AK Herculis is a contact binary of spectral class F with a number of obvious peculiarities such as a displacement of secondary eclipse from phase 0 . P 500, unequal heights of the maxima, and a possibly sinusoidal (≃60 yr) period variation. The light curve is variable and shows erratic short-term behavior, although the latter effects are not unusual for a short period contact binary. Here we present new photoelectric light curves in theB andV passbands and derive several new times of minima. The new observations are compared graphically with seven earlier light curves. We analyzed the new observations and those by Bookmyer (1972) by the method of differential corrections. We find that the temperature difference between components is too large (≃1000 K) for AK Her to be considered a normal W UMa-system, but too small to allow a detached configuration obeying the mass-luminosity law. The solutions of both series of observations clearly indicate a contact configuration. Other unusual findings include a bolometric albedo greater than 2, and rather low values for the limb-darkening coefficients. We offer some possible reasons for these peculiarities. The mass ratio seems well determined from the photometry and is about 0.26, which is much lower than assumed in most earlier papers. The brightness and color of the known visual companion, ADS 10408B, were found by the differential corrections program. They agree, within observational error, with values found directly at the 200-in. telescope by Eggen, despite the fact that the B component contributes only a few percent of the light of the multiple system. The suggestion by Schmidt and Herczeg (1959) that a fourth body is responsible for the sinusoidal period variation seems stronger now than when first proposed. To encourage a search for this hypothetical component, we have tabulated expected values for its angular separation and relative mass with respect to the eclipsing pair, as well as the amplitude of the expected variable proper motion of the eclipsing pair. Discovery as a visual component seems impossible with present means, but the variable proper motion should be detectable. Thus we suggest that ADS 10408A (AK Her) be observed astrometrically. If a search for the predicted variable proper motion proves negative, mass transfer would remain the only plausible mechanism to explain the period change. We advance a different reason than did Schmidt and Herczeg to explain the phase lag of secondary eclipse.