INTRODUCTION

Pupils with developmental language disorder (DLD) present significant deficits in language acquisition that affect both production and comprehension (Bishop et al., 2017). These language deficits affect social and emotional development and can put them at risk of school failure. In general, when these pupils’ different linguistic problems are studied, the focus tends to be more on the expressive dimension of language and less on the receptive aspects. All this translates into a greater body of scientific research on expressive deficits related, for example, to verbal inflection markers in languages such as English or to the very limited structural length of sentences and frequent grammatical errors (Coloma et al., 2020) found in Spanish-speaking pupils.

A similar situation occurs with the implementation of intervention programmes in pupils with DLD, that is, there is a clear predominance of programmes aimed at improving aspects related to language production, with objectives linked to the improvement of receptive skills present to a lesser extent (Law et al., 2004). At the same time, there is a lack of intervention programmes promoting the development of executive functions in preschoolers. In the present study, we will emphasize the need to pay greater attention to the optimization of the linguistic and executive functions underlying oral language comprehension.

It is not easy to clarify the concept of language comprehension. Bishop's (1997) proposal provides a general framework for its analysis. In this model, comprehension is understood as a process during which information is transformed into different types of representations—phonological, lexical, morphological and syntactic—until the meaning is gleaned from one's existing general knowledge. It could be added that this is a system with two levels that includes both referential and inferential comprehension skills. Referential comprehension involves understanding the meaning of words and sentences. Word comprehension involves linking a sequence of sounds with the lexical entries of the mental dictionary previously represented in a subject's memory. Sentence comprehension, for its part, involves (syntactic and semantic) combinatorial processes of words to form more complex units. In addition, the representation of words involves the use of sublexical units (phonemes, syllables and morphemes), while sentence comprehension usually involves lexical and supralexical units (phrases) as well as some sublexical units such as grammatical morphemes. A critical step occurs when one moves from comprehending simple sentences to understanding discourse, that is, to the level of inferential comprehension. At this point, one must be able to form links between the different statements in order to build a mental model, that is, it is necessary to integrate the information into a coherent conceptual structure that unites the different statements, incorporating them into one's general knowledge. Here, the subject must establish connections between what he hears, the linguistic context and his knowledge of the world (Desmarais et al., 2013). In addition, this whole process is facilitated by the efficient deployment of two types of skills. In the first place, there are inferential skills that play a crucial role in understanding discourse in both the local domain of sentence processing and the more global or situational domain in which the discourse is situated. Both levels must be interrelated to achieve and maintain discourse coherence and to enrich the representation of meaning. Second, executive functions can be understood as the set of mental abilities that allow goal-setting, planning the steps necessary to achieve goals and self-monitoring to maximize performance. They represent the superordinate level of the cognitive hierarchy and include tasks such as working memory (the capacity to maintain information in mind during the implementation of another cognitive task), planning (the capacity of formulation, evaluation and selection of the required actions to reach the objective), semantic fluency (the capacity to name words belonging to the same semantic category), inhibition (the capacity to stop behaviour or to suppress irrelevant information for the performance of a task) and cognitive flexibility (the capacity to modify an answer according to the stimulus offered). In this context, one must consider how the influence of executive functions on oral language comprehension skills contributes to the building of inferences, construction of mental models (i.e., abstraction), maintenance of previous structures (i.e., working memory), integration of ideas (i.e., cognitive flexibility) and removal of irrelevant information (i.e., inhibition). Working memory, in particular, is essential for oral language comprehension, because noun and verb phrases tend to be situated apart from each other and need to be maintained and attached in the proper order to be comprehended. There is much research linking the adequate development of executive functions with language skills as early as preschool. Verbal working memory and semantic fluency acquire special relevance in this context. Both have been linked to vocabulary acquisition and are good predictors of pupils’ oral language comprehension skills. A notable body of research has pointed out significant executive deficits in pupils with DLD (Henry et al., 2012). Lastly, Chien (2020) recently found significant improvements in participants’ curriculum-based receptive vocabulary and oral comprehension after a comprehensive intervention programme that included training in inhibitory control and working memory.

