When a briefly presented disk is followed by a briefly presented ring, obse
rvers may report not seeing the disk, an effect known as "metacontrast." Th
ree experiments showed that the ability of observers to distinguish such a
disk-ring pair from a flickering ring depended on how quickly they responde
d. When responding slowly, observers were least accurate at disk-ring stimu
lus onset asynchronies (SOAs) of about 40-50 ms. Shorter and longer SOAs re
sulted in improved performance. When responding quickly, however, performan
ce improved monotonically with SOA. These findings are discussed with refer
ence to a 2-stage model of metacontrast, in which early responses reflect a
stage of processing before surface completion is carried out (so only cont
our information is available). The technique of experimentally controlling
response latency may prove useful in tracing the temporal development of ot
her percepts.