Journal Logo

Original Studies

An Efficient Bedside Measure Yields Prognostic Implications for Language Recovery in Acute Stroke Patients

Keator, Lynsey M. MA, CCC-SLP*; Faria, Andreia V. MD, PhD; Kim, Kevin T. BA*; Saxena, Sadhvi MS, MHS*; Wright, Amy E. BA*; Sheppard, Shannon M. PhD, CCC-SLP*; Breining, Bonnie L. PhD*; Goldberg, Emily MS, CCC-SLP*; Tippett, Donna C. MA, MPH, CCC-SLP*,‡,§; Meier, Erin PhD, CCC-SLP*; Hillis, Argye E. MD, MA*,‡,∥

Author Information
Cognitive and Behavioral Neurology: September 2020 - Volume 33 - Issue 3 - p 192-200
doi: 10.1097/WNN.0000000000000238


As the leading cause of disability in the United States (Benjamin et al, 2017), stroke is a major health concern. Poststroke survivor rates are rising (Fisher et al, 2014), increasing demands on the health care system (Breitenstein et al, 2017) due to longer hospital stays (Ellis et al, 2012) and extended rehabilitation services (Dickey et al, 2010; Pederson et al, 2004).

Many stroke survivors (21–38%) suffer from aphasia (Engelter et al, 2006), a language disorder resulting from damage to the left-hemisphere (LH) regions supporting language. Aphasia affects verbal expression, auditory comprehension, reading, and/or writing abilities (Parr et al, 1997). Not only does it influence one’s ability to communicate with family and friends (Hemsley and Code, 1996), but aphasia also drastically decreases one’s education and employment opportunities, thus resulting in a poor quality of life (Franzén-Dahlin et al, 2010; Hilari et al, 2003). Due to the impact of aphasia on one’s quality of life and the significant costs associated with health care for individuals with aphasia (IWA) (Dickey et al, 2010), early identification of aphasia has important implications for prognostication and treatment planning.

Typically, individuals suspected of experiencing a stroke are assessed within minutes or hours of hospital admission. Despite variable individual performance during this acute phase, the logistical demands of the health care system require clinicians to make quick diagnostic decisions (Johnson et al, 1998). Initial severity of one’s language deficits is an important factor to consider when determining a prognosis for IWA (Lazar et al, 2008; Plowman et al, 2012); consequently, it is critical that the presence and severity of aphasia are adequately and accurately evaluated in the acute phase (Nouwens et al, 2015). Poststroke language intervention, most commonly provided by speech-language pathologists, has been found to optimize patient outcomes (Godecke et al, 2012), especially when compared to individuals who did not receive aphasia intervention (Brady et al, 2016; Breitenstein et al, 2017).

Traditionally, in an acute clinical care setting, poststroke language function is evaluated using a myriad of clinical measurements and diagnostic assessments (Rohde et al, 2018); however, many assessments are too demanding, time consuming, and/or cumbersome for IWA to complete in the initial stage of recovery (El Hachioui et al, 2017). In addition, speech-language pathologists often are not available to conduct a detailed linguistic evaluation in the first hours and days post stroke (El Hachioui et al, 2017).

Neuroimaging studies have demonstrated the site and volume of lesion to be important factors in predicting an individual’s recovery from stroke (Heiss et al, 2003; Watila and Balarabe, 2015). The associations between lesion site and aphasia that have been identified in these studies have contributed to our understanding of the loss of language function in the context of lesion characteristics (Saur et al, 2006). However, these studies relied on assessments such as stroke scales like the National Institutes of Health Stroke Scale (NIHSS) (Brott et al, 1989) or brief screening tests like the Frenchay Aphasia Screening Test (Enderby et al, 1987) and Language Screening Test (Flamand-Roze et al, 2011) that do not capture language abilities in natural contexts (LaPointe, 2011). Additionally, an individual’s clinical presentation, especially in terms of language function, is unstable immediately after a stroke and changes rapidly (El Hachioui et al, 2017), thereby suggesting that extensive testing may not be efficient or necessarily productive in the acute setting. Therefore, it is critical to identify an efficient screening test for aphasia that can be administered not only by speech-language pathologists (the primary health care providers involved in aphasia rehabilitation), but also by other health professionals such as neurologists.

A previous study by our group (Agis et al, 2016) investigated the relationship between patient performance on a picture description task (Cookie Theft, originally from the Boston Diagnostic Aphasic Examination [Goodglass et al, 2001] and now part of the NIHSS [Brott et al, 1989; Lyden et al, 2004]; see Figure 1 in the Supplemental Digital Content 1, and brain imaging to determine if such a relationship may provide relevant clinical information above and beyond that provided by the NIHSS language score (which rates aphasia severity qualitatively on a scale from 0 to 2). The study included patients with LH or right-hemisphere (RH) damage and a healthy, age-matched control group. Results suggested that performance on the Cookie Theft picture description task contributes independent information that is not captured in the NIHSS score. Because this picture description task is already included in the NIHSS, no additional administration time is needed. Agis et al (2016) emphasized that the task requires minimal training, <2 minutes to administer, and only a few additional minutes to analyze.

