The effect of a computerized cognitive
flexibility training on cognitive performance
A cognitive plasticity comparison study between stroke patients
and healthy elderly
M. Steffers Student nr.: 10752668 Date: 20 October 2015
Master’s Thesis: Brain & Cognition Supervisor: Mw. Drs. R.M. van de Ven Second Assessor: Dhr. Dr. S. Van der Werf
Abstract
:
There is an increasing amount of evidence suggesting that the central nervous system is capable of change and adaptation throughout the lifespan. Cognitive plasticity describes this resilience of cognition. This cognitive plasticity and cognitive abilities are thought to decline with age. However, there are studies suggesting that this might not be true in a substantial amount of the aging population. On another subject, brain trauma alters this cognitive plasticity and is responsible for cognitive rehabilitation. Training your brain may prevent cognitive decline and enhance cognitive rehabilitation in groups who are vulnerable to cognitive decay. Several studies show promising results in different vulnerable groups like elderly and stroke patients. However, little comparison studies between these two groups in the population. The current study investigates the effect and difference in effect of a computerized cognitive flexibility training between stroke patients and healthy elderly. The subjects train for 12 consecutive weeks with either a cognitive flexibility training or an active control condition. A third group of only stroke patients is on the waitlist for 12 consecutive weeks and will start training after this waiting period. We found a slight improvement on working memory related task performance in both groups and training conditions. No differences between healthy elderly and stroke patients were found. Moreover, both the cognitive flexibility training as the active control training produced the same effect. The waitlist group showed no cognitive improvement. The results in this study show promising results in reversing the negative
cognitive effects of age and brain injury by enhancing the working memory capacity and working memory related performance. However, the effects are limited and therefore not yet clinical relevant. To maximize the effect of the training program more elaborate research is necessary.
Introduction
Human aging is often associated with cognitive decline and eventually with cognitive impairments (Buitenweg, Murre & Ridderinkhof, 2012). Another common cause of cognitive impairment is a Cerebral Vascular Accident (CVA). Cognitive impairments are measured in approximately one third of the chronic stroke patients (van de Ven, Schmand, Groet, Celtman & Murre, 2015). These two large groups combined make cognitive decline and impairment eminent in a large proportion of the human population. These kind of problems can greatly interfere with activities and quality of daily life. Given the vast expansion of elderly in the population and the increased longevity expectation, there is a pressing need to increase self-independence. Even so, in CVA patients cognitive problems are so evident that rehabilitation programs are needed to minimize the impact on daily life.
Fortunately, the brain is thought not to be a static entity. It has the capacity to alter its structure and function dynamically throughout the lifespan (Pascual-leone, et al., 1999; Grady et al, 2006). Moreover, after brain damage, like a CVA, changes occur in the structure of the brain. Some of these changes are the result of the injury itself and some of these changes are due to reorganization of neural pathways in response to the injury (Ramachandran, 2005). This resilience of the brain throughout the lifespan and after an CVA is referred to as plasticity, with the resiliency in cognition being referred to as cognitive plasticity (Schupf, et al., 2004). After a CVA or with new demands of the environment, the brain alters the acquisition of new knowledge and the processing efficiency. These alterations change the concurrent range of performance and functioning, making the brain suitable for the new situation, hence cognitive plasticity (Lövdén, Bäckman, Lindenberger, Schaefer & Schmiedek, 2010).
In the current literature it is commonly found that younger people have higher properties of cognitive plasticity than older people, hence the commonly made notion that there is an effect of age on the resiliency of the brain (Singer, Lindenberger & Baltes, 2003). Moreover, aging is associated with changes in the brain that will limit cognitive functioning (Lustig et al., 2009). On the contrary, research indicates that a substantial fraction of aged individuals show only modest losses of cognitive abilities, or even maintain functioning with age (Gallagher, Bizon, Hoyt, Helm & Lund, 2003; Schupf, et al., 2004). Therefore it can be argued that cognitive plasticity remains with age. A possible explanation for this discrepancy comes from the hypothesis of Mattson and colleagues (2002). They hypothesized that neurons and glia cells respond to environmental stressors in aging by either adapting or succumbing. In this hypothesis, adapting is associated with successful aging and succumbing with maladaptive aging.
For stroke patients the reduction of cognitive function is partly due to neuronal death in the infarcted tissue and to cell dysfunction in the areas surrounding the infarct. Dysfunction in more remote areas from the infarcted tissue is believed to be due to diaschisis: the loss of function of a brain area connected to a remote injured area, causing a physiological imbalance (Witte, Bidmon, Schiene, Redecker & Hagemann, 2000). According to Wieloch and Nikolich (2006), functional recovery after a cardio vascular accident(CVA) involves three phases. In the first phase, which lasts to two or three weeks after the insult, there is reversal of diaschisis; cell repair is activated. This phase is also referred to as the spontaneous recovery phase. In the second phase, the properties of existing neuronal pathways are changed, which is called functional cell plasticity. In the third phase, neuroanatomical plasticity leads to the formation of new connections. Wieloch and Nikolich (2006) states that both phase two and three are involved in normal learning, which they claim as the driving force during functional recovery.
The nature of cognitive decline and impairment differs between stroke patients and healthy elderly, although it can be argued that in both groups the brain might be capable of change and adaptation to the environment. As mentioned earlier, it is thought that the brain is not a static entity, nowadays there is a trend to make use of this resiliency of the human brain by training cognition. This to help maintain or recover abilities with age or after a brain injury. An important issue on this topic is that of near and transfer. Transfer defines the degree in which the progress made in the training environment can be extended to a different context. With near transfer referring to training effects to domains proximal to the
trained skills and far transfer referring to training effects to domains more distal of the trained skills (Buitenweg, Murre & Ridderinkhof, 2012).
