Early Language Access and STEAM Education: Keys to Optimal Outcomes for Deaf and Hard of Hearing Students

Abstract

This paper offers an overview of a large study of language and cognitive development in deaf and hard of hearing children. Specifically, we investigated how acquiring a signed or spoken language (language modality) and when a child’s access to language begins (i.e., at birth or later in development) influence cognitive development. We conducted in-person behavioral assessments with 404 children 3–10 years old (280 deaf and hard of hearing; 124 typically hearing). The tasks measured a range of abilities along a continuum of how strongly they depend on language input, such as general vocabulary and number words (strongly dependent) vs. skills such as tracking sets of two to three objects and standardized ‘nonverbal’ picture-similarity tasks (relatively independent of language). Overall, the timing of children’s access to language predicted more variability in their performance than language modality. These findings help refine our theories about how language influences development and suggest how a STEAM pedagogical approach may ameliorate the impacts of later access to language. These results underscore children’s need for language early in development. That is, deaf and hard of hearing children must receive fully accessible language input as early as possible through sign language, accompanied by hearing technology aimed at improving access to spoken language, if desired.

1. Introduction

Encompassing and integrating the domains of Science, Technology, Engineering, Arts, and Mathematics (STEAM) education is receiving increased attention, given its potential to enhance student engagement and learning (e.g., Lin & Tsai, 2021; Perignat & Katz-Buonincontro, 2019). STEAM is envisioned to enrich traditional STEM education, to empower teachers from different areas to collaborate to develop interdisciplinary curricula, and to creatively and effectively address many serious and complex societal problems (Taylor, 2016). Myriad definitions for STEAM exist. Perignat and Katz-Buonincontro (2019) conducted an integrative review of 44 STEAM studies and categorized them into empirical, practical, and pedagogical. They identified various purposes for STEAM approaches, as well as a range of multi-, trans-, and interdisciplinary criteria used in the literature. Their analysis also revealed a consensus view that creativity and innovation are major components of STEAM, though they found no agreed-upon measures of creativity and a lack of learning outcomes for areas related to problem-solving and arts education.

1.1. The Promise of STEAM for Deaf and Hard of Hearing Learners

Our perspective on STEAM is a pedagogical one, though not in terms of a specific pedagogy like the one proposed by Lin and Tsai (2021). We view STEAM as an integrated approach to education and learning that sparks student curiosity and invites learners to make connections across varied disciplines. In this paper we will draw connections between children’s access to language and the impact that effective access to language has on the development of domains that are usually thought to depend on language, as well as those that are generally considered to depend less on language, as described below. To foreshadow our findings and interpretations of the results, we propose that an integrated, interdisciplinary STEAM pedagogical approach could be a way to ameliorate the negative consequences that affect deaf and hard of hearing children who have not experienced early and full access to language. We explore possible classroom practices that could be informed by these findings in the context of an integrated and interdisciplinary STEAM perspective (e.g., Wolbers et al., 2020; Dostal & Wolbers, 2014; Nunes & Moreno, 2002; Monson et al., 2021; Clerc Center, 2024; Gardiner-Walsh, 2024). Finally, in line with a number of other researchers (e.g., W. C. Hall, 2017; Clark et al., 2020; Napoli et al., 2015; Zhang et al., 2024; Humphries et al., 2024), we conclude that early access to language supports age-appropriate development in deaf and hard of hearing children, in all domains.

We provide an overview of a large research project studying the linguistic and cognitive development of deaf, hard of hearing, and hearing children. This project, among the largest of its kind, focused on mathematics but also studied related skills that touch on the other STEM disciplines as well as other components of STEAM education. For example, tasks that strongly depend on language include assessments of vocabulary, which contributes to children’s ability to think critically and reason about arguments (Dunn & Dunn, 2007; Adams Costa et al., 2019). Examples of skills that are thought to depend less on language include nonverbal IQ, here measured via a picture similarity task (Elliott, 2007). This task assesses children’s ability to attend to visually presented objects and patterns and make inferences and category judgments about them. This set of skills is strongly implicated in the Arts component of the STEAM disciplines. An example of a cross-cutting set of skills, executive functioning, is measured via parent reports using the Behavior Rating Inventory of Executive Function—Preschool Version (known as the BRIEF-P, Gioia et al., 2003). Executive functions include a child’s ability to plan and organize their activities to achieve a goal and to inhibit irrelevant activities, as well as emotional self-regulation and cognitive flexibility, and have been shown to be influenced by children’s language experience (Botting et al., 2017; M. L. Hall et al., 2019; Kronenberger & Pisoni, 2020). Thus, while the focus of the study is on language and mathematics, the project investigated a wide range of skills and abilities that are pertinent to all of the STEAM areas.

1.2. The Importance of Early Experiences

Our approach to understanding how language and cognition (and ultimately, education) are related is grounded in a developmental psychology framework. This perspective holds that children’s early experiences are powerful, though not necessarily deterministic, influences on their development. The groundbreaking longitudinal study of every child born on the island of Kauai in 1955 conducted by Werner and Smith (2001) is a paragon example. For each of the 698 children born that year, the researchers documented the birth conditions as well as the home environments and outcomes for over 40 years. Despite documenting negative birth experiences (e.g., lack of oxygen, difficult birth) and unsupportive home environments (e.g., neglect, abuse) for a subset of the children, approximately one-third of those children reached young adulthood without demonstrating significant negative outcomes (e.g., mental health issues, encounters with police/the law). Werner and Smith documented many characteristics of the children that lead to this resilience, which underscored their conclusion that even accumulating many negative experiences early in life (now termed adverse childhood experiences) does not necessarily determine a child’s outcome. We emphasize children’s resilience and the potential for a STEAM pedagogical approach that recognizes and centers deaf and hard of hearing children’s experiences to support their learning.

1.3. Relationships Between Language and Cognition

As mentioned earlier, the research project takes as centrally important early access to language: language facilitates concept building and knowledge acquisition. Here we describe a few ways this relationship has been thought about in the literature and potential hypotheses and mechanisms. The first is a proposal, summarized in Gentner (2016) that proposes that language functions as a kind of internal “tool kit” that enhances our cognitive abilities. Gentner puts forth specific processes by which language achieves this, including analogical comparison and abstraction. Analogical comparison allows children to “structurally align” language and concepts; a child hearing relational language (e.g., “top” or “middle” to describe a shelf in a vertical box) will use analogical comparison to create a structural alignment that spatially maps the relational meaning to the part of the object being described (this task is described in more detail in a later section). A second type of relationship proposed in the literature is that language facilitates the child’s processing and/or manipulation of information, as characterized in Holmer et al.’s (2016) Developmental Ease of Language Understanding (D-ELU) model, an extension of their earlier model based on adults. The model essentially holds that ease of access to linguistic/lexical representations (for example, the lexical items in the count sequence) will influence the ways in which linguistic information can be harnessed to support cognitive functions, such as working memory and short-term memory. These systems are critical for the temporary storage and manipulation of information needed for complex cognitive and language tasks, including comprehension, learning, and reasoning (Baddeley, 2003).

A third view holds that specific components of language itself provide symbols that are necessary to form certain kinds of concepts. This view is instantiated in the semantic bootstrapping hypothesis put forth by Carey (2009) to account for the origins of children’s concepts of quantities larger than four. We will describe this view in more detail in a later section, as it directly relates to the research project and set of studies we will present here. Briefly, Carey (2009) hypothesizes that the exact representation of small sets of quantities (less than or equal to three) and approximate representation of large sets (greater than three) do not require language. However, precisely representing large quantities (that is, four or more), does require language. Specifically, Carey suggests that language related to numbers, specifically a count sequence, is necessary to develop representations of quantities larger than four. Difficulty representing quantities larger than four poses clear barriers to achieving competence in even basic arithmetic and mathematical skills. Because typically hearing children universally learn the count sequence with relatively little variability, it is difficult to gain traction with this question by only studying hearing children. We turn now to a characterization of the language development of deaf and hard of hearing children, where the variability in their experience offers insight into this theoretical question, as well as potential guidance for optimizing their development.

