We present H I observations and analyses of the kinematics of 24 satellite-
primary galaxy pairs with projected separations between 4.9 and 240 kpc. Th
e satellites have masses of less than 3% of their primary spirals. Two esti
mates for the masses of the primaries are available, one from their rotatio
n curves and one from the orbital properties of the satellites. Defining ch
i as the ratio of these two mass estimates, it is a measure of the presence
, or absence, of a significant halo. The chi-distribution for these 24 pair
s is presented and the selection effects are discussed. Moreover, we show t
hat the chi-distribution of more numerous pairs, with projected separations
of less than 200 kpc, identified by Zaritsky et al., after adopting select
ion criteria quite different from ours, is similar to our chi-distribution.
We show that the observational biases have a negligible effect; the biased
and unbiased distributions are essentially identical. In order to understa
nd this distribution, N-body calculations were executed to simulate the dyn
amical behavior of relatively low mass satellites orbiting primary disk gal
axies with and without extended halos. The models and the real galaxies wer
e "observed" in the same fashion. In addition, we made a partially analytic
al analysis of the behavior of orbits in a logarithmic potential. We find t
hat a "generic" model, characterized by a single disk/halo combination, can
not reproduce the observed P(chi) distribution. However, a simple two-compo
nent population of galaxies, composed of not more than 60% with halos and 4
0% without halos, is successful, if galaxies have dimensions of order 200 k
pc. If galaxies are considerably larger with sizes extending to 400 kpc or
more, the constraints become more onerous. No generic model can describe th
e full range of the observed P(chi), particularly if the distribution for r
(p) < 200 kpc is compared with that for r(p) > 200 kpc. Regardless of the m
ix of orbital eccentricities, neither pure halo, nor canonical (disk and ha
lo masses are comparable within the disk radius) models will work. A multic
omponent approximation to reality can be constructed for which the canonica
l model must be mixed with a small fraction of systems essentially devoid o
f a massive dark halo. Only by including these complexities can the full ra
nge of P(chi) be modeled with any degree of success over all radial extents
. We show that dynamical friction cannot be ignored in these explorations a
nd that the average mass of a galaxy is in the range of (1-5) x 10(12) M.,
with a mass-to-luminosity ratio of at most a few hundred. This is insuffici
ent to close the universe.