Archaeological evidence relating to the Roman period was found in large numbers on tracks of the motorways A28, A16 and A29. All did not benefit from an excavation, even from an “evaluation” (assessment excavation), and the operations which resulted in a publication are rare. It remains of these important works an extensive, largely unknown and heterogenous documentation. This documentation was inventoried, formalized and criticized, in the frame of the Action Collective de Recherches « Quinze ans d’archéologie préventive sur les grands tracés lineaires en Picardie », a part of a large national archaeological project. We reported this preliminary work in the first chapter, by presenting the documentation, its state, its potential and its limits. The set, made up with 57 settlement sites, was considered a relatively representative sample of the regional settlement. We tried to characterize and distribute them between the five or six principal identified forms of occupation: villae, single farms, open habitat, ancillary settlements, or road establishments. Characterizing incompletely excavated sites was very difficult: one of the chosen ways was to make an inventory of the different types of feature: ditches, timber buildings -on posts and on more or less massive foundations-, stone buildings, cellars and little excavated store rooms, sunken huts, burning structures and burials (chapter 2). After a put into perspective in the regional history-graphical context, the last chapter sets out the main contributions of the ACR to the questions of pattern of occupation and its dynamics, to understanding the different forms of settlement: some -already well-known as villae- others -single farms, small hamlets, marginal settlement and agricultural appendices-, completely underestimated. Road landscape is the subject of a development to put back the few establishments discovered in a general prospect. The chapter ends with the presentation of the most meaningful results obtained in understanding the field-systems and their ditches, and by a prospect of what it is possible to say about the countryside landscapes, from environmental data.