The effect of inattention and cognitive load on unpleasantness judgments of environmental sounds

Applied Acoustics 164 (2020) 107278

Contents lists available at ScienceDirect

Applied Acoustics journal homepage: www.elsevier.com/locate/apacoust

The effect of inattention and cognitive load on unpleasantness judgments of environmental sounds Jochen Steffens a,⇑, Franz Müller b, Melanie Schulz b, Samuel Gibson b a b

Düsseldorf University of Applied Sciences, Institute of Sound and Vibration Engineering (ISAVE), Münsterstraße 156, 40476 Düsseldorf, Germany Technische Universität Berlin, Audio Communication Group, Einsteinufer 17c, 10587 Berlin, Germany

a r t i c l e

i n f o

Article history: Received 9 November 2019 Received in revised form 17 January 2020 Accepted 17 February 2020

a b s t r a c t Evaluations of our acoustical environment depend on sound characteristics as well as on factors attributable to the person and situation. The aim of the study was to investigate the role of the listener’s attentional state and cognitive load, temporal characteristics of environmental sounds, and potential interaction effects on overall retrospective sound evaluations. Sixty-three test participants were presented six two-minute recordings of environmental sounds, whereby the temporal position of one occurring sound peak or valley was varied. The experimental group was distracted from the test sounds and cognitively loaded by performing the Stroop Test during the sound presentation and retrospectively rated the sounds in terms of loudness and pleasantness. In contrast, the control group attentively listened to the sounds and rated them during and after listening. Results showed that participants in the experimental group rated the sounds on average as 6.7% more pleasant than the participants of the control group. Moreover, previous findings on the influence of the peak position on the retrospective judgments (recency effect) were replicated; however, no effect of the temporal position of sound valleys was found. Retrospective judgments in the control group could further be predicted by the averaged momentary judgment, the linear trend over time as well as the momentary judgment during the peak and the end. Findings thus emphasize the need of taking into account human (in)attention and memory processes when assessing complex environmental sounds. Ó 2020 Elsevier Ltd. All rights reserved.

1. Introduction Human evaluations of our acoustical environment depend on multiple factors, such as features of the sound, psychological characteristics of the listener or the context in which the evaluation takes place. While the influence of sound properties on perception and evaluation has been widely investigated in psychoacoustical research [5,10], the role of the listener’s state (e.g., attentional processes) has received less attention. Also, it is widely unknown how sound- and listener-related variables interact in the evaluation process. Therefore, the major aim of the current study was to investigate the influence of attention and memory processes in the listener, the characteristics of the sound and their interactions on retrospective pleasantness judgments of environmental sounds. 1.1. Listener-related influences on attention The perception and evaluation of sound in a certain situation depend on a listener’s focus of attention. Attention can be ⇑ Corresponding author. E-mail address: [email protected] (J. Steffens). https://doi.org/10.1016/j.apacoust.2020.107278 0003-682X/Ó 2020 Elsevier Ltd. All rights reserved.

described as the process of selecting information that is important for current action and perceptual goals and suppressing irrelevant information. The focus of attention can either be directed deliberately (e.g., to a certain speaker) or steered automatically in response to a salient sensory input (focusing on the cause of the stimulus). We follow a cognitive load theory of attention in which the perception or rejection of an (acoustical) stimulus is assumed to depend on the level of cognitive load (defined as the used amount of working memory resources) associated with a certain distracting task [16]. That means that the extent to which unattended information is perceived depends on the perceptual load of the attended task. Recent studies have supported the aforementioned assumptions by showing that cognitive load negatively affects auditory detection [20] and speech recognition [17]. One method for studying divided attention is the dual-taskparadigm. This paradigm involves comparing conditions in which persons perform two tasks simultaneously with those where only one single task has to be executed [2]. With this technique, the degree of interference between the two tasks can be measured. Kahneman [11] suggests that interference will arise even when no mechanisms regarding perception or response are shared between concurrent activities. This, however, is assumed to be only

