Neurobiology of Learning and Memory 114 (2014) 51–57
Contents lists available at ScienceDirect
Neurobiology of Learning and Memory journal homepage: www.elsevier.com/locate/ynlme
The effect of practice on random number generation task: A transcranial direct current stimulation study Fioravante Capone a,b,⇑, Gianluca Capone c, Federico Ranieri a,b, Giovanni Di Pino a,b, Gianluca Oricchio d, Vincenzo Di Lazzaro a,b a
Institute of Neurology, Campus Bio-Medico University, Via Alvaro del Portillo 200, 00128 Rome, Italy Fondazione Alberto Sordi – Research Institute for Ageing, Rome, Italy Department of Economic Geography, Urban and Regional Research Centre Utrecht (URU), Utrecht University, The Netherlands d Campus Bio-Medico University, Via Alvaro del Portillo 200, 00128 Rome, Italy b c
a r t i c l e
i n f o
Article history: Received 8 January 2014 Revised 22 April 2014 Accepted 27 April 2014 Available online 6 May 2014 Keywords: Random number generation Practice tDCS Cognitive plasticity Executive functions Brain stimulation
a b s t r a c t Random number generation (RNG) is a procedurally-simple task related to specific executive functions, such as updating and monitoring of information and inhibition of automatic responses. The effect of practice on executive functions has been widely investigated, however little is known on the impact of practice on RNG. Transcranial direct current stimulation (tDCS) allows to modulate, noninvasively, brain activity and to enhance the effects of training on executive functions. Hence, this study aims to investigate the effect of practice on RNG and to explore the possibility to influence it by tDCS applied over dorsolateral prefrontal cortex. Twenty-six healthy volunteers have been evaluated within single session and between different sessions of RNG using several measures of randomness, which are informative of separable cognitive components servicing random behavior. We found that repetition measures significantly change within single session, seriation measures significantly change both within and between sessions, while cycling measures are not affected by practice. tDCS does not produce any additional effect, however a sub-analysis limited to the first session revealed an increasing trend in seriation measure after anodal compared to cathodal stimulation. Our findings support the hypothesis that practice selectively and consistently influences specific cognitive components related to random behavior, while tDCS transiently affects RNG performance. Ó 2014 Elsevier Inc. All rights reserved.
1. Introduction Random number generation (RNG) is a procedurally-simple task widely exploited to investigate cognitive function. Subjects are instructed to say aloud, in a random order, the numbers from 1 to 9, in synchrony with a pacing stimulus. Despite its apparent simplicity, RNG involves several mental processes. Indeed, it requires adopting the correct strategy, on the basis of the selection of appropriate responses and suppression of those that appear to violate the instructions and the subject’s concept of randomness, monitoring the output and eventually modifying the strategy of Abbreviations: RNG, random number generation; tDCS, transcranial direct current stimulation; DLPFC, dorsolateral prefrontal cortex; TMS, transcranial magnetic stimulation; REP, repetition; POKER, poker; GAP, median repetition gap; SDD, standard deviation of digits; TPI, turning point index; RUNS, runs; AC, combined adjacency; CST, total count score. ⇑ Corresponding author at: Institute of Neurology Campus Bio-Medico University, Via Alvaro del Portillo 200, 00128 Rome, Italy. Fax: +39 06 225411936. E-mail address: [email protected] (F. Capone). http://dx.doi.org/10.1016/j.nlm.2014.04.013 1074-7427/Ó 2014 Elsevier Inc. All rights reserved.
production (Jahanshahi et al., 1998). In particular, RNG is strongly related to specific executive functions, such as updating and monitoring of information, and inhibition of prepotent responses of counting or cycling through the set of numbers (Miyake et al., 2000). Executive functions, such as dual-task performance (Bherer et al., 2005), information updating (Dahlin, Nyberg, Bäckman, & Neely, 2008), task switching (Kramer, Hahn, & Gopher, 1999) and inhibition of prepotent responses (Wilkinson & Yang, 2012) can be trained and this training mostly reflects the establishment of cognitive plasticity (Karbach & Schubert, 2013). Although the ascertained role of RNG in assessing executive functions, little research faced RNG-induced plasticity. To our knowledge, only few studies have examined the effect of practice in RNG task, providing contrasting results (Evans & Graham, 1980; Jahanshahi, Saleem, Ho, Dirnberger, & Fuller, 2006; Peters et al., 2007). Carrying out the RNG task requires the activation of different brain regions, like the left dorsolateral prefrontal cortex (DLPFC), the anterior cingulate cortex, the superior parietal cortex, the right
52
F. Capone et al. / Neurobiology of Learning and Memory 114 (2014) 51–57
inferior frontal cortex, and the cerebellar hemispheres (Jahanshahi, Dirnberger, Fuller, & Frith, 2000). Functional imaging (Jahanshahi et al., 2000) and transcranial magnetic stimulation (TMS) (Jahanshahi et al., 1998) studies indicate that the left DLPFC plays a critical role in this network by exerting an inhibitory influence over temporal–parietal cortex, in order to suppress counting. Interestingly, TMS-induced modulation of the left DLPFC affects the RNG performances in a stimulation frequency-dependent fashion: the tendency to produce ascending or descending ordered sequences (counting bias) is significantly reduced after inhibitory (1 Hz) repetitive TMS, while is significantly increased after excitatory (10 Hz) stimulation (Knoch, Brugger, & Regard, 2005). Transcranial direct current stimulation (tDCS) is a safe and reliable technique that allows to modulate, non-invasively, brain functions and their plasticity. In this technique, a weak current is applied constantly over time to increase (anodal stimulation) or decrease (cathodal stimulation) the excitability of the neuronal populations underlying the active electrode (Nitsche et al., 2008). It has been demonstrated that tDCS can produce transient behavioral changes and influence cognitive functions (LevasseurMoreau & Fecteau, 2012). When performed in parallel with training protocols, it showed to be able to enhance the protocol effects. In particular, tDCS over the frontal regions, combined with cognitive training, seems to be able to modulate executive functions and improve the ability to inhibit responses, as Ditye et al. demonstrated by means of the Stop Signal Task (Ditye, Jacobson, Walsh, & Lavidor, 2012). In the present study, we evaluated the effect of practice on RNG performance to induce training plasticity. To this aim, we employed several different measures of randomness, because each of them is considered to reflect the efficiency of separable cognitive components servicing RNG behavior (Towse & Neil, 1998). Moreover, we tested whether tDCS could influence the training-induced plasticity. In the light of the specific role played by the left DLPFC, we applied either anodal or cathodal stimulation over this area during the RNG task. Previous evidence indicates that performance improvements can occur within-session (on-line effects), as well as between training sessions (off-line effects), i.e., the performance at the beginning of a given session could be different from the performance at the end of the previous session (Robertson, PascualLeone, & Miall, 2004). To this purpose, we evaluated the effect of practice and tDCS within single sessions of RNG and between sessions of RNG separated by at least 48 h. 2. Material and methods 2.1. Participants 26 Healthy subjects (9 male and 17 female) have been enrolled in the study. They had no previous history of neurological or psychiatric disorders and were not taking any medication at the time of the assessment. All the subjects were naive to the RNG task and were not explicitly informed of the experimental variables tested. The mean age was 28 years (range 23–49, SD 5.05). Participants signed a written informed consent prior to the participation in this study, which was approved by the Local Ethics Committee. The research was completed in accordance with the Helsinki Declaration. 2.2. Design A within-subject repeated measures design was used. There were three conditions: RNG coupled with sham tDCS; RNG with left anodal/right cathodal tDCS; RNG with right anodal/left cathodal tDCS. All participants took part in all conditions. The order
