We studied behavioral variant frontotemporal dementia (bvFTD) using object alternation (OA) as a novel probe of cognition. This task was adopted from animal models and is sensitive to ventrolateral-orbitofrontal and medial frontal function in humans. OA was administered to bvFTD patients, normal controls, and a dementia control group with Alzheimer disease (AD). Two other frontal lobe measures adopted from animal models were administered: delayed response (DR) and delayed alternation (DA). Brain volumes were measured using the semiautomatic brain region extraction method. Compared with the normal controls, bvFTD patients were significantly impaired on OA and DR. For OA and DR, sensitivities and specificities were 100% and 51.5% (cutoff=22.5 errors) and 9.5% and 98% (cutoff=1.5 errors), respectively. Negative predictive value (NPV) for OA was 100% at all prevalence rates. Comparing AD with bvFTD, there were no significant differences on OA, DR, or DA. Nevertheless, positive predictive value (PPV) and NPV were good at all prevalence rates for OA (cutoff=36.5 errors) and DA (cutoff=6 errors); PPV was good for DR (cutoff=9 errors). Error scores above cutoffs favored diagnosis of AD. Performance on OA was significantly related to medial frontal gray matter atrophy. OA, together with DR and DA, may facilitate assessment of bvFTD as a novel probe of medial frontal function.