2

J. Steffens et al. / Applied Acoustics 164 (2020) 107278

the case if the load imposed by each of the activities is high enough. One example mentioned by Kahneman in this context is the combination of driving and conversing. The ability of humans to execute both tasks simultaneously is maintained until the demands of driving become critical, ultimately leading to an interruption of the conversation. Studies using the dual-task paradigm have repeatedly confirmed the aforementioned assumptions (see [15], for a review). Recent research in the field of acoustics has applied dual-task experimental paradigms to measure listening effort expended during speech understanding, demonstrating that interfering secondary tasks, such as driving a vehicle or recognizing visual patterns [30] can negatively affect speech recognition (see [9], for a review). The listener’s attention in a real-life situation is, of course, not always focussed on the acoustical environment, especially when it comes to non-speech sounds. Therefore ‘focussed’ laboratory studies might introduce a bias when predicting annoyance reactions in mundane life by results from laboratory studies. Several studies on noise effects therefore aimed at re-creating mundane activities and attentional processes by, for example, letting participants read magazines [3,27,29] or medical literature [19]. In the field of product sound design, Steffens [22] let participants interact with household appliances under test and observed that their sounds were perceived as more pleasant compared to a laboratory experiment, in which only sounds of the devices were presented. 1.2. Stimulus-related effects on attention When discussing the role of (in)attention and cognitive load on sound perception and evaluation, one must also consider the role of the stimulus itself leading to the concept of saliency. This stimulus-driven (‘bottom-up’) mechanism postulates that some features in a scene are conspicuous based on their context and, thus are salient and attract attention [13]. This takes into account the fact that sounds in the evolutionary sense have a warning function against potential hazards. Auditory saliency is governed by the loudness of the sound as well as by the temporal and spectral contrast to competing sounds (e.g., [12,13]). However, it is not only the ‘objective’ physical properties that contribute to the saliency of a sound. Saliency can also be learned and be of personal meaning to the individual, such as hearing your own name, or may simply arise from a violation of a person’s expectation [14]. Studies have shown that computational saliency models can reliably predict changes in ratings of urban soundscape pleasantness [4,7,26]. Furthermore, various studies in the field of psychoacoustics have demonstrated the important role of loudness peaks (i.e., salient events) on overall judgments of time-varying environmental sounds [18,6]. This, however, seems to be dependent on the temporal position of the salient event. Here, it has been shown that, due to memory effects, sound events occurring towards the end of a sound play a more important role than other sections when it comes to an overall retrospective judgment (recency effect; [28]). The fact that retrospective judgments can be predicted only by the unweighted combination of peak and end was established by Kahneman and colleagues as the so-called ’peak-end rule’ (e.g., [8]). In this context, Steffens et al. [24] observed that a weighted combination of peak and end can explain a large proportion of variance in daily retrospective judgments of environmental sounds. 1.3. The current study The major aim of the current study was to investigate the influence of inattention in the course of an increased cognitive load caused by secondary task on unpleasantness judgments of envi-

