Use of narratives to assess language disorders in an inpatient pediatric psychiatric population

Abstract

A large proportion of child psychiatry patients have undiagnosed language disorders. Adequately developed language is critical for psychotherapy and cognitive-behavioral therapies. This study investigated (1) whether assessment of oral narratives would identify language impairments in this population undetected by assessment of only core language abilities, and (2) the extent to which measures of cognition, working memory, emotional distress, and social function differentially predict core language and narrative development. Results showed that (1) more than twice as many children were identified with language impairment when both narrative and core language assessment were used, and (2) core language comprehension and complex verbal working memory were the strongest predictors of narrative production, while core language comprehension, a less complex working-memory task, and social skills best predicted narrative comprehension. Emotional distress did not predict either. The results emphasize the importance of evaluating child psychiatry patients’ language, using both core language and narrative measures.

Keywords

Childhood psychiatric disorders oral narratives language assessment narrative assessment working memory

Introduction

Language ability, particularly narrative skill, is important for both psychiatric diagnosis and psychotherapeutic intervention. Storytelling knowledge underlies how we understand ourselves within the social-emotional context and temporal-causal framework (Losh & Capps, 2003; Wyer, 2004) and even how we organize our cognition and memory (Schank, 1995). We make sense of our lives and relationships with others by knowing and telling our stories. Bruner (2004) described narratives as self-telling for self-making, an act integral to the creation of self. Psychiatric diagnosis requires well-developed language competence, at the textual level, as well as the basic lexical and grammatical levels, as individuals are interviewed and tell their own life stories. In the process of psychotherapy, a person puts together memories of his or her life into a reorganized, more coherent whole and may use narratives to externalize problems and separate them from the self (Chaika, 2000). Therefore, language ability, particularly narrative skill, is highly relevant for the child psychiatry population. Because narratives are a means for both psychotherapeutic evaluation (Emde, Wolf, & Oppenheim, 2003) and intervention (Angus & McLeod, 2004; Chaika, 2000; Gardner, 1993), it is important that disorders in narrative generation and understanding are identified, taken into account, and remediated (Losh & Capps, 2003).

The comorbidity of language disorders and psychiatric disorders is well established. Researchers recognize the high incidence of language impairment in children with psychiatric disorders (Beitchman, Cohen, Konstantaras, & Tannock, 1996; Cohen, Barwick, Horodezky, Vallence, & Im, 1998; Toppelberg & Shapiro, 2000) and the risk of psychiatric disorders for children with communication impairments (Conti-Ramsden & Botting, 2008). Between 60% and 70% of children referred to both inpatient and outpatient clinics for psychiatric services have been diagnosed with speech and/or language impairment (Giddan, Milling, & Campbell, 1996), with 40% of these disorders previously unsuspected (Cohen et al., 1998). Longitudinal investigations have also found an increase in psychiatric disorders in later life for children identified earlier with speech or language impairment (e.g., Beitchman et al., 2001; Clegg, Hollis, Mawhood, & Rutter, 2004). In a metareview of the previous 10 years of research related to comorbid childhood communication and psychiatric disorders, Toppelberg and Shapiro (2000) concluded that early detection of language processing problems may be crucial in the prevention of psychiatric disorders. Early identification and remediation of language impairment may also ameliorate later delinquent behavior (Brownlie et al., 2004). Since verbal communication is foundational to psychotherapeutic and cognitive-behavioral therapies, in order to support the therapeutic process and mitigate negative long-term developmental and psychosocial outcomes, it is imperative that language disorders be identified and remediated as early as possible (Cohen et al., 1998).

Beitchman and colleagues (1996) proposed that communication difficulties experienced by child psychiatry patients may be related to underlying specific language impairments that are masked by the children’s social-emotional and behavioral disorders. A factor contributing to masking language deficits, resulting in under-referral rates, may be that at least some of the children have core language skills (phonology, semantics, and sentence-level morphosyntax) within the average range, while their language ability is less well developed at the more complex discourse level. These higher-level skills can be more difficult to detect and may not appear compromised during a typical psychiatry interview.

Investigations of the language skills of children in the psychiatry population to date have evaluated representational aspects of language, including the comprehension and production of syntax and semantics, but rarely pragmatic or discourse features (Beitchman et al., 1996; Cohen et al., 1998). The production of connected discourse, usually assessed with narrative tasks, has more recently received attention for clinical evaluation of language impairment (Justice, Kaderavek, Ukrainetz, Eisenberg, & Gillam, 2006) and has been shown to be predictive of overall outcomes in a language-impaired population (Botting, Faragher, Smikin, Knox, & Conti-Ramsden, 2001). One investigation of the narratives told by children with either language or psychiatric disorders, or both (Vallance, Im, & Cohen, 1999), found deficits in language and information structure as well as information flow and cohesion for children who had both language and psychiatric disorders.

Storytelling is a dynamic task requiring multiple, simultaneous levels of processing and integration of linguistic, cognitive, experiential, contextual, and social-emotional information (McCabe & Peterson, 1991). Narrators must first understand experience as temporally and causally related events that occur sequentially in a spatial and socio-emotional context. They must recognize goals and obstacles to accomplishing those goals. They must then have the language in which to encode these contextually embedded events, which includes vocabulary, verb forms (tense and aspect), conjunctions, and adverbials, as well as the coordination of these forms in hierarchically ordered and related clauses. A narration must make the narrator’s perspective and response and the context of the story (who, what, when, where) as well as the event sequence understandable to the listener. Thus, storytellers must do considerable advance hierarchical planning, holding the story components in memory and manipulating them while the individual sentences and total narrative are planned, encoded, and narrated. While doing this, they must also monitor the listener for attention and understanding and make adjustments during the narration to accommodate these. The complexity of coordinating these multiple factors leads to a protracted span of narrative development through childhood. The chaotic nature of family life for individuals with psychosocial disorders makes it even more likely that children with such disturbances find it difficult to make sense of their life experiences and are at risk for specific deficits in narrative ability. Narrative tasks may thus identify more high-level language-use impairments in children with psychosocial disorders than are identified by measures of core language skills alone.

