Neurones with low best frequency (less than 2 kHz) and sensitive to interaural phase differences were recorded in the dorsal part of the central nucleus of the cat inferior colliculus. Best frequency tone (S) and N0ise (N) bursts were delivered binaurally via closed field sound systems either in phase (S0, N0) at both ears or inverted at one ear (SII, NII). The responses to tone + noise bursts in the stimulus configurations S0N0, SIINII and SIIN0 and noise bursts (N0 and NII) were compared. The latter two tone + noise configurations (S0NII and SIIN0) give a binaural masking level difference (BMLD) of up to 15 dB in psychophysical experiments, i.e. an increase in noise level is necessary to mask perception of the tone. Most cells responded better to in phase stimuli (here called 0 cells). A minority of cells responded better to out of phase stimuli (here called II cells). Each cell's response was correlated with the level of acoustic stimulus (tone or masker) in the preferred configuration and not with the BMLD situation: for the 0 cells, the responses were therefore maximal for S0N0 and minimal for SIIN II. For II cells, the gradation was reversed: the response to SIINII was maximal and that to S0N0 minimal. For both populations, the responses to S0NII and SIIN0 were intermediate between the S0N0 and SIINII responses. Cells that responded best to S0NII or SIIN0, i.e. cells selectively coding BMLD, were not found. This was also true for the synchronized spike rates of those cells showing phase locked responses to the stimulus frequency. Some cells appeared to be strongly suppressed by the addition of an non-preferred masker (i.e. in the configuration that resulted in less response to a noise-alone burst; e.g. N II for the 0 cells). Other cells were more suppressed by the addition of a preferred masker (N0 for the 0 cells). The difference in the number of spikes evoked by the tone + N0ise and the N0ise burst was analyzed according to signal detection theory and neuronal masked threshold determined. Some 0 cells showed lower thresholds in the configuration S0NII whereas others had higher thresholds in this configuration. This correlated with the binaural suppression effects noted above: when the noise in the preferred configuration (N0) gave more suppression, the threshold was lower for S0N II; when NII gave more suppression the threshold was higher for S0NII. Over the whole population, these effects cancelled out and the neuronal threshold was N0t significantly affected by the BMLD configuration. A possible BMLD substrate was found in the signal to N0ise ratio of the cell responses. When the response to a tone + N0ise burst was divided by that to a noise alone burst for each stimulus configuration, the response to the preferred signal (e.g. S0 for 0 cells) was higher when the noise configuration producing a demasking in psychophysical experiments (NII for 0 cells) was applied. It seems that single cells in the central nucleus of the inferior colliculus cannot code BMLDs. This task would have to be carried out by higher order cells integrating the output of an array of 0 or II cells. CompariS0n of the signal/noise relationships in these two channels would allow the tone to be coded in either of the BMLD positive configurations S0N II or SIIN0.