Taking an entirely different approach. Kress H980) shows how law itself is an important envi- ronmental attribute: law classifies space so that some activities are criminal only if they occur in certain places. These same distinctions also influence law enforcement activity so that crim- inal behavior carried out in private is much more likely to go unnoticed than is the same activity in a public space (cf. Stinchcombe I966). ln a different vein the introduction ofa behav- ioral approach to the geography of crime has served to open up not only some new data sources but also the possibility of integrated analysis of crime and control. Carter and Hill (I979). for example. have investigated the influ- cnce of the perceived risk of getting caught on burglars‘ selection of targets. Hakim and Ren- gert (l98l) have studied interjurisdictional dif- ferences in crime rates resulting from differ- ences among districts in police strategies (“crime spillover"). But these various analyses are still based upon traditional images of crimi- nality because they rely on official statistics to identify both crime rates and criminals. They remain rooted at a phenomenal level of analysis. and in the case of the work on crime spillover are restricted to statistical empiricism.