Expanding on the 2016 investigation by Agis and colleagues, we tested the hypotheses that objective measures on a picture description task administered within 48 hours post stroke would (a) predict language recovery, (b) estimate LH lesion volume and location, and (c) correlate with other commonly administered bedside language assessments.

We hypothesized that objective measures from the Cookie Theft picture description task, a narrative language task, would predict an individual’s communication outcomes at the chronic time point (>6 months post stroke). We also hypothesized that performance on the task would be associated with acute vascular and lobar damage in the LH and that performance measures from the task would correlate with other, more time-consuming and cumbersome language assessments that are typically administered bedside in the acute phase of stroke recovery.

To our knowledge, no study has examined the value of narrative speech measures from acute stroke patients to predict language performance in the chronic stage of recovery. The predictive value of this picture description task, and correlations with existing assessments, may substantiate the clinical importance of a reliable yet rapid bedside measure for acute stroke patients that can be administered by a variety of health care professionals.



We recruited 90 patients (M age = 59.06 ± 11.74 years; 24.90% women; M education = 13.54 ± 3.05 years) with an MRI-confirmed LH ischemic stroke. Recruitment was completed consecutively as part of a larger project. This sample was obtained by screening an additional 353 patients on the stroke service. Of these, 271 were ineligible (due to hemorrhage, previous stroke, or neurologic disease affecting the brain; time post stroke; poor level of consciousness; left handedness; and/or lack of English proficiency), and 82 declined. Inclusion criteria were as follows: clinical evidence of acute hemispheric stroke, able to complete testing within 48 hours of symptom onset, able to provide informed consent, premorbid proficiency in English, age ≥18 years, and premorbidly right handed. Exclusion crtieria included previous symptomatic stroke or history of neurologic disease.

Of the 90 patients initially recruited, 86 (95%; M age = 59.03 ± 11.93 years; 29.06% women; M education = 13.62 ± 3.05 years) completed testing acutely (within 48 hours of admission and 24 hours of their initial MRI scan) at Johns Hopkins University Hospital. Four of the patients (4.4%) were unable to complete testing acutely due to fatigue or discharge from the facility, but they were tested at the chronic time point (>6 months post stroke) and included in behavioral analyses.

Of the 86 patients who were tested acutely, 25 (28%; M age = 58.83 ± 13.90 years; 32% women, M education = 14.52 ± 2.98) were also tested at the chronic time point. This number represents those who were able and willing to return for behavioral assessments and MRI. Sixty-one patients did not return for a second assessment for a variety of reasons (declined participation, new onset of exclusion criteria, or not yet 6 months post stroke). In total (including the four patients who were not tested acutely), 29 patients completed the testing at the chronic time point (32% of the initial 90; M age = 58.93 ± 13.17; 24.03% women; M education = 14.16 ± 3.08). Demographic data for the individuals who were assessed acutely (n = 86) and chronically (n = 29) are provided in Table 1. Figure 1 shows a lesion overlap map for 88 of the patients (2 scans were unavailable at the time of our analysis).

TABLE 1 - Study Participants’ Demographic Information
Testing Age (Years) Sex (Female), n (%) Education (Years)
Acute (n = 86) 59.03 ± 11.93 25 (29.06) 13.62 ± 3.05
Chronic (n = 29) 58.93 ± 13.17 7 (24.03) 14.16 ± 3.08
Data are presented as M ± SD unless otherwise indicated.

Lesion overlap map for 88 of the 90 study participants (2 scans were unavailable at the time of analysis). The color indicates the number of participants with lesion damage at a particular location. The upper boundary (n = 10) of the color scale represents the highest lesion overlap.

A cohort of 35 healthy, age-matched controls (M age = 60.6 ± 10.1) from the Agis and colleagues (2016) study was used to compare the performance of patients to the performance of individuals with typical language function.

All of the study participants were right handed and were premorbidly proficient speakers of English. According to electronic medical records and clinical assessment, participants did not have a history of a previous symptomatic stroke or neurologic disease nor was there indication of reduced arousal or ongoing sedation. Inclusion and exclusion criteria for all of the participants were consistent with those in Agis and colleagues (2016).

All of the study participants independently provided informed consent or indicated a decision maker to provide consent. The study protocol was approved by the Johns Hopkins Medicine Institutional Review Board and was planned, conducted, and recorded according to the principles of the Declaration of Helsinki and its later amendments.

Behavioral Assessments

Test administration included the Cookie Theft picture description task, the 30-item short form of the Boston Naming Test (BNT; Williams et al, 1989), the Hopkins Action Naming Assessment (HANA; Breining et al, 2015), and selected subtests of the Western Aphasia Battery—Revised (WAB–R; Kertesz, 2007). Not all of the participants completed all of the assessments at the acute time point due to fatigue or discharge from the hospital. Of the 86 who were tested acutely, 86 completed the Cookie Theft picture description task, 81 completed the BNT, 69 completed the HANA, and 33 completed the WAB–R. Of the 29 participants who were tested at the chronic time point, 29 completed the Cookie Theft picture description task, 25 completed the BNT, and 23 completed the HANA. The WAB–R was not administered at follow-up.

