Address correspondence to: Dr DR Welch, The Jake Gittlen Cancer Research Institute (C7810), Penn State University College of Medicine, 500 University Drive, Hershey, PA 17033–0850, USA. Tel:(+ 1) 717 531 5633; Fax:(+ 1) 717 531 5634; E-mail: drw9@ psu. edu. from development of metastasis. Not all cells within a tumor are able to metastasize (reviewed in [37]). Metastatic cells are a subset within a heterogeneous tumorigenic population that are endowed with additional capabilities to those required for uncontrolled growth. Therefore, to fully understand the metastatic cell, one must not only understand the steps involved in the conversion of cells from normal to neoplastic, but must superimpose the complexities of the many steps involved in metastatic cascade. It stands to reason, then, that if the objective is to understand metastasis, appropriate models designed expressly for this purpose will be required. Moreover, the appropriate use of these models is required. The purpose of this article is to highlight some of the idiosyncracies of metastasis assays and to identify potential pitfalls that can lead to spurious results and/or improper interpretation. This paper was prompted by an incident at a recent meeting. One of my postdoctoral fellows had developed a series of cell clones from a human malignant melanoma. One clone was particularly metastatic; so, shortly after publishing its description, several investigators requested it from us. One of them approached us at that meeting and said,‘Those cells you sent us don’t metastasize.’My postdoc was horrified at the thought that her results could not be reproduced. We knew that the cells had behaved as expected in replicate experiments in our lab as well as elsewhere. Nonetheless, upon our return, we immediately retested the cells in vivo. Fortunately, they behaved as expected. Subsequently, we found that the other investigator had slightly modified the conditions for growing the cells, maintaining the animals, length of assay and methods for quantifying metastases.‘Trivial’experimental parameters were altered by the other investigator with a catastrophic result. The investigator has since repeated the in vivo study according to our protocols and replicated our results. No other incident in the laboratory better illustrated the need to reproduce experiments exactly before changing a single parameter. This incident also highlighted how little we actually know about the many variables in metastasis research. Further, it exposed the hazards of assuming that those desiring to study metastasis understand the complexities involved. In this vein, I decided to compile a list of conditions that are critical to consider when planning or executing metastasis-related studies. This had been done by Josh Fidler in a chapter published in 1978 [38]. At that time, Fidler wrote,‘Too many times investigators cannot reproduce published results because of their failure to conform to exact [emphasis added] experimental conditions.’He then added that,‘Sometimes... subtle and totally unsuspected factors may profoundly influence the outcome of otherwise wellplanned, well-executed experiments.’Some things have not changed. Some details have been discovered in the interim and an updated review was needed. As I began this project, I was reminded of his exhortations to study the history of our field. In doing so, many mistakes could potentially be avoided. Or as Santayana put it,‘Those who cannot remember the past are condemned to repeat it.’[39] I hope that this attempt to summarize and to evaluate techniques employed for in vivo metastasis studies is useful for new and established investigators, just as Fidler’s paper was for me early …