In this paper, I have shown that the scientific community has a deeply ambiguous attitude towards the presence of subjectivity in research. While both desiring and proclaiming objectivity, working scientists routinely use subjective criteria in their everyday research. The justification is pragmatic, and entirely reasonable: it is impossible for working scientists to deal with the plethora of new results and theories that constantly present themselves in any other way. However, mindful of past abuses in the history of science, the scientific community remains committed to keeping the presence of subjectivity in the research enterprise to a minimum.
This commitment has led to the widespread adoption of techniques for statistical inference that appear to be "objective". Known as frequentist methods, they have become central to the research enterprise, with their outcomes - P-values and 95 per cent confidence intervals - becoming a sine qua non for acceptance by leading science journals. As I have shown, however, these textbook methods are neither objective nor reliable indicators of either effect size or statistical significance of research findings. By failing to take into account the intrinsic plausibility of the hypothesis under test, frequentist methods are capable of greatly exaggerating both the size and the significance of effects which are in reality the product of mere chance.
The implicit recognition of these failings by scientific community is evidenced by the way in which essentially identical results from the supposedly "objective" frequentist methods are interpreted in entirely different ways, according to the subjective belief of researchers. Thus, a large and "highly statistically significant" result in parapsychology will be ignored, while a small and statistically non-significant link between passive smoking and cancer will be deemed to "add considerably" to the case against environmental tobacco smoke.