the three conditions was pseudo-randomized across participants. The interval between the sessions was at least 48 h. Participants were blind to conditions. 2.3. Random number generation Each session consisted of four consecutive runs. On each run, participants were asked to say aloud a sequence of 100 numbers, each number ranging from 1 to 9, as much random as possible at a metronomic rate of 1.2 Hz. The first two runs were performed without stimulation and were separated by 1 min; once the second run ended, tDCS (either sham or real) was started to last until the end of last run. After 5 min, two more RNG runs, separated by a 1 min interval, were performed during the stimulation. The concept of randomness was explained to the participant using the analogy of picking pieces of paper numbered 1–9 out of a hat. All subjects were naive to the RNG task. 2.4. Measures of randomness A multiplicity of measures to quantify deviation from randomness has been employed in the literature. Towse and Neil have identified four factors by means of a principal component analysis of 16 measures (Towse & Neil, 1998). Ginsburg and Karpiuk (1994) and Peters et al. (2007) have summarized, by means of a factor analysis, several measures in three main factors: a seriation factor, associated to the inhibition of prepotent responses, a cycling factor, associated to the successful monitoring of previous output, and a repetition factor, associated to output inhibition. In our analysis, we employed two measures for each factor identified from the work by Peters and colleagues and added the two measures by Towse and Neil and Jahanshahi and colleagues that seem to cover better the inhibition of prepotent responses (Jahanshahi et al., 2006; Peters et al., 2007; Towse & Neil, 1998). 2.4.1. Repetition measures Such measures refer to the repetition of the same digit within a certain interval. Repetition (REP) counts the number of identical, adjacent digits in the sequence. Poker (POKER) is the number of repetitions that occur between two and five successive digits. 2.4.2. Cycling measures Such measures refer to the attempt to use every possible alternative before repeating any digit. The Median Repetition Gap (GAP) is the median of the intervals between two identical digits. The Standard Deviation of Digits (SDD) is the standard deviation of the frequencies distribution of the nine possible digits. 2.4.3. Seriation measures Such measures refer to the tendency to use the natural order of numbers in the sequence. The Turning Point Index (TPI) is the ratio between observed and expected turning points (in percentage form). A turning point is any point in the sequence where there is a shift from an ascending to a descending subsequence (or vice versa). Runs (RUNS) measure the variance of the lengths of ascending subsequences. The Combined Adjacency (AC) is the ratio (in percentage form) between the number of adjacent pairs of digits (also called Series) and the total number of pairs of digits in the sequence. The Total Count Score (CST) is given by the sum of squared lengths of ascending and descending subsequences in steps of one and two. 2.5. Transcranial direct current stimulation tDCS was applied by a battery-driven constant-current stimulator (Eldith-NeuroConn GmbH, Ilmenau, Germany) via
F. Capone et al. / Neurobiology of Learning and Memory 114 (2014) 51–57
conductive-rubber electrodes, placed in two saline-soaked sponges (5 7 cm) and current output was monitored by a built-in ampere meter. tDCS was performed according to the paradigm described in Fecteau et al. (2007). For left anodal/right cathodal stimulation, the anode was placed over the left F3 (international EEG 10/20 system) and the cathode electrode over the right F4. For right anodal/left cathodal stimulation, the polarity was reversed: the anode was placed over F4 and the cathode electrode over F3. All along the active stimulation, participants received a constant current of 2 mA. Subjects stimulated with tDCS are able to consciously perceive only the beginning of the stimulation, through a sense of itching that localizes underneath the electrode lasting few seconds. The sham stimulation was carried out placing the electrodes in the same position as for active stimulation, but in order to keep subjects blind toward the applied stimulation, the stimulator was kept activated only for 30 s. This procedure made the participants feel the initial itching sensation associated with tDCS, but without receiving any active current for the rest of the stimulation period (Gandiga, Hummel, & Cohen, 2006) and did not allowed participants to distinguish between real and sham tDCS conditions (see Fig. 1). 2.6. Data analysis We analyzed the on-line (within-session) and off-line (between sessions) effects of practice and of tDCS on the RNG measures of randomness. A graphical representation of the effect of practice and tDCS is provided in Figs. 2 and 3, respectively. Fig. 2 shows the average and standard deviation values of the RNG measures across sessions and runs; it also depicts the expected value from a computer-generated pseudorandom series. Fig. 3 shows the average values of the RNG indicators across stimulation types. In our statistical analysis, we used two-way repeated measures ANOVA, separately for each RNG measure. In many cases, Mauchly’s test indicated that the assumption of sphericity was violated, invalidating regular F statistics. Therefore, we employed a multilevel model approach (Goldstein, 2011).
53
Our model can be written in the following compact form:
yijk ¼ b0 þ
X bN xNijk þ v k þ ujk þ eijk N
where y represents our dependent variable (a measure of randomness) and the bN are fixed effect coefficients of the corresponding variables xN. The random structure of the model includes random intercepts for individuals (vk), for sessions within individuals (ujk), and for runs or stimulation type within sessions (eijk). The analysis has been performed using the statistical software R (version 2.15.2 http://www.r-project.org/). In order to analyze practice effects, we include as fixed effect variables session (off-line effects), runs (on-line effects) and their interaction. In Table 1, for each measure of randomness, we report the results of the full specification (Full), the best specification selected according to the Akaike Information Criterion (Sel) if different from the full specification, and the best specification considering session and run as continuous rather than categorical variables (Cts) that allows more flexibility in the choice of parameters. We also report a likelihood ratio test that compares each model to a baseline model including only the random structure. The last five columns report the values of the contrasts (Full, Sel) and parameters (Cts) coefficients. In all cases, we use maximum-likelihood estimation. In order to analyze tDCS effects, we included session and stimulation type as fixed effects, and we estimated random intercepts between different stimulation types (rather than runs). In Table 2 we report contrasts coefficients for stimulation type: we contrast the two runs before stimulation vs. the two runs after stimulation (Pre vs. Post), the runs with sham stimulation vs. the runs with active stimulation (Sham vs. Stim), and the runs with cathodal stimulation vs. the runs with anodal stimulation (Cat vs. An). As in the previous case, we report a likelihood ratio test vs. the baseline model. Since practice effects could have washed out the tDCS effect, we also performed our analysis by restricting the sample to the first session only.
Fig. 1. Study design. A within-subject repeated measures design was used. There were three conditions: RNG with sham tDCS; RNG with left anodal/right cathodal tDCS; RNG with right anodal/left cathodal tDCS. The order these conditions was pseudo-randomized across participants. The interval between the sessions was at least 48 h. Every session consisted of four consecutive runs. On each run, participants were asked to say the numbers 1–9 in a sequence as random as possible (100 trials at a metronomic rate of 1.2 Hz). The first two runs were performed without stimulation and were separated by 1 min; then tDCS (sham or real) was started and after 5 min, two more RNG runs separated by 1 min, were performed during stimulation.
54
F. Capone et al. / Neurobiology of Learning and Memory 114 (2014) 51–57
Fig. 2. Effect of practice on random number generation performance within-session and between sessions. (A) REP = Repetition; (B) POKER; (B) GAP = Median repetition gap; (D) SDD = standard deviation of digits; (E) TPI = turning point index; (F) RUNS; (G) AC = combined adjacency; and (H) CST = total count score. The error bars represent standard deviations. For each measure, data from a pseudorandom series are also shown for visual comparison (dotted lines).
Fig. 3. Effect of tDCS on random number generation performance. (A) REP = Repetition; (B) POKER; (C) GAP = Median repetition gap; (D) SDD = standard deviation of digits; (E) TPI = turning point index; (F) RUNS; (G) AC = combined adjacency; and (H) CST = total count score. PRE = first two runs within each session; SHAM = runs with sham stimulation; CAT = runs with cathodal stimulation; AN = runs with anodal stimulation.
55
F. Capone et al. / Neurobiology of Learning and Memory 114 (2014) 51–57 Table 1 Multilevel model: off-line and on-line practice effect. D.V.
Model
AIC
LR test
Off-line effect Linear
Repetition REP
POKER
Cycling GAP
SDD
Seriation TPI RUNS
AC
CST
On-line effect Quadr.
Linear
Quadr.
0.13
0.64** 0.64** 0.29**
0.22
1.08* 1.08* 1.37*
Full Sel Cts
1345.72 1333.1 1329.79
16.77 13.39** 14.71**
0.42
Full Sel Cts
1910.97 1900.62 1897.63
11.41 5.77 8.76*
0.49
Full Sel Cts
653.92 639.85 642.4
8.75 6.82+ 0.28
0.04
Full Sel Cts
810 801.62 799.01
Full Cts
2277.72 2268.95
88.25** 83.01**
6.66** 8.16**
0.75
5.84** 9.09**
Full Sel Cts
288.84 286.92 277.1
41.28** 31.21** 37.02**
0.16** 0.16** 0.24**
0.04 0.04
0.15** 0.15** 0.18**
2024.43 2023.6 2016.19
**
97.14 85.97** 91.38**
**
1.25 1.25
**
2934.47 2928.28 2923.3
**
Full Sel Cts Full Sel Cts
10.76 1.14 5.76
65.38 59.57** 62.55**
0.3
0.76
Interaction
0.17 0.17
0.01
0.36 0.36
1.4+ 0.44+
0.01
0.02 0.02
0.02 0.02
0.04
0.02 0.02
0.02
0.04
0.32*
0.02 0.08 0.08 0.2
0.2+
5.45 5.45** 5.91** **
14.9 14.9** 16.88**
2.6 2.6
0.1*
4.02 4.02** 5.82** **
14.79 14.79** 24.66**
1.49+ 0.74+
4.37** 1.38** 0.17**
0.01 0.01
0.05** +
0.95 0.95+ 0.48+ *
5.2 5.2* 2.6*
2.6** 0.82** 8.02+ 2.53+
Model specifications: Full includes session, run, and their interaction as categorical variables; Sel reports the model with the lowest Akaike Information Criterion; Cts reports the model with the lowest Akaike Information Criterion when session and run are included as continuous variables. All specification include random intercepts for individuals, for sessions within individuals, and for runs within sessions. ** p < 0.01. * p < 0.05. + p < 0.1.
