How Speaker Configuration of Recorded Media Affects Subjective Ratings for Listeners With and Without Hearing Loss Authors: Todd A. Ricketts, Rebecca Krivoshey, Javier Santos-Garrido, Haiping Huang, Erin M. Picou Published: 2026 ABSTRACT Purpose:The purpose of this study was to evaluate the effects of loudspeaker configuration and relative dialogue-to-background sound levels on subjective ratings of speech understanding and sound quality of recorded media for listeners with and without hearing loss.Method:Three groups of 20 adults participated, including younger and older adults with normal hearing and older adults with mild sloping to moderately severe sensorineural hearing loss. Participants rated their perceived speech understanding and sound quality in response to 10-s-long recorded media clips in audio-only and audiovisual formats. Three loudspeaker configurations were used: television speakers, stereo speakers, and five-channel surround sound. The dialogue-to-background ratio was manipulated to be either approximately +7 or +12 dB.Results:Participants in all three groups rated speech understanding and sound quality as better with more favorable dialogue-to-background ratios. Older listeners with hearing loss consistently rated their speech understanding as lower than either group with normal hearing. For listeners with hearing loss, surround sound was consistently rated among the poorest, and the stereo speaker configuration was rated among the best. Furthermore, the sound quality disadvantages found for the television speakers in young listeners without hearing loss were generally not evident in older listeners with or without hearing loss.Conclusions:The results did not support the use of five-channel surround sound configuration as a method to improve perceived speech understanding in listeners with hearing loss. However, they do support individualized clinical recommendations, as the surround sound was not universally the lowest rated configuration for all participants.Supplemental Material: https://doi.org/10.23641/asha.32691750 The majority of adults with hearing loss are dissatisfied with their ability to understand dialogue in recorded media (e.g., television shows and movies) due to the deleterious effect of background sounds (e.g., audience laughter, crowd noise, sound effects, and background music; Armstrong, 2016; Mathers, 1991; Strelcyk & Singh, 2018). Furthermore, adults with greater degrees of self-reported hearing loss report more problems watching recorded media compared to those with self-reported normal hearing or less self-reported hearing loss (Armstrong, 2016; Strelcyk & Singh, 2018). Dissatisfaction while watching recorded media is a critical issue for general well-being because it is a common leisure activity for older adults. Survey data suggest that approximately 85% of older adults with hearing loss rate hearing well when watching television as “very important” or “extremely important” (Strelcyk & Singh, 2018). In addition, adults over 65 years of age spend three times more waking time watching television compared to younger adults (Depp et al., 2010). In a study of adults aged 25–85 years with mild-to-moderate sensorineural hearing loss, “television” was the most frequently reported listening situation over a 2-week evaluation period (Christensen et al., 2024). Unfortunately, current assistive technologies are not fully addressing the difficulties adults with hearing loss experience while watching recorded media at home (e.g., television or streamed movies). While hearing aid use can improve audibility, the majority of hearing aid owners report that they are no more than moderately satisfied with their devices when listening to the television (Picou, 2022; Strelcyk & Singh, 2018). Furthermore, adoption rates of wireless streamers designed to connect to media delivery systems (including television) are relatively low, although slowly growing (Picou, 2022). Therefore, other solutions might be necessary to address the listening needs of adults with hearing loss. One potential solution is to modify the audio configuration used to deliver sounds in recorded media. Two potential options for modifying the audio configurations of home media environments include changing the loudspeaker configuration and modifying the dialogue-to-background ratio (DBR). Loudspeaker Configuration Television sound delivery has evolved over time from simplistic mono or stereo speakers to sophisticated surround sound systems using more than 10 audio channels (Zhang et al., 2017). Most commercial media produced in the last several years have been initially mastered using one of several surround sound techniques (Gaston & Sanders, 2008; Zhang et al., 2017). Surround sound simply refers to sound mastery and delivery techniques that are intended to make the listener feel that they are surrounded by sound. Often, multiple sound channels are presented from multiple loudspeakers that surround a listener. For the past several years, one of the most commonly available surround sound configurations has been 5.1, with five channels of primary sound content intended to surround the listener (center, left, right, left rear, and right rear) and one low-frequency effects channel. The center channel is intended to be delivered from a loudspeaker directly in front of the listener and often carries much of the dialogue. The other four surrounding loudspeakers might also carry dialogue but are often primarily used to convey motion and present other sounds from spatially separate locations. The low-frequency effects channel carries very low frequency sound effects typically intended to be delivered through a subwoofer. In addition to the loudspeaker configuration (e.g., 5.1), surround sound methods are also commonly referred to by the specific codec applied (i.e., the specific encoding and decoding method used in the digital representation of the multichannel sound). Current popular codecs include DTS, Dolby Digital, Dolby Atmos, and DTS:X. In more recent years, novel sound delivery schemes have been developed, aimed at improving the listener's control over the relationship between dialogue and background sounds, potentially benefiting listeners with hearing loss (Armstrong, 2016; Fuchs & Oetting, 2013; Shirley et al., 2017; Shirley & Oldfield, 2015). One particularly promising scheme assigns dialogue and other sounds as defined “objects” that can be independently manipulated. Rather than only allowing speaker-specific level control, object-based audio allows manipulation of the relative level of individual sound objects. This level control is applied to objects even when they span multiple speakers while still maintaining the relative sound positional information (Armstrong, 2016; Coleman et al., 2018; Shirley et al., 2017; Shirley & Oldfield, 2015). This object-based manipulation of the relative levels of dialogue and nondialogue sounds presents a possible solution