 Research
 Open Access
 Published:
Algebra dissociates from arithmetic in the brain semantic network
Behavioral and Brain Functions volume 18, Article number: 1 (2022)
Abstract
Background
Mathematical expressions mainly include arithmetic (such as 8 − (1 + 3)) and algebra (such as a − (b + c)). Previous studies have shown that both algebraic processing and arithmetic involved the bilateral parietal brain regions. Although previous studies have revealed that algebra was dissociated from arithmetic, the neural bases of the dissociation between algebraic processing and arithmetic is still unclear. The present study uses functional magnetic resonance imaging (fMRI) to identify the specific brain networks for algebraic and arithmetic processing.
Methods
Using fMRI, this study scanned 30 undergraduates and directly compared the brain activation during algebra and arithmetic. Brain activations, singletrial (itemwise) interindividual correlation and meantrial interindividual correlation related to algebra processing were compared with those related to arithmetic. The functional connectivity was analyzed by a seedbased region of interest (ROI)toROI analysis.
Results
Brain activation analyses showed that algebra elicited greater activation in the angular gyrus and arithmetic elicited greater activation in the bilateral supplementary motor area, left insula, and left inferior parietal lobule. Interindividual singletrial brainbehavior correlation revealed significant brainbehavior correlations in the semantic network, including the middle temporal gyri, inferior frontal gyri, dorsomedial prefrontal cortices, and left angular gyrus, for algebra. For arithmetic, the significant brainbehavior correlations were located in the phonological network, including the precentral gyrus and supplementary motor area, and in the visuospatial network, including the bilateral superior parietal lobules. For algebra, significant positive functional connectivity was observed between the visuospatial network and semantic network, whereas for arithmetic, significant positive functional connectivity was observed only between the visuospatial network and phonological network.
Conclusion
These findings suggest that algebra relies on the semantic network and conversely, arithmetic relies on the phonological and visuospatial networks.
Background
Mathematical expressions are frequently conceived and represented in arithmetic and algebra. Arithmetic is concretely represented as the combination of digits and operators (such as 8−(1 + 3)). Contrastingly, algebra includes abstract operations with letters and operators (such as a−(b + c)) and has more abstract representation than arithmetic in form [1]. Although previous studies demonstrated the dissociation between algebra and arithmetic [2,3,4,5,6,7,8], the neural mechanisms underlying this dissociation remain unclear. The current study used functional magnetic resonance imaging (fMRI) to examine how algebraic processing dissociates from arithmetic processing in brain networks.
Arithmetic in the brain
In the arithmetic expression “4 × 5−9,” the subexpression “4 × 5” is recursively combined with “9” by subtraction. It has been proposed that the arithmetic operations both involved numerical quantity calculation and orderrule of arithmetic operations [9]. Neuroimaging and neuropsychological studies have revealed that arithmetic processing is subserved by the visuospatial network, which typically includes the bilateral intraparietal sulcus (IPS) and bilateral inferior and superior parietal lobules [10,11,12,13,14]. Several studies have shown that arithmetic activated the bilateral parietal lobules also support number processing [14,15,16,17]. Approximately 73% of the behavioral variance of visuospatial magnitude processing has been attributed to the change in cortical thickness in the right superior parietal cortex, and approximately 55% of the behavioral variance of children’s arithmetic abilities has been attributed to the change in cortical folding in the right IPS [18]. Patients with injuries in the parietal cortex typically experience arithmetic impairment [19, 20] and impaired visuospatial performance [21].
During arithmetic computation, the phonological system also supports the coding of visual or auditory numbers, the results of numerical calculations were temporarily stored in the phonological loop of working memory [22, 23]. fMRI studies showed that arithmetic processing relies on the phonological network, which typically includes the bilateral supplementary motor areas and the precentral gyrus [15, 24,25,26,27,28]. For example, Zhou et al. [27] reported that both addition and multiplication are supported by a broad neural system involving the supplementary motor areas, precentral gyrus, bilateral IPS, and middle frontal gyrus. One recent study showed that exact arithmetic involves more phonological or verbal processing localized in the left Rolandic operculum, precentral gyrus, and supplementary motor area than approximate arithmetic [25]. Neuropsychological investigations have also shown that patients with infarction in the left frontal lobe (including the precentral gyrus) have difficulty retrieving arithmetic facts [29, 30].
Algebra in the brain
Algebra (such as a−(b + c)) contains abstract operations with letters and operators, which is similar to arithmetic in procedural knowledge [9, 31]. Algebra is similar to arithmetic in form, but the former involves more abstract representation than the latter. Evidence from several domains of cognitive neuroscience has illustrated the association and dissociation between the processing of algebra and arithmetic.
Both algebra and arithmetic processing is associated with regions in the parietal lobes. As mentioned above, arithmetic processing typically activate the bilateral IPS [10,11,12,13,14, 28]. Similarly, compared to language syntax processing, algebraic processing typically recruits the bilateral parietal brain regions around the IPS more, to represent numerical magnitude [32,33,34]. It dissociated from the typical brain activation in classic language processing including the left inferior frontal gyrus and posterior and anterior middle temporal gyrus [12, 13, 35, 36]. Algebraic operations did not recruit any more language resources than did simple reading (i.e., the grammar task); in contrast, they did rely on areas previously linked to arithmetic cognition.
Algebraic processing involves more abstract language symbolic; thus, it dissociated from arithmetic in the application of conceptual and semantic knowledge [12, 32, 37, 38]. For example, Geary et al. [38] found that algebraic cognition was related to semantic memory systems of addition facts in ninth graders. Neuroimaging studies further implied that the brain network’s response to algebra is also associated with language semantic processing. A metaanalysis study of 120 functional neuroimaging studies suggested that the semantic processing system actually covers a wide range of brain regions including seven leftlateralized regions: posterior inferior parietal lobe (angular gyrus), middle temporal gyrus, fusiform and parahippocampal gyri, dorsomedial prefrontal cortex, inferior frontal gyrus, ventromedial prefrontal cortex, and posterior cingulate gyrus [39]. In a task involving algebraic transformation using nonarithmetic symbols (“artificial algebra”), adults showed activation in the left posterior parietal and left prefrontal regions [37]. When both expert mathematicians and nonmathematicians made semantic judgements regarding algebraic statements, a bilateral network involving the prefrontal, parietal, and inferior temporal regions was recruited [32]. Recent studies have shown that compared to simple numerical processing and arithmetical computation, algebraic terms elicited greater activation in the typically semantic network such as left angular gyrus, left middle temporal gyrus, and left inferior frontal gyrus [12, 40]. These regions typically overlap with the typically semantic network [39]. These findings suggested that algebra is dissociated from arithmetic in brain activation.