As stated by Lövdén, Bäckman, Lindenberger, Schaefer & Schmiedek (2010), a training program is effective if improvements affect many domains of functioning relevant for maintaining competence and autonomy in everyday life”. It is commonly advertised that training programs improve cognition and that these programs help maintaining competence and autonomy in everyday life. Although studies show promising results on the topic of brain-training, the currently available training programs generally fail to display fundamental transfer (Lustig et al., 2009). According to Buitenweg, Murre and Ridderinkhof (2012), plasticity is key to attain long term retention and transfer of a brain-training program. In other words, the brain has to respond to the training as it would do to new demands of the environment. In a study done by Günter, Schäfer, Holzner, & Kemmler (2010) elderly in a retirement home participated in a 14-week computer assisted cognitive training program. Their research indicated that cognitive training can achieve long-term improvements in cognition. The researchers state that this kind of training could prevent and treat cognitive deficits in older people. To link these results with the earlier described hypothesis of Mattson and colleagues (2002), about successful and maladaptive aging, this training program can be seen as an environmental stressor causing the neurons and glia cells to adapt instead of succumb. This can result in enhancement of cognitive performance. Possibly succumbing will occur if the stress produces by the environment is too high for the brain to adapt successfully. A training program could play a substantial role in stimulating the brain to successfully adapt to the demands of the training. Possibly a training program that is too demanding negatively influences the effect of the training. As mentioned earlier, recovery in stroke patients occurs in three distinct phases. According to Wieloch and Nikolich (2006), enriching the environment is an excellent way to enhance functional recovery in stroke patients. This induces multiple biological effects in the brain that could account for the positive effect on recovery. This positive effect is only found within the spontaneous recovery phase of several weeks (Wieloch & Nikolich, 2006; Westerberg, et al., 2009). The last two phases of recovery in stroke patients are similar to normal learning. Therefore it would be logical that a training program is beneficial for stroke patients as well. Although it is not ruled out that biological changes after a CVA still alter the process in the long term. Evidence for the effect of a cognitive training program after a CVA comes from a study done by Westerberg and colleagues (2009). In their research, 18 participants were randomized to either the treatment or the passive control condition. Participants in the treatment condition trained with a computerized working memory training program for five consecutive weeks, with a minimal of four timer per week. Westerberg and colleagues found evidence that, one to three years after a stroke, intensive training van improve working memory and attentional performance. In addition they found that the results can be extended to cognitive functioning in daily life, as the participants reported subjective improvements.
As described in the review done by Buitenweg, Murre and Ridderinkhof (2012), cognitive processes can be stimulated with more variability in a training program. This variability requires the subjects to integrate multiple cognitive domains rather than training separate cognitive mechanisms. They state that cognitive domains are neurologically and behaviourally intertwined. Possibly, a training program results in cognitive improvement if the program includes frequent switching between various training tasks (van de Ven, Schmand, Groet, Celtman & Murre, 2015; Buitenweg, Murre & Ridderinkhof, 2012). The Training Project Amsterdam Senior and Stroke (TAPASS) aims to make use of this cognitive flexibility training paradigm to improve cognition in both stroke patients and healthy elderly (van de Ven, Schmand, Groet, Celtman & Murre, 2015). The current study is part of the TAPASS research project, and aims to investigate and compare the effect of this training program in both healthy elderly and stroke patients. The key-question of the current research is identifying if the computerized cognitive flexibility training enhances cognition, that is, if is the training program shows fundamentals of transfer. Secondly, the objective is to identify the difference in effect of this training program between stroke patients and healthy elderly. It is hypothesized that long term biological processes alter the process of recovery. These
biological processes are thought to positively alter the recovery, the same as in the phase of spontaneous recovery, although to a less extend. This would be resulting in a higher training effect for CVA patients than for healthy elderly.
Furthermore, it is hypothesized that the environmental stress induced by the training influences the amount of cognitive improvement. In other words, the more challenging the training would be, the more gain in cognitive performance. Moreover, if the training will be too challenging, the environmental stress will be too great and neurons and glia cells will succumb instead of adapt, causing little effect of the training. This effect is possibly the same for both elderly and stroke patients.
In this study there are two different training conditions, a more challenging intervention condition and a less challenging active control condition. First pre-test and post-test differences will be compared to see if either or both of the training programs are effective, defined by several aspects of cognition including working memory, executive functioning, processing speed, and attention. Then the effects of the training will be compared between elderly and stroke patients. The difficulty level of the training will be used to see if a more challenging environment will account for more cognitive plasticity.
Because it is hypothesized that biological factors account for functional recovery in stroke patients, it is investigated if the training effect is due to the training program or due to spontaneous recovery. Therefore, the amount of cognitive plasticity that cannot be attributed to the training, will be controlled for by including a waitlist group. Additionally this waitlist group is necessary to control for test-retest or learning effects that can alter the results.
Methods
(Patient) sampleThe current study is part of the TAPASS study project. This double blind randomized control study with an experimental intervention group, active control condition and waitlist control condition was carried out at the University of Amsterdam. Both CVA patients and healthy elderly were recruited. The total sample was divided into two groups: the CVA patients group and the healthy elderly group. The healthy elderly were recruited via online advertisement, radio commercials and local newspapers. The CVA patients were recruited from several Dutch rehabilitation centres and rehabilitation departments of hospitals. The desired sample sizes were estimated based on a similar study done by Günter, Schäfer, Holzner, & Kemmler (2010). In this study a computerized cognitive training program was tested in a retirement home. Their sample consisted of 20 elderly participants. They found significant effects of the training.
The participants were randomly assigned to each of the conditions. Additionally the waitlist control condition was only present in the CVA patients sample. In the CVA group patients had experienced their stroke between five months and five years ago and were between the age of 42-73. The healthy elderly group was between the age of 60-78. The sampling distribution and characteristics are displayed in table 1. The TAPASS study was approved by the medical Ethical Review Board of the VU University medical Centre, Amsterdam. Participants received travel expenses and were granted lifetime access to the training program.
In-and exclusion criteria. In the CVA patients group, patients were be included if they suffered a CVA
between five months and five years ago the last five years and were between 30 and 80 years old. Additionally patients had to have had received rehabilitation therapy as inpatient or outpatient. Patients needed to have objectified cognitive impairments, assessed by an experienced clinician. The healthy elderly could be included if they were between the age of 60 and 85. Both the healthy elderly and CVA patients were excluded if they suffered from psychiatric disorders, other neurological disorders besides a CVA, were colour-blind or used strong sedatives. All participants were required to have daily access to a
computer with an internet connection and be able to use the computer independently. Furthermore participants had to be mentally and medically fit enough to complete 12 weeks of training. Their mental state for inclusion was evaluated by using the Telephone Interview Cognitive Status (TICS). A score of
≤30 on the TICS is suggestive for cognitive impairments. To screen out participants with too severe cognitive impairments, participants with a score of <26 were excluded (Brandt, Spencer & Folstein, 1988).
Table 1.
Sampling distribution and characteristics.
Groups Condition Number of participants Age Time since stroke (days)
Male Female Total Mean SD
CVA patients Intervention 13 8 21 58.33 907.52 469.88
Active control 14 6 20 61.65 866.40 447.54
Waitlist 11 9 20 54.85 772.33 470.99
Total 38 23 66 58.28 852.34 458.17
Healthy elderly Intervention 8 17 25 64.72
Active control 7 12 19 66.56
Total 15 28 43 65.49
Total sample 53 51 104
Materials
Intervention and control conditions. The study consisted of an intervention training, an active
control training and a waitlist training. Both training conditions consisted of 30 minutes training per day, five times per week for 12 consecutive weeks. The training was presented online and done at home. There was no supervisor/trainer present, however a trainer could be contacted via e-mail in case a participant experienced difficulties or had questions. Furthermore, both training groups were contacted once per one or two weeks.
