Dyslexia is a specific learning disability that is characterized by severe and persistent difficulties in the acquisition of word-level reading and spelling skills, in the absence of sensory or neurological deficits and inadequate teaching (Peterson and Pennington, 2012). Having (a family risk of) dyslexia has been associated with lower performance in multiple areas and skills related to literacy (e.g., reading comprehension), including vocabulary development (Vellutino et al., 2004; Snowling and Melby-Lervåg, 2016). However, vocabulary deficits are not necessarily attested in children with dyslexia (Catts et al., 2005). To further understand the relationship between dyslexia, word learning, and vocabulary, the current study reports on a word learning experiment in children with and without dyslexia. Specifically, children were asked to repeat, identify, and produce novel words of low phonological complexity, with phonological cues that were either consistent or inconsistent with grammatical class (i.e., verb or noun). These phonological cues served to investigate whether children with and without dyslexia are sensitive to phonological cues and use them to support word learning.

Vocabulary learning forms the foundation for the development of higher-order language skills, such as morphological awareness and syntactic skills (e.g., Bates and Goodman, 2001; Torppa et al., 2010), as well as word reading and reading comprehension (Perfetti et al., 2010; Hulme et al., 2015; Suggate et al., 2018; van Viersen et al., 2018). Children with dyslexia have been found to show lower vocabulary levels already from very early ages (e.g., Scarborough, 1990; Torppa et al., 2010; Duff et al., 2015; van Viersen et al., 2017). However, when looking at the developmental pathways of these children, findings point toward a delay in vocabulary growth instead of deviant growth trajectories. Lower initial growth rates in both receptive and expressive vocabulary are followed by a weaker deceleration of growth toward the end of the growth spurt between the ages of 17 and 35 months, causing children to largely close the gap (van Viersen et al., 2017). The preschool/preliteracy vocabulary delay is thus not necessarily persistent.

Differences in vocabulary levels may arise again after reading onset, for example due to less exposure to print (Kush et al., 2005; Mol and Bus, 2011; Hulme and Snowling, 2016). However, such deficits are not attested across-the-board (e.g., Cunningham and Stanovich, 1991; van Viersen et al., 2018; Adlof et al., 2021). Furthermore, some studies have even suggested that vocabulary can be a protective or mitigating factor in older populations with dyslexia (Cavalli et al., 2016; Wiseheart and Altmann, 2018; van Viersen et al., 2019). These findings do not point toward a persistent vocabulary deficit in dyslexia.

Although findings on vocabulary outcomes in dyslexia are mixed, they seem consistent in that children with dyslexia display verbal word-learning difficulties (Thomson and Goswami, 2010; Alt et al., 2017, 2019; Kimppa et al., 2018; Adlof et al., 2021; Malins et al., 2021). Word learning requires constructing phonological representations that are connected to semantic representations (Gupta and Tisdale, 2009). For children with dyslexia, constructing these detailed phonological representations might be challenging: many children with dyslexia have a deficit in phonological processing, which has been proposed as an important underlying cause of dyslexia (see Ramus et al., 2003; Vellutino et al., 2004; Protopapas and Parrila, 2019, for a recent overview). This phonological deficit refers to the difficulties children with dyslexia have in segmenting, storing, reproducing, and manipulating phonological information. Indeed, studies have found relationships between phonological processing (such as phoneme awareness) and vocabulary size (e.g., Gathercole et al., 1999; Rispens and Baker, 2012; Abel and Schuele, 2014) as well as between phonological processing (nonword repetition) and novel word learning (Gathercole et al., 1999; Adlof and Patten, 2017).

