1. Problem: talking about quantities or measuring talk-rate

In early childhood education two assignments circulate that fit in the same sentence and do not describe the same operation. The first asks the adult to talk about quantities, relations, and space while the play or the routine is already under way: to take up the “taller” of a tower, to ask whether there is a cup for each person, to name far and near when a child pushes a car. The second asks for a number about talk: how many mathematical utterances came out per hour, in what percentile the “mathematical language” landed, whether the number talk script was recited in full, whether a detector marked the room as rich or poor. Talking about quantities is an exchange. Measuring talk-rate is bookkeeping. They can occur on the same Monday. Only the first is the craft.

It is worth naming the construct before naming the market. Early mathematical conversation —math talk, mathematical language in interaction, and the variant called number talk when it is reduced to counting aloud— is not a glossary stuck onto the child. It is situated talk: cardinality, comparison, magnitude, spatial relations, part and whole, enough and not enough, said in turns between a child and an adult who can wait, reformulate, and give the word back. Mathematical language as the child’s knowledge and math talk as an exchange are not the same variable. King and Purpura (2021) place mathematical language in the position of a possible mediator between direct numeracy activities at home and numeracy skills; they do not treat it as an hourly rate for surveillance. Confusing lexicon, exchange, and rate is the first confusion in this field.

The second confusion is one of product. Five substitutes are on offer that fit on a slide and do not fit in the block area: a speech detector that returns a math talk rate; a generative system that drafts number talk scripts or a list of closed questions, usable in any room at any hour; a dashboard of utterances per hour, or a percentile of mathematical language, designed so that a coordinator can compare educators; a chatbot or a robot presented as the one who “does math talk” while the adult attends to something else; a fidelity list whose score orders teachers. None of those objects is, by its form, a turn. The leap —from the rate, the script, the dashboard, the substitute, or the ranking to the claim that conversation has already occurred— is not authorized by the literature this review reads.

There is an economy of the shortcut with two mechanisms. The first is displaced authorship: kneeling beside the tower, hearing “more,” and deciding whether the next turn speaks of height, of blocks that are missing, or of whether it fits on the shelf takes attention; reading a generated script, or checking whether the detector raised the rate, takes seconds and leaves a number. The platform offers the second operation under the name of the first. The second is false closure: the dashboard fills when the hour has accumulated utterances; the conversation stays open because the play continues tomorrow and the child’s word has changed. For the detector, finished means counted. For the craft, finished means that quantitative language was extended inside an activity the child was sustaining. This article reconstructs that non-equivalence without estimating an effect of the detector on development: the corpus does not contain that trial and the figure is not manufactured. Empirical finding, framework, and design inference are distinguished. Counting mathematical words can serve a study. It does not occupy the place of the turn.

2. Method: narrative review 2021–2026

A critical narrative review was conducted, not a meta-analysis and not a synthesis with a protocol for a flow of exclusions. The purpose was not to estimate a homogeneous effect of math talk or to combine coefficients from home, classroom, and professional learning —coefficients this text does not invent— but to sustain a category argument: what counts as mathematical conversation in early childhood education and what remains at the ceiling of the artifact. The window is 2021–2026. DOI verification in Crossref corresponds to 2 October 2026. Each source was read for the object it studies: mathematical talk of those who care for children, scaffolding, home numeracy, books, environment intervention, mediation in play, learners of more than one language, beliefs, pedagogical knowledge, feedback, certification examinations, or professional learning.

Inclusion criteria: a DOI verifiable in the list for this kit; relevance to early quantitative language in play, routines, books, or professional learning; the possibility of marking the reading as an empirical finding, a framework, or a design inference. Twenty-four of the sources verified in fuentes.md were used. The cut followed coverage of three contexts —home, classroom, professional learning— and not an effect size. No authors, journals, or DOIs outside that list were added, nor product catalogs or platform reports without an article. Empirical finding means that the work observes, interviews, surveys, or intervenes, and that this review asserts only what the design and the accessible results allow, without reestimating models. Framework means that the work organizes principles and is not a trial of the artifact ceiling. Design inference means a consequence for practice that this article proposes and that must not be read as a validated norm.