The training tasks were derived from the existing website www.braingymmer.com and were designed to be motivating and visually attractive. Feedback was provided based on personal scores, on a tree star rating scale. The two training conditions and waitlist condition are described in further detail below.
Intervention training. The intervention group received a cognitive flexibility training, containing nine
tasks selected to mainly train attention, reasoning and working memory. Each training session consisted of 10 tasks, meaning that one task was executed twice during a training session. Participants switched to a new task every three minuets. Tasks were presented directly after each other to promote frequent switching between tasks. The order of the tasks ensured that tasks designed to train the same cognitive domain were never immediately after each other.
The intervention training was adaptive, meaning that the difficulty of the tasks was modified based on individual performance on the training tasks. When participants had received enough points on a level, they were obliged to go to the next level. There were up to 20 levels available. The games were as described below, clustered per cognitive domain they were primarily designed to train.
Working memory.
Moving Memory. This game was a variation of the classic game ‘memory’. Participants are required to
find matching pairs of figures on faced down tiles. The cards are numbered. Each time a correct combination is found the cards are shuffled. Participants need to remember the location of the cards as well as which figure is underneath which number..
Toy Shop. During this game participants had to memorize items on a shopping list. After a short period
of time, had to order these items from a virtual toy store. Higher levels included more items on the wish list and more distractors.
Multi Memory. In this game the participant saw tiles with figures in different colours at a specific
location on the screen. The participant was required to memorize the figures, colours and location of the tiles. The participant was required to reproduce the cards by dragging colours and figures to the right place on the screen, when the figures had disappeared. The number of figures to remember and reconstruct gradually increased at higher levels.
Reasoning.
Square Logic. During this game the participant saw a number of blocks with the numbers 1, 2 and 3 on
them. They were required to eliminate all of the blocks by stacking the blocks on top of each other. However, there were a couple of rules. Blocks could only be stacked if the number on them differed by one (e.g. 1 on 2, 2 on 3 but 3 not on 1 and 1 not on 1). The second rule was that block could only be stacked if they were next to each other, horizontally, vertically and diagonally. Higher levels included more blocks.
Out of Order. During this game the participant saw a row with tiles with different figures in different
colours and shapes. The goal was to order the tiles in a way that each tile neighbours a tile with at least one similarity. For example, the neighbouring tiles needed to be the same colour, or the same shapes, or the shape had to have the same filling, or they had to have the same amount of shapes. The possible similarities were: shape (oval, rectangle, wave or bevelled rectangle); number of figures (1, 2 or 3), colour (blue, green, grey, red of purple) and fillings (striped, empty or fully coloured). This task had to be performed under time pressure. Higher levels contained more tiles.
Patterned Logic. This time limited game consisted of a row with coloured tiles with figures on them. The
figures and colours were ordered in a specific way, with a few gaps. The tiles necessary to complete the sequence were displayed in the centre of the screen. The participant had to place the missing tiles correctly in the sequence so the pattern was completed.
Attention.
Mind the Mole.. During this game the screen was filled with carrots. On top of the carrots, molehills
appeared. These molehills moved in a specific manner. If the molehill changed in movement, the participant had to click on it as soon as possible to remove the molehill. If the participant had clicked too soon or too late, carrots will disappear.
Birds of a Feather . During this game participant saw birds in different colours. The participant was
required to count the amount of light-blue birds with pointed beaks under time pressure. The higher the level, the more similar the distractor birds and the to be counted birds became. This game made, besides attention a strong appeal at visual searching strategies and overview capacities.
Pattern Matrix. This was a time limited game in which participants had to mentally rotate block patterns
to find pairs of two equal patterns. The higher the level, the more difficult the patterns became and the more patterns where presented. This game, primarily designed to train attention, made a strong appeal at visual perception and mental rotation.
Active control training. The active control training consisted of four tasks. Participants switched
between tasks every 10 minutes and performed three tasks per session. The active control training was not adaptive, nine of the 20 available levels were selected at front. Therefore the active control training was believed to not trigger improvement on cognition and to be less challenging than the intervention training.
Participants trained at one of these levels per week while the level was increased each week during the first five weeks, from week 6 the level every two weeks increased. If a participant did not master the level, the participant had to continue training on that level until one star was achieved. It was ensured that the active control training was sufficiently challenging, but not too difficult. Further more the switching component was not available in the active control training. The four different training tasks are described in detail below.
Training tasks.
Pay Attention. This was a time limited game, designed for improvement of divided attention and
response speed. In this game, the screen was filled with diamonds and mines. The participants had to protect their diamonds. On the screens, mines started flashing red. This was the moment the participants had to click on the mines in order to defuse them. If participants clicked too early or too late on the mines, they would explode and destroyed the diamonds surrounding the mine.
Grid Tracks. This game was designed for the improvement of sustained attention. The participants saw
a couple of blocks with stars on them. In the beginning of a round, some stars turned blue. Participants were required to focus on the blocks with a blue star on them. The stars turned white again and the blocks started moving. Participants had to follow the blocks which had a blue star on them. When the blocks had stopped moving, participants had to click on the blocks which initially contained a blue star.
Sliding Search. This was a game for processing speed. The participant saw 6 equivalent pictures at the
top of the screen. At the bottom of the screen, pictures passed through the screen. The participant was required to drag the matching picture from the bottom of the screen on top of the passing picture, before this picture disappeared.
Fuzzle. This is a game for visual perception and visual working memory. The participant looked at a
photograph to memorize it. After this, the picture was shattered into pieces and the participant had to set the pieces in the correct order so the photograph was whole again.
Waitlist control condition. The waitlist control group did not engage in a training program during the
first twelve weeks. After this waiting period, all participants in this condition received the intervention training.
Outcome measures. A Neuropsychological testing battery was conducted at two different time
points, to measure the training effects on cognition. This testing battery consisted of six tests, which are described in to detail in the next section of this thesis.
Trail Making Test (TMT) from the Delis-Kaplan executive function system (D-Kefs). The TMT was derived from the D-kefs. This test provides information on visual search, processing speed, mental flexibility, scanning and executive functioning. This test consisted of five parts: visual scanning, number sequencing, letter sequencing, number-letter sequencing and motor speed. The test-retest reliability During the first part, visual scanning, participants had to cross out the number three on a paper with different numbers and letters. The second part consisted of connecting numbers in ascending order. The third part consisted of connecting letters in alphabetical order. During the fourth part, participants had to sequence connecting the letters and numbers in alphabetical and ascending order (1-A-2-B-3-C etc.). This part relied a lot on participants their mental flexibility. The last part consisted of following a line connecting empty dots, to measure participants their motor speed.
