Research in Developmental Disabilities 35 (2014) 3534–3542
Contents lists available at ScienceDirect
Research in Developmental Disabilities
Differential verbal working memory effects on linguistic production in children with Specific Language Impairment Andrea Marini a,b,*, Cinzia Gentili c, Massimo Molteni c, Franco Fabbro a,b a
Department of Human Sciences, University of Udine, Udine, Italy IRCCS ‘‘E. Medea: La Nostra Famiglia’’, San Vito al Tagliamento, Pn, Italy c IRCCS ‘‘E. Medea: La Nostra Famiglia’’, Bosisio Parini, Lc, Italy b
A R T I C L E I N F O
A B S T R A C T
Article history: Received 26 June 2014 Accepted 28 August 2014 Available online 18 September 2014
Deficits in verbal working memory (vWM) have often been reported in children with Specific Language Impairments (SLIs) and might significantly contribute to their linguistic difficulties. The linguistic and narrative skills of a group of children with diagnosis of SLI were compared to those of a group of children with typical development. The linguistic assessment included a comprehensive analysis of their lexical, grammatical and narrative abilities. Overall, the participants with SLI had difficulties at all three levels of linguistic processing. The effect of vWM was marginal on lexical processing, significant on grammatical structuring, and null on narrative construction. ß 2014 Elsevier Ltd. All rights reserved.
Keywords: Specific Language Impairment Language Verbal working memory Non word repetition
1. Introduction The label Specific Language Impairment (SLI) applies to a cluster of relatively frequent developmental disorders characterized by language delay in children with otherwise normal physical and intellectual development (Leonard, 1998; Newbury, Fisher, & Monaco, 2010). SLI is not a homogenous condition (Laws & Bishop, 2003). As language is a complex cognitive function resting on the interaction between different levels of processing, its development may be affected in different ways. A comprehensive linguistic assessment should include all of these levels of processing and address the delicate issue of their reciprocal interconnections. Generally speaking, language can be analyzed from two major points of view: 1) a within-sentence or microlinguistic level, which focuses on lexical (i.e., phonetic, phonological, morphological and semantic) and grammatical (i.e., morphosyntactic and syntactic) skills; 2) a between-sentence or macrolinguistic level, which focuses on pragmatic (i.e., the ability to generate inferences so to contextualize the literal meanings of words and sentences, and understand the communicative intentions of the interlocutors) and discourse (i.e., the ability to establish linguistically cohesive and conceptually coherent ties) skills. Increasing evidence shows that these levels (and their interrelations) can be simultaneously assessed by adopting a multi-level approach to narrative language analysis (e.g., Marini, Andreetta, del Tin, & Carlomagno, 2011) and that this ensures a comprehensive understanding of the linguistic symptoms observed in both adult patients and children with developmental and/or acquired disorders (e.g., Marini, Tavano, & Fabbro, 2008; see also Botting, 2002 for further considerations on the usefulness of narrative as a tool for the assessment of both linguistic and pragmatic impairments). For example, Marini, Tavano and Fabbro (2008) showed that the use of a narrative production
* Corresponding author at: Department of Human Sciences, University of Udine, Italy. Tel.: +39 3355393224. E-mail address: [email protected] (A. Marini). http://dx.doi.org/10.1016/j.ridd.2014.08.031 0891-4222/ß 2014 Elsevier Ltd. All rights reserved.
A. Marini et al. / Research in Developmental Disabilities 35 (2014) 3534–3542
3535
task might highlight the presence of lexical and grammatical disturbances in children with SLI who had scored within normal range on traditional linguistic assessment tests assessing the same abilities. Accumulating evidence suggests that children with SLI might experience also additional disturbances that are not specifically linguistic. For example, Vugs, Hendriks, Cuperus, and Verhoeven (2014) have recently showed that a large cohort of children with SLI aged 4–5 years scored worse than a group of age-matched peers with typical development on tests assessing executive functions (i.e., inhibition, shifting, emotional control, and planning/organization) and both verbal and visuospatial working memory. These results are not an isolate finding. Indeed, difficulties in the ability to select appropriate responses, inhibit irrelevant ones, plan strategies (e.g., Finneran, Francis, & Leonard, 2009; Henry, Messer, & Nash, 2011) as well as impairments in verbal working memory (vWM) have been frequently observed in children with SLI (Ellis Weismer, Evans, & Hesketh, 1999; Montgomery, 2006) and might affect their linguistic development and functioning. The concept of working memory refers to a set cognitive processes that allow to temporarily store and manipulate limited information (e.g., Cowan, Nugent, Elliott, Ponomarev, & Saults, 2005). According to the phonological storage deficit hypothesis, a difficulty in storing the incoming phonological information might contribute to their linguistic impairments (Archibald & Gathercole, 2007; Bishop, 2006). A reduced vWM capacity may hamper the generation of adequate phonological representations in the mental lexicon, thus leading to difficulties in their recognition when acting as listeners and in their production when involved in language production. These disturbances might negatively affect the child’s linguistic development: a difficulty in keeping track of phonological and/or lexical items in short-term memory and eventually process them in working memory might result in slowed vocabulary acquisition (e.g., Gathercole & Baddeley, 1990) and trigger a range of linguistic difficulties (Graf Estes, Evans, & Else-Quest, 2007). Interestingly, the few children with SLI with normal levels of vWM have also better performance on tests assessing linguistic skills and produce longer utteraces (e.g., Botting & Conti-Ramsden, 2001). If present, limitations in vWM might affect the children’s efficiency in daily communicative interactions and may lead to the production of shorter utterances. For example, Duinmeijer, de Jong, and Scheper (2012) showed that the performance of a group of children with SLI on the digit span correlated positively with the mean lenght of their utterances (MLU). Furthermore, a deficit in verbal working memory may trigger the selection of inappropriate words that, in turn, might determine qualitatively inferior levels of organization at the text level (macrolinguistic organization). Unfortunately, the interconnections among these apparently different skills (i.e., working memory, lexical selection, and macrolinguistic organization) cannot be detected by traditional tests assessing production skills in isolation (e.g., naming tests, fluency tasks, etc) but might emerge clearly when applying a multi-level approach to the analysis of language samples obtained on narrative production tasks. For these reasons, the present study focuses on two major issues: (1) the need for a comprehensive assessment of linguistic functions in children with SLI and (2) the opportunity to consider the role potentially played by verbal working memory in linguistic processing in these children with all the implications for appropriate linguistic assessment and remediation. Namely, we analyzed in detail the linguistic and narrative skills of a group of schoolage children with SLI and compared their performance with that of a cohort of children with typical development. Importantly, as the levels of verbal working memory might have affected the linguistic performance of the children included in the study, we hypothesized that, when controlling for working memory in the statistical analyses, those group-related differences on language tasks that were heavily biased by vWM capacity would disappear. We also hypothesized that such a comprehensive analysis would allow us to provide a clearer picture of the linguistic disorders that characterize these children and of their potential interconnections at different levels of processing. 2. Methods 2.1. Participants Sixty-four Italian-speaking participants were included in the study. They formed an experimental and a control group (see Table 1). The experimental group was formed by 32 children with diagnosis of SLI with mixed expressive-receptive disorder (ICD-10 diagnosis: F80.2) aged between 7 and 11.11 years old (mean 8 years and 8 months; standard deviation, SD, 1.6). None of them had mental retardation, psychotic symptoms, brain lesions, auditory difficulties or comorbid dyslexia. The General (verbal and performance) intelligence of SLI participants was assessed by administering the Wechsler scales (Wechsler, 1993). All participants were required to have a Performance Quotient (PIQ) of 85 or above this threshold. The average performance of the SLI participants on Verbal and Performance Intelligence Quotient (IQ) was matched against
Table 1 Means (and standard deviations) showing demographic data of the two groups of participants and their performance on the non-word repetition task.
Age Education Sex Non-word repetition*
SLI (N = 32)
TLD (N = 32)
8.77 (1.58) – range: 7–11.11 3.2 (1.6) – range: 1st–6th grade M = 19 13.34 (2.21)
8.61 (1.41) – range: 7–11.11 3.5 (1.3) – range: 1st–6th grade M = 17 14.69 (1.00)
SLI: children with Specific Language Impairment; TLD: children with Typical Language Development. * Group-related difference on this task was significant (p < 05).
