Evaluating the Contribution of Executive Functions to Language Tasks in Cognitively Demanding Contexts

Abstract

Purpose

The purpose of this study was to determine the degree to which the executive functions of attention switching and inhibition predicted performance on language subtests from the Temple Assessment of Language and (Verbal) Short-Term Memory in Aphasia (TALSA; N. Martin, Minkina, Kohen, &Kalinyak-Fliszar, 2018) across 3 interval conditions (no delay, 5-s delay, and 5-s filled delay), which was designed to add a memory and executive load to language tasks.

Method

This study was a post hoc experimental design. Participants included 27 people with aphasia who were administered 5 subtests from the TALSA (Naming, Word Repetition, Nonword Repetition, Category Judgment, and Rhyming Judgment), which were selected to evaluate input and output levels of processing in the 3 interval conditions listed above. Three executive tasks were administered to evaluate inhibition (Simon and Flanker tasks) and attention switching (number–letter shifting).

Results

Independent variables were proportion correct on each TALSA task in 3 separate time conditions, and predictor variables were efficiency on the Simon task and number–letter shifting task. Linear regression modeling was completed, which revealed that inhibition was a significant predictor of proportion correct for Word Repetition and Category Judgment in the 5-s filled interval condition.

Conclusions

Our findings indicate that inhibition plays a role in completing tasks that require lexical and/or semantic processing in cognitively demanding conditions. Attention switching was not a significant predictor for any task. These results are an important step toward creating methods to evaluate executive skills in the context of language production.

Supplemental Material

https://doi.org/10.23641/asha.9765107

Executive abilities are high-level cognitive skills that impact functional communication (Fridriksson, Nettles, Davis, Morrow, & Montgomery, 2006; Purdy, 2002) and are often thought of as “control processes” that regulate cognitive systems such as language (Lambon Ralph & Fillingham, 2007; Miyake, Emerson, & Friedman, 2000). As such, researchers recognize that impaired executive skills contribute to breakdown of language abilities in people with aphasia (PWA; Frankel, Penn, & Ormond-Brown, 2007; Kuzmina & Weekes, 2017; Hula & McNeil, 2008; Miyake, Emerson, et al., 2000; Ramsberger, 2005). Other reports indicate that strong executive skills can support the language system in PWA (Penn, Frankel, Watermeyer, & Russell, 2010; Ramsberger, 2005). Therefore, it is of clinical importance to develop assessment tools that are sensitive to breakdowns in executive functions and to determine how those breakdowns interact with language performance (Frankel et al., 2007; Kuzmina & Weekes, 2017; Miyake, Emerson, et al., 2000). The purpose of this study was to determine how well performance on executive tasks evaluating inhibition and attention-switching abilities predicted performance on traditional language assessments in varying conditions. Doing this accomplished two goals. First, we aimed to determine if nonverbal executive processes contribute to performance on traditional language assessment tasks. Second, and more specifically, we aimed to show that executive function abilities contributed to performance on language measures under conditions of added memory and executive load. Accomplishing these goals can be clinically meaningful as a way to provide insight into language assessment tasks that are sensitive to executive processing abilities.

Historically, aphasia has been viewed as a loss of language representations; however, current psycholinguistic models are more consistent with aphasia being a loss of access to and retrieval of those representations rather than a loss of the linguistic knowledge itself (Hula & McNeil, 2008; McNeil, 1982; McNeil & Pratt, 2001). This change in the characterization of aphasia has resulted in increased research that attempts to identify and evaluate the mechanisms that support linguistic access and retrieval, such as short-term memory, working memory, and executive functions. Short-term memory is hypothesized to be integral in maintaining activation of semantic, lexical, and phonological representations over time for processing of single words and utterances (N. Martin & Saffran, 1992, 1997; N. Martin, Saffran, & Dell, 1996). Executive functions are hypothesized to control component processes, such as inhibition and attention (Conway & Engle, 1994; Hula & McNeil, 2008; McNeil, Odell, & Tseng, 1991; Wright & Fergadiotis, 2012). Additionally, executive abilities, such as inhibition, have been identified as important for resolving competition during lexical selection (R. C. Martin & Allen, 2008; Schnur, Schwartz, Brecher, & Hodgson, 2006).

In line with current psycholinguistic models of lexical access/retrieval, the Temple Assessment of Language and (Verbal) Short-Term Memory in Aphasia (TALSA; N. Martin, Minkina, Kohen, & Kalinyak-Fliszar, 2018) is a newly developed test that aims to evaluate language and verbal short-term memory abilities in PWA. In addition to evaluating language performance, some subtests in the TALSA assess the ability to activate and maintain activation of language representations under conditions of increased memory load, implemented as a 5-s interval before a response or between two items to be compared. For example, the Rhyming Judgment subtest requires the participant to determine if two words or two nonwords rhyme or do not rhyme. In one condition, a stimulus is presented, which is then followed by a silent 5-s interval (5-s UF) before another stimulus is presented. In another condition, the 5-s interval is filled (5-s F) with the participant reading aloud randomly generated numbers appearing on the screen before a second stimulus is presented for comparison with the first. Whereas the unfilled 5-s interval adds memory load, the 5-s F condition adds both a memory and an executive component by requiring participants to switch attention and inhibit irrelevant stimuli.

