An Analysis of Right Hemisphere Stroke Discourse in the Modern Cookie Theft Picture

Abstract

Purpose:

Adults with right hemisphere damage demonstrate differences in connected speech compared to controls, but systematic, quantitative methods to capture these differences are lacking. The current study aimed to (a) investigate if measures using the Modern Cookie Theft picture description would identify discourse differences in acute right hemisphere stroke, and (b) examine if discourse differences were associated with documented cognitive impairment.

Method:

Eighty-four participants completed the Modern Cookie Theft picture description within 5 days of right hemisphere stroke. Descriptions were analyzed for multiple microlinguistic characteristics. Medical charts were retrospectively reviewed for documented presence of cognitive impairment.

Results:

Individuals with acute right hemisphere stroke produced fewer content units, total syllables, and lower left–right content unit ratios compared to controls, indicating a paucity of informativeness. Presence of cognitive impairment was associated with fewer content units produced.

Conclusions:

Multiple measures of microlinguistic discourse characteristics differentiated adults with right hemisphere stroke from controls, highlighting variations in both the quantity and quality of connected speech. Findings continue to underscore the contribution and correlation between cognitive skills and discourse performance. Future work is needed to assess the relationship between particular cognitive domains and discourse production as well as to investigate longitudinal changes to discourse production during stroke recovery.

Supplemental Material:

https://doi.org/10.23641/asha.20778541

A stroke in the right hemisphere (RH) of the brain may result in impairments of cognition as well as communication, negatively impacting a person's ability to participate in effective spoken discourse. Deficits in spoken discourse are a hallmark of communication impairments in individuals with RH stroke. Individuals with RH stroke are a heterogeneous group and have variable performance on discourse tasks. RH discourse has been described as egocentric, tangential, verbose or laconic, lacking emotional words, less informative, less relevant, confabulatory, and having poor global coherence (the ability to maintain a topic for the duration of conversation) compared to healthy controls (Agis et al., 2016; Blake, 2006; Joanette et al., 1986; Marini, 2012; Marini et al., 2005; Rogalski et al., 2010; Uryase et al., 1991). Such heterogeneity, poses a challenge for clinicians on objectively identifying discourse impairments. The cognitive–linguistic underpinnings for these deficits are not yet well understood.

Discourse production can provide insight into not only language status but also the interplay of language and cognition on such tasks. Communication is a complex goal-directed behavior that involves multiple cognitive–linguistic processes. To communicate effectively, one must not only have intact word retrieval and auditory comprehension but also adequate sustained attention, memory, and executive functioning during conversation to recognize the topic, formulate a response, recall relevant information, organize thoughts, recognize and revise errors, and attend to linguistic and pragmatic features, such as tone of voice and facial expressions. Although the cognitive basis for discourse impairments is not fully understood, associations between cognitive skills—including attention, memory, and processing speed—and abnormal discourse have been observed (Hillis Trupe & Hillis, 1985; Rogalski et al., 2010).

A number of theoretical models hold promise to understand discourse impairments following RH stroke; however, research remains inconclusive as they may not all be relevant to the wide variety of presentations or have not been rigorously examined or compared (Sherratt & Bryan, 2012). One promising theory suggests that specific cognitive impairments may impact language production at the discourse level (Blake et al., 2002; McDonald, 2000; Sherratt & Bryan, 2012). Deficits in attention and processing speed have been associated with impairments in global coherence (Rogalski et al., 2010). Impairments in discourse post-RH stroke may also be secondary to impairments in executive functioning, which is associated with the right frontal lobe. As executive functions are the higher level processes involved in successful completion of goal-directed behavior, which includes planning, organizing, inhibition, and self-regulation, it makes sense that discourse that is tangential, verbose, or perhaps bizarre may be a result of RH stroke (McDonald, 2000). A second model suggests that the RH plays a role in synthesizing language and, thus, deficits in language production may be in part due to impaired comprehension, difficulty making sense of the context, and impaired plausibility, or ability to comprehend world knowledge (Hough et al., 1990; McDonald, 2000; Sherratt & Bryan, 2012). These deficits may then lead to poor organization of discourse production or impairments in macrostructure.

