The Impact of COVID-19 in Brazil Through Educational Neuroscience Lens

1. Introduction

Schooling is a human experience with a strong impact on cognitive, neurological, and social development. Enrolling and remaining in formal education brings profound changes in child development [1], such as the strong promotion of cognitive control that occurs with elementary schooling in children aged 5 to 7 years [2]. In addition, there are specific effects, such as the impact of learning to read on neurological functioning and cognitive skills like phonological processing and visual perception [3,4,5], as well as the impact of learning mathematics and related concepts, such as geometry, object recognition, and the automatization of calculation processes, on neurological and cognitive functioning [6,7,8].

Although educational quality is crucial for all, children from lower socioeconomic status are at higher risk of receiving education of lower quality. Higher socioeconomic status often provides access to better educational resources, supportive learning environments, and greater motivation, all of which enhance academic performance [9]. Conversely, students from lower socioeconomic backgrounds face limited access to learning materials, stimulation at home, and higher exposure to stress, which can hinder their academic and cognitive success [10]. In the last decades, educational neuroscience has shown how brain function and structure, shaped by socioeconomic status, are one of the main pathways for the impact of socioeconomic status on education [11] and cognitive outcomes [10]. Those results are especially relevant for countries in the Global South, such as Brazil, in which poverty rates are high and educational outcomes are low.

The COVID-19 pandemic and the consequent disruption of in-person education was one of the major recent events with a strong impact on schooling [12]. The impact was due not only to lack of schooling but also because in the cases in which schooling continued in a virtual form, the quality of education was not guaranteed. For example, Engzell et al. found that students made little or no progress in learning during the closure [13]. During the pandemic, educational neuroscience became crucial for identifying the neurocognitive effects of schooling disruption as well as intervention targets. Specifically, educational neuroscience during remote teaching helped to explore educational practices [14], to improve educational practices [15], train teachers [16], and to measure the impact of educational disruption on neurodevelopment [17].

Although all children suffered the consequences of the pandemic in the short and the long term, the most vulnerable populations suffered more, even when some compensatory measures were taken [18]. Logically, children who do not have access to the internet and a large-screen device, have poorly educated parents, and do not have a quiet space to study were more affected [19]. In consequence, school closures in all countries are likely to have sharpened existing disparities as socioeconomic status is positively associated with children's ability to adjust to remote learning [20].

About this last point, overall studies have found that educational suspension had effects on cognition as well as on academic achievement. For example, a systematic review analyzed articles published between March 1, 2020, and April 30, 2021, to investigate the general effects of school closures on student performance in elementary and high school [21]. The authors highlighted that, due to remote learning, students' development of academic competencies and skills could be affected by up to one-tenth of a standard deviation, potentially resulting in a lifetime income loss of 2.7% to 4.6%. As a consequence, the study suggested a global economic loss of 2.56 billion euros or 1.3% of future gross domestic product. Importantly, although the COVID-19 pandemic interrupted regular schooling across almost the entire world, a more intense effect was observed on countries with fewer financial resources. For example, while 95% of 15-year-olds in Denmark, Slovenia, Norway, Poland, and Iceland had access to a computer and internet in 2018, only about 50% of 15-year-olds in Mexico had access to a computer and internet for remote learning [20].

Specifically in Brazil, an Emergency Remote Learning model was implemented, and about 5.5 million Brazilian children had limited or no access to education during this period [22]. According to data from INEP [23], Brazil was one of the countries where schools remained closed the longest, with an average of 279 days closed in 2020 and an additional 103 days in 2021.

There were few studies conducted in Brazil analyzing the impact of the pandemic. One group of studies analyze the scores in standardized tests of academic achievement obtained after the school closure ended. For example, in the 2022 Programme for International Student Assessment (PISA), which evaluates 15-year-old students' knowledge in mathematics, reading, and science, Brazilian students showed declines in all three subjects [24]. During remote learning, the assessment showed that 38% of students (OECD average: 34%) had difficulty understanding school assignments at least once a week, and about 30% (OECD average: 24%) struggled to find someone to help them with schoolwork. In another study, data from the 2021 Basic Education Assessment System (Saeb) showed declines in student performance, correlated with the impact of the COVID-19 pandemic on Brazilian education [25]. Results indicated that the greatest drop occurred in written language in the second year of elementary school, where the average score fell from 750 points in 2019 to 725.9 points in 2021, about half a standard deviation. In mathematics, there was also a decline, though less significant, from 750 to 741.6 points. In later school years, the biggest declines occurred in mathematics, with fifth- and ninth-grade students showing score drops of 11 and 7 points, respectively, compared to 2019. In Portuguese, there was a drop of 6.6 points for fifth graders and 2.2 points for ninth graders [23].

