- A form of publication bias - studies that fail in some respect to meet their author's expectations in respect, for example, of problem solved, or statistically significant results obtained. They are less likely to be submitted for publication than those that are more 'positive'. It is sometimes an indication of a crippling perfectionism on the part of the author. In other cases, extreme perversity can result. For example, suppose that the null hypothesis (i.e. that the association being studied does not exist) is indeed the case. But 5 per cent of studies show by chance that a statistically significant relationship exists. These are submitted - and published - leaving the 95 per cent of studies in which the null was not rejected gathering dust in researchers' file drawers and people doing systematic reviews concluding wrongly that the null is false. Publication bias occurs whenever the probability of a study being published depends on the statistical significance of its results and can have a seriously distorting effect on the conclusions of systematic reviews and meta-analyses.