Differences in narrative discourse ability in traumatic brain injury and a neurotypical comparison group: Influence of age, education, and sex

Abstract

Objective:

To examine the influence of traumatic brain injury (TBI) on narrative discourse comprehension and production of individuals with TBI compared to a non-brain-injured (NBI) group. Further, this work investigated the potential influence of socio-demographic factors (i.e., age, education, sex) on narrative discourse performance.

Design:

Case-control design

Setting:

General community

Participants:

A total of 44 adults were grouped into two age- and education-matched cohorts. The TBI group comprised 21 participants with mild-moderate cognitive impairment, and the NBI comprised 23 participants.

Interventions:

Not applicable

Main Outcome Measures:

Narrative discourse outcomes were measured by two comprehension measures (Discourse Comprehension Test, picture story comprehension task) and two production measures (story grammar, story completeness). Four univariate analyses of covariance (ANCOVA) were performed with age, education, and sex as covariates. The ANCOVAs were performed using parameter estimates with robust standard errors, and boxplots were constructed to understand differences between cohorts for each discourse measure.

Results:

No sociodemographic variables (i.e., the covariates) were significant for any of the four discourse measures. The NBI group outperformed the TBI group on the DCT (p = .003, 95% CI [1.46, 6.82], partial η² = .20), PSC (p = .005, 95% CI [.61, 3.26], partial η² = .18), and story completeness (p = .005, 95% CI [.43, 2.26], partial η² = .21). Effect sizes were large across these three discourse measures. No significant difference between groups was found for story grammar (p = .440, 95% CI [−.11, .25], partial η² = .02).

Conclusions:

TBI disrupts both narrative discourse comprehension and production ability. Findings underscore the need to evaluate both narrative comprehension and production performance following TBI. Age, education, and sex did not influence discourse outcomes.

Discourse Functioning following Traumatic Brain Injury

Impaired discourse production is a hallmark of traumatic brain injury (TBI), (1, 2) impacting functioning across life domains.(3, 4) Breakdowns in narrative discourse are common (5) and have been documented across a range of measures. At the global (i.e., macrolinguistic) level, individuals with TBI often have difficulty organizing stories (6, 7) and managing story content.(8–10)

Narrative comprehension in TBI has received less attention. The few studies that have examined narrative comprehension suggest that TBI can disrupt understanding of spoken stories, particularly for inferences.(11, 12) Investigations of reading ability provide further evidence of discourse comprehension deficits following TBI.(13, 14) Individuals with TBI struggle with understanding paraphrases of written expository text,(14) and systematic review identified deficits in reading comprehension and speed, suggesting protracted comprehension processes after TBI.(13)

Most studies investigating narrative ability following TBI have focused on storytelling without examining story comprehension. A recent TBI study identified moderate-to-large correlations between narrative discourse comprehension and production, suggesting that challenges in understanding stories co-occur with storytelling problems.(15) This finding has paralleled findings for studies of other discourse genres in TBI. Individuals with TBI who have difficulties with producing expository discourse are likely to have problems understanding expository text.(14) However, in a large-scale study, the TBI group performed more poorly than neurotypical peers on narrative production but comparably on narrative comprehension measures.(1) Given the limited research on discourse comprehension and TBI and equivocal findings, further inquiry into narrative comprehension ability in TBI is warranted.

Factors Potentially Influencing Discourse Performance

The influence of sociodemographic factors on cognitive-linguistic functioning in TBI has not been well-studied. Sex, race, and ethnicity are frequently underreported, which pose threats to external validity and hinder a more holistic understanding of communication outcomes.(16) In light of this, sociodemographic factors that have been found to impact cognitive-linguistic functioning in neurotypical populations merit further examination, specifically age, education, and sex.

Multiple aspects of discourse are noted to change with age in neurotypical adults (17–20) (e.g., coherence,(21) episodic information,(22) informativeness.(23)) In a study of conversational ability, although younger adults performed better than older adults overall on the evaluated dimensions (e.g., language, pragmatics, comprehension), older adults scored higher on two variables – initiating conversation and word-finding.(24) While many aspects of discourse may diminish with age, some aspects seem to improve.

