Experiment 3 comprised an exploratory analysis of white matter correlates of word learning. When considering the immediate consequences of word learning, speeded recognition accuracy scores significantly correlated with FA in the bilateral AF and the right UF, such that children with higher speeded recognition accuracy had higher FA in these white matter tracts. These correlations are broadly in line with previous studies linking these tracts to various language measures (e.g., Catani & Thiebaut de Schotten, 2008; Mabbott et al., 2009; Romeo et al., 2018; Su et al., 2018; Urger et al., 2015; Von Der Heide et al., 2013). It should be noted that these correlations were coupled with positive correlations between speeded recognition response time and FA in the same tracts, most likely due to a speed-accuracy trade-off (i.e., children who responded more accurately also responded more slowly). Despite observing no correlations between pre-test cued recall accuracy and FA in any of the ROIs, over-nap changes in cued recall accuracy were significantly (and positively) associated with FA in the right AF. Children with higher FA in this tract had greater increases in explicit memory of the novel words following the nap. Given that a comparable analysis of the wake group was not feasible, the behavioural changes cannot be exclusively attributed to sleep. Nevertheless, it can be concluded that offline changes in memory for novel words are associated with white matter integrity of the right AF.
This study examined whether daytime naps, like periods of nocturnal sleep, benefit the consolidation of novel words, as well as their integration into the lexical network. This was investigated in two age groups: young adults and children aged 10-12 years old. To examine if children's word learning benefits more from naps than adults (similar to previously reported enhanced effects of nocturnal sleep in children e.g., see James et al., 2017 for a review), the two age groups were compared on nap sleep parameters and on change in word learning performance across the nap. Finally, an exploratory experiment conducted for the child group aimed to identify white matter correlates of word learning.
Daytime naps played a protective role in novel word Learning.
 For children and adults, a 90-minute daytime nap led to the maintenance of explicit memory of novel words, as measured via a cued recall task, whereas an equivalent wake period led to forgetting. Whilst adults were also slower at recognising novel words following wake compared to sleep, there was no effect of sleep on speeded recognition accuracy for either age group (potentially due to higher levels of initial performance leaving less room for improvement). Furthermore, neither age group demonstrated evidence of lexical integration (as indexed by lexical competition between the known and trained words) following a nap. This falls in contrast to studies examining overnight sleep and word learning in children and adults, which demonstrated sleep-associated improvements in explicit memory, as well as engagement in competition (Fletcher et al., 2020; Smith et al., 2018; Tamminen et al., 2010, 2013; Weighall et al., 2017). There were also no statistically significant correlations between over-nap changes in explicit memory or lexical integration and sleep parameters known to support systems consolidation (Born & Wilhelm, 2012; Diekelmann & Born, 2010; Rasch & Born, 2013). Notably, previous studies of both children and adults, using similar novel word learning paradigms, have found improvements in recall of novel words following a full night of sleep, evidence of lexical integration, and associations between these changes and sleep spindle parameters (Henderson et al., 2012; Smith et al., 2018; Tamminen et al., 2010). Thus, while the current results support a protective effect of daytime naps relative to wake for explicit memory in children and adults, there was no evidence that the naps played an active role in strengthening and integrating newly learned words. Possibly then, the 90- minute naps experienced here were beneficial for maintaining new memories of novel words, perhaps through guarding against forgetting that might occur during wake as a consequence of interference (Wixted, 2004), but not for supporting active systems consolidation, at least at the age groups of focus and when daytime naps are not habitual (consistent with Tamminen et al., 2017). This account explains why retention of information after a nap was still superior to wake and why the nap did not promote strengthening or integration of the new words. 