A second group of correlational studies analyzed what ages were most impacted. For example, Alves et al. carried out a longitudinal study with students from the second to the fifth grade of elementary school in Minas Gerais [26], aiming to assess reading fluency during the remote learning period. The findings indicated that second-grade students suffered the greatest negative impact due to school closures and achieved lower results in March 2022 compared to the same period in the previous year (the beginning of the COVID-19 pandemic). This result suggests that the younger the children the most impacted their academic achievement. This is possibly because older students, in the final years of elementary school or in high school, seem to be more familiar with technology and, therefore, more autonomous [27,28,29].

A third type of studies addressed the impact of the pandemic, comparing the pre-pandemic norms with the post-pandemic scores obtained by high socioeconomic status children. For example, in a study conducted with children in grades 1 to 5 of a private elementary school in a city of medium to high socioeconomic status in the interior of the state of São Paulo [30], the results showed significantly lower performance compared to the pre-pandemic normative sample. The declines were observed in arithmetic for 3rd, 4th, and 5th grades and in writing for 2nd, 4th, and 5th grades.

To worsen the situation, in addition to the deficits observed in academic learning, some studies have also suggested cognitive deficits, specifically in EF [31,32,33]. EF refers to a set of cognitive processes related to the control and management of cognition and behavior [34]. These include basic components such as inhibition (IN), working memory (WM), and cognitive flexibility (CF), as well as complex components like planning [35]. EF is strongly related to academic success, affecting the learning process, and, in the long term, lower EF levels are predictors of academic difficulties and school dropout [36,37,38,39]. Brazilian studies have also demonstrated the relationship between EF and school subjects [40,41,42], such as mathematics, reading, and writing [39,43,44].

Campos, Seabra and Carreiro longitudinally followed students in the early years of elementary school before and during the pandemic, at the beginning and end of 2020 [45]. They observed an increase in complaints about PE-related problems among students. In another Brazilian study, after schools reopened, 117 students from 1st and 2nd grades in a public school had their neuropsychological functions assessed. The results pointed to a high prevalence of children in the literacy phase showing alert states or deficits in orientation, memory, attention, language, visuospatial skills, arithmetic skills, and verbal fluency [46].

In this context, it is evident that school is a highly stimulating environment for human development, and therefore, ensuring access to education and the quality of teaching is essential [47,48]. Despite these findings, there are still no studies in Brazil aimed at describing potential EF and academic losses across different school years to identify the most affected grades and thus enable more targeted intervention efforts. Therefore, this study aimed to contribute to map executive functioning and reading, writing, and math skills in Brazilian students from 1st to 9th grade after the pandemic, comparing them with the pre-pandemic normative data of the assessment instruments used to determine possible performance discrepancies. Importantly, we will explore that from an educational neuroscience perspective. The scarcity of national studies on this topic in the Global South and the importance of educational monitoring as part of an evidence-based policy justify this study, whose results may contribute to guiding future projects designed to mitigate some of the learning losses caused by the remote learning period.

2. Materials and Methods

2.1. Participants

Initially, 193 elementary school children from three public schools in two cities of the Greater São Paulo area participated in the study, with 72 students from the city of Santo André and 111 students from the city of Embu das Artes. The sample was selected for convenience. All children from these schools were invited to participate; however, certain diagnoses were used as exclusion criteria based on school records, including Attention Deficit Hyperactivity Disorder, Intellectual Disability, Autism Spectrum Disorder, and Specific Learning Disorder.

Subsequently, after analyzing the results, participants with an intellectual level potentially indicative of undiagnosed intellectual disability were excluded. Specifically, exclusion criteria included scores below the 5th percentile on the Raven's Colored Progressive Matrices test, for participants aged 6 to 10 years; and an intelligence quotient below 70, according to results from the Wechsler Abbreviated Scale of Intelligence [WASI], for participants aged 11 and older. Based on this criteria, fifteen participants were excluded from the sample.

