Effect of Digital Highlighting on Reading Comprehension Given Text-to-Speech Technology for People with Aphasia

Abstract

Background:

Many people with aphasia have a strong desire to participate in reading activities despite persistent reading challenges. Digital reading devices and text-to-speech (TTS) technology are increasing in popularity and have the potential to help people with aphasia. Systematic investigation of modifiable TTS features provides a means of exploring this potential.

Aims:

This study’s aim was to evaluate the effect of digital highlighting synchronised with TTS auditory and written output on reading comprehension by people with aphasia and to determine their highlighting preferences.

Methods & Procedures:

This work was registered with clinicaltrials.gov and assigned the clinical trial registry number 01446r prior to initiation of data collection. Twenty-five adults with aphasia read and listened to passages presented in three synchronised highlighting conditions: sentence highlighting, single word highlighting, and no highlighting. Participants answered comprehension questions, selected most and least preferred conditions, and provided feedback explaining highlighting preferences.

Outcome & Results:

Comprehension accuracy did not vary significantly across presentation conditions, but participants preferred either single word or sentence highlighting over no highlighting.

Conclusions:

Neither word nor sentence highlighting benefitted or hindered comprehension by people with aphasia as a group, but individual differences may occur. Clinicians should attend to personal preferences when implementing digital highlighting as a reading support strategy.

Digital reading devices and text-to-speech (TTS) technology are readily available and increasingly popular among people of all ages. For some adults with aphasia, using TTS technology facilitates comprehension of written material by presenting written and auditory content simultaneously (Knollman-Porter, Wallace, Brown, Hux, Hoagland, & Ruff, 2019; Wallace, Knollman-Porter, Brown, & Hux, 2019). An important aspect of this benefit is that TTS systems have presentation features (e.g., voice, rate, text highlighting) that are modifiable to match individual comprehension needs and preferences.

Rehabilitation specialists can modify traditional reading materials to meet aphasia-friendly formatting standards by providing abundant white space, using large and standard fonts, and simplifying sentence structure and vocabulary (Brennan, Worrall, & McKenna, 2005). Although helpful, making these modifications can be time-consuming for clinicians or caregivers, may not match the unique reading preferences of a person with aphasia, and may not provide adequate benefit to allow independent participation in daily reading activities. Advances in TTS technology provide an alternative method of individualizing written content presentation. One TTS feature—digital highlighting of single words or sentences—is the focus of the current research. Highlighting words or sentences simultaneously as they are spoken may improve attention to the text, thereby facilitating comprehension; however, researchers have yet to evaluate systematically this possibility with respect to people with aphasia.

Cognitive and Linguistic Contributions to Reading

Reading is a complex task requiring the coordination of multiple bottom-up and top-down cognitive and linguistic processes (Cutting, Materek, Cole, Levine, & Mahone, 2009; Hook & Crawford-Brooke, 2015; Rashotte, MacPhee, & Torgesen, 2001). Bottom-up processing involves identifying printed words through decoding or whole word recognition (Hook & Crawford-Brooke, 2015; Rashotte et al., 2001). In contrast, top-down processing promotes comprehension through constructing and extracting meaning from contextual and structural cues in the text (Cutting et al., 2009; Hook & Crawford-Brooke, 2015). Skilled readers are efficient in coordinating bottom-up and top-down processes to decode and comprehend text (Gibson, 2006; Johnson & Rayner, 2007; Rayner & Johnson, 2005; Samuels & Flor, 1997). They do this by retrieving and categorizing word meanings into grammatical units while engaging higher cognitive processes to understand text structure and self-monitor comprehension (Kendeou, van den Broek, Helder, & Karlsson, 2014; Oakhill & Cain, 2012; Oakhill & Yuill, 2013; Perfetti, Satfura, & Adlof, 2013). The result is creation of a cohesive and logical mental model of a text that is revised and refined as new information is presented (Jiang & Farquharson, 2018).

Cognitive processes such as attention, memory, and executive function are essential for successful processing of written content (Catts, Fey, Zhang, & Tomblin, 1999; 1990; Sesma, Mahone, Levine, Eason, & Cutting, 2009; Swanson, 1999). As such, researchers have documented a relation between poor attention and poor reading comprehension (Brock & Knapp, 1996; Martinussen & Mackenzie, 2015). This relation appears to reflect the negative effects of impaired allocation of cognitive resources on the integration of ideas and construction of mental models (Kendeou et al., 2014). Similarly, researchers have found that intact working memory is critical for coordinating and manipulating the construction of mental models while reading (Cain, Oakhill, & Bryant, 2004). Hence, people with good working memory tend to have better reading comprehension than people with poor working memory (Cain et al., 2004; Carretti, Borella, Cornoldi, & de Beni, 2009; Seigneuric & Ehrlich, 2005; Seigneuric, Ehrlich, Oakhill, & Yuill, 2000).