The primary outcome measure used was the time necessary for the fourth part, the alternation condition, corrected for performance on part one and two. This measure was believed to reflect
participants abilities of mental flexibility. Normative data was available to adjust test results for age and educational differences (Delis, Kaplan & Kramer, 2001). The test-retest reliability for this part was r .55, with Cronbach’s alphas for the age groups 40-49, 50-59, 60-69 and 70-79 respectively .74, .81, .80 and .60 (Delis, Kaplan & Kramer, 2001).
Dutch translation of the Rey auditory Verbal Learning Test (RAVLT). The RAVLT is a
memory test designed to measure short-term auditory verbal memory, rate of learning and retention of information. This test consisted of 15 unrelated one syllable Dutch words. Through a tape recorder, these 15 words were played in 25 seconds total. The participants were required to repeat as many words as possible. This process was done five times in total. After 15 minutes, there was a delayed recall part in which participants were asked to list the words they could remember.
Dutch psychological tests are evaluated by the ‘commissie testaangelegenheid Nederland’ (COTAN). According to the COTAN, the reliability is good r ≥ .80, as well as the construct validity r ≥ .30, (Evers, Vliet-Mulder & Groot, 2000). The correct words, as well as the double words and the wrong words were scored. The primary outcome measures were the total amount of words imprinted over the five trials, and the total amount of words remembered in the delayed recall trial.
Paced Auditory Serial Addition Test (PASAT). Cicerone (1997) observed that the PASAT had
two components; a component relating to the processing ability required to complete the task and a component regarding to the speed of information processing. According to Madigan and colleagues (2000), sustained attention, working memory and simultaneously performing several cognitive tasks under specific time constraints, are cognitive functions needed for successful completion of the PASAT. Therefore, the PASAT is a neuropsychological test sensitive to measuring working memory, information processing, and sustained and divided attention.
During this test, participants heard a number ranging from one till six every 3.2 seconds. The participants were required to add the number they heard, to the previous number. For example, they heard the number three and 3.2 seconds later the number two. The answer the participants has to give is the number 5. They were required to do 60 sums. After this, another block of the same task was presented only the interval between the numbers was be 2.8 seconds.
The PASAT shows high levels test-retest reliability with correlations ranging between r .76 and r .95 (Tombaugh, 2006). However, the PASAT is highly susceptible to practice effects (Tombaugh, 2006). Although the pre- and post-test were twelve weeks apart, interpreting these results has to be done while keeping practice effects in mind. The primary outcome measure was the total amount correct on both blocks.
Digit Symbol Substitution Test (DSST). This test is part of the Wechsler Adult Intelligence
Scale(Wechsler, 2000). This test is one of the most widely used instruments for describing the performance of younger and older adults in aging studies (Hoyer, Stawski, Wasylynshyn & Verhaeghen, 2004). According to Hoyer and colleagues (2004), the DSST seems to serve as a robust marker for describing sample characteristics in studies of age differences. In addition, scores on the DSST have been shown to exhibit strong correlations with measures that involve processing speed (Laux & Lance, 1985). During this test, participants had to draw hieroglyphic-like symbols in empty boxes, placed under a digit. The correct symbol for the digit could be found in the look up table. The outcome measure was the number of empty boxes completed in 90 seconds. According to the COTAN, the reliability of this test was good r ≥ .80, and the construct-validity satisfactory with correlations ranging between r .20 and r .29 (Evers, Vliet-Mulder & Groot, 2000). The primary outcome measure was the number of empty boxes the participant was able to fill in. Normative data was available to adjust results for age and educational differences (Wechsler, 2000).
Verbal category and letter fluency test. This is a commonly used neuropsychological test in
experimental settings. The participant has to generate as many words as possible in one minute, according to a specific rule. Letter and category tasks were used to asses semantic knowledge, retrieval abilities, executive functioning and verbal productivity (Gladsjo, Shuman, Evans, Peavey, Miller & Heaton, 1999; Schmand, Groenink & van den Dungen, 2008). There is a specific pattern of age differences in performance found on category fluency tasks. That is, older participants tend to generate fewer words than younger participants on category fluency tasks (Kozora & Cullum, 1995).Both the category and letter fluency tests had to versions which were counterbalanced over participants such that at follow-up they did the other version
For the letter fluency task, participants had to generate as many words as possible within one minuet. The words had to begin with a specific letter. Both versions consisted of three trials. In the first version, the words had to begin with the P, than a G and the third time the R. In the second version the letters were different, namely K, O and M
The category fluency was basically the same as the letter fluency task, only the words did not have to begin with a specific letter, but had to fall in a specific category. Most participants had already performed the commonly used versions of the fluency task, with categories as animals and occupation, four different categories were used. For the first version, the categories were male names and supermarket articles. For the second version the categories were female names and city names. For the category fluency task, a switch condition was added. Participants were asked to alternate between naming a word from category one and naming a word from category two.
According to the COTAN the classic form of the category fluency task shows a satisfactory reliability(.80≥r < .90) and a satisfactory construct-validity (.20 ≥ r < .29) (Evers, Vliet-Mulder & Groot, 2000). The reliability of the letter fluency task is high, r .80, with a parallel- form reliability of r .78 (Schmand, Groenink & van den Dungen, 2008).The primary outcome measure for the category and letter fluency tasks were the number of correct words mentioned. For the letter fluency task, normative data was available to correct test results for age and educational differences (Benton & Hamsher, 1989).
Letter-Number Sequencing (LNS). This task is part of the Wechsler Adult Intelligence Scale
(Wechsler, 2000). This is a task to asses working memory span, sustained attention, concentration and metal manipulation. According to
Ryan, Sattles and Lopez (2000) t
his task shows great differences across age groups. The performance of this task declines with age.During this task the participant heard a sequence of letters and numbers. The participants is required to repeat the sequence and to put the letters and numbers in ascending order. The primary outcome measure was the total number of sequences the participant was able to repeat. According to the COTAN, the reliability of this test was good r ≥ .80, and the construct-validity satisfactory with correlations ranging between r .20 and r .29 (Evers, Vliet-Mulder & Groot, 2000).
Procedure. The current research is part of the Training Project Amsterdam Seniors and Stroke
(TAPASS). The following description will only include the part relevant for the current study. The participants were divided into two groups, that is a stroke patients group and a healthy elderly group. Participants in the stroke group were randomly assigned to the either the intervention training, active control training or the waitlist condition. Participants in the healthy elderly group were randomly assigned to the intervention training condition or the active control training condition. The randomization took place after it was decided that participants met the inclusion criteria.
The baseline measurement took place just before the start of the training. A trained graduate clinical Neuropsychology student conducted the neuropsychological assessment. Both the participant as the assessor were kept blind of the training condition. The test was conducted in a controlled environment at
the University of Amsterdam. After the baseline measurement, training instructions and participants started the training period.