3. Results
4. Discussion
Repetition measures showed significant on-line changes induced by practice (Table 1: Linear on-line effect column). REP mildly decreased within each session (Fig. 2A) while POKER increased with a more pronounced effect in the first session (Fig. 2B). Cycling measures did show neither an off-line nor an on-line effect (Fig. 2C and D). Both the off-line and the on-line effects of practice on the four seriation measures were significant. Indeed, RUNS, AC, and CST decreased within each session, as well as between sessions (Fig. 2F–H), while TPI increased with a similar pattern (Fig. 2E). Moreover, non-linear effects were also present both within (Table 1: Quadratic on-line effect column), and between sessions (Table 1: Interaction column). It means that the amount of the effects of practice was higher earlier, while decreased with time. tDCS did not produce any additional significant effect (Fig. 3), beyond the above-mentioned on-line practice effects emerging from the pre–post contrast (comparing the first two runs of each session vs. the last two runs) in repetition and seriation measures (Table 2: Pre vs. Post column). Among the cycling measures, GAP shows a statistically significant increase after stimulation, either cathodal or anodal (Table 2: Sham vs. Stim column), but the overall model is not significant (Table 2: LR Test column). When restricting the analysis only to Session 1, we can observe a trend (p < 0.1) toward an increase in CST after anodal stimulation, compared to cathodal (Table 2: Cat vs. An column).
4.1. Practice-induced changes in measures of randomness The present study provides strong evidence that practice induces changes in specific features of human RNG performance. RNG behavior depends on separable cognitive components reflected by different measures of randomness. These measures have been grouped by previous works (Ginsburg & Karpiuk, 1994; Peters et al., 2007) in three main factors: seriation, related to the inhibition of prepotent responses; cycling, associated to the successful monitoring of previous output, and repetition, related to output inhibition. We found that RNG practice produces a significant change in the measures of seriation (decrease in RUNS, AC, and CST and in increase in TPI) and repetition (decrease in REP and increase in POKER), while the parameters of cycling remain stable across time. Thus, our data show that training in the RNG task specifically improves the ability to suppress habitual responses. Such finding is in line with previous demonstrations of traininginduced plasticity in other tasks involving inhibition of prepotent responses. In 1994, Dulaney and Rogers, examining the effect of practice on a multiple-item version of the Stroop color-word task, found that interference declined with practice (Dulaney & Rogers, 1994). Thus, they argued that a word reading suppression response contributes to the observed practice-related improvement in Stroop task performance. Interestingly, the pattern of practicerelated improvement in interference was similar to the one observed in the present study for seriation and repetition measures,
56
F. Capone et al. / Neurobiology of Learning and Memory 114 (2014) 51–57
4.2. Selective effect of practice on measures of randomness
Table 2 Multilevel model: the effects of tDCS. D.V.
Model
LR test
Pre vs. Post
Sham vs. Stim
18.51+ 4.13
0.15** 0.17+
0.11 0.04
0.16 0.33
Full sample Session 1
14.95 3.25
0.26+ 0.46
0.25 0.8
0.39 0.3
Full sample Session 1
14.93 1.29
0.01 0.01
Full sample Session 1
13.64 3.95
0.01 0.07
0.06 0.22
0.02 0.09
64.25** 16.71**
1.42** 2.16**
0.11 0.07
0.1 1.52
Repetition REP Full sample Session 1 POKER
Cycling GAP SDD
Seriation TPI Full sample Session 1
0.09* 0.06
Cat vs. An
0.02 0.1
RUNS
Full sample Session 1
37.86** 9.25*
0.04** 0.06**
0.03 0.03
0.04 0.06
AC
Full sample Session 1
73.28** 17.18**
0.96** 1.42**
0.34 0.7
0.27 0.69
CST
Full sample Session 1
47.84** 12.58**
3.19** 4.5**
0.54 0.47
2.02 10.86+
Model specifications: Full sample includes session and stimulation, with the following planned contrasts: Pre vs. Post (the two runs before stimulation vs. the two runs after stimulation); Sham vs. Stim (the runs with sham stimulation vs. the runs with active stimulation); Cat vs. An (the runs with cathodal stimulation vs. the runs with anodal stimulation). The random structure includes random coefficients for individuals, for sessions within individuals, and for stimulation type within sessions. Session 1 includes only data from the first session. The random structure includes random coefficients for individuals, and for stimulation type within individuals. ** p < 0.01. * p < 0.05. + p < 0.1.
with the greatest effect occurring early in practice (Table 1: Quadratic on-line effect column and Interaction column). While the impact of practice has been widely demonstrated in Stroop task performance (Edwards, Brice, Craig, & Penri-Jones, 1996), only few studies have investigated such impact on RNG task. A practice-related effect was showed by Evans and Graham (1980). They examined 17 healthy volunteers who performed the RNG, in a dual-task context and found a progressive improvement in task performance. However, this finding was not confirmed by Jahanshahi and coworkers, who investigated the effect of increasing the rate of the paced generation and the effect of practice on several measures of randomness (Jahanshahi et al., 2006). They found that faster rates made numbering less random, while the repetition of the task did not produce any effect, concluding that RNG performance is a highly controlled and demanding process and so, impervious to practice. However, some methodological aspects should be considered to compare the results of the present study with those findings. In Jahanshahi et al. (2006), practice effect was evaluated by asking 13 healthy subjects to generate numbers randomly at the rate of 0.5 Hz (i.e., one response every 2 s) on 10 consecutive trials, separated by short breaks of a few minutes. Moreover, a significant part of this sample (9 out of 13 subjects) was not naive to the RNG task, being they also enrolled in the initial experiment where they had to perform the repetition of RNG at six different rates. Hence, in accord with our results showing stronger amelioration in the initial phase of practice, the presence of a previous training could partially explain the lack of practice-related improvement observed in the subsequent experiment. In addition, the influence of rate of the pacing stimulus during RNG task (0.5 Hz of Jahanshahi et al. compared to 1.2 Hz of our study) may also account for the different results.
Previous studies have demonstrated that, compared to computer-generated pseudorandom series, healthy volunteers show particular biases in RNG performance (Ginsburg & Karpiuk, 1994; Jahanshahi et al., 1998; Spatt & Goldenberg, 1993). In particular, human beings tend to avoid repeating the same number (repetition avoidance) and to produce ascending or descending ordered sequences (counting bias). Our data confirmed these findings (dotted lines in Fig. 2) and showed that the effects of practice on these peculiarities of human RNG follow opposite trends. Indeed, we observed an increase in repetition avoidance and a reduction of counting bias. Moreover, the effect of practice manifested different time-courses for repetition and seriation measures; it was transient (within session) on the formers, while persistent (both within and between sessions) on the latters. These findings could appear contrasting, because more successful RNG performance should result from greater suppression of both these biases. One possible hypothesis sees counting bias as a more clear violation of randomness, thus more easily detected and controlled than repetition avoidance. Alternatively, we can speculate that the reason why repetition avoidance and counting bias are differentially influenced by practice lays in the fact that such biases refer to different and partially separable executive functions. This is supported by the results of the studies that analyzed changes of RNG performance under different types of manipulation. For instance, Terhune and Brugger demonstrated that post-hypnotic amnesia suggestion influences repetition avoidance with no effect on counting bias (Terhune & Brugger, 2011), while Knoch and colleagues found that repetitive TMS influences selectively counting in a frequency-dependent manner (Knoch et al., 2005). Moreover, Anzak and coworkers showed that repetition is specifically correlated to the activity of the subthalamic nucleus (Anzak et al., 2012). Thus, the results of the present study, taken together with previous evidence of a dissociated behavior of repetition and seriation measures, support the idea that repetition avoidance may be a function of the capacity of keeping track of recent responses and comparing them to a conception of randomness, while counting bias may express the ability to suppress stereotyped sequences (Miyake et al., 2000). 4.3. tDCS-induced changes in measures of randomness To date, this is the first study that evaluated the effect of tDCS on RNG performance. The overall analysis revealed that neither cathodal nor anodal stimulation influences the training-induced plasticity. Indeed, the application of tDCS over left DLPFC does not produce any additional significant effect, beyond the abovementioned practice effects in repetition and seriation measures. Moreover, these practice-induced changes are not significantly influenced by tDCS. However, it may be speculated that, the effect of tDCS, although present, may be not enough strong to arise on top of the wide effect exerted by practice. In this case our analysis may not be able to evidence such effect. In order to investigate this possibility, we performed a further sub-analysis by restricting the sample to the first session only. This analysis revealed a trend toward an increase in CST after anodal stimulation compared with cathodal stimulation, suggesting that tDCS could produce a specific but transient effect on RNG performance by increasing counting bias. Such finding is consistent with previous studies showing that a single session of repetitive TMS over the left DLPFC can influence human RNG behavior by specifically affecting counting bias (Jahanshahi et al., 1998; Knoch et al., 2005). In particular, Knoch et al. (2005) found a frequency-dependent effect of repetitive TMS: counting bias was reduced after 1 Hz stimulation and increased after 10 Hz stimulation. Moreover, the selective effect
F. Capone et al. / Neurobiology of Learning and Memory 114 (2014) 51–57
of tDCS on CST is in good agreement with previous works showing that tDCS can influence the ability to inhibit automatic responses (Hsu et al., 2011; Jacobson, Javitt, & Lavidor, 2011). However, the tDCS-induced changes showed in the present paper should be regarded very cautiously because they are evident only in the sub-analysis limited to the first session. Moreover, few additional methodological aspects should be considered in the interpretation of these results. First of all, previous works have demonstrated that both intensity and duration of exposure are critical factors in determining the efficacy of tDCS (Nitsche et al., 2008). For instance, Ditye et al. (2012) found a significant effect of tDCS by using anodal stimulation (1.5 mA) over the right inferior frontal gyrus for 15 min daily for five consecutive days. Accordingly, we cannot rule out that different stimulation paradigms could have been more effective. Second, the study design should be also considered. We employed a within-subjects design in which the stimulation (anodal, cathodal, sham) varied at each session. Such design is very useful to evaluate the effect of tDCS within-session (on-line effect), but it could fail to find an off-line effect (between training sessions). Indeed, previous studies have found an off-line effect both on cognitive (Martin et al., 2013) and on motor training (Reis et al., 2009) when the same kind of tDCS was repeated over multiple sessions. Further studies, specifically designed to address these issues, will clarify the effects of tDCS on RNG performance and the neural correlates behind them.
