Lay Summary
In this article, we focus on child speech production and development. We present a technique used in ordinary school settings to investigate how children tell stories, which uses graph analysis to understand speech repetition patterns. We discuss data provided by a series of studies using graph analysis, which show that it can predict children’s reading and linguistic skills. In general, children who produce more connected narratives, with fewer repetitions, become better readers and develop better linguistic skills later on. Studies like the ones reported here are important for informing intervention programs, research that turns scientific findings into practical applications and public educational policies.
Introduction
Acquiring oral language is a natural process for humans, and our brains have evolved to learn linguistic cues (Dehaene, 2009). Unless children have particular forms of genetic variants, such as Developmental Language Disorder (DLD), they are likely to achieve a high degree of skill understanding and producing language.
From the final stages of pregnancy to the end of our lives, evolution has allowed us to acquire oral language without any kind of formal instruction. Humans have been capable of communicating with each other since the first hominids, even if communication happened, in the beginning, in a simplistic manner. Nonetheless, with evolution, it has been possible to systematize the sounds, and humans developed the ability to think symbolically. Despite the debate whether language is an innate faculty of the human cognitive system or a byproduct of human evolution and its interaction with the environment, it is known that language is used to support thoughts and, to do so, everyone must acquire it (Traxler, 2023).
1. Child speech production and development
Despite the variability of the input they receive, babies can develop oral language based only on the input they receive from their caregivers, without any formal instruction, and this starts very early, when they are still in the womb. The fetus’ auditory system can perceive environmental input in the third trimester of pregnancy. Specifically, the auditory stimuli that the fetus is exposed to provide the basic outline for the prosodic features of the fetus’ native language. Because the mother’s voice is generated in and propagated through the mother’s body, it is, on average, the loudest thing the fetus is exposed to. Prosodic characteristics of speech (like accent, for example) are available to the fetus and can help the child identify important components of the speech signal. As soon as a fetus learns something about the prosody of their native language, they will gain substantial benefits when it comes to learning their first language (Traxler, 2023).
From birth on, phonemes constitute the building blocks of words in the baby’s developing language. Thus, one basic task that the infant must solve is to figure out how the inventory of speech sounds is organized. Once again, prosody plays a big role. Infants make use of prosodic features to identify boundaries between words in fluent speech. In addition, infants notice patterns in complex stimuli and use those patterns to analyze speech stimuli and identify important subcomponents, including words. They can use several strategies to segment and understand the functioning of the language and how it is organized. First, they need to identify the spoken sounds, noticing what the possible phonemes of the specific language are that the given context will allow them to learn.
But how do we know how babies process language before they can speak? Among many possible ways to investigate the language acquisition phenomenon in babies, researchers have developed an experimental technique that measures the applied pressure on pacifiers given to babies while different conditions are being tested. This enables scientists to observe what babies are paying attention to when acquiring language (Traxler, 2023). The pioneering study employing this technique was conducted by Eimas et al. (1971), who investigated the perception of phonemes by newborns. A specific sound was played repeatedly and then replaced with another phoneme (for example, /ba/ was replaced by /pa/). The results showed that newborn babies can already distinguish between the two sounds, allowing them to assimilate possible phonemes in the language being acquired.
In addition to differentiating between phonemes, infants need to identify how these phonemes are organized in words as well. According to Traxler (2023), this ability appears at around 6 months of age. One of the strategies they possibly use is prosody, since this linguistic feature of speech can assist in understanding the beginning and ending of probable words. Babies may be able to infer which sound is the limit of a word according to the intonation and emphasis (stressed syllable) that adults make. Later on, children also make use of phonological information in the target language, identifying the patterns of occurrence of phonemes and estimating the possibility of a certain syllabic sound occurring in the language being acquired (Traxler, 2023). Once phoneme and word notions are acquired, children begin to learn the meaning of words (lexicon) and, finally, the way in which they are organized within a sentence (syntax).
