We report on our search for microlensing toward the Large Magellanic Cloud
(LMC). Analysis of 5.7 yr of photometry on 11.9 million stars in the LMC re
veals 13-17 microlensing events. A detailed treatment of our detection effi
ciency shows that this is significantly more than the similar to2-4 events
expected from lensing by known stellar populations. The timescales ((t) ove
r cap) of the events range from 34 to 230 days. We estimate the microlensin
g optical depth toward the LMC from events with 2 < <(t)over cap> < 400 day
s to be <tau>(400)(2) = 1.2(-0.3)(+0.4) x 10(-7), with an additional 20% to
30% of systematic error. The spatial distribution of events is mildly inco
nsistent with LMC/LMC disk self-lensing, but is consistent with an extended
lens distribution such as a Milky Way or LMC halo. Interpreted in the cont
ext of a Galactic dark matter halo, consisting partially of compact objects
, a maximum-likelihood analysis gives a MACHO halo fraction of 20% for a ty
pical halo model with a 95% confidence interval of 8%-50%. A 100% MACHO hal
o is ruled out at the 95% confidence level for all except our most extreme
halo model. Interpreted as a Galactic halo population, the most likely MACH
O mass is between 0.15 and 0.9 Mo, depending on the halo model, and the tot
al mass in MACHOs out to 50 kpc is found to be 9(-3)(+4) x 10(10) M., indep
endent of the halo model. These results are marginally consistent with our
previous results, but are lower by about a factor of 2. This is mostly due
to Poisson noise, because with 3.4 times more exposure and increased sensit
ivity to long-timescale events, we did not find the expected factor of simi
lar to4 more events. In addition to a larger data set, this work also inclu
des an improved efficiency determination, improved likelihood analysis, and
more thorough testing of systematic errors, especially with respect to the
treatment of potential backgrounds to microlensing. We note that an import
ant source of background are supernovae (SNe) in galaxies behind the LMC.