5. Conclusions Our study demonstrates that practice consistently influences specific features of RNG performance and suggests that tDCS could transiently modify RNG behavior. These findings raise a number of important issues that will be relevant for forthcoming research: among them, the transferability of an improvement in RNG performance to other tasks that have not been trained and the usefulness of non-invasive brain stimulation techniques for enhancing the training-induced plasticity. Acknowledgments This research received no specific grant from any funding agency, commercial or not-for-profit sectors. The authors declare no potential conflicts of interest relating to the subject of this report. References Anzak, A., Gaynor, L., Beigi, M., Foltynie, T., Limousin, P., Zrinzo, L., et al. (2012). Subthalamic nucleus gamma oscillations mediate a switch from automatic to controlled processing: A study of random number generation in Parkinson’s disease. NeuroImage. Bherer, L., Kramer, A. F., Peterson, M. S., Colcombe, S., Erickson, K., & Becic, E. (2005). Training effects on dual-task performance. Are there age-related differences in plasticity of attentional control? Psychology and Aging, 20, 695. Dahlin, E., Nyberg, L., Bäckman, L., & Neely, A. S. (2008). Plasticity of executive functioning in young and older adults: Immediate training gains, transfer, and long-term maintenance. Psychology and Aging, 23, 720. Ditye, T., Jacobson, L., Walsh, V., & Lavidor, M. (2012). Modulating behavioral inhibition by tDCS combined with cognitive training. Experimental Brain Research, 219, 363–368.
57
Dulaney, C. L., & Rogers, W. A. (1994). Mechanisms underlying reduction in Stroop interference with practice for young and old adults. Journal of Experimental Psychology: Learning, Memory, and Cognition, 20, 470. Edwards, S., Brice, C., Craig, C., & Penri-Jones, R. (1996). Effects of caffeine, practice, and mode of presentation on Stroop task performance. Pharmacology Biochemistry and Behavior, 54, 309–315. Evans, F. J., & Graham, C. (1980). Subjective random number generation and attention deployment during acquisition and overlearning of a motor skill. Bulletin of the Psychonomic Society. Fecteau, S., Knoch, D., Fregni, F., Sultani, N., Boggio, P., & Pascual-Leone, A. (2007). Diminishing risk-taking behavior by modulating activity in the prefrontal cortex: A direct current stimulation study. The Journal of Neuroscience, 27, 12500–12505. Gandiga, P. C., Hummel, F. C., & Cohen, L. G. (2006). Transcranial DC stimulation (tDCS): A tool for double-blind sham-controlled clinical studies in brain stimulation. Clinical Neurophysiology, 117, 845–850. Ginsburg, N., & Karpiuk, P. (1994). Random generation: Analysis of the responses. Perceptual and Motor Skills, 79, 1059–1067. Goldstein, M. (2011). Multilevel statistical models (4th ed.). Chichester, UK: Wiley. Hsu, T.-Y., Tseng, L.-Y., Yu, J.-X., Kuo, W.-J., Hung, D. L., Tzeng, O. J., et al. (2011). Modulating inhibitory control with direct current stimulation of the superior medial frontal cortex. Neuroimage, 56, 2249–2257. Jacobson, L., Javitt, D. C., & Lavidor, M. (2011). Activation of inhibition: Diminishing impulsive behavior by direct current stimulation over the inferior frontal gyrus. Journal of Cognitive Neuroscience, 23, 3380–3387. Jahanshahi, M., Dirnberger, G., Fuller, R., & Frith, C. (2000). The role of the dorsolateral prefrontal cortex in random number generation: A study with positron emission tomography. Neuroimage, 12, 713–725. Jahanshahi, M., Profice, P., Brown, R. G., Ridding, M. C., Dirnberger, G., & Rothwell, J. C. (1998). The effects of transcranial magnetic stimulation over the dorsolateral prefrontal cortex on suppression of habitual counting during random number generation. Brain, 121, 1533–1544. Jahanshahi, M., Saleem, T., Ho, A. K., Dirnberger, G., & Fuller, R. (2006). Random number generation as an index of controlled processing. Neuropsychology, 20, 391. Karbach, J., & Schubert, T. (2013). Training-induced cognitive and neural plasticity. Frontiers in Human Neuroscience, 7. Knoch, D., Brugger, P., & Regard, M. (2005). Suppressing versus releasing a habit: Frequency-dependent effects of prefrontal transcranial magnetic stimulation. Cerebral Cortex, 15, 885–887. Kramer, A. F., Hahn, S., & Gopher, D. (1999). Task coordination and aging: Explorations of executive control processes in the task switching paradigm. Acta Psychologica, 101, 339–378. Levasseur-Moreau, J., & Fecteau, S. (2012). Translational application of neuromodulation of decision-making. Brain Stimulation, 5, 77–83. Martin, D. M., Liu, R., Alonzo, A., Green, M., Player, M. J., Sachdev, P., et al. (2013). Can transcranial direct current stimulation enhance outcomes from cognitive training? A randomized controlled trial in healthy participants. International Journal of Neuropsychopharmacology, 16(9), 1927–1936. Miyake, A., Friedman, N. P., Emerson, M. J., Witzki, A. H., Howerter, A., & Wager, T. D. (2000). The unity and diversity of executive functions and their contributions to complex ‘‘frontal lobe’’ tasks: A latent variable analysis. Cognitive Psychology, 41, 49–100. Nitsche, M. A., Cohen, L. G., Wassermann, E. M., Priori, A., Lang, N., Antal, A., et al. (2008). Transcranial direct current stimulation: State of the art 2008. Brain Stimulation, 1, 206–223. Peters, M., Giesbrecht, T., Jelicic, M., Merckelbach, H., American Psychiatric Association, H., Anastasi, A., et al. (2007). The random number generation task: Psychometric properties and normative data of an executive function task in a mixed sample. Journal of the International Neuropsychological Society, 13, 626–634. Reis, J., Schambra, H. M., Cohen, L. G., Buch, E. R., Fritsch, B., Zarahn, E., et al. (2009). Noninvasive cortical stimulation enhances motor skill acquisition over multiple days through an effect on consolidation. Proceedings of the National Academy of Sciences of the United States of America, 106(5), 1590–1595. Robertson, E. M., Pascual-Leone, A., & Miall, R. C. (2004). Current concepts in procedural consolidation. Nature Reviews Neuroscience, 5, 576–582. Spatt, J., & Goldenberg, G. (1993). Components of random generation by normal subjects and patients with dysexecutive syndrome. Brain and Cognition, 23, 231–242. Terhune, D. B., & Brugger, P. (2011). Doing better by getting worse: Posthypnotic amnesia improves random number generation. PLoS ONE, 6, e29206. Towse, J. N., & Neil, D. (1998). Analyzing human random generation behavior: A review of methods used and a computer program for describing performance. Behavior Research Methods, Instruments, & Computers, 30, 583–591. Wilkinson, A. J., & Yang, L. (2012). Plasticity of inhibition in older adults: Retest practice and transfer effects. Psychology and Aging, 27, 606.
