Competing Speech Perception in Middle Age

Abstract

Purpose

This research forum article summarizes research from our laboratory that assessed middle-aged adults' ability to understand speech in the presence of competing talkers.

Method

The performance of middle-aged adults on laboratory-based speech understanding tasks was compared to that of younger and older adults.

Results

Decline in the ability to understand speech in complex listening environments can be demonstrated in midlife. The specific auditory and cognitive contributors to these problems have yet to be established.

Conclusion

There is evidence that the ability to understand a target speech message in the presence of competing speech messages changes relatively early in the aging process. The nature and impact of these changes warrant further investigation.

Older adults often experience substantial problems understanding speech when listening conditions deteriorate. Background noise makes communication in many situations difficult for older listeners, especially when this noise consists of voices of other talkers (e.g., Desjardins & Doherty, 2013; Helfer, Chevalier, & Freyman, 2010; Helfer & Freyman, 2008, 2014; Humes & Coughlin, 2009; Humes, Lee, & Coughlin, 2006; Lee & Humes, 2012; Rossi-Katz & Arehart, 2009; Tun & Wingfield, 1999). Competing speech situations require individuals to use both peripheral and cognitive resources in order to communicate successfully, as the multiple speech sources must be segregated, messages not of interest must be ignored, and attention often must be switched in midconversation.

When do these challenges begin? For most individuals, problems coping in complex listening situations come on gradually. It is not unusual for middle-aged adults to self-refer for hearing assessment, and in fact, people in this age range (even those who have normal pure-tone thresholds) report more hearing-related problems than do younger adults (Demeester et al., 2012). Results of large-scale speech-in-noise screening programs confirm these reports, as a relatively large proportion of middle-aged adults need more advantageous signal-to-noise ratios to achieve criterion levels of performance (Nachtegaal et al., 2009).

The source of functional midlife hearing problems has yet to be determined. Pure-tone threshold elevation often begins in middle age (e.g., Morrell, Gordon-Salant, Pearson, Brant, & Fozard, 1996), which undoubtedly contributes to problems understanding speech in any type of background noise. Changes in suprathreshold peripheral processing of sounds also could contribute to problems understanding speech. For example, suboptimal temporal processing may cause listeners to be less able to take advantage of fluctuations in a competing speech signal, leading to reduced ability to glimpse portions of the to-be-attended message (e.g., George, Festen, & Houtgast, 2006; Takahashi & Bacon, 1992). In fact, several studies have noted abnormal temporal processing in middle-aged listeners (e.g. Grose & Mamo, 2010; Leigh-Paffenroth & Elangovan, 2011; Lister, Besing, & Koehnke, 2002). The ability to segregate sounds (of obvious importance in competing speech situations) also may change by midlife (e.g., Alain & McDonald, 2007; Lewald & Hausmann, 2013). Additionally, cognitive changes that could influence the ability to attend to one message while ignoring another have been documented in middle-aged individuals (e.g., Gunstad et al., 2006).

Speech Perception in Middle-Aged Adults: Results From Our Laboratory

Study 1: Speech Perception and Temporal Processing

The first exploration of speech perception in middle age from our laboratory examined speech understanding and temporal processing in groups of younger (19–22 years) and middle-aged (45–55 years) women (Helfer & Vargo, 2009). Speech understanding was measured in the presence of two types of maskers: a steady-state speech-shaped noise and a competing speech masker consisting of two same-sex talkers. The target and masking signals were presented in both spatially coincident and spatially separated conditions. The Gaps-in-Noise test (Musiek et al., 2005) was used to assess temporal processing. Although performance between the two groups of participants was equivalent in the presence of steady-state noise and when target and masking speech were spatially separated, middle-aged individuals performed significantly more poorly than younger adults in the spatially coincident speech-in-speech condition. Notably, speech recognition scores obtained in that condition were significantly associated with performance on the Gaps-in-Noise test. While high-frequency pure-tone thresholds (which were slightly elevated in the middle-aged listeners, with an average threshold at 4–8 kHz of 17 dB HL) were significantly correlated with speech understanding in steady-state noise, they were not related to speech understanding in competing speech or to Gaps-in-Noise test performance.

