The effectiveness of cognitive flexibility training on reasoning, attention, and memory in stroke patients

(1)

The Effectiveness of Cognitive Flexibility Training on Reasoning,

Attention, and Memory in Stroke Patients

Roos Heijerman

Department of Psychology: Brain and Cognition, University of Amsterdam

student ID: 10456392

e-mail address : roos.heijerman@student.uva.nl internal supervisor: Renate van de Ven

(2)

Patients recovering from stroke very often experience cognitive complaints. Knowing which factors influence early recovery and which cognitive functions show persisting deficits in the long-term might result in a better prognostication and interventions might be directed towards diminishing the rate of cognitive impairment. Cognitive ‘brain’ training is often thought to be effective in improving cognitive impairment. However, scientific proof of these effects in untrained tasks (i.e. transfer effects) seems scarce. In addition, factors thought to be associated with clinical outcome after stroke, are lesion location and baseline functioning. To gain more insight in cognitive functioning of patients after stroke, this study focuses on reasoning, attention and memory performance after cognitive training. The present study investigated the effectiveness of a cognitive flexibility training, as compared with an active control group. Furthermore, it was explored whether lesion location and baseline performance determine stroke outcome. Results revealed no significant difference in performance on any of the three cognitive domains between the cognitive flexibility training and the active control group. Furthermore, no proof was found for the influence of lesion location and baseline performance on stroke outcome. Because of low statistical power, it is suggested for future research to investigate the effectiveness of the cognitive flexibility training in a bigger sample.

Keywords: stroke, cognitive flexibility training, attention, reasoning, memory, lesion location, hemisphere, baseline performance

(4)

1. Introduction

Cognitive deficits after stroke are very common. Neuropsychological problems often occurring after stroke include reduced capacity for new learning, slowed information processing and disruptions in complex integrative functions (Millis, et al., 2001; Ruttan, Martin, Liu, Colella, & Green, 2008). Even years after stroke, these neuropsychological deficits can cause problems in functioning in daily life activities, disruptions of social networks and difficulties in returning to work (Ruttan, et al., 2008). Therefore, it is important to gain insight in the long-term course of cognitive functioning after stroke. Knowing which factors are associated with early recovery and what cognitive functions show persisting deficits in the long-term might result in better prognostication and interventions might be directed towards diminishing the rate of cognitive impairment (Bentley, et al., 2014; Millis, et al., 2001; Ruttan, et al., 2008).

One promising way of ameliorating the cognitive impairments in healthy individuals, is by using cognitive ‘brain’ training games. Although scientific evidence concerning the improvement of general mental capacity after brain training is sparse (Jaeggi, Buschkuehl, Jonides, & Shah, 2011; Nouchi, et al., 2012), there is accumulating evidence that certain cognitive trainings are effective (Jaeggi, et al., 2011). In a large sample size of healthy individuals (N>1000), Owen et al. (2010) showed significant cognitive improvement after a six week training period. However, their study provided no evidence that training-related improvements may generalize to other tasks involving similar cognitive functions (Owen, et al., 2010). In another study, Jaeggi et al. (2011) investigated these ‘transfer’ effects for a specific working memory training in children, and found that those who considerably improved on the training task also showed a performance increase on untrained fluid intelligence tasks, in contrast to the control group (Jaeggi, et al., 2011). Furthermore, in an elderly sample, Nouchi et al. (2012) demonstrated transfer effects of a four week cognitive training on the improvement of executive functions and processing speed, but not for global cognitive status and attention. However, their results need to be replicated in larger samples, and the long-term effects and relevance in daily-life experience are uncertain (Nouchi, et al., 2012).

Similarly, the evidence for transfer effects after cognitive training in stroke patients is also somewhat inconsistent. In a study by Pyun et al. (2009), the authors evaluated a twelve week individualized home programme for improvement in cognitive functioning. Their programme focussed on attention, memory and executive and higher cognitive functions of stroke patients.

(5)

Although participants showed significant improvement in Instrumental Activity of Daily Living (IADL) performance, domain-specific cognitive neuropsychological assessment showed improvement on some tests but these improvements were not significant (Pyun, et al., 2009). Middleton and colleagues (1991) compared two different computer based cognitive rehabilitation therapies in patients with acquired brain injury, focussing on reasoning/logical thinking and attention/memory, respectively. Their results revealed that both treatment groups showed significant improvements on cognitive functioning. However, they failed to demonstrate any unique and specific treatment effects (Middleton, Lambert, & Seggar, 1991). Other studies specifically investigated the effectiveness of a computerized working memory training in patients with stroke and acquired brain injury. They revealed significant improvements on non-trained working memory and attention tasks after training as compared with a control group (Lundqvist, Grundström, Samuelsson, & Rönnberg, 2010; Westerberg, et al., 2007). Furthermore, they also suggest the training effect was transferred to working-memory related activities in daily life (Lundqvist, et al., 2010). De Luca et al. (2014) compared the effectivity of a pc-cognitive training in addition to conventional treatment with a control group which only performed conventional rehabilitation in stroke patients. The training focussed on cognitive abilities as attention, language, memory and executive functions. The study found greater cognitive improvement in the experimental group as compared to the control group, with significant differences for nearly all neuropsychological tests performed. However, their results need to be replicated in larger and more homogenous samples (De Luca, et al., 2014).

Although these results seem promising, factors influencing the effectivity of cognitive ‘brain’ training games are still unclear. In their review, Green and Bavelier (2008) propose that the manner in which task difficulty is progressed, the motivational state of the learner, the type of feedback provided by the training, and variability are important factors determining the effectiveness of training (Green & Bavelier, 2008). In accordance, Buitenweg, Murre and Ridderinkhof (2012) also proposed the importance of tailoring the progression of the level of task difficulty, and the variability of the tasks. Furthermore, they postulate that flexibility (i.e. task switching) and novelty are important factors determining the effectiveness of the training (Buitenweg, Murre, & Ridderinkhof, 2012).