Contents lists available at ScienceDirect
Neurobiology of Learning and Memory journal homepage: www.elsevier.com/locate/ynlme
The effect of practice on random number generation task: A transcranial direct current stimulation study Fioravante Capone a,b,⇑, Gianluca Capone c, Federico Ranieri a,b, Giovanni Di Pino a,b, Gianluca Oricchio d, Vincenzo Di Lazzaro a,b a
Institute of Neurology, Campus Bio-Medico University, Via Alvaro del Portillo 200, 00128 Rome, Italy Fondazione Alberto Sordi – Research Institute for Ageing, Rome, Italy Department of Economic Geography, Urban and Regional Research Centre Utrecht (URU), Utrecht University, The Netherlands d Campus Bio-Medico University, Via Alvaro del Portillo 200, 00128 Rome, Italy b c
a r t i c l e
i n f o
Article history: Received 8 January 2014 Revised 22 April 2014 Accepted 27 April 2014 Available online 6 May 2014 Keywords: Random number generation Practice tDCS Cognitive plasticity Executive functions Brain stimulation
a b s t r a c t Random number generation (RNG) is a procedurally-simple task related to specific executive functions, such as updating and monitoring of information and inhibition of automatic responses. The effect of practice on executive functions has been widely investigated, however little is known on the impact of practice on RNG. Transcranial direct current stimulation (tDCS) allows to modulate, noninvasively, brain activity and to enhance the effects of training on executive functions. Hence, this study aims to investigate the effect of practice on RNG and to explore the possibility to influence it by tDCS applied over dorsolateral prefrontal cortex. Twenty-six healthy volunteers have been evaluated within single session and between different sessions of RNG using several measures of randomness, which are informative of separable cognitive components servicing random behavior. We found that repetition measures significantly change within single session, seriation measures significantly change both within and between sessions, while cycling measures are not affected by practice. tDCS does not produce any additional effect, however a sub-analysis limited to the first session revealed an increasing trend in seriation measure after anodal compared to cathodal stimulation. Our findings support the hypothesis that practice selectively and consistently influences specific cognitive components related to random behavior, while tDCS transiently affects RNG performance. Ó 2014 Elsevier Inc. All rights reserved.
1. Introduction Random number generation (RNG) is a procedurally-simple task widely exploited to investigate cognitive function. Subjects are instructed to say aloud, in a random order, the numbers from 1 to 9, in synchrony with a pacing stimulus. Despite its apparent simplicity, RNG involves several mental processes. Indeed, it requires adopting the correct strategy, on the basis of the selection of appropriate responses and suppression of those that appear to violate the instructions and the subject’s concept of randomness, monitoring the output and eventually modifying the strategy of Abbreviations: RNG, random number generation; tDCS, transcranial direct current stimulation; DLPFC, dorsolateral prefrontal cortex; TMS, transcranial magnetic stimulation; REP, repetition; POKER, poker; GAP, median repetition gap; SDD, standard deviation of digits; TPI, turning point index; RUNS, runs; AC, combined adjacency; CST, total count score. ⇑ Corresponding author at: Institute of Neurology Campus Bio-Medico University, Via Alvaro del Portillo 200, 00128 Rome, Italy. Fax: +39 06 225411936. E-mail address: [email protected] (F. Capone). http://dx.doi.org/10.1016/j.nlm.2014.04.013 1074-7427/Ó 2014 Elsevier Inc. All rights reserved.
production (Jahanshahi et al., 1998). In particular, RNG is strongly related to specific executive functions, such as updating and monitoring of information, and inhibition of prepotent responses of counting or cycling through the set of numbers (Miyake et al., 2000). Executive functions, such as dual-task performance (Bherer et al., 2005), information updating (Dahlin, Nyberg, Bäckman, & Neely, 2008), task switching (Kramer, Hahn, & Gopher, 1999) and inhibition of prepotent responses (Wilkinson & Yang, 2012) can be trained and this training mostly reflects the establishment of cognitive plasticity (Karbach & Schubert, 2013). Although the ascertained role of RNG in assessing executive functions, little research faced RNG-induced plasticity. To our knowledge, only few studies have examined the effect of practice in RNG task, providing contrasting results (Evans & Graham, 1980; Jahanshahi, Saleem, Ho, Dirnberger, & Fuller, 2006; Peters et al., 2007). Carrying out the RNG task requires the activation of different brain regions, like the left dorsolateral prefrontal cortex (DLPFC), the anterior cingulate cortex, the superior parietal cortex, the right
52
F. Capone et al. / Neurobiology of Learning and Memory 114 (2014) 51–57
inferior frontal cortex, and the cerebellar hemispheres (Jahanshahi, Dirnberger, Fuller, & Frith, 2000). Functional imaging (Jahanshahi et al., 2000) and transcranial magnetic stimulation (TMS) (Jahanshahi et al., 1998) studies indicate that the left DLPFC plays a critical role in this network by exerting an inhibitory influence over temporal–parietal cortex, in order to suppress counting. Interestingly, TMS-induced modulation of the left DLPFC affects the RNG performances in a stimulation frequency-dependent fashion: the tendency to produce ascending or descending ordered sequences (counting bias) is significantly reduced after inhibitory (1 Hz) repetitive TMS, while is significantly increased after excitatory (10 Hz) stimulation (Knoch, Brugger, & Regard, 2005). Transcranial direct current stimulation (tDCS) is a safe and reliable technique that allows to modulate, non-invasively, brain functions and their plasticity. In this technique, a weak current is applied constantly over time to increase (anodal stimulation) or decrease (cathodal stimulation) the excitability of the neuronal populations underlying the active electrode (Nitsche et al., 2008). It has been demonstrated that tDCS can produce transient behavioral changes and influence cognitive functions (LevasseurMoreau & Fecteau, 2012). When performed in parallel with training protocols, it showed to be able to enhance the protocol effects. In particular, tDCS over the frontal regions, combined with cognitive training, seems to be able to modulate executive functions and improve the ability to inhibit responses, as Ditye et al. demonstrated by means of the Stop Signal Task (Ditye, Jacobson, Walsh, & Lavidor, 2012). In the present study, we evaluated the effect of practice on RNG performance to induce training plasticity. To this aim, we employed several different measures of randomness, because each of them is considered to reflect the efficiency of separable cognitive components servicing RNG behavior (Towse & Neil, 1998). Moreover, we tested whether tDCS could influence the training-induced plasticity. In the light of the specific role played by the left DLPFC, we applied either anodal or cathodal stimulation over this area during the RNG task. Previous evidence indicates that performance improvements can occur within-session (on-line effects), as well as between training sessions (off-line effects), i.e., the performance at the beginning of a given session could be different from the performance at the end of the previous session (Robertson, PascualLeone, & Miall, 2004). To this purpose, we evaluated the effect of practice and tDCS within single sessions of RNG and between sessions of RNG separated by at least 48 h. 2. Material and methods 2.1. Participants 26 Healthy subjects (9 male and 17 female) have been enrolled in the study. They had no previous history of neurological or psychiatric disorders and were not taking any medication at the time of the assessment. All the subjects were naive to the RNG task and were not explicitly informed of the experimental variables tested. The mean age was 28 years (range 23–49, SD 5.05). Participants signed a written informed consent prior to the participation in this study, which was approved by the Local Ethics Committee. The research was completed in accordance with the Helsinki Declaration. 2.2. Design A within-subject repeated measures design was used. There were three conditions: RNG coupled with sham tDCS; RNG with left anodal/right cathodal tDCS; RNG with right anodal/left cathodal tDCS. All participants took part in all conditions. The order