It was previously believed that microstructural elements of language (e.g., phonology, morphology) were less impacted compared to macrostructural elements (e.g., coherence, topic maintenance, and inferencing; Johns et al., 2008; Rogalski et al., 2010; Schneider et al., 2021). However, emerging research suggests that microstructure may not be as well preserved as initially thought. Research suggests that individuals with an RH stroke utilized significantly more words in an inferential picture description task as compared to controls and used more pictured nouns than inferential nouns (Mackisack et al., 1987). In another study, authors presented participants with unfamiliar objects and probed them to ask at least three questions to help identify the function of the objects. The authors found that individuals with RH stroke asked significantly more content-based questions using “what” compared to neurologically healthy controls who used a wider variety of question types (Minga et al., 2020). Additionally, individuals with acute RH stroke produced fewer content units (CUs), a measure of informativeness, and more syllables per CU (SyllCU), a measure of communication efficiency, than healthy controls, when describing the Cookie Theft picture (Agis et al., 2016). Low CUs and high SyllCU have been associated with low scores on attention and memory tests, and both CUs and SyllCU were associated with lesion location and size, and provided information regarding higher level cortical functions such as attention, integration, and topic maintenance, in addition to hemispatial neglect (Agis et al., 2016; Hillis Trupe & Hillis, 1985). Thus, these studies illustrate that microstructural elements of discourse may not be spared following RH stroke and that impairments in microstructure are associated with specific cognitive impairments.

A variety of stimuli and measures have been used to assess discourse in individuals with acquired neurological impairments, including RH stroke (Bryant et al., 2016). For example, Minga et al. (2021) established a protocol to assess discourse in those with RH stroke that consists of probes for free speech (i.e., open-ended questions regarding communication difficulties), conversational discourse, descriptive discourse (i.e., picture description-using the original Cookie Theft picture), narrative discourse (i.e., cat rescue picture and Cinderella story), procedural discourse (i.e., how to make a peanut butter and jelly sandwich), and question asking. Assessing production across a variety of discourse contexts can help identify communication deficits that may not be obvious in one context over another. For instance, Marini et al. (2005) utilized three different discourse tasks with individuals with RH stroke. The first task involved reading stories aloud and retelling them to a naïve listener. The second involved telling a story based on a series of cartoon-like pictures portraying an event. In the third task, participants had to sequence cartoon-like pictures into a logical order prior to describing the scene. They found that individuals with RH damage had impairments in coherence and lacked informativeness on picture description but not in linguistically based story retell. The authors suggested that this finding may reflect difficulty organizing and retrieving information for visual or picture-based content. Thus, authors suggest that using pictures or visual stimuli to generate discourse may provide a more comprehensive depiction of discourse production in those with RH damage compared to linguistically based tasks. In summary, as communication is a complex task, using a variety of stimuli to elicit discourse is likely to provide a more comprehensive understanding of discourse and impairments in spoken language.

One commonly used stimulus for picture description is the Cookie Theft picture (Goodglass et al., 2001). Speech-language pathologists (SLPs) may use this picture in isolation or as part of the Boston Diagnostic Aphasia Examination (BDAE) to assess language abilities post stroke in the acute to chronic stage of recovery (Goodglass et al., 2001). The Cookie Theft picture is also used by neurologists to quantify language abilities as part of the National Institutes of Health Stroke Scale to determine appropriate treatment for stroke and assess treatment effectiveness (T. Brott et al., 1989; T. G. Brott et al., 1992; National Institute of Neurological Disorders and Stroke rt-PA Stroke Study Group, 1995). A number of studies have analyzed Cookie Theft picture descriptions for microlinguistic elements (Agis et al., 2016; Hillis Trupe & Hillis, 1985; Yorkston & Beukelman, 1980). In a study of RH discourse, Hills Trupe and Hillis (1985) found that adults with RH damage produced Cookie Theft picture descriptions characterized by paucity of speech (i.e., low CUs) as well as tangential speech and digressions (i.e., high SyllCU) compared to typically aging adults and adults with aphasia. These findings were upheld by another study of RH discourse production using the Cookie Theft picture by Agis et al. (2016). Authors found that not only did individuals with RH stroke produce less content as measured by fewer CUs, they also produced fewer CUs per minute, more SyllCU, and a lower ratio of left–right CU (LRCU) than healthy controls (Agis et al., 2016). Differences in CU per minute and SyllCU suggest impairments in communication efficiency, whereas lower LRCU suggests presence of hemispatial neglect. These measures were also correlated with RH lesion locations including the posterior frontal lobe, temporal lobe, and inferior parietal cortex. Additionally, lesion volume was negatively associated with CUs and LRCU and positively correlated with SyllCU (Agis et al., 2016). These two previous studies have thus established that the Cookie Theft picture description, with the above measures, is sensitive to language differences as an RH stroke as to healthy adults.

