Interpreting the declaration of Helsinki (2008): "must", "should" and different kinds of obligation.

The Declaration of Helsinki is widely regarded as the preeminent code of research ethics. Revised six times since 1964, the versions differ in their substantive requirements, and also in the way that obligations are expressed, especially regarding the use of the prescriptors "should" and "must". The 2000 version contained roughly two-thirds "should" versus one-third " must". But this ratio was inversed in the final 2008 version--although in its penultimate draft practically all occurrences of "must" had been replaced with "should". We consider and analyze the significance of these variations for policy and practice. We argue that the Declaration can plausibly be viewed as 'soft law'. In interpreting it in legislative and jurisdictional contexts the terms "should" and "must" cannot be seen as synonymous. Even if the soft-law claim is rejected, and the Declaration is viewed as providing ethical guidance only, the question of how to interpret "should" and "must" remains. We explore three possible interpretations: categorical versus hypothetical requirements; perfect versus imperfect obligations; and aspiration versus obligation. We conclude that the most plausible way of understanding the distinction is in relation to the strength of the categorical obligations which the Declaration seeks to set out.