Compared to healthy individuals, stroke patients might cope with additional factors influencing their cognitive functioning. Location of the lesion is thought to be such a factor. Because functional brain responses after stroke are anatomically specific (Grefkes & Fink, 2011), lesion anatomy is expected to be an important predictor of recovery after stroke

(6)

(Bentley, et al., 2014). In their study, Cheng et al. (2014) found that injury to areas of the corticospinal tract often lead to motor impairments. These impairments were related to the degree of disability or dependence in daily life activities, and correlated with higher functional impairment (Cheng, et al., 2014). Furthermore, they found that damage to the right inferior parietal lobe was associated with specific components of neglect, which is believed to strongly influence functional outcome after stroke. Similarly, lesions in the left superior and middle temporal cortex were associated with aphasia, which has been demonstrated to be a risk factor for poor clinical outcome (Cheng, et al., 2014). Their findings illustrate that the distinct influence of structural brain lesions on recovery involves specific brain regions depending on hemisphere side. Bentley et al. (2014) also investigated the influence of lesion location on early recovery. They found that high recovery rates were principally found with lesions more superficial and anterior. Low recovery rates were associated with lesions to bilateral parietal, right insula and brainstem (Bentley, et al., 2014). Their results show that lesion anatomy is a relatively strong determinant of early recovery, and they furthermore argue that it is unlikely that lesion size can explain their results because their model covaried for this factor (Bentley, et al., 2014).

In accordance, in a review by Rosso and Samson (2014) it is stated that site of the lesion predicts outcome better than size did. They point out that the initial volume of the lesioned area is only important in the sense that the larger the volume, the more likely it is that important and strategic areas are infarcted (Rosso & Samson, 2014). Some studies do argue that lesion size is significantly correlated with clinical outcome after stroke (Khan, Goddeau, Zhang, Moonis, & Henninger, 2014; Lev, et al., 2001) Their data support the notion that the reason for overall good outcome in untreated patients is likely related to smaller clot burden leading to spontaneous recanalization resulting in smaller infarct volumes. However, Voght et al. (2012) do point out that although functions in the brain that are being compromised by stroke are indeed located in specific areas, the relative importance of one function lost is not well correlated to the volume of the tissue.

Similarly, Bentley et al. (2014) suggest that associations of lesion location with clinical outcome are likely to be confounded by effects of lesion location on baseline. Studies investigating training effects in healthy elderly and in patients with stroke indeed found differences in improvement for different baseline levels of cognitive functioning (Desmond, Moroney, Sano, & Stern, 1996; Johansson & Tornmalm, 2012; Peretz, et al., 2011). High baseline was correlated with less improvement after training as compared to low baseline scores. These studies found that individuals suffering from cognitive impairment, as reflected

(7)

by lower baseline score, show greater benefits from cognitive training (Desmond, et al., 1996; Johansson & Tornmalm, 2012; Peretz, et al., 2011). Peretz et al. (2011) ascribes this finding in terms of the ceiling effect, in which patients with low baseline levels have more room for improvement, whereas patients with high baseline scores have less opportunity to improve because they already have high scores.

Taken together, it is still not clear whether cognitive training in stroke patients – and in healthy individuals – is effective in the sense that it transfers to untrained cognitive abilities. Furthermore, it is unknown to what extend factors as lesion anatomy and baseline performance influence improvement after cognitive training in stroke patients. This study focuses on the performance of stroke patients on three cognitive domains (i.e. reasoning, attention and memory) before and after following an online cognitive or active control training. Specifically, it was investigated whether a twelve-week online cognitive flexibility training improves reasoning, attention and memory performance in stroke patients. Furthermore, independent of the effectiveness of the training, this study also tried to identify baseline factors influencing a good clinical outcome in stroke patients. Thus, a second aim of this study was to explore whether baseline performance (high versus low) and lesion location (left, right or both hemispheres) influence outcome after a twelve-week cognitive or active control training (irrespective of any group differences) in stroke patients.

Concerning the main research question, it is hypothesized that the twelve-week cognitive flexibility training would improve the scores on the selected cognitive domains of reasoning, attention and memory in stroke patients more than those in patients who followed an active control training for the same amount of time. Furthermore, it was hypothesized that low baseline cognitive scores would be associated with greater improvement on the selected cognitive domains after training, whereas high baseline scores would result in less improvement. In addition, because functional brain responses after stroke are anatomically specific, it was expected that lesion anatomy would influence stroke outcome.

2. Methods

2.1. Participants

All participants (N = 51: male = 32) included in this study were patients between 30 and 80 years old (M =59.29, SD = 7.88), who had had a stroke within the last five years and received rehabilitation as inpatient or outpatient. Participants were recruited through rehabilitation centres (e.g. Heliomare and Reade) or by advertisements in newsletters of Dutch associations

(8)

for stroke patients. Patients who were interested in participating and fulfilled inclusion criteria received written information about the study. Inclusion criteria were presence of objective cognitive dysfunctions as demonstrated by neuropsychological assessment, or by a neurologist, physiatrist, psychologist, or another experienced clinician; at study entry patients must still have cognitive complaints; patients had to be mentally fit (as defined by the Telephone Interview Cognitive Status (TICS) score > 26) and physically fit enough to be able to complete twelve weeks of training; patients had to be able to understand the training instructions. Furthermore, participants must have daily access to a computer with sound and Internet connection, and must have some basic computer skills (e.g. opening emails, clicking on links, being able to smoothly use the mouse). Exclusion criteria were neurodegenerative disease; epilepsy; serious psychiatric disorder (e.g. a history of multiple psychotic episodes, acute psychosis, acute major depression); diagnosed learning disability (e.g. mental retardation); any other disease resulting in severe cognitive dysfunction; alcohol or drug dependency; severe colour blindness, aphasia, or neglect; vision or hearing problems.

Anonymity was guaranteed and patients could withdraw from the study at any moment without any consequences. The research was approved by the Medical Ethical Committee of the Vrije Universiteit from Amsterdam.

2.2. Procedure

Patients, who were interested in participating in the study, were asked to fill in an online screening form and they were contacted by phone for a quick cognitive screening. Also, they were asked for permission to access their rehabilitation and hospital files for information about their type of stroke. When participants were still interested and fulfilled the inclusion criteria, they were randomly assigned to either the cognitive flexibility training, or the active control training. Thus, the study was a double blind, randomized controlled intervention study with an active control group. Participants were asked to come to the University of Amsterdam (UvA) twice. Between these visits, participants trained online at home for 30 minutes a day, five times a week for twelve weeks. On both visits the same face-to-face neurological assessment was administered, and participants had to complete some online neurological tasks. In addition, during their first visit participants received a more detailed instruction about the training and signed the informed consent. Furthermore, they were shown how to log on to the website and were able to practice with the games. During the training, participants were regularly contacted and were asked about their progression, motivation, whether they experienced any problems or had any questions.