We recorded the participants’ descriptions of the Cookie Theft picture description task at the time of assessment for subsequent transcription and analysis by a speech-language pathologist (L.M.K.). A standard prompt, consistent with task administration for the Boston Diagnostic Aphasic Examination, was used: “Tell me everything you see going on in this picture.” There was no time limit for answering.

The participants’ picture descriptions were scored for connected speech content (total content units [CU; words/phrases or their exact equivalents]); communicative efficiency (syllables/content unit; Syll/CU); and ratio of left to right content units (L:R CU ratio; ratio of the number of CU produced from the left side of the picture to the number of CU produced from the right side of the picture), which is a measure of cortical dysfunction. Total time spent by the speech-language pathologist analyzing the picture description was estimated to be 15 minutes per participant.

Normed data and a scoring system exist for this picture description task and include a list of published CU (eg, washing/doing dishes; see Table 1 in the Supplemental Digital Content 2 for a table of published CU, mentioned by healthy controls (Craig et al, 1993; Yorkston and Beukelman, 1980). We used this information to score the participants’ picture descriptions for CU, Syll/CU, and L:R CU ratio. Consistent with the published scoring conventions, each CU was scored only once, even if the participant mentioned it many times or in different ways (ie, mom, mother, mama), and 1 point was awarded for each correctly named CU.

A previous analysis of picture descriptions indicated that IWA produce fewer CU and more Syll/CU than healthy controls, indicating inefficient communication (Yorkston and Beukelman, 1980). The L:R CU ratio quantifies contralesional hemispatial neglect for stroke patients with both LH and RH damage (Agis et al, 2016). Previous studies have demonstrated that right neglect is captured by this ratio (Agis et al, 2016) and may be as common after an LH stroke as left neglect is after an RH stroke (Kleinman et al, 2007).

The short form of the BNT prompts patients to name 30 objects (presented as line drawings), with 1 point awarded for every correct answer (score range = 0–30). Objects range from high-frequency items such as bed to low-frequency items such as abacus. If an individual is unable to independently name an object, the administrator provides a phonemic cue. Correct responses elicited in response to a cue are not included in the total number of correct responses.

The HANA is used to evaluate action naming. Patients are prompted to name 30 actions (eg, run, spill, whisper; presented as line drawings), with 1 point awarded for every correct answer (score range = 0–30). These stimuli are matched in number and frequency to items on the BNT. If patients name an object instead of an action, they are reinstructed to name an action. Correct responses elicited in response to phonemic cues are not included as a correct response. The total raw score is calculated from the total number of correct responses.

The WAB–R assesses expressive and receptive language function across the domains of spontaneous speech (ie, information content and fluency, grammatical competence, and paraphasias), auditory verbal comprehension (ie, yes/no questions, auditory word recognition, and sequential commends), repetition, and naming and word finding (ie, object naming, word fluency, sentence completion, responsive speech). Part 1 of the WAB–R was administered and scored according to the manual. For the 33 patients who completed the WAB–R acutely, the Aphasia quotient was used as a measure of aphasia severity.

Neuroimaging Data

All of the participants participated in an MRI scan acutely, including diffusion-weighted imaging. Technicians, blinded to behavioral data, identified the boundaries of ischemia/infarct, which was defined as hyperintense on diffusion-weighted imaging (and hypointense on apparent diffusion coefficient sequences). A radiologist (A.V.F.) reviewed the stroke delineations and calculated the total volume of stroke using ROIEditor software ( The least diffusion-weighted imaging (b value = 0) was used for brain mapping.

In order to minimize the effects of the intensity abnormalities (due to the stroke) in the imaging mapping, the lesioned area was replaced by the “healthy” tissue from the contralateral hemisphere. This “artificial” b value = 0 image was mapped to the JHU-MNI-b0 atlas (Mori et al, 2008; by a series of linear transformations and by large deformation diffeomorphic metric mapping (Ceritoglu et al, 2013). Using the resultant transformation fields, two types of maps (lobar, based on classical anatomy, and vascular, based on vascular territories) were inversely warped from the Johns Hopkins University template to each individual’s brain (Ceritoglu et al, 2013; Djamanakova et al, 2014; Oishi et al, 2009), thereby performing the automated segmentation. Total lesion volume and percent of damage in each cortical lobe (frontal, parietal, temporal, occipital), thalamus, basal ganglia, and each vascular territory (anterior cerebral artery, middle cerebral artery [MCA], posterior cerebral artery) were calculated.

Statistical Analysis

We used SPSS (Version 25) to perform all analyses. We used paired t tests to evaluate the differences between the language performance of the patients who were tested at the acute time point (n = 86) and those who were tested at the chronic time point (n = 29). Unpaired Welch t tests were used to account for unequal sample sizes (Levene test for equality of variances, P = 0.000) in order to compare demographics of those who were tested at the chronic time point (n = 29) and those tested acutely (n = 61). We also used unpaired Welsh t tests to compare language performance of our clinical population with that of the controls. We used multivariable linear regressions to predict communication recovery (change in CU from acute to chronic time point divided by acute CU) and communication outcome (CU at >6 months post stroke). Multivariable linear regressions were also used to determine if these two language measures (communication recovery and communication outcome) predicted total lesion volume and location. Pearson correlations were calculated between picture description task measure and separately with lesion volume and language battery scores (BNT, HANA, WAB–R).