At this point, it should be noted that pupils with DLD exhibit problems that affect the different levels of representation of oral language: phonological, lexical, grammatical and discursive (Coloma et al., 2020); verbal working memory and semantic fluency; and inferential abilities (Henry et al., 2012). This would explain the frequent appearance of oral language comprehension problems, which obviously constitute a clear risk factor for future difficulties in reading, especially when it comes to extracting the meaning of written texts (Desmarais et al., 2013). Consequently, it is essential to design intervention programmes that enrich all the skills involved in the early stimulation of language comprehension (Tarvainen et al., 2020). The need is more than justified, since there is little evidence of the effectiveness of therapy in improving receptive language skills (Law et al., 2004). Some studies have not found any significant results (Boyle et al., 2010), while others have discovered clear improvements in specific aspects such as, for example, receptive vocabulary, sentence structure, and the understanding of sentences (Tarvainen et al., 2021).

Based on the above, the present research has been designed to test the following hypothesis: Pupils diagnosed with DLD and those with TD will show greater gains in oral language comprehension, measured in terms of the tasks concepts and following directions, word classes—receptive, sentence structure, comprehension of paragraphs and narratives, working memory, and semantic fluency following the intervention programme than a control group of pupils with typical language development and a control group of pupils diagnosed with DLD. As a secondary objective, we expect that verbal working memory and semantic fluency have a positive relation with post-treatment gains in the variables used to assess oral language comprehension in this study.

METHODS Design

The study applied a pretest treatment–post-test design for a treatment group of pupils with DLD. To complete the design, a non-equivalent experimental group (of pupils with typical development—TD) was included, as well as two control groups (one equivalent and one non-equivalent). The dependent variables were three subtests of the receptive component of the CELF-4-Spanish (Semel et al., 2006: Concepts and Following Directions, Word Classes—Receptive and Sentence Structure); Comprehension of Paragraphs and Narratives; and Verbal Working Memory and Semantic Fluency. The chronological age and non-verbal IQ suitability of all the pupils to be included in this research were checked. The pretest assessment protocol was then dispensed to the selected pupils. Next, the intervention programme was implemented. Finally, the post-test was carried out. Both the tests and the intervention were administered in the children's schools. Prior authorization was obtained from schools and families using the corresponding informed consent form. Compliance with ethical standards was also positively evaluated by the university ethics committee.

Participants

A total of 99 children enrolled in schools on the island of Tenerife (Canary Islands, Spain) participated in the study. All participants were randomly assigned; they were divided into four groups: (1) a group of pupils with DLD receiving treatment (DLD-T); (2) a group of pupils with DLD not receiving treatment (DLD-C); (3) a group of pupils with typical language development receiving treatment (TD-T); and (4) a group of pupils with typical language development not receiving treatment (TD-C). Table 1 shows the descriptive statistics by group for age and non-verbal IQ.

TABLE 1. Descriptive statistics on age and non-verbal IQ by group Group Male:female Age (years) Non-verbal IQ Range Mean (SD) Range Mean (SD) DLD-C 14:11 5.2–6.3 5.6 (0.3) 80–106 96 (7) TD-C 14:11 5.2–6.3 5.7 (0.3) 89–113 111 (6) DLD-T 15:10 5.3–6.2 5.7 (0.3) 80–106 98 (8) TD-T 15:9 5.2–6.3 5.8 (0.3) 80–120 107 (8) Note: DLD-C, developmental language disorders control group; TD-C, typical development control group; DLD-T, developmental language disorders treatment group; and TD-T, typical development treatment group.

The normality of age was verified using the Kolmogorov–Smirnov test (z = 0.08; d.f. = 99; p = 0.174). An analysis was carried out to ensure that the groups did not differ on these variables. As a preliminary step, the homogeneity of the variances was determined using Levene's test (F(3,95) = 0.6; p = 0.591). The analysis of variance (ANOVA) did not show significant differences (F(3,95) = 3.0; p = 0.520; η2 = 0.01).

The matrix subtest of the K-BIT Intelligence Test was used to assess non-verbal IQ in the sample (Kaufman & Kaufman, 2000). The normality of non-verbal IQ was verified using the Kolmogorov–Smirnov test (z = 0.10; d.f. = 99; p = 0.098). To verify that the groups were matched in this variable, a hypothesis test was performed. As a preliminary step, the homogeneity of the variances was determined using Levene's test (F(3,95) = 1.9; p = 0.139). the ANOVA did not show significant differences (F(3,95) = 5.1; p = 0.097; η2 = 0.04).

