More than 50 yr ago, field studies recorded the same-sex pairs (and trios) of penguins displaying to each other during the mating season, using behavior patterns typical of heterosexual mating displays. Such observations led to a hypothesis that due to a lack of sex recognition pairing occurs at random with respect to sex, an idea countered by the argument that sex recognition is highly accurate. No quantification of same-sex mating displays has tested the frequency of such displays in penguins or tested the hypothesis of random display partners with respect to sex. During their mating season, we studied displaying and paired king penguins, Apenodytes patagonicus, at Kerguelen Island and sexed them using a DNA marker, to quantify any occurrence of this behavior. Indeed, same-sex courtship displays were common (28.3% of 53 displaying pairs), the great majority of which were between males. Some homosexually displaying males eventually paired with females, but such males were significantly slower in heterosexual pairing than males that did not display homosexually. In two extraordinary cases, same-sex pairs learned each other’s calls, an essential step in the pairing process. The frequency of such pairs was much lower than among displaying couples, significantly so for males. Finally, the frequency of homosexually displaying pairs was significantly lower than expected from random assortment of displaying birds, for both males and females. We examined possible explanations for same-sex display and its biological significance. A population sex-ratio bias in favor of males and high concentration of male sex hormones may help to explain non-reproductive homosexually displaying pairs.