Using soundscape simulation to evaluate compositions for a public space sound installation^a)

In urban planning, sound is often considered as an environmental burden that should be mitigated. To address the deleterious effect noise exposure has on public health,¹ most environmental policies focus on noise control procedures (e.g., Refs. 2 and 3) Yet, sound plays a complex role in the way cities are experienced, and reducing sound levels alone does not necessarily lead to an improved quality of life.^4,5 Rather than being seen as a nuisance that has to be mitigated, sound can instead be considered as a resource in relation to other urban planning considerations, through the soundscape approach.⁶ Soundscape [defined by the International Standard Organization (ISO) as the “acoustic environment as perceived or experienced and/or understood by a person or people, in context”⁷] enables more complex representations of sound and allows one to envisage both the positive and negative outcomes it can have on the quality of urban environments.⁸ 

In this regard, a growing body of literature has focused on implementing and documenting design plans to preserve or improve existing soundscapes through soundscape interventions.^9,10 Specifically, there is increasing evidence that the deliberate introduction of new sound elements to existing acoustic environments can benefit urban public spaces. Some of these studies demonstrated that added sounds could positively affect people's behavior, for instance by fostering social interactions,^11–14 increasing duration of stay and favor activities such as chatting or eating/drinking,^15,16 and even by affecting walking pace ^17,18 or crowd density and walking patterns.¹⁹ Other field studies showed that added sounds can improve soundscape evaluation, through global assessments ^20,21 or by increasing evaluations on variables such as eventfulness and excitement,^22, pleasantness,^23, calmness ²⁴ or even by reducing the perceived sound level. In laboratory settings, the effects of adding natural sounds such as birds or streams on soundscape evaluation have been extensively investigated, studies showing for instance that such sounds could be evaluated as preferable,²⁵ reduce perceived loudness and increase pleasantness,²⁶ increase soundscape quality,²⁷ and eventfulness.²⁸ Otherwise, a recent study by Oberman and colleagues evaluated the impact of three sound art interventions on soundscape measurement and showed different impacts for each intervention on perceived pleasantness, calmness, excitement, and appropriateness.²⁹ 

A broad range of methods have been used to evaluate the perceptual and affective attributes of soundscape interventions. Among them, soundscape scales have been widely spread and refined in recent years. The most broadly used protocol is the Swedish Soundscape Quality Protocol (SSQP),³⁰ featured as the method A in the ISO/TS 12913-2:2018.³¹ It comprises a set of scales based on principal components underlying soundscape evaluation, pleasantness and eventfulness, established by Axelsson and co-workers³² The method proposed in the standard also includes a scale relative to appropriateness.³¹ Otherwise, Payne and Guastavino proposed the Perceived Restorativeness Soundscape Scale (PRSS) to assess the restorative potential of sound environments in terms of being-away, extent-coherence, compatibility, and fascination,³³ while Welch et al. developed a set of seventeen semantic differentials to measure the affective properties and the qualities of soundscape, though a study involving creative writing.³⁴ The use of both the SSQP and the PRSS led to statistically significant results when evaluating or comparing sound art interventions (e.g., Refs. 23, 24, and 29). However, the study led by Oberman and colleagues showed that the SSQP alone could provide ambiguous responses regarding the impact of sound interventions on the eventfulness component and could be optimized.²⁹ If soundscape scales provide a subjective evaluation of soundscape among a set of predetermined criteria, more in-depth information about the various associations, emotions and feelings associated with the perception of sound environments in the presence of sound art can be obtained through qualitative methods such as ethnographical research (e.g., Ref. 35) or open-ended interviews (e.g., Ref. 11). Ultimately, quantitative and qualitative data collection methods can be integrated together through methodological triangulation.^31,36

To collect soundscape data, laboratory and in situ methods coexist, each having their own advantages and limitations.³² Laboratory experiments involve the simulation or reproduction of soundscapes and provide more control on the sound environment. Among the existing soundscape reproduction or simulation techniques, Ambisonics (see Ref. 37) has been increasingly used in recent years and is usually considered ecologically valid.^39–41 Generally, studies involving the simulation of soundscape interventions artificially integrate added sounds to a pre-recorded sound environment,^26–29 and similar technologies have been proposed as soundscape simulation tools to help professionals of the built environment anticipate the impact of urban design decisions on soundscape,^41,42 or as a tool for soundscape composition.⁴³ Still, the only laboratory study evaluating the impact of sound art interventions on soundscape recorded them in situ and compared them with recording positions in which they were not audible through a virtual soundwalk approach.²⁹ 

Overall, studies on the impact of added sounds on soundscape mostly focus on the introduction of either natural sounds or generic music to existing sound environments, and the few studies on the impact of sound art installations on soundscape ^{13,22,24,29,35,44} were systematically carried out a posteriori, i.e., once the sound installations were already deployed. If such studies revealed the potential for sound art to improve urban soundscapes, the methods used provide little room for sound artists to implement perceptual feedback within the compositional process. Yet, people's reception and perception has been an essential consideration for sound artists from the very emergence of sound installation art (see Ref. 44 for a review). Similar to the way soundscape researchers emphasize context,^7,46 the design of public space sound installations is usually thought of in relation to a multitude of site-specific aspects, including perception (e.g., Refs. 47 and 48). Hence, evaluating sound installations' impact on soundscape before deployment through soundscape simulation would benefit sound artists by informing their composition at the early stages of creation process.

