The Cleveland Family Speech and Reading Study: A Review of Long-Term Outcomes Linking Phenotypes and Genotypes for Speech Sound Disorders

Abstract

Purpose:

This article reviews findings from the Cleveland Family Speech and Reading Study, a 25-year longitudinal study of speech sound disorders (SSDs) of unknown etiology. The study sought to describe the long-term outcomes for individuals with early SSD, including those diagnosed with childhood apraxia of speech (CAS), and to identify genetic findings for endophenotypes and phenotypes associated with SSD.

Method:

The data included in this article were collected from families (N = 296) with a child diagnosed with SSD. Probands and available siblings were assessed on a battery of speech, language, and literacy measures at preschool (4;0–6;11 [years;months]), school age (7;0–11;11), and adolescence/young adulthood (12–25 years). Families (N = 272) provided DNA samples, including 73 probands with CAS. Genome-wide association studies (N = 435) and whole genome sequencing (N = 101) analyses were conducted on a subset of the sample.

Results:

Poorer outcomes for adolescents/adults are associated with oral motor skills, an early history of motor difficulties, comorbid language impairment, and persistent speech sound errors. Children with SSD had both rare genetic and common variants that were correlative and may be causal. Several of these variants occurred de novo in the CAS proband with or without a family history of SSD, suggesting that de novo variants play a larger role in CAS and SSD than anticipated.

Conclusions:

Some children with early SSD demonstrate persistent speech, language, and literacy difficulties in adolescence and adulthood, while others show partial or full resolution. Family health history questionnaires do not capture long-term outcomes; thus, finer long-term assessments are needed to identify phenotypes in older children. Genetic data suggest considerable locus heterogeneity. Having a compendium of variants for SSD, including CAS, will help practitioners move toward precision medicine with deep phenotyping supporting accurate diagnoses and personalized interventions.

Over the past 3 decades, genome-wide association studies (GWAS) and whole genome sequencing (WGS) have identified a genetic contribution to many speech and language disorders (Chan et al., 2024; Hildebrand et al., 2022; Morgan et al., 2024). Genetic studies of speech and language disorders have examined individuals with known genetic syndromes and individuals with speech and language disorders of unknown etiology (nonsyndromic). However, translating these advances into clinical practice has been a slow process. Lauretta et al. (2023) identified potential barriers to the use of genetics by speech-language pathologists (SLPs). These included a lack of confidence in their genetic knowledge, uncertainty about the relevance of genetics to guide clinical practice, and the SLP's level of experience with genetics. In contrast, factors that enabled the use of genetics included recognizing the value of a genetic diagnosis, receiving support from other health care professionals in treating genetic conditions, and developing ongoing relationships with genetic services.

This example reflects the principles of precision, or personalized medicine, which involves tailoring treatments to an individual's genetic composition and environmental factors rather than using a standard approach for all patients. Precision medicine is the science wherein individual variability in genetics, environment, or lifestyle profiles can impact treatment of and response to disease. Based on this knowledge, therapies and medications (drugs) can be targeted at the individual level to help alleviate symptoms of a disease or disorder. The most direct example of precision medicine is when a treatment is directed at the gene or its product, or the biological pathway involved in the condition. Precision medicine encompasses both diagnostic (predictive) elements and treatment (therapeutic) components. By identifying mutations in genes that may cause a disorder (e.g., mutations in FOXP2), a pipeline is created for the diagnostic component of precision medicine as it relates to speech sound disorder (SSD).

Despite the promise of precision medicine, its application to speech and language disorders remains complex. For example, research involving individuals with genetic syndromes has provided detailed phenotyping related to the syndrome and identified variations among individuals with the same syndrome. In this way, the spectrum of phenotypic variability in a specific syndrome can be described and may provide a roadmap for intervention. However, many children on the caseloads of pediatric SLPs have not been diagnosed with a syndrome, and genetic testing has not been performed. Studies examining individuals with speech and language disorders of unknown etiology have attempted to find genetic variants that may influence speech and language skills (Chan et al., 2024; Formicola et al., 2024; Stein et al., 2020). For those who undergo genetic testing, findings do not always provide clear relationships between the genotype and the observed phenotype. Furthermore, genetic heterogeneity may be observed as individuals with similar phenotypes may present with different genetic variants, and multiple genes may contribute to the speech and language disorder within a single individual (Peter et al., 2023). Complicating matters, studies examining individuals with speech and language disorders of unknown etiology may rely on historical reports of past disorders, which can be unreliable (Saul et al., 2017). Direct testing of individuals to determine the phenotypes associated with genetic findings is hampered because standardized tests may not include an older individual's age range (Fidler et al., 2011; McGregor et al., 2020).

These findings emphasize the need for better education and collaboration between SLPs and health care providers with experience in treating genetic conditions. They also highlight the potential value of using genetic information to guide clinical practice. For example, if a child is diagnosed with a known genetic syndrome, early speech and language intervention may be guided by the characteristics of the syndrome. One illustration of the value of using genetic information to support intervention is providing parent training for infants with galactosemia, a metabolic disease associated with a high risk of speech and language impairment (LI). Peter et al. (2021, 2023) designed an intervention program to increase speech perception, babbling output, and pragmatic abilities. Intervention began as early as 2 months of age, well before conventional treatment usually begins. The provision of this early intervention had positive and long-lasting effects.

To advance the application of precision medicine for SSD, it is important to understand the genetic and phenotypic variability associated with the disorder and the various approaches used to investigate it. Below, we briefly describe large- and small-scale genetic studies of SSD to illustrate the strengths and limitations of the different research approaches to SSD, including a description of the diverse phenotypes and endophenotypes associated with SSD. Next, we describe the Cleveland Family Speech and Reading Study (CFSRS), one of the few longitudinal studies of SSD of unknown etiology that identified phenotypes and genetic influences on speech and language disorders. We summarize our findings from the CFSRS regarding adolescent and young adult follow-up, which focused on differences between those with persistent SSD and those whose SSD had resolved. Finally, we discuss genetic findings from the CFSRS and clinical implications for SLPs and the research community.