When the training period or waiting period had finished, participants returned to the University of Amsterdam follow up measurement took place. The follow up measurement took place in the same environment and with the same assessor. When participants were assigned to the waitlist control condition, training instructions were given after the follow up measurement and participants started the training period.
Statistical analysis. The dataset was analysed by using IBM SPSS 21 for Macintosh. P-values of .05 or
lower were considered significant. Outliers were examined by using Cook’s and Mahalanobis distances. To standardize the scores of the tests, norm scores provided by the testing manual are used for the DSST, NDS, TMT and the RAVLT. Based on these norms Z-scores could be calculated which were used for the analyses. For the PASAT and fluency tests, Z-scores were calculated by using the mean and standard deviation of the sample split by group. For the follow up Z-scores the same mean an standard deviation of the baseline measurement were used. Additionally a composite score was calculated of the z-scores of all the individual test. The training effect was defined by the difference between the baseline measurement and the follow up measurement for each individual test and for the composite score.
The main analysis investigates the effectiveness of the training. Additionally the in effectiveness between groups and conditions were investigated. This was done with a repeated measures MANOVA with two levels(baseline measurement and follow-up measurement). Data of both groups and conditions were included, with exception of the waitlist condition data. The measurements were the Z-scores of the different individual tests and the Z-scores of the composite score. The within subjects factor was time(i.e. the time between the baseline measurement and the follow-up measurement). The between subjects factors were group and condition. After the main analysis the waitlist condition is being analysed with the same repeated measures MANOVA.
Challenge as a predictor for cognitive plasticity. As hypothesized, challenge induced by the environment could be influencing cognitive plasticity. Therefore it was expected that the more challenge was experienced, the more cognitive plasticity could be induced; enhancing the training effect. Probably, the amount of challenge induced by the training was not only depending on the difficulty level of the training, but also on the level of cognition at the baseline measurement. In other words, the effect of challenge induced by the training on the training effect is moderated by baseline cognition. Moreover, it is likely that the relationship between baseline cognition and the training condition on cognitive plasticity was not linear. There could have been an optimum point for the training effect. The expected relationship is visualised in figure 1.
To test this moderation hypothesise several analysis were carried out, with the training effect of the composite score as the dependent continuous variable, training group as categorical dichotomous independent variable and baseline cognition as continuous independent moderator.
In the first step, correlations were carried out to investigate the data at a group level. Normality of the data was assessed to decide which correlations were appropriate. In the next step, multiple quadratic regression analysis was carried out to further investigate the data at an individual level. The predictor for the training effect was the composite score at baseline, with the models split for group and cognition. By splitting the regression models for group and condition, it could be investigated is the hypothesized relationship as visualised in figure 1 could be observed in the data. Thus if baseline cognition influenced the training effect produced by either one of the training conditions.
The primary goal of the quadratic regression analysis is investigating the mechanism behind the training effect. Because after the training, not all tasks improved significantly, the same analysis was
repeated with a new composite score with only tasks that improved significantly after the training (LNS, DSST and RAVLT direct-recall). The same holds for a new composite score of the training effect.
Figure 1: Hypothesized quadratic relationship between cognitive performance at baseline and the
effectiveness of the training
.
Results
Based on Mahalanobis distance (18.57) one participant from the healthy elderly group could be considered an outlier. However this participant showed no irregularities during the neuropsychological assessment and during the training. Cooks distance (0.003) indicates that the possible outlier is not largely influencing the data. Therefore analysis were conducted twice: first with the possible outlier and later without. These outputs were compared and no differences could be found. Reported results include all participants. The assumption for normality has been assessed by using the Shapiro-Wilk test of normality. It has been found that not all measures meat the assumption of normality. These measures and the corresponding statistics can be found in table 2. These results influence the correlation statistics that was used. Box’s M test indicates that the within-group covariance matrices are equal, M(210, 5544) = 492.76,
p>.05. There were only two levels of the within-subjects factor. Therefore the assumption of sphericity
has been met. 0 2 4 6 8 10 12 0 2 4 6 8 10 12 14 16 18 20 T rai ni ng e ff ec t
Cognitive performance at baseline
Active control training Intervention training
Table 2.
significant non-normal distributed measures
Group Condition Measure W df p
CVA Active control
LNS pre-test .87 20 <.05
LNS post-test .89 20 <.05
PASAT pre-test .91 20 <.05
Fluency semantic single post-test .96 20 <.01
RAVLT direct recall post-test .90 20 <.05
Intervention PASAT pre-test .93 21 <.05
Healthy elderly
Active control
DSST pre-test .89 19 <.05
DSST post-test .86 19 <.05
Fluency semantic single pre-test .88 19 <.05 Fluency semantic switch pre-test .87 19 <.05 Fluency semantic switch post-test .80 19 <.01 Intervention
TMT post-test .90 25 <.05
PASAT pre-test .91 25 <.05
PASAT post-test .85 25 <.01
Composite cognition at baseline .89 25 <.05
Main Analysis: repeated measures MANOVA
It was hypothesized that the intervention training positively influenced the training effect. The training effect in the active control condition would be smaller. It was expected that no differences between the baseline measurement and follow-up measurement would be observed in the waitlist condition. Additionally it was hypothesized that the training effect of both training conditions would be larger for the CVA group than for the healthy elderly group.
When analysing the data without the waitlist control condition, there was a significant main effect of time on the composite score of cognition, F(1, 81) = 25.71, p<.01 r = .24. However, no significant interaction has been found between time and the training condition F(1, 81) = .049, p>.05 r = .001. Thus the active control condition produced the same effect as the intervention training. The same holds for the interaction between time and groups F(1, 81) = .01, p>.05 r =<.01 and time, groups and condition, F(1, 81) = .01, p>.05 r <.01.
For the individual tasks in this analysis, there was also a main effect of time found on the performance on the PASAT, F(1,81) = 20.12, p<.01, r = .20 with no additional interactions with group F(1, 81) = 1.05,
p>.05 r = .01. or condition, F(1, 81) = 0.8, p>.05 r = <.01. The main effect of time on the performance of
the LNS was also significant with F(1, 81) = 16.25, p<.01, r = .18. There were no interactions with condition, F(1, 81) = .02, p>.05, r = <. 01 and with group, F(1, 81) = .38, p>.05 r = <.01. There is a main effect of time found on the performance on the direct recall part of the RAVLT, F(1, 81) = 11.59, p<.01, r = .13 but no interactions with condition, F(1, 81) = 1.23, p>.05, r = .01 and group, F(1, 81) = 1.43, p>.05,
r = .02. Lastly there was a main effect found of time on the performance of the DSST, F(1, 81) = 18.95, p<.01, r = .19 but no interactions with condition, F(1, 81)=1.69, p>.05, r = .02 or group, F(1, 81)<.01, p>.05, r<.01. These results indicate that performance on the PASAT, RAVLT, DSST and LNS increase
over time. These effects do not differ between groups and conditions.