The construction of detailed phonological representations is required for storing words in long-term memory, and long-term phonological representations in turn support short-term memory performance (Gathercole and Baddeley, 1990; Gathercole et al., 1999). For instance, typically developing children and children with dyslexia recall novel words better when they have phonologically dense neighborhoods. However, children with dyslexia show poorer overall levels of serial recall and are more likely to substitute novel words with real words (Thomson et al., 2005). Thomson and colleagues attribute such lexicalization errors of children with dyslexia to poorly specified phonological representations: these poorer representations make it harder to reconstruct decaying traces of novel words. Children with dyslexia have also been found to perform more poorly on visual-verbal paired associate learning (PAL) tasks, in which children are required to learn an association between a novel spoken word and a visual referent (Vellutino et al., 1995; Mayringer and Wimmer, 2000; Messbauer and de Jong, 2003; Elbro and Jensen, 2005; Warmington and Hulme, 2012). Performance is especially poor when children are asked to produce new labels (Litt et al., 2013; Kalashnikova and Burnham, 2016; Alt et al., 2017). These findings have generally been explained by difficulties with encoding and retrieving detailed phonological representations of novel words for the purpose of production, rather than general difficulties with paired associate learning per se (Litt and Nation, 2014). Constructing or maintaining novel phonological representations also depends on words in long-term memory and related factors such as word-likeness, neighborhood density, and phonotactic probability (e.g., Storkel, 2001, 2004).

The phonological form of a word needs to be learnt in order to acquire vocabulary. The phonological form of a word can also provide cues about the grammatical structure of a word. In English, for instance, nouns tend to have more syllables than verbs, and fewer consonants per syllable. Also, bisyllabic nouns are likely to have stress on the first syllable, while bisyllabic verbs have stress on the second syllable (Kelly, 1992). Corpus analyses also show that both English and Dutch child-directed speech contain phonological cues that can lead to categorization of verbs and nouns (Monaghan et al., 2007). If a language learner is sensitive to such phonological cues, this may facilitate word learning and strengthen the link between phonological and semantic representations.

There is evidence that children are sensitive to phonological “typicality” (Cassidy and Kelly, 2001; Fitneva et al., 2009). For instance, 3- to 6-year-old children were more likely to associate monosyllabic (i.e., verb-like) novel words with actions and three-syllabic (i.e., noun-like) words with objects (Cassidy and Kelly, 2001). Importantly, children were more likely to remember words in which phonology was consistent with word class. In a similar experiment, Fitneva et al. (2009) created eight noun-like and eight verb-like pseudowords. Seven-year-old children were presented with the novel word (e.g., “skik”) and saw two pictures. They were told that the target referred either to an object (noun) or action (verb). In the consistent condition, the phonological cue agreed with the grammatical category (e.g., a noun-like nonword referring to an object), whereas in the inconsistent condition this was not the case (e.g., a noun-like nonword referring to an action). Children had to repeat the word and point to a picture. The results showed that in the initial phase of learning, children tended to choose a referent on the basis of phonological cues. In later stages of learning (after feedback), the phonological cue was found to have an effect on verb learning (i.e., children recalled verbs more accurately in the consistent condition). The study by Fitneva et al. (2009) thus indicates that phonological cues to word class can aid verb learning.

Similarly, in Dutch—the language under investigation in the current study—phonological cues have been found to distinguish between nouns and verbs: native adult speakers of Dutch were able to use phonological information to categorize nonword stems as nouns or verbs (Don and Erkelens, 2008). For instance, a bisyllabic target with final syllable containing schwa (“fallem”) was likely to be categorized as a noun, whereas a monosyllabic target with a superheavy syllable (“pluig”) was more likely to be categorized as a verb. Furthermore, on the basis of a corpus search, Don and Erkelens found that children between 2;0 and 3;6 rarely shifted words from nouns to verbs or vice versa. However, when they did shift, they changed the words to fit the phonological frame of the new category (i.e., from a noun to a verb using a phonological structure fitting a verb). In a computational study, segmental phonological cues could also successfully predict word class in Dutch (Durieux and Gillis, 2001). These findings suggest that these phonological cues may aid word learning in Dutch children.

It is yet unknown whether children with dyslexia are able to use phonological cues for word class to facilitate word learning. There are some indications that this could be problematic for this group of children. First, the phonological deficit implies the construction of less detailed phonological representations. This could hamper the implicit recognition of phonological patterns in relation to word categories. Second, pattern detection in general has been reported to be more difficult in children with dyslexia: studies on implicit and statistical learning have found that groups with dyslexia show lower outcomes than groups without dyslexia (e.g., Vicari et al., 2003; Pavlidou et al., 2010; see Lum et al., 2013, for a meta-analysis). In a recent word learning study, young adults with dyslexia were less able to use implicit knowledge of cross-situational statistics (Kligler et al., 2023). As a consequence of such an implicit learning deficit, children with dyslexia might be less sensitive to phonological or syntactic cues that could foster word learning through efficient categorization of new words.

