched for initial diﬃculty in the orthographic neighbourhood density and bigram fre- ﬁrst test session. This beneﬁt is in line with evidence quency based on the new set to conﬁrm that these suggesting an enhanced role for sleep in children’s did not diﬀer between word lists, and retained the memory consolidation (Peiﬀer et al., 2020; Wilhelm amended version for both Experiment 1 and 2 here to facilitate developmental comparisons. et al., 2012; Wilhelm et al., 2013), linked to a higher pro- 2. Only 18 of the 24 items were also entries in the Florida portion of the slow neural oscillations that are associated Free Association Norms. These indicated that the two with memory consolidation processes. However, other sets would likely diﬀer in semantic neighbourhood studies have also found superior memory retention in density by this measure, with high-density concepts children across shorter periods that do not contain having more associates (M = 17.33) than low-density sleep (Bishop et al., 2012; Smalle et al., 2018), suggesting concepts (M = 12.22; p = .05). 16 E. JAMES ET AL. 3. Note that the pre-registration refers to a signiﬁcant Bates, D., Kliegl, R., Vasishth, S., & Baayen, H. (2015a). eﬀect of semantic neighbourhood density for cued Parsimonious mixed models. arXiv Preprint ArXiv, 1506, form recall in an initial adult experiment (Experiment 04967. https://doi.org/10.48550/arXiv.1506.04967 S1). This was due to an error in which test session was Bates, D., Maechler, M., Bolker, B., & Walker, S. (2015b). Fitting entered into analyses as a continuous rather than categ- linear mixed-eﬀects models using lme4. Journal of orical predictor. Statistical Software, 67(1), 1–48. https://doi.org/https://doi. 4. One participant did not complete 2/3 deﬁnitions tests, org/10.18637/jss.v067.i01 and was excluded from this analysis. Bishop,D.V., Barry,J.G., &Hardiman, M. J. (2012). Delayed retention of new word-forms is better in children than adults regardless of language ability: A factorial two-way study. PloS one, 7(5), e37326. https://doi.org/10.1371/journal.pone. Acknowledgment An earlier version of this manuscript was published in the ﬁrst Borovsky, A., Ellis, E. M., Evans, J. L., & Elman, J. L. (2016). Lexical author’s PhD thesis, and we would like to thank Professor leverage: Category knowledge boosts real-time novel word Dorothy Bishop for suggesting further exploratory analyses. recognition in 2-year-olds. Developmental Science, 19(6), 918–932. https://doi.org/10.1111/desc.12343 Brysbaert, M., Warriner, A. B., & Kuperman, V. (2014). Concreteness ratings for 40 thousand generally known Funding English word lemmas. Behavior Research Methods, 46(3), E.J was supported by an Economic and Social Research Council 904–911. https://doi.org/10.3758/s13428-013-0403-5 1 + 3 studentship in conducting this study. The research was Buchanan, L., Westbury, C., & Burgess, C. (2001). Characterizing additionally supported ESRC grant ES/N009924/1 awarded to semantic space: Neighborhood eﬀects in word recognition. L.M.H. and M.G.G. Psychonomic Bulletin & Review, 8(3), 531–544. https://doi. org/10.3758/BF03196189 Christensen, R. H. B. (2015). Ordinal-Regression models for Data availability statement ordinal data. R Package Version 2015.6-28. http://www. cran.r-project.org/package=ordinal. All materials, data, and analyses are available at https://osf.io/ Clay, F., Bowers, J. S., Davis, C. J., & Hanley, D. A. (2007). 35ftn/. Teaching adults new words: The role of practice and conso- lidation. Journal of Experimental Psychology: Learning, Memory, and Cognition, 33(5), 970. https://doi.org/10.1037/ Disclosure statement 0278-7393.33.5.970 Coltheart, M. (1981). The MRC psycholinguistic database. The No potential conﬂict of interest was reported by the author(s). Quarterly Journal of Experimental Psychology Section A, 33 (4), 497–505. https://doi.org/10.1080/14640748108400805 Davies, R. A., Arnell, R., Birchenough, J. M., Grimmond, D., & ORCID Houlson, S. (2017). Reading through the life span: Individual diﬀerences in psycholinguistic eﬀects. Journal of Emma James http://orcid.org/0000-0002-5214-0035 Experimental Psychology: Learning, Memory, and Cognition, 43(8), 1298. https://doi.org/10.1037/xlm0000366 Davis, M. H., & Gaskell, M. G. (2009). A complementary systems References account of word learning: Neural and behavioural evidence. Philosophical Transactions of the Royal Society B: Biological Antony, J. W., Ferreira, C. S., Norman, K. A., & Wimber, M. (2017). Sciences, 364(1536), 3773–3800. https://doi.org/10.1098/ Retrieval as a fast route to memory consolidation. Trends in rstb.2009.0111 Cognitive Sciences, 21(8), 573–576. https://doi.org/10.1016/j. Dumay, N., & Gaskell, M. G. (2007). Sleep-associated changes in tics.2017.05.001 the mental representation of spoken words. Psychological Ashworth, A., Hill, C. M., Karmiloﬀ-Smith, A., & Dimitriou, D. Science, 18(1), 35–39. https://doi.org/10.1111/j.1467-9280. (2014). Sleep enhances memory consolidation in children. 2007.01845.x Journal of Sleep Research, 23(3), 304–310. https://doi.org/ Engelthaler, T., & Hills, T. T. (2017). Feature biases in early word 10.1111/jsr.12119 learning: Network distinctiveness predicts age of acquisition. Baayen, R. H., Piepenbrock, R., & Gulikers, L. (1995). The CELEX Cognitive Science, 41,120–140. https://doi.org/10.1111/cogs. lexical database (release 2). Distributed by the Linguistic Data Consortium, University of Pennsylvania. Forster, K. I., & Forster, J. C. (2003). Dmdx: A windows display Balota, D. A., Yap, M. J., Hutchison, K. A., Cortese, M. J., Kessler, program with millisecond accuracy. Behavior Research B., Loftis, B., Neely, J. H., Nelson, D. L., Simpson, G. B., & Methods, Instruments, & Computers, 35(1), 116–124. https:// Treiman, R. (2007). The English lexicon project. Behavior doi.org/10.3758/BF03195503 Research Methods, 39(3), 445–459. https://doi.org/10.3758/ Gais, S., Lucas, B., & Born, J. (2006). Sleep after learning aids BF03193014 memory recall. Learning & Memory, 13(3), 259–262. https:// Barr, D. J., Levy, R., Scheepers, C., & Tily, H. J. (2013). Random doi.org/10.1101/lm.132106 eﬀects structure for conﬁrmatory hypothesis testing: Keep Gaskell, M. G., & Marslen-Wilson, W. D. (1997). Integrating form it maximal. Journal of Memory and Language, 68(3), 255– and meaning: A distributed model of speech perception. 278. https://doi.org/10.1016/j.jml.2012.11.001 LANGUAGE, COGNITION AND NEUROSCIENCE 17 Language and Cognitive Processes, 12(5-6), 613–656. https:// Behavior Research Methods, 44(4), 978–990. https://doi.org/ doi.org/10.1080/016909697386646 10.3758/s13428-012-0210-4 Goossens, N. A., Camp, G., Verkoeijen, P. P., & Tabbers, H. K. Kuznetsova, A., Brockhoﬀ, P. B., & Christensen, R. H. (2017). (2014). The eﬀect of retrieval practice in primary school lmerTest package: Tests in linear mixed eﬀects models. vocabulary learning. Applied Cognitive Psychology, 28(1), Journal of Statistical Software, 82(13), 1–26. https://doi.org/ 135–142. https://doi.org/10.1002/acp.2956 10.18637/jss.v082.i13 Grondin, R., Lupker, S. J., & McRae, K. (2009). Shared features Lampe, L. F., Hameau, S., & Nickels, L. (2022). Semantic variables dominate semantic richness eﬀects for concrete concepts. both help and hinder word production: Behavioral evidence Journal of Memory and Language, 60(1), 1–19. https://doi. from picture naming. Journal of Experimental Psychology: org/10.1016/j.jml.2008.09.001 Learning, Memory, and Cognition, 48(1), 72–97. https://doi. Hameau, S., Nickels, L., & Biedermann, B. (2019). Eﬀects of org/10.1037/xlm0001050 semantic neighbourhood density on spoken word pro- Mak, M. H., Hsiao, Y., & Nation, K. (2021). Anchoring and contex- duction. Quarterly Journal of Experimental Psychology, 72 tual variation in the early stages of incidental word learning (12), 2752–2775. https://doi.org/10.1177/ during reading. Journal of Memory and Language, 118, 1747021819859850 104203. https://doi.org/10.1016/j.jml.2020.104203 Henderson, L. M., Weighall, A., & Gaskell, G. (2013). Learning Mak, M. H., & Twitchell, H. (2020). Evidence for preferential new vocabulary during childhood: Eﬀects of semantic train- attachment: Words that are more well connected in seman- ing on lexical consolidation and integration. Journal of tic networks are better at acquiring new links in paired- Experimental Child Psychology, 116(3), 572–592. https://doi. associate learning. Psychonomic Bulletin & Review, 27(5), org/10.1016/j.jecp.2013.07.004 1059–1069. https://doi.org/10.3758/s13423-020-01773-0 Henderson, L. M., Weighall, A. R., Brown, H., & Gaskell, M. G. Marian, V., Bartolotti, J., Chabal, S., & Shook, A. (2012). (2012). Consolidation of vocabulary is associated with Clearpond: Cross-linguistic easy-access resource for phono- sleep in children. Developmental Science, 15(5), 674–687. logical and orthographic neighborhood densities. PloS one, https://doi.org/10.1111/j.1467-7687.2012.01172.x 7(8), e43230. https://doi.org/10.1371/journal.pone.0043230 Hills, T. T., Maouene, M., Maouene, J., Sheya, A., & Smith, L. McClelland, J. L. (2013). Incorporating rapid neocortical learn- (2009). Longitudinal analysis of early semantic networks. ing of new schema-consistent information into complemen- Psychological Science, 20(6), 729–739. https://doi.org/10. tary learning systems theory. Journal of Experimental 1111/j.1467-9280.2009.02365.x Psychology: General, 142(4), 1190. https://doi.org/10.1037/ Hoover, J. R., Storkel, H. L., & Hogan, T. P. (2010). A cross-sec- a0033812 tional comparison of the eﬀects of phonotactic probability McClelland, J. L., McNaughton, B. L., & Lampinen, A. K. (2020). and neighborhood density on word learning by preschool Integration of new information in memory: New insights children. Journal of Memory and Language, 63(1), 100–116. from a complementary learning systems perspective. https://doi.org/10.1016/j.jml.2010.02.003 Philosophical Transactions of the Royal Society B: Biological Hulme, R. C., & Rodd, J. M. (2021). Learning new word meanings Sciences, 375(1799), 20190637. https://doi.org/10.1098/rstb. from story reading: The beneﬁt of immediate testing. PeerJ, 2019.0637 9, e11693. https://doi.org/10.7717/peerj.11693 McClelland, J. L., McNaughton, B. L., & O’Reilly, R. C. (1995). Why James, E., Gaskell, M. G., & Henderson, L. M. (2019). Oﬄine con- there are complementary learning systems in the hippo- solidation supersedes prior knowledge beneﬁts in children’s campus and neocortex: Insights from the successes and fail- (but not adults’) word learning. Developmental Science, 22(3), ures of connectionist models of learning and memory. e12776. https://doi.org/10.1111/desc.12776 Psychological Review, 102(3), 419. https://doi.org/10.1037/ James, E., Gaskell, M. G., & Henderson, L. M. (2020a). Sleep- 0033-295X.102.3.419 dependent consolidation in children with comprehension McGregor, K. K., Licandro, U., Arenas, R., Eden, N., Stiles, D., and vocabulary weaknesses: It’ll be alright on the night? Bean, A., & Walker, E. (2013). Why words are hard for Journal of Child Psychology and Psychiatry, 61(10), 1104– adults with developmental language impairments. Journal 1115. https://doi.org/10.1111/jcpp.13253 of Speech, Language, and Hearing Research, 56(6), 1845– James, E., Gaskell, M. G., Pearce, R., Korell, C., Dean, C., & 1856. https://doi.org/10.1044/1092-4388(2013/12-0233) Henderson, L.-M. (2020b). The role of prior lexical knowl- McRae, K., Cree, G. S., Seidenberg, M. S., & McNorgan, C. (2005). edge in children’s and adults’ incidental word learning Semantic feature production norms for a large set of living from illustrated stories. Journal of Experimental Psychology: and nonliving things. Behavior Research Methods, 37(4), Learning, Memory, and Cognition, 47(11), 1856– 547–559. https://doi.org/10.3758/BF03192726 1869. https://doi.org/10.31234/osf.io/vm5ad Mirman, D., & Magnuson, J. S. (2008). Attractor dynamics and James, E., Gaskell, M. G., Weighall, A., & Henderson, L. (2017). semantic neighborhood density: Processing is slowed by Consolidation of vocabulary during sleep: The rich get near neighbors and speeded by distant neighbors. Journal richer? Neuroscience & Biobehavioral Reviews, 77,1–13. of Experimental Psychology: Learning, Memory, and Cognition, https://doi.org/10.1016/j.neubiorev.2017.01.054 34(1), 65. https://doi.org/10.1037/0278-7393.34.1.65 Kumaran, D., Hassabis, D., & McClelland, J. L. (2016). What learn- Nelson, D. L., McEvoy, C. L., & Schreiber, T. A. (2004). The ing systems do intelligent agents need? Complementary University of south ﬂorida free association, rhyme, and learning systems theory updated. Trends in Cognitive word fragment norms. Behavior Research Methods, Sciences, 20(7), 512–534. https://doi.org/10.1016/j.tics.2016. Instruments, & Computers, 36(3), 402–407. https://doi.org/ 05.004 10.3758/BF03195588 Kuperman, V., Stadthagen-Gonzalez, H., & Brysbaert, M. (2012). Peiﬀer, A., Brichet, M., De Tiège, X., Peigneux, P., & Urbain, C. Age-of-acquisition ratings for 30,000 English words. (2020). Re-epithelialization and immune cell behaviour in 18 E. JAMES ET AL. an ex vivo human skin model. Scientiﬁc Reports, 10(1), 1–9. Language, and Hearing Research, 49(6), 1175–1192. https:// https://doi.org/10.1038/s41598-019-56847-4 doi.org/10.1044/1092-4388(2006/085) Perry, L. K., Axelsson, E. L., & Horst, J. S. (2016). Learning what to Storkel, H. L., Bontempo, D. E., Aschenbrenner, A. J., Maekawa, remember: Vocabulary knowledge and children’s memory for J., & Lee, S.-Y. (2013). The eﬀect of incremental changes in object names and features. Infant and Child Development, 25 phonotactic probability and neighborhood density on (4), 247–258. https://doi.org/10.1002/icd.1933 word learning by preschool children. Journal of Speech, Pexman, P. M. (2020). How does meaning come to mind? Four Language, and Hearing Research, 56(5), 1689–1700. https:// broad principles of semantic processing. Canadian Journal doi.org/10.1044/1092-4388(2013/12-0245) of Experimental Psychology/Revue Canadienne de Tamminen, J., Davis, M. H., Merkx, M., & Rastle, K. (2012). The Psychologie Expérimentale, 74(4), 275–283. https://doi.org/ role of memory consolidation in generalisation of new lin- 10.1037/cep0000235 guistic information. Cognition, 125(1), 107–112. https://doi. Pexman, P. M., Hargreaves, I. S., Siakaluk, P. D., Bodner, G. E., & org/10.1016/j.cognition.2012.06.014 Pope, J. (2008). There are many ways to be rich: Eﬀects of Tamminen, J., & Gaskell, M. G. (2013). Novel word integration in three measures of semantic richness on visual word recog- the mental lexicon: Evidence from unmasked and masked nition. Psychonomic Bulletin & Review, 15(1), 161–167. semantic priming. Quarterly Journal of Experimental https://doi.org/10.3758/PBR.15.1.161 Psychology, 66(5), 1001–1025. https://doi.org/10.1080/ Pexman, P. M., & Yap, M. J. (2018). Individual diﬀerences in 17470218.2012.724694 semantic processing: Insights from the calgary semantic Tamminen, J., Lambon Ralph, M. A., & Lewis, P. A. (2013). The decision project. Journal of Experimental Psychology: role of sleep spindles and slow-wave activity in integrating Learning, Memory, and Cognition, 44(7), 1091. https://doi. new information in semantic memory. The Journal of org/10.1037/xlm0000499 Neuroscience, 33(39), 15376–15381. https://doi.org/10. Qualtrics (2014). Qualtrics [software]. Provo, UT, USA: Qualtrics. 1523/JNEUROSCI.5093-12.2013 Rabovsky, M., Schad, D. J., & Rahman, R. A. (2016). Language pro- Tamminen, J., Payne, J. D., Stickgold, R., Wamsley, E. J., & duction is facilitated by semantic richness but inhibited by Gaskell, M. G. (2010). Sleep spindle activity is associated semantic density: Evidence from picture naming. Cognition, with the integration of new memories and existing knowl- 146,240–244. https://doi.org/10.1016/j.cognition.2015.09.016 edge. Journal of Neuroscience, 30(43), 14356–14360. R Core Team. (2015). R: A language and environment for stat- https://doi.org/10.1523/JNEUROSCI.3028-10.2010 istical computing. R Foundation for Statistical Computing, Tham, E. K., Lindsay, S., & Gaskell, M. G. (2015). Markers of auto- Vienna, Austria. http://www.R-project.org/. maticity in sleep-associated consolidation of novel words. Rezlescu, C. (2015). Testable: The web-based platform for creat- Neuropsychologia, 71, 146–157. https://doi.org/10.1016/j. ing, running and sharing behavioural experiments. neuropsychologia.2015.03.025 Association for Psychological Science Teaching institute Tse, D., Langston, R. F., Kakeyama, M., Bethus, I., Spooner, P. A., conference, New York, 21–24 May. Wood, E. R., Witter, M. P., & Morris, R. G. (2007). Schemas and Smalle, E. H., Page, M. P., Duyck, W., Edwards, M., & Szmalec, A. memory consolidation. Science, 316(5821), 76–82. https:// (2018). Children retain implicitly learned phonological doi.org/10.1126/science.1135935 sequences better than adults: A longitudinal study. van der Kleij, S. W., Rispens, J. E., & Scheper, A. R. (2016). The Developmental Science, 21(5), e12634. https://doi.org/10. eﬀect of phonotactic probability and neighbourhood 1111/desc.12634 density on pseudoword learning in 6- and 7-year-old chil- Storkel, H. L. (2001). Learning new words. Journal of Speech, dren. First Language, 36(2), 93–108. https://doi.org/10. Language, and Hearing Research, 44(6), 1321–1337. https:// 1177/0142723715626064 doi.org/10.1044/1092-4388(2001/103) Wechsler, D. (2011). WASI-II: Wechsler Abbreviated Scale of intel- Storkel, H. L. (2009). Developmental diﬀerences in the eﬀects of ligence. Pearson. phonological, lexical and semantic variables on word learn- Wickham, H. (2009). Ggplot2: Elegant graphics for data analysis. ing by infants. Journal of Child Language, 36(02), 291–321. Springer Science & Business Media. https://doi.org/10.1017/S030500090800891X Wilhelm, I., Prehn-Kristensen, A., & Born, J. (2012). Sleep-depen- Storkel, H. L., & Adlof, S. M. (2009). The eﬀect of semantic set size dent memory consolidation – what can be learnt from chil- on word learning by preschool children. Journal of Speech, dren? Neuroscience & Biobehavioral Reviews, 36(7), 1718– Language, and Hearing Research, 52(2), 306–320. http://jslhr. 1728. https://doi.org/10.1016/j.neubiorev.2012.03.002 pubs.asha.org/data/Journals/JSLHR/929279/jslhr_52_2_306. Wilhelm, I., Rose, M., Imhof, K. I., Rasch, B., Büchel, C., & Born, J. pdf. https://doi.org/10.1044/1092-4388(2009/07-0175) (2013). The sleeping child outplays the adult’s capacity to Storkel, H. L., Armbrüster, J., & Hogan, T. P. (2006). convert implicit into explicit knowledge. Nature Diﬀerentiating phonotactic probability and neighborhood Neuroscience, 16(4), 391-393. https://doi.org/10.1038/nn. density in adult word learning. Journal of Speech, 3343 http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Language Cognition and Neuroscience Taylor & Francis http://www.deepdyve.com/lp/taylor-francis/word-learning-in-the-context-of-semantic-prior-knowledge-evidence-of-X19p9VfbaG

Loading next page...

References (74)

H. Storkel, Jonna Armbrüster, T. Hogan (2006)
Differentiating phonotactic probability and neighborhood density in adult word learning.
Journal of speech, language, and hearing research : JSLHR, 49 6
James McClelland, B. McNaughton, R. O’Reilly (1995)
Why there are complementary learning systems in the hippocampus and neocortex: insights from the successes and failures of connectionist models of learning and memory.
Psychological review, 102 3
Matthew Mak, Hope Twitchell (2020)
Evidence for preferential attachment: Words that are more well connected in semantic networks are better at acquiring new links in paired-associate learning
Psychonomic Bulletin & Review, 27
Solène Hameau, L. Nickels, B. Biedermann (2019)
Effects of semantic neighbourhood density on spoken word production
Quarterly Journal of Experimental Psychology, 72
H. Storkel (2001)
Learning new words: phonotactic probability in language development.
Journal of speech, language, and hearing research : JSLHR, 44 6
S. Gais, Brian Lucas, J. Born (2006)
Sleep after learning aids memory recall.
Learning & memory, 13 3
Tomas Engelthaler, Thomas Hills (2017)
Feature Biases in Early Word Learning: Network Distinctiveness Predicts Age of Acquisition.
Cognitive science, 41 Suppl 1
Ray Grondin, S. Lupker, K. McRae (2009)
Shared Features Dominate Semantic Richness Effects for Concrete Concepts.
Journal of memory and language, 60 1
Matthew Mak, Yaling Hsiao, K. Nation (2020)
Anchoring and contextual variation in the early stages of incidental word learning during reading
Journal of Memory and Language
S. Gunawardena, W. Rivera, Epstein, C Heathcock, B. Finkelstein, T. Aoki, C. Poulter, D. Jarstfer, C. Zhang, J. Poulter, Am, Chem, B Blagg, M. Jarstfer, D. Rogers, B Bierl-Leonhardt, D. Moreno, M. Schwarz, J. Fargerlund, J. Plimmer, L Tarshis, M. Yan, J. Sacchettini, A Hemmerlin, S. Rivera, H. Erickson, H Erickson, L Altman, R. Kowerski, D. Laungani, J. Am, D. Tse, R. Langston, M. Kakeyama, I. Bethus, P. Spooner, E. Wood, M. Witter, R. Morris
References and Notes Supporting Online Material Materials and Methods Figs. S1 to S5 Tables S1 to S3 Electron Impact (ei) Mass Spectra Chemical Ionization (ci) Mass Spectra References Schemas and Memory Consolidation Research Articles Research Articles Research Articles Research Articles
D. Bates, M. Machler, B. Bolker, S. Walker (2014)
Fitting Linear Mixed-Effects Models Using lme4
Journal of Statistical Software, 67
Matthew Davis, M. Gaskell (2009)
A complementary systems account of word learning: neural and behavioural evidence
Philosophical Transactions of the Royal Society B: Biological Sciences, 364
K. Forster, J. Forster (2003)
DMDX: A Windows display program with millisecond accuracy
Behavior Research Methods, Instruments, & Computers, 35
Ines Wilhelm, Michael Rose, Kathrin Imhof, B. Rasch, C. Büchel, J. Born (2013)
The sleeping child outplays the adult's capacity to convert implicit into explicit knowledge
Nature Neuroscience, 16
Ines Wilhelm, A. Prehn-Kristensen, Jan Born, Jan Born (2012)
Sleep-dependent memory consolidation – What can be learnt from children?
Neuroscience & Biobehavioral Reviews, 36
H. Storkel, Daniel Bontempo, A. Aschenbrenner, Junko Maekawa, Su-Yeon Lee (2013)
The effect of incremental changes in phonotactic probability and neighborhood density on word learning by preschool children.
Journal of speech, language, and hearing research : JSLHR, 56 5
J. Tamminen, M. Ralph, P. Lewis (2013)
The Role of Sleep Spindles and Slow-Wave Activity in Integrating New Information in Semantic Memory
The Journal of Neuroscience, 33
Jill Hoover, H. Storkel, T. Hogan (2010)
A Cross-Sectional Comparison of the Effects of Phonotactic Probability and Neighborhood Density on Word Learning by Preschool Children.
Journal of memory and language, 63 1
James McClelland (2013)
Incorporating rapid neocortical learning of new schema-consistent information into complementary learning systems theory.
Journal of experimental psychology. General, 142 4
D. Bates, R. Kliegl, S. Vasishth, H. Baayen (2015)
Parsimonious Mixed Models
arXiv: Methodology
(1995)
The CELEX lexical database (release 2). Distributed by the Linguistic Data Consortium
(2015)
Testable: The web-based platform for creat
L. Buchanan, Chris Westbury, C. Burgess (2001)
Characterizing semantic space: Neighborhood effects in word recognition
Psychonomic Bulletin & Review, 8
(2011)
WASI-II: Wechsler Abbreviated Scale of intelligence
R. Irizarry (2019)
ggplot2
Introduction to Data Science
K. McRae, George Cree, Mark Seidenberg, C. McNorgan (2005)
Semantic feature production norms for a large set of living and nonliving things
Behavior Research Methods, 37
E. Tham, Shane Lindsay, M. Gaskell (2015)
Markers of automaticity in sleep-associated consolidation of novel words
Neuropsychologia, 71
(2015)
Ordinal-Regression models for ordinal data. R Package Version 2015.6-28
Sanne Kleij, J. Rispens, A. Scheper (2016)
The effect of phonotactic probability and neighbourhood density on pseudoword learning in 6- and 7-year-old children
First Language, 36
Alexandra Kuznetsova, P. Brockhoff, R. Christensen (2017)
lmerTest Package: Tests in Linear Mixed Effects Models
Journal of Statistical Software, 82
P. Pexman, Ian Hargreaves, Paul Siakaluk, Glen Bodner, Jamie Pope (2008)
There are many ways to be rich: Effects of three measures of semantic richness on visual word recognition
Psychonomic Bulletin & Review, 15
J. Antony, C. Ferreira, K. Norman, M. Wimber (2017)
Retrieval as a Fast Route to Memory Consolidation
Trends in Cognitive Sciences, 21
K. McGregor, Ulla Licandro, Richard Arenas, N. Eden, Derek Stiles, Allison Bean, E. Walker (2013)
Why words are hard for adults with developmental language impairments.
Journal of speech, language, and hearing research : JSLHR, 56 6
J. Tamminen, Gareth Gaskell (2013)
Novel word integration in the mental lexicon: Evidence from unmasked and masked semantic priming
Quarterly Journal of Experimental Psychology, 66
D. Barr, R. Levy, Christoph Scheepers, Harry Tily (2013)
Random effects structure for confirmatory hypothesis testing: Keep it maximal.
Journal of memory and language, 68 3
Nicole Goossens, Gino Camp, P. Verkoeijen, H. Tabbers (2014)
The Effect of Retrieval Practice in Primary School Vocabulary Learning
Applied Cognitive Psychology, 28
D. Bishop, J. Barry, M. Hardiman (2012)
Delayed Retention of New Word-Forms Is Better in Children than Adults Regardless of Language Ability: A Factorial Two-Way Study
PLoS ONE, 7
H. Storkel, S. Adlof (2009)
The effect of semantic set size on word learning by preschool children.
Journal of speech, language, and hearing research : JSLHR, 52 2
L. Tyler, M. Gaskell, W. Marslen-Wilson (1997)
Integrating Form and Meaning: A Distributed Model of Speech Perception.
Language and Cognitive Processes, 12
E. James, Gareth Gaskell, A. Weighall, L. Henderson (2017)
Consolidation of vocabulary during sleep: The rich get richer?
Neuroscience & Biobehavioral Reviews, 77
Arielle Borovsky, Erica Ellis, Julia Evans, J. Elman (2016)
Lexical leverage: category knowledge boosts real-time novel word recognition in 2-year-olds.
Developmental science, 19 6
J. Tamminen, Jessica Payne, R. Stickgold, E. Wamsley, M. Gaskell (2010)
Sleep Spindle Activity is Associated with the Integration of New Memories and Existing Knowledge
The Journal of Neuroscience, 30
A. Ashworth, C. Hill, A. Karmiloff-Smith, D. Dimitriou (2014)
Sleep enhances memory consolidation in children
Journal of Sleep Research, 23
N. Dumay, M. Gaskell (2007)
Sleep-Associated Changes in the Mental Representation of Spoken Words
Psychological Science, 18
J. Tamminen, Matthew Davis, Marjolein Merkx, K. Rastle (2012)
The role of memory consolidation in generalisation of new linguistic information
Cognition, 125
D. Balota, M. Yap, K. Hutchison, Michael Cortese, Brett Kessler, B. Loftis, J. Neely, D. Nelson, G. Simpson, R. Treiman (2007)
The English Lexicon Project
Behavior Research Methods, 39
E. Smalle, M. Page, W. Duyck, M. Edwards, A. Szmalec (2018)
Children retain implicitly learned phonological sequences better than adults: a longitudinal study.
Developmental science, 21 5
M. Brysbaert, Amy Warriner, V. Kuperman (2014)
Concreteness ratings for 40 thousand generally known English word lemmas
Behavior Research Methods, 46
E. James, M. Gaskell, Rhiannon Pearce, Caroline Korell, Charlotte Dean, L. Henderson (2021)
The role of prior lexical knowledge in children's and adults' incidental word learning from illustrated stories.
Journal of experimental psychology. Learning, memory, and cognition
R. Hulme, J. Rodd (2021)
Learning new word meanings from story reading: the benefit of immediate testing
PeerJ, 9
James McClelland, B. McNaughton, Andrew Lampinen (2020)
Integration of New Information in Memory: New Insights from a Complementary Learning Systems Perspective
bioRxiv
Thomas Hills, M. Maouene, J. Maouene, Adam Sheya, Linda Smith (2009)
Longitudinal Analysis of Early Semantic Networks
Psychological Science, 20
E. James, M. Gaskell, L. Henderson (2019)
Offline consolidation supersedes prior knowledge benefits in children's (but not adults') word learning.
Developmental science, 22 3
Leoni Lampe, Solène Hameau, L. Nickels (2021)
Semantic variables both help and hinder word production: Behavioral evidence from picture naming.
Journal of experimental psychology. Learning, memory, and cognition
A. Rakita, N. Nikolić, M. Mildner, J. Matiasek, A. Elbe-Bürger (2020)
Re-epithelialization and immune cell behaviour in an ex vivo human skin model
Scientific Reports, 10
D. Kumaran, D. Hassabis, James McClelland (2016)
What Learning Systems do Intelligent Agents Need? Complementary Learning Systems Theory Updated
Trends in Cognitive Sciences, 20
H. Wickham (2009)
ggplot2 - Elegant Graphics for Data Analysis
L. Henderson, A. Weighall, Helen Brown, M. Gaskell (2012)
Consolidation of vocabulary is associated with sleep in children.
Developmental science, 15 5
Robert Davies, Ruth Arnell, Julia Birchenough, Debbie Grimmond, Sam Houlson (2017)
Reading Through the Life Span: Individual Differences in Psycholinguistic Effects
Journal of Experimental Psychology: Learning, Memory, and Cognition, 43
H. Storkel (2008)
Developmental differences in the effects of phonological, lexical and semantic variables on word learning by infants*
Journal of Child Language, 36
F. Clay, J. Bowers, C. Davis, D. Hanley (2007)
Teaching adults new words: the role of practice and consolidation.
Journal of experimental psychology. Learning, memory, and cognition, 33 5
P. Pexman, M. Yap (2018)
Individual Differences in Semantic Processing: Insights From the Calgary Semantic Decision Project
Journal of Experimental Psychology: Learning, Memory, and Cognition, 44
L. Perry, E. Axelsson, Jessica Horst (2016)
Learning What to Remember: Vocabulary Knowledge and Children's Memory for Object Names and Features.
Infant and Child Development, 25
V. Kuperman, H. Stadthagen-González, M. Brysbaert (2012)
Age-of-acquisition ratings for 30,000 English words
Behavior Research Methods, 44
D. Mirman, J. Magnuson, Nicole Landi, J. Rueckl, Whitney Tabor, Ken Mcrae, Lori Buchanan
Attractor Dynamics and Semantic Neighborhood Density: Processing Is Slowed by near Neighbors and Speeded by Distant Neighbors We Thank Ann Kulikowski for Her Help with Data Collection And
P. Pexman (2020)
How does meaning come to mind? Four broad principles of semantic processing.
Canadian journal of experimental psychology = Revue canadienne de psychologie experimentale
V. Kuperman, H. Stadthagen-González, M. Brysbaert (2012)
Age-of-acquisition ratings for 30 thousand English words
D. Nelson, C. McEvoy, Thomas Schreiber (2004)
The University of South Florida free association, rhyme, and word fragment norms
Behavior Research Methods, Instruments, & Computers, 36
M. Coltheart (1981)
The MRC Psycholinguistic Database
Quarterly Journal of Experimental Psychology, 33
R. Team (2014)
R: A language and environment for statistical computing.
MSOR connections, 1
L. Henderson, A. Weighall, Gareth Gaskell (2013)
Learning new vocabulary during childhood: effects of semantic training on lexical consolidation and integration.
Journal of experimental child psychology, 116 3
Milena Rabovsky, D. Schad, R. Rahman (2016)
Language production is facilitated by semantic richness but inhibited by semantic density: Evidence from picture naming
Cognition, 146
E. James, M. Gaskell, L. Henderson (2019)
Sleep-dependent consolidation in children with comprehension and vocabulary weaknesses: it'll be alright on the night?
Journal of child psychology and psychiatry, and allied disciplines
Viorica Marian, James Bartolotti, S. Chabal, Anthony Shook (2012)
CLEARPOND: Cross-Linguistic Easy-Access Resource for Phonological and Orthographic Neighborhood Densities
PLoS ONE, 7

Publisher: Taylor & Francis
Copyright: © 2023 The Author(s). Published by Informa UK Limited, trading as Taylor & Francis Group
ISSN: 2327-3801
eISSN: 2327-3798
DOI: 10.1080/23273798.2022.2102198
Publisher site: See Article on Publisher Site

Abstract

LANGUAGE, COGNITION AND NEUROSCIENCE https://doi.org/10.1080/23273798.2022.2102198 REGULAR ARTICLE Word learning in the context of semantic prior knowledge: evidence of interference from feature-based neighbours in children and adults a b b b b Emma James , M. Gareth Gaskell , Gráinne Murphy , Josie Tulip and Lisa M. Henderson a b Department of Experimental Psychology, University of Oxford, Oxford, UK; Department of Psychology, University of York, York, UK ABSTRACT ARTICLE HISTORY Received 8 April 2022 The presence of phonological neighbours facilitates word-form learning, suggesting that prior Accepted 7 July 2022 phonological knowledge supports vocabulary acquisition. We tested whether prior semantic knowledge similarly beneﬁts word learning by teaching 7-to-10-year-old children (Experiment 1) KEYWORDS and adults (Experiment 2) pseudowords assigned to novel concepts with low or high semantic Vocabulary; word learning; neighbourhood density according to feature norms. Form recall, deﬁnition recall, and semantic prior knowledge; semantic categorisation tasks were administered immediately after training, the next day, and one week neighbourhood density; later. Across sessions, pseudowords assigned to low-density (versus high-density) semantic consolidation; development neighbourhood concepts elicited better word-form recall (for adults) and better meaning recall (for children). Exploratory cross-experiment analyses demonstrated that the neighbourhood inﬂuence was most robust for recalling meanings. Children showed greater gains in form recall than adults across the week, regardless of links to semantic knowledge. While the results suggest that close semantic neighbours interfere with word learning, we consider alternative semantic dimensions that may be beneﬁcial. Our ability to learn new words remains important across The role of prior knowledge in new word acquisition the lifespan, enabling us to communicate about new can be examined both in terms of the initial learning of a things we experience. Encountering an unfamiliar food new word and its long-term retention in vocabulary, as from a diﬀerent cuisine or a rare animal at the zoo, for dissociated within a complementary learning systems instance, requires us to learn its association with a new framework (Davis & Gaskell, 2009; McClelland et al., combination of sounds, encode its key features, and 1995). According to this model, vocabulary knowledge determine how it relates to concepts that we already is stored in a distributed manner in neocortical regions know. For example, we might learn that a pomelo is a of the brain, permitting eﬃcient language processing type of fruit that is similar to a grapefruit, but larger and and communication (Gaskell & Marslen-Wilson, 1997). sweeter. In using what we know about grapefruits to For a new word to become part of this system, it must support new learning, we bring a broader understanding do so in such a way to avoid disrupting existing connec- of citrus fruit features—inferring that a pomelo likely has a tions. The complementary learning systems model pro- waxy peel, pips, and juicy ﬂesh. However, the availability poses that initial lexical representations are formed of this related knowledge will vary across new concepts using the hippocampal memory system, with the new we encounter, and it is not clear the extent to which it form-meaning mapping not yet fully integrated with helps or hinders memory for the new word. We addressed existing neocortical vocabulary (Davis & Gaskell, 2009). this question by examining how children’sand adults’ This system allows knowledge of a new word to be learning of novel concepts is aﬀected by the density of acquired rapidly without disrupting existing connec- associated semantic neighbourhoods. Determining the tions, and the new representation can then support role of prior knowledge has important implications for slower integration of the new word into neocortical- theories of word learning, as well as for understanding based networks. how word learning changes as semantic knowledge The strengthening of neocortical connections is pro- accumulates across development. posed to result from repeated reactivations of the CONTACT Emma James emma.james@psy.ox.ac.uk Department of Experimental Psychology, University of Oxford, Anna Watts Building, Oxford, OX2 6GG, UK Supplemental data for this article can be accessed http://doi.org/10.1080/23273798.2022.2102198. © 2022 The Author(s). Published by Informa UK Limited, trading as Taylor & Francis Group This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. 2 E. JAMES ET AL. hippocampal representation, which can occur “oﬄine” One way of examining the inﬂuence of prior knowl- during sleep. In line with this, increased competition is edge on word learning is to manipulate the psycholin- observed between new and familiar words in lexical pro- guistic properties of the to-be-learned words. By cessing tasks following sleep but not wake in both adults quantifying the similarities between new items and real (Dumay & Gaskell, 2007) and children (Henderson et al., words that would be known to the learner in diﬀerent 2012), suggesting that sleep enhances integration with ways, we can examine whether these relationships neocortical vocabulary. These sleep-associated consoli- predict memory for new words before and/or after oppor- dation processes are further associated with improved tunities for consolidation. In examining word-form simi- accuracy and eﬃciency in recalling the new word- larity, many studies have demonstrated that novel forms (Ashworth et al., 2014; Gais et al., 2006; James words with many phonological neighbours (i.e. words et al., 2020a; Tamminen et al., 2010). Deliberate retrieval that diﬀer by a single sound) are more readily learned practice may also enhance recall of new word knowl- than words with fewer phonological neighbours. This edge (e.g. Hulme & Rodd, 2021), proposed to operate beneﬁt has been found for pre-school children (Hoover via similar principles of hippocampal reactivation et al., 2010;Storkel, 2009;Storkel et al., 2013), school- (Antony et al., 2017). These ﬁndings converge on the aged children (James et al., 2019;van derKleij et al., importance of memory reactivation processes in sup- 2016), and adults(Jameset al., 2019; Storkel et al., porting vocabulary acquisition, and highlight the impor- 2006), suggesting that individuals can access prior knowl- tance of understanding factors that inﬂuence this edge to support learning across the lifespan (although longer-term consolidation of new word knowledge as note that this support may be less apparent in incidental well as those that support initial encoding. word learning; James et al., 2020b). From a complemen- One factor proposed to support lexical consolidation tary learning systems perspective, James et al. (2019) is the learner’s prior linguistic knowledge (James et al., found that this early beneﬁt from phonological neigh- 2017). Recent progress in complementary learning bours was reduced one week later, once words with systems theory has examined how the speed of neocor- fewer neighbours had opportunities to beneﬁt from tical learning may be inﬂuenced by memory schema oﬄine consolidation processes. These ﬁndings support (e.g. Kumaran et al., 2016; McClelland, 2013; McClelland the proposal that items related to prior word-form knowl- et al., 2020), driven by evidence that new information edge may be integrated into neocortical systems more related to existing knowledge becomes integrated into rapidly, and are therefore less dependent on oﬄine reac- neocortical networks very rapidly (Tse et al., 2007). For tivation processes to support long-term memory. example, McClelland (2013) demonstrated that a In this study, we turn our attention towards semantic neural network trained with structured knowledge of neighbours to develop a broader understanding of how birds and their semantic properties (e.g. grow, move, diﬀerent aspects of prior knowledge inﬂuence memory ﬂy) showed more rapid learning of a typical new for new vocabulary. Learning the meaning of a new example that shared these properties (a cardinal) than word is a crucial part of vocabulary acquisition itself, an atypical example (a penguin, which cannot ﬂy but and also supports the long-term retention of new can swim). This model demonstrates how information word-form knowledge (Henderson et al., 2013). that capitalises upon existing knowledge can be inte- However, while psycholinguistic measures that capture grated into neocortical networks without requiring semantic neighbourhoods have well-documented inﬂu- repeated hippocampal reactivations (Tse et al., 2007). ences on the processing of known words (see Pexman, From a language learning perspective then, it follows 2020, for a recent review), they have received relatively that individuals with rich lexical networks will likely little attention in experimental studies of word learning. have more relevant structures to draw upon to Network approaches to modelling language acquisition support new word learning, and thus consolidation support that semantic neighbours broadly predict voca- into neocortical networks is proposed to proceed more bulary growth, ﬁnding that new words are more likely to rapidly. Capitalising upon existing knowledge may be be acquired when they are highly connected to known one means by which individuals with good vocabulary ones (Engelthaler & Hills, 2017; Hills et al., 2009), knowledge acquire new vocabulary at faster pace than although semantically distinct words may be acquired those with weak vocabulary, thereby increasing the earlier (Engelthaler & Hills, 2017). Yet only two studies “vocabulary gap” in performance across development to our knowledge have examined the learning and (James et al., 2017). However, the nature of this lexical retention processes that might underlie these inﬂu- support is not well-speciﬁed, and it is not clear how ences: Tamminen et al. (2013) in a study of sleep-associ- word learning might be inﬂuenced by diﬀerent aspects ated consolidation in adults, and Storkel and Adlof of linguistic knowledge. (2009) in a study of preschool children. In contrast to LANGUAGE, COGNITION AND NEUROSCIENCE 3 the beneﬁts seen for word-form neighbours, the ﬁndings processes enhance engagement of the new words of both studies point towards interference from related with existing semantic knowledge. semantic knowledge when learning new words. Thus, unlike for word-form neighbours, it appears from Tamminen et al. (2013) taught adults pseudowords as the existing research that connections to semantic knowl- names for novel concepts with either sparse or dense edge may interfere with new word acquisition. However, semantic connections. These novel concepts were semantic relationships can be conceptualised along created from existing concepts with either low- or high- several diﬀerent dimensions (Hameau et al., 2019), and density semantic neighbourhoods, as quantiﬁed by the diﬀerent measures have been shown to contribute number of associates provided in free association norms unique variance in predicting performance in speeded (Nelson et al., 2004). To make each concept novel, they word recognition tasks (Pexman et al., 2008). Tamminen added a novel feature (e.g. bee whose sting feels pleasant, et al. (2013) derived their measures of semantic neigh- crab that has a beak). Participants were tested on their bourhood from free-association norms, which were knowledge of the new pseudowords immediately after created by asking participants to produce the ﬁrst word learning, the next day, and one week later—allowing for that came to mind when presented with the target an assessment of the inﬂuence of prior knowledge (Nelson et al., 2004). Free-association norms are con- before and after opportunities for consolidation. Across sidered a language-based measure that typically reﬂects all test sessions, participants had poorer performance for the co-occurrence of concepts in spoken and/or written high- versus low-density concepts in a synonym judge- language, and can be broadly distinguished from object- ment task, which required them to identify which of based metrics that draw upon the content of the concepts three familiar words was associated with the trained themselves (Buchanan et al., 2001). For example, feature word. This perhaps indicates that participants experi- production norms are collected by asking participants enced interference from existing knowledge in learning to list features of target concepts (McRae et al., 2005), pro- the new concepts. However, there was no inﬂuence of ducing measures of semantic richness (i.e. the number of the density of prior semantic knowledge on explicit features produced) and common features between con- recall of the new word-forms or their meanings, cepts. Semantic neighbourhoods conceptualised in this suggesting that this interference might arise from way may arguably be more beneﬁcial to a new learner lexical processing during the synonym judgement than linguistic co-occurrence. For example, the target task rather than during learning per se. In line with “bird” leads to “cat” as a more frequent lexical associate this, responses only slowed to high-density novel than “robin” (Nelson et al., 2004), yet knowledge of items in a speeded semantic categorisation task one birds is intuitively more useful when learning about a week later, suggesting that a longer period of consoli- robin than a cat. In this study, we test the hypothesis dation enhanced competition between new and that semantic prior knowledge deﬁned by object-based known concepts as they became better integrated in similarities will facilitate word learning and consolidation. memory (similar to the way in which lexical compe- This proposal garners support from studies of early tition was observed following sleep in the studies language learning that capture the variability in knowl- described above). Thus, dense existing semantic knowl- edge that preschool children bring to the task. For edge elicited interference when processing the novel example, Borovsky et al. (2016)demonstrated that concepts in the context of familiar words, but there infants were more able to learn and recognise new was no evidence that memory for the new words them- words from categories that they had more knowledge selves was aﬀected by links to semantic knowledge. about compared to categories for which they had lower However, evidence of semantic interference in estab- levels of existing knowledge. Similarly, Perry et al. (2016) lishing new word representations comes from a study found that preschool children with larger shape-based with children. Storkel and Adlof (2009) quantiﬁed noun vocabularies were more likely to remember object semantic set sizes of non-objects by collecting free shapes during word learning. These studies support the associations from presented line drawings. In a sub- proposal that semantic knowledge relevant to a new con- sequent word-learning task, preschool children were cept’s properties may aid acquisition, although they more accurate in identifying nonwords associated with cannot address questions of longer-term memory as chil- objects from small semantic set sizes, suggesting that dren were only tested on the same day as learning. —at least in young children—connections to existing semantic knowledge can interfere in learning and/or The present study remembering new information. In this study, the eﬀect only emerged at a delayed test one week later, We used a similar design to Tamminen et al. (2013), but oﬀering further support to the hypothesis that oﬄine instead drew upon shared features as an object-based 4 E. JAMES ET AL. measure of semantic relationships to test whether example, Davies et al. (2017) showed that eﬀects of psy- semantic neighbourhood density can beneﬁt word cholinguistic variables on lexical processing decline learning (rather than the associative measure used pre- across the lifespan as the lexical system accumulates viously). We taught participants pseudowords and experience and maximises learning eﬃciency. We associated deﬁnitions, formed by adding a novel present two experiments to explore these possibilities feature to known concepts from low and high feature across children aged 7–10 years (Experiment 1) and density neighbourhoods. We tested explicit recall of adults (Experiment 2). While the developmental studies the pseudowords and deﬁnitions immediately after described above typically assessed semantic prior learning, the next day, and one week later, to examine knowledge inﬂuences in pre-school children, selecting the inﬂuence of existing knowledge on word learning a school-aged group allowed us to use the same exper- before and after opportunities for consolidation. We imental tasks for children and adults, thus facilitating also used a speeded semantic categorisation task to developmental comparisons. Further, these age groups index integration of the new words into neocortical overlap with those that have been examined in studies vocabulary, capitalising on previous ﬁndings that of phonological neighbours in word learning (James known words with high semantic neighbourhood et al., 2019; James et al., 2020b), allowing comparisons density are responded to more quickly in this task than with diﬀerent types of prior knowledge inﬂuence. words with low semantic density (Mirman & Magnuson, Our main research questions were as follows: First, is 2008). Tamminen et al. (2013) found that implicit seman- explicit memory for novel words helped or hindered by tic neighbourhood eﬀects for trained pseudowords links to existing semantic knowledge during word learn- emerged only after a period of consolidation, consistent ing, as deﬁned by object-based norms? Second, can with other studies that have found implicit semantic newly trained concepts acquire the lexical properties activation to emerge after one/more periods of sleep of their neighbours, beneﬁting from rich semantic con- (Clay et al., 2007; Tham et al., 2015). nections in speeded reaction time tasks? Third, what is One motivation for understanding the role of seman- the time course of this engagement with semantic tic prior knowledge is to understand how word learning knowledge? By using both explicit and implicit measures might change across development. The studies of new word knowledge, we aimed to capture initial described above span a broad age range from preschool encoding processes and those that indicate integration children to adults, yet there is a lack of direct develop- with existing knowledge, proposed to require periods mental comparisons. A study of known words taken of oﬄine consolidation. Finally, across experiments we from linguistic corpora suggested that young infants also explored whether children and adults are diﬀerently start by learning words from sparse semantic neigh- inﬂuenced by semantic knowledge during word learn- bours but increasingly beneﬁt from dense neighbours ing, reﬂecting diﬀerences in the amount of prior knowl- across the preschool years (Storkel, 2009). However, edge or their sensitivity to it during learning. In the ﬁrst these developmental diﬀerences have not been tested experiment with children, we thus tested three exper- experimentally, and no studies to our knowledge have imental hypotheses: 1) A large number of shared fea- considered the inﬂuence of semantic neighbours in tures will facilitate explicit aspects of word learning, as school-aged children. There are two possibilities here: demonstrated by superior performance in recall and rec- ﬁrst, school-aged children may show smaller semantic ognition tasks; 2) Novel concepts that share lots of fea- density eﬀects than adults given that our selected tures with existing concepts should show a reaction measure is based on adult norms, and children may time advantage when compared to novel concepts not have yet acquired rich enough knowledge about that share fewer features, in a speeded semantic categ- concepts to have such extreme diﬀerences in low- orisation task; and 3) Across tasks, eﬀects of neighbour- versus high-density items. Speaking to this proposal, hood density (i.e. better recall/recognition for high Pexman and Yap (2018) found that adults with better density items; a density eﬀect in speeded semantic cat- existing vocabulary knowledge showed more sensitivity egorisation) will emerge only after a night’s sleep (24- to semantic neighbourhood density in speeded hour test) or longer period of consolidation (week responses during a semantic categorisation task com- follow-up test). pared to adults with lower vocabulary knowledge, suggesting that the knowledge learners bring to the Experiment 1 task is highly relevant. Alternatively, children might show larger semantic density eﬀects under the possi- Experiment 1 can be considered exploratory in the sense bility that an underdeveloped system may be more sen- that it was not pre-registered, and that the sample size sitive to the inﬂuence of existing knowledge. For was determined opportunistically (school availability LANGUAGE, COGNITION AND NEUROSCIENCE 5 within the timeframe of the study). However, our analy- pronounceable (Appendix 1), and these were split into sis plan and key hypothesis tests were consistent with a two lists matched on length, number of orthographic pre-registered adult study being conducted in parallel neighbours, and bigram frequency (Marian et al., (Experiment S1, http://osf.io/3vnsg; detailed below). 2012). A subset of 16 pseudowords were selected for Experiment 1 with children. Novel concepts were created by taking an existing Experiment 1 methods base concept and adding an additional feature (further details below). For example, a gorilla (base concept) Participants that has green skin (added feature). Half of the base con- Two whole classes of children took part in the study, cepts were animals, for purposes of the semantic categ- recruited via two schools in North Yorkshire. The result- orisation task. Critically, the base concepts were selected ing sample included 51 children (25 male) aged 7–10 for having high (n = 12) or low (n = 12) semantic neigh- years (M = 8.67 years). One additional child was excluded bourhood density according to the McRae et al. (2005) from analyses due to hearing diﬃculties. Two of the 2 feature norms. In this ﬁrst step towards examining included children were absent on the second day of feature-based semantic neighbourhoods in word learn- testing, and thus only contributed data for two out of ing, we took a broad approach to deﬁning semantic the three follow-up tests. neighbourhood density. First, we selected for the The study was approved by the Research Ethics Com- number of features of each item (also termed semantic mittee of the Department of Psychology, University of richness), as has consistently been shown to facilitate York. Consent was obtained from the school head tea- lexical processing (Pexman, 2020). Second, we con- chers. Parents were fully informed about the study and sidered the density of the semantic neighbourhood by were given the opportunity to opt their child out of selecting for low versus high intercorrelational feature taking part. density—the extent to which the listed features co- occurred in other normed concepts. This metric is described in detail within the database documentation Design and procedure (McRae et al., 2005): pairs of features are considered sig- Children completed a single training session in a whole- niﬁcantly correlated if they share ≥ 6.5% of their variance class setting, which lasted approximately 45 min. Test within the database (i.e. they often co-occur together in sessions were then conducted individually in a quiet the 541 normed concepts), and the proportion of signiﬁ- setting outside the classroom at three time points: the cantly correlated feature pairs is calculated for each same day (T1), the next day (T2), and one week later concept. We predicted that this co-occurrence would (T3). Standardised assessments of vocabulary and non- also support learning, being indicative of many shared verbal ability (matrix reasoning) from the Wechsler features that could support processing (Grondin et al., Abbreviated Scale of Intelligence II (Wechsler, 2011) 2009) and representing many existing connections were also collected during these sessions for descriptive between concepts. purposes. The mean t-scores were within the average Our selected low-density base concepts had fewer range for both matrix reasoning (M = 46.53, SD = 9.73) features listed in the norms (≤ 16), and fewer of these and vocabulary (M = 59.71, SD = 11.92). listed features (≤ 14%) co-occurred in other normed concepts. High-density base concepts had more features overall (≥ 18) and more of these (≥ 25%) also co- Stimuli occurred in other concepts. The two groups of stimuli Pseudowords were initially selected using the English were otherwise well matched on measures of frequency, Lexicon Project (Balota et al., 2007) according to the fol- age of acquisition, imageability, concreteness and word lowing criteria: 5–6 letters long, no orthographic neigh- length (Table 1). A pilot study of these base concepts bours, and a nonword rejection Z-score of −0.45 to 0.45 with adults supported a reaction time beneﬁt for high- (i.e. an average range response time for rejection in a density concepts in a semantic categorisation task lexical decision task). These criteria were used to (mean diﬀerence = 14 ms; t(70) = 2.56, p = .01). ensure that the word-forms were well-matched across The added features that made each concept novel conditions, and to minimise alternative sources of varia- were also selected from the McRae et al. (2005) norms, bility in the speeded semantic categorisation task. and each occurred only once in the norms to minimise Twenty-four bisyllabic pseudowords were selected in the inﬂuence of additional semantic neighbourhoods. total such that each began with diﬀerent vowels or con- The features were drawn from a range of perceptual, sonant clusters and were judged to be easily behavioural and functional categories, which were 6 E. JAMES ET AL. Table 1. Properties of stimuli in the low and high semantic neighbourhood density conditions. a a b c c d e No. of features % features correlated AoA Frequency Log10 freq Imageability Concreteness No. of phonemes Low 12.75 5.58 5.14 16.41 1.02 607.56 4.89 4.33 High 18.92 40.00 5.17 16.38 1.13 616.00 4.89 4.50 p <.001* <.001* .96 1 .56 .61 .90 .79 a b c d e McRae et al. (2005). Kuperman et al. (2012). CELEX English linguistic database (Baayen et al., 1995). MRC Psycholinguistic database (Coltheart, 1981). Brys- baert et al. (2014). *signiﬁcant diﬀerence between low vs high semantic density items at p < .05. matched in type across low- and high-density base con- Form repetition cepts (Appendix 1). To ensure that these combinations Children heard each new word-form spoken by the of base concepts and features did not diﬀer in plausi- experimenter, with its orthographic form projected on bility across low- and high-density conditions, 58 the PowerPoint at the front of the classroom. They adults completed online ratings of how plausible they repeated the pseudoword aloud twice, and sub- would ﬁnd each item in a children’s storybook. High- sequently copied it into their workbooks. and low-density items did not diﬀer in plausibility (ps >.2). A subset of 16 items (8 per density condition) Deﬁnition repetition were selected for Experiment 1. A single ﬁxed set of Children were introduced to the deﬁnition of each pseu- pseudoword-concept pairings was used for Experiment doword, and again repeated it aloud twice. 1 (with counterbalancing of pseudowords across density conditions introduced in Experiment 2). Drawing task Children were given 30 s per item to draw a picture of the new concept, designed to help them to engage with its diﬀerent features. These were not scored or ana- Training tasks lysed further. The training tasks were conducted with the class as a whole. Children were given workbooks to support their Meaning matching learning, and were guided through a number of tasks After the workbooks had been collected, further learning using a PowerPoint presentation projected at the front and feedback took place via a multiple choice quiz. In the of the classroom (see Figure 1). The ﬁrst three tasks ﬁrst round, a pseudoword and three possible options for were completed for each item in turn (form and its deﬁnition were presented on screen, and children had deﬁnition repetition, drawing), followed by the to show their answer by raising one, two or three ﬁngers. meaning matching task for all items. In total, children In the second round, the deﬁnition was presented and the heard each new word-form nine times, and each children had to choose the correct word-form to match. deﬁnition six times. Each item was presented once in each round, with the correct answer provided after each one. Test tasks Children completed the test tasks individually with the experimenter. There were four test tasks to assess diﬀerent aspects of word knowledge. All test tasks were presented using DMDX software v5.1.3.4 (Forster & Forster, 2003), with item order randomised. The tasks were presented in the following ﬁxed order. Cued form recall Children were presented with the ﬁrst consonant(s) and vowel of the word (both aurally and visually), and were asked to speak the remainder of the word. Children Figure 1. Schematic of the training tasks used in the learning phase. were encouraged to attempt partial responses even if Note: For each item, children ﬁrst repeated the new word aloud twice and they were not sure of the answer, and the experimenter wrote it down, before repeating the deﬁnition twice. They were given 30 transcribed the responses for scoring on the basis of s to draw a picture of the new concept. After completing these learning tasks for all items, children completed two rounds of multiple choice quizzes. whole word accuracy (0, 1). LANGUAGE, COGNITION AND NEUROSCIENCE 7 Form recognition contributing to model ﬁt(p <.2). We then used a Children were presented with auditory and orthographic forward “best-path” approach to test for the inclusion of presentations of the pseudoword alongside a corre- appropriate random slopes (Barr et al., 2013). The results sponding foil in which the ﬁnal vowel was changed presented are from the most complex model supported (see Appendix 1). Both of the written stimuli remained by the data. The data and analysis scripts are available on screen for up to 7 s, or until the child had selected on the OSF (https://osf.io/35ftn). Figures were made their answer with a key press response. using ggplot2 (Wickham, 2009). Speeded semantic categorisation Experiment 1 results Children were presented with each word-form visually Cued form recall and auditorily, and were asked to make speeded judge- ments about whether or not the concept was an animal Children recalled a mean proportion of .20 (SD = .40) of the using a key press response. They were asked to respond word-forms at T1, and performance improved substantially as quickly and accurately as possible, and each trial ter- over tests (Figure 2A; delay1: β = 0.95, SE = 0.05, Z = 21.10, p minated after a response or 7 s. To allow for adjustment <.001). Recall continued to improve between T2 (M = .51, to the task and response format, the experimental task SD = .50) and T3 (M = .80, SD = .40; delay2: β = 0.91, SE = was preceded by 24 practice trials using existing 0.07, Z = 13.35, p <.001).There wasnoinﬂuence of seman- English words, providing feedback for erroneous tic neighbourhood density in recall of word-forms, alone or responses. We analysed both the accuracy (0, 1) and in interaction with test session (ps>.6; Table 2). the response time (ms) for correct trials. Form recognition Cued meaning recall Children were given an auditory and visual presentation Children could successfully recognise the new word-forms of each word-form, and asked to provide as much of the at above chance levels at T1 (M= .83, SD = .38), and deﬁnition as they could remember. Verbal responses improved at subsequent tests (T2: M = .92, SD = .28; T3: were transcribed by the experimenter. A total of two M = .94, SD = .24). This eﬀect of test session was statistically points could be awarded per item for correctly recalling signiﬁcant across both contrasts (delay1: β = 0.39, SE = 0.05, the base concept and the added feature. Z= 8.07,p < .001; delay2: β = 0.21, SE = 0.10, Z = 2.08, p = .037), again demonstrating signiﬁcant improvements in form knowledge across the week. As with the recall of Analyses word-forms,there wasnoinﬂuence of semantic neigh- Data were analysed in R (R Core Team, 2015), using lme4 bourhood density on their recognition (ps > .18; Table 2). (Bates et al., 2015b) and ordinal (Christensen, 2015)to ﬁt mixed eﬀects models. For each dependent variable, we Cued meaning recall initially ﬁtted a model with ﬁxed eﬀects of test session, semantic density, and their interaction. Fixed eﬀects Children scored an average of .36 out of a maximum of 2 were deviance coded to enable interpretation of each points for each item at T1 (SD = .76). There were no sig- predictor in relation to the overall mean. Test session niﬁcant changes in performance across test sessions (ps is a three-level factor, and we set two orthogonal con- > .36; Table 3), but there was a signiﬁcant diﬀerence in trasts to interpret the data: delay1 tested for diﬀerences memory for words from diﬀerent semantic neighbour in memory performance without versus with opportu- conditions (β = −0.48, SE = 0.18, Z = −2.62, p = .009). nities for consolidation (T1 vs. T2&T3); delay2 tested for Children were better at recalling deﬁnitions with low continued changes across the week (T2 vs. T3). For semantic neighbourhood density (M = .47, SD = .84) models with discrete dependent variables, Wald’sZ than high semantic neighbourhood density (M = .26, was used to determine statistical signiﬁcance. For reac- SD = .67; Figure 2C). There was no evidence of an inter- tion times, we report signiﬁcance computed using the action between test session and semantic neighbour- lmerTest package (Kuznetsova et al., 2017). hood density (pruned from the ﬁnal model; p = .687). In light of earlier convergence issues in attempting to ﬁt maximal models (Barr et al., 2013), we adopted a parsimo- Semantic categorisation nious modelling approach for these experiments (Bates et al., 2015a). We ﬁrst ﬁtted a model with our ﬁxed Accuracy eﬀects of interest and random intercepts for participants Performance was very low on the semantic categoris- and items, and then pruned away the interaction if not ation task (M = .59, SD = .49). Neither test session nor 8 E. JAMES ET AL. Figure 2. Explicit recall performance by semantic neighbourhood condition and test session. Note: Proportion correct for (A) Form recall in Experiment 1; (B) Form recall in Experiment 2; (C) Meaning recall in Experiment 1; and (D) Meaning recallin Experiment 2. Individual points mark average participant recall for each condition for each test session. Error bars denote 95% conﬁdence intervals. semantic neighbourhood density inﬂuenced accuracy The data were log-transformed to remediate issues of on this task (all ps > .4; Table A2-1). skewness in model ﬁtting. We also removed responses < 200 ms or that were ≥ 2.5 standard deviations above each participant’s condition mean. We analysed RTs to Reaction time correct responses only, leaving 49.05% of original trials. We were cautious in analysing the RT data considering Responses were slowest at T1 (M = 2154 ms, SD = that performance accuracy was so low in this task, and 1211 ms) compared to later test points (β = −0.07, SE = removed participants who were at/below chance perform- 0.01, t = −7.66, p < .001), but the decrease in response ance (n = 11). This left 40 participants in the analysis, who times between the T2 (M = 1833ms, SD = 1150 ms) and ranged from .52-.83 in categorisation accuracy (M =.63). T3 (M = 1696 ms, SD = 977 ms) tests were not statistically Table 2. Final analysis models for Experiment 1: Form tasks. a b Cued form recall Form recognition β SE Z p β SE Z p (Intercept) 0.02 0.28 0.07 .946 2.89 0.25 11.38 <.001 delay1 0.95 0.05 21.10 <.001 0.39 0.05 8.07 <.001 delay2 0.91 0.07 13.35 <.001 0.21 0.10 2.08 .037 density 0.07 0.25 0.28 .780 −0.29 0.22 −1.34 .181 Note: (a) Analysis based on from 2416 observations across 51 participants and 16 items. The ﬁnal model included by-participant and by-item intercepts. The two-way interaction between time and density was pruned from the model with no reduction in model ﬁt χ = 0.88, p = .64; (b) Analysis based on from 2416 observations across 51 participants and 16 items. The ﬁnal model included by-participant and by-item intercepts, as well as by-participant random slopes for the eﬀect of density. The two-way interaction between time and density was pruned from the model with no reduction in model ﬁt χ = 1.76, p = .41. LANGUAGE, COGNITION AND NEUROSCIENCE 9 Table 3. Final analysis model for Experiment 1: Cued meaning concepts. This hindrance was observed at the immediate recall. test and did not change at the delayed tests. High- rela- β SE Z p tive to low-density semantic knowledge activated 0|1 2.11 0.27 7.69 <.010 during encoding may thus interfere in forming the new 1|2 2.19 0.27 7.99 <.001 representation, and/or could make the novel concept delay1 0.01 0.04 0.22 .827 delay2 −0.07 0.08 −0.9 .368 harder to retrieve amongst its competitors at test. density −0.48 0.18 −2.62 .009 In Experiment 2, we examined the contribution of Note: Analysis based on from 2416 observations across 51 participants and semantic neighbourhood density to word learning in 16 items. The ﬁnal model included by-participant and by-item intercepts. The two-way interaction between time and density was pruned from the adults. In an adult study carried out in parallel to Exper- model with no reduction in model ﬁt χ = 0.75, p = .69. iment 1, we found no inﬂuence of semantic neighbour- hood density in a comparable word learning experiment signiﬁcant (β = −0.03, SE = 0.02, t = −1.84, p = .066). (Experiment S1, available on the OSF at https://osf.io/ There was no inﬂuence of semantic neighbourhood ksdfu/). Speciﬁcally, adults did not show interference from dense semantic neighbourhoods in recalling the density on reaction times (ps > .14; Table A2-2). word meanings, as we observed for children. However, adults showed much higher levels of recall performance than children, and diﬀerent training tasks were used Experiment 1 discussion across the two experiments. To facilitate developmental Experiment 1 examined how children learn and remem- comparisons, we repeated the experiment with adults ber pseudowords paired with novel concepts. Children using the same training procedures as Experiment 1, recalled 20% of the pseudowords on the same day as but we reduced the number of exposures during train- learning, but showed substantial improvements across ing to ensure comparable levels of performance the week: averaging 51% and 80% at the day and between age groups. week follow-up tests, respectively. However, they were much poorer at learning the word meanings: they Experiment 2 showed low accuracy in both the meaning recall (18%) and semantic categorisation (59%) tasks, which neither Three hypotheses were pre-registered on the Open improved nor declined with repeated tests. This increase Science Framework (http://osf.io/yk3d5): 1) Cued recall in recall for word-forms is consistent with previous for word-forms will improve over time, consistent with ﬁndings of an oﬄine consolidation and/or retrieval prac- Experiment 1 (and Experiment S1), and with extant evi- tice beneﬁt for this aspect of word knowledge, and adds dence supporting strengthening of novel word-forms to growing evidence that deﬁnition recall does not by delayed tests; 2) Where a neighbourhood density beneﬁt from the same opportunities for reactivation eﬀect emerges, we predict that low-density items will (James et al., 2020a; Tamminen et al., 2012; Tamminen be better learned than high-density items—consistent & Gaskell, 2013). However, participants may also have with our ﬁndings from Experiment 1 (and non-signiﬁcant beneﬁted from opportunities to re-encode the word- numerical diﬀerences in Experiment S1); and 3) If the forms (but not meanings) during repeated tests in this absence of a density eﬀect in the deﬁnitions task for study, facilitating improvements in this aspect of word adults in Experiment S1 was driven by their higher per- knowledge across the week. formance, then we would expect a neighbourhood Our primary research questions related to the new density eﬀect to emerge at lower performance levels words’ engagement with existing semantic knowledge, in this task. However, if the absence of the density as indicated by performance diﬀerences related to eﬀect is driven by adults’ learning eﬃciency (relative to semantic neighbourhood density. We found that existing the enhanced sensitivity of developing learners to semantic knowledge can inﬂuence new vocabulary acqui- semantic competitors), we would expect no eﬀect of sition in school-aged children: they were better at recal- density in the deﬁnitions task for adults regardless of ling novel semantic concepts from low- versus high- performance levels. density semantic neighbourhoods. However, recall of word-forms appeared unaﬀected by these semantic manipulations. Thus, in line with the processing interfer- Experiment 2 methods ence observed for language-based neighbourhoods pre- Participants viously in adults (Tamminen et al., 2013), dense feature- based semantic neighbourhoods also appear to elicit 70 participants were recruited via the University of York interference observable in children’s learning of new Psychology Department participant pool according to 10 E. JAMES ET AL. the following criteria: native monolingual English speak- with children. Participants circled their meaning match- ers, aged 18-35, with normal or corrected-to-normal ing answers (1, 2, or 3) in an additional training booklet. hearing and vision, and no reading or language dis- orders. Three participants did not complete more than Test tasks one of the three follow-up sessions, and were excluded from analyses. Thus, the ﬁnal sample consisted of 67 par- The four test tasks were programmed for participants ticipants (14 male), with a mean age of 20.33 years (SD = to complete online from home, in the same ﬁxed 2.54). Nine participants contributed only partial data (2/3 order described above. Adults were provided with sessions) having missed the ﬁnal session. only the written cues, and gave typed responses for Participants received either £10 or course credit for the recall task. Given that we were most interested in their time. The study was approved by the Department vocabulary learning (rather than orthographic learning of Psychology Research Ethics Committee at the Univer- speciﬁcally), answers were scored according to whether sity of York. they read as phonologically correct (e.g. attee or atty instead of attie; chiypod instead of chipod). The form and deﬁnition recall tasks were hosted online using Design and procedure Qualtrics (Qualtrics, 2014). A link within the survey took participants to the form recognition and semantic To make Experiment 2 as comparable as possible to categorisation tasks, which were programmed using Experiment 1, we conducted training in a group setting Testable (Rezlescu, 2015) to enable response time lasting approximately 45 min. Three test sessions were recordings. then completed online according to the same schedule: the same day (T1), next day (T2), and one week later (T3). Participants were asked to complete the ﬁrst test Analyses session within 2 h of training, and complete each sub- Analyses were conducted as in Experiment 1. sequent session at a similar time (by 6pm at the latest). All sessions completed on the correct day were included in the analyses. Although we did not implement speciﬁc Experiment 2 results attention checks in the online tests, inspection of task per- formance conﬁrms that participants were engaged with Cued form recall the activities (i.e. recognition task performance was The proportion of word-forms recalled on the same day always well above chance, ≥ 67%). No standardised of learning (M = .21, SD = .40) was highly comparable to assessments were collected for Experiment 2. Experiment 1 (M = .20, SD = .40), suggesting a similar level of diﬃculty for children and adults. Recall improved signiﬁcantly at the delayed tests (delay1: β = 0.39, SE = Stimuli 0.03, Z = 13.32, p < .001; Figure 2B), but continued The full set of 24 items were used for the adults. We improvements between T2 (M = .36, SD = .48) and T3 additionally incorporated two elements of counterba- (M = .38, SD = .49) were not statistically signiﬁcant (p lancing for this experiment to ensure that idiosyncratic = .122; Table 4). diﬀerences in the stimuli were not responsible for the There was a small but statistically signiﬁcant eﬀect of neighbourhood density eﬀects. The two versions of the density (β = −0.11, SE = 0.05, Z = −2.254, p = .025): word- stimuli altered the set of pseudowords and novel fea- forms associated with low neighbourhood density con- tures assigned to each density condition. cepts were better recalled (M = .33, SD = .47) than those associated with high-density concepts (M = .29, SD = .46). This density eﬀect did not change over time, Training tasks and the interaction was pruned from the ﬁnal model (p = .383). The training tasks were identical to Experiment 1, except with form and deﬁnition repetitions reduced to one per item. Only one round of meaning matching was admi- Form recognition nistered, presenting each deﬁnition once with three options for its word-form on each occasion. This meant A technical issue meant that T1 form recognition and that participants had ﬁve exposures to the new word- semantic categorisation data from the ﬁrst set of partici- forms in total, and only two exposures to the deﬁnitions, pants was not saved from Testable (n = 9), and this issue intended to reduce adults’ performance levels in line also aﬀected a later session for two participants. LANGUAGE, COGNITION AND NEUROSCIENCE 11 Table 4. Final analysis models for Experiment 2: Form tasks. a b Cued form recall Form recognition β SE Z p β SE Z p (Intercept) −1.11 0.24 −4.62 <.001 3.38 0.26 12.89 <.001 delay1 0.39 0.03 13.32 <.001 0.11 0.04 2.59 .010 delay2 0.07 0.05 1.55 .122 −0.04 0.08 −0.56 .578 density −0.11 0.05 −2.24 .025 0.21 0.11 1.97 .049 Note: (a) Analysis based on from 4608 observations across 67 participants and 24 items. The ﬁnal model included by-participant and by-item random intercepts, and by-participant random slopes for the eﬀect of density. The two-way interaction between time and density was pruned from the model with no reduction in model ﬁt χ = 1.92, p = .38; (b) Analysis based on from 4200 observations across 65 participants and 24 items. The ﬁnal model included by-participant and by-item random intercepts, and by-item random slopes for the eﬀect of density. The two-way interaction between time and density was pruned from the model with no reduction in model ﬁt χ = 0.10, p = .95. Unfortunately it was not possible to replace these par- Semantic categorisation ticipants due to timing constraints, and our main Accuracy hypotheses related to the explicit recall measures for Accuracy was generally very low (M = .57, SD = .50), and this experiment. We removed any participants who did did not change across the course of the week (ps > .35). not have data from at least two of the three sessions, There was also no signiﬁcant eﬀect of neighbourhood leaving 65 participants for these analyses. density (p = .508; Table A2-3). Recognition of the new word-forms was much higher than participants’ ability to recall them. Performance was lowest at the ﬁrst test point (M = 0.91, SD = 0.29; delay1: Reaction time β = 0.11, SE = 0.04, Z = 2.59, p = .010), but there were no At this low level of performance, 16 participants were further changes in performance between the day (M excluded from RT analyses on the basis of chance-level = .94, SD = .24) and week (M = .93, SD = .26; p = .578) performance (note that this exclusion was not tests. There was a small but signiﬁcant eﬀect of neigh- speciﬁed in the pre-registration due to an oversight). bourhood density (β = 0.21, SE = 0.11, Z = 1.97, p This left 49 participants in the analysis, who ranged = .049): performance was slightly higher for high- from .51-.74 in categorisation accuracy (M = .60). Only density items (M = .93, SD = .26) than low-density (M 44.98% of the data was retained after data trimming = .92, SD = .27). However, there was no evidence of an (as above), and so caution is needed in interpreting interaction with test session (pruned from ﬁnal model; these data. Modelling was carried out on the log-trans- p = .951; Table 4). formed data, and showed only a decrease in reaction time across test sessions: participants were slowest at the ﬁrst test (M = 1202 ms, SD = 495 ms; delay1: β = Cued meaning recall −0.08, SE = 0.01, t = −6.45, p < .001), and continued to Participants scored an average of 0.39 (SD = 0.78) points improve between the day (M = 1017 ms, SD = 430 ms) per item at the ﬁrst test, which did not change over time and week (M = 911 ms, SD = 403 ms) memory tests (ps > .7; Figure 2D; Table 5). Whilst this level of perform- (delay2: β = −0.05, SE = 0.02, t = −2.71, p = .010). There ance was highly comparable to Experiment 1 (M= .36, was no eﬀect of neighbourhood density (p = .344; SD = .76), recall of meanings was not aﬀected by the Table A2-4). semantic neighbourhood density of the concepts in adult participants (p = .704). Experiment 2 discussion In Experiment 2, we examined whether semantic neigh- Table 5. Final analysis model for Experiment 2: Cued meaning bourhood density inﬂuences adults’ word learning. To recall. draw comparisons with children in Experiment 1, we β coeﬃcient SE Z p used more items and fewer exposures to the new Fixed eﬀects stimuli to create a similar level of task diﬃculty 0|1 1.86 0.22 8.63 <.001 1|2 1.98 0.22 9.18 <.001 between the two groups. Adults recalled a comparable delay1 −0.01 0.03 −0.33 .738 proportion of the stimuli across the diﬀerent tasks to delay2 −0.02 0.05 −0.29 .773 density −0.07 0.19 −0.38 .704 children, but note that the overall information learned Note: Analysis based on from 4536 observations across 66 participants and was still higher for adults as they were provided with 24 items. The ﬁnal model included by-participant and by-item intercepts. more items (24 vs. 16). Consistent with the results of The two-way interaction between time and density was pruned from the model with no reduction in model ﬁt χ = 0.01, p = .995. Experiment 1, memory for new word-forms improved 12 E. JAMES ET AL. over repeated tests distributed across the week, whereas Z = −2.22, p = .027), showing a beneﬁt for recalling deﬁnition knowledge remained stable. meanings associated with low-density semantic neigh- Like children, adults were inﬂuenced by semantic bourhoods. There was no eﬀect of age group, alone or neighbourhood density in recalling the new items. in interaction with any other variable. However, while children had been inﬂuenced by the semantic manipulation in recalling the word meanings, General discussion adults showed this eﬀect in recalling the word-forms— despite no explicit demands on accessing semantic The two experiments showed that semantic prior knowl- knowledge in these tasks. For word-form recall, the edge inﬂuences new word learning. We used an object- eﬀects of semantic neighbourhood density were based metric to test the hypothesis that more shared similar in direction to those observed Experiment 1, features could facilitate learning, in contrast to previous demonstrating a disadvantage for high-density items. studies that found semantic interference from language- However, this was also accompanied by a small beneﬁt based measures of relatedness. However, this was not for recognising high-density items. In contrast to chil- the case: both children and adults showed interference dren, adults were not inﬂuenced by semantic density from dense feature-based neighbourhoods in recalling in recalling the novel meanings. the new items. This semantic interference emerged for the task drawing upon meaning knowledge for children and form knowledge in adults, although cross-exper- Additional exploratory analyses iment analyses indicated that these task diﬀerences The results suggest that both children and adults experi- may not be robust. In the following discussion, we enced interference from high-density semantic neigh- focus ﬁrst on the nature of semantic inﬂuences during bourhoods, but there were group diﬀerences in how word learning, before considering possible developmen- this interference manifested in the diﬀerent measures tal diﬀerences and implications for word learning more of word learning. We conducted additional exploratory broadly. analyses to assess whether these patterns of perform- ance for explicit recall of word-forms and meanings The inﬂuence of semantic neighbours during were statistically diﬀerent between the two word learning experiments. For each of the two recall measures, we ﬁtted a mixed The results are consistent with the few previous studies eﬀects model with ﬁxed eﬀects of session, density, and that have manipulated the availability of semantic group (children versus adults), with all interaction neighbours during word learning, ﬁnding that memory terms. Random eﬀects were speciﬁed as above, and for new words can be hindered by links to denser the full model tables can be found in Appendix 2 semantic neighbourhoods (Storkel & Adlof, 2009; Tam- (Tables A2-5 and A2-6). For the word-form recall task, minen et al., 2013). Our ﬁndings add two key contri- there were main eﬀects of test session (delay1: β = butions to the literature here: ﬁrst, that school-aged 0.68, SE = 0.03, Z = 25.07, p < .001; delay2: β = 0.50, SE = children are similarly aﬀected by semantic neighbours 0.04, Z = 12.10, p < .001), reﬂecting the improvements as preschool children and adults; and second, that seen across the week in each experiment. There was feature-based conceptualisations of semantic neigh- also a main eﬀect of group (β = −0.58, SE = 0.16, Z = bourhoods inﬂuence learning as well as associative −3.55, p < .001), with adults performing worse than chil- (language-based) metrics. Counter to predictions that dren, but this was in the context of signiﬁcant inter- semantic inﬂuences would emerge at delayed tests— actions with test session (group*delay1: β = −0.29, SE = following increased opportunities for the new represen- 0.03, Z = −10.98, p < .001; group*delay2: β = −0.43, SE = tations to integrate with existing knowledge (Davis & 0.04, Z = −10.44, p < .001). Pairwise contrasts for each Gaskell, 2009; McClelland et al., 1995)—the eﬀects test session showed that the two groups did not diﬀer emerged immediately after training for tasks assessing at the ﬁrst test point (p = .942), but that children increas- explicit knowledge of the new forms and/or meanings, ingly outperformed adults in their likelihood of recalling and did not change across the week. This early the word-forms at T2 (β = −0.88, SE = 0.33, Z = −2.62, p inﬂuence of existing knowledge may be due to the = .009) and T3 (β = −2.60, SE = 0.34, Z = −7.60, p < .001). nature of the training task, given that related concepts There was no eﬀect of density, alone or in interaction were explicitly incorporated during encoding (i.e. the with any other variable. base concepts were named in the deﬁnitions, and par- For the meaning recall task, only the main eﬀect of ticipants used their prior knowledge of these concepts density was statistically signiﬁcant (β = −0.27, SE = 0.12, to draw the items). A key question for future studies is LANGUAGE, COGNITION AND NEUROSCIENCE 13 thus whether the time course of semantic inﬂuence derived, and the extent to which they capture relevant would diﬀer if these similarities were not made explicit semantic relationships for supporting learning. Feature during encoding, requiring learners to infer similarity norms are created by asking participants to list features with known concepts using images or feature descrip- of diﬀerent concepts, but these reports are biased tions alone. towards salient and distinctive features; participants Why then do dense semantic networks lead to poorer are less likely to report the ordinary features that they memory performance for new concepts in this context? share with many other concepts. Thus, the metrics we One possible explanation is that the co-activation of used to deﬁne semantic neighbourhood density may multiple related concepts during encoding leads to not capture the vast array of highly familiar features competition or interference in processing. Recent known and shared for certain concepts, which may be ﬁndings from the lexical processing literature indicate beneﬁcial when learning new related concepts. To that the high-density disadvantage may relate to the explore this possibility further, we conducted some speciﬁc metrics we chose when designing our stimuli. additional analyses (Appendix 3) using an alternative We selected base concepts that varied in both semantic metric from the McRae et al. (2005) feature norms as a richness (the number of reported features) and density predictor of performance across experiments: the pro- (the co-occurrence of those features in other concepts), portion of distinctive features (i.e. the proportion of based on early evidence that shared features facilitate the base concept’s listed features that were not listed lexical processing (Grondin et al., 2009). Since then, for other concepts). This metric was not signiﬁcantly cor- several studies have demonstrated that our selected related to either the number of features reported or the measures might have opposing inﬂuences, particularly percentage of correlated features used in initial selection in studies of word production (Hameau et al., 2019; of the stimuli, suggesting that it captures a diﬀerent Lampe et al., 2022; Rabovsky et al., 2016). According to semantic dimension. The results showed that items Lampe et al. (2022), an abundance of semantic features with high feature distinctiveness were slightly harder leads to stronger lexical activation that supports faster to recall (66%) than items with fewer distinctive features and more accurate responses during picture naming (71%), suggesting that atypicality may hinder concept tasks, whereas high intercorrelational density more memory. Thus, there may be a beneﬁt for semantic strongly activates related concepts that cause interfer- prior knowledge in word learning that was not well-cap- ence. Applied to the present ﬁndings, the co-activation tured by our design. of a large number of related concepts when learning It is not possible to dissociate between these possible the high-density items may have led to interference in theoretical explanations with the present results, but establishing the new semantic representation and/or they highlight two important aspects to consider for when performing the recall tasks, with further research future studies: ﬁrst, that multiple semantic dimensions required to pinpoint the locus of this eﬀect. should be examined simultaneously to understand An alternative (not mutually exclusive) possibility is their inﬂuences on learning (similar to recent analyses that a dense network of co-occurring features makes it for word recognition, e.g. Lampe et al., 2022); and more challenging to integrate the highly distinctive second, the need to distinguish between the availability feature that made each concept novel (i.e. the new of existing knowledge (here, the base concept) and the concept is relatively more atypical of existing knowl- ease at which new information can be incorporated edge). Framed in this way, our stimuli perhaps more into existing networks (the novel feature). Speaking to closely align with computational models of learning aty- this distinction, studies that do not require the inte- pical category exemplars (McClelland et al., 2020): across gration of new semantic information ﬁnd the opposite both conditions, learners had the same amount of new pattern of results to those presented here: pseudowords information—a single feature—to integrate with exist- are more readily remembered when paired with existing ing knowledge. However, when the known concept concepts from higher density semantic networks (Mak & comes from a dense neighbourhood, this additional Twitchell, 2020). Thus, both the availability of semantic feature can be considered more atypical of existing con- knowledge and the need to integrate new information cepts. Thus, it becomes more challenging to integrate are key considerations in understanding how prior this novel information than when there are fewer knowledge can inﬂuence new word learning. related concepts, requiring more extensive opportu- Finally, it is important to consider that the initial chal- nities for learning and reactivation than were oﬀered lenge of learning concepts from high-density neigh- by the present study. bourhoods may yet translate to longer-term A third possibility for the density disadvantage relates processing beneﬁts over time and further exposures. to the way in which the feature norms themselves were Although we originally set out to examine semantic 14 E. JAMES ET AL. integration with the semantic categorisation tasks, we aﬀects the resources available to encode or retrieve later reduced the number of stimulus exposures in the associated word-forms, given that this activation is adults’ learning phase to aid in interpreting diﬀerences experience-dependent (Pexman, 2020). On the converse, in semantic inﬂuences between children and adults strengths in explicit learning may mean that adults can when performing at a similar level of diﬃculty. Thus, overcome semantic competition in the deﬁnitions task. perhaps unsurprisingly, we did not see any inﬂuence These possibilities warrant further investigation in of semantic neighbourhood density on speeded proces- studies designed and powered to examine developmen- sing as we initially had predicted (and as was observed tal diﬀerences. by Tamminen et al., 2013). We consider that the present results thus reﬂect a relatively early stage of Encoding and consolidation processes in word new word knowledge, and that the new word meanings learning were not well-consolidated into vocabulary during the course of the experiment. With additional learning Moving beyond the inﬂuence of semantic neighbours, opportunities, dense semantic neighbourhoods may the results are consistent with previous studies demon- yet provide a beneﬁcial role in new word knowledge. strating improvements in word-form knowledge with In line with this possibility, Mak et al. (2021) demon- repeated tests across a week period (e.g. Henderson strated that words encountered across semantically et al., 2013; James et al., 2019; Storkel, 2001; Tamminen diverse texts are more poorly learned than those et al., 2010). These improvements are consistent with the encountered in a single semantic context in the ﬁrst hypothesis that new word knowledge is strengthened instance, but that diversity comes to beneﬁt word “oﬄine”, via opportunities for hippocampal reactivation knowledge after an initial period of stabilisation has during sleep (Davis & Gaskell, 2009; McClelland et al., occurred. Thus, with more opportunities for training 1995), and/or that performance improves with repeated over a longer period of time, the disadvantages seen retrieval practice (Goossens et al., 2014). A ﬁrst notable for learning high-density concepts in the present study ﬁnding here is that participants demonstrated clear may translate to a longer-term processing beneﬁt for gains in word-form knowledge but not meaning knowl- the new words, in line with the lexical processing advan- edge, despite comparably low initial performance in tage observed for the base concepts themselves. these tasks. A likely key factor in this diﬀerence in gains is that participants were re-exposed to the word- forms at each test point (i.e. in the form recognition Developmental diﬀerences in semantic inﬂuences task, and as a cue for the deﬁnition recall task), We tested children and adults using the same exper- whereas there was no further re-exposure to the novel imental paradigm, providing an insight into the meanings. With further opportunities for (re-)encoding inﬂuence of semantic knowledge on word learning the word-forms, it is perhaps no surprise that such across development. It was clear that both groups impressive gains were seen across the course of the accessed related semantic knowledge during the exper- week. iment, marked by superior recall of items associated with However, we also consider the possibility that diﬀer- low- versus high-density semantic neighbourhoods. Yet ences in gains are additionally inﬂuenced by the there was some indication of a developmental diﬀerence extent to which diﬀerent aspects of new word knowl- in how these neighbourhood eﬀects manifest in task edge can build on existing representations, in line with performance. In Experiment 1, children were inﬂuenced a complementary learning systems perspective. That is, by neighbourhood density only in their recall of mean- this discrepancy is similarly observed in studies that ings and not word-forms. For adults in Experiment 2 equate opportunities to re-encode form and meaning however (and non-signiﬁcantly in Experiment S1), neigh- aspects of new word knowledge, suggesting that bourhood density eﬀects were most apparent in the repeated exposures may not be the only contributing form recall measure—despite no requirement for factor. For example, the repeated tests in James et al. semantic knowledge to be retrieved for task success. (2020a) incorporated a single re-exposure of the word- These task diﬀerences were not anticipated and may form (as a cue for meaning recall) and training image be spurious—indeed, it is important to stress that the (as a cue for a picture naming task) at three test points cross-experiment analyses did not ﬁnd clear evidence over 24 h. While word-form recall improved after sleep, of developmental diﬀerences in neighbourhood deﬁnition recall remained stable across periods of eﬀects. However, we can consider that perhaps only wake and sleep despite repeated opportunities to re- the mature lexical-semantic system activates semantic encode semantic features. In the context of the knowledge so automatically during learning that it present experiment, the word-forms represented a LANGUAGE, COGNITION AND NEUROSCIENCE 15 relatively arbitrary combination of sounds, proposed to that multiple mechanisms may contribute to enhanced be dependent on the hippocampal system at encoding vocabulary consolidation during this period. Indeed, it and thus most reliant on reactivation to support neocor- could be that children beneﬁt more than adults from tical consolidation (James et al., 2019). On the converse, the retrieval practice or re-encoding opportunities at the word meanings created for this experiment were each test point. A valuable next step here will be to novel variants of known concepts, directly building on examine whether developmental diﬀerences remain existing representations in both semantic neighbour- when the test delay is a between-subjects manipulation, hood conditions. Recent computational studies have with diﬀerent groups completing only a single test conceptualised the neocortical system as being prior either the same day, the next day, and one week later. knowledge-dependent (Kumaran et al., 2016; McClelland, Understanding these developmental diﬀerences in 2013; McClelland et al., 2020), indicating that neocortical longer-term memory processes, and whether we can learning can occur rapidly in the context of existing capitalise upon them to support vocabulary develop- knowledge without requiring hippocampal reactivation. ment, presents an exciting avenue for future research. Thus, while semantic aspects of new word knowledge may sometimes beneﬁt from oﬄine consolidation (e.g. McGregor et al., 2013), the scope for capitalising upon Conclusions existing semantic connections may render this eﬀect It is well-established that prior knowledge aﬀects new less robust across studies than the beneﬁts observed learning. This study built upon previous studies of pho- for word-form memory. Whether this explanation nological knowledge in vocabulary learning to show that would hold beyond the confound of repeated word- semantic neighbours also aﬀect the acquisition of new form tests within the present design remains an open words. Further, these semantic inﬂuences can be cap- question, which could be better addressed by using a tured by object-based metrics, as well as the associative between-subjects manipulation of test delay. semantic dimensions used in previous studies (Storkel & The second notable ﬁnding is that children showed Adlof, 2009; Tamminen et al., 2013). We found that by greater improvements in word-form knowledge across training pseudowords and associated novel concepts the week than adults. We emphasise caution in inter- with close semantic neighbours (i.e. diﬀering in a preting this result: ﬁrst because it was the result of an single feature), children and adults found it harder to exploratory analysis, and second because of some meth- learn and/or remember new words associated with odological diﬀerences between the two experiments (i.e. dense feature neighbourhoods. Given that these the spoken versus written modality of the recall test). ﬁndings somewhat contradict the semantic density However, superior long-term retention of word-forms beneﬁts observed in studies of known words, we in children has been demonstrated across several pre- propose that time and/or experience, semantic dimen- vious studies (James et al., 2019; Smalle et al., 2018), sion, and semantic distance should each be thoroughly including those using identical training and test para- examined to understand the role that prior semantic digms across the two groups (James et al., 2020b). A knowledge plays in vocabulary acquisition. valuable contribution of the present study is that interpretation of these group diﬀerences is often con- founded by adults’ relative strength in initial encoding: Notes do children beneﬁt more by delayed tests because of developmental diﬀerences in memory processes, or is 1. Experiment S1 used original items selected from the it simply that there is more scope for improvement English Lexicon Project. Due to experimenter error, when initial learning is weak? We found here that chil- slight variants of four of the items were used in Exper- iment 1 (i.e., attie, bryat, shamal, vorgol instead of dren continue to show delayed beneﬁts relative to attay, bryet, shimal, vorgal). However, we recomputed adults even when matched for initial diﬃculty in the orthographic neighbourhood density and bigram fre- ﬁrst test session. This beneﬁt is in line with evidence quency based on the new set to conﬁrm that these suggesting an enhanced role for sleep in children’s did not diﬀer between word lists, and retained the memory consolidation (Peiﬀer et al., 2020; Wilhelm amended version for both Experiment 1 and 2 here to facilitate developmental comparisons. et al., 2012; Wilhelm et al., 2013), linked to a higher pro- 2. Only 18 of the 24 items were also entries in the Florida portion of the slow neural oscillations that are associated Free Association Norms. These indicated that the two with memory consolidation processes. However, other sets would likely diﬀer in semantic neighbourhood studies have also found superior memory retention in density by this measure, with high-density concepts children across shorter periods that do not contain having more associates (M = 17.33) than low-density sleep (Bishop et al., 2012; Smalle et al., 2018), suggesting concepts (M = 12.22; p = .05). 16 E. JAMES ET AL. 3. Note that the pre-registration refers to a signiﬁcant Bates, D., Kliegl, R., Vasishth, S., & Baayen, H. (2015a). eﬀect of semantic neighbourhood density for cued Parsimonious mixed models. arXiv Preprint ArXiv, 1506, form recall in an initial adult experiment (Experiment 04967. https://doi.org/10.48550/arXiv.1506.04967 S1). This was due to an error in which test session was Bates, D., Maechler, M., Bolker, B., & Walker, S. (2015b). Fitting entered into analyses as a continuous rather than categ- linear mixed-eﬀects models using lme4. Journal of orical predictor. Statistical Software, 67(1), 1–48. https://doi.org/https://doi. 4. One participant did not complete 2/3 deﬁnitions tests, org/10.18637/jss.v067.i01 and was excluded from this analysis. Bishop,D.V., Barry,J.G., &Hardiman, M. J. (2012). Delayed retention of new word-forms is better in children than adults regardless of language ability: A factorial two-way study. PloS one, 7(5), e37326. https://doi.org/10.1371/journal.pone. Acknowledgment An earlier version of this manuscript was published in the ﬁrst Borovsky, A., Ellis, E. M., Evans, J. L., & Elman, J. L. (2016). Lexical author’s PhD thesis, and we would like to thank Professor leverage: Category knowledge boosts real-time novel word Dorothy Bishop for suggesting further exploratory analyses. recognition in 2-year-olds. Developmental Science, 19(6), 918–932. https://doi.org/10.1111/desc.12343 Brysbaert, M., Warriner, A. B., & Kuperman, V. (2014). Concreteness ratings for 40 thousand generally known Funding English word lemmas. Behavior Research Methods, 46(3), E.J was supported by an Economic and Social Research Council 904–911. https://doi.org/10.3758/s13428-013-0403-5 1 + 3 studentship in conducting this study. The research was Buchanan, L., Westbury, C., & Burgess, C. (2001). Characterizing additionally supported ESRC grant ES/N009924/1 awarded to semantic space: Neighborhood eﬀects in word recognition. L.M.H. and M.G.G. Psychonomic Bulletin & Review, 8(3), 531–544. https://doi. org/10.3758/BF03196189 Christensen, R. H. B. (2015). Ordinal-Regression models for Data availability statement ordinal data. R Package Version 2015.6-28. http://www. cran.r-project.org/package=ordinal. All materials, data, and analyses are available at https://osf.io/ Clay, F., Bowers, J. S., Davis, C. J., & Hanley, D. A. (2007). 35ftn/. Teaching adults new words: The role of practice and conso- lidation. Journal of Experimental Psychology: Learning, Memory, and Cognition, 33(5), 970. https://doi.org/10.1037/ Disclosure statement 0278-7393.33.5.970 Coltheart, M. (1981). The MRC psycholinguistic database. The No potential conﬂict of interest was reported by the author(s). Quarterly Journal of Experimental Psychology Section A, 33 (4), 497–505. https://doi.org/10.1080/14640748108400805 Davies, R. A., Arnell, R., Birchenough, J. M., Grimmond, D., & ORCID Houlson, S. (2017). Reading through the life span: Individual diﬀerences in psycholinguistic eﬀects. Journal of Emma James http://orcid.org/0000-0002-5214-0035 Experimental Psychology: Learning, Memory, and Cognition, 43(8), 1298. https://doi.org/10.1037/xlm0000366 Davis, M. H., & Gaskell, M. G. (2009). A complementary systems References account of word learning: Neural and behavioural evidence. Philosophical Transactions of the Royal Society B: Biological Antony, J. W., Ferreira, C. S., Norman, K. A., & Wimber, M. (2017). Sciences, 364(1536), 3773–3800. https://doi.org/10.1098/ Retrieval as a fast route to memory consolidation. Trends in rstb.2009.0111 Cognitive Sciences, 21(8), 573–576. https://doi.org/10.1016/j. Dumay, N., & Gaskell, M. G. (2007). Sleep-associated changes in tics.2017.05.001 the mental representation of spoken words. Psychological Ashworth, A., Hill, C. M., Karmiloﬀ-Smith, A., & Dimitriou, D. Science, 18(1), 35–39. https://doi.org/10.1111/j.1467-9280. (2014). Sleep enhances memory consolidation in children. 2007.01845.x Journal of Sleep Research, 23(3), 304–310. https://doi.org/ Engelthaler, T., & Hills, T. T. (2017). Feature biases in early word 10.1111/jsr.12119 learning: Network distinctiveness predicts age of acquisition. Baayen, R. H., Piepenbrock, R., & Gulikers, L. (1995). The CELEX Cognitive Science, 41,120–140. https://doi.org/10.1111/cogs. lexical database (release 2). Distributed by the Linguistic Data Consortium, University of Pennsylvania. Forster, K. I., & Forster, J. C. (2003). Dmdx: A windows display Balota, D. A., Yap, M. J., Hutchison, K. A., Cortese, M. J., Kessler, program with millisecond accuracy. Behavior Research B., Loftis, B., Neely, J. H., Nelson, D. L., Simpson, G. B., & Methods, Instruments, & Computers, 35(1), 116–124. https:// Treiman, R. (2007). The English lexicon project. Behavior doi.org/10.3758/BF03195503 Research Methods, 39(3), 445–459. https://doi.org/10.3758/ Gais, S., Lucas, B., & Born, J. (2006). Sleep after learning aids BF03193014 memory recall. Learning & Memory, 13(3), 259–262. https:// Barr, D. J., Levy, R., Scheepers, C., & Tily, H. J. (2013). Random doi.org/10.1101/lm.132106 eﬀects structure for conﬁrmatory hypothesis testing: Keep Gaskell, M. G., & Marslen-Wilson, W. D. (1997). Integrating form it maximal. Journal of Memory and Language, 68(3), 255– and meaning: A distributed model of speech perception. 278. https://doi.org/10.1016/j.jml.2012.11.001 LANGUAGE, COGNITION AND NEUROSCIENCE 17 Language and Cognitive Processes, 12(5-6), 613–656. https:// Behavior Research Methods, 44(4), 978–990. https://doi.org/ doi.org/10.1080/016909697386646 10.3758/s13428-012-0210-4 Goossens, N. A., Camp, G., Verkoeijen, P. P., & Tabbers, H. K. Kuznetsova, A., Brockhoﬀ, P. B., & Christensen, R. H. (2017). (2014). The eﬀect of retrieval practice in primary school lmerTest package: Tests in linear mixed eﬀects models. vocabulary learning. Applied Cognitive Psychology, 28(1), Journal of Statistical Software, 82(13), 1–26. https://doi.org/ 135–142. https://doi.org/10.1002/acp.2956 10.18637/jss.v082.i13 Grondin, R., Lupker, S. J., & McRae, K. (2009). Shared features Lampe, L. F., Hameau, S., & Nickels, L. (2022). Semantic variables dominate semantic richness eﬀects for concrete concepts. both help and hinder word production: Behavioral evidence Journal of Memory and Language, 60(1), 1–19. https://doi. from picture naming. Journal of Experimental Psychology: org/10.1016/j.jml.2008.09.001 Learning, Memory, and Cognition, 48(1), 72–97. https://doi. Hameau, S., Nickels, L., & Biedermann, B. (2019). Eﬀects of org/10.1037/xlm0001050 semantic neighbourhood density on spoken word pro- Mak, M. H., Hsiao, Y., & Nation, K. (2021). Anchoring and contex- duction. Quarterly Journal of Experimental Psychology, 72 tual variation in the early stages of incidental word learning (12), 2752–2775. https://doi.org/10.1177/ during reading. Journal of Memory and Language, 118, 1747021819859850 104203. https://doi.org/10.1016/j.jml.2020.104203 Henderson, L. M., Weighall, A., & Gaskell, G. (2013). Learning Mak, M. H., & Twitchell, H. (2020). Evidence for preferential new vocabulary during childhood: Eﬀects of semantic train- attachment: Words that are more well connected in seman- ing on lexical consolidation and integration. Journal of tic networks are better at acquiring new links in paired- Experimental Child Psychology, 116(3), 572–592. https://doi. associate learning. Psychonomic Bulletin & Review, 27(5), org/10.1016/j.jecp.2013.07.004 1059–1069. https://doi.org/10.3758/s13423-020-01773-0 Henderson, L. M., Weighall, A. R., Brown, H., & Gaskell, M. G. Marian, V., Bartolotti, J., Chabal, S., & Shook, A. (2012). (2012). Consolidation of vocabulary is associated with Clearpond: Cross-linguistic easy-access resource for phono- sleep in children. Developmental Science, 15(5), 674–687. logical and orthographic neighborhood densities. PloS one, https://doi.org/10.1111/j.1467-7687.2012.01172.x 7(8), e43230. https://doi.org/10.1371/journal.pone.0043230 Hills, T. T., Maouene, M., Maouene, J., Sheya, A., & Smith, L. McClelland, J. L. (2013). Incorporating rapid neocortical learn- (2009). Longitudinal analysis of early semantic networks. ing of new schema-consistent information into complemen- Psychological Science, 20(6), 729–739. https://doi.org/10. tary learning systems theory. Journal of Experimental 1111/j.1467-9280.2009.02365.x Psychology: General, 142(4), 1190. https://doi.org/10.1037/ Hoover, J. R., Storkel, H. L., & Hogan, T. P. (2010). A cross-sec- a0033812 tional comparison of the eﬀects of phonotactic probability McClelland, J. L., McNaughton, B. L., & Lampinen, A. K. (2020). and neighborhood density on word learning by preschool Integration of new information in memory: New insights children. Journal of Memory and Language, 63(1), 100–116. from a complementary learning systems perspective. https://doi.org/10.1016/j.jml.2010.02.003 Philosophical Transactions of the Royal Society B: Biological Hulme, R. C., & Rodd, J. M. (2021). Learning new word meanings Sciences, 375(1799), 20190637. https://doi.org/10.1098/rstb. from story reading: The beneﬁt of immediate testing. PeerJ, 2019.0637 9, e11693. https://doi.org/10.7717/peerj.11693 McClelland, J. L., McNaughton, B. L., & O’Reilly, R. C. (1995). Why James, E., Gaskell, M. G., & Henderson, L. M. (2019). Oﬄine con- there are complementary learning systems in the hippo- solidation supersedes prior knowledge beneﬁts in children’s campus and neocortex: Insights from the successes and fail- (but not adults’) word learning. Developmental Science, 22(3), ures of connectionist models of learning and memory. e12776. https://doi.org/10.1111/desc.12776 Psychological Review, 102(3), 419. https://doi.org/10.1037/ James, E., Gaskell, M. G., & Henderson, L. M. (2020a). Sleep- 0033-295X.102.3.419 dependent consolidation in children with comprehension McGregor, K. K., Licandro, U., Arenas, R., Eden, N., Stiles, D., and vocabulary weaknesses: It’ll be alright on the night? Bean, A., & Walker, E. (2013). Why words are hard for Journal of Child Psychology and Psychiatry, 61(10), 1104– adults with developmental language impairments. Journal 1115. https://doi.org/10.1111/jcpp.13253 of Speech, Language, and Hearing Research, 56(6), 1845– James, E., Gaskell, M. G., Pearce, R., Korell, C., Dean, C., & 1856. https://doi.org/10.1044/1092-4388(2013/12-0233) Henderson, L.-M. (2020b). The role of prior lexical knowl- McRae, K., Cree, G. S., Seidenberg, M. S., & McNorgan, C. (2005). edge in children’s and adults’ incidental word learning Semantic feature production norms for a large set of living from illustrated stories. Journal of Experimental Psychology: and nonliving things. Behavior Research Methods, 37(4), Learning, Memory, and Cognition, 47(11), 1856– 547–559. https://doi.org/10.3758/BF03192726 1869. https://doi.org/10.31234/osf.io/vm5ad Mirman, D., & Magnuson, J. S. (2008). Attractor dynamics and James, E., Gaskell, M. G., Weighall, A., & Henderson, L. (2017). semantic neighborhood density: Processing is slowed by Consolidation of vocabulary during sleep: The rich get near neighbors and speeded by distant neighbors. Journal richer? Neuroscience & Biobehavioral Reviews, 77,1–13. of Experimental Psychology: Learning, Memory, and Cognition, https://doi.org/10.1016/j.neubiorev.2017.01.054 34(1), 65. https://doi.org/10.1037/0278-7393.34.1.65 Kumaran, D., Hassabis, D., & McClelland, J. L. (2016). What learn- Nelson, D. L., McEvoy, C. L., & Schreiber, T. A. (2004). The ing systems do intelligent agents need? Complementary University of south ﬂorida free association, rhyme, and learning systems theory updated. Trends in Cognitive word fragment norms. Behavior Research Methods, Sciences, 20(7), 512–534. https://doi.org/10.1016/j.tics.2016. Instruments, & Computers, 36(3), 402–407. https://doi.org/ 05.004 10.3758/BF03195588 Kuperman, V., Stadthagen-Gonzalez, H., & Brysbaert, M. (2012). Peiﬀer, A., Brichet, M., De Tiège, X., Peigneux, P., & Urbain, C. Age-of-acquisition ratings for 30,000 English words. (2020). Re-epithelialization and immune cell behaviour in 18 E. JAMES ET AL. an ex vivo human skin model. Scientiﬁc Reports, 10(1), 1–9. Language, and Hearing Research, 49(6), 1175–1192. https:// https://doi.org/10.1038/s41598-019-56847-4 doi.org/10.1044/1092-4388(2006/085) Perry, L. K., Axelsson, E. L., & Horst, J. S. (2016). Learning what to Storkel, H. L., Bontempo, D. E., Aschenbrenner, A. J., Maekawa, remember: Vocabulary knowledge and children’s memory for J., & Lee, S.-Y. (2013). The eﬀect of incremental changes in object names and features. Infant and Child Development, 25 phonotactic probability and neighborhood density on (4), 247–258. https://doi.org/10.1002/icd.1933 word learning by preschool children. Journal of Speech, Pexman, P. M. (2020). How does meaning come to mind? Four Language, and Hearing Research, 56(5), 1689–1700. https:// broad principles of semantic processing. Canadian Journal doi.org/10.1044/1092-4388(2013/12-0245) of Experimental Psychology/Revue Canadienne de Tamminen, J., Davis, M. H., Merkx, M., & Rastle, K. (2012). The Psychologie Expérimentale, 74(4), 275–283. https://doi.org/ role of memory consolidation in generalisation of new lin- 10.1037/cep0000235 guistic information. Cognition, 125(1), 107–112. https://doi. Pexman, P. M., Hargreaves, I. S., Siakaluk, P. D., Bodner, G. E., & org/10.1016/j.cognition.2012.06.014 Pope, J. (2008). There are many ways to be rich: Eﬀects of Tamminen, J., & Gaskell, M. G. (2013). Novel word integration in three measures of semantic richness on visual word recog- the mental lexicon: Evidence from unmasked and masked nition. Psychonomic Bulletin & Review, 15(1), 161–167. semantic priming. Quarterly Journal of Experimental https://doi.org/10.3758/PBR.15.1.161 Psychology, 66(5), 1001–1025. https://doi.org/10.1080/ Pexman, P. M., & Yap, M. J. (2018). Individual diﬀerences in 17470218.2012.724694 semantic processing: Insights from the calgary semantic Tamminen, J., Lambon Ralph, M. A., & Lewis, P. A. (2013). The decision project. Journal of Experimental Psychology: role of sleep spindles and slow-wave activity in integrating Learning, Memory, and Cognition, 44(7), 1091. https://doi. new information in semantic memory. The Journal of org/10.1037/xlm0000499 Neuroscience, 33(39), 15376–15381. https://doi.org/10. Qualtrics (2014). Qualtrics [software]. Provo, UT, USA: Qualtrics. 1523/JNEUROSCI.5093-12.2013 Rabovsky, M., Schad, D. J., & Rahman, R. A. (2016). Language pro- Tamminen, J., Payne, J. D., Stickgold, R., Wamsley, E. J., & duction is facilitated by semantic richness but inhibited by Gaskell, M. G. (2010). Sleep spindle activity is associated semantic density: Evidence from picture naming. Cognition, with the integration of new memories and existing knowl- 146,240–244. https://doi.org/10.1016/j.cognition.2015.09.016 edge. Journal of Neuroscience, 30(43), 14356–14360. R Core Team. (2015). R: A language and environment for stat- https://doi.org/10.1523/JNEUROSCI.3028-10.2010 istical computing. R Foundation for Statistical Computing, Tham, E. K., Lindsay, S., & Gaskell, M. G. (2015). Markers of auto- Vienna, Austria. http://www.R-project.org/. maticity in sleep-associated consolidation of novel words. Rezlescu, C. (2015). Testable: The web-based platform for creat- Neuropsychologia, 71, 146–157. https://doi.org/10.1016/j. ing, running and sharing behavioural experiments. neuropsychologia.2015.03.025 Association for Psychological Science Teaching institute Tse, D., Langston, R. F., Kakeyama, M., Bethus, I., Spooner, P. A., conference, New York, 21–24 May. Wood, E. R., Witter, M. P., & Morris, R. G. (2007). Schemas and Smalle, E. H., Page, M. P., Duyck, W., Edwards, M., & Szmalec, A. memory consolidation. Science, 316(5821), 76–82. https:// (2018). Children retain implicitly learned phonological doi.org/10.1126/science.1135935 sequences better than adults: A longitudinal study. van der Kleij, S. W., Rispens, J. E., & Scheper, A. R. (2016). The Developmental Science, 21(5), e12634. https://doi.org/10. eﬀect of phonotactic probability and neighbourhood 1111/desc.12634 density on pseudoword learning in 6- and 7-year-old chil- Storkel, H. L. (2001). Learning new words. Journal of Speech, dren. First Language, 36(2), 93–108. https://doi.org/10. Language, and Hearing Research, 44(6), 1321–1337. https:// 1177/0142723715626064 doi.org/10.1044/1092-4388(2001/103) Wechsler, D. (2011). WASI-II: Wechsler Abbreviated Scale of intel- Storkel, H. L. (2009). Developmental diﬀerences in the eﬀects of ligence. Pearson. phonological, lexical and semantic variables on word learn- Wickham, H. (2009). Ggplot2: Elegant graphics for data analysis. ing by infants. Journal of Child Language, 36(02), 291–321. Springer Science & Business Media. https://doi.org/10.1017/S030500090800891X Wilhelm, I., Prehn-Kristensen, A., & Born, J. (2012). Sleep-depen- Storkel, H. L., & Adlof, S. M. (2009). The eﬀect of semantic set size dent memory consolidation – what can be learnt from chil- on word learning by preschool children. Journal of Speech, dren? Neuroscience & Biobehavioral Reviews, 36(7), 1718– Language, and Hearing Research, 52(2), 306–320. http://jslhr. 1728. https://doi.org/10.1016/j.neubiorev.2012.03.002 pubs.asha.org/data/Journals/JSLHR/929279/jslhr_52_2_306. Wilhelm, I., Rose, M., Imhof, K. I., Rasch, B., Büchel, C., & Born, J. pdf. https://doi.org/10.1044/1092-4388(2009/07-0175) (2013). The sleeping child outplays the adult’s capacity to Storkel, H. L., Armbrüster, J., & Hogan, T. P. (2006). convert implicit into explicit knowledge. Nature Diﬀerentiating phonotactic probability and neighborhood Neuroscience, 16(4), 391-393. https://doi.org/10.1038/nn. density in adult word learning. Journal of Speech, 3343

Journal

Language Cognition and Neuroscience – Taylor & Francis

Published: Feb 7, 2023

Keywords: Vocabulary; word learning; prior knowledge; semantic neighbourhood density; consolidation; development

Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Word learning in the context of semantic prior knowledge: evidence of interference from feature-based neighbours in children and adults

Word learning in the context of semantic prior knowledge: evidence of interference from feature-based neighbours in children and adults

Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Word learning in the context of semantic prior knowledge: evidence of interference from feature-based neighbours in children and adults

Word learning in the context of semantic prior knowledge: evidence of interference from feature-based neighbours in children and adults

References (74)

Abstract

Journal

Recommended Articles

There are no references for this article.

Our policy towards the use of cookies