Effects of noise sensitivity and listening effort on perceptual ratings of background noise

Even though methods to assess individual noise sensitivity were established more than 45 years ago (Weinstein, 1978), the psychological underpinnings of this personality trait are still rather unclear (Williams , 2021; Kliuchko , 2016). Previous research indicates that noise sensitivity appears to be unrelated to actual sensory sensitivity in terms of auditory acuity (Ellermeier , 2001; Stansfeld, 1992) and would be better understood as an affective predisposition that leads to higher unpleasantness ratings (Ellermeier , 2001).

In fact, Kliuchko (2016) even found evidence for less efficient cognitive processing of auditory information in noise-sensitive subjects, supposedly because they are not as capable of making accurate predictions of the sensory input. However, as these effects were specific to time variant characteristics of noise, Kliuchko et al. did not consider it to account for general negative affective reactions to environmental sound.

One possible explanation for the connection between these difficulties in auditory processing and this negative predisposition to noise might be found in the interference of noise sensitivity with performing cognitive tasks in noisy environments. A body of research with sometimes contradictory findings implies that whether noise sensitivity affects performance depends on the type of task, for instance, whether it is about short term memory, arithmetic problem-solving (Belojević , 1992), spelling, or grammar (Weinstein, 1974). Also, the type of background noise is a crucial aspect, with understandable speech by a limited number of speakers usually yielding the most interference (Braat-Eggen , 2020; Ellermeier , 2001; Visentin , 2023a).

Even if task performance remains unaffected, noise-sensitive people have been found to perceive tasks as more difficult (Visentin , 2023b) and stressful and show higher levels of saliva cortisol (indicating stress and negative affect) when performing cognitive tasks in noise (Persson Waye , 2002; Ljungberg and Neely, 2007).

The less efficient processing of time-variant auditory input by noise-sensitive subjects reported by Kliuchko (2016) might at least partially account for a higher consumption of attentional resources by the environment which in turn increases the effort subjects have to invest and ultimately deteriorates task execution. The resulting frustration and experience of stress provide reasonable explanations for increased annoyance by noise.

These processes would be bound to situations where a considerable amount of mental effort has to be invested to complete a task, yet noise-sensitivity is mostly conceptualized as a stable personality trait (Ellermeier , 2001), and the question remains, how the negative reaction to noise behaves in situations with lower task load or no task at all.

The aim of this research is thus to further explore this interplay between noise sensitivity, the mental effort required to complete a task in the presence of environmental sounds, and affective ratings of this noise. In this context, three hypotheses regarding noise sensitivity, mental effort, performance and noise evaluation were formulated:

• H₁: More noise-sensitive subjects exhibit higher levels of effort when performing a task.

• H₂: Noise sensitivity decreases participants' task performance.

• H₃: Noise sensitivity moderates the effect of mental effort on perceived unpleasantness of environmental noise.

The three hypotheses were tested using the data of a laboratory experiment on task performance in the presence of noise, which was part of a larger study on the validity of listening experiments in virtual reality with regard to increased cognitive load. While the role of virtual reality was practically irrelevant to the research described here, the experimental design, including a variation of task load, however, provided suitable data to investigate our research questions.

The sample comprised 61 participants (11 female, 48 male, two divers) with a mean age of 31.1 years (standard deviation, SD = 12.3 years). The sample was divided into four experimental conditions in a 2 × 2 between-subjects design (see Table 1).

First, the sample was split into two groups conducting the experiment either in a real or a virtual visual environment. Participants were seated at a desk in a regular office at the Institute of Computer and Communication Technology (ICCT) in Cologne, and in the virtual reality group, they wore a head-mounted display (HTC Vive Pro 1) showing a virtual duplicate of this office. Then, within both groups, the sample was split into a high cognitive load and a low cognitive load condition.

In both load conditions, participants listened to several acoustical scenes of a typical office environment for a duration of 40 s each. Afterwards, the scenes hat to be rated in terms of loudness and unpleasantness using pseudo-continuous sliders with 100 scale values from 0 to 4. The sliders were labeled with five equidistant labels from “not at all” to “extremely” taken from ISO/TS 12913-2:2018 (ISO, 2018).

