Efficient learning produces spontaneous neural repetition suppression in prefrontal cortex

Depósito de Investigación de la Universidad de Sevilla https://idus.us.es/ This is an Accepted Manuscript of an article published by Elsevier in Behavioural Brain Research, available at https://doi.org/10.1016/j.bbr.2009.12.026 Under a CC-BY-NC-ND license Please cite this article in press as: León-Carrión J, et al. Efficient learning produces spontaneous neural repetition suppression in prefrontal cortex. Behav Brain Res (2010), doi:10.1016/j.bbr.2009.12.026 ARTICLE IN PRESS UNCORRECTED PROOF GModel BBR62801–7 Behavioural Brain Research xxx (2009) xxx–xxx 1 Contents lists available at ScienceDirect Behavioural Brain Research journal homepage: www.elsevier.com/locate/bbr Research report1 Efficient learning produces spontaneous neural repetition suppression in prefrontal cortex 2 3 Jose León-Carrióna,b,∗, Meltem Izzetogluc, Kurtulus Izzetogluc, Juan Francisco Martín-Rodrígueza,b, Jesús Damas-Lópezb, Juan Manuel Barroso y Martina, María Rosario Domínguez-Moralesb 4 5 aDepartment of Experimental Psychology, University of Seville, Spain6 bCenter for Brain Injury Rehabilitation (C.RE.CER.), Torneo 23, Seville, Spain7 cSchool of Biomedical Engineering, Science and Health Systems, Drexel University, Philadelphia, PA, USA8 9 article info 10 11 Article history: 12 Received 19 May 200913 Received in revised form 16 December 2009 14 15 Accepted 21 December 200916 Available online xxx 17 Keywords: 18 Neural repetition suppression19 Learning20 Dorsolateral prefrontal cortex21 Near-infrared spectroscopy 22 Working memory23 abstract Our study focuses on the physiological effects of repetition on learning and working memory using an adaptation of Luria’s Memory Word-Task (LMWT). We assess the hemodynamic response in dorsolateral prefrontal cortex (DLPFC) of 13 healthy subjects while completing LMWT. Free word recalls were acquired at the beginning, middle and end of the task. Behavioral results showed that all subjects could recall the complete word list by the 10th trial, which was considered as successful task accomplishment. We observed an attenuation of stimulus-evoked neural activity in prefrontal neurons. Our findings show that the temporal integration of efficient verbal learning is mediated by a mechanism known as neural repetition suppression (NRS). This mechanism facilitates cortical deactivation in DLPFC once learning is successfully completed. This cortical reorganization is interpreted as a progressive optimization of neural responses to produce a more efficient use of neural circuits. NRS could be considered one of the natural mechanisms involved in the processes of memory learning. © 2009 Elsevier B.V. All rights reserved. 1. Introduction24 The capacity to learn and remember reflects the adaptive, plas-25 tic ability of the neural system to change in response to experience.26 Learning is the process of acquiring new information that leads to 27 memorization when motivated information persists. Memory is a28 consequence of learning, a product of persistent functional and bio-29 chemical changes in the relationship between neurons. Learning is 30 a change in the strength of precise neural circuits as a function of 31 practice procedures. These neural changes are different from those32 produced by non-motivated training. Practice and repetition pro-33 duce a restructuring – an anatomical shift – of functional activation34 inbrainareasassociatedwiththestimulus to be learned [1–3]. Prac-35 tice as a learning process may modulate hemodynamic activity in36 the cerebral cortex, producing an increase, decrease or functional 37 reorganization of brain activity [4,5]. 38 Repetition is a natural learning method, refreshing information39 several times before it becomes permanently and temporar-40 ily accessible. When a stimulus is repeated, the neural activity 41 ∗Corresponding author at: Human Neuropsychology Laboratory, School of Psychology, Department of Experimental Psychology, C/Camilo José Cela s/n, University of Seville, Seville, Spain. Tel.: +34 95 457 4137; fax: +34 95 437 4588. E-mail address: [email protected] (J. León-Carrión). produced with each repetition is normally reduced through a 42 mechanism known as neural repetition suppression (NRS) (see 43 Grill-Spector et al. [42] for a review). This mechanism may occur in 44 different brain regions. The reduced level of hemodynamic changes 45 associated with this repetition has also been reported in neu- 46 roimaging studies. We will use NRS to refer to decreased neural 47 responses following stimulus repetition. 48 Learning by repetition requires repetition and the will to 49 remember. It is a voluntary process. One cannot learn solely by 50 means of a priming effect, which does not require the will to retain. 51 What makes learning by repetition different is that the person’s 52 deliberate aim is to keep information in memory (working mem- 53 ory),temporarilyorpermanently.Inlearningbyrepetition,subjects 54 are asked to maintain the repeated information in memory because 55 they will have to report on it later. We define learning by repeti- 56 tion as the voluntary process of keeping information in memory, 57 without deliberate manipulation, by means of periodic repetition. 58 While repetition is a learning mechanism in and of itself, the volun- 59 tary cognitive effort that learning by repetition requires attention 60 through motivation. According to Baddeley [6], learning by simple 61 repetition is unlikely to succeed unless the learner plays an active 62 part and puts some effort into the learning process (although with 63 sufficient practice, the demand for attention becomes minimal). 