Two of the groups (DLD-T and DLD-C) were selected by convenience sampling, since the pupils had to meet specific selection criteria. To select the pupils for the DLD groups, an initial evaluation was carried out in all the schools on Tenerife, in collaboration with the school management and psycho-pedagogical guidance teams. These professionals were asked to refer all pupils showing possible signs of DLD, that is, comprehension or expression problems in one or more components of language, but especially in morphosyntax and semantics, or pupils showing unresolved language difficulties since starting school. A total of 147 pupils were referred, and a language evaluation protocol was administered to them to confirm the diagnosis. This consisted of a language test, specifically the Clinical Evaluation of Language Fundamentals CELF-4-Spanish (Semel et al., 2006). For participant diagnosis, we used an indicator derived from the three main test indices (Core Language, Expressive Language and Receptive Language), with a composite score < 77.5 (–1.5 SD) obtained for at least one. This administration of the evaluation protocol led to the selection of a sample of 50 pupils diagnosed with DLD. A total of 65 participants were excluded from the study for presenting simple language delay, that is, a slight chronological delay in development characterized more by phonological difficulties than by structural deficits, and 32 participants were excluded for not completing the tests, due to repeated absences or lack of cooperation.

Pupils in the TD groups were selected by means of discretional sampling to ensure that the four groups were as similar as possible in other variables that might influence the results. A total of 50 pupils with TD were selected from the classmates of the pupils with DLD. To respect the proportion of boys and girls in all groups, each sex was randomized separately, that is, the girls were assigned to the groups randomly and the same was done with the boys. The participants in this group had no language difficulties and were being educated within the usual parameters. One pupil was excluded for not completing the tests, due to repeated absences.

Therefore, the final sample consisted of 99 pupils with different social backgrounds, enrolled in both public and private schools and residing in both rural and urban areas.

Assessment materials

A group of speech and language therapists was created to run the initial and final assessments. The distribution was counterbalanced to ensure that they did not assess the same participants at both assessments. Furthermore, the evaluators did not know in which group the children belonged.

Concepts and Following Directions, Word Classes—Receptive and Sentence Structure subtests

The subtests of the CELF-4-Spanish (Semel et al., 2006) that assess oral comprehension were selected: Concepts and Following Directions, Word Classes—Receptive and Sentence Structure. The Concepts and Following Directions subtest assesses the ability (1) to interpret spoken instructions of increasing length and complexity that contain concepts requiring logical operations; (2) to recall the names, characteristics and order of a series of named objects; and (3) to identify from among the illustrations provided the objects that have been named. The Word Classes—Receptive subtest assesses the ability to comprehend the shared and unshared meanings of associated words. Finally, the Sentence Structure subtest is used to assess the pupil's capacity to interpret sentences of increasing length and complexity and select the images that illustrate their referential meaning.

The psychometric properties of the three subtests were adequate in terms of both the internal consistency as measured by Cronbach's alpha (α = 0.888) and for each subtest (Concepts and Following Directions: α = 0.836; Word Classes—Receptive: α = 0.799; and Sentence Structure: α = 0.854). Since this parameter is biased by the number of elements included in the calculation, the composite reliability (McDonald's ω) and the average variance extracted (AVE) were also calculated, obtaining excellent reliability overall (ω = 0.965 and AVE = 0.721) and good reliability for the subtests (Concepts and Following Directions: ω = 0.769 and AVE = 0.625; Word Classes—Receptive: ω = 0.775 and AVE = 0.653; and Sentence Structure: ω = 0.768 and AVE = 0.623). In both cases, correct answers were scored with 1 point.

Comprehension of paragraphs and narratives

Two tasks were completed, with correct responses scoring 1 point in both cases. The first task consists of reading a short paragraph out loudly and clearly to pupils once, at a steady, leisurely pace, one word per second:

Saturday was a fun day. Juana and her family went to Juana's soccer game. Juana was the team goalkeeper. The game was tied. At the last moment a player from the other team tried to score a goal, but Juana stopped her. Her friends congratulated her at the end of the game.