1.4. Early Language Experience Is Important Regardless of Language Modality

The timing and quality of access to both spoken and signed language input can be variable for deaf and hard of hearing children who are born to hearing parents who do not (yet) sign. Hearing technology has the potential to increase children’s access to spoken language, though there is always a period of time early in development during which an infant or young child will lack access to the surrounding spoken language, that is, the period before hearing aids are fitted or the period prior to activation of a cochlear implant or other hearing technology. Much research (reviewed in, for example, Werker & Gervain, 2013) has documented the early language acquisition of typically hearing infants. Even before they produce spoken language, hearing infants have learned a great deal about the rhythm of their language (prosody), which sounds and contrasts are part of their language (phonology), which sounds can be combined (phonotactics), as well as recognizing a large number of high-frequency words. Children who are exposed to a sign language from birth acquire visual language on a timetable that is very comparable to that shown by typically hearing children acquiring a spoken language, if not earlier and faster (e.g., Lillo-Martin & Henner, 2021; Newport & Meier, 1985). That is, children exposed to language in either modality from birth show parallels in the ages at which they reach the milestones of babbling, first words, a 10-word vocabulary, and producing two-word utterances.

A robust literature demonstrates that, regardless of modality, later access to language is associated with lower language proficiency. Overall, children who receive cochlear implants show lower performance on spoken language measures relative to their typically hearing peers, as well as highly variable outcomes. While earlier ages of implantation are generally associated with better spoken language outcomes, most early-implanted children’s expressive and receptive language does not reach the average level of performance observed in typically hearing peers (e.g., Niparko et al., 2010; Dettman et al., 2016). Later access to sign language leads to similar negative impacts and is associated with decrements in recognizing individual signs (Mayberry & Fischer, 1989) as well as in understanding simple (Cheng & Mayberry, 2025; Cheng et al., 2023) and complex sentences (Mayberry et al., 2024). This evidence also demonstrates that these negative effects persist into adulthood, even after decades of daily sign language use. Their work further shows that later access to a first language leads to changes in brain development and structure (Cheng et al., 2019). The corresponding literature on the long-term effects of later access to spoken language is much smaller. In this body of work, “long-term” often refers to participants who received cochlear implants in early childhood and who are tested when they are adolescents. One exception is a study by Moura et al. (2023) that studied cochlear implant users who were implanted before or after the age of 3 years. Testing the participants 25 years after they received their implants, the researchers found that individuals who had been implanted earlier had better spoken language outcomes than did the individuals who were implanted after the age of 3 years. However, no comparisons to typically hearing adults or to standard scores were reported. Thus, it remains unclear what the level of performance was for even the higher-performing, earlier-implanted group.

1.5. Researchers Must Consider Early Language Experience

Many studies report that deaf (and hard of hearing, for some studies) children show lower performance in other STEAM-relevant domains in addition to language, covering a range of skills such as mathematics, implicit sequential pattern learning (e.g., Conway et al., 2009), and executive functioning (e.g., Hintermair, 2013). A detailed line of work conducted by Pagliaro and Kritzer (2013) shows that differences between deaf, hard of hearing, and hearing children in basic mathematics performance can already be observed in the preschool years. As mentioned, most studies that compare the number understanding of deaf and hard of hearing with typically hearing children report delays. However, Zarfaty et al. (2004) found that preschool-aged deaf and hard of hearing children performed as well as or better than typically hearing children in recreating sets of two to four objects.

Given the earlier discussion of the persistent negative impact of later access to language on language processing and understanding, we argue that such persistent impacts are likely not limited to the linguistic domain. For example, Kelly and Mousley (2001) compared deaf and hearing college students on the same math problems presented in two formats: graphically (i.e., with minimal language content) and as language-based word problems. The groups performed similarly on the problems presented in the graphic format and on the least difficult word problems. However, the hearing group outperformed the deaf group on the more difficult word problems, despite the deaf students possessing reading levels that were adequate for comprehending the language of the word problems. The authors noted two other differences between the groups: first, that the deaf students made more errors than the hearing students, and that only students in the deaf group left word problems blank. Their interpretation of these differences invoked multiple factors, including deaf students’ lack of engagement or confidence and differing levels of concentration, that may have led to the increased error rate.

Bull et al. (2017) compared the acuity of magnitude representation in deaf, hard of hearing, and hearing students between the ages of 5 and 13 years of age. This acuity indexes an individual’s sensitivity to differences between the quantities of two sets of objects. That is, an individual who can reliably indicate that a set of 15 objects is greater than a set of 10 objects has greater acuity than an individual who can only reliably distinguish between sets containing 20 objects vs. 10 objects. Bull and colleagues found that the hearing students outperformed the deaf and hard of hearing students, showing greater sensitivity to these set size differences. While the authors in both studies noted the age and hearing levels of the deaf and hard of hearing students (Bull additionally reports age of onset of deafness and preference for signed/spoken communication), neither study included any information about the participants’ early language experiences. Thus, we cannot know what role this factor may have played in the deaf and hard of hearing students’ performance.

Other studies have examined the relationship between language access, language production, and cognitive functions in deaf children and adults in the areas of spatial reasoning and spatial orientation, finding that their early language experience (or relative lack of access to language) directly related to their spatial performance. Gentner et al. (2013) compared deaf homesigning and hearing children in Turkey on a non-linguistic spatial mapping task. The children were presented with a box containing three shelves, with an attractive “winner” card placed on one of the shelves—either the top, middle, or bottom shelf. The child was then shown an identical box with three shelves (in which the cards were not visible) and had to indicate which shelf contained the “winner” card. “Top,” “middle” and “bottom” are examples of relational terms. Such terms form part of the cognitive ‘toolkit’ proposed by Gentner (2016) that invite children to engage in analogical reasoning. In this case spatial alignment allows children to form a spatial mapping between the language and the object being discussed and to select the correct shelf. However, the homesigners were deaf children who had not acquired spoken language and had also not been exposed to sign language. The researchers also found that the deaf children had not developed their own gestures to express such spatial relationships. They significantly underperformed the hearing children. Further supporting the role of language in this task is the result that hearing children who received spatial language in their instructions outperformed hearing children who did not, even after a delay of two days, when they were asked to play the same game again, but without instructions given the second time. Apparently the structural alignment they achieved during their first encounter with the task and instructions persisted after the initial testing session.

Pyers et al. (2010) found a similar relationship between production of spatial language and performance on a spatially guided search task among adult signers of Lengua de Señas Nicaragüense (LSN, or Nicaraguan Sign Language). Signers who produced spatial language when describing the location of an object in a room relative to a colored wall (i.e., the key is to the left of the red wall) were able to find the object even after they had been disoriented. In contrast, signers who did not use spatial language to describe the location of the object were not able to choose the correct location of the object they had previously seen hidden before they were disoriented by being blindfolded and turned around by the researchers. This latter example is particularly revealing because all of the participants were deaf and used LSN from similarly early ages. Indeed, some of the signers served as language models for other signers. Thus we can be more confident in pointing to the differences in participants’ spatial language to explain their differential performance on the search task.