When an artifact —a rate detector, a generative script, a dashboard, a chatbot or robot without an adult, a checklist turned into a ranking— lacks, in the corpus, a trial that equates it to dialogic mediation, it is discussed as a category ceiling. No N, d, r, AUC, fidelity percentages, or percentiles of “math language” are invented. The sample figures that appear below are in the works themselves and are attributed. The criteria in section 8 are design hypotheses.

3. State of the art: math talk and mathematical language

It is useful to separate four strata that the market in “automatic math talk” flattens into a single label. The first is the construct: mathematical conversation as an exchange of quantitative, relational, and spatial language, not as a token count. The second is the home: numeracy activities, scaffolding, game format, prompts, the text of the book, the beliefs of the person who cares for the child, and variation across contexts. The third is the classroom: intervention environments, the quality of mediation in play, mathematical activity with the youngest children, books, dialogic reading, and mathematical language with those who are learning in more than one language. The fourth is professional learning: enjoyment and confidence, feedback and knowledge, teachers’ perspectives, certification examinations, beliefs about what the child already brings, and semiotic resources in professional learning. Category inference, not a finding about a detector: a percentile does not observe the tower; a turn with an adult present can.

Mathematical language and math talk touch and do not substitute for each other. King and Purpura (2021) study 125 children aged between 3.12 and 5.26 years, assessed in fall and spring, and ask whether mathematical language mediates the relation between the frequency of direct numeracy activities at home, as reported by families, and numeracy skills. Status: empirical finding from a mediation design. This review does not reestimate the indirect effect. What the design authorizes is a distinction: doing activities at home and having mathematical language available is not the same as displaying a rate of utterances. Ouyang and Chan (2025) examine the interrelations among home numeracy activities, children’s attitudes toward mathematics, and achievement. Status: empirical finding of relations among home practice, affect, and achievement. Inference: a dashboard of utterances can rise while the attitude toward mathematical play goes out, and the corpus does not authorize treating that possibility as if the percentile had measured it.

Elgavi and Hamo (2025) offer principles derived from neuroscience for early childhood educators. Status: framework, not a trial of mathematical conversation and not an evaluation of a device. The list of principles is not reconstructed here beyond what the title declares —principles for educators, not a metric of talk— and they are not converted into a license for an artificial tutor. Payne (2024) tests an intervention named as a math talk learning environment. Status: empirical finding whose object is an environment, not a rate. It is not invented whether the environment “worked” as a coefficient: what is retained is that, in that literature, math talk names conditions of teaching. Turan et al. (2026) study a mathematical-language intervention and the mathematical development of preschoolers who are learning in more than one language. Status: empirical finding with a specified population. Category inference: mathematical language can be the content of an intervention with people; it is not, for that reason, a monolingual count of utterances per hour. The state of the art in this corpus measures relations, mediations, environments, feedback, and beliefs. It does not contain a trial that equates a detector, a generative script, an administrative percentile, a robot without an adult, or a checklist ranking to dialogic mediation. When a ceiling is discussed later, a published null effect will not be being summarized. An equivalence that the market asserts and that these works do not establish will be being refused.

4. Findings: home and caregivers

The home is not a backdrop to mathematical conversation: it is one of the places where quantitative language is extended or thinned, and this block does not treat it as a single rate. Yang et al. (2025) compare the math talk of American and Chinese fathers and mothers during numeracy activities and during routines, and ask whether beliefs matter. Status: empirical finding. It is not used here to rank cultures. The design separates numeracy activity and routine: dressing, eating, or putting away are not noise around the “real” lesson. A script written for the carpet does not see that separation; a single rate per hour erases it. Schnieders and Schuh (2022) take scaffolding, mathematical talk, and game format together. Status: empirical finding. It is not invented which format wins. The consequence comes first: the format conditions which turns fit. A closed bank of questions, generated without seeing the play, treats every activity as if it admitted the same phrases. Scaffolding is a step that is offered and withdrawn; it is not a list crossed off.

