In this paper, the importance of the interaction between theory, obser
vations and models for the analysis of climatic changes is illustrated
for three timescales: 10(4) yr (those of Quaternary ice ages), 10(2)
yr (the Holocene) and IO? yr (the Little Ice Age). Using paleoclimate
results to help explain past vegetation patterns, or using climates re
constructed from the fossil record to verify climate models produces t
he most acceptable results when attention is paid to the spatial and t
emporal scales of action of both systems. The long discussion of the p
ollen-climate ''equilibrium question'' showed that models underlying r
esults such as pollen diagrams, migration fronts or isopoll maps impos
e limits on the interpretations obtained from the fossil record. Our i
nterpretations of the sensitivity of vegetation to changes in climate
and the nature of climatic impacts on vegetation is determined in part
by the structures we impose on the data.