No significant effects were found on any of the other tasks and any of the interaction effects between time, condition and group. In the waitlist condition there was a significant effect found of time on the performance of the PASAT, F(1, 19) = 6.49, p <.05, r = .26. Thus participants in the waitlist control condition improved on performance of the PASAT. Differences between the baseline measurement and the follow up measurement for the training conditions are visualized in figure 2 for the training conditions and figure 3 for the waitlist conditions. Since no interactions between the training effect and group and training conditions were significant, the groups and training conditions are combined in these charts. The mean and standard deviations of the training effects are shown in table 3.
Table 3.
Mean and standard deviation of the Z-score difference between the baseline measurement and the follow-up measurement (i.e. the training effect) of both groups and training conditions combined.
Measure minimum maximum M SD
Composite score -.48 .90 .16** .28 DSST -1.30 1.30 .21** .45 LNS -2.00 4.30 .41** .93 TMT -3.00 2.30 .84 .84 PASAT -2.51 2.65 .35** .69 Letter fluency -1.45 3.55 .10 .75
Category fluency switch condition -2.54 4.67 -.03 1.12
Category fluency single condition -3.54 1.84 -.13 .92
RAVLT direct recall -1.80 2.30 .29** .79
RAVLT delayed recall -3.10 3.50 -.14 1.17
**. Training effect is significant at the .01 level.
Figure 2. Visualisation of the difference between the test performance before and after the training.
Because no differences between groups and training conditions were found, results shown are with both groups and training conditions combined. Note: **. Difference is significant at the .01 level.
Figure 3. Visualisation of the difference between the test performance before and after the waiting period.
Note: *. = Difference is significant at the .05 level
Explorative analysis: Correlations between baseline cognition and training effect
In a more explorative nature correlations were carried out to investigate the relations between the training effect on different Neuropsychological test and the composite score of the baseline measurement. Because not all data is distributed normally, as can be seen in table 2, the non-parametric spearman correlations were carried out. All the different training effects were correlated with baseline cognition, split by group and condition. There was a significant negative relationship found between the high switch trainings effect of the DSST with baseline cognition in the stroke patients group, ρ = -.47, p (two-tailed) <.05. In other words, when the composite score of the baseline measurement increased, the training effect of the difference in performance on the DSST before and after the training decreases. No other significant correlations were found as can be seen in table 4.
In addition correlations were carried out to investigate the relation between the number of days since the stroke and the training effect. Against the expectations, no significant correlations were found as can be found in table 4.
Table 4.
Spearman’s rho with a 2-tailed significance test. Correlations between the different neuropsychological tests and the composite score of the baseline measurement.
*. Correlation is significant at the .05 level.
Explorative analysis: Composite score of cognition as a predictor for the training effect
Quadratic regression analysis were carried out to further investigate the data at an individual level. It was hypothesized that there is a quadratic trend between the composite score at baseline and the training effect. Because the two training conditions possibly differ in the amount of challenge they induce, there will be differences in this quadratic trend. The predicted trend is visualized in figure 1.
In the healthy elderly group the composite score at baseline was found not to be a significant predictor F(2, 41) = .36, p = .67 for the training effect. When looking separately at the training conditions in the healthy elderly group, the quadratic regression coefficient remained insignificant for the active control training, F(2, 16) = 2.42 , p=.12 and the intervention training, F(2, 22) = .18, p= .84.
Group Condition Measure Correlation with composite
score at baseline
CVA
Control
Composite score training effect -.11
RAVLT direct recall training effect -.09
RAVLT delayed recall training effect .06
Semantic fluency training effect -.04
Phonetic fluency training effect .03
PASAT training effect -.18
TMT training effect -.26
LNS training effect .44
DSST training effect -.17
Intervention
Composite score training effect -.05
RAVLT direct recall training effect .02
RAVLT delayed recall training effect .11
Semantic fluency training effect -.10
Phonetic fluency training effect -.17
PASAT training effect .23
TMT training effect .11
LNS training effect -.36
DSST training effect -.47*
Healthy
Control
Composite score training effect -.28
RAVLT direct recall training effect -.17
RAVLT delayed recall training effect -.26
Semantic fluency training effect -.07
Phonetic fluency training effect -.07
PASAT training effect -.81
TMT training effect .12
LNS training effect .34
DSST training effect .05
Intervention
Composite score training effect -.01
RAVLT direct recall training effect .04
RAVLT delayed recall training effect -.24
Semantic fluency training effect .24
Phonetic fluency training effect -.12
PASAT training effect -.01
TMT training effect -.08
LNS training effect .10
In the stroke patients group the composite score at the baseline measurement was also found not to be a significant predictor for the training effect, F(2, 38) = .03, p = .98. Similarly, when analysing the training conditions separately the quadratic regression coefficient remained insignificant for the active control training, F(2, 17) = .04 , p=.96 and the intervention training, F(2, 18) = .05, p= .95.
Because after the training not all tasks improved significantly, the relationship between the composite score at baseline and the training effect might still be present. This relationship might become apparent when analysing only the tasks the training seems to have an influence on. Therefore, the same quadratic regression analysis was repeated with a new composite score of the baseline measurement. This composite score consists of only tasks that improved significantly after the training (LNS, DSST and RAVLT direct recall). The same holds for a new composite score of the training effect.
The quadratic regression was found not to be significant for the healthy elderly, F(2, 41) = 1.64, p>.05 and the stroke patients group, F(2, 38) = 2.21, p>.05. with both training conditions combined. The linear trend was found to be significant for the stroke patients group, with both training conditions combined F(1, 39) = 4.49, p<.05, R2=.10. However, the proportion explained variance appeared to be very minimal
(10%).
When splitting the groups further into the training conditions, the quadratic regression trend for the healthy elderly in active control training reveals significance, F(2, 16) = 5.038, p<.05, R2=39. This is
notably different from the healthy elderly in the intervention training condition where the regression is insignificant F(1, 17) = .08, p>.05, R2<.01. In other words, 39% of the observed variance in the active
control training for the healthy elderly group can be explained by the composite score of the baseline measurement. The same is appears not to be true for the intervention training. The difference between the two conditions is visualized in figure 4 and 5.
In the stroke patients group, the quadratic regression of the active control training condition is insignificant, F(2, 17) = .25, p>.05. The same holds for the quadratic regression in the intervention training condition, F(2, 18) = 2.91, p=.80, R2 = .23. Notably the linear regression is significant F(1, 19) =
6.11, p<.05, R2 = .242. This regression slope is visualized in figure 6. The quadratic regression in the
waitlist control condition became insignificant with the new composite scores, F(2, 17) = 2.21, p>.05.
Figure 4. Significant quadratic trend for the active control training condition in the healthy elderly group
with the new composite scores. The new baseline measurement and training effect composite scores consist of the DSST, LNS and RAVLT-direct recall Z-scores.