There were no significant differences in demographic data between the patients who were tested acutely and those who were tested at the chronic time point. There was, however, a significant difference in Cookie Theft picture description CU: Patients performed significantly better on the task at the chronic time point compared with the acute time point (t24 = −5.35; P < 0.001; 95% CI: −6.68, −2.96). Similarly, patients performed significantly better on the BNT at the chronic time point compared with the acute time point (t22 = −2.41; P = 0.025; 25% CI: −7.77, −0.58). No significant difference was found between scores on the HANA at the two time points (Table 2).

TABLE 2 - Differences in Behavioral Performance
Behavioral Changes From Acute to Chronic Time Point for Patients Tested at Both Time Points (n = 25)
Time Point CU Syll/CU L:R CU Ratio BNT HANA
Acute 6.28 ± 4.94 8.58 ± 9.86 0.93 ± 0.71 17.61 ± 10.75 18.94 ± 9.78
Chronic 11.1 ± 6.04* 6.75 ± 4.23 1.18 ± 0.55 21.78 ± 7.76* 21.71 ± 6.68
Behavioral Differences Between Acute Patient Cohort and Healthy Age-matched Controls
Patients (n = 86) 7.72 ± 4.95* 18.71 ± 9.78* 17.73 ± 9.44*
Control (n = 35) 16 ± 3.7 27.5 ± 2.4 23.92 ± 3.46
*Significant at P<0.05.
†Patients performed with lower scores than controls for the picture description task (CU) and BNT.
BNT = Boston Naming Test. CU = content units. HANA = Hopkins Assessment of Naming Actions.

Acutely, patients’ CU was significantly lower than that of the age-matched controls (t121 = −10.15; P < 0.0001; 95% CI: −9.89, −7.22). At the acute time point, patients’ mean score on the BNT was significantly lower than that of the controls (t120 = −5.45; P < 0.0001; CI: −11.97, −5.59). Patients’ mean score on the HANA at the acute time point was also significantly lower than that of the controls (t108 = −3.98; P < 0.0001; 95% CI: −9.28, −3.11) (Table 2).

Communication Recovery

Communication recovery was determined by the change in total CU production from the acute time point to the chronic time point: chronic CU − acute CU/baseline CU. Communication recovery negatively correlated with acute CU (r = −0.50; P = 0.015) and was predicted by acute CU, age, and LH lesion volume; F3,18 = 3.98; P = 0.024; r2 = 0.40. After adjusting for other variables, acute CU independently predicted communication recovery (t = −2.5; P = 0.021; 95% CI: −1.42, −0.130). Importantly, there was a significant correlation between acute CU and chronic CU: r = 0.68; P = 0.0002. Acute CU explained 46.3% of the variance in chronic CU and predicted chronic CU performance (t = 4.45, P < 0.0001; 95% CI: 0.446–1.219). However, there was no significant correlation between acute CU and change in CU: chronic CU − acute CU; r = –0.18; P = 0.38.

Communication Outcome

Communication outcome (ie, picture description task measures at the chronic time point) was determined by the number of CU at the chronic time point. CU outcome was predicted by acute CU, age, and lesion volume (F3,19 = 4.08; P = 0.02; r2 = 0.39), but only acute CU independently predicted communication outcome after adjusting for other variables (t = 3.198; P = 0.005; 95% CI: 0.23–1.11) (Table 3).

TABLE 3 - Multivariable Linear Regression: Predictors of Communication Outcome
Predictor Coefficient Standard Error t P 95% Confidence Interval
Acute content units 0.672* 0.210 3.198 0.005 0.232–1.111
Age 0.042 0.072 0.588 0.564 –0.108–0.193
Total lesion volume 0.000012 0.000 0.332 0.743 0.000–0.000
*Significant at P<0.05.

Total lesion volume was predicted by acute CU, age, and education (F3,58 = 2.83; P = 0.046, r2 = 0.13), and acute CU independently predicted the total lesion volume when controlling for other variables (t = –2.85; P = 0.006; 95% CI: −5162.7, −904.13). Percent of damage to the left MCA territory was also best predicted by a model that included acute CU, age, and education (F3,48 = 3.64, P = 0.019, r2 = 0.19), but only acute CU independently predicted left MCA percent of damage (t = –3.04; P = 0.004; 95% CI: −1.59, −0.32).

Higher acute CU (better performance) negatively correlated with total LH lesion volume (r = –0.40; P < 0.0001), percent of damage to the frontal (r = –0.40; P = 0.002) and temporal (r = –0.45; P < 0.0001) lobes, and percent of damage to the left MCA region (r = –0.44; P < 0.0001).

Correlations between acute CU on the picture description task and lesion characteristics are provided in Tables 4 (lobar lesion volume) and 5 (vascular territory lesion volume). Chronic CU only correlated with percent of damage to the temporal lobe (r = –0.43; P < 0.048).

