Atmospheric neutrinos, which are the decay products of hadrons
produced by cosmic ray interactions in the atmosphere, have
been detected in a number of underground experiments.
Although the predictions for individual flux components, i.e.,
and 
, are uncertain by
at least 20% [53], the ratio 
is much cleaner, with various calculations agreeing at the 5% level.
The Kamiokande and IMB experiments [54] have both observed a statistically significant deviation of r from the expected value, as indicated in Table 3. The value quoted is determined from the ratio of muons to electrons produced within the detector, compared to the theoretical expectation. The Soudan II data is consistent, though with larger statistical errors. Earlier results from Frejus and NUSEX do not show signs of a deviation, but again have large statistical uncertainties.
Table: Ratios 
observed by
recent experiments. The first (second) uncertainty is statistical
(systematic).
The small value of r observed by Kamiokande and IMB suggests
the possibility of the disappearance of 
or the
appearance of extra 
. In particular, the results could
be accounted for by 
or 
oscillations
with 
eV
and near
maximal mixing (
).
The oscillation interpretation has recently been supported by
the observation by Kamiokande of an anomaly in r for
multi-GeV events [55], which is consistent with their
earlier sub-GeV sample (and which, incidentally, excludes the
interesting possibility of a positron excess due to proton
decay [56], 
.).
Also, the multi-GeV data exhibit a zenith angle distribution
which is suggestive of oscillations, though the statistics are
not compelling.
However, there are
caveats. In particular, (a) the anomaly has not been observed by
all groups. (b) There are possible uncertainties due to
the interaction cross sections in the detector and particle identification.
However, at the energies involved it is unlikely that there would
be significant differences between the 
and 
cross sections, and the preliminary results from a KEK beam test do
not show any signs of particle mis-identification for Kamiokande.
(c) The IMB collaboration has also analyzed the ratio of throughgoing
to stopping muons. No anomaly is observed, excluding the lower
part of the 
range, e.g., 
eV
,
suggested by r.
(d) IMB has also measured the absolute flux of upward muons.
No anomaly was observed, nominally excluding the interesting
parameter range. However, this conclusion relies on the
theoretical calculation of the absolute 
flux, and also
involves uncertainties from the deep inelastic scattering cross
section [53].
One can regard the atmospheric neutrino anomaly
as a strong suggestion for neutrino oscillations. However, confirmation
will probably require long baseline oscillation experiments, which
are sensitive to small 
and large mixings.
Experiments sensitive to 
oscillations
are proposed or suggested for Brookhaven, Fermilab, CERN,
and KEK. There are also several proposals for long baseline experiments at
reactors, which, however, are only sensitive to 
disappearance.