Color polymorphisms have provided classical examples of how frequency-dependent selection maintains genetic variation in natural populations. Here we tested for the first time, the hypothesized adaptive function of a female-specific color polymorphism in odonates to lower male harassment towards females generally. Under conditions controlling for sex ratio, population density and morph frequency, we also tested two major frequency-dependent selection hypotheses for the maintenance of the polymorphism. Using groups of captive Enallagma hageni, whose females are either green or a male-like blue, we varied morph frequency at two sex ratios. We quantified sexual harassment towards females by visual observations, and by the presence of dust on females that was transferred from dusted males. Per capita harassment rate for the female-monomorphic treatments did not differ from that of the female-polymorphic treatments. At a male-biased sex ratio, per capita harassment rate towards blue, but not green females increased with morph frequency, providing partial support for frequency-dependent selection resulting from male learning of female morphs. Even at high frequency, green females were not harassed more than blue, contrary to the prediction that males should always recognize green females as mates. Moreover, frequency-dependent harassment towards blue females was not detectable using harassment measured with dust evidence, which greatly underestimated the incidence of sexual harassment. Our findings identified problems with the use of insectaries and the dusting technique to quantify male sexual harassment towards females, as well as with a past insectary experiment on Ischnura elegans that failed to demonstrate frequency-dependent harassment.