Eason et al. (2021) observe 50 dyads of adults and children aged 2 to 4 years in pretend play, and code numerical statements and prompts that invite talk about numbers. The prompts were infrequent. Statements and prompts were uniquely related to the quantity and the diversity of the child’s number words, analyzed separately. Status: empirical finding. It is not reported which coefficient was larger. Design inference: the prompt matters because it opens, for the child, the occasion to advance language inside the play the child is already sustaining, not because it adds a question to the counter. Pretend play is here the setting of the prompt, not the topic of this review. DePascale et al. (2021) separate what a detector would fuse. Spontaneous focus on number is related to the mathematical ability of adults and of children; that focus did not significantly predict math talk; ability was related to talk about more advanced ideas and about numbers in context. Status: empirical finding, and the nuance is the finding. Attending to number is not conversing mathematics. Huang (2025) examines, in play, relations among the adult’s mathematical language, the child’s, and early skills. Status: empirical finding; signs and magnitudes are not reestimated. Play is not decoration for the item: it is the place where both languages are studied together.

Wang et al. (2024) manipulate the text of the stories and take as the outcome the math talk of adults and of children. Status: empirical finding. Direction and size are not invented. The object is retained: the page can be a prompt for the exchange if someone reads with the child. A generator that replaces the adult does not reproduce that condition; it produces a soliloquy in the shape of a question. Melzi et al. (2025) study the math talk of Latino caregivers across contexts and its relation to child outcomes. Status: empirical finding. The contexts are not ordered here from greater to lesser richness. The design forces one to see that this talk is neither a stable trait nor a single percentile: it changes with context. Haktanir and Ivrendi (2026), with 316 families in Turkey, find moderate home support for mathematics and for preparatory literacy during the pandemic, with a positive and moderate correlation between the two, and they compare that support with pre-COVID-19 studies. Status: empirical finding. It is not invented whether support went up or down.

5. Findings: classroom and teaching practice

In the classroom the construct changes hands: it is no longer only the person who cares for the child at home, but a practice that organizes an environment, mediates play, responds to the youngest children, uses books, or sustains reading. Payne (2024) does not describe a microphone. The study tests a math talk learning environment. Status: empirical finding from an intervention. An effect size is not invented. What is retained is that, in that tradition, math talk is an environment the educator sustains with materials, time, and talk. A dashboard can be installed in a poor environment and still mark “activity.” The environment is the condition; the rate is, at most, a residue. Karlsson et al. (2026) study the quality of mediation in play-responsive teaching. They compare mathematics teaching without play and play-responsive teaching: the first proved rich in mathematics made available, through the structured use of materials; in the second, child agency and responsiveness stood out more than the quantity of mathematics offered. They complement each other. Status: qualitative empirical finding. Inference: quality is not a rate. Content is needed and participation is needed, and neither is read off a percentile.

MacDonald (2026) observes, in six Australian services, babies and children aged 2 to 40 months, for up to two days per site. The study describes counting, measuring, locating, designing, playing, and explaining, and scenes in which educators respond to opportunities in everyday life. Status: qualitative empirical finding. Mathematical conversation does not begin on the carpet at age four and is not reduced to numerals. A number talk checklist for counting objects aloud does not see measuring, locating, or explaining with a child who does not yet sustain a long sentence; a detector trained to hear “three” does not either. Zhang et al. (2026) film six teachers in a kindergarten in China and interview them with stimulated recall, within a framework of motive and of everyday and scientific concepts. The activity the child starts alone is not enough: what matters is how conditions are created. Status: qualitative empirical finding. Creating conditions is not reciting a script. Hardy (2025) studies systematic instruction with mathematics books. Status: empirical finding. The book is a medium inside adult instruction, not a file of questions for a device. Ergenekon and Işıkoğlu (2026) present dialogic reading as support for literacy and for early mathematics. Status: an object of turns around a text; an effect is not reestimated. Someone asks, waits, and returns to the page.