Genetic Studies and SSD

In the modern era, genetic study designs for SSD, as shown in Appendix B (Table B1), are possible with any type of phenotypic and biospecimen data collection (Yehia & Eng, 2019). The structure of the data and sample size drive the framework of the analysis. Data collection has focused on case series, single families or family collections, larger-scale cohort studies, and, more recently, the extraction of medical records (Magielski et al., 2024). Each type of design can provide some information on SSD and childhood apraxia of speech (CAS) etiology; however, depending on the design, the information may not have an immediate clinical impact or may be more helpful in understanding the underlying biology of the disorder(s). For example, the phenotype data for larger-scale cohort studies and the extraction of medical records may be based on self-report, questionnaires, and/or medical records, which are not based on direct testing, and the information may not be reliable.

Most large studies of SSD rely on combining data across cohorts and are hampered by challenges with data harmonization—the integration of different assessments, dialects, and vocabularies by participants' age and sex, which can make exact cross-comparisons across studies difficult. For example, a few large cohort studies such as the Avon Longitudinal Study of Parents and Children (ALSPAC) and the Adolescent Brain Cognitive Development (ABCD) study have preplanned large-scale data collections. The ALSPAC study recruited 14,000 pregnant women and followed their children for over 2 decades to examine environmental and genetic factors on their development (Fraser et al., 2013). However, because the focus was not specific to speech and language skills, it is difficult to integrate the data sets with studies using standardized speech and language assessments. Similarly, although, the ABCD study (Karcher & Barch, 2021) collected a wide variety of behavioral, neuroimaging, and environmental data from a large cohort of U.S. children aged 9–10 years (N = 11,880), it was not designed to assess SSD or LI specifically and lacked the standardized assessments typically used in speech and language research, making data harmonization difficult.

In contrast, genetic studies of rare disorders, such as CAS, include a relatively small number of participants compared to population-based studies. Such studies allow for deep phenotyping, which is defined as comprehensive testing of individual components of a phenotype or endophenotype (Robinson, 2012). Phenotypes are observable traits such as speech and language skills. A phenotype may be related to genetic or environmental influences (National Human Genome Research Institute, n.d.). Endophenotypes are objectively measurable biophysiological, neuroanatomical, cognitive, or neuropsychological parameters closely associated with a behavioral trait and useful in detecting genetic influences (Gottesman & Gould, 2003). Deep phenotyping studies using this approach provide clinically meaningful results. For example, deep phenotyping of speech production in the K.E. family led to the discovery of the FOXP2 gene (Lai et al., 2001). Another gene, CNTNAP2, was found to be a neurodevelopment gene and was associated with multiple phenotypes, including autism, intellectual disability, schizophrenia, epilepsy, sequential learning, phonology, and LIs (Poliak et al., 1999; Strauss et al., 2006; Whitehouse et al., 2011).

In summary, as shown by the CFSRS and others (Rvachew & Matthews, 2024; Shriberg et al., 1994; Stackhouse, 1992), SSDs have considerable phenotypic diversity with developmental changes based on age, all of which may not be captured in larger studies. While larger studies provide more statistical confidence in outcomes, they may only capture a few facets of this multifactorial disorder. Small focused collections with deep phenotyping provide the test bed to identify the most pertinent, perhaps novel, phenotypes. These phenotypes may then be assessed in larger cohorts with a reduced number of tests, validating their importance as key features of a disorder.

Phenotypes and Endophenotypes Associated With Genetic Studies of SSD

Genetic studies of speech and language disorders have employed diverse phenotypes, different age groups of probands, and various diagnostic criteria for determining affection status. In young children, ages birth to 5 years, phenotypes may be based on historical reports of speech/language delays or disorders by parents and direct assessment by SLPs. In older children (6–12 years) or adolescents/young adults (12+ years), identification of phenotypes may be more difficult as the speech and language disorder may have resolved; historical reports may not be accurate; and difficulties in related areas, such as literacy, may not have been identified. Additionally, while some older individuals may exhibit persistent speech sound errors on later-developing sounds, which are readily observed, others may have difficulty producing multisyllabic words or complex sentences, which are not as apparent and may go undetected (Lewis et al., 2019).

For genetic studies, endophenotypes related to SSD may be more useful than composite test scores, often used to diagnose phenotypes. Endophenotypes are presumed to assess one skill of a clinical phenotype and are thought to be more directly related to genes. For example, one class of endophenotypes related to SSD is phonological processing skills, composed of phonological awareness, phonological memory, and rapid automatized naming (RAN). These skills reflect an individual's ability to recognize, store, retrieve, and manipulate speech sounds and are thought to underlie the broader phenotype of SSD.

Understanding the various endophenotypes associated with a speech or language disorder may result in an earlier, more precise diagnosis of a disorder and enable targeted intervention. One such endophenotype is phonemic awareness, a metalinguistic skill that involves recognizing, isolating, and manipulating phonemes. It is a skill that can be targeted early in intervention and plays a critical role in the development of reading and writing skills (Lewis et al., 2011). As shown in Figure 1, structural equation modeling was used to examine how endophenotypes predict school-age language and literacy outcomes. The arrows represent hypothesized causal pathways. Phonological awareness skills, especially phonemic awareness, had the greatest impact on school-age outcomes of spelling, written language, reading, and spoken language. Endophenotypes predicted school-age literacy over and above what was predicted by a clinical diagnosis of SSD or LI. Findings suggest that these endophenotypes and common genetic influences affect early childhood SSD and later school-age literacy skills. See Lewis et al. (2011) for details.

Figure 1.

The relationship between endophenotypes and school-age phenotypes. The figure shows the relationships among endophenotypes and their impact on phenotypes assessed at school age. These findings are based on a structural equation model that examined the impact of speech sound disorder (SSD) and language impairment (LI) on school-age outcomes (Lewis et al., 2011). The arrows represent hypothesized causal pathways. The numbers above the line indicate regression coefficients estimated from the model. Bolded arrows and numbers represent pathways that are statistically significant at p < .05. Early phonological awareness skills impact school-age literacy. Other skills, such as speeded naming and vocabulary, are related to spoken language. Oral motor skills were related to articulation and written language, possibly due to poor handwriting.