Importantly, differences in performance between deaf and hard of hearing children and typically hearing children are often attributed to their deafness, rather than to their variable experiences with language (see Gottardis et al., 2011; Santos & Cordes, 2022 for reviews and discussion). However, these discrepancies in outcomes are not inevitable, and indeed disappear when the methodological shortcomings of prior research are effectively addressed in research design and in interpretation of the results. These shortcomings include failing to recognize that many deaf and hard of hearing children experience language environments that do not offer adequate support. Indeed, previous research (M. L. Hall et al., 2018; Marshall et al., 2015) that takes into account the language experiences of deaf and hard of hearing children found that those who were exposed to a sign language from birth by their deaf or hard of hearing parents demonstrated similar executive functioning skills as typically hearing age peers. Similarly, researchers who are sensitive to the cultural milieu of deaf and hard of hearing children will interpret data in a way that values and respects the resilience and experiences of deaf and hard of hearing children. Henner et al. (2021) provide an excellent example of this contextualization in their examination of the relationship between children’s ASL abilities and their scores on the Northwest Evaluation Association: Measures of Academic Progress (NWEA MAP): Mathematics Test. Holcomb et al. (2024) offer a rich exploration of these issues from the perspective of multiple deaf co-authors, discouraging researchers from unquestioningly accepting the performance of monolingual, higher-SES, typically hearing white children as “the norm”.

1.6. Early Experiences Influence Development Even When Children’s Behavior Is Rudimentary

Early experiences, even in infancy, can influence development in other areas as well. For example, Slaughter et al. (2011) found very early evidence for sensitivity to the concept of one-to-one correspondence, that is, tagging an object (usually via a point or touch) while producing counting words in the correct sequence (e.g., the words “one”, “two”, “three”, etc.) in English. Importantly, when applying the counting procedure, each item in a set can be tagged only once with a point or a count word. The researchers found that typically hearing 18-month-old infants in Japan preferred to look at videos in which a woman’s hand pointed to each of six fish accompanied by the appropriate Japanese counting words over the same videos accompanied by English words (which the infants did not know) or beeps. They conclude that infants extract abstract knowledge from their experiences even when their own productions are sparse or nonexistent (18-month-old infants typically cannot produce many number words). These kinds of findings support the inferences we and other researchers make about the impact of early language experiences, even when we only test children when they are older.

1.7. Systems for Representing Quantity

While the set of skills assessed in the project are broad, many of them center on how children represent quantity and the role of language experience in developing these representations. Therefore, in this section, we provide background information about the comparative (cross-species) and developmental psychology research on how quantity is represented in human and non-human animals and the role of language in each type of representation. The literature supports the existence of two subsystems for representing quantity in human infants, each with its own limitations (Mou & Vanmarle, 2014). The first is known as the Object Tracking System, which allows individuals to precisely distinguish quantities up to four, based on a perceptual system (i.e., without counting, which is a language-based tool). The limitation of this subsystem is its relatively small upper limit (four). The second is the Approximate Number System, which allows individuals to approximately distinguish sets of objects larger than four; the acuity of this system depends on the ratio of the numbers of objects in each set, and on the child’s age. Infants can reliably distinguish sets with a ratio of 1:2 (e.g., 7 vs. 14 objects) and individuals’ acuity improves over development, with adults able to distinguish sets with a 7:8 ratio (e.g., sets of 14 vs. 16 objects) (Halberda & Feigenson, 2008; Xu & Spelke, 2000). Here the limitation is in the approximate, rather than precise or exact, nature of the representation. Demonstrations that humans and a wide range of non-human animals tend to represent and use the Approximate Number System (also known as the Analogue Magnitude System) in similar ways (Gallistel, 1990; Piazza, 2010) suggest that this system does not depend on language.

Not covered by these two subsystems is a way to precisely represent quantities larger than four. Carey (2009) proposed that such representations require the acquisition of a sequence of ordered symbols in which each symbol is associated with a quantity (e.g., “six”), and the next symbol in the sequence refers to a quantity that is one more than the preceding symbol (e.g., “seven”). Because this system is based on acquiring a natural human language with a counting sequence, it is limited to humans and only to those humans whose languages feature counting sequences that refer to the natural numbers. Thus this criterion excludes speakers of some languages: Pirahã (Gordon, 2004) and Mundurukú (Pica et al., 2004) have sparse numeral vocabulary; that is, they do not have words that correspond to the meanings of numbers such as 2, 3, 4, etc. Correspondingly, speakers of such languages have been shown to struggle to exactly represent sets larger than three or four items (see citations above).

1.8. Overview of the Study of Language and Math

The current work offers an overview and synthesis of one of the largest studies of language and cognitive development in deaf and hard of hearing children in the US, with attention to numerical cognition and the various skills that have been found to affect its development. Of particular interest to educators of deaf and hard of hearing children, we focus on the issue of disentangling the many factors that have been conflated in prior work with this population. Specifically, our research questions are the following: How do (1) language modality (i.e., whether the language a child is learning is signed or spoken) and (2) the timing of a child’s language experience (i.e., whether language access begins at birth or some point later in development) influence performance in several domains of cognitive development. Our research design, described in more detail below, allows us to separate the effects of hearing level, use of a sign or spoken language, and the timing of a child’s access to language. This design represents a major contribution to the literature, because many or all of these factors have been conflated in prior studies with deaf and hard of hearing children.

We conducted a wide range of tasks to assess children’s skills in various domains; in the “Tasks” section, we will briefly describe each skill and how it was assessed. Some of these abilities are widely agreed to depend on language input, such as general vocabulary. Similarly, acquiring the meanings of the words and signs in the count sequence, that is, understanding that the sign for ‘seven’ refers to exactly seven items, also depends on language. Executive functioning is a more general set of skills; however, these have also been shown to be influenced by language experience in early childhood. Two skills that rely on understanding how counting leads to identifying a property of a set (the cardinality principle) and mapping among different ways of representing quantity, are also language-dependent. In contrast, other tasks measure skills that have been considered to be independent of language, such as nonverbally tracking small sets of objects (those containing two or three items), deciding which of two sets of objects is larger, and standardized ‘nonverbal’ picture-matching tasks. Figure 1 summarizes the tasks and lists them in decreasing order based on how strongly they are thought to depend on language.

The Research Team and Approach

The research team was led by a hearing coda who is a developmental psychologist and sign language linguist; the team included both deaf and hearing team members. The research coordinator who oversaw all recruitment efforts was a fluent hearing signer who came from a deaf signing family. The tasks and instructions were chosen, adapted, or developed by the principal investigator (coda) in conjunction with deaf and hearing graduate students and research assistants. Importantly, deaf team members who are proficient signers interacted directly with all of the signing deaf and hard of hearing children who participated, giving them instructions, answering their questions, and encouraging them in the tasks. Hearing team members who could sign served as camera operators and in support roles. No interpreters were used in any stage of the project.

2. Materials and Methods

2.1. Participants

This project assessed how language experiences influence children’s number knowledge. Specifically, by studying deaf, hard of hearing, and hearing children this project aimed to understand the effects of language modality (signed or spoken language) and the timing of language exposure (early or later) on children’s number cognition. The larger project was divided into two parts, Project One and Project Two, each focusing on different age groups and assessing various aspects of number knowledge. Project 1 focused on the developmental trajectory of the acquisition of the count word/signs and their meanings by young children ages 3–7 (Figure 2). Specifically, we tested their knowledge of the count sequence, magnitude approximation, object tracking, and mastery of the cardinality principle—that is, the understanding that the last number in a count represents the total number of items in the set. While these studies tend to be conducted with preschool-age children, we extended the age range due to the pervasiveness of later access to language experienced by deaf and hard of hearing children. Project 2 focused on how children aged 5–10 learn how to ‘translate’ among the different representational formats used for quantities (Figure 3). These formats include displays of objects; the count words/signs; Arabic numerals; and the written number words in English. Because children generally begin to acquire Arabic numerals and written number words in kindergarten and first grade, the range for this set of tasks was slightly older. Children could participate in both sets of tasks, but the overall project was not designed as a longitudinal task. Forty-four children participated in both Project 1 and Project 2.

