This is the dataset (available in .txt and .csv format) containing the data (demographic data and measures/scores) that go with the published paper Janse & Andringa (2021) on older adults’ recognition of words taken from fast and spectrally reduced speech. In this study, we investigated which cognitive, linguistic and hearing-related abilities predicted older adults' recognition of words taken from fast speech. We investigated through structural-equation modelling which predictors related to word recognition success. The fast speech was presented to older adults in two listening conditions: clear (unprocessed) and filtered (low-pass filtered speech). We also investigated whether word recognition success was differentially predicted by the individual abilities for the two listening conditions. The dataset therefore contains, for the 105 older participants in this study, their age, hearing ability, scores on several memory, processing speed and vocabulary measures, as well as their word recognition accuracy in clear and filtered conditions. As the paper is available to anyone (Open Access), more information on how the measures were calculated can be found there.