The relationship between a sound installation and its existing environment can take many forms, depending on the artistic intention and on site-specific considerations. Through this variety, common composition strategies and issues have been theorized in the literature (see Ref. 45). Livingston proposed a taxonomic division between strategies for adding sounds in public spaces: integrated/site-specific/background (added sounds that subtly blend in with the existing sound environment so that they can stay unnoticeable) versus oppositional/borrowed/foreground (added sounds that are clearly noticeable).⁴⁹ Similarly, Botteldooren et al. proposed three design imperative for soundscape design: backgrounded (the introduced sounds stay unnoticed), supportive (the added sounds enhance the existing experience), and focused (the added sounds become a point of interest).⁵ Many creators also investigated the notion of non-energetic masking (also called informational masking) where added sounds purposely distract listeners' attentions from other sources,^{35,44,50–52} and the phenomena has also been studied in the soundscape literature (e.g., Refs. 27, 29, and 53). The existence of shared composition strategies does not imply that there is an obvious way to operationalize them, and approaches can be as diverse as there are sound artists (for instance, see how two approaches to generate oppositional sounds may differ in Refs. 50 and 52). Nevertheless, recent works showed that different artistic propositions could lead to different perceptual impacts on soundscape,²⁹ and further work is needed to better understand the link between sound art composition strategies and their impact on soundscape.

The present study was conducted in the context of a research-creation collaboration between the authors in this paper around the permanent sound art installation Niches Acoustiques. Created by sound artist Nadine Schütz (the other authors are researchers in the fields of soundscape, music technology, and sound design), this laureate project of Budget Participatif de la Ville de Paris will lead to the planned, permanent deployment of the sound installation on the forecourt of the new Judicial Court of Paris, France. The overall intention of Niches Acoustiques is to create an appeasing, beneficial, and varied auditory foreground which reduces the perceived dominance of annoying and monotonous noises while opening up the courthouse's forecourt to an urban narrative which connects it to the urban neighborhood. The title of the installation, Niches Acoustiques, is borrowed from the bioacoustics Niche Hypothesis, according to which the co-existence of diverse species, particularly in densely populated areas, is fostered by the spectral and temporal differentiation of their vocalization patterns.^54,55 The “Acoustic Niches” sound installation project interprets, activates, and transposes this principle in a (psychoacoustically informed) spectrotemporal means to modify soundscape perception by adding distinct sonic ambiences on the forecourt, in a non-intrusive way, with low volumes of the added sounds, and through a “complementary composition” approach. In this context, the notion of auditory foreground refers to added sounds which are acoustically, semantically, and/or spatially distinct from the pre-existing sound environment. The collaboration (including the present study) aims at informing the composition of Niches Acoustiques and evaluating the impact of the sound art intervention through soundscape evaluations at different stages of the composition process. We report here on a laboratory study involving the simulation of compositional sketches of the sound installation (in the form of short excerpts) in the presence of a reproduction of the forecourt's existing sound environment using a soundscape simulation tool developed and validated in a previous study.⁵⁶ To compare the impact of the different sound installation sketches on soundscape evaluation, participants familiar with the forecourt of the Judicial Court were invited to evaluate each excerpt with a set of semantic differential scales and were then asked to respond to a semi-structured interview. The present study addresses the following research questions:

• RQ1: How do public space users evaluate everyday city soundscape modified by the presence of sound art?

• RQ2: How do public space users' soundscape evaluations vary for different sound art composition strategies?

Considering that the composition of a sound installation is highly sensitive to site-specific considerations as well as to its artistic intention, answering the second research question required to provide a classification of composition strategies that can be generalized to other sound art interventions. To do so, the methodology applied by the sound artist during the creation of the sound installation sketches was structured so that they could be gathered into broader composition strategies, and we report here on the impact of these composition strategies on soundscape rather than on an excerpt-to-excerpt analysis. To allow for their comparison, these strategies were subsequently positioned within an Abstract (sounds that cannot be ascribed to any real or imagined provenance)/Referential (recorded sounds that suggest or at least do not hide the source to which they belong) dichotomy, as proposed by Landy.^57,58 Our intuition, prior to the experiment, was that Abstract sounds would be perceived as being more oppositional because of their unexpected nature while Referential sounds would be perceived as being more integrated because they could more easily blend in with the existing sound environment.

To wrap up, the goal of this study is not to impose compositional principles to sound artists or to replace the artistic intention—each sound installation having its own artistic statement and design goals—but rather to systematically evaluate the impact of sound art composition strategies that are broad enough to be transferred to other sound art interventions. The present study also aims at investigating the productivity of the proposed research-creation methodology, specifically assessing the relevance of the soundscape simulation tool for evaluating sound art interventions. Meanwhile, the proposed methodology is intended to help inform the design of Nadine Schütz's Niches Acoustiques sound installation by anticipating its impact on soundscape evaluation.