(9)

2.3. Materials

2.3.1. Instruments and outcome measures

The cognitive training was completed at home on a computer with Internet access, via the website of uva.braingymmer.com. Before the start, as well as after completion, of the twelve week cognitive flexibility or active control training several neuropsychological tests were administered. The following three cognitive domains were assessed: reasoning, attention, and (working) memory. For each of the cognitive domains, composite scores were calculated by averaging the z-scores of two tests. Z-scores for pre- and post-training were calculated according to the unit-weighted method, and by using the means and standard deviations from baseline scores.

2.3.1a. Reasoning

For measuring reasoning abilities, composite scores were calculated on Raven’s Coloured Progressive Matrices (CPM) and the Tower of London (ToL). In Raven’s CPM participants are presented with a matrix of geometric figures, with one entry missing. Subjects have to complete the design by correctly selecting the missing part from a number of options (Raven, 1989). Correlations with other intelligence test are reported to be moderate to high, between r =.5 and .9 (Raven, 1989) and test-retest reliability is reported between r =.59 and .79 (Raven et al. 1998 as cited in (Bouma, Mulder, Lindeboom, & Schmand, 2012). In the present study, 20 matrices were presented to the participants, and the primary outcome measure was the total number of correctly answered items. This score was converted into a z-score.

The Tower of London (Shallice, 1982) requires subjects to rearrange coloured beads on rods from a given start position to a target position, in a limited number of moves. Test-retest reliability of the ToL has been reported to be fairly strong, between r =.70 and .81 (Culbertson & Zillmer, 1998; Lemay, Bedard, Rouleau, & Tremblay, 2004), and moderate correlations with other measures of executive functions are reported, between r =.25 and .48 (Anderson, Anderson, & Lajoie, 1996). In this study, an online version of the Tower of London was used, of which reliability and validity was unknown. Participants were presented with 10 problems ranging from 2- to 7-move test problem configurations, and were given a maximum time of two minutes to solve the problem before starting with the next one. Subjects were given no repeated trials for failed problems, thus maintaining task novelty. The primary outcome measures of the Tower of London in this study were (1) the total number of solved problems within 2 minutes, (2) the problem-solving time, that is, the interval (in milliseconds) between the presentation of a problem to its solution, and (3) number of moves needed for all trials

(10)

together. Administration and scoring methods for the Tower of London vary among studies (Anderson, et al., 1996; Culbertson & Zillmer, 1998; Lemay, et al., 2004; Shallice, 1982). Berg and Byrd (2002) point out the importance of multiple measures of ToL performance, and recommend that at minimum there should be a time related measure and a ‘success or accuracy’ measure (Berg & Byrd, 2002). Success or accuracy can be defined as the percentage solved within the time limit regardless of the number of extra moves (Culbertson & Zillmer, 1998). Therefore, in the present study, z-scores were first calculated for the three primary outcomes measures. Because a higher total problem solving time and a higher total number of moves means worse performance, these z-scores were reversed such that higher scores reflect better performance. Next, these three z-scores were averaged to represent the outcome for the ToL.

2.3.1b. Attention

To measure attention, a composite score of the Pased Auditory Serial Addition Test (PASAT) and Trail Making Test (TMT) was calculated. In the PASAT (Gronwall, 1977) subjects are presented with two series of 61 digits at a speed of 3.2 and 2.8 seconds, and are asked to add up the number they just heard with the immediately preceding number. The PASAT is reported to have a high internal validity with Cronbach’s α =.96, and it has a high test-retest reliability of r =.93-.97 (Egan, 1988; McCaffrey, 1995 as cited in (Sherman, Strauss, & Spellacy, 1997). Furthermore, the PASAT loads highly (r =.75) on the attention/concentration factor as identified in previous studies of the Wechsler Adult Intelligence Scale-Revised (WAIS-R), a well known measure to test intelligence in adults and older adolescents (Crawford, Obonsawin, & Allan, 1998). In the present study, participants were required to make 60 summations for both the subtest presented at 3.2 and 2.8 seconds. Thus, the maximum score for both subtests was 120. The percentage of total correct responses was calculated, and converted into z-scores.

The Delis-Kaplan Executive Function System (D-KEFS) TMT requires the subjects to draw lines sequentially, connecting numbers, letters, or alternating between numbers and letters in ascending and alphabetical order, respectively. A moderate to strong split-half reliability has been reported, between r =.50 and .80, and is has a test-retest reliability of r =.77 (Homack, Lee, & Riccio, 2005). Internal validity of the D-KEFS has been reported fairly low, between r =.10 and .3. Correlations with other measures of executive function are moderate to high, r = .31 to .59 (Strauss, Sherman, & Spreen, 2006). The outcome measure of the D-KEFS TMT was the time (in seconds) required to complete each version of the task.

(11)

For each task version a scaled score was calculated, based on norm scores from a population sample of 1750 subjects corrected for age, gender, education, race/ethnicity and geographic region, according to the D-KEFS scoring manual. This study used the alternating condition, corrected for the number and letter condition, to compute the z-scores from.

2.3.1c. Memory

Working memory was measured with the Rey Auditory Verbal Learning Task (RAVLT) and the Corsi-block tapping test, of which a composite score was calculated. In the RAVLT, participants are presented with 15 words and have to remember as many of them as they can. Participants are given five immediate trials, and one delayed recall trial after 20 minutes. It has a high internal consistency with a Cronbach’s α of .80-.83 and a high parallel form reliability of r =.72-.80 (Bouma, et al., 2012). Primary outcome measures for the RAVLT used in this study were (1) the number of correctly remembered words for each of the five trials, (2) the total number of remembered words across the five trials (Sum Score), (3) the maximum number of correct remembered words (Maximum Score), and (4) the number of correctly remembered words after 30 minutes (Delayed Recall). To take into account participants’ remembering abilities as well as their memory capacity, a Forgetting Rate for each participant was first calculated by subtracting the Delayed Recall Score from Maximum Score, and dividing it by the Maximum Score. This score was transformed to a z-score. Because a higher Forgetting Rates represents worse performance, this z-score was first reversed such that higher scores reflect better performance. Second, the Sum Score was also transformed to a z-score. These two z-scores were then averaged to represent the outcome for the RAVLT.

