General properties of spatial weights models, in particular Markovian
properties, are systematically investigated The role of stationary spa
tial distribution, interpretable as an importance-centrality or promin
ence index, is emphasized. Spatial interaction models, and among them
the gravity model, are classified with respect to the time reversal an
d aggregation invariance properties obeyed by the associated spatial w
eights. Nine examples, involving connectivity, flows and distance deca
y analysis, integral geometry, and Dirichlet-Voronoi tessellations ill
ustrate the main concepts, with a particular geometrical emphasis, and
show how traditional, heuristic ingredients aimed at defining spatial
weights can be recovered from general models.