People with aphasia often have both attention and working memory deficits that may affect decoding and reading comprehension efforts (Friedmann & Gvion, 2003; Wright & Shisler, 2005). In accordance with resource allocation theory, the language processing and production challenges of people with aphasia reflect insufficient attention capacity and/or inefficient allocation of attentional resources (Hula & McNeil, 2008; McNeil, Odell, & Tseng, 1991; Murray, 1999). Some people with aphasia have attention problems that make inhibiting irrelevant stimuli difficult when reading (Martin & Allen, 2008). These attention problems affect reading comprehension by decreasing creation of a mental model and integration of ideas. Some researchers have evaluated the contributions of attention to reading efforts in individuals with aphasia and suggest that targeting attention deficits through direct attention training may lead to improvements in reading comprehension (Coelho, 2005; Sinotte & Coelho, 2007). Work by Lee and Sohlberg (2013) further supports this notion revealing that computer-based treatments aimed at directly remediating attention deficits (e.g., Attention Process Training) combined with facilitation of metacognition may increase attention and allocation of resources during reading activities for individuals with mild aphasia. Given the potential benefit of direct attention training to enhance reading efforts for individuals with aphasia, it stands to reason that compensatory strategies which enhance attention to written text may also support reading comprehension efforts.

Working memory deficits of people with aphasia may impede the updating and coordination of incoming information (Baddeley, 2000; Ericsson & Kintsch, 1995), again negatively affecting text comprehension. Because attention is fundamental to memory (Baddeley, Wilson & Watts, 1995; Huppert & Piercy, 1982; Posner & Petersen, 1990), impaired utilization of working memory when reading may stem from poor attention to the written text.

Benefits of Text Highlighting

Given that restorative interventions aimed at attention may facilitate increased reading abilities, it stands to reason that compensatory strategies supporting selective and sustained attention may also serve to increase visual attention to text, thereby increasing comprehension. Synchronizing digital text highlighting with auditory TTS output may benefit people with aphasia by improving attention to text during the reading process. In turn, increased attention may promote improved working memory to process and construct a mental model of presented material.

The intuitive appeal of this rationale has prompted research about using digitally highlighted TTS with children with developmental reading problems; however, findings from such studies have not always borne expected benefits. For example, Ayres, Langone, Douglas, Mead, and Bell (2008) studied the effects of highlighted TTS output on comprehension by students with moderate intellectual disabilities. Findings indicated that, for the group as a whole, digital highlighting did not improve reading comprehension significantly; however, some individual participants experienced improved comprehension. More recent research by Keelor (2017) confirmed the finding of no significant comprehension benefit for a group of students with reading difficulties when processing passages with digitally highlighted TTS output. However, to explore this finding further, Keelor, Creaghead, Silbert, Breit-Smith, and Horowitz-Kraus (2018) investigated the relation among measures of reading comprehension and measures of language and executive function. Results revealed a significant positive relation between language skills and reading comprehension given TTS presentation either with or without digital highlighting; no comprehension accuracy difference occurred between highlighted and non-highlighted TTS presentations. Thus, students with reading problems who benefitted most from TTS output were those for whom language functioning was a relative strength; the addition of digital highlighting did not further increase the benefit gained from TTS. When used as a study strategy, highlighting results in more time spent attending to text and subsequently primes cognitive processes associated with comprehending content (Ponce & Mayer, 2014).

Whether people with aphasia will experience the same or disparate outcomes when provided TTS output with synchronised highlighting remains to be determined. Content presentation characteristics such as color, size, and orientation have been frequently studied across age groups and provide highly salient clues to alter visual attention patterns (Wolfe & Horowitz, 2017). Highlighting features in TTS systems provide color cues for each presented word or sentence, and thus, may serve as an important feature to enhance visual attention to text for individuals with aphasia while simultaneously reducing cognitive demands.

Current Study

Researchers have documented comprehension benefits for some adults with aphasia given simultaneous written and auditory presentation (Knollman-Porter et al., 2019; Wallace et al., 2019) and have also established that people with aphasia benefit from aphasia-friendly formatting that uses large font sizes and has abundant white space (Brennan, Worrall, & McKenna, 2005). Given these findings, investigating the addition of synchronous highlighting to TTS is of interest as this type of visuographic text manipulation may benefit individuals with aphasia. Such manipulation may enhance attention to written words and increase comprehension beyond that evidenced with multimodal text presentation alone or with other visuographic modifications suggested as aspects of aphasia-friendly formatting. Additionally, people with aphasia may experience varying degrees of benefit when text highlighting occurs at word versus sentence levels. Therefore, the primary purpose of the current study was to examine the comprehension accuracy of adults with chronic aphasia when presented with simultaneous written and spoken TTS output of newspaper articles that varied in accordance with three conditions: no highlighting, single word highlighting, and sentence highlighting. A secondary purpose was to explore the highlighting preferences of people with aphasia. Because we know people with different aphasia profiles have distinct language skills and deficits (Helm-Estabrooks, Albert, Nicholas, 2014; DeDe, 2013), a third research question related to examining comprehension accuracy differences between people with fluent versus nonfluent aphasia. Specific research questions included:

• How does comprehension accuracy by people with aphasia vary when processing written content presented with TTS output and no highlighting, synchronised single word highlighting, or synchronised sentence highlighting?

• What are the highlighting preferences of people with aphasia when processing written content presented with TTS output?