The present study uses a soundscape simulation tool previously developed and validated through listening tests. Information about the tool development and validation is presented in Ref. 56. The simulation consists of the reproduction of higher-order ambisonics (HOA) field recordings of the sound environment on site, along with the auralization of added sounds, yielding composition sketches of the sound installation using a 3D acoustic model of the site (simulating early reflections and late reverberation) converted to HOA streams. The resulting soundscape is presented over a loudspeaker array in a listening room for soundscape evaluation using semantic differential scales (Fig. 1). All components of the simulation, from HOA encoding and auralization to playback and graphical user interface are implemented in Cycling '74 Max.⁵⁹ 

The data on the sound environment of the Judicial Court of Paris' forecourt was collected during a measurement campaign in spring 2021, detailed in Refs. 56 and 60. We conducted punctual HOA recordings and sound level measurements throughout the public space across five sessions covering different activity levels (weekday morning, afternoon and evening; weekend morning and evening). During each session, 5-min recordings were made across measurement points gridding the square (Fig. 2). At each position, we measured equivalent sound pressure and third-octave levels with a B&K 2250 sound level meter together with fourth-order ambisonics recordings with an mh Acoustics em32 Eigenmike.^37,61,62 All measurements were oriented towards the direction opposite to the Judicial Court, at a height of 1.3 m.

The listening test focuses on the comparison of various sound installation sketches. To ensure smooth transitions between these conditions, we designed a continuous baseline sound environment by concatenating fourth order HOA separate excerpts from the measurement campaign (not to be confused with the Referential excerpts presented in Sec. II B). The excerpts were selected to ensure that they were representative of the public space's average level of activity, spatially close enough to each other (see the included positions in Fig. 2), and did not contain salient sounds so that participants focus on the added sounds during the listening tests (see Ref. 57 for more detail). Excerpts, selected during joint listening sessions with two of the authors, ranged from 30 s to around 2 min. A total of 38 excerpts were crossfaded in fully random orders using python's reathon library to generate Reaper scripts.⁶³ In other words, a Baseline using the same 38 excerpts was generated with a different, randomized order for each participant so that they would listen to a different superposition of the background recordings and the added sounds, and to ensure that the observed effects would be independent of the temporal evolution of the background sound environment. A 3-s crossfade between excerpts was applied to provide short yet smooth and unnoticeable transitions. In total, the Baseline lasted around 45 min, and was looped in the experiment. A 40-min sound level measurement of the calibrated Baseline (without the 4 dB padding mentioned Sec. II A 5) was conducted in the listening room using a B&K 2250. We found a L_Aeq,40min of 61.9 dBA and a L_A10-L_A90 of roughly 6 dBA, confirming that the chosen excerpts were representative of an average level of activity in the parvis while remaining sufficiently stable.⁶⁰ 

The auralization method is only summarized here, a complete description of the procedure is available in Ref. 56. Initially in the form of monophonic excerpts, the added sounds were spatialized using IRCAM's EVERTims framework⁶⁴ integrated in Max's spat∼ library.⁶⁵ The 3D model of the forecourt from which the auralization is based on includes the main surfaces of the forecourt, the position of the sound sources (the future sound installation's speakers), as well as of the listener (see Fig. 2). Upon reception of the 3D model, EVERTims computes a list of image sources that correspond to the early reflections of the space, while the late reverberation is simulated with a Feedback Delay Network.^64,66 The output of the auralization unit is ultimately encoded into fourth order HOA streams with spat∼.³⁷ As physical parameters such as the reverberation time of the court were missing, we fined-tuned the auralization through analytical listening with sound experts, and validated it in a preliminary listening test, as reported in.⁵⁶ Finally, the mastering of the auralized added sounds was operated in two steps. First, all excerpts' loudness was normalized with pyloudnorm, a python implementation of the ITU-R BS.1770-4 standard for loudness measurement.⁶⁷ Then, the gain was adjusted by ear in the presence of the simulated sound environment for each of the excerpts by author N.S. through a dedicated Max patch, to mimic the protocol that will be carried out during the sound installation's deployment.

The experiment was conducted at IRCAM's studio 4, an acoustically treated listening room, over a hemispherical dome of 24 Amadeus PMX 4 speakers ⁶⁸ placed on four height levels beginning at ear level (Fig. 3). Encoding and decoding parameters were chosen during joint listening sessions including three of the authors, who compared in situ listening with the reproduced sound environment. The Eigenmike 32 signals were encoded into a fourth-order HOA stream with spat ∼ using Tikhonov regularization.³⁷ At the output of the system, the auralized added sounds and the sound environment—both in the form of fourth-order HOA streams—were summed and decoded with spat∼ using energy preserving method with max-r_E weighting function.⁶⁹ 

The 5-min A-weighted equivalent levels (L_Aeq) values captured during the measurement campaign were compared with similar measurements in the listening room's sweet spot at a height of 1.3 m to calibrate the reproduction levels of the baseline sound environment.⁶² However, we decided to reduce the reproduced soundscape sound level by 4 dB with respect to in situ measurements, as the level of the reproduced soundscape was perceived by the authors as higher than in situ. Observed in several laboratory studies (e.g., Refs. 30, 70, and 71) this phenomenon could be related to the relatively high sound levels in situ, with 5-min L_Aeq typically ranging from 60 to 70 dBA.⁶⁰ A 4-dB reduction was unanimously perceived as the best match to replicate the perceived loudness in situ.

The artistic design goals of the sound installation have been presented in the Introduction. This section focuses on the composition strategies for the sketches (i.e., the sound installation's sound materials) evaluated in the present study. The primary sound content of the new auditory foreground introduced by the Niches Acoustiques sound art installation is inspired by the site of the installation and by the artistic intention to reinforce its urban relatedness; they combine sonic fragments recorded in the forecourts' urban surroundings, which are edited into new imaginary sonic landscape scenes and supplemented by their own “musical shadows.” The present study focuses on these two types of sound materials, the original Referential sonic landscape scenes and their more Abstract musical shadows, derived from the original scenes through two different editing procedures corresponding to two degrees of (preprocessed) abstraction: medium-abstraction and total-abstraction.