After rating loudness and unpleasantness, participants were presented with a list of 12 sound sources from which they had to pick the ones that were presented in the scene. This list contained the 3–7 correct sounds per scene and distractors that were randomly selected from a pool of 20 wrong answers. There were eight different acoustical scenes, each of which was presented twice, resulting in a total of 16 trials per participant.

Including questionnaires on noise sensitivity and mental effort before and after the experiment, respectively (see Sec. 2.4), the entire experiment took about 20 min.

In the low-load condition, participants had no task but to listen while the acoustical scene was played. They were instructed to sit back in relax and just observe either the audiovisual virtual environment or just the acoustical scene, depending on the experimental condition.

In the high-load condition, a dual-task paradigm was applied where participants performed the Stroop Color Word Interference test (Stroop, 1935) while listening. In this test, participants were presented a word for a color (e.g., “blue”) written in either the same or a different font color (e.g., “red”). Their task was to click a button corresponding to the font color and to complete as many trials as possible. After the participants gave the response, there was a 100 ms blank screen followed by the next trial. If no response was given after three seconds, the next trial started automatically and the missed trial was stored as a false response. The trials were randomly selected from a list of trials with even numbers of congruent (word matches the font color) and incongruent (word does not match the font color) color-word combinations.

Participants were explicitly instructed to devote all their attention to completing the Stroop test and not to the sounds presented. As an incentive to focus on the Stroop test, the five participants with the most correctly completed trials were rewarded a 25€ voucher. After the experiment, the same number of vouchers was randomly distributed among the participants in the low-load condition.

The acoustical scenes consisted of sound sources typically found in an office, including a conversation between two employees, a ticking clock, keyboard typing, sounds from a printer, a coffee maker, a fan, and traffic from outside the building. Three to seven of these sources were combined in different ways to create eight different acoustical scenes. The scenes were presented at average levels from 50.0 to 65.2 dBA. Table 2 shows all scenes with the included sounds and the average L_Aeq level.

The scenes were presented as binaural renderings via open headphones (Sennheiser HD600 Pro, Sennheiser, Wedemark, Germany) and dynamically auralised using a Polhemus Fastrak head tracker and the SoundScape Renderer framework (Ahrens , 2008). For the auralization, binaural room impulse responses (BRIRs) for each source position and different head orientations were measured at the participants' seat using a Neumann KU100 dummy head (Neumann, Berlin, Germany) and a Genelec 8020D loudspeaker (Genelec, Lisalmi, Finland). In the acoustical scenes, the sound sources were placed at different positions inside and outside the office and the loudspeaker was moved to each position for the BRIR recordings. The BRIRs were then convolved with anechoic recordings of the corresponding sound source.

Prior to the experiment, noise sensitivity was assessed using a short version of the Lärmempfindlichkeitsfragebogen, LEF (“noise sensitivity questionnaire”), by Zimmer and Ellermeier (1998), comprising nine items. All items were averaged to calculate an individual noise sensitivity score.

At the end of the experiment, participants rated their invested effort by means of the NASA task load index (TLX) by Hart and Staveland (1988), which consists of self-reports on the experienced Mental Demand, Physical Demand, Temporal Demand, the levels of Effort and Frustration, and the own Performance in a task. All items were averaged to a composite task load index. However, the item Performance was excluded, because it was significantly correlated with only two out of five other items, namely, Physical Demand (r = −0.259, p = 0.044) and Temporal Demand (r = −0.335, p = 0.008).

Figure 1 depicts the ratings for loudness and unpleasantness of the eight acoustical scenes averaged across all experimental conditions and the two ratings per participant for each acoustical scene. Overall, unpleasantness ratings are slightly higher and both loudness and unpleasantness ratings tend to increase with the level of the acoustical scenes (see Table 1). A noteworthy exception is Scene 4, which has the second highest loudness ratings and, together with Scene 8, is the only scene that includes human speech.

The test-retest reliability calculated as the correlation between the first and second rating of each acoustical scene was moderate, with a reliability score of 0.591 for loudness and a score 0.540 for unpleasantness, respectively.

As a performance measure in the sound recognition task, the relative numbers of correctly selected sounds were calculated and averaged across all scenes and presentations. In the low load conditions, the average portion of correctly remembered sounds was 55.1% (SD, 23.2%) in the real environment and 54.5% (SD, 25.7%) in the virtual environment. In the high load condition, the average portion was 45.7% (SD, 23.6%) in the real environment and 43.4% (SD, 22.5%) in the virtual environment.

