New scientific paradigms often generate an early wave of enthusiasm among researchers and a barrage of studies seeking to validate or refute the newly proposed idea. All else being equal, the strength and direction of the empirical evidence being published should not change over time, allowing one to assess the generality of the paradigm based on the gradual accumulation of evidence. Here, I examine the relationship between the magnitude of published quantitative estimates of parasite–induced changes in host behaviour and year of publication from the time the adaptive host manipulation hypothesis was first proposed. Two independent data sets were used, both originally gathered for other purposes. First, across 137 comparisons between the behaviour of infected and uninfected hosts, the estimated relative influence of parasites correlated negatively with year of publication. This effect was contingent upon the transmission mode of the parasites studied. The negative relationship was very strong among studies of parasites which benefit from host manipulation (transmission to the next host occurs by predation on an infected intermediate host), i.e. among studies which were explicit tests of the adaptive manipulation hypothesis. There was no correlation with year of publication among studies on other types of parasites which do not seem to receive benefits from host manipulation. Second, among 14 estimates of the relative, parasite–mediated increase in transmission rate (i.e. increases in predation rates by definitive hosts on intermediate hosts), the estimated influence of parasites again correlated negatively with year of publication. These results have several possible explanations, but tend to suggest biases with regard to what results are published through time as accepted paradigms changed.