Well they do, and they might even be seriously periodic - just strongly affected by sexual abuse/activity. It is difficult to obtain data for other reasons also; for example not wanting to remember and subsequently reflect upon experience (generally leading to conscious desire for objectification).
Observed periodicity appears to be monthly (and perhaps half-monthly). The weak periodicity observed could however be explained by other factors, including experience of ambient objectification, synchronisation with ambient female cycle, regular build up and subsequent release, etc. Selection bias is unlikely considering this is all of the data I have. There is a likely bias as resultant of the peak detection algorithms applied (stacking and centring of data). But it is definitely worth some more research. Unfortunately it is a more difficult experiment to conduct in this day and age. I wonder how many psychologists would have risked conducting such an experiment given their current definition of mental health. Well to be fair, perhaps there is good data out there already... I suspected they might be somewhat periodic years ago, but it has taken yet more years to start recording them. Recording a first data point is never very exciting (considering it means nothing by itself - it is just an isolated date).
There was a time in history when people didn't need to talk about this, let alone publish (- ever wondered why psychology started?), but with a non-functional state actively subjecting its younger population to abuse, it is probably worth recognising reality for what it is given it is so rarely encountered.
For reference, the average period is taken to be 30.05 days. This is measured peak to peak across a) all peaks > 1 WD per night (taking the final peak / maximum within each cycle) + b) isolated peaks (of 1 WD). Peak generation / stacking is applied within a 1 week interval (up to 7 days). Data is taken across 3 independent periods. Approximate data marked with tilde (~). It probably requires a K-S test versus a simulation of the average (or fourier analysis) to validate the likelihood of a coincidental versus real correspondence.
Long live intellectualisation.