In the Stroop Test, the mean score of correctly answered trials was 97.7% (SD, 3,9%) in the real environment and slightly increased to 98.3% (SD: 2.7%) in the virtual environment.

Figure 2 depicts the subjective effort by means of the average NASA TLX score for each condition. Effort ratings in the high load conditions including the Stroop test are evidently higher than in the low load conditions, while the distribution is rather similar in the real environment and VR version of the experiment.

Prior to testing the hypotheses, several analyses of variance (ANOVAs) were calculated to test for test effects of both the test environment (real vs virtual) and the two load conditions (high load/Stroop test vs low load/no Stroop test) on each participants' average task performance as well as on their average loudness and unpleasantness ratings. As the performance in the Stroop test was only present in the high load condition, this variable was only tested for differences regarding the environment.

Results revealed that the environment did not have a significant effect on the relative number of correct trials in the Stroop test performance (F = 0.52, p = 0.474, η_p² = 0.014). Moreover, the number of correctly identified sound sources did not differ significantly between the environment groups (F = 3.17, p = 0.080, η_p² = 0.052), but with regard to the load condition (F = 10.33, p = 0.002, η_p² = 0.151). The same applied to subjective effort, which was not significantly affected by the environment (F = 0.37, p = 0.548, η_p² = 0.006), but which was significantly higher in the high-load than in the low-load condition (F = 20.15, p < 0.001, η_p² = 0.258).

A two-way multivariate analysis of variance (MANOVA) further showed that neither loudness nor unpleasantness ratings were significantly affected by the environment (loudness, F = 0.158, p = 0.692; unpleasantness, F = 0.015; p = 0.902; η_p² = 0.021) or the load condition (loudness, F = 0.17, p = 0.683; unpleasantness: F = 0.49, p = 0.488; η_p² = 0.012). Also, the inconsistency for loudness and unpleasantness did neither significantly differ across environments (loudness, F = 0.072, p = 0.790; unpleasantness, F = 0.54, p = 0.465; η_p² = 0.070) nor load conditions (loudness, F = 0.75, p = 0.391; unpleasantness, F < 0.01, p = 0.990; η_p² = 0.054).

Since the environment had no noteworthy impact on any of the variables of interest for this study, it will not be further addressed in the remainder of this paper, and all analyses will be limited to differences regarding the load condition.

Results revealed a significant positive correlation between noise sensitivity and subjective effort of r = 0.259 (p = 0.046), confirming H₁ in terms of subjectively experienced task load.

In the next step, the relationship between task performance and the individual noise sensitivity was investigated. Results indicate that the number of correct trials in the Stroop test was not significantly associated with noise sensitivity (r = −0.062, p = 0.706).

Regarding the number of correctly remembered sound sources of the presented acoustic scenes, a significant relation with noise sensitivity was observed only in the high-load condition (r = −0.341, p = 0.034). This finding suggests that participants with higher noise sensitivity were less able to recall the presented sound sources while performing the Stroop test. By contrast, in the low-load condition, this correlation was small and non-significant (r = 0.197, p = 0.393).

Therefore, it can be stated again that our hypothesis H₂ was only partially confirmed in the high-load condition. However, recalling fewer sound sources in this dual-task paradigm, which can be understood as decrease in task performance in the secondary task, provides evidence for increased mental effort (Kahneman, 1973). Therefore, this observation provides further support for our hypothesis H₁ that the invested mental effort was higher among noise-sensitive participants.

To test H₃, noise sensitivity was tested as a moderator in a regression model to predict loudness and unpleasantness ratings, averaged for each participant over all acoustical scenes from an interaction between noise sensitivity and mental effort. While subjective effort did have a significant effect on both loudness (β = 0.265, p = 0.046) and unpleasantness ratings (β = 0.445, p < 0.001), the interaction effect was not statistically significant in both models (loudness, β = −0.207, p = 0.091; unpleasantness, β = −0.080, p = 0.483), thus showing no support of H₃ by our data.

However, a mediation analysis suggests that the positive relationship between individual noise sensitivity and mean ratings of both loudness and unpleasantness (in comparison to the sample average) was fully mediated by the subjective effort of the participants (see Fig. 3).