The participants' ages ranged from 6 to 14 years (M=10; SD=5.65), with 55.62% female and 44.38% male. Regarding grade level, 8.44% were in 1st grade, 10.11% in 2nd and 3rd grades, 12.92% in 4th grade, 14.04% in 5th grade, 12.36% in 6th grade, 11.80% in 7th grade, 10.67% in 8th grade, and 9.55% in 9th grade.

Data collection occurred in the second semester of 2023 (from September to November) and the first semester of 2024 (from May to July). Table 1 presents the school grade of the study participants during the pandemic period (2020 and 2021) and at the time of data collection (2023 and 2024).

The city of Santo André has an average monthly salary for formal workers of approximately US$ 625.92 [49]. In the 2023 Basic Education Development Index (IDEB), the city scored 6.5 out of 10 points in the early years of elementary education in public schools and 5.2 out of 10 points in the final years [23]. Embu das Artes has an average monthly salary for formal workers of approximately US$ 650.03 [49]. The IDEB for the early years of elementary education in public schools is 6.2 out of 10 points in the early years and 5.0 out of 10 points in the final years [23]. Both cities have IDEB close to the average for the State of São Paulo (6.2 in the initial grades and 5.1 in the final grades).

In March 2020, in-person classes were suspended for all children. Regular school attendance was resumed in February 2022, following government guidelines.

2.2. Instruments

The following instruments were used to evaluate child cognitive performance:

- Digit Span Task (DST): Assesses short-term verbal memory (Forward Order - FO) and working memory (Backward Order - BO). The examinee's task is to repeat sequences of digits in either forward or backward order. The test was administered individually, and the measures used were total correct responses (sum of both orders). This task is standardized for the Brazilian population and there is evidence of validity for ages 4 to 10 years [50]. In this study, this instrument was used with children up to 10 years old. For this task the minimum possible raw score is zero and the maximum possible raw score is 16.

- Digit Span Subtest from the Wechsler Intelligence Scale for Children - WISC IV: Assesses short-term verbal memory (Forward Order - FO) and working memory (Backward Order - BO). The examinee's task is to repeat sequences of digits in either forward or backward order. The test was administered individually, and the measures used were total correct responses (sum of both orders). This task is standardized for the Brazilian population and has demonstrated validity for individuals aged 6 years to 16 years and 11 months [51]. In this study, this instrument was used with children aged 11 and older. For this task the minimum possible raw score is zero and the maximum possible raw score is 16.

- Free Verbal Fluency (FVL): Assesses linguistic-mnemonic processes related to lexical search and components of executive functions, such as cognitive flexibility, inhibition, working memory, and search strategy organization. In this individually administered task, the examinee must say as many words as possible within a given time (two and a half minutes), respecting the guidelines to keep their eyes closed and not say proper names or numbers. This task is standardized for the Brazilian population and its validity was probed with individuals aged 6 to 12 years [52]. In this study, for participants older than the normative sample, data corresponding to age 12 were used only to determine whether, even when compared to younger children, there could be indicators of difficulty in these cognitive skills. For this instrument, the minimum possible raw score is zero and the maximum raw score is equivalent to the number of correct answers.

- Hayling Test (HT): The HT is divided into two parts, A and B, which assess, respectively, linguistic aspects and EF (such as cognitive flexibility and inhibitory control), as well as processing speed. In this individually administered task, the participant must, in Part A, complete the sentences with words that fit appropriately in the context, as quickly as possible. In Part B, the objective is to complete the sentences with words unrelated to the context presented. There are two versions, both with evidence of validity and standardization for the Brazilian population: Children’s Hayling Test (CHT), used with individuals aged 6 to 12 years [53], and Adult’s Hayling Test (AHT), from age 13 onwards [54]. Part A is only a basal analysis to assess EF in Part B, in this study we used only measures related to Part B. The measures used were execution time in Part B, which is the sum of the reaction time in seconds of all items, and the number of errors in Part B. For CHT the minimum possible raw score is zero and the maximum possible raw score is 10, but for AHT the minimum possible raw score is zero and the maximum possible raw score is 15.

School achievement was evaluated with the following instruments.