3536
A. Marini et al. / Research in Developmental Disabilities 35 (2014) 3534–3542
general population norms and results are expressed as Intelligence Quotient scores (mean 100, SD 15). Overall, the mean Fullscale Intelligence Quotient (FIQ) score (mean = 86) was within 1 SD below average, with a relative discrepancy between the Verbal Scale (mean = 79: more than one SD below normal range) and the Performance Scale (mean = 94: within normal range). The control group was formed by 32 participants with Typical Language Development (TLD) aged between 7 and 11.11 years old (mean 8 years and 6 months; SD 1.4). They were selected from a larger sample of 1300 children recruited for the standardization of the ‘‘Batteria per la Valutazione del Linguaggio in Bambini dai 4 ai 12 anni’’ (BVL_4-12, Marini, Marotta, Bulgheroni, & Fabbro, 2014), a comprehensive battery of tests designed to tap three modalities (i.e., repetition, comprehension, and production) along several levels of processing: from phonetics to discourse. They performed within normal range on the Raven’s progressive matrices (Raven, 1938), in the non-word repetition subtest of the Prove di Memoria e Apprendimento per l’Eta` Evolutiva (PROMEA; Vicari, 2007), and on the forward and backward digit spans’ subtests of the Wechsler Scales (Wechsler, 1993). Furthermore, they had average school performance in language and reading. In a preliminary interview, their teachers confirmed that they had normal cognitive and learning development. According to school records and parents’ reports none of them had a known history of psychiatric or neurological illness, learning disabilities, hearing or visual loss. The two groups had similar chronological age and education (age: t(62) = .437; p = 664; education: t(62) = .886; p = .379). However, an independent-samples t-test showed that the SLI participants produced fewer correct repetitions on the Non-Word Repetition subtest of the BVL-4-12 (t(62) = 3.136; p < 003; d = .766). Interestingly, performance on tests of non word repetition is considered a measure of vWM and a reliable marker of SLI (Conti-Ramsden & Botting, 2001) as it differentiates children with SLI already at the age of 2.6 years (Chiat & Roy, 2007). All participants were selected among middle class families in regions in North-East Italy (Lombardia, Veneto and Friuli Venezia Giulia). Children with diagnosed hearing impariments, neurological disorders, austistic spectrum disorder or ADHD were not included in the study. In both groups the level of education of the parents ranged from high school to Master’s degree. For the experimental group, data collection was performed during standard linguistic evaluation at the IRCSS ‘‘E. Medea’’, Ass.ne ‘‘La Nostra Famiglia’’ in San Vito al Tagliamento (Pn) and Bosisio Parini (Lc). The study received institutional ethics’ approval by the Ethics Committee of the Research Institute IRCCS ‘‘E. Medea’’. All parents released their informed consent to the participation of their children to the study and to the treatment of the data. 2.2. Procedures of linguistic assessment The linguistic assessment was delivered by trained speech-therapists or developmental psychologists in a quiet room at the Research Institute IRCCS ‘‘E. Medea’’ (for children with SLI) or in their schools (for children with TD). The linguistic skills of the participants were assessed by administering the BVL_4-12 (Marini et al., 2014). The administration of these tests occurred in two sessions of approximately 45 min each and the children’s performance was audio-recorded. In the first session children received a cohort of tests assessing linguistic and narrative production. In the second session they received tests assessing repetition and comprehension skills. For the specific purposes of this study we will report the performance of these children on tests assessing lexical (Section 2.2.1), grammatical (Section 2.2.2), and narrative (Section 2.2.3) production abilities. 2.2.1. Assessment of lexical production skills The participants’ lexical productive skills were evaluated by administering tasks assessing naming, semantic fluency and narrative discourse production. In the naming subtest of the BVL 4-12 children were asked to tell the name of a series of stimuli depicted in sheets. The stimuli have been selected so to represent words with different (1) frequencies in spoken Italian (words with low, medium and high frequency), (2) grammatical class (verbs and nouns), and (3) semantic features (action verbs and nouns pertaining to 16 semantic categories). Each correct answer was assigned 1 point. The maximum score for this test is 67. The subtest of semantic fluency required the children to produce as fast as they could a series of words according to a specific semantic criterion within 1 min. Namely, in this version of the test the participants were asked to produce as many words as they could that pertained to the semantic category ‘‘animals’’ and then to the semantic category ‘‘objects in the house’’. This test offers a good way to assess the extension of a child’s vocabulary. Additional measures assessing lexical productive skills were elicited by administering a cartoon story made of six drawings presented on the same page (the ‘‘Nest Story’’: Paradis, 1987). Each storytelling was tape-recorded and transcribed verbatim by two independent coders. Two independent raters analyzed these transcriptions calculating the number of words and the time of each storytelling. These two parameters (i.e., words produced and time elapsed during the storytelling) allowed to derive a measure of speech rate considered in terms of words per minute (Marini, Andreetta, et al., 2011). This measure provides additional information on the children’s lexical efficiency, that is their ability to select words while producing a narrative discourse. Furthermore, a Percentage of Semantic Paraphasias was calculated by dividing the number of semantic paraphasias produced by each child by the number of words. These values were multiplyed by 100. A semantic paraphasia was scored whenever a target word had been replaced by a semantically related one (e.g., fiore [Flower] instead of albero [Tree]). 2.2.2. Assessment of grammatical production skills The children’s grammatical skills were assessed in terms of mean lenght of utterance (MLU) in the Nest Story description task, morphosyntactic organization and syntactic processing. The former aspect was analyzed by deriving by each story
A. Marini et al. / Research in Developmental Disabilities 35 (2014) 3534–3542
3537
description the total number of utterances calculated following the criteria detailed in Marini, Andreetta, et al. (2011). The MLU was then calculated by dividing the total number of words by the number of utterances. Morphosyntactic organization was analyzed by administering a test of sentence completion and by deriving from the story tellings produced by the participants a % of paragrammatic errors to words. The sentence completion subtest of the BVL_4-12 assesses the child’s ability to process derivational and inflectional moprhology to produce grammatically well-formed sentences. Namely, each child listened to a sentence (the model; e.g., Marco apre la porta [Marco opens the door]) and then heard the beginning of a second sentence (the prompt), which (s)he was asked to complete by assigning the correct morphemes to the verb (the target). In our example, the prompt is Anche noi. . . (in English: We also. . .) and the child was expected to complete the target sequence as follows: Anche noi apriamo la porta (in English: We also open the door). The test is made of 14 pairs of model sentences and prompts with different levels of grammatical complexity. Each correct answer was assigned 1 point with a maximum score of 14. The other measure tapping the participants’ morphosyntactic skills, i.e., % of paragrammatic errors, was derived by the analysis of the narrative speech samples produced by the participants while telling the Nest Story. Paragrammatic errors include the substitution of free morphemes, which in Italian involve mainly the substitution of function words (for example, batte da una porta ‘‘he is knocking from a door’’ – da instead of a), and of bound morphemes (for example, Questo e` una coppia ‘‘this [masc] is a couple [fem]’’ – should be questa). The analysis of the participants’ syntactic processing skills was assessed by deriving a measure of grammatical completeness from the descriptions of the Nest Story and by administering a test of syntactic comprehension, respectively. The % of Complete Sentences was derived by dividing the number of utterances that had been scored as well-formed from a grammatical point of view produced during the description of the Nest Story by the total number of utterances produced during the story telling (Marini, Andreetta, et al., 2011). An utterance was considered grammatically complete if all the arguments required by the verb had been correctly inserted and no omissions or substitutions of free or bound morphemes could be found.
2.2.3. Assessment of narrative production skills The children’ narrative skills were assessed in terms of textual organization and levels of informativeness. As for textual organization at intersentential level, the macrolinguistic measures included a Percentage of Global Coherence Errors. Errors of global coherence included the production of utterances that may be tangential, conceptually incongruent with the story, propositional repetitions, or simple fillers. An utterance is considered tangential when it contains a derailment in the flow of discourse with respect to the information previously provided in preceding utterances as in the following sequence produced during the description of the Picnic picture:/It is a picnic/I like picnics/I have made several picnics in my life/. Here, the second and the third utterances are scored as tangential, as they provide irrelevant information triggered by a specific idea portrayed in the stimulus picture. An utterance is considered conceptually incongruent when it includes ideas not directly addressed by the stimulus. Let us consider the following sequence produced during the description of the Cookie Theft picture:/the children are trying to get the cookies/the TV is out/. In this example, the second utterance is scored as conceptually incongruent as in the picture stimulus there is no TV. A propositional repetition is a sequence where the speaker simply repeats ideas that have been already mentioned. Therefore, any new information is provided (e.g., the second utterance in the sequence:/the man is walking on the sidewalk/he is on the sidewalk/). Finally, a filler utterance is scored whenever the speaker produces an utterance that does not provide any additional information like in the following sequence:/the man and the woman are eating/my god, and now?/ah, yes, I get it/. The last two sequences are scored as filler utterances. The Percentage of Global Coherence Errors was calculated by dividing the number of global coherence errors by the number of utterances and multiplying this value by 100. The informative skills of the children were assessed in terms of percentage of thematic units identified in each story telling and percentage of informative words in the sample. A thematic unit is considered one of the main ideas or details in the story that have been previously identified. The number of thematic units produced by each child was considered an index of the participants’ abilities to derive information from the picture stimuli. Therefore, we could derive a Percentage of Thematic Selection by dividing the thematic units identified by each child in the Nest Story by the total number of possible thematic units present in the story and multiplying this value by 100. In order to derive the Percentage of Lexical Informativeness for each storytelling we identified the lexical information units (LIUs), i.e., those content and function words that were not only phonologically well-formed but also appropriate from a grammatical and pragmatic point of view (e.g., Marini, 2012). Therefore, all those words that had been classified as semantic or verbal paraphasias, fillers, paragrammatic errors or forming tangential or extraneous utterances (i.e., utterances that were somehow deviating from the gist of the story) were excluded from the LIUs’ count. The Percentage of Lexical Informativeness was then obtained by dividing the number of LIUs by the number of words and multiplying this value by 100. The scoring procedure was performed independently by two raters on the stories produced by the participants and then compared. High interjudge reliability for the story narrative analyses has been previously documented (Marini & Urgesi, 2012). An interrater reliability analysis using Kappa statistics was performed to determine consistency among raters. Acceptable inter-rater reliability was defined as k .70 (Carletta, 1996).
