The rapid acquisition of language, influenced by linguistic and cognitive resources and shaped by psychosocial factors, is commonly observed in children aged 24–30 months (Bornstein and Putnick, 2012; Walker et al., 2011), across several languages (Braginsky et al., 2019). However, research has demonstrated variability in the early vocabularies of typically developing children (TD) (Fenson et al., 1994), Late Talkers (LTs) (Rescorla, 2011), and those at risk for a neurodevelopmental disorder including Autism Spectrum Disorder (ASD) (Rescorla and Safyer, 2013) and Developmental Language Disorder (DLD) (Leonard, 2014; Acosta Rodríguez et al., 2017).

In TD children, vocabulary development is influenced by child and parental factors as early as age two (Nylund et al., 2021). Variability in vocabulary acquisition may arise from cognitive and linguistic factors such as social cues, statistical learning mechanisms, context learning, and word frequency affecting learning nouns, verbs, adjectives, and functional words (Braginsky et al., 2019). While most TD children exhibit lexical growth between 24 and 30 months, LTs experience delays in the onset of first words and word combinations, with slow vocabulary growth after 24 months, despite having no hearing loss and intact cognitive abilities (Dale et al., 2003; Rescorla and Dale, 2013). LTs continue to exhibit delays at age three, with unusually small vocabularies (Fisher, 2017). Producing fewer than 100 words by 30 months can indicate early language delays (Rescorla et al., 2000), although variability is expected between persistent and non-persistent LTs (Auza and Murata, 2021; Rescorla, 2011; Desmarais et al., 2010).

ASD, characterized by social interaction deficits and repetitive behaviors, may or may not involve language delays (World Health Organization, 2024). Approximately 75% of ASD individuals are verbal (Lord et al., 2004; Sigman and McGovern, 2005; Tager-Flusberg and Kasari, 2013), yet they exhibit variability in vocabulary production (Rescorla and Safyer, 2013), sometimes producing fewer words than TD children at 12 months (Righi et al., 2014). Children with ASD may produce similar percentages of nouns, verbs, and functional words compared to TD children, but they may also demonstrate delayed onset of first words, limited functional word use, and reduced morphological production (Marini et al., 2020; Oetting and Hadley, 2017). Furthermore, they may lack word combinations or have short Mean Length of Utterances (MLU) (Charman et al., 2003) similar to persistent LTs that progress to DLD (Chilosi et al., 2019; Leonard, 2014; Rescorla, 2011).

In the absence of cognitive, motor, neurological, or hearing problems (Leonard, 2014), DLD is characterized by difficulties in understanding or producing language and using it in context for communication (World Health Organization, 2024). Typically, children with DLD experience delays in the onset of first words and word combinations (Fisher, 2017; McGregor, 2020). However, variability in language production is also anticipated, given the heterogeneity in language features associated with DLD (Bishop, 2017).

Therefore, comparing diverse groups of children is crucial to delineate unique and shared features among different conditions, refining diagnostic criteria among them, and elucidating expressive lexical profiles (Ellis-Weismer et al., 2011; Lord et al., 2004; Luyster et al., 2007). Analyzing vocabulary production across diverse groups involves examining those categories commonly used by TD children, revealing variability as the norm rather than the exception (Fenson et al., 1994). Instead of focusing on a single pattern that overlooks the inherent variability in language use across individuals (Bates et al., 1994; Bates et al., 1988), variability should be used as a key to understanding the process of language acquisition. Each child’s developmental trajectory is indeed influenced by a complex interplay of factors, resulting in unique patterns even within clinical groups. Variability underscores the importance of personalized approaches to understanding and supporting each child’s needs. By acknowledging and studying variability, we can better tailor interventions and support strategies that cater to individual strengths and challenges in language development. Emphasizing this variability also encourages a broader perspective on developmental patterns across clinical groups, promoting nuanced research and clinical practices.

From a descriptive perspective, differences in overall word production and MLU3 are observed between children with atypical development and TD children. Specifically, the median values for overall word production were: TD (M = 289, IQR = 166.5, 412); LTs (M = 34, IQR = 17.5, 72); ASD (M = 47, IQR = 3, 160.5); and DLD (M = 74, IQR = 45, 196.5). For MLU3, the median values were: TD (M = 2.7, IQR = 2, 3.6); LTs (M = 0, IQR = 0, 2); ASD (M = 0, IQR = 0, 2.5); and DLD (M = 0, IQR = 0, 2.7) (See Supplementary Table S1).