The current investigation considered whether oral narratives provide critical diagnostic information for child psychiatry patients, contributing to the identification of children with language impairments overlooked by traditional assessment of isolated core language skills (semantics, syntactic form, and phonology). To better understand the relationship between psychosocial and language disorders, we also considered the relative contribution of general cognitive factors, social skills, working memory, and social-emotional states to the children’s core language and narrative ability. Since narrative ability has been associated with social-emotional, cognitive, and linguistic knowledge, understanding the relationship of these factors to narrative production will be central to planning intervention for children with comorbid psychiatric and language disturbances.

We hypothesized that evaluation of children’s comprehension and production of oral narratives would identify language impairments in the child psychiatry population undetected by assessment of core language skills alone. To further evaluate how narratives could be used diagnostically in this population, we also tested whether the children’s narratives could be differentiated from those of same-age children outside the child psychiatry population. Finally, we assessed the relationships between the children’s narratives and each of the following clinical measures: cognition, working memory, emotional distress, and social function.

Method

Data analyzed for this investigation were part of the clinical assessment of children admitted for a four-week diagnostic period to the Child Psychiatry Inpatient Unit in a children’s hospital. The children displayed a variety of internalizing and externalizing behaviors, as well as developmental and learning disabilities (see Table 1). Multidisciplinary assessments were conducted during this period by psychiatry, psychology, speech-language pathology, occupational therapy, and education professionals. The assessments yielded measurements of the children’s language, cognition, working memory, emotional distress, and social skill, which were analyzed to address the research questions.

Table 1.

Summary of psychiatric history of participant cohort.

Mean age
First psychiatric concerns	3;5
First psychiatric assessment	6;6
Admission and assessment	10;1
Psychiatric parameter	Proportion of children
Family history
Learning disability	.20
Addictions	.40
Mental health disorders	.80
Discharge diagnoses
ADHD	.70
ODD	.30
Anxiety disorder	.50
Mood disorder	.30
Developmental disorder	.40
Attachment disorder	.20
Mean number of mental health diagnoses: 3.6 per child
Medication on discharge
Neuroleptics	.40
SSRI	.30
Anticonvulsant	.10
Stimulant	.30
Other^a	.50

ADHD: attention deficit hyperactivity disorder; ODD: oppositional defiant disorder; SSRI: selective serotonin reuptake inhibitor.

Includes atomoxetine, clonidine, tricyclics, melatonin.

Participants

Over the course of a two-year period, 48 children, 6–12 years old (42 boys and 6 girls, mean age 10;1) were recruited from those referred by the psychiatrist for speech-language pathology assessment. Participants were three 6-year-olds, six 7-year-olds, four 8-year-olds, seven 9-year-olds, five 10-year-olds, thirteen 11-year-olds, and ten 12-year-olds. The mean number of Diagnostic and Statistical Manual–TR (DSM-TR; 2000) psychiatric diagnoses (Table 1) was 3.6 per child. To improve sample homogeneity, the following selection criteria were used:

exposure to English for at least five years;

Perceptual Reasoning Index (PRI) score above 85 on the Wechsler Intelligence Scale for Children–IV (WISC-IV);

responses to pure-tone audiometry 20 dB or less at 500, 1000, 2000, 4000, or 6000 Hz; and

absence of acute psychotic symptoms.

Measures

Core language: Clinical Evaluation of Language Fundamentals–4

The Clinical Evaluation of Language Fundamentals–4 (CELF-4; Semel, Wiig, & Secord, 2003), standardized for ages 5–21 years, is a test of language production and comprehension skills at the word and sentence level. The subscale scores include Receptive (CELF-R) and Expressive Language (CELF-E), combined to yield CELF-O (Overall).

Narrative language: The Test of Narrative Language

The Test of Narrative Language (TNL; Gillam & Pearson, 2004) is a standardized assessment tool (ages 5–2 years) that evaluates children’s auditory comprehension and oral production of stories. There are three oral narrative-production formats: story retell, story formulation from a sequence of pictures, and story creation from a single picture (Alien Story). Narratives are scored on several dimensions: grammatical use, referential and cohesion factors, story grammar elements (e.g., setting, event sequences, resolution of plot), dialogue between characters, and description of temporal and causal relationships. The two subscale scores are Narrative Comprehension (TNL-R) and Oral Narration (TNL-E), yielding a combined score, TNL-O.

Narrative productivity and complexity

The Index of Narrative Microstructure was calculated according to the procedures described by Justice et al., 2006, as a post hoc analysis to determine whether children in the child psychiatry population told shorter or less complex Alien Stories (TNL) than normally developing children. It measures:

productivity: total number of words, number of different words, mean length of utterance in words and morphemes; and

complexity: number of T-units (comparable to main clauses) with at least one dependent clause, use of coordinating conjunctions, and use of subordinating conjunctions.

Cognitive ability: Wechsler Intelligence Scale for Children–IV

The WISC-IV (Wechsler, 2003; standardized for ages 6–16) assesses cognitive ability in four areas. The Verbal Comprehension Index (VCI) measures verbal cognitive ability and the PRI was used as a measure of non-verbal cognitive ability. The WISC-IV was not re-administered if testing had occurred within the previous two years. Canadian normative data were used for the current study.

