Math skills entail adding, subtracting, multiplying, and dividing symbolic numbers efficiently, and are developed through childhood and adolescence (Siegler et al., 2011; Siegler and Braithwaite, 2017). These skills constitute an essential element of the K-12 school curriculum and have been positively associated with academic achievement (Gómez-Velázquez et al., 2015). Thus, failure to adequately develop such skills may have deleterious effects on an individual’s daily life.

Arithmetic fact processing involves basic and complex cognitive processes and develops over time (e.g., Geary et al., 2008; Geary, 2010; Meyer et al., 2010). Multiple studies have identified the basic cognitive processes which predict math skills, noting that cognitive inhibition, set-shifting (van der Sluis et al., 2004; Wei et al., 2018), working memory (Geary, 2010; Xenidou-Dervou et al., 2017; Wei et al., 2018), visuospatial skills, and executive functions (Kahl et al., 2021) explain children’s ability to solve arithmetic problems and incorporate arithmetic facts, suggesting that children require a higher order cognitive to reach certain math skills.

Recent studies have proposed a role for environmental factors, noting that sex (Hyde, 2016; Palomares-Ruiz and García-Perales, 2020), education (Palomares-Ruiz and García-Perales, 2020), or weight status also affect mathematical skill levels. Several studies have highlighted a negative relationship between body mass index (BMI) and math achievement scores (Crosnoe, 2004; Castelli et al., 2007; Cottrell et al., 2007; Manes, 2015). Further, longitudinal studies using standardized tests (e.g., Woodcock-McGrew-Werder mini-battery of achievement, Woodcock et al., 1994) reported a negative association between weight status and math scores (Datar et al., 2004; Gable et al., 2012; Kranjac, 2015; Li and O’Connelly, 2015; Martin et al., 2017). Kranjac (2015) reported that the math skills of children with obesity were independent of the level of self-efficacy (i.e., the ability to organize and execute behaviors needed to attain chosen goals, Bandura, 1997). Participants with obesity had lower scores by age than their normal-weight peers, indicating a direct negative effect of obesity on math skills.

Numerous studies have noted problems in executive functions such as inhibitory control (Mamrot and Hanć, 2019; Alatorre-Cruz et al., 2021), attention (Maayan et al., 2011; Blanco-Gómez et al., 2015; Ayoub-Meo et al., 2019), and working memory (WM, Maayan et al., 2011; Schwartz et al., 2013) in children with obesity. Neuroimaging studies have attributed these problems to decreased gray matter volume or density in multiple brain regions (Alosco et al., 2014; Borsini et al., 2015; Ou et al., 2015; Perlaki et al., 2018; Laurent et al., 2020; Ronan et al., 2020). Although the cognitive processes affected in children with obesity overlap those predicting math skills (van der Sluis et al., 2004; Geary, 2010; Xenidou-Dervou et al., 2017; Wei et al., 2018; Kahl et al., 2021), it is possible that there is an interaction between weight status and developmental age, (Siegler et al., 2011; Siegler and Braithwaite, 2017), which will affect the consolidation of cognitive processes associated with each arithmetic processing stage [i.e., encoding, math strategy, and verification (Campbell, 2005)].

Prior to consolidation of math skills, children must have an adequate association between mental representations and numerical magnitudes to calculate arithmetic facts [i.e., encoding stage (Dehaene and Cohen, 1995; Siegler et al., 2011; Siegler and Braithwaite, 2017; Malone et al., 2019)]. Moreover, there is a shift in the strategy used to solve problems in childhood, from a reliance on procedural strategies to a greater reliance on retrieval strategies (Cragg and Gilmore, 2014; Cragg et al., 2017). Most studies exploring arithmetical development have shown that strategy development is not an abrupt shift, but rather a change in the frequency of strategy choices used by children at different time points [overlapping waves theory (Siegler, 1996)], with some studies suggesting that the procedural strategy gradually evolves to one of retrieval (Geary, 2011; Peters and De Smedt, 2018). However, there is substantial evidence that strategy selection depends on both arithmetic development and the type of arithmetic problem (Zhou et al., 2006; Van Beek et al., 2014). For example, it is expected that children will solve addition or multiplication problems using a retrieval strategy, while subtraction will require procedural strategies (Zhou et al., 2006; Barrouillet et al., 2008; Van Beek et al., 2014). Development of an efficient strategy is reflected in the subject’s performance, particularly in response times (RTs) (Cragg et al., 2017; Proverbio and Caminati, 2019). For example, it has been suggested that young adults have a greater mastery of the resolution of arithmetical facts than children because young adults display shorter RTs, even when both age groups have the same response accuracy and trigger the same kind of math strategy (Cragg et al., 2017). We might suppose that when two children solving the same math problem differ in their RTs, they are using a different strategy or have different mastery levels using the same strategy.

