Background
Attention Deficit Hyperactivity Disorder (ADHD) is a common childhood disorder characterized by symptoms of inattention, hyperactivity and impulsivity, that significantly impact many aspects of behavior as well as performance, both at school and at home [
1,
2]. Understanding the mechanisms by which such behavioral problems can develop may have important implications on early assessment, prevention, and treatment.
Developmental difficulties rarely occur in isolation [
3]. A close relationship between the development of Inattention/Hyperactivity (IH) symptoms and language skills has been consistently reported [
4,
5]. Cross-sectional studies found that children with ADHD have an increased prevalence of language impairments [
6,
7]. Several difficulties in linguistic skills have been reported among children with ADHD, particularly with regards to expressive language skills: phonology, vocabulary, syntax and pragmatic [
8‐
11]. Although data on this are somewhat inconsistent [
4], children with ADHD may also have deficits in receptive language skills [
12]. However, in longitudinal studies the association between early IH symptoms with later language skills has been found to be weak or absent [
5,
13]. Several authors have suggested that language difficulties could precede the development of ADHD and represent an early expression of the disorder [
14,
15].
Conversely, cross sectional studies found that children with language impairments have an elevated prevalence of ADHD [
15‐
18] as well as deficits in selective attention tasks, in particular in the auditory modality [
19]. Longitudinal studies have reported that early language difficulties are associated with later IH symptoms during the preschool [
13,
14] and school periods [
5,
20,
21], even when prior levels of IH symptoms are accounted for. Recent results of longitudinal studies support a causal role of language difficulties in the development of IH symptoms [
13]. Difficulties in language skills may be associated with ineffective use of self-directed speech for self-regulation, which may subsequently lead to IH symptoms (
Hypothesis 1). Following 120 children at 30, 36, and 42 months of age, Petersen et al. [
13] reported that the relationship between early language skills and later IH symptoms was mediated by language-based self-regulation during the preschool period. This result suggests that language functions (i.e., private or inner speech) may support behavioral and attentional control [
22]. Nevertheless, two other hypotheses for the association between early language skills and later IH symptoms have been proposed. The link between language skills and behavioral problems may be mediated by interpersonal difficulties (
Hypothesis 2) [
23‐
25], as poor language skills may interfere with socialization which may then lead to IH symptoms [
25]. Like all neurodevelopmental disorders, language disorders and ADHD are known to share some etiological factors (such as genetic or pre- and postnatal environmental factors) [
3,
6,
26]. A last hypothesis is that the common vulnerability has a sequential expression during the development by impacting first on language skills and later on behavior (
Hypothesis 3), creating the illusion of a directional effect between early language skills and later ADHD symptoms (i.e., heterotypic continuity [
27]).
Rather surprisingly, few of the previous studies [
18] have examined which aspects of early language skills are most strongly associated with the development of IH symptoms. Snowling et al. [
18] reported that children’s expressive language impairment at 5.5 years was the language profile most strongly associated with ADHD in adolescence. Researchers have called for more longitudinal studies to explore the association between language difficulties and IH symptoms and specify the underlying developmental processes [
28].
The preschool years are a crucial period in children’s psychological development. Previous studies support a significant instability of language skills between 3 and 5.5 years [
29]. For some children, the onset of behavioral, emotional and/or social problems occurs during this period [
30]. Addressing the stated research questions in preschoolers rather than in older children is of utmost importance since influences with respect to long-lasting outcomes may be more determinant during the first years of life, as suggested by the Developmental Origin of Health and Disease Hypothesis [
31,
32].
In the present study, we use data from a large (
N = 1459) prospective mother-child cohort to test bidirectional relationships between children’s language skills and inattention/hyperactivity (IH) symptoms between 3 and 5.5 years. We expect to replicate previous longitudinal studies [
13,
14], which found an asymmetrical association between language skills and IH symptoms during the preschool period (i.e., the association between language skills and IH symptoms was stronger than the reverse). If the influence of early language difficulties on the development of IH symptoms is mediated by an ineffective use of self-directed speech, language tests tapping into expressive language skills should be most strongly associated with later IH symptoms (
Hypothesis 1). Additionally, we also sought to test whether the association might be mediated by interpersonal difficulties (
Hypothesis 2) and whether shared pre- and postnatal environmental factors might explain both language skills and IH symptoms (
Hypothesis 3).
Results
Structure of language skills
The Confirmatory Factor Analysis model including 2 single latent factors representing language skills measured respectively by the five measures of language skills at 3 years and the five measures of language skills at 5.5 years provided an excellent fit to the data: CFI = 0.992, TLI = 0.988 and RMSEA = 0.031 (95 % CI [0.021, 0.041]). Both latent variables provide a general index of language skills, encompassing phonology, syntax, lexicon and conceptual knowledge, using both receptive and expressive modalities.
The stability of language skills between 3 and 5.5 years in our models was found to be high (β = 0.76). This result supports the idea that the latent variables for language at 3 years and 5.5 years reflect similar constructs.
Structural equation model
Our structural equation models displayed excellent fit to the data (see fit indices in Table
2).
