A survey for low-luminosity emission-line galaxies is described in whi
ch deep CCD images have been obtained through five narrow-band (DELTA(
lambda) = 80 angstrom) filters. The survey covers 0.7 square degrees t
o a depth corresponding to a monochromatic magnitude M6600 = 21.2 in t
he emission line plus continuum. The survey is sensitive to objects wi
th strong Halpha emission out to a redshift of z = 0.054, as well as h
igher redshift galaxies with [O III], Hbeta, or [O II] emission. Two h
undred ninety-three emission-line galaxy candidates were selected base
d on a (line plus continuum)/continuum ratio of 1.5 or greater. These
candidates were observed spectroscopically with a multifiber spectrogr
aph. Eleven low-redshift (z < 0.054) emission-line galaxies were confi
rmed with moderate to strong Halpha emission and M(B) in the range - 1
7.8 to - 12-9. Thirty-six additional objects are emission-line galaxie
s at higher redshift, including 11 in which only one emission line is
seen. Nine of the low-redshift emission-line galaxies form an essentia
lly complete sample, and two methods are explored to calculate the lum
inosity function. While the range of luminosity sampled is small enoug
h that trade-offs between slope and normalization of the luminosity fu
nction are possible, the combination of this survey with the study of
Salzer allows fairly restrictive limits on these two quantities. The b
est-fit Schechter function to the combination of these two data sets h
as slope alpha = - 1.35 and normalization, log phi = -3.15 mag-1 Mpc-
3. In heavy-element abundance, the objects found in this survey span t
he range defined by known dwarf emission-line galaxies, with the most
metal-poor object having an O/H abundance about 0.09 of the solar valu
e. Despite the fact that the survey technique employed is sensitive to
the presence of ultra-low metal-abundance galaxies, no such objects w
ere found. One group of higher redshift emission-line galaxies, those
with [O III] lambda5007 shifted into the Halpha region, has relevance
to the interpretation of faint blue galaxy counts. These objects are e
xamples of the type of galaxy that has been proposed to account for th
e excess counts at B = 22 and fainter. We find that the number discove
red in the survey is exactly that predicted by the local luminosity fu
nction, i.e., there is no evidence for strong evolution in the number
density of these objects, in disagreement with that interpretation.