for consumers to directly improve DBRs. Examples of current object-based audio codecs include Dolby Atmos and DTS:X. While the majority of modern, commercially available, recorded media have audio that is recorded and mixed in surround sound, these signals can be downmixed into fewer channels, for example, stereo (two channels) or mono (a single channel). Because the audio output after downmixing is altered, the listening experience may also change. For example, previous work examined relative preferences of listeners with hearing loss for television dialogue presented via a dedicated center loudspeaker as part of a surround sound configuration compared with a downmixed stereo presentation via two loudspeakers (Shirley, 2013). On average, the speech clarity was rated slightly but significantly higher for the dedicated center channel loudspeaker used in surround sound than for a stereo presentation. Shirley attributed these differences to reduced crosstalk artifacts (i.e., frequency-specific errors in the stereo signal presentation timing relative to the intended presentation position). The physical loudspeaker components used can also affect the listening experience because of resultant system differences in dynamic range, frequency bandwidth, linear and nonlinear distortion, and so forth (Holman, 2012). Due to large differences in cost and convenience, as well as individual preferences, those watching recorded media at home might be listening to configurations ranging from elaborate home theater systems with multiple high-quality loudspeakers presented in sound-treated rooms to small, built-in stereo television speakers with limited bandwidth. Also common, but beyond the scope of this investigation, are “virtual surround sound” techniques, which provide the illusion of being surrounded by sound via loudspeakers built into a television or an external “sound bar.” A survey of 515 adults over the age of 60 years revealed that the majority (78%) reported using the default “built-in” speakers in the television and less than 10% reported using surround sound (Strelcyk & Singh, 2018). Typically, such default television loudspeakers are down-firing or backfiring when integrated into flat-panel televisions. One potential explanation for dissatisfaction with recorded media listening at home is the use of default television loudspeakers. It is unclear whether using loudspeaker configurations other than the default television loudspeakers (i.e., surround sound or separate stereo loudspeakers) can improve speech recognition and ratings of sound quality for recorded media. DBR Another way to improve access to recorded media at home could be to modify the ratio of the dialogue to the nondialogue (or background) sounds by either modifying how the materials are initially mastered or through user controls using one of the object-based codecs described above. The background sounds in recorded media would be expected to be a limiting factor for understanding dialogue (speech), especially for older adults with hearing loss. This speculation is based on evidence that adults with hearing loss (Dubno et al., 1984; Moore et al., 2014) and adults who are elderly (Helfer et al., 2017) have more difficulty understanding speech in noise than young adults with normal hearing. While background sounds have the potential to make it more difficult to access dialogue, full elimination of these sounds is not desirable because they contribute significantly to the total experience. For example, background sounds, such as film scores and soundtracks, human voices, and sound effects, have been shown to help the viewer feel immersed and make the experience feel more realistic (Larsson et al., 2010; Pavic et al., 2023). In addition, music in film contributes to the emotional response to the program, which serves to reinforce the narrative, in addition to the feeling and mood of the scene (Larsson et al., 2010; Pavic et al., 2023; Tan et al., 2013). Furthermore, some background sounds may enhance comprehension by drawing attention or reinforcing the story (Armstrong, 2016). However, the improvements in plot comprehension related to inclusion of nondialogue background sounds found in listeners with mild-to-moderate hearing loss are not typically present in listeners with more severe hearing loss (Shirley et al., 2017). Therefore, one goal of sound mixing for television and movies is to balance the importance of dialogue recognition against desired immersion and emotional impact for the total experience. However, it is assumed that mixing decisions are typically made for general audiences, without consideration of the potential effects of hearing loss that are present in many listeners. In response to expressed difficulties when listening to recorded media, providing a “clean audio” channel containing dialogue to assist listeners with difficulty hearing was first proposed more than 30 years ago (Meares, 1991). Since that time, there has been considerable work aimed at optimizing relative sound levels for listeners, commonly referred to as the “Clean Audio Project” (Armstrong, 2016; Fuchs & Oetting, 2013; Shirley & Ward, 2021). One goal of this work was to identify the preferred relative levels of the surround sound channels containing dialogue (primarily front) compared to the other sound channels for listeners with hearing loss. However, the outcome of this work has not been straightforward, in part, because of the great variety in the type and effects of sounds other than dialogue routinely included in recorded media. Given the variety of nondialogue sounds and their relative positive and negative effects on groups and individual listeners, it is perhaps not surprising that a clear definition of what constitutes the optimal “clean audio” signal has been elusive. Indeed, the work to date has demonstrated that the same relative sound levels (sometimes referred to as the “preferred mix ratio”) will not be optimal for all listeners and all listening situations; individual listeners sometimes exhibit different preferences even when intelligibility is not compromised (Fuchs & Oetting, 2013). Consequently, further exploration of preferences for the relative levels of dialogue and nondialogue sounds through currently common loudspeaker configurations is important to better inform personalization with the goal of addressing listening difficulties with recorded media. Purpose The purpose of this study was to evaluate the effects of loudspeaker configuration on subjective ratings of speech understanding and sound quality of recorded media for listeners with and without hearing loss. In this investigation, two speaker configurations that are quite different but commonly available were selected (flat-panel down-firing television speakers and a five-channel surround-sound system). A third speaker configuration (stereo presentation from two of the speakers and the amplifier used in the surround sound configuration) was