Education and socioeconomic status (SES) can also influence discourse performance. A study examining SES and storytelling in native English speakers with TBI found that unskilled workers had poorer cohesive adequacy than skilled or professional workers.(25) Race/ethnicity were not reported. Narratives by adults with high school education are more informative and coherent than those by adults with less education.(26) Even a few extra years of education can shape language functioning. Adults with illiteracy without formal education score lower on oral language tasks, including narrative discourse, compared to those with low education (i.e., one to four years).(27) These studies suggest that even small increases in education enhance cognitive-linguistic functioning and higher education levels may contribute to resilience of discourse ability following TBI.

Although there is little information on sex and discourse ability in TBI, sex has emerged as a potential factor in neurotypical discourse studies. Findings suggest that women tend to tell longer, more elaborative stories, whereas men tend to tell shorter, more factually oriented stories.(28, 29) These differences have been observed across narrative contexts (e.g., story retellings, recounts, autobiographical narratives), suggesting that sex can influence narrative ability. Caution is warranted in generalizing these findings as samples sizes were relatively small. Replication with larger samples is necessary to validate these preliminary findings.

Given that sociodemographic factors can affect discourse performance in neurotypical individuals, examination of these factors may provide further insights into understanding discourse functioning in TBI. Hence, the aims of this study were to 1) examine differences in narrative discourse comprehension and production in TBI compared to a non-brain-injured (NBI) comparison group, and 2) assess the potential influence of age, education, and sex on narrative discourse ability. We hypothesized that 1) the TBI group would have poorer performance across discourse measures and 2) age, education, and sex would have a significant influence on discourse outcomes.

Methods

This study was part of a larger research project investigating narrative discourse in TBI, comprising multiple studies of the same TBI group. Although prior studies of the project(15, 30) applied the same discourse measures as this study, they only examined discourse outcomes in the TBI group. Unique to this study are the involvement of a neurotypical comparison group and the questions of interest.

Study design

This study implemented case-control design with two age- and education-matched cohorts: 1) TBI and 2) no history of brain injury (NBI). Data collection occurred over one to three sessions (about 3 hours total) within a two-week period. Tasks were administered by a speech-language pathologist and presented via E-Prime software on a computer. Participants were block randomized to one of two task sequences as a counter-balance measure to control for order effects.

Recruitment

Individuals with TBI were recruited through a university, community hospitals, and a state brain injury organization. NBI participants were recruited through the university, community, and from families and friends of the TBI group. This study received approval from the university Institutional Review Board (IRB) with inclusion criteria as follows: (a) minimum 16 years of age, (b) single closed-head injury as determined by medical record, c) minimum 3-months post-injury, (d) mild-moderate cognitive impairment, e) native English speaker, and e) hearing and vision screening pass. Exclusion criteria entailed neurological or psychiatric illness, motor speech disorder, or aphasia.

Determination of mild-moderate cognitive impairment was based on performance on 1) the Ranchos Los Amigos Level of Cognitive Functioning Scale – Revised (LCFS)(31) and 2) Galveston Orientation and Amnesia Test (GOAT)(32). Participants had a minimum LCFS score of VII (automatic, appropriate, minimal assistance for daily living skills) and minimum GOAT score of 75 to ensure adequate orientation, attention, and behavior and novel learning capability. Glasgow Coma Scale scores (GCS),(33) if available, were reported to estimate TBI severity. See Supplemental Material S1 for screening and injury characteristics.

Discourse Measures

Discourse comprehension

Discourse Comprehension Test (DCT).

The DCT,(34) a standardized assessment, evaluated auditory/verbal story comprehension on five stories, which were simultaneously presented orally and in writing. Thereafter, eight yes/no questions were presented about the story, varying in the quality of requested information (i.e., salience and explicitness). The DCT score was the total number of correct responses (range: 0 to 40).

Picture Story Comprehension (PSC).

A PSC task evaluated understanding of visual/non-verbal stories and adapted picture sequences from the book, Sector 7.(35) The stimuli were developed to achieve equivalence with the story retelling stimuli (described in the Discourse Production section) regarding number of frames, main characters, visual complexity, and events. After stimuli presentation, participants answered questions about the story.

PSC questions, like the DCT questions, varied in salience and explicitness, and consisted of 20 questions (Supplemental Material S2). The PSC score was the number of correct responses (range: 0 to 20).