• What differences exist between people with fluent versus nonfluent aphasia in the pattern of benefit experienced given different highlighting conditions or in their highlighting preferences when processing written content with TTS output?

Methods

Participants

Study participants included 14 males and 11 females with chronic aphasia. They ranged in age from 35 to 78 years (M = 59.20, SD = 11.15), were between ten and 266 months post-stroke or stable encephalopathy (M = 116.68, SD = 72.90), and had completed between 12 and 19 years of education (M = 14.76, SD = 2.30). All participants were right hand dominant prior to acquiring aphasia and spoke American English as their primary language. Participants passed a hearing screening confirming perception of 1000Hz, 2000Hz, and 4000Hz tones presented at 40dB in at least one ear. Two participants (i.e., Q and W) regularly wore hearing aids and, thus, did not complete hearing screenings; both had received audiology services within the past year and demonstrated adequate hearing of conversational speech with aids in place. All participants passed a visual acuity screening requiring accurate identification of their first name in each of ten sets of five names; all names appeared in black, 24-point, Times New Roman font on a laptop computer. Participant demographic information appears in Table 1.

Table 1.

Participant Demographic Information

Participant	Race	Gender	Age (years)	Education level (years)	Time post-onset (months)	Currently receiving SLP1 services	Employment status	Living status
A	White	Male	52	12	158	No	Retired	Independent
B	White	Female	74	16	240	No	Retired	Spouse
C	White	Female	64	12	241	No	Retired	Spouse
D	White	Female	73	12	88	No	Retired	Spouse
E	White	Female	35	18	51	No	Disability	Spouse
F	White	Male	50	16	135	No	Disability	Spouse
G	White	Female	52	16	131	No	Disability	Children
H	Black	Male	57	14	91	No	Disability	Parent
I	White	Female	60	12	83	No	Retired	Spouse
J	White	Male	47	12	118	No	Part-time	Parent
K	Black	Female	52	16	75	No	Disability	ECF2
L	White	Female	73	12	132	No	Retired	Partner
M	White	Male	71	18	266	No	Retired	Independent
N	White	Male	57	16	108	Yes	Retired	Spouse
O	White	Female	58	18	23	Yes	Volunteer	Independent
P	White	Female	49	16	93	Yes	Retired	Parent
Q	White	Male	78	18	68	Yes	Retired	Spouse
R	White	Male	73	15	38	No	Retired	Spouse
S	White	Male	65	16	23	Yes	Retired	Spouse
T	White	Male	47	16	185	Yes	Part-time	Independent
U	White	Female	44	12	25	No	Retired	Parent
V	White	Male	55	12	156	Yes	Volunteer	Family
W	Black	Male	71	16	196	No	Retired	Spouse
X	White	Male	60	12	183	Yes	Part-time	Independent
Y	White	Male	63	16	10	Yes	Retired	Spouse

¹SLP = speech-language pathology

²ECF = extended care facility.

We assessed participants’ language and cognitive abilities through standardized batteries and subtests administered prior to study initiation. Language assessments included the Aphasia Quotient portion of the Western Aphasia Battery – Revised (WAB-R; Kertesz, 2006); the Comprehension of Spoken Sentences, Spoken Paragraphs, and Written Sentences subtests of the Comprehensive Aphasia Test (CAT; Swinburn, Porter, & Howard, 2004); and the Paragraph Factual subtest of the Reading Comprehension Battery for Aphasia – 2^nd Edition (RCBA-2; LaPointe & Horner, 1998). To gather information about cognitive skills, we administered the Cognitive Linguistic Quick Test + (CLQT+; Helm-Estabrooks, 2017). Testing results for each participant appear in Table 2.

Table 2.

Participant Standardized Test Scores

			WAB-R	CLQT + Domains					CAT Subtests			RCBA
Participant	Aphasia type	Fluency type	Aphasia Quotient (100)	Attention (215)	Memory (185)	Executive Function (40)	Language (37)	Visuo- spatial (105)	Spoken Sentence (32)	Spoken Paragraph (4)	Written Sentence (32)	Paragraph Factual (10)
A	Anomic	Fluent	66.6	165	142	22	22	88	16	1	13	7
B	Broca's	Nonfluent	27	136	66	8	1	66	21	2	22	9
C	Conduction	Fluent	72.3	197	144	25	26	93	25	4	20	9
D	Transcortical Sensory	Fluent	67.1	156	106	13	12	72	22	3	24	7
E	Anomic	Fluent	92.2	203	160	27	27	99	23	4	22	10
F	Conduction	Fluent	69.6	200	144	32	26	102	28	4	26	9
G	Broca's	Nonfluent	47.9	182	108	24	12	95	20	1	22	9
H	Broca's	Nonfluent	67.8	192	146	24	23	94	25	3	28	10
I	Broca's	Nonfluent	64.6	176	162	22	27	86	16	4	24	10
J	Wernicke's	Fluent	64.2	188	126	25	23	91	19	2	20	7
K	Wernicke's	Fluent	49.2	124	109	9	25	53	16	2	16	7
L	Broca's	Nonfluent	55.5	125	92	18	17	68	19	4	20	9
M	Anomic	Fluent	81.3	176	145	19	27	29	26	4	26	9
N	Broca's	Nonfluent	34.9	164	77	16	7	76	16	2	12	3
O	Broca's	Nonfluent	86.2	134	136	17	28	52	19	4	17	10
P	Broca's	Nonfluent	48.3	184	88	19	11	92	16	2	14	9
Q	Conduction	Fluent	60.7	124	57	13	13.5	51	15	2	15	5
R	Broca's	Nonfluent	52.6	73	53	20	7	60	28	4	24	9
S	Global	Nonfluent	19.6	81	50	21	0	63	2	0	3	4
T	Broca's	Nonfluent	59.7	147	82	17	16	67	16	3	10	4
U	Broca's	Nonfluent	15.6	172	72	17	2	83	23	4	19	9
V	Conduction	Fluent	45	155	90	22	16	67	10	3	9	7
W	Anomic	Fluent	97.2	176	127	21	23	83	26	3	18	4
X	Broca's	Nonfluent	61.5	192	91	26	15	94	15	2	13	8
Y	Conduction	Fluent	80.6	201	158	31	29.5	99	18	3	24	10