Though the Cookie Theft picture demonstrates great clinical utility, there was a need to update the stimulus to improve testing validity. As the original Cookie Theft picture appears outdated and representative of traditional gender bias, the new picture is in color and has been updated to include a man washing dishes while the children are attempting to steal cookies. It also features a woman outside mowing the lawn while talking on the cell phone, a cat chasing birds, and a dog in the kitchen. Berube et al. (2019) established normative data on this updated Modern Cookie Theft (MCT) picture stimuli. Using orthographic transcripts from 50 healthy controls, norms were established for CUs, number of syllables (Syll), SyllCU, and left to right content unit ration (LRCU). As described in Berube et al. (2019), the normative data distinguish individuals with aphasia due to left hemisphere (LH) stroke from healthy controls. However, it is yet to be determined if these measures can identify individuals with discourse impairments secondary to RH stroke.

The vast majority of discourse research in RH stroke is completed in the subacute to chronic phase, after there may have been substantial recovery and reorganization of structure–function relationships (Blake, 2006; Joanette et al., 1986; Marini, 2012; Marini et al., 2005; Minga et al., 2021; Rogalski et al., 2010; Schneider et al., 2021; Uryase et al., 1991). There has been limited research in the acute stage of RH stroke. This gap in the literature needs to be addressed because acute deficits reveal the functions for which the RH is critical (before functional reorganization has occurred during stroke recovery). Additionally, the lack of empirical evidence about characteristics of acute discourse production skills post-RH stroke negatively impacts clinical management at that recovery stage. What is considered disordered versus typical variation in discourse production remains to be determined.

In this study, we aimed (a) to investigate discourse production abilities in a sample of adults with acute RH stroke and (b) to assess if clinically diagnosed cognitive impairment was associated with measures of discourse quantity and quality. Our hypotheses are threefold.

1. Measures from the MCT will differentiate those with acute RH stroke from controls.

2. Results of the MCT analysis will reflect both paucity (low CUs, low Syll) and inefficient speech (high SyllCU) as previously reported in the literature using the Cookie Theft picture (Agis et al., 2016; Hillis Trupe & Hillis, 1985).

3. As discourse is a complex cognitive process involving attention, memory, and executive functioning, we predict that paucity and/or inefficient speech identified on the MCT will be associated with clinically diagnosed cognitive impairment (Blake et al., 2002; McDonald, 2000; Rogalski et al., 2010; Sherratt & Bryan, 2012).

Method

In this retrospective analysis, we collected data from individuals who sustained an RH stroke and consented to participate in our longitudinal study approved by the Johns Hopkins School of Medicine Institutional Review Board. Individuals were included if they had clinical evidence of an acute ischemic RH stroke (by computed tomography or magnetic resonance imaging of the brain), who provided informed consent or whose legal guardian provided consent, were 18 years of age or older, did not have a history of previous neurological disease or injury (excluding previous stroke or TIA), reported normal or corrected-to-normal hearing and vision, and had a premorbid proficiency in English. Individuals who had a previous stroke were only included in their previous stroke was small, and the patient had no residual deficits. As previous studies have shown that history of previous stroke does not impact language abilities, we did not expect that these individuals would perform any differently than those with a first-time stroke (Goldberg et al., 2021). Individuals consent to completing language testing and imaging 1–5 days, 3 months, 6 months, and 12 months poststroke. This study includes only those individuals who completed the MCT picture description within 5 days of RH stroke onset.

We utilized the MCT picture description protocol previously established by our group (see Berube et al., 2019). The first 90 s of each recording were transcribed orthographically by a research-trained SLP or trained research assistant. Transcriptions were then analyzed for Syll, CUs, SyllCU, and LRCU ratio and compared to previously established norms for these measures on the same task (Berube et al., 2019). We will briefly summarize development of MCT norms, but additional details can be found in Berube et al. (2019).

Syll

Syll count were collected and analyzed to measure quantity of speech output. CUs then served as a proxy measure of overall amount of speech output. All utterances, including revisions, repetitions, and nonword utterances, were included in the Syll counts, whereas utterances indicating the participant was finished with the task (i.e., “I think that's it” or “That's all”) were not included. This methodology was informed by procedures established by Nicholas and Brookshire (1993).

CUs

Words/phrases were designated as CUs for the MCT if they were included in the picture descriptions of at least three healthy controls from the norms, established and presented in Berube et al. (2019), in order to eliminate infrequent or uncommon outliers only mentioned by one or two persons. Based on this criterion, 96 CU's were identified. CUs were classified as left-sided, ride-sided, or neither/both sides (e.g., “chaos,” “house,” or “little.”; see Berube et al., 2019) depending on the CUs corresponding location in the picture (see Table 1). For instance, the CUs “woman,” “mowing,” and “lawn” would be classified as right CUs since the image of a woman mowing the lawn occurs on the right side of the picture. CUs mentioned multiple times were only counted once. CUs measure the amount of information given.