N. Martin et al. (2018) reported normative data for the TALSA battery. Their outcomes revealed that PWA demonstrated differing patterns of accuracy in the 1-s UF condition and the 5-s UF delay condition (e.g., greater accuracy after a 5-s response delay for some and reduced accuracy for others), but average performance in the 5-s F condition was consistently worse than the 1-s UF condition or the 5-s UF condition. Additionally, PWA demonstrated a larger discrepancy in the 5-s F condition than a control group indicating that they were more sensitive to the increased executive demands of the 5-s F condition. Similar trends have been reported in dual-task studies, in which PWA perform significantly worse in dual-task conditions than non–dual-task conditions and demonstrate a greater accuracy “cost” than age-matched controls (Hula & McNeil, 2008; Murray, Holland, & Beeson, 1997, 1998).

During the 5-s F condition, the participant sees/hears a stimulus, reads aloud random numbers appearing on a screen, and then completes the linguistic task. Performance in the [5-s] F condition appears consistent with reports of dual-task effects in aphasia; however, this condition does not represent a pure dual task. Instead of requiring participants to complete two tasks at once with multiple switches between stimuli, the participant has to switch attention from the test stimuli to a new task (while potentially rehearsing the stimuli) and then return attention back to the test stimuli to respond. In the [5-s] F condition, the filled interval could disrupt participants' ability to maintain activation of the lexical item because rehearsal is suppressed by the distractor condition (reading random numbers aloud). Therefore, one of the aims of this study is to determine how executive functions contribute to the language performance of PWA in the different interval conditions (1-s UF, 5-s UF, 5-s F) of the TALSA.

Executive functions modulate the cognitive subprocesses required to complete complex cognitive tasks. They can be discussed and evaluated as basic executive skills (e.g., inhibition, attention switching, and working memory updating) or as more complex functions such as goal-directed behavior and planning. Research suggests that PWA often have some impairment of executive skills (Purdy, 2002), and due to the inherent interconnectedness of cognitive functions (Miyake, Emmerson, et al., 2000), it is clinically relevant to develop assessments that evaluate executive skills within a linguistic context to determine how they can impact language performance (McNeil et al., 2004). Inhibition, attention switching, and updating working memory are three basic executive skills that are important for language processing (Miyake, Emerson, et al., 2000). In this study, we focused on two of the functions described by Miyake, Friedman, et al. (2000): inhibition and attention switching. These two basic skills were selected based on the task demands of the 5-s F condition of the TALSA. During the 5-s F condition, participants must switch attention between test stimuli and distractors and inhibit irrelevant stimuli to respond to the test stimuli. Thus, we were interested in how performance on attention switching and inhibition tasks would predict performance on specific subtests of the TALSA.

Inhibition requires that a prepotent or automatic response be suppressed when necessary. For example, when completing a task designed to evaluate inhibition, such as the Stroop, the participant must override their automatic response to read a word so that they can name the color in which the word is written. Overriding the automatic response slows the response to the weaker stimuli, resulting in an interference effect (Cohen, Dunbar, & McClelland, 1990; Pompon, McNeil, Spencer, & Kendall, 2015; Wiener, Tabor Connor, & Obler, 2004). Attention switching was described by Monsell (1996) as the ability to shift attention between mental sets or tasks that require focus on a relevant task while an irrelevant set is not actively engaged. In typical attention-shifting tasks (e.g., number–letter shifting), the type of stimuli that is engaged changes over the trials, so that stimuli are focused on or suppressed based on task rules. Therefore, attention-switching tasks also require the ability to overcome proactive interference from prior competing tasks (Wylie & Allport, 2000) and negative priming, which can impair processing of stimuli that were previously suppressed (Waszak, Hommel, & Allport, 2005).

Research has shown that, compared to age-matched controls, PWA are more sensitive to task interference (R. C. Martin & Allen, 2008; Pompon et al., 2015), are less able to inhibit irrelevant stimuli (Pompon, et al., 2015), and have greater difficulty in switching attention between tasks (LaPointe & Erickson, 1991). Differences have also been reported in the executive abilities of people with fluent and nonfluent aphasia. Kuzmina and Weekes (2017) found that people with nonfluent aphasia performed worse across verbal (e.g., auditory control task) and nonverbal (e.g., Flanker task) cognitive tasks than people with fluent aphasia. Such findings shed light on how executive skills can support language function and the importance of including cognitive–linguistic skills in aphasia assessment.

The 5-s F condition was created to evaluate how memory and executive load impact language performance; however, there is growing evidence that executive skills also contribute to typical language assessment tasks. Kuzmina and Weekes (2017) found that nonverbal cognitive control was correlated with language comprehension (e.g., action comprehension and sentence comprehension) and verbal cognitive control was significantly correlated with naming ability (e.g., picture naming) in PWA. Allen, Martin, and Martin (2012) suggested that impaired verbal inhibition could be related to semantic language tasks, resulting in reduced ability to inhibit irrelevant representations, increased competition during lexical selection, and reduced ability to resolve ambiguity (e.g., when phrases or sentences are ambiguous, more competitors may become active). Additionally, complex executive functions (measured by Wisconsin Card Sorting Test and Tower of Hanoi; Heaton, Chelune, Talley, Kay, & Curtiss, 1993) were correlated with semantic processing tasks (e.g., Peabody Picture Vocabulary Test and Pyramids and Palm Trees; Simon, 1975) in PWA. These findings are relevant to the current study, which seeks to evaluate the contributions of attention switching and inhibition to language tasks from the TALSA battery with and without added executive load and provide further insight into the role of executive skills in language.