also included to examine possible effects related to speaker type (i.e., driver size/bandwidth) and number of presentation channels, separately. The effects of DBR were also evaluated, focusing on two DBRs that were expected to result in relatively high ratings of speech recognition performance (+12 dB DBR) and also lower ratings of speech recognition performance (+7 dB DBR). This manipulation was applied because of the possibility that speaker configuration preference might interact with DBR (e.g., sound quality preference may be dominated by speech recognition at low DBRs and may be dominated by alternative factors at high DBRs if speech recognition is acceptable). In addition, this manipulation was of interest because of the ability for DBRs to be controlled (e.g., at the mixing level, indirectly through user manipulation of the relative center channel level or directly through advanced auditory object-based codecs). Because recorded media inherently often include visual components, study stimuli included recorded media with and without video (auditory–visual and auditory-only stimuli, respectively). Subjective ratings of speech recognition performance and sound quality were evaluated to capture two important elements that might be differentially affected by stimulus presentation choices. Participants included younger and older adults with normal hearing, in addition to older adults with hearing loss. Method Participants A total of 60 adults in three groups (each n = 20) participated in this study. The first two groups included younger (aged 18–35 years, M = 24.4 years, SD = 4.9) and older (aged 50–79 years, M = 69.8 years, SD = 5.4) listeners with normal-hearing thresholds defined as frequency-specific air-conduction thresholds of no worse than 25 dB HL from 500 to 4000 Hz. The third group included older adults (aged 50–81 years, M = 70.2 years, SD = 5.7) with mild sloping to moderately severe air-conduction hearing thresholds. Maximum better ear hearing thresholds through 4000 Hz were limited to no worse than 65 dB HL in this group during recruitment, because it was of interest to evaluate listeners who watch television without hearing aids; only participants who denied experience with hearing aids were included. The better ear thresholds for the two groups of older adults can be seen in Supplemental Material S1. All participants in the third group had maximum better ear hearing thresholds through 1000 Hz of 40 dB HL. Participants with hearing loss had normal middle ear pressure (−150 to 50 daPa) and normal middle ear compliance (0.3–1.8 ml), or an air–bone gap of less than 15 dB, consistent with sensorineural hearing loss. All participants were native speakers of American English by self-report. Study procedures were approved by the Vanderbilt University Institutional Review Board (No. 161938). Participants were paid for their time on a per session basis. Stimulus Selection, Preparation, and Presentation It was of interest to capture a range of different materials because the overall experience and preference for a particular loudspeaker configuration (e.g., stereo vs. surround sound) are complex and individualized. The listener experience can depend on many factors including the specific material, type of background sound, and even the affective (emotional) response to the material, as has been shown for listeners with normal hearing (Kerins, 2013; Moiragias & Mourjopoulos, 2023). Although controlled speech-in-noise tests would ensure consistent DBR within and across clips, commercial television shows and movie content, which provide word-to-word variability in talker and background sound type, were chosen to better preserve realism. Criteria for inclusion as test material included (a) continuous dialogue for at least 60 s during which the dialogue and nondialogue (background) levels were relatively constant (based on subjective judgment); (b) the source material was available in 5.1 surround format and included audible (target) dialogue only in the center channel; (c) root-mean-square (RMS) average dialogue levels in the center channel that were at least 14 dB higher than all other signals in the center channel; (d) inclusion of some dialogue with foreign accents; (e) a range of background sound types including music, sound effects, and background conversation; and (f) dialogue with male or female talkers with a variety of speaking rates. Use of a range of different materials was desirable to improve generalization of the results, in part, because foreign dialects, noise in the listening environment, poor diction, and high speech rates can also negatively affect dialogue recognition (Armstrong, 2016; Strelcyk & Singh, 2018). In order to assess sample eligibility and prepare the stimuli for possible inclusion in the final set of test stimuli, samples were obtained via the HDMI and optical audio outputs of a Sony PlayStation 3. The 5.1 audio and video output were digitally extracted via an Elgato HD60 S+ external capture card, Echo AudioFire 12 audio interface, and Vegas Pro video editing software (Version 14.0, Build 252, 64 bits). After screening more than 100 clips for possible inclusion from popular television and movie offerings available on commercial media streaming services, a total of twelve 60-s clips (six audio and six audiovisual) were digitized to form the final test materials. For each stimulus set (audio, audiovisual), the six clips had the general form of (a) British-accented male and female speech in background of classical music, (b) American-accented male speech in a background of modern music, (c) American-accented male and female speech in a background of modern music, (d) American-accented female speech in a background of cafeteria noise, (e) a conversation among several American-accented talkers in a noisy restaurant, and (f) a conversation among several British- and American-accented talkers in background noise. After the clips were selected and time edited to meet study criteria, the audio was modified via sound editing software (Adobe Audition, Version 2017.1.1, Build 10.1.1.11) to manipulate presentation level and DBR within and across all clips while maintaining the natural sound characteristics as much as possible. Quantification of DBR Since this study required manipulation of relative dialogue and background levels, the calibration was accomplished based on levels measured from concatenated and frequency-filtered versions of the test materials, which were then applied to final test materials that were not frequency filtered or concatenated (as described below). Specifically, signals were first bandpass filtered from 353 to 5657 Hz to eliminate the frequency ranges for which there is little important speech information. This filtering results in a frequency range encompassing approximately 89% of important speech information (ANSI/ASA S3.5-1997 [R2020; American National Standards Institute, 2020]). Then, any