Discourse Production

For discourse production, participants viewed a wordless 16-frame picture story, Old MacDonald Had An Apartment House.(36) Following stimuli removal, participants retold the story, which was recorded and transcribed. The transcripts were segmented into T-units for subsequent organizational and content analyses.

Story Grammar.

The story grammar measure was the proportion of T-units within episode structure, which assesses narrative organization and has demonstrated sensitivity to TBI.(37, 38) Stories were analyzed for episodes. A complete episode included: 1) an initiating event, 2) the character’s attempt to achieve a goal, and 3) a direct consequence. A partial episode contained two of the three components. The story grammar score was the number of T-units within episodes divided by total T-units (range 0 to 1.00).

Story Completeness.

The story completeness measure assessed essential story content and has demonstrated sensitivity to TBI.(9, 39). Each story was evaluated for the presence of five critical components established in other studies.(9, 39) The story completeness score was the total number of critical components produced (range: 0 to 5).

Reliability.

To check for potential bias and determine reliability of discourse analyses, 18 percent of transcripts (n = 8) and 25 percent (n = 11) were re-analyzed for story grammar and story completeness, respectively. Story grammar intra-rater and inter-rater reliability were 93% and 84%, respectively. Story completeness intra-rater and inter-rater reliability were each 96%.

Covariates

Age, education, and sex were ascertained through medical record review when available and confirmed through self-report. All participants were queried [e.g., “What is your (age/level of education in years/sex)?”].

Data Analysis

Sample Size

The a priori determined sample size was 54 participants to detect a medium effect size between groups with statistical power of .80. Given recruitment challenges (e.g., administrative IRB delays, screening failures), we did not achieve the target sample size. However, smaller studies have found statistical significance for small-to-moderate or greater effect sizes with 20 to 30 participants.(14, 38, 40)

Statistical analyses were performed with Statistical Package for the Social Sciences version 29 software. Means and standard deviations were calculated for all measures except sex. An independent samples t-test was used to compare age and education between groups. Univariate analyses of covariance (ANCOVAs) were performed to identify differences between groups on the discourse variables of interest (DCT, PSC, story grammar, and story completeness) with age, education, and sex as covariates. ANCOVAs were implemented using parameter estimates with robust standard errors to account for non-normality in the data. Boxplots were generated in R (R Core Team, 2020) with RStudio version 4.5.1 using the R package, “ggplot2” version 2.2.0.9000,(41) to understand the score distribution between groups.

Results

Descriptive Statistics

The TBI and NBI groups comprised 21 and 23 participants, respectively. See Table 1 and Supplemental Material S3 and S4 for participant sociodemographic characteristics. Groups did not differ on age (p =.40) or education (p = .09). Typical of the TBI population,(42) there were more males in the TBI group (71%) compared to the NBI group (48%).

Table 1.

Participant Sociodemographic Characteristics

Variable	NBI *		TBI *
	n	%	n	%
Gender
Males	11	48%	15	71%
Females	12	52%	6	29%
Race/Ethnicity
Caucasian	21	91%	16	76%
African-American	2	9%	4	19%
Hispanic	0	0%	1	5%
	M (SD)	Range	M (SD)	Range
Age (Years)	35.45 (15.68)	18–67	39.19 (13.05)	20–68
Education (Years)	13.78 (2.26)	12–18	15.10 (2.77)	12–21

^*NBI = Non-brain-injured comparison group; TBI = Traumatic brain injury group

Means and standard deviations for the discourse measures are presented in Table 2.

Table 2.

Descriptive Statistics and Univariate ANCOVAs of Discourse Performance

	TBI Group		NBI Group		t	Partial η2
Measure	M [95% CI]	SD	M [95% CI]	SD
Discourse Comprehension
DCT*	33.62 [31.76, 34.73]	4.39	37.04 [35.97, 38.80]	1.82	3.13†	.20
PSC*	16.33 [15.25, 17.05]	2.06	17.91 [17.22, 18.94]	1.88	2.96†	.18
Discourse Production
Story Grammar*	.64 [.52, .75]	.32	.70 (.59, .81)	.16	0.79	.02
Story Completeness*	3.33 (2.63, 3.83)	1.74	4.48 (4.00, 5.14)	.73	2.99†	.19

^*DCT = Discourse Comprehension Test (Brookshire & Nicholas, 1993; auditory/verbal story comprehension); PSC = Picture Story Comprehension (visual/non-verbal story comprehension); Story grammar = proportion of T-units in episode structure in story retelling; Story completeness = number of critical components in story retelling

^†p < .01

Analyses of Covariance

TBI and NBI groups differed significantly on three discourse measures after controlling for age, education, and sex. No sociodemographic variables were significant covariates for any discourse measure (p>.10 for all).