WAB-R Aphasia Quotient scores ranged from 15.60 to 97.20 (M = 59.49, SD = 20.86). In accordance with cut-off scores provided in the WAB-R manual, two participants exhibited very severe aphasia (i.e., scores ≤ 25), six exhibited severe aphasia (i.e., scores between 26 and 50), 12 exhibited moderate aphasia (i.e., scores between 51 and 75), and four exhibited mild aphasia (i.e., scores between 76 and 93.8; Kertesz, 2006). The remaining participant (i.e., W) achieved an aphasia quotient above the cut-off for an aphasia diagnosis per WAB-R standards but had a clinical diagnosis of aphasia and displayed language challenges on other administered measures (see Table 2). Twelve participants displayed characteristics of fluent aphasia, and 13 participants displayed characteristics of non-fluent aphasia.

Materials

Study materials included written and auditory presentations of edited articles from newspapers, researcher-generated comprehension questions about each article, Kurzweil 3000^® software and ApowerREC^® screen recording software for stimulus development, laptop computers with E-Prime^® 3.0 software for stimulus presentation, and a Canon HF R700 or R800 digital video camera to record participant comments.

Newspaper articles.

Experimental stimuli included 39 edited newspaper articles (i.e., 36 experimental articles and three practice articles). All stimulus articles came from US newspapers available online and met the following criteria: (a) did not report news local to the region in which participants lived, (b) did not report national news, (c) did not report general knowledge information (e.g., World Series results), and (d) only used acronyms assumed to be comprehensible without explanation to all participants (e.g., United States = US). We purposefully selected general interest stories for use as experimental stimuli to reduce the potential influence of personal motivation for and background knowledge of story topics as a means of increasing internal validity.

We edited the newspaper articles to be of consistent length and readability for research purposes (Knollman-Porter et al., 2019; Hux et al., 2020). Furthermore, we sought to create stimuli consistent with materials people with aphasia were likely to encounter in real-world settings while maintaining feasible reading levels and lengths for people with language challenges. To determine an appropriate grade-level for experimental articles, we evaluated the Flesch-Kincaid grade-level (Flesch, 1948) of 69 randomly selected, unedited newspaper articles. These articles had grade-level equivalencies ranging from 6.5 to 13.0 (M = 9.93, SD = 1.69). Based on this information, we stipulated a grade equivalency between 9.0 and 11.0 for each stimulus article (M = 9.99, SD = .57). In addition, we edited stimulus articles to make each between 180 and 220 words in length (M = 202.28, SD = 12.56) which was consistent with short articles found in online newspapers. Editing articles to meet the length and grade level criteria involved removing complete sentences only; we made no single word or phrase substitutions or deletions and did not rearrange any sentences. Each edited article contained no more than two quotes. Finalized edited articles ranged from eight to 15 sentences in length (M = 10.72, SD = 1.86).

Recordings.

We used the Kurzweil 3000^® software to create digital video screen displays of experimental and practice articles. Kurzweil 3000^® is adaptive educational software designed to provide support to people with reading challenges. We chose this software because it allows a user to customize the type of highlighting (i.e., single word, sentence, or paragraph). Each article appeared on the system with 24-point, Times New Roman font centered on a white background; highlighting appeared as a yellow bar surrounding the desired text during highlighting conditions. Within the Kurzweil 3000^® settings, we set each article to be read by the computer-generated David voice from the Windows platform at a speed of 145 words per minute. We chose the David voice because people with aphasia can comprehend it at this speaking rate and prefer it to some other computer-generated voices (Hux et al., 2017). We chose the speaking rate of 145 words per minute because, as a group, people with aphasia prefer this rate to one that is substantially faster or slower (Hux et al., 2020). Of note, when creating stimuli for the sentence-length highlighting condition, the Kurzweil 3000^® highlighting function sometimes extended beyond a single sentence or stopped mid-sentence. To maintain consistency across stimuli and ensure highlighting of each sentence in its entirety, we added white periods between sentences on nine occasions and removed periods in abbreviations on eight occasions.