ronmental sounds. Based on the literature described, we hypothesized that cognitively loaded and thus inattentive participants would rate sounds as significantly more pleasant and less loud than attentive participants (H1), due to limited perceptual and cognitive resources. With regard to the temporal position of salient events (peaks), we assumed that the later a peak (i.e. a sound event whose level is higher than the rest) is positioned in a sound scenario, the louder and more unpleasant the sound scenario will be judged, regardless of the attentional demands of the participant (H2a). Analogously, we expected that the later a valley (i.e. a sound event whose level is lower than the rest) occurs in a sound scenario, the lower the overall unpleasantness and loudness would be rated (H2b). Regarding the relationship between momentary and retrospective judgments, we aimed at replicating and extending the findings by Steffens et al. [24] who showed that retrospective pleasantness judgments can be predicted by the averaged momentary judgment, the linear trend as well as the momentary judgment during the peak and the end (H3). Finally, potential interactions between sound characteristics and the listener’s attentional state were explored without formulating specific hypotheses. 2. Method 2.1. Participants The study was conducted in accordance with the Helsinki Declaration and involved 63 participants aged between 20 and 63 years (27% female; mean age: 29.5 years, SD = 7.9). Twenty participants were contacted via the TU Berlin mailing list of active study participants and received 10 Euro as compensation. The remaining 43 participants were students of the TU Berlin who could acquire credit points for participating. Requirements for participation were normal hearing, sufficient knowledge of German and no color blindness, obtained by self-reports of the participants. The data set of one participant was excluded from the evaluation due to the fundamentally incorrect execution of the task. 2.2. Experimental design Participants were randomly assigned to an inattentive (experimental) and an attentive (control) group, both listening to nine sounds with a duration of 2 min. Participants of the inattentive group conducted the Stroop Color-Word Interference Test during the presentation of the test sounds in order to constantly distract them from the acoustical stimuli. Participants in the control group were asked to perform momentary judgments during the presentation of the sounds, meaning to continuously adjust a slider between ’very quiet’ and ’very loud’ to reflect the perceived loudness of the sounds over time. This was done to continuously draw participants’ attention to the test sounds. After the presentation of each sound, participants in both groups had to rate the loudness and pleasantness of the sounds via two sliders with 101 steps, ranging from ’very quiet’ to ’very loud’ and ’very pleasant’ to ’very unpleasant’. The whole experiment lasted about 30 min. 2.3. Materials 2.3.1. Sound stimuli Sound files taken from the General Series 6000 sound database were used and edited to compose nine different two-minute sound scenarios. Six of the scenarios, referred to as test sounds, consisted of a constant ‘base’ sound interrupted by another twenty-second sound ‘event’ that differed in timbre and level (see Table 1). The combination of the difference in timbre and level was used to

3

J. Steffens et al. / Applied Acoustics 164 (2020) 107278 Table 1 Description of the sound scenarios used in the study. Sound

Peak type

Peak position

Base

Peak/valley

Base/peak level [dBA]

1 2 3 4 5 6 7 8 9

peak peak peak valley valley valley – – –

early middle late early middle late – – –

Distant road traffic + birdsong Distant road traffic + birdsong Distant road traffic + birdsong Construction sounds Construction sounds Construction sounds Japanese Speech Café + Music Birdsong + Train Rattle

Construction sounds Construction sounds Construction sounds Distant road traffic Distant road traffic Distant road traffic – – Lawn mower

65/85 65/85 65/85 76/59 76/59 76/59 72/– 68/– 73/85

played at the beginning of the experiment, and the two filler sounds were always placed in the third and sixth position. The sounds and instructions were presented to the participants using a laptop computer with Matlab software, an external audio interface (Focusrite Scarlett 6i6), and circumaural headphones (DT 770 Pro).

effectively provoke an attentional shift in the participants during the experiment. The test sounds were divided into two different event types into groups of three sounds. In the first group, distant road traffic sound with birdsong (base) was interrupted by louder construction sounds (peak event). Analogously, in the second group, construction sounds (base) were stopped so that distant road traffic sound became audible (valley events). In both groups, the peaks and valleys occurred after 10 s (early), 50 s (middle), or 90 s (late) in the respective test sounds (see Fig. 1). After running several pre-tests, the overall level of the sounds with valley events was reduced to make the much longer construction sounds tolerable for participants. However, the level difference between base and event was kept between 17 and 20 dB(A) for every test sound. In terms of psychoacoustical loudness (according to DIN 45631/ A1), peak (sounds 1–3) and valley event scenarios (sounds 4–6) achieved an overall time-averaged value of 31.5 and 24.8 sone, respectively. Overall time-averaged sharpness scores (according to DIN 45692) were 1.5 acum for the peak event scenarios and 1.4 acum for the valley event scenarios. These psychoacoustical metrics were computed using HEAD acoustics ArtemiS SUITE. To distract the participants from the structure of this 3  2 factorial design and to reduce anticipation errors, one training and two filler sounds were used in addition to the test sounds. These three sounds (Table 1; no. 7–9), however, were not included in the data analysis. The test sounds were successively presented to both groups in a randomized order; only the training sound was