Understanding the neural processes that underlie behavioral and developmental trajectories of spoken language in infancy is crucial to a range of psycholinguistic hypotheses for how children later learn other linguistic skills, such as reading. Nonetheless, although the biological bases of language acquisition are essential, one cannot underestimate or neglect the role of experience in this process.
2. The impact of experience on early speech development
The role of the environment is essential in language development. Infants learn about their first language by listening to people talk. Typically, when we talk to infants, the way we speak changes radically: the pitch of our voice increases, we speak in shorter sentences, we speak more clearly and distinctly, and we vary our pitch and our loudness much more than we do when we speak to adults. This collection of speech properties goes by the name of infant-directed speech or child-directed speech (and was previously referred to as motherese). Research on infant-directed speech indicates that, although it may not be necessary for children to learn their first language, it helps them master some aspects of speech comprehension and word learning (Traxler, 2023).
Most child-directed speech is provided by children’s parents or caregivers, who are important agents in their language learning trajectory, together with the quality and quantity of exposure that their environment offers. This influence is translated into factors such as the size of the baby’s vocabulary and the speed with which the baby can process incoming speech. Importantly, these two abilities show a symbiotic relationship, which has long-term, cumulative effects; that is, the more words the child knows, the faster they can recognize familiar words; the faster they can recognize familiar words, the more resources they have left over to pay attention to and learn from other parts of the speech signal. Research has mapped this virtual cycle, children who are faster at recognizing familiar words at 6 months old have a bigger vocabulary at 24 months old than children who are slower at recognizing familiar words at 6 months old (Traxler, 2023). In other words, differences between children who are fast and slow processors can be explained largely as a kind of practice effect and are impacted by socioeconomic differences.
Socioeconomic disparities are associated with differences in cognitive development and disparities in brain structure, more prominently in regions supporting language, reading, executive functions and spatial skills (Noble et al., 2015). Differences in processing efficiency between different infants can also be attributed in part to how much adults talk to them, and there are vast differences between and within different socioeconomic groups in the amount of speech that gets directed toward infants (Noble et al., 2015). Maternal education also appears to matter a great deal. Independent of socioeconomic status, mothers with more formal education speak more to their children than mothers with less formal education.
As we have seen so far, there is wide consensus about the interplay between nature and nurture when it comes to child language development. In this sense, research in the field must consider both the neurobiological underpinnings and the effects of experience when investigating child speech.
3. Neuroimaging techniques: potential and limitations
Until recently, investigations of the neurobiology of early language development had been hindered by the lack of age-appropriate neuroimaging techniques. This has started changing rapidly due to technological advances such as functional near infrared spectroscopy (fNIRS). These non-invasive tools have allowed a deeper understanding of how the developing brain deals with language. Recently, the establishment of neuroimaging methodological protocols for infant experimental studies has consolidated research pathways, leading to a more robust understanding of this line of research. That means that neural pathways and trajectories have been identified that demonstrate a reliable association with different dimensions of language development. Understanding this relationship, grounded on the neurobiology of early language production and comprehension, has the potential to yield new models of developmental trajectories.
However, neuroimaging techniques are still not widely accessible or affordable, especially in the case of scientists doing research in WILD countries[1] like Brazil. The alternative generally involves integrating advances in neuroimaging (e.g., fNIRS, fMRI, ERPs) and more ecological study designs, consisting of behavioral techniques and translational perspectives to address fundamental questions about early language acquisition and developmental trajectories.
Examples of behavioral techniques that can be combined with neuroimaging techniques include network science. Network-based approaches offer a complementary perspective, shifting the focus from neurological pathways to the structural organization of children’s cognitive and overt linguistic behavior. Network-based models (e.g., semantic, phonological, and neural) have become powerful tools for examining how language knowledge is organized and developed across populations. Findings from network approaches converge on the idea that the structure of a child’s language networks offers more than a quantitative index of how many words are known; it also provides a qualitative lens on how words are connected and how learning unfolds over time.