Table 1.

Content units (CUs).

Content unit classification	Details
Left sided CU	about to fall; ready to fall over; falling off; going to fall; tipping; unbalanced; losing his balance; toppling over; falling backwards
	boy
	children; kids; child
	climbing; tripping; climbed up
	cookie jar; jar; container of cookies, jar of cookies
	cookies; treats
	daughter
	dog
	dropped; dropping; knock them down
	eating; licking; picking up; sticking out his tongue to eat; picking up; to eat it
	getting into; getting; grabbing; stealing; breaking into; seeking; trying to get
	Girl
	grasping to hang on; holding on to, holding
	help; helping
	laughing; laughing her head off
	out of the cabinet; on a shelf; in the cupboard; from the top shelf; from the cabinet
	reaching up, asking for
	shorts
	sister; sis
	socks
	son
	standing there; standing in; standing on; on; standing next to
	stool; step stool; chair; three-legged stool
	striped; stripes
	younger; young
Right sided CU	belt
	birds
	blue; bluish; navy blue, tail blue
	cat
	curtains
	cutting; mowing; mows; on/using lawnmower; yardwork; outside work; with the lawn mower; pushing a motorized miller; pushing a mower
	dad; father
	dishes; plates; dinner plate; bowl
	dishwasher
	flower garden; flower bed; garden of flowers; flowers; tulips; bed of flowers
	flying around; flying away; flying up
	grass; lawn; yard; backyard; garden
	houses; housing; buildings; apartment buildings; home; apartment houses; property
	husband
	man; guy
	mom; mother; mommy
	nice day; nice outside; clear day; beautiful sunny day; beautiful day
	not paying a lot of attention; oblivious; didn't realize; distracted; doesn't know that it's going to go badly; fails to notice; not concentrating; completely clueless; not watching what she is doing
	outside; out; in the background; behind
	overflowing; running out; running over; spilling; spilling on; spilled; pouring out of; flooding; flowing over; overrunning; dribbling over
	pants; jeans; trousers
	picket fence; fence; fenced in
	playing; chasing; having their way; preying on; running after; trying to catch; following; ready to spring; attacking
	sink
	sponge
	spotted; polka dots; spots
	suds; bubbles; soap; sudsy; soapy
	talking on cell phone; phone to ear; speaks on the phone; on the phone; on the telephone; hanging on that cell phone gabbing away; speaking on the cell phone; on the cell phone talking to somebody
	washing (dishes); doing (dishes); scrubbing; cleaning
	watching; looking out; looking through
	water
	wife
	window; kitchen window
	woman; lady
Neither/both CU	afternoon
	big
	black
	blonde
	brown
	chaos, mayhem
	family, family's
	gray; grayish
	green
	hair
	hand
	happy; joyously
	inside; interior
	kitchen
	light; lighter
	little; small
	long, longish
	making quite a mess; slopping; a bit of a mess
	mistake; by mistake
	off the floor; from the floor; on the floor; on the kitchen floor; all over the floor
	one
	open
	orange
	picture
	pink; pinkish
	red
	skirt; dress
	slippers; shoes; booties; pumps; boots
	smiling
	tan; beige
	three
	top; shirt; blouse; jersey
	turquoise
	two; couple of
	wearing; dressed in
	white
	yellow; yellowish

Note. See Berube et al. (2019).

SyllCU

SyllCU was calculated by dividing the total number of Syll by the total CUs to serve as a measure of speech efficiency. SyllCU measures higher than the norm indicate more inefficient speech (i.e., using more syllables to express information), whereas SyllCU lower than the norm may indicate more telegraphic speech (i.e., using fewer syllables to express information).

LRCU

LRCU was collected and analyzed to quantify presence or severity of hemispatial inattention/neglect. Smaller LRCU values suggested left-sided inattention whereas larger values indicated more right-sided inattention.