Turan et al. (2026) place a mathematical-language intervention with preschoolers who are developing more than one language. Status: empirical finding. A counter calibrated to a single language can classify as silence a quantitative relation said in the other language or half in each. Translanguaging is not reopened here as a topic: what is noted is that the construct, read with this intervention, is not monolingual by nature. A percentile in a single language is not a neutral measure of that room.

An anonymized editorial illustration is in order, not a trial. In an unidentified municipal center, a room of 3- to 5-year-olds, one morning was observed: about forty minutes of block construction and about twenty of snack. A girl says that her tower is “más alta” (taller). The educator kneels, brings her own tower closer, and asks whether the girl’s fits in the window of the shelf or sticks out. She waits. At snack she asks whether there is enough for each person to take a cup. Utterances were not counted and a checklist was not applied.

6. Findings: professional learning and pedagogical knowledge

Professional learning shows why a fidelity ranking does not replace the craft. Gray and Harris (2024) listen to early childhood students in England on enjoyment, confidence, and the experience of teaching mathematics to children under five. They read the interviews with an ecological theory: the school and childhood biography shapes attitude, with consequences for initial preparation. Status: qualitative empirical finding. A checklist score applied to someone who is beginning punishes that biography instead of working with it. Aumann et al. (2026), with 48 educators and 140 children in German kindergartens, find that process feedback was associated with mathematical development; person feedback and measured pedagogical knowledge did not show significant effects in that model. Status: empirical finding. It does not authorize saying that knowledge “does not matter” in general. It authorizes saying that, in mathematics embedded in the everyday, the comment that follows the child’s attempt did work that the knowledge score, there, did not do. A checklist that praises (“you are good at mathematics”) resembles person feedback. One that marks “asked a question” without hearing the process is not process feedback.

Demir (2026) interviews 27 preschool teachers in southeastern Turkey, with children aged 3 to 6 years. All recognized the importance of early mathematics; many described limited pedagogical knowledge, attributed to university preparation, to resources, and to lack of support. Status: qualitative empirical finding. The gap is one of preparation, not of a missing application. A generated script can cover it for a morning and leave it intact the following year. Li and Oppenzato (2026) ask whether certification examinations require mathematics or pedagogical content knowledge for early childhood. Status: an analysis of requirements, not a trial of child achievement. The distribution across states is not invented. If the examination asks for the arithmetic of the person who teaches and not for how one talks about quantities with a four-year-old, the system reproduces the confusion between knowing mathematics and knowing how to converse it.

Papic and Papic (2026) survey 325 educators in classrooms of 3- to 5-year-olds and observe the environment in a subsample of 102. Confidence in pedagogical knowledge and in the capacity to help children learn mathematics was high. More than half did not agree that most children arrive with some mathematical skills. The observed quality of the environment was described as minimal —on the order of 3 on a scale of 7— and its association with beliefs or confidence was weak. Status: empirical finding. Declared confidence is not enough for an environment, and believing that the child arrives without mathematics pushes talk toward interrogation. A fidelity score can be high in a thin room. Quane et al. (2026) report professional learning with key-word signs for early mathematics: 20 signs of position and direction —up, down, through, around— accompanied by the voice. Status: empirical finding about a semiotic resource.

7. Contrast: artifacts that do not converse mathematics

Coding mathematical talk is legitimate research: Eason et al. (2021), DePascale et al. (2021), Yang et al. (2025), Melzi et al. (2025), and Wang et al. (2024) do it, with different designs. The ceiling appears when that count leaves the study and is installed as surveillance, as a substitute for the adult, or as a ranking. In this corpus there is no trial that equates a rate detector, a script generator, a dashboard of utterances per hour, a chatbot or robot without an adult, or a checklist score, to dialogic mediation. They are discussed as a category ceiling. An N of “rooms with a detector” is not manufactured, nor a correlation between percentile and learning.