The present study
Although previous studies demonstrated the dissociation between algebra and arithmetic [2, 4,5,6,7,8, 12, 32, 35, 37, 38], the neural bases of the dissociation between algebraic processing and arithmetic is still unclear. This study uses fMRI to identify the specific neural bases for the dissociation between algebraic and arithmetic processing. The arithmetic and algebraic tasks used in this study were matched in surface form and difficulty. In the present study, algebraic used the expressions of the operational principles for addition, subtraction, multiplication, and division (such as a−(b + c)). To match the algebraic task, we used the expressions requiring arithmetic procedures (such as 8−(1 + 3)) rather than that requiring the simple retrieval of arithmetic facts. For each problem of the algebraic and arithmetic tasks, two algebraic expressions or two arithmetic expressions were presented simultaneously on the screen, one at the top and one at the bottom. Participants were asked to judge whether the results of the two expressions were equal.
Prior studies have indicated that arithmetic involves numerical quantity operations and phonological processing [10,11,12,13,14, 28]. Contrastingly, the application of semantic knowledge is involved in algebraic processing [12]. Therefore, we hypothesized that algebra relies more on the semantic network and arithmetic involves the phonological and visuospatial networks based on the aforementioned reports. It is important to clarify whether algebra relies on visuospatial or semantic networks. If algebra and arithmetic rely on different brain networks, it has strong implications for education and teaching. Distinct approaches should be used to learn algebra and arithmetic. A verbalization approach emphasizing semantic comprehension was used to learn algebra, and visualization and phonological approaches were used to learn arithmetic [41]. To test this hypothesis, we compared the fMRI data regarding brain activation for algebra with that for arithmetic, and we conducted a singletrial (itemwise) interindividual correlation analysis and a traditional meantrial interindividual correlation analysis to examine the brainbehavior correlations [25, 42].
Methods
Participants
Thirty righthanded undergraduate students (15 male and 15 female; mean age = 21.92 years, range: 18–25 years) were recruited to participate in this study from Beijing Normal University, China. They majored in a wide range of disciplines except for mathematics. These participants reported having no history of neurological disorders or head injuries. The experiment was fully explained to the participants before informed consent was obtained. This study was approved by the Institutional Review Board of the State Key Laboratory of Cognitive Neuroscience and Learning at Beijing Normal University.
Materials
Each problem of the algebraic and arithmetic tasks included two expressions of the twostep operation for addition, subtraction, multiplication, and division. Two expressions were presented synchronously in white against a black background (with a red, green, and blue value of 0, 0, 0), one at the top of the screen and one at the bottom of the screen (Fig. 1). Participants were asked to judge whether the results of the two expressions were equal. The problems of the algebraic task were presented in lowercase letters (i.e., a, b, c) and operators (i.e., + , −, × , ÷). All equal problems were selected from the operational principles acquired by students in primary school, mainly including the associative law, commutative law, and distributive law. Because there were few twostep operational principles and the difficulty level need to be matched, the algebraic task only contained 16 problems. Half of the problems were equal problems, and each expression has been used in one equal problem and one unequal problem. The previous fMRI studies of mathematical processing have suggested that using 16 trials or less can yield stable and reliable results [40, 43]. To match the algebraic task, we designed the same number of problems for the arithmetic task. All problems were presented in digits (i.e., 1–9) and operators (i.e., + , −, × , ÷), which required judging whether two expressions were equal by numerical calculation. The average length of the problems was controlled between two tasks, and all problems of the algebraic and arithmetic tasks can be seen in Appendix.
The stimulus presentation and behavioral data recordings were programmed using Eprime software (Version 1.1, Psychology Software Tools, Inc., www.pstnet.com) on a Pentium 4 laptop. Stimuli were projected onto a translucent screen placed at the back of the magnet bore. Participants viewed the screen through a mirror mounted on the head coil at 30 cm from the eyes.
Procedure
Prior to scanning, the participants underwent a training session with the same type of materials as a formal experiment to ensure that they understood the instructions of this experiment. After that, participants were required to complete the experimental tasks in the scanner. The experiment used a block design because it has high efficiency, which can increase the statistical power than eventrelated design of similar timing [44,45,46].
The scanning session was organized into two runs, each lasting 192 s. Each run consisted of four experimental blocks (two experimental blocks for each task) and four baseline blocks (Fig. 1). Thus, there were a total of four blocks for each condition. Each block with four trials lasted 24 s, and each trial lasted 6 s. The balanced Latin square design [47] was used to counterbalance the order effect of the two experimental tasks, and the presentation order of trials was random. There was a 1 min rest period after each run.
For each trial in the experimental blocks, two expressions were presented synchronously on the screen, one at the top and one at the bottom. Participants were asked to judge whether the two expressions were equal. For each trial in the baseline blocks, two arrows were presented synchronously on the screen, one at the top and one at the bottom. Participants were asked to judge whether the two arrows pointed in the same direction. The arrow judgement was the baseline to remove consistent or inconsistent judgements and the left or right finger motor response, which are the common judgement and response patterns across all the conditions. Each trial lasted 6 s, and a fixation cross was displayed to fill the remaining time if the participant did not use all 6 s. Half of the participants responded to the trials by pressing a key on a response box on their left with their left index finger when the two mathematical expressions or arrows were equal. The remaining participants responded to the trials by pressing a key on a response box on their right with their right index finger. Both accuracy and speed were emphasized.
fMRI data acquisition
Images were obtained using a Siemens (Munich, Germany) 3T Trio MRI scanner using a standard eightchannel head coil. After automatic shimming of the magnetic field, threedimensional highresolution T1 anatomical images were acquired for coregistration with the functional images. Next, functional volumes were acquired using a multiple slice T2*weighted echo planar imaging (EPI) sequence with the following parameters: inplane resolution = 3.125 × 3.125 mm^{2}; repetition time = 2000 ms; echo time = 30 ms; flip angle = 90°; field of view = 240 × 240 mm^{2}; matrix dimensions = 64 × 64; field of view = 200 mm; and slice thickness = 4 mm. The entire brain was imaged in 32 slices.