- School Performance Test II (TDE II): The TDE II aims to assess basic skills in reading, writing, and arithmetic. There is evidence of validity for this test, which has also been standardized for the Brazilian population, with students from the first to the ninth year of elementary school [55]. Below is a brief description of each part:

• Writing Subtest: Consists of a list of 40 words that are read one by one to the students, and they must write them down. There is one version for 1st to 4th grade and another for 5th to 9th grade. The test was administered collectively, with time being measured, and scoring was interrupted after 10 consecutive errors or "don't know" responses. The measures used were the score and the duration of the task. For this task the minimum possible raw score is zero and the maximum possible raw score is 40, in both versions.
• Reading Subtest: Composed of a list of 36 words for 1st to 4th grade, and 33 words for 5th to 9th grade, which the examinee must read aloud individually. The time was recorded, and scoring was interrupted after 10 consecutive errors or "don't know" responses. The measures used were the score and the duration of the task. For students in grades 1 to 4 the minimum possible raw score is zero and the maximum possible raw score is 36, but for students in grades 5 to 9 the minimum possible raw score is zero and the maximum possible raw score is 33.
• Arithmetic Subtest: Consists of a list of 37 exercises for 1st to 5th grade, and 43 for 6th to 9th grade, which the examinee must solve. The test was administered individually for children from 1st to 3rd grade and collectively for other participants, according to the manual's guidelines. The time was recorded, and scoring was interrupted after 10 consecutive errors or "don't know" responses. The measures used were the score and the duration of the task. For students in grades 1 to 5 the minimum possible raw score is zero and the maximum possible raw score is 37, but for students in grades 6 to 9 the minimum possible raw score is zero and the maximum possible raw score is 43.

To check inclusion criteria we administered the following tests:

- Raven’s Coloured Progressive Matrices (CPM): The CPM assesses general intelligence through visual puzzles designed to evaluate pattern recognition. It is organized in three parts, which are arranged in increasing difficulty. There is evidence of validity and it was standardized for the Brazilian population, specifically for individuals aged 5 to 11 years [56]. In this study, this instrument was administered to children from 1st to 5th grade to screen for significant intellectual deficits as an exclusion criterion.

- Wechsler Abbreviated Scale of Intelligence (WASI): This is a brief intelligence assessment tool. The Vocabulary and Matrix Reasoning subtests were individually administered, to estimate general cognitive functioning. There is evidence of the validity for this test, which is also standardized for the Brazilian population, with normative study covering individuals aged 6 to 89 years [57]. In this study, it was used for children from 6th grade onwards to screen for significant intellectual deficits as exclusion criteria.

2.3. Data Collection Procedures

The project was approved by the Research Ethics Committee of the proposing institution. Authorization was obtained from the school directors, the students' guardians and the students themselves assented for participation in the research. The assessments were divided into three sessions, held at the school in a suitable location, and were carried out individually or in groups, following the application guidelines of each test. At the conclusion of the analysis, individual data and the overall research results were shared with the respective guardians of the students. The application followed a fixed order, first session: WASI or Raven's Coloured Progressive Matrices (according to the student's age at the time of the test), FVF, Hayling Test, Digit Tasks and Reading (from TDE-II); second session: individual (1st to 3rd grade) and collective (from 4th grade) application of the Writing Subtest (TDE-II); third session: individual (1st to 3rd grade) and collective (from 4th grade) application of the Arithmetic Subtest (TDE-II).

2.4. Data Analysis Procedures

Descriptive statistics (frequency, percentage, mean, and standard deviation) were used to analyze sociodemographic data and characterize the participants. Additionally, mean and standard deviation were employed to describe data across age groups or school grades in the pre-pandemic period (based on instrument norms) and the post-pandemic period (based on data collected in this study), considering each task and dependent variable separately [58].

To examine potential differences between the sample in this study and the normative samples established in the pre-pandemic period, as described in the test manuals, one-sample t-tests were conducted [58]. Given the small sample size for each school grade and age, comparative analyses also included confidence intervals for the differences, estimated using the bootstrapping resampling procedure (1,000 resamplings, BcA - Bias-Corrected and Accelerated, with a 95% confidence interval). The BcA bootstrapping method was used to mitigate issues related to the lack of normal data distribution due to the small sample size, improve the accuracy of confidence intervals, and correct biases and asymmetries in the distribution of estimates [59].

