The nature of abacus is one kind of working memory training: a functional MRI and behavior study.
John Yung-Sung Cheng, MD; Chun-Yen Chang, PhD
國際研討會名稱 : STEM 2014, 12-15th, July, the University of British Columbia in Vancouver.
Abstract With aid of educational neuroscience, we can take a different perspective on the traditional Chinese tool for calculation. The aims of our unique longitudinal study are to evaluate the relatively short term effect of six-months abacus-based mental calculation (AMC) training on the functional connectivity network, and to investigate the possibility of AMC training effect transfer to the untrained working memory task. Abacus experts have demonstrated extraordinary potential of mental calculation in the past cross-sectional studies. Non-experts show activity in the prefrontal and perisylvian areas, while experts, with long-term training of at least three to five years, show more activation over premotor and parietal regions.The trained children advanced 9.11 percentile on the WISC-IV exam’s working memory subsection (from 81.19 to 90.30, p = 0.037 < 0.05). Resting state MRI revealed increased connectivity between the left inferior frontal area and the left inferior parietal lobe. The combined results suggested six months of AMC training can generalize to untrained working memory exams, and a midway paradigm shift can be expected along the fronto-parietal circuitry within the relatively long time required to become an expert. Compared to the common practice of task repetition, AMC is not only an arithmetic operation, but an appropriate and interesting tool to improve working memory capabilities by exercising visuo-motor and visuo-spatial skills.
Keywords: abacus, abacus-based mental calculation, functional connectivity, resting state, working memory, working memory training
72
Goals and objectives
The goals of our study are as following: (1) to evaluate the relatively short term effect of six-months abacus-based mental calculation (AMC) training on the functional connectivity network, and (2) to prove the possibility of AMC training effect transfer to the untrained working memory task.
Theoretical framework
With a history tracing back various millenniums in different cultures, the abacus is an essential tool to human civilization growth and development. It is also a significant symbol in Chinese culture for trade and calculation. Inscription of Chinese abacus on the Representative List of the Intangible Cultural Heritage of Humanity was announced on December 2013 by UNESCO for its cultural significance and
"training in abacus-based mental arithmetic is thought to improve a child’s attention span, memory and mental capability." (1) Abacus experts have demonstrated extraordinary potential of mental calculation (2), and even beginner youths are often expected to improve in attention and mental focus abilities. Past imaging studies show a significant difference between experts and non-experts in brain functional areas when performing mathematical tasks (2-5). Non-experts show activity in the prefrontal and perisylvian areas, while experts with long-term training show more activation over premotor and parietal regions. Most abacus-related studies in the past have cross-sectional designs, comparing abacus experts with non-experts. However, in the current Chinese social norm, with the availability of more active or creative extracurricular courses, very few children have the opportunity or willingness to pursue further abacus training after the initial one to two years, far less than the necessary training to become an expert. Therefore, we are curious about the effects of short-term abacus training on functional and behavioral levels. More importantly, how soon can we expect change after initializing training on the functional level and the behavioral level? Children in Chinese society are encouraged to study for at least six months to one year to become acquainted with abacus-based mental calculation (AMC) and express significant behavioral improvement. The Chinese Abacus Association in Taiwan states that in past experience, the greatest level of behavioral improvement is shown after six months to one year. Irwing et al. suggested in 2008 a general intelligence improvement after receiving only 34 weeks of abacus training(6).
Hu et al. suggested in 2011 potential enhancement on white matter integrity after abacus training for 3 years (7). Takeuchi et al. showed in 2013 cognitive training can have probable effects on intrinsic brain activity and connectivity (8). In this study, we hypothesize potential midway paradigm drift, significant change in visuo-motor and visuo-spatial regional connectivity, and working memory behavioral improvement
73
within a six-month abacus training period. After deploying standardized tests, functional MRI (fMRI), and resting state MRI (rsMRI), we evaluate the validity and scale of this change on both the behavioral level and the functional level under a relatively short-term training curriculum.