Existing executive function batteries such as the Behavioural Assessment of the Dysexecutive Syndrome (Wilson, Alderman, Burgess, Emslie, & Evans, 1996), the Delis–Kaplan Executive Function System (Delis, Kaplan, & Kramer, 2001), and the Functional Assessment of Verbal Reasoning and Executive Strategies (MacDonald, 1998) were created and normed on populations with traumatic brain injury, those with acquired brain injury, and neurotypical adults. Evaluating executive functions in PWA is often difficult due to the complex language demands of executive function batteries, including these tests and others (see Stroop test, verbal fluency tasks, and digit spans forward and backward). As a result, nonverbal assessments of executive function (e.g., Tower of London, Shallice, 1982; Wisconsin Card Sorting Task, Heaton et al., 1993; Comprehensive Trail-Making Test, Reynolds, 2002) are often implemented in clinical and research settings (see Keil & Kaszniak, 2002). Very few assessments have included executive function skills in batteries designed and normed for PWA (for an exception, see the Cognitive Linguistic Quick Test; Helm-Estabrooks, 2001). The long-term goal of the current research is to develop assessments that are sensitive to both language and executive function abilities in PWA. Doing this would provide insight into how these skills interact and influence each other during communication activities.

In addition to its influence on specific language tasks, evidence indicates that executive ability is related to functional communication (Fridriksson et al., 2006), discourse production (Frankel et al., 2007), conversation (Ramsberger, 2005), and rehabilitation outcomes (Nicholas, Sinotte, & Helm-Estabrooks, 2005; Ramsberger, 2005) in PWA. Thus, it is of interest to identify clinical measures that are sensitive to executive processing influences on language performance. In the current study, we were interested in determining how well nonverbal executive tasks evaluating inhibition and shifting predict performance in three different interval conditions (1-s UF, 5-s UF, and 5-s F) for the following TALSA subtests: Naming, Word Repetition, Nonword Repetition, Category Judgment, and Rhyming Judgment. These tasks were selected because they represent a range of the subtests on the TALSA and cover both input and output levels of processing. The specific goals of this study were to evaluate (a) if performance on attention switching and inhibition contributes significantly to performance in the 5-s filled condition and (b) if executive function ability contributes to language performance on the five TALSA subtests without the added executive load (1-s UF and 5-s UF vs. 5-s F condition). We hypothesized that measures of inhibition and attention shifting would predict performance in the 5-s F condition for all five tasks due to the interfering stimulus present in this condition. Although some reports suggest that nonverbal executive skills do contribute to language tasks, such as sentence comprehension (Kuzmina & Weekes, 2017) and conversational success (Ramsberger, 2005), we did not anticipate that the nonverbal executive tasks would significantly contribute to language performance in the unfilled conditions (1-s UF, 5-s UF) of the TALSA. The reason we did not expect executive contributions to be readily apparent on the unfilled test conditions we evaluated was due to their relative simplicity. In the unfilled conditions, each task was presented in a way that required only short-term maintenance of activation (for 1 or 5 s) before completing the linguistic operation. The filled context required maintenance of the activated representation while dealing with a second task (naming randomly generated numbers for 5 s), which we expected would require recruitment of executive functions. The long-term goal of this research is to develop assessment tasks that are sensitive to language and executive performance for PWA. The results from the current study will add evidence on how executive functions contribute to language performance in aphasia and the utility of the 5-s F condition for identifying the possible presence of executive function impairment in PWA.

Method

Design

This study was completed post hoc, with data collected from a larger ongoing project to develop and norm the TALSA battery (N. Martin et al., 2018).

Participants

All participants were diagnosed with aphasia due to left-hemisphere injury and were recruited from the greater Philadelphia area. Participants included 27 native English speakers diagnosed with aphasia based on the Western Aphasia Battery–Revised (WAB-R; Kertesz, 2006; see Table 1 for demographics). Average age of participants was 53.04 years, with a range of 32–72 years. Months postonset ranged from 6 to 301, with an average of 67.59. Aphasia quotients ranged from 33.6 to 94.3, and the following aphasia types were present: Broca's, conduction, transcortical sensory, transcortical motor, and anomic (diagnoses based on WAB-R). Data from five control participants are also included in the supplemental material for this article. Supplemental Material S1 includes demographic information, and Supplemental Materials S2 and S3 include language and executive function performance measures, respectively. Study procedures were approved by the Temple University Institutional Review Board, and all participants consented to participating in this research. Language and cognitive measures used for the current study were collected as part of an ongoing project. A global language test (WAB-R; Kertesz, 2006) and the executive function tasks were administered as background testing prior to administration of the TALSA.

Table 1.

Demographic information for participants.

