The flash-lag effect refers to the phenomenon in which a flash adjacent to a continuously moving object is perceived to lag behind it. To test three previously proposed hypotheses (motion extrapolation, positional averaging, and differential latency), a new stimulus configuration, to which the three hypotheses give different predictions, was introduced. Instead of continuous motion, a randomly jumping bar was used as the moving stimulus, relative to which the position of the flash was judged. The results were visualized as a spatiotemporal correlogram, in which the response to a flash was plotted at the space-time relative to the position and onset of the jumping bar. The actual human performance was not consistent with any of the original hypotheses. However, all the results were explained well if the differential latency was assumed to fluctuate considerably, its probability density function being approximated by Gaussian. Also, the model fit well with previously published data on the flash-lag effect.