Another important endophenotype linked to SSD and its common comorbidity, literacy development, is phonological encoding. Difficulty with phonological encoding, often measured by multisyllabic word repetition, may be related to deficits in phonological memory, which impact learning to read (Catts, 1986). While speech sound production may be a phenotype useful for accurately identifying young children with SSD, the endophenotypes that impact literacy skills may be more useful in determining affection status for older children, adolescents, and adults. Comorbidities of SSD, such as reading disorder (RD) and LI, are best explained in terms of multiple endophenotypes related to multiple phenotypes. As described below, the CFSRS evaluated multiple phenotypic and endophenotypic traits of speech, language, and co-occurring conditions such as dyslexia, following individuals into adolescence and young adulthood.

The CFSRS

The CFSRS is a 25-year study that examined the phenotypic and genetic bases of SSD (Lewis et al., 2019, 2024); participants were enrolled from 1989 to 2014. Children with a diagnosis of SSD were referred to the study from the clinical caseloads of SLPs in Ohio. Families (N = 296) with a child with an SSD were enrolled in the study. One hundred and sixteen of these families had a child diagnosed with CAS. DNA samples were provided by 73 families with a CAS proband and 199 families with other SSDs.

Most participants in the CFSRS were tested at least twice, beginning at preschool and again at school age, and followed into adolescence/young adulthood. Testing was conducted individually in two approximately 3-hr sessions in a speech research laboratory at Case Western Reserve University or, at a parent's request, in a quiet and adequately lit room in the family's home. These studies were approved by the Case Western Reserve Institutional Review Board: Study Number 20240534, “Speech and Reading Study,” and Study Number 20230490, “Phonology Genetics.” Informed consent and assent were obtained from all participants prior to testing. See Table 1 for a summary of participants in the CFSRS.

Table 1.

Description of participants in the Cleveland Family Speech and Reading Study (CFSRS).

Total number of families enrolled in the CFSRS	296
Total number of probands in the CFSRS	296
Total number of siblings in the CFSRS	543
Number of siblings with SSD	194
Number of siblings no SSD	335
Number of siblings unknown SSD status	14
Number of probands with 1 time point, 2 time points, 3 time points, 4 or more time points	n = 114 one visit n = 91 two visits n = 74 three visits n = 17 four or more visits

Note. Participants were enrolled in the CFSRS between 1989 and 2014. All probands and available siblings were tested at each follow-up visit. Parents completed a family history questionnaire and developmental questionnaire at each visit. SSD = speech sound disorder.

General Inclusion Criteria for CFSRS

Study participants in the CFSRS study met the following general inclusion criteria: fewer than six episodes of otitis media prior to the age of 3 years, being a monolingual English speaker, and absence of a genetic syndrome or a history of neurological disorders other than CAS. An additional inclusion criterion was documentation of normal intelligence as defined by a prorated performance IQ of at least 80 on the Wechsler Preschool and Primary Scale of Intelligence–Revised (Wechsler, 1989), the Wechsler Intelligence Scale for Children–Third Edition (Wechsler, 1991), or the Wechsler Intelligence Scale for Children–Fourth Edition (Wechsler, 2003). Eligible children were also required to pass a pure-tone audiometric screening at 20 dB HL bilaterally.

Inclusion Criteria for CAS

A subset of children referred to the CFSRS had a history of a CAS diagnosis by SLPs. A review of parent intake questionnaires indicated that the diagnosis of CAS occurred between 2.5 and 5 years of age, which aligns with the age range commonly cited for a diagnosis of CAS (American Speech-Language-Hearing Association [ASHA], 2007). Upon entry into the study, a diagnosis was confirmed by the ASHA-certified SLP serving on the research team with experience in motor speech disorders and reconfirmed by a second ASHA-certified SLP. Although the majority of participants with CAS were enrolled prior to the ASHA 2007 CAS consensus criteria, the criteria used to confirm a diagnosis of CAS in the CFSRS were mapped onto the ASHA criteria and the Murray et al. (2014) CAS indicators for studies prior to 2007.

A subset of children referred to the CFSRS had a history of a CAS diagnosis by SLPs. A review of parent intake questionnaires indicated that the diagnosis of CAS occurred between 2.5 and 5 years of age, which aligns with the age range commonly cited for a diagnosis of CAS (American Speech-Language-Hearing Association [ASHA], 2007). Upon entry into the study, a diagnosis was confirmed by the ASHA-certified SLP serving on the research team with experience in motor speech disorders and reconfirmed by a second ASHA-certified SLP. Although the majority of participants with CAS were enrolled prior to the ASHA 2007 CAS consensus criteria, the criteria used to confirm a diagnosis of CAS in the CFSRS were mapped onto the ASHA criteria and the Murray et al. (2014) CAS indicators for studies prior to 2007.

A subset of children referred to the CFSRS had a history of a CAS diagnosis by SLPs. A review of parent intake questionnaires indicated that the diagnosis of CAS occurred between 2.5 and 5 years of age, which aligns with the age range commonly cited for a diagnosis of CAS (American Speech-Language-Hearing Association [ASHA], 2007). Upon entry into the study, a diagnosis was confirmed by the ASHA-certified SLP serving on the research team with experience in motor speech disorders and reconfirmed by a second ASHA-certified SLP. Although the majority of participants with CAS were enrolled prior to the ASHA 2007 CAS consensus criteria, the criteria used to confirm a diagnosis of CAS in the CFSRS were mapped onto the ASHA criteria and the Murray et al. (2014) CAS indicators for studies prior to 2007.