Participant	Age	MPO	Educ	WAB	Aphasia type	Etiology
FS	53	12	12	70.6	Conduction	L intracerebral hemorrhage
QH	57	22	18	84.9	Anomic	L intracranial hemorrhage
UT	53	17	14	91	Anomic	L MCA CVA with basal ganglia involvement
KL1	59	30	14	92.4	Anomic	L thalamic CVA
SX	47	192	14	92.8	Anomic	L MCA CVA
EC	54	107	17	83.5	Anomic	L frontal parietal CVA, subcortical involvement
DD	60	70	16	55.6	Broca's	L frontal parietal CVA
SL	55	109	12	89	Anomic	L parietal AVM
GI	47	100	12	70	Anomic	L MCA CVA
HI	53	25	13	65.3	Conduction	L frontal CVA and old R corona radiata infarct
KL	34	13	14	93.3	Anomic	L CVA
EL	46	144	12	94.3	Anomic	L CVA
MI	56	72	19	71.5	Transcortical sensory	L CVA: posterior, temporal, occipital
CM	47	6	10	89.3	Anomic	L CVA
EC	61	301	18	62.5	Broca's	L CVA cerebral aneurysm
NH	67	48	12	49.9	Broca's	L CVA
UN	72	14	17	33.8	Broca's	L CVA
TB	40	69	12	92.2	Anomic	L MCA infarct and watershed area of L MCA/PCA
TB4	52	36	12	66.7	Conduction	L temporal infarct, L anterior thalamus and hypothalamus
LT	32	12	16	88.6	Conduction	L CVA
EH	46	64	13	81.4	Broca's	L CVA
KU	67	6	15	90.5	Anomic	L CVA
DC	49	23	14	33.6	Broca's	L CVA
KM	67	193	18	80.3	Transcortical motor	L CVA
CN	49	10	10	76.3	Broca's	L CVA
UP	48	37	14	88.4	Anomic	L CVA
FL	61	93	12	58.1	Broca's	L CVA and acute disseminated encephalomyelitis

Note. MPO = months post onset; Educ = years of education; WAB = Western Aphasia Battery; L = left; MCA = middle cerebral artery; CVA = cerebral vascular accident; AVM = arteriovenous malformation; R = right; PCA = posterior cerebral artery.

For the purposes of this project, five TALSA subtests were administered in three conditions (1-s UF, 5-s UF, and 5-s F). In the 1-s UF condition, the participant would see or hear a stimulus and then respond (name, compare a different stimuli, etc.) after 1-s (see Figure 1A). In the 5-s UF condition, participants would see/hear the stimuli and then respond after a silent 5-s delay (see Figure 1B). In the 5-s F condition, participants would see/hear the stimuli followed by a 5-s delay in which they would read randomly generated numbers (1–9) that appeared on a computer screen, after which they would respond to the stimuli (see Figure 1C). All TALSA tests were programmed and presented with E-Prime software. See N. Martin et al. (2018) for TALSA administration and stimuli details. The following TALSA subtests were administered:

Figure 1.

(A) Example of stimuli in the 1-s unfilled condition of Picture Naming. Participants see a picture and then name it after 1 s (no delay). (B) Example of stimuli in the 5-s unfilled condition of Word Repetition. After stimuli are presented, the participant waits 5 s and then repeats the word (5-s UF). (C) Example of stimuli in the 5-s filled condition of Category Judgment. Stimulus word is presented aurally and visually followed by a 5-s filled delay in which the participant reads random numbers off a screen. Then, a second stimulus is presented, and the participant determines if the two words are in the same category by hitting “yes” or “no” buttons on a keyboard.

1. Picture Naming: This task is intended to evaluate the semantic–lexical–phonological pathways required to name pictures. The stimuli include three sets of 30 pictures for the three different time interval conditions. The three sets of pictures were matched for number of syllables (1–3) and frequency (Pastizzo & Carbone, 2007). To do this, words were classified as high (> 25) or low (< 25) frequency and were equally divided among the three sets. Pictures appear for 2 s, followed by a cue (beep) to name the item. The cue occurs after a 1-s, 5-s, or 5-s filled interval.

2. Word and Nonword Repetition: In these two subtests, participants are asked to repeat real words that are one to three syllables and then, in a separate subtest, repeat nonwords that are phonotactically legal and derived from the real word stimuli. Word Repetition can require the participant to access phonological, lexical, and semantic levels of word representation, and Nonword Repetition taps more specifically into the phonological route. Word repetition stimuli included 15 high-frequency and high-imageability words (5 one-syllable, 5 two-syllable, and 5 three-syllable words). Nonwords were balanced and created by altering one to two phonemes from the word repetition stimuli. Stimuli were recorded in a human voice and presented via E-Prime software. Once participants heard the stimuli, they responded after hearing a beep. The cue occurs after a 1-s, 5-s, or 5-s filled interval.

3. Category Judgment: The purpose of the category judgment task is to evaluate a person's knowledge of category membership. There is a word (verbal semantics) and picture (conceptual semantics) version of the test, and we used the word version for this study. The test stimuli included 60 sets of word pairs (20 for each time condition) that fall into five possible categories (i.e., animals, transportation, vegetables, furniture, and musical instruments). During this test, the participant simultaneously hears and sees a word followed by the time interval (1-s UF, 5-s UF, 5-s F) condition of the test being administered. After the time interval, a second word is presented, and the participant has to immediately determine if the second word was in the same category as the first by hitting a key for yes or no.

4. Rhyming Judgment: This task evaluates input phonological processing by requiring participants to determine if two words rhyme. Stimuli included 20 word pairs (in each condition), 10 that rhyme and 10 that do not rhyme. All stimuli were one syllable and were presented via a recorded female voice. Participants were presented with a word aurally and then another word. The participant was then prompted to determine if the words rhymed by selecting “yes” (they rhyme) or “no” (they don't rhyme) on a keyboard. Based on the time interval, participants would hear the words with 1 s between (1-s UF), with 5 s between (5-s UF), or with a 5-s filled condition (5-s F) between words to be compared.

In addition, we used three tasks from the Turku Executive Function Battery (Soveri, Rodríguez-Fornells, & Laine, 2011): two that evaluated inhibition (Simon task and flanker task, Figure 2A and 2C respectively) and one that assessed attention shifting abilities (number–letter shifting task, Figure 2B).