Children were recruited from schools for the deaf across the United States and elementary schools in the northeast region of the United States. The study included children who ranged from mild, moderately, and profoundly deaf and included children with unilateral hearing. Because this was the first large-scale study to compare numerical cognition development in deaf and hard of hearing children with that in typically hearing peers, children with additional disabilities were excluded (e.g., autism, intellectual disability). In a large study examining mathematical skills, Henner et al. (2021) found that children who were suspected of having a learning disability performed better than children who had been identified as having a learning disability; both groups performed less well than children who were not identified as having a learning disability. We acknowledge that the appropriate identification of learning and other disabilities (Gentry et al., 2025) and the learning and development of DeafDisabled children are in urgent need of attention from researchers (Bowen & Probst, 2023).

Children were categorized into four groups based on the modality of their language (signed ASL or spoken English) and timing of their access to language (early or later): ASL Early, ASL Later, English Early, and English Later (see Bandurski & Gałkowski, 2004; M. L. Hall et al., 2018; Marshall et al., 2015, who use comparable criteria in their designs). Early access to language meant children were exposed to language from birth from their parent(s), while later-exposed children were exposed to language beginning sometime after birth. The ASL Early group included deaf and hard of hearing children who attended programs for deaf and hard of hearing children where the primary language was ASL and had at least one deaf or hard of hearing signing parent. The ASL Later group consisted of deaf or hard of hearing children who attended the same signing programs; however, their parents were both hearing and were not fluent signers at the time of testing. Because children in these two groups were recruited from the same programs serving deaf and hard of hearing children, they are fairly comparable in terms of cultural and environmental factors. We should note that the language experience of many of the children in the ASL Later group began with exposure to spoken language only and that many of them used hearing aids, cochlear implants, or other hearing technology. However, in a pattern commonly observed with deaf and hard of hearing children whose parents initially choose to expose them only to spoken language, at some point they began attending a signing educational program (Holcomb et al., 2024; Mauldin, 2016, 2019).

The English Early group was made up of typically hearing children whose primary language was spoken English. Lastly, the English Later group included deaf or hard of hearing children who attended specialized programs focused on listening and speaking approaches to learning English. All children in this group used at least one type of hearing technology (e.g., hearing aid(s), cochlear implant(s), or bone-anchored hearing aid(s) (BAHA)). While the children in the two English groups did not attend the same schools, family socioeconomic status (SES) was included in our analyses, and we also provide the racial and ethnic backgrounds of all participants in the published study reports.

We recognize that deaf and hard of hearing children are a heterogeneous group with a wide range of language and educational experiences. Consequently, we designed the study and our approach carefully to avoid the biases and methodological shortcomings of studies that compare deaf or hard of hearing children with typically hearing children described by Holcomb et al. (2024) and to avoid perpetuating harmful, deficit-based narratives. Our goal with this design was for the two early-exposed groups to be as comparable as possible (that is, deaf or hard of hearing children acquiring ASL from parents at home and hearing children acquiring spoken English at home). A consequence of this coarse, binary distinction (access from birth vs. beginning at a later point in development) is a great deal of variability in the later-exposed groups, with some of those children beginning to have access to language at ages that are still considered early for deaf and hard of hearing children, such as prior to their first birthday. Many children who enter signing educational programs at early ages have excellent outcomes, performing comparably to their peers who have deaf or hard of hearing signing parents (see for example, Finton et al., 2024; Henner et al., 2016). Therefore, including such children in the later-exposed group will boost the performance of that group. Any differences we still observe between the children exposed from birth at home with parents and this later-exposed group reveal the additional benefits of exposure from birth as well as exposure to ASL at home and at school. Thus, this categorization works against our hypothesis about the importance of early access to language, with the same logic applying to the children in the English Later group.

To sum up, this design maximizes the comparability of the language experiences of deaf and hard of hearing children whose access to ASL begins from birth at home and typically hearing children whose access to spoken English begins from birth at home, and offers a rare opportunity to compare these groups directly. Finally, this design offers the strongest test of the hypothesis that later access to language, and not deafness or use of a sign language, is associated with any performance differences we observe. In some of our studies, we have included a continuous variable for the timing of language access. However, it is difficult to summarize the complexity of a child’s language experience with one number. Therefore, in the future, we suggest researchers use tools like the Language Access Profile Tool (LAPT) developed by M. L. Hall and De Anda (2022).

2.2. Tasks

The current paper summarizes findings from tasks included in Projects 1 and 2. We will take each task in turn, briefly describing the materials and procedure, and noting whether it is a standardized assessment. All tasks, procedures, and instructions were carefully designed, adapted and piloted by a team of deaf and hearing fluent and native signers to be presented in either ASL or English, fully considering the needs of deaf and hard of hearing children who experience later exposure to language. The procedure for each task varied; here we will briefly describe the materials and what children had to do for each task. For sample sizes and other details please refer to the publications listed in

Table 1. Participant sample sizes for all reported tasks; tasks are presented in descending order of hypothesized dependence on language skills.

Project	Skill (Task)	Relevant Paper	ASL Early n	ASL Later n	English Early n	English Later n	Overall N
One	Vocabulary (PPVT-IV)	(Carrigan & Coppola, 2020)	Not included	Not included	47	28	55
One	Count Sequence (Highest Count)	Current paper	42	35	64	51	192
One	Executive Function (BRIEF-P)	(Goodwin et al., 2021)	26	28	46	23	123
One	Number-Knower Level/Cardinality (Give-a-Number)	(Carrigan et al., in preparation)	41	34	62	39	176
Two	Mapping Across Number Representations	(Walker et al., 2023)	43	52	47	46	188
One	Nonverbal Object Tracking (Mr. Elephant)	(Quam et al., under review)	27	28	57	41	153
One	Approximate Number System (Panamath)	(Santos et al., in preparation)	43	39	67	52	201
One	Nonverbal IQ (DAS Picture Similarities)	(Quam & Coppola, 2023)	41	31	41	31	144

. Figure 1 provides the list of included tasks, the underlying skill they assess, a picture representing the task, and whether the skills are expected to depend on language.

These tasks were selected because we deemed them to be the most appropriate ones to yield the types of data needed to address our research questions. Namely, whether children’s language experiences affect the development of their number knowledge, including the count list, the cardinality principle, the ability to discriminate between large sets, to track objects, and to understand different representations of quantity. The selection of the remaining tasks was guided by the prior literature on those skills, which suggests that language ability, nonverbal IQ, and executive functioning influence one or more of the number-related abilities.

2.2.1. Peabody Picture Vocabulary Test, Fourth Edition

The PPVT-IV (Dunn & Dunn, 2007) is a standardized assessment that measures receptive vocabulary. Children are presented with a series of pictures and then hear a word from the examiner and they are to select the picture that best corresponds to the word. In our study (Carrigan & Coppola, 2020), only deaf and hard of hearing children who were acquiring spoken English were included. Because the overall design included children who use ASL, however, we adapted the standard presentation of stimuli so that children who used ASL as well as spoken English would have comparable visual components. That is, we presented videos of an experimenter saying the spoken English words to the hearing children in the English Early group and to the deaf and hard of hearing children in the English Later group.

2.2.2. Highest Count

The Highest Count task tests children’s knowledge of the count sequence. We first asked children to produce the count words or signs up to 10, and then, if they succeeded, up to 20 (this was the abstract, or “without objects” condition). We then provided sets of 10 and 20 fish arranged on a board to offer children objects to anchor their counting (the “with objects” condition) and followed the same procedure. This procedure is commonly used in the developmental psychology literature and has previously been used with deaf and hard of hearing children who use spoken and/or sign language (Secada, 1984; Leybaert & Van Cutsem, 2002).

