This study presents the results of two experiments conducted to investigate how the amount of variation between target and distractor objects in a visual scene influences referential overspecification. We hypothesized that as this variation gets higher, speakers tend to include more redundant information in their target descriptions. The results showed that this was indeed the case. We suggest that scene variation causes speakers to make use of quick heuristics when selecting the content of their referring expressions, and discuss the implications of these findings for computational models that automatically generate referring expressions.