Meta-analyses of school-aged children with DLD demonstrate moderate group-level deficits in nonverbal cognitive performance compared with typically developing peers. Longitudinal cohort studies further indicate that even when nonverbal reasoning or language skills are within age expectations in preschool years, a substantial subgroup shows a relative developmental lag by school age or adolescence. Children and adolescents with DLD showed significantly slower processing speed than typically developing peers across both verbal and nonverbal tasks. These results are based on a single meta-analysis, while the other included studies focused on nonverbal reasoning and language development without direct measurement of processing speed. The overall quality of evidence is moderate, with consistent findings across study designs, but limitations include heterogeneity of measures and predominance of observational data.
| Reference | Study type | Population | Intervention and comparison | Outcomes | Risk of bias «Additional comments for included studies...»2 |
|---|---|---|---|---|---|
| RCT=randomized controlled trial; SR=systematic review; MA=meta-analysis; DLD= developmental language disorder; SLI=specific language impairment; TD=typically developing; LD= language disorder | |||||
| «Zapparrata NM, Brooks PJ, Ober T. Developmental La...»1 | Systematic review and meta-analysis | Children and adolescents with DLD, mean age 8.9 years (range: 4.3–22.7 years), recruited from multiple countries and clinical/research settings. 46 studies, n=812 (DLD) and n=870 (TD) | Observational comparisons between children with DLD and TD peers; no intervention. | Processing speed and nonverbal task performance (reaction/response time measures). | Moderate |
| «Gallinat E, Spaulding TJ. Differences in the perfo...»2 | Systematic review and meta-analysis | Children with SLI, age-matched to TD peers, across 131 studies (1995–2012), n=138 (SLI) and n=138 (TD). | Comparison of nonverbal cognitive test performance between SLI and TD peers. | Nonverbal IQ / nonverbal reasoning test scores. | Moderate |
| «Griffiths S, Kievit RA, Norbury C. Mutualistic cou...»3 | Longitudinal cohort study | Longitudinal cohort of children aged 7–13 years (n=501), with poor language at school entry. |
Observational longitudinal modelling of vocabulary and nonverbal reasoning development. | Change in nonverbal reasoning and language ability over time. | Low-moderate |
| «Botting N. Language, literacy and cognitive skills...»4 | Longitudinal cohort study | Young adults (n=83) aged 24 years with a history of DLD, originally recruited at age 7 from specialist language units in the UK; compared with age-matched TD peers. | Observational longitudinal follow-up; comparison between DLD group and age-matched TD peers. | Language ability, literacy skills, and nonverbal cognitive ability assessed longitudinally from adolescence to adulthood. | High |
| «Durkin K, Mok PL, Conti-Ramsden G. Core subjects a...»5 | Observational cohort study | 176 eleven-year-old children with SLI in England, compared with national norms. | Observational cohort study examining predictors of academic and cognitive outcomes. | Language ability, performance IQ, and school achievement at the end of primary school. | Moderate |
| «Lewis BA, Freebairn L, Tag J, ym. Adolescent outco...»6 | Prospective longitudinal cohort study | Adolescents (n=170) aged 11–18 years with histories of early childhood speech sound disorder (SSD) with and without comorbid language impairment, compared with TD peers (n=146). | Prospective longitudinal cohort comparison. | Language ability, literacy outcomes, and nonverbal cognitive ability in adolescence. | Moderate |
| Reference | Comments |
|---|---|
| «Zapparrata NM, Brooks PJ, Ober T. Developmental La...»1 | In the article, the term DLD has been applied as a descriptor across clinical groups,
children with language impairments whose nonverbal IQ score are 70 or higher receive
a diagnosis of DLD. The meta-analysis included observational reaction time studies
with heterogeneous processing speed tasks (verbal and nonverbal). Moderators included
task (simple RT, choice RT, naming, congruent/baseline conditions of interference
