Traditionally, the neuronal representation of interaural timedifference (ITD) is often modeled using a delay-line mechanism. Incontrast to delay-line models which inherently decode ITD, themodel proposed here represents interaural phase differences in form ofthe relative activity between neuron populations located in the twobrain hemispheres. Therefore, the localization information is onlyencoded in specific activities in the two hemispheres. It cannot bedirectly determined but has to be decoded first.This study proposes a decoder based on likelihoodcalculations. Instead of decoding a single phase difference for agiven hemispheric activity-balance, the likelihood that the observedactivity was due to any possible phase difference is calculated. It isshown that applying the decoder separately to different frequencychannels, results in a visualization very similar to that of across-correlogram used in delay-line models. It is also shown that thedecoder is able to account for psychoacoustic data, including thebandwidth-dependent lateralization of 1.5 ms delayed noise thatpreviously imposed a problem on two-channel models.