silence in excess of 5 ms was removed. Finally, the relative RMS levels of the dialogue in the center channel were compared to the nondialogue sounds from the other four channels. Segments without speech energy in the center channel were removed prior to making this comparison. The resulting RMS level comparison was referred to as DBR. Test Material Level Adjustment The test materials (unfiltered and previously unmodified) were then modified based on the measured DBRs. Specifically, the level of all sound channels in each clip was first adjusted in unison so that the loudness (Loudness Units relative to Full Scale re: ITU-R BS.1770-4; International Telecommunication Union [2015]) for the unaltered center channel (dialogue) was equivalent across clips. The low-frequency effects channel was then removed (the low-frequency range of these signals preclude a meaningful effect on dialogue audibility). The total (unfiltered) level of the other four channels was then adjusted to manipulate DBR as measured within approximately 2- to 3-s blocks while maintaining dialogue loudness. We intentionally allowed DBR to vary within these blocks to preserve some natural variation in DBR. Furthermore, the exact duration of these blocks varied somewhat because DBR was manipulated over entire dialogue words or phrases rather than artificially manipulating DBR midword. Test Conditions There were six total conditions (2 DBRs × 3 loudspeaker configurations) each for the audio-only and audiovisual presentations. The audio-only and audiovisual stimuli were different clips. During the audio-only stimulus presentations, no visual cues were provided to participants. During the audiovisual presentations, the television or movie scenes from the original clips were displayed. Pilot testing revealed that potential participants consistently provided subjective feedback after listening for 10 s or less. Consequently, 10-s segments were chosen for evaluation (necessitating samples based on a minimum of 60 s per clip for the six conditions). The level of the four nondialogue channels was adjusted, so the DBR was +7 or +12 dB. While the choice of these two DBRs was somewhat arbitrary, our pilot data revealed that both +7- and +12-dB DBRs were commonly observed in the media clips screened for potential study inclusion. In addition, these ratios fell well within the total range of 0 to +30 DBR observed during screening, and they avoided extremely challenging or extremely easy dialogue understanding conditions (based on pilot testing). For example, pilot work revealed that some listeners with hearing loss reported that DBRs poorer than +7 were unacceptably poor. The same DBRs were used for the audio-only and audiovisual presentations. Three loudspeaker configurations were evaluated: (a) embedded stereo television speakers (television), (b) external stereo speakers (stereo), and (c) a five-channel surround sound speaker system (surround sound). To accommodate the stereo presentation conditions, the final calibrated five-channel test materials were downmixed to stereo versions using the Adobe Audition sound editing software. Materials were then calibrated for presentation at a 70-dBA level of the dialogue for each loudspeaker configuration to ensure a loud but realistic television listening level and to also ensure that materials were equivalent in relative audibility across loudspeaker configurations. Test Environment The test materials were output using a laptop computer (Dell) and either five or two channels of an Echo FireWire12 audio interface. Video materials, when present, were output via HDMI to a Sharp Roku TV 5316X (Model LC-32LB481U) with software Version 7.0.0. Audio materials were output from the audio interface to the television via HDMI or to the direct audio input (passthrough) of a Yamaha Natural Sound AV Receiver (RX-V463) for stereo and surround sound loudspeaker configurations. Tannoy Precision 6P loudspeakers were used for presentation of the center channel materials and the stereo loudspeaker materials. Four loudspeakers (Definitive Technologies Model BP-2X) were utilized for presentation of the other four channels of materials for the five-channel surround sound conditions. All participants were evaluated in a double-walled sound-attenuating booth. Participants were seated in the center of the booth, with the television and loudspeakers placed approximately 42-in. from the center of the seating position. The television delivered the video signal for all relevant test conditions and was placed on a wooden shelf directly in front of the participants. The television included two down-firing speakers that were activated for the stereo television audio test conditions. All advanced audio processing features within the television were disabled (i.e., virtual surround sound), and sound delivery was set to standard stereo. Consistent with the recommended 5.1 surround sound configuration (ITU-R BS.775), four loudspeakers were placed at ±30° and ±110° angles relative to the listener, and the center channel speaker was placed directly in front of participants, directly below the television. Two loudspeakers placed at ±30° were used for the external stereo speaker conditions. Procedure Prior to data collection, participants provided informed consent and underwent hearing evaluation using pure-tone air-conduction audiometry and immittance measures. During data collection, the six audio-only clips or six audiovisual clips were each presented in random order within each of two blocks. The presentation block order (audio-only or audiovisual) was counterbalanced within each of the three participant groups. Within each clip, the patients listened to 10-s segments sequentially. The segments were presented in order; however, each of the six listening conditions (3 speaker configurations × 2 DBRs) was applied to each 10-s segment randomly, without replacement. A unique randomization was applied for every participant and condition. Consequently, each participant provided survey responses 12 times (corresponding to the six separate audio-only and audiovisual test clip materials). Prior to testing, participants were provided the following instructions: “You will hear short segments of a television show or a movie. Please listen to each segment carefully. As you're listening, think about how it sounds. How was the sound quality? How much of the speech of the main talker(s) could you understand? You will listen to 6 segments in a row. They are all from the same show or movie. You will be asked to rate your understanding of the speech and what you think of the sound quality after each segment. We will then go on to a different clip from a different show or movie and repeat the process a total of 12 times.” After listening to each 10-s segment, participants completed a subjective survey on a computer tablet regarding the sound quality of the segment and the ease of understanding of the target dialogue. The