The detector counts pieces that those works separate: statement and prompt (Eason et al., 2021), spontaneous focus and talk (DePascale et al., 2021), mathematics made available and participation (Karlsson et al., 2026). A rate can rise by force of “how many are there?” without a situated prompt or a response from the child. The generator stumbles on another cut. Wang et al. (2024) treat the text as a condition of the exchange between adult and child, not as a replacement for the adult. Schnieders and Schuh (2022) make talk depend on the format. A closed bank, written before the activity is seen, has the form of mathematical language and does not have contingency. The dashboard and the percentile flatten what Yang et al. (2025), Melzi et al. (2025), Ouyang and Chan (2025), and Haktanir and Ivrendi (2026) separate: routine and activity, contexts, attitude, support at home. An administrative number is not home numeracy.

The chatbot or the robot that “does math talk” without an adult removes the condition the classroom does not let go of: educators’ response (MacDonald, 2026), a sustained environment (Payne, 2024), reading in turns (Ergenekon and Işıkoğlu, 2026), books with instruction (Hardy, 2025), conditions created by teachers (Zhang et al., 2026), language that may not be only one (Turan et al., 2026). Pronouncing numerals does not inherit those findings. The checklist turned into a ranking fails on the side of professional learning. Aumann et al. (2026) separate species of feedback. Papic and Papic (2026) detach confidence from the environment. Demir (2026) and Gray and Harris (2024) place the limit in preparation and in biography. Li and Oppenzato (2026) show that certification can ask for the wrong knowledge. Quane et al. (2026) show a turn that the item “said the question” does not see if it was a sign.

Reading schema. Each item opposes an operation of the craft to an artifact. It is not a scale, it does not order rooms, and it does not summarize an effect.

  1. Situated prompt. The question takes up the play or the routine the child is already sustaining. Against that, a bank of closed questions, generated before the activity is seen.
  2. Turn that extends. The adult takes up the child’s word and adds a quantitative or spatial relation. Against that, a count of utterances per hour.
  3. Adult in the exchange. The person who teaches or who cares for the child sustains the turn. Against that, a chatbot or a robot that “does math talk” without adult mediation.
  4. Situated judgment. One looks at the quality of the exchange and at process feedback, not at declared confidence. Against that, a fidelity checklist turned into a ranking.
  5. Context, not a single number. Talk changes across play, routine, book, and language. Against that, a percentile of mathematical language for surveillance.
Schema, not a fabricated datum. Category ceiling between dialogic mediation and the artifacts that do not converse mathematics. It does not report N, d, r, AUC, or a trial percentile.

The ceiling admits subordinate pieces. A microphone can serve to code prompts, as in Eason et al. (2021).

8. Discussion: criteria for genuine mathematical conversation

The four criteria that follow are a design inference of this article. They are not a validated norm or an inspection standard, and they must not be printed as the checklist the previous section refuses. If a coordination uses them to rank rooms, it repeats the error under another letterhead. They are anchored in home, classroom, and professional learning, and they are declared as hypotheses.

Prompts situated in play and in routines. Eason et al. (2021) show that the question is not magic: statements and prompts are related to children’s numerical language, and prompts are scarce. The occasion matters. Yang et al. (2025) allow that occasion to be a routine —dressing, putting away— and not only a mathematics corner. Schnieders and Schuh (2022) require that the question fit the format of the play. Wang et al. (2024) make the page a condition of the exchange, not a teleprompter. In the illustration with 3- to 5-year-olds, “does it fit or does it stick out?” is situated because the tower was already there. Hypothesis: the referent of the prompt is the child’s action in course. A script usable in any room, at any minute, with any material, fails even if it contains a number. Generating ten variants does not situate it.