In the Corsi-block test, participants have to remember a sequence of blocks indicated by the experimenter. This study used an online version of the Corsi-block test, in which the blocks were highlighted sequentially and participants have to click on the blocks in the same order immediately after the sequence was presented. Sequences vary from 2 to 8 blocks per trial, and participants were given two attempts per trial. The task was terminated when participants failed to reproduce the correct sequence on both trials. Test-retest reliability of the Corsi-block test has been reported to be moderate, r =.38 (Sala, Gray, Baddeley, Allamano, & Wilson, 1999), but no reliability or validity of the online version is available. The primary outcome measures for the Corsi-block test in this study were (1) the longest completed sequence, (2) number of correct trials, and (3) the number of incorrect trials. Previous studies report many variations in scoring methods, including the percentage correctly repeated

(12)

sequences, percentage correct by position, the span limit, use of partial credit, and combining forward and backward span scores. In a study by Kessels, van Zandvoort, Postma, Kappelle and de Haan (2000) they compared two scoring methods, that is, the Block Span and the Total Score. Block Span was defined as the length of the last correctly repeated sequence and the Total Score was the product of the Block Span and the number of correctly repeated sequences until the test was discontinued. According to the authors this latter score takes into account the performance on both trials of an equal length, and is considered more reliable than the Block Span alone. Indeed, their results indicated that the Total Score was the more sensitive measure (Kessels, van Zandvoort, Postma, Kappelle, & de Haan, 2000). Therefore, in this study a z-score was calculated for the longest completed sequence. The percentage of correct repeated sequences was also converted to a z-score. The percentage rather than the absolute number of correctly repeated sequences was used, because the number of total presented sequences differed among participants. These two z-scores were averaged to represent the z-score for each participant on the Corsi-block test.

2.3.2 Training

The intervention training consisted of ten tasks, selected to train the cognitive domains of attention, reasoning and working memory. The cognitive flexibility training group trained on ten tasks per day for approximately three minutes per task, and tasks were presented directly after each other to assure cognitive flexibility. The active control training condition trained on only three out of four selected tasks per day, and switched between tasks approximately every ten minutes. Importantly, the training of the intervention group was adaptive to their level of performance, whereas in the active control group the level of the tasks was adapted according to a predefined schedule. A more detailed description of the training games and schedule can be found in the Appendix.

2.4. Analyses

Comparative statistics were performed using IBM SPSS Statistics 22.0 software. Before analyses, data was screened for outliers, and normality, homogeneity of variances, and homogeneity of intercorrelations were tested. The main goal of this study was to examine the effect of the cognitive flexibility training in stroke patients. Therefore, pre-training and post-training scores on the selected tasks were compared for all cognitive domains (attention, reasoning, and memory). To this extent, a 2x2 split-plot analysis of variance (ANOVA) was carried out. Time (i.e. pre- and post-training scores on reasoning, attention, and memory performance), was the within-subject factor. Group (cognitive training versus active control

(13)

training) was the between subjects factor. Furthermore, to explore the influence of baseline performance (low versus high) and lesion location (left hemisphere, right hemisphere or both) these variables were also entered as between-subjects factors independent of any possible training effects. Low baseline was defined as participants scoring below average on the selected cognitive domain pre-training, whereas high baseline was defined as participants scoring above average. Significance levels were set at α = 0.05 (two-tailed), Bonferroni’s correction was used where multiple comparisons were made.

3. Results

3.1. Participants

A total of 51 individuals enrolled in the study (32 males). Participants were randomly assigned to either the cognitive flexibility training (n = 25, 14 males, mean age = 58.88, SD = 8.23) or the active control training (n = 26, mean age = 59.69, SD = 7.86). There were no significant differences in sex (p = 0.338) or age (p = 0.717) between the two groups. Missing values are either because participants dropped out before finishing the total training, or because they could not complete all cognitive tests at either of the two time points. Table 1. shows the number of participants from which data was available for analyses. Outliers were defined according to the outlier labeling rule (Hoaglin & Iglewicz, 1987). Because results revealed no difference when outliers were removed from analysis, the results reported here are those including the outliers.

Table 1.

Number of participants per condition Active control

training

Cognitive flexibility

training Total

Enrolled in training 25 26 51

Finished total training 17 12 29

T0 & T1 Reasoning 20 15 35

T0 & T1 Attention 17 14 31

T0 & T1 Memory 18 13 31

(14)

3.2. Training effects

To compare the effect of the training on each cognitive domain, three separate split-plot ANOVA’s were performed with Time (i.e. pre- and post-scores on each cognitive domain) as the within subjects factor and Condition (cognitive flexibility training versus active control training) as the between subjects factor. Analysis revealed no significant interaction effect between time and condition for reasoning scores (F(1, 33) = 1.347, p = 0.254, ηp2 = 0.039).

Therefore, the change in reasoning performance over time did not differ between the high switch training condition and the active control training. Furthermore, reasoning performance did not differ significantly between the high switch group (M = -0.170, SE = 0.180) and the active control group (M = -0.164, SE = 0.156; main effect Condition: F(1, 33) = 0.001, p = 0.982, ηp2 < 0.001). However, there was a significant main effect for time. Reasoning

performance scores were significantly lower after twelve weeks (M = -0.357, SD = 0.780) as compared with before the start of the training (M = 0.024, SD = 0.742; main effect Time: F(1, 33) = 10.403, p = 0.003, ηp2 = 0.240). Figure 1. shows the marginal means with standard

errors of the reasoning scores for both groups at the two time points.

Regarding attention performance, the change in scores was larger for the active control condition (mean difference post – pre-training = 0.712) than for the high switch condition (mean difference post – pre-training = -0.032). However, this effect was only of borderline

Figure 1. Reasoning performance pre- and post-training per condition (means and standard errors). There is a significant decrease in reasoning performance after 12 weeks of training (p= 0.003). No significant difference between groups was found.

(15)

significance (interaction effect Time x Condition: F(1, 29) = 3.739, p = 0.063, ηp2 = 0.114),

and when correcting for multiple comparisons this effect was even less significant. Also the main effect of Condition was not significant (F(1, 29) = 1.070, p = 0.31, ηp2 = 0.036).

Furthermore, the main effect of Time (i.e. attention pre- and post-training scores) was not significant (M = -0.368, SD = 1.331; M = -0.0075, SD = 1.298; F(1, 29) = 3.126, p = 0.088, ηp2 = 0.097). Figure 2. shows the marginal means with standard errors of the attention scores

for both groups at the two time points.