64 Motivation has a strong effect on learning. Attention to a stimu- 65 lus is sufficient to cause its retrieval, whether or not the subject 66 0166-4328/$ – see front matter © 2009 Elsevier B.V. All rights reserved. doi:10.1016/j.bbr.2009.12.026 Please cite this article in press as: León-Carrión J, et al. Efficient learning produces spontaneous neural repetition suppression in prefrontal cortex. Behav Brain Res (2010), doi:10.1016/j.bbr.2009.12.026 ARTICLE IN PRESS UNCORRECTED PROOF GModel BBR62801–7 2J. León-Carrión et al. / Behavioural Brain Research xxx (2009) xxx–xxx Fig. 1. (Bottom) image representing the fNIRS probe positioning in the human forehead. Four sources and 10 detectors configure a total of 16 information channels covering the whole probed surface. (Top) our fNIRS probe is designed to image DLPFC bilaterally [24]. Probed Brodmann areas include 9, 10 and 46. intends to retrieve it [7,8]. Voluntary processes are intentional and67 controlled.68 Different studies have demonstrated that functional hemody-69 namic changes occur after learning as neural activity in prefrontal70 cortex increases and decreases at specific times. However, the role 71 of time in maintaining working information in the brain has yet72 to be determined. Prefrontal cortex (PFC) has been found to be73 part of the cortical network associated with practice and learning 74 [9]. This cortical area has been linked to visual-motor [10,11] and75 verbal learning processes [1,2,12]. However, neuroimaging studies76 have been inconclusive in characterizing its role in these processes. 77 Within PFC, the dorsolateral prefrontal cortex (DLPFC) has been78 associated with executive processes, namely working memory and 79 maintenance of ongoing information. Working memory is the first80 temporal integrative function of PFC, playing a critical role in the81 temporal organization of learning [13–15]. Our previous studies82 have shown that the neural result of maintaining an arousing stim- 83 ulus in working memory is a period of increased activation in84 DLPFC. This increase is not ongoing, since a decrease must occur85 at some point due to task achievement [15,16]. 86 In our present study, we focus on the physiological effects of87 repetition on learning and working memory. We use an adaptation88 of Luria’s Memory Word-Task (LMWT) to study functional hemody-89 namic changes related to learning by verbal repetition. Functional 90 near infrared spectroscopy (fNIRS) is used to test the hypotheses91 that repeated verbal presentation of information will produce an92 increase in DLPFC activation only during learning, followed by a93 decrease in activation once the information has been learned.94 In LMWT, a list of ten completely unrelated words is read95 aloud to the subject for he or she to memorize. This procedure96 is repeated ten times, while hemodynamic levels of oxyHb and 97 deoxyHbarerecorded.WecomparedtheDLPFC hemodynamic acti-98 vation pattern produced during a first period of word list learning 99 by repetition, with that produced during a second period when the100 list was already learned but the words were still being repeated. In 101 terms of haemoglobin concentrations, we hypothesize an increase 102 in oxyHb and a decrease in deoxyHb in the DLPFC, reflecting activa- 103 tionofthisarea[17,18],duringtasklearningphase,thatis, when the 104 subject is not yet able to recall the complete word list. Conversely, a 105 significant drop in the regional-DLPFC oxyHb concentration, along 106 with a significant increase in regional deoxyHb, is also expected, 107 reflecting deactivation of our region of interest. 108 2. Materials and methods 109 2.1. Subjects 110 Thirteen healthy, right-handed volunteers (4 male and 9 female), ranging in 111 age from 23 to 43 (mean 25.91, standard deviation 5.47), participated in the cur- 112 rent study. IQ was estimated using the Barona method [19]; IQ averages were 113 117.13 for men and 115.37 for women. Written informed consent was obtained 114 from all subjects. Protocol was in accordance with the Declaration of Helsinki 115 (http://www.wma.net/e/policy/b3.htm) and approved by the Local Ethics Commit- 116 tee. 117 2.2. fNIRS description 118 fNIRS is an optical functional neuroimaging technique, developed according 119 to the method designed by Chance and Leigh [20]. Both human [21–25] and ani- 120 mal studies [26–28] have tested the validity of this method. fNIRS is designed to 121 detect changes in the concentration of haemoglobin molecules, the oxygen carrier 122 of red corpuscles. The fNIRS light source emits a light that penetrates the brain. 123 The proportion of light reflected by the corresponding molecules (i.e., oxygenated 124 haemoglobin—oxyHb) and registered by the fNIRS detectors, determines how much 125 oxygenated blood is present. Different studies have reported a high positive cor- 126 relation between PET CBF measurements and fNIRS oxyHb, while a good spatial 127 agreement has been found in studies that employed simultaneous fMRI and fNIRS Q1 128 (Kato et al., 2006; [29,30]). 