Next, a retrospective chart review was completed by the first author (S.B.), an SLP with over 8 years of clinical experience in the medical setting, for documentation of cognitive impairment. Based on this chart review, cognitive impairment was marked as present or absent since varying assessment procedures were used across participants. The first author read medical records of the patient's hospitalization for acute stroke, including notes by SLPs, occupational therapists, physical therapists, neurologists, and physical medicine and rehabilitation physicians. If the patient had a previous stroke, the medical chart for that admission was not reviewed in the current study. Cognitive impairment was marked as present if the individual had formal cognitive testing indicating impairment, such as on the Cognitive Linguistic Quick Test (CLQT; Helm-Estabrooks, 2001) or Repeatable Battery for the Assessment of Neuropsychological Status (RBANS; Randolph, 2012); a cognitive screening indicating concern and warranting continued testing for cognitive impairment, such as on the Montreal Cognitive Assessment (MOCA; Nasreddine et al., 2005) or Saint Louis University Mental Status Exam (SLUMS; Tariq et al., 2006); or if a provider described cognition as impaired. Cognitive assessment was completed within the first week of admission.

Charts for all 84 participants were reviewed. Twenty-six individuals completed the RBANS, 13 completed the CLQT, three completed the SLUMS, and nine completed the MoCA. On the RBANS, if a person scored lower than 80 on the total index score, a person was classified as having cognitive impairment (Randolph, 2012). If the scale score fell in the “low average” range with a total index score of 80–89, a person was classified as having cognitive impairment based on the SLPs diagnosis. For the CLQT, the composite severity rating was used to classify cognitive impairment, with a score below 3.5 suggesting cognitive deficits (Helm-Estabrooks, 2001). Scores below 26 on the MOCA and below 27 on the SLUMS were classified as cognitive impairment (Nasreddine et al., 2005; Tariq et al., 2006). Twenty-two patients had documentation from an occupational therapist commenting on impaired cognition and four by physical therapists. Thirteen participants had documentation from an SLP with evaluation and treatment for a diagnosis other than cognition (i.e., dysphagia or dysarthria).

Results

Statistical analyses were completed using Stata/IC 15.1 for Windows. A total of 84 participants (36 females, 48 male) with RH damage due to stroke were included in this study. Participants' mean age was 64.67 ± 13.16 years, mean years of education was 13.92 ± 3.38, and 70 (88.61%) were right-handed. Forty-eight (57.14%) were White/Caucasian, 33 (39%) were Black/African American, and three (4%) identified their race as other (see Table 2). Sixteen (7%) had a history of a previous stroke. We included individuals with previous strokes if the previous stroke was small and had no persisting symptoms, as previous research has found that a history of previous stroke does not necessarily significantly contribute to language impairment in individuals with a new LH stroke unless the previous stroke substantially contributed to total lesion volume from both strokes (Goldberg et al., 2021).

Table 2.

Demographics of individuals with right hemisphere (RH) stroke.

Group	Age (years), M (SD)	Education* (years), M (SD)	Sex* (F), n (%)	Handedness* (R), n (%)	Race* n (%)
RH stroke group (n = 84)	64.67 (13.16)	13.92 (3.38) n = 73	36.0 (42.0)*	70 (79.0) (n = 79)	White: 48 (57.0); Black: 33 (39.0); other: 3 (4.0)
Healthy controls (n = 50)	58.9 (18.2)	16.0 (0.7)	36.0 (72.0)*	36 (82.0) (n = 44)	White: 37 (79.0); Black; 5 (10.6); other: 5 (10.6) (n = 47)

Note. F = female; R = right handed.

^*Statistically significant.

In order to assess group demographic difference, we compared demographics and MCT measures between our RH group and the healthy controls from Berube et al. (2019). Two-way independent-samples t test equal variances assumed showed that there were no significant differences in age, t = −1.4675; df(129), p = .14, though there was a significant difference in years of education, t = 4.4104; df(121), p < .001. Using Pearson chi-square analyses, significant differences were found between groups in terms of frequency of handedness (p = .012), sex (p = .001), and race (p = .001). The healthy controls from Berube et al. (2019) had a higher frequency of participants who were ambidextrous, female, and had more years of education, and more individuals in the control sample were White. We also tested whether there was an association between having a previous stroke and abnormal discourse measures. There were no significant differences in age (p = .430), sex (p = .768), education (p = .859), race (p = .831), or handedness (p = .753) between those with a history of a previous stroke and those without.

Between-Group Comparisons: Overall Output (Syll), Informativeness (Cus), Efficiency (Syllcu), and Presence of Hemispatial Neglect (LRCU)

Results for Hypotheses 1 and 2

We hypothesized that measures from the MCT would identify individuals with communication differences resulting from RH stroke and that the results would reflect both paucity and inefficient speech as previously reported in the literature (Agis et al., 2016; Blake, 2006; Hillis Trupe & Hillis, 1985).