Statistical analysis of the fMRI data
Individual MRI datasets were analyzed using SPM12 software (Wellcome Department of Imaging Neurosciences, University College London, UK; http://www.fil.ion.ucl.ac.uk/spm). All volumes were realigned to the first volume and spatially normalized to a common value to correct for whole brain differences over time. Images were then smoothed using an isotropic Gaussian kernel of 4 mm and highpass filtered at a cutoff of 128 s.
Brain activation analysis
After preprocessing, parameterestimated images for individual participants across the whole brain were calculated. Group analyses with random effects were conducted using the univariate mass analysis of variance (ANOVA) on the brain activation maps of all participants with material type as the independent variable, the separate ANOVA was calculated for each voxel of the brain map. Then, the brain activation for each type of material relative to fixation was calculated. Brain activations for the two types of materials were compared. We used the stringent threshold at clusterlevel p < 0.05 false discovery rate (FDR)corrected for multiple comparisons at the wholebrain level, with underlying voxellevel p < 0.001 uncorrected. We also reported the lenient threshold at voxellevel p < 0.05 uncorrected.
Brainbehavior correlation analysis
The brainbehavior correlation analyses are the individual differences analyses, which are different from the withinsubject analyses by contrasting different experimental conditions. The withinsubject analyses are most sensitive to processes that are consistent across all participants, including those that are necessary for performing a task. In contrast, individual differences analyses of the brainbehavior correlation are most sensitive to processes that can be recruited in varying degrees to produce incremental gains in performance [48]. Although there was no withinsubject difference of behavioral results between algebraic and arithmetic tasks, the brainbehavior correlation analyses can also be used based on the individual differences within each task. Using the withinsubject analyses by contrast brain activation of different experimental conditions and individual differences analyses by brainbehavior correlation together can help provide an accurate and representative neural mechanism of cognitive processing [48, 49].
Previous studies have proven that singletrial (itemwise) analyses also applied to the data of block design [50, 51]. Thus, we first use the singletrial (itemwise) interindividual correlation analysis that has been reported in previous fMRI and eventrelated potential (ERP) studies that used to examine the brainbehavior correlations [25, 42]. First, the correlation between the brain activation maps and the reaction times (RTs) was determined for each trial. Then, a onesample ttest on the correlation coefficients obtained for all trials of one type of processing against zero was performed. The analysis was performed separately for algebra and arithmetic tasks. The singletrial correlation has been reported as more effective than the traditional meantrial correlation as it can filter out much of the noise that exists after the first step [42].
The traditional meantrial interindividual correlation was also used, and the results were compared with those of the singletrial correlation. The mean brain activation map and the mean RT for each type of processing for each participant were determined. Then, a correlation analysis between the mean brain activation map and the RT for all participants was performed. To compare the results of the singletrial and meantrial correlation analyses, we used the stringent threshold at clusterlevel p < 0.05 FDRcorrected with underlying voxellevel p < 0.001 uncorrected, and the lenient threshold at voxellevel p < 0.05 uncorrected simultaneously.
Further, we defined seven functional regions of interest (ROIs) based on results from previous studies for the semantic, phonological, and visuospatial networks. Based on the metaanalysis of 120 functional neuroimaging studies of semantic processing [39], we defined four ROIs of the semantic network in the left hemisphere that included the middle temporal gyrus (Montreal Neurological Institute (MNI) coordinates [− 45, − 21, − 15]), inferior frontal gyrus (MNI coordinates [− 51, 24, 0]), dorsomedial prefrontal cortex (MNI coordinates [− 6, − 48, 30]), and angular gyrus (MNI coordinates [− 57, − 45, 30]). Based on the fMRI studies of phonological processing for arithmetic [27], we defined two ROIs of the phonological network in the left hemisphere that included the precentral gyrus (MNI coordinates [− 45, 4, 30]) and supplementary motor area (MNI coordinates [− 3, 3, 58]). Based on the fMRI studies of visuospatial processing for arithmetic [52], we defined one ROI of the visuospatial network in the left superior parietal lobule (MNI coordinates [− 32, − 68, 56]). Each ROI was a sphere with a radius of 9 mm. These ROIs were used to compare the levels of brainbehavior correlations between algebra and arithmetic.
Functional connectivity analysis
The functional connectivity was analyzed by a seedbased ROItoROI analysis in the CONN toolbox using SPM preprocessed data [53]. The ROIs were consistent with the ROIs of the brainbehavior correlation analysis, including four ROIs of the semantic network (i.e., middle temporal gyrus, inferior frontal gyrus, dorsomedial prefrontal cortex, and angular gyrus), two ROIs of the phonological network (i.e., precentral gyrus and supplementary motor area), and one ROI of the visuospatial network (i.e., superior parietal lobule) in the left hemisphere. In the firstlevel analyses, Pearson’s correlation coefficients were calculated between the time courses of each pair of ROIs for each subject, and these correlation coefficients were converted to normally distributed zscores using Fisher’s transform. In the secondlevel analysis, the onesample ttest on the zscores against zero was performed separately for algebra and arithmetic tasks. We used the significant threshold at p < 0.05 FDRcorrected.
