Oxygen nonstoichiometry of perovskite-type oxide solid solution La1-xSrxMnO3 divided by d (x = 0-0.5) is summarized in a plot of oxygen content, 3 + d, vs. log P(O-2) at 873-1273 K. For x less than or equal to 0.4, two plateaus are observed in the plot, one around the stoichiometric point as to the lattice site, i.e. (3 + d) = 3.00, and the other at low temperatures and high P(O-2) at which (3 + d) was saturated. At the latter plateau, the mean Mn valence was +3.3-3.4 independent of x and the saturated oxygen content decreased Linearly with increasing x. A defect model is proposed to rationalize the (3 + d) vs. log P(O-2) relationship as well as the reported electronic properties between the two plateaus, which is based on the following assumptions. (i) Because the trivalent cation vacancies cause a large electronic imbalance and local lattice distortion, they do not stay close to each other and close to the Sr-La' site. To express the spacing among these defects, we introduce the concept of a vacancy excluding space around each cation vacancy and Sri,. (ii) With the formation of cation vacancies, a nonbonding oxygen 2p level is formed by the oxide ions around the vacancies. The formed level is narrow and the mobility of electrons in this level is low. We assume that the conductive e(g)(up arrow) level and the low mobility t(2g)(down arrow) level of Mn are very close to each other and that these levels are lower than the nonbonding oxygen 2p level. This nonbonding oxygen 2p level serves as the hole-trap. The (3 + d) vs, log P(O-2) relationships between the two plateaus can be explained by assuming that the volume of the vacancy excluding space around a trivalent cation vacancy and that around Sr-La' are nine and three unit cells, respectively, of pseudo-cubic perovskite-type ABO(3). Applying the proposed electronic structure, nonstoichiometry of the oxygen-deficient composition could be explained by the random distribution of oxide-ion vacancies. However, analysis based on the defect equilibrium could not identify the cation vacancy site, whether it is a La site or a Mn site or both a La and Mn site, because the results are almost identical whatever the cation vacancy site. Considering the reported neutron diffraction analysis results, we report a detailed analysis placing the cation vacancies predominantly on the La site.