129 Our fNIRS device (NIM, Inc., Philadelphia, PA) calculates relative changes to 130 baseline values of oxygenated (oxyHb) and deoxygenated haemoglobin (deoxyHb) 131 molecules by means of a continuous wave (CW) spectroscopy system, which applies 132 light to tissue at constant amplitude. CW-type instruments rely on simplified 133 assumptions regarding the tissue being probed and the changes occurring inside 134 the sampling volume. Absolute concentration changes are very difficult to gauge 135 Please cite this article in press as: León-Carrión J, et al. Efficient learning produces spontaneous neural repetition suppression in prefrontal cortex. Behav Brain Res (2010), doi:10.1016/j.bbr.2009.12.026 ARTICLE IN PRESS UNCORRECTED PROOF GModel BBR62801–7 J. León-Carrión et al. / Behavioural Brain Research xxx (2009) xxx–xxx 3 Fig. 2. Task completion and fNIRS data acquisition protocol. Firstly, a 5-s baseline is recorded. Then, subjects listens to the word list during approximately 16 s. This period is followed by a 20-s silence. Subjects then asked to recall the list in trials 1, 5 and 10. Recall period was not included in the statistical analysis. WLR: word reading period; SI: silence interval. given that the real pathlength of light photons is unknown and cannot be measured 136 or inferred.137 The NIM probe is 17.5 cm long and 6.5 cm wide. It contains four light sources 138 surrounded by ten detectors, for a total of 16 channels of data acquisition, covering139 an area of 14 cm ×3.5 cm on the forehead. A source–detector distance of 2.5 cm 140 provides a penetration depth of 1.25 cm. The probe is positioned so that the line of141 sources is set at the line of fronto-polar electrodes [FP1–FP2] (in the International142 10–20 system). This is designed to image cortical areas that correspond to DLPFC143 [24]. DLPFC generally occupies the upper and side regions of the frontal lobes. It144 is comprised of BA 9 and 46. Area 9 occupies the dorsal region of lateral PFC and145 extends medially to the paracingulate of humans. Area 46 is generally located at the146 anterior end of the middle frontal sulcus. The fronto-polar PFC, BA 10, is a region147 positioned above the Orbito Frontal Cortex (OFC), inferior to Area 9, and anterior to148 Area 46, serving as a junction point between the OFC and DLPFC (Krawczyk, 2002)149 (see Fig. 1). A complete data acquisition cycle lasts approximately 330 ms, making150 the temporal resolution approximately 3 Hz.151 2.3. Assessment of verbal learning by repetition152 Luriaˇ ıs Memory Word Test (LMWT) [31], adapted and computerized by León-153 Carrión et al. ([32]; software available at www.neurobirds.com), is a measure with154 extensive clinical use which explores and assesses aspects of verbal memory and155 learning processes. It consists of a list of 10 words – read aloud to the subject with156 one-second intervals between each word – which the subject is asked to learn in ten157 consecutive trials. Subjects were informed that they had ten trials to complete this158 task and that the word list would be read at the beginning of each trial. The exper-159 imenter could not give feedback on the subject’s performance during the course of160 the task.161 Table 1 Mean oxyHb concentrations during the first and second halves of the test and statistical significance of the comparisons between them. Channel 1st half (oxyHb) 2nd half p 1 .038 (.023) −.065 (.035) .031* 2 .066 (.029) −.064 (.035) .012* 3 .045 (.019) −.060 (.038) .025* 4 .065 (.034) −.059 (.045) .018* 5 .063 (.040) −.071 (.049) .068 6 .053 (.035) −.082 (.048) .014* 7 .055 (.040) −.059 (.052) .120 8 .027 (.029) −.080 (.044) .021* 9 .028 (.044) −.044 (.045) .307 10 .009 (.043) −.057 (.043) .201 11 .029 (.053) −.093 (.042) .121 12 .011 (.042) −.078 (.049) .036* 13 .059 (.030) −.079 (.039) .041* 14 .055 (.040) −.074 (.042) .033* 15 .072 (.029) −.072 (.037) .012* 16 .027 (.038) −.062 (.045) .244 *Significant at 0.05 level. Table 2 Mean deoxyHb concentrations during the first and second halves of the test and statistical significance of the comparisons between them. Channel 1st half (deoxyHb) 2nd half p 1−.030 (.015) .039 (.018) .011* 2−.035 (.018) .043 (.017) .006* 3−.030 (.017) .047 (.016) .009* 4−.008 (.011) .028 (.014) .041* 5−.012 (.017) .039 (.020) .147 6−.038 (.014) .031 (.010) .004* 7−.040 (.014) .055 (.032) .016* 8−.038 (.013) .026 (.014) .003* 9−.060 (.018) .055 (.029) .008* 10 −.037 (.021) .040 (.015) .012* 11 −.037 (.022) .025 (.013) .033* 12 −.054 (.021) .025 (.017) .005* 13 −.049 (.015) .042 (.017) .000* 14 −.023 (.019) .027 (.013) .041* 15 −.043 (.027) .050 (.020) .030* 16 −.054 (.021) .020 (.018) .010* *Significant at 0.05 level. In the classic version of LMWT, subjects must recall the word list immediately 162 after each word list reading. This test was adapted for our research purposes and 163 recording settings: each word list reading (WLR) period was followed by a 20-s 164 silence interval (SI) to allow the subject enough time to store the information. Sub- 165 jects were informed that they would be asked to recall as many words as possible 166 after certain trials, to assess the subject’s level of engagement to the task. Our study 167 Fig. 3. Representation of oxyHb mean concentration for 1st and 2nd half of the test. In the 1st half of the test (left), the mean oxyHb concentration is over 0 in almost every area of DLPFC, showing a decrease below 0 in the 2nd half (right). Please cite this article in press as: León-Carrión J, et al. Efficient learning produces spontaneous neural repetition suppression in prefrontal cortex. Behav Brain Res (2010), doi:10.1016/j.bbr.2009.12.026 ARTICLE IN PRESS UNCORRECTED PROOF GModel BBR62801–7 4J. León-Carrión et al. / Behavioural Brain Research xxx (2009) xxx–xxx Fig. 4. (A) Bar graphs represent changes in the mean concentration of oxyHb and deoxyHb in each hemispheric prefrontal area (channels 1–4, 6 and 8 of the left DLPFC and channels 12–15 of right hemisphere) from the 1st half of the test to the 2nd half. A significant decrease in mean oxyHb and an increase in deoxyHb were obtained in all of these channels. (B) A typical time course of oxyHb (blue) and deoxyHb (red) changes across trials (subjects achieved recall of almost the complete word list in the 5th trial). Temporal course of the oxyHb is fluctuant but over 0 in the 1st half of the test, and below 0 when the word list is