Looking at memory performance, there was no difference in the change in scores between the active control condition and the high switch condition (interaction effect Time x Condition: Pillai’s Trace = 0.091; F(1, 29) = 2.886, p = 0.10, ηp2 = 0.091), and when adjusting

for multiple comparisons this effect was even less significant. In addition, there was no main effect for Condition (F(1, 29) = 0.509, p = 0.481, ηp2 = 0.017) or for time, that is, memory

scores pre- and post-training (F(1, 29) = 0.050, p = 0.824, ηp2 = 0.002). Figure 3. shows the

marginal means with standard errors of the memory scores for both groups at the two time points.

Figure 2. Attention performance pre- and post-training per condition (means and standard errors). There is no significant difference in attention performance after 12 weeks of training. No significant difference between groups was found.

(16)

3.3. Possible factors influencing stroke outcome

Because the results revealed no significant effect of the training, data from both groups were pooled to explore the possible influence of baseline performance (high versus low) and lesion location (left, right or both hemispheres) on the scores of each cognitive domain over time. Again, Time (pre- post scores on each cognitive domain) was the within subjects factor and either Baseline (high vs. low) or Hemisphere (left, right or both) were the between subjects factors.

Results for the effect of Baseline performance on the change scores for each cognitive domain are shown in Figure 4. Examining the effect of baseline performance on attention scores, results revealed a significant interaction effect (interaction effect Attention x Baseline: F(1, 29) = 5.269, p = 0.029, ηp2 = 0.154), indicating that the change in attention scores

differed for the low and high baseline performers. Post-hoc analyses indicated that participants with low baseline attention scores significantly improved after twelve weeks (mean difference post – pre-training = 0.916, Wilks’ λ = 0.757, F(1,29) = 9.286; p = 0.005, ηp2 = 0.243), whereas high baseline performers showed no significant increase (mean

difference post – pre-training = 0.035, Wilks’ λ = 0.999, F(1,29) = 0.021; p = 0.886, ηp2 =

0.001). However, when corrected for multiple comparisons, this interaction effect was not significant anymore. Looking at reasoning performance revealed that participants with a high

Figure 3. Memory performance pre- and post-training per condition (means and

standard errors). There is no significant difference in memory performance after 12 weeks of training. No significant difference between groups was found.

(17)

baseline score declined more (mean difference post – pre-training = -0.531, SE = 0.137) than participants with low baseline scores (mean difference post – pre-training = -0.128, SE = 0.178). However, this result was only marginally significant (interaction effect Reasoning x Baseline: Pillai’s Trace = 0.89, F(1, 33) = 3.216, p = 0.082, ηp2 = 0.089). Analyses further

revealed no significant effect of baseline performance on the change in memory performance (interaction effect Memory x Baseline: F(1, 29) = 2.259, p = 0.144, ηp2 = 0.072).

(18)

Results for the effect of Hemisphere are shown in Figure 5. Analyses indicate no significant interaction effects for either of the cognitive domains (interaction effect Reasoning x Hemisphere: F(2, 30) = 0.516, p = 0.602, ηp2 = 0.033; interaction effect Attention x

Hemisphere: F(2, 26) = 1.061, p = 0.361, ηp2 = 0.075; interaction effect Memory x

Hemisphere: F(2, 27) = 0.318, p = 0.73, ηp2 = 0.023). Thus, the change in scores on reasoning,

attention and memory performance did not differ between participants with a left sided or right sided lesion, or when both hemispheres were lesioned. Furthermore, no significant main effects for Hemisphere were found for reasoning, attention or memory (main effect Reasoning: F(2, 30) = 0.776, p = 0.469, ηp2 = 0.049; main effect Attention: F(2, 26) = 1.563,

p = 0.229, ηp2 = 0.107; main effect Memory: F(2, 27) = 1.299, p = 0.289, ηp2 = 0.0088). Thus,

(19)

4. Discussion

The aim of the present study was to investigate the effectiveness of an online cognitive flexibility training as compared to an active control training in patients suffering from stroke. Participants trained for a period of twelve weeks, for 30 minutes a day, five days a week on games aimed at improving reasoning, attention and memory performance. Independent from the effectiveness of this training, a secondary aim of this study was to explore whether baseline performance and lesion location influence stroke outcome. Performance on reasoning, attention and memory abilities each were measured with the composite score of two neuropsychological tests per cognitive domain. It was hypothesized that participants receiving the cognitive flexibility training would improve more than participants who received the active control training. Furthermore, it was hypothesized that participants with low baseline scores in any of the cognitive domains would benefit more from the training than would participants who already had high baseline scores. Since functional brain responses after stroke are anatomically specific, lesion location was also thought to influence stroke outcome on reasoning, attention and memory performance.

Examining the effects of the training, results do not confirm the hypotheses. The change in pre- and post-training scores on any of the cognitive domains did not differ significantly between the cognitive flexibility training and the active control training. There was also no significant difference in reasoning, attention or memory performance between the two groups. Furthermore, for attention and memory performance there was no difference in pre- training

Figure 5. Reasoning, attention and memory performance pre- and post-training for each lesion location (means and standard errors). No main effect of Hemisphere was found. There was no interaction between Hemisphere and Time.

(20)

scores as compared with at the end of the training. Remarkably, for reasoning ability participants scored significantly lower after twelve weeks of playing brain training games than before the start of their training. Post hoc analyses revealt that participants scored significantly lower on Ravens’s PCM post-training as compared with the pre-training. Unfortunately, there seems to be no obvious explanation for this finding, therefore it is interesting to find out whether this result will be comfirmed in a different group of patients. The fact that the present study did not find any significant transfer effects of the training is not entirely unexpected or inconsistent with previous research. In a review of Van Heugten, Wolters, Gregório, and Wade (2012), investigating the effectiveness of cognitive rehabilitation, they showed that only 54% of the studies (53 out of 95) demonstrated a significant effect of the experimental treatment. In 23% of the studies there was no difference between the experimental and control condition, and in another 20% the effects were partial (i.e., effects were not present for the primary or all outcome measures, or treatment altered function in some domains more than others) (van Heugten, Wolters Gregório, & Wade, 2012). The absence of any transfer effects in the present study can be explained in several ways.