TABLE 4 - Acute Picture Description Measures Correlate With Lobar Lesion Volume and Location
Total Lesion Volume Left Frontal Lobe Percent Damage Left Parietal Lobe Percent Damage Left Temporal Lobe Percent Damage
r P r P r P r P
Acute content units –0.40*** <0.0001 –0.40** 0.002 –0.29 0.17 –0.45*** <0.0001
*Significant at P < 0.05.
**Significant at P < 0.01.
***Significant at P < 0.001.

TABLE 5 - Acute Picture Description Measures Correlate With Lesion Volume in Vascular Territories
ACA Percent Damage MCA Percent Damage PCA Percent Damage
r P r P r P
Acute content units –0.17 0.22 –0.44*** <0.0001 –0.23 0.08
*Significant at P < 0.05.
**Significant at P < 0.01.
***Significant at P < 0.001.
ACA = anterior cerebral artery. MCA = middle cerebral artery. PCA = posterior cerebral artery.

Correlations With Other Language Measures


At the acute time point, quantitative measures of the Cookie Theft picture description task (CU, Syll/CU, L:R CU ratio) predicted aphasia severity, as measured by the WAB–R Aphasia quotient: F3,29 = 17.83; P < 0.0001; r2 = 0.65. After controlling for Syll/CU and L:R CU ratio, CU independently predicted WAB–R Aphasia quotient (t = 6.8; P < 0.0001; 95% CI: 3.09–5.75). Acute CU correlated with the following: WAB–R Aphasia quotient (r = 0.78; P < 0.0001); the Fluency, Grammatical Competence, and Paraphasias subtest of the WAB–R (r = 0.54; P = 0.017); HANA scores (r = 0.61; P < 0.0001); and BNT scores (r = 0.65; P = 0.0001). Communicative efficiency (Syll/CU) negatively correlated with BNT scores (r = –0.23; P = 0.041), and L:R CU ratio positively correlated with HANA scores (r = 0.25; P = 0.04). Table 6 shows the correlations between the Cookie Theft picture description measures and the other language tests.

TABLE 6 - Narrative Measures Correlate With Language Measures
Acute Narrative Measures Correlate With Language Measures
Language Measure r P r P r P
Content units 0.78 <0.0001*** 0.65 <0.0001*** 0.61 <0.0001***
Syll/CU 0.03 0.88 –0.23 0.041* –0.22 0.07
L:R CU ratio 0.29 0.12 0.20 0.76 0.25 0.04*
Chronic Narrative Measures Correlate With Language Measures
Language Measure r P r P
Content units 0.57 <0.001** 0.61 <0.001**
Syll/CU –0.168 0.422 –0.013 0.955
L:R CU ratio 0.41 0.043* 0.32 0.138
*Significant at P < 0.05.
**Significant at P < 0.01.
***Significant at P < 0.001.
AQ = aphasia quotient. BNT = Boston Naming Test. HANA = Hopkins Action Naming Assessment. L:R CU ratio = ratio of left to right content units. Syll/CU = syllables per content unit. WAB–R = Western Aphasia Battery—Revised.


At the chronic time point, CU correlated with BNT (r = 0.57; P = 0.003) and HANA (r = 0.61; P = 0.002) scores, and the L:R CU ratio correlated with BNT scores (r = 0.41; P = 0.04). Importantly, acute picture description task measures also correlated with BNT (r = 0.55; P = 0.008) and HANA (r = 0.64; P = 0.002) scores at the chronic time point.


Findings from the current study support our initial hypotheses. Results showed that acute CU in a picture description task predicts communication recovery (change in CU/baseline CU) and picture description content in the chronic stages of recovery. Furthermore, our results support use of the Cookie Theft picture description task as an effective and efficient bedside measure for IWA in the acute setting (Agis et al, 2016). Cookie Theft picture description measures are associated with acute vascular and lobar damage in the LH language centers, and content on this measure correlates with other, more time-consuming language assessments that are typically administered by speech-language pathologists at the bedside. The number of CU produced acutely provides prognostic information that is critical to health care providers, as well as to family and caregivers, in an acute care setting.

When considered in the context of a previous investigation by our group (Agis et al, 2016), our results indicate that an extensive analysis of discourse is not needed for efficient and effective use of this tool clinically. In sum, by simply “checking off” CU produced by an individual within the first 48 hours after a stroke, clinicians across a variety of disciplines can determine aphasia presence and severity in order to make appropriate clinical decisions and educate patients and families regarding poststroke recovery.

As expected, communication recovery negatively correlated with acute CU, suggesting that patients who perform poorly on the picture description task have the most room for improvement as they recover. Moreover, CU recovery was best predicted by acute CU, age, and LH lesion volume; recovery was independently predicted by acute CU after adjusting for other variables. In a model including acute CU, age, LH lesion volume, and percent of damage to MCA, communication outcome was also predicted; however, in this model, MCA and LH lesion volumes independently predicted outcomes after controlling for other variables.