the three conditions was pseudo-randomized across participants. The interval between the sessions was at least 48 h. Participants were blind to conditions. 2.3. Random number generation Each session consisted of four consecutive runs. On each run, participants were asked to say aloud a sequence of 100 numbers, each number ranging from 1 to 9, as much random as possible at a metronomic rate of 1.2 Hz. The first two runs were performed without stimulation and were separated by 1 min; once the second run ended, tDCS (either sham or real) was started to last until the end of last run. After 5 min, two more RNG runs, separated by a 1 min interval, were performed during the stimulation. The concept of randomness was explained to the participant using the analogy of picking pieces of paper numbered 1–9 out of a hat. All subjects were naive to the RNG task. 2.4. Measures of randomness A multiplicity of measures to quantify deviation from randomness has been employed in the literature. Towse and Neil have identified four factors by means of a principal component analysis of 16 measures (Towse & Neil, 1998). Ginsburg and Karpiuk (1994) and Peters et al. (2007) have summarized, by means of a factor analysis, several measures in three main factors: a seriation factor, associated to the inhibition of prepotent responses, a cycling factor, associated to the successful monitoring of previous output, and a repetition factor, associated to output inhibition. In our analysis, we employed two measures for each factor identified from the work by Peters and colleagues and added the two measures by Towse and Neil and Jahanshahi and colleagues that seem to cover better the inhibition of prepotent responses (Jahanshahi et al., 2006; Peters et al., 2007; Towse & Neil, 1998). 2.4.1. Repetition measures Such measures refer to the repetition of the same digit within a certain interval. Repetition (REP) counts the number of identical, adjacent digits in the sequence. Poker (POKER) is the number of repetitions that occur between two and five successive digits. 2.4.2. Cycling measures Such measures refer to the attempt to use every possible alternative before repeating any digit. The Median Repetition Gap (GAP) is the median of the intervals between two identical digits. The Standard Deviation of Digits (SDD) is the standard deviation of the frequencies distribution of the nine possible digits. 2.4.3. Seriation measures Such measures refer to the tendency to use the natural order of numbers in the sequence. The Turning Point Index (TPI) is the ratio between observed and expected turning points (in percentage form). A turning point is any point in the sequence where there is a shift from an ascending to a descending subsequence (or vice versa). Runs (RUNS) measure the variance of the lengths of ascending subsequences. The Combined Adjacency (AC) is the ratio (in percentage form) between the number of adjacent pairs of digits (also called Series) and the total number of pairs of digits in the sequence. The Total Count Score (CST) is given by the sum of squared lengths of ascending and descending subsequences in steps of one and two. 2.5. Transcranial direct current stimulation tDCS was applied by a battery-driven constant-current stimulator (Eldith-NeuroConn GmbH, Ilmenau, Germany) via
F. Capone et al. / Neurobiology of Learning and Memory 114 (2014) 51–57
conductive-rubber electrodes, placed in two saline-soaked sponges (5 7 cm) and current output was monitored by a built-in ampere meter. tDCS was performed according to the paradigm described in Fecteau et al. (2007). For left anodal/right cathodal stimulation, the anode was placed over the left F3 (international EEG 10/20 system) and the cathode electrode over the right F4. For right anodal/left cathodal stimulation, the polarity was reversed: the anode was placed over F4 and the cathode electrode over F3. All along the active stimulation, participants received a constant current of 2 mA. Subjects stimulated with tDCS are able to consciously perceive only the beginning of the stimulation, through a sense of itching that localizes underneath the electrode lasting few seconds. The sham stimulation was carried out placing the electrodes in the same position as for active stimulation, but in order to keep subjects blind toward the applied stimulation, the stimulator was kept activated only for 30 s. This procedure made the participants feel the initial itching sensation associated with tDCS, but without receiving any active current for the rest of the stimulation period (Gandiga, Hummel, & Cohen, 2006) and did not allowed participants to distinguish between real and sham tDCS conditions (see Fig. 1). 2.6. Data analysis We analyzed the on-line (within-session) and off-line (between sessions) effects of practice and of tDCS on the RNG measures of randomness. A graphical representation of the effect of practice and tDCS is provided in Figs. 2 and 3, respectively. Fig. 2 shows the average and standard deviation values of the RNG measures across sessions and runs; it also depicts the expected value from a computer-generated pseudorandom series. Fig. 3 shows the average values of the RNG indicators across stimulation types. In our statistical analysis, we used two-way repeated measures ANOVA, separately for each RNG measure. In many cases, Mauchly’s test indicated that the assumption of sphericity was violated, invalidating regular F statistics. Therefore, we employed a multilevel model approach (Goldstein, 2011).
53
Our model can be written in the following compact form:
yijk ¼ b0 þ
X bN xNijk þ v k þ ujk þ eijk N
where y represents our dependent variable (a measure of randomness) and the bN are fixed effect coefficients of the corresponding variables xN. The random structure of the model includes random intercepts for individuals (vk), for sessions within individuals (ujk), and for runs or stimulation type within sessions (eijk). The analysis has been performed using the statistical software R (version 2.15.2 http://www.r-project.org/). In order to analyze practice effects, we include as fixed effect variables session (off-line effects), runs (on-line effects) and their interaction. In Table 1, for each measure of randomness, we report the results of the full specification (Full), the best specification selected according to the Akaike Information Criterion (Sel) if different from the full specification, and the best specification considering session and run as continuous rather than categorical variables (Cts) that allows more flexibility in the choice of parameters. We also report a likelihood ratio test that compares each model to a baseline model including only the random structure. The last five columns report the values of the contrasts (Full, Sel) and parameters (Cts) coefficients. In all cases, we use maximum-likelihood estimation. In order to analyze tDCS effects, we included session and stimulation type as fixed effects, and we estimated random intercepts between different stimulation types (rather than runs). In Table 2 we report contrasts coefficients for stimulation type: we contrast the two runs before stimulation vs. the two runs after stimulation (Pre vs. Post), the runs with sham stimulation vs. the runs with active stimulation (Sham vs. Stim), and the runs with cathodal stimulation vs. the runs with anodal stimulation (Cat vs. An). As in the previous case, we report a likelihood ratio test vs. the baseline model. Since practice effects could have washed out the tDCS effect, we also performed our analysis by restricting the sample to the first session only.
Fig. 1. Study design. A within-subject repeated measures design was used. There were three conditions: RNG with sham tDCS; RNG with left anodal/right cathodal tDCS; RNG with right anodal/left cathodal tDCS. The order these conditions was pseudo-randomized across participants. The interval between the sessions was at least 48 h. Every session consisted of four consecutive runs. On each run, participants were asked to say the numbers 1–9 in a sequence as random as possible (100 trials at a metronomic rate of 1.2 Hz). The first two runs were performed without stimulation and were separated by 1 min; then tDCS (sham or real) was started and after 5 min, two more RNG runs separated by 1 min, were performed during stimulation.
54
F. Capone et al. / Neurobiology of Learning and Memory 114 (2014) 51–57
Fig. 2. Effect of practice on random number generation performance within-session and between sessions. (A) REP = Repetition; (B) POKER; (B) GAP = Median repetition gap; (D) SDD = standard deviation of digits; (E) TPI = turning point index; (F) RUNS; (G) AC = combined adjacency; and (H) CST = total count score. The error bars represent standard deviations. For each measure, data from a pseudorandom series are also shown for visual comparison (dotted lines).
Fig. 3. Effect of tDCS on random number generation performance. (A) REP = Repetition; (B) POKER; (C) GAP = Median repetition gap; (D) SDD = standard deviation of digits; (E) TPI = turning point index; (F) RUNS; (G) AC = combined adjacency; and (H) CST = total count score. PRE = first two runs within each session; SHAM = runs with sham stimulation; CAT = runs with cathodal stimulation; AN = runs with anodal stimulation.
55
F. Capone et al. / Neurobiology of Learning and Memory 114 (2014) 51–57 Table 1 Multilevel model: off-line and on-line practice effect. D.V.
Model
AIC
LR test
Off-line effect Linear
Repetition REP
POKER
Cycling GAP
SDD
Seriation TPI RUNS
AC
CST
On-line effect Quadr.
Linear
Quadr.
0.13
0.64** 0.64** 0.29**
0.22
1.08* 1.08* 1.37*
Full Sel Cts
1345.72 1333.1 1329.79
16.77 13.39** 14.71**
0.42
Full Sel Cts
1910.97 1900.62 1897.63
11.41 5.77 8.76*
0.49
Full Sel Cts
653.92 639.85 642.4
8.75 6.82+ 0.28
0.04
Full Sel Cts
810 801.62 799.01
Full Cts
2277.72 2268.95
88.25** 83.01**
6.66** 8.16**
0.75
5.84** 9.09**
Full Sel Cts
288.84 286.92 277.1
41.28** 31.21** 37.02**
0.16** 0.16** 0.24**
0.04 0.04
0.15** 0.15** 0.18**
2024.43 2023.6 2016.19
**
97.14 85.97** 91.38**
**
1.25 1.25
**
2934.47 2928.28 2923.3
**
Full Sel Cts Full Sel Cts
10.76 1.14 5.76
65.38 59.57** 62.55**
0.3
0.76
Interaction
0.17 0.17
0.01
0.36 0.36
1.4+ 0.44+
0.01
0.02 0.02
0.02 0.02
0.04
0.02 0.02
0.02
0.04
0.32*
0.02 0.08 0.08 0.2
0.2+
5.45 5.45** 5.91** **
14.9 14.9** 16.88**
2.6 2.6
0.1*
4.02 4.02** 5.82** **
14.79 14.79** 24.66**
1.49+ 0.74+
4.37** 1.38** 0.17**
0.01 0.01
0.05** +
0.95 0.95+ 0.48+ *
5.2 5.2* 2.6*
2.6** 0.82** 8.02+ 2.53+
Model specifications: Full includes session, run, and their interaction as categorical variables; Sel reports the model with the lowest Akaike Information Criterion; Cts reports the model with the lowest Akaike Information Criterion when session and run are included as continuous variables. All specification include random intercepts for individuals, for sessions within individuals, and for runs within sessions. ** p < 0.01. * p < 0.05. + p < 0.1.