CFSRS Deep Phenotyping

The CFSRS assessed a wide range of phenotypes and endophenotypes for different facets of speech, language, and other comorbid conditions, such as dyslexia. As shown in Table 2, the assessed phenotypes included speech sound production, spoken language, and written language. For a complete description of the test battery, see Lewis et al. (2004) for preschool and school-age measures and Lewis et al. (2015) for adolescent/young adult measures. Endophenotypes hypothesized to contribute to these phenotypes were phonological awareness, phonological memory, speeded naming, vocabulary, and motor speech skills. At each visit (early childhood, school age, and adolescent/young adulthood), age-appropriate measures were administered. As new versions of standardized tests became available, they were incorporated into the test battery.

Table 2.

Measures of endophenotypes and phenotypes at early childhood and school age/adolescent follow-up.

Endophenotype	Time 1 measures (early childhood)	Time 2 and 3 measures (school-age/adolescence)
Phonological awareness	CTOPP Elision and Blending Words	CTOPP Elision and Blending Words
Phonological memory	Nonsense word repetition	Nonsense word repetition
Speeded naming	CTOPP Rapid Color Naming	CTOPP Rapid Color Naming
Vocabulary	PPVT	PPVT
Motor speech skills
Oral motor	Robbins & Klee Oral Motor Protocol	Fletcher Time-by-Count
Syllable sequencing	Syllable Repetition Task	Syllable Repetition Task
Phenotypes	Time 1 measures (early childhood)	Time 2 and 3 measures (school-age/adolescence)
Speech sound production	GFTA Multisyllabic Word Repetition	GFTA Multisyllabic Word Repetition
Spoken language	CELF-P:2 or TOLD-P	CELF-3
Written language	Not assessed at Time 1	WRMT-R (Word Attack and Word ID); TWS-3; TOWRE-2

Note. CTOPP = Comprehensive Test of Phonological Processing; PPVT = Peabody Picture Vocabulary Test; GFTA = Goldman–Fristoe Test of Articulation; CELF-P:2 = Clinical Evaluation of Language Fundamentals–Preschool: 2nd Edition; TOLD-P = Test of Language Development–Primary; CELF-3 = Clinical Evaluation of Language Fundamentals–Third Edition; WRMT-R = Woodcock Reading Master Test–Revised; TWS-3 = Test of Written Spelling–3rd Edition; TOWRE-2 = Test of Word Reading Efficiency–2nd Edition.

Summary of Adolescent/Young Adult Outcomes in the CFSRS

Few studies have longitudinally tracked individuals with SSD of unknown etiology into adolescence and beyond. Below, we present our findings of the adolescent/young adult follow-up studies comparing individuals with persistent errors to those without persistent errors on phenotype and endophenotype measures. Next, we present our findings for a subset of participants who were diagnosed with CAS in early childhood. The data presented in the studies below were previously reported by Lewis et al. (2019, 2024). For full details of these studies, please refer to these sources.

Persistence of Participants with SSD

Using early childhood assessments, participants were assigned to the following groups: SSD-only, SSD and LI, and No SSD. Siblings of participants with no history of SSD, LI, or CAS comprised the No SSD group. At the adolescent/young adult follow-up, participants were assigned to the following four groups based on whether they had errors in conversational speech and their performance on the multisyllabic word repetition task: Low Multisyllabic Word Repetition (MSW) group (n = 33) had no errors in conversational speech but performed less than or equal to −1.5 SDs from the mean on the multisyllabic word, Persistent SSD group (n = 41) demonstrated speech errors in conversation and performed less than or equal to −1.5 SDs from the mean on the multisyllabic word, Resolved SSD group (n = 105) performed greater than −1.5 SDs from the mean on the multisyllabic word, and No SSD group (n = 137 siblings) had no errors in conversation and scored greater than 1.5 SDs from the mean on the multisyllabic word.

Adolescent/young adult groups were compared using an analysis of variance (ANOVA) based on early childhood classifications. To account for multiple testing, we conservatively corrected for 18 ANOVA and set the α level at .0028. Significant group effects were followed by Tukey's post hoc comparisons to determine how the groups differed from one another.

Longitudinal follow-up of 179 individuals enrolled in the CFSRS at 4–6 years of age and followed at 11–18 years showed significant differences on all measures by adolescence based on group assignment (Resolved SSD, Low MSW, and Persistent SSD) at p < .001 (see Table 3). Tukey's post hoc comparisons showed that the Low MSW and Persistent SSD groups performed more poorly than the Resolved SSD group on all measures. All three SSD groups scored more poorly on the Word Attack and the nonword repetition (NWR) measures than the No SSD group. The Low MSW group performed better than the Persistent SSD group on the NWR and Fletcher time-by-count tasks, suggesting that the Persistent SSD group may have poorer oral motor skills. Adolescents with persistent SSD also had higher rates of comorbid LI and RD than those with no history of SSD or those whose SSD had resolved by adolescence (Lewis et al., 2015). Better phenotyping of older children and adolescents is essential for developing targeted interventions for these at-risk individuals. For details of this study, see Lewis et al. (2015).

Table 3.

Adolescent outcomes for adolescents with histories of SSD-only, SSD + LI, and No SSD at early childhood.

Variable	Adolescent outcome groups				F	p value
	No SSD n = 137	Resolved SSD n = 105	Low MSW n = 33	Persistent SSD n = 41
Early Childhood Group	No SSD = 137 (100%)	SSD-only = 65 (62%) SSD + LI = 40 (38%)	No SSD = 3 (9%) SSD-only = 5 (15%) SSD + LI = 25 (76%)	No SSD = 6 (15%) SSD-only = 7 (17%) SSD + LI = 28 (68%)	N/A	N/A
Male:female ratio	68:69	72:33	22:11	29:12	N/A	N/A
Age (SD)	13.96 (2.09)	14.02 (1.93)	13.53 (1.93)	13.43 (1.77)	0.20	.818
NWR,a,b,c,d,e,f M (SD)	77.96 (16.27)	71.10 (16.78)	49.36 (19.81)	36.90 (22.36)	65.62	< .001
Fletcher,c,e,f M (SD)	−0.36 (0.82)	−0.26 (0.69)	0.04 (0.91)	0.60 (1.21)	12.84	< .001
Word Attack,a,b,c,d,e M (SD)	105.66 (11.81)	100.93 (9.05)	84.03 (19.05)	86.73 (14.47)	42.08	< .001
Word ID,b,c,d,e M (SD)	104.60 (13.84)	100.90 (10.17)	81.78 (20,86)	84.77 (13.78)	38.01	< .001

Note. SSD = speech sound disorder; LI = language impairment; MSW = multisyllabic word repetition; N/A = not applicable; NWR = Nonword Repetition; Fletcher = Fletcher Time-by-Count Test of Diadochokinetic Syllable Rate; Word Attack = Word Attack Subtest of the Woodcock Reading Mastery Tests–Revised; Word ID = Word Identification Subtest of the Woodcock Reading Mastery Tests–Revised.