We captured three audio and video recordings of each article using ApowerREC^® software. The three recordings per article corresponded with three experimental conditions: no highlighting, synchronised single word highlighting, and synchronised sentence highlighting.

Comprehension questions.

We generated six basic comprehension questions for each article. All questions appeared as incomplete sentences with the final word or phrase deleted. Each question was factual in nature and used verbiage consistent with that of the article. Each of the six questions related to a piece of information located directly in the passage (i.e., semantic, factual information); no questions related to inferential information from a passage. According to the Structure Building Framework and Construction-Integration Model of Reading, such questions are designed to capture surface level understanding of text relying on literal and exact structural and semantic representation of content (e.g.,, comprehension of words, syntactic relations, and meaning of a sentence; Gernsbacher, 1997; Kintsch, 1998; Kucheria, Sohlberg, Yoon, Fickas, & Prideaux, 2018; Rupp, Ferne, & Choi, 2006). An example story and related comprehension question appear in the Appendix.

Researchers have utilized a variety of question types when evaluating reading comprehension of adults with acquired brain injury (e.g., free recall, sentence verification; e.g., Sohlberg, Griffiths, and Fickas, 2014). We chose to evaluate basic reading comprehension using a multiple-choice question format as it relies heavily on content recognition (i.e., cued recall) rather than free content recall. Multiple-choice question format serves to reduce memory demands and thus, we could ensure that the task was more highly related to reading comprehension abilities than to memory impairments that participants may have experienced. Furthermore, the use of multiple-choice questions was necessary given the expressive language deficits demonstrated by the majority of participants in this study; written choice strategies are widely used with individuals with aphasia as a means of supporting language (Garrett & Beukelman, 1995). We presented response choices associated with each question in multiple choice format in which a vertical list appearing after the sentence stem contained the target response and three foil responses. Target responses occurred in each of the four answer positions approximately the same number of times.

We completed a dependency analysis following procedures used in previously published work (Hux et al., 2020). To complete the analysis, we recruited 75 adults (44 females) between 21 and 91 years of age (M = 50.89, SD = 17.11) with a minimum of a high school education. All individuals who completed dependency analyses confirmed through self-report no history of learning disability, cognitive impairment, or acquired neurological deficit (e.g., seizure disorder). This validation procedure served to ensure that comprehension questions reflected the factual story content and could not be answered correctly (i.e., were not guessable) without reading the passage. Participants responded with less than 40% accuracy to each comprehension question without prior reading of the corresponding passage.

Stimulus presentation.

We used 17-inch Dell touchscreen laptop computers to present experimental stimuli using E-Prime^® software. Use of E-Prime^® 3.0 allowed for controlled and systematic stimulus presentation along with data logging of response accuracy as participants progressed through experimental sessions. To ensure each participant saw all 36 experimental articles in a randomised order, we created three stimulus sets with four articles in each condition (e.g., four articles with single word highlighting, four articles with full sentence highlighting, and four articles with no highlighting), totaling 12 articles per set. We programmed E-Prime^® to present the four articles within each condition in a randomised order across participants. During each of the three sessions, participants completed the experimental task with a different E-Prime^® stimulus set, allowing for exposure to all stimulus articles across the course of the study. Presentation of passage text within E-Prime allowed for simultaneous highlighting of each subsequent word or sentence (dependent on condition) with audio output.

Preference selection materials.

We used the practice articles from each condition embedded into PowerPoint^© slides to facilitate participants’ expression of their presentation condition preferences.

Procedures

Institutional Review Boards at both universities at which data collection occurred approved the study materials and procedures prior to participant recruitment and data collection. Prior to completion of this study, this work was registered with clinicaltrials.gov and assigned the clinical trial registry number 01446r. Given that study participants exhibited chronic aphasia and some had testing available from prior participation in research studies or clinical services, we accessed existing records when possible and did not repeat testing. If assessment results were more than one year old or were not available, we administered the WAB-R during the first session or during an additional testing session, as needed.

Stimulus articles and comprehension question presentations.

A researcher read aloud the written instructions displayed on the laptop screen prior to beginning each experimental session. A practice article matching each subsequent condition appeared on the screen allowing a participant to adjust the volume of the computer-generated speech output, practise using the touchscreen to respond to comprehension questions, and demonstrate understanding of task instructions. After responding to practice article questions, participants touched on a green Go icon to advance the presentation software to the first experimental article. The associated auditory recording and text highlighting, as appropriate for a given condition, began one second after appearance of the article text. As soon as the auditory rendition of the article ended, the written text disappeared, and the first comprehension question appeared on the screen.

Participants could not refer back to an article when answering comprehension questions. A researcher read each question and the corresponding response options aloud unless a participant expressed a desire to read them independently. Participants selected an answer by touching a circle to the left of one of the possible responses. They could change an answer multiple times before progressing to the next question; however, once they indicated either verbally or gesturally a final response selection, the researcher progressed the screen to the next question, and participants could not go back to a previous question. As desired, participants could request breaks between articles or conditions.