To address the first research question, which focuses on the vocabulary production patterns across the four groups, we employed Generalized Linear Mixed Models (GLMM). Based on the optimal model, we reported the main effects and interaction terms, estimated regression coefficients, standard errors, t-values, and p-values (Supplementary Tables S2, S3.1–3.5). The same results were also presented and visualized as profile plots (Figure 1).

As shown in Figure 1, TD children, as expected, scored significantly higher in all vocabulary categories compared to LTs, ASD and DLD. Notably, across all groups, the categories Routines and Sounds-Onomatopoeias consistently scored higher than Nouns, Verbs, Adjectives-Adverbs, and Functional Words. Nouns and Adjectives-Adverbs had relatively higher scores. Children with DLD notably showed high scores in Sounds-Onomatopoeias.

The results of a more detailed analysis of the aforementioned profiles using GLMM are summarized in Supplementary Tables S2, S3.1–3.5. Supplementary Table S2 shows the comparison of scores between the six lexical categories in each of the four groups. In addition, Supplementary Tables S3.1–3.5 report the main effects of the four groups for each vocabulary category, as well as the interaction effects between vocabulary categories for the following pairs: Nouns vs. Verbs; Verbs vs. Adjectives/Adverbs; Adjectives/Adverbs vs. Functional Words; Functional Words vs. Routines; Routines vs. Sounds-Onomatopoeias.

The average scores for LTs, children with ASD, and those with DLD were all significantly lower than those for TD children. Specifically, the estimated average score differences were − 38.2 for LTs (p < 0.001), −28.9 for ASD (p < 0.001), and − 27.5 for DLD (p < 0.001) (Supplementary Table S3.1). These results highlight significant differences in vocabulary production across the four groups.

Regarding the main effects of vocabulary categories, Verbs showed significantly lower scores compared to Nouns (reference category) (estimate = −16.7, p < 0.001). In contrast, no significant differences were found for Adjectives-Adverbs (estimate = −3.5, p > 0.05). Functional Words had significantly lower scores (estimate = −18.5, p < 0.001), while Routines had significantly higher scores (estimate = 6.4, p < 0.05). Sounds-Onomatopoeias showed the highest scores (estimate = 20.4, p < 0.001), indicating clear differences across vocabulary categories.

With respect to the interaction effects, significant interactions were observed in specific vocabulary categories among the groups. For example, for the DLD group, a significant interaction was observed in Sounds-Onomatopoeias (estimate = 13.0, p = 0.020, Supplementary Table S3.5), indicating that this vocabulary category is particularly prominent in children with DLD compared to other vocabulary types.

To answer the second research question, the distribution of vocabulary production scores was observed and quantiles were calculated (Figure 2). Examination of the equivalence/non-equivalence of these distributions revealed statistically significant differences in non-equivalence across all six categories. However, for Routines, while the overall null hypothesis that the distributions are equal across the four groups was rejected with a p-value slightly below 0.05, the Holm-Bonferroni correction for pairwise multiple comparisons showed no significant differences in variability between groups. For Nouns, Adjectives-Adverbs, and Functional Words, the TD group exhibited greater variability compared to the LTs group. For Verbs, the TD group showed greater variability than the other groups. For Sounds-Onomatopoeias, the scores distribution slightly differed between the TD and ASD groups and between the LT and ASD groups (Supplementary Table S4).

Regarding the third research question, a hierarchical cluster analysis revealed six distinct patterns of vocabulary development across the six categories, identifiable as clusters. Among them, three were in the high performance range and the other three were in the low performance range. The first three clusters were named as follows: cluster 1 “Highest vocabulary,” cluster 2 “Moderately high,” (with focus on Routines/ Sounds-onomatopoeias), and cluster 3 “Moderately high,” (with focus on Adjective-Adverbs). Similarly, the three low-score clusters had three levels: cluster 4 “Moderately Low,” (especially high on Sounds-onomatopoeias), cluster 5 “Moderately Low,” (with focus on Nouns), and cluster 6 “Lowest vocabulary,” (with focus on Routines and Sounds-onomatopoeias, Figure 3).

