A model for chromoluminance pattern detection and pedestal effects is descr
ibed. This model has five stages. The stimulus is first processed by the co
ne array and then by color-spatial linear operators. The outputs of the lin
ear operators may be expressed as weighted sums of cone contrasts over spac
e. There are three opposite sign-pairs of linear spatial operators in the m
odel. Their spectral:tuning at each point in space is similar to the lumina
nce, green/red and blue/yellow mechanisms in color opponent models, but the
ir sensitivity to cone inputs varies as a function of space. The operators
in each pair are the same except that the signs of the cone inputs in one a
re the opposite of those in the other. A non-linear response operator follo
ws each linear operator. It receives two inputs, one excitatory and the oth
er divisive inhibitory. The excitatory input is the half-wave rectified out
put of one of the linear operators. The inhibitory input is a non-linear su
m of all linear operator outputs. The non-linear response operator raises t
he excitatory input to a power, and divides it by the inhibitory input plus
a constant to produce the response. The detection variable is computed by
combining the difference in response to target-plus-pedestal and pedestal a
lone across the three non-linear operators. The model accounts well for the
large data set presented in the companion paper and is generally consisten
t with other results in the literature. The spectral sensitivities of,the i
nferred chromoluminance pattern mechanisms are similar to those obtained wi
th different methods. The data set is shown to be inconsistent with several
other models. (C) 2000 Elsevier Science Ltd. All rights reserved.