When my wife, a rational agent, is preparing to welcome our monthly visitor, she is confronted by a preliminary wave of unwelcome hormones.  The proximate effect of these hormones is to make her a tad more grouchy than she normally is about otherwise mundane events. But because my wife is a rational agent, presumably she is able to forecast this effect and account for it.  In other words, when she has the impulse to feel perturbed about some minor calamity she reasons that her impulse is likely an artificial response brought on by the fluctuating chemistry in her brain.  And this reasoning would lead her to moderate her emotional response.

Indeed I have witnessed my dear wife execute exactly these calculations.  When this happens, the household is always most appreciative.  Sadly, it doesn’t always happen.

My theory therefore is that what is happening is not a uniform variation in the hormonal level, but rather a spike in hormonal volatility that forces my wife into a signal-extraction problem that is inherently prone to error.  For example, when her absent-minded husband leaves the lights on in the kitchen and she detects, in response, an oncoming alteration in her mood, she must determine whether this minor offence is something she would ordinarily be upset about.  My theory is that hormone volatility makes it hard to know exactly the current hormonal level and therefore difficult to back out the baseline appropriate degree of aggravation (in this case, i would argue, none at all.)

And thus hormones, an otherwise purely nominal variable, can have real (cyclical) effects.