"Noisy, crowded offices" was fourth, reminding me of our previous conversations (about 2/3 of the way down this posting, and this posting too).
Back to that paper I mentioned yesterday, "Building an Information System for Collaborative Researchers: A Case Study from the Brain-tumor Research Domain" by Mamrak, Boyd, and Ordonez (Fresno State people can click here for full text). The paper's been a great source of quotes for me to draw on to illustrate problems with requirements (I quoted from it in my response to Rita Vick's article). Here's an excerpt from Mamrak et al.:
A third clinician was genuinely interested in using our information system. However, every time we met with him to determine his data input and query requirements, his needs would shift from our last meeting. This process extended over several years: with every presentation of a fast prototype, and even when the first version of the system was implemented and installed, at every meeting with him we came away with several pages of notes indicating changes that he still required. When we expressed our concerns to him about our inability to converge on a common understanding of his needs, he constantly assured us that we were 'getting very close' and that the next time we would surely have it 'right.'