Figure 5. insignificant quadratic trend for the intervention condition in the healthy elderly group with the
new composite scores. The new baseline measurement and training effect composite scores consist of the DSST, LNS and RAVLT-direct recall Z-scores. This is notably different from the trend line in the active control training condition in the healthy elderly group.
Figure 6. Significant linear trend and insignificant quadratic trend for the intervention training in the stroke
patients group with the new composite scores. The new baseline measurement and training effect composite scores consist of the DSST, LNS and RAVLT-direct recall Z-scores.
Discussion
In this study, the effect of a computerized cognitive training was tested within healthy elderly and CVA patients. The results indicate that, after the training period, there was an slight improvement found on selective the DSST, LNS, RAVLT direct recall and on the composite score of cognition. However, there were no differences found of this effect between the intervention training and the active control training. Moreover, no difference in training effect were found between healthy elderly and stroke patients. These improvements on performance were not present in the waitlist group, except for improvement on the PASAT. Performance on this task was equal for the waitlist condition as for both training conditions. Therefore the improvement of this task is more likely to be due to learning and test-retest effects than to be due to the training. This is consistent with the current literature, that the PASAT is susceptible to practice effects (Tombaugh, 2006).
The waitlist condition was only present in the CVA group, therefore in this group it seems likely that the found improvements are due to both training programs. In the healthy elderly group, such a control is lacking. Because there were no indications for differences between the CVA patients group and the healthy elderly group on improvement of cognition, it seems likely that in both groups the found improvement is due to the training effect.
Not all tasks improved after the training. So what does it mean that only the DSST, LNS and RAVLT direct recall improve? To begin with discussing improvement on the DSST. This is a widely used instrument to describe age differences. Younger people tend to score better than older people (Hoyer, Stawski, Wasylynshyn & Verhaeghen, 2004). As described earlier, the DSST shows a strong correlation with measures that involve processing speed (Laux & Lance, 1985). Moreover, according to Hoyer and colleagues (2004), the DSST can be an indicator for cognitive decline in aging. The finding that the performance on this task improves after the training period could be an indication that the processing speed of participants is improving. Even so, an indication that both training programs trigger the reversal of cognitive decline in this domain, although the effects are small.
The next task that showed improvement is the LNS. This working memory span, sustained attention, concentration and mental manipulation test. This is a task designed to measure working memory span, sustained attention, concentration and mental manipulation. Interestingly, just as the DSST, the LNS shows decline with age (Ryan, Sattles & Lopez, 2000). This could be another indication that the negative effects of age on cognition are being slightly reversing after either one of the training programs.
The last task on which participants showed an improvement in performance on after the training is the RAVLT. However, this improvement is not found in the delayed recall phase. The delayed recall phase depends on different aspects of memory than the direct recall phase. It seems that that working memory and imprinting improves, but that consolidation and reproduction are not affected by the training.
To summarize the results thus far, it seems like there is evidence that supports that the performance of three tasks improves after following a training. All three involve working memory. Working memory can be defined as the ability to hold and manipulate information during a short time period. It makes us able to respond based on that brief internal representation (Westerberg et all., 2007). Westerberg and colleagues (2007) investigated the effect of a working memory training on stroke patients. Their results confirm the findings of the current study, that the training caused a significant improvement of working memory capacity. Working memory problems are very often present in stroke patients (Westerberg et all., 2007) and working memory capacity declines with age (Salthouse, 1991). In addition, Salthouse (1991) found that age related cognitive differences in other aspects were reduced by statistically controlling for working memory deterioration. Many of the age differences in cognition appear therefore to be mediated by age-related reductions in working memory (Salthouse, 1991). Training the capacity of working memory could therefore be a step in reversing the negative effects of age on cognition and in extend to the long term effects of a CVA.
Another finding was that the overall improvement was found to be significant. That is, the composite score of all the tasks improved after the training. Possibly because of the improvement of the above
mentioned task, but also due to the summation of small insignificant improvements on the other tasks. All these small insignificant effects combined could lead to a more significant score. However, the question remains if this improvement on average can be extended to a subjective improvement of cognition, or that it remains just the summation of all the little effects. Therefore, it is recommended that more extensive research in the subjective feeling of cognitive improvement is done.
As mentioned earlier, the results show that both the intervention training as the active control training produce the same gain on selective neuropsychological tasks. This is a discrepancy with the hypothesize made in the review done by Buitenweg, Murre and Ridderinkhof (2012). They hypothesized that cognitive processes can be stimulated with more variability in a training program. This variability requires the subjects to integrate multiple cognitive domains rather than training separate cognitive mechanisms. Furthermore, they state that a training program might result in cognitive improvement if the program includes frequent switching between various training tasks. It seemed that especially working memory capacity could benefit from the computerized cognitive flexibility program, but that a higher variability and switching component did not contribute to the training effect.
Perhaps the switching component used in the intervention training was not enough to make the training more effective. Other studies, such as the study done by Minear and Shah (2008) demonstrated that frequent switching between two simple tasks can improve executive functioning. A similar effect of task switching was found in the study done by Karbach and Kray (2009). Both studies included a task switching paradigm in which participants switched between simple tasks every trial. Perhaps the task switching component in the intervention training used in this study failed to show the desired effect, because the switching occurred to slowly. In addition, in the task switching paradigm used by Karbach and Kray (2009) and Minear and Shah (2008), participants switched back and forth between the tasks. In the current intervention training, participants switched, but only performed a task once per training session. Therefore it is recommend that future research include a task switching component in which fewer tasks have to be performed during one training session. Than a switching paradigm as used by Minear and Shah (2008) and Karbach and Kray (2009) can be implemented in the currently used intervention training. The next research topic was investigating the relation between the baseline cognitive abilities and the training outcome. It was expected that the lower the cognitive abilities at baseline, the higher the training effect. The detailed quadratic expectation is visualized in figure 1. No significant correlations other than the relation between the DSST and baseline cognition in the intervention training condition in the CVA group was found. Moreover, the composite score cognition could not fundamentally predict the training effect.
It can be argued that to explore if baseline cognition has anything to do with the training effect, it can be informative to only look at the tasks that were affected by the training. It was hypothesized that the intervention training would be more challenging than the active control training. The performance level on the cognitive tests at baseline, in combination with the level would account for the challenge induced by the environment. If a participant scores high at baseline, the training would be less challenging than for a participant that scores low at baseline. However, no differences in training effect between the training conditions were found, indicating that both training programs elicit the same effect. When comparing the amount of variance that can be explained by the baseline composite score, it was found that only the quadratic regression trend of baseline cognition could predict the training effect, but only in the healthy elderly group who received the active control training. This line predicts, that low cognitive scores are associated with a relatively high training effect, average scores with a relatively low effect and higher scores with a relatively high training effect. For as in the intervention training, the cognitive score at baseline is not associated with the training effect.