First of all, it is argued that time after stroke is an important factor influencing recovery of cognitive function after stroke. Kotila et al. (1984) found that cognitive recovery after stroke was maximal within the first three months post-stroke. Cognitive improvement on some domains was also evident after one year (Desmond, et al., 1996; Kotila, Waltimo, Niemi, Laaksonen, & Lempinen, 1984), but no further improvement was found after two years (Desmond, et al., 1996). Time since stroke for most patients included in the present study was more than one year ago. Therefore, these patients might be in the chronic phase. It might be important to discriminate between the chronic type and patients in the post-acute phase to find an effect of the training.

Secondly, some research suggests that restoration of cognitive functions after brain injury is not expected to occur (see for a review (Geusgens, Winkens, van Heugten, Jolles, & van den Heuvel, 2007; Wilson, 2000). If indeed cognitive functions cannot be restored, no transfer effects on the neuropsychological tests will be detected after a cognitive training aiming to improve cognition. Therefore rehabilitation should focus on teaching compensatory strategies (i.e. strategy training) aimed at teaching patients new internal and external strategies and techniques to compensate for their problems in every day life (Geusgens, et al., 2007).

Thirdly, some remarks can be made about the outcome measures and the assessment of cognitive improvement in the present study. The use of composite scores in the present study might have resulted in loss of important information, and might thus be less specific or

(21)

detailed. Although rationally combined measures can be useful in research and clinical situations, these measures risk the loss of critical information if not carefully considered (Berg & Byrd, 2002). Furthermore, it could be that the way improvement was measured in this study was not sensitive enough to detect any changes in cognitive abilities. From some neuropsychological tests it is known that they are susceptible to ceiling effects (e.g. ToL; (Reizner, Song, & Levin, 2002)). Indeed, 16% of the participants solved 9 out of 10 trials on baseline and another 60% solved all trials during baseline measurement.

Fourth, it might be important to critically evaluate the specificity, duration and flexibility aspects of the current training. It could have been that the selected games used to train participants’ cognitive performance were too specific. According to Nouchi et al (2011), the mechanism of transfer effects can be explained by recent hypothesis which proposes that transfer effects can be induced if the process during both training and transfer task overlap and involve similar brain regions. It could have been that the games only elicited activation in very specific brain areas required for the task at hand, and therefore the mechanism for transfer effects would not apply. However, this seems unlikely, because the games were carefully selected for their ability to train reasoning, attention and memory. Another explanation could be that the intensity or duration of the training would not allow for improvement on the selected tasks in stroke patients. This is also unlikely, for there are studies that already show improvement in cognitive performance after a 4 to 6 week training period (Jaeggi, et al., 2011; Johansson & Tornmalm, 2012; Nouchi, et al., 2012; Olesen, Westerberg, & Klingberg, 2003). However, the extent and magnitude to which these training effects apply to near and far transfer effects, still varies considerably among these studies (e.g. type of tasks used, type of cognitive domain investigated, number of participants investigated). Finally, it could have been that the flexibility manipulation in the cognitive flexibility training was ineffective. Flexibility in the training was ensured by switching between games every other three minutes, as compared with every ten minutes for the active control training. It could be that the switch cost was too high and participants were unable to really train within those three minutes.

Looking at the effect of baseline performance on the change in scores for attention abilities, results revealed that participants with low baseline scores significantly improved after twelve weeks, whereas participants with high baseline scores did not show any significant improvement over time. However, when this interaction effect was corrected for multiple comparisons it was not significant anymore. Furthermore, looking at reasoning abilities, it seemed that participants with the high baseline scores declined more than participants with

(22)

low baselines. However, this effect was not significant (p = 0.067). For memory abilities the change in scores before and after training was independent of baseline performance. The fact that the current study could not replicate earlier findings that low baseline performance was associated with greater improvement on any of the cognitive domains can be explained by the low power of the analyses (< 0.45). Earlier research is fairly consistent on the effect of low baseline performance on cognitive improvement in stroke patients, a finding that is ascribed to the ceiling effect (Desmond, et al., 1996; Johansson & Tornmalm, 2012; Peretz, et al., 2011). In addition, because of the fact that the training did not have any significant effects (no improvement was found whatsoever), it could have been more difficult to detect any possible effects of baseline functioning.

For the influence of lesion location on reasoning, attention or memory performance this study also did not find any proof. The change in scores (i.e. the effect of time) for either of the cognitive domains was independent of lesion site (left hemisphere, right hemisphere or both). Furthermore, lesion location did not influence reasoning, attention or memory performance. One possible explanation could be the low observed power of the analyses (for all tests < 0.215). The fact that the left hemisphere would be the dominant hemisphere and be associated with higher recovery rates in the long-term (Bentley, et al., 2014; Desmond, et al., 1996) was not supported by the present study. However, in their article, Grefkes et al. (2011) pointed out the importance of ‘functional integration’ and state that “a connectivity-based system perspective seems to be much closer to the neurobiology underlying brain function in both physiological and pathological conditions compared with approaches assigning specific behaviours (or clinical symptoms) to anatomically segregated regions”. Stroke-induced lesions may not only affect the connectivity between the cortex and spinal cord, but can also affect the interactions between different areas more distant from the lesion. It is suggested that functional outcome after stroke can be predicted by how both hemispheres are coupled (Grefkes & Fink, 2011). Thus, it might be irrelevant to look at a specific brain site, but rather the pattern of brain activation must be examined to determine risk factors in stroke outcome. Furthermore, in the context of ‘functional specialization’, the left versus right distinction might have been too general. For example, memory performance in this study was measured by combining the z-scores of the RAVLT and the Corsi-block tapping task. It might be expected that performance on the RAVLT relies more heavily on the left hemisphere for this test measures verbal memory. Conversely, the Corsi-block tapping task measures visuospatial short-term memory, and is found to rely more on the right hemisphere (Kessels, et al., 2000). Hence, the distinction between left and right hemisphere might be inconvenient.

(23)

Taken together, the present study found no effect of the cognitive flexibility training on reasoning, attention or memory performance. Independent of the effectivity of this training, this study also failed to show any significant influence of baseline scores or lesion location on cognitive performance. A major limitation of this study is probably the low power from statistical analyses; it cannot be concluded with certainty whether the absence of training effects and the absence of any correlation with lesion location and baseline performance on cognition are due to the small sample size or because there really is no effect. In addition, effects sizes were also fairly small, making it more difficult to detect any specific effects of the treatment. Therefore, it is suggested that this study should be replicated with more participants. Furthermore, this study chose to use composite scores to measure reasoning, attention and memory performance in order to make conclusions about overall cognitive abilities and transfer effects. It might also be interesting for future research to look for the effect of the training on a more detailed level per neuropsychological test to make more specific conclusions about training effectivity.