A. Marini et al. / Research in Developmental Disabilities 35 (2014) 3534–3542
3538
3. Results 3.1. Analysis of lexical performance The group-related differences on the measures assessing lexical production (i.e., Naming, Semantic Fluency, Speech Rate, Percentage of Semantic Paraphasias produced in the Nest Story description task) were analyzed with a series of independentsamples t-tests. The level of statistical significance was set at p < .0125 (.05/4 dependent variables) after Bonferroni correction for multiple comparisons. Effect sizes were calculated in terms of Cohen’s d (Cohen, 1988). This is computed as the difference between the mean of the control group and that of the group of participants with SLI divided by the pooled sample standard deviation. The role played by verbal working memory on linguistic performance was controlled by performing an additional series of One-Way ANCOVAs with the scores at the NWR subtest of the BVL_4-12 as covariate. Table 2 reports the results of these analyses showing the level of group-related significance before (t-tests) and after covariation (ANCOVAs). When verbal working memory was not included in the analysis as covariate, the independent samples t-tests showed that the group of children with SLI had impaired performance on all of these measures: Naming (t(59) = 2.920; p < .005; d = .75); Speech Rate (t(61) = 3.070; p < .003; d = .77); Semantic Fluency (t(62) = 2.667; p < .010; d = .67), Percentage of Semantic Paraphasias (t(44) = 4.290; p < 001; d = 1.08). Interestingly, when the analyses were rerun controlling for NWR, only the group-related differences in Semantic Fluency and Speech Rate were no longer significant (p = .163 and p = .036, respectively), while those pertaining Naming (p < .009) and the Percentage of Semantic Paraphasias (p < .001) survived. 3.2. Analysis of grammatical performance The group-related differences on the measures assessing grammatical skills (i.e., MLU, Sentence Completion, % Complete Sentences, % Paragrammatic Errors) were analyzed with a series of independent-samples t-tests. Also in this case the level of statistical significance was set at p < .0125 (.05/4 dependent variables) after Bonferroni correction for multiple comparisons. Effect sizes were calculated in terms of Cohen’s d (Cohen, 1988). The role played by verbal working memory on linguistic performance was controlled by performing an additional series of One-Way ANCOVAs with the scores at the NWR subtest of the BVL_4-12 as covariate. Table 3 reports the results of these analyses showing the level of group-related significance before (t-tests) and after covariation (ANCOVAs). When verbal working memory was not included in the analysis as covariate, the independent samples t-tests showed that the group of children with SLI had impaired performance on all of these measures: MLU (t(61) = 2.934; p < .005; d = .74); Sentence Completion (t(62) = 2.679; p < 009; d = .67); % Paragrammatic Errors (t(61) = 3.562; p < .001; d = .89); % Complete Sentences (t(61) = 3.756; p < 001; d = .95). Interestingly, when the analyses were replicated by performing a series of One-Way ANCOVAs with the scores at the NWR subtest of the BVL_4-12 as covariate the group-related differences on almost all of these measures disappeared: MLU (p = .038), Sentence Completion (p = .155), and % Paragrammatic Errors (p = .029). The only measure that was not significantly affected by vWM was the % of Complete Sentences on the Nest Story description task (p < .010) (see Table 3). This might be explained in two ways: one possibility is that the vWM difficulties of the participants with SLI affected their ability to correctly process the grammatical information required to complete and understand sentences in traditional tasks as well as their ability to produce morphologically sound words on a narrative description task, but not their ability to produce well-formed sentences on the same narrative production task; a second possibility is that this apparently awkward result might depend on a different factor, such as the potential interconnections between micro- and macrolinguistic aspects of linguistic processing. As the first possibility seems unlikely, we explored the potential relationship between the % of Complete Sentences and the macrolinguistic measures that might have affected the production of complete sentences in the narrative task (i.e., % of Global Coherence Errors and % of Lexical Informativeness) by using Pearson product-moment correlation coefficient. Interestingly, these correlations were highly significant (% Global Coherence Errors: r = .402; p < 025; % of Lexical Informativeness: r = .571; p < 001). 3.3. Analysis of narrative performance The group-related differences on the measures assessing narrative skills (i.e., % Global Coherence Errors, % Lexical Informativeness, % Thematic Selection) were analyzed with a series of independent-samples t-tests. Also in this case the level Table 2 Results of the analysis of Lexical skills in the groups of participants with SLI and typical language development. The table reports also significance levels before (t-tests) and after (ANCOVAs) covariation. Lexical skills
SLI
TLD
Naming Semantic Fluency Speech Rate % Semantic paraphasias
58.89 (3.41) 23.63 (6.61) 88.97(26.11) 1.68 (1.52)
61.47 28.50 108.07 .39
(3.46) (7.96) (23.21) (.75)
t-Value
df
p-Value
Cohen’s d
NWR Covariate (p-value)
2.920 2.667 3.070 4.290
59 62 61 44
.005* .010* .003* .001*
.75 .67 .77 1.08
.009§ .163 .036 .001§
SLI: children with Specific Language Impairment; TLD: children with Typical Language Development. * Group-related differences significant after Bonferroni correction for multiple comparisons (p < .0125). § Group-related differences still significant after covariation for NWR.
A. Marini et al. / Research in Developmental Disabilities 35 (2014) 3534–3542
3539
Table 3 Results of the analysis of grammatical skills in the groups of participants with SLI and typical language development. The table reports also significance level before (t-tests) and after (ANCOVAs) covariation. Grammatical skills MLU Sentence Completion % Paragrammatic errors % Complete Sentences
SLI 4.74 (1.36) 10.28 (2.98) 1.58 (1.83) 49.39 (20.97)
TLD 5.72 (1.27) 12.03 (2.19) .33 (.75) 67.65 (17.51)
t-Value 2.934 2.679 3.562 3.756
df 61 62 61 61
p-Value *
.005 .009* .001* .001*
Cohen’s d
NWR Covariate (p-value)
.74 .67 . 89 .95
.038 .155 .029 .010§
SLI: children with Specific Language Impairment; TLD: children with Typical Language Development. * Group-related differences significant after Bonferroni correction for multiple comparisons (p < .0125). § Group-related differences still significant after covariation for NWR.
of statistical significance was set at p < .0166 (.05/3 dependent variables) after Bonferroni correction for multiple comparisons. Effect sizes were calculated in terms of Cohen’s d (Cohen, 1988). The role played by verbal working memory on linguistic performance was controlled by performing an additional series of One-Way ANCOVAs with the scores at the NWR subtest of the BVL_4-12 as covariate. Table 4 reports the results of these analyses showing the level of group-related significance before (t-tests) and after covariation (ANCOVAs). When verbal working memory was not included in the analysis as covariate, the independent samples t-tests showed that the group of children with SLI had impaired performance on % Global Coherence Errors (t(61) = 3.580; p < .001; d = .90) and % Lexical Informativeness (t(61) = 3.806; p < .001; d = .96). The only measure on which the group of participants with SLI was not impaired was % Thematic Selection (t(62) = 1.808; p = .075; d = .45). Interestingly, when the analyses were replicated by performing a series of One-Way ANCOVAs with the scores at the NWR subtest of the BVL_4-12 the group-related differences survived for both % Global Coherence Errors (p < .001) and % Lexical Informativeness (p < .003) (see Table 4). Also in this case, in order to analyze the potential relationship among the macrolinguistic measures where the participants with SLI were found impaired a series of correlational analyses were conducted using Pearson product-moment correlation coefficient. Interestingly, the correlation between % Global Coherence Errors and % Lexical Informativeness was highly significant (r = .597; p < 001). 3.4. Discussion This study investigated linguistic and narrative abilities in children with Specific Language Impairment compared to a group of children with typical language development. Furthermore, the effect potentially exerted by verbal working memory on their linguistic production skills was examined by exploring the role of vWM on lexical, grammatical and narrative processing, respectively. The results showed that the participants with SLI had difficulties at all three levels of linguistic processing and that vWM apparently affected their linguistic skills in different ways. This effect was marginal on lexical processing, significant on grammatical structuring, and null on narrative construction. Before controlling for the effects of vWM on linguistic performance, the children with SLI showed generalized linguistic difficulties with the only, notable, exception of % Thematic Selection. The narrative speech samples produced by the participants with SLI were characterized by a high incidence of global coherence errors and a significantly reduced amount of informative words (i.e., % Lexical Informativeness). The production of global coherence errors might be interpreted as a consequence of their lexical difficulties. Indeed, the macrolinguistic analysis showed that they produced significantly more errors of global coherence than the group of healthy participants. A qualitative inspection of these errors revealed that the children with SLI did not produce tangential or conceptually incongruous utterances. Rather, they produced propositional repetitions and filler utterances. Together with the adequate score of thematic selection, these observations suggest that they had understood the gist of the story, could select all expected concepts, being even able to refrain from introducing irrelevant (i.e., tangential or conceptually inappropriate) propositions. Our data suggest that these children did not have problems in identifying the target concepts and that the macrolinguistic impairment might be an epiphenomenon of a microlinguistic one. Indeed, the relevant difficulties experienced by the participants with SLI on measures of lexical production (Naming, Semantic Fluency, Speech Rate and % of Semantic Paraphasias) suggest an impairment in the processes of both lexical selection and access. Furthermore, their generalized grammatical difficulties (measured in terms of MLU, % of Paragrammatic Errors, % Complete Sentences, and Sentence Completion) highlight the presence of additional problems also Table 4 Results of the analysis of narrative skills in the groups of participants with SLI and typical language development. The table reports also significance level before (t-tests) and after (ANCOVAs) covariation. Narrative skills
SLI
TLD
% Global Coherence Errors % Lexical Informativeness % Thematic Selection
14.10 (11.45) 75.23 (11.53) 31.12 (11.35)
5.44 (7.36) 86.03 (10.98) 36.20 (11.12)
t-Value
df
p-Value
3.580 3.806 1.808
61 61 62
.001* .001* .075
SLI: children with Specific Language Impairment; TLD: children with Typical Language Development. * Group-related differences significant after Bonferroni correction for multiple comparisons (p < .0166). § Group-related differences still significant after covariation for NWR.