2.2.3. Behavior Rating Inventory of Executive Function—Preschool Version

The BRIEF-P (Gioia et al., 2003) is a standardized parent-report questionnaire that assesses children’s executive functioning. Executive function refers to behaviors that allow an individual to overcome more automatic or established responses to achieve a goal (e.g., Garon et al., 2008). For this 63-item questionnaire, parents answered how frequently a specific behavior has been a problem for their child during the last six months. Its subscales include Emotional Control, Working Memory, and Plan/Organize. Specific examples of prompts rated by parents include “When given two things to do, [my child] only remembers the first” (from the Working Memory subscale; probes a child’s ability to hold information in mind for later use) and “When instructed to clean up, [my child] puts things away in a disorganized, random way” (from the Plan/Organize subscale; measures a child’s ability to plan and implement their goal-directed behaviors).

2.2.4. Give-a-Number

Give-a-Number (also called Give-N) (Wynn, 1990; see Wege et al., 2024 for a methodological review) assesses children’s understanding of the cardinal principle. Knowing the cardinal principle refers to the notion that the last number word or sign used after correctly applying the counting principles represents a property of a set of objects. The task is presented in the context of a game in which the child is helping fish go swimming with their friends in a “pond” (a bowl). In the canonical titrated version of the Give-a-Number task, children are first asked to place one toy fish into a bowl and to check that they have placed the correct quantity. If successful, the experimenter then increases the requested set size by one, until the child succeeds on a predetermined stopping point (generally six). This task is considered the gold standard task for assessing children’s number knowledge and is widely used with hearing children to assess their understanding of the cardinal principle, as is a related task, the How Many? task. Give-a-Number has been used with deaf and hard of hearing children who use spoken language (Shusterman et al., 2022).

Because we expected variability among our participant groups, we opted to administer the task using a fixed order of trials, given in one or two stages, depending on the child’s performance. Children first saw 3 trials each of the quantities one through six (18 trials total); children were classified as “cardinal-principle knowers” if they accurately produced sets for both quantities 5 and 6 on at least 2 out of 3 trials. If children succeeded on the quantities up to 6, we also tested them on 5 quantities between 7 and 20. As in the canonical version of the task, children were asked to check the quantity of fish they placed in the bowl and to correct it if the quantity did not match the requested quantity.

2.2.5. Mapping

The Mapping task examines children’s ability to “map” or “link” different representational formats of quantities (i.e., Arabic numerals, signed and spoken number words, and sets of objects). This skill is critical for developing computational fluency, the ability to execute simple arithmetic quickly and automatically, which lays the foundation for intermediate and advanced mathematics. Children are presented with a target quantity in one representational format and are asked to choose the matching quantity in another format from four options (e.g., map the Arabic numeral “1” to the same quantity from four number signs (THREE, FIVE, ONE, or TWO). This task was adapted by our team from one used with younger hearing children by Hurst et al. (2017). Our adaptation tested quantities 3 through 9 and featured 9 practice trials and 42 test trials presented in a fixed random order. We developed non-linguistic instructions that showed the child what to do (vs. the experimenter explaining to them what to do). Children watched a video of a child completing the task (third-person perspective). The video then showed a child’s hand pointing straight ahead at the screen to indicate their response, offering a first-person perspective to the child who was about to engage in the task.

2.2.6. Mr. Elephant

Mr. Elephant tests children’s ability to track quantities of objects without using language. This version was adapted from the task developed by Shusterman et al. (2017). In this task, the experimenter drops a given number of balls into a large wooden Mr. Elephant toy and either the same number of balls, or one fewer ball, exits his trunk. For example, on a given trial, if the experimenter dropped 4 balls in, either 3 or 4 balls would come out. On each trial, children are asked to indicate whether any balls remained inside Mr. Elephant. Children participated in practice trials involving 1 and 2 balls and then completed 6 test trials in which the target quantity ranged from 2 to 7; trials were presented in a fixed random order.

2.2.7. Panamath

Panamath (Halberda et al., 2008) assesses the acuity of an individual’s Approximate Number System (ANS) and is widely used in the literature. This task is also framed as a game in which children are ‘dot collectors’ with the goal of collecting as many dots as possible. Children see a computer display divided into two sides, with each side featuring a familiar character from Sesame Street and a set of colored dots (one color per side). The dots are presented for a short time and the child points to indicate which side contains more dots. The quantities presented are well into the “large” range (greater than 8 objects) and cannot be counted during the relatively short presentation time. Children received visual feedback after each trial (a smile face indicating correct responses, and a sad face indicating incorrect responses). Children saw a total of 60 trials in 3 min; the specific quantities presented depended on the child’s age.

2.2.8. Picture Similarities Task in the Differential Abilities Scale, Second Edition

The Picture Similarities task in the Differential Abilities Scale (DAS-II; Elliott, 2007) is a standardized instrument designed to assess nonverbal IQ. In this task, children are asked to match the target image on a card handed to them by the experimenter to a related image from a set of four in the test binder (see Figure 1). The related image in early trials is an identity match, that is, the image on the card is identical in size and shape to the one given in the test book (e.g., the child would place the card with the hexagon under the hexagon shape in the test book). As the trials progress, the differences between the target image and the intended related image grow. The later trials require more abstract reasoning about the relations between the target image and the correct response. We recognize the limitations of the vast majority of standardized instruments that are not normed with deaf and hard of hearing children; we chose the Picture Similarities subtest because it does not require linguistic instruction and also does not contain any printed words or numerals.

2.3. Procedure

The study was conducted in accordance with the Declaration of Helsinki, and the protocol was approved by the Institutional Review Board of the University of Connecticut (H13-105) on 1 February 2016. Parents received colorful, engaging brochures written in plain-text English and Spanish that summarized the research questions and invited their children to participate. The brochures included contact information—via telephone, videophone, and email—enabling communication with the research team in English, ASL, and Spanish. Parents participated by reviewing and signing the consent form, completing a background questionnaire that asked about the etiology of the child’s deafness, hearing levels, family language use, and demographic information. Parents also completed the BRIEF-P assessment of their child’s executive functioning.

All assessments were conducted in person in a private and distraction-free room. The research team generally spent multiple days at the schools. The individual testing the child was fluent in the child’s preferred language (i.e., either by a deaf individual who was a fluent signer or a hearing individual fluent in English). All team members who interacted with deaf and hard of hearing signing children were fluent signers. Tasks and instructions were carefully designed and piloted by a team of deaf and hearing fluent and native signers to be presented in either ASL or English, fully considering the needs of deaf and hard of hearing children who experience later exposure to language. The experimenter engaged the child in brief, friendly conversation to build rapport and confirm their willingness to participate. Both projects began with lower-stakes tasks to ease the child into the session: Project 1 started with a counting task, while Project 2 began with the mapping task involving non-linguistic instructions that showed the child what to do.

The team followed a general task order, but the research team remained responsive to each child’s attentional state and occasionally substituted tasks to maintain engagement. Children were offered breaks upon request or when they appeared to benefit from one. In total, the tasks for each project took approximately 60 min to complete. Some children finished in a single session, while others returned for a second session. The research team remained flexible and responsive to students’ and teachers’ needs, adjusting around class schedules and lunch periods as needed. All sessions were videorecorded and scored offline.