Because the improved tasks involve working memory improvement, the active control training would improve working memory best in patients with a below or above average cognition score. It remains unclear and counter intuitive why this relationship appears in this particular pattern, only in the active control training with healthy elderly. The proportion explained variance however is moderate and the
number of participants included in this analysis relatively low. Therefore, the found trend could be coincidence, a product of chance.
In the CVA group, the training effect in the active control training condition could not be predicted by baseline cognition. In the intervention training on the other hand, the training effect could be predicted by the new composite score at baseline. The found relationship was linear. Again the proportion explained variance was moderate. This relationship indicates that the better the performance at baseline, the lower the training effect. This is as hypothesized, although it was hypothesized that this trend would be visible in the healthy elderly too.
Moreover the results fail to show the hypothesized relationship between the cognitive scores at baseline and the training effect. This is inconsistent with the findings of Greenwood (2007). He states that cognitive deficits induce changes in processing strategies that drive plasticity, leading to increased regional activation. This could lead to an increase of the training effect. Based on this statement, it could be argued that that the cognitive deficits in the sample were too little to find this effect. Additionally only participants who completed the training program were included into the analysis. The participants who failed to continue the training program could be the participants that find the training too difficult. Therefore the current findings could be biased.
Conclusion
It becomes clear that the results do not support the hypothesis that differences in the mechanism of cognitive plasticity occur between chronic CVA patients and healthy elderly. No differences in training effect were found between the two groups, indicating that a training program affects both groups equally. In addition the relationship between the cognitive performance at baseline and the amount of cognitive improvement does not show a concrete difference between CVA patients and healthy elderly.
Both the intervention training as the active control training were found to be effective on tasks that involve in working memory capacity and processing speed; supporting the results of Westerberg and colleagues (2009). The improved tasks after the training program were tasks generally used to describe cognitive decline with age. However, the results show only small improvements on these tasks, therefore the training programs are not yet clinically relevant. The results form an indication that steps can be made in the reversal of the negative effects of aging and the long term effects of a stroke. More elaborate research into this topic is recommended. Ideally with a sample containing participants with large differences in cognitive abilities. This in combination with different levels of difficulty of the same training, would provide more information about the (optimal) relationship between challenge and training gain. Moreover, the training program failed to show improvements on other modalities, therefore more elaborate research in designing training tasks and long term effect studies are recommended to further improve the training program. By improving multiple modalities the objective and subjective cognitive gain could be maximized, translating into improvement of quality of life.
References
Axelrod, B. N., & Henry, R. R. (1992). Age-related performance on the Wisconsin Card Sorting, Similarities, and Controlled Oral Word Association Tests. The Clinical Neuropsychologist, 6, 16-26. Benton, A. L., & Hamsher, K. (1989). Multilingual aphasia examination. Iowa City: AJA associates.
Brandt, J., Spencer, M., & Folstein. M. (1988). The telephone interview for cognitive status. Neuropsychiatry,
Neuropsychology, and Behavioural Neurology, 1, 111-117.
Cicerone, K. D. (1997). Clinical sensitivity of four measures of attention to mild traumatic brain injury.
The Clinical Neuropsychologist, 11, 266–272.
Delis, D. C., Kaplan, E., & Kramer,. J. (2001). Delis-Kaplan executive function system. San Antonio: Psychological Corporation.
Evers, A., Vliet-Mulder, J. C., & Groot, C. J. (2000). Documentatie van tests en testresearch in Nederland,
Deel I en II: testbeschrijvingen. Amsterdam/Assen: NIP/van Gorcum.
Faul, F., Erdfelder, E., Lang, A.-G., & Buchner, A. (2007). G*Power 3: A flexible statistical power analysis program for the social, behavioral, and biomedical sciences. Behavior Research Methods, 39, 175-191.
Gallagher, M., Bizon, J. L., Hoyt, E. C., Helm, K. A., & Lund, P. K. (2003). Effects of aging on the hippocampal formation in a naturally occurring animal model of mild cognitive impairment. Experimental Gerontology, 38, 71–77
Gladsjo, J. A., Schuman, C. C., Evans, J. D., Peavy, G. M., Miller, S. W., & Heaton, R. K. (1999). Norms for letter and category fluency: Demographic corrections for age, education, and Ethnicity. Psychological
Assessment, 6¸147-178.
Grady, C. L., Springer, M. V., Hongwanishkul, D., McIntosh, A. R., & Winocur, G. (2006). Age related changes in brain activity across the adult lifespan. Journal of Cognitive Neuroscience, 18, 227-241 Greenwood, P.M. (2007). Functional plasticity in cognitive aging: review and hypothesis.
Neuropsychology, 21, 657-673.
Günther, V. K., Schäfer, P., Holzner, B. J. & Kemmler, G. W. (2003). Long-term improvements in cognitive performance through computer-assisted cognitive training: a pilot study in a residential home for older people. Agin & Mental Health, 7, 200-206.
Hoyer, W. J., Stawski, R. S., Wasylyshyn, C. & Verhaeghen, P. (2004). Adult age and Digit Symbol Substitution performance: A meta-analysis. Psychology and Aging, 19, 211-214.
Karbach, J., & Kray, J. (2009). How useful is executive control training? Age differences in near and far transfer of task-switching training. Developmental Science, 12, 978-990.
Kozora, E., & Cullum, C. M. (1995). Generative naming in normal aging: Total output and qualitative changes using phonemic and semantic constraints. The Clinical Neuropsychologist, 9, 313- 325.
Laux, L. F., & Lane, D. M. (1985). Information processing components of substitution test performance. Intelligence, 9, 111–136.
Lövdén, M., Bäckman, L., Lindenberger, U., Schaefer, S., & Schmiedek, F. (2010). A theoretical framework for the study of adult cognitive plasticity. Psychological Bulletin, 136, 659-676.
Lustig, C., Shah, P., Seidler, R., & Reuter-Lorenz, P. A. (2009). Aging, training, and the brain: a review and future directions. Neuropsychology Review, 19, 504-522
Madigan, N. K., DeLuca, J., Diamond, B. J., Tramontano, G., & Averill, A. (2000). Speed of information processing in traumatic brain injury: Modality specific factors. Journal of Head Trauma Rehabilitation, 15, 943–956.
Mattson, M.P., Duan, W., Chan, S.L., Cheng, A., Haughey, N., Gary, … Furukawa, K. (2002).
Neuroprotective and neurorestorative signal transduction mechanisms in brain aging: modification by genes, diet and behavior. Neurobiology of Aging, 23, 695-705.
Minear, M., & Shah, P. (2008). Training and transfer effects in task switching. Memory & Cognition, 36, 1470-1483.