80 70 60

90

Level [dB(A)]

90

Level [dB(A)]

Level [dB(A)]

90

80 70 60

0

40

80

120

0

40

80

120

0

70 60

80 70 60

80

Time [s]

120

80

120

90

Level [dB(A)]

Level [dB(A)]

80

40

Time [s]

90

40

70

Time [s]

90

0

80

60

Time [s]

Level [dB(A)]

2.3.2. Stroop Color-Word Interference Test The Stroop Color-Word Interference Test is a concentration task in which a list of differently colored words is presented to a participant [25]. The words in the list describe different colors. However, the word information does not match the color of the word. The task of the participant is either to read the words or to name their coloring. In the present study, we used the Stroop Test to distract the participants during the presentation of the test sounds and to ensure a constant cognitive load. The computer-implemented Stroop Test was used in a slightly modified form. The participants were presented with lists of 15 words. These lists repeatedly contained the words blue, green, yellow and red in random order. Although the order was random, it was ensured that color did not occur several times in succession. The limitation was put in place to prevent the occurrence of easily recognizable patterns during the test and, thus, to prevent fluctuations in test performance. Four buttons were positioned below the word list. Each button was labeled with one of the colors in black lettering. The task of the test was to assign the coloring of the

80 70 60

0

40

80

Time [s]

120

0

40

80

120

Time [s]

Fig. 1. Level-Time-Diagram of the used test sounds with peaks (first row) and valleys (second row) and event positions early, middle and late (from left to right).

J. Steffens et al. / Applied Acoustics 164 (2020) 107278

current word (indicated by a grey frame) as quickly and conscientiously as possible by pressing the corresponding button. After each assignment, the frame indicated the next word in the sequence. After assigning the last word in the list, a new list was generated and the frame was reset to the first position. The test was paused at the end of each two-minute test sound, and the participants were asked to evaluate the overall loudness and pleasantness of the acoustical scenario. In addition to the evaluation criteria, the number of correct assignments and response time between the assignments were recorded (not reported in this paper). It was assumed that the performance of the participants (number of correct answers and response time) could improve over the course of the test. This could have affected the cognitive load and thus the subsequent assessment of the sounds. In order to keep the performance as constant as possible, each participant completed a training sequence under supervision. This training sequence consisted of the Stroop Test accompanied by the training sound. Due to the rather simple task, a training period of two minutes was considered appropriate in order to reduce possible learning effects to a minimum while also keeping the cognitive load right before the test at a low level.

100 peak

90

Unpleasantness [%]

4

valley

80

70

60

50 inattentive group attentive group

40 early

middle

late

early

middle

late

Event type/Event position Fig. 2. Means and 95% confidence intervals for the factor unpleasantness. The data was divided into inattentive (bright) and attentive group (dark) sets and grouped by event type and event position.

2.4. Data analysis Data processing and analysis were performed using IBM SPSS 25.0. Before the main analyses, the relationship between the rated loudness and pleasantness was examined. Due to their high correlation (r = 0.717; p < .001), both variables were collapsed into one factor ‘unpleasantness’ by means of a Principal Component Analysis. This factor was scaled in a way that 100% unpleasantness corresponded to the ratings ’very loud’ and ’very unpleasant’, 0% to ’very soft’ and ’very pleasant’ and 50% to a neutral rating. Hypotheses were tested by means of linear mixed-effects models including a random intercept for each participant and using restricted maximum likelihood estimates of variance components and Type III Analysis of Variance via Satterthwaite’s degrees of freedom method. For all analyses, the significance level a was set to 0.05. 3. Results 3.1. Intraclass correlation coefficient Before testing the hypotheses, a null model including only a random intercept for each participant was computed to obtain the Interclass Correlation Coefficient (ICC). This analysis showed that 27.4% of the overall variance of the judgments was explained by person-related differences (ICCtotal = 0.274, ICCexp. = 0.154; ICCcontrol = 0.356). The ICCs for both groups revealed that sound evaluations in the attentive group varied more than twice as much across individuals (35.6%) than they did in the inattentive group (15.4%). 3.2. Influence of (in)attention and cognitive load In the next step, linear mixed-effects models were computed to test the hypotheses formulated in the introductory section. Results of a linear mixed-effects model with the experimental condition as independent variable confirmed our first hypothesis (H1) that participants in the inattentive group would perceive the test sounds as more pleasant than participants in the attentive group, F (1,60) = 7.17; p = .010, R2marginal = 0.039, D = 6.6%. Fig. 2 depicts the means and 95% confidence intervals for the factor unpleasantness for both test groups (each with i = 31 observations) and grouped by peak type and peak position.