Figure 2.

(A) Example of Simon task stimuli. (B) Example of number–letter task stimuli. (C) Example of flanker task stimuli.

Inhibition: Simon Task

During this task, a blue or red square appears on the right or left side of the computer screen. The participant is prompted to press a key on the right or left side of the keyboard, corresponding to the color of the square. While doing this, they must inhibit the location of the square. The test includes 30 congruent trials (e.g., when the square and response key are on the same side of the screen) and 30 incongruent trials (e.g., when the square and the correct response key are not on the same side), for a total of 60 trials. Each trial begins with a fixation cross in the center of the screen, which is present for 800 ms. The squares are present for 5,000 ms (or until the participant responds), followed by a blank screen for 1,000 ms. To interpret the results of this task, a change score was created by subtracting the reaction time (RT) on incongruent from the RT on congruent correct trials.

Inhibition: Flanker Task

During the flanker task, arrows are presented on a screen, and participants must press a button corresponding with the direction that the middle arrow is pointing (e.g., <<<<<, <<><<). When the direction of the middle arrow is the opposite direction of the other arrows (e.g., <<><<), the participant must respond to the direction of the middle arrow and inhibit a response to the other arrows (i.e., incongruent; Eriksen & Eriksen, 1974). The flanker task includes 100 trials with eight practice items. Stimuli were presented for 7,000 ms, and there was 1,000 ms between trials. Similar to the Simon task, RT for incongruent trials was subtracted from the RT for congruent tasks on only correct trials for data analysis.

Attention Switching: Number–Letter Shifting Task

During the number–letter shifting task, participants see alpha–numeric combinations that appear in a box above and below a line. When the stimuli are in a box above the line, the task is to determine if the number is even or odd. If the boxed stimuli are below the line, the task is to determine if the letter is a consonant or a vowel. The number–letter shifting task included three blocks. The first block (Block 1) includes 32 trials with equal numbers of even and odd numbers that all appear in the upper square (above the line), and participants have to respond if the number is even or odd. In the second block (Block 2), there are 32 trials, and all of the stimuli appear in a box under the line so the task was to determine if the letter is a consonant or a vowel. In the third block (Block 3), there are 32 switch trials and 48 nonswitch trials. Switch trials occur when there is a change from the stimuli appearing above or below the line (and making a decision about number or letter). If the combination is below the line, the participant's task is to determine if the letter is a consonant or a vowel. If the combination is above the line, the participant indicates if the number is even or odd (Rogers & Monsell, 1995). At the beginning of each trial, a fixation cross appears for 1,000 ms followed by two small boxes in the center of the screen, with one of them above a line and the other below a line. A number–letter combination appears in one of the boxes, and the stimuli remain on the screen for 3,000 ms or until the participant responds. The variable of interest for the number–letter shifting task was the difference between average RT for correct responses in Blocks 1 and 2 (no change trials) and the average RT for correct responses in Block 3 (change trial). Creating this “change” score allowed us to, as much as possible, isolate the time cost of the attention switching present in Trial 3. Only 25 of the 27 participants were able to complete this task (Participants NH and SX did not complete the number–letter shifting task).

Results

Language Performance

Participants were administered five subtests from the TALSA in three interval conditions (1-s UF, 5-s UF, and 5-s F), and performance on each subtest is described below (see Table 2).

Table 2.

Average performance on Temple Assessment of Language and (Verbal) Short-Term Memory in Aphasia tasks.

Subtest	M	SD	Range
Naming
1-s UF	0.717	0.239	0.07–1.0
5-s UF	0.686	0.254	0.03–0.97
5-s F	0.724	0.229	0.03–0.97
Word Repetition
1-s UF	0.831	0.205	0.33–1.0
5-s UF	0.831	0.215	0.0–1.0
5-s F	0.676	0.264	0.0–1.0
Nonword Repetition
1-s UF	0.545	0.296	0.0–0.93
5-s UF	0.419	0.257	0.0–0.93
5-s F	0.201	0.241	0.0–0.73
Category Judgment
1-s UF	0.943	0.063	0.8–1.0
5-s UF	0.929	0.084	0.70–1.0
5-s F	0.829	0.115	0.50–1.0
Rhyming Judgment
1-s UF	0.879	0.108	0.6–1.0
5-s UF	0.893	0.103	0.65–1.0
5-s F	0.826	0.093	0.6–0.95

Note. 1-s UF = 1-s unfilled; 5-s UF = 5-s unfilled; 5-s F = 5-s filled.

Picture Naming. In the 1-s UF condition, average proportion correct was .72 (SD = .24) for the 27 participants with aphasia. In the 5-s UF condition, the mean proportion correct was .69 (SD = .26). Unexpectedly, accuracy went up in the 5-s F condition with a mean proportion correct of .72 (SD = .23).

Word and Nonword Repetition. Mean performance on word repetition in the 1-s UF and 5-s UF conditions was the same (.83). Proportion correct was .68 (SD = .26) in the 5-s F condition. In nonword repetition, mean proportion correct was .55 (SD = .30) in 1-s UF, .42 (SD = .26) in 5-s UF, and .20 (SD = . 24) in the 5-s F interval.

Category Judgment. Mean proportion correct was .94 (SD = .06) in the 1-s UF condition and .93 (SD = .08) in the 5-s UF condition. Proportion correct was lower in the 5-s F condition (M = .83, SD = .12).