Methodology
A total of forty-three healthy right-handed children participated in this study. The subjects are split into two groups: abacus training (ABA, n=21, 17 females, mean age
= 9.61 years) and reading control (CON, n=22, 16 females, mean age = 10.57). Two abacus-based mental calculation (AMC) teachers were recruited to provide training for the ABA group over a six-month period. The group received 3.5 hours of training per week (2 hours of teacher-student lessons and 1.5 hours of practice homework) for 25 weeks. The CON group performed general reading exercises for 3.5 hours per week for 25 weeks. The total effective training time was 87.5 hours. The subjects were examined with fMRI, rsMRI functional connectivity (FC), and behavioral exams before and after the six-month training period. Abacus rating exams for the ABA group were provided at the end of the training period by the Chinese Abacus Association. The subjects were asked to perform a simple serial addition task (SSA) during the fMRI scan. The task initialized with the presentation of a plus-sign visual stimulus of 15 seconds, followed by a control or task period of 24 seconds. The control block presented ten eight-digit random numbers presented in series with frequency of 0.5 Hz and subjects were required to gaze at the numbers without calculation. The task session presented ten random one-digit numbers with the same frequency and the subjects were instructed to sum the presented digits. At the end of each task block, the subjects are asked to select by mouse-click the correct sum from a queue of two numbers. During the rsMRI scan, subjects were instructed to keep their eyes closed, remain clear-minded, and think of nothing in particular. The fMRI images were acquired with a 1.5 T MAGNETOM Avanto system (Siemens) using a T2*-weighted gradient-echo echo-planar imaging (EPI) sequence (TR/TE/FA = 3000 ms/50 ms/90°. For each subject, 104 volumes of 20 axial slices per volume were acquired. Resting state acquisitions were performed with the same sequence except TR = 2000 ms and total volume = 180. After each image acquisition, the WISC-IV exam’s working memory subtasks (including digit span, letter-number sequencing, and arithmetic subtests) were deployed to examine the subjects’ working memory capabilities, and an anxiety score questionnaire was used to evaluate the level of anxiety the subjects felt towards mathematical processing and towards the MRI examination. Statistical analyses of functional images for regions of significant change were accomplished using a multi-step process using SPM5 (9) and REST (10).
The EPI data was realigned and spatially normalized into MNI template, then
74
smoothed with a Gaussian kernel of 6 mm. Functional activations were obtained by modeling the data with GLM and group results were analyzed using one sample t-test.
For rsMRI, after preprocessing, FC images underwent seed-based correlation analysis.
Left prefrontal seeds were selected from the results of a task-related fMRI pilot study (11) (Table 1) including left fronto-insular cortex and left Broca area (LBA.) The correlated results were transformed to approximate Gaussian distribution using Fisher’s z transformation. Between-group and within-group comparisons were assessed by unpaired two-sample and paired two-sample t-tests respectively. For regression analysis, the ROIs were selected from activated prefrontal and parietal clusters determined using the anatomical automatic labeling (AAL) template (12), including left inferior frontal gyrus (LIFG) and left angular gyrus (LAG).
Results
In the serial addition task, ABA group demonstrated significantly stronger decreased reaction time in comparison to CON group (ABA p = 0.011 < 0.05 and CON p = 0.258, 17.35% and 9.33% decrease respectively). ABA group also expressed a 15.87% decrease in anxiety to mathematic processing, while CON group expressed an 8.81% increase.In working memory expression, ABA group (n = 21) showed significant change (p = 0.037 < 0.05) and advanced 9.11 percentile on the WISC-IV exam’s working memory subsection (from 81.19 to 90.30), and the effect size is medium (Cohen's d = 0.43). The CON group (n = 22) revealed no significant change (p = 0.238) and advanced 4.83 percentile (from 82.68 to 87.51) The ABA group expressed improvement in the letter-number sequencing subtest, making a 9.60%
increase (9.43 to 10.33,) while the CON group made a 5.00% improvement (10.00 to 10.50.) However, the CON group demonstrated greater growth in the arithmetic subtest from 8.82 to 9.32, while the ABA group made a slight improvement from 9.43 to 9.71.
75
Discussion
In the past decade, several reports of AMC have demonstrated distinct brain activations differences between experts and novices (2-5). On functional MRI results, non-experts show activity in the prefrontal and perisylvian areas, while experts with long-term training show more activation over premotor and superior parietal regions.In recent years, structural and functional connectivity were deployed to study brain intrinsic integrity in children with 3-year training of the AMC (7,13). Increased integrity in white matter tracts were observed over right premotor cortex, corpus callosum, and left occipito-temporal junction. The findings also suggested long-term abacus training may improve working memory capacity (7). Resting state MRI (rsMRI) revealed enhanced connectivity between right supplementary motor area (SMA), and right inferior frontal gyrus (IFG), suggesting more extensive engagement of visuo-spatial-attention networks in trained subjects while performing numerical working memory task (13). Multiple cortices on functional MRI were involved, including IFG, SMA, posterior superior parietal gyrus (PSP), and superior occipital gyrus, and it implies the long-term abacus training effects could be transferred to the untrained working memory task.