4. Alternative approaches: Graph-based Analysis of Child Narrative Speech
The investigation of children’s oral abilities is relevant in many ways, since it can simultaneously provide information about features of their linguistic development as well as their level of competence at the discourse level in naturalistic contexts. Oral narrative skills in kindergarten have been associated with later-acquired reading outcomes (e.g., Catts; Adlof; Weismer, 2006). In addition, research findings also show that young children’s oral narrative skills predict their social and behavioral competence, corroborating the fundamental role of oral narrative abilities to literacy development and overall academic performance (Huang et al., 2022).
Among non-semantic network approaches, speech graphs offer a novel technique that focuses on structural patterns and recurrence rather than focusing on meaning or grammar. In this approach, words are represented as nodes and their sequential connections as edges, creating a network-like map of how a story unfolds (Mota et al., 2023). The resulting speech graph yields two unique metrics: Largest Strongly Connected Component (LSC) and Repeated Edges (RE). These are non-language specific, and provide a complementary lens for examining connectedness, with the potential to reveal subtle disruptions in narrative organization that are characteristic of speech in early child development. Figure 1 illustrates the oral narrative protocol used in this line of research.
Figure 1. Figure 1. Example of an oral narrative protocol. Source: the authors. ALT: Diagram showing an oral narrative protocol: a child first observes a set of pictures, plans a story, tells the story, and the narrative is then represented as a speech graph.
Narratives are particularly useful for speech graph analysis because, unlike casual conversation, they place greater cognitive demands on the speaker. Narrative production requires the mental organization of ideas before verbal expression, making it a complex and revealing context for examining how thoughts are connected over time (McGregor, 1997). In such a context, speech graph analysis can differentiate between short-range recurrence, or repeated associations between the same pair of words (e.g., “a boy, a boy”), and long-range connectedness, which refers to the number of nodes within a connected component. For example, in the sequence “The boy with the cap on his head is the boy I spoke to yesterday” words are indirectly connected through multiple associations, providing a measure of global connectedness (Mota et al., 2016).
While most linguistic metrics provide a holistic analysis of cohesion by capturing complexity across the whole narrative, speech graph analysis offers an alternative by quantifying structural organization and recurrence patterns within speech, providing insights independent of meaning or grammatical form. Taken together, speech graph features provide a promising approach to studying narrative structure, demonstrating how connectedness can be measured in stories.
In children, the increase in language complexity is generally associated with a more complex mental organization informing cognitive development. Based on the idea that children’s narrative skills reveal how they organize their thoughts, a graph-theoretical based approach - in which each node represents a word, and each directed edge represents the temporal order of consecutive words - is quite informative in naturalistic settings in different contexts.
Previous studies applying speech graph analyses have reported that the oral narrative structure changes from a short to a long-range recurrence pattern as soon as a child starts to read, increasing connectedness. In Mota et al. (2016), narratives from seventy-six second-grade children (ages from 6 to 8 years old) were collected in the middle of the school year, together with the intelligence quotient (IQ), theory of mind (ToM), as well as reading and math scores acquired from a national assessment four months later. The results reveal that the Largest Strongly Connected Component (LSC) predicted the national reading exam performance five months later. Furthermore, the results reveal that the association with reading performance was independent of IQ and ToM performances. In a subsequent study with the same participants in the third grade, Mota et al. (2020) reported that exclusively verbal short-term memory was associated with connectedness measures (LSC). In addition, the correlation between short-term memory and reading fluency was significant only in the second grade, pointing to specific dynamics associated with the reading acquisition process. More recently, Malcorra et al. (2024) conducted a longitudinal study with 253 children (first and second graders), which corroborated the results in Mota et al. (2016), showing that speech connectedness predicted performance in phonological awareness, reading comprehension, and word reading accuracy 3–4 months prior to full literacy.