At the same time, however, it could be that phonological pattern recognition is not problematic for children with dyslexia, as phonological output is affected more than recognition (e.g., Swan and Goswami, 1997; Ramus and Szenkovits, 2008), meaning that the patterns themselves might be recognized (implicitly). Furthermore, findings on statistical learning are far from conclusive, as relationships between statistical learning and literacy are not always attested (Schmalz et al., 2019; van Witteloostuijn et al., 2019, 2021) and not all studies report poorer statistical learning in people with dyslexia (Kelly et al., 2002; He and Tong, 2017; Staels and van den Broeck, 2017; West et al., 2021). Steeper learning curves have even been reported, suggesting more successful statistical learning for this group (Bennett et al., 2008). It is therefore an open question whether phonological cues are used for grammatical categorization to the same extent by children with and without dyslexia.

As far as we are aware, the only available evaluation of the use of cues for grammatical categorization in word learning of children with dyslexia comes from Gilliver and Byrne (2009). In three experiments, they looked into form class cues contributing to noun learning of English-speaking 4- to 5-year-olds. The reading risk status of these preschoolers was established on the basis of phonological awareness, letter recognition, the Rhyme and Final Sound test and vocabulary. In the first experiment, children had to repeat six (three-syllable) names for novel creatures and identify the correct picture in a recognition task (e.g., “Which one is schmalenork?”). There was a correlation between reading risk and recall; children with better recall had a lower risk for reading problems, but no such correlation was evident for recognition. In the second experiment, children were presented with (one/two-syllable) novel words, which could refer to a proper name (“Point to the one that is daxy”) or a category name (“Point to the one that is a daxy”). Performance on this task was not related to reading risk status, which means children at risk were equally sensitive to form class cues (i.e., signaling the difference between proper names and category names). In the final experiment, children were presented with a short story in which novel proper and count nouns were presented, leading to a final sentence (e.g., “This is a fomp called zikt”). After the story, they were asked to recall both the proper names (“Do you remember who this creature is?”) and the category name (“Do you remember what kind of creature this is?”). Children with a lower reading risk were better able to recall the targets and were better able to provide the word class.

Gilliver and Byrne (2009) take their findings to relate to the capacity of phonological and/or working memory: children at risk performed more poorly at recalling novel words, which requires more detailed phonological representations—and possibly greater processing demands—compared to recognition. Although the correlations between risk status and the learning outcomes (recall and word class accuracy) were weak-moderate only, these findings warrant further research on the ability of children with dyslexia to use cues for grammatical categorization and the type of cues they may be able to use. Furthermore, they speak to the influence of increasing task demands (e.g., production) and deficits in underlying skills that may affect the quality of phonological representations of new words.

The aim of the present study was to move beyond general aspects of literacy and phonology involved in word learning in children with and without dyslexia and assess to what extent word learning is influenced by phonological cues for grammatical class and (deficits in) phonological (literacy-related) and semantic (language-related) skills. Word learning was evaluated across different phases (repetition, identification, and naming). These different response types place different demands on the required level of detail of phonological representations of new words. The novel words were either consistent with their word class (a verb-like monosyllabic item in a verb frame, referring to an action; a noun-like bisyllabic item in a noun frame, referring to an object) or inconsistent (a noun-like item in a verb frame, referring to an action and vice versa). In the present study, the following research questions were addressed:

Given that children with dyslexia are less able to form detailed phonological representations, we hypothesized lower performance of this group, specifically on the naming (production) of novel words during word learning (RQ1). In addition, for RQ2 we explored whether phonological cues might affect word learning outcomes for children with dyslexia. Previous research has found that phonological cues affect children's categorization of words (e.g., Fitneva et al., 2009). Such cues also seem to be available in Dutch (Don and Erkelens, 2008). However, so far, only one study has addressed the possible role of phonological cues to word class in children with a risk of reading problems. These findings did not point to different use of form class cues (Gilliver and Byrne, 2009). It is therefore an open question whether such cues play a different role for children with and without dyslexia. Finally, with respect to RQ3, we expected that both literacy-related (phonological awareness, rapid automatized naming, verbal memory, nonword reading) and language-related (vocabulary and sentence repetition) skills would contribute to word learning in children with and without dyslexia, especially when more detailed phonological representations are required (i.e., in the naming phase).