learned. The inverse pattern can be observed for deoxyHb. Q3 (For interpretation of the references to color in this figure legend, the reader is referred to the web version of the article.) Please cite this article in press as: León-Carrión J, et al. Efficient learning produces spontaneous neural repetition suppression in prefrontal cortex. Behav Brain Res (2010), doi:10.1016/j.bbr.2009.12.026 ARTICLE IN PRESS UNCORRECTED PROOF GModel BBR62801–7 J. León-Carrión et al. / Behavioural Brain Research xxx (2009) xxx–xxx 5 requested free-recall verbal answers only after the 1st, 5th and 10th trials. No feed- 168 back on the subject’s performance was provided after these trials. In order to avoid169 deliberate learning in these trials, we did not inform the subject as to when the recall 170 would be requested.171 2.4. Experimental procedure172 Sixteen-channel fNIRS data was collected from 13 subjects while they completed173 LMWT. The experimental protocol began with a 5-s baseline while subjects were 174 resting. The experimenter then read the preset word list to the subject, which lasted175 approximately 16 s. This was followed by a 20-s silence. The list was then read again, 176 or in trials 1, 5 and 10, the subject was asked to recall the words. He or she was given177 an indefinite period of time to do so, while the experimenter collected his or her178 answers (Fig. 2).179 2.5. fNIRS data processing180 We filtered the raw fNIRS data with a previously developed and adjusted finite181 impulse response low-pass filter (0.14–0.17 Hz) to eliminate possible heart pulsa-182 tion, respiration artifacts, high frequency equipment noise, etc. [24].183 Data on changes in oxyHb and deoxyHb relative to the initial 5-s baseline were184 calculated using modified Beer-Lambert law. In this study, we assumed a differen- 185 tial pathlength factor of 6, which is considered valid for measurements of the adult186 human brain [33]. We fit a first order quadratic polynomial to the oxyHb measure- 187 ments and removed its contents from those recordings to eliminate possible low188 frequency signal drifts.189 After epoch extraction, we performed baseline correction (normalization) by190 extracting the mean of 3 s of data prior to each epoch from the full epoch data which 191 followed. Thus, we could observe hemodynamic changes related to the procedure 192 of interest where its effects may be reduced and compare the differences in these193 changes among different procedures. The next step was to separate the first period194 of the test (trials 1–5) from the second (trials 6–10). The following step involved195 eliminating outliers by removing trials that generated epoch mean values 2.5 stan-196 dard deviations away from each subject’s grand average. After outlier elimination, 197 trial averages for the first and second periods were calculated.198 We deliberately omitted an analysis of the recall phase for various reasons:199 firstly, a free-recall verbal response has no time limit, making the analysis of inter-200 subject averages unsuitable. Secondly, given the modality of the response, the201 vocalization of the subject could produce artifacts in the signal. Only behavioral202 results were used to monitor the word list learning process.203 2.6. Statistical analysis 204 Two 2-way ANOVAs for repeated measures (16 channels ×2 periods) were con-205 ducted. The dependent variables were oxyHb and deoxyHb concentration levels.206 Post hoc tests (Duncan’s test) were performed to study the contribution of DLPFC207 regions to the processes here studied. Spearman’s rank correlations were also calcu-208 lated. We calculated the difference in words recalled in the first period (5th trial–1st209 trial) and correlated this variable with the difference in oxyHb concentration lev-210 els and the difference in deoxyHb levels for these trials. The same procedure was211 followed for the second period (10th trial–5th trial).212 3. Results213 3.1. Behavioral results214 In trials 1, 5 and 10, we recorded all subject performances to215 assure that they had learned the complete word list. The percent- 216 age of subjects who recalled the complete word list varied across217 trials (Chi-Squared = 25.2; d.f. = 4; p< 0.0001). Only 7.69% (Adjusted218 residual [AR] = 1.4) recalled 10 words in the 1st trial, whereas 76.9% 219 recalled the complete word list in the 5th trial. All subjects recalled220 the 10 words in the 10th trial.221 3.2. Hemodynamic results222 OxyHb results showed significant main effects of periods223 (F[1,12] = 6.31, p= 0.027). Post hoc analyses showed a higher 224 level of oxyHb concentration during the first period of LMWT225 (M= 0.044 ␮M) than during the second (M=−0.069 ␮M) (Fig. 2).226 The remaining comparison and interaction did not reach signif-227 icance (both p’s > 0.9). DeoxyHb results showed a main effect of228 period (F[1,12] = 20.68, p< 0.001). Post hoc analysis showed a lower 229 level of deoxyHb concentration during the first period of the230 test (M=−0.037 ␮M; corrected p< 0.01) than during the second231 (M= 0.037 ␮M) (Fig. 3). The remaining comparisons did not reach 232 significance (all p’s > 0.12). 233 Significant interactions between channels and period were 234 also detected for oxyHb (F[15,180] = 2.01; p= 0.016) and deoxyHb 235 (F[15,180] = 1.81; p= 0.036). Post Hoc analyses also showed that 236 channels with significant highest differences between the first and 237 second period of the test were 2, 3, 4, 13, 14, 15 and 16 (all p’s < 0.05). 238 These channels would correspond to BA 45, 46, 9 and 10 bilaterally. 