Contamination of the environment by radionuclides is usually estimated using soil and grass sampling. However, radionuclides are often not homogeneously distributed in soils. In the alpine Mercantour region (Western Alps, France) a large heterogeneity in Chernobyl 137Cs deposition has been previously observed. Here we report additional 137Cs results together with new 90Sr and Pu data for soil, grass, milk, and cheese samples. The results show that radioisotopes from nuclear weapons tests fallout are more homogeneously distributed than Chernobyl 137Cs. Further, we observe that the 137Cs and 90Sr contents are less variable in milk samples than in grass or soil samples. This can be attributed to the homogenization effect of cow vagrancy during grazing. Hence milk seems to be a more robust sample than soil or grass to evaluate the extent of contamination on a regional scale. We explore this idea by comparing own unpublished 90Sr results and 90Sr results from the literature to establish the relationship between altitude of grazing and contamination of soil and milk for Western Europe. There is a significant positive correlation between soil contamination and altitude and an even closer correlation between milk 90Sr activity (A) and altitude (h): A = A0 + e(k x h) where A0 is the expected activity of milk sampled at sea level (A0 = 0.064 +/- 0.014 Bq g(-1) Ca) and h is the altitude of grazing, k being a constant (k = 0.95 x 10(-3) +/- 0.11 x 10(-3) m(-1) Bq g(-1) Ca). The fact that there is less scattering in the relationship for the 90Sr(milk)-altitude than for 90Sr(soil)-altitude suggests, again, that milk is a well-suited sample for environmental survey. The relationship between the altitude of grazing and the 90Sr content of milk and cheese can also be used to assess the authenticity of dairy products.