Results
Behavioral results
The mean RT was 2653 ms for algebraic task and 2855 ms for arithmetic task. The mean accuracy was 87.6% and 86.8%, respectively. RTs and accuracy were analyzed with a repeated measure ANOVA. The main effect of stimulus type was not significant for RTs (F (1, 29) = 2.55, p = 0.116, η^{2} = 0.042) or accuracy (F (1, 29) = 1.10, p = 0.754, η^{2} = 0.002).
Brain activations
The brain activations for algebra and arithmetic relative to baseline are displayed in Fig. 2. Algebra and arithmetic elicited significant activation in a broad array of brain regions, including the bilateral parietal, frontal, and occipital gyri. The differences in brain activation between algebra and arithmetic are shown in Fig. 2 and Table 1. Algebra had greater activation in the bilateral angular gyrus relative to arithmetic. In contrast, arithmetic elicited greater activation in the bilateral supplementary motor areas, left insula, and left inferior parietal lobule than algebra.
Brainbehavior correlations
The singletrial brainbehavior correlations for algebra and arithmetic are displayed in Fig. 3 and Table 2. At the threshold of clusterlevel p < 0.05 FDRcorrected with underlying voxellevel p < 0.001 uncorrected, significant positive brainbehavior correlations were observed in the bilateral middle temporal gyri, bilateral inferior frontal gyri, bilateral dorsomedial prefrontal cortices, and left angular gyrus for algebra. Significant negative brainbehavior correlations were observed in the bilateral superior parietal lobules for algebra. For arithmetic, significant positive brainbehavior correlations were observed in the right inferior temporal gyrus and negative brainbehavior correlations were observed in the bilateral superior parietal lobules, bilateral middle occipital gyri, bilateral precentral gyri, left supplementary motor area, left insula, and bilateral cerebella. Figure 3 shows the number of voxels of the singletrial brainbehavior correlations for algebra and arithmetic in the ROIs under the threshold of voxellevel p < 0.05 uncorrected. Algebra elicited a larger number of voxels in four ROIs of the semantic network, including the middle temporal gyrus, inferior frontal gyrus, dorsomedial prefrontal cortex, and angular gyrus. Arithmetic elicited a greater brainbehavior correlation and a larger number of voxels in three ROIs of the phonological and visuospatial networks, including the precentral gyrus, supplementary motor area, and superior parietal lobule.
The traditional meantrial brainbehavior correlations for algebra and arithmetic are displayed in Fig. 4. Only at the lenient threshold of voxellevel p < 0.05 uncorrected, significant positive brainbehavior correlations were observed in the bilateral middle temporal gyri and bilateral dorsomedial prefrontal cortices, and negative brainbehavior correlations were observed in the bilateral superior parietal lobules for algebra. Significant positive brainbehavior correlations were observed in the bilateral inferior temporal gyri, and negative brainbehavior correlations were observed in the bilateral superior parietal lobules, bilateral middle occipital gyri, bilateral precentral gyri, and bilateral cerebella for arithmetic. Figure 4 shows the number of voxels of the meantrial brainbehavior correlations for algebra and arithmetic for each ROI under the threshold of voxellevel p < 0.05 uncorrected.
Functional connectivity
The functional connectivity between seven ROIs of the semantic, phonological, and visuospatial networks for algebra and arithmetic are shown in Fig. 5 and Table 3. At the threshold of p < 0.05 FDRcorrected, the positive functional connectivity within the semantic network (i.e., middle temporal gyrus, inferior frontal gyrus, dorsomedial prefrontal cortex, and angular gyrus) and phonological network (i.e., precentral gyrus and supplementary motor area) were observed for algebra and arithmetic, respectively. In addition, for algebra, the significant positive functional connectivity was observed between the visuospatial network (i.e., superior parietal lobule) and semantic network (i.e., inferior frontal gyrus and dorsomedial prefrontal cortex). For arithmetic, the significant positive functional connectivity was only observed between the visuospatial network (i.e., superior parietal lobule) and phonological network (i.e., precentral gyrus).
Discussion
The present study aimed to clarify the neural dissociation between algebra and arithmetic. Betweengroup brain activation analyses revealed that algebra resulted in greater activation in the bilateral angular gyrus, and arithmetic elicited greater activations in the bilateral supplementary motor area, left insula, and left inferior parietal lobule. Interindividual singletrial brainbehavior correlation analyses showed significantly positive brainbehavior correlations in the semantic network, including the bilateral middle temporal gyri, bilateral inferior frontal gyri, bilateral dorsomedial prefrontal cortices, and left angular gyrus for algebra. For arithmetic, the significantly negative brainbehavior correlations were observed in the phonological network, including the bilateral precentral gyrus, and left supplementary motor area, and in the visuospatial network, including the bilateral superior parietal lobules. For algebra, there was significantly positive functional connectivity between the visuospatial network and semantic network. For arithmetic, there was significantly positive functional connectivity between the visuospatial network and phonological network. These findings suggest that algebra relies more on the semantic network and arithmetic relies more on the phonological and visuospatial networks.
Semantic network in algebra
The current study found that algebra relies more on the semantic network, including the bilateral angular gyrus, bilateral middle temporal gyri, bilateral inferior frontal gyri, bilateral dorsomedial prefrontal cortices, and left angular gyrus, than arithmetic. These regions overlap with the semantic network according to the metaanalysis of functional neuroimaging study by Binder et al. that was based on word processing [39]. Previous studies have revealed that the angular gyrus was critical for conceptual combination independent of the modality of the semantic content integrated [54,55,56], which indicated that the left angular gyrus was sensitive to compositional behavior of twoword phrases or tone sequences. For algebra processing, components including letters and operators were integrated to algebraic expressions similar to pure syntactic representation. The composition of operation and integration of algebra involved the conceptual combination. It elicited the activation the angular gyrus is to support the representation of integrated semantic information.