Working Memory Index

A subscale of the WISC-IV, the Working Memory Index (WMI) involves tasks in which the child is asked to repeat or reorder either numbers (Digit Span) or letters and numbers (Letter Number).

Complex verbal working memory

The Competing Language Processing Task (CLPT; Gaulin & Campbell, 1994; standardized for 6–12 years) is a measure of auditory–verbal working memory, requiring the delayed recall of the last words in a series of sentences with an intervening semantic task (answering true/false questions) related to the same information. It requires the child to hold the words in memory while processing other linguistic information.

Emotional distress

The Beck Youth Inventories (Beck, Beck, Jolly, & Steer, 2005), for children 7–14 years old, are five self-report inventories used assess symptoms of depression, anxiety, anger, disruptive behavior, and self-concept.

Social skills: Adaptive Behavior Assessment Scales–II, Social Index

The Adaptive Behavior Assessment Scales–II (ABAS-II; Harrison & Oakland, 2004) is a checklist completed by the parent or caregiver to assess a child’s (5–21 years) functional adaptation in the environment. The Social Index evaluates leisure and social skills.

Procedures and analysis

Each child’s psychiatric history was taken during the initial interview by the admitting psychiatrist. The clinical/experimental measures were administered and scored according to published standard procedures by a certified psychologist, speech-language pathologist, or occupational therapist, as appropriate for each measure. All measures were scored by the administering professional according to the method specified in the test manual. In order to make comparisons among measures, all test scores were converted to a standard metric with a mean of 100 and a standard deviation of 15. Three scores were calculated for both the CELF-4 and the TNL: Overall (O), Expressive (E), and Receptive (R). For the purpose of the study, a score on any measure lower than one standard deviation below the mean was used to identify a child as being in the clinical range. Analysis included two-tailed t-tests, correlations, and step-wise forward linear regressions to examine the relationships between narratives and other variables.

The three narratives told by each child for the TNL were audio taped on a digital recorder and transcribed into a text file by the speech-language pathologist.¹ A second trained listener blind to the research questions reviewed 20% of the audio recordings and their transcripts for transcription reliability and independently scored the Alien Story narrations, for which the scoring is the most subjective of the three narrative-production tasks. Reliability between scorers, calculated as a proportion of point-by-point comparisons, was .91. This is comparable to the inter-scorer reliability of .90 reported by Gillam and Pearson (2004). The transcripts were then coded independently by the first and second authors according to the Index of Microstructure Guidelines (Justice et al., 2006) and disagreements were resolved. Finally, counts were entered into Justice et al.’s factor score formulas to yield Productivity and Complexity scores for each child.

Results

Both standard and converted test scores are shown in Table 2. As a group, the children scored more than one standard deviation below the mean on only three measures: narrative production (TNL-E; x = 6.60, M = 10 ± 3), the CLPT (x = −1.71, M = 0 ± 1), and social skills (ABAS-II; x = 70.77, M = 100 ± 15). The percentages of children who scored more than one standard deviation below the mean were: TNL-E − 54%, CLPT − 69%, Social Skills - 84%.

Table 2.

Overall group means and standard deviations (SD) for clinical measures.

Measures	Standardized averagetest scores	Experimental group test scores	Experimental group converted scaled scores
CELF-4-R	100 (15)	94 (15)	94 (15)
CELF-4-E	100 (15)	90 (13)	90 (13)
CELF-4-O	100 (15)	91 (13)	91 (13)
TNL-R	10 (3)	9 (3)	93 (15)
TNL-E	10 (3)	7 (3)	83 (13)^b
TNL-O	100 (15)	86 (15)	86 (15)
CLPT	0 (1)	−1.7 (1)	78^b
WISC-IV–Verbal	100 (15)	94 (11)	94 (11)
WISC-IV–Perceptual	100 (15)	97 (14)	97 (14)
WISC-IV–WMI	100 (15)	88 (12)	88 (12)
ABAS–Social Index	100 (15)	70 (10)	70 (10)^b
Beck–Self-concept^a	50 (10)	50 (12)	100
Beck–Depression	50 (10)	53 (10)	101
Beck–Anxiety	50 (10)	51 (12)	100
Beck–Disruptive	50 (10)	52 (8)	102
Beck–Anger	50 (10)	50 (10)	99

CELF-4: Clinical Evaluation of Language Fundamentals-4 (Semel, Wiig, & Secord, 2003), R: Receptive, E: Expressive, O: Overall; TNL: Test of Narrative Language (Gillam & Pearson, 2004), R: Receptive, E: Expressive, O: Overall; CLPT: Competing Language Processing Task (Gaulin & Campbell, 1994); WISC-IV: Wechsler Intelligence Scale for Children–IV (Wechsler, 2003), WMI: Working Memory Index. ABAS: Adaptive Behaviour Assessment Scale–II (Harrison & Oakland, 2004).

Beck Youth Inventories (Beck et al., 2005).

Below-average performance.

Core versus narrative language performance

Overall narrative-language skills in this sample were more impaired than core language skills on a two-tailed t-test (p < .01). This was mainly accounted for by the significantly more impaired oral narration (TNL-E) as compared to core language production (CELF-E). There was no difference between narrative and core language comprehension (TNL-R and CELF-R; see Table 3).

Table 3.

Differences between mean scores on the CELF-4 and TNL.