survey included the following questions: (a) How would you rate the overall sound quality? (from 1 to 13, with 1 being very good, 7 being average, and 13 being very bad) and (b) How much of the speech could you understand? (from 1 to 13, with 1 being all of it, 7 being about half of it, and 13 being none of it). No other labeling was provided. The scaling system was developed for this study as a compromise between the ease of use of a small number of categories and the desirable properties of scaling structures expected to support parametric analysis (Norman, 2010; Sullivan & Artino, 2013). Participants also answered questions related to their emotional response to sound following each presentation. Data Analysis The two subjective rating outcomes (sound quality and speech understanding) were analyzed separately for the audio-only and audiovisual presentation conditions after averaging subjective ratings within each participant, across all six clips. Prior to statistical analysis and averaging, the data were subjected to rationalized arcsine transform to normalize variance across the rating range as variance decreases for conditions for which ratings were near ceiling or floor (Studebaker, 1985). All analyses were conducted in R language for statistical computing (Version 4.1.3; R Core Team, 2022). Linear mixed-effects models were created using the “lmer” function of the lme4 package (Bates et al., 2015) and analyzed using the “anova” function in base R. For each outcome, full models were constructed with three fixed factors and their interactions: loudspeaker (television, stereo, surround sound), group (young normal hearing, older normal hearing, older hearing loss), DBR (+7 DBR, +12 DBR), and one random intercept (participant). Significant main effects and interactions were explored using the “emmeans” function of the emmeans package (Lenth, 2019), adjusting for false discovery rate (Benjamini & Hochberg, 1995). Results Sound Quality Audio-Only Conditions Analysis of the subjective ratings of sound quality in response to audio-only clips revealed significant main effects of speaker, F(2, 250) = 26.12, p < .0001, partial η2 = .17; DBR, F(1, 250) = 160.31, p < .0001, partial η2 = .39; and group, F(2, 50) = 7.74, p < .001, partial η2 = .24; in addition, there was a significant Speaker × Group interaction, F(4, 250) = 12.85, p < .0001, partial η2 = .17. The other two-way interactions (DBR × Group: F(2, 250) = 1.00, p = .37; DBR × Speaker: F(2, 250) = 2.62, p = .08) were not statistically significant, nor was the three-way interaction, F(4, 250) = 1.04, p = .39. These results indicate that DBR affected ratings of sound quality (approximately 12 rationalized arcsine units [rau] lower in the +7 dB than the +12 DBR), but the effects of group membership or loudspeaker configuration did not depend on the DBR. Given the 13-point scale and the expanded range provided by rau transform, 12 rau is approximately equivalent to a difference of one rating category. Figure 1 displays the average sound quality ratings for the three loudspeaker configurations for each of the three participant groups (audio-only presentation conditions). Follow-up testing of the significant Loudspeaker × Group interaction demonstrated that in young adults with normal hearing, sound quality was significantly poorer for the television loudspeakers than for either the stereo (estimated marginal mean difference [EMMD] = 25.3 rau, 95% confidence interval [CI] [18.5, 32.2], p < .0001) or surround sound (EMMD = 17.9 rau, 95% CI [11.1, 17.9], p < .0001). For older adults with normal hearing, sound quality was significantly better for stereo than surround sound (EMMD = 12.3 rau, 95% CI [4.3, 19.1], p < .0001). In contrast, there were no significant differences in sound quality ratings between any of the three loudspeaker conditions for older listeners with hearing loss. Mean subjective sound quality ratings in rationalized arcsine units (rau) for audio clips reflecting the significant Group × Loudspeaker interaction. All ratings for each participant were averaged across all six audio clips evaluated prior to analysis. ***p < .001. Audiovisual Conditions Analysis of the subjective ratings of sound quality in response to audiovisual clips revealed significant main effects of speaker, F(2, 260) = 15.28, p < .0001, partial η2 = .11, and DBR, F(1, 260) = 246.79, p < .00001, partial η2 = .49. In addition, there was a significant Speaker × Group interaction, F(4, 260) = 7.53, p < .0001, partial η2 = .10, and a significant Group × DBR interaction, F(2, 260) = 4.28, p = .015, partial η2 = .03. The other two-way interaction (DBR × Speaker: F(2, 260) = 2.72, p = .07) and the three-way interaction (F(4, 260) = .24, p = .92) were not statistically significant. Follow-up testing of the significant Speaker × Group interaction for the audiovisual conditions (see Figure 2, left panel) revealed similar trends to those found for the audio-only conditions, with a few exceptions. Specifically, young adults with normal hearing rated sound quality as significantly poorer for the television loudspeakers than for either stereo (EMMD = 15.5 rau, 95% CI [9, 22], p < .0001) or surround sound (EMMD = 13.3 rau, 95% CI [6.7, 19.8], p < .0001). However, the sound qualities for stereo and surround sound were not rated as significantly different from each other for young adults with normal hearing. Older listeners with normal hearing rated sound quality significantly poorer for surround sound than stereo (EMMD = 10.4 rau, 95% CI [3.6, 17.3], p = .0012). Older listeners with hearing loss also rated sound quality lower with surround sound than stereo (EMMD = 10.4 rau, 95% CI [3.3, 17.5], p = .0018). Mean subjective sound quality ratings in rationalized arcsine units (rau) for audiovisual clips. The left panel reflects the significant Loudspeaker × Dialogue-to-Background Ratio (DBR) interaction, and the right panel reflects the Group × DBR interactions. All ratings for each participant were averaged across all six auditory–visual clips evaluated prior to analysis. *p < .05, ***p < .001. Follow-up testing of the significant Group × DBR interaction (see Figure 2, right panel) demonstrated that sound quality was rated significantly better for the +12 DBR than the +7 DBR for all three groups of listeners (p < .0001); however, the magnitude of this difference was larger for older listeners with hearing loss than for either of the groups with normal hearing (EMMD = 27.2 rau compared to 18.5 and 18.7 rau for the older and the younger group, respectively). This result suggests that listeners with hearing loss benefit more from an improved DBR than do listeners with normal hearing in either age group. Subjective Speech Understanding Audio-Only Conditions The pattern of results for speech understanding was broadly consistent with the subjective ratings of sound quality described above. Analysis of the understanding ratings in the audio-only