The problems and questions are many:
1. The clinical symptoms are variable, difficult to quantitate, and enormous in number. The symptoms cover emotions, sexual feelings, mood states, behavioral changes, and somatic complaints. Despite multiple questionnaires, we are still not convinced that there exists a reliable, objective method for observing and measuring symptoms that are experienced internally, rather than manifested via external behavior.
2. The discrepancy between retrospective and prospective accounts regarding cyclic changes is now well-documented and recognized. Women use menses as a marker of time, and unpleasant, easily remembered experiences are attributed to an easily recognized signpost. If women in our culture have been conditioned to expect symptoms in the premenstrual phase and have been taught to expect fluid retention, pain, and emotional reactions, that is precisely what will be reported. Our lives are rhythmical. Day alternates with night. There are sleeping and waking, being hungry and being full, the circadian rhythms of our glands, and the ultimate rhythm: the sexual cycle. It is the most natural thing to seek a rhythm for our behavior.
The Ruble study is now a classic. In this study, 44 undergraduates at Princeton University were deliberately deceived about which phase of the menstrual cycle they were experiencing. A bogus electroencephalogram, complete with electrodes attached to the head, was heralded as a new technique capable of predicting the date of menstruation. Subjects were told they were either premenstrual (due in 1-2 days) or intermenstrual (due in 7-10 days). Only those women who were led to believe that their period would begin in 2 days reported significantly higher symptom ratings on pain, water retention, and eating habit changes. This was interpreted as a reflection of stereotypic expectations.
3. Is there a specific syndrome? A syndrome must have a specific pathophysiology; specific signs and symptoms can be documented; and a specific treatment achieves a beneficial response. Not a single one of these criteria can be met. One of the basic problems is that we have lumped everything into PMS, contain behavioral changes, somatic complaints, and psychological problems, implying the existence of a specific syndrome. Part of the problem is that all the tools of research reflect the way the author of the tool conceptualizes PMS, which in turn is based upon the background and training of the author.
4. The experimenter expectancy effect has to be properly controlled. Subjects tend to comply with what they deem to be the experimenter's hypothesis. This has been studied in regard to PMS. and no significant difference in PMS symptomatology can be demonstrated when the purpose of the study is disguised, and in addition the responses can be influenced by positive or negative manipulations. This relates to findings of negative mood changes when subjects are asked to assess their menstrual distress retro spectively.

Studies are complicated by high placebo responses. Clinical studies of premenstrual syndrome typically demonstrate a 30-50% response to placebo and, if a positive effect is anticipated by the subjects, up to 80%. Only well-designed, double-blind, placebo-controlled, randomized trials yield reliable data.