OBJECTIVETo explore the influence of verbal working memory load on associative neural networks.

METHODSTwenty-one subjects were required to complete a verbal delayed matching-to-sample task under the condition of low (3 items) or high (5 items) working memory load (WML). The 19-channels event-related potentials (ERP) were analyzed with statistical parametric mapping.

RESULTSA significant difference in working memory capacity (WMC) was found between low WML and high WML groups [2.48∓0.30 vs 3.30∓0.76; t(20)=5.950, P=0.000]. Statistical parametric mapping revealed that during the encoding stage, the effects of WML appeared in succession in the right ventral attention network (rVAN), the dorsal attention network, and the language areas in the left hemisprere. During the maintenance stage, the effects WML occured in the rVAN acompanied by either DAN or left frontal-temporal regions. CONCLUSIONS;onclusions When the WML is beyond the WMC, the rVAN may participate in the prevention of interference among items and in the activation of long-term memory.