3. Results

For each skill, we examined whether children’s performance was influenced by the age at which they gained access to their primary language, that is whether they were exposed to it in their home starting from birth (referred to as the “early” group) or whether their access to language began at a later point in their development (referred to as the “later” group). The second major factor we examined was the modality of their current language use, that is, whether they primarily used a sign language (i.e., ASL) or a spoken language (i.e., spoken English). For each task, we included in our analytical models relevant factors that have been suggested in prior literature to influence performance on that task, such as age and family SES.

3.1. Peabody Picture Vocabulary Test (Vocabulary)

Using a linear regression run in SPSS version 28, we found that language timing significantly predicted children’s PPVT raw scores, as did their age. Specifically, children who were exposed to spoken English early outperformed those who were exposed to spoken English later via hearing aids and cochlear implants, by about 11.5 points. SES did not predict PPVT raw scores.

3.2. Highest Count

We entered into our analysis the highest number the child could count to, in either condition (with or without objects). A linear regression model run in R found that children who were exposed from birth to their home language could count higher than children who were exposed to their language later in development. Unsurprisingly, a child’s age also predicted their counting ability. There were no differences between the deaf and hard of hearing children who used ASL or the hearing children who used spoken English regarding how high they could correctly count.

3.3. BRIEF-P (Executive Function)

The raw scores of the children in the ASL Early and English Early groups did not differ (Welch’s t-test) (Goodwin et al., 2021). A linear regression model run in SPSS found that earlier language exposure, SES, and age all independently contributed to better executive functioning (Goodwin et al., 2021). This study also examined the timing of auditory exposure and did not find a relationship between auditory exposure and executive function.

3.4. Give-a-Number (Cardinal Principle)

We used the child’s first response, prior to their response given after they checked the quantity of fish they originally placed in the bowl. Using a linear regression model run in R, we found that ASL supported cardinal principle achievement as well as spoken English did (Carrigan et al., in preparation). That is, we found no differences between deaf and hard of hearing children who were acquiring ASL from birth and hearing children who were learning spoken English from birth (Carrigan et al., in preparation). We also found that whether a child was deaf, hard of hearing, or hearing; used a sign or spoken language; or had higher or lower SES was not associated with their cardinal principle knowledge (Carrigan et al., in preparation).

3.5. Mapping (Conversion Among Representational Formats)

Using a linear regression model run in R, we found that children’s ability to map between different numerical representations was not predicted by their language modality; children whose primary language was English or ASL performed similarly (Walker et al., 2023). However, the timing of language exposure was a significant predictor of mapping performance. Specifically, children who were exposed to language early outperformed those exposed to language later. Age was also a significant predictor, with older children showing superior mapping abilities compared to younger children.

3.6. Mr. Elephant (Non-Linguistic Object Tracking)

Using a logistic regression model run in R, we found that the timing of children’s access to language and SES, but not a child’s age or language modality, was associated with children’s ability to track sets of two and three objects (Quam et al., under review).

3.7. Panamath (Approximate Number System Acuity)

Using a linear regression model run in SPSS, we found that language modality was a significant predictor of Approximate Number System (ANS) acuity (Santos et al., in preparation). Children whose exposure to language began at birth performed better than later-exposed children, and English users outperformed ASL users. For English users, the effects of age and timing of language exposure on ANS acuity disappeared when vocabulary was included in the analysis, as vocabulary was significantly associated with ANS acuity. For ASL users, age and timing of language exposure remained significant in addition to vocabulary in predicting ANS acuity (Santos et al., in preparation).

3.8. Picture Similarities (Nonverbal IQ)

Using a linear regression model run in R, we found that the timing of language access predicted much more variability in children’s performance than the modality of the language being learned by the child (ASL or English) (Quam & Coppola, 2023).

4. Discussion

First, we return to the theoretical underpinnings regarding how much each of these skills has been hypothesized to depend on language. We will also review the findings from prior studies, including whether the skill has been previously studied in deaf and hard of hearing children. We will note the contribution of our design to understanding whether prior findings might have been due to the timing of children’s exposure to language, followed by a discussion of how we interpret the findings overall, and conclude with the implications of these findings for STEAM educators who work with deaf and hard of hearing children.

4.1. Peabody Picture Vocabulary Test

Prior work with deaf and hard of hearing children (Adams Costa et al., 2019) using the PPVT-V found that children who were acquiring spoken English via cochlear implants scored significantly lower than matched control children with typical hearing, on average approximately 14.6 points lower. Consistent with the findings of Adams Costa and colleagues, using a linear regression run in SPSS, we found that language timing significantly predicted children’s PPVT raw scores, as did their age. Specifically, hearing children who were exposed to language from birth outperformed those who were exposed to language later via hearing aids and cochlear implants, by a similar degree as found in the earlier study that used the traditional methodology in which the experimenter said the word aloud to the child.

4.2. Highest Count

Acquiring the count sequence is a specialized subset of vocabulary and is also clearly dependent on language. The only prior study comparing the counting skills of deaf and hard of hearing children learning ASL and hearing children learning English is an unpublished dissertation (Secada, 1984). Secada found that both deaf and hard of hearing children and their typically hearing peers exhibit similar patterns of number development when provided with accessible language from birth. Leybaert and Van Cutsem (2002) found that hearing children learning French could count higher than deaf children learning Belgian French Sign Language and/or French. They attribute this difference to a number of factors, including educational quality and fewer opportunities for incidental learning but not to the deaf children’s later access to language. They report that only 3 of their 21 deaf participants had deaf parents, and they do not report whether those parents signed. In our study of 64 hearing children and 128 deaf and hard of hearing children, we found that the children exposed to language early could count significantly higher than the children whose language access began later in development.

4.3. BRIEF-P

The BRIEF has been found to be influenced by children’s language ability and experiences. While there has been a great deal of work on executive functioning in deaf and hard of hearing children (e.g., Botting et al., 2017; M. L. Hall et al., 2019; Kronenberger & Pisoni, 2020), very few studies included both deaf, hard of hearing, and hearing children, as well as children who were acquiring sign language and spoken language. Prior studies also rarely systematically controlled for the timing of children’s access to language and the impact of the timing of auditory exposure on EF. Further, none of the prior work studied preschool-age deaf and hard of hearing children. By using our experimental design to account for these differences in children’s experiences, we found that the timing of children’s access to language predicted children’s executive functioning skills but that auditory experience did not.

4.4. Give-a-Number

Because this skill entails understanding the meanings of the words or signs of the count sequence, this skill is also thought to depend on language. When children are beginning to learn to count, they pass through a stage in which they recite the words or signs without understanding the corresponding quantities (as in the Highest Count task described earlier). The cardinal principle constitutes a major milestone in the development of numerical cognition and reflects children’s understanding that the last number word produced when counting refers to the cardinal value of the set of objects. Furthermore, number word knowledge supports the development of number concepts for both basic (1–6) and larger quantities, emphasizing the importance of number-related language in fostering quantity understanding.

The study conducted by Leybaert and Van Cutsem (2002) described earlier compared 21 deaf children with 28 typically hearing children on a “set-creation” task that resembles the Give-a-Number task. They asked children to create sets containing between 3 and 14 items. They found no significant differences between the groups; however, because they matched the children based on school year and not chronological age, the deaf children were on average 1 year older than the hearing children. While this choice is understandable given the researchers’ concern about potentially disadvantaging the deaf children given the lower amount and quality of schooling they received, it makes it difficult to assess a true developmental parallel between the groups in the achievement of the cardinal principle. In a larger study, Shusterman et al. (2022) compared 44 deaf and hard of hearing children who were learning spoken English (but no sign language) with 79 typically hearing children with comparable age ranges. They found that the deaf children showed similar patterns of development as did the hearing children, but that the deaf children’s development was delayed. Specifically, the odds of being a cardinal-principle knower were 17 times higher for the hearing children (early-exposed) than for the deaf children (later-exposed), controlling for age. They also found that this difference between the groups was fully mediated by the children’s language skills.

