This paper discusses problems connected with the unplanned effects upon man, his foodor resource-organisms and wildlife, of releasing naturally occurring and man-made chemical substances into the environment through human activities. The chief reason for studying this subject is the danger to human health, well-being and amenity which may be caused by unforeseen side-effects of a chemical substance during its production, transport, use and disposal. Potential harm to man is normally perceived and assessed by society on economic, social and cultural grounds. These provide the motivation for scientific research on the effects of chemical substances on the environment. Ideally, scientists should be able to predict likely effects of a chemical directly on man or indirectly via food, crops, livestock, wildlife or climate before the substance is released, enabling a more realistic cost-benefit appraisal to be made. A first approximation to predicting a potentially harmful substance may involve the following criteria: toxic in larger amounts; biologically non-essential; accumulating in tissue with age; environmentally persistent; biochemically active; not stored in inactive deposits in tissues much faster than ingested; more harmful to certain genotypes than others of the target species; environmentally mobile in the biogeochemical cycles; not forming highly stable inert compounds in nature; increasing in the environment during the last 100 generations of the target organism. A good deal is already known about the effects of chemical substances on man during their production, transport and use, from the study of occupational medicine, toxicology, pharmacology, etc., and to a lesser extent their effects on resource-organisms, from veterinary science and plant pathology. Much less is known about the effects of a chemical upon wildlife species, following its disposal in the environment. This last area of knowledge needs improving because ecological cycles and food chains may deliver the potentially hazardous chemical from affected wildlife back to man. Moreover, wildlife can often be used as indicators of environmental states and trends for a potentially harmful substance, giving an early warning of future risks to man. We are also frequently ignorant of how far wildlife may be supportive to human well-being: as a food-base for an important resource-species, e.g. fisheries or grazing animals; as a key species maintaining the stability of economically valuable ecosystems; as predators of crop or livestock pests; as a species involved in mineral recycling or biodegradation; as an important amenity. Failure to recognize the mutually interactive roles of man, resource-species, wildlife organisms and climate in the biosphere and their different tolerances to chemical substances has hindered the development of a unitary environmental management policy embracing all four biosphere components. Although a good deal is already known about the influence of molecular structure on the toxicity to human beings of drugs and certain other chemicals, much less is known about the influence of molecular structure on the environmental persistence of a chemical. For wildlife, persistence is probably the most important criterion for predicting potential harm because there is inevitably some wild species or other which is sensitive to any compound and any persistent chemical, apparently harmless to a limited number of toxicity-test organisms, will eventually be delivered by biogeochemical cycles to a sensitive target-species in nature. This means that highly toxic, readily biodegradable substances may pose much less of an environmental problem, than a relatively harmless persistent chemical which may well damage a critical wild species. The study of chemical effects in the environment resolves itself into a study of (a) the levels of a substance accumulating in air, water, soils (including sediments) and biota (including man), and (b) when the threshold action-level has been reached, effects produced in biota which constitute a significant adverse response (i.e. environmental dose-response curve). In order to predict trends in levels of a chemical, much more information is needed about rates of injection, flow and partitioning between air, water, soils and biota; and loss via degradation (environmental balance-sheets). These dynamic phenomena are governed by the physico-chemical properties of the molecule. Fluid mechanics and meteorology may in future provide the conceptual and technical tools for producing predictive models of such systems. Most of our knowledge of effects derives from acute toxicology and medical studies on man, but since environmental effects are usually associated with chronic exposure, studies are being increasingly made of long-term continuous exposure to minute amounts of a chemical. The well-known difficulty of recognizing such effects when they occur in the field is aggravated by the fact that many of the effects are non-specific and are frequently swamped by similar effects deriving from exposure to such natural phenomena as famines, droughts, cold spells, etc. Even when a genuine effect is recognized, a candidate causal agent must be found and correlated with it. This process must be followed by experimental studies, unequivocally linking chemical cause and adverse biological effect. All three stages are difficult and costly, and it is not surprising that long delays are often experienced between the recognition of a significant adverse effect and a generally agreed chemical cause. There is often ample uncertainty to allow under-reaction as well as over-reaction to potential hazards, both backed up by 'scientific' evidence.