Rhyming Judgment. Mean proportion correct was similar in the 1-s UF condition (M = .88, SD = .11) and the 5-s UF condition (M = .89, SD = .10). In the 5-s F condition, average performance was .83 (SD = .09).

Data Analysis

Nonparametric Spearman rho correlation analysis was completed to determine if executive function variables were correlated prior to entering them into the regression model. The correlation analysis revealed that the Simon and flanker variables were significantly correlated (correlation coefficient = .448). Both the flanker and Simon variables are meant to capture inhibition; therefore, the flanker variable was not included in the regression analyses to avoid redundancy. The Simon RT difference score variable was selected over the flanker RT difference score due to previous evidence that performance on the Simon task was a significant predictor of language tasks with a working memory load (N. Martin, Kohen, Kalinyak-Fliszar, Soveri, & Laine, 2012). Importantly, the executive variables we used were change scores for the Simon task and the number–letter shifting task (see Table 3 for RT data). The variable for the Simon task was difference between average RT for correct trials in the congruent and incongruent conditions. The number–letter shifting variable was the difference between the average RT for correct responses in Blocks 1 and 2 (no change trials) and the average RT for correct responses in Block 3 (change trial). Calculating this change score allowed us to evaluate the executive cost of inhibiting a response (Simon) or switching attention (number–letter RT).

Table 3.

Means and standard deviations of reaction time (RT) for executive function measures.

Task	M (ms)	SD (ms)
Simon task
Congruent correct RT	1,104.32	276.66
Incongruent correct RT	1,182.02	315.55
Flanker task
Congruent correct RT	1,294.22	467.94
Incongruent correct RT	1,529.38	679.81
Number–letter shifting
RT in Trials 1 and 2	1,391.13	448.41
RT in Trial 3	2,472.67	845.83

Prior to entering variables into the linear regression models, data were normalized and transformed into z scores. Linear regression modeling using simultaneous entry of predictor variables was then completed. Dependent variables were proportion correct on Naming, Word Repetition, Nonword Repetition, Category Judgment, and Rhyming Judgment in three conditions (1-s UF, 5-s UF, and 5-s F). Predictor variables (independent variables) were the RT difference scores between incongruent and congruent correct trials on the Simon task and RT difference scores for the number–letter shifting task. Because of the nature of these two variables, we expected positive relationships between the dependent variables and the predictor variables. After running the linear regression models, t tests were used to determine if the individual contribution of each predictor was significant.

Naming

None of the models was significant (see Table 4).

Table 4.

Summary of regression analyses—Naming.

Dependent variable	R 2	Adj. R 2	SE	F value of ANOVA	p value of ANOVA	Intercept	Predictor variables	B	Standardized β for predictor	t statistic and p value of predictor
1-s UF	.19	.13	0.93	F(2, 22) = 2.72	.09	−.02	None	n/a	n/a	n/a
5-s UF	.14	.06	0.96	F(2, 22) = 1.78	.19	.00	None	n/a	n/a	n/a
5-s F	.17	.09	0.93	F(2, 22) = 2.31	.12	.02	None	n/a	n/a	n/a

Note. Adj. = adjusted; ANOVA = analysis of variance; 1-s UF = 1-s unfilled; n/a = not applicable; 5-s UF = 5-s unfilled; 5-s F = 5-s filled.

Word and Nonword Repetition

Word Repetition. For word repetition, the 5-s F model was significant with an adjusted R ² value of .217, F(2, 22) = 4.333, p = .026. The t statistic revealed that the Simon task was a significant contributor to the model with t = 2.380, p = .026. The models for word repetition in the 1-s UF and 5-s UF conditions were not significant (see Table 5).

Table 5.

Summary of regression analyses—Word and Nonword Repetition.

Dependent variable	R 2	Adj. R 2	SE	F value of ANOVA	p value of ANOVA	Intercept	Predictor variables	β	Standardized β	t statistic and p value
Word 1-s UF	.01	−.08	1.01	F(2, 22) = 0.144	.87	.05	None	n/a	n/a	n/a
Word 5-s UF	.07	−.01	1.04	F(2, 22) = 0.830	.45	−.00	None	n/a	n/a	n/a
Word 5-s F	.28	.22	0.88	F(2, 22) = 4.333	.03*	−.03	Simon	0.42	.43	t = 2.38, p = .03*
Nonword 1-s UF	.04	−.05	0.99	F(2, 22) = 0.407	.67	.06	None	n/a	n/a	n/a
Nonword 5-s UF	.02	−.08	1.01	F(2, 22) = 0.168	.85	.00	None	n/a	n/a	n/a
Nonword 5-s F	.09	.00	0.92	F(2, 22) = 1.037	.37	−.06	None	n/a	n/a	n/a

Note. Adj. = adjusted; ANOVA = analysis of variance; 1-s UF = 1-s unfilled; n/a = not applicable; 5-s UF = 5-s unfilled; 5-s F = 5-s filled.

^* p < .05.

Nonword Repetition. None of the models was significant (see Table 5).

Category Judgment

The model for Category Judgment in the 5-s F condition was significant with an adjusted R ² value of .321, F(2, 22) = 6.664, p = .005. The individual contribution of Simon RT was a significant predictor with t = 3.363 and p = .003. The 1-s UF and 5-s UF models were not significant (see Table 6)

Table 6.

Summary of regression analyses—Category Judgment.

