Recent ISO data have allowed, for the first time, observationally based est
imates for source confusion in mid-infrared surveys. We use the extragalact
ic source counts from ISOCAM in conjunction with K-band counts to predict t
he confusion resulting from galaxies in deep midinfrared observations. We s
pecifically concentrate on the near-future Space Infrared Telescope Facilit
y (SIRTF) mission, and calculate expected confusion for the Infrared Array
Camera (IRAC) on board SIRTF. A defining scientific goal of the IRAC instru
ment will be the study of high-redshift galaxies using a deep, confusion-li
mited wide-field survey at 3-10 mum. A deep survey can reach 3-mu Jy source
s with reasonable confidence in the shorter wavelength IRAC bands. Truly co
nfusion-limited images with the 8 mum will be difficult to obtain because o
f practical time constraints, unless infrared galaxies exhibit very strong
evolution beyond the deepest current observations. We find L* galaxies to b
e detectable to z = 3-3.5 at 8 mum, which is slightly more pessimistic than
found in 1999 by Simpson & Eisenhardt.