Turns that extend quantitative language. A mathematical word is not enough. To extend is to move from “more” to a relation the child can take up again: taller, it fits, there is enough for each person, around, farther. Karlsson et al. (2026) prevent a choice between content and participation: structured materials make mathematics available; play-responsive teaching puts agency in play. The turn that extends does both things at the scale of one exchange. Aumann et al. (2026) bring it close to process feedback, not to the label placed on the person. MacDonald (2026) recalls that, with the youngest children, extending can be measuring, locating, or explaining. Quane et al. (2026) recall that it can go in the hand, where the word counter does not look. Hypothesis: a loop of “how many are there?” raises a rate and may extend nothing. Extension is recognized in the child’s next turn. It is an observational hypothesis, not a rubric with points.

Coherence between home and classroom, without surveilling families. Melzi et al. (2025) make context a variable of the talk of the person who cares for the child. Haktanir and Ivrendi (2026) describe support, not a score of homes. Ouyang and Chan (2025) include attitude. King and Purpura (2021) ask about language as a mediator of activities, not about minutes in an application. Payne (2024) places the environment on the classroom side. Coherence is not the same script in the kitchen and on the carpet. It is that each adult can recognize the quantitative language of the other place without turning it into a percentile. Hypothesis: someone who understands that the wording of a story conditions talk (Wang et al., 2024) can lend a book. They cannot, with this corpus, send the family a percentile of mathematical language.

Do not replace mediation by the rate or by the dashboard. The artifacts in the box fail together when the number, the script, or the device occupies the place of the turn. Papic and Papic (2026) separate confidence and environment, and show that not believing the child arrives with mathematics changes talk. Li and Oppenzato (2026) show an examination that can ask for the wrong knowledge. Elgavi and Hamo (2025) do not offer a neuroscientific shortcut to the artificial tutor. Demir (2026) and Gray and Harris (2024) place the work in preparation and in biography. Hypothesis: there is mathematical conversation when an adult who is present extends quantitative language in play or routine, with contingency, and the institution does not replace that judgment with a rate.

Four tensions block the slogan. Counting prompts in order to understand them (Eason et al., 2021) is not surveillance of educators: research coding and an administrative metric are not the same use.

9. Limits

This review is narrative. It does not combine effects and it does not produce a flow of exclusions with numbers of records. No d, r, AUC, fidelity percentages, or percentiles are invented. The figures that are attributed —125 children in King and Purpura (2021); 50 dyads in Eason et al. (2021); 48 educators and 140 children in Aumann et al. (2026); 27 teachers in Demir (2026); 316 families in Haktanir and Ivrendi (2026); 325 educators and a subsample of 102 in Papic and Papic (2026); six services and ages from 2 to 40 months in MacDonald (2026); six teachers in Zhang et al. (2026); 20 signs in Quane et al. (2026)— come from those works. Where the result is described, the description follows the reading of those texts and not a reestimation. Where the result is not restated —the direction of the effect in Wang et al. (2024), Payne (2024), Hardy (2025), Turan et al. (2026), Huang (2025), Ouyang and Chan (2025), Yang et al. (2025), and Schnieders and Schuh (2022), and the sense of the pre-pandemic comparison in Haktanir and Ivrendi (2026)— it is because this review stays with the object of the design and does not manufacture the sign of the coefficient.

The samples are situated: Germany, Turkey, Australia, China, Latino caregivers, preschoolers who are learning more than one language, students in England, certification requirements in a system of states. There is no universal effect of math talk. Several 2026 texts are cited by the year of print publication when it diverges from the year of online release, and without a volume when the record does not carry one. The morning of the tower and the snack is an anonymized editorial illustration: it is not a datum, it has no N, and it must not be cited as a study. The artifacts that are criticized were not evaluated as commercial products. Elgavi and Hamo (2025) were read as a framework, without reconstructing their principles one by one. The activity framework used by MacDonald (2026) is not turned into an added reference. This text is not a review of pedagogical documentation, nor of translanguaging, nor of role play, nor of open-ended materials, nor of risk in play. Those objects have another cut. The criteria in section 8 are design hypotheses. They do not authorize an inspection or a new score.