Dependent variable	R 2	Adj. R 2	SE	F value of ANOVA	p value of ANOVA	Intercept	Predictor variables	β	Standardized β for predictor	t statistic and p value of predictor
1-s UF	.18	−.10	.93	F(2, 22) = 2.33	.12	.01	None	n/a	n/a	n/a
5-s UF	.15	.07	.91	F(2, 22) = 1.93	.17	.07	None	n/a	n/a	n/a
5-s F	.38	.32	.68	F(2, 22) = 6.66	.005*	.06	Simon RT	0.46	.57	t = 3.36, p = .003*

Note. Adj. = adjusted; ANOVA = analysis of variance; 1-s UF = 1-s unfilled; n/a = not applicable; 5-s UF = 5-s unfilled; 5-s F = 5-s filled; RT = reaction time.

^* p < .05.

Rhyming Judgment

None of the models was significant (see Table 7).

Table 7.

Summary of regression analyses—Rhyming Judgment.

Dependent variable	R 2	Adj. R 2	SE	F value of ANOVA	p value of ANOVA	Intercept	Predictor variables	β	Standardized β for predictor	t statistic and p value of predictor
1-s UF	.05	−.03	1.03	F(2, 22) = 0.66	.54	.03	None	n/a	n/a	n/a
5-s UF	.10	.02	.90	F(2, 22) = 1.22	.32	.06	None	n/a	n/a	n/a
5-s F	.18	.10	.89	F(2, 22) = 1.27	.30	.06	None	n/a	n/a	n/a

Note. Adj. = adjusted; ANOVA = analysis of variance; 1-s UF = 1-s unfilled; n/a = not applicable; 5-s UF = 5-s unfilled; 5-s F = 5-s filled.

Discussion

The aims of this study were to determine if performance on five language tasks from the TALSA was predicted by efficiency of inhibition and attention switching, especially in the 5-s F condition, which was designed to add memory and executive load to traditional language assessment tasks. Our results revealed partial support for our hypotheses. Performance on inhibition and/or attention switching predicted proportion correct in the 5-s F condition for two of the five TALSA subtests we evaluated. Executive skills did not predict performance in the 1-s UF or 5-s UF condition, in line with our predictions. Our aims, findings, and language performance results will be discussed below, followed by clinical implications, limitations, and future directions.

Aim 1: To Determine if Executive Function Ability Contributes to Language Performance on the Five TALSA Subtests Without the Added Executive Load (1-s UF and 5-s UF)

Our first aim was to determine the extent to which executive function ability predicted performance on traditional language tasks from the TALSA in the 1-s UF and 5-s UF conditions. TALSA subtests evaluating naming, word and nonword repetition, rhyming judgment, and category judgment were analyzed for this study. These subtests were selected because they encompass input and output levels of processing. Naming requires access to the semantic, lexical, and phonological routes to accurately retrieve and produce the lexical items of pictured objects. Word Repetition can also access phonological, lexical, and semantic levels as part of input and output processing of verbal stimuli. Nonword Repetition and Rhyming Judgment rely primarily on phonological abilities, and the Category Judgment (using words) subtest is meant to evaluate verbal semantic processing via categorization.

The findings from this study revealed that efficiency of nonverbal inhibition (Simon task) and attention switching (number–letter shifting) did not predict performance in unfilled interval (1-s UF, 5-s UF) conditions of the TALSA. This finding is consistent with our expectations that nonverbal executive functions would not predict performance in tasks that occur in simpler contexts that require just the short-term maintenance of a stimulus. Some reports indicate that nonverbal executive abilities can contribute to language performance, but those typically refer to complex language tasks, such as sentence comprehension, action comprehension (Kuzmina & Weekes, 2017), and conversation (Ramsberger, 2005), that are more likely to recruit executive functions than the single-word operations we evaluated.

Aim 2: Determine if Performance on Attention Switching and Inhibition Contributes Significantly to Performance in the 5-s Filled Condition

The purpose of the 5-s F condition of the TALSA was to determine how added memory and executive load impacted performance on traditional language assessment tasks. Our findings revealed that executive processes, specifically inhibition, were predictors of success in the 5-s F condition for two of the five subtests we evaluated, lending partial support for our hypotheses. Inhibition (Simon task) was a significant predictor for performance on the Category Judgment task and the Word Repetition task, signifying that efficient inhibition was an indicator of success in the 5-s filled condition of these tasks.

Attention switching was not a significant predictor for any of the tasks we evaluated. This finding was unexpected because, at face value, the attention-switching task appears to be closely aligned with the demands of the 5-s F condition, which requires participants to switch from linguistic stimuli to distractor and then back to the test stimuli. However, it is possible that the 5-s filled condition actually taps into an ability to efficiently complete multiple tasks similar to what is required by dual-task paradigms. Miyake, Friedman, et al. (2000) reported that attention switching did not significantly predict dual-task performance in a latent variable analysis and suggested that the ability to complete dual tasks may represent an executive function that is somewhat different than the three identified as important for language processing (inhibition, attention switching, working memory updating). Allen et al. (2012) also reported that attention shifting was not significantly correlated with language or short-term memory measures.