The Referential scenes were based on field recordings (not to be confused with the baseline, they are unrelated to the measurement campaign described Sec. II A 1) in various streets and public spaces surrounding the Forecourt, using stereo and directional mono microphones. These recordings were first cut into semantic and sonic units, differentiating between more continuous textures and more punctual events. Given the intended brevity of the composition excerpts, only a few of these units were used, selected for their semantic and sonic variety. The resulting Referential scenes were composed following a semantic and associative approach, and grouped into two different sound types: narrative scenes, combining various sound events with widely varying dynamics (instruments-paths, birds-games and voices, see for instance Mm. 1), and natural tones, which present simpler, more continuous sound textures (articulate waters, and wind-rustling foliage).

The medium-abstraction applies resonance effects on the referential compositions using pitch-based harmonic resonators tuned to pure octave chords or triad chords with some deviating notes, bandpass filters inverse to the Baseline's average spectral energy profile, and spectral freezing. The total-abstraction results from a melodic audio-to-midi-based synthesis using alternatively Ableton Live 11 Suite's⁷² built-in Audio-to-MIDI conversion function and Max's IM-I analyzer plugin (mapped to a E minor triad), directing the performance of a virtual instrument (a virtual bell piano included in Ableton Live 11's core library).

In addition to this evaluation of three distinct positions within the artistically imagined referential-to-abstract continuum, the study also explored another composition technique, combining the basic referential compositions with their different abstractions in hybrid, steady mixes—simple combination of excerpts using signal addition—or fluctuating mixes—mixes that are constantly fluctuating between the referential scene and its abstract counterpart using an envelope automation in the form of a triangle wave with a period of roughly 40 s (0.025 Hz). These mixes correspond to the ultimate artistic intention to alternate between these different types of sounds in the final installation, based on environmental data captured on-site by acoustic and meteorological sensors.

Altogether, 26 installation excerpts (composed added sounds) resulting from the above-described composition strategies, each 95 s long, were presented to the participants. Five concrete, Referential excerpts were chosen based on samples from the original field recordings, and were either event-based (narrative scenes, see for instance Mm. 1) or texture-based (natural tones).

To reach a reasonable experiment duration, we had to make a selection from all the possible compositional variations derived from these five Referential compositions. Four selected excerpts made with four different resonant and spectral manipulations applied on the Referential excerpts presented a medium-abstraction compositional position (Resonated excerpts, see Fig. 4 and Mm. 2). The four manipulations alternate between and combine in different ways bandpass filters, pitch-based harmonic resonators and spectral freezing.

In comparison, total-abstraction was presented by two Synthesized excerpts, created with two different synthesizers (virtual instruments) based on a pitch-based audio-to-midi conversion of the concrete excerpts (see Fig. 4 and Mm. 3).

The remaining mixed excerpts are combinations of Referential and mid-abstraction (Resonated) excerpts and of Referential and total-abstraction (Synthesized) excerpts, always maintaining the interplay of original referential sound and the abstraction derived from it. These compositions explore the two above-described mixing techniques, fluctuating mixes and steady mixes (see Fig. 4, Mm. 4, Mm. 5, and Mm. 6).

The final selection of 26 excerpts comprises five Referential excerpts, four Resonated excerpts, two Synthesized excerpts, five mixes fluctuating between Referential and Resonated excerpts (Ref/Res Fluctuating), five mixes fluctuating between Referential and Synthesized excerpts (Ref/Syn Fluctuating), and five steady mixes superimposing Referential and Synthesized excerpts (Ref/Syn Steady). To characterize the acoustic levels of the excerpts, a series of 1min30s-sound level measurements was conducted in the listening room using a B&K 2238, in the presence of the added sounds alone. Mean L_Aeq1min30s and L_AFmax are reported in Table I and show that levels have similar ranges across composition strategies, except for Referential excerpts which have lower sound levels.

Twenty participants were recruited for the evaluation, including 9 Judicial Court workers (age = 44.1 ± 10.2) and 11 residents (age = 44.8 ± 13.5). All of them were familiar with the studied space to ensure ecological validity and self-reported normal hearing. Residents were recruited by distributing flyers and displaying posters in the neighborhood, while workers were recruited through an email sent from the borough to the Judicial Court's mail list. All participants reported using the public space several times a month, while a majority used it almost daily (see Table II). They received a 20€ compensation for their participation.

Participants were continuously exposed to the baseline sound environment described Sec. II A 2 and evaluated it in the presence of added sounds pertaining to the different composition paradigms described Sec. II B. There was a total of 28 unique conditions, including the 26 compositions described in Sec. II B as well as 2 conditions with no added sounds, including only the background sound environment at different, random moments in the experiment (Baseline). In total, the experiment featured 6 composition strategies (Referential, Resonated, Synthesized, Ref/Res Fluctuating, Ref/Syn Fluctuating, and Ref/Syn Steady) for a total of 34 excerpts, including 6 duplicates (1 × Referential, 2 × Ref/Syn Fluctuating, 2 × Ref/Res Fluctuating, 1 × Ref/Syn Steady) to measure test-retest reliability.