10. Conclusions

A math talk rate detector, a generator of number talk scripts or of closed questions, a dashboard of utterances per hour or a percentile of mathematical language for surveillance, a chatbot or a robot that “does math talk” without an adult, and a fidelity checklist turned into a teacher ranking do not constitute support for mathematical conversation in early childhood education. The craft is the dialogic mediation of the person who teaches and of the person who cares for the child, able to extend quantitative language within play and within routines. The home shows it as beliefs, format, prompts, text, context, and support, not as a rate. The classroom shows it as environment, quality of mediation, response to everyday mathematics, books, and language with those who are learning in more than one language. Professional learning shows it as biography, process feedback, pedagogical knowledge, certification, beliefs, and multimodal resources.

Where the works do not measure a detector, this article does not claim that they have refuted it with a coefficient. Where they measure prompts, mediation, feedback, environment, or support, it does not translate them into “the room already converses mathematics” because a number went up. The criteria —situated prompt, turn that extends, coherence between home and classroom without surveillance, refusal to replace mediation by the rate— are design hypotheses for reading a practice or a product. They are not a norm. Turning them into a ranking would be a return to the artifact through the door of the discussion. Talking about quantities, with an adult who hears and gives the word back, is the craft. Measuring talk-rate can be a method of study. It is not the conversation, and it must not be presented as what it is not.

Laboratorio Editorial de NEXTECH.IA / Ingeniero Mitre.