239 These channels also showed a lower deoxyHb concentration during 240 the first period of the test (all p’s < 0.05). Fig. 4 shows mean differ- 241 ences for significant channels between the two test periods, as well 242 as typical temporal courses for oxyHb and deoxyHb. Tables 1 and 2 243 display these differences and associated statistical comparisons. 244 Correlation analyses showed significant positive correlations 245 between oxyHb and memory performance in channels 1–4 and 246 12–16 for the first period (Spearmann’s ’s ranging from 0.54 to 247 0.73; p’s ranging from 0.03 to 0.005). In these channels, significant 248 negative correlations were found between memory performance 249 and deoxyHb (’s ranging from −0.59 to −0.78; p’s ranging from 250 0.02 to 0.003). For the second period, significant negative corre- 251 lations were found between oxyHb and memory performance in 252 channels 1–3 and 11–15 (’s ranging from −0.51 to −0.66; p’s rang- 253 ing from 0.04 to 0.013). Positive correlations between deoxyHb and 254 memory performance were also detected for this period in chan- 255 nels 1–4 and 12–16 (’s ranging from 0.54 to 0.61; p’s ranging from 256 0.03 to 0.014). 257 4. Discussion 258 We obtained one clear and significant result: during the learning 259 process, an increase in activation takes place in right and left DLPFC, 260 which then decreases or ceases when the learning is complete 261 (Fig. 2). What we observed was the existence of Neural Repetition 262 Suppression (NRS), defined as an attenuation of stimulus-evoked 263 neural activity attributed to intrinsic and automatic processes in 264 cortical neurons [34–36]. Our finding shows NRS in DLPFC after 265 effective verbal learning, an adaptation of hemodynamic activity 266 in DLPFC during multiple repetitions. The correlations between 267 subject memory recall and fNIRS activation evidences the neuro- 268 physiological substrates related to LWMT (see Tables 1 and 2). The 269 correlation analysis demonstrates memory performance during the 270 first and second part of the task and associated hemodynamic 271 activity in DLPFC. The physiological correlates of reduced memory 272 load after repetition is NRS. Our data also reinforces the findings 273 of Garrido et al. [37] that auditory learning is associated with 274 repetition-dependent plasticity in the human brain. 275 In order to provide cognitive meaning to our results, we must 276 first understand the biological significance behind the two different 277 hemodynamic activities produced during the task: oxyHB increase 278 and decrease. In a previous fNIRS study that sought to establish the 279 relationships between prefrontal cortex (PFC) and cognitive con- 280 trol, we found that oxygen availability in superior PFC is linked 281 to an increase in metabolism associated with attention level and 282 effectiveness of cognitive control [16]. Matsui et al. [38] studied 283 the hemodynamic response of the prefrontal area during words 284 memory learning and found that during the task, oxyHb concentra- 285 tions increased and deoxyHb concentrations decreased. Recently 286 Molteni et al. [39] found that working memory load increased brain 287 oxygenation during the first half of a memory task, which persisted 288 throughout the central resting period. They also found that decreas- 289 ing working memory load was coupled with oxyHb decrease. 290 An increase in oxygenation in right DLPFC during learning has 291 been related to greater demands on cognitive control processes, 292 with the goal of increasing neural and cognitive efficiency [40].293 According to Büchel et al. [41], activation in specialized cortical 294 areas changes with time during learning. Furthermore, the time 295 Please cite this article in press as: León-Carrión J, et al. Efficient learning produces spontaneous neural repetition suppression in prefrontal cortex. Behav Brain Res (2010), doi:10.1016/j.bbr.2009.12.026 ARTICLE IN PRESS UNCORRECTED PROOF GModel BBR62801–7 6J. León-Carrión et al. / Behavioural Brain Research xxx (2009) xxx–xxx course of these plastic changes is highly correlated to individual296 learning performance, reflecting the progressive optimization of297 neural responses elicited by the task. The psychobiological mean-298 ing of this finding is that NRS must be present in successful verbal299 learning to maximize the effectiveness and accuracy of learning,300 and also to free up space in working memory. This optimization301 of neural responses stems from a verbal learning process requiring302 time and repetition of the material to be learned to be effective.303 Four mechanisms that may play a role in NRS were reported by304 Grill-Spector et al. [42]. Firing-rate adaptation, a sign of “reduced305 excitability”, occurs briefly due to higher potassium currents that306 increase membrane conductance. Synaptic Depression, a short-307 lived drop in the efficiency of synaptic activity, is due to drops308 in “pre-synaptic neurotransmitter release”. Long-term Depression,309 a sign of changes in plasticity due to “correlated pre- and post-310 synaptic activity”, can last hours and tends to involve various311 stages (including protein synthesis). Finally, long-term potentia-312 tion, also long-lasting and involving various stages, is a sign of313 higher efficiency of synaptic activity. According to Baldeweg [43]314 and Friston [44], the effect of repetition supports evidence of a315 hierarchical sensory system that uses predictive coding, receiv- 316 ing bottom-up sensory input (evidence) from below and top-down317 input (prediction) from above. According to these authors, pre-318 diction error is reduced during stimulus repetition, producing an319 adjustment of the connection strength between levels through320 synaptic plasticity. The consequence of behavioral facilitation and321 reduced neural activity