This study investigates the potentials and limits of sociolinguistic research on language shift. Starting from a position that the ultimate goal of the research must be to create a general theory of language shift of predictive power, the author examines the explanatory potential of current mainstream research methodology now regarded as canonical in the practice of research. He argues for the view that, for the purposes of the research goal mentioned, the arsenal of social psychology may prove more fruitful than sociologically-based correlational-global analysis methodology. There are, however, two necessary conditions to this. On the one hand, we cannot be satisfied with a mere additive consideration of the "subjective" psychological factors in addition to the "objective" factors of language shift. Instead, there is a need for a general change in point of view. On the other hand, sociolinguistics needs to show greater care in treating terms, notions, and theories borrowed from social psychology in a methodologically more precise way than is reflected in today's research practice.