condition revealed significant main effects of loudspeaker, F(2, 250) = 20.52, p < .0001, partial η2 = .14; DBR, F(1, 250) = 234.07, p < .0001, partial η2 = .48; and group, F(2, 50) = 63.50, p < .0001, partial η2 = .72. In addition, there was a significant Loudspeaker × Group interaction, F(4, 250) = 4.08, p = .0032, partial η2 = .06, and a significant Speaker × Dialogue-to-Background interaction, F(2, 250) = 9.38, p < .0001, partial η2 = .07. The other two-way interaction (DBR × Group: F(2, 250) = .23, p = .79) and the three-way interaction (F(4, 250) = .75, p = .56) were not statistically significant. Follow-up testing of the significant Speaker × Group interaction (see Figure 3, left panel) demonstrated that young adults with normal hearing rated their understanding to be significantly better for the stereo than for either the television loudspeakers (EMMD = 13.3, 95% CI [6.8, 9.7], p < .0001) or surround sound (EMMD = 11.17 rau, 95% CI [4.7, 17.6], p < .0001). Older adults with normal hearing rated their understanding to be significantly lower for the surround sound than stereo (EMMD = 11.3 rau, 95% CI [4.8, 17.8], p < .0001). Older listeners with hearing loss rated their understanding to be significantly lower for surround sound than for either the television loudspeakers (EMMD = 8.7 rau, 95% CI [3.2, 15.3], p = .005) or stereo (EMMD = 7.43, 95% CI [2.7, 14.1], p = .01). Mean subjective speech understanding ratings in rationalized arcsine units (rau) for audio-only clips. The left panel reflects the significant Loudspeaker × Dialogue-to-Background Ratio (DBR) interaction, and the right panel reflects the Loudspeaker × DBR interactions. All ratings for each participant were averaged across all six audio clips evaluated prior to analysis. *p < .05, ***p < .001. Follow-up testing of the significant Speaker × DBR interaction (see Figure 3, right panel) demonstrated that understanding was rated significantly better for the +12 DBR than the +7 DBR for all loudspeaker conditions (p < .0001); however, the magnitude of this difference was smaller for the television loudspeakers than for either stereo or surround sound (EMMD = 11.7 rau vs. 23.6 and 23.2 rau for the stereo and surround sound loudspeakers, respectively). This result suggests that changing the DBR had a smaller effect on rated speech understanding when using television loudspeakers than when using the other two loudspeaker configurations when visual cues were not present. Examining this interaction with a focus on loudspeaker configuration, for the +7 DBR, rated understanding was significantly lower for surround sound than for either the television (p < .001, EMMD = 10 rau, 95% CI: 4.7 to 15.4) or stereo (p < .001, Mdifference = 9.8 rau, 95% CI [4.5, 15.1]). For the +12 DBR, rated understanding was significantly higher for stereo than for either the television loudspeakers (p < .001, EMMD = 11.6 rau, 95% CI [6.3, 16.9]) or surround sound (p < .001, Mdifference = 10.1 rau, 95% CI [4.8, 15.5]). These results suggest that ratings of speech understanding were lowest for surround sound when the DBR was poor and highest for stereo when the DBR was favorable. Audiovisual Conditions Analysis of the ratings of understanding in the audiovisual conditions revealed significant main effects of loudspeaker, F(2, 260) = 5.86, p = .0032, partial η2 = .04; DBR, F(2, 260) = 286.56, p < .0001, partial η2 = .47; and group, F(2, 52) = 23.14, p < .0001, partial η2 = .52. In addition, there was a significant Group × DBR interaction, F(2, 260) = 20.69, p < .0001, partial η2 = .14, and a significant Loudspeaker × DBR interaction, F(2, 260) = 3.79, p = .024, partial η2 = .03. The other two-way interaction (Loudspeaker × Group: F(2, 260) = 1.85, p = .11) and the three-way interaction (F(4, 260) = 1.67, p = .16) were not statistically significant. Follow-up testing of the significant Group × DBR interaction (see Figure 4, left panel) demonstrated that understanding was rated significantly better for the +12 DBR than the +7 DBR for all three groups of listeners (p < .0001); however, the magnitude of this difference was larger for older listeners with hearing loss than for either of the groups with normal hearing (EMMD = 24.5 rau vs. 11 and 12.1 rau for the older group and the younger group, respectively). This result is consistent with sound quality ratings of audiovisual stimuli and suggests that changing the DBR is most impactful for listeners with hearing loss. Mean subjective speech understanding ratings in rationalized arcsine units (rau) for audiovisual clips. The left panel reflects the significant Group × Dialogue-to-Background Ratio (DBR) interaction, and the right panel reflects the Loudspeaker × DBR interactions. All ratings for each participant were averaged across all six audio clips evaluated prior to analysis. ***p < .001. When examining this interaction with a focus on groups, rated understanding was significantly lower for older listeners with hearing loss than younger or older listeners without hearing loss (p = .0001); however, these differences were larger for the +7 DBR than the +12 DBR (EMMD = 32.5 and 32.0 rau vs. 20.1 and 18.5 rau for the +7 and +12 DBR, respectively). This result suggests that having unfavorable DBR was more impactful for older listeners with hearing loss than for either group with normal hearing. Follow-up testing of the significant Loudspeaker × DBR interaction (see Figure 4, right panel) demonstrated that significantly higher ratings of understanding for the +12 DBR over the +7 DBR conditions were measured for all loudspeaker configurations (p < .0001); however, differences were larger for surround sound than for either the television or stereo loudspeakers (EMMD = 19.3 rau vs. 15 and 13.2 rau for the television and stereo loudspeakers, respectively). This result suggests that changing the DBR had a larger effect on rated speech understanding when using surround sound than when using the other two loudspeaker configurations when visual cues were present. When instead examining this interaction with a focus on loudspeaker configuration, rated understanding was significantly lower for surround sound than for either the television loudspeakers (p = .01, EMMD = 4.2 rau, 95% CI [0.27, 8.1]) or stereo loudspeakers (p < .001, Mdifference = 7 rau, 95% CI [3.1, 10.9]), but only for the poorer (+7) DBR. Discussion The purpose of this study was to evaluate the effects of loudspeaker configuration on subjective ratings of speech understanding and sound quality of recorded media for listeners with and without hearing loss. The effect of changing the DBR and the potential interaction between DBR and loudspeaker configuration ratings were also evaluated to explore the potential for improving difficulties with speech recognition and sound quality of recorded media, particularly in listeners with hearing loss. The results demonstrated a few consistent and expected findings. Specifically, mean ratings