The mechanism of the irrelevant speech effect in reading: a systematic review and three-level meta-analysis.

Readers are frequently exposed to background speech while reading, which slows reading speed and reduces comprehension accuracy-a phenomenon known as the irrelevant speech effect (ISE). The Semantic Interference Hypothesis proposes that only intelligible speech disrupts reading, whereas the Phonological Interference Hypothesis predicts that both intelligible and unintelligible speech cause disruption. However, empirical findings have not provided unequivocal support for either account. The present study addressed this issue by using a three-level meta-analysis. After a systematic search and screening, 30 studies were included, yielding 606 effect sizes. The results showed that, compared to the silence condition, intelligible speech slowed down the whole reading processing and prolonged lexical processing. Unintelligible speech also slowed down reading and affected post-lexical stages of lexical processing, as reflected in longer go-past time and total fixation duration. Importantly, intelligible speech exerted a stronger disruptive effect than unintelligible speech, by increasing reading time and delaying lexical processing across all stages. These results suggest that both intelligible and unintelligible speech interfere with reading, but intelligible speech causes greater disruption than unintelligible speech. The Phonological and the Semantic Interference Hypotheses only provide partial accounts for these findings, whereas the Process Interference Hypothesis offers a more comprehensive explanation.