To investigate our first hypothesis, we compared MCT measures of the 84 individuals with RH stroke to the norms established by Berube et al. (2019; see Table 3). We completed between-group comparisons using independent-samples t test (equal variances assumed) with multiple comparisons corrected via Bonferroni correction procedures (corrected alpha ≤ .0125). Significant differences were observed in CU, t(132) = 9.0233, p < .001, and Syll t(132) = 7.2179; p < .001 (see Figures 1 and 2). Adults with acute RH stroke produced fewer CUs and Syll compared to healthy adult controls, indicating that adults with RH stroke produced less overall speech and less information in their speech compared to controls. Of note, there were a few outliers who produced CU and Syll above the range of the RH group. These participants did not have evidence of cognitive impairment on clinical chart review; thus, we suspect that these participants did not experience cognitive-communication following stroke. The two groups were not different in SyllCU, t(132) = −0.8443; p ≤ .40, nor LRCU, t(132) = 0.2767; p < .55 (see Figures 3 and 4). To contextualize this finding, there were no significant differences in measures of efficiency (SyllCU) or hemispatial neglect/inattention (LRCU) between adults with acute RH stroke and healthy controls. Transcripts and measure examples for impairments in CUs, Syll, SyllCU, and LRCU for individuals with RH stroke and healthy controls can be found in Supplemental Material S1.

Table 3.

Between-group comparison of Modern Cookie Theft discourse measures.

Group	CUs, M (SD)	Syll, M (SD)	SyllCU, M (SD)	LRCU ratio, M (SD)
RH stroke group	17.32 (9.32)*	107.39 (65.14)*	6.27 (2.58)	0.69 (0.49)
Healthy controls	33.5 (11.4)*	197 (87.2)*	5.8 (1.5)	0.74 (0.49)

Note. CUs = content units; Syll = syllables; SyllCU = syllables per CU; LRCU = left to right content unit ratio; RH = right hemisphere.

^*Statistically significant.

Figure 1.

Between-group differences: CU. CU = content units; RH = right hemisphere.

Figure 2.

Between-group differences: Syll. Syll = syllables; RH = right hemisphere.

Figure 3.

Between-group differences: SyllCU. SyllCU = syllables per CU; RH = right hemisphere.

Figure 4.

Between-group differences: LRCU. CU = content units; LRCU = left to right CU ratio; RH = right hemisphere.

Results for Hypothesis 3

Since previous work suggests a relationship between discourse production and cognitive processing, we assessed the correlation between presence/absence of cognitive impairment (via chart review) and measures of CUs, Syll, SyllCU, and LRCU. We predicted that differences identified on the MCT would be associated with cognitive impairment. Of the 84 participants with RH stroke, 54 (64.29%) had cognitive impairment documented in their medical record. Presence of cognitive impairment was associated with abnormal CUs, Pearson X ²(1) = 9.04; p = .003 (see Figure 5). There were no associations between cognitive impairment and abnormal Syll, Pearson X ²(1) = 3.24; p = .072; high SyllCU, Pearson X ²(1) = 0.1716; p = .68; and low LRCU, Pearson X ²(1) = 2.12; p = .137.

Figure 5.

Frequency of cognitive impairment and abnormal discourse measures in individuals with RH stroke. RH = right hemisphere; CU = content units; Syll = syllables; SyllCU = syllables per CU; LRCU = left to right content unit ratio; * = statistically significant compared to the normative sample.

Within-Group Comparisons: Normal and Abnormal MCT Measures in RH Stroke

Participants with acute RH stroke were then classified as demonstrating typical or atypical discourse production. We identified the frequency of MCT measures greater than or less than 1.5 SDs from the mean of healthy controls for individuals with acute RH stroke as this cutoff is frequently used to determine disordered versus typical performance (Antonsson et al., 2021; Ostrand & Gunstad, 2021). For Syll, 24 (28.57%) participants with RH stroke produced syllables significantly below normal. For CUs, 38 (45.24%) individuals with RH stroke had an abnormally low number of CUs. Sixteen (19.05%) individuals with RH stroke had an above average SyllCU, suggesting verbosity. In regard to LRCU, 12 (14.29%) participants with RH stroke showed abnormally low LRCU, suggesting presence of left neglect.

We compared frequency of abnormal measure between those with and without history of a previous stroke using independent-samples t test (equal variances assumed) with multiple comparisons corrected via Bonferroni correction procedures (corrected alpha ≤ .0125). There were no significant differences in frequency of abnormal CU, Pearson X ²(1) = 0.0299; p = .863; Syll, Pearson X ²(1) = 0.5048; p = .477; SyllCU, Pearson X ²(1) = 0.6617; p = .416; or LRCU, Pearson X ²(1) = .0198; p = .888, between those who had a previous stroke and those who did not. Using two-way independent-samples t test equal variances assumed, there were no significant differences in age or education for participants with RH stroke with atypical versus typical CU, Syll, SyllCU, or LRCU (all p values > .05), nor were there significant associations between these measures and race by Pearson chi-square analysis (all p values > .05). However, there was a significant association between sex and low Syll, Pearson X ²(1) = 5.29; p = .021, and low LRCU, Pearson X ²(1) = 5.91; p = .015 (see Tables 4 and 5).