CELF-4		TNL
	Mean (SD)		Mean (SD)	T (2-tailed)	df	p-value
CELF-O	91.48(13.33)	TNL-O	85.69(14.69)	3.04	37	.004*
CELF-R	94.00(14.65)	TNL-R	93.13(14.90)	0.41	37	.450
CELF-E	90.20(13.50)	TNL-E	83.02(13.44)	3.27	37	.002*

Significant at p < .005.

Twelve of the 48 participants scored in the clinical range on the CELF-O. These 12 also scored more than one standard deviation below the mean on the TNL-E, but only nine of them were in the clinical range on the TNL-O due to higher scores on the TNL-R. The TLN–O identified 13 additional children in the clinical range. In short, the core language measure identified 12 of the participants (.25), as having language impairment production, whereas adding the TNL-O as a measure of language impairment yielded a total of 25 of the 48 children (.52), more than twice as many. When focusing specifically on impairment of language production, 15 children (.31) scored in the clinical range on the CELF-E and another 16 children (.33) met clinical-range criteria on the TNL-E, for a total of 31 children (.64) showing impairment. Again, including the narrative measure identified more than twice as many children with language deficits.

Productivity and complexity of oral narratives

Scores on the Index of Narrative Microstructure, available for 47 subjects, were compared to normative z-scores reported by Justice et al. (2006). Productivity scores were not significantly different from the normative sample (p = .45), but complexity ratings were significantly lower than average in our cohort of psychiatric patients (p < .001). In other words, the psychiatric patients in this study generated narratives of average length (z-score: M = .05), but below average complexity (z-score: M = –.58). Although there was a negative correlation between productivity and complexity, it did not reach statistical significance (p >.05; r = –.30, p = .07).

Correlation of cognitive, social, and emotional measures with language measures

All language measures showed moderate to high correlations with one another (r = .43 to .87, p < .001) with the exception of core language production (CELF-E) and narrative-production skills (TNL-E), which showed a significant but low correlation (r = .35, p < .05) (see Table 4).

Table 4.

Correlations of language, cognition, social skills, and emotional distress.

	Core language measures			Oral narrative measures
	CELF-O	CELF-R	CELF-E	TNL-O	TNL-R	TNL-E
Core language – CELF-4
CELF-Overall
CELF-Receptive	0.86***
CELF-Expressive	0.94***	0.71***
Oral narrative – Test of narrative language
TNL-Overall	0.56***	0.56***	0.52***
TNL-Receptive	0.51***	0.49***	0.52***	0.87***
TNL-Expressive	0.43**	0.46***	0.35*	0.84***	0.46***
Competing language processing task
CLPT	0.40**	0.39**	0.33*	0.50***	0.45***	0.41**
Cognitive abilities – WISC-IV
FS/GAI	0.52***	0.51***	0.52***	0.44**	0.41**	0.33*
Verbal comprehension	0.64***	0.61***	0.62***	0.53***	0.45***	0.45***
Perceptual reasoning	0.39**	0.44**	0.33*	0.36*	0.29*	0.32*
Working memory (WMI)	0.56***	0.48***	0.55***	0.45**	0.41**	0.33*
Social skills – Adaptive behavior assessment scales, second edition
ABAS-2	0.03	0.09	0.13	0.48**	0.47**	0.34*
Emotional measures – Beck Youth Inventories
Self-concept	0.16	0.14	0.16	0.13	0.11	0.10
Depression	−0.29	−0.19	−0.36*	−0.09	−0.15	0.03
Anxiety	−0.22	−0.13	−0.24	0.15	−0.17	−0.08
Anger	−0.22	−0.25	−0.20	0.14	−0.19	−0.01
Disruptive behavior	−0.28	−0.26	−0.28	−0.34*	−0.34*	−0.20

p < .05; ** p < .01; *** p < .001.

CELF-4: Clinical Evaluation of Language Fundamentals (Semel, Wiig, & Secord, 2003); TNL: Test of Narrative Language (Gillam & Pearson, 2004); WISC-IV: Wechsler Intelligence Scale for Children–IV (Wechsler, 2003), FS/GAI: full-scale/general ability index. ABAS: Adaptive Behaviour Assessment Scale–II (Harrison & Oakland, 2004).

Cognitive measures were also significantly correlated with language measures, with perceptual reasoning showing low-to-moderate correlations and verbal comprehension showing high correlations. Relatively higher correlations were shown between verbal comprehension and core language than between verbal comprehension and narrative scores. WISC-IV Perceptual Reasoning showed low correlations with narrative comprehension and production (r = .29, p < .05; r = .32, p < .05) and moderate and low correlations with core language comprehension and production (r = .44, p < .01; r = .33, p < .05). In contrast, WISC-IV Verbal Comprehension was highly correlated with core language comprehension and production (r = .61, p < .001; r = .62, p < .001, respectively) and only moderately correlated with narrative comprehension and production (r = .45, p < .001; r = .45, p < .001).

Both working-memory scores (WMI and CLPT) were significantly correlated with language measures; however, they showed differing patterns with core language and narrative scores. Whereas the WMI showed stronger correlations with core language abilities, the CLPT showed relatively stronger correlations with narrative measures. The WMI showed strong correlations with core language comprehension and production on the CELF-4 (r = .48, p < .001; r = .55, p < .001, respectively) but lower correlations with narrative comprehension and production (r = .41, p < .01; r = .33, p < .05). In contrast, the CLPT verbal working-memory task showed moderate correlation with narrative comprehension (r = .45, p < .001) and narrative production (r = .41, p < .01), but lower correlation with core language comprehension and production (r = .39, p < .01; r = .33, p <.05, respectively). The CLPT and WMI were unrelated (r = –.16, p = .28).

