A Cortical Surface-Based Meta-Analysis of Human Reasoning

Abstract

Recent advances in neuroimaging have augmented numerous findings in the human reasoning process but have yielded varying results. One possibility for this inconsistency is that reasoning is such an intricate cognitive process, involving attention, memory, executive functions, symbolic processing, and fluid intelligence, whereby various brain regions are inevitably implicated in orchestrating the process. Therefore, researchers have used meta-analyses for a better understanding of neural mechanisms of reasoning. However, previous meta-analysis techniques include weaknesses such as an inadequate representation of the cortical surface's highly folded geometry. Accordingly, we developed a new meta-analysis method called Bayesian meta-analysis of the cortical surface (BMACS). BMACS offers a fast, accurate, and accessible inference of the spatial patterns of cognitive processes from peak brain activations across studies by applying spatial point processes to the cortical surface. Using BMACS, we found that the common pattern of activations from inductive and deductive reasoning was colocalized with the multiple-demand system, indicating that reasoning is a high-level convergence of complex cognitive processes. We hope surface-based meta-analysis will be facilitated by BMACS, bringing more profound knowledge of various cognitive processes.

Keywords: Bayesian meta-analysis of the cortical surface (BMACS); functional magnetic resonance imaging; inductive and deductive reasoning; integrated nested Laplace approximation (INLA); log-Gaussian Cox process.

Figure 1

Overview of BMACS. (A) Searching the PubMed database and incorporating additional studies from previous meta-analyses, we first included 996 candidate studies for our meta-analysis of reasoning. (B) After following the suggestions by Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA; Moher et al. 2009), 76 studies were filtered for the meta-analysis. (C) The acquired foci were then located in the standard MNI coordinate system. (D) Subsequently, the foci were mapped onto the fsaverage surface for each reasoning process and hemisphere, respectively, resulting in the final dataset of 1413 foci from 74 studies (iL, inductive Left; iR, inductive Right; dL, deductive Left; dR, deductive Right). (E) Using LGCPs, BMACS estimated reasoning-specific maps such as inductive reasoning (orange-red) and deductive reasoning (teal-cyan). (F) Finally, the results were mapped onto the fs_LR surface (Van Essen et al. 2012) and thresholded using exceedance probability for visualization. Darker colors correspond to regions with above 95% exceedance probability (highly related regions), and lighter colors correspond to regions with above 50% exceedance probability (moderately related regions).

Figure 2

Reasoning-specific maps above 95% exceedance probability. The maps represent brain areas that are highly related to neural process of reasoning, which correspond to regions with above 95% exceedance probability. (A) The conjunction map of both inductive and deductive reasoning depicts the regions that were observed in both reasoning-specific maps. (B) The map illustrates cortical surface areas related to inductive reasoning. (C) The map highlights cortical surface areas related to deductive reasoning. Borders and names of parcels were indicated with 180 parcels per hemisphere, following the multimodal parcellation of the Human Connectome Project (Glasser et al. 2016a). In the Results section, we described regions following the anatomical labeling by Glasser et al. (2016a), where they grouped the 180 parcels into 22 broader regions for readers’ readability.

Figure 3

Reasoning-specific maps from MKDA. Colored areas represent the results specific to inductive reasoning (A), and deductive reasoning (B), being estimated from MKDA that were significant at P < 0.05. The results were mapped onto the fs_LR surface for visualization (Van Essen et al. 2012).

Figure 4

Reproducing the effects from a working memory dataset. The effects of covariates and distribution of random effects were reproduced using BMACS. (A) Sample size had no significant effect (: 0.01, 95% HDI: [−0.08, 0.09]) on working memory tasks. (B) Age was negatively correlated (: −0.15, 95% HDI: [−0.24, −0.06]) with performance on working memory tasks. (C) The distribution of random effects matched the distribution in the original study (Samartsidis et al. 2019). The red lines in (A) and (B) illustrate 95% HDIs reported in the original study. The red dot in (B) displays the posterior mean. The red dashed line in (C) represents the prior Gamma distribution used in the original study.

Figure 5

Successful recovery of true parameters. To test whether BMACS could accurately estimate the model parameters, we generated a new dataset from arbitrarily set parameters (i.e., true parameters) and attempted to recover these parameters using BMACS. Most of the parameters were recovered, showing that the true parameters (red lines) were located within the estimated 95% highest density intervals (black lines), except a slight underestimation of a variance parameter (variance.sf.dR). iL, inductive Left; iR, inductive Right; dL, deductive Left; dR, deductive Right; variance.sf/range.sf, variance and range parameters that describe the characteristics of spatial fields with regard to the Matérn covariance structure (Lindgren and Rue 2015).