Table 4.

p values (by chi-square or t test) for within-group comparison of demographics of individuals with right hemisphere stroke with abnormal versus normal Modern Cookie Theft discourse measures.

Variable	Low CU	Low Syll	High Syll/CU	Low LRCU	Cognitive impairment
Age	.273	.961	.157	.543	.6886
Education	.026	.969	.660	.0382	.6722
Sex	.100	.021*	.936	.015*	.393
Handedness	.637	.287	.304	.913	.332
Race	.624	.966	.812	.596	.133
Cognitive impairment	.003*	.072	.679	.137

Note. CU = content unit; Syll = syllable; LRCU = left/right content unit ratio.

^*Statistically significant.

Table 5.

Pairwise correlations for Modern Cookie Theft measures, age, and education corrected for multiple comparisons.

Variable	Age	Education	CU	Syll	SyllCU	LRCU
Age	1.0000
Education	−.1020	1.0000
CU	−.1142	.1688	1.0000
Syll	−.0202	.2120	.8406	1.0000
SyllCU	.0824	−.0135	−.0562	.3878	1.0000
LRCU	.0917	.0079	.2707	.2126	−.0236	1.0000

Note. CU = content unit; Syll = syllable; SyllCU = Syllables per CU; LRCU = left/right content unit ratio.

Multivariable linear regressions were calculated to predict LRCU based on sex and age and to predict Syll based on sex and age. There was no statistical significance found for neither LRCU, F(2, 81) = 0.35, p = .7081, with an R ² of .0085, nor Syll, F(2, 81) = 0.77, p = .4644, with an R ² of .0188. Thus, sex was not correlated with LRCU or Syll, independent of age.

Discussion

Although it has long been appreciated that people with RH damage poststroke experience difficulties with some aspects of communication, their deficits are less easily quantified and characterized than communication problems after LH stroke. Functional imaging studies of language processing in neurotypicals consistently show activation in RH homologues of the language network associated with both expressive and receptive language tasks (e.g., Buckner et al., 1995; Hocking et al., 2009; Vartanian & Goel, 2005), but the role of the RH during these tasks is rarely discussed. Previous work in stroke populations has found that picture description provides samples of discourse that can be reliably compared to that of healthy controls to help capture expressive language differences (Agis et al., 2016; Andreetta et al., 2012; Berube et al., 2019; Nicholas & Brookshire, 1993; Yorkston & Buekelman, 1980). Picture description is well suited for characterizing discourse as individuals with RH stroke express fewer of the concepts mentioned by neurotypical adults and sometimes produce more words and syllables in doing so, adding concepts that are not considered relevant by healthy controls (Agis et al., 2016; Hillis Trupe & Hillis, 1985; Blake, 2005; Mackenzie et al., 1997). We confirmed these differences in people with acute RH stroke before reorganization or recovery. These results provide additional support that the “normal” roles of the RH during language include attention, integration, and selection/maintenance of salient concepts.

Many of our findings regarding picture description differences after RH stroke have been previously described using other stimuli, including the original Cookie Theft picture from the BDAE (Agis et al., 2016; Hillis Trupe & Hillis, 1985; Joanette et al., 1986; Marini, 2012; Marini et al., 2005; Uryase et al., 1991). Both the original Cookie Theft picture and the MCT have been successful in identifying paucity of speech using CUs and inefficient speech using SyllCU (Agis et al., 2016; Hillis Trupe & Hillis, 1985). Berube et al. (2019) had both healthy controls and persons with aphasia who describe both the MCT and original Cookie Theft picture. They found that there was a high correlation between the two pictures in measures of CUs, Syll, and SyllCU. Additionally, individuals produced significantly more CUs in their descriptions of the MCT as compared to the original Cookie Theft picture. However, novel aspects of our study are that we used the MCT picture and that patients were studied acutely after stroke. The MCT has several advantages over the original, including a greater variety of concepts elicited (including colors) and absence of gender stereotypes. Studying acute rather than chronic stroke has the advantage of identifying deficits before they resolve or are modified over the course of recovery as patients relearn or learn to compensate for lingering disabilities. For example, in the acute stage, decreased verbal output and informativeness may represent a loss or impaired ability to provide information pertinent to the topic. In the chronic stage, these deficits may evolve into verbose speech characteristics, which demonstrates a compensatory behavior to increase the amount of salient information given.