Given that math strategy selection is not reflective of math skill level, event-related potential (ERP) studies have focused on specific components of mathematical problem solving such as the time at which the participant verifies a presented result, where that the individual’s math skill is reflected not only in the behavioral performance but also in the electrophysiological pattern. For example, Proverbio and Caminati (2019) reported that adults who were skilled math performers had shorter RTs, greater response accuracy and an electrophysiological pattern characterized by shorter latencies in a frontocentral positivity occurring between 300 and 500 ms (P300 component) than individuals with poorer math skills in arithmetic. Additionally, participants with poorer skills displayed larger P300 amplitude for incorrect than correct responses compared with the higher-performing participants. The P300 component is generally associated with attention allocation, and its amplitude is affected by increased task complexity (Polich, 2007). Proverbio and Caminati (2019) concluded that lower-performing participants had a poorer representation of arithmetic solutions and more difficulty retrieving that information because they were less able to reject incorrect solutions due to their posited solution strategy.

Based on the idea that adults have greater math skills than children, ERP studies comparing these age groups reveal that brain-electrical activity associated with arithmetic skills depends on age and the kind of arithmetical problem. Prieto-Corona et al. (2010) compared ERPs of children and adults during a multiplication task. Both age groups displayed a negativity between 250 and 450 ms in centro-parietal regions (arithmetic N400 component) and a late positive component (LPC) that appears between 500 and 675 ms. Group differences were observed in the arithmetic N400 component, with children displaying a larger amplitude and longer latency with a greater spatial distribution than adults. There is general agreement that arithmetic facts are stored in semantic networks (Kutas and Hillyard, 1980), with N400 amplitude modulations attributed to the relatedness or memory representations associated with the problem and its possible solutions (Niedeggen et al., 1999). Thus, Prieto-Corona et al. (2010) concluded that children displayed greater cognitive effort in retrieving the solution from long-term memory than adults. Following this line of thought, Zhou et al. (2011) compared adults and children on addition and multiplication tasks, finding different results; the arithmetic N400 was only observed for addition, while a left anterior negativity was reported for multiplication (N400, Kutas and Hillyard, 1980). The age groups differed in both ERP components, with children displaying smaller arithmetic N400 amplitude in the multiplication condition and greater arithmetic N400 amplitude in the addition condition than the adults. Based on their earlier work in adults Zhou et al. (2007), proposed that children rely more than adults on a quantitative manipulation supported on brain networks associated with manipulation of numerical magnitude while the smaller ERP amplitude during multiplication reflects a less stable verbal processing system for solving multiplication problems.

Problem complexity effects on age-related differences in performance include increased RTs and error rates as complexity increases, resulting in the split effect of strategy selection (Imbo and Vandierendonck, 2007; Van Beek et al., 2014). The split effect is defined as the use of retrieval strategies for simple problems and procedural strategies for more complex problems (Imbo and Vandierendonck, 2007). This effect is associated with an electrophysiological pattern in which adolescents have a greater frontocentral negativity between 250 and 500 ms (N280 component) for large than small additions and larger LPC in right sensors for small than large additions (Van Beek et al., 2014). In this study, the N280 component seems to serve a function similar to the arithmetic N400 component observed in adults. The greater N280 amplitude was interpreted to reflect greater engagement of attention and working memory resources for large than small addition, while the LPC amplitude was interpreted as a marker of the split effect.