Each condition presentation began with a practice article followed by three practice questions. After the practise, participants performed the experimental task with four articles presented with highlighting matching the practice condition. The process repeated for each condition, resulting in exposure to 12 experimental articles and responses to the associated comprehension questions during a single session. Participants completed two additional sessions using the same procedures within two weeks of the initial session. Thus, in total, participants responded to question sets about 36 unique articles, 12 in each of the three experimental conditions.

Participant preference.

Participants ranked each of the three highlighting conditions from most to least preferred after completing all three experimental sessions. We prompted participants with questions to determine their condition preference ranking (e.g., Which did you like best? Which did you like least?). Participants provided verbal and gestural (e.g., thumbs up) explanations, as possible given speech and language deficits, for their rankings.

Data Analysis

Variables of interest included: (a) comprehension accuracy for each highlighting condition and (b) participants’ preference rankings of conditions.

Comprehension accuracy.

We performed a repeated measures analysis of variance (ANOVA) to identify any significant accuracy difference among the single word, sentence, and no highlighting conditions. Within each condition, we also computed independent sample t-tests to identify significant accuracy differences between participants with fluent versus nonfluent aphasia.

Participant condition preference.

We determined condition preference by totaling the number of participants who selected each highlighting condition as their most and least preferred at the end of the third experimental session. We transcribed all participant verbalisations and salient gestures/facial expressions regarding highlighting preferences as well as any explanations participants offered about reading strategies used to facilitate experimental task performance. We examined the rationales provided by participants for each condition and coded responses as expressing solely positive, solely negative, mixed reviews, or indifferent feelings.

Results

Comprehension Accuracy

Participants achieved the highest average comprehension accuracy score in the single word highlighting condition (range: 38.89 – 90.28, M = 68.07, SD = 13.82), followed by the full sentence highlighting condition (range: 38.89 – 88.89, M = 67.02, SD = 14.50) and no highlighting condition (range: 40.28 – 93.06, M = 65.5, SD = 14.82). Figure 1 represents both individual and full group data. The majority of participants did not show individual differences across conditions; however, some minimal differences in performance were noted. Computation of a repeated measures ANOVA revealed no significant accuracy difference across conditions, F_(2,48) = 1.311, p = .274.

Figure 1.

Participants’ percent accuracy across conditions.

Participants with fluent aphasia (n = 12) achieved greatest comprehension accuracy in the single word highlighting condition, followed by no highlighting and sentence highlighting conditions (see Table 3). Participants with nonfluent aphasia (n = 13) achieved greatest comprehension accuracy in the sentence highlighting condition, followed by single word highlighting and no highlighting conditions. We computed independent-samples t-tests to compare comprehension accuracy between the fluent and nonfluent participant groups for each highlighting condition. The tests revealed no significant difference between groups for any highlighting conditions.

Table 3.

Participant Accuracy Across Fluency Groups

Condition	Participant group	Range	Mean	Standard deviation	t value (23)	p value
Word	Nonfluent	38.89-84.72	66.24	13.52	−.683	.501
	Fluent	51.39-90.28	70.06	14.46
Sentence	Nonfluent	43.06-83.33	66.88	16.99	−.049	.962
	Fluent	38.89-88.89	67.17	12.47
None	Nonfluent	40.28-77.78	63.68	12.29	−.633	.533
	Fluent	43.06-93.06	67.48	17.50

Condition Preferences and Perceptions

Participants ranked highlighting conditions according to preference following completion of the third experimental session (See Table 4). As a group, participants expressed greatest preference for the sentence highlighting condition, followed by word highlighting, and no highlighting. The least preferred condition was the no highlighting condition. Visual inspection of the data did not reveal patterns relative to participant condition preference and accuracy in a given condition.

Table 4.

Number of Participants with Fluent versus Nonfluent Aphasia Preferring each Highlighting Condition

Participant group	Most preferred condition			Least preferred condition
	Word	Sentence	None	Word	Sentence	None
Fluent	3	7	2	4	2	6
Nonfluent	6	6	1	3	1	9
Combined groups	9	13	3	7	3	15

Note. Bold font indicates condition most and least preferred by the greatest number of participants.

Participants with fluent aphasia (n = 12) expressed greatest preference for the sentence highlighting condition and least preference for the no highlighting condition. Participants with nonfluent aphasia (n = 13) expressed equal preference for the sentence and word highlighting conditions; they expressed least preference for the no highlighting condition.

Participant Explanations for Highlighting Preferences

Participants provided rationales, as possible given expressive language limitations, for their highlighting preference selections. Differing opinions emerged regarding the benefits and detriments associated with each highlighting condition.

Sentence highlighting.

Participants most frequently preferred the sentence highlighting condition; however, comments regarding the condition differed across participants. Fourteen participants expressed solely positive comments about sentence highlighting, five expressed solely negative comments, two expressed mixed opinions, and four expressed indifference. Positive comments focused on how sentence highlighting increased the ability to follow the auditory presentation of the story and comprehend the text. In contrast, distractibility and a decreased ability to keep up with the highlighted text were provided as rationales for not liking sentence highlighting. Representative participant comments about the sentence highlighting condition appear in Figure 2.