control tasks), stimulus type (linguistic/nonlinguistic), stimulus modality (auditory/nonauditory),
and response modality (verbal/nonverbal). Potential sources of bias include publication bias and small-study effects. Pooled estimates may be influenced by variability in task demands and unmeasured confounders such as attention, motor speed, or comorbid neurodevelopmental conditions. The proportion of heterogeneity (I2) attributed to true differences between the studies was 44%. The sensitivity analyses confirmed that the results did not change across the full range of possible rho values. With the two outlier effects from Oliveira et al. (2021) included in the analysis, there was evidence of publication bias (z = 3.40, p < 0.001). However, with the outliers removed, the funnel plot demonstrated symmetry of effects, with no evidence of publication bias (z = 1.45, p = 0.15). Hence, the two outlier effects were dropped from meta-analysis. The lack of a significant moderating effect of publication year suggests that the findings were not tied to specific diagnostic criteria (e.g., cutoff scores on nonverbal intelligence tests). Lack of information about comorbid diagnoses. |
| «Gallinat E, Spaulding TJ. Differences in the perfo...»2 | Included studies were required to meet the following criteria: (a) contained a conventional
operational definition or a clinical diagnosis of specific language impairment; (b)
examined monolingual, English-speaking children; (c) were written in English; (d)
were grouped to distinguish between SLI and TD groups; (e) matched the SLI and TD
groups for chronological age; and (f) administered a nonverbal IQ test to the SLI
and TD participants. The authors ensured that each included study used an explicit definition of SLI or a clinical diagnosis; however, they did not analyse or harmonise which language tests were used to define SLI, which cutoff criteria were applied, how strictly exclusion criteria (e.g. nonverbal intelligence, neurological factors) were enforced. Studies used varying diagnostic criteria for specific language impairment and a wide range of nonverbal cognitive tests, introducing clinical and methodological heterogeneity. Nonverbal IQ measures may partly reflect processing speed or working memory demands, resulting in indirectness. Residual confounding was not consistently controlled. Sensitivity analysis was performed and the effect size was relatively stable. |
| «Griffiths S, Kievit RA, Norbury C. Mutualistic cou...»3 | Findings are based on longitudinal latent change score modelling, which assumes measurement invariance across time. The current analysis uses data from receptive vocabulary and nonverbal reasoning assessments conducted when the children were in Year 3, Year 6 and Year 8. Language disorder was defined explicitly in the study (poor language at school entry within a longitudinal cohort), but not in terms of clinical, test-specific diagnostic criteria. Estimates depend on model specification, and causal inferences cannot be drawn. Attrition over follow-up may introduce bias, although the large sample size mitigates this risk. |
| «Botting N. Language, literacy and cognitive skills...»4 | Clinically recruited sample (language units); substantial attrition across the longitudinal
follow-up; use of language measures beyond their normative age range (CELF-4 at age
24); different nonverbal IQ instruments across time points, making developmental change
harder to interpret; modest sample size for adult group comparisons. Participants identified with primary language difficulties at age 7, with exclusion of neurological disorders, hearing impairment, and intellectual disability; DLD status in adulthood inferred from childhood diagnosis rather than re-established using adult cutoffs. Long-term follow-up from childhood to adulthood is subject to substantial attrition, with retained participants likely representing a higher-functioning subgroup. Recruitment from specialist language units limits generalisability. Changes in assessment tools across developmental stages may affect comparability over time. |
| «Durkin K, Mok PL, Conti-Ramsden G. Core subjects a...»5 | Historical dataset (late 1990s); reliance partly on teacher assessments rather than