Cohen’s d .90 .96 .45
NWR Covariate (p-value) .001§ .003§ .733
3540
A. Marini et al. / Research in Developmental Disabilities 35 (2014) 3534–3542
in the ability to extract from the activated lexical items all the necessary morphological and morphosyntactic information that is required to generate adequate syntactic representations of the intended message. Noteworthy, 20% of all words were lexical repetitions and lexical fillers, mainly personal comments about the story or repetitions of previously introduced ideas. Most of these lexical repetitions and fillers clustered in repetitive and filler utterances that, in turn, increased the amount of global coherence errors. It seems, then, that a difficulty in accessing specific lexical entries (a microlinguistic process) may have affected the ability to keep narrative coherence (a macrolinguistic process). Interestingly, also the reduction in lexical informativeness may be linked to a disturbance in bottom-up processes originating in the process of lexical retrieval rather than from top-down processes concerning the conceptualization of the story. Importantly, the significant negative correlation between the percentage of lexical informativeness and the percentage of global coherence errors suggests that the speech samples were progressively less informative with increasing production of propositional repetitions and filler utterances that, as we have seen earlier, are probably linked to the reduced levels of lexical retrieval. Overall, then, our general impression is that the production of repetitive and filler utterances, rather than a sign of a problem in the macrolinguistic organization of their narrative discourse, might reflect a strategy to cope with a lexical problem: while in the phase of message generation, their word selection difficulties may have triggered the production of filler or repeated utterances that allowed them to skip prolonged pauses perceived as more annoying than simply ‘‘filling the gap’’. This might have also led to the production of fewer complete sentences as shown by the significant correlation between the % of Errors of Global Coherence and the % of Complete Sentences. This represents an interesting example of how a multilevel procedure for discourse analysis might shed light not only on selective (i.e., lexical, grammatical, narrative) aspects of message production, but also on the complex interactions occurring among micro- and macrolinguistic processes. Once verbal working memory was controlled for, some of these group-related differences disappeared. At the lexical level, the covariation showed that vWM affected minimally this level of linguistic processing in the participants with SLI (i.e., Speech Rate on the narrative description task and performance on the Semantic Fluency task). Semantic Fluency is usually assumed to tap the extension of a child’s mental lexicon through a complex procedure involving the ability to monitor the execution of the test, inhibit the production of words that can be either semantically unrelated with the target category or mere repetitions of previously produced items. Therefore, this test cannot be considered as a purely linguistic task. Its execution is likely affected not only by the extension of the child’s lexical repertoire, but also by other factors such as executive skills and verbal working memory. Indeed, our results confirm the important role of vWM in this task. Likely, vWM is critical to keep track of the words that have already been produced so to avoid useless repetitions. Noteworthy, the influence of verbal working memory on this task may not be limited to the procedures that allow its execution. Indeed, verbal working memory is thought to significantly contribute to the process of lexical learning, especially in younger children with SLI (e.g., Ellis & Sinclair, 1996; Martin & Gupta, 2004; Montgomery, Magimairaj, & Finney, 2010). Therefore, the influence of such a critical cognitive function on this task might have been both direct and indirect. It may have directly affected the execution of the task (via the monitoring component of working memory). However, it might have also indirectly altered the performance as these children have smaller lexical repertoires because of their working memory difficulties. Similar considerations may apply also to Speech Rate. If vWM’s influence on lexical processing was minimal, its influence on grammatical processing was highly significant. Indeed, vWM affected all of the grammatical measures considered in this study with the only exception of the % of Complete Sentences produced on the narrative production task. This might be explained in two ways: one possibility is that the vWM difficulties of the participants with SLI affected their ability to correctly process the grammatical information required to complete and understand sentences in traditional tasks as well as their ability to produce morphologically sound words on a narrative description task, but not their ability to produce wellformed sentences on the same narrative production task; a second possibility is that this apparently awkward result might depend on the potential interconnections between micro- and macrolinguistic aspects of linguistic processing. As the first possibility seems unlikely, we explored the potential relationship between the % of Complete Sentences and the % of Global Coherence Errors and found that this correlation was significant. This suggests that vWM plays a major role in grammatical processing in children with SLI and highlights the importance of including a narrative production task in the clinical assessment of linguistic symptoms in these individuals. A final consideration regards the null impact of vWM on measures of macrolinguistic organization. As we have already noted, it is likely that these errors were the epiphenomenon of a lexical impairment rather than reflecting a macrolinguistic impairment. In conclusion, these findings suggest that SLI is a complex developmental disorder whose symptoms are linked to the interaction among several factors. While supporting the claims about the important role played by vWM in determining at least specific aspects of the linguistic profile of children with SLI (e.g., Archibald & Gathercole, 2007; Bishop, 2006), the results from the current study suggest that vWM might not be the core symptom of SLI (Lum, Conti-Ramsden, Page, & Ullman, 2011). These results increase our understanding of the effects exerted by impairments in verbal working memory on linguistic performance. This is coherent with the observation that, even if frequent in children with SLI (e.g., Duinmeijer et al., 2012; Montgomery, 2006; Montgomery et al., 2010), limitations in verbal working memory are not always present (Archibald & Joanisse, 2009). Furthermore, the present study also showed that narrative discourse in children with SLI may present signs of both micro- and macrolinguistic impairment. However, our data suggest that the macrolinguistic impairment might be just an epiphenomenon of the underlying deficit in lexical processing that, from a structural point of view, leads to the frequent interruption of the utterances and determines a reduction of the levels of sentence completeness, while from a functional point of view determines the selection of strategies to cope with the impairment and leads to the production of lexical fillers and repetitions. Once clustered in propositional repetitions and filler utterances, these
A. Marini et al. / Research in Developmental Disabilities 35 (2014) 3534–3542
3541
uninformative words lower the levels of global coherence across the narratives. And this will eventually determine a significant reduction of the levels of informativeness. Interestingly, this complex interplay between micro- and macrolinguistic processes can be found, even if with different characteristics, also in patients with other aethiologies. For example, in a group of non-aphasic persons with traumatic brain injury the production of errors of cohesion (a macrolinguistic measure) determined a reduction in speech rate (a microlinguistic measure) suggesting that the latter may have been determined by the frequent interruptions in the flow of thoughts rather than more specific linguistic deficits (Marini, Galetto, et al., 2011). Similarly, in a group of schizophrenic participants the massive production of global coherence errors triggered the production of microlinguistic errors in terms of semantic paraphasias (Marini, Spoletini, et al., 2008). A final remark regards the limitations of this study. Indeed, due to the small number of participants, the conclusions outlined above are based on the results of bivariate correlations. However, the relationships among these measures may not necessarily be linear. Therefore, further studies are needed with larger groups of participants in order to perform more sophisticated statistical analyses, such as factorial designs or regression models that will allow researchers to deepen the understanding of the complex interaction among micro- and macrolinguistic processes in children with SLI. Acknowledgements This work was supported by a grant from IRCCS ‘‘E. Medea’’ (Ricerca Corrente 2013, Italian Ministry of Health) to A. M. The authors wish to thank Daniela Milan, Paola Casonato, Nicoletta Drigo, Marcella Zamparini and all those speech therapists who actively helped us in the selection and recruitment of the children who participated to this study. References Archibald, L. M. D., & Gathercole, S. E. (2007). Nonword repetition in specific language impairment: More than a phonological short-term memory deficit. Psychonomic Bullettin, 14(5), 919–924. Archibald, L. M. D., & Joanisse, M. (2009). On the sensitivity and specificity of nonword repetition and sentence recall to language and memory impairments in children. Journal of Speech, Language and Hearing Research, 52, 899–914. Bishop, D. V. M. (2006). What causes specific language impairment in children? Current Direction in Psychological Science, 15(5), 217–221. Botting, N. (2002). Narrative as a tool for the assessment of linguistic and pragmatic impairments. Child Language Teaching and Therapy, 18, 1–12. Botting, N., & Conti-Ramsden, G. (2001). Non-word repetition and language development in children with specific language impairment. International Journal of Language and Communication Disorders, 36, 421–432. Carletta, J. (1996). Assessing agreement on classification tasks: The kappa statistic. Computational Linguistics, 22, 249–254. Chiat, S., & Roy, P. (2007). The preschool repetition test: An evaluation of performance in typically developing and clinically referred children. Journal of Speech and Hearing Research, 50, 429–443. Cohen, J. (1988). Statistical power analysis for the behavioral sciences (2nd ed.). NJ: Lawrence Erlbaum. Conti-Ramsden, G., & Botting, N. (2001). Psycholinguistic markers for specific language impairment (SLI). Journal of Child Psychology and Psychiatry, 42, 741–748. Cowan, N., Nugent, L., Elliott, E., Ponomarev, I., & Saults, S. (2005). The role of attention in the development of short-term memory: Age differences in the verbal span of apprehension. Child Development, 70, 1082–1097. Duinmeijer, I., de Jong, J., & Scheper, A. (2012). Narrative abilities, memory and attention in children with a specific language impairment. International Journal of Language and Communication Disorders, 47(5), 542–555. Ellis, N. C., & Sinclair, S. G. (1996). Working memory in the acquisition of vocabulary and syntax: Putting language in good order. The Quarterly Journal of Experimental Psychology, 49, 234–250. Ellis Weismer, S., Evans, J., & Hesketh, L. (1999). An examination of verbal working memory capacity in children with specific language impairment. Journal of Speech, Language and Hearing Research, 42, 1249–1260. Finneran, D. A., Francis, A. L., & Leonard, L. B. (2009). Sustained attention in children with specific language impairment. Journal of Speech, Language and Hearing Research, 52, 915–929. Gathercole, S., & Baddeley, A. (1990). Phonological memory deficits in language disordered children: Is there a causal connection? Journal of Memory and Language, 29, 336–360. Graf Estes, K., Evans, J. L., & Else-Quest, N. M. (2007). Differences in the nonword repetition performance of children with and without specific language impairment: A meta-analysis. Journal of Speech Language and Hearing Research, 50, 177–195. Henry, L. A., Messer, D. J., & Nash, G. (2011). Executive functioning in children with specific language impairment. Journal of Child Psychology and Psychiatry, 53(1), 37–45. Laws, G., & Bishop, D. V. M. (2003). A comparison of language abilities in adolescents with Down Syndrome and children with Specific Language Impairment. Journal of Speech, Language and Hearing Research, 46, 1324–1339. Leonard, L. B. (1998). Children with specific language impairment. Cambridge, MA: MIT Press. Lum, J. A., Conti-Ramsden, G., Page, G., & Ullman, D. M. T. (2011). Working, declarative and procedural memory in specific language impairment. Cortex, 48(9), 1138–1154. Marini, A. (2012). Characteristics of narrative discourse processing after damage to the right hemisphere. Seminars in Speech and Language, 33(1), 68–78. Marini, A., Andreetta, S., del Tin, S., & Carlomagno, S. (2011). A multi-level approach to the analysis of narrative language in aphasia. Aphasiology, 25(11), 1372–1392. Marini, A., Galetto, V., Zampieri, E., Vorano, L., Zettin, M., & Carlomagno, S. (2011). Narrative language in traumatic brain injury. Neuropsychologia, 49, 2904–2910. Marini, A., Marotta, L., Bulgheroni, S., & Fabbro, F. (2014). Batteria per la Valutazione del Linguaggio in Bambini dai 4 ai 12 anni. Firenze, Italy: Giunti OS. Marini, A., Spoletini, I., Rubino, I. A., Ciuffa, M., Bria, P., Martinotti, G., et al. (2008). The language of schizophrenia: An analysis of micro and macro-linguistic abilities and their neuropsychological correlates. Schizophrenia Research, 105, 144–155. Marini, A., Tavano, A., & Fabbro, F. (2008). Assessment of linguistic abilities in Italian children with specific language impairment. Neuropsychologia, 46, 2816–2823. Marini, A., & Urgesi, C. (2012). Please get to the point! A cortical correlate of linguistic informativeness Journal of Cognitive Neuroscience, 24(11), 2211–2222. Martin, N., & Gupta, P. (2004). Exploring the relationship between word processing and verbal short-term memory: Evidence form associations and dissociations. Cognitive Neuropsychology, 21, 213–228. Montgomery, J. W. (2006). Real-time language processing in school-age children with specific language impairment. International Journal of Language and Communication Disorders, 41, 275–291. Montgomery, J. W., Magimairaj, B. M., & Finney, M. C. (2010). Working memory and specific language impairment: An update on the relation and perspectives on assessment and treatment. American Journal of Speech-Language Pathology, 19, 78–94. Newbury, D. F., Fisher, S. E., & Monaco, A. P. (2010). Recent advances in the genetics of language impairment. Genome Medicine, 2, 6.
3542
A. Marini et al. / Research in Developmental Disabilities 35 (2014) 3534–3542
Paradis, M. (1987). The assessment of bilingual aphasia. Hillsdale, NJ: Erlbaum. Raven, J. C. (1938). Progressive Matrices: A Perceptual Test of Intelligence. London: HK Lewis. Vicari, S. (2007). Prove di Memoria e Apprendimento per l’Eta` Evolutiva (PROMEA). Firenze, Italy: Giunti OS. Vugs, B., Hendriks, M., Cuperus, J., & Verhoeven, L. (2014). Working memory performance and executive function behaviors in young children with SLI. Research in Developmental Disabilities, 35, 62–74. Wechsler, D. (1993). WISC—R. Scala d’intelligenza Wechsler per bambini riveduta. Firenze, Italy: Giunti OS.