Social skills, as measured by the Social Index of the ABAS-II, showed no relationship with core language or cognitive abilities. On the other hand, social skills correlated positively with narrative language, particularly overall narrative ability (TNL-O; r = .48, p < .01) and narrative comprehension (TNL-R; r = .47, p < .01). Most of the emotional measures did not correlate significantly with language or cognitive measures. The exceptions were significant correlations between Depression and core language production, and Disruptive Behavior and narrative production.

Predicting narrative and core language performance

A series of step-wise forward linear multivariate regressions was conducted to examine the relationships between core language or narrative scores and other clinical variables. The social, emotional, working memory, and cognitive variables were entered into the regression formula to predict core language and narrative production and comprehension performance (see Table 5). The following prediction formulas were generated:

narrative production was predicted by core language comprehension (β = .587, p < .001) and the CLPT (β = .397, p = .003);

narrative comprehension was predicted by core language comprehension (β = .317, p = .069), social skills (β = .421, p = .006), and the WMI (β = .337, p = .046);

core language production was predicted by the VCI of the WISC-IV (β = .447, 579, p = .001), the WMI (β = .508, p = .001), and the Social Index score of the ABAS-II (β = .276, p = .022); and

core language comprehension was predicted by narrative production (β = .471, p = .001), WISC-IV VCI (β = .413, p = .000), Anger (β = –.212, p = .028), and the WMI (β = .230, p = .032).

Table 5.

Step-wise regression predictions.

9 a. Step-wise regression predicting narrative production
Model	Predictor	R ²	R² Change
1	CELF-R	.506	.724
2	CLPT	.663	.814
9 b. Step-wise regression predicting narrative comprehension
Model	Predictor	R ²	R² Change
1	CELF-R	.395	.629
2	Social	.510	.714
3	WMI	.584	.764
9 c. Step-wise regression predicting core language expression
Model	Predictor	R ²	R² Change
1	VCI-WISC-4	.422	.649
2	WMI	.640	.800
3	Social	.709	.842
9 d. Step-wise regression predicting core language comprehension
Model	Predictor	R ²	R² Change
1	TNL-E	.524	.724
2	VCI	.728	.853
3	Anger	.771	.878
4	WMI	.812	.901

CELF-4: Clinical Evaluation of Language Fundamental–4 (Semel, Wiig, & Secord, 2003), O: Overall, R: Receptive, E: Expressive. TNL: Test of Narrative Language (Gillam & Pearson, 2004), O: Overall, R: Receptive, E: Expressive. CLPT: Competing Language Processing Task. WMI: Working Memory Index. VCI: Verbal Comprehension Index.

Discussion

We investigated whether (1) evaluation of children’s oral narratives would identify language impairment not indicated by core language assessment alone, (2) measures of narrative productivity and complexity would differentiate children in the psychiatric population from those outside this population, and (3) measures of cognition, working memory, emotional distress, and social function related differentially to core language and narrative skills in a child psychiatry population.

Core language and oral narrative as indicators of language impairment

For the children in our cohort, the mean scores for less complex core language production and comprehension were in the average range, whereas the mean narrative-production scores on the TNL were in the clinical range. The discrepancy between core language and narrative may be a factor in the high proportion of children in this population who have been identified with speech or language disorders that were previously unsuspected (Cohen et al., 1998). In keeping with previous reports, a one-year retrospective review of records on our own inpatient child psychiatry unit found that 70% of the children identified with language impairment during their admission assessment had either not been previously assessed for speech-language development or had been found to have average core language skills. These children perhaps had acquired foundational language skills sufficient for functional communication and produced verbal output at a rate and complexity not noticeably different from their peers, particularly with the overlay of social or emotional disturbance, yet had impaired discourse skills difficult to detect in the typical psychiatric interview, psychotherapy session, or classroom setting. Interestingly, Cohen and Horodezky (1998) reported that, of the children referred to a psychiatric clinic solely for emotional-behavioral issues, comprehensive assessments revealed that one third had only language impairment without a psychiatric diagnosis.

Our primary hypothesis was strongly supported: more than twice as many children were identified as having language impairment when scores on both the narrative and core language assessments were considered, as opposed to the core language test alone. Because narrative assessment is not currently a routine part of assessment of children admitted to psychiatric units, it is important to consider the validity of equating low performance on a narrative measure with language impairment in this group of children before making a general recommendation. A stringent review of 43 standardized language tests by Spaulding, Plante, and Farinella (2006) concluded that the CELF-4 and TNL were among very few tests that showed a level of sensitivity and specificity sufficient to identify children with language impairment. Because the CELF-4 has a longstanding reputation as a valid and reliable measure of language impairment, the strong relationship between core language and narrative in our sample – both comprehension and production – supports the use of the TNL to identify language deficits.

Productivity and complexity of narratives

One explanation for the higher number of children with low oral narration scores than low core language scores could be that the children in this cohort simply talked less and therefore produced less well-developed narratives. Children in psychiatric populations are frequently described as having a paucity of language. Computation of the Index of Microstructure (Justice et al., 2006) for the children in our study indicated that, while these children’s productivity scores were commensurate with those of Justice et al.’s typically developing children, their linguistic complexity scores were significantly lower; that is, while the amount of output is average for our psychiatric population, the complexity score of their verbal output was significantly lower in comparison to same-aged peers without language or psychiatric problems. These results support our second hypothesis: measures of complexity, but not productivity, differed between psychiatric and non-psychiatric populations

In our cohort there was a suggestion of a trade-off between productivity and complexity scores, with a weak negative correlation between the two variables that did not reach statistical significance. The fact that the cohort had average core language skills but lower complexity scores for their oral narratives suggests that, while the children may have acquired some more complex syntactic forms, the increased processing resources required to produce a coherent narrative decreased their ability to apply this knowledge contextually. Using a different scoring scheme, other researchers have found a similar disparity between allocation of processing resources to either language content (related to productivity) or form (related to complexity) in narratives of children with specific language impairment (Johnston, 2007; Moser, 2003).