Concerning H2a, results confirmed our assumption that the peak position had a significant effect on the unpleasantness ratings for the high peak scenarios, F(2,122) = 12.50, p < .001, R2marginal = 0.069. As expected, the later high peaks occurred in a sound, the more unpleasant the respective sound scenario was evaluated by the participants. As illustrated by Fig. 2, unpleasantness ratings increased over the peak positions ‘‘early”, ‘‘middle” and ‘‘late” from 55.2% over 64.5% to 68.5% (early vs. late: p = .001; middle vs. early: p = .007) for the inattentive group and from 65.7% over 67.7% to 74.6% (early vs. late: p = .031) for the attentive group. No effect, however, was observed related to the interaction between experimental condition and the peak position (p = .248). Also, no significant differences in the valley scenarios between peak positions (H2b) were observed, F(2,122) = 0.88, p = .417. 3.3. Prediction of retrospective judgments In the next step, we tested whether the aforementioned findings by Steffens et al. [24] could be replicated and extended to a shorter time scale (2 min) and extended with regard to loudness judgments (H4). This was done only for the attentive control group where both momentary and retrospective loudness judgments were performed by the participants. The averaged momentary judgments, as performed by the participants of the control group, are depicted in Fig. 3. For each participant and stimulus, we computed the four variables Average (i.e. the momentary judgments averaged over time), the linear Trend over time (representing the standardized coefficient of a time regression), the Peak (max) (maximum value occurring during the momentary judgment) and the End (momentary judgments averaged across the last five seconds of each stimulus) (see [23], for formulas and more detailed descriptions). The obtained variables (as well as the dependent variable) were zstandardized to obtain their relative importance by means of the standardized regression coefficient b. To test our hypothesis, we used these four variables as independent variables in one linear mixed-effects model to predict retrospective judgments. As hypothesized, results revealed a significant effect for each of the variables, confirming H3. The statistics of the regression model (estimates of fixed effects) are reported in Table 2. It becomes evident that Peak and Average

5

25 0 -25 -50 0

40

80

50

Momentary judgment

50

Momentary judgment

Momentary judgment

J. Steffens et al. / Applied Acoustics 164 (2020) 107278

25 0 -25 -50

120

0

40

25 0 -25 -50 40

0 -25 -50

120

0

80

120

50 25 0 -25 -50 0

40

Time [s]

80

40

80

120

Time [s] Momentary judgment

50

0

25

Time [s] Momentary judgment

Momentary judgment

Time [s]

80

50

120

Time [s]

50 25 0 -25 -50 0

40

80

120

Time [s]

Fig. 3. Averaged momentary loudness judgments (control group) of the six test sounds over time.

Table 2 Results of a linear mixed-effects model (estimates of fixed effects) predicting retrospective loudness judgment (control group) (H3). Predictors

b

SE

df

t

p

95% CI

Intercept Average Trend Peak (max) End

0.00 0.30 0.11 0.30 0.25

0.08 0.09 0.05 0.07 0.08

28.21 174.96 154.18 171.50 166.53

0.00 3.23 2.34 4.52 2.98

>0.999 0.001 0.021 <0.001 0.003

[ 0.17 0.17] [0.12 0.48] [0.02 0.20] [0.17 0.43] [0.08 0.41]