In the case of children with developmental language disorder (DLD), Perez et al. (2025) showed that the analysis of children’s narratives with speech graphs may provide a language-independent complement to standardized tools for identifying speech-level disruptions in atypical populations. The study examined longitudinal changes in oral narrative retelling among children with and without DLD and combined speech graphs analysis, semantics and morphosyntax scores over one year to differentiate typically developing (TD) and DLD groups. Forty-two children (21 TD, 21 DLD) matched by age, gender and IQ completed a set of Semantics and Morphosyntax tests and narrative retelling tasks at two time points. Composite scores for semantics and morphosyntax were analyzed alongside speech graph measures of long-range (uninterrupted utterances) and short-range (repetitions) connectedness. Results show that typically developing children achieved higher semantic and morphosyntactic composite scores than peers with DLD, and showed greater linguistic gains across time. Higher composite scores predicted greater long-range connectedness, reflecting stronger narrative cohesion. Children with DLD produced narratives with significantly lower long-range connectedness and higher short-range repetition, indicating fragmented speech organization.
Taken together, the studies reviewed here evidence the power of the speech graphs technique in predicting children’s linguistic skills, in clinical and non-clinical samples, by assessing their speech production at an early age. Although none of the studies included children’s socioeconomic status as an independent variable, we can hypothesize, based on what we discussed in Section 2, that living conditions, including unpredictability and scarcity, would interact with cognitive variables in predicting children’s linguistic skills (Noble et al., 2015).
5. Conclusion: The future of speech graphs investigations
Speech graphs provide a safe and scalable way to evaluate educational policies, since the stories may be recorded and transcribed automatically using an AI-based algorithm. Each narrative is represented computationally by a word recurrence graph, which is analyzed to determine academic development. In a typical school environment, graph attributes indicate academic development, according to Mota et al. (2016, 2020). Therefore, this protocol can be a valuable, language-independent tool for education professionals. As a monitoring tool, researchers (and also teachers) can plot information on the narrative structure obtained through graphs on a normative curve representing the evolution of oral narrative complexity throughout the school years. With this curve, they can evaluate the development of the user’s communication skills compared to their peers in the same or different school years, providing a global perspective of their ability to generate narratives.
In a translational framework, aiming to bridge research in the laboratory and applications in educational settings, speech graph models may inform accommodations and interventions that address delays and disorders in language development, ranging from early speech development to reading acquisition. Thus, investigations into the neurobiology of language development, considering aspects that are increasingly focusing on translational research, highlight the relevance of this field for practitioners as well as interdisciplinary and novice researchers. In all cases, complying with ethical and transparency norms, including, for example, how to handle sensitive data from minors and how to document automated analyses, are of paramount importance to guarantee healthy scientific and educational practices.
In general, the studies reviewed in this article point to a feasible, low-cost, and inclusive assessment based on storytelling (as it does not require the child to be able to read), enabling the screening and tracking of cognitive development related to reading acquisition on a large scale. For Latin American countries such as Brazil, scalable solutions could be crucial to planning public educational policies based on data. Public and policy domains may benefit from more translational studies such as the ones reported here, which aim to be illustrative of the field, rather than exhaustive.
Future studies could examine the relationship between language connectedness and various cognitive and linguistic attributes in a more diverse sample, including families of different socioeconomic status and educational levels, and examine whether it would be possible to replicate the results in a more heterogeneous sample.
In addition, future studies could evaluate other measures based not only on transcriptions but also on acoustic signals and facial expressions, which could help us understand communication in greater depth, as well as the role of a child’s temperament, considering that the ability to narrate a story may be challenging for shy children.
In conclusion, the relevance of this article lies within the need for further understanding of the neurocognitive processes that ground language development and the optimization of language investigation and intervention programs. In that sense, we hope that the present piece has offered a humble contribution to important discussions in the field.
Acknowledgments
The authors thank all the parents, educational professionals and children for their generous participation in the studies.
