Objective To establish if MRI evidence of hippocampal atrophy (HcA) is an independent surrogate of EEG criteria for the diagnosis of medial temporal lobe (MTL) epilepsy (MTLE). Background MRI evidence of HcA has been shown to correlate with mesial temporal sclerosis (MTS), intracranial evidence of MTL seizure onset, and outcome after temporal lobectomy. The reported rate of discordance between scalp ictal EEG recordings and MRI evidence of unilateral HcA ranges from rare to moderate. We examined the surface and depth ictal EEG findings of patients with HcA, as detected by volumetric MRI, to clarify their significance in detecting areas of epileptogenicity in this group of patients. Methods From a group of patients with refractory epilepsy, we identified 119 patients with HcA (97 with unilateral and 13 with bilateral HcA, 9 with HcA and mass lesion). MRI volumetric studies were used to obtain Hc ratios. Absolute volumes were used to detect bilateral atrophy. Surface and depth EEG recordings were analyzed for localization of ictal abnormalities, and their distribution was compared for concordance with the location of HcA. Surgical outcome was reviewed. Results Of the 110 patients with isolated HcA, 63 had surgery; 82% of ictal depth EEG onsets were concordant with the atrophic Hc, and 72% ictal surface EEG onsets were concordant. Four patients with concordant EEG and HcA failed to achieve seizure control with resection of the atrophic Hc. Furthermore, 3 patients with discordant EEG and HcA had resection of the non-atrophic Hc with excellent results. Among the 47 non-operated patients, 54% had discordant or unlocalized ictal depth EEG results and 52% had discordant ictal surface EEG. Four of the 9 lesional patients with HcA had excellent outcome after lesionectomy without hippocampectomy. Conclusion The presence of HcA is not an independent predictor of the site of epileptogenesis.