^aNo SSD group differs from Resolved SSD group.

^bNo SSD group differs from Low MSW group.

^cNo SSD group differs from Persistent SSD group.

^dResolved SSD group differs from Low MSW group.

^eResolve SSD group differs from Persistent SSD group.

^fLow MSW group differs from Persistent SSD group.

Persistence of Participants With CAS

A second study was conducted with adolescents/young adults who had an early history of CAS (n = 32). Although 13 participants with CAS did not participate in the adolescent/young adult follow-up, these participants did not differ from those who were followed in oral motor skills; socioeconomic status, as measured by the Four Factor Index of Social Class (Hollingshead, 2011); or age assessed at the school-age visit. The CAS subset of participants in the follow-up study represents children with the same demographic profile and baseline characteristics as the full CAS cohort. We do not believe that we are limiting the generalizability (external validity) of our results as the follow-ups were conducted irrespective of the persistence status. Our statistical tests are conservative because it is likely that a greater number of children who did not participate in follow-up testing were remediated, and our results would have been more significant had we been able to include that data. At adolescence, participants were classified as persistent (n = 19) or nonpersistent (n = 13). Persistent speech errors were defined as a score of less than the 16th percentile on the Goldman–Fristoe articulation test and/or greater than four errors in conversational speech. Four participants with persistent errors in early adolescence had resolved by 16 years of age.

Welch's two-sample t test was used to compare the persistent and nonpersistent groups for variables that were normally distributed (i.e., Word Identification and Word Attack subtests of the Woodcock Reading Mastery Tests and the Fletcher time-by-count). The Wilcoxon–Mann–Whitney nonparametric test was used to compare groups that were not normally distributed (i.e., MSW, NWR, Elision, and RAN).

Results showed that significantly more persistent than nonpersistent participants had low scores on the Fletcher time-by-count measure (p = .008), suggesting that persistent participants had poorer oral motor skills than nonpersistent participants (Lewis et al., 2024). In addition, the persistent participants demonstrated significantly more auditory-perceptual features of dysarthria, as identified by Iuzzini-Seigel et al. (2022), such as hypernasality and imprecise articulation, than the nonpersistent participants (Lewis et al., 2024). Regardless of the persistence of speech errors, difficulties were observed in multisyllabic word repetition, phonological processing, and literacy for both groups. As shown in Figure 2, both the persistent and nonpersistent groups had mean and median scores below a standard score of 90 on the Word Identification and Word Attack subtests, suggesting that both groups struggled with reading skills.

Figure 2.

Comparison of persistent and nonpersistent participants on endophenotypes. The figure shows the group performance on measures. Mean scores of groups are indicated by the dotted lines. Median scores of groups are indicated by solid lines. As shown in the graph, the persistent speech sound disorder group scored below the nonpersistent group on word identification, elision, rapid color naming, nonword repetition (NWR), multisyllabic word repetition (MSW), and the Fletcher time-by-count, with only the rapid color naming and the Fletcher time-by-count reaching statistical significance. Scores of the nonpersistent group fell below a standard score of 100 for all measures, suggesting continued weakness in these endophenotypes.

Genetic Studies in CFSRS

Genetic Analyses and Findings of the CFSRS Participants

We used multiple genetic methods in our analyses, including GWAS (Benchek et al., 2021) and WGS (Chan et al., 2024), to identify novel genes and to replicate previously identified genes associated with speech, language, and literacy disorders. See Appendix A for a summary of the techniques used for genetic testing. We utilized WGS and microarrays and discovered that unusual structural variants (SVs) were present in SSD cases (Chan et al., 2024); rare single nucleotide variants (SNVs) were also observed (unpublished data). We detected 19 very large SVs, mostly deletions, considered to be karyotypically visible, and five extreme loss (or gain) of function SNVs. Our data show that CAS, and more broadly SSD, represents a disparate group of neurologic conditions with diverse genetic etiology (i.e., affecting cell types or neural tissue). Thus, each family has its own mutation and genes with little overlap among them. For example, we identified two independent individuals with CAS who have a de novo deletion > 600 kb in 16p11.2 spanning 28 genes associated with various developmental disorders, including autism. We also detected three different chromosome 4 deletions associated with diverse neurologic phenotypes. Using bioinformatics and a search of the literature for other reports, we are reasonably confident of the association between CAS and deletions on chromosomes 2q24.3, 6p12.3-p12.2, 11q23.2-q23.3, and 16p11.2. Weaker evidence was present for deletions on 2q11.2 and 22q11.1. Importantly, we identified novel deletions that are not commonly associated with SSD. While we discovered these SVs in children with CAS, we also found them in children with other SSDs, illustrating genetic heterogeneity and variable expressivity. We created a compendium of possible loci (with or without syndromic features) as a basis for generating hypotheses about the genetic architecture of SSD (Chan et al., 2024). Several individuals with the 16p11.2 deletion had developed the deletion de novo, not inheriting this stretch of the chromosome from either parent (see Figure 3).

Figure 3.

Example of a de novo deletion in a child with CAS. The figure shows the pedigree of a child with CAS who presented with a 16p11.2 de novo deletion. The parents and sister reported no speech and language difficulties and did not have this mutation. ADHD = attention-deficit/hyperactivity disorder; CAS = childhood apraxia of speech; LD = language disorder; WGS = whole genome sequencing.