In algebra, formula derivations of the relationship between the letter and the operator (such as the commutative or associative laws) require knowledge of mathematical principles. Thus, algebra is similar to mathematical principles. Previous studies have shown that the mathematical principle processing based on conceptual knowledge is subserved by a semantic network [12, 40, 51, 57]. Liu et al. [51] reported that mathematical principles (e.g., exchanging the position of two operands in addition does not change their sum) elicited greater activation of the semantic network in the left middle temporal gyrus, left inferior frontal gyrus, and left angular gyrus than arithmetic computation (e.g., when the number 8 is first divided by 4 and then multiplied by 3, the final result is 6). The key role of the semantic network in mathematical principles suggests an important role of conceptual knowledge in algebra.
Furthermore, algebra requires abstract conceptual knowledge, and arithmetic involves concrete conceptual knowledge. Recent studies have suggested that concepts are organized in the brain according to semantic categories [58] such as animate versus inanimate [59] or concrete versus abstract [60] and that the classical semantic network is important for abstract conceptual knowledge [61,62,63]. A metaanalysis of studies regarding abstract concepts revealed that abstract concepts tend to produce stronger activation in the semantic network than concrete concepts [63]. By contrast, previous studies emphasized that abstract algebra concepts induced the nonlinguistic cortical network dissociated from the classical semantic network [32, 33]. The subjective representation of the concreteness of algebra by expert mathematicians relied more on mental imagery, which differs from nonmathematical concepts in these studies [32, 33]. However, for laymen, algebra relies on the classical semantic network, suggesting a close association between algebra and language processing.
Phonological and visuospatial networks in arithmetic
The present study found that arithmetic relied more on the phonological network, including the bilateral precentral gyri and supplementary motor areas, than algebra. It is hypothesized that arithmetic problems are solved by fact retrieval, in which the facts are stored as verbal codes [11, 24, 29], In this regard, arithmetic is related to phonological processing ability and elicits activity in the bilateral precentral gyri that supports phonological processing. On the other hand, arithmetic has greater demand on the phonological working memory, which supports the temporary storage and updating of intermediate results in the multistep arithmetic [64]. These results are in line with models of arithmetic operations that outline a phonological processing network [25, 28, 65].
Our results also indicate that arithmetic results in greater activations in the visuospatial networks including the bilateral superior parietal lobules, which is consistent with the previously reported role of the visuospatial network in arithmetic [10,11,12,13,14, 28]. These regions are responsible for visuospatial information from the symbolic form of the arithmetic expression and reflect the visuospatial processing of concrete symbolic and quantitative representations in arithmetic. Numbers can elicit a visuospatial response, as the quantity of a number can be represented on a mental number line from left to right. This processing is evidenced by the spatial numerical association of response codes effect, which describes that people respond faster to small numbers with the left hand and large numbers with the right hand [66, 67]. Arithmetic can be completed using the visuospatial processing of a dynamic mental number line, as addition leads to a right shift and subtraction leads to a left shift in visuospatial processing [68,69,70].
Significance of the singletrial brainbehavior correlation
Neuroimaging researchers often use interindividual brainbehavior correlations to explore the associations between the brain and human behaviors [71,72,73]. The correlation analysis investigates how interindividual differences in brain functions relate to interindividual behavioral performances and can help provide an accurate and representative neural mechanism of cognitive processing [49].
In this study, the traditional meantrial brainbehavior correlations approach yielded correlations with p < 0.05, which is not considered significant in fMRI studies. The meantrial correlation approach may not efficiently remove enough of the noise and does not result in a high correlation coefficient [74]. To improve the disadvantages of the traditional meantrial correlation approach, we used the singletrial brainbehavior correlation approach previously reported in fMRI and ERP studies [42, 72]. The singletrial brainbehavior correlation approach avoids the influence of noise on the statistical analyses. This correlation approach first removes much of the noise (or residual activity) when determining the correlation for each trial and then conducts a ttest for the weak yet possibly stable correlation coefficients with the majority of the noise filtered. The results of this study showed significantly positive brainbehavior correlations in the semantic network for algebra and significantly negative brainbehavior correlations in the phonological and visuospatial networks for arithmetic (clusterlevel p < 0.05 FDRcorrected with underlying voxellevel p < 0.001 uncorrected). In algebra processing, letters and operators were integrated to semantic concept combination. Brain activation increase in semantic network with longer response times because more neurons responding to conceptual combination are recruited for the successful performance of the task [75]. However, arithmetic processing relied more on procedural knowledge that was similar to automatic processing [76]. Brain activation decreases with longer response times because of more efficient use of brain pathways.
Conclusion
Algebra is the abstract form of arithmetic and activates the semantic network, including the bilateral middle temporal gyri, bilateral inferior frontal gyri, bilateral dorsomedial prefrontal cortices, and left angular gyrus. In contrast, arithmetic activates the phonological network, including the bilateral precentral gyri and supplementary motor areas, and the visuospatial network, including the bilateral superior parietal lobules. These results suggest that the semantic network in the brain supports algebra. It deepened our understanding of the relationship between algebraic processing and arithmetic processing. Whereas previous studies suggested that algebraic processing activated the nonsemantic cortical network in expert mathematicians [32, 33], our results highlighted the important role of the classical semantic network in algebraic processing. These findings have strong implications for education and teaching. In schoollevel education in general, algebra learning might take advantage of semantic knowledge.
Availability of data and materials
The data are currently not publicly available due to participant privacy, but they are available from the corresponding author upon reasonable request. Every request will be reviewed by the institutional review board of the State Key Laboratory of Cognitive Neuroscience and Learning at Beijing Normal University.
Abbreviations
 IPS:

Intraparietal sulcus
 fMRI:

Functional magnetic resonance imaging
 EPI:

Echo planar imaging
 ANOVA:

Analysis of variance
 ERP:

Eventrelated potential
 RTs:

Reaction times
 MNI:

Montreal Neurological Institute
 ROI:

Region of interest
References
 1.
Tolar TD, Lederberg AR, Fletcher JM. A structural model of algebra achievement: Computational fluency and spatial visualization as mediators of the effect of working memory on algebra achievement. Educ Psychol. 2009;29(2):239–66.
 2.
Bryant P, Christie C, Rendu A. Children’s understanding of the relation between addition and subtraction: Inversion, identity, and decomposition. J Exp Child Psychol. 1999;74(3):194–212.