We subsequently examined the relationship between these six clusters and the four groups, revealing a highly significant association (χ² = 52.55, d.f. 15, p < 0.001). The correspondence analysis captured nearly all the variability in two dimensions (Dimension 1: 80%, Dimension 2: 17.6%), as illustrated in Figure 4. Notably, three clusters strongly correlated with TD children, while the remaining three clusters corresponded to LTs, children with DLD, and ASD.

Specifically, 73% of TD children were associated predominantly with high-score clusters 1, 2, and 3 (primarily 1). For LTs, 98% were linked to clusters 4, 5, and 6 (mainly 6). Regarding DLD, 89% were primarily associated with clusters 3, 4, 5, and 6 (especially 4). The ASD group was associated with all clusters, with a predominant association with cluster 6.

Dimension 1 straightforwardly reflects positive values indicating high scores in vocabulary categories and association with TD children, whereas negative values correspond to lower scores and association with the three groups of children with language delays. Dimension 2’s interpretation is less definitive, but distinguishes the ASD group (negative values) from the DLD and LTs groups (positive values).

Furthermore, clusters associated with the DLD group exhibited relatively high scores, particularly in Sounds-onomatopoeias, followed by clusters associated with LTs. In contrast, clusters associated with ASD were characterized by lower scores.

This study aimed to analyze the lexical profiles in four groups of children with typical and atypical development by examining their word production across six categories. While existing English literature emphasizes variability in language acquisition (Bates et al., 2017; Dale and Goodman, 2004; Fenson et al., 1994) and vocabulary categories, this study makes a unique contribution by focusing on the variability of expressive language skills in Spanish-speaking children with typical and atypical language development.

In relation to the first hypothesis, the analysis of the profiles of the six vocabulary categories (Nouns, Verbs, Adjectives-Adverbs, Functional Words, Routines, and Sounds-Onomatopoeias) across the four groups (TD, LTs, ASD, and DLD) revealed significant similarities and differences between the groups. As anticipated, the TD group achieved the highest median scores, indicating robust and typical development across all vocabulary categories. Secondly, our findings suggest that the use of vocabulary categories in Spanish-speaking TD children is similar to those described in TD children acquiring other languages such as English or Italian (Bates et al., 1994; Caselli et al., 1995; D'Amico et al., 2001; D'Odorico and Fasolo, 2007; Rescorla et al., 2014). The order and rates of the acquisition of vocabulary categories, are highly similar regardless of the target language’s typology. Serving as a baseline for comparison, the TD group stands out from the clinical groups due to its high production of vocabulary categories.

Among the clinical groups, one of the similarities was that they showed low production trends. LTs showed the lowest scores compared to those with TD, followed by ASD and DLD. The group with TD exhibited the highest scores in Routines and Sounds-onomatopoeias, compared to Nouns, Verbs, Adjectives-Adverbs, and Functional words. The consistent use of Sounds-onomatopoeias as the most produced category suggests that these sound-based expressions play a significant role in easing language production during early language development. This is likely due to their transitional status during early expressive vocabulary development (Auza and Murata, 2021; Caselli et al., 1995; MacRoy-Higgins et al., 2016). Sounds-onomatopoeias may stand in place of Nouns for children with language difficulties, as challenges in remembering words, particularly Nouns, can impair language processing abilities and comprehension, leading to lower overall vocabulary production (MacRoy-Higgins et al., 2016; Ellis-Weismer et al., 2013).

Some interesting interactions between the vocabulary categories and the four groups were observed in terms of profile differences. A strong interaction effect was observed between the Sounds-onomatopoeias and the DLD group, which showed markedly higher values than the other vocabulary categories. This is in contrast to the profile patterns of the other groups, highlighting their unique dependence on Sounds-onomatopoeias. These findings underscore the disproportionate retention or influence of certain vocabulary categories in each clinical group, pointing to a unique language profile.