We confirmed our predictions that the MCT picture description task could identify discourse differences following acute RH stroke. As those with RH stroke are a heterogeneous group (Tompkins, 2008; Weed, 2008), we found that some of the abnormalities we identified were present in only a subset of individuals with RH stroke. Although group means of some abnormal characteristics were not statistically significant from norms, it may be the case that some individuals (perhaps those with specific lesion sites) had significant differences compared to controls. For example, 16 (19%) individuals with RH stroke had a higher total SyllCU ratio when compared to the healthy controls, which may indicate inefficient or verbose speech secondary to repetitions, irrelevant content, or digressions (Agis et al., 2016; Hillis Trupe & Hillis, 1985). However, group-level analyses indicated no statistical differences between them. Likewise, 12 (14%) individuals had an abnormally low LRCU, indicating the presence of left hemispatial neglect. Women (who also were older than men) more frequently showed this pattern and also had lower Syll relative to men. However, it is probable that lesion site or volume was more likely to account for differences in deficits across individuals, although we did not have adequate imaging data to evaluate this hypothesis. Future studies are planned to determine the sites of lesions associated with specific abnormalities revealed with the MCT (see Agis et al., 2016, for lesions associated with specific deficits in describing the original cookie theft picture).

As discourse is a complex goal-directed behavior involving attention, memory, and executive functioning along with other linguistic processes, such as lexical retrieval and syntactic planning, we hypothesized that individuals clinically diagnosed with cognitive deficits following RH stroke may demonstrate differences in discourse as seen on the MCT. We found that presence of cognitive impairment was associated with low CUs. This finding is consistent with previous research that reported paucity of speech as measured by CU in the original Cookie Theft picture descriptions was associated with attention and memory scores (Hillis Trupe & Hillis, 1985). Low CU may be associated with deficits in sustained attention, (working) memory, or executive functioning, which may interrupt or diminish task perseverance due to loss of set and early task abandonment. Or, it may be the case that participants become distracted, resulting in production of excess and unrelated speech without awareness or initiation of redirection, as deficits in processing speed, working memory, attention, and executive functioning have been associated with impairments in topic maintenance (Johns et al., 2008; Rogalski et al., 2010; Zimmerman et al., 2011). Such features have been recognized in previous studies describing RH discourse going to either extreme of verbosity or paucity (Hillis Trupe & Hillis, 1985). As individuals with RH stroke often have deficits in visuospatial skills, it cannot be ruled out that such differences may have negatively impacted performance on a picture description task. However, there was no significant difference in LRCU, a measure of hemispatial neglect, between the RH group and the norm. Thus, hemispatial neglect is less likely, but general visual processing and attention deficits, such as during scanning, may have negatively impacted overall performance rather than performance isolated to one visual field. In the future, we plan to examine which specific cognitive process(es) are associated with differences of MCT discourse measures.

One limitation to our study was the absence of uniform testing procedures to identify the presence or absence of cognitive impairment. As our study was retrospective and utilized clinical data to identify presence of cognitive impairment, we could not determine if specific cognitive functions were associated with low CUs. In the future, we will utilize uniform cognitive measures to gain insight into which specific cognitive processes are associated with measures of discourse production quantity and quality. A second limitation was that our sample of RH and healthy controls were significantly different in sex, years of education, race, and handedness, as our healthy controls were more often ambidextrous, female, had more years of education, and were White. It is important in the future that we establish a more diverse, representative normative sample by which to compare performance of patient populations for determining differences in discourse. Additionally, our sample only included the acute stroke time point, and we did not have subsequent data from a subacute or chronic time point. We were unable to assess if the MCT is sensitive to discourse differences beyond the acute stage, but this is another future direction of our continued work.

In conclusion, the MCT picture description task and measures can identify individuals with communication differences following an RH stroke. On average, individuals with RH stroke tended to produce fewer CU and fewer syllables, suggestive of paucity of speech and less relevant information. Some individuals with acute RH stroke also produced higher SyllCU and/or lower LRCU than controls, though this was not significant. Lower CUs was associated with presence of cognitive impairment, though additional research is needed to identify which aspects of cognition are related to deficient content in discourse production.

Funding Statement

This work was supported by National Institutes of Health/National Institute on Deafness and Other Communication Disorders grants R01 DC05375, P50 DC014664, and R01DC015466 awarded to Argye E. Hillis.