We also found that children with SSD displayed varying levels of deficits in vocabulary knowledge, reading ability, and language scores and examined these phenotypes using GWAS (Benchek et al., 2021). We were able to find new loci and replicate prior results. Our strongest replication signal was observed for the ATP2C2 locus, with the trait of single word reading, showing shared genetic etiology with reading phenotypes. Children with CAS were included in the GWAS, so these loci may be shared between SSD and CAS. Longitudinal follow-up of children with SSD may demonstrate reading deficits among those who carry this variant. Using cross-trait analyses, we identified new loci. The strongest signal was at the IFI6 locus. Fine-mapping is needed to determine the clinical impact of this locus. For example, finding rare protein-coding changes in genes near these signals may help pinpoint specific loci and help us to refine the genetic architecture of SSD. It is also possible that the variants discovered at the GWAS loci may regulate distant genes and will require alternative approaches (Long et al., 2025).

Summary and Conclusions

This article reviewed the CFSRS and long-term outcomes for adolescents and young adults who participated in the CFSRS. Phenotypes and endophenotypes for SSD of unknown etiology that are useful for genetic studies were presented. Genetic findings from the CFSRS using participants diagnosed with CAS were also discussed. Below, we outline the implications of these findings for clinical practice and future directions.

Family History Questionnaires Do Not Fully Define Affectation Status

A comprehensive speech and language evaluation typically includes a developmental questionnaire and/or a parent interview. Parents and caregivers are sometimes asked about the affectation status of other family members, including themselves, in a family health history questionnaire. Family history can also be recorded by drawing a family tree or pedigree. Family history is easily accessed, inexpensive, and a tool for those without access to genetic testing. It may also lead to greater sharing of information among family members who may be separated geographically (Wildin et al., 2021). While family history data are useful for collecting information about both rare and common disorders and allow for early assessment and preventive interventions, it may not always provide an accurate assessment of familial disorders. The accuracy of the family health questionnaire or pedigree may vary depending on the informant's recall and the availability of records to corroborate the information. For example, in the CFSRS, comparisons of parent self-report of childhood disorders and direct testing revealed that parent report did not match direct testing for 10% of reading difficulties and 18% of language difficulties. Another issue with the use of questionnaires to determine affectation status is that there is no standard questionnaire employed across studies, making it difficult to confirm reports of a disorder and understand the criteria used for classification. This is especially relevant for older children, adolescents, and adults who may no longer present with a disorder.

The CFSRS adopted a multi-method approach for determining an individual's phenotype that combined questionnaire data and direct testing, including testing of adult parents, which may be best practice for genetic studies. Additional suggestions for improving the accuracy of family history questionnaires for genetic studies include the following:

• have immediate family members complete the questionnaire prior to the scheduled testing to allow time to contact relatives to verify affection status of family members,

• interview the parent and review the questionnaire in person to ensure the informant understands the questions and to obtain details about the diagnoses of family members,

• draw the family pedigree with the parent providing confirmation of the relationship of family members to the proband and identifying the relationships among members (such as adoption, half-siblings, etc.), and

• update the pedigree at each follow-up.

Although family history questionnaires are important to establish affectation status, standardized testing is a reliable and valid way to diagnose a disability (Paul et al., 2018). However, few measures are available for phenotyping older children, adolescents, and adults.

Lack of Measures to Assess Phenotypes and Endophenotypes Across the Lifespan

Few studies have examined the long-term outcomes for individuals with childhood histories of SSD, with and without comorbid LI. However, several studies suggest that difficulties may persist into adolescence and adulthood (Flipsen, 2002; Johnson et al., 2010; Lewis et al., 2015, 2019, 2024). Long-term outcomes for individuals with a history of a severe SSD, such as CAS, may provide insight into phenotypes that can differentiate those with favorable outcomes from those with persistent deficits. However, phenotyping adolescents and adults may be difficult due to the lack of standardized measures with normative data for speech skills at older ages. For example, single-word articulation tests composed of isolated and phonetically less complex words, such as the Goldman–Fristoe Test of Articulation (Goldman & Fristoe, 1986, 2000), may not capture SSDs in older individuals. Adolescents/young adults might be more accurately assessed using measures that evaluate speech production in more phonetically and linguistically complex contexts where they may exhibit difficulties with syllable sequencing that are not observed in existing standardized measures developed for younger children. Unfortunately, there is a lack of normative data for these repetition tasks, limiting both their clinical and research utility in older populations.

Nevertheless, even in the absence of normative data for older individuals, tasks that tap underlying skills, such as multisyllable word repetition, may still be valuable for genetic studies. As demonstrated in our genetic studies and those of others, endophenotypes are more closely related to genes than composite scores (Barry et al., 2007). For genetic studies, composite scores from standardized tests that assess multiple skills may not be as useful in detecting genetic variants as endophenotypes. For example, some adolescents and adults may experience difficulties with reading and spelling due to poor phonological awareness skills. Others may have comorbid LIs that contribute to poor reading and spelling skills. Both may score comparably on standardized tests of reading and spelling yet have different underlying etiologies.

Phenotyping older children, adolescents, and adults is essential for both genetic studies and clinical practice. Accurate identification of affected family members is necessary for identifying genetic contributions to disorders. Determining the genotype of affected family members may lead to targeted interventions and early diagnosis and intervention of younger family members. Moreover, genetic studies may reveal a syndrome that has not yet been identified within the family. Finally, assessment of adolescents and adults may reveal factors associated with long-term outcomes for individuals with early histories of speech and language disorders. This may allow for earlier identification of younger children at risk for persistent difficulties so that these students may be monitored into adolescence.

