The responses of 30 V1 complex cells were recorded using a complete set of transiently presented, oriented stimuli of different contrasts. A back-propagation neural network was used to predict the multivariate visual stimuli from the neuronal responses on a trial-by-trial basis. For single neurons, the strength of the response was much better at predicting the orientation of a visual stimulus than its contrast. Using the temporal modulation of the response improved the ability to predict the contrast of a stimulus without affecting the ability to predict the orientation. Removing stimulus latency from the responses, by time-shifting the individual responses an amount equal to the average latency, significantly reduced the ability to predict stimulus contrast, demonstrating that the response latency is reliable enough, even for a single neuron and a single trial, for it to be used to help determine stimulus contrast. Pooling the responses from a group of 11 neurons demonstrated that small groups of neurons can accurately represent multivariate stimuli in a single trial.