Participants were seated at the sweet spot of the speaker dome with a rotating chair fixed to the floor (Fig. 5—left) and evaluated the excerpts through a Max interface displayed on a 21.5 inches monitor using an external mouse (Fig. 5—right). In presence of the background sound environment, participants were first presented with a photograph and an aerial view of the studied site (Fig. 6) for 40 s, while being asked to try to recall the space in their memory. They were then asked to listen to the 34 excerpts and evaluate their soundscape through a set of continuous scales (Fig. 5—right and Table III). All excerpts and all scales were presented in a fully randomized order. Within each trial, each excerpt was presented to the participants for 15 s before the questionnaire appeared to ensure they listened and acclimated to the soundscape. They could then answer the questionnaire for 75 s before the end of the excerpts that lasted a total of 95 s. However, they were able to skip to the next excerpt with a dedicated button (Fig. 5—right, top right corner) that appeared after 50 s (including the initial 15 s), provided that they had filled all scales. A 10-s transition was set to smoothly switch between excerpts either when participants used the skip button or at the end of the excerpt. Depending on the participant, trials lasted between 60 s and 100 s for a total testing time between roughly 40 min and 1 h.

The participant ran a practice trial with the experimenter before starting the experiment, to help them familiarize themselves with the task. An optional break was automatically triggered at the halfway point of the experiment (after the 17th excerpt).

At the end of the experiment, the experimenter conducted a semi-structured interview⁷³ with the participants through an interview guide of six questions (Table IV). Participants were recorded with a Zoom H4N pro.⁷⁴ Interviews lasted from roughly 10 to 30 min. The entire study lasted around 90 min in total.

Participants were asked to evaluate soundscapes across nine continuous semantic differential scales (Table III). The elaboration of the questionnaire is the result of shared reflection between the sound artist (author N.S.) and the scientific team (all the other authors) to provide an evaluation tool both suitable for comparing soundscape interventions and relevant with regard to the design goals of the installation, while ensuring questionnaire brevity. In short, we needed to investigate not only how the added sounds could affect the perceived affective quality of the Forecourt's soundscape,³⁶ but also their emotional impact,³⁴ as well as how they could evoke novelty (i.e., less familiar soundscapes) and variation in an urban soundscape dominated by traffic, as it is one of the goals of the future sound installation. The questionnaire does not comprise scales related to the soundscape's eventfulness, as it is believed to be more related to human sounds, especially its French translation.^75,76 Rather, we used questions relative to variation and emergence, in an attempt to capture the attributes of eventfulness that are not related to notions of human activities or liveliness, but instead to more analytical dimensions related to the perceived number of sources and their dynamics.⁷⁶ The order of the scales was fully randomized between participants.

The experiment was followed by semi-structured interviews based on six questions (Table IV). The goal of the interviews was to obtain interpretative guidance on the results obtained with the scales, but also to identify participants' opinions on the added sounds and their relationship with the forecourt of the Judicial Court. Quotes reported in the results were translated from French by author V.F.

Statistical analyses were computed in r 4.0.3 with rstudio 2022.12.0 + 353 for Windows, with a statistical significance level of 0.05. Prior to the analysis, ratings were collapsed for each participant across duplicate conditions with the mean value, including the two conditions with no added sounds. The data were highly non-normal, whether univariate or multivariate. For this reason and because of the relatively small sample size, we decided to conduct semi-parametric and non-parametric analyses when relevant. To investigate the two research questions, we conducted two types of statistical analysis.

To validate the questionnaire instrument and to determine components underlying soundscape judgements in the presence of sound art while increasing interpretability of the results [RQ1], we ran a principal components analysis (PCA) on the 9 items with oblique rotation (oblimin) using the psych package version 2.0.9.⁷⁸ Prior to the PCA, the Kaiser-Meyer-Olkin verified the sampling adequacy KMO = 0.76 (“middling,” according to Ref. 80) and all KMO values for individual items were above 0.6, which is above the acceptable limit of 0.5. Subsequent analyses were made from the resulting components of the PCA, shown in italics.

To investigate the effect of the added sounds on soundscape evaluation and to compare composition strategies [RQ2], we conducted two semi-parametric repeated measures MANOVAs with the components from the PCA as dependent variables and the composition strategy as independent variables using the multRM function from the manova.rm package, version 0.5.3.⁸⁰ The first MANOVA was conducted with the composition strategy excluding mixes (Baseline; Referential; Resonated; Synthesized) as a within-subject factor to compare the influence of added sounds along the referential/abstract paradigm. The second MANOVA was conducted on all composition strategies (Referential; Synthesized; Ref/Res Fluctuating; Ref/Syn Fluctuating; Ref/Syn Steady) in the within-subject factor to evaluate the influence of the combination of abstract and referential sounds on soundscape evaluation. Since the covariance matrix was singular in some cases and because of the small sample size, we used the Modified ANOVA-type statistic (MATS) and wild bootstrap resampling method for p-values, with 10 000 iterations.⁸¹ For both tests, follow-up semi-parametric repeated measures ANOVA were conducted with the same independent variables using the ANOVA-type statistics (ATS) and wild bootstrap resampling also with 10 000 iterations from the RM function in the MANOVA.RM package. Finally, we ran post hoc Wilcoxon signed rank exact tests to compare each condition, with Holm p-value correction using the r package stats.⁸² For each Wilcoxon test, we report on p-values in addition to the r effect size estimated using the package rstatix.^83,84 For the MANOVAs and subsequent analyses, data were collapsed for each participant with the mean value corresponding to each condition.