Furthermore, effect size was calculated using Cohen’s d [60] to quantify the magnitude of the differences, applying the following criteria: d > 0.80 = large effect; d = 0.50 = medium effect; d = 0.20 = small effect. A significance level of p < 0.05 was considered statistically significant. All analyses were conducted using IBM SPSS Statistics software (version 27.0).

3. Results

Table 2, Table 3, Table 4, Table 5, Table 6, Table 7, Table 8 and Table 9 present the participants' performance in terms of descriptive statistics (mean and standard deviation), as well as the normative data, and the t-tests results. The executive function measures are presented by age, while the academic performance measures are presented by school year because the respective test manuals grouped the normative sample data this way.

Regarding working memory performance comparisons, assessed using the Digit Span Test (see Table 2), significant differences were observed at all ages except for participants aged 8 and 9 years. The confidence interval in the bootstrap analysis supported these findings. Notably, students aged 6 and 7 years performed significantly better than the normative samples of the instrument. The effect size in these comparisons ranged from large to medium, respectively. In contrast, students aged 10 to 14 years performed significantly worse than the pre-pandemic norms, with effect sizes ranging from small to medium.

Regarding verbal fluency, when comparing the current mean with the normative mean of the Verbal Fluency Test (Table 3), statistically significant differences were observed in the results of participants aged 9 to 14 years. These differences were corroborated by the confidence interval in the bootstrap analysis, with the current sample showing poorer performance compared to pre-pandemic norms across all ages. In all those cases, the effect sizes ranged from medium to large, with the most pronounced differences observed in students aged 11 to 14 years. It is important to note that, for verbal fluency measures, the norms for 12-year-olds were used to compare participants aged 13 and 14 due to the absence of norms for this age group. Even when compared to younger age norms, the present sample of 13- and 14-year-olds performed worse than the normative mean.

In terms of errors on the Hayling Test (see Table 4), which measures cognitive flexibility and inhibitory control, statistically significant differences were observed between participants aged 7, 9, 10, 11, 12, 13, and 14, with the present study sample showing lower performance (i.e., higher number of errors) compared to pre-pandemic normative means. The effect size also ranged from medium to large, indicating the relevance of the differences. In the group of 6-year-old participants, although the p-value (p = 0.056) did not reach the conventional statistical significance threshold (p < 0.05), the bootstrap analysis, combined with a large effect size, suggests potential practical significance. As mentioned previously, this result indicates that the small sample size (n = 8) may have compromised the statistical power of the analysis.

Regarding the comparisons involving the analysis of reaction time from the Hayling B Test, which also provides measures of cognitive flexibility and inhibitory control, the results revealed significant differences in the participants aged 10 to 12 years (Table 5). In these groups, the participants presented faster performance compared to the normative means. These results were corroborated by the confidence intervals obtained in the bootstrap analyses. In these comparisons, the effect sizes ranged from medium to large. For the 6-year-old group, although the p value (p = 0.171) did not reach the conventional level of statistical significance (p < 0.05), the analysis of the bootstrap confidence interval and the average effect size indicate the presence of a relevant difference. This discrepancy may reflect a Type II error, possibly due to the small sample size (n = 8), which compromises the statistical power of the analysis [61].

Writing ability was assessed by the Writing Subtest of the TDE-II. As shown in Table 6, students from 5th to 8th grade showed statistically significant differences between the means, with weaker post-pandemic performance. These results were corroborated by the confidence intervals of the bootstrap analysis. The effect size ranged from medium to large in these comparisons.

For the word reading score, assessed using the TDE-II Reading Subtest, students from 5th to 8th grades performed significantly lower than expected when compared to pre-pandemic norms, which was corroborated by the bootstrap analysis confidence interval and the observed mean effect sizes (Table 7). In the 4th grade, although the p-value (p = 0.058) did not reach the conventional level of statistical significance, the bootstrap analysis, combined with a robust effect size, indicates a potential practical significance. This finding suggests that the sample size limitation may have influenced the results of the t-test analysis and, therefore, that the bootstrapping method was necessary to confirm the results. Regarding the execution time in this reading subtest (Table 8), only the 2nd grade performed faster compared to the normative mean, a result that was corroborated by the bootstrap analysis. The effect size of this comparison was large.