Most past studies focus on cross-sectional analysis of abacus experts and non-experts, but neglect the potential midway paradigm shift (paradigm drift) within the relatively long time required to become an expert. But, how soon can we expect changes after initializing training on the functional level and the behavioral level? The Chinese Abacus Association in Taiwan states that in past experience, the initial
76
greatest level of behavioral improvement can be expected in training periods between six months to one year, implying functional changes may be in place before physical changes become detectable. Stigler's study in 1986 revealed children after one year AMC training could develop the ability of mental abacus (14). Irwing's report in 2008 suggested a general intelligence improvement after receiving only 34 weeks of abacus training (6). Therefore, we assume that 6 months is the minimum period required for the change.
Paradigm drift in mental calculation after abacus training
In the local Chinese and Japanese cultural background, speed and accuracy of mental calculation is crucial to daily life and an important aspect of academic success.
The “number sense” greatly depends on the parietal, prefrontal, and cingulate areas, with the HIPS playing a key part in the manipulation of quantity (18). Neuroimaging studies revealed the activation of different neural pathways between untrained children mental calculation and trained experts: the mental calculation task in children mainly activate prefrontal and perisylvian zones of the left hemisphere, with differences shown while performing simple and complex calculation; in sharp contrast, when children are trained in AMC, the activation shifts focus to visuo-motor and visuo-spatial pathways. This is consistent with our knowledge of how AMC is trained in its steps: with each cognitive focus, a change in the underlying functional network can be witnessed in the end results.
With knowledge of the training methods and its end result reflected in neuroimaging, we ask ourselves: what changes can be expected in the relatively short-term training of six months? Research show functional differences between three-year experts and untrained subjects, physical properties can be expected to change after long-term training, and general improvements are noted in time as short as thirty-four weeks. But the greatest behavioral improvements are suggested within six months to one year of training, and we proposed the intrinsic functional circuits should rightfully reflect this change.
Past studies suggest a shift of activation from frontal lobes to parietal lobes was observed with the gain of arithmetic competence, and intra-parietal shift of activation from HIPS to angular gyrus were also reported (19). Evidence for developmental change in mental arithmetic revealed a shift of activations with ages from prefrontal cortex to left inferior parietal lobe. The increased functional specialization of left inferior parietal cortex in mental arithmetic was accompanied by decreased dependence on memory and attention resources with development (20).
Our results demonstrated shift in activation region consistent with our expectation along the fronto-parietal network, but not to the extent of further parietal regions, such as angular gyrus and posterior superior parietal cortex, as suggested by
77
studies involving long-term trained experts. We propose the current and eventual changes to follow a paradigm “drift” along the circuitry, beginning at the prefrontal sectors and moving gradually to the parietal. An increase in functional connectivity between LBA and the left inferior parietal gyrus shown in our rsMRI results suggest the circuitry improves after AMC training.
Our results also revealed a midpoint for this eventual shift that may provide deeper understanding into AMC training’s effect on the learning brain. But, what is the possible direction of future paradigm shift? PSP was proposed to be recruited for mental abacus (3,4), and the activations in angular gyrus increased with the gain of arithmetic knowledge (19). Therefore, we propose two different directions for end-term change: to shift cranially towards the horizontal intra-parietal sulcus (HIPS) and the posterior superior parietal region, or to shift dorsally towards the angular gyrus.
Working memory training
Working memory (WM) is one’s ability to keep limited information in queue and able to be manipulated over a short period of time. This specific cognitive ability is of crucial importance in many different cognitive skills or tasks, and improvement of WM capacity has been proven in several studies (8,21,22,23). WM training-related changes on brain activation were observed in frontal, parietal cortex and basal ganglia (21,22), and it was even proposed intrinsic brain activity and connectivity can be altered by cognitive training (8).
Improvement of implicit WM training can always be expected based on the repetitive training and adjustments of tasks. It is more important to evaluate the potential transfer of WM training to other non-trained tasks. The generalization of WM training effects to other non-trained tasks has been difficult (24,25), but there are promising studies demonstrating transfer after WM training to non-trained tasks can be achievable (22,23,26,27). In the current study, we rediscover a key element of Chinese culture, evaluate the validity and effectiveness of AMC training as a tool to augment WM capacity, and verify the possibility of training effect transfer to the untrained working memory test.
The effects of training in brain plasticity are undeniable, but the future focus should be for which training regimens may result in the best transfer effects, and to whom it is most beneficial (23). AMC training can be considered as an effective method of WM training with a time constraint. Despite its limitations and time-consuming training regimen, it provides stronger training results and retains higher transfer potential than conventional WM training. In modern era, with calculators that can be operated by the simplest of minds, we propose to view the abacus no longer as a tool for trade or computation, but as a toy to learn through play.
