Colour is more important than form in driving the scan path during active visual feature conjunction search. A model shows how feature-based attention in V4 provides information to LIP allowing it to represent behaviourally relevant locations and attract attention to target coloured locations. Attentional effects in V4 develop such that an initial spatial attention focus, scaled according to stimulus density, becomes object-based later in the response. Thus, this is the first model to replicate, with temporal precision, different attentional effects observed in single-cell studies of V4 and IT, in addition to the systems level scan path behaviour.