Relations of other measures with language measures

Core language and narrative language measures

Since narrative production was the only language area for which the group mean was in the clinical range, we were interested in which skills contributed to this deficit. Regression analysis revealed that the significant predictors of expressive narrative scores were core language comprehension (CELF-R) and the more complex verbal working-memory task (CLPT). The CELF-R was a predictor of both narrative comprehension and production. Because both the TNL-E and CLPT scores were below average in this population, a question is what underlies the impairment in both narrative production and verbal working memory. Interestingly, core language comprehension (rather than core language production) predicted narrative production. In fact, the regression models demonstrated a strong reciprocal relationship between CELF-R and TNL-E. The scaffolding of language comprehension seems to be the more critical factor in developing adequate narrative production, which could be related to underlying cognitive ability or working memory. The regression results are consistent with the prevalent finding that foundational language comprehension is a predictor of language outcomes in later childhood and adulthood (Flax, Realpe-Bonilla, Roesler, Chouchury, & Benasich, 2009).

Cognition and language

Verbal cognitive skills measured by the WISC-IV VCI were significantly correlated with both core language (CELF-R and CELF-E) and narrative skills (TNL-R and TNL-E), but in the regression model, VCI scores only predicted CELF-R and CELF-E performance. The correlation between VCI scores and language tasks was expected; however, the finding that VCI predicts core language, but not narrative ability, was not anticipated. It is possible that core language skills rather than narrative skills are more directly dependent upon the abilities required for the verbal cognitive tasks. One implication is that assessing verbal cognitive abilities alone is insufficient to predict language skills related to higher-level discourse.

Working memory and language

As discussed earlier, narrative performance was related to working-memory abilities. Demands of the narrative task are complex and multilayered, including reliance on core language structures, hierarchical plot development, temporal and causal relationships, story-protagonist motivations, perspective taking, and social comprehension. Results of this investigation suggest an association between the types of auditory working-memory tasks on the WISC-IV WMI and the CLPT and the language tasks on the CELF and the TNL. The strongest correlations were between the CELF and WMI scores and the TNL and CLPT scores. The correlation between the CLPT and the WMI was .166; no relationship was demonstrated.

Similar relationships were also evident in the regression models, in which the WISC-WMI was predictive of CELF-R and CELF-E performance, while the CLPT was a predictor of TNL-E skills. As a measure of verbal working memory, the CLPT is a more dynamic task that requires the recruitment and integration of more linguistic and working-memory resources than simply isolated word storage and sequential memory. Recent evidence, such as the results of the studies by Montgomery and Evans (2009) and Montgomery, Polunenko, and Marinellie (2009) supports this interpretation.

According to Montgomery et al. (2009), the CLPT is a measure of that component of working memory designated as ‘attentional capacity and control’ in Baddely’s (1998) model. Montgomery refers to this central executive part of working memory as ‘attentional resource capacity/allocation’. The Digit Span task is a measure of the storage buffer component of working memory, which temporarily holds incoming information while it is processed cognitively and is necessary for and associated with simple sentence comprehension. Our results are congruent with this dual component interpretation of working memory related to initial storage in the buffer zone and later more complex processing.

The WMI and CLPT task requirements are similar to those in a study by Moser (2003), who used an auditory–verbal storage-only task and an auditory–verbal storage-plus-processing task to assess the ability of 7–11-year-old children with language impairment. She found a significant correlation between narrative-production tasks on the TNL and the auditory storage-plus-processing task but not the storage-only task. The WMI and Moser’s storage-only tasks may require more isolated resources or ‘local memory’, which involves storage and recall of information with manipulation of concrete, unrelated letters and numbers. The CLPT and Moser’s storage-plus-processing tasks require what may be regarded as a more comprehensive or ‘global memory’, which includes syntactic decoding and semantic interpretation while storing and retrieving words from memory.

The CLPT has been criticized as an inadequate measure of complex working memory because processing demands do not change along with storage demands when additional sentences are presented (Montgomery & Evans, 2009). Experimental measures of complex auditory–verbal working memory have improved rapidly in research over the last decade, but clinical protocols are not available. In our study, the CLPT was a better choice than the WMI for measuring complex verbal working memory in relation to narrative tasks. Consistent with previous research (Ellis-Weismer, Evans, & Hesketh, 1999; Montgomery & Evans, 2009; Montgomery et al., 2009; Moser, 2003), the current investigation suggests that complex auditory–verbal working-memory tasks, but not simple ones, are useful in elucidating the relationship between working memory and language production, specifically higher-level discourse such as narrative. It would be informative to test the relationship in the child psychiatry population with a research instrument that overcomes shortcomings of current published tests. Recent research on children’s working memory (Colozzo, 2009) suggests that it is possible to identify strategies children use in working-memory tasks. It would be clinically useful to determine whether such strategies can be taught to children who are having difficulty with these tasks.

In summary, the core language tasks on the CELF, at the word and sentence level, are predicted by the WMI scores related to local memory requirements, while narrative-production skills are better predicted by CLPT scores, which require a broader integration of linguistic information or global memory. In one respect, our results contradicted those by Montgomery et al. (2009) in that the WMI – but not the CLPT – was a predictor of narrative comprehension on the TNL-R. This might be due to the nature of the items on the TNL-R, which more often involve recall of specific content rather than the planning and integration of ideas required in narrative production or narrative comprehension in other contexts.