Few studies have analyzed ERP components in children or preteens, and none have compared differences in mathematical processing in obese and normal weight, despite the recognized association between childhood obesity and poorer mathematical skills. It would be expected that obese and non-obese preteens have a consolidated mental representation and understanding of numerical magnitudes and an efficient strategy for solving arithmetic problems that includes appropriate use of the two types of math strategy. As childhood obesity has been associated with poor behavioral performance, a better understanding of the specific processing stages affected in this population is needed.

In this study, we assessed the math abilities of obese and non-obese preteens. We stratified participants by performance on a mathematical pre-test and then compared their performance, brain-electrical activity (ERPs), and ERP sources during the verification stage of addition and subtraction operations. We expected to find worse performance in participants with obesity than non-obese participants, with participants with obesity exhibiting lower response accuracy and longer RTs than their normal-weight peers. Three consistent ERP components were analyzed: the P150, P300, and arithmetic N400. We expected to find worse performance in participants with obesity than their normal weight peers, with electrophysiological responses consistent with greater cognitive effort, difficulties with attention allocation and with answer retrieval. Thus, we hypothesized that there would be differences between weight and mathematical skill groups in the P300, reflecting difficulty in allocating attentional resources, and N400, where longer latency and greater amplitude would reflect problems with answer retrieval. We did not expect to see differences between groups or conditions for the P150 component. We also expected that preteens with obesity with high or low math skills would show worse performance and an electrophysiological pattern associated with a greater cognitive effort than their normal-weight peers. Given that preteens between 9 and 10 years old should efficiently solve the simplest operations, such as addition, we expected differences between weight groups only in subtraction condition.

The goal of this study was to identify behavioral and electrophysiological differences between obese and non-obese participants with low or high math skill levels. Given that obesity has been associated with poorer mathematical skills, we hypothesized that obese preteens would exhibit worse behavioral performance and greater amplitude and longer latency of P300 and arithmetic N400 components on simple math problems associated with more cognitive effort in attention allocation and answer retrieval in both math skills level. Behaviorally, we observed no differences in accuracy or RT during the experimental math task between obese and non-obese participants, which is contrary to our hypothesis. One explanation for this result is that our math task did not require significant cognitive effort, given that all participants performed at a high level. Alternatively, in accord with the diffusion model, longer RTs may be used as a strategy to improve accuracy (Ratcliff and Smith, 2004; Starns and Ratcliff, 2010). Our experimental design provided all participants with sufficient time to calculate and verify their response, which may have positively affected their behavioral performance and reduced our ability to identify differences between groups.

While differences were not seen in task performance, there were significant differences between preteens with obesity and their normal-weight peers on psychometric tests. Regardless of math skill level, differences in psychometric tests favored the non-obese group. This finding is in accord with previous studies describing poorer executive functions (Maayan et al., 2011; Blanco-Gómez et al., 2015; Ayoub-Meo et al., 2019; Alatorre-Cruz et al., 2021) and lower math skills in children with obesity (Crosnoe, 2004; Castelli et al., 2007; Cottrell et al., 2007; Manes, 2015). Our study adds to this literature by indicating that these differences persist even in obese preteens with higher math skills.

In electrophysiological results, we hypothesized there would be differences between weight groups in the most reported ERPs; P300 and the arithmetic N400 components. In keeping with these expectations, we consistently observed a positivity between 200 and 400 ms on fronto-central locations, which we identify as the P300 (Polich, 2007). In our task, we suggest that the P300 component is associated with attentional processing as described in adults (Proverbio and Caminati, 2019). We further propose that this component could be associated with attention allocation (Dimoska et al., 2006; Ramautar et al., 2006; Polich, 2007) as our task requires preteens to evaluate three different responses and to select the correct response while inhibiting their response to incorrect answers. In addition, we also observed a negative wave between 300 and 500 ms on central and parietal ROIs in both experimental conditions (i.e., addition and subtraction), which we identify as the arithmetic N400 component.