In the four consecutive SEMs, IH symptoms at 5.5 years were significantly predicted by language skills at 3 years (standardized estimate in Model 4: β =−0.12, SE = 0.04,
p-value = 0.002). Language skills at 5.5 years were not associated with SDQ IH symptoms scores at 3 years (standardized estimate in Model 4: β = 0.04, SE = 0.03,
p-value = 0.105) (Table
2 and Additional file
1: Figure S2).
Next, we tested whether the association between language skills at 3 years and IH symptoms at 5.5 years was mediated by peer relationship problems or prosocial behavior (Hypothesis 2). We found no evidence of such mediation (peer relationship problems: Wald test = 0.48; p-value = 0.490; prosocial behavior: Wald test = 0.24; p-value = 0.621).
Beyond the effect of the latent variable representing language skills at 3 years on IH symptoms at 5.5 years, there were no significant direct effect of language tests at 3 years on IH symptoms at 5.5 years. The ranks of the total effects were the following: comprehension of instructions (β =−0.13), sentence repetition (β =−0.11), picture naming (β =−0.08), word and nonword repetition (β =−0.04) and semantic fluency (β =−0.03).
Sensitivity analyses
Examining language skills and IH symptoms when they were dichotomized to reflect potentially clinically significant problems (language skills: < -1 SD; SDQ IH symptoms scores > 6; i.e., 85th percentile) did not alter the significance of our results. Children with IH symptoms at 5.5 years which may reflect clinically significant problems (SDQ IH symptoms score above 6; 16.2 % of our sample) had significantly lower scores on language skills at 3 years (Model C: standardized estimate = -0.12;
p-value = 0.021; Table
3). One language test at 3 years (comprehension of instructions) was significantly associated with clinically significant IH symptoms at 5.5 (Models C5; Table
3; β = -0.11,
p-value = 0.030). The dichotomized (at < - 1 SD) language score at 5.5 years was not associated with SDQ IH symptoms scores at 3 years (Model F; standardized estimate = 0.03;
p-value = 0.601; Additional file
1: Table S2).
As individual differences in language skills (β = 0.77) were found to be more stable than individual differences in IH symptoms (β = 0.47), we conducted a sensitivity analysis a) removing the effect of language skills at 3 years on language skills at 5.5 years and b) removing the effect of IH symptoms at 3 years on IH symptoms at 5.5 years in Model 4. Under these conditions, the effect of language skills at 3 years on IH symptoms at 5.5 years (β = -0.18, SE = 0.08, p-value < 0.001) was much greater than the effect of IH symptoms at 3 years on language skills at 5.5 years (β = -0.01, SE = 0.02, p-value = 0.792) (Wald test of the difference = 7.61; p-value = 0.006), implying that the effect of language skills at 3 years on IH symptoms at 5.5 years is unlikely to be explained by differences in cross-time stability.
After verifying that the measurement parameters of the latent variables were sex-invariant, we used a multiple-group structural equation model stratified by sex and found no significant sex differences (Wald test = 0.32; p-value = 0.575) in the effects of language skills at 3 years on IH symptoms at 5.5 years (males: β = -0.09, SE = 0.06, p-value = 0.115; females: β = -0.14, SE = 0.06, p-value = 0.020). Yet, males had significantly higher SDQ IH symptoms scores at 3 and 5.5 years (βmales - βfemales = 0.21, p-value < 0.001; and 0.29, p-value < 0.001; respectively) and significantly lower language skills at 3 years but not at 5.5 years (βmales - βfemales = -0.33, p-value < 0.001; and -0.10, p-value = 0.135; respectively).
Discussion
Prior studies indicate high levels of comorbidity between ADHD and language impairment [
7,
16,
17], highlighting the importance of longitudinal research in testing different hypotheses on the nature of these associations. Based on a large (
N = 1459), prospective, mother-child cohort, our study confirms the asymmetrical relationship between language skills and IH symptoms during the preschool period. Early language skills not only predict later language skills but also later IH symptoms. Our results are consistent with prior findings [
13,
14].
Regarding the specific nature of the influence of early language on later IH symptoms, we found that the comprehension of instructions tests at 3 years was most strongly related to IH symptoms at 5.5 years (Models 4: β = -0.13 and C5 : β = -0.11,
p-value = 0.030). Moreover, a nonsignificant trend was observed between the sentence repetition test at 3 years and IH symptoms at 5.5 years (Model C3: β = -0.09,
p-value = 0.109 and Model 4: β = -0.11). Interestingly, the other three language tests at 3 years, which only involved single words (semantic fluency, word and nonword repetition, and picture naming), were not associated with IH symptoms at 5.5 years. Thus, these results suggest that early syntactic ability is the language domain most strongly associated to the development of IH symptoms. At first sight, these results seem to differ from those reported by [
18], which mentioned expressive language deficits as the main precursor of inattention symptoms, in children with a history of speech-language impairment. Yet, results reported in their Figure 1 suggest that receptive language was also affected in these children. The comparison between the two studies is hindered by the fact that Snowling et al. [
18] have chosen to group language variables into receptive and expressive components, rather than into word-level
versus sentence-level, or than reporting the results of each test. Thus our results may be more similar than their reporting suggests. Overall, our finding that sentence- rather than word-level language skills predict IH is consistent with the view that language-based self-regulation mediates this relationship (
Hypothesis 1), if one takes the plausible view that language-based regulation requires formulating propositional phrases, as opposed to just single words.
