This dataset includes the intelligibility responses of native (Canadian) and non-native (Dutch and Spanish) listeners to plain and Lombard speech produced by native and non-native English (native Dutch) speakers. The speakers produced utterances, from which we extracted keywords which were mixed with noise and presented to the listeners. The speakers produced plain speech (speech in quiet) and Lombard speech (speech in noise). This dataset was used to investigate the Lombard intelligibility benefit for non-native Lombard speech.