Sources of variability in the prospective relation of language to social, emotional, and behavior problem symptoms: Implications for developmental language disorder.

Sampling Weights and Missing Data

Inverse probability weighting is often utilized in large longitudinal studies to yield population representative samples by calculating weights that account for selective participation and adjusting them for inevitable nonresponse in subsequent waves. Likewise, inverse probability weighting is utilized in SCALES, as detailed elsewhere ( Gooch et al., 2019 ; Vamvakas et al., 2019 ). In brief, weights were constructed as the inverse of the probability of inclusion in the study from a logistic regression model fit to the entire screened population of 6,459 monolingual children attending mainstream schools; this model estimated the probability of inclusion for in-depth assessment at Year 1, with predictor variables including sex, season of birth, and scores on the Children’s Communication Checklist—Short. These weights were further adjusted for differential nonresponse/missing data at Year 3 by estimating a second logistic regression model fit to 529 children selected for in-depth assessment at Year 1. This model utilized predictors of missingness on the SDQ data ( n = 125) including IDACI rank score, SDQ total difficulties score, pupils on school role, percentage of children in school with special educational needs, and percentage receiving free school meals. The final sampling weights were the multiplication of the inverse of the probabilities from both logistic models ( Vamvakas et al., 2019 ). There were no differences between children with ( n = 363) and without ( n = 125) teacher-rated SDQ on age, sex, nonverbal cognition, DLD status, SDQ scores at reception, or SES (IDACI rank scores; online Supplemental Materials S2 ). Hence, the weighted models are representative of the monolingual cohort from which this sample was drawn.

Consent Procedures

Consent procedures and study protocol were developed in consultation with Surrey County Council and approved by the Research Ethics Committee at Royal Holloway, University of London, where this study, The Surrey Communication and Language in Education Study, originated. Ethical approval for continued data storage and analysis is provided by the University College London Research Ethics Committee (9733/002). For the screening phase, opt-out consent was employed as data could be provided anonymously; 20 families opted out. In the second phase, written, informed consent for two episodes of direct assessment, including teacher report of child language and behavior, was obtained from parents or legal guardians of participants. Prior to assessment in Year 3, families received an additional information sheet and the option to withdraw from the study; 18 families withdrew, five moved abroad, three could not be contacted, and three provided insufficient data at test for diagnostic classification. Of the 29 children (19 male) not included in follow-up, 22 had been classified as “typically developing” in Year 1 and had no evidence of language, learning, or behavioral difficulties.

Receptive/Expressive One-Word Picture Vocabulary Tests (R/EOWPVT-4; Martin & Brownell, 2011 )

ROWPVT and EOWPVT require word-to-picture matching and picture naming tests, respectively, with possible scores ranging from 0–190. Test–retest reliability is .97 for both measures, and internal consistency for ages 5 to 8 years is excellent (Cronbach’s alpha = .94–.97).

Test of Reception of Grammar—Short Form ( Bishop, 2003 )

Forty of the original 80 test items were included in which children heard a sentence such as “The ball that is red is on the pencil” and were asked to select the corresponding picture out of a choice of four. If a child answered incorrectly on six consecutive items, then the test was discontinued. Scores for this task range from 0 to 40, with excellent agreement between short and long forms in pilot testing, r (17) = .88.

School-Age Sentence Repetition Imitation Test—English (SASIT-E32)

The SASIT-E32 ( Marinis et al., 2011 ) asks the child to repeat prerecorded sentences of increasing length and grammatical complexity, played over headphones (possible score 0–32). Interrater reliability for scoring is excellent, .98 ( Chiat & Roy, 2013 ).

Assessment of Comprehension and Expression 6–11 ( Adams et al., 2011 )

In narrative recall, the child was asked to listen to a prerecorded story with accompanying pictures displayed on a laptop computer. After listening to the story, the child was asked to tell the story in their own words with the pictures displayed. The child was awarded 1 point for a maximum of 35 propositions accurately retold. Internal consistency is adequate (Cronbach’s alpha = .73) for children aged 6 to 11 years.

A bespoke measure of narrative comprehension was constructed in which the child was asked to answer 12 (six literal and six inference) questions about the story. Answers were scored on a 3-point scale (0 for an incorrect/no response, 1 for a partially correct response, and 2 points for a complete and accurate response) with a total possible score of 24. All scoring was done by consensus to ensure rater consistency. For all aforementioned language measures, test scores at Year 1 (ages 5–6) were utilized in this study.

Nonverbal Ability

Nonverbal ability (NVIQ) was measured using block design and matrix reasoning subtests. These were from the Wechsler Preschool and Primary Scales of Intelligence (3rd U.K. ed.; Wechsler, 2003 ) in Year 1 (ages 5–6) and the Wechsler Intelligence Scales for Children (4th U.K. ed.; Wechsler, 2004 ) in Year 3 (ages 7–9).

Socioeconomic Status

The IDACI scores were derived from household postcodes and provide an estimate of socioeconomic deprivation ( McLennan et al., 2011 ). Deprivation is defined as households receiving income support, jobseeker’s allowance, working or disabled person’s tax credits, or national asylum support whose equalized income is 60% below national median before housing costs.