Additional Information
Conflict of Interest
Natália Bezerra Mota works at Motrix, a technology startup. Janaina Weissheimer declares no conflict of interest.
Statement of Data Availability
Data sharing is not applicable to this article, as no new data were created or analyzed in this study.
AI Usage Statement
The authors declare that no AI tool was used.
Funding Sources
JW receives a grant as productivity researcher at the National Council of Technological and Scientific Development - CNPq (n. 308917/2022-7).
References
ALVES, Marcus Vinícius; EKUNI, Roberta; HERMIDA, Maria Julia; VALLE-LISBOA, Juan (ed.). Cognitive sciences and education in non-WEIRD populations: a Latin American perspective. Cham: Springer Nature, 2022.
CATTS, Hugh W.; ADLOF, Suzanne M.; WEISMER, Susan Ellis. Language deficits in poor comprehenders: A case for the simple view of reading. Journal of Speech, Language, and Hearing Research, v. 49, n. 2, p. 278-293, 2006. https://doi.org/10.1044/1092-4388(2006/023).
DEHAENE, Stanislas. Reading in the brain: The science and evolution of a human invention. New York: Viking, 2009.
EIMAS, Peter D.; SIQUELAND, Einar R.; JUSCZYK, Peter; VIGORITO, James. Speech perception in infants. Science, v. 171, n. 3968, p. 303-306, 1971. https://doi.org/10.1126/science.171.3968.303.
HUANG, Becky H.; BEDORE, Lisa M.; RAMÍREZ, Rica; WICHA, Nicole. Contributions of oral narrative skills to English reading in Spanish-English Latino/a dual language learners. Journal of Speech, Language, and Hearing Research, v. 65, n. 2, p. 653-671, 2022. https://doi.org/10.1044/2021_JSLHR-21-00105.
MALCORRA, Bárbara; RIBEIRO, Marina; JENSEN, Luísa; GOMES, Giovana; MELETTI, Tamara; MOTA, Natália Bezerra. Speech connectedness predicts reading performance three months in advance: a longitudinal experiment. NPJ Science of Learning, v. 9, art. 35, 2024. https://doi.org/10.1038/s41539-024-00248-4.
MCGREGOR, Karla K. The nature of word-finding errors of preschoolers with and without word-finding deficits. Journal of Speech, Language, and Hearing Research, v. 40, n. 6, p. 1232-1244, 1997. https://doi.org/10.1044/jslhr.4006.1232.
MOTA, Natália Bezerra; CALLIPO, Renata; LEITE, Lígia; TORRES, Ana Raquel; WEISSHEIMER, Janaina; BUNGE, Silvia A.; COPELLI, Mauro; RIBEIRO, Sidarta. Verbal short-term memory underlies typical development of ‘thought organization’ measured as speech connectedness. Mind, Brain, and Education, v. 14, n. 1, p. 51-60, 2020. https://doi.org/10.1111/mbe.12208.
MOTA, Natália Bezerra; WEISSHEIMER, Janaina; FINGER, Ingrid; RIBEIRO, Marina; MALCORRA, Bárbara; HÜBNER, Lilian. Speech as a Graph: Developmental Perspectives on the Organization of Spoken Language. Biological Psychiatry: Cognitive Neuroscience and Neuroimaging, v. 8, n. 10, p. 985–993, 2023. https://doi.org/10.1016/j.bpsc.2023.04.004.
MOTA, Natália Bezerra; WEISSHEIMER, Janaina; MADRUGA, Beatriz; ADAMY, Nery; BUNGE, Silvia A.; COPELLI, Mauro; RIBEIRO, Sidarta. A naturalistic assessment of the organization of children’s memories predicts cognitive functioning and reading ability. Mind, Brain and Education, v. 10, n. 3, p. 184–195, 2016. https://doi.org/10.1111/mbe.12122.