reached similarly high b values, being the most important predictors, followed by End and Trend. The overall model explains almost half of the variance in the retrospective loudness judgments (R2marginal = 0.497). Accordingly, we compared the predictive power of several ’submodels’ mentioned in the literature. The first one was mentioned by Steffens and Guastavino [23] and consists of the arithmetic mean of the momentary judgments over time and their linear trend (Average + Trend). Two more models include both only the momentary judgment during the peak and the at the end (Peakend rule). While Fredrickson and Kahneman [8] proposed an unweighted combination of the two elements as a valid predictor for overall retrospective judgments, Steffens et al. [24] found that a weighted combination with a higher weight for the momentary judgment at the end explained more variance. Also in this study, results revealed the variance explained by the fixed effects was highest for a prediction model consisting of a weighted combination of peak and end (R2marginal = 0.474), followed by a model including an unweighted combination of peak and end (R2marginal = 0.451) and a model consisting of average and trend (R2marginal = 0.395). 4. Discussion This study investigated the role of (in) attention and cognitive load and interactions with sound characteristics on the evaluations of complex environmental sounds. Results confirmed our assumption that participants whose attention was directed to sounds presented in an experiment rated them as less unpleasant than those distracted and cognitively loaded by a secondary task. This finding

is in line with our hypothesis and highlights the need for taking into account attentional processes when investigating sound evaluations. In addition, our findings suggest that studies in noise research and psychoacoustics focussing participants’ attention on the sounds under test might overestimate human loudness perceptions and annoyance reactions in mundane life. This might, in particular, be the case in situations in which people’s cognitive load is governed by a certain activity, such as driving or cooking. The Stroop Test used in the experiment, however, is a specific reading task in which participants might internally ‘hear’ the words they are reading, leading to an activation of the auditory cortex [31] and in turn to a suppression of the actual environmental sounds. Therefore, further research is needed to investigate whether the influence of inattention and cognitive load holds true also for non-verbal tasks. Moreover, studies should aim to replicate the findings with sounds of longer duration to allow for higher generalizability of the observed effects to mundane perception processes. Our study further replicated previous findings on the role of recency effects for negative sound events. The recency effect suggests that it is not only relevant whether a certain sound event occurs, but also when this happens. This emphasizes the role of long-term memory processes in sound evaluations. It is worth mentioning that, in our experiment, the recency effect affected both the attentional and the inattentional group in a similar way. This finding indicates that, despite the shifted focus of attention and the resulting constant differences in overall unpleasantness judgments, participants in both groups seemed to rely on similar information when forming these judgments. Contrary to our

6

J. Steffens et al. / Applied Acoustics 164 (2020) 107278

hypothesis, however, a recency effect was not observed for evaluations of sounds with valleys. This finding might be explained by a negativity bias meaning that negative information is processed more thoroughly than positive information [1,21]. We therefore assume that peaks have a higher likelihood to be memorized and recalled in the course of a retrospective judgment than valleys. The importance of peaks in combination with the recency effect (’peak-end rule’) is further highlighted when relating features of the momentary judgments to the overall retrospective loudness judgments in the control group. Our results revealed that only relying on the information related to the peak and the end explains almost half of the variance in the retrospective judgments. Even though the weighted combination of peak and end slightly outperformed the unweighted combination as originally proposed by Fredrickson and Kahneman [8], this small advance in variance explanation has to be considered in the context of potential overfitting associated with adjusting the regression coefficients for this specific dataset. Besides, the predictive power of the peak-end rule in this study (as opposed to other studies, such as by Steffens and Guastavino [23]) could be related to the specific characteristics of the presented sounds. In the present study, a rather strong perceptual contrast between peak and the other elements of the sounds was intended and achieved which is assumed to function as a boundary condition for the validity of the peak-end rule. However, besides this heuristic element in retrospective judgments suggested by the peak-end rule, some degree of ’cognitive averaging’ over time as indicated by the effect of the averaged momentary judgment still seems to be in place. This corroborates findings by Fiebig and Sottek [6] and Steffens et al. [24] who also found that both averaged momentary judgments and the peak contribute to the prediction of overall loudness and pleasantness judgments. Finally, the effect of the linear trend on retrospective judgments also highlights the role of the temporal dynamics of an acoustical scenario and further suggests that participants prefer positive trends (i.e. sounds are perceived as more pleasant over time) over negative ones.

