Auditory lexical decision in the wild

The present report describes a version of the MALD (Massive Auditory Lexical Decision) database which investigates listener performance in the often crowded TELUS World of Science—Edmonton (TWOSE) on tablets with headphones, with most participants experiencing varying levels of uncontrolled distraction. A total of 533 listeners participated in the experiment (ages 4–86; 51% female; 81.8% native English). Stimuli were a randomly selected subset of 2000 words and 2000 pseudowords extracted from the MALD project (Tucker et al., 2018), recorded by a single speaker. The stimuli were further randomly divided into 20 lists, each containing 100 words and 100 pseudowords, with no practice stimuli. Each participant was presented with a single list, with the entire experimental session usually lasting between five and ten minutes. TWOSE-MALD participants, unsurprisingly, perform worse than listeners in a laboratory setting in terms of both accuracy and response latencies. The relationship between the standard predictors and response latency remains the same. We also find age related effects indicating that accuracy increases rapidly (from ages 4 to 14) and slowly plateaus, while response latencies rapidly decrease until participants reach their early twenties, after which a steady increase is noted as participants’ age increases.