Introduction: Activation maps acquired with perfusion-based FMRI are known to be more specific to the parenchyma and less sensitive to proximal draining veins compared to BOLD [1]. Here, the specificity of ASL activation maps in the nonhuman primate was investigated at high spatial resolution. Since the choice of labeling parameters can affect the specificity of the measured activation, the influence of the post-labeling-delay (PLD) and labeling-time (LT) on perfusion-based activation maps was evaluated. Methods: Measurements were performed on anesthetized monkeys using a vertical 7T/60cm Bruker Biospec system. Single-shot, multi-slice GE-EPI images were acquired at 0.75x0.9x2mm3 resolution (64x64x7, TR/TE=5300 /14ms). The preparation module for ASL consisted of a constant labeling period with a 2.5 mT/m gradient followed by a PLD of 200, 500, 800, and 1200ms. CBF images were derived by subtraction of label from control images and BOLD maps were derived from the control scans. For visual stimulation a flickering LED array was used (42s ON/ 42s OFF). Specificity was evaluated by calculating the activation profiles over the depth of cortex from maps of percent change and cross-correlation [2]. Results: The profiles for CBF percentage-change show a peak at 1.2 mm depth (Fig. 1a), which corresponds to anatomical layer IV. For increasing PLD, the profile of the functional activation narrowed and the activation at the surface decreased from 16.6±6% to 7.4±6% for PLD of 200ms and 1200ms, respectively (p<0.05). The BOLD percent-change profiles show largest activation at the surface (Fig. 1b). The profiles of the cross-correlation for fCBF and BOLD resembled a bellshaped curve with a peak at ~0.75mm depth (BOLD shown in Fig 1b). A Gaussian was fit to the profiles, resulting in a significantly smaller FWHM for CBF than BOLD (p<0.05). The difference in the percentage-change and cross-correlation profiles can be explained based on the profile of the coefficient of variation (CoV), (illustr