Dissecting the chain of information processing and its interplay with neurochemicals and fluid intelligence across development

1. Mean Drift Rate calculated across all the trials. 2. Boundary separation calculated across all the trials. 3. Non-decision time calculated across all the trials. 4. The three parameters of the alerting network (Task 1). The alerting network was calculated by subtracting the diffusion parameter during the double-cue condition from the same diffusion parameter in the no-cue condition. 5. The three parameters of the orienting network (Task 1). The orienting network was calculated by subtracting the diffusion parameter during the spatial cue condition from the same diffusion parameter in the central cue condition. 6. The three parameters of the executive network (Task 1). The executive network was calculated by subtracting the diffusion parameter during the congruent condition from the same diffusion parameter in the incongruent condition. 7. The three diffusion parameters of the distance effect obtained from contrasting “far” trials vs. “near” trials (Task 2). 8. The three diffusion parameters of the distance effect obtained from contrasting 135 degrees trials vs. 45 degrees trials (Task 3). 9. The three diffusion parameters of the SNARC effect obtained from contrasting low SNARC trials vs high SNARC trials (Task 2).

Regarding the column “Ord”, here we run three models (i) M1 which featured the intercept and the linear fit, (ii) M2 which featured the intercept, the linear fit and the quadratic fit and (iii) M3 which featured the intercept, the linear fit, the quadratic fit and the cubic fit. Following that, we assigned three p-values, (i) the p-value of the linear fit from M1, (ii) the p-value of the quadratic fit from M2, and (iii) the p-value of the cubic fit from M3. If none of these p-values was less than .05, the “Ord” value is N/A. If only the p-value of the linear fit from M1 is significant then “Ord” is 1, if the p-value of the quadratic fit from M2 is significant but the p-value of the cubic fit from M3 is not significant then “Ord” is 2, and if the p-value of the cubic fit from M3 is significant then “Ord” is 3. Essentially, the “Ord” column indicates the highest order fit that significantly contributes to the data above and beyond the less higher order fit/s. The variables (both for the first and the second assessment) are sorted based on the Pearson’s p-value of the first assessment.

The first column denotes the task (i.e., Task 1, Task 2, or Task 3), the second column denotes the assessment where “A1” concerns the first assessment analyses corresponding to Equation 3, “A2” concerns the second assessment analyses corresponding to Equation 4, and “Prediction” concerns predicting behaviour during the second assessment based on neurochemicals during the first assessment corresponding to Equation 5. The third column has three names separated by underscores, the first name corresponds to the region and the neurochemical used where IPS=intraparietal sulcus and MFG=middle frontal gyrus, and GLU=glutamate, GABA=gamma-Aminobutyric acid and NAA=N-acetylaspartate, the third name corresponds to the diffusion parameter that was used as the dependent variable, and the second name corresponds to the way each diffusion parameter was calculated where overall=the diffusion parameter was calculated across all the trials, AL=alerting network, OR=orienting network, EX=executive network, DISTANCE=the effect of distance, SNARC=the effect of SNARC. For the rest of the columns (df=degrees of freedom, β=standardized coefficient, P_BO=bootstrapped P-value), where β (column 5) and P_BO (column 6) correspond to the neurochemical*age interaction and β (column 7) and P_BO (column 8) correspond to the main effect of the neurochemical. The column “max_VIF” shows the maximum variance inflation factor assessing multicollinearity, and the column “SW” and “SW_P” shows the Shapiro-Wilk statistic and Shapiro-Wilk P-value, respectively, assessing the normality of the residuals in each model. The column “int_R²” is the adjusted R-squared of the model, and the column “non_int_R²” is the adjusted R-squared of the same model, but when omitting the interaction predictor and the column “delta_R²” is the difference between “int_R²” and “non_int_R²”. Of note, the variance inflation factor in the “Prediction” analyses was calculated, including all predictors apart from the predictor “age during the second assessment”, as this predictor is bound to be positively correlated to the predictor “age during the first assessment”.

To assess whether behaviour during the second assessment (i.e., A2) was predicted by neuroimaging measures during the first assessment (i.e., A1) while controlling for behaviour during the first assessment (i.e., A1), we employed SF6-eq1 as can be seen below which is a variant of (Equation 5) that additionally includes behaviour during the first assessment as can be seen below highlighted in bold. The first column denotes the task (i.e., Task 1, Task 2 or Task 3), the second column (i.e., “Prediction”) concerns predicting behaviour during the second assessment based on neurochemicals during the first assessment. The third column has three names separated by underscores, the first name corresponds to the region and the neurochemical used where IPS=intraparietal sulcus and MFG=middle frontal gyrus, and GLU=glutamate, GABA=gamma-Aminobutyric acid and NAA=N-acetylaspartate, the third name corresponds to the diffusion parameter that was used as the dependent variable, and the second name corresponds to the way each diffusion parameter was calculated where overall=the diffusion parameter was calculated across all trials, AL=alerting network, OR=orienting network, EX=executive network, DISTANCE=the effect of distance, SNARC=the effect of SNARC. For the rest of the columns (df=degrees of freedom, β=standardized coefficient, P_BO=bootstrapped P-value), where β (column 5) and P_BO (column 6) correspond to the neurochemical*age interaction and β (column 7) and P_BO (column 8) correspond to the main effect of the neurochemical. The column “int_R²” is the adjusted R-squared of the model, and the column “non_int_R²” is the adjusted R-squared of the same model, but when omitting the interaction predictor and the column “delta_R²” is the difference between “int_R²” and “non_int_R²”.

At the start of each entry, the dependent variable (Y), the independent variable (X), the mediator (M), and the moderator (W) are defined and highlighted in bold. 1. IPS GABA moderated mediation model with visuomotor connectivity as the mediator. 2. IPS glutamate moderated mediation model with visuomotor connectivity as the mediator. 3. IPS GABA moderated mediation model with visuomotor processing as the mediator. 4. IPS glutamate moderated mediation model with visuomotor processing as the mediator.

As expected, positive correlations of the three diffusion parameters were obtained across the three different tasks assessing attention (Task 1, ANT), digit comparison (Task 2, DC) and mental rotation (Task 3, MRT).

We conducted exploratory factor analyses by adding as input the non-decision time parameters of the three tasks (i.e., three variables as input) after controlling for age, and only one factor was extracted (i.e., eigenvalue>1). As can be seen in the table below, this factor (extraction method: Principal Component Analysis, rotation method: none) was consistently positively related to the non-decision time (Ter) of each of the three tasks.

These additional analyses support that both predictors (i.e., IPS glutamate*age and IPS GABA*age) were independently significant in tracking task performance (non-decision time in Task 1, Task 2 and when Task 1-3 were combined) even when both were added in the same multiple regression model.