For interpretative guidance on the quantitative results, follow-up interviews where transcribed and analyzed using nvivo 1.7.1 for Windows,⁸⁵ using open coding to identify emerging themes.⁸⁶ 

The 20 participants rated the Baseline soundscape as mildly pleasant (Mdn = 57.0) and neither soothing nor arousing (Mdn = 50.2). The reproduction of the forecourt's soundscape was perceived as representative of the forecourt of the Judicial Court as participants rated the Baseline soundscape as appropriate with the forecourt (Mdn = 71.5), familiar (Mdn = 84.7) and reflecting the character of the space (Mdn = 77.2). Six participants stated in the follow-up interviews that the experiment was realistic and representative of the forecourt (e.g., “I closed my eyes and I really had the sensation that I was on the forecourt”), while only one participant questioned its realism (“I don't know whether car noises where artificial or not”). The systematic exclusion of salient or disruptive sounds from the Baseline's concatenated excerpts is reflected in participants' ratings, as they rated it as being stable over time (Mdn = 35.7) and with an average emergence (Mdn = 49.5). Conversely and despite the Baseline's perceived Sound Level being mildly low (Mdn = 61.2), six participants described the listening experiment as globally loud, while one participant stated that it was quieter than in situ. Table V shows the full list of values for the Baseline condition.

To assess the test-retest reliability of the participants' ratings, the values obtained for the 6 excerpts that appeared twice were correlated using Pearson's coefficient of correlation. The test-retest reliability was found to be poor. The mean correlation across the 9 items was 0.5 with a range of 0.41–0.60. Results were similar when comparing the two baseline conditions with a mean correlation of 0.56 across all scales, with the exception of the scales Varied (r = 0.21) and Coherent (r = 0.84). However, Cronbach's α values for internal consistency suggested that the scales were reliable to some extent (Table VI).

An initial analysis was done to obtain eigenvalues for each component in the data. Based on the scree plot and in accordance with Kaiser's criterion (eigenvalues > 1.0), three components were retained that explained 74% of the variance.

The items that cluster on the same components suggest that component 1 (Pleasant) represents the soundscape's appreciation (29%), component 2 (Familiar) is associated with the character, familiarity and appropriateness of the soundscape (28%), while component 3 (Varied) is linked to the variety and emergence of the soundscape (16%). See Table VI for component loadings. Both the soundscape's appreciation and character had high reliabilities (Cronbach's α = 0.84 and 0.83, respectively). However, the soundscape's variety had relatively low reliability (Cronbach's α = 0.60). These values suggest that the participants' ratings were reliable despite the poor test-retest scores. All items mostly load on only one component with a complexity ranging from 1.0 to 1.3, except for Coherent, which loaded both on soundscape's appreciation and character with a complexity of 2.

The three components emerging from the data show similarities with some of the components first identified by Axelsson et al.⁷⁵ and later confirmed by Tarlao et al.⁷⁷ See Sec. IV for further comparisons. These components will be used in Secs. III C–III E to compare the composition sketches.

In this section, we compare the three composition strategies (Referential, Resonated, and Synthesized excerpts) with the Baseline condition. The overall repeated-measure MANOVA on these four conditions shows a significant effect of composition strategy on the three components (MATS ≈ 98.16, p < 0.001). Follow-up repeated measures ANOVAs show a significant effect of composition strategy on all components (Pleasant: ATS ≈ 4.5, p ≈ 0.009; Familiar: ATS ≈ 29.5, p < 0.001; Varied: ATS ≈ 16.7, p < 0.001).

Descriptive statistics are shown in Table X and in Fig. 7, while results of the post hoc tests are shown in Table VII. Together, they reveal that

• For the Pleasant component, the Synthesized excerpts were significantly less appreciated than the Referential condition. However, there were no significant differences between the other conditions, despite moderate effect sizes suggesting that the Synthesized excerpts may be less appreciated than the Referential excerpts (r = 0.49) and the Resonated excerpts (r = 0.40).

• For the Familiar component, all conditions were significantly different from one another except for the Resonated and Synthesized excerpts. Specifically, all excerpts were perceived as being strongly less familiar than the Baseline. In addition, both Resonated and Synthesized excerpts were perceived as being less familiar than Referential excerpts.

• For the Varied component, the Baseline condition was perceived as being less varied than any other condition. Despite being significant, effect sizes suggest that Resonated (r = 0.49) and Synthesized excerpts (r = 0.43) were perceived as being more varied than Referential excerpts.

If we did not observe the impact of composition strategies on soundscape appreciation except for Synthesized excerpts which were less appreciated that Referential excerpts, results suggest that all composition strategies substantially affected soundscape familiarity and variety. Specifically, the Resonated and Synthesized excerpts more strongly affected the familiarity compared to the Referential excerpts, while effect sizes suggest that they also more strongly affected the perceived variety.

In this section, we only report on results relative to hybrid composition strategies (i.e., mixes between referential and abstract compositions) as the Referential, Resonated, and Synthesized excerpts have been discussed before. The overall repeated-measure MANOVA on the conditions excluding the Baseline (Concrete; Referential; Synthesized; Ref/Res Fluctuating; Ref/Syn Fluctuating; Ref/Syn Steady) shows a significant effect of composition strategy on the three components (MATS ≈ 41.65, p < 0.001). Follow-up repeated measures ANOVAs show a significant effect of excerpt on all components (Pleasant: ATS ≈ 3.8, p = 0.007; Familiar: ATS ≈ 5.4, p = 0.005; Varied: ATS ≈ 6.6, p = 0.002).