Social skills and language

In our population there was a significant correlation between narrative and social skills but not between core language and social skills. This is not surprising, because creating and telling a story involve both (1) understanding social relations that motivate the goals and actions of story protagonists and (2) awareness of listener needs that allow the story to be comprehensible and interesting. In contrast, core language includes more concrete constructs of language form and content. The relationship between narrative ability and social skills has been identified previously in the autistic population (Losh & Capps, 2003), which our results suggest may be a relationship that exists more broadly in the child psychiatry population. The study of narrative ability appears to provide a fruitful avenue for exploring the relationship between language and social skills in a variety of clinical populations. Consideration may also be given to whether interventions targeting narratives might also improve social function.

Emotional factors and language processing

Sirianni (2004) hypothesized negative effects of emotional factors on language processing, and Toppelberg et al. (2002) found a strong connection between emotional/behavior problem scores and language skills in one child psychiatry population. This should be especially true for narrative performance, in which plot motivation and recipient design both require understanding and communication of emotions. This expected relationship was not found in our investigation for the most part, with the exception of low negative correlations between Depression and core language production and Disruptive Behavior and narrative production. None of the subscale mean scores for the cohort were in the clinical range, despite severe emotional problems in this inpatient population. Our conclusion is that this measure did not reliably evaluate the emotional state of the children with complex psychiatric disorders in our cohort.

Conclusions and clinical implications

Narrative assessment is an important component of identifying language impairment in the child psychiatry population. More than half the children in our complex population not previously diagnosed with language impairment were identified as having impaired language when higher-level discourse skills, measured by narrative ability, were tested in addition to core language abilities. Because communication skills are central to psychiatric diagnosis and the psychotherapeutic process, it is imperative to carefully consider the need for assessment of more complex contextual language in addition to more foundational syntactic and semantic information.

In this population of children with multiple comorbid psychiatric disorders, performance on the more complex narrative tasks could not be predicted from the assessment of core language production or the VCI of the WISC-IV, but was predicted by core language comprehension in combination with a complex verbal working-memory task. Assessment of working-memory skills related to discourse-level language processing needs to involve tasks that include more integrative global aspects of verbal working memory and gradients of both storage and processing characteristic of the central executive component of working memory rather than simply discrete sequential memory. This is likely a fruitful domain for future research.

Including assessment of oral narrative skills with the psychiatrically referred population is likely to increase the identification of language-processing deficits, which involve the integration of linguistic, social, emotional, and cognitive abilities. As Cohen (2001) points out, it is important not to ascribe communication difficulties to a psychologic process, such as inattention or emotional distress, when they should be attributed to a primary language impairment. Children with multiple psychiatric comorbidities require multiple interventions, and the identification and treatment of language impairment can be expected to contribute to improved therapeutic outcomes.

Footnotes

Acknowledgements

We would like to thank those who have contributed to this project, including our statistician Boris Kuzeljevic, the entire Child Psychiatry Inpatient Team, and, in particular, the children and families who participated.

Funding

This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

Notes

Author biographies

Patsy Pearce is a Speech and Language Pathologist on the Child Psychiatry Unit at BC Children’s Hospital. Her research interests have included neurolingistic disorders in children and linguistic outcomes of high risk infants.

Carolyn Johnson is an Assistant Professor Emerita in the School of Audiology and Speech Sciences at the University of British Columbia. Her research areas are the development of children’s discourse/pragmatic ability and phonological development by bilingual children.

Patricia Manly is a Psychologist formerly on the Child Psychiatry Unit at BC Children’s Hospital and currently the program coordinator of Hamber House Adolescent Day Treatment Program in Vancouver British Columbia. Her research interests include interpersonal attribution and depression, as well as neurobehavioural sequelae of interpersonal trauma.

Jake Locke is a Child and Adolescent Psychiatrist at British Columbia Children’s Hospital and Associate Professor in the Department of Psychiatry at the University of British Columbia. His current research interests include mindfulness based interventions for depressed teens and for parents of children with ADHD.

References

Angus

L. E.

McLeod

(Eds.) (2004). The handbook of narrative and psychotherapy. London, UK: Sage Publications.

Baddely

(1998). Human memory: Theory and practice. Hove, UK: Psychology Press.

Beck

Jolly

Steer

(2005). Beck Youth Inventories-II. San Antonio, TX: Pearson.

Beitchman

Cohen

Konstantareas

Tannock

(Eds.) (1996). Language, learning and behaviour disorders: Developmental, biological and clinical perspectives. Cambridge, NY: Cambridge University Press.

Beitchman

Wilson

Johnson

C. J.

Atkinson

Young

Adlaf

Douglas

(2001). Fourteen-year follow-up of speech/language-impaired and control children: Psychiatric outcome. Journal of the American Academy of Child and Adolescent Psychiatry, 40, 75–82.

Botting

Faragher

Smikin

Knox

Conti-Ramsden

(2001). Predicting pathways of specific language impairment: What differentiates good and poor outcome? Journal of Child Psychology and Psychiatry, 42, 1013–1020.

Brownlie

Beitchman

Escobar

Young

Atkinson

Johnson

C. J.

Douglas

(2004). Early language impairment and young adult delinquent and aggressive behaviour. Journal of Abnormal Child Psychology. 32, 453–467.

Bruner

(2004). The narrative creation of self. In Angus

L. E.

McLeod

(Eds.), The handbook of narrative and psychotherapy. London, UK: Sage Publications.