standardized tests alone; heterogeneous educational placements; absence of a typically
developing control group (comparison to national norms instead); use of the SLI construct. Children recruited from language units; explicit exclusion of autism, neurological disorders, hearing impairment, and general learning disability; non-verbal IQ measured and treated as an exclusionary criterion. Outcomes were assessed at a single school-age time point, limiting conclusions about developmental trajectories. Academic outcomes may be influenced by educational context and support received, introducing confounding. Findings are based on one national education system, limiting external validity. |
| «Lewis BA, Freebairn L, Tag J, ym. Adolescent outco...»6 | The cohort included heterogeneous subgroups (speech sound disorder with and without
language impairment), reducing internal comparability. Clinically referred US sample; primary focus on speech sound disorder rather than language disorder; complex subgrouping (SSD-only, SSD+LI, persistent vs resolved); LI defined cross-sectionally rather than developmentally; limited cross-cultural generalisability. Language impairment defined as ≤ 8 scaled score on two or more language subtests, with non-verbal IQ > 80 required; LI treated as a comorbid condition rather than a primary developmental disorder. Longitudinal outcomes are subject to confounding by socioeconomic factors and baseline cognitive ability. Nonverbal reasoning was not a primary outcome, resulting in indirect evidence for the main review question. |
Results
| Reference | Number of studies and number of patients (I/C) | Follow-up time | Relative effect |
|---|---|---|---|
| I=intervention; C=comparison; CI=confidence interval, nonverbal reasoning = visual reasoning | |||
| «Gallinat E, Spaulding TJ. Differences in the perfo...»2 | 131 studies; 138 SLI and 138 TD | Cross-sectional | SMD −0.693 (SE = 0.16; 95% CI [−1.049, −0.336]), p = 0.0012 |
| «Griffiths S, Kievit RA, Norbury C. Mutualistic cou...»3 | N=501 | 5 years | Vocabulary predicted later growth in nonverbal reasoning (β ≈ 0.23) |
| «Botting N. Language, literacy and cognitive skills...»4 | ≈ 84 DLD / 88 peers | 16–24 years | Subgroup (~30%) showed persistent nonverbal IQ lag |
| «Durkin K, Mok PL, Conti-Ramsden G. Core subjects a...»5 | N = 176 | Age 11 years | Performance IQ predicted outcomes (β ≈ 0.40) |
| «Lewis BA, Freebairn L, Tag J, ym. Adolescent outco...»6 | N = 316 | 11–18 years | Lower nonverbal IQ associated with poorer outcomes (p < 0.01) |
| Level of evidence: moderate The quality of evidence is downgraded due to (study limitations, indirectness, imprecision). |
|||
| Reference | Number of studies and number of patients (I/C) | Follow-up time | Relative effect |
|---|---|---|---|
| I=intervention; C=comparison; CI=confidence interval | |||
| «Gallinat E, Spaulding TJ. Differences in the perfo...»2 | 131 studies | Cross-sectional | Language impairment co-occurred with nonverbal deficits |
| «Griffiths S, Kievit RA, Norbury C. Mutualistic cou...»3 | N=501 | 5 years | Vocabulary predicted later growth in nonverbal reasoning (β ≈ 0.23) |
| «Botting N. Language, literacy and cognitive skills...»4 | ≈ 84 DLD / 88 peers | 7–24 years | Large persistent language deficits into adulthood (d ≈ 0.8–1.0) |
| «Durkin K, Mok PL, Conti-Ramsden G. Core subjects a...»5 | N = 176 | Age 11 years | Language ability predicted English, Maths, and Science (p < 0.001) |
| «Lewis BA, Freebairn L, Tag J, ym. Adolescent outco...»6 | N = 316 | 11–18 years | Early language impairment predicted adolescent outcomes (p < 0.001) |
| Level of evidence: moderate The quality of evidence is downgraded due to study limitations, indirectness, imprecision. |
|||
| Reference | Number of studies and number of patients (I/C) | Follow-up time | Relative effect (95% CI) |
|---|---|---|---|
| «Zapparrata NM, Brooks PJ, Ober T. Developmental La...»1 | 46 studies; 812 DLD / 870 controls | Cross-sectional | SMD g = 0.47 (95% CI 0.38–0.55), p < 0.001 |
| Level of evidence: moderate The quality of evidence is downgraded due to (study limitations, indirectness and possible publication bias). |
|||
| I= intervention; C=comparison; CI=confidence interval, nonverbal reasoning = visual reasoning | |||