Additionally, the low use of verbs and functional words indicated challenges in their production (Jackson-Maldonado et al., 1993; Marini et al., 2020; Oetting and Hadley, 2017; Tardif et al., 1999). Previous research has shown that the lower use of verbs may result from the lower use of adjectives-adverbs and functional words (Bates et al., 1994), which can impact word combinations. Moreover, if a child’s verb repertoire is limited, concern should be raised for those children who remain in the lowest 10%, using about half the verbs that typical children produce at 24 months. The low use of Verbs also affects MLU, as the verb repertoire supports the acquisition of sentence structure (Hadley et al., 2016). Recent studies have demonstrated that linguistic structure is crucial for acquiring non-nominal lexical categories, emphasizing its role over conceptual complexity (Braginsky et al., 2019). Cognitive demands also apply to adjectives and functional words, which rely on the prior acquisition of categories such as nouns and verbs (Bates et al., 1994). Overall, while common profile patterns were observed among the four groups, the interaction effects revealed that each group has distinct characteristics in their vocabulary production. In particular, the differences in verbs and functional words in the LTs and ASD groups (with effect sizes ranging from moderate to large), as well as in sounds-onomatopoeias in the DLD group (with a large effect size), were especially notable.

Regarding vocabulary production variability, the interquartile range provides additional insights and partly corroborates our second hypothesis. While the ASD and DLD groups show greater variability, the LTs displayed low variability. Our results reveal that within a clinical condition, some children can produce as many words in certain vocabulary categories, such as nouns, verbs, and routines, as TD children, while in other categories, they produce few or none. While the variability in sounds-onomatopoeias is broad but similar across all groups, the variability of other vocabulary categories differs. In nouns and adjectives-adverbs, the ASD and DLD groups exhibit similar patterns, with greater dispersion than LTs. This suggests that the ASD and DLD group include more individuals who produce either few or many words (greater dispersion) in these categories, whereas LTs mostly produce few words (low dispersion). For verbs, the variability differs, with each group showing distinct levels of dispersion: TD shows the greatest dispersion while LTs have the lowest dispersion, making the TD group distinguishable from the clinical groups. Despite the cognitive demands and dependency on other words that make verb production challenging in various languages (Childers and Tomasello, 2006; Tomasello et al., 1997), studies have shown that TD children as young as 2 years can produce verbs in languages like English or Spanish, or even as young as 1 year and 3 months in languages like Korean (Choi and Gopnik, 1995). Findings on Spanish acquisition in both monolingual and bilingual children indicate early usage of different types of verbs with expanded verbal morphology and low error rates (Gathercole et al., 1999; Ingram et al., 2008; Serra et al., 2001; Serrati Sellabona et al., 2004; Silva-Corvalán and Montanari, 2008). Indeed, cognitive and linguistic demands may affect variability in verb production, as evidenced by our results. A different panorama is observed with functional words. Despite the DLD group being older, their use of functional words shows low variability very much alike to LTs, indicating difficulties in producing this category. Thus, distinguishing between groups based on vocabulary categories requires considering both the number of words produced and their dispersion. For instance, verb variability does not help distinguish between ASD and DLD, while pronounced low dispersion in this category distinguishes LTs, indicating that most of the children do not produce or produce few verbs. In functional words, the low variability of LTs aligns with the DLD group and, to a lesser degree, with ASD children. Our results also showed that the latter group can produce functional words similarly to TD children, although most children with ASD may not produce them due to morphosyntactic difficulties (Marini et al., 2020; Oetting and Hadley, 2017). Variability in routines was common across all four groups of children, particularly in the ASD group. Interestingly, children with ASD exhibited more routines than LTs. This phenomenon may be partly attributed to the rote learning mechanism that many children with ASD develop early on, which occurs even more frequently than in TD children (Wivell, 2017).

Our analysis highlighted differences in the language abilities of various groups of children. The variability in word production in typical and atypical children serves as an indicator of their increasing language proficiency. This variability reflects individual differences observed among different clinical groups. Some children learn words slowly and may remain at risk for language disorders (Bates, 2004), while others acquire words quickly and start using them at an early age (Fenson et al., 1994).Variability has been explored in many studies (e.g., Fenson et al., 1994; Fernald and Marchman, 2011; Huttenlocher et al., 2010) as it provides insights into the unique developmental paths of both typical and atypical children, allowing for tailored interventions and better understanding of vocabulary acquisition processes within each clinical condition.