Like Shusterman and colleagues, our experimental design examined differences in when during development children began accessing language and, in addition to children who were acquiring spoken English, also included deaf and hard of hearing children who were acquiring ASL. We found no differences between deaf and hard of hearing children who were acquiring ASL from birth and hearing children who were learning spoken English from birth (Carrigan et al., in preparation).

4.5. Mapping

As described earlier, this task was adapted from one used with younger hearing children by Hurst et al. (2017); no prior work has systematically examined this skill in deaf and hard of hearing children. This task is hypothesized to depend on language given that robust knowledge of the number words and signs underlies a child’s ability to convert those linguistic representations into the other representational formats (i.e., arrays of dots and Arabic numerals). We found that neither hearing level nor language modality influenced performance, though the timing of language exposure was a significant predictor of mapping performance. Specifically, early-exposed children outperformed those exposed to language later. As expected, age also significantly predicted performance, with older children demonstrating better mapping abilities than younger children (Walker et al., 2023).

4.6. Mr. Elephant

This task tests children’s ability to track quantities of objects without using language and is considered to depend less on language than the earlier skills assessed. Such non-linguistic abilities have been hypothesized to support the development of language-based numerical knowledge, such as acquiring the count sequence and understanding cardinality, both major milestones in number cognition. For the trials involving only two or three balls, this task is hypothesized to not depend on language. Such non-linguistic abilities have been hypothesized to support the development of language-based numerical knowledge, such as acquiring the count sequence and understanding cardinality, both major milestones in number cognition. Shusterman and colleagues found that typically hearing children who had mastered the cardinal principle, as well as children who had not (and who are referred to as “subset knowers”) showed comparable performance for tracking quantities of balls from one to four. All of the typically hearing participants in that study had experienced full access to language from birth.

We found that children exposed to language early in development performed better than children exposed to language later in life. This finding is not consistent with the hypothesis that this ability is independent of language (Quam et al., under review).

4.7. Panamath

Because this task assesses children’s ability to discriminate between sets based on their approximate magnitude rather than assessing exactly how many objects are being presented, it is considered to not depend on language skills. One prior study conducted by Bull et al. (2017) reported a comparison of 75 deaf and hard of hearing children with 75 children with normal hearing between the ages of 5 and 13 years (mean age 9 years). They found that while the ANS acuity of both groups of children improved with age, the deaf and hard of hearing children performed worse at every age point relative to the typically hearing children. That is, the deaf and hard of hearing children were less likely to be able to correctly indicate which side of the screen contained more circles. In combination with results from their other prior studies, they interpret this finding to indicate that both groups of children represent numerical information in similar ways. They suggest that the difference in performance may arise for two reasons: children with hearing loss (1) process numerical information less efficiently than their hearing counterparts (as evidenced by the SNARC task (which tests the association between space and numerical magnitude) and (2) represent numerical information less accurately (based on a number-line task). The authors report many characteristics of the children with hearing loss who participated in the study, such as level of hearing loss, hearing device used, communication preference (sign, spoken, mixed), SES, and parental education. However, they do not report any information about when during development the children began to access their first language. Assuming that their sample reflects the distribution observed in the US and many other countries, it is likely that only a small percentage (less than 10%) of their participants had access to either spoken or sign language from birth. Thus, their conclusion that “there is clearly something about having hearing loss that puts children at risk for lower ANS acuity” may instead be related to a later age of access to language.

Unlike the prior studies from the large project reported here, in addition to the timing of language exposure, we found that language modality was a significant predictor of Approximate Number System (ANS) acuity. As in Bull et al. (2017), age was also a significant predictor (Santos et al., in preparation). The modality finding was unexpected, as there was no reason to expect using ASL would hinder ANS acuity.

4.8. Picture Similarities

This task is designed to assess nonverbal IQ, which is not thought to depend on language. As described earlier, we carefully trained our team to interact with deaf and hard of hearing children who may have experienced language deprivation in ways that allowed them to demonstrate their knowledge. Even with these efforts, we found that children whose access to language began later in development performed significantly worse than did children whose access to language began at birth (Quam & Coppola, 2023). Reesman et al. (2014) evaluated the instruments that are most frequently used with deaf and hard of hearing children on five criteria. These criteria include whether guidance regarding accommodations is provided for using the task with deaf and hard of hearing children; guidance regarding interpretation of results; and empirical assessment of potential bias. Their review “confirms that in many cases there is a lack of information and evidence to appropriately answer important questions related to issues of possible test bias and other issues germane to test validity” when using assessments designed for and validated with hearing children only.

4.9. Overall Discussion

This study was designed to examine the impact of different language experiences on the development of number cognition in, deaf, hard of hearing, and hearing children between 3 and 10 years of age. In sum, we found that a child’s language experience affects their cognitive development in multiple domains. To return to our research questions, we found that (1) language modality, that is, whether children were learning a signed or spoken language, largely did not affect their performance. Addressing our second research question, we found that deaf, hard of hearing, and typically hearing children whose access to either a signed or spoken language began at birth all showed comparable performance. These two findings suggest that hearing level and use of a sign language do not themselves explain differences that have been previously reported in the literature. Importantly, we found that children whose access to language began at some point after birth, showed lower performance on average than those whose access began at birth. The negative impact of the timing of language access held regardless of whether the child was learning a spoken or signed language.

As described earlier, we included several factors and skills that have been demonstrated to affect number cognition, with the idea of controlling for those in our analyses. These included the child’s executive functioning, nonverbal abilities, and the family’s SES. However, instead of being able to assess the contribution of each of these skills to children’s number cognition development, we found independent negative effects of later access to language on those skills themselves. Further, we also found almost universal negative impacts of later exposure to language (regardless of language modality) on task performance for quantity-related skills, such as nonverbal object tracking and the Approximate Number System, that are hypothesized to not depend on language. Crucially, most of these tasks have only been reported on in children with typical hearing who have experienced full access to language from birth.

We now return to the three possibilities we laid out earlier as explanations for the relationship between the development of language and cognition: (1) Gentner’s (2016) hypothesis that language functions as a ‘tool kit’; (2) Holmer et al.’s (2016) Developmental Ease of Language Understanding (D-ELU) model; and (3) Carey’s (2009) bootstrapping hypothesis, which applies to number concepts but not to other aspects of cognitive development. In our view, the results from the Give-a-Number study directly support Carey’s bootstrapping hypothesis. Further, the impact of not learning the count sequence at the appropriate time in development certainly has direct implications for children’s mapping abilities and may also contribute to a negative impact on their ability to represent small quantities, as measured by the Mr. Elephant task. Our findings also broadly support the ideas put forward by Gentner and Holmer and colleagues, that is, that language supports and enhances other cognitive abilities. However, we do not think that our results can distinguish between the notions of language as a cognitive tool kit or the D-ELU model.

4.10. Sign Language Guarantees Early Language Acquisition for Deaf and Hard of Hearing Children

In principle, early and full access to either signed or spoken language should support optimal linguistic and cognitive development in deaf and hard of hearing children. However, in practice, deaf and hard of hearing children as a group are not achieving levels of mastery of spoken language comparable to their typically hearing peers of the same age (Dettman et al., 2016, as well as works cited in M. L. Hall et al., 2019). This disparity persists despite earlier identification and amplification, the most up-to-date assistive hearing technology (such as modern cochlear implants), and participation in early intervention services. Another complicating factor regarding the spoken-language-only approach is that by the time it is apparent that a deaf or hard of hearing child is not meeting milestones for spoken language development, the optimal window for acquiring sign language has narrowed. It is now well documented that signed and spoken languages are acquired and processed by highly overlapping brain structures and regions and that the sensitive period for language acquisition applies to sign languages as well as spoken languages. Thus, given this body of knowledge, the negative impact of delays in providing access to sign language are avoidable and unnecessary. W. C. Hall (2017) discusses the protective influence of early access to sign language, which promotes “the healthy growth of all developmental domains through a fully-accessible first language foundation.”