Contents lists available at ScienceDirect
Research in Developmental Disabilities
Differential verbal working memory effects on linguistic production in children with Specific Language Impairment Andrea Marini a,b,*, Cinzia Gentili c, Massimo Molteni c, Franco Fabbro a,b a
Department of Human Sciences, University of Udine, Udine, Italy IRCCS ‘‘E. Medea: La Nostra Famiglia’’, San Vito al Tagliamento, Pn, Italy c IRCCS ‘‘E. Medea: La Nostra Famiglia’’, Bosisio Parini, Lc, Italy b
A R T I C L E I N F O
A B S T R A C T
Article history: Received 26 June 2014 Accepted 28 August 2014 Available online 18 September 2014
Deficits in verbal working memory (vWM) have often been reported in children with Specific Language Impairments (SLIs) and might significantly contribute to their linguistic difficulties. The linguistic and narrative skills of a group of children with diagnosis of SLI were compared to those of a group of children with typical development. The linguistic assessment included a comprehensive analysis of their lexical, grammatical and narrative abilities. Overall, the participants with SLI had difficulties at all three levels of linguistic processing. The effect of vWM was marginal on lexical processing, significant on grammatical structuring, and null on narrative construction. ß 2014 Elsevier Ltd. All rights reserved.
Keywords: Specific Language Impairment Language Verbal working memory Non word repetition
1. Introduction The label Specific Language Impairment (SLI) applies to a cluster of relatively frequent developmental disorders characterized by language delay in children with otherwise normal physical and intellectual development (Leonard, 1998; Newbury, Fisher, & Monaco, 2010). SLI is not a homogenous condition (Laws & Bishop, 2003). As language is a complex cognitive function resting on the interaction between different levels of processing, its development may be affected in different ways. A comprehensive linguistic assessment should include all of these levels of processing and address the delicate issue of their reciprocal interconnections. Generally speaking, language can be analyzed from two major points of view: 1) a within-sentence or microlinguistic level, which focuses on lexical (i.e., phonetic, phonological, morphological and semantic) and grammatical (i.e., morphosyntactic and syntactic) skills; 2) a between-sentence or macrolinguistic level, which focuses on pragmatic (i.e., the ability to generate inferences so to contextualize the literal meanings of words and sentences, and understand the communicative intentions of the interlocutors) and discourse (i.e., the ability to establish linguistically cohesive and conceptually coherent ties) skills. Increasing evidence shows that these levels (and their interrelations) can be simultaneously assessed by adopting a multi-level approach to narrative language analysis (e.g., Marini, Andreetta, del Tin, & Carlomagno, 2011) and that this ensures a comprehensive understanding of the linguistic symptoms observed in both adult patients and children with developmental and/or acquired disorders (e.g., Marini, Tavano, & Fabbro, 2008; see also Botting, 2002 for further considerations on the usefulness of narrative as a tool for the assessment of both linguistic and pragmatic impairments). For example, Marini, Tavano and Fabbro (2008) showed that the use of a narrative production
* Corresponding author at: Department of Human Sciences, University of Udine, Italy. Tel.: +39 3355393224. E-mail address: [email protected] (A. Marini). http://dx.doi.org/10.1016/j.ridd.2014.08.031 0891-4222/ß 2014 Elsevier Ltd. All rights reserved.
A. Marini et al. / Research in Developmental Disabilities 35 (2014) 3534–3542
3535
task might highlight the presence of lexical and grammatical disturbances in children with SLI who had scored within normal range on traditional linguistic assessment tests assessing the same abilities. Accumulating evidence suggests that children with SLI might experience also additional disturbances that are not specifically linguistic. For example, Vugs, Hendriks, Cuperus, and Verhoeven (2014) have recently showed that a large cohort of children with SLI aged 4–5 years scored worse than a group of age-matched peers with typical development on tests assessing executive functions (i.e., inhibition, shifting, emotional control, and planning/organization) and both verbal and visuospatial working memory. These results are not an isolate finding. Indeed, difficulties in the ability to select appropriate responses, inhibit irrelevant ones, plan strategies (e.g., Finneran, Francis, & Leonard, 2009; Henry, Messer, & Nash, 2011) as well as impairments in verbal working memory (vWM) have been frequently observed in children with SLI (Ellis Weismer, Evans, & Hesketh, 1999; Montgomery, 2006) and might affect their linguistic development and functioning. The concept of working memory refers to a set cognitive processes that allow to temporarily store and manipulate limited information (e.g., Cowan, Nugent, Elliott, Ponomarev, & Saults, 2005). According to the phonological storage deficit hypothesis, a difficulty in storing the incoming phonological information might contribute to their linguistic impairments (Archibald & Gathercole, 2007; Bishop, 2006). A reduced vWM capacity may hamper the generation of adequate phonological representations in the mental lexicon, thus leading to difficulties in their recognition when acting as listeners and in their production when involved in language production. These disturbances might negatively affect the child’s linguistic development: a difficulty in keeping track of phonological and/or lexical items in short-term memory and eventually process them in working memory might result in slowed vocabulary acquisition (e.g., Gathercole & Baddeley, 1990) and trigger a range of linguistic difficulties (Graf Estes, Evans, & Else-Quest, 2007). Interestingly, the few children with SLI with normal levels of vWM have also better performance on tests assessing linguistic skills and produce longer utteraces (e.g., Botting & Conti-Ramsden, 2001). If present, limitations in vWM might affect the children’s efficiency in daily communicative interactions and may lead to the production of shorter utterances. For example, Duinmeijer, de Jong, and Scheper (2012) showed that the performance of a group of children with SLI on the digit span correlated positively with the mean lenght of their utterances (MLU). Furthermore, a deficit in verbal working memory may trigger the selection of inappropriate words that, in turn, might determine qualitatively inferior levels of organization at the text level (macrolinguistic organization). Unfortunately, the interconnections among these apparently different skills (i.e., working memory, lexical selection, and macrolinguistic organization) cannot be detected by traditional tests assessing production skills in isolation (e.g., naming tests, fluency tasks, etc) but might emerge clearly when applying a multi-level approach to the analysis of language samples obtained on narrative production tasks. For these reasons, the present study focuses on two major issues: (1) the need for a comprehensive assessment of linguistic functions in children with SLI and (2) the opportunity to consider the role potentially played by verbal working memory in linguistic processing in these children with all the implications for appropriate linguistic assessment and remediation. Namely, we analyzed in detail the linguistic and narrative skills of a group of schoolage children with SLI and compared their performance with that of a cohort of children with typical development. Importantly, as the levels of verbal working memory might have affected the linguistic performance of the children included in the study, we hypothesized that, when controlling for working memory in the statistical analyses, those group-related differences on language tasks that were heavily biased by vWM capacity would disappear. We also hypothesized that such a comprehensive analysis would allow us to provide a clearer picture of the linguistic disorders that characterize these children and of their potential interconnections at different levels of processing. 2. Methods 2.1. Participants Sixty-four Italian-speaking participants were included in the study. They formed an experimental and a control group (see Table 1). The experimental group was formed by 32 children with diagnosis of SLI with mixed expressive-receptive disorder (ICD-10 diagnosis: F80.2) aged between 7 and 11.11 years old (mean 8 years and 8 months; standard deviation, SD, 1.6). None of them had mental retardation, psychotic symptoms, brain lesions, auditory difficulties or comorbid dyslexia. The General (verbal and performance) intelligence of SLI participants was assessed by administering the Wechsler scales (Wechsler, 1993). All participants were required to have a Performance Quotient (PIQ) of 85 or above this threshold. The average performance of the SLI participants on Verbal and Performance Intelligence Quotient (IQ) was matched against
Table 1 Means (and standard deviations) showing demographic data of the two groups of participants and their performance on the non-word repetition task.
Age Education Sex Non-word repetition*
SLI (N = 32)
TLD (N = 32)
8.77 (1.58) – range: 7–11.11 3.2 (1.6) – range: 1st–6th grade M = 19 13.34 (2.21)
8.61 (1.41) – range: 7–11.11 3.5 (1.3) – range: 1st–6th grade M = 17 14.69 (1.00)
SLI: children with Specific Language Impairment; TLD: children with Typical Language Development. * Group-related difference on this task was significant (p < 05).