5. Conclusion In our study, we demonstrated that psychological processes associated with the listener, namely (in-)attention, cognitive load, and memory, significantly affect unpleasantness ratings of complex sound scenarios. These findings have strong implications for research on noise effects and psychoacoustical metrics. For example, both laboratory and field studies focussing participants’ attention on test sounds might overestimate loudness and unpleasantness judgments in real-life in which people are usually focussed on their current activity. Amongst others, this might be relevant for practitioners in sound engineering when designing their products and/or when determining tolerance thresholds for acoustical product characteristics. A manufacturer of washing machines, for example, might focus on designing the crucial temporal sound elements within a washing or spinning cycle. This could be the beginning when the user is (attentively) interacting with the machine or parts of the spinning cycle where the auditory saliency reaches its maximum. In addition, when designing acoustical events, positive trends over time are to be preferred. A methodological recommendation could be further to integrate mundane tasks in experimental designs, as done in this study or others mentioned in the paper. Alternatively, one might subtract the ‘inattentional bias’ from obtained loudness or pleasantness ratings, as determined in this study. This bias, however, presumably depends on the saliency of the sound itself as well as on the task-at-hand during which the sound typically occurs in daily life and thus will be subject to future research.

CRediT authorship contribution statement Jochen Steffens: Conceptualization, Methodology, Validation, Formal analysis, Resources, Writing - original draft, Writing review & editing, Supervision, Project administration. Franz Müller: Software, Validation, Formal analysis, Investigation, Resources, Data curation, Writing - original draft, Writing - review & editing, Visualization. Melanie Schulz: Software, Formal analysis, Investigation, Resources, Data curation, Writing - original draft. Samuel Gibson: Software, Formal analysis, Investigation, Resources, Data curation, Writing - original draft, Writing - review & editing, Visualization. Declaration of Competing Interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. References [1] Baumeister RF, Bratslavsky E, Finkenauer C, Vohs KD. Bad is stronger than good. Rev General Psychol 2001;5(4):323–70. [2] Broadbent DE, editor. Perception and communication. Elmsford (NY): Pergamon Press, Inc.; 1958. [3] de Coensel B, Botteldooren D, Berglund B, Nilsson ME, de Muer T, Lercher P. Experimental investigation of noise annoyance caused by high-speed trains. Acta Acust United Acust 2007;93(4):589–601. [4] De Coensel B, Botteldooren D. A model of saliency-based auditory attention to environmental sound. Proceedings of 20th international congress on acoustics, 2010. ICA 2010. CD-ROM, Sydney, Australia. [5] Fastl H, Zwicker E. Psychoacoustics: facts and models. 3rd ed. Berlin: Springer; 2007. [6] Fiebig A, Sottek R. Contribution of peak events to overall loudness. Acta Acust United Acust 2015;101(6):1116–29. [7] Filipan K, Bockstael A, De Coensel B, Schönwiesner M, Botteldooren D. A novel auditory saliency prediction model based on spectrotemporal modulations. Proceedings of the 22nd international congress on acoustics, 2016. Buenos Aires, Argentina. [8] Fredrickson BL, Kahneman D. Duration neglect in retrospective evaluations of affective episodes. J Pers Soc Psychol 1993;65(1):45–55. [9] Gagne J P, Besser J, Lemke U. Behavioral assessment of listening effort using a dual-task paradigm: a review. Trends Hear 2017;21. 2331216516687287. [10] Genuit K, Fiebig A. Psychoacoustics and its benefit for the soundscape approach. Acta Acust United Acust 2006;92(1):952–8. [11] Kahneman D. Attention and effort. Englewood Cliffs (NJ): Prentice-Hall; 1973. [12] Kalinli O, Narayanan SS. A saliency-based auditory attention model with applications to unsupervised prominent syllable detection in speech. Proceedings of InterSpeech. CD-ROM, 2007. [13] Kayser C, Petkov CI, Lippert M, Logothetis NK. Mechanisms for allocating auditory attention: an auditory saliency map. Curr Biol 2005;15(21):1943–7. https://doi.org/10.1016/j.cub.2005.09.040. [14] Knudsen EI. Fundamental components of attention. Annu Rev Neurosci 2007;30(1):57–78. https://doi.org/10.1146/annurev.neuro.30.051606.094256. [15] Koch I, Poljac E, Müller H, Kiesel A. Cognitive structure, flexibility, and plasticity in human multitasking-An integrative review of dual-task and taskswitching research. Psychol Bull 2018;144(6):557–83. https://doi.org/ 10.1037/bul0000144. [16] Lavie N, Hirst A, de Fockert JW, Viding E. Load theory of selective attention and cognitive control. J Exp Psychol Gen 2004;133(3):339–54. [17] Mattys SL, Wiget L. Effects of cognitive load on speech recognition. J Mem Lang 2011;65(2):145–60. https://doi.org/10.1016/j.jml.2011.04.004. [18] Namba S, Kuwano S. Relation between overall noisiness and instantaneous judgment of noise and the effect of background noise level on noisiness. J Acoust Soc Japan 1980;1(2):99–106. [19] Öhrström E, Björkman M, Rylander R. Noise annoyance with regard to neurophysiological sensitivity, subjective noise sensitivity and personality variables. Psychol Med 1988;18(3):605–13. [20] Raveh D, Lavie N. Load-induced inattentional deafness. Attention Perception Psychophys 2015;77(2):483–92. https://doi.org/10.3758/s13414-014-0776-2. [21] Rozin P, Royzman EB. Negativity bias, negativity dominance, and contagion. Personality Soc Psychol Rev 2001;5(4):296–320. [22] Steffens J. Realism and ecological validity of sound quality experiments on household appliances. In: Kang J, Chourmouziadou K, Sakantamis B Wang, Hao Y, editors. Soundscape of European cities and landscapes: (e-book). pp. 132– 135. [23] Steffens J, Guastavino C. Trend effects of momentary and retrospective soundscape evaluations. Acta Acust United Acust 2015;101:713–22.

