Visual search is often used to probe the temporal resolution of visual attention. Standard theories posit the involvement of a “saliency map” to select the target position and focus attention. This suggests that during serial (or “difficult”) search, there is iterative feedback from the saliency map to lower areas, reflecting the exploration of candidate locations by attention until the target is found. In a previous experiment (Dugué et al., PLOSOne, 2011), we applied double-pulses (25ms interval) of Transcranial Magnetic Stimulation (TMS) over V1 at different delays post-stimulus and demonstrated a specific effect on attentional target selection at 300-325ms. This delay was interpreted as the average delay necessary (in this specific task) to find the target. The aim of this new study is to determine whether V1 is involved at other delays during this task, confirming the presence of iterative feedback between V1 and higher-level areas. We applied doublepulses of TMS (sub-threshold): one pulse fixed at 312.5ms (between 300 and 325ms, see previous experiment) and the second one at variable post-stimulus delays from 112.5 to 437.5ms (25ms steps). Thanks to a phosphene-mapping procedure, the visual stimuli were presented either at the retinotopic location corresponding to the TMS-targeted area, or at the opposite location. Subjects (n=10) performed a serial search task: finding a T among Ls. We simultaneously recorded EEG to relate behavioural results to cortical oscillations. The target detection performance curves revealed a periodic modulation, with TMS-specific impairments recurring at post-stimulus delays compatible with theta frequencies (5.7Hz). EEG results confirmed that the phase of pre-stimulus spontaneous EEG oscillations in the theta frequency range (5-9Hz) covaried with target detection. These results argue in favour of a periodic causal influence of V1 during an attentional search, possibly reflecting successive cycles (~6Hz) of a ‘select-and-focus’ iterative loop between lower- and higher-level areas. Acknowledgement: This research was funded by a EURYI award to RV.