 3.
Cappelletti M, Kopelman M, Morton J, Butterworth B. Dissociations in numerical abilities revealed by progressive cognitive decline in a patient with semantic dementia. Cogn Neuropsychol. 2005;22:771–93.
 4.
Dehaene S, Cohen L. Cerebral pathways for calculation: double dissociation between rote verbal and quantitative knowledge of arithmetic. Cortex. 1997;33(2):219–50.
 5.
Pesenti M, Depoorter N, Seron X. Noncommutability of the N + 0 arithmetical rule: a case study of dissociated impairment. Cortex. 2000;36:445–54.
 6.
Rasmussen C, Ho E, Bisanz J. Use of the mathematical principle of inversion in young children. J Exp Child Psychol. 2003;85:89–102.
 7.
Sherman J, Bisanz J. Evidence for use of mathematical inversion by threeyearold children. J Cogn Dev. 2007;8:333–44.
 8.
Delazer M, Domahs F, Lochy A, Karner E, Benke T, Poewe W. Number processing and basal ganglia dysfunction: a single case study. Neuropsychologia. 2004;42(8):1050–62.
 9.
Ernest P. A model of the cognitive meaning of mathematical expressions. Br J Educ Psychol. 1987;57:343–70.
 10.
Nakai T, Okanoya K. neural evidence of crossdomain structural interaction between language and arithmetic. Sci Rep. 2018. https://doi.org/10.1038/s41598018312798.
 11.
Dehaene S, Piazza M, Pinel P, Cohen L. Three parietal circuits for number processing. Cogn Neuropsychol. 2003;20:487–506.
 12.
Zhang H, Chen C, Zhou X. Neural correlates of numbers and mathematical terms. Neuroimage. 2012;60(1):230–40.
 13.
Thioux M, Pesenti M, Costes N, De Volder A, Seron X. Taskindependent semantic activation for numbers and animals. Brain Res Cogn Brain Res. 2005;24:284–90.
 14.
Arsalidou M, Taylor MJ. Is 2+2=4? Metaanalyses of brain areas needed for numbers and calculations. Neuroimage. 2011;54(3):2382–93.
 15.
Andin J, Fransson P, Dahlstrom O, Ronnberg J, Rudner M. The neural basis of arithmetic and phonology in deaf signing individuals. Lang Cogn Neurosci. 2019;34(7):813–25.
 16.
Bugden S, Price GR, McLean DA, Ansari D. The role of the left intraparietal sulcus in the relationship between symbolic number processing and children’s arithmetic competence. Dev Cogn Neurosci. 2012;2(4):448–57.
 17.
Bugden S, Woldorff MG, Brannon EM. Shared and distinct neural circuitry for nonsymbolic and symbolic doubledigit addition. Hum Brain Mapp. 2019;40(4):1328–43.
 18.
Kuhl U, Friederici AD, Friederici AD, Emmrich F, Brauer J, Wilcke A, et al. Early cortical surface plasticity relates to basic mathematical learning. Neuroimage. 2020;204:116235.
 19.
Ashkenazi S, Henik A, Ifergane G, Shelef I. Basic numerical processing in left intraparietal sulcus (IPS) acalculia. Cortex. 2008;44(4):439–48.
 20.
Baldo JV, Dronkers NF. Neural correlates of arithmetic and language comprehension: a common substrate? Neuropsychologia. 2007;45(2):229–35.
 21.
Sigmundsson H, Anholt SK, Talcott JB. Are poor mathematics skills associated with visual deficits in temporal processing? Neurosci Lett. 2010;469(2):248–50.
 22.
Zhou XL, Li MY, Li LNA, Zhang YY, Cui JX, Liu J, et al. The semantic system is involved in mathematical problem solving. Neuroimage. 2018;166:360–70.
 23.
Frst AJ, Hitch GJ. Separate roles for executive and phonological components of working memory in mental arithmetic. Memory Cognition. 2000;28(5):774–82.
 24.
Pollack C, Ashby NC. Where arithmetic and phonology meet: the metaanalytic convergence of arithmetic and phonological processing in the brain. Dev Cogn Neurosci. 2018;30:251–64.
 25.
Li MY, Tan YX, Cui JX, Chen CS, Dong Q, Zhou XL. The semantic network supports approximate computation. Neuropsychology. 2019;33(6):842–54.
 26.
Wu TH, Chen CL, Huang YH, Liu RS, Lee HJ. Effects of longterm practice and task complexity on brain activities when performing abacusbased mental calculations: a PET study. Eur J Nucl Med Mol Imaging. 2009. https://doi.org/10.1007/s0025900809490.
 27.
Zhou X, Chen C, Zang Y, Dong Q, Chen C, Qiao S, et al. Dissociated brain organization for singledigit addition and multiplication. Neuroimage. 2007;35(2):871–80.
 28.
Cheng DZ, Xiao Q, Chen Q, Cui JX, Zhou XL. Dyslexia and dyscalculia are characterized by common visual perception deficits. Dev Neuropsychol. 2018;43(6):497–507.
 29.
Cheng D, Wu H, Yuan L, Xu R, Chen Q, Zhou X. Modalitydependent or modalityindependent processing in mental arithmetic: evidence from unimpaired auditory multiplication for a patient with left frontotemporal stroke. J Int Neuropsychol Soc. 2017;23(8):692–9.
 30.
Tohgi H, Saitoh K, Takahashi S, Takahashi H, Utsugisawa K, Yonezawa H, et al. Agraphia and acalculia after a left prefrontal (F1, F2) infarction. J Neurol Neurosurg Psychiatry. 1995;58:629–32.
 31.
McCloskey M, Caramazza A, Basili A. Cognitive mechanisms in number processing and calculation: evidence from dyscalculia. Brain Cogn. 1985;4:171–96.
 32.
Amalric M, Dehaene S. Origins of the brain networks for advanced mathematics in expert mathematicians. Proc Natl Acad Sci USA. 2016;113(18):4909–17.
