The traditional approach to toxicity testing is to consider dose (concentration)-effect relationships at arbitrarily fixed exposure durations which are supposed to reflect ‘acute’ or ‘chronic’ time scales. This approach measures the proportion of all exposed individuals responding by the end of different exposure times. Toxicological databases established in this way are collections of endpoint values obtained at fixed times of exposure. As such these values cannot be linked to make predictions for the wide range of exposures encountered by humans or in the environment. Thus, current toxicological risk assessment can be compromised by this approach to toxicity testing, as will be demonstrated in this paper, leading to serious underestimates of actual risk. This includes neonicotinoid insecticides and certain metallic compounds, which may require entirely new approaches.