J. Steffens et al. / Applied Acoustics 164 (2020) 107278 [24] Steffens J, Steele D, Guastavino C. Situational and person-related factors influencing momentary and retrospective soundscape evaluations in day-today life. J Acoust Soc Am 2017;141(3):1414–25. https://doi.org/10.1121/ 1.4976627. [25] Stroop JR. Studies of interference in serial verbal reactions. J Exp Psychol 1935;18(6):643. [26] Tordini F, Bregman AS, Cooperstock JR. Toward an improved model of auditory saliency. Proceedings of the international conference on auditory display (ICAD) 2013, 2013. Lodz, Poland. [27] Torija AJ, Ruiz DP, de Coensel B, Botteldooren D, Berglund B, Ramos-Ridao Á. Relationship between road and railway noise annoyance and overall indoor sound exposure. Transp Res Part D: Trans Environ 2011;16(1):15–22. https:// doi.org/10.1016/j.trd.2010.07.012.

7

[28] Västfjäll D. The ‘‘end effect” in retrospective sound quality evaluation. Acoust Sci Tech 2004;25(2):170–2. [29] White K, Bronkhorst AW, Meeter M. Type of activity and order of experimental conditions affect noise annoyance by identifiable and unidentifiable transportation noise. J Acoust Soc Am 2018;143(4):2165. https://doi.org/ 10.1121/1.5031019. [30] Wu Y-H, Aksan N, Rizzo M, Stangl E, Zhang X, Bentler R. Measuring listening effort: driving simulator versus simple dual-task paradigm. Ear Hear 2014;35 (6):623–32. https://doi.org/10.1097/AUD.0000000000000079. [31] Yao B, Belin P, Scheepers C. Silent reading of direct versus indirect speech activates voice-selective areas in the auditory cortex. J Cognit Neurosci 2011;23(10):3146–52.