after stimulus repetition seems to be sup-322 ported by long-term potentiation between neurons. The greater323 the number of repetitions, the faster and earlier memory trace pro-324 cessing is detected, diminishing hemodynamic response until no325 activation exists. This corresponds to the facilitation model pro- 326 posed by Grill-Spector et al. [42] which assumes that synaptic327 potentiation causes faster processing and that facilitation will be328 more pronounced when top-down information is processed.329 Decreases in extent or intensity of activation have also been330 found in studies related to task practice. Hebb [45] found that331 the ability to reproduce a short list of verbal items from memory332 improved immediately following presentation. This phenomenon,333 known as priming, has been studied to understand how prior expo- 334 sure to a stimulus can facilitate its subsequent identification and 335 classification. Priming is also mediated by repetition suppression336 and has been reported in imaging studies. Priming and memory337 by repetition are different processes. Priming does not require the338 same degree of encoding as does memory, occurring when an item339 is encoded and repeated. When a previously encoded item that340 reappears is neither the focus of attention, nor needed to com- 341 plete the task at hand, its recent processing is not influential, as342 is the case in priming [35]. Some authors have found similarities343 between priming and the physiological properties of decreased344 neural responses occurring with item repetition. According to Say-345 ala et al. [46] and Jansma et al. [47], the primary goal for underlying346 activation decreases is increased neural efficiency, and a more effi-347 cient use of specific ‘neural circuits’. According to Poldrack [4], 348 decreases in activation represent a contraction of the neural repre-349 sentation of the stimulus. While learning a task through practice, a350 decrease in the hemodynamic activity of DLPFC (such as we found)351 can reflect a good performance [47]. 352 Our findings are consistent with the accepted evidence that353 DLPFC is highly involved in working memory. Active maintenance354 of information load in working memory recruits DLPFC [48–50].We 355 found higher bilateral levels of oxyHb in DLPFC during the learn- 356 ing process than during post-learning, suggesting that this region 357 is involved in working memory. Working memory (WM) allows358 humans to maintain a limited amount of information in an active359 state for a brief period of time [51–53] facilitating efficient encoding 360 of information [54]. The timely activation of neurons in PFC is cru-361 cial for efficient learning. Our data shows that effective learning by 362 repetition produces a reduced demand on WM through decreased 363 DLPFC oxygenation as a consequence of a shift from controlled 364 to automatic processing. Our data is consistent with other behav- 365 ioral studies reporting that practice of a cognitive task increases 366 performance speed, reduces response variability, and lowers error 367 rate. This reflects the transition from controlled to automatic pro- 368 cessing. It also indicates that automatic and controlled processing 369 have the same functional anatomical substrate, but differ in effi- 370 ciency. According to Koch et al. [55], the transition from controlled 371 to automatic processing WM is associated with exponential signal 372 decreases in task-relevant regions, suggest that temporal changes 373 in brain activation patterns can be attributed to enhanced efficiency 374 of information processing as a result of cognitive practice. 375 Our finding of a NRS effect in DLPFC associated with verbal 376 learning is a first among studies in the literature. Büchel et al. [41] 377 suggested that a decrease in activation during learning enhances 378 response selectivity through time-dependent changes in effective 379 connectivity at synaptic levels. Our findings provide evidence that 380 these decrements in DLPFC activation allow processing speed to 381 increase and performance to improve [56].382 In our study, we considered learning successful when all sub- 383 jects (100%) could recall the complete word list. This occurred 384 in the 10th trial. However, due to our experimental protocol and 385 technique, we could not identify the trial where each subject had 386 accomplished the task. Our behavioral data showed that the entire 387 sample, as a homogenous group, accomplished the task in the last 388 trial. LMWT is designed to assess clinical populations, particularly 389 people with brain injury. Healthy subjects usually perform effi- 390 ciently in this task, whereas learning rates are significantly worse 391 in patients with memory difficulties [57]. Our data suggests that 392 LMWT could be a valuable tool for memory assessment in patients 393 with neurological damage. 394 In conclusion, our data also points out that NRS is necessary not 395 only to complete or close learning, but to keep DLPFC free and avail- 396 able for engagement in other tasks upon demand. The NRS found in 397 our study may be the first step in the temporal integration of learn- 398 ing; first the representation of new information is maintained in 399 DLPFC during learning; then it is stored in a more permanent place 400 for future use. NRS seems to be a natural physiological phenomenon 401 that occurs when an the repetition of an item is motivated, regard- 402 less of its behavioral significance This is an important finding for 403 educational purposes, as it reinforces the psychobiological role of 404 repetition in learning. Our results also have implications for serial 405 studies using neuroimaging and for treatment of memory disor- 406 ders in patients with acquired brain injury, but further studies are 407 needed in this regard. 