Real-time word recognition in challenging listening conditions is defined by two primary dimensions, each of which exhibit within-subjectstability and link to outcomes

All listeners must contend with the fact that speech is often heard in sub-optimal, challenging conditions. While much is known about the compensatory auditory mechanisms listeners might use in these contexts, these may not be universal. Spoken word recognition offers an intriguing locus for context-dependent adaptation that could apply to all challenging conditions; indeed, research has found that spoken word recognition changes in these contexts, but what these changes represent and whether they have meaningful effects on outcomes was unknown. Here, we build on recent work by applying a Principal Component Analysis approach to eye-tracking data, demonstrating that word recognition primarily varies along two dimensions across different types of challenging listening conditions (noise and vocoding). Furthermore, we show that individual listeners are fairly consistent in how they configure word recognition across challenge types, and that these configurations are predicted by domain-general factors. Finally, we find that these changes to word recognition offer indirect benefits for performance outcomes, but only for a subset of listeners. Together, this work provides insight into how word recognition operates in and compensates for challenging contexts.