Quantitative analysis of this picture description contributes valuable information about the lesion volume and location. Primarily, acute CU provides information about lesion location, total lesion volume, and percent of damage to the left frontal and temporal lobes and left MCA territory. Chronic CU negatively correlated only with temporal lobe lesion volume. These results are in agreement with those of a previous study identifying the correlation of this measure with lesion volume (Agis et al, 2016). The negative correlation between acute CU and total lesion volume suggests that patients with larger infarcts demonstrate meager content in picture description compared with patients with smaller infarcts. Overall, correlations between lesions and behavioral outcomes at the acute stage of recovery showed that patients with larger lesions (particularly involving the left frontal and temporal lobes and the left MCA territory) have poorer language outcomes. While these results are not surprising, they show that this 1- to 2-minute bedside test is a sensitive measure of these relationships.

CU on a picture description task correlates with more cumbersome language batteries (eg, WAB–R) and confrontation naming tasks (eg, BNT and HANA) that are typically used in the acute setting. The results of our study support a previous finding (Agis et al, 2016) that an aphasia severity rating can be obtained from the Cookie Theft picture description task, which takes only a few minutes to administer. In addition to aphasia severity, CU, Syll/CU, and L:R CU ratio correlate with confrontation object and action naming. This positive correlation between the picture description task and naming assessments highlights the importance of CU in the acute phase because they are representative of the individual’s overall naming ability. However, this task has many advantages over traditional naming tests like the BNT and HANA. It takes less time to administer (<2 minutes) (Agis et al, 2016); is a better representation of communication function; and is less sensitive to education, socioeconomic status, and cultural differences. For example, many individuals in urban areas with low socioeconomic status are unfamiliar with the pictures of objects like a sphinx, trellis, and so on, that are assessed in the BNT. Likewise, English speakers from other parts of the United States or other countries may use different names than the “correct” response for some pictures on the BNT and the HANA. The picture description task allows CU to be expressed with a variety of words or phrases with the same meaning.

Syll/CU is a measure of communicative efficiency that has been shown to capture irrelevant CU in the spoken output of patients after an RH stroke (Barbieri and De Renzi, 1989). The negative correlation between communicative efficiency, as measured by Syll/CU and BNT scores at acute time point, indicates that lower BNT scores are associated with higher Syll/CU. This result is expected, as a higher Syll/CU score represents inefficient communication with more interjections, perseverations, and circumlocutions that contribute to the total syllables, but not to the CU measure.

The L:R CU ratio is an important calculation to consider because it quantifies hemispatial neglect in stroke patients. Previous studies have demonstrated that right neglect may be as common after an LH stroke as left neglect is after an RH stroke (Agis et al, 2016; Kleinman et al, 2007). In addition, a significant left:right bias in patients with an LH stroke as compared to controls has been previously shown (Agis et al, 2016). Our results showed that the acute L:R CU ratio correlates with HANA scores acutely and with BNT scores at the chronic time point.

To complement a previous investigation from our group (Agis et al, 2016), these results reveal the clinical importance of a picture description task for health care professionals across all rehabilitation settings, but particularly in the acute stage post stroke. Previous neuroimaging studies used lesion characteristics (Heiss et al, 2003; Saur et al, 2006; Watila and Balarabe, 2015) to contribute to our understanding of brain–behavior relationships; however, these investigations did not report on the natural discourse of individual language profiles, especially acutely. Such information is dependent on bedside testing and clinical assessments of language (LaPointe, 2011). The Cookie Theft picture description task is an efficient and reliable bedside assessment and is already part of the NIHSS, which is administered to all stroke patients in certified stroke centers in the United States and internationally. Therefore, incorporating it into the acute evaluation does not add additional administration time. For clinicians who are familiar with the NIHSS, the additional administration and scoring for CU requires just a few minutes and minimal training (Agis et al, 2016). Furthermore, because the training is quick (Agis et al, 2016), health care professionals who are unfamiliar with traditional language batteries or the NIHSS can effectively navigate this assessment tool to glean critical information.

Clinically, there is a need for a simple screening tool to expedite the identification of aphasia and to provide prognostic information that contributes to the individual’s acute medical needs and rehabilitation planning. Assessing narratives such as those produced as part of the Cookie Theft picture description task offers an ecologically valid option for evaluating communication (Ballard and Thompson, 1999). From a research perspective, screening tools would assist clinical researchers to screen for aphasia in large stroke populations (El Hachioui et al, 2017). In addition to providing critical linguistic and neuroanatomical information in the acute stages of stroke recovery, this tool can be implemented in outpatient settings or other evaluative sites for patients in the subacute and chronic phases of stroke recovery to determine aphasia severity and assess naming abilities, communication efficacy, and patterns of neglect.

Study Limitations

A primary limitation of this study is the limited sample of patients who were tested at the chronic time point. This may have resulted in inadequate power to detect additional associations at follow-up. Additionally, the WAB–R was not administered at the chronic time point for a variety of reasons. Namely, many of the patients we included had not yet returned for a follow-up assessment at the chronic time point. Finally, we chose to include picture naming, not the WAB–R, as an outcome measure because, although it is not a measure of overall aphasia severity, naming severity is typically consistent with overall aphasia severity, and naming is typically used as a primary outcome measure in studies about aphasia (Fridriksson et al, 2018). Future studies should incorporate additional measures of language at acute and chronic time points. Despite these limitations, the results of this study offer an efficient measure for assessing aphasia.