3. Results
4. Discussion
Repetition measures showed significant on-line changes induced by practice (Table 1: Linear on-line effect column). REP mildly decreased within each session (Fig. 2A) while POKER increased with a more pronounced effect in the first session (Fig. 2B). Cycling measures did show neither an off-line nor an on-line effect (Fig. 2C and D). Both the off-line and the on-line effects of practice on the four seriation measures were significant. Indeed, RUNS, AC, and CST decreased within each session, as well as between sessions (Fig. 2F–H), while TPI increased with a similar pattern (Fig. 2E). Moreover, non-linear effects were also present both within (Table 1: Quadratic on-line effect column), and between sessions (Table 1: Interaction column). It means that the amount of the effects of practice was higher earlier, while decreased with time. tDCS did not produce any additional significant effect (Fig. 3), beyond the above-mentioned on-line practice effects emerging from the pre–post contrast (comparing the first two runs of each session vs. the last two runs) in repetition and seriation measures (Table 2: Pre vs. Post column). Among the cycling measures, GAP shows a statistically significant increase after stimulation, either cathodal or anodal (Table 2: Sham vs. Stim column), but the overall model is not significant (Table 2: LR Test column). When restricting the analysis only to Session 1, we can observe a trend (p < 0.1) toward an increase in CST after anodal stimulation, compared to cathodal (Table 2: Cat vs. An column).
4.1. Practice-induced changes in measures of randomness The present study provides strong evidence that practice induces changes in specific features of human RNG performance. RNG behavior depends on separable cognitive components reflected by different measures of randomness. These measures have been grouped by previous works (Ginsburg & Karpiuk, 1994; Peters et al., 2007) in three main factors: seriation, related to the inhibition of prepotent responses; cycling, associated to the successful monitoring of previous output, and repetition, related to output inhibition. We found that RNG practice produces a significant change in the measures of seriation (decrease in RUNS, AC, and CST and in increase in TPI) and repetition (decrease in REP and increase in POKER), while the parameters of cycling remain stable across time. Thus, our data show that training in the RNG task specifically improves the ability to suppress habitual responses. Such finding is in line with previous demonstrations of traininginduced plasticity in other tasks involving inhibition of prepotent responses. In 1994, Dulaney and Rogers, examining the effect of practice on a multiple-item version of the Stroop color-word task, found that interference declined with practice (Dulaney & Rogers, 1994). Thus, they argued that a word reading suppression response contributes to the observed practice-related improvement in Stroop task performance. Interestingly, the pattern of practicerelated improvement in interference was similar to the one observed in the present study for seriation and repetition measures,
56
F. Capone et al. / Neurobiology of Learning and Memory 114 (2014) 51–57
4.2. Selective effect of practice on measures of randomness
Table 2 Multilevel model: the effects of tDCS. D.V.
Model
LR test
Pre vs. Post
Sham vs. Stim
18.51+ 4.13
0.15** 0.17+
0.11 0.04
0.16 0.33
Full sample Session 1
14.95 3.25
0.26+ 0.46
0.25 0.8
0.39 0.3
Full sample Session 1
14.93 1.29
0.01 0.01
Full sample Session 1
13.64 3.95
0.01 0.07
0.06 0.22
0.02 0.09
64.25** 16.71**
1.42** 2.16**
0.11 0.07
0.1 1.52
Repetition REP Full sample Session 1 POKER
Cycling GAP SDD
Seriation TPI Full sample Session 1
0.09* 0.06
Cat vs. An
0.02 0.1
RUNS
Full sample Session 1
37.86** 9.25*
0.04** 0.06**
0.03 0.03
0.04 0.06
AC
Full sample Session 1
73.28** 17.18**
0.96** 1.42**
0.34 0.7
0.27 0.69
CST
Full sample Session 1
47.84** 12.58**
3.19** 4.5**
0.54 0.47
2.02 10.86+
Model specifications: Full sample includes session and stimulation, with the following planned contrasts: Pre vs. Post (the two runs before stimulation vs. the two runs after stimulation); Sham vs. Stim (the runs with sham stimulation vs. the runs with active stimulation); Cat vs. An (the runs with cathodal stimulation vs. the runs with anodal stimulation). The random structure includes random coefficients for individuals, for sessions within individuals, and for stimulation type within sessions. Session 1 includes only data from the first session. The random structure includes random coefficients for individuals, and for stimulation type within individuals. ** p < 0.01. * p < 0.05. + p < 0.1.
with the greatest effect occurring early in practice (Table 1: Quadratic on-line effect column and Interaction column). While the impact of practice has been widely demonstrated in Stroop task performance (Edwards, Brice, Craig, & Penri-Jones, 1996), only few studies have investigated such impact on RNG task. A practice-related effect was showed by Evans and Graham (1980). They examined 17 healthy volunteers who performed the RNG, in a dual-task context and found a progressive improvement in task performance. However, this finding was not confirmed by Jahanshahi and coworkers, who investigated the effect of increasing the rate of the paced generation and the effect of practice on several measures of randomness (Jahanshahi et al., 2006). They found that faster rates made numbering less random, while the repetition of the task did not produce any effect, concluding that RNG performance is a highly controlled and demanding process and so, impervious to practice. However, some methodological aspects should be considered to compare the results of the present study with those findings. In Jahanshahi et al. (2006), practice effect was evaluated by asking 13 healthy subjects to generate numbers randomly at the rate of 0.5 Hz (i.e., one response every 2 s) on 10 consecutive trials, separated by short breaks of a few minutes. Moreover, a significant part of this sample (9 out of 13 subjects) was not naive to the RNG task, being they also enrolled in the initial experiment where they had to perform the repetition of RNG at six different rates. Hence, in accord with our results showing stronger amelioration in the initial phase of practice, the presence of a previous training could partially explain the lack of practice-related improvement observed in the subsequent experiment. In addition, the influence of rate of the pacing stimulus during RNG task (0.5 Hz of Jahanshahi et al. compared to 1.2 Hz of our study) may also account for the different results.
Previous studies have demonstrated that, compared to computer-generated pseudorandom series, healthy volunteers show particular biases in RNG performance (Ginsburg & Karpiuk, 1994; Jahanshahi et al., 1998; Spatt & Goldenberg, 1993). In particular, human beings tend to avoid repeating the same number (repetition avoidance) and to produce ascending or descending ordered sequences (counting bias). Our data confirmed these findings (dotted lines in Fig. 2) and showed that the effects of practice on these peculiarities of human RNG follow opposite trends. Indeed, we observed an increase in repetition avoidance and a reduction of counting bias. Moreover, the effect of practice manifested different time-courses for repetition and seriation measures; it was transient (within session) on the formers, while persistent (both within and between sessions) on the latters. These findings could appear contrasting, because more successful RNG performance should result from greater suppression of both these biases. One possible hypothesis sees counting bias as a more clear violation of randomness, thus more easily detected and controlled than repetition avoidance. Alternatively, we can speculate that the reason why repetition avoidance and counting bias are differentially influenced by practice lays in the fact that such biases refer to different and partially separable executive functions. This is supported by the results of the studies that analyzed changes of RNG performance under different types of manipulation. For instance, Terhune and Brugger demonstrated that post-hypnotic amnesia suggestion influences repetition avoidance with no effect on counting bias (Terhune & Brugger, 2011), while Knoch and colleagues found that repetitive TMS influences selectively counting in a frequency-dependent manner (Knoch et al., 2005). Moreover, Anzak and coworkers showed that repetition is specifically correlated to the activity of the subthalamic nucleus (Anzak et al., 2012). Thus, the results of the present study, taken together with previous evidence of a dissociated behavior of repetition and seriation measures, support the idea that repetition avoidance may be a function of the capacity of keeping track of recent responses and comparing them to a conception of randomness, while counting bias may express the ability to suppress stereotyped sequences (Miyake et al., 2000). 4.3. tDCS-induced changes in measures of randomness To date, this is the first study that evaluated the effect of tDCS on RNG performance. The overall analysis revealed that neither cathodal nor anodal stimulation influences the training-induced plasticity. Indeed, the application of tDCS over left DLPFC does not produce any additional significant effect, beyond the abovementioned practice effects in repetition and seriation measures. Moreover, these practice-induced changes are not significantly influenced by tDCS. However, it may be speculated that, the effect of tDCS, although present, may be not enough strong to arise on top of the wide effect exerted by practice. In this case our analysis may not be able to evidence such effect. In order to investigate this possibility, we performed a further sub-analysis by restricting the sample to the first session only. This analysis revealed a trend toward an increase in CST after anodal stimulation compared with cathodal stimulation, suggesting that tDCS could produce a specific but transient effect on RNG performance by increasing counting bias. Such finding is consistent with previous studies showing that a single session of repetitive TMS over the left DLPFC can influence human RNG behavior by specifically affecting counting bias (Jahanshahi et al., 1998; Knoch et al., 2005). In particular, Knoch et al. (2005) found a frequency-dependent effect of repetitive TMS: counting bias was reduced after 1 Hz stimulation and increased after 10 Hz stimulation. Moreover, the selective effect
F. Capone et al. / Neurobiology of Learning and Memory 114 (2014) 51–57
of tDCS on CST is in good agreement with previous works showing that tDCS can influence the ability to inhibit automatic responses (Hsu et al., 2011; Jacobson, Javitt, & Lavidor, 2011). However, the tDCS-induced changes showed in the present paper should be regarded very cautiously because they are evident only in the sub-analysis limited to the first session. Moreover, few additional methodological aspects should be considered in the interpretation of these results. First of all, previous works have demonstrated that both intensity and duration of exposure are critical factors in determining the efficacy of tDCS (Nitsche et al., 2008). For instance, Ditye et al. (2012) found a significant effect of tDCS by using anodal stimulation (1.5 mA) over the right inferior frontal gyrus for 15 min daily for five consecutive days. Accordingly, we cannot rule out that different stimulation paradigms could have been more effective. Second, the study design should be also considered. We employed a within-subjects design in which the stimulation (anodal, cathodal, sham) varied at each session. Such design is very useful to evaluate the effect of tDCS within-session (on-line effect), but it could fail to find an off-line effect (between training sessions). Indeed, previous studies have found an off-line effect both on cognitive (Martin et al., 2013) and on motor training (Reis et al., 2009) when the same kind of tDCS was repeated over multiple sessions. Further studies, specifically designed to address these issues, will clarify the effects of tDCS on RNG performance and the neural correlates behind them.
