Speech-in-noise research typically distinguishes between energetic masking (EM: interference between target and masker at the periphery) and informational masking (IM: interference higher in the auditory pathway). IM can itself be broken down into “lower-level” (e.g. acoustic, spatial) and “higher-level” (e.g. linguistic) IM. We use the term “informational interference” (inf-int) to refer to higher-level IM which is influenced by long-term linguistic knowledge.

Unlike EM, inf-int is poorly understood, in part because it is extremely difficult to manipulate inf-int without altering EM or lower-level IM. However, one apparently key feature of inf-int is that it often involves linguistic factors, such as whether or not a masker includes intelligible speech. It is generally believed that intelligible maskers, which involve both EM and IM, are more detrimental to target perception than acoustically-similar but unintelligible maskers, which involve only EM. However, this masker intelligibility effect has typically been demonstrated using connected speech maskers. As a result, it is difficult to determine which specific characteristics of the masker speech underlie any observed effects.

In the current series of studies (total N = 360), we tested the masker intelligibility effect using word list maskers. In their unmanipulated state, these maskers contain lexical and semantic information but lack syntax and sentence-level prosody. Our results suggest that intelligible word list maskers can actually be less detrimental to target perception (or at least no worse) than acoustically-matched, unintelligible equivalents, including time-reversed and noise-vocoded word lists. These findings suggest that the locus of inf-int is unlikely to be at the lexical-semantic level, and may instead reside in masker characteristics associated with connected speech, such as sentence-level syntax and/or prosody. By systematically varying our maskers along a range of parameters, we have taken steps towards isolating the specific characteristics contributing to inf-int. These results highlight the difficulty of consistently characterising and empirically quantifying IM during speech-in-noise listening.