Our study suggests that health care professionals, with minimal task-specific training, can glean critical cognitive–linguistic information from a quick bedside language assessment. Results emphasize the efficacy of a picture description task to capture an accurate measure of aphasia severity and the percentage of damage to the language cortex. This finding is important because longer tests or batteries are cumbersome at bedside, and patients within the first 48 hours of stroke often cannot tolerate long language tests. This short assessment provides an efficient acute stroke assessment and prognostication of expressive communication improvements.


The authors gratefully acknowledge the participation and support of the study participants and their families.


Agis D, Goggins MB, Oishi K, et al. 2016. Picturing the size and site of stroke with an expanded National Institutes of Health Stroke Scale. Stroke. 47:1459–1465. doi:10.1161/STROKEAHA.115.012324
Ballard KJ, Thompson CK. 1999. Treatment and generalization of complex sentence production in agrammatism. J Speech Lang Hear Res. 42:690–707. doi:1092-4388/99/4203-0690
Barbieri C, De Renzi E. 1989. Patterns of neglect dissociation. Behav Neurol. 2:13–24. doi:10.3233/BEN-1989-2102
Benjamin EJ, Blaha MJ, Chiuve SE, et al. 2017. Heart disease and stroke statistics—2017 update: a report from the American Heart Association. Circulation. 135:e146–e603. doi:10.1161/CIR.0000000000000485
Brady MC, Kelly H, Godwin J, et al. 2016. Speech and language therapy for aphasia following stroke. Cochrane Database Syst Rev. 1:CD000425. doi:10.1002/14651858.CD000425.pub4
Breining BL, Tippett DC, Davis C, et al2015. Assessing dissociations of object and action naming in acute stroke. Paper presented at the Clinical Aphasiology Conference, May 2015, Monterey, California.
Breitenstein C, Grewe T, Flöel A, et al. 2017. Intensive speech and language therapy in patients with chronic aphasia after stroke: a randomised, open-label, blinded-endpoint, controlled trial in a health-care setting. Lancet. 389:1528–1538. doi:10.1016/S0140-6736(17)30067-3
Brott T, Adams HP Jr, Olinger CP, et al. 1989. Measurements of acute cerebral infarction: a clinical examination scale. Stroke. 20:864–870. doi:10.1161/01.STR.20.7.864
Ceritoglu C, Tang X, Chow M, et al. 2013. Computational analysis of LDDMM for brain mapping. Front Neurosci. 7:151. doi:10.3389/fnins.2013.00151
Craig HK, Hinckley JJ, Winkelseth M, et al. 1993. Quantifying connected speech samples of adults with chronic aphasia. Aphasiology. 7:155–163. doi:10.1080/02687039308249503
Dickey L, Kagan A, Lindsay MP, et al. 2010. Incidence and profile of inpatient stroke-induced aphasia in Ontario, Canada. Arch Phys Med Rehabil. 91:196–202. doi:10.1016/j.apmr.2009.09.020
Djamanakova A, Tang X, Li X, et al. 2014. Tools for multiple granularity analysis of brain MRI data for individualized image analysis. Neuroimage. 101:168–176. doi:10.1016/j.neuroimage.2014.06.046
El Hachioui H, Visch-Brink EG, de Lau LM, et al. 2017. Screening tests for aphasia in patients with stroke: a systematic review. J Neurol. 264:211–220. doi:10.1007/s00415-016-8170-8
Ellis C, Simpson AN, Bonilha H, et al. 2012. The one-year attributable cost of poststroke aphasia. Stroke. 43:1429–1431. doi:10.1161/STROKEAHA.111.647339
Enderby PM, Wood VA, Wade DT, et al. 1987. The Frenchay Aphasia Screening Test: a short, simple test for aphasia appropriate for non-specialists. Int Rehab Med. 8:166–170. doi:10.3109/03790798709166209
Engelter ST, Gostynski M, Papa S, et al. 2006. Epidemiology of aphasia attributable to first ischemic stroke: incidence, severity, fluency, etiology, and thrombolysis. Stroke. 37:1379–1384. doi:10.1161/01.STR.0000221815.64093.8c
Fisher A, Martin J, Srikusalanukul W, et al. 2014. Trends in stroke survival incidence rates in older Australians in the new millennium and forecasts into the future. J Stroke Cerebrovasc Dis. 23:759–770. doi:10.1016/j.jstrokecerebrovasdis.2013.06.035
Flamand-Roze C, Falissard B, Roze E, et al. 2011. Validation of a new language screening tool for patients with acute stroke: the Language Screening Test (LAST). Stroke. 42:1224–1229. doi:10.1161/STROKEAHA.110.609503
Franzén-Dahlin Å, Karlsson MR, Mejhert M, et al. 2010. Quality of life in chronic disease: a comparison between patients with heart failure and patients with aphasia after stroke. J Clin Nurs. 19:1855–1860. doi:10.1111/j.1365-2702.2010.03219.x