NOBLE, Kimberly; HOUSTON, Suzanne; BRITO, Natalie; BARTSCH, Hauke; KAN, Eric; KUPERMAN, Joshua M.; AKSHOOMOFF, Natacha; AMARAL, David G.; BLOSS, Cinnamon S.; LIBIGER, Ondrej; SCHORK, Nicholas J.; MURRAY, Sarah S.; CASEY, Berdine J.; CHANG, Linda; ERNST, Thomas M.; FRAZIER, Jean A.; GRUEN, Jeffrey R.; KENNEDY, David N.; VAN ZIJL, Peter; MOSTOFSKY, Stewart; KAUFMANN, Walter E.; KENET, Tal; DALE, Anders; JERNIGAN, Terry L.; SOWELL, Elizabeth R. Family income, parental education and brain structure in children and adolescents. Nature Neuroscience, v. 18, n. 5, p. 773–778, 2015. https://doi.org/10.1038/nn.3983.
PEREZ, Reinaldo Cabrera; WEISSHEIMER, Janaina; BEDORE, Lisa; PEÑA, Elizabeth. Longitudinal speech graph analysis of narrative retelling in bilingual Spanish-English children with and without developmental language disorder. SSRN, 2025. Preprint. https://doi.org/10.2139/ssrn.5801342.
TRAXLER, Matthew J. Introduction to psycholinguistics: understanding language science. 2. ed. Oxford: Wiley-Blackwell, 2023.
Review
DOI: https://doi.org/10.25189/2675-4916.2026.V7.N1.ID971.R
Editorial Decision
EDITOR 1: Mirela Cunha Cardoso Ramacciotti
ORCID: https://orcid.org/0000-0002-3109-8376
AFFILIATION: Pontifícia Universidade Católica do Rio Grande do Sul, Rio Grande do Sul, Brasil.
-
EDITOR 2: Caroline Handley
ORCID: https://orcid.org/0009-0007-8435-0778
AFFILIATION: Western Kentucky University, Kentucky, Estados Unidos.
-
ASSESSMENT: The article "Speech Production in Child Language Development" offers a timelyand highly relevant contribution to the intersection of psycholinguistics, networkscience, and education. The study fulfills the main objective of this theme issue,on the Neurobiology of Early Language Development, by addressing fundamentalquestions on early language acquisition and developmental trajectories. They dothat by showing how network-based approaches may complement neuroimagingstudies. As the authors argue, such approaches provide a deeper understanding ofthe structural organization of children’s cognitive and linguistic overt behavior.
The article by Weissheimer and Mota provides a comprehensive review of speechproduction and child language development, introducing Speech Graph Analysisas an innovative, ecological technique for early assessment. The authorseffectively bridge neurobiological perspectives with translational educationalapplications, offering a robust argument for the use of graph-theoretical models topredict literacy and identify linguistic disruptions in both typical and neurodiversepopulations. By positioning speech as a structural network (graph), the studymoves beyond traditional semantic and grammatical metrics to quantify narrativeorganization through long-range connectedness and short-range recurrence.
One of the strongest points of the article is its contextualization within "WILD"(Worldwide, In situ, Local, Diverse) countries, highlighting the necessity for lowcost, scalable assessment tools in regions like Brazil. The integration oflongitudinal data demonstrating the predictive power of graphs for reading skills 3–6 months in advance provides compelling evidence for the technique's utility.
For the aforementioned reasons, we recommend the article for publication in thetheme issue on Neurobiology of Early Language Development.
Rounds of Review
REVIEWER 1: Gutemberg Raposo da Silva Ferreira
ORCID: https://orcid.org/0000-0002-3736-6095
AFFILIATION: Universidade Federal do Tocantins, Tocantins, Brasil.
-
REVIEWER 2: Pietra Da Ros Roig da Silva
ORCID: https://orcid.org/0000-0002-2603-9750
AFFILIATION: Pontifícia Universidade Católica do Rio Grande do Sul, Rio Grande do Sul, Brasil.