ANCOVAs for the discourse measures (Table 2) and corresponding boxplots demonstrated significant differences on both discourse comprehension measures. The NBI group significantly outperformed the TBI group on the DCT (auditory/verbal story comprehension), t(39) = 3.13, p = .003, 95% CI [1.46, 6.82] with partial η² = .20, indicating a large effect. The NBI group was tightly clustered towards the higher end of scores (median: 37) compared with the wider distribution for the TBI group (median: 34; Figure 1). There was little overlap in the interquartile ranges (IQR) between the groups. The TBI group also had two outliers at the lower end.

Figure 1. DCT performance across groups.

Boxplots comparing DCT scores, a narrative comprehension measure, between TBI and NBI groups.

Similarly, the NBI group scored significantly higher than the TBI group on PSC (visual/non-verbal story comprehension), t(39) = 2.96, p = .005, 95% CI [.61, 3.26] with partial η² = .18, indicating a large effect. The NBI group had a narrower range of performance (median:19; Figure 2). Differences are further evidenced by the TBI group’s median score of 17, which corresponded to the cut-off for the first quartile in the NBI group.

Figure 2. PSC performance across groups.

Boxplots comparing PSC scores, a narrative comprehension measure, between TBI and NBI groups.

Significant differences were identified on one of the two discourse production measures. The NBI group scored significantly higher than the TBI group on story completeness, t(39) = 2.99, p = .005, 95% CI [.43, 2.26] with partial η² = .19, representing a large effect. NBI group PSC scores were highly skewed towards the higher end (median: 5, the ceiling score; Figure 3). Although the TBI group had a median of 4, the distribution spanned the entire scoring range.

Figure 3. Story completeness performance across groups.

Boxplots comparing story completeness scores, a narrative production measure, between TBI and NBI groups.

By contrast, NBI and TBI groups performed comparably on story grammar, t(39) = .79, p = .440, 95% CI [−.11, .25] with partial η² = .02, indicating a small effect. The distribution was similar with equivalent medians between groups (Figure 4). However, the IQR was narrower in the NBI group, and there were three outlier scores of 0 in the TBI group.

Figure 4. Story grammar performance across groups.

Boxplots comparing story grammar scores, a narrative production measure, between TBI and NBI groups.

Discussion

This study examined narrative discourse performance of individuals with TBI and compared their comprehension and production to neurotypical peers. Further, the study investigated the potential influence of age, education, and sex on discourse outcomes. Findings generally supported the hypothesis that narrative discourse ability differed between groups as evidenced by the NBI group outperforming the TBI group on three of four discourse measures.

The NBI group scored higher than the TBI group on both story comprehension measures. These results were robust and align with previous studies that have examined discourse comprehension and narratives following TBI.(11, 14, 43) These findings differ from an earlier large-scale study on discourse ability in TBI,(1) which identified differences in storytelling ability but not in story comprehension between TBI and NBI groups. The prior TBI cohort comprised individuals who were more than 35 years post-injury whereas the majority of the current TBI cohort had substantially fewer years post-injury. A possible explanation for the divergent findings may be that individuals who have had TBI for decades have learned to better compensate for their discourse comprehension difficulties than those who have lived with TBI for less time. Additionally, the current results suggest that story comprehension deficits occur across presentation modalities (auditory/verbal, visual/non-verbal), indicating the possibility of a general narrative comprehension problem (e.g., problems understanding the underlying story schema)(44, 45) rather than a language-specific impairment.

