An experimentalist's challenge: when artifacts of intervention interact with treatments

Abstract

A common, usually mphcit, assumptlon of the present epistemology of expenmental natural science is the following so long as any necessary expenrnental intervention is identically apphed to each treatment then the effects of arlfacts of that mtervention wdl be constant across all treatments This constancy of artlfacts of intervention thus allows unbiased assessment of differences between and among treatments because effects of any such artifacts contnbute equally to each treatment and therefore cancel out in contrasts Unfortunately, this assumphon will b e vlolated whenever a n art~fact of intervention and the expenmental treatment lnteract As one illustrahon of the lack of appreciation of this assumption, we review those manne ecological studies that employ tethenng of mobile prey organisms as a technique by which to assess the relative Intensity of predatlon as a function of changing habitat That review reveals that (1) only 55% of the 22 studies even include discuss~on of artlfacts of tethenng (2) only 9 % acknowledge the possibility that the magnitude or direction of the between-habltat difference in predation as estimated from mortality of prey on tethers could be Inaccurate if the artifactual enhancement of predatlon rate induced by tethenng is not constant across habltats and (3) no study actually tests the assumptlon that tethenng artlfacts are independent of habltat (the expenmental treatment) If different consumers are present in different proportions in the habltats being compared, as is often the case, it is possible and even likely that the magnitude of artifactual enhancement of predation induced by tethenng otherwise mobde organisms wdl fail to remaln constant across habitats This represents but one example of a general lack of recognition that artifacts of expenrnental intervention may interact with treatments a concern that applies also to use of enclosures and cages in field experiments and aquana and contamers in the laboratory Because the expenrnental intervention that induces the art~fact is typically essential for conduct of the expenment clever indlrect techniques may be needed to allow expenmentahsts to assess the importance of nonaddihve artlfacts of intervention