-
ROUND 1
REVIEWER 1
30/06/2026 | 18:45
This manuscript introduces an innovative and low-cost way to monitor how children develop speech and reading skills over time. By turning children’s spoken stories into visual maps (graphs) of how words are connected, the authors show that it is possible to predict, months in advance, which children are likely to have stronger or weaker reading and language abilities. In simple terms, children who tell more connected stories, with fewer short and repetitive word sequences, tend to become better readers and develop stronger linguistic skills later on.The potential impact is considerable, especially for educational and clinical settings in low‑resource contexts. Because this method does not require children to read and can be applied in everyday school situations, it may help teachers, speech therapists, psychologists, and policymakers to detect learning risks early and to plan targeted interventions and public policies grounded in empirical data. Overall, the findings are supported by converging evidence from several longitudinal studies with both typically developing and neurodiverse children, which gives the conclusions strong credibility.
General Assessment:This manuscript offers a clear, well-grounded critical perspective on child speech production and language development, with a particular focus on graph-based analysis of children’s narrative speech. The claims and conclusions are supported by a coherent synthesis of prior empirical studies showing that speech connectedness measures can predict later reading and linguistic outcomes. Overall, the piece has strong potential impact for researchers and practitioners interested in early assessment, intervention, and educational policy, especially in contexts where high-cost neuroimaging methods are not feasible.Public Review:The article’s main strength is how it connects core questions in psycholinguistics and developmental science, how children acquire and organize spoken language, with an ecological, school-based assessment technique that can be applied at scale. The manuscript is successful in presenting speech graph analysis in an accessible way for readers who may not be familiar with graph theory, while still preserving the conceptual rigor needed for an academic audience. The explanation of key graph metrics, such as long-range connectedness (e.g., Largest Strongly Connected Component) and short-range recurrence (e.g., repeated edges), helps readers understand what exactly is being measured and why these measures matter for narrative organization.The review of evidence is a second major contribution. By discussing studies in typically developing children alongside work with neurodiverse populations and children with developmental language disorder, the manuscript demonstrates that speech graph analysis is not just an abstract computational tool but a method with practical diagnostic and predictive value. Importantly, the work frames graph-based speech measures as complementary to more traditional linguistic and cognitive assessments, rather than as a replacement, which is a balanced and convincing positioning.A particularly relevant aspect of the manuscript is its attention to feasibility and equity. The discussion of constraints faced by researchers and schools in “WILD” settings (including Brazil) provides important context for why scalable, low-cost approaches are necessary. The paper makes a compelling case that narrative-based protocols can support early screening and tracking of skills linked to literacy development, with possible downstream benefits for interventions and policy design. The conclusion effectively situates this line of research within translational goals, emphasizing how early speech measures can inform educational accommodations and public planning.
-
REVIEWER 2
30/06/2026 | 15:05
The manuscript is well written, theoretically anchored, and methodologically robust. It addresses a highly relevant issue in child speech development by proposing a systematic and analytically rigorous framework for examining recurrence patterns in children’s speech production. The study demonstrates conceptual clarity and internal coherence, aligning its research questions, analytical procedures, and interpretations in a consistent manner.The study holds significant potential impact within the fields of language sciences, language development, and education. By introducing a quantifiable and replicable model for analyzing structural and recurrence patterns in child speech, it enhances methodological precision in speech analysis and expands the analytical toolkit available to researchers. Its language-independent design further increases its relevance, enabling cross-linguistic comparisons and interdisciplinary dialogue. Beyond academia, the proposed approach may contribute to evidence-based clinical assessment and inform educational practices related to language development and emergent literacy.The manuscript will be of interest to scholars in linguistics, psycholinguistics, and cognitive science, as well as to speech-language pathologists and educators working with child language development across neurotypical and neurodivergent populations. Additionally, it may be particularly valuable to researchers investigating early indicators of language delay, narrative development, and literacy acquisition.By offering a scalable, low-cost, and methodologically transparent tool for quantifying recurrence patterns in speech production, the graph node analytical models constitute a meaningful contribution to the field. The findings are supported by systematic and replicable analyses, strengthening the study’s relevance for research on narrative structure and developmental trajectories in child language.