Study 2: Hearing, Cognition, and Speech Perception

Another study from our laboratory was designed to examine connections among hearing thresholds, cognitive ability, and speech understanding in younger, middle-aged, and older listeners (Helfer & Freyman, 2014). We manipulated the number of same-sex masking talkers (one or two) as well as confusability of the target speech and masking speech (via altering the syntactic similarity between the to-be-attended and to-be-ignored speech streams). Cognitive abilities measured included auditory short-term and working memory, inhibitory ability, and processing speed. The middle-aged participants in this study performed more similarly to older participants for a single competing talker and more similarly to younger listeners in the two-talker masker condition. However, significant differences were found between middle-aged and younger listeners in all speech-masking conditions, indicating that age-related changes for this type of speech perception task occurred early in senescence. The effect of syntactic similarity was comparable across groups, suggesting that aging does not bring out increased susceptibility to this aspect of confusability between target and masker. Elevated high-frequency pure-tone thresholds, smaller auditory short-term memory spans, and slower processing speed were associated with poorer speech recognition performance among the middle-aged and older participants. Unexpectedly, the proportion of masker errors (that is, when participants responded with a word in the to-be-ignored sentence instead of one from the to-be-attended sentence) was greater for participants with better hearing, longer short-term memory spans, and better inhibitory ability. Of note was that self-perceived hearing problems (measured with selected questions from the Speech, Spatial and Qualities of Hearing Scale; Gatehouse & Noble, 2004) were greater for the middle-aged participants in this study as compared to the older participants.

Study 3: Lexical Factors in Competing Speech Perception

Our most recent work in this area (Helfer & Jesse, 2015) explored how lexical characteristics (specifically, neighborhood density and frequency of word occurrence) of to-be-attended speech (target) and to-be-ignored speech (masker) affected performance of younger, middle-aged, and older adults. Individuals in this study heard pairs of sentences in which the frequency or neighborhood density of words in the target or masking speech stream was manipulated. To-be-attended and to-be-ignored sentences were presented from a single loudspeaker located directly in front of the listener. Participants repeated back the target sentence, which was cued by both the first word (the target sentence always started with the name “Theo”; the masking sentence began with “Victor” or “Michael”) and the target talker's voice, which was consistent throughout the experiment. Participants also completed a battery of cognitive tests similar to the one described earlier for Study 2.

Results of this experiment confirmed that to-be-attended words that occurred frequently or were from sparse lexical neighborhoods were easier to understand than were rare words or those from dense lexical neighborhoods, even in the presence of a competing speech message. Previous studies (Dirks, Takayanagi, Moshfegh, Noffsinger, & Fausti, 2001; Revill & Spieler, 2012; Sommers & Danielson, 1999; Spieler & Balota, 2000; Taler, Aaron, Steinmetz, & Pisoni, 2010) have found that lexical frequency in to-be-attended speech is more important for older than for younger listeners. However, in our study, the effect of word frequency in the to-be-attended speech stream was most influential for middle-aged participants (as compared to either older or younger participants) and was related to selected cognitive abilities (specifically, inhibitory ability and processing speed). Remarkably, lexical characteristics of the masking speech stream also influenced performance, as speech recognition ability was poorer in the presence of to-be-ignored words that were easy to recognize (that is, that came from sparse lexical neighborhoods or that occurred frequently in the lexicon) as compared to performance with hard-to-recognize masking words.

Stimuli in this study had a consistent target talker and a consistent masking talker; hence, we expected that participants would be able to incidentally learn the two voices, leading to improved performance over the course of the study (e.g., Felty, Buchwald, & Pisoni, 2009; Johnsrude et al., 2013). Figure 1 shows percent correct performance as a function of trial, with data averaged over the first 50 trials, the middle 50 trials, and the last 50 trials. It can be observed that some learning occurred for all listener groups, particularly over the first 50 trials. Of note is that, during the initial trials of the study, the speech recognition ability of middle-aged listeners was inferior to that of younger participants listening at a 3-dB poorer signal-to-noise ratio; this difference disappeared as the experiment progressed. However, younger adults listening at the same signal-to-noise ratio as middle-aged participants scored consistently better throughout the experiment.

Figure 1.

Speech recognition performance in the presence of a single-talker speech masker, as a function of trial order. Older and middle-aged listeners had a signal-to-noise ratio of −1 dB, as did one group of younger participants (Y −1). The other group of younger participants had a signal-to-noise ratio of −4 dB (Y −4). Data are from Helfer and Jesse (2015).

Conclusion

Results of studies described in this research forum article, as well as work from other researchers (e.g., Başkent, van Engelshoven, & Galvin, 2014; Dubno et al., 2008; Ruggles, Bharadwaj, & Shinn-Cunningham, 2012), provide evidence that substantial changes in speech recognition ability can occur in midlife. These problems are most easily observed in difficult listening situations. The reasons behind this early aging effect have yet to be determined but are likely a combination of lack of audibility due to initial stages of presbycusis, changes in suprathreshold processing of speech sounds, and subtle cognitive decline.

Funding Statement

Work discussed in this research forum article was supported by National Institute on Deafness and Other Communication Disorders Grants R01 01625 (awarded to Richard Freyman) and R01 012057 (awarded to Karen S. Helfer).