As predicted, we found differences between weight groups in latency, amplitude, and source analyses of ERP components in the subtraction condition, with some of these differences dependent on the participant’s level of math skill. We expect that preteens have mastered simple addition, but that they would still be learning how to solve subtraction problems efficiently. Our findings are consistent with the expectation that they are gaining new skills in subtraction while evincing mastery over simple addition problems. If mastery of simple addition problems is considered a developmental milestone, then obese preteens seem to have passed this milestone. However, this group may face new mathematical challenges differently than their healthy-weight peers, as suggested by longitudinal studies in which participants with obesity seem to differ from their normal weight peers in arithmetical development throughout childhood and adolescence (Datar et al., 2004; Gable et al., 2012; Kranjac, 2015; Li and O’Connelly, 2015; Martin et al., 2017). Our finding that performance on subtraction problems in obese preteens was dependent on math skill level is novel and provides new information on differences in math performance between participants with obesity and normal weight preteens.

In preteens with obesity, we found that weight status negatively correlated with attentional control, with obese groups exhibiting longer P300 latency than their normal-weight peers in the subtraction condition regardless of math skill level. Longer P300 latency has been associated with a lower capacity to allocate attention (Dimoska et al., 2006; Ramautar et al., 2006; Polich, 2007). Therefore, these findings support the idea that executive functions are affected in the obese population (Maayan et al., 2011; Blanco-Gómez et al., 2015; Ayoub-Meo et al., 2019; Alatorre-Cruz et al., 2021). Moreover, this result match with Proverbio and Caminati’s (2019)’ finding. They found that participants with poorer math skills displayed longer P300 latency during incorrect and correct responses, which was not observed in participants with higher math skills, such modulation was interpreted as a poorer representation of arithmetic solutions and more difficulty retrieving that information. Thus, we can add to the interpretation that a longer P300 latency in obese preteens is associated with poorer representation of arithmetic solutions and more difficulty retrieving that information. It may also be reflecting a poorer attention control in the identification of correct solutions.

Based on the working memory system model (Baddeley, 2003), attention control and working memory are successive steps in complex processing tasks such as learning. Therefore, difficulties in one processing step will affect the next step. We thus suppose that problems in attention control will affect the retrieval of solutions from working memory and be reflected in greater amplitude and longer latency of arithmetic N400 component. Contrary to our hypothesis, we found shorter arithmetic N400 latency in right central and central ROIs for participants with obesity regardless of math skill level. Previous studies reported that in arithmetical processing, left areas are recruited when the problem requires subject calculates the result, while right and left areas are activated when a task’s mental demand is higher (Arsalidou et al., 2018). Therefore, we suggest that the longer arithmetic N400 latency in right-brain regions observed in participants with obesity might reflect that subtraction condition demands high neural resources to left hemisphere, thus they must recruit additional right-brain areas to retrieve the solutions from working memory.

Regarding the differences between weight groups that depended on math skill level, our findings were also unexpected. In participants with low math skill level, a smaller arithmetic N400 amplitude for obese than non-obese participants in left parietal and centro-parietal ROIs. An explanation for this result is that our findings did not match previous studies because we used a different experimental paradigm. These studies reported: (1) greater arithmetic N400 amplitude associated with a greater cognitive effort recovering an unrelated response (Niedeggen et al., 1999; Prieto-Corona et al., 2010), and (2) smaller arithmetic N400 amplitude associated with increasing age (Prieto-Corona et al., 2010; Zhou et al., 2011). Moreover, comparisons between children and adults might reflect the development of brain areas specializing in math processing. Adults are expected to require fewer neural resources than observed in children, resulting in a smaller arithmetic N400 amplitude (For review Peters and De Smedt, 2018).