 33.
Amalric M, Dehaene S. A distinct cortical network for mathematical knowledge in the human brain. Neuroimage. 2019;189:19–31.
 34.
Monti MM, Parsons LM, Osherson DN. Thought beyond language: neural dissociation of algebra and natural language. Psychol Sci. 2012;23(8):914–22.
 35.
Cappelletti M, Butterworth B, Kopelman M. Spared numerical abilities in a case of semantic dementia. Neuropsychologia. 2001;39:1224–39.
 36.
Cheng D, Zhou A, Yu X, Chen C, Jia J, Zhou X. Quantifier processing can be dissociated from numerical processing: evidence from semantic dementia patients. Neuropsychologia. 2013;51(11):2172–83.
 37.
Anderson JR, Qin YL, Sohn MH, Stenger VA, Carter CS. An informationprocessing model of the BOLD response in symbol manipulation tasks. Psychon Bull Rev. 2003;10(2):241–61.
 38.
Geary DC, Hoard MK, Nugent L, Rouder JN. Individual differences in algebraic cognition: relation to the approximate number and semantic memory systems. J Exp Child Psychol. 2015;140:211–27.
 39.
Binder JR, Desai RH, Graves WW, Conant LL. Where is the semantic system? A critical review and metaanalysis of 120 functional neuroimaging studies. Cereb Cortex. 2009;19(12):2767–96.
 40.
Liu J, Yuan L, Chen CS, Cui JX, Zhang H, Zhou XL. The semantic system supports the processing of mathematical principles. Neuroscience. 2019;404:102–18.
 41.
Zhou X, Zeng, JY. Threecomponent mathematics for students. Infant Child Dev. 2021. https://doi.org/10.1002/icd.2283.
 42.
Zhou XL, Li MY, Zhou HT, Li LN, Cui JX. Itemwise interindividual brainbehavior correlation in task neuroimaging analysis. Front Neurosci. 2018;12:17.
 43.
SuárezPellicioni M, Booth JR. Fluency in symbolic arithmetic refines the approximate number system in parietal cortex. Hum Brain Mapp. 2018;39(10):3956–71.
 44.
Amaro E Jr, Barker GJ. Study design in fMRI: basic principles. Brain Cogn. 2006;60(3):220–32.
 45.
Friston KJ, Zarahn E, Josephs O, Henson RN, Dale AM. Stochastic designs in eventrelated fMRI. Neuroimage. 1999;10(5):607–19.
 46.
Miezin FM, Maccotta L, Ollinger JM, Petersen SE, Buckner RL. Characterizing the hemodynamic response: effects of presentation rate, sampling procedure, and the possibility of ordering brain activity based on relative timing. Neuroimage. 2000;11(6):735–59.
 47.
Bradley JV. Complete counterbalancing of immediate sequential effects in a latin square design. J Am Stat Assoc. 1958;53(282):525–8.
 48.
Yarkoni T, Braver T. Cognitive neuroscience approaches to individual differences in working memory and executive control: conceptual and methodological issues. 2010. p. 87–107.
 49.
Lebreton M, Bavard S, Daunizeau J, Palminteri S. Assessing interindividual differences with taskrelated functional neuroimaging. Nat Hum Behav. 2019;3(9):897–905.
 50.
DodellFeder D, KosterHale J, Bedny M, Saxe R. fMRI item analysis in a theory of mind task. Neuroimage. 2011;55(2):705–12.
 51.
Liu J, Zhang H, Chen C, Chen H, Cui J, Zhou X. The neural circuits for arithmetic principles. Neuroimage. 2017;147:432–46.
 52.
Dehaene S, Spelke E, Pinel P, Stanescu R, Tsivkin S. Sources of mathematical thinking: behavioral and brainimaging evidence. Science. 1999;284(5416):970–4.
 53.
WhitfieldGabrieli S, NietoCastanon A. Conn: a functional connectivity toolbox for correlated and anticorrelated brain networks. Brain Connectivity. 2012; 2(3): 12541.
 54.
Boylan C, Trueswell JC, ThompsonSchill SL. Compositionality and the angular gyrus: a multivoxel similarity analysis of the semantic composition of nouns and verbs. Neuropsychologia. 2015;78:130–41.
 55.
Lumaca M, Vuust P, Baggio G. Network analysis of human brain connectivity reveals neural fingerprints of a compositionality bias in signaling systems. Cereb Cortex. 2021. https://doi.org/10.1093/cercor/bhab307.
 56.
Price AR, Bonner MF, Peelle JE, Grossman M. Converging evidence for the neuroanatomic basis of combinatorial semantics in the angular gyrus. J Neurosci. 2015;35(7):3276–84.
 57.
Saxe AM, McClelland JL, Ganguli S. A mathematical theory of semantic development in deep neural networks. Proc Natl Acad Sci USA. 2019;116(23):11537–46.
 58.
Huth AG, de Heer WA, Griffiths TL, Theunissen FE, Gallant JL. Natural speech reveals the semantic maps that tile human cerebral cortex. Nature. 2016;532(7600):453.
 59.
Caramazza A, Shelton JR. Domainspecific knowledge systems in the brain: the animateinanimate distinction. J Cogn Neurosci. 1998;10(1):1–34.
 60.
Binder JR, Westbury CF, McKiernan KA, Possing ET, Medler DA. Distinct brain systems for processing concrete and abstract concepts. J Cogn Neurosci. 2005;17(6):905–17.
 61.
Hoffman P, Binney RJ, Ralph MAL. Differing contributions of inferior prefrontal and anterior temporal cortex to concrete and abstract conceptual knowledge. Cortex. 2015;63:250–66.
 62.
SkipperKallal LM, Mirman D, Olson IR. Converging evidence from fMRI and aphasia that the left temporoparietal cortex has an essential role in representing abstract semantic knowledge. Cortex. 2015;69:104–20.
 63.
Wang J, Conder JA, Blitzer DN, Shinkareva SV. Neural representation of abstract and concrete concepts: a metaanalysis of neuroimaging studies. Hum Brain Mapp. 2010;31(10):1459–68.
