The temporal resolution of cochlear-nerve fibres in the European starling was determined with sinusoidally amplitude-modulated noise stimuli similar to those previously used in a psychoacoustic study in this species (Klump and Okanoya, 1991). Temporal modulation transfer curves (TMTFs) were constructed for cochlear afferents allowing a direct comparison with the starling's behavioural performance. On average, the neuron's detection of modulation was less sensitive than that obtained in the behavioural experiments, although the most sensitive cells approached the values determined psychophysically. The shapes of the neural TMTFs generally resembled low-pass or band-pass filter functions, and the shapes of the averaged neural functions were very similar to those obtained in the behavioural study for two different types of stimuli (gated and continuous carrier). Minimum integration times calculated from the upper cut-off frequency of the neural TMTFs had a median of 0.97 ms with a range of 0.25 to 15.9 ms. The relations between the minimum integration times and the tuning characteristics of the cells (tuning curve bandwidth, Q10 dB-value, high- and low-frequency slopes of the tuning curves) are discussed. Finally, we compare the TMTF data recorded in the starling auditory nerve with data from neurophysiological and behavioural observations on temporal resolution using other experimental paradigms in this and other vertebrate species.