were poorer for the lower (+7 dB) than the higher (+12 dB) DBRs. That is, all listener groups reported that lower levels of background sounds, relative to the dialogue levels, resulted in better speech understanding and better sound quality regardless of loudspeaker configuration and stimulus modality (audio-only or audiovisual clips). These findings are consistent with many past studies demonstrating decreases in speech recognition and sound quality with decreasing signal-to-noise ratios (e.g., B. C. J. Moore & Tan, 2003; Pavlovic, 1987). In addition and consistent with speech recognition performance data, the group of older listeners with hearing loss consistently rated speech understanding lower than either the younger or older listeners with hearing in the normal range (e.g., Dubno & Dirks, 1993). Other data trends in the current study related to loudspeaker configuration were more nuanced and depended on rating type (sound quality or perceived speech recognition) and, in some cases, DBR; each is discussed separately in the following sections. Sound Quality The subjective sound quality results revealed that young adults with normal hearing rated the sound quality for television loudspeakers significantly poorer than for either the stereo or surround sound conditions. This pattern was evident both with and without visual information present and for both DBRs evaluated. This finding is broadly consistent with previous data demonstrating that “natural” sound quality is inversely related to sound bandwidth and, in particular, to a lack of energy in the lower frequencies (B. C. J. Moore & Tan, 2003). The small loudspeakers commonly integrated into the flat-panel televisions often produce much lower sound levels in the lowest and highest frequencies (i.e., narrower bandwidth) in comparison to the much larger external loudspeakers (such as those used in the current study). In contrast to young listeners, however, older listeners with and without hearing loss rated the sound quality of the television loudspeakers similar to the stereo loudspeakers, both with and without visual information and across both DBRs. These findings support that the inferiority of the rated sound quality associated with built-in television speakers, as demonstrated by young adults with normal hearing, might not be evident in older listeners. Consequently, the use of larger, higher quality speakers with greater frequency bandwidth may not lead to improved sound quality for older adults when watching recorded media at home. Interestingly, despite the increased potential for immersion and previous work suggesting a preference for a dedicated center channel instead of listening in stereo (Shirley, 2013), the results of this study did not generally support sound quality advantages for a surround sound configuration compared to stereo loudspeakers for any of the listener groups. Indeed, when significant differences were present (e.g., all auditory–visual conditions), the sound quality ratings were higher with stereo than with surround sound loudspeakers. The reasons for this somewhat surprising result are unclear but might be partially attributable to some methodological choices described below. Speech Understanding Subjective speech understanding ratings broadly followed the observed trends found for the sound quality, both within and across groups. Specifically, when significant differences were present, ratings of speech recognition in the surround sound condition were lowest or tied for lowest across conditions. Conversely, speech recognition ratings in the stereo condition were highest or tied for highest across conditions. Furthermore, while decreasing DBR led to lower ratings of speech understanding as expected, the negative effects were generally smallest when listening to television loudspeakers and the largest for surround sound. As expected, when group interactions were present, ratings from the group of older listeners with hearing loss were the most negatively affected by decreases in DBR. Speech understanding ratings were not significantly different for the three loudspeaker configurations at the better (+12 dB) DBR with audiovisual clips. This finding is not surprising and suggests that subjective speech understanding differences as a function of loudspeaker configuration are eliminated when understanding is easy (rated performance at or near ceiling). The poorer ratings of speech understanding and sound quality in the older listeners with hearing loss identified in the current study are consistent with the behaviorally measured performance demonstrated in several previous studies. Therefore, this study provides further evidence that the negative auditory and cognitive changes associated with aging (Arjmandi et al., 2024; Divenyi & Haupt, 1997; McAuliffe et al., 2012; Vermeire et al., 2016) have negative effects on self-perceived listening during important everyday activities such as the experience of watching recorded media. The results of the current study were somewhat surprising given previous research demonstrating significantly higher speech clarity ratings for a dedicated center channel loudspeaker (as used in the surround sound condition) than for the stereo condition (Shirley, 2013). While the reasons for these disparate findings are unknown, there were several methodological differences between these investigations including the use of a range of commercial five-channel surround sound clips for which the DBR was manipulated in the current study. In addition, this previous study was also different from the current study with regard to the equipment used, including the loudspeakers and the test environment, as well as the distance between the speakers and the listener. Previous research reviews have shown that preferences for surround sound loudspeakers compared to stereo loudspeakers can be affected by a wide range of factors including the specific listening materials evaluated (cf. Kerins, 2013, for a review). However, given that a range of commercial clips using both audio-only and audiovisual clips were evaluated, the current study results do not support the use of surround sound instead of other loudspeaker configurations as a beneficial solution for improving difficulties understanding recorded media for older adults with hearing loss. Instead, improvements in DBRs (e.g., through object-based mixing) could improve the experience of listening to recorded media. Such audio-based technological interventions warrant further investigation to explore how they affect not only speech recognition and sound quality but also immersion and emotional experiences. Clinical Implications The current study has some clinical implications related to suggestions for improving the recorded media experience at home for listeners with and without hearing loss. First, it is clear that better DBRs result in better sound quality and perceived speech understanding. In addition, the