Managers must understand the effects of stressors on ecosystems in order to identify thresholds of harm. To be meaningful, however, thresholds will usually need to be defined for situations where multiple stressors are operating. We investigated the individual and combined effects of the principal stressors (elevated nutrient concentration, increased streambed sediment cover, and decreased flow due to water abstraction) operating in native tussock grassland streams converted to pasture in New Zealand. We used three approaches – a survey of 32 small streams, manipulation of stressors in a factorial design in an experiment involving 9 real streams, and finally multiple stressor manipulations in an artificial channel experiment. We investigated the consequences for benthic community composition and our results pointed to particularly marked effects of increased sediment on the streambed. Most community metrics responded to at least one of the stressors. But in many cases we detected complex synergistic or antagonistic interactions among the stressors. It is not unusual, in other words, for the consequences of stressors to be unpredictable on the basis of knowledge of single effects. If managers only consider the effects of individual stressors their assessment of risk may be higher or lower than reality.