408 Uncited references Q2 409 [58–60].410 Acknowledgements 411 This study has been supported by the Spanish Ministry of Sci- 412 ence and Innovation grant (PSI-2008-04218/PSIC) and the Center 413 for Brain Injury Rehabilitation (C.RE.CER) grant at Seville, Spain. 414 References 415 [1] Raichle ME, Fiez JA, Videen TO, MacLeod AM, Pardo JV, Fox PT, et al. Practice- 416 related changes in human brain functional anatomy during nonmotor learning. 417 Cereb Cortex 1994;4:8–26. 418 [2] Petersen SE, van Mier H, Fiez JA, Raichle ME. The effects of practice on the func- 419 tional anatomy of task performance. Proc Natl Acad Sci USA 1998;99:853–60. 420 [3] Petersson KM, Elfgren C, Ingvar M. Dynamic changes in the functional anatomy 421 of the human brain during recall of abstract designs related to practice. Neu- 422 ropsychologia 1999;37:567–87. 423 Please cite this article in press as: León-Carrión J, et al. Efficient learning produces spontaneous neural repetition suppression in prefrontal cortex. Behav Brain Res (2010), doi:10.1016/j.bbr.2009.12.026 ARTICLE IN PRESS UNCORRECTED PROOF GModel BBR62801–7 J. León-Carrión et al. / Behavioural Brain Research xxx (2009) xxx–xxx 7 [4] Poldrack RA. Imaging brain plasticity: conceptual and methodological issues—a 424 theoretical review. Neuroimage 2000;12:1–13.425 [5] Garavan H, Kelley D, Rosen A, Rao SM, Stein EA. Practice-related functional acti-426 vation changes in a working memory task. Microsc Res Tech 2000;51:54–63. 427 [6] Baddeley A. Working memory, thought, and action. New York: Oxford Univer-428 sity Press; 2007.429 [7] Kahneman D, Treisman A. Changing views of attention and automaticity. In:430 Parasuraman R, Davies R, editors. Varieties of attention. New York: Academic431 Press; 1984. p. 29–61.432 [8] Zbrodoff NJ, Logan GD. On the autonomy of mental processes: a case study of433 arithmetic. J Exp Psychol Gen 1985;115:118–30.434 [9] Fuster JM. The prefrontal cortex. Anatomy, physiology and neuropsychology of435 the frontal lobe. Philadelphia, PA: Lippincott-Raven; 1997 p. 84.436 [10] Debaere F, Wenderoth N, Sunaert S, Van Hecke P, Swinnen SP. Changes in 437 brain activation during the acquisition of a new bimanual coodination task.438 Neuropsychologia 2004;42:855–67.439 [11] Landau SM, Schumacher EH, Garavan H, Druzgal TJ, D’Esposito M. A functional440 MRI study of the influence of practice on component processes of working441 memory. Neuroimage 2004;22:211–21.442 [12] Fletcher P, Büchel C, Josephs O, Friston K, Dolan R. Learning-related neuronal443 responses in prefrontal cortex studied with functional neuroimaging. Cereb444 Cortex 1999;9:168–78.445 [13] Goldman-Rakic PS. Architecture of the prefrontal cortex and the central exec-446 utive. Ann N Y Acad Sci 1995;769:71–83. 447 [14] Fuster JM. The prefrontal cortex—an update: time is of the essence. Neuron448 2001;2:319–33.449 [15] Leon-Carrion J, Damas J, Izzetoglu K, Pourrezai K, Martín-Rodríguez JF, Barroso 450 y Martin JM, et al. Differential time course and intensity of PFC activation for451 men and women in response to emotional stimuli: a functional nearinfrared452 spectroscopy (fNIRS) study. Neurosci Lett 2006;403:90–5.453 [16] León-Carrion J, Damas-López J, Martín-Rodríguez JF, Domínguez-Roldán JM, 454 Murillo-Cabezas F, Barroso Y, et al. The hemodynamics of cognitive control: 455 the level of concentration of oxygenated hemoglobin in the superior prefrontal456 cortex varies as a function of performance in a modified Stroop task. Behav457 Brain Res 2008;193:248–56.458 [17] Strangman G, Franceschini MA, Boas DA. Factors affecting the accuracy of459 near-infrared spectroscopy concentration calculations for focal changes in oxy- 460 genation parameters. Neuroimage 2003;18:865–79.461 [18] Huppert TJ, Hoge RD, Diamond SG, Franceschini MA, Boas DA. A temporal com-462 parison of BOLD, ASL, and NIRS hemodynamic responses to motor stimuli in463 adult humans. Neuroimage 2006;29:368–82.464 [19] BaronaA, Reynolds CR, Chastain R. A demographically based indexofpremorbid465 intelligence. J Consult Clin Psychol 1984;52:885–7. 466 [20] Chance B, Leigh Jr JS. Oxygen intermediates and mixed valence states of467 cytochrome oxidase: infrared absorption difference spectra of compounds468 A, B, and C of cytochrome oxidase and oxygen. Proc Natl Acad Sci USA469 1977;74:4777–80.470 [21] Chance B. NMR and time-resolved optical studies of brain imaging. Adv Exp471 Med Biol 1993;333:1–7.472 [22] Chance B, Luo Q, Nioka S, Alsop DC, Detre JA. Optical investigations of physiol-473 ogy: a study of intrinsic and extrinsic biomedical contrast. Philos Trans R Soc 474 Lond B Biol Sci 1997;352:707–16.475 [23] Hoshi Y, Tamura M. Dynamic multichannel near-infrared optical imaging of476 human brain activity. J Appl Physiol 1993;75:1842–6.477 [24] Izzetoglu M. Functional near-infrared imaging. IEEE Trans Neural Netw 478 2005;13:153–9. 479 [25] Kato H, Izumiyama M, Koizumi H, Takahashi A, Itoyama Y. Near-infrared480 spectroscopic topography as a tool to monitor motor reorganization after hemi-481 paretic stroke: a comparison with functional MRI. Stroke 2002;33:2032–6. 