(24)

References

Anderson, P., Anderson, V., & Lajoie, G. (1996). The Tower of London test: Validation and standardization for pediatric populations. The Clinical Neuropsychologist, 10(1), 54-65.

Bentley, P., Kumar, G., Rinne, P., Buddha, S., Kallingal, J., Hookway, C., et al. (2014). Lesion locations influencing baseline severity and early recovery in ischaemic stroke. European Journal of Neurology.

Berg, W. K., & Byrd, D. L. (2002). The Tower of London Spatial Problem-Solving Task: Enhancing clinical and research implementation. Journal of Clinical and Experimental Neuropsychology, 24(5), 586-604.

Bouma, A., Mulder, J., Lindeboom, J., & Schmand, B. (2012). Handboek neuropsychologische diagnostiek. Amsterdam: Pearson Assessment.

Buitenweg, J. I., Murre, J. M., & Ridderinkhof, K. R. (2012). Brain training in progress: a review of trainability in healthy seniors. Frontiers in Human Neuroscience, 6, 183. Cheng, B., Forkert, N. D., Zavaglia, M., Hilgetag, C. C., Golsari, A., Siemonsen, S., et al.

(2014). Influence of stroke infarct location on functional outcome measured by the Modified Rankin Scale. Stroke.

Crawford, J. R., Obonsawin, M. C., & Allan, K. M. (1998). PASAT and components of WAIS-R performance: Convergent and discriminant validity. Neuropsychological Rehabilitation, 8(3), 255-272.

Culbertson, W. C., & Zillmer, E. A. (1998). The Tower of LondonDX: A standardized approach to assessing executive functioning in children. Archives of Clinical Neuropsychology, 13(3), 285-301.

De Luca, R., Calabrò, R. S., Gervasi, G., De Salvo, S., Bonanno, L., Corallo, F., et al. (2014). Is computer-assisted training effective in improving rehabilitative outcomes after brain injury? A case-control hospital-based study. Disability and Health Journal, 7(3), 356-360.

Desmond, D. W., Moroney, J. T., Sano, M., & Stern, Y. (1996). Recovery of cognitive function after stroke. Stroke, 27(10), 1798-1803.

Geusgens, C. A. V., Winkens, I., van Heugten, C. M., Jolles, J., & van den Heuvel, W. J. A. (2007). Occurrence and measurement of transfer in cognitive rehabilitation: A critical review. Journal of Rehabilitation Medicine, 39(6), 425-439.

Green, C. S., & Bavelier, D. (2008). Exercising your brain: A review of human brain plasticity and training-induced learning. Psychological Aging, 23(4), 692-701.

Grefkes, C., & Fink, G. R. (2011). Reorganization of cerebral networks after stroke: New insights from neuroimaging with connectivity approaches. Brain, 134(Pt 5), 1264-1276.

Gronwall, D. M. A. (1977). Paced auditory serial-addition task: A measure of recovery from concussion. Perceptual and Motor Skills, 44, 367-373.

Hoaglin, D. C., & Iglewicz, B. (1987). Fine-tuning some resistant rules for outlier labeling. Journal of the American Statistical Association, 82(400), 1147-1149.

Homack, S., Lee, D., & Riccio, C. A. (2005). Test Review: Delis-Kaplan Executive Function System. Journal of Clinical and Experimental Neuropsychology, 27, 599-609.

Jaeggi, S. M., Buschkuehl, M., Jonides, J., & Shah, P. (2011). Short- and long-term benefits of cognitive training. Proceedings of the National Academy of Sciences, 108(25), 10081-10086.

(25)

Johansson, B., & Tornmalm, M. (2012). Working memory training for patients with acquired brain injury: Effects in daily life. Scandinavian Journal of Occupational Therapy, 19(2), 176-183.

Kessels, R. P. C., van Zandvoort, M. J. E., Postma, A., Kappelle, L. J., & de Haan, E. H. F. (2000). The Corsi block-tapping task: Standardization and normative data. Applied Neuropsychology, 7(4), 252-258.

Khan, M., Goddeau, R. P., Jr., Zhang, J., Moonis, M., & Henninger, N. (2014). Predictors of outcome following stroke due to isolated M2 occlusions. Cerebrovascular Diseases Extra, 4(1), 52-60.

Kotila, M., Waltimo, O., Niemi, M. L., Laaksonen, R., & Lempinen, M. (1984). The profile of recovery from stroke and factors influencing outcome. Stroke, 15(6), 1039-1044. Lemay, S., Bedard, M.-A., Rouleau, I., & Tremblay, P.-L. G. (2004). Practice effect and

test-retest reliability of attentional and executive tests in middle-aged to elderly subjects. The Clinical Neuropsychologist, 18(2), 284-302.

Lev, M. H., Segal, A. Z., Farkas, J., Hossain, S. T., Putman, C., Hunter, G. J., et al. (2001). Utility of perfusion-weighted CT imaging in acute middle cerebral artery stroke treated with intra-arterial thrombolysis: Prediction of final infarct volume and clinical outcome. Stroke, 32(9), 2021-2028.

Lundqvist, A., Grundström, K., Samuelsson, K., & Rönnberg, J. (2010). Computerized training of working memory in a group of patients suffering from acquired brain injury. Brain Injury, 24(10), 1173-1183.

Middleton, D. K., Lambert, M. J., & Seggar, L. B. (1991). Neuropsychological Rehabilitation: Microcomputer-assisted treatment of brain-injured adults. Perceptual and Motor Skills, 72(2), 527-530.

Millis, S. R., Rosenthal, M., Novack, T. A., Sherer, M., Nick, T. G., Kreutzer, J. S., et al. (2001). Long-term neuropsychological outcome after traumatic brain injury. The Journal of Head Trauma Rehabilitation, 16(4), 343-355.

Nouchi, R., Taki, Y., Takeuchi, H., Hashizume, H., Akitsuki, Y., Shigemune, Y., et al. (2012). Brain training game improves executive functions and processing speed in the elderly: A randomized controlled trial. PLoS One, 7(1), e29676.

Olesen, P. J., Westerberg, H., & Klingberg, T. (2003). Increased preftontal and parietal activity after training of working memory. Nature neuroscience, 7(1).

Owen, A. M., Hampshire, A., Grahn, J. A., Stenton, R., Dajani, S., Burns, A. S., et al. (2010). Putting brain training to the test. Nature, 465(7299), 775-778.

