Parasitic-phase sea lampreys (Petromyzon marinus) are difficult to study in the wild. A series of laboratory studies (1984-1995) of single attacks on lake trout (Salvelinus namaycush), rainbow trout (Oncorhynchus mykiss), and burbot (Lota lota) examined host size selection; determined the effects of host size, host species, host strain, and temperature on host mortality; and estimated the weight of hosts killed per lamprey. Rainbow trout were more able and burbot less able to survive attacks than lake trout. Small sea lampreys actively selected the larger of two small hosts; larger sea lampreys attacked larger hosts in proportion to the hosts' body sizes, but actively avoided shorter hosts (a?? 600 mm) when larger were available. Host mortality was significantly less for larger (43-44%) than for smaller hosts (64%). However, the yearly loss of hosts per sea lamprey was less for small hosts (range, 6.8-14.2 kg per sea lamprey) than larger hosts (range, 11.4-19.3 kg per sea lamprey). Attacks at the lower of two temperature ranges (6.1-11.8A?C and 11.1-15.0A?C) did not significantly reduce the percentage of hosts killed (54% vs. 69%, p > 0.21), but longer attachment times at lower temperatures reduced the number of hosts attacked (33 vs. 45), and produced the lowest loss of hosts (6.6 kg per sea lamprey). Low temperature appeared to offset other factors that increase host mortality. Reanalysis of 789 attacks pooled from these studies, using forward stepwise logistic regression, also identified mean daily temperature as the dominant factor affecting host mortality. Observations in Lakes Superior, Huron, and Ontario support most laboratory results.