References

  1. Aumann, L., Puca, R. M., y Gasteiger, H. (2026). Effects of early childhood teacher knowledge and feedback on children’s early mathematical development in German kindergartens. Early Childhood Education Journal, 54(2), 619–629. https://doi.org/10.1007/s10643-025-01859-1
  2. DePascale, M., Prather, R., y Ramani, G. B. (2021). Parent and child spontaneous focus on number, mathematical abilities, and mathematical talk during play activities. Cognitive Development, 59, Article 101076. https://doi.org/10.1016/j.cogdev.2021.101076
  3. Demir, M. (2026). Preschool teachers’ pedagogical knowledge, awareness and perspectives of early childhood mathematics education. Early Years, 46(3), 506–522. https://doi.org/10.1080/09575146.2025.2524616
  4. Eason, S. H., Nelson, A. E., Dearing, E., y Levine, S. C. (2021). Facilitating young children’s numeracy talk in play: The role of parent prompts. Journal of Experimental Child Psychology, 207, Article 105124. https://doi.org/10.1016/j.jecp.2021.105124
  5. Elgavi, O., y Hamo, P. (2025). Math on the brain: Seven principles from neuroscience for early childhood educators. Early Childhood Education Journal, 53(8), 2837–2848. https://doi.org/10.1007/s10643-024-01656-2
  6. Ergenekon, E., e Işıkoğlu, N. (2026). Dialogic reading: A way to support early literacy and math skills in preschoolers. Early Childhood Education Journal. https://doi.org/10.1007/s10643-026-02358-7
  7. Gray, L., y Harris, M. (2024). Enjoyment, confidence, and broader experiences of teaching mathematics to the under-fives: Accounts from students of Early Childhood Studies. European Early Childhood Education Research Journal, 32(5), 895–908. https://doi.org/10.1080/1350293x.2024.2331219
  8. Haktanir, H., e Ivrendi, A. (2026). Home support for mathematics and early literacy activities in early childhood period: Comparison with pre-COVID-19 findings. Early Years, 46(3), 599–612. https://doi.org/10.1080/09575146.2025.2552411
  9. Hardy, J. K. (2025). The effects of systematic instruction with math books to teach math to preschoolers. Early Childhood Education Journal, 53(5), 1577–1589. https://doi.org/10.1007/s10643-024-01686-w
  10. Huang, H.-H. (2025). Math talk in play contexts: Relations between parent and child math language and early math skills. Early Childhood Education Journal, 53(7), 2719–2729. https://doi.org/10.1007/s10643-024-01783-w
  11. Karlsson, L., Palmér, H., y Carlsen, M. (2026). Quality of preschool teachers’ mediation of early years mathematics in play-responsive teaching activities. Educational Studies in Mathematics, 122(2), 251–270. https://doi.org/10.1007/s10649-025-10470-5
  12. King, Y. A., y Purpura, D. J. (2021). Direct numeracy activities and early math skills: Math language as a mediator. Early Childhood Research Quarterly, 54, 252–259. https://doi.org/10.1016/j.ecresq.2020.09.012
  13. Li, X., y Oppenzato, C. (2026). Math or math pedagogical content knowledge for early childhood teachers? The math requirements on certification exams across the States. Early Years, 46(3), 537–550. https://doi.org/10.1080/09575146.2025.2529908
  14. MacDonald, A. (2026). Everyday mathematical activities of babies and toddlers in early childhood education settings. Early Childhood Education Journal, 54(3), 1041–1051. https://doi.org/10.1007/s10643-025-01899-7
  15. Melzi, G., Soria, P. M., y Mesalles, V. (2025). Latine caregiver math talk across contexts and its relation to child math outcomes. Early Childhood Research Quarterly, 72, 133–142. https://doi.org/10.1016/j.ecresq.2025.02.013
  16. Ouyang, X., y Chan, W. W. L. (2025). The interrelations between home numeracy activities, children’s attitudes toward mathematics, and their mathematics achievement. Early Childhood Research Quarterly, 70, 19–29. https://doi.org/10.1016/j.ecresq.2024.08.008
  17. Papic, M. M., y Papic, C. (2026). Is confidence in mathematics pedagogy enough? Exploring early childhood teachers’ mathematics beliefs and confidence. Early Childhood Education Journal, 54(2), 675–691. https://doi.org/10.1007/s10643-024-01840-4
  18. Payne, T. (2024). The math talk learning environment: Testing an early childhood math intervention. Early Childhood Research Quarterly, 66, 224–233. https://doi.org/10.1016/j.ecresq.2023.10.012
  19. Quane, K., Trewartha, B., Gaunt, L., Wadsworth, Z., y Stigwood, E. (2026). Introducing the use of KWS in mathematics in early childhood: Insights from professional learning. Early Childhood Education Journal. https://doi.org/10.1007/s10643-026-02249-x
  20. Schnieders, J. Z.-Y., y Schuh, K. L. (2022). Parent-child interactions in numeracy activities: Parental scaffolding, mathematical talk, and game format. Early Childhood Research Quarterly, 59, 44–55. https://doi.org/10.1016/j.ecresq.2021.10.004
  21. Turan, E., Van den Broek, D., Peters, E., De Smedt, B., y Torbeyns, J. (2026). Effects of a mathematical language intervention on mathematical development in preschool dual language learners. Learning and Instruction, 102, Article 102269. https://doi.org/10.1016/j.learninstruc.2025.102269
  22. Wang, M., Vasilyeva, M., y Laski, E. V. (2024). Words matter: Effect of manipulating storybook texts on parent and child math talk. Early Childhood Research Quarterly, 69, 65–77. https://doi.org/10.1016/j.ecresq.2024.07.002
  23. Yang, Q. T., Jeffries, I., Xie, Z., Star, J. R., Harris, P. L., y Rowe, M. L. (2025). American and Chinese parents’ math talk during numeracy and routine activities: Do parental beliefs matter? Early Childhood Research Quarterly, 72, 284–294. https://doi.org/10.1016/j.ecresq.2025.04.002
  24. Zhang, W., Li, L., y Disney, L. (2026). How kindergarten teachers in China create conditions for children’s mathematics concept learning. Early Years, 46(3), 424–442. https://doi.org/10.1080/09575146.2025.2499874