 64.
Ding Y, Liu RD, Liu H, Wang J, Zhen R, Jiang R. Effects of working memory, strategy use, and singlestep mental addition on multistep mental addition in chinese elementary students. Front Psychol. 2019. https://doi.org/10.3389/fpsyg.2019.00148.
 65.
StanescuCosson R, Pinel P, van de Moortele PF, Le Bihan D, Cohen L, Dehaene S. Understanding dissociations in dyscalculia—a brain imaging study of the impact of number size on the cerebral networks for exact and approximate calculation. Brain. 2000;123:2240–55.
 66.
Dehaene S, Bossini S, Giraux P. The mental representation of parity and number magnitude. J Exp Psychol. 1993;122:371–96.
 67.
Viarouge A, Hubbard EM, Dehaene S. The organization of spatial reference frames involved in the SNARC effect. Q J Exp Psychol. 2014;67(8):1484–99.
 68.
Knops A, Thirion B, Hubbard EM, Michel V, Dehaene S. Recruitment of an area involved in eye movements during mental arithmetic. Science. 2009;324(5934):1583–5.
 69.
McCrink K, Dehaene S, DehaeneLambertz G. Moving along the number line: operational momentum in nonsymbolic arithmetic. Percept Psychophys. 2007;69(8):1324–33.
 70.
Pinhas M, Fischer M. Mental movements without magnitude? A study of spatial biases in symbolic arithmetic. Cognition. 2008;109:408–15.
 71.
Beaty RE, Kenett YN, Christensen AP, Rosenberg MD, Benedek M, Chen Q, et al. Robust prediction of individual creative ability from brain functional connectivity. Proc Natl Acad Sci USA. 2018;115(5):1087–92.
 72.
Li M, Cheng D, Lu Y, Zhou X. Neural association between nonverbal number sense and arithmetic fluency. Hum Brain Mapp. 2020;41(18):5128–40.
 73.
Sommerauer G, Grass KH, Grabner RH, Vogel SE. The semantic control network mediates the relationship between symbolic numerical order processing and arithmetic performance in children. Neuropsychologia. 2020. https://doi.org/10.1016/j.neuropsychologia.2020.107405.
 74.
Meyer GJ, Finn SE, Eyde LD, Kay GG, Moreland KL, Dies RR, et al. Psychological testing and psychological assessment—a review of evidence and issues. Am Psychol. 2001;56(2):128–65.
 75.
Mohamed MA, Yousem DM, Tekes A, Browner N, Calhoun VD. Correlation between the amplitude of cortical activation and reaction time: a functional MRI study. AJR Am J Roentgenol. 2004;183(3):759–65.
 76.
Mason RA, Just MA. Neural representations of procedural knowledge. Psychol Sci. 2020;31(6):729–40.
Acknowledgements
Not applicable.
Funding
This research was supported by grants from the Natural Science Foundation of Beijing (Grant Number: 5212004), National Natural Science Foundation of China (Grant Number: 31700977, 31671151, 31600896), the 111 Project (Grant Number: BP0719032), and the Advanced Innovation Center for Future Education (Grant Number: 27900110631111).
Author information
Affiliations
Contributions
DZ, MY and XL conceived and designed the study. DZ, MY and LY performed the experiment and analyzed the data. DZ and MY interpreted the data and drafted the manuscript. JX, LW, NY, XZ, XJ and XL critically revised the manuscript for important intellectual content. All authors read and approved the final manuscript
Corresponding author
Ethics declarations
Ethics approval and consent to participate
This study was approved by the Institutional Review Board of the Institute of Cognitive Neuroscience and Learning at Beijing Normal University. Consent to participate was obtained from each participant.
Consent for publication
Not applicable.
Competing interests
The authors declare that they have no competing interests.
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Appendix
Appendix
Materials used in the present study.
Algebraic task  Arithmetic task  

Expression 1  Expression 2  Expression 1  Expression 2 
a−b−c  a−(b + c)  9–3−2  8−(1 + 3) 
a ÷ b × c  (a × c) ÷ b  8 ÷ 4 × 3  (9 × 4) ÷ 6 
a × (b + c)  a × b + a × c  2 × (6 + 2)  4 × 3 + 4 × 1 
a ÷ b ÷ c  a ÷ (b × c)  8 ÷ 2 ÷ 2  6 ÷ (3 × 1) 
a−b + c  a + (c−b)  9–6 + 3  8−(7−5) 
a + (b + c)  (a + c) + b  7−(4 + 2)  (8−4)−3 
a × (b × c)  (a × c) × b  8−(3 + 3)  9−(5 + 2) 
a × b ÷ c  a × (b ÷ c)  (4 ÷ 2) × 2  8 ÷ (4 ÷ 2) 
a(b + c)  a ÷ b × c  8−(1 + 3)  8 ÷ 4 × 3 
a × b + a × c  a−b + c  4 × 3 + 4 × 1  9−6 + 3 
a × (b ÷ c)  a−b−c  (9 × 4) ÷ 6  (8−4)−3 
a + (c−b)  a × (b + c)  8−(7−5)  2 × (6 + 2) 
(a × c) × b  a ÷ b ÷ c  9−(5 + 2)  8 ÷ (4 ÷ 2) 
a ÷ (b × c)  a × (b × c)  (4 ÷ 2) × 2  8−(3 + 3) 
(a × c) ÷ b  (a + c) + b  8 ÷ 2 ÷ 2  9−3−2 
a + (b + c)  a × b ÷ c  7−(4 + 2)  6 ÷ (3 × 1) 
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.
About this article
Cite this article
Cheng, D., Li, M., Cui, J. et al. Algebra dissociates from arithmetic in the brain semantic network. Behav Brain Funct 18, 1 (2022). https://doi.org/10.1186/s12993022001864
Received:
Accepted:
Published:
DOI: https://doi.org/10.1186/s12993022001864
Keywords
 Algebra
 Arithmetic
 Semantic network
 Phonological network
 Visuospatial network