Figure 2.

Participant opinions about sentence highlighting.

Single word highlighting.

Several participants made positive comments regarding the single word highlighting condition even though it was not most preferred by the group as a whole. Participants’ opinions differed about benefits and detriments of single word highlighting. Twelve participants expressed exclusively positive comments regarding the condition, seven expressed exclusively negative comments, two expressed mixed opinions, and four expressed feelings of indifference. Positive comments centered on how single word highlighting increased the ability to follow the auditory presentation of the story and comprehend the text. In contrast, distractibility and feelings of agitation were rationales for not liking sentence highlighting. Representative participant comments about the single word highlighting condition appear in Figure 3.

Figure 3.

Participant opinions about single word highlighting.

No highlighting.

Participants expressed mixed opinions about the no highlighting condition even though the group as a whole least preferred it. Eight participants expressed only positive comments about no highlighting, ten expressed only negative comments, two expressed mixed opinions, and five expressed feelings of indifference. Positive comments focused on how no highlighting was most familiar, similar to reading at home, and aided text comprehension. In contrast, difficulty following along with the digitised voice and decreased text comprehension were rationales for not preferring the no highlighting condition. Participant comments regarding the no highlighting condition appear in Figure 4.

Figure 4.

Participant opinions about the no highlighting condition.

Discussion

Impaired comprehension abilities after aphasia are often chronic and affect a variety of activities (Holland, 2007; Parr 1995). Still, people with aphasia have a strong desire to regain reading skills and engage with written materials conveying personally relevant content (Knollman-Porter, Wallace, Hux, Brown, & Long, 2015). Consequently, exploring methods of facilitating independent processing of written content by people with aphasia is a clinically relevant goal.

The purpose of this study was to evaluate effects on the comprehension of people with aphasia by adding synchronized digital highlighting to TTS output. We also examined the preferences of people with aphasia about highlighting options and any differences between people with fluent versus nonfluent aphasia. The findings may inform clinicians about potential benefits and detriments associated with implementing highlighting features available via TTS technology and the importance of considering both comprehension accuracy and personal preference when selecting support strategies.

Comprehension Accuracy

No significant accuracy difference emerged across highlighting conditions for participants as a whole nor between the fluent and nonfluent aphasia groups. These findings are consistent with those from literature regarding the use of highlighting with children with developmental reading challenges (Ayres et al., 2008; Keelor, 2017; Keelor et al., 2018). Hence, our assumption that highlighting would mediate the attention problems of people with aphasia and thereby improve reading comprehension was not borne out by the findings. Furthermore, this pattern occurred even though most participants liked one or both highlighting condition more than the no highlighting condition and believed it helped in tracking and comprehending the text.

Extant literature suggests that dual modality presentation (i.e., having simultaneous access to written and auditory content) improves comprehension for some people with aphasia (Knollman-Porter et al., 2019; Wallace et al., 2019). Further supplementing input with digital highlighting does not appear either to help or hinder comprehension, however. One possible reason for this is that the text highlighting occurred at a rate inconsistent with the speed with which the person with aphasia could process the content. Support for this contention comes from participant comments about feeling forced to focus on all highlighted words rather than the words to which they believed they needed to attend or not being able to keep up with the highlighting. By forcing tracking at too fast or too slow a speed, the highlighting may have sharpened attention to written content yet did not provide an advantage to comprehension. If this is true, only by allowing individualized control of the progression of highlighting would a person with aphasia experience a benefit.

Another possibility for the lack of significant comprehension differences across conditions is that the distraction associated with sequential text highlighting over-taxed the limited cognitive resources available to people with aphasia. Several researchers (e.g., Hula & McNeil, 2008; McNeil, Odell, & Tseng, 1991; Murray, 1999) have speculated that people with aphasia either have reduced capacity or inefficient allocation of cognitive resources. If the addition of highlighting created a situation in which the amount of visual stimulation exceeded the processing limits of a person with aphasia, no additional benefit beyond that provided by dual modalities would be expected; in fact, a decrease in comprehension might occur in some instances. Again, some participant comments provided support for this possibility, especially with respect to the distracting nature of single word highlighting.

Highlighting Preference

Understanding reading behaviors and preferences post-stroke from the perspective of the person with aphasia is necessary given the complexity of reading and the vast impact reading can have on participation (Webster, Morris, Malone, & Howard, 2020). Participants, as a group, most preferred either single word or sentence highlighting (n = 22) over no highlighting (n = 3); the majority (n = 15) least preferred the no highlighting condition. Respecting the preferences of people with aphasia when recommending support strategies is imperative, especially given the lack of any significant comprehension differences across highlighting conditions. Previous researchers have emphasized the importance of considering the opinions of people with aphasia to identify their goals and close the gap between what they want, need, and currently can do (Byng & Duchan, 2005; Duchan & Black, 2001; Helm-Estabrooks, Albert, & Nicholas, 2014; LPAA Project Group, 2000). Selecting and implementing desired support strategies increases the likelihood of consistent use, which in turn leads to additional practise and greater likelihood of achieving functional gains (Dalemans, De Witte, Wade, & Van Den Heuvel, 2008; Wepman, 1953; Worrall et al., 2011). Thus, given that significant comprehension differences were not revealed in the current group of study participants, clinicians may consider deferring to client preference when deciding whether to implement highlighting as a comprehension support strategy.