Creating Standardized Measures for Phenotyping and a Variant Database for Use in Precision Medicine

The Human Phenome Project was proposed in Freimer and Sabatti (2003). The goal of the project was to standardize measures for assessing phenotypes and to correlate them to genotypes and responses to treatments. The use of standard measures and terminology would allow data sharing and combining data sets for genetic analyses. Standard criteria for diagnosing speech and language disorders may assist clinicians in subtyping disorders and allow for a detailed description of a phenotype. For example, ASHA has provided some guidance for diagnostic criteria of disorders, such as CAS (ASHA, 2007). These diagnostic criteria have improved the accuracy of diagnosing young children with CAS. However, diagnostic criteria for older children, adolescents, and adults have yet to be determined. Deep phenotyping is needed to identify individuals most at risk and provide a better understanding of the spectrum of the disorder by comparison to other individuals with CAS. The genetics of CAS have been shown to be heterogeneous, underscoring the necessity of deep phenotyping to differentiate subtypes of CAS.

Creating a database of variants that are causal will be necessary for the practice of precision medicine (Fisher et al., 1998; Morgan et al., 2024). For some conditions, testing only a single gene is sufficient to identify the problematic variants. However, SSDs are very heterogeneous at the genetic level, and examining the entire genome will be necessary to find the most impactful variants.

Clinical Implications for SLPs

SLPs need to become better educated in the application of genetics to clinical practice. A survey study found that while SLPs and audiologists indicated that genetics may be useful in clinical practice, they require more training in genetics (Peter et al., 2019). To increase an SLP's comfort level regarding genetics, Lauretta et al. (2023) suggest incorporating requirements for genetics and genomics knowledge into the requirements for professional certification as an SLP. Additionally, making continuing education development pertaining to genetics and clinical practice more available would also be beneficial. To support clinical application, collaboration with other professionals, such as genetic counselors, will further support an SLP's understanding of genetics and may contribute to improved diagnosis, prognosis, and intervention. Individuals with severe and persistent disorders may be better identified earlier. An understanding of subtypes within diagnostic groups may lead to more focused therapy. Deep phenotyping may create a spectrum of deficits for speech and language disorders so that variability within a diagnostic category may be recognized.

Clinical Implications of Genetic Findings

Children with SSDs may carry both common and rare genetic variants. Some of these variants are causal, while others may be merely correlative proxy markers. Novel common variants in genes were identified via our GWAS (e.g., IFI6 region). We also confirmed some genes discovered by other groups (e.g., ATP2C2, CNTNAP2). Replication of previously discovered genetic findings is the cornerstone of human genetics. Through WES or WGS, rare variants can be carefully followed to determine if they are causal.

We discovered that many CAS events may be due to structural genomic variation (e.g., deletions, duplications, and other chromosomal rearrangements; Chan et al., 2024). We also discovered that de novo variants play a significant role in CAS. These de novo variants include both SVs (Chan et al., 2024) and rare protein-coding variants. Case reports and case series support our conclusion (Fisher et al., 1998; Lai et al., 2000, 2001; Peter et al., 2017, 2019; Shriberg et al., 2008). We recommend that all individuals with CAS undergo genetic screening at a minimum using cytoarrays or broader scans via WES or WGS. Genetic screening in research may lead to the discovery of new genes, while genetic screening in a clinical setting will require careful interpretation of the results with patients (AlAbdi et al., 2023; Long et al., 2025).

Limitations and Future Directions

Several limitations and the need for future research should be noted. First, our sample size is relatively small. Therefore, data sharing is essential to create large enough data sets needed for new genetic technologies. Also, we did not examine the implementation of genetic findings into clinical practice. Exploring the usefulness of genetic findings is required to advance clinical practice. Studies are needed to explore ways to integrate genetic information for precision diagnosis, prognosis, and therapy. Research is also needed to extend phenotype assessment into adulthood. Our adult sample size was too small for statistical analyses. Future studies may examine adults with genetic variants and describe long-term outcomes.

While the NWR and multisyllabic word repetition tasks distinguished individuals with persistent errors from nonpersistent speech errors, these measures were not standardized on a control sample without any family history of SSD. Future studies may standardize some of the endophenotype measures that we have found useful in genetic studies. Future studies may also include imaging and other methods to examine how genetic variants are expressed in the brain.

Author Contributions

Barbara A. Lewis: Conceptualization, Data curation, Methodology, Writing – original draft. Gabrielle J. Miller: Conceptualization, Data curation, Methodology, Writing – original draft. Penelope Benchek: Conceptualization, Data curation, Methodology, Writing – review & editing. Catherine Stein: Conceptualization, Data curation, Methodology, Writing – review & editing. Sudha K. Iyengar: Conceptualization, Data curation, Methodology, Writing – original draft, Writing – review & editing.

Data Availability Statement

The data that support the findings of this article are available on request from the corresponding author. The data are not publicly available due to the institutional review board's restrictions on the research participants' permission.

Appendix A

Techniques for Genetic Testing

There are a number of genetic technologies that can assay (i.e., test or analyze) different components of the human genome (reviewed in Porubsky & Eichler, 2024), some of which are listed in Table B1. Typically, these range from DNA microarrays to short-read whole exome sequencing (WES) and short- and long-read whole genome sequencing (WGS).

Standard genetic testing is conducted in two different contexts. The first is in the clinical setting, in which patients seen in medical settings are evaluated for genetic conditions or variants. This type of testing has very rigorous standards for interpreting variants that may be causal. Usually, there is significant prior evidence for the causal relationship between the phenotype and the genotype behind the report. Many developmental disorders and syndromes fall into this category. However, for a speech sound disorder (SSD) such as childhood apraxia of speech (CAS) without a reported syndrome or prior genetic testing reported by parents or physicians, there is a void in knowledge. In the United States, the American Academy of Medical Genetics and Genomics (ACMG) produces guidelines for variant interpretation (Richards et al., 2015). The ACMG has significant global outreach, and its guidelines have been adopted by many other countries, such as Canada, the United Kingdom, Australia, and many places in Europe.