In this study, we aimed to further understand the relationship between dyslexia, word learning, and vocabulary. We tested whether children with dyslexia showed similar word learning outcomes as typical readers, and whether children with dyslexia were equally sensitive to phonological cues for word class. Monosyllabic (‘verb-like') and bisyllabic (‘noun-like') novel words were either presented as verbs (e.g., “I voek”) or nouns (e.g., “a voek”), paired with a picture of an action (for verbs) or object (for nouns). The phonological structure of the targets was thus either consistent or inconsistent with word class (verb or noun). We first assessed whether differences existed between children with and without dyslexia on the repetition, identification, and/or naming of novel words in word learning (RQ1). Subsequently, we assessed to what extent the phonological cues to word class aided word learning in children with and without dyslexia (RQ2). Finally, we examined which literacy- and language-related skills contributed to word learning across reading levels (RQ3).

The hypothesis that children with dyslexia show lower performance in word learning specifically on the naming (production) of novel words was confirmed. Groups did not differ in repetition (immediate recall) or identification (phonological recognition) of the novel words. However, children with dyslexia had more difficulty in the naming (production) phase, in which they were asked to recall the target after a delay. Children with dyslexia made more errors overall, although the distribution of errors did not differ in the two groups.

Repetition and recognition are arguably less demanding than naming (production or recall). This finding is consistent with previous studies, in which word learning deficits are typically found for phonological recall but not immediate recall (repetition) or recognition (e.g., Gilliver and Byrne, 2009; Litt and Nation, 2014). As was found in previous studies, children with dyslexia generally perform well on the repetition of simple nonwords, only showing poorer performance when nonwords are three to four syllables long (de Bree et al., 2007). Moreover, children with dyslexia only seem to have difficulties with the identification of novel words when foils are phonologically highly similar, such as changes in final voicing (Alt et al., 2017). Importantly, processing demands increase when a higher level of phonological skills and detail of representations are required. For instance, in a sample of English-speaking second graders, Alt et al. (2017) found similar performance on novel word recognition in children with dyslexia compared to typically developing peers. Children with dyslexia showed lower performance on mispronunciation detection (i.e., requiring access to a detailed novel phonological form), but only when words were longer (four vs. two syllables) or phonologically similar to one another (similar sounding foils had different final phonemes, e.g., “gompav” vs. “gompaf”). In addition, when novel words had to be named (i.e., requiring production of the novel phonological form), children with dyslexia showed lower performance when words were more phonologically similar.

In a follow-up study, Alt et al. (2019) showed that naming deficits for children with dyslexia (7-9 years) were related to phonological rather than semantic learning, again showing effects of word length and phonological similarity. In these studies, as noted by Alt et al., the similar sounding foils may have influenced children's phonological representations in the naming task. In the current study, differences in naming accuracy were observed with short (one and two-syllable) targets, in the absence of similar sounding foils. The specific difficulties with naming (phonological recall) suggest that phonological representations were sufficient for immediate recall (repetition) but were not maintained over a short period of time, as has been observed by others (Litt and Nation, 2014). However, it is noteworthy that we were still able to find differences between children with dyslexia and typical readers in the naming phase after repeated exposure (4–5 times) to the new words. It is conceivable that children with dyslexia are disadvantaged by a delay between exposure and naming, or by the exposure to foils in the identification phase. However, earlier studies also observed group differences in “fast mapping” or PAL procedures, in which a production probe is typically presented immediately after repeated exposure of a novel word (Litt et al., 2013; Kalashnikova and Burnham, 2016; Alt et al., 2017).