Regarding arithmetic, a skill assessed by the Arithmetic Subtest of the TDE-II, 8th and 9th grade students performed worse than the pre-pandemic normative averages, a result corroborated by the confidence interval obtained in the bootstrap analysis, with large effect sizes for both groups (Table 9). In the group of 4th grade students, although the p-value (p = 0.066) did not reach the conventional threshold of statistical significance (p < 0.05), both the bootstrap confidence interval and the effect size suggest the presence of practical relevance, with lower performance in the current sample. This finding indicates that the reduced sample size may have compromised the statistical power, making it difficult to detect significant differences even in the face of a trend of lower performance than the standardized data. Likewise, the groups of 1st, 2nd, and 5th grade students presented non-significant p-values; however, subsequent analyses – through bootstrap confidence intervals and effect sizes (d ranging from 0.410 to 0.484) – suggest that the sample limitation may have masked the practical relevance of the results [60], which indicate superior performances in the current sample compared to pre-pandemic norms.

Table 10 presents a comparative summary of the significant results obtained in this study when comparing the post-pandemic data with pre-pandemic normative data. Executive function measures are presented by age, while academic performance measures are presented by school year, since the respective test manuals grouped the normative data of the sample in this way. It should be noted that for score measures, the higher the value, the better the performance (top of the table). Conversely, for error and time measures, the higher the value, the worse the performance (bottom of the table).

4. Discussion

The present study compared child cognition (short-term verbal memory, working memory, verbal fluency, cognitive flexibility, inhibitory control, processing speed) and cognitive achievement (reading, writing, and arithmetic) performance after the COVID-19 pandemic with the norms established for these variables before, by age and school grade. Specifically, our results showed EF and academic losses across 1st to 9th school years in Brazil, a population underrepresented in educational neuroscience literature.

Overall, we have found that in the majority of the tests, children showed lower scores after the pandemic than before, especially between 10 and 13 years (i.e., between 5th and 8th grade). No difference was observed at age 8 (i.e., 3rd grade). This general result is in line with previous studies showing that the suspension of in-person education impacted negatively on child cognitive development and academic achievement [12,13,17,21], and also with studies conducted in Brazil [23,24,25,26]. As others have proposed, limitations such the lack of computers or internet access as well as the difficulty understanding school assignments jointly with the problems to find someone to help with schoolwork, were reported in Brazil as potential explanations for the decrease in achievement [24,25]. However, the findings showing a higher number of differences in the oldest children and the higher grades contrast with studies conducted in the country. For example, Alves and collaborators [26] analyzed children across 2nd to 5th grades of elementary school and found an impact of classes-suspension in reading fluency only in the 2nd grade children, but they analyzed a sample coming from private high socioeconomic-status schools. Instead, the schools in our study were from middle-to-low socioeconomic status contexts. It has been proposed that older students, in the final years of elementary school or in high school, seem to be more familiar with technology and, therefore, more autonomous, taking most advantage from virtual classes [27,28,29]. However, if socioeconomic conditions impede access to computers or to the internet, technology familiarity would not have any effect. Future studies should analyze how the family socioeconomic level can interact with the student’s age to produce different impacts on reading.

One strength of our study is the number of tests implemented, and it is also an example of how educational neuroscience has contributed to a very detailed evaluation of the impact of the pandemic. In an analysis of the findings for each test, our results revealed a significant decline in the average verbal fluency performance, especially among participants aged 9 to 14 years, and remarkably, when compared to younger age norms. This result aligns with investigations conducted in Brazil, which reported an increase in complaints of difficulties related to EF among students in the early years of Elementary School [45]. Furthermore, Ribeiro, Celeste, and Reis [46] found that 48.72% of children enrolled in the 1st and 2nd years of Elementary School, after the pandemic period, presented scores indicating warnings or deficits in verbal fluency. In sum, our results suggest that class-suspension is associated with lower vocabulary (number of words acquired by individuals).