Differences were found on one of the two discourse production measures, story completeness. The NBI group produced more critical components than the TBI group. This finding is consistent with those in the TBI literature demonstrating frequent disruption to informational aspects of discourse (e.g., omission of critical information).(46, 47)

Story grammar performance, a measure of narrative organization, was equivalent between groups. This is at odds with findings that have consistently identified story grammar deficits in people with TBI using the same (3-component coding) or similar measures (i.e., 7-component coding).(9, 25, 38) In the largest studies(9, 25) that have implemented the same measure, means for the TBI and NBI groups were comparable to those in this study. However, the standard deviation for the current TBI group was greater than for the previous studies, indicating greater variability. A possible explanation for the discrepant findings is that this study may have been underpowered for story grammar evaluation and that a larger sample may have produced findings more congruent with previous studies. The three outlier scores of 0 (Figure 4) suggest that, in a larger study, there may be a greater representation of scores in the lower range. Alternatively, perhaps chronicity was a factor given half the sample was several years post-injury. Paralleling discourse comprehension in the aforementioned large-scale study of TBI, (1) the individuals who have lived with TBI longer in this study may have developed strategies over time to better organize their stories.

No sociodemographic variables significantly influenced discourse outcomes as measured in this study. Although participants ranged in age, no individuals were older than 70 years. Studies that have identified age-related communication changes have typically included adults in their 80s and older.(22, 48) Had there been more older individuals, an effect of age on discourse performance may have been observed. Lack of a sex effect may have been due to the type of narrative used. Although sex differences have been reported for story retellings, these differences have often emerged in the expression of biographical stories.(28, 29) Perhaps the structured nature of the story retelling task precluded more personal expression where sex differences may more likely manifest. Surprisingly, education did not have an effect on discourse outcomes as there was a range of education levels in both groups, and education level has been associated with cognitive-linguistic performance.(27) An earlier study(25) found that cohesive adequacy was influenced by SES level but not story grammar, suggesting that education may affect narrative ability in TBI in specific aspects. Thus, measure selection may be critical to identifying the influence of factors, like sex and education.

Despite the nonsignificant covariate findings, it is logical that individual factors, including sociodemographics, would shape one’s communication. The study constraints (see Study Limitations) may have precluded significant findings. From a clinical standpoint, understanding how these factors influence discourse could potentially reduce bias in assessment (e.g., typical v. atypical discourse performance for one’s age). Additionally, such insights could enhance personalization of treatment and, ultimately, improve healthcare equity in managing cognitive-communication disorders after TBI.

Study limitations

This study had various limitations. It may have been underpowered for examining story grammar and effects of sociodemographic variables on discourse. In addition, the scope was limited by using global measures of discourse production and examining one type of narrative expression. Findings for story retelling may not generalize to other story types. This study did not examine the effect of the age range, chronicity, injury severity, and other relevant variables on discourse performance in TBI (e.g., race/ethnicity, examiner characteristics, co-morbid mental health diagnoses) or their interactions, which may have provided further insights. For example, understanding the relationship between injury-related characteristics and cognitive-communicative performance may help disentangle puzzling findings, such as that for story grammar and sociodemographics in this study. Lastly, a formal cognitive assessment would have complemented cognitive functioning measures to provide a more complete description of cognitive ability.

Conclusions

This study is one of the few studies in the TBI literature to examine discourse comprehension ability in tandem with discourse production ability, demonstrating that individuals with TBI have difficulty with both story comprehension and aspects of storytelling. The findings underscore the need to evaluate both discourse comprehension and production following TBI. Future research examining sociodemographic factors in discourse should consider measures and discourse elicitation task carefully. Investigation of other factors, such as psychiatric and neurologic comorbidities and chronicity, may also prove fruitful in furthering our understanding of the sources of change and variability in discourse outcomes in TBI.

Abbreviations:

DCT

Discourse Comprehension Test

GCS

Glasgow Coma Scale

GOAT

Galveston Orientation and Amnesia Test

IRB

Institutional Review Board

IQR

inter-quartile range

LCFS

Ranchos Los Amigos Level of Cognitive Functioning Scale – Revised

ANCOVA

analysis of co-variance

NBI

non-brain-injured or no brain injury

PSC

Picture Story Comprehension

SES

socio-economic status

TBI

traumatic brain injury

Data Availability Statement

The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request and a signed shared data agreement.