Descriptive statistics for each cognitive and school performance measure are shown in Table 1. Significant correlations were observed between school performance and the cognitive measures, except for the TAC measures. High correlations were observed between school performance and language measures (PCFO and PPVT), moderate correlations between school performance and general intelligence (Raven), and moderate to low correlations between school performance and EF measures (moderate correlations with TMTB and MTA, and low with MTV). The coefficients are shown in Table 2.

Table 1

Note: PCFO: Phonological Awareness Test Oral Production (Prova de Consciência Fonológica por Produção Oral); PPVT: Peabody Picture Vocabulary Test; TMTB: Trail Making Test, Part B; TAC: Cancellation Attention Test (Teste de Atenção por Cancelamento); MTA: Auditory Working Memory Test (Teste de Memória de Trabalho Auditiva); MTV: Visual Working Memory Test (Teste de Memória de Trabalho Visual).

Table 2

Note: Significant findings are highlighted in bold.

The measures (variables) to be included in the regression analysis were selected based on the correlation analysis. The hierarchical model was chosen to test whether the language and EF measures could contribute to academic performance even after controlling for intelligence. The resulting explanatory models are presented in an integrated manner in Table 3 (all with satisfactory adjustment (p < 0.001).

Table 3

Note: Significant findings are highlighted in bold.

In the first analysis, the test Raven, included as an explanatory variable in the first step (model 1), explained 34% of variance in student performance. In the second step, the inclusion of the language variables increased the explanatory power of the model to 59% (second model). Both language variables were important predictors of school performance, and with their inclusion, intelligence prediction was no longer significant. In the third step, the EF variables were added, generating a very consistent model (model 3), able to explain up to 66% of school performance. In this model, PA, vocabulary, cognitive flexibility, and (although marginally significant) auditory working memory have statistically significant regression coefficients; the intelligence measure still had no significant contribution.

In the second analysis, the first step was similar to that of the previous regression (model 1). However, in the second step, the EF skills were included generating a model (2b) able to explain up to 51% of the variance in school performance, with all variables included (Raven, TMTB-sequences, MTA) contributing with statistically significant regression coefficients. Finally, in the third step, the language variables were added, yielding similar results to those of the third step of the previous regression (model 3). It is worth mentioning that with the inclusion of the language measures, the performance of working memory was no longer a significant predictor of school performance (as it was in the second step), possibly due to the shared variance with the language measures.

The results show that, according to the literature, school performance is related to different cognitive abilities, intelligence (Almeida et al., 2008; Dodonova & Dodonov, 2012; Lemos et al., 2008; Oliveira & Soares, 2011; Spinath et al., 2010), oral language (Capovilla & Dias, 2008; Dias & Seabra, 2012; Mokhtari & Niederhauser, 2013; Piccolo & Salles, 2013; Skibbe et al., 2008; Tenório & Ávila, 2012), and EF (Blair & Diamond, 2008; Diamond, 2013; Fonseca et al., 2015; Piccolo & Salles, 2013; Seabra et al., 2014). Attention was the only one that did not correlate with children school performance. It is likely that in 5th grade, when there are more complex and challenging concepts, attention (a more basic skill), according to a cancellation assessment, is not sufficiently discriminative (or as discriminative as the other skills) among students with better or worse academic performance; it is important to highlight that a non-clinical sample of children was evaluated. Overall, the results show that children with higher level of general intelligence, larger vocabulary and better PA, greater cognitive flexibility, and strong working memory skills tend to have better grades.

Based on this initial finding, the subsequent regression analyses explored explanatory models of school performance based on the cognitive abilities investigated. The first model showed that general intelligence is able to significantly explain the performance of children, which was expected, according to the review of previous studies in this area (Almeida et al., 2008; Lemos et al., 2008; Dodonova & Dodonov, 2012; Oliveira & Soares, 2011; Spinath et al., 2010; Zampieri et al., 2012). It is important to remember that the study sample did not include participants with intellectual disabilities; therefore, the analyses took into account the less extreme variations in this skill. Thus, general intelligence, when evaluated isolatedly, was a significant predictor of students’ performance at the end of the 5th grade.

After controlling the effect of intelligence, the inclusion of language skills and EF increased the explanatory power of the model, showing that both domains make unique contributions to school performance. It is worth highlighting, however, that with the inclusion of language variables (model 2a) the contribution of intelligence was no longer significant, i.e., PA and vocabulary, in that order, were better predictors of school performance than intelligence in the sample studied. This finding may be related to the age and educational level of the participants. In other words, oral language in general, and, specifically, PA and vocabulary are important predictors (regarded as precursors) of reading proficiency (Capovilla & Dias, 2008; Cutting et al., 2009; Dias & Seabra, 2012; Kairaluoma et al., 2013; Mokhtari & Niederhauser, 2013; Scarborough, 2009; Seabra & Dias, 2012a; Skibbe et al., 2008). Reading skills, in turn, are developed during the elementary years (EF-I in Brazil), and they are also related to school performance (Dias et al., 2015). Therefore, it is possible that the reading skills were not yet consolidated in this sample of 5th graders and can be mediating this result. In older students, who already have good reading skills, this ability, and consequently oral language abilities, may exert a less significant effect on school performance.