3536
A. Marini et al. / Research in Developmental Disabilities 35 (2014) 3534–3542
general population norms and results are expressed as Intelligence Quotient scores (mean 100, SD 15). Overall, the mean Fullscale Intelligence Quotient (FIQ) score (mean = 86) was within 1 SD below average, with a relative discrepancy between the Verbal Scale (mean = 79: more than one SD below normal range) and the Performance Scale (mean = 94: within normal range). The control group was formed by 32 participants with Typical Language Development (TLD) aged between 7 and 11.11 years old (mean 8 years and 6 months; SD 1.4). They were selected from a larger sample of 1300 children recruited for the standardization of the ‘‘Batteria per la Valutazione del Linguaggio in Bambini dai 4 ai 12 anni’’ (BVL_4-12, Marini, Marotta, Bulgheroni, & Fabbro, 2014), a comprehensive battery of tests designed to tap three modalities (i.e., repetition, comprehension, and production) along several levels of processing: from phonetics to discourse. They performed within normal range on the Raven’s progressive matrices (Raven, 1938), in the non-word repetition subtest of the Prove di Memoria e Apprendimento per l’Eta` Evolutiva (PROMEA; Vicari, 2007), and on the forward and backward digit spans’ subtests of the Wechsler Scales (Wechsler, 1993). Furthermore, they had average school performance in language and reading. In a preliminary interview, their teachers confirmed that they had normal cognitive and learning development. According to school records and parents’ reports none of them had a known history of psychiatric or neurological illness, learning disabilities, hearing or visual loss. The two groups had similar chronological age and education (age: t(62) = .437; p = 664; education: t(62) = .886; p = .379). However, an independent-samples t-test showed that the SLI participants produced fewer correct repetitions on the Non-Word Repetition subtest of the BVL-4-12 (t(62) = 3.136; p < 003; d = .766). Interestingly, performance on tests of non word repetition is considered a measure of vWM and a reliable marker of SLI (Conti-Ramsden & Botting, 2001) as it differentiates children with SLI already at the age of 2.6 years (Chiat & Roy, 2007). All participants were selected among middle class families in regions in North-East Italy (Lombardia, Veneto and Friuli Venezia Giulia). Children with diagnosed hearing impariments, neurological disorders, austistic spectrum disorder or ADHD were not included in the study. In both groups the level of education of the parents ranged from high school to Master’s degree. For the experimental group, data collection was performed during standard linguistic evaluation at the IRCSS ‘‘E. Medea’’, Ass.ne ‘‘La Nostra Famiglia’’ in San Vito al Tagliamento (Pn) and Bosisio Parini (Lc). The study received institutional ethics’ approval by the Ethics Committee of the Research Institute IRCCS ‘‘E. Medea’’. All parents released their informed consent to the participation of their children to the study and to the treatment of the data. 2.2. Procedures of linguistic assessment The linguistic assessment was delivered by trained speech-therapists or developmental psychologists in a quiet room at the Research Institute IRCCS ‘‘E. Medea’’ (for children with SLI) or in their schools (for children with TD). The linguistic skills of the participants were assessed by administering the BVL_4-12 (Marini et al., 2014). The administration of these tests occurred in two sessions of approximately 45 min each and the children’s performance was audio-recorded. In the first session children received a cohort of tests assessing linguistic and narrative production. In the second session they received tests assessing repetition and comprehension skills. For the specific purposes of this study we will report the performance of these children on tests assessing lexical (Section 2.2.1), grammatical (Section 2.2.2), and narrative (Section 2.2.3) production abilities. 2.2.1. Assessment of lexical production skills The participants’ lexical productive skills were evaluated by administering tasks assessing naming, semantic fluency and narrative discourse production. In the naming subtest of the BVL 4-12 children were asked to tell the name of a series of stimuli depicted in sheets. The stimuli have been selected so to represent words with different (1) frequencies in spoken Italian (words with low, medium and high frequency), (2) grammatical class (verbs and nouns), and (3) semantic features (action verbs and nouns pertaining to 16 semantic categories). Each correct answer was assigned 1 point. The maximum score for this test is 67. The subtest of semantic fluency required the children to produce as fast as they could a series of words according to a specific semantic criterion within 1 min. Namely, in this version of the test the participants were asked to produce as many words as they could that pertained to the semantic category ‘‘animals’’ and then to the semantic category ‘‘objects in the house’’. This test offers a good way to assess the extension of a child’s vocabulary. Additional measures assessing lexical productive skills were elicited by administering a cartoon story made of six drawings presented on the same page (the ‘‘Nest Story’’: Paradis, 1987). Each storytelling was tape-recorded and transcribed verbatim by two independent coders. Two independent raters analyzed these transcriptions calculating the number of words and the time of each storytelling. These two parameters (i.e., words produced and time elapsed during the storytelling) allowed to derive a measure of speech rate considered in terms of words per minute (Marini, Andreetta, et al., 2011). This measure provides additional information on the children’s lexical efficiency, that is their ability to select words while producing a narrative discourse. Furthermore, a Percentage of Semantic Paraphasias was calculated by dividing the number of semantic paraphasias produced by each child by the number of words. These values were multiplyed by 100. A semantic paraphasia was scored whenever a target word had been replaced by a semantically related one (e.g., fiore [Flower] instead of albero [Tree]). 2.2.2. Assessment of grammatical production skills The children’s grammatical skills were assessed in terms of mean lenght of utterance (MLU) in the Nest Story description task, morphosyntactic organization and syntactic processing. The former aspect was analyzed by deriving by each story
A. Marini et al. / Research in Developmental Disabilities 35 (2014) 3534–3542
3537
description the total number of utterances calculated following the criteria detailed in Marini, Andreetta, et al. (2011). The MLU was then calculated by dividing the total number of words by the number of utterances. Morphosyntactic organization was analyzed by administering a test of sentence completion and by deriving from the story tellings produced by the participants a % of paragrammatic errors to words. The sentence completion subtest of the BVL_4-12 assesses the child’s ability to process derivational and inflectional moprhology to produce grammatically well-formed sentences. Namely, each child listened to a sentence (the model; e.g., Marco apre la porta [Marco opens the door]) and then heard the beginning of a second sentence (the prompt), which (s)he was asked to complete by assigning the correct morphemes to the verb (the target). In our example, the prompt is Anche noi. . . (in English: We also. . .) and the child was expected to complete the target sequence as follows: Anche noi apriamo la porta (in English: We also open the door). The test is made of 14 pairs of model sentences and prompts with different levels of grammatical complexity. Each correct answer was assigned 1 point with a maximum score of 14. The other measure tapping the participants’ morphosyntactic skills, i.e., % of paragrammatic errors, was derived by the analysis of the narrative speech samples produced by the participants while telling the Nest Story. Paragrammatic errors include the substitution of free morphemes, which in Italian involve mainly the substitution of function words (for example, batte da una porta ‘‘he is knocking from a door’’ – da instead of a), and of bound morphemes (for example, Questo e` una coppia ‘‘this [masc] is a couple [fem]’’ – should be questa). The analysis of the participants’ syntactic processing skills was assessed by deriving a measure of grammatical completeness from the descriptions of the Nest Story and by administering a test of syntactic comprehension, respectively. The % of Complete Sentences was derived by dividing the number of utterances that had been scored as well-formed from a grammatical point of view produced during the description of the Nest Story by the total number of utterances produced during the story telling (Marini, Andreetta, et al., 2011). An utterance was considered grammatically complete if all the arguments required by the verb had been correctly inserted and no omissions or substitutions of free or bound morphemes could be found.
2.2.3. Assessment of narrative production skills The children’ narrative skills were assessed in terms of textual organization and levels of informativeness. As for textual organization at intersentential level, the macrolinguistic measures included a Percentage of Global Coherence Errors. Errors of global coherence included the production of utterances that may be tangential, conceptually incongruent with the story, propositional repetitions, or simple fillers. An utterance is considered tangential when it contains a derailment in the flow of discourse with respect to the information previously provided in preceding utterances as in the following sequence produced during the description of the Picnic picture:/It is a picnic/I like picnics/I have made several picnics in my life/. Here, the second and the third utterances are scored as tangential, as they provide irrelevant information triggered by a specific idea portrayed in the stimulus picture. An utterance is considered conceptually incongruent when it includes ideas not directly addressed by the stimulus. Let us consider the following sequence produced during the description of the Cookie Theft picture:/the children are trying to get the cookies/the TV is out/. In this example, the second utterance is scored as conceptually incongruent as in the picture stimulus there is no TV. A propositional repetition is a sequence where the speaker simply repeats ideas that have been already mentioned. Therefore, any new information is provided (e.g., the second utterance in the sequence:/the man is walking on the sidewalk/he is on the sidewalk/). Finally, a filler utterance is scored whenever the speaker produces an utterance that does not provide any additional information like in the following sequence:/the man and the woman are eating/my god, and now?/ah, yes, I get it/. The last two sequences are scored as filler utterances. The Percentage of Global Coherence Errors was calculated by dividing the number of global coherence errors by the number of utterances and multiplying this value by 100. The informative skills of the children were assessed in terms of percentage of thematic units identified in each story telling and percentage of informative words in the sample. A thematic unit is considered one of the main ideas or details in the story that have been previously identified. The number of thematic units produced by each child was considered an index of the participants’ abilities to derive information from the picture stimuli. Therefore, we could derive a Percentage of Thematic Selection by dividing the thematic units identified by each child in the Nest Story by the total number of possible thematic units present in the story and multiplying this value by 100. In order to derive the Percentage of Lexical Informativeness for each storytelling we identified the lexical information units (LIUs), i.e., those content and function words that were not only phonologically well-formed but also appropriate from a grammatical and pragmatic point of view (e.g., Marini, 2012). Therefore, all those words that had been classified as semantic or verbal paraphasias, fillers, paragrammatic errors or forming tangential or extraneous utterances (i.e., utterances that were somehow deviating from the gist of the story) were excluded from the LIUs’ count. The Percentage of Lexical Informativeness was then obtained by dividing the number of LIUs by the number of words and multiplying this value by 100. The scoring procedure was performed independently by two raters on the stories produced by the participants and then compared. High interjudge reliability for the story narrative analyses has been previously documented (Marini & Urgesi, 2012). An interrater reliability analysis using Kappa statistics was performed to determine consistency among raters. Acceptable inter-rater reliability was defined as k .70 (Carletta, 1996).