The evidence base in support of a bilingual and bicultural approach that includes both a signed and spoken language (e.g., Clark et al., 2020) is expanding rapidly and includes contributions from the fields of early intervention and education, as well as developmental psychology and linguistics, and has garnered support from researchers and professionals who primarily work with deaf children acquiring spoken language with hearing technology. As noted earlier, sign languages are acquired by deaf children with adequate language exposure along a similar timeline to that of hearing children acquiring spoken language (e.g., Lillo-Martin & Henner, 2021). Further, exposure to sign language prior and after cochlear implantation increases language and cognitive skills in deaf children (Delcenserie et al., 2024), and deaf and hard of hearing infants whose hearing parents began signing with them in early infancy have comparable vocabularies as those with deaf and hard of hearing signing parents (Caselli et al., 2021). Deaf children who acquire a signed and spoken language from birth can develop age-appropriate spoken language skills (Davidson et al., 2014) and may even outperform their monolingual deaf peers (Hassanzadeh, 2012). A paper co-authored by specialists in both sign and spoken language development in deaf children discussed the risks and benefits of encouraging that all deaf children learn sign language (Napoli et al., 2015). A recent research synthesis and meta-analysis found that sign language was positively correlated with deaf and hard of hearing students’ spoken and written language development (Zhang et al., 2024). Moreover, family use of sign language can signal acceptance of their deaf and hard of hearing child and may confer social and emotional health benefits (Kushalnagar et al., 2020).

These findings naturally lead to the question of how we can best support development in all deaf and hard of hearing children. We suggest a multi-pronged approach that includes educating professionals regarding the negative consequences of later access to language and encouraging educational researchers to develop pedagogical strategies targeted to deaf and hard of hearing children and specific to each domain that may be affected (one model to consider in this regard is the Strategic and Interactive Writing Intervention (SIWI) approach (e.g., Wolbers et al., 2020; Dostal & Wolbers, 2014). Another aspect of this approach must be advocating for policy recommendations that ensure full and early access to language for all deaf and hard of hearing children, such as the changes suggested by Humphries et al. (2024). Providers, communities, and policymakers must work together to reduce and dismantle barriers to equitable hearing health care and intervention and support services, which include geographical area, parental educational and socioeconomic status, race/ethnicity, and insurance coverage, among other factors (see, for example, Bush et al., 2017; Meinzen-Derr et al., 2022; Mauldin, 2016).

4.11. Limitations

Despite the strength of the findings reported here, we must consider the limitations and methodological issues of this approach. One issue is that there currently exist limited tools and methods that allow researchers to directly compare the performance of children learning spoken English with that of children acquiring ASL. Further, there is a natural limit on the sample sizes of signing deaf and hard of hearing children, first due to the low incidence of childhood deafness and second because only about 20% of deaf and hard of hearing PreK-12 students are educated in a sign language (U.S. Department of Education, 2020).

Another factor to consider is that a large number of deaf and hard of hearing children in the Later ASL group started out as children who were learning spoken English exclusively. Like many deaf and hard of hearing children in the US who did not make progress with a spoken-English-only approach, at some point they switched to a program that used ASL as the language of instruction (M. L. Hall et al., 2019). For this and associated reasons, direct comparisons between the subgroups are not appropriate, because the children in each group differ in systematic ways. For the children in the Later English group, access to language tends to begin earlier in development than for the children in the Later ASL group, but the quality of the spoken language signal is likely degraded relative to what children with typical hearing receive. Lastly, the analyses reported here are based on a binary distinction between children whose access to language either began at birth or did not. We made this decision for two reasons: in part because our methodology did not allow us to obtain precise measures of exactly when in development a child’s access to language became substantive, and also because the group of deaf and hard of hearing children whose access to ASL began at birth is the most comparable to the language learning environment of the typically hearing children, who also began learning language at home, from their parents. A continuous measure of the timing of language access, combined with measures of the quality and quantity of language input, would likely yield more insights into the relationships between language and cognitive development.

4.12. Implications for STEAM Education

The wide range of domains and skills affected by later access to language summarized in this overview underscores its potential negative impact on STEAM areas. One way to support STEAM education is to ensure that all deaf and hard of hearing children acquire a language foundation that allows them to make the connections among its component disciplines. Thus, we hope that this work contributes to the conceptual basis for an integrated STEAM curriculum that addresses the issues faced by deaf and hard of hearing students whose early interactions and environments have not been fully accessible to them. At the same time, a STEAM pedagogical approach that centers deaf and hard of hearing children, engages them, and strives to highlight the connections among various disciplines offers opportunities to offset the negative impacts experienced by some deaf and hard of hearing students. Deaf and hard of hearing students are highly attuned to visual dimensions of their environments, and we encourage practitioners to leverage this quality in designing curricula aimed at all deaf and hard of hearing students, which could especially benefit those who have experienced later access to language. For example, in the domain of mathematics, Zarfaty et al. (2004) found that preschool-aged deaf children spontaneously took advantage of the spatial cues presented in a task that tested their representation and discrimination of quantity. Nunes and Moreno (2002) used aspects of this approach in developing their intervention for deaf students in elementary school. Specifically, they promoted the presentation of word problems involving transformations over time using visual elements such as diagrams. These diagrams reduced the cognitive demands of remembering sequences of events, a skill that the researchers had identified as a particular difficulty for deaf and hard of hearing students.

Some examples of STEAM teaching approaches, guidelines, and materials are specifically geared toward deaf and hard of hearing students at the elementary and secondary levels. The PLAYground Online Camp for Deaf Children (Monson et al., 2021) is an eight-week virtual summer camp for K-8 students based at the Metro Deaf School in Minnesota. The webinar STEAM is Everywhere (Clerc Center, 2024) is a resource for teachers of secondary school students that demonstrates how to teach STEAM, specifically electronics, in classrooms designed for deaf and hard of hearing students. Finally, Gardiner-Walsh (2024) provides detailed guidelines for teachers and classroom assistants at the secondary level who want to collaborate on offering STEAM education in their classrooms.

5. Conclusions

In sum, this large study of close to 400 deaf, hard of hearing, and typically hearing children found that access to signed or spoken language from birth provides a solid foundation for age-appropriate cognitive development across a range of skills. We also found that children who experienced later access to either spoken or sign language tended, on average, to perform worse than their deaf and hard of hearing and typically hearing peers on these tasks. The notion that later and less access to language might explain reported discrepancies between hearing and deaf and hard of hearing children’s performance in tasks related to numerical cognition has been proposed by many researchers (e.g., Gottardis et al., 2011; Santos & Cordes, 2022). However, this project is the first to systematically test this hypothesis and find support for it.

These results have both strong theoretical and practical implications. In terms of theory, these findings offer an opportunity to begin refining our theories regarding which aspects of language structure and experience are most influential in promoting development in the different domains encompassed by STEAM. They also highlight the potential effects of language experience on domains previously considered to be independent of language.

In terms of early intervention and educational practice, these results point strongly to the need for children to experience accessible language early in development. That is, deaf and hard of hearing children must receive exposure to fully accessible sign language input as early as possible, in addition to any hearing technology aimed at improving their access to spoken language. Finally, we urge collaboration among pedagogical specialists with expertise in deaf education and experience teaching deaf and hard of hearing children to develop STEAM curricula that engage, value, and support deaf and hard of hearing children.