Accordingly, although there is evidence that the PA is a predictor of reading proficiency, even in adolescents (Kairaluoma et al., 2013), its relationship with school performance found in the present study was stronger than expected. Thus, it is possible that reading proficiency (in terms of word recognition) was not yet fully developed in the sample studied, which would explain the still major role played by PA in the performance of those students. This fact may reflect a sampling bias or a problem in most Brazilian schools, where children reach the last grade of elementary school (EF-I in Brazil) without complete mastery of the written code. Another hypothesis is that the PA measure used has other requirements, which will be discussed further ahead.

Another noteworthy finding is related to the contribution of EF to the model. As expected, there was a unique contribution of EF to school performance, showing that these skills are indeed of great importance for the learning process (Blair & Diamond, 2008; Diamond, 2013; Fonseca et al., 2015; Seabra et al., 2014). Cognitive flexibility, proved to be specially linked to school performance. This may suggest that the students which are more able to handle two or more sources of information simultaneously (such as paying attention to the teacher while taking notes) or those who are able to consider different approaches (e.g., trying different alternatives to solve a problem) are those who get better grades.

Comparing the contributions of oral language and EF, it was observed that the initial inclusion of language skills generated a model with higher explanatory power (model 2a) than that of the model with the initial inclusion of the EF skills (Model 2b). Moreover, in third step of the regressions, when all skills were included, working memory did not have a significant contribution to the final model, as it was expected (T. Alloway & Alloway, 2010). One explanation for this finding is a possible overlap of the measures used, i.e., the PA tasks require, to some extent, processing or mental manipulation of information (for example, when blending the sounds /g/, /a/, /t/, /o/, the child needs to hold all of these sounds in memory and blend them to form the word “gato”, in Brazilian Portuguese. Similarly, in English it would be blending /c/, /a/, /t/ to form the word “cat”). Thus, it is likely that PA mediates part of the auditory working memory load related to school performance. This hypothesis is confirmed by analyzing model 2b, in which, before the inclusion of the language skills, auditory working memory contributed significantly to school performance prediction. Model 3, therefore, includes (possibly embedded in PA) the working memory load.

It is worth highlighting that although the final model is able to explain 66% of the variance in school performance, there is still some unexplained variance. In other words, approximately 44% of the variance in school performance could not be explained by intelligence, oral language, and EF. This result supports the conclusion reported by Oliveira et al. (2009) that there are many cognitive components involved in the learning process and thus in school performance. Future studies should explore other possible variables that can explain student performance.

Therefore, the present study provides evidence that although general intelligence is related to school performance, oral language and EF seem to have a greater contribution to the performance of students at the end of elementary school (EF-I in Brazil). At this school level, in which the basic academic skills are still being consolidated, children with better PA, vocabulary, and cognitive flexibility skills perform more competently at school and get higher grades. However, it should be pointed out that the present study had some limitations such as small number of participants, evaluation of only one school grade, and type of school. Thus, caution is required in generalizing the findings since aspects of the development and specific demands of the school grade investigated can influence these results. Therefore, future studies should extend this research to encompass different age groups and grade levels. Another limitation is related to the variables selected for this study, especially oral language, since PA and vocabulary may have greater effect during earlier grade levels. Accordingly, it is important that future studies extend these findings and investigate the role of other skills, such as syntactic and discourse skills, in school performance. Despite these limitations, the results found lead to the development of some hypotheses that can be tested in future research, such as that better oral language skills could lead to greater reading proficiency, and that greater cognitive flexibility will help students adequately cope with the demands of school work, resulting in higher grades.

Experimental studies should confirm this assumption; however these findings can assist in the design and implementation of intervention programs aimed at promoting cognitive development and impacting school performance. Thus, in addition to intelligence, these programs should include the promotion of language and executive skills. Similarly, the evaluation of these skills can be used to identify children that are at risk of school failure. These issues can be addressed and more extensively explored in future research.

Given the evidence of poor performance of Brazilian students, the present study can provide a useful contribution to this area of research and to teachers and other professionals involved in education by identifying important skills related to school performance at the end of elementary school (EF-I in Brazil). This study showed that intelligence, oral language, and EF are associated with school performance. Moreover, it was also found that oral language, more specifically PA and vocabulary, and EF, especially cognitive flexibility, are stronger predictors of school performance of 5th graders than intelligence. Thus, it can be said that these skills should be the target of intervention programs during the first years of formal schooling. This becomes even more interesting due to the availability of intervention programs to promote these skills, including in Brazil (Dias & Mecca, 2015).

Accordingly, it is suggested that professionals devote closer attention to the stimulation and development of these skills. As for the researchers, new research questions are proposed for future studies: (e.g., Can other language skills, such as discursive skills, more complex ones, be more strongly associated to school performance? What skills are the most important and in which levels of education? Which are the long-term effects of stimulation of language and executive skills in childhood?). A more extensive body of evidence can, in the future, guide the development of public policies and practices to promote cognitive development and, consequently, impact student performance.