Sensor-level network characteristics associated with arithmetic tasks varying in complexity were esti-mated using tools from modern network theory. EEG signals from children with math difficulties (MD)and typically achieving controls (NI) were analyzed using minimum spanning tree (MST) indices derivedfrom Phase Lag Index values – a graph method that corrects for comparison bias. Results demonstratedprogressive modulation of certain MST parameters with increased task difficulty. These findings wereconsistent with more distributed network activation in the theta band, and greater network integration(i.e., tighter communication between involved regions) in the alpha band as task demands increased.There was also evidence of stronger intraregional signal inter-dependencies in the higher frequencybands during the complex math task. Although these findings did not differ between groups, severalMST parameters were positively correlated with individual performance on psychometric math tasksinvolving similar operations, especially in the NI group. The findings support the potential utility of MSTanalyses to evaluate function-related electrocortical reactivity over a wide range of EEG frequencies inchildren.