482 [26] Hoshi Y, Kobayashi N, Tamura M. Interpretation of near-infrared spectroscopy483 signals: a study with a newly developed perfused rat brain model. J Appl Physiol484 2001;90:1657–62.485 [27] Fuster J, Guiou M, Ardestani A, Cannestra A, Sheth S, Zhou YD, et al. Near- 486 infrared spectroscopy (NIRS) in cognitive neuroscience of the primate brain.487 Neuroimage 2005;26:215–20.488 [28] Onoe H, Watanabe Y, Tamura M, Hayaishi O. REM sleep-associated489 haemoglobin oxygenation in the monkey forebrain studied using near-infrared490 spectrophotometry. Neurosci Lett 1991;129:209–13.491 [29] HockC,VillringerK, Müller-Spahn F, Wenzel R, Heekeren H, Schuh-Hofer S, et al.492 Decrease in parietal cerebral hemoglobin oxygenation during performance of493 a verbal fluency task in patients with Alzheimer’s disease monitored by means 494 of near-infrared spectroscopy (NIRS)—correlation with simultaneous rCBF-PET 495 measurements. Brain Res 1997;755:293–303.496 [30] Villringer K, Minoshima S, Hock C, Obrig H, Ziegler S, Dirnagl U, et al. Assessment497 of local brain activation. A simultaneous PET and nearinfrared spectroscopy498 study. Adv Exp Med Biol 1997;413:149–53. [31] Christensen A-L. Luriaˇ ıs neuropsychological investigation. Copenhagen: 499 Murksgaard; 1975. 500 [32] León-Carrión J, Taaffe PJ, Barroso y Martín JM. Neuropsychological assessment 501 of persons with acquired brain injury. In: León-Carrión J, Zitnay GA, von Wild 502 K, editors. Brain injury treatment: theories and practices. England and USA: 503 Taylor & Francis; 2006. p. 275–312. 504 [33] Kubota Y, Toichi M, Shimizu M, Mason RA, Findling RL, Yamamoto K, et al. 505 Prefrontal hemodynamic activity predicts false memory—a near-infrared spec- 506 troscopy study. Neuroimage 2006;31:1783–9. 507 [34] Desimone R. Neural mechanisms for visual memory and their role in attention. 508 Proc Natl Acad Sci USA 1996;93:13494–9, 16. 509 [35] Wiggs CL, Martin A. Properties and mechanisms of perceptual priming. Curr 510 Opin Neurobiol 1998;8:227–33. 511 [36] Summerfield C, Trittschuh EH, Monti JM, Mesulam MM, Egner T. Neural rep- 512 etition suppression reflects fulfilled perceptual expectations. Nat Neurosci 513 2008;11:1004–6. 514 [37] Garrido MI, Kilner JM, Kiebel SJ, Friston K. Evoked brain responses are generated 515 by feedback loops. Proc Natl Acad Sci USA 2008;104:20961–6. 516 [38] Matsui M, Tanaka K, Yonezawa M, Kurachi M. Activation of the prefrontal cor- 517 tex during memory learning: near-infrared spectroscopy study. Psychiatry Clin 518 Neurosci 2007;61:31–8. 519 [39] Molteni E, Butti M, Bianchi AM, Reni G. Activation of the prefrontal cortex dur- 520 ing a visual n-back working memory task with varying memory load: a near 521 infrared spectroscopy study. Conf Proc IEEE Eng Med Biol Soc 2008:4024–7. 522 [40] Rajah MN, Ames B, D’Esposito M. Prefrontal contributions to domain-general 523 executive control processes during temporal context retrieval. Neuropsycholo- 524 gia 2008;46:1088–103. 525 [41] Büchel C, Coull JT, Friston KJ. The predictive value of changes in effective con- 526 nectivity for human learning. Science 1999;283:1538–41. 527 [42] Grill-Spector K, Henson R, Martin A. Repetition and the brain: neural models of 528 stimulus-specific effects. Trends Cogn Sci 2006;10:14–23. 529 [43] Baldeweg T. Repetition effects to sounds: evidence for predictive coding in the 530 auditory system. Trends Cogn Sci 2006;10:93–4. 531 [44] Friston K. A theory of cortical responses. Philos Trans R Soc Lond B Biol Sci 532 2005;360:815–36. 533 [45] Hebb DO. Distinctive features of learning in the higher animal. In: Delafresnaye 534 JF, editor. Brain mechanisms and learning. Oxford: Blackwell; 1961. p. 37–46. 535 [46] Sayala S, Sala JB, Courtney SM. Increased neural efficiency with repeated perfor- 536 mance of a working memory task is information-type dependent. Cereb Cortex 537 2006;16:609–17. 538 [47] Jansma JM, Ramsey NF, Slagter HA, Kahn RS. Functional anatomical correlates 539 of controlled and automatic processing. J Cogn Neurosci 2001;13:730–43. 540 [48] Cohen JD, Perlstein WM, Braver TS, Nystrom LE, Noll DC, Jonides J, et al. Tem- 541 poral dynamics of brain activation during a working memory task. Nature 542 1997;386:604–8. 543 [49] Postle BR, D’Esposito M. What-Then-Where in visual working memory: an 544 event-related fMRI study. J Cogn Neurosci 1999;11:585–97. 545 [50] Rypma B, Prabhakaran V, Desmond JE, Glover GH, Gabrieli JD. Load-dependent 546 roles of frontal brain regions in the maintenance of working memory. Neu- 547 roimage 1999;9:216–26. 548 [51] Owen AM, Evans AC, Petrides M. Evidence for a two-stage model of spatial 549 working memory processing within the lateral frontal cortex: a positron emis- 550 sion tomography study. Cereb Cortex 1996;6:31–8. 551 [52] D’Esposito M, Ballard D, Aguirre GK, Zarahn E. Human prefrontal cortex 552 is not specific for working memory: a functional MRI study. Neuroimage 553 1998;8:274–82. 554 [53] Smith EE, Jonides J. Storage and executive processes in the frontal lobes. Science 555 1999;283:1657–61. 556 [54] D’Esposito M, Postle BR, Rypma B. Prefrontal cortical contributions to work- 557 ing memory: evidence from event-related fMRI studies. Exp Brain Res 558 2000;133:3–11. 559 [55] Koch K, Wagner G, von Consbruch K, Nenadic I, Schultz C, Ehle C, et al. Tempo- 560 ral changes in neural activation during practice of information retrieval from 561 short-term memory: an fMRI study. Brain Res 2006;1107:140–50. 562 [56] Jonides J, Naveh-Benjamin M, Palmer J. Assessing automaticity. Acta Psychol 563 (Amst) 1985;60:157–71. 564 [57] Altepeter TS, Adams RL, Buchanan WL, Buck P. Luria memory words test and 565 Wechsler memory scale: comparison of utility in discriminating neurologically 566 impaired from controls. J Clin Psychol 1990;46:190–3. 567 [58] Chance B. Computer hardware and software. Philadelphia, PA: Near Infrared 568 Monitoring (NIM) Inc.; 2003. 569 [59] Fuster JM, Alexander GE. Neuron activity related to short-term memory. Sci- 570 ence 1971;173:652–4. 571 [60] Fuster JM. Unit activity in prefrontal cortex during delayed-response 572 performance: neuronal correlates of transient memory. J Neurophysiol 573 1973;36:61–78. 574