Regarding the scores obtained in working memory, we found a similar trend: a significant decline was observed when comparing the current average with the normative average, particularly among older students. The decrease in working memory associated with the pandemic has been observed in previous studies [46,62] and it has been suggested that might be linked with the anxiety produced by the lockdown [62]. Conversely, the fact that students aged 6 and 7 achieved scores higher than the pre-pandemic norm, contrasted with the findings of the study conducted in Brazil by Ribeiro, Celeste, and Reis [46], whose results indicated that 62.4% of the children in the first and second years of Elementary School showed scores indicative of deficit in working memory. Remarkably, a study conducted in Ethiopia compared test results applied to groups before and after school closures, demonstrating higher scores in working memory and non-verbal intelligence in the group assessed after the return to in-person school activities [63]. Future studies should clarify the reason for these differences.

In relation to inhibitory control, with the exception of 8-year-old students, all other participants had a higher number of errors than the expected normative average. Conversely, they showed faster performance than expected according to the guidelines established before the pandemic, but this is not interpreted as an improvement given that the higher speed is associated with a worse achievement in the task. The high number of errors is in line with previous evidence showing similar results [45]. Daily screen use has been suggested as one of the factors decreasing inhibition in American school-age children [63], but in Brazil, only 31% of public elementary school students had a computer/ tablet and broadband at home [64], and therefore, more evidence is needed to explain the relationship between these factors in Global South countries.

Regarding educational outcomes, our results generally replicate previous findings [13]: the pandemic negatively affected academic achievement across various educational areas, mainly in 5th and 8th grades. Studies [see 18 for a review] indicate that prolonged educational disruption has resulted in attenuated learning gains. Further, if that impact was significant in countries from the Global North [13], and the most remarkable effects were for those already at risk for educational disparities [18,19], we might expect a bigger effect in countries from the Global South such as Brazil, which is actually what we found. Among the factors explaining those effects, the direct lack of educational opportunities, as well as the stressors produced by the lockdown were suggested as important variables [17].

Specifically, we found that in reading and writing tests, children had lower scores after the pandemic, between 5th and 8th grade. It is important to emphasize that these students, during the pandemic, were in the process of literacy development and/or in the early years of elementary education. Our results corroborate the ones obtained by Santos [30] which revealed significantly lower performance compared to pre-pandemic standards among 2nd, 4th, and 5th-grade students from high socioeconomic status schools, showing that in this area the pandemic impacted across all economic levels. We have also observed that 2nd grade students were faster post-pandemic, although that result might be due to particularities of the evaluated sample such as the homogeneity (the standard deviation is small in comparison with standard deviation of 1st year students as well as with the norm). However, the general trend indicated that the pandemic affected negatively reading and writing achievements. Brazilian studies conducted with children in the early years of elementary school also highlighted learning gaps related to reading comprehension and fluency [26]. About writing, it has been suggested that the use of digital platforms did not allow for the stimulation of students' motor skills [65]. This situation had already been anticipated by authors who discussed the potential risks associated with low-quality education resulting from remote learning [17] and by studies showing similar results with big sample sizes [13,63].

Regarding arithmetic skills, the results are the opposite: at some grades, they improved after the pandemic, but in other ones, they decreased. Specifically, at 1st, 2nd and 5th grade, children improved after the pandemic. Instead, 4th, 8th, and 9th-grade students showed a significant decline in performance when compared to pre-pandemic normative averages. This last result agrees with the one obtained by Santos [30], which identified a decline in the academic performance of students in 3rd, 4th, and 5th grades, when comparing their results to the instrument's normative standards. However, the result of the improvement for the first years is unexpected, and also contradictory with studies involving big samples [13,66]. Future studies should confirm our results.

In summary, the results obtained in the assessment of academic skills are generally consistent with data from national [24] and international [25] educational evaluations, which indicated a decline in student performance. This phenomenon is closely associated with the impact of the Covid-19 pandemic, particularly in countries such as Brazil which kept educational institutions closed for extended periods and faced significant challenges in implementing remote learning [23]. In the face of humanitarian crises and natural disasters, the disruption of access to the educational system can lead to long-term consequences, even with the implementation of compensatory measures [18].

The contribution of this work was to add data on the effects of the pandemic on children from the Global South. The strength is that the administration of various tests led us to explain in high detail the impact, according to age or grade, and to the cognitive or educational process. However, this study also has limitations, especially the low sample size. Although statistical procedures (bootstrapping) were implemented to mitigate issues related to the small sample size, it is important to state that these results are not generalizable (we used a convenience sample) and that future studies are needed to determine exactly the long-term impact of the pandemic on child cognitive and educational outcomes.