78
Conclusions and significance of the study
Within the relatively short (6 months) training of abacus-based mental calculation, significant changes were observed in fronto-parietal network and working memory behavioral expression. Functional connectivity analysis with rsMRI revealed increased connectivity between the left inferior frontal area and the inferior parietal lobe. Most importantly, the AMC training effect can be general ized to affect the untrained tasks of the WISC-IV exam’s working memory subsection.
But are these effects persistent in the long run? And furthermore, what is the required amount of training before one completes the paradigm shift to “transform”
into the mental model as that of experts? And even though the present study shows AMC training effects generalized to untrained working memory tasks, but there is still a long way to go before the leap is made to complex cognitive abilities such as reading, arithmetic, or general intelligence. We recommend future AMC-related studies place focus on the training conditions and target subjects most suitable to produce lasting effects on the mental scheme. Additionally, the evaluation of transfer effects should not be to general skill assessments involving multiple complex cognitive abilities but to specific cognitive traits.
In conclusion, this study offers reappraisal of the abacus, a traditional Chinese tool, and its training in the modern setting. Compared to the common practice of task repetition, AMC training is a more complex and interesting method of cognitive training with a stronger sense of self-accomplishment through its step-by-step training curriculum. Our results also suggested AMC training may increase learners’
confidence in arithmetic tasks and alleviate anxiety. It has potentially long-term positive psychological effects in future learning. AMC is not only an arithmetic operation, but it is also an appropriate tool to improve working memory capabilities by exercising visuo-motor and visuo-spatial skills. With aid of educational neuroscience, we can take a new perspective on the traditional Chinese tool for calculation in the modern age.
79
References
1. UNESCO Culture Sector - Intangible Heritage. Retrieved 2013 Dec 6 from http://www.unesco.org/culture/ich/en/RL/00853
2. Tanaka S. (2002) NeuroReport 13 (2002) 2187–2191.
3. Hanakawa T. (2003) Neuroimage 19 (2003) 296–307.
4. Chen F.Y. (2006) Neuroscience Letters 403: 46–51
5. Chen C.L. (2006) Nuclear Instruments and Methods in Physics Research A 569:
567–571
6. Irwing P. (2008) Personality and Individual Differences. 2008;45(7):694-696.
7. Hu, Y. (2011) Human Brain Mapping 32:10–21 8. Takeuchi, H. (2013) Cortex 49: 2106-2125
9. Statistical Parametric Mapping 5, Wellcome Trust Centre for Neuroimaging, UCL, UK. Retrieved 2013 Jan 24 from http://www.fil.ion.ucl.ac.uk/spm/
10. Song X.W. (2011) REST: A Toolkit for Resting-State Functional Magnetic Resonance Imaging Data Processing. PLoS ONE 6(9): e25031.
11. Pilot study, ISMRM 2013, Cheng Y.S.
12. Tzourio-Mazoyer N. (2002). NeuroImage 15 (1): 273–289.
13. Li Y. (2013) Brain Research 1539 (2013) 24-33.
14. Stigler J. W. (1986) American Journal of Education Vol. 94, No. 4, pp. 447-479 15. Hatano G. (1977). Cognition, 5, 47-55.
16. Hatano G. (1987). Developmental Psychology. 1987, Vol. 23, No. 6,832-838 17. Reisberg, D. (1984). Journal of Experimental Psychology: Learning, Memory, and
Cognition, 10, 203-221.
18. Dehaene S. (2003) Cogn. Neuropsychol. 20 (2003) 487–506.
19. Zamarian L. (2009) Neurosci Biobehav Rev. 2009 Jun;33(6):909-25.
20. Rivera S.M. (2005) Cereb Cortex. 15(11):1779-90.
21. Olesen P.J. (2004) Nature Neurosci. Vol. 7, No. 1 (Jan., 2004), pp. 75-79 22. Klingberg T. (2010) Trends Cogn Sci. 2010 Jul;14(7):317-24.
23. Jaeggi S.M. (2011) Proc Natl Acad Sci USA 108:10081-10086.
24. Owen A.M. (2010) Nature 465:775–778.
25. Melby-Lervåg M. (2013) Developmental Psychology 2013, Vol. 49, No. 2, 270-291
26. Jaeggi SM (2008) Proc Natl Acad Sci USA 105:6829–6833.
27. Kundu B. (2013) J Neurosci. 2013 May 15;33(20):8705-15.
80