Fridriksson J, Rorden C, Elm J, et al. 2018. Transcranial direct current stimulation vs sham stimulation to treat aphasia after stroke: a randomized clinical trial. JAMA Neurol. 75:1470–1476. doi:10.1001/jamaneurol.2018.2287
Godecke E, Hird K, Lalor EE, et al. 2012. Very early poststroke aphasia therapy: a pilot randomized controlled efficacy trial. Int J Stroke. 7:635–644. doi:10.1111/j.1747-4949.2011.00631.x
Goodglass H, Kaplan E, Barresi B. 2001. Boston Diagnostic Aphasia Examination, 3rd ed. Baltimore, Maryland: Lippincott Williams & Wilkins.
Heiss WD, Thiel A, Kessler J, et al. 2003. Disturbance and recovery of language function: correlates in PET activation studies. NeuroImage. 20(suppl 1):S42–S49. doi:10.1016/j.neuroimage.2003.09.005
Hemsley G, Code C. 1996. Interactions between recovery in aphasia, emotional and psychosocial factors in subjects with aphasia, their significant others and speech pathologists. Disabil Rehabil. 18:567–584.
Hilari K, Wiggins RD, Roy P, et al. 2003. Predictors of health-related quality of life (HRQL) in people with chronic aphasia. Aphasiology. 17:365–381. doi:10.1080/02687030244000725
Johnson AF, Valachovic AM, George KPJohnson AF, Jacobson BH. 1998. Speech-language pathology practice in the acute care setting: a consultative approach. Medical Speech-language Pathology: A Practitioner’s Guide, 2nd ed. New York, New York: Thieme Medical Publishers; 96–130.
Kertesz A. 2007. Western Aphasia Battery—Revised. San Antonio, Texas: Psychological.
Kleinman JT, Newhart M, Davis C, et al. 2007. Right hemispatial neglect: frequency and characterization following acute left hemisphere stroke. Brain Cogn. 64:50–59.
LaPointe L. 2011. Handbook of Aphasia and Brain-based Cognitive-language Disorders. New York, New York: Thieme Medical Publishers.
Lazar RM, Speizer AE, Festa JR, et al. 2008. Variability in language recovery after first-time stroke. J Neurol Neurosurg Psychiatry. 79:530–534. doi:10.1136/jnnp.2007.122457
Lyden P, Claesson L, Havstad S, et al. 2004. Factor analysis of the National Institutes of Health Stroke Scale in patients with large strokes. Arch Neurol. 61:1677–1680. doi:10.1001/archneur.61.11.1677
Mori S, Oishi K, Jiang L, et al. 2008. Stereotaxic white matter atlas based on diffusion tensor imaging in an ICBM template. Neuroimage. 40:570–582. doi:10.1016/j.neuroimage.2007.12.035
Nouwens F, Visch-Brink EG, Van de Sandt-Koenderman MM, et al. 2015. Optimal timing of speech and language therapy for aphasia after stroke: more evidence needed. Expert Rev Neurother. 15:885–893. doi:10.1586/14737175.2015.1058161
Oishi K, Faria A, Jiang H, et al. 2009. Atlas-based whole brain white matter analysis using large deformation diffeomorphic metric mapping: application to normal elderly and Alzheimer’s disease participants. Neuroimage. 46:486–499. doi:10.1016/j.neuroimage.2009.01.002
Parr S, Byng S, Gilpin S, et al. 1997. Talking About Aphasia: Living With Loss of Language After Stroke. Buckingham, United Kingdom: Open University Press.
Pederson PM, Vinter K, Olsen TS. 2004. Aphasia after stroke: type, severity and prognosis. The Copenhagen Aphasia Study. Cerebrovasc Dis. 17:35–43. doi:10.1159/000073896
Plowman E, Hentz B, Ellis C Jr. 2012. Post-stroke aphasia prognosis: a review of patient-related and stroke-related factors. J Eval Clin Pract. 18:689–694. doi:10.1111/j.1365-2753.2011.01650.x
Rohde A, Worrall L, Godecke E, et al. 2018. Diagnosis of aphasia in stroke populations: a systematic review of language tests. PLoS One. 13:e0194143. doi:10.1371/journal.pone.0194143
Saur D, Lange R, Baumgaertner A, et al. 2006. Dynamics of language reorganization after stroke. Brain. 129(pt 6):1371–1384. doi:10.1093/brain/awl090
Watila MM, Balarabe SA. 2015. Factors predicting post-stroke aphasia recovery. J Neurol Sci. 352:12–18. doi:10.1016/j.jns.2015.03.020
Williams BW, Mack W, Henderson VW. 1989. Boston Naming Test in Alzheimer’s Disease. Neuropsychologia. 27:1073–1079. doi:10.1016/0028-3932(89)90186-3
Yorkston KM, Beukelman DR. 1980. An analysis of connected speech samples of aphasic and normal speakers. J Speech Hear Disord. 45:27–36. doi:10.1080/02699206.2017.1334092

stroke; aphasia; clinical assessment; language; rehabilitation

Supplemental Digital Content

Copyright © 2020 Wolters Kluwer Health, Inc. All rights reserved.