A. Marini et al. / Research in Developmental Disabilities 35 (2014) 3534–3542
3538
3. Results 3.1. Analysis of lexical performance The group-related differences on the measures assessing lexical production (i.e., Naming, Semantic Fluency, Speech Rate, Percentage of Semantic Paraphasias produced in the Nest Story description task) were analyzed with a series of independentsamples t-tests. The level of statistical significance was set at p < .0125 (.05/4 dependent variables) after Bonferroni correction for multiple comparisons. Effect sizes were calculated in terms of Cohen’s d (Cohen, 1988). This is computed as the difference between the mean of the control group and that of the group of participants with SLI divided by the pooled sample standard deviation. The role played by verbal working memory on linguistic performance was controlled by performing an additional series of One-Way ANCOVAs with the scores at the NWR subtest of the BVL_4-12 as covariate. Table 2 reports the results of these analyses showing the level of group-related significance before (t-tests) and after covariation (ANCOVAs). When verbal working memory was not included in the analysis as covariate, the independent samples t-tests showed that the group of children with SLI had impaired performance on all of these measures: Naming (t(59) = 2.920; p < .005; d = .75); Speech Rate (t(61) = 3.070; p < .003; d = .77); Semantic Fluency (t(62) = 2.667; p < .010; d = .67), Percentage of Semantic Paraphasias (t(44) = 4.290; p < 001; d = 1.08). Interestingly, when the analyses were rerun controlling for NWR, only the group-related differences in Semantic Fluency and Speech Rate were no longer significant (p = .163 and p = .036, respectively), while those pertaining Naming (p < .009) and the Percentage of Semantic Paraphasias (p < .001) survived. 3.2. Analysis of grammatical performance The group-related differences on the measures assessing grammatical skills (i.e., MLU, Sentence Completion, % Complete Sentences, % Paragrammatic Errors) were analyzed with a series of independent-samples t-tests. Also in this case the level of statistical significance was set at p < .0125 (.05/4 dependent variables) after Bonferroni correction for multiple comparisons. Effect sizes were calculated in terms of Cohen’s d (Cohen, 1988). The role played by verbal working memory on linguistic performance was controlled by performing an additional series of One-Way ANCOVAs with the scores at the NWR subtest of the BVL_4-12 as covariate. Table 3 reports the results of these analyses showing the level of group-related significance before (t-tests) and after covariation (ANCOVAs). When verbal working memory was not included in the analysis as covariate, the independent samples t-tests showed that the group of children with SLI had impaired performance on all of these measures: MLU (t(61) = 2.934; p < .005; d = .74); Sentence Completion (t(62) = 2.679; p < 009; d = .67); % Paragrammatic Errors (t(61) = 3.562; p < .001; d = .89); % Complete Sentences (t(61) = 3.756; p < 001; d = .95). Interestingly, when the analyses were replicated by performing a series of One-Way ANCOVAs with the scores at the NWR subtest of the BVL_4-12 as covariate the group-related differences on almost all of these measures disappeared: MLU (p = .038), Sentence Completion (p = .155), and % Paragrammatic Errors (p = .029). The only measure that was not significantly affected by vWM was the % of Complete Sentences on the Nest Story description task (p < .010) (see Table 3). This might be explained in two ways: one possibility is that the vWM difficulties of the participants with SLI affected their ability to correctly process the grammatical information required to complete and understand sentences in traditional tasks as well as their ability to produce morphologically sound words on a narrative description task, but not their ability to produce well-formed sentences on the same narrative production task; a second possibility is that this apparently awkward result might depend on a different factor, such as the potential interconnections between micro- and macrolinguistic aspects of linguistic processing. As the first possibility seems unlikely, we explored the potential relationship between the % of Complete Sentences and the macrolinguistic measures that might have affected the production of complete sentences in the narrative task (i.e., % of Global Coherence Errors and % of Lexical Informativeness) by using Pearson product-moment correlation coefficient. Interestingly, these correlations were highly significant (% Global Coherence Errors: r = .402; p < 025; % of Lexical Informativeness: r = .571; p < 001). 3.3. Analysis of narrative performance The group-related differences on the measures assessing narrative skills (i.e., % Global Coherence Errors, % Lexical Informativeness, % Thematic Selection) were analyzed with a series of independent-samples t-tests. Also in this case the level Table 2 Results of the analysis of Lexical skills in the groups of participants with SLI and typical language development. The table reports also significance levels before (t-tests) and after (ANCOVAs) covariation. Lexical skills
SLI
TLD
Naming Semantic Fluency Speech Rate % Semantic paraphasias
58.89 (3.41) 23.63 (6.61) 88.97(26.11) 1.68 (1.52)
61.47 28.50 108.07 .39
(3.46) (7.96) (23.21) (.75)
t-Value
df
p-Value
Cohen’s d
NWR Covariate (p-value)
2.920 2.667 3.070 4.290
59 62 61 44
.005* .010* .003* .001*
.75 .67 .77 1.08
.009§ .163 .036 .001§
SLI: children with Specific Language Impairment; TLD: children with Typical Language Development. * Group-related differences significant after Bonferroni correction for multiple comparisons (p < .0125). § Group-related differences still significant after covariation for NWR.
A. Marini et al. / Research in Developmental Disabilities 35 (2014) 3534–3542
3539
Table 3 Results of the analysis of grammatical skills in the groups of participants with SLI and typical language development. The table reports also significance level before (t-tests) and after (ANCOVAs) covariation. Grammatical skills MLU Sentence Completion % Paragrammatic errors % Complete Sentences
SLI 4.74 (1.36) 10.28 (2.98) 1.58 (1.83) 49.39 (20.97)
TLD 5.72 (1.27) 12.03 (2.19) .33 (.75) 67.65 (17.51)
t-Value 2.934 2.679 3.562 3.756
df 61 62 61 61
p-Value *
.005 .009* .001* .001*
Cohen’s d
NWR Covariate (p-value)
.74 .67 . 89 .95
.038 .155 .029 .010§
SLI: children with Specific Language Impairment; TLD: children with Typical Language Development. * Group-related differences significant after Bonferroni correction for multiple comparisons (p < .0125). § Group-related differences still significant after covariation for NWR.
of statistical significance was set at p < .0166 (.05/3 dependent variables) after Bonferroni correction for multiple comparisons. Effect sizes were calculated in terms of Cohen’s d (Cohen, 1988). The role played by verbal working memory on linguistic performance was controlled by performing an additional series of One-Way ANCOVAs with the scores at the NWR subtest of the BVL_4-12 as covariate. Table 4 reports the results of these analyses showing the level of group-related significance before (t-tests) and after covariation (ANCOVAs). When verbal working memory was not included in the analysis as covariate, the independent samples t-tests showed that the group of children with SLI had impaired performance on % Global Coherence Errors (t(61) = 3.580; p < .001; d = .90) and % Lexical Informativeness (t(61) = 3.806; p < .001; d = .96). The only measure on which the group of participants with SLI was not impaired was % Thematic Selection (t(62) = 1.808; p = .075; d = .45). Interestingly, when the analyses were replicated by performing a series of One-Way ANCOVAs with the scores at the NWR subtest of the BVL_4-12 the group-related differences survived for both % Global Coherence Errors (p < .001) and % Lexical Informativeness (p < .003) (see Table 4). Also in this case, in order to analyze the potential relationship among the macrolinguistic measures where the participants with SLI were found impaired a series of correlational analyses were conducted using Pearson product-moment correlation coefficient. Interestingly, the correlation between % Global Coherence Errors and % Lexical Informativeness was highly significant (r = .597; p < 001). 3.4. Discussion This study investigated linguistic and narrative abilities in children with Specific Language Impairment compared to a group of children with typical language development. Furthermore, the effect potentially exerted by verbal working memory on their linguistic production skills was examined by exploring the role of vWM on lexical, grammatical and narrative processing, respectively. The results showed that the participants with SLI had difficulties at all three levels of linguistic processing and that vWM apparently affected their linguistic skills in different ways. This effect was marginal on lexical processing, significant on grammatical structuring, and null on narrative construction. Before controlling for the effects of vWM on linguistic performance, the children with SLI showed generalized linguistic difficulties with the only, notable, exception of % Thematic Selection. The narrative speech samples produced by the participants with SLI were characterized by a high incidence of global coherence errors and a significantly reduced amount of informative words (i.e., % Lexical Informativeness). The production of global coherence errors might be interpreted as a consequence of their lexical difficulties. Indeed, the macrolinguistic analysis showed that they produced significantly more errors of global coherence than the group of healthy participants. A qualitative inspection of these errors revealed that the children with SLI did not produce tangential or conceptually incongruous utterances. Rather, they produced propositional repetitions and filler utterances. Together with the adequate score of thematic selection, these observations suggest that they had understood the gist of the story, could select all expected concepts, being even able to refrain from introducing irrelevant (i.e., tangential or conceptually inappropriate) propositions. Our data suggest that these children did not have problems in identifying the target concepts and that the macrolinguistic impairment might be an epiphenomenon of a microlinguistic one. Indeed, the relevant difficulties experienced by the participants with SLI on measures of lexical production (Naming, Semantic Fluency, Speech Rate and % of Semantic Paraphasias) suggest an impairment in the processes of both lexical selection and access. Furthermore, their generalized grammatical difficulties (measured in terms of MLU, % of Paragrammatic Errors, % Complete Sentences, and Sentence Completion) highlight the presence of additional problems also Table 4 Results of the analysis of narrative skills in the groups of participants with SLI and typical language development. The table reports also significance level before (t-tests) and after (ANCOVAs) covariation. Narrative skills
SLI
TLD
% Global Coherence Errors % Lexical Informativeness % Thematic Selection
14.10 (11.45) 75.23 (11.53) 31.12 (11.35)
5.44 (7.36) 86.03 (10.98) 36.20 (11.12)
t-Value
df
p-Value
3.580 3.806 1.808
61 61 62
.001* .001* .075
SLI: children with Specific Language Impairment; TLD: children with Typical Language Development. * Group-related differences significant after Bonferroni correction for multiple comparisons (p < .0166). § Group-related differences still significant after covariation for NWR.