Then five questions are asked:

What was Saturday like?

Where did Juana and her family go?

What happened at the end of the game?

How did the game end?

Why did Juana's friends congratulate her?

The second task involves telling a story, ‘Tito, the greedy dog’, with the help of nine picture cards (for the story, see the Appendix). The following literal and inferential questions were then asked:

What animal is the story about?

Where does the story take place?

Why did the food disappear?

Who is spoken of in the story when it is said ‘she fed him and took him out for a walk’?

What did Tito do when the food disappeared?

Why couldn't he get the food that was inside the burrow?

Why was Tito still hungry?

What did Tito do to enter the burrow?

Why did Tito share his food?

Why do you think that in the end Tito always left a little food in the garden?

What do you think happened after Tito and the mouse shared the food? What did Tito and the mouse learn?

Working memory

The Backward Digit Span subtest of the WISC-IV (Wechsler, 2003) was used. In this test, the evaluator says an increasingly long series of digits that the participants have to repeat in reverse order. Two elements are presented for each level, and at least one must be produced correctly to proceed to the next level. The dependent variable was the longest span or list of elements correctly reproduced by the participant.

Semantic fluency

The Controlled Oral Word Association Test (COWAT) (Benton & Hamster, 1983) was used. This enabled the evaluation of semantic fluency (SF). The participants were asked to name as many animals as they knew. The time limit for each category was 60 s.

Procedure

The pretest was administered prior to the intervention programme, at the end of the 2017–18 school year. The programme was then implemented during the 2018–19 school year (from the first week of November to the first week of April), with the post-test administered in the second half of April.

An ANOVA was performed for each dependent variable using the pretest measures, which allowed us to evaluate the initial differences between the groups and thus establish the baseline. As shown in Table 2, for all variables significant differences were found, with a large effect size.

TABLE 2. Analyses of variance (ANOVAs): Main effect of pretest and orthogonal contrasts Subtest F(3,95) η2 DLD-C versus TD-C DLD-C versus DLD-T DLD-C versus TD-T TD-C versus DLD-T TD-C versus TD-T DLD-T versus TD-T F(1;97) η2 F(1;97) η2 F(1;97) η2 F(1;97) η2 F(1;97) η2 F(1;97) η2 PN 15.6*** a 0.33 21.9*** 0.19 2.15 0.02 36.4*** 0.28 10.3** 0.10 2.0 0.02 21.0*** 0.18 CFD 76.1*** a 0.71 121.6*** 0.57 1.2 0.01 124.5*** 0.57 103.0*** 0.52 0.0 0.00 101.3*** 0.52 WC-R 50.2*** a 0.62 53.3*** 0.36 2.1 0.02 143.1*** 0.60 7.5** 0.07 2.9 0.02 56.1*** 0.37 SS 84.7*** a 0.73 109.0*** 0.53 1.1 0.01 211.2*** 0.69 43.0*** 0.31 1.6* 0.01 114.3*** 0.55 VWM 23.1*** 0.42 29.7*** 0.24 0.2 0.00 34.6*** 0.27 34.4*** 0.27 0.2 0.00 39.6*** 0.29 SF 11.1*** 0.26 18.0*** 0.16 0.0 0.00 14.0*** 0.13 19.2*** 0.19 0.2 0.00 15.2*** 0.14 Note: aWelch's F. DLD-C, developmental language disorders control group; TD-C, typical development control group; DLD-T, developmental language disorders treatment group; TD-T, typical development treatment group; PN, Comprehension of Paragraphs and Narratives; CFD, Concepts and Following Directions; WC-R, Word Classes—Receptive; SS, Sentence Structure; VWM, Verbal Working Memory; and SF, Semantic Fluency. **p ≤ 0.01; and ***p ≤ 0.001.

In all variables, the results for the two groups of pupils with DLD (DLD-T and DLD-C) were significantly lower than those of the two groups of pupils with typical development (TD-T and TD-C), with a large effect size, while the equivalent groups, DLD-T versus DLD-C on the one hand and TD-T versus TD-C on the other, did not show any differences between them.