In our study, we compared same-age preteens with the assumption that they were at a similar stage of cognitive development. Hence, we suggest that the differences we report between weight groups are related to the ability to efficiently activate neural networks associated with math skill processing. In children with poor math skills, neuroimaging studies have reported decreased network activation (Berteletti et al., 2014), and problems in modulation of brain responses as the complexity of arithmetic problems is increased (De Smedt et al., 2011; Ashkenazi et al., 2012). The electrophysiological pattern we report in obese preteens might reflect reduced neural-network activation during the processing of the retrieval of solutions from working memory. In addition, this possible explanation might also be supported on the topographical location of the arithmetic N400 component where the weight groups differed. Given that: (1) greater arithmetic N400 amplitude was observed in the left parietal region for non-obese participants, (2) greater arithmetic N400 amplitude was correlated with higher executive functions scores (i.e., TM scores), and (3) left brain areas should be recruited when the arithmetic problem is easy (Arsalidou et al., 2018). We suggest that non-obese participants might be spending more brain resources in the left hemisphere because they did not recruit additional areas as obese participants did to solve a complex problem. Therefore, differences between weight groups observed arithmetic N400 components may reflect problems in the retrieval of solutions from working memory for the participants with obesity.

In participant with high math skill level, we did not find differences between weight groups in the behavioral performance of our math task. Our weight groups only differed in source-level activation of the P300 component, with the obese group showing a hypoactivation of the left SPL. The posterior parietal cortex supports visuospatial or attentional processing (Vandenberghe and Gillebert, 2009), and plays an important role in the manipulation of visual-spatial or auditory-verbal information in working memory (Koenigs et al., 2009). For arithmetic processing, SPL is associated with three circuits for number processing. In particular, the SPL supports attentional orientation of the mental number line (Dehane et al., 2003; Arsalidou et al., 2018). The function of this brain area in arithmetic processing is still discussed, but it has been proposed that parietal, temporal, and occipital areas play an important role in the representation of figurative/object schemes during the calculation of arithmetic facts (Pascual-Leone and Johnson, 2005), with right parietal areas activated when a task’s mental demand is too high for the left hemisphere to trigger automatic schemas, while left parietal areas recruited when the problem require that the subject calculates the result, involving the child’s mental attentional capacity, thus a harder-relational figurative knowledge (Arsalidou et al., 2018).

In our study, we presume that subtraction operations require greater attentional orientation of the mental number line and thus harder-relational figurative knowledge, therefore, the preteens should recruit left parietal areas. The hypoactivation on left SPL observed in our participants with obesity may reflect a difficulty in brain response modulation with the increased arithmetical complexity of subtraction problems as is observed in children with poorer math skills or dyscalculia (De Smedt et al., 2011; Ashkenazi et al., 2012), which may be associated with an inefficient allocation of attentional orientation and weaker-relational schemes to solve the subtraction operation. We suggest that these results support the need for additional studies specifically focused on attentional resource allocation in obese adolescents during problem solving tasks.

Our behavioral and electrophysiological results might indicate that lower mathematical skills previously reported in obese preteens are not directly associated with weight status. Rather, weight status seems to interact with the cognitive processes needed for math processing.

Interestingly, regardless of math skill level, differences between preteens with obesity and their normal-weight peers were related to the allocation of attentional resources and working memory processes. Thus, obese preteens seem to have worse executive control to solve math problems than their normal-weight peers, supporting the dependency relationship between weight status and these cognitive processes.

Our results support previously reported results in suggesting that participants with obesity face a disadvantage with respect to their normal-weight peers that is reflected in a delay in the development of math skills. This disadvantage may be due to poorer attention control and the retrieval of solutions from working memory that could be compensated by other cognitive processes, additional training, or recruitment of additional brain areas.

The original contributions presented in the study are publicly available. This raw data can be found here: https://openneuro.org/datasets/ds004019.

The studies involving human participants were reviewed and approved by the University of Arkansas for Medical Sciences Institutional Review Board. Written informed consent to participate in this study was provided by the participants’ legal guardian/next of kin.

GA-C, SS, and LL-P contributed to the conception and design of the study. HD and DH organized the database and applied the experimental task. GA-C and DW performed the statistical analysis. GA-C and LL-P wrote the first draft of the manuscript. All authors contributed to manuscript revision, read, and approved the submitted version.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.