In general terms, six clusters of word production were identified, confirming our third hypothesis: three clusters exhibited high production profiles and three exhibited low production profiles. This suggests a typical and well-distributed language development profile, varying from low but age-expected production to medium or high production across all vocabulary categories. TD children exhibited a diverse and balanced distribution across three clusters of high production, indicating three levels of word production. Variations among different vocabulary categories are expected; other studies have shown that they may involve different learning mechanisms, such as concreteness for nouns, frequency for predicates, or linguistic structure for functional words (Braginsky et al., 2019). This result is consistent with the literature, which has stated that language production is highly variable among TD children (Bates et al., 2017). Concerning the three low production profiles, LTs were predominantly associated with these clusters, showing three levels of word use, with a strong presence in clusters five and six, which were the lowest of all the clusters. They appear to lack visible production strategies, suggesting a delayed profile in language development. This result indicates that LTs are, also in some way, heterogeneous in lexical production, consistent with findings from previous studies (Dale et al., 2003; Desmarais et al., 2010; Perry and Kucker, 2019). In Desmarais et al. (2010), heterogeneity among LTs was also observed, with three different clusters, one being more affected, particularly in the number of words. Accurately characterizing LTs and others at risk for language delays is essential for precise diagnosis and targeted interventions. This requires specific statistical methods that account for the inherent heterogeneity in populations with language delays (Perry and Kucker, 2019). In our study, the association of LTs with clusters five and six suggests a significant delay, with greater language difficulties compared to other groups. This may indicate that some of these children could be classified in the future as presenting with DLD, especially given the low use of verbs and functional words found in these clusters. The ASD group showed associations with all clusters; notably, 30% were associated with a typical language cluster, consistent with previous findings (Charman et al., 2003; Ellis-Weismer et al., 2011; Luyster et al., 2007). Our finding aligns with Vogindroukas et al. (2022), in that ASD can belong to either a typical or an atypical language profile. This highlights the importance of defining clinical subgroups within this spectrum.

Nonetheless, the highest percentage of individuals with ASD (39%) were comprised in the lowest cluster, which also encompassed a high proportion of LTs (41%). This indicates that while the lowest cluster is representative of the ASD group, it also overlaps with LTs. Regarding these findings, Rescorla and Safyer (2013), suggested that lexical composition in ASD is delayed more than atypical. This perspective aligns with a “dimensional account” (Jiménez et al., 2021) of language delay, implying that language disorders in ASD primarily reflect quantitative differences along a single dimension. Nonetheless, Lazenby et al. (2016) found that by 12 months, children later diagnosed with ASD had low receptive and expressive language scores but were also producing and understanding certain words in an “unexpected way.” This finding supports the “categorical account” (Jiménez et al., 2021) which posits that children with ASD have different language profiles. Our findings align with both perspectives, as we observed distinct vocabulary category discrepancies between LTs and ASD, highlighting the discriminatory potential of specific categories. Most of the children in the DLD group (71%) were linked to clusters of low production, predominantly cluster four. However, another third (30%) was associated with clusters of high production, possibly due to being the oldest among the children. The similarities in lexical profiles between the LTs and DLD group suggest shared lexical production traits, possibly indicating parallel lexical acquisition trajectories in verbs and functional words. Previous research has supported the idea that vocabulary composition is a critical factor in differentiating children with language delays (Ellis-Weismer et al., 2011; Jiménez et al., 2021; Perry et al., 2023). Regarding verbs, persistent LTs that evolve into DLD can be detected with the use of this category (Auza and Murata, 2021; Hadley et al., 2016). These profiles of association with clusters can provide insights into the linguistic strengths and challenges within each group. It is essential to consider that individual differences can contribute to this variability in children with TD and with clinical conditions.

This analysis highlighted distinct lexical profiles among TD, LTs, ASD, and DLD groups, showing variability across vocabulary categories. Rather than simply confirming lower vocabulary levels in clinical groups, the study suggests that the variability within each category may reflect the unique characteristics of each group. This variability offers additional insight into the characterization of these groups, contributing to a deeper understanding of heterogeneity in early language development across clinical populations.