Colour adaptation modifies the long-wave versus middle-wave cone weights and temporal phases in human luminance (but not red-green) mechanism.

1. The human luminance (LUM) mechanism detects rapid flicker and motion, responding to a linear sum of contrast signals, L' and M', from the long-wave (L) and middle-wave (M) cones. The red-green mechanism detects hue variations, responding to a linear difference of L' and M' contrast signals. 2. The two detection mechanisms were isolated to assess how chromatic adaptation affects summation of L' and M' signals in each mechanism. On coloured background (from blue to red), we measured, as a function of temporal frequency, both the relative temporal phase of the L' and M' signals producing optimal summation and the relative L' and M' contrast weights of the signals (at the optimal phase for summation). 3. Within the red-green mechanism at 6 Hz, the phase shift between the L' and M' signals was negligible on each coloured field, and the L' and M' contrast weights were equal and of opposite sign. 4. Relative phase shifts between the L' and M' signals in the LUM mechanism were markedly affected by adapting field colour. For stimuli of 1 cycle deg-1 and 9 Hz, the temporal phase shift was zero on a green-yellow field (approximately 570 nm). On an orange field, the L' signal lagged M' by as much as 70 deg phase while on a green field M' lagged L' by as much as 70 deg. The asymmetric phase shift about yellow adaptation reveals a spectrally opponent process which controls the phase shift. The phase shift occurs at an early site, for colour adaptation of the other eye had no effect, and the phase shift measured monocularly was identical for flicker and motion, thus occurring before the motion signal is extracted (this requires an extra delay). 5. The L' versus M' phase shift in the LUM mechanism was generally greatest at intermediate temporal frequencies (4-12 Hz) and was small at high frequencies (20-25 Hz). The phase shift was greatest at low spatial frequencies and strongly reduced at high spatial frequencies (5 cycle deg-1), indicating that the receptive field surround of neurones is important for the phase shift. 6. These temporal phase shifts were confirmed by measuring motion contrast thresholds for drifting L cone and M cone gratings summed in different spatial phases. Owing to the large phase shifts on green or orange fields, the L and M components were detected about equally well by the LUM mechanism (at 1 cycle deg-1 and 9 Hz) when summed spatially in phase or in antiphase. Antiphase summation is typically thought to produce an equiluminant red-green grating. 7. At low spatial frequency, the relative L' and M' contrast weights in the LUM mechanism (assessed at the optimal phase for summation) changed strongly with field colour and temporal frequency. 8. The phase shifts and changing contrast weights were modelled with phasic retinal ganglion cells, with chromatic adaptation strongly modifying the receptive field surround. The cells summate L' and M' in their centre, while the surround L' and M' signals are both antagonistic to the centre for approximately 570 nm yellow adaptation. Green or orange adaptation is assumed to modify the L and M surround inputs, causing them to be opponent with respect to each other, but with reversed polarity on the green versus orange field (to explain the chromatic reversal of the phase shift). Large changes in the relative L' and M' weights on green versus orange fields indicate the clear presence of the spectrally opponent surround even at 20 Hz. The spectrally opponent surround appears sluggish, with a long delay (approximately 20 ms) relative to the centre.