In the macaque monkey, both ventral premotor cortex (area F5) and primary motor cortex (M1) are part of a cortico-cortical circuit controlling grasp. F5 neurones encode the selection of various types of grasping actions, and also respond to visual presentation of graspable objects. F5 may exert its actions on the hand through cortico-cortical connections to M1, which has a major corticospinal output to hand and arm muscles. Here we compared local field potential (LFP) activity in F5 and M1 in an awake, behaving macaque performing a visuomotor task. LFPs provide a population measure of synaptic activity within an area. Surgeries to implant a head restraint device and cortical recording chambers were performed under deep anaesthesia, induced with 10 mg kg^-1 ketamine i.m. and maintained with 2-2.5% isoflurane in 50:50 O₂:N₂O; full aseptic procedures were observed and antibiotics and analgesics given postoperatively. The task was as follows: six differently shaped objects were mounted individually on horizontal, spring-loaded shuttles, on a carousel. When two touch pads were depressed by the monkey, one object was presented visually. After a variable delay, the monkey was cued to reach for, grasp, pull and hold the object for 1s. LFPs were recorded simultaneously on multiple single microelectrodes in both F5 and M1 during 30 recording sessions; in all 144 LFPs were analysed. During both object observation and object grasp LFP power was dominated by 18-35 Hz (beta) power. LFPs were classified using post hoc analysis of the total beta power present during both object observation and steady grasp, (Student-Neumann-Keuls test, P1 object), or untuned (no significant tuning for different objects). The total beta power in each LFP for each object was measured, and objects ranked from ‘best’ to ‘worst’, the best being the most preferred object. Both areas showed distinct object-related tuning. During object presentation, a higher proportion of F5 LFPs showed tuning than did those in M1. During grasp, this was reversed. In general, M1 LFPs were more likely to be highly tuned than F5 LFPs (all comparisons, χ² test, P<0.05). The two areas had similar patterns of object preferences as a whole, but each LFP recording showed a characteristic order. The order of preference for the objects was not conserved from observation to grasp, either for individual LFPs or for whole areas. Thus 18-35 Hz LFP power in F5 is object-tuned prior to movement onset as well as during movement execution, and this parallels single neurone data reported previously. F5 LFPs exhibit a broader range of preferred objects than M1 and more sites showed tuning during observation; although there is some pre-movement tuning in M1, the predominant effect is during grasp, when all M1 LFPs were tuned to some extent. These results are consistent with a role of F5 in selecting an appropriate action, which is executed via M1.