Clinical Application

The lack of significant comprehension benefit, yet strong preference either for single word or sentence highlighting, complicates the provision of strategy use recommendations for people with aphasia. Clinicians may want to evaluate comprehension accuracy with and without digital highlighting before implementing it as a support strategy. If assessment results reveal no significant benefit or detriment to reading comprehension, deferring to a person’s preference is appropriate. However, because synchronised digital text highlighting differs substantially from traditional reading, exposing people to this support on multiple occasions before soliciting a preference and implementing a specific option may be wise.

Limitations

This study included participants with varying aphasia severities and profiles. Because the number of participants in each aphasia severity group varied substantially, and statistical power for across group comparisons was limited, we chose not to complete statistical analyses regarding the impact of aphasia severity on comprehension and preference. Similarly, the small number of study participants who achieved scores below the cut-off for the Comprehension of Spoken Paragraphs subtest of the Comprehensive Aphasia Test did not allow for completion of statistical analyses regarding the relation between standardized testing performance and experimental accuracy. Studies with more homogenous participant populations or larger numbers of participants exhibiting various aphasia profiles may reveal differences for whom digital text highlighting is or is not beneficial.

We selected newspaper articles to ensure that experimental stimuli simulated functional, real-world texts readily available to people with aphasia. However, to enhance internal validity, we modified the stimuli to ensure comparable length and reading grade level across articles. This resulted in the study materials being representative of some, but not all, newspaper articles. To present these articles in each experimental condition, we used Kurzweil 3000^®; on some occasions, we had to alter article formatting to ensure appropriate highlighting using this system. Presentation of unedited newspaper articles and use of different digital reading software might alter study results both in terms of reading comprehension and preference.

Future Directions

We administered a broad cognitive assessment to each participant prior to study initiation; however, additional, in-depth evaluation of specific cognitive abilities (e.g., attention, working memory) was not performed. Thus, we could not consider fully the contribution of cognitive impairments on participants’ reading comprehension abilities when presented with TTS and digital highlighting. Administering an assessment battery with more sensitivity to key cognitive processes may have provided a means of analyzing more fully possible contributors to the obtained results. Future researchers may wish to consider this when exploring further this type of reading support. Also, assessing the attentional abilities of people with aphasia can be challenging because language and cognitive deficits influence one another (Connor & Fucetola, 2011; Helm-Estabrooks, 2002; Keil & Kaszniak, 2002); hence, nonverbal assessments may be preferable to verbal ones. The Conners’ Continuous Performance Test (CPT-2; Conners, 2002) and the Test of Variables of Attention (TOVA; Leark, Dupuy, Greenberg, Corman, & Kindschi, 1998-1999) are examples of attention assessments that do not require oral responses and may be suitable for people with aphasia (Connor & Fucetola, 2011).

An important consideration when studying the effects of digital highlighting on reading comprehension is the extent to which people actually track highlighted words or sentences as they are spoken aloud. Using eye-tracking technology may prove helpful in monitoring the eye movements of a person with aphasia while reading, as researchers have done with other populations (e.g., Ikeshita, Yamaguchi, Morioka, Yamazoe, 2018). Results from studies implementing such procedures may provide clinicians with further direction about optimal methods of supporting people with aphasia when engaged in reading tasks.

Conclusion

The results did not yield a significant difference in the comprehension accuracy of people with aphasia across the three highlighting conditions. However, the majority of participants preferred one of the highlighting conditions over the no highlighting condition, even though some found the highlighting to be distracting or irritating. Overall, the findings suggest that clinicians should individualize the use of supports when creating a reading comprehension plan for a person with aphasia, taking into consideration both comprehension accuracy and individual preference.

Appendix

Example Article

Award Winning Tree in Ludington

The Official Tree Board of Ludington, Michigan presented a certificate to Noah and Amelia Robinson for two noble fir trees on the front lawn of their residence on North Haas Street. The trees are the first within the city limits to enter the Prestigious Tree Program administered by the tree board.

The noble fir trees were estimated to be planted in September of 1907 by William and Helen Davis. The 112-year-old trees stand nearly 80 feet tall and each have a diameter of about three and a half feet round. The front lawn where the trees are planted is known as the most photographed spot in Ludington.

The Prestigious Tree Program is open to all species of trees. To enter the program, trees must be healthy trees located within the United States and must meet one or more of the tree board’s criteria. Ludington’s noble fir trees were admitted to the program for being greater than 50 years old as determined by the opinion of an Arborist and for being considered Landmark Trees. Residents are encouraged to submit their trees to the Ludington Tree Board as a part of the “Keep Ludington Beautiful” Project. The project is dedicated to engaging all citizens in every aspect of keeping Ludington clean and beautiful.

The Official Tree Board of Ludington presented Noah and Amelia Robinson a certificate for:

1. Noble fir trees

2. Sycamore Maple trees

3. English Walnut trees

4. White Oak trees