Descriptive statistics are shown in Table IX and Fig. 8, while results of the post hoc tests are shown in Table VIII. Together, they reveal that

• For the Pleasant component, there were no significant differences between conditions, despite moderate effect sizes suggesting that Ref/Syn Fluctuating and Ref/Syn Steady excerpts were less pleasant than Referential excerpts (r = 0.61 and r = 0.49, respectively) and that Ref/Res Fluctuating and Ref/Syn Steady excerpts were more pleasant than Synthesized excerpts (r = 0.45 and r = 0.47, respectively).

• For the Familiar component, Ref/Syn Fluctuating and Ref/Syn Steady excerpts were significantly less familiar than Referential excerpts. Despite being significant, moderate effect sizes suggest that the other mix strategy (Ref/Res Fluctuating) was also being perceived as less familiar than Referential excerpts (r = 0.48). Effect sizes also suggest that the Ref/Res Fluctuating excerpts were perceived as more familiar than Resonated excerpts (r = 0.44), that the Ref/Syn Steady and Ref/Res Fluctuating excerpts were more familiar than the Synthesized excerpts (r = 0.47 and r = 0.43, resp.), and finally that Ref/Res Fluctuating excerpts were more familiar than the Ref/Syn Fluctuating excerpts (r = 0.48).

• For the Varied component, all three mix strategies (Ref/Res Fluctuating, Ref/Syn Fluctuating and Ref/Syn Steady) were significantly more varied than Referential excerpts. Despite being significant, moderate effect sizes suggest that all three mix strategies were also perceived as being more varied than Resonated excerpts (r = 0.54, r = 0.55, and r = 0.42, respectively).

Results indicate that soundscapes with mixed excerpts might be more pleasant than those with Synthesized excerpts and less pleasant than those with Referential excerpts, although we could not identify significant differences. Mixed excerpts were also less familiar and more varied than those with Referential excerpts, while effect sizes suggest that they tended to be more familiar than those with Synthesized and Resonated excerpts, as well as more varied than those with Resonated excerpts. Otherwise, we did not detect significant differences between the ratings of the three hybrid composition strategies across all components, although effect sizes suggest that Ref/Res Fluctuating excerpts might have been perceived as more familiar than Ref/Syn Fluctuating excerpts. Overall, results suggest that soundscapes with hybrid compositions fall in between purely referential and purely abstract compositions in terms of pleasantness and variety, while they were perceived as being more varied than Referential excerpts, and potentially than Resonated excerpts.

During the interviews, all twenty participants referred to at least one referential sound such as birdsong (N = 17), wind/rain (N = 9), kids playing (N = 14), or water (N = 13) while sixteen participants mentioned abstract sounds through diverse associations (e.g., “some kind of music,” “electroacoustic music,” “electronic sounds,” “metallic noise,” “crystal sounds”). These mentions provide interpretative guidance on the quantitative results. Since it is unclear as to which specific condition the participants actually referred to, the qualitative exploration that follows is only associated with Referential and Abstract composition strategies when relevant. It should be noted that Referential sounds are more often mentioned than Abstract sounds, likely because it is easier to describe identifiable sounds (in terms of sound source or action producing sound) than unidentifiable sounds as the latter cannot be ascribed to a specific cause.⁸⁷ 

Interviews reveal that all participants had a positive opinion of at least one of the referential sounds, most often referring to nature sounds such as birdsong (N = 13, e.g., “I liked the sound of the birds. It struck me, yeah, it struck me.”), or wind/rain (N = 12, e.g., “the rustling leaves, I enjoyed them”), while nine participants also had a negative opinion of at least one of the referential sounds, most often water (N = 6, e.g., “the sounds from the water, they were a little unpleasant”). Opinion on Abstract sounds was given by only thirteen participants and was positive for four participants (e.g., “there were tones that pleased the ear”), mixed for five participants (e.g., “for the electronic sounds[…], they were aggressive. […] Some other were more balanced, I could consider them.”), and negative for four participants (“the metallic noise would be unpleasant”). Overall, Referential composition strategies were more often described positively than negatively during the interviews, and participants particularly enjoyed natural sounds such as birdsongs and wind or rain. However, the difference in Pleasant ratings between Referential excerpts and the Baseline did not reach statistical significance. Conversely, Abstract strategies were described equally often positively and negatively, but among them, Synthesized excerpts were rated significantly lower than the Baseline and lower than most of the other conditions on the Pleasant component. These results illustrate that interviews can provide more nuanced evaluations than scales alone. They also suggest that people agree more easily on what they do not like than on what they like when describing sound art interventions.

To the question: “Were there any remarkable, out of the ordinary soundscapes during your listening?,” all participants mentioned at least one of the Referential sounds and half of them mentioned at least one of the Abstract sounds. Fourteen participants stated that at least one of the added sounds was unfamiliar or surprising. Among them, five participants reflected on the potential of added sounds to transport them elsewhere (e.g., “there was one that was completely different, that took us completely elsewhere”) or to remind them of nearby parks (e.g., “the fact that we could hear [birds], it reminded me of the park next door”), highlighting the restorative potential of added sounds. Conversely, nine participants were concerned about the added sounds' appropriateness. Among those that explained why, some felt that the purpose of site was incompatible with sound art (e.g., “I had some trouble imagining how you could hear music next to a judicial court. I do not know, maybe it is a misconception because it would be very pleasant.”; “I do not find [the added sounds] appropriate for the place, which is majestic, it is the judicial court. […] It represents Justice.”), while others thought it might be disruptive (e.g., “they may not be adapted for someone who wants to read or sit next to it”).