One possibility is that sleep parameters need to reach a particular threshold before active systems consolidation is observed. Consistent with this, Piosczyk et al. (2013) found that a daytime nap promoted memory consolidation of wordpairs in 16-year-old females, but only when the nap was characterised by high sigma power. Another explanation for why effects that emerge across nocturnal sleep were not seen here, is because levels of REM sleep were low. Our focus here was on ascertaining whether the previously reported correlations between changes in word learning over nocturnal sleep and SWS and spindle parameters would be observed over a nap here. However, it has also been shown that vocabulary learning might benefit from REM sleep that occurs after SWS during an afternoon nap (Batterink, Westerberg, & Paller, 2017). During nocturnal sleep, SWS and REM sleep alternate cyclically, with SWS naturally followed by REM sleep (Rasch & Born, 2013). Various theoretical models have proposed that SWS and REM play complementary roles in memory consolidation. Active systems accounts argue that whilst systems consolidation occurs during SWS (repeatedly activating newly encoded memories and driving the integration of new memories into the network of pre-existing long-term memories), REM sleep acts to stabilize the transformed memories by enabling undisturbed synaptic consolidation (Diekelmann & Born, 2010). According to the sequential hypothesis (e.g., Ambrosini & Giuditta, 2001), SWS works to eliminate non-adaptive memory traces, with subsequent REM sleep strengthening the remaining traces. Walker and Stickgold (2010) further postulate that REM sleep plays an integrative function, forming rich associative links between new and existing knowledge. These theories are supported by nap studies, for instance Mednick, Nakayama, and Stickgold (2003) found that only naps containing REM improved performance on a visual discrimination task, with naps without REM leading to maintained performance. Thus, nocturnal sleep, rich with SWS and REM sleep, might therefore lead to more opportunities for spontaneous reactivation of newly learned words. For example, a recent study demonstrated that toddlers that had both a daytime nap and night time sleep benefitted the most for generalisation of novel categories, compared to those that did not nap or were tested 4-hrs after the nap (Werchan, Kim, & Gomez, 2021). 
This raises an important point: Post-training naps may be particularly powerful for enhancing longer-term retention when they are combined with nocturnal sleep. It has been shown that when 6-7-year-olds napped soon after mirror discrimination training this led to significantly after gains in reading fluency than compared to not napping after training, even when performance was tested after nocturnal sleep (Torres et al., 2021). These results support the claim that the delay between training and sleep onset is a critical factor, particularly in younger learners (e.g., Backhaus et al., 2008; James et al., 2020; Walker et al., 2020), and crucially, that the protective effects of a nap may work in combination with nocturnal sleep, as a useful memory enhancer. 
Developmental differences in nap architecture and over-nap change in word learning.
Despite adults napping for longer than children and spending more time in N2 and REM sleep, children exhibited significantly (and proportionally) more SWS. This is consistent with previous reports of greater nocturnal SWS in children in this age range compared to adults (e.g., Wilhelm et al., 2013). Spindle density (both fast and slow) and sigma power (fast only) were also higher in adults than in children. Higher sigma power in adults aligns with previous findings demonstrating that sigma frequency increases linearly with age (Zhang et al., 2021) and plateaus in adulthood (Purcell et al., 2017). Spindle density, on the other hand, increases from childhood to adolescence, peaks at age 15 (Zhang et al., 2021), and then declines from middle to late adulthood (Purcell et al., 2017). Given that the children tested in the current study were 10-12 years old, it is likely that they had not yet reached peak spindle density, which could explain the higher spindle density found in the young adults. Overall then, the comparison between adult and child nap architecture is in line with previous studies of nocturnal sleep (Wilhelm et al., 2013; Zhang et al., 2021). 
Behaviourally, children showed a small increase in cued recall accuracy across the nap, while accuracy decreased slightly on this task in adults. This difference in over-nap change in cued recall between the two age groups was significant, albeit small numerically. Whilst children spent more time in SWS than adults, this group difference cannot be simply attributed to increased opportunities for spontaneous reactivation of the new memories during SWS in children, given that there were no correlations between over-nap change in cued recall and time in SWS or delta power in either group. It is possible that the children's levels of alertness benefited more from the nap than adults, accounting for their better post-nap performance for cued recall. However, as children were slower on the post-nap PVT than adults (but had similar major lapses),4 we think this to be an unlikely explanation. 
