In an era in which mortality risk for complex cardiac surgery is reducing, there is an increased focus on reducing morbidity. Brain injury following cardiac surgery remains important and represents a spectrum through grave irrecoverable global injury, focal irreversible and reversible neurological deficit, transient post-operative neurological dysfunction and more subtle neurocognitive change [1,2]. In this issue, Miyairi et al. [3], report neurocognitive outcomes in two series of patients, 28 undergoing coronary artery bypass surgery (CABG) and 46 undergoing aortic arch surgery utilizing retrograde cerebral perfusion (RCP), of which 19 underwent RCP for durations O60 min. This study is the first of its kind to compare neurocognitive outcome between RCP and a standard CABG group, i.e. a group undergoing cardiopulmonary bypass but not requiring a period of hypothermic circulatory arrest. From our own experience [4,5] thoracic aortic patients are a difficult group to study neurocognitively and this cohort of 46 aortic patients is the largest group of RCP patients to be studied so far. The authors are, therefore, to be congratulated on the performance of a difficult clinical task. Moreover, their study demonstrates no mortality and no reported stroke or transient neurological dysfunction.The authors report a significantly worse neurocognitive outcome in patients undergoing RCP O60 min compared to CABG and equivalent outcome to CABG in patients undergoing shorter RCP durations. These data lead to two important inferences; first, we might infer that RCP !60 min is neurologically safe and causes no more brain injury than a standard CABG procedure, and second, we can infer that RCP cannot afford extended brain protection during hypothermic circulatory arrest (HCA) and that like HCA alone, the risk of brain injury following RCP increases after 60 min of arrest/RCP duration.Is the first inference correct? Miyairi et al. present their data as the difference in mean Z-scores between preand post-operative performance for each test performed. They do not report the group incidence of significant neurocognitive change by accepted criteria (e.g. O20% or O1 SD decline in 2 or more tests.), the group change in raw scores for each test or an overall neurocognitive change score based on the Z score analysis. A cursory inspection of the early testing interval for CABG and short-RCP groups suggest that the S-RCP group either deteriorated more or improved less in 9 of 11 tests and that overall neurocognitive deterioration was greater than for CABG (Fig. 1). Thus, the reason for failing to detect a difference at the early interval in S-RCP patients probably represents a type II statistical error in an underpowered study.Despite this reservation, there appears to be very little true difference, if any, in neurocognitive outcomes between CABG and S-RCP at 4-6 months. There is a clear suggestion that early change has improved in the S-RCP group but persists in the L-RCP group and may be irreversible; a finding supported by previous studie...