We did not find support for the hypothesis that children with dyslexia may benefit less from phonological cues for word class in word learning than their typical reading peers. Dutch children—with or without dyslexia—do not seem to show sensitivity to phonological cues for word class in word learning. This lack of sensitivity might reflect the fact that Dutch phonological cues for word class are probabilistic rather than deterministic. For instance, while bisyllabic items were all ‘noun-like' (mapier- papier “paper”), monosyllabic items were more ambiguous (voek is similar to verb stems such as zoek “search” as well as monosyllabic nouns such as boek “book”). It is also possible that the nature of the word learning task (i.e., exposing children to consistent and inconsistent items in sentence frames) might have obscured any effect of phonological cues for word class. As we did not test children's initial ‘guess' to word class for the novel words, we cannot determine whether phonological cues would have played a role before learning. A main effect of phonological cue was not found either, indicating that children performed similarly on the production of monosyllabic (‘verb-like') and bisyllabic (‘noun-like') targets, even though these differed in neighborhood density. It should be noted, however, that the effect size for this main effect was small (while other effect sizes were close to zero) and the current sample size provided limited power to detect such an effect as significant. Given earlier results for Dutch (Don and Erkelens, 2008), perhaps such cues are only used for initial categorization and easily overridden. In Dutch, morphological or syntactic cues might be more useful than phonological cues for word categorization (Erkelens, 2009). Given that words were offered in sentence frames (“a” vs. “I”), this information may have overridden the weak phonological cues. Fitneva et al. (2009) observed effects of consistency only on verb learning. As suggested by these authors, this fits computational analyses of English, which show that nouns benefit from distributional information while phonological cues are more useful for verbs (Monaghan et al., 2007). In sum, even if phonological cues could theoretically be used for word learning, they were not strong enough in our study to find substantial effects. There is no evidence to suggest that children with dyslexia are less sensitive to phonological cues for word class than typical readers. This also aligns with recent findings indicating that children with dyslexia do not show specific impairments in statistical learning (e.g., Schmalz et al., 2019; van Witteloostuijn et al., 2019; West et al., 2021).

The hypothesis that performance on word learning is explained by language- and literacy-related skills in both children with and without dyslexia was confirmed. Both oral language skills (receptive vocabulary) and literacy-related skills (phonological awareness and nonword reading) were found to contribute to word learning in the naming phase. These results are in line with previous findings that vocabulary and phonological representations are strongly related (Melby-Lervåg et al., 2012). It should, however, be stressed that our findings stem from concurrent data, and do not rely on a longitudinal study of the influence of phonology (phonological form and phonological cues) on word learning.

In sum, the current findings are in line with previous studies on the phonological deficit (Swan and Goswami, 1997; Ramus and Szenkovits, 2008) and studies that show phonological difficulties contribute to word learning (problems) in children with dyslexia (Thomson and Goswami, 2010; Litt and Nation, 2014; Alt et al., 2017, 2019; Adlof et al., 2021). Taken together, this study shows that retrieving and producing phonological forms of novel words is challenging for children with dyslexia. We found no evidence of an implicit learning deficit, as neither group used phonological cues for word class. Possibly, the lack of an additional impact of the phonological deficit on word categorization may act as a buffer for language acquisition. This indicates that word learning and word reading are important to consider as separate aspects in relation to dyslexia.

In future work, it would be interesting to assess the role of orthography in novel word learning of children with dyslexia. Children with dyslexia may benefit from orthography, although to a lesser extent than typical readers (Baron et al., 2018; Alt et al., 2019). Finally, it is unclear to what extent children would have benefited from more exposure to the novel words, as has been found for children with developmental language disorders (Gray, 2004). It is likely that children with dyslexia need more exposure to novel words to form sufficiently detailed and robust phonological representations.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The studies involving humans were approved by the Ethics Committee of Utrecht University. The studies were conducted in accordance with the local legislation and institutional requirements. Written informed consent for participation in this study was provided by the participants‘ legal guardians/next of kin.

SvV: Formal analysis, Methodology, Visualization, Writing – original draft, Writing – review & editing. AK: Conceptualization, Data curation, Investigation, Methodology, Project administration, Supervision, Writing – review & editing. EdB: Conceptualization, Data curation, Investigation, Methodology, Project administration, Supervision, Writing – review & editing.