Chaika

(2000). Linguistics, pragmatics and psychotherapy. London, UK: Whurr Publishers.

10.

Clegg

Hollis

Mawhood

Rutter

(2004). Developmental language disorders—a follow-up in later adult life: Cognitive, language and psychosocial outcomes. Journal of Child Psychology and Psychiatry, 46, 128–149.

11.

Cohen

(2001). Language impairment and psychopathology in infants, children, and adolescents. London, UK: Sage Publications.

12.

Cohen

Barwick

Horodezky

Vallance

(1998). Language, achievement, and cognitive processing in psychiatrically disturbed children with previously identified and unsuspected language impairments. Journal of Child Psychology and Psychiatry, 39, 865–877.

13.

Cohen

Horodezky

(1998). Prevalence of language impairments in psychiatrically referred children at different ages: Preschool to adolescence [Letter to the editor]. Journal of the American Academy of Child and Adolescent Psychiatry, 35, 461–262.

14.

Colozzo

(2009). Children’s use of rehearsal to remember pictures and words: Do self-report, observation, and stimulus effects tell the same story? Unpublished doctoral dissertation, University of British Columbia, Vancouver, Canada.

15.

Conti-Ramsden

Botting

(2008). Emotional health in adolescents with and without a history of specific language impairment (SLI). Journal of Child Psychology and Psychiatry, 49, 516–525.

16.

Diagnostic and Statistical Manual-TR (2000). American Psychiatric Association, Washington, D.C.

17.

Ellis-Weismer

Evans

Hesketh

(1999). An examination of verbal working memory capacity in children with specific language impairment. Journal of Speech. Language and Hearing Research, 42, 1249–1260.

18.

Emde

Wolf

Oppenheim

(Eds.) (2003). Revealing the inner worlds of young children—The MacArthur story stem battery. New York, NY: Oxford University Press.

19.

Flax

Realpe-Bonilla

Roesler

Chouchury

Benasich

(2009). Using early standardized language measures to predict later language and early reading outcomes in children at high risk for language-learning impairments. Journal of Learning Disabilities, 42, 61–75.

20.

Gardner

(1993). Storytelling in psychotherapy with children. London, UK: Jason Aronson.

21.

Gaulin

Campbell

(1994). Procedure for assessing verbal working memory in normal school-age children: Some preliminary data. Perceptual and Motor Skills, 79, 55–64.

22.

Giddan

Milling

Campbell

(1996). Unrecognized language and speech deficits in preadolescent psychiatric inpatients. American Journal of Orthopsychiatry, 66, 85–92.

23.

Gillam

Pearson

(2004). Test of narrative language. Austin, TX: Pro-Ed.

24.

Harrison

Oakland

(2004). Adaptive Behavior Assessment System–II. Amsterdam. NL: Elsevier.

25.

Johnston

(2007). Thinking about child language. Greenville, NC: Thinking Publications.

26.

Justice

Kaderavek

Ukrainetz

Eisenberg

Gillam

(2006). The Index of Narrative Microstructure: A clinical tool for analyzing school-age children’s narrative performances. American Journal of Speech-Language Pathology, 15, 177–191.

27.

Losh

Capps

(2003). Narrative ability in high-functioning children with autism or Asperger’s syndrome. Journal of Autism and Developmental Disorders, 33, 239–251.

28.

Mayer

(1969). Frog where are you? New York, NY: Penguin.

29.

McCabe

Peterson

(Eds.) (1991). Developing narrative structure. Hillsdale, NJ: Lawrence Erlbaum Associates.

30.

Montgomery

Evans

(2009). Complex sentence comprehension and working memory in children with specific language impairment. Journal of Speech, Language, and Hearing Research, 52, 269–288.

31.

Montgomery

Polunenko

Marinellie

(2009). Role of working memory in children’s understanding spoken narrative: A preliminary investigation. Applied Psycholinguistics, 30, 485–509.

32.

Moser

(2003). Beyond storage: Working memory and specific language impairment. Unpublished master’s thesis, University of British Columbia, Vancouver, Canada.

33.

Schank

(1995). Tell me a story: Narrative and intelligence. Evanston, IL: Northwestern University Press.

34.

Semel

Wiig

Secord

(2003). Clinical Evaluation of Language Fundamentals-4. Toronto, Canada: Harcourt Canada.

35.

Sirianni

(2004). Psychological stress and language processing in school-aged children. Journal of Speech-Language Pathology and Audiology, 28, 112–121.

36.

Spaulding

Plante

Farinella

(2006). Eligibility criteria for language impairment: Is the low end of normal always appropriate? Speech, Language, and Hearing Services in Schools, 37, 61–72.

37.

Toppelberg

C. O.

Medrano

Peña Morgens

Nieto-Castañon

(2002). Bilingual children referred for psychiatric services: Associations of language disorders, language skills, and psychopathology. Journal of the American Academy of Child & Adolescent Psychiatry, 41(6), 712–722.

38.

Toppelberg

Shapiro

(2000). Language disorders: A 10-year research update review. Journal of the American Academy of Child and Adolescent Psychiatry, 39, 143–152.

39.

Vallance

Cohen

(1999). Discourse deficits associated with psychiatric disorders and with language impairments in children. Journal of Child Psychology and Psychiatry, 40, 693–704.

40.

Wechsler

(2003). Wechsler Intelligence Scale for Children-IV. New York, NY: Psychological Corp.

41.

Wyer

(2004). Social comprehension and judgment: The role of situation models, narratives, and implicit theories. Mahwah, NJ: Lawrence Erlbaum Associates.