Noise Increases Listening Effort in Normal-Hearing Young Adults, Regardless of Working Memory Capacity

As listening conditions worsen (e.g. background noise increases), additional cognitive effort is required to process speech. The existing literature is mixed on whether and how cognitive traits like working memory capacity moderate the amount of effort that listeners must expend to successfully understand speech. Here, we validate a dual-task measure of listening effort (Experiment 1) and demonstrate that for normal-hearing young adults, effort increases as steady-state masking noise increases, but working memory capacity is unrelated to the amount of effort expended (Experiment 2). We propose that previous research may have overestimated the relationship between listening effort and working memory capacity by measuring listening effort using recall-based tasks. The present results suggest caution in making the general assumption that working memory capacity is related to the amount of effort expended during a listening task.

Node Ordering for Rescalable Network Summarization (or, the Apparent Magic of Word Frequency and Age of Acquisition in the Lexicon)

How can we “scale down” an n-node network G to a smaller network, with k « n nodes, so that G (approximately) maintains the important structural properties of G? There is a voluminous literature on many versions of this problem if k is given in advance, but one’s tolerance for approximation (and the resulting value of k) will vary. Here, then, we formulate a “rescalable” version of this approximation task for complex networks. Specifically, we propose a node ordering version of graph summarization: permute the nodes of G so that the subgraph induced by the first k nodes is a good size-k approximation of G, averaged over the full range of possible sizes k. We consider as a case study the phonological network of English words, and discover two natural word orders (word frequency and age of acquisition) that do a surprisingly good job of rescalably summarizing the lexicon.

What Accounts for Individual Differences in Susceptibility to the McGurk Effect?

The McGurk effect is a classic audiovisual speech illusion in which discrepant auditory and visual syllables can lead to a fused percept (e.g., an auditory /bɑ/ paired with a visual /gɑ/ often leads to the perception of /dɑ/). The McGurk effect is robust and easily replicated in pooled group data, but there is tremendous variability in the extent to which individual participants are susceptible to it. In some studies, the rate at which individuals report fusion responses ranges from 0% to 100%. Despite its widespread use in the audiovisual speech perception literature, the roots of the wide variability in McGurk susceptibility are largely unknown. This study evaluated whether several perceptual and cognitive traits are related to McGurk susceptibility through correlational analyses and mixed effects modeling. We found that an individual’s susceptibility to the McGurk effect was related to their ability to extract place of articulation information from the visual signal (i.e., a more fine-grained anal- ysis of lipreading ability), but not to scores on tasks measuring attentional control, processing speed, working memory capacity, or auditory perceptual gradiency. These results provide support for the claim that a small amount of the variability in susceptibility to the McGurk effect is attributable to lipreading skill. In contrast, cognitive and perceptual abilities that are commonly used predictors in individual differences studies do not appear to underlie susceptibility to the McGurk effect.

Measuring Listening Effort: Convergent Validity, Sensitivity, and Links With Cognitive and Personality Measures

Purpose: Listening effort (LE) describes the attentional or cognitive requirements for successful listening. Despite substantial theoretical and clinical interest in LE, inconsistent operationalization makes it difficult to make generalizations across studies. The aims of this large-scale validation study were to evaluate the convergent validity and sensitivity of commonly used measures of LE and assess how scores on those tasks relate to cognitive and personality variables. Method: Young adults with normal hearing (N = 111) completed 7 tasks designed to measure LE, 5 tests of cognitive ability, and 2 personality measures. Results: Scores on some behavioral LE tasks were moderately intercorrelated but were generally not correlated with subjective and physiological measures of LE, suggesting that these tasks may not be tapping into the same underlying construct. LE measures differed in their sensitivity to changes in signal-to-noise ratio and the extent to which they correlated with cognitive and personality variables. Conclusions: Given that LE measures do not show consistent, strong intercorrelations and differ in their relationships with cognitive and personality predictors, these findings suggest caution in generalizing across studies that use different measures of LE. The results also indicate that people with greater cognitive ability appear to use their resources more efficiently, thereby diminishing the detrimental effects associated with increased background noise during language processing.

Keep Listening: Grammatical Context Reduces But Does Not Eliminate Activation of Unexpected Words

To understand spoken language, listeners combine acoustic-phonetic input with expectations derived from context (Dahan & Magnuson, 2006). Eye-tracking studies on semantic context have demonstrated that the activation levels of competing lexical candidates depend on the relative strengths of the bottom-up input and top-down expectations (cf. Dahan & Tanenhaus, 2004). In the grammatical realm, however, graded effects of context on lexical competition have been predicted (Magnuson, Tanenhaus, & Aslin, 2008), but not demonstrated. In the current eye-tracking study, participants were presented with target words in grammatically unconstraining (e.g., “The word is . . . ”) or constraining (e.g., “They thought about the . . .”) contexts. In the grammatically constrained, identity-spliced trials, in which phonetic information from one token of the target was spliced into another token of the target, fixations to the competitor did not differ from those to distractors. However, in the grammatically constrained, cross-spliced trials, in which phonetic information from the competitor was cross-spliced into the target to increase bottom-up support for that competitor, participants fixated more on contextually inappropriate competitors than phonologically unrelated distractors, demonstrating that sufficiently strong acoustic-phonetic input can overcome contextual constraints. Thus, although grammatical context constrains lexical activation, listeners remain sensitive to the bottom-up input. Taken together, these results suggest that lexical activation is dependent upon the interplay of acoustic-phonetic input and top-down expectations derived from grammatical context.

Assessing the Effects of “Native Speaker” Status on Classic Findings in Speech Research

It is common practice in speech research to only sample participants who self-report being “native English speakers.” Although there is research on differences in language processing between native and non-native listeners (see Lecumberri et al., 2010 for a review), the majority of speech research that aims to establish general findings (e.g., testing models of spoken word recognition) only includes native speakers in their sample. Not only is the “native English speaker” criterion poorly defined, but it also excludes historically underrepresented groups from speech perception research, often without attention to whether this exclusion is likely to affect study outcomes. The purpose of this study is to empirically test whether and how using different inclusion criteria (“native English speakers” vs. “non-native English speakers”) affects several well-known phenomena in speech perception research. Five hundred participants completed word (N = 200) and sentence (N = 300) identification tasks in quiet and in moderate levels of background noise. Results indicate that multiple classic findings in speech perception research—including the effects of noise level, lexical density, and semantic context on speech intelligibility—persist regardless of “native English” speaking status. However, the magnitude of some of these effects differed across participant groups. Taken together, these results suggest that researchers should carefully consider whether L1/LX status is likely to affect outcomes, and make decisions about inclusion criteria on a study-by-study basis.