Peretz, C., Korczyn, A. D., Shatil, E., Aharonson, V., Birnboim, S., & Giladi, N. (2011). Computer-based, personalized cognitive training versus classical computer games: A randomized double-blind prospective trial of cognitive stimulation. Neuroepidemiology, 36(2), 91-99.

Pyun, S.-B., Yang, H., Lee, S., Yook, J., Kwon, J., & Byun, E.-M. (2009). A home programme for patients with cognitive dysfunction: A pilot study. Brain Injury, 23(7-8), 686-692.

Raven, J. (1989). The Raven Progressive Matrices: A review of national norming studies and ethnic and socioeconomic variation within the United States Journal of Educational Measurement, 26(1), 1-16.

Reizner, R. D., Song, J., & Levin, H. S. (2002). Raising the ceiling: The Tower of London - Extended version. Developmental neuropsychology, 21(1), 1-14.

Rosso, C., & Samson, Y. (2014). The ischemic penumbra: The location rather than the volume of recovery determines outcome. Currunt Opinion Neurology, 27(1), 35-41. Ruttan, L., Martin, K., Liu, A., Colella, B., & Green, R. E. (2008). Long-term cognitive

(26)

timed and untimed tests at 1 and 4.5 or more years after injury. Archives of Physical Medicine and Rehabilitation, 89(12, Supplement), 69-76.

Sala, S. D., Gray, C., Baddeley, A., Allamano, N., & Wilson, L. (1999). Pattern span: A tool for unwelding visuo-spatial memory. Neuropsychologia, 37, 1189-1199.

Shallice, T. (1982). Specific impairments of planning. PhilosophicalTransactions of the Royal Society of London 298, 199-209.

Sherman, E. M. S., Strauss, E., & Spellacy, F. (1997). Validity of the Paced Auditory Serial Addition Test (PASAT) in adults referred for neuropsychological assessment after head injury. The Clinical Neuropsychologist, 11(1), 34-45.

Strauss, E., Sherman, E. M. S., & Spreen, O. (2006). A compendium of neurospychological tests: Administration, norms, and commentary. (3rd ed.): Oxford University Press. van Heugten, C., Wolters Gregório, G., & Wade, D. (2012). Evidence-based cognitive

rehabilitation after acquired brain injury: A systematic review of content of treatment. Neuropsychological Rehabilitation, 22(5), 653-673.

Westerberg, H., Jacobaeus, H., Hirvikoski, T., Clevberger, P., Östensson, M. L., Bartfai, A., et al. (2007). Computerized working memory training after stroke: A pilot study. Brain Injury, 21(1), 21-29.

Wilson, B. (2000). Compensating for Cognitive Deficits Following Brain Injury. Neuropsychology Review, 10(4), 233-243.

(27)

Appendix

Description of the training games

Cognitive flexibility training

The cognitive flexibility training consists of 9 different games. Each game has 20 levels, and per level three stars can be awarded. Participants are instructed to keep training on the current level when they have only one star. When they have obtained two stars, they are free to choose whether to train on the current level, or switch to the level above. With three stars, participants are asked to train on the next level. In this way, it is assured that task difficulty is tailored to each participant’s performance level.

Attention games Pattern Matrix

In this game participants have to mentally rotate figures with certain patterns to find pairs. This game is under time pressure. The higher the level, the more difficult the patterns become and the more patterns will be presented per trial.

Birds of a feather

In this game participants are instructed to count the number of blue birds presented between distracter birds. This game is under time pressure. The higher the level, the more similar the target and distracter birds become.

Brainfreeze

In this game participants are presented with a number of squares continuously changing colours. Participants have to click on each square to freeze the image as soon as the colours match their neighbouring square. The higher the level, the more squares will be presented.

Mind the mole

Participants are presented with a field of carrots. A mole appears in this field. As soon as the movement pattern (flashing, up/downwards, left to right) of the mole start to change,

(28)

Reasoning games Square logic

Participants are presented with blocks containing numbers (e.g. 1 to 3). Blocks have to be placed on each other (the bottom block will disappear) such that only one block will remain. Blocks can only be placed on other blocks next to them and only when they are exactly one number higher or lower. This task is under time pressure. The higher the level, the more blocks and numbers will be presented.

Out of order

Participants are presented with cards, containing figures differing in colour, shape, filling, and number. The cards need to be arranged in such a way that they match with their neighbour figure on at least one of their characteristics. This task is under time pressure. The higher the level, the more cards will be presented.

Patterned logic

Participants will be presented with a pattern build from tiles based on colour and a figure mark. Participants have to complete the missing tiles in the pattern. The higher the level, to more gaps participants have to fill.

Memory games Toy shop

Participants are presented with a shopping list containing pictures of items. They have to remember those items and collect these items from a store. The higher the level, the more items will have to be remembered.

Multi memory

Participants will be presented with several pictures in different colours and shapes. They have to reconstruct these figures after they have disappeared from the screen.

Moving memory

Participants are presented with tiles containing a number on the top, and a figure at the back. They have to find pairs with matching figures. However, after a pair has been found, the

(29)

remaining tiles change position. Thus, figures can only be remembered based on the numbers of the tiles. The higher the level, the more tiles are presented.

Active control training

The active control training consists of three games per day. All games consist of 20 levels, but the highest level participants will eventually train on is level nine. Each week, the level on which participants had to train was defined according to a predefined schedule, thus, training was not adapted to their personal abilities. In the first six weeks participants start to train on the first level, and each week they are asked to train on one level above. In the seventh week participants are asked to train on level six again, in week eight participants train on level seven. Weeks nine and ten are trained on level eight, and in weeks eleven and twelve participants are asked to train on level nine.

Attention games Fuzzle

Participants have to reconstruct a fractured picture.

Sliding search

Pictures will slide over the lower part of the screen. Participants have to match the pictures from the lower part with those from the upper part of the screen.

Pay attention

Different squares will appear on the screen. As soon as they change colour participants have to click on the square and it will disappear. The higher the level, the more squares will be presented.

Grid tracks

Participants are presented with a number of blocks. Some of them contain a blue star. When these blue stars disappear, participants have to mentally follow the trajectory of those blocks. When the blocks stop moving, they have to identify the location of the block that initially had blue stars.

The effectiveness of cognitive flexibility training on reasoning, attention, and memory in stroke patients