Among the other hypotheses that have emerged to explain the directional relationship between early language skills and later ADHD symptoms, our results do not support the interpersonal difficulties hypothesis (
Hypothesis 2) as the main explanation. Indeed, the link between language skills at 3 years and IH symptoms at 5.5 years was not found to be mediated by interpersonal difficulties. This result differs from the study conducted by Menting et al. [
25], but is not necessarily contradictory, since a) their study was conducted between ages 6 and 10 whereas our study was conducted during the preschool period and b) their study specifically examined mediation by peer rejection whereas our study examined mediation by broader aspects of interpersonal difficulties (including peer rejection).
Of course, other mediators could be considered, such as working memory which is known to be poorer in ADHD [
58‐
60]. Unfortunately, this study did not include any measure of executive function. The language tests that were used, although they sometimes involved short-term memory (Word and nonword repetition (ELOLA), Sentence repetition (NEPSY) and to a lesser extent Comprehension of instructions (NEPSY)), were relatively light on working memory. Therefore, it is not possible in our study to further investigate the mediating role of working memory, and future studies would benefit in examining it.
Our results do not either support the idea that asymmetrical relationships might reflect the effect of environmental factors becoming manifest in different domains at different ages (Hypothesis 3). We found a small decrease (14 %) in the estimates of the effect of language skills on later IH symptoms when comparing unadjusted models to models that were adjusted for a broad range of pre- and postnatal factors. Thus, the asymmetrical relationship is essentially unaffected by the effects of a broad range of environmental factors.
Overall, our results suggest that language difficulties in the syntactic domain precede the development of IH symptoms during the preschool period. One explanation would be that language difficulties represent an early marker of ADHD, i.e., an early expression of the disorder [
14,
15]. Another explanation would be that early language difficulties in the syntactic domain may play a causal role in the development of IH symptoms. Indeed, in line with prior findings [
10], children with ADHD may experience unexpected difficulties comprehending more complex information than children without ADHD. They may also have difficulties formulating sophisticated self-directed instructions to regulate their own behavior.
Strengths and limitations
Strengths of our study are the longitudinal design, the large sample size and the usage of validated language tests and questionnaires.
One possible limitation of our analysis is that IH symptoms were assessed using behavior rating scales completed by parents (SDQ), and could reflect reporting bias. More than one source of information and particularly preschool teacher’s ratings of IH symptoms would have also been useful, as the child’s ability to attend and concentrate and remain at his/her desk or place in the circle is usually more fully tested in the preschool setting [
61]. In addition it is also possible that the parent may be rating the child’s difficulty in following instructions and verbally stated demands – that is problems understanding or retaining what the parent is asking or demanding - as a symptom of inattention and hyperactivity. Further studies will have to confirm our findings by measuring IH symptoms through teachers’ or other non-parental raters. Second, some SDQ scores (SDQ emotional symptoms and peer relationship problems at 3 and 5.5 years; SDQ peer relationship problems at 3 years) and language tests (semantic fluency and picture naming at 3 years) had relatively low internal consistency (<0.70), as is often the case with scales assessing complex constructs based on a limited number of items (
e.g., 5 for the SDQ scores). Third, our study was not suited to determine whether verbal self-regulation mediates the effects of language skills on later IH symptoms because no direct measurement of self-regulation skills was available in our study. Fourth, developmental trajectories of children’s language and IH symptoms are complex and intertwined. Further studies focusing on individual trajectories of language and behavioral development are warranted. Finally, the rate of maternal depression in this sample was relatively high. However it must be underlined that the cut-off of the CES-D (i.e., 16) used to define depression at 3 and 5 years was chosen to increase sensitivity to detect mothers at high risk of having clinical depression (including subthreshold forms) [
62], which may have an impact on maternal reporting of child behavior problems [
63].
Acknowledgements
- We are indebted to the participating families, the midwife research assistants (L Douhaud, S Bedel, B Lortholary, S Gabriel, M Rogeon, and M Malinbaum) for data collection, the psychologists (Marie-Claire Cona and Marielle Paquinet) and P Lavoine, J Sahuquillo and G Debotte for checking, coding, and data entry.
- Members of the EDEN mother-child cohort study group are as follows: I. Annesi-Maesano, JY. Bernard, J. Botton, M.A. Charles, P. Dargent-Molina, B. de Lauzon-Guillain, P. Ducimetière, M. de Agostini, B. Foliguet, A. Forhan, X. Fritel, A. Germa, V. Goua, R. Hankard, B. Heude, M. Kaminski, B. Larroque (decease) (means that the member of the EDEN study is deceased), N. Lelong, J. Lepeule, G. Magnin, L. Marchand, C. Nabet, F. Pierre, R. Slama, M.J. Saurel-Cubizolles, M. Schweitzer, O. Thiebaugeorges.