"The tradition of identifying urban centres in different aspects of urban study is as old as urban modelling itself. The use of the attributes of the CBD and several subcentres instead of the characteristics of all land parcels or zones can be seen as a higher level of analysis. On the other hand, old technological limitations to consideration of smaller territorial units are being successfully overcome, and research efforts can be directed to more detailed spatial analysis. The question is: do we still need generalisation, i.e. urban centres identification, or it is preferable to operate at meso- or even micro-geographical levels in real estate prices modelling? We address this question at the geographical level of Traffic Analysis Zones. The paper is devoted to the formal identification of service employment centres in the Lyon Urban Area and their use in a hedonic model of apartment prices, which includes more than forty apartment attributes and location attributes as control variables. The application of the tradition of centres consideration is compared with an ìobjectiveî centrality index and a ìsubjectiveî accessibility index calculated for each zone. The aim is to find out, which of the three concepts better fits a regression model of apartment prices. Travel time by different transportation modes is considered. In hedonic modelling, both global and geographically weighted OLS regressions are used accompanied by measures of spatial autocorrelation and spatial variability. A surprising result is that the zone traditionally considered as being the CBD appears not to be as important for apartment prices as two of the identified subcentres. The other result is that the goodness-of-fit of both global and local hedonic models with these two subcentres is slightly better than that with an accessibility index, which in turn is slightly better than that with a centrality index.""