The data from the interviews was consistent with participants' ratings, where both Referential and Abstract composition strategies had a strong impact on soundscape familiarity when compared to the Baseline. The interviews also confirm the close relationship between familiarity and appropriateness found in the PCA and illustrate the challenge of proposing a sound installation that gives people the impression of being away while still being appropriate for the site. As a participant pointed out: “On the one hand, it is totally inappropriate because unfortunately it is uncommon to hear sound art in public spaces […]. On the other hand, it would be appropriate because it would allow us to disconnect from the soundscapes we are used to hearing”.

The propensity of added sounds to emerge from the soundscape or to bring variety was not directly discussed in the interviews, except for a few rare mentions (e.g., a participant when talking about the added sounds: “it is pretty varied”). However, eight participants stated that the added sounds had a masking effect on the existing sound environment, referring to either Referential sounds (e.g., “there was some kind of a rain sound that soothed everything and was very enveloping”), Abstract sounds (e.g., “sometimes the electroacoustic music masks the traffic noise”), mixed excerpts (e.g., “The water sounds with the birds and the gong-like sound […]; the mix between them covers the annoying background noise, we almost do not hear it anymore”) or as a whole (e.g., “I was hearing the sounds and I thought: well, that's good, it compensates well, it envelops well the sounds from the cars”).

These direct references provide evidence that sound art has the potential to provide a form of non-energetic masking from unwanted noise in the existing environment, as is often discussed by sound artists (e.g., Refs. 48, 49, 51, and 52) and was recently observed in a few studies.^20,29,88 Though non-energetic masking was not directly measured, it is likely related to the rise in soundscape variety and emergence caused by both Referential and Abstract composition strategies during the listening test.

Participants did not refer to the mixes between Abstract and Referential excerpts, except for the one above-quoted mention. Rather, their description of the soundscapes was either general or focused on a specific sound. It suggests that the combination of composition strategies did not raise specific concerns or strong opinions, explaining why the evaluation of hybrid composition strategies fell in between that of Abstract and Referential compositions regarding pleasantness and familiarity.

In answer to the first research question [RQ1], our results suggest that the evaluation of everyday city soundscape modified by the presence of sound art can be described in terms of three components: Pleasant, Familiar, and Varied. The first component, Pleasant, was found to explain 29% of the variance in soundscape measures and is analogous to the pleasantness component used in the model presented in the ISO 12913-2:2018, proposed by Axelsson et al. and validated by Tarlao et al.^36,75,77 This component allowed us to measure the impact of sound art interventions on soundscape appreciation. If appropriateness was loaded on a factor associated with pleasantness in Tarlao and colleagues' model,⁷⁷ we found it associated with a different component, Familiar, that explains 28% of the variance. Similar to the familiarity identified by Axelsson et al.⁷⁵ while explaining a greater amount of variance, it played a significant role in the comparison of composition strategies. Our study suggests that assessing soundscape's familiarity is necessary when evaluating and comparing sound art interventions as it allows one to gauge how their inclusion can evoke novelty in a familiar soundscape, specifically how they can lead to less familiar soundscapes by disrupting the rhythms that dominate the urban experience, without a positive or negative connotation.⁸⁹ This also corroborates Oberman and colleagues' observation that sound art interventions could influence the appropriateness of the sound environment.²⁹ The third component, Varied, was found to explain 16% of the variance, and seems to be inversely related to monotonous observed by Tarlao et al.^90, Varied was related to the variety and number of sources but was believed to be less related to the presence of sources denoting human activities than eventfulness.⁷⁶ This component provided information on the propensity of sound art interventions to rise above the existing soundscape and can be meaningful to position sound art within an integrated/oppositional dichotomy.⁴⁹ Further research is required to assess the comparative performance of scales relative to variety and eventfulness to describe variation in soundscape, and their relationship with sounds from human activity.

Together, these results suggest that a model for soundscape evaluation solely based on pleasantness and eventfulness as proposed in the ISOs 12913-2:2018 and 12913-3:2019 may be incomplete to assess sound art installations and/or unconventional soundscape interventions and situations. Although it provides adequate characterizations of soundscapes,⁷⁷ this two-dimensional model does not allow to evaluate how soundscape interventions can (positively) reshape or reconfigure urban soundscapes, i.e., their impact on familiarity or on variety, even though these aspects are often critical to sound artists (e.g., see Ref. 90)

To answer the second research question [RQ2], we compared the impact of different sound art composition strategies on quantitative and qualitative soundscape measurement. The sound design methodology proposed by the author N.S. allowed us to gather sketches of the sound installation in the form of excerpts into six composition strategies that we positioned within an Abstract (sounds that cannot be ascribed to any real or imagined provenance)/ Referential (recorded sounds that suggest or at least do not hide the source to which they belong) dichotomy.⁵⁸ If all added sounds led to less familiar and more varied soundscapes, results show that Abstract sounds more affected soundscape familiarity and were perceived as more varied than Referential sounds. These effects were similar for Synthesized excerpts (note-to-midi abstractions of the Referential excerpts, see section B) than Resonated effects (filtering of the Referential excerpts), while the evaluation of hybrid composition strategies (mixes between Referential and Synthesized or Resonated excerpts) fell in between that of purely Abstