Spatial deployment of attention in visual search: new evidence against a strict parallel model