5. Conclusions Our study demonstrates that practice consistently influences specific features of RNG performance and suggests that tDCS could transiently modify RNG behavior. These findings raise a number of important issues that will be relevant for forthcoming research: among them, the transferability of an improvement in RNG performance to other tasks that have not been trained and the usefulness of non-invasive brain stimulation techniques for enhancing the training-induced plasticity. Acknowledgments This research received no specific grant from any funding agency, commercial or not-for-profit sectors. The authors declare no potential conflicts of interest relating to the subject of this report. References Anzak, A., Gaynor, L., Beigi, M., Foltynie, T., Limousin, P., Zrinzo, L., et al. (2012). Subthalamic nucleus gamma oscillations mediate a switch from automatic to controlled processing: A study of random number generation in Parkinson’s disease. NeuroImage. Bherer, L., Kramer, A. F., Peterson, M. S., Colcombe, S., Erickson, K., & Becic, E. (2005). Training effects on dual-task performance. Are there age-related differences in plasticity of attentional control? Psychology and Aging, 20, 695. Dahlin, E., Nyberg, L., Bäckman, L., & Neely, A. S. (2008). Plasticity of executive functioning in young and older adults: Immediate training gains, transfer, and long-term maintenance. Psychology and Aging, 23, 720. Ditye, T., Jacobson, L., Walsh, V., & Lavidor, M. (2012). Modulating behavioral inhibition by tDCS combined with cognitive training. Experimental Brain Research, 219, 363–368.
57
Dulaney, C. L., & Rogers, W. A. (1994). Mechanisms underlying reduction in Stroop interference with practice for young and old adults. Journal of Experimental Psychology: Learning, Memory, and Cognition, 20, 470. Edwards, S., Brice, C., Craig, C., & Penri-Jones, R. (1996). Effects of caffeine, practice, and mode of presentation on Stroop task performance. Pharmacology Biochemistry and Behavior, 54, 309–315. Evans, F. J., & Graham, C. (1980). Subjective random number generation and attention deployment during acquisition and overlearning of a motor skill. Bulletin of the Psychonomic Society. Fecteau, S., Knoch, D., Fregni, F., Sultani, N., Boggio, P., & Pascual-Leone, A. (2007). Diminishing risk-taking behavior by modulating activity in the prefrontal cortex: A direct current stimulation study. The Journal of Neuroscience, 27, 12500–12505. Gandiga, P. C., Hummel, F. C., & Cohen, L. G. (2006). Transcranial DC stimulation (tDCS): A tool for double-blind sham-controlled clinical studies in brain stimulation. Clinical Neurophysiology, 117, 845–850. Ginsburg, N., & Karpiuk, P. (1994). Random generation: Analysis of the responses. Perceptual and Motor Skills, 79, 1059–1067. Goldstein, M. (2011). Multilevel statistical models (4th ed.). Chichester, UK: Wiley. Hsu, T.-Y., Tseng, L.-Y., Yu, J.-X., Kuo, W.-J., Hung, D. L., Tzeng, O. J., et al. (2011). Modulating inhibitory control with direct current stimulation of the superior medial frontal cortex. Neuroimage, 56, 2249–2257. Jacobson, L., Javitt, D. C., & Lavidor, M. (2011). Activation of inhibition: Diminishing impulsive behavior by direct current stimulation over the inferior frontal gyrus. Journal of Cognitive Neuroscience, 23, 3380–3387. Jahanshahi, M., Dirnberger, G., Fuller, R., & Frith, C. (2000). The role of the dorsolateral prefrontal cortex in random number generation: A study with positron emission tomography. Neuroimage, 12, 713–725. Jahanshahi, M., Profice, P., Brown, R. G., Ridding, M. C., Dirnberger, G., & Rothwell, J. C. (1998). The effects of transcranial magnetic stimulation over the dorsolateral prefrontal cortex on suppression of habitual counting during random number generation. Brain, 121, 1533–1544. Jahanshahi, M., Saleem, T., Ho, A. K., Dirnberger, G., & Fuller, R. (2006). Random number generation as an index of controlled processing. Neuropsychology, 20, 391. Karbach, J., & Schubert, T. (2013). Training-induced cognitive and neural plasticity. Frontiers in Human Neuroscience, 7. Knoch, D., Brugger, P., & Regard, M. (2005). Suppressing versus releasing a habit: Frequency-dependent effects of prefrontal transcranial magnetic stimulation. Cerebral Cortex, 15, 885–887. Kramer, A. F., Hahn, S., & Gopher, D. (1999). Task coordination and aging: Explorations of executive control processes in the task switching paradigm. Acta Psychologica, 101, 339–378. Levasseur-Moreau, J., & Fecteau, S. (2012). Translational application of neuromodulation of decision-making. Brain Stimulation, 5, 77–83. Martin, D. M., Liu, R., Alonzo, A., Green, M., Player, M. J., Sachdev, P., et al. (2013). Can transcranial direct current stimulation enhance outcomes from cognitive training? A randomized controlled trial in healthy participants. International Journal of Neuropsychopharmacology, 16(9), 1927–1936. Miyake, A., Friedman, N. P., Emerson, M. J., Witzki, A. H., Howerter, A., & Wager, T. D. (2000). The unity and diversity of executive functions and their contributions to complex ‘‘frontal lobe’’ tasks: A latent variable analysis. Cognitive Psychology, 41, 49–100. Nitsche, M. A., Cohen, L. G., Wassermann, E. M., Priori, A., Lang, N., Antal, A., et al. (2008). Transcranial direct current stimulation: State of the art 2008. Brain Stimulation, 1, 206–223. Peters, M., Giesbrecht, T., Jelicic, M., Merckelbach, H., American Psychiatric Association, H., Anastasi, A., et al. (2007). The random number generation task: Psychometric properties and normative data of an executive function task in a mixed sample. Journal of the International Neuropsychological Society, 13, 626–634. Reis, J., Schambra, H. M., Cohen, L. G., Buch, E. R., Fritsch, B., Zarahn, E., et al. (2009). Noninvasive cortical stimulation enhances motor skill acquisition over multiple days through an effect on consolidation. Proceedings of the National Academy of Sciences of the United States of America, 106(5), 1590–1595. Robertson, E. M., Pascual-Leone, A., & Miall, R. C. (2004). Current concepts in procedural consolidation. Nature Reviews Neuroscience, 5, 576–582. Spatt, J., & Goldenberg, G. (1993). Components of random generation by normal subjects and patients with dysexecutive syndrome. Brain and Cognition, 23, 231–242. Terhune, D. B., & Brugger, P. (2011). Doing better by getting worse: Posthypnotic amnesia improves random number generation. PLoS ONE, 6, e29206. Towse, J. N., & Neil, D. (1998). Analyzing human random generation behavior: A review of methods used and a computer program for describing performance. Behavior Research Methods, Instruments, & Computers, 30, 583–591. Wilkinson, A. J., & Yang, L. (2012). Plasticity of inhibition in older adults: Retest practice and transfer effects. Psychology and Aging, 27, 606.
