Verbal memory deficits remain a major complication of dominant hemisphere temporal lobectomy for epilepsy. The extent of this deficit was assessed preoperatively and 1 month and 1 year postoperatively with the Wechsler Verbal Memory Scale (WMSV) in 14 adults undergoing left temporal lobectomy. Intraoperative localization of language and verbal memory was also performed by electrical stimulation mapping. The WMSV score decreased an average of 22% at 1 month (13 cases), and 11% at 1 year (10 cases), even though in the majority of cases the medial extent of the resections had been significantly modified as a result of preoperative memory changes in response to intracarotid amobarbital perfusion testing. Memory decline was greater in patients who were not seizure-free, and correlated with the lateral (but not the medial) extent of the resection. The memory deficit could be predicted intraoperatively with 80% accuracy from the relationship of the resection to sites identified by electrical stimulation mapping as essential to naming or input or storage aspects of memory. This technique was applied prospectively in two additional cases with left temporal epileptic foci and complete verbal memory loss with left hemisphere amobarbital inactivation. These resections were tailored to avoid the essential naming and memory sites; the WMSV score increased 1 month postoperatively in both cases. This study identifies a lateral cortical component for verbal memory. Sites essential for that component can be localized intraoperatively with stimulation mapping; when they are spared in a resection, verbal memory deficit following dominant hemisphere temporal lobectomy can be prevented even in high-risk cases.