Finally, a potential influencing role of participants' motivation on the mediation effects of subjective effort was tested. Even though the experimental load condition is not explicitly addressed in the mediation analyses, the subjective effort differed considerably across the high load and low load condition (see Fig. 2). This finding might be attributable not only to the Stroop test, but also to the incentive being bound to task performance in the high load condition. Thus, we recalculated the mediation analyses including only data from the high load condition. This analysis revealed, that in both models the mediation paths from noise sensitivity to effort (β = 0.253, p = 0.037 in both models) and from effort to stimulus rating remained significant (loudness: β = 0.429, p = 0.013; unpleasantness: β = 0.482, p = 0.009), whereas the direct paths did not (loudness: β = 0.060, p = 0.672; unpleasantness: β = −0.138, p = 0.351). Thus, both H₃ and H₁ assuming a significant positive relationship between noise sensitivity and mental effort were confirmed regardless of differences in motivation due to the presence or absence of an incentive.

The effects of noise sensitivity on task performance, mental effort, and noise evaluation were investigated in a laboratory study by comparing ratings of acoustic scenes of an office presented while participants completed the Stroop test to ratings of these stimuli when they were presented without an additional task.

Although the performance in the primary task (i.e., the Stroop test) was not affected by noise sensitivity, both self-reports and a lower task performance in the secondary task (remembering the different sound in the acoustical scene) indicated higher mental effort among more noise-sensitive participants. This corroborates previous findings (Ljungberg and Neely, 2007; Visentin , 2023b) and could be explained by a more inefficient processing of auditory information observed by Kliuchko (2016).

Furthermore, increased mental load apparently led to higher loudness and unpleasantness ratings by noise sensitive subjects. This result is interesting as it essentially suggests that noise-sensitive participants exhibited more negative sound evaluation only when they were investing higher mental effort. It might be that, in terms of level and type of noise, the acoustical environment was not disturbing enough to trigger “noise-sensitive reactions” in the low load condition without the Stroop test. As both sound level and the composition of environmental sounds were designed to resemble a realistic office environment, these effects can be expected to apply to real-life situations and yield two noteworthy implications.

First, as stated in Sec. 1, situations where completing a task requires mental effort might be considered a link between inefficient cognitive processing of noise and a negative affective predisposition to noise, as noise-sensitive subjects become aware on the effect noise has on their ability to execute cognitively demanding tasks.

Second, in this context, noise sensitivity seems to depend on situational factors, although it is often considered a stable personality trait (Ellermeier , 2001; Williams , 2021). While the noise sensitivity questionnaire (NoiSeQ) by Schütte (2007) addresses this issue by defining different subscales, the most established inventory by Weinstein (1978) as well as the LEF by Zimmer and Ellermeier (1998) that was applied here treat noise sensitivity as a singular trait. The NoiSeQ was not used here, because the inventory comprises 35 different attributes (as the data used here are part of a larger study that did not have noise sensitivity as its primary objective).

As the data used here are part of an experiment that was primarily designed to test the validity of experiments in virtual environment involving high cognitive load, there are some limitations to this study. First, working memory capacity, which was shown to be a crucial factor regarding individual performance in the Stroop test (Long and Prat, 2002; Shipstead and Broadway, 2013), was not measured here.

Second, for a profound analysis on the impact of different situations (here, performing the Stroop test vs just listening to the sounds), even sample sizes in both load conditions would be necessary. To circumvent this issue, the hypothesis testing was mostly based on measures of subjective effort, with lower and higher scores representing both conditions, respectively, rather than using these conditions as predictors themselves.

Given these limitations, further research is needed to disentangle the relationship between noise sensitivity, efficiency in auditory processing and working memory capacity, and how these increase the mental effort invested in completing cognitive task in the presence of environmental sounds. Here, future studies could provide deeper insights on whether negative reactions of noise-sensitive subjects to noise are indeed primarily a result of momentary mental effort, which, with time, might consolidate in a more stable permanent affective predisposition against noise.

This study was funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) – 532148125 and supported by the central publication fund of Hochschule Düsseldorf University of Applied Sciences.

This research was not supported by any external funding and the authors confirm that there are no conflicts of interest to disclose.

This study was approved as compliant with the Declaration of Helsinki by the Commission for Responsibility in Research at the Technische Hochschule Köln (Application number THK-2023-0003). At the beginning of the experiment, participants provided informed consent regarding the processing of their data.

The data that support the findings of this study are available from the corresponding author upon reasonable request.