Clinical genetic testing is federally regulated in the United States under the authority of the Food and Drug Administration, the Centers for Medicare and Medicaid, and the Federal Trade Commission (see https://www.genome.gov/about-genomics/policy-issues/Regulation-of-Genetic-Tests). As far as we are aware, clinical genetic testing, specifically for SSD/CAS using the standard panels, is not common. Some parents with children seeking clearer answers may have had prior genetic testing documented in their medical records. Alternatively, the statewide newborn screening programs in the United States (https://www.cdc.gov/newborn-screening/about/index.html) may have identified children for specific genetic issues, such as inborn errors of metabolism (e.g., galactosemia), which, in turn, were associated with SSD (Verhoef et al., 2024).

A second context for genetic testing operates in the research community, where original discoveries are being produced linking novel genetic markers and SSD/CAS. Some recent studies performed genetic testing for endophenotypes associated with SSD and common genomic variants via DNA microarrays and genome-wide association studies (GWAS; Verhoef et al., 2024). For example, the Cleveland Family Speech and Reading Study (CFSRS) conducted a GWAS for endophenotypes of SSD using 148 families. We identified multiple significant variants that we followed bioinformatically (i.e., using computer-based analyses). However, a weakness of GWAS is that it may not lead to a causal gene on the first pass. Even if results are replicated, the genetic variant may only be a proxy for another nearby variant and may point to a different causal gene. Many different methods have been used to narrow down the results to specific causal genes at a particular chromosomal location (Strober et al., 2025).

The results of GWAS are often summarized as a risk score, such as a polygenic risk score (PRS). For a multidimensional trait such as SSD, genetic risk can be borne by thousands of markers across the genome, each contributing a small amount of individual risk but jointly providing a reasonable risk predictor. A PRS is a sum of the effects of many genes to determine an individual's genetic risk for a disorder, such as speech and language disorders, compared to other individuals. PRSs can be used to classify children of varying severity and subtypes of SSD. They may also assist clinicians in identifying children most at risk for persistent speech problems and other comorbid conditions. PRSs have many practical applications, one of which is to stratify cohorts by various endophenotypes, particularly when the measure is not available on all individuals, thus replacing the phenotypic construct with the derived genetic assessment. However, to produce a PRS that can be generalized and has a reproducible effect in the test population requires a significant amount of validation (Maamari et al., 2025).

The oldest report of a rare variant (generally < 1%–2% in the population) is the single nucleotide variant in the FOXP2 gene in a family known as the K.E. family (Fisher et al., 1998). This finding was based on a combination of technologies, which included linkage mapping, sequencing, and cytogenetic arrays, as evidence came from the inheritance of this mutation in the K.E. family and several isolated cases with translocations that interrupted FOXP2. Contemporary approaches performed WGS or WES in small- to medium–sized samples. Rare variants were analyzed, with investigators focusing on variants that likely affect protein structure or function (i.e., classic mutations; Yasmin et al., 2024). Other studies examined individuals with a well-characterized deletion on chromosome 16p11.2 and found that the majority of subjects had a highly penetrant and severe form of CAS (Fedorenko et al., 2016; Raca et al., 2013); other neurodevelopmental conditions are also associated with this deletion. The CFSRS recently published work showing the role of copy number variants in CAS, including several cases with the chromosome 16p11.2 deletion (Chan et al., 2024).

In 20+ years of data collection, we have not found many families with two siblings with CAS; thus, de novo variants (i.e., variants that are new to the proband and not inherited from parents) may play a very important role in CAS. For example, we observed that several of the 16p11.2 deletions were de novo (Fedorenko et al., 2016; Raca et al., 2013). Infrequently, CAS has been reported in cytogenetic case reports, but most of the cases do not have detailed phenotyping (Boyar et al., 2001; Fanizza et al., 2014; Shriberg et al., 2008). Greater integration of deep phenotyping with genetic data will be essential to advance our understanding of the genetic architecture of CAS and related SSDs.

Appendix B

Study Designs used to Find Genetic Causes of Speech and Language Traits

Study design type	Technologies used for assessment	Newer technologies used for assessment
Syndromic patients with known or easily discoverable genetic variants where large chromosomal deletions/duplications or rearrangements are expected	Fluorescent in situ hybridization (FISH), chromosomal microarray analysis (CMA), and cytogenetic analysis (karyotyping)	Whole genome sequencing (WGS), may need long-read WGS
Single cases or case series of unknown origin where large chromosomal deletions/duplications or rearrangements are expected	Fluorescent in situ hybridization (FISH), chromosomal microarray analysis (CMA), and cytogenetic analysis (karyotyping)	Whole genome sequencing (WGS), may need long-read WGS
Single family or multiple families with SSD*,#	Fluorescent in situ hybridization (FISH), chromosomal microarray analysis (CMA), and cytogenetic analysis (karyotyping)	Whole exome sequencing (WES), Whole genome sequencing (WGS); may eventually need long-read WGS if cases remain unsolved
Medium to large-sized case–control or case-cohort studies (> 1,000 cases and a ratio of at least 1:5 for cases versus controls)	DNA microarray analysis	Whole exome sequencing (WES), Whole genome sequencing (WGS)

Note. To return results to patients, the testing must be performed in a Clinical Laboratory Improvement Amendments (CLIA) certified laboratory following the American College of Medical Genetics and Genomics Practice Guidelines (Ishida et al., 2024; https://www.acmg.net/ACMG/Medical-Genetics-Practice-Resources/Practice-Guidelines.aspx).

^*Having parental biospecimens and data allows for determination if the variant is de novo.

^#Multiple families, case–control and case-cohort designs can also be used for genome-wide association studies, WES and WGS in a research setting for new variant discovery.

Publisher Note: This article is part of the Forum: Research Symposium on Genetics in Communication Sciences and Disorders.

Funding Statement

This article stems from the 2024 Research Symposium at the ASHA Convention, which was supported by National Institute on Deafness and Other Communication Disorders Award R13DC003383. Research reported in this publication was supported by National Institute on Deafness and Other Communication Disorders Awards DC000528 to Barbara A. Lewis and R01DC012380 to Sudha K. Iyengar. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.