Inhibition predicted performance on two of the five subtests we evaluated: Word Repetition and Category Judgment. Previous literature has reported that inhibition/cognitive control contributes to language tasks such as sentence comprehension, action comprehension (Kuzmina & Weekes, 2017), lexical selection (Schnur et al., 2006), and semantic judgment tasks with an added memory load (N. Martin et al., 2012). Allen et al. (2012) also reported that complex executive function ability was correlated with semantic processing tasks, which they hypothesized could be due to the relational reasoning demands of semantic and executive tasks. In this study, inhibition contributed to tasks that required lexical and/or semantic-level processing. These trends indicate that executive skills play a role in tasks that require lexical semantic processing in the 5-s F condition. It is possible that the significant contribution of inhibition in these two models was related to lexical competition in tasks that had the potential to activate multiple semantic concepts (R. C. Martin & Allen, 2008; Schnur et al., 2006) and was not a significant component for tasks that primarily accessed phonological processing abilities. Although this finding was not considered in our hypotheses, it is consistent with other reports that semantic tasks rely on inhibitory processes to resolve competition and inhibit distractors (R. C. Martin & Allen, 2008). Additionally, the performance on the two phonological tasks we selected (Rhyming Judgment, Nonword Repetition) was not predicted by the executive tasks used in this study. This finding indicates that the phonological tasks, which do not necessarily require lexical or semantic-level processing, are not as demanding on the executive system. These findings are in line with those reported by N. Martin et al. (2012) in which inhibition was not a significant predictor of performance on a rhyming task with added memory load. Additional research is needed to explore this issue.

Language Performance on TALSA Subtests

Across the five TALSA subtests we evaluated in the three time intervals, our findings were mostly consistent with N. Martin et al. (2018). As a group, participants demonstrated similar performance in the 1-s UF and 5-s UF conditions across language subtests. In the 5-s F condition, accuracy decreased, except for the Naming subtest in which accuracy minimally increased (see Table 2). This finding was unexpected, but there are a few possible explanations. The first could be related to the test stimuli, which were different across the three conditions (1-s UF, 5-s UF, and 5-s F). It is possible that the 30 naming stimuli used in the 5-s F condition were easier than the other two conditions for this particular group of participants, although this trend was not observed with a larger group of participants (N. Martin et al., 2018). We anticipated that the distractor in the 5-s F condition would impact participants' ability to access/maintain the lexical items of the pictures. However, this did not appear to be the case in this study or in that of N. Martin et al. (2018). It is possible that the visual representation (picture) of the naming stimuli was maintained instead of the word's linguistic representation in the filled interval condition, which made this subtest less susceptible to the interference task. In the remaining subtests we evaluated (e.g., Category Judgment, Rhyming Judgment, Word and Nonword Repetition), the stimuli were presented aurally, increasing the likelihood of rehearsal to maintain activation of the lexical item and the potential for the interference condition to disrupt that rehearsal.

Limitations

The current study was limited by the post hoc design, which impacted our task selection but provided us with an opportunity to answer our preliminary questions. Additionally, finding tasks that represent the highly intertwined components of the executive system is difficult. In this study, we were limited to nonverbal tasks evaluating attention switching and inhibition. However, our findings provide support that inhibition contributes to performance of some language tasks, in the context of distraction (5-s F). Of the tasks examined in this study, those that engage input lexical and semantic representations (Category Judgment and Word Repetition) were predicted by performance on inhibition tasks. These results provide us with the groundwork to continue investigating how executive processes contribute to language in PWA.

The participants reported in this article are heterogeneous in relation to aphasia type, severity, lesion location, and time postonset. As a result, the group performance was variable, which impacted our results. The language tasks reported here are a part of the TALSA battery, which is designed to evaluate language and short-term memory across a range of PWA. We felt that, at this stage, it was important that our research be inclusive of all types of aphasia and ranges of severity due to the lack of evidence that aphasia severity is related to nonverbal cognitive functions, such as the executive tasks used in this study (Fridriksson et al., 2006; Helm-Estabrooks, 2002; van Mourik, Verschaeve, Boon, Paquier, & Harskamp, 1992). However, future research may require more restricted participant selection.

Clinical Implications

Evaluating language abilities in cognitively demanding conditions provides insight into how distracting contexts can impact communication. Research suggests that executive functions contribute to rehabilitation outcomes (Nicholas et al., 2005; Ramsberger, 2005). As such, evaluating executive abilities can be clinically meaningful in rehabilitation settings. However, within the current medical climate, assessment time is limited, and developing tasks that are sensitive to both language and executive impairment could provide clinicians with information on both abilities in a time efficient way. Additionally, it is important to evaluate executive functions and language together as opposed to isolating them, because the interaction of these abilities can impact performance on communication tasks. The current study evaluated how well performance on nonverbal executive tasks predicted accuracy in language tasks during different delay conditions with varying loads. Our results indicated that inhibition, but not attention switching, did contribute to some language tasks that require multiple levels of language processing (lexical and semantic). These findings support our next steps, which are to determine how well the 5-s F condition of the TALSA is able to identify executive impairments in PWA and its potential to be used as a task of verbal executive skills in aphasia assessment.

Future Directions

The results of this research are important because they add to the evidence on how executive tasks contribute to specific language abilities in PWA. Although our results are restricted to nonverbal executive tasks, our future research will evaluate verbal executive processing to determine how closely it relates to the 5-s F condition of the TALSA and language performance. Our long-term goals are to develop verbal executive tasks that are sensitive to both language and executive functioning abilities for PWA that can be used to evaluate how cognitively demanding contexts impact language performance.

Funding Statement

Research reported in this publication was supported by National Institute on Deafness and Other Communication Disorders Grants R01DC01924 and R21DC008782, awarded to Nadine Martin. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.