DBR benefits are larger for the listeners with hearing loss. This suggests that one recommendation for listeners who have difficulty watching recorded media would be to improve DBRs. Currently, improving DBRs needs to be completed at the mixing stage when using traditional surround sound techniques, given that research to date from the clean audio project suggests that channel level control is insufficient. However, improving the DBR could also be done at home, if the media was mixed and controlled through object-based codecs (such as Direct X). The results of this study suggest that providing end-user object-based control of the DBR could significantly improve ratings of sound quality and speech understanding, with benefits that are even larger for adults with hearing loss than for those with normal hearing. While user-controlled, object-based audio is available in some media and appears to provide a potential personalization solution, it is unclear when it will be widely available through the most common media service providers or whether average consumers will readily adopt it. However, this study provides additional evidence that object-based audio could be beneficial for enhancing the media experience for adults with hearing loss. The current study provides some insights into optimal loudspeaker configuration but does not allow for definitive recommendations. Although there were significant differences between conditions in this study, the differences varied by signal modality (audio only, audiovisual), DBR (+7, +12 dB), and listener group. The ratings are notably variable, suggesting that there is no clear single recommendation for loudspeaker configuration for all situations and groups. Despite the variability, the surround sound configuration was the least likely to offer the highest ratings of speech understanding performance or sound quality. The stereo loudspeakers tended to result in higher ratings than other configurations, although not always. Therefore, recommending surround sound system use to all adults who are struggling with recorded media viewing would not be evidence based, although the variability in the data suggests that, for some people, a surround sound system might result in the highest ratings. The exploration of the individual variability in the current study warrants future investigations. Limitations One limitation to the current study was that the low-frequency effects channel was eliminated for all loudspeaker configurations. Although important speech information is not present in these very low frequencies and no effect on the speech intelligibility rating results would be expected, the inclusion of the low-frequency effects channel might have resulted in different sound quality preference ratings. Although young listeners with normal hearing rated the television loudspeakers as having significantly worse sound quality compared to the other loudspeaker configurations, despite the absence of the low-frequency effects signal, the contributions of low-frequency effects channel warrant further investigation. In addition, the pattern of sound quality results identified herein might be affected by participants' definition of sound quality within the current experimental design. For example, some of the potential sound quality advantages for surround sound related to immersion might have been limited by the short duration of the test segments and were also not directly evaluated through an outcome measure. Related to this, the current study was limited to clips with dialogue only present in the center surround sound channel. It is unclear whether similar results would be found for materials that include moving dialogue (i.e., panned across multiple channels/loudspeakers) or that are in other channels (e.g., left or right). Future work related to overall preference and immersion with longer stimuli, different sound quality dimensions, and different types of dialogue may be warranted. Finally, the current study did not allow for full exploration of the separate effects of age and hearing loss, as there was no group of young adults with hearing loss in the study. Moreover, the interpretation of aging effects is confounded by differences in life experiences. For example, younger participants grew up with different technologies available to them (e.g., wireless streaming to earbuds from mobile devices) than did older participants (e.g., broadcast television and radio). Based on the technologies available to each group in adulthood, one might expect older participants to rate surround sound as higher, as that configuration might have been more representative of their listening experiences in adulthood. Although the reverse was observed (i.e., younger listeners were more likely than older ones to rate surround sound favorably), the contribution of lifetime listening experiences to an individual listener's ratings of sound quality and speech understanding to recorded media warrants further consideration. Conclusions Overall, many of the findings in this investigation were consistent with previous research. That is, lower (less favorable) DBRs reduced subjective ratings of sound quality and speech recognition performance. Furthermore and as expected, the group of younger listeners without hearing loss rated the sound quality of the television loudspeaker configuration significantly lower than the two configurations using larger loudspeakers. However, sound quality differences across loudspeaker configurations were less evident with older listeners. More importantly, there were also some unexpected findings. Specifically, use of a surround sound configuration was not supported as resulting in better ratings than default television loudspeakers or a stereo loudspeaker configuration. The results indicate that, clinically, surround sound would not be an evidence-based recommendation for most patients with hearing loss. However, the ability for patients to manipulate the DBR while watching recorded media at home has the potential to improve the sound quality and perceived speech understanding. Although current technological solutions, such as object-based audio mixing, are limited in availability, changes in DBR are a more robust way to address difficulties in listening to recorded media for adults with hearing loss than changes in loudspeaker configuration. Future work is warranted to continue to explore interventions for improving the recorded media experience at home because it is a common and important activity for adults with and without hearing loss. Data Availability Statement The data that support the findings of this study are available from the corresponding author upon reasonable request. Supplementary Material Individual (colors) and mean (black lines) better ear, audiometric thresholds as a function of frequency for older participants with (near) normal hearing (right panel) and older participants with hearing loss (left panel).