NINDS director Story Landis leads a very long author list on this Perspective piece in Nature, here. The article is open-access, so there should be no difficulty in taking a look, even for our readers without subscription. Note particularly the guidelines proposed in Box 1.
The piece comes about as a result of a meeting held at NIH in June of 2012, the aim of which, was to improve the reproducibility of findings from animal studies, particularly those used in preclinical research. The conclusions, while pretty predictable, are entirely appropriate–as a journal editor, I face the continuous challenge of getting our authors to avoid some of the obvious traps (like pseudo-replicates).
Note that these guidelines are not intended for the exploratory research that often precedes the design and execution of hypothesis-testing experiments that could be published.