Cohen’s d .90 .96 .45
NWR Covariate (p-value) .001§ .003§ .733
3540
A. Marini et al. / Research in Developmental Disabilities 35 (2014) 3534–3542
in the ability to extract from the activated lexical items all the necessary morphological and morphosyntactic information that is required to generate adequate syntactic representations of the intended message. Noteworthy, 20% of all words were lexical repetitions and lexical fillers, mainly personal comments about the story or repetitions of previously introduced ideas. Most of these lexical repetitions and fillers clustered in repetitive and filler utterances that, in turn, increased the amount of global coherence errors. It seems, then, that a difficulty in accessing specific lexical entries (a microlinguistic process) may have affected the ability to keep narrative coherence (a macrolinguistic process). Interestingly, also the reduction in lexical informativeness may be linked to a disturbance in bottom-up processes originating in the process of lexical retrieval rather than from top-down processes concerning the conceptualization of the story. Importantly, the significant negative correlation between the percentage of lexical informativeness and the percentage of global coherence errors suggests that the speech samples were progressively less informative with increasing production of propositional repetitions and filler utterances that, as we have seen earlier, are probably linked to the reduced levels of lexical retrieval. Overall, then, our general impression is that the production of repetitive and filler utterances, rather than a sign of a problem in the macrolinguistic organization of their narrative discourse, might reflect a strategy to cope with a lexical problem: while in the phase of message generation, their word selection difficulties may have triggered the production of filler or repeated utterances that allowed them to skip prolonged pauses perceived as more annoying than simply ‘‘filling the gap’’. This might have also led to the production of fewer complete sentences as shown by the significant correlation between the % of Errors of Global Coherence and the % of Complete Sentences. This represents an interesting example of how a multilevel procedure for discourse analysis might shed light not only on selective (i.e., lexical, grammatical, narrative) aspects of message production, but also on the complex interactions occurring among micro- and macrolinguistic processes. Once verbal working memory was controlled for, some of these group-related differences disappeared. At the lexical level, the covariation showed that vWM affected minimally this level of linguistic processing in the participants with SLI (i.e., Speech Rate on the narrative description task and performance on the Semantic Fluency task). Semantic Fluency is usually assumed to tap the extension of a child’s mental lexicon through a complex procedure involving the ability to monitor the execution of the test, inhibit the production of words that can be either semantically unrelated with the target category or mere repetitions of previously produced items. Therefore, this test cannot be considered as a purely linguistic task. Its execution is likely affected not only by the extension of the child’s lexical repertoire, but also by other factors such as executive skills and verbal working memory. Indeed, our results confirm the important role of vWM in this task. Likely, vWM is critical to keep track of the words that have already been produced so to avoid useless repetitions. Noteworthy, the influence of verbal working memory on this task may not be limited to the procedures that allow its execution. Indeed, verbal working memory is thought to significantly contribute to the process of lexical learning, especially in younger children with SLI (e.g., Ellis & Sinclair, 1996; Martin & Gupta, 2004; Montgomery, Magimairaj, & Finney, 2010). Therefore, the influence of such a critical cognitive function on this task might have been both direct and indirect. It may have directly affected the execution of the task (via the monitoring component of working memory). However, it might have also indirectly altered the performance as these children have smaller lexical repertoires because of their working memory difficulties. Similar considerations may apply also to Speech Rate. If vWM’s influence on lexical processing was minimal, its influence on grammatical processing was highly significant. Indeed, vWM affected all of the grammatical measures considered in this study with the only exception of the % of Complete Sentences produced on the narrative production task. This might be explained in two ways: one possibility is that the vWM difficulties of the participants with SLI affected their ability to correctly process the grammatical information required to complete and understand sentences in traditional tasks as well as their ability to produce morphologically sound words on a narrative description task, but not their ability to produce wellformed sentences on the same narrative production task; a second possibility is that this apparently awkward result might depend on the potential interconnections between micro- and macrolinguistic aspects of linguistic processing. As the first possibility seems unlikely, we explored the potential relationship between the % of Complete Sentences and the % of Global Coherence Errors and found that this correlation was significant. This suggests that vWM plays a major role in grammatical processing in children with SLI and highlights the importance of including a narrative production task in the clinical assessment of linguistic symptoms in these individuals. A final consideration regards the null impact of vWM on measures of macrolinguistic organization. As we have already noted, it is likely that these errors were the epiphenomenon of a lexical impairment rather than reflecting a macrolinguistic impairment. In conclusion, these findings suggest that SLI is a complex developmental disorder whose symptoms are linked to the interaction among several factors. While supporting the claims about the important role played by vWM in determining at least specific aspects of the linguistic profile of children with SLI (e.g., Archibald & Gathercole, 2007; Bishop, 2006), the results from the current study suggest that vWM might not be the core symptom of SLI (Lum, Conti-Ramsden, Page, & Ullman, 2011). These results increase our understanding of the effects exerted by impairments in verbal working memory on linguistic performance. This is coherent with the observation that, even if frequent in children with SLI (e.g., Duinmeijer et al., 2012; Montgomery, 2006; Montgomery et al., 2010), limitations in verbal working memory are not always present (Archibald & Joanisse, 2009). Furthermore, the present study also showed that narrative discourse in children with SLI may present signs of both micro- and macrolinguistic impairment. However, our data suggest that the macrolinguistic impairment might be just an epiphenomenon of the underlying deficit in lexical processing that, from a structural point of view, leads to the frequent interruption of the utterances and determines a reduction of the levels of sentence completeness, while from a functional point of view determines the selection of strategies to cope with the impairment and leads to the production of lexical fillers and repetitions. Once clustered in propositional repetitions and filler utterances, these
A. Marini et al. / Research in Developmental Disabilities 35 (2014) 3534–3542
3541
uninformative words lower the levels of global coherence across the narratives. And this will eventually determine a significant reduction of the levels of informativeness. Interestingly, this complex interplay between micro- and macrolinguistic processes can be found, even if with different characteristics, also in patients with other aethiologies. For example, in a group of non-aphasic persons with traumatic brain injury the production of errors of cohesion (a macrolinguistic measure) determined a reduction in speech rate (a microlinguistic measure) suggesting that the latter may have been determined by the frequent interruptions in the flow of thoughts rather than more specific linguistic deficits (Marini, Galetto, et al., 2011). Similarly, in a group of schizophrenic participants the massive production of global coherence errors triggered the production of microlinguistic errors in terms of semantic paraphasias (Marini, Spoletini, et al., 2008). A final remark regards the limitations of this study. Indeed, due to the small number of participants, the conclusions outlined above are based on the results of bivariate correlations. However, the relationships among these measures may not necessarily be linear. Therefore, further studies are needed with larger groups of participants in order to perform more sophisticated statistical analyses, such as factorial designs or regression models that will allow researchers to deepen the understanding of the complex interaction among micro- and macrolinguistic processes in children with SLI. Acknowledgements This work was supported by a grant from IRCCS ‘‘E. Medea’’ (Ricerca Corrente 2013, Italian Ministry of Health) to A. M. The authors wish to thank Daniela Milan, Paola Casonato, Nicoletta Drigo, Marcella Zamparini and all those speech therapists who actively helped us in the selection and recruitment of the children who participated to this study. References Archibald, L. M. D., & Gathercole, S. E. (2007). Nonword repetition in specific language impairment: More than a phonological short-term memory deficit. Psychonomic Bullettin, 14(5), 919–924. Archibald, L. M. D., & Joanisse, M. (2009). On the sensitivity and specificity of nonword repetition and sentence recall to language and memory impairments in children. Journal of Speech, Language and Hearing Research, 52, 899–914. Bishop, D. V. M. (2006). What causes specific language impairment in children? Current Direction in Psychological Science, 15(5), 217–221. Botting, N. (2002). Narrative as a tool for the assessment of linguistic and pragmatic impairments. Child Language Teaching and Therapy, 18, 1–12. Botting, N., & Conti-Ramsden, G. (2001). Non-word repetition and language development in children with specific language impairment. International Journal of Language and Communication Disorders, 36, 421–432. Carletta, J. (1996). Assessing agreement on classification tasks: The kappa statistic. Computational Linguistics, 22, 249–254. Chiat, S., & Roy, P. (2007). The preschool repetition test: An evaluation of performance in typically developing and clinically referred children. Journal of Speech and Hearing Research, 50, 429–443. Cohen, J. (1988). Statistical power analysis for the behavioral sciences (2nd ed.). NJ: Lawrence Erlbaum. Conti-Ramsden, G., & Botting, N. (2001). Psycholinguistic markers for specific language impairment (SLI). Journal of Child Psychology and Psychiatry, 42, 741–748. Cowan, N., Nugent, L., Elliott, E., Ponomarev, I., & Saults, S. (2005). The role of attention in the development of short-term memory: Age differences in the verbal span of apprehension. Child Development, 70, 1082–1097. Duinmeijer, I., de Jong, J., & Scheper, A. (2012). Narrative abilities, memory and attention in children with a specific language impairment. International Journal of Language and Communication Disorders, 47(5), 542–555. Ellis, N. C., & Sinclair, S. G. (1996). Working memory in the acquisition of vocabulary and syntax: Putting language in good order. The Quarterly Journal of Experimental Psychology, 49, 234–250. Ellis Weismer, S., Evans, J., & Hesketh, L. (1999). An examination of verbal working memory capacity in children with specific language impairment. Journal of Speech, Language and Hearing Research, 42, 1249–1260. Finneran, D. A., Francis, A. L., & Leonard, L. B. (2009). Sustained attention in children with specific language impairment. Journal of Speech, Language and Hearing Research, 52, 915–929. Gathercole, S., & Baddeley, A. (1990). Phonological memory deficits in language disordered children: Is there a causal connection? Journal of Memory and Language, 29, 336–360. Graf Estes, K., Evans, J. L., & Else-Quest, N. M. (2007). Differences in the nonword repetition performance of children with and without specific language impairment: A meta-analysis. Journal of Speech Language and Hearing Research, 50, 177–195. Henry, L. A., Messer, D. J., & Nash, G. (2011). Executive functioning in children with specific language impairment. Journal of Child Psychology and Psychiatry, 53(1), 37–45. Laws, G., & Bishop, D. V. M. (2003). A comparison of language abilities in adolescents with Down Syndrome and children with Specific Language Impairment. Journal of Speech, Language and Hearing Research, 46, 1324–1339. Leonard, L. B. (1998). Children with specific language impairment. Cambridge, MA: MIT Press. Lum, J. A., Conti-Ramsden, G., Page, G., & Ullman, D. M. T. (2011). Working, declarative and procedural memory in specific language impairment. Cortex, 48(9), 1138–1154. Marini, A. (2012). Characteristics of narrative discourse processing after damage to the right hemisphere. Seminars in Speech and Language, 33(1), 68–78. Marini, A., Andreetta, S., del Tin, S., & Carlomagno, S. (2011). A multi-level approach to the analysis of narrative language in aphasia. Aphasiology, 25(11), 1372–1392. Marini, A., Galetto, V., Zampieri, E., Vorano, L., Zettin, M., & Carlomagno, S. (2011). Narrative language in traumatic brain injury. Neuropsychologia, 49, 2904–2910. Marini, A., Marotta, L., Bulgheroni, S., & Fabbro, F. (2014). Batteria per la Valutazione del Linguaggio in Bambini dai 4 ai 12 anni. Firenze, Italy: Giunti OS. Marini, A., Spoletini, I., Rubino, I. A., Ciuffa, M., Bria, P., Martinotti, G., et al. (2008). The language of schizophrenia: An analysis of micro and macro-linguistic abilities and their neuropsychological correlates. Schizophrenia Research, 105, 144–155. Marini, A., Tavano, A., & Fabbro, F. (2008). Assessment of linguistic abilities in Italian children with specific language impairment. Neuropsychologia, 46, 2816–2823. Marini, A., & Urgesi, C. (2012). Please get to the point! A cortical correlate of linguistic informativeness Journal of Cognitive Neuroscience, 24(11), 2211–2222. Martin, N., & Gupta, P. (2004). Exploring the relationship between word processing and verbal short-term memory: Evidence form associations and dissociations. Cognitive Neuropsychology, 21, 213–228. Montgomery, J. W. (2006). Real-time language processing in school-age children with specific language impairment. International Journal of Language and Communication Disorders, 41, 275–291. Montgomery, J. W., Magimairaj, B. M., & Finney, M. C. (2010). Working memory and specific language impairment: An update on the relation and perspectives on assessment and treatment. American Journal of Speech-Language Pathology, 19, 78–94. Newbury, D. F., Fisher, S. E., & Monaco, A. P. (2010). Recent advances in the genetics of language impairment. Genome Medicine, 2, 6.
3542
A. Marini et al. / Research in Developmental Disabilities 35 (2014) 3534–3542
Paradis, M. (1987). The assessment of bilingual aphasia. Hillsdale, NJ: Erlbaum. Raven, J. C. (1938). Progressive Matrices: A Perceptual Test of Intelligence. London: HK Lewis. Vicari, S. (2007). Prove di Memoria e Apprendimento per l’Eta` Evolutiva (PROMEA). Firenze, Italy: Giunti OS. Vugs, B., Hendriks, M., Cuperus, J., & Verhoeven, L. (2014). Working memory performance and executive function behaviors in young children with SLI. Research in Developmental Disabilities, 35, 62–74. Wechsler, D. (1993). WISC—R. Scala d’intelligenza Wechsler per bambini riveduta. Firenze, Italy: Giunti OS.