Intervention procedure

For the description of the intervention programme, we followed the recommendations made by Hoffmann et al. (2014). The objective of the training was to improve language comprehension by working on a series of essential components at the word, sentence and discourse level, together with verbal working memory and semantic fluency. It should be noted that a crucial component of the programme has been the inclusion of discourse and the emphasis on inferential comprehension, in line with previous proposals such as those of the Nuffield Early Language Intervention (NELI) programme (Bowyer-Crane et al., 2008; Hulme et al., 2020; West et al., 2021) and especially that formulated by Dawes et al. (2019). Indeed, retelling and story generation activities that favoured narrative cohesion and coherence, supported by open-ended questions and visual support (pictures, drawings, icons, and graphic organizers) and pictographic planning (Spencer & Petersen, 2020) were implemented. When the students heard the story, a series of activities aimed at promoting comprehension were carried out before the retelling of the story. For example, low-frequency vocabulary was reviewed, semantic and concept maps were made with meaningful words, what they heard in the story was related to their prior knowledge, the story was summarized highlighting the main ideas, literal and inferential questions were asked, and central aspects of the story were discussed. In addition, inferential comprehension was stimulated by implementing a series of activities supported by visual material, graphic resources and multiple-choice answers (Monfort and Monfort, 2016). The programme was implemented by a total of 45 preschool teachers and 30 speech and language therapists. They were given 20 h of training. In addition to being provided with a detailed dossier with all the materials, they received training in a practical workshop format. They were given two notebooks. The first was called the teacher's notebook, and it included a short introduction to the main objectives of the intervention for 5-year-old children. Examples were also provided of the correct use of the intervention techniques, as shown in Table 3. The second notebook, called the student's notebook, contained most of the material required for carrying out the different activities. During the intervention, the teachers received weekly visits from members of our research group; during these visits, possible doubts were resolved and explicit support was provided in the classroom context. In addition, over the course of the programme, four additional plenary meetings were held to verify the reliability of the programme and evaluate the process. Later, the sessions were recorded. Subsequently, two independent evaluators verified, through a checklist, that the sessions were in line with the established schedule, with a very good agreement (κ = 0.927).