The manuscript would benefit from a few refinements that may further strengthen its clarity, precision, and alignment with current terminology in the field.First, the authors may wish to reconsider the use of the term Specific Language Impairment (SLI). Current consensus in the field increasingly favors the term Developmental Language Disorder (DLD), which is already employed in Section 4 of the manuscript. Ensuring terminological consistency throughout the text would align the manuscript with contemporary standards in language development research. Similarly, the expression “genetic defect” may be considered outdated in current neurodevelopmental discourse. Alternatives such as “genetic variants,” “genetic disorders,” or “genetic influences” would better reflect current terminology and the multifactorial understanding of DLD.Second, greater terminological precision is recommended in the discussion of phonological constructs. In the sentence “Children need to identify how sounds (phonemes) are organized in words as well,” the terms “sounds” and “phonemes” appear to be used interchangeably. However, these terms are not equivalent. “Sounds” generally refer to phones, i.e., the concrete acoustic/articulatory realizations of speech, whereas phonemes are abstract, contrastive units in the phonological system of a language. Although phonemes are realized as phones and the concepts are closely related, they are theoretically distinct. For the sake of terminological precision, I would suggest replacing “sounds (phonemes)” with simply “phonemes”, if the intention is to refer to abstract phonological units in order to ensure conceptual accuracy.Additionally, the manuscript would benefit from a brief discussion of the advantages and limitations of integrating neuroimaging techniques, particularly fNIRS, within the proposed framework. Including a concise reflection potentially at the end of Section 3 would strengthen the methodological transparency of the work and further contextualize its contributions.Overall, these refinements would further enhance the manuscript’s clarity, conceptual precision, and impact, supporting its potential value for both research and applied contexts.
-
ROUND 2
REVIEWER 1
30/06/2026 | 01:45
This manuscript provides a comprehensive review of speech production and child language development, introducing Speech Graph Analysis as an innovative, ecological technique for early assessment. The authors effectively bridge neurobiological perspectives with translational educational applications, offering a robust argument for the use of graph-theoretical models to predict literacy and identify linguistic disruptions in both typical and neurodiverse populations.
The article "Speech Production in Child Language Development" offers a timely and highly relevant contribution to the intersection of psycholinguistics, network science, and education. By positioning speech as a structural network (graph), the study moves beyond traditional semantic and grammatical metrics to quantify narrative organization through long-range connectedness and short-range recurrence. One of the strongest points of the text is its contextualization within "WILD" (Worldwide, In situ, Local, Diverse) countries, highlighting the necessity for low-cost, scalable assessment tools in regions like Brazil. The integration of longitudinal data demonstrating the predictive power of graphs for reading skills 3–6 months in advance provides compelling evidence for the technique's utility. The narrative flow is logical, transitioning from birth-stage auditory processing to advanced discourse-level organization, making the complex topic of graph theory accessible to a broader academic audience.
-
REVIEWER 2
30/06/2026 | 10:31
Dear authors and editor,
The requested revisions have been thoroughly and satisfactorily incorporated into the most recent version of the manuscript. The authors successfully addressed the points raised and improved the overall clarity and quality of the text. I appreciate their willingness, responsiveness, and attention to detail in implementing the suggested changes, which have contributed positively to the manuscript’s development.
While the current version represents a clear improvement, some aspects could still be further refined. In particular, although Sections 1 and 2 present a range of empirical studies on child language development, the manuscript would benefit from a more explicit identification and delineation of the theoretical frameworks under review. Given that the abstract states that the article aims to “briefly review theories of child speech production and development,” a clearer articulation of specific theoretical models would strengthen the conceptual grounding and ensure closer alignment between the stated objectives and the overall structure of the review. This is a relatively minor point and could be considered in future related work.
PDF
