Purpose: Shi and Sánchez (2010) developed models to predict the optimal test language for evaluating Spanish/English (S/E) bilinguals' word recognition. The current study intended to validate their conclusions in a separate bilingual listener sample. Method: Seventy normal-hearing S/E bilinguals varying in language profile were included. Participants were presented with English monosyllabic and Spanish bisyllabic words in quiet and in speech-spectrum noise (+6 and 0 dB SNR). Relative success on the 2 tests was indicated by the difference in z score for each test. Results: The current group of participants was comparable to participants in Shi and Sánchez (2010) in regard to their language background and test performance. Previously developed models fit current data well. Age of English acquisition (AOA[subscript E]) yielded more consistent models across test conditions than language dominance (LD). New models incorporating average proficiency rating (across domains) and relative proficiency rating (across languages) yielded best prediction. Conclusions: Shi and Sánchez's (2010) models incorporating key linguistic variables (AOA[subscript E], LD, and average/relative proficiency ratings) can help clinicians predict bilingual clients' relative success in word recognition in Spanish versus English. These models are appropriate for current clinical work with S/E bilinguals in metropolitan areas similar to New York City.