Previous work on intrasentential codeswitching has noted that switches between determiners and their noun complements are frequent in both Spanish-English and Welsh-English data. Two major recent theories of codeswitching, the Matrix Language Frame model and a Minimalist Program approach, make potentially competing predictions regarding the source language of the determiner in these mixed nominal constructions. In this paper we evaluate the predictions of each theory with reference to comparable sets of Spanish-English and Welsh-English codeswitching data. Mixed nominal constructions are extracted to test the compatibility of these data with the predictions, taking into account coverage and accuracy. We find that the data are broadly consistent with each set of predictions but do not find statistically significant differences between the accuracy of the predictions of the two theories. We examine in detail the counterexamples to the predictions of each theory to see what further factors may influence codeswitching patterns between determiners and their nouns, and also discuss the differences in observed patterns in the data from each language pair. (Contains 5 notes and 2 tables.)