TABLE 3. Materials and procedures Activities, techniques and materials Lexical–semantic Coloured cards showing objects and simple and complex actions. Comprender el lenguaje haciendo ejercicios (Aguado et al., 2003) Puppets, treasure chest and journal for the Lexicon Pirate game (Motsch & Ulrich, 2012) Activities involved naming, categorization, classification, semantic families, semantic maps, synonyms and antonyms, definitions, recall, lexical access, and functional lexicon. Teachers and therapists provided support through intervention techniques such as phonological awareness: ma-ri-po-sa; s-o-l; (ca)-fe (but-ter-fly; s-u-n; (co)-ffee); contextual expansion of words: production of words in complex linguistic contexts (sentences, paragraphs, etc.); bootstrapping, or the provision of phonetic and semantic clues. ‘La comen los conejos …’, ‘La zana …’ (‘Rabbits eat them’, ‘Ca…’); visual organizers: semantic maps and diagrams; open questions Morphology Story: Why does the polar bear have a short tail? Cards with drawings. Blackboard, cards and magnifying glasses for the Detective game (Zoski & Erickson, 2016) Activities involved inflectional, derivative and compositional morphology. Teachers and therapists employed two intervention techniques for morphology. First, modelling was used, that is, presenting a correct model and waiting for it to be repeated at some point, spontaneously, by the pupil. For example, the adult would say: ‘He is very happy (muy feliz). He is extremely happy (felicísimo). He is very quiet (muy tranquilo). He is … (tranquilísimo)’. If the pupil did not respond, the teacher added ‘tranquilísimo’. The second technique was bootstrapping or external facilitation. For example, if the pupil said: ‘I don't know what that is called’, the adult replied with a semantic bootstrapping, ‘the child is very relaxed, very calm (muy relajado, muy tranquilo)’. And if he did not answer, he was offered a phonological bootstrapping, ‘the child is tranquilí …’, so that the pupil would answer: ‘tranquilísimo’ Syntax Following orders with Simon Says. Comprender el lenguaje haciendo ejercicios (Aguado et al., 2003) Activities involved building sentences, completing and ordering sentences, detecting errors in sentences, identifying whether sentences have the same meaning, and following instructions. Intervention techniques used included recasting, that is, where the professional responded to the child's immature or incorrect expressions with a recast containing a restructured grammatical form. For example, if the child said: ‘Fat elephant eats flower’, the professional replied: ‘A very fat elephant eats a flower’; concatenation, in which the professional would ask questions to complete parts of the statement, then produce the complete statement for the child. For example, the professional would say: ‘Tell me, who do you see in this picture?’ and the pupil would reply ‘An elephant’, to which the professional would respond: ‘And what is it doing?’. The answer might be ‘Eating a flower’. The professional would say: ‘Let's say it all together, and the pupil would say: ‘A very fat elephant is eating a flower’; imitation, where the correct model is presented to the pupil for repetition Narrative discourse Vignettes with short stories. Symbols Story The three hungry mice Activities involved retelling and generation of personal and invented stories. Intervention techniques included recast, imitation and open questions. Inferences Images based on Más allá de las palabras (Monfort et al., 2016) Activities involved different types of predictive inference (e.g., showing an image of a birthday cake with candles and asking, ‘What do you think will happen here?’); logical inference (e.g., showing images of footprints and saying, ‘If I see these tracks, … who has walked by?’); logical–cultural inference (e.g., showing an image and saying, ‘We went to Teide to see the snow, we can infer that it is …’); linguistic inference (e.g., saying, ‘Listen to this sentence, think, and then answer the question: While swimming, Felipe found a mussel on the rocks. Where is Felipe?’); and pragmatic inference (showing an image of a dog with an empty food bowl and asking, ‘What is the dog waiting for?’). Intervention techniques consisted of asking different types of questions based on inference class Verbal working memory Activities involved completing two sentences with a word and recalling the words used, e.g., saying, ‘In autumn, the leaves fall from the … (trees)’; ‘When we get sick, we go to the … (doctor)’; repetition of series of letters and numbers, in original and reverse order (3-4-1-7; z-e-g-s); presenting combinations of letters and numbers and recalling first the numbers and then the letters (A-2-L-5-J) Semantic fluency Activities included saying words belonging to different categories (animals, food, colours) as quickly as possible; saying everything that comes to mind about a word list; saying a sentence that includes an element presented in an image (frog, lion, watermelon)

All sessions were run in person at the schools. A total of 95 sessions of 60 min each were held, during the 2018–19 school year, following the same sequence and with identical materials. In our case, we used a multitiered systems of support (MTSS) model, with the pupils in the two experimental groups, DLD-T and TD-T, working through all three levels of practice. The pupils in the control groups belonged to different classrooms or schools than the pupils in the experimental groups. The first 4 days of the week, each teacher worked in the context of the regular classroom, combining situations with all the pupils (level 1) and in small groups of three pupils each (level 2), while on Fridays the pupils with DLD-T and TD-T participating in the study went, together with a third classmate with TD (a different one each time), to the speech and language therapy room to work with the speech and language therapist (level 3) to internalize the skills being practiced with the teacher and complete tasks stimulating inferences, semantic fluency, and verbal working memory.

The teaching process made use of games and employed plenty of visual materials to encourage participation. The programme content and activity format can be found in Table 3. Prior authorization was requested from educational centres and families.

Data analysis

An ANOVA was performed for each dependent variable with the pretest–post-test difference to determine if there were differential gains after the intervention. As a preliminary step for all the ANOVAs performed, the homogeneity of variances was determined using Levene's test. In the contrasts that presented heterogeneity, the robust Welch's test was used. Orthogonal contrasts were performed as post-hoc comparisons in those evaluations that showed significant differences, to identify which groups showed these differences. Generalized η2 was used as an indicator of the effect size, for both the main effects and for the simple effects of the ANOVAs. An η2 around 0.01 is generally considered a small effect, an η2 around 0.06 indicates a medium effect, and an η2 > 0.14 is a large effect. Finally, a correlation analysis was carried out between post-treatment gains in the executive functions (working memory and semantic fluency) and gains in comprehension variables (oral language skills). All analyses were performed with SPSS v25.

RESULTS

Figure 1 shows the means and standard deviations (error bars) of the gains due to the intervention of the four groups in each task.