The enhanced protective role of naps for children relative to adults (coupled with no correlations with SWS or spindle parameters) raises the possibility that previously reported enhanced benefits of consolidation over nocturnal sleep in children might not be solely due to extra amounts of SWS. It has been claimed that reactivation of newly acquired memories can occur without sleep, during periods of wakeful rest, and memory benefits associated with wakeful rest have been attributed to the absence of interfering information (Dewar et al., 2012, 2014; Mednick, Cai, Shuman, Anagnostaras, & Wixted, 2011; Wixted, 2004). Studies have demonstrated benefits of wakeful rest for word-pair recall in children (e.g., Martini, Martini, & Sachse, 2021), and thus one potential explanation for the present results is that children may benefit more from wakeful rest than adults; however, this remains an open question since no studies to our knowledge have directly addressed whether children show greater effects of wakeful rest relative to adults. It also remains unclear whether the present findings could be explained by developmental differences in susceptibility to proactive or retroactive interference (i.e., such that children capitalised from the lack of interference more than the adults). Preliminary findings suggest that retroactive interference effects remain robust across development but that adults can be more resilient to the effects of retroactive interference than children (e.g., Darby & Sloutsky, 2013; Fatania & Mercer, 2017), however, the literature is sparse on whether there are developmental differences in proactive or retroactive interference in the context of novel word learning. For example, if children are at increased risk of interference when attempting to recall newly learned information, one possibility is that a nap allows the developing system to be “reset”, decreasing susceptibility to such interference and enhancing performance. However, this remains highly speculative. Aside from having pedagogical value, addressing these questions will be theoretically important in deepening our understanding of the various routes by which newly learned words become established in long-term language networks.
Structural brain correlates of word learning in children.
The final contribution of this study was to explore the structural neuroanatomical correlates of the word learning process, both immediately after learning and short-term changes that occurred over a nap opportunity. Faster speeded recognition of novel words immediately after learning was associated with higher fractional anisotropy in the bilateral AF and the right UF. This suggests that children with higher white matter integrity in these brain regions are also better at recognising new words immediately after training. These results align with previous research on various aspects of language use and language learning (Farah et al., 2020; Lebel, Benischek, et al., 2019; Mabbott et al., 2009; Romeo et al., 2018; Su et al., 2018), solidifying the importance of these tracts as part of the developing language learning network. Previous studies have linked the integrity of these tracts to a broad range of language measures in children (e.g., expressive language ability, Farah et al., 2020; auditory-verbal memory, Mabbott et al., 2009; conversational experience, Romeo et al., 2018; rate of vocabulary growth, Su et al., 2018); however, the present results importantly add to this existing literature by demonstrating a role for the AF and UF in recognising novel words, pointing to an underlying learning mechanism for the speeded recognition of new word forms.
Another finding of note here was although cued recall accuracy at immediate testing was not related to any of the ROIs, over-nap change in recall accuracy was positively associated with FA in the right AF. This finding supports the role of the right AF in the retention of novel words, as opposed to initial encoding processes. The fact that these correlations emerged in the right hemisphere is perhaps surprising in the context of the commonly reported left lateralization of the language network (e.g., Romeo et al., 2018; Sreedharan et al., 2015; Su et al., 2018; Urger et al., 2015). However, there have been other studies demonstrating associations with language learning in the right hemisphere (e.g., Farah et al., 2020; Romeo et al., 2018; Sreedharan et al., 2015; Su et al., 2018). Takashima et al. (2019) found that when processing newly learned words, young children (8-10 years old) had greater right lateralized activation in the inferior frontal gyrus (IFG), while older children (14-16 years old) had greater activation in the left IFG. This is in line with studies demonstrating a shift from right to left lateralization as a result of cognitive maturation, as well as increased lexical knowledge or familiarity (Sugiura et al., 2011; Szaflarski et al., 2006; Takashima et al., 2019). Possibly, as children get older and their lexical knowledge increases, a similar shift to left lateralised structural correlates of word learning would be observed.
In sum, models of word learning distinguish between processes that help us to quickly acquire new words from the environment and processes that support their long-term consolidation into existing vocabulary, suggesting that variability might emerge at multiple stages of new word acquisition (see James et al., 2017). Here, these models are expanded by identifying structural neuroanatomical correlates of the word learning process, both immediately after learning and over a nap, importantly identifying different neuroanatomical correlates at each point. Such correlates will have use for future studies which track the time course of word learning over longer time periods, including over cumulative periods of sleep. In adults, white matter integrity has been shown to be associated with offline gains in motor memory, with nocturnal sleep spindle density mediating this relationship (Mander et al., 2017; Vien et al., 2019). Therefore, the identification of white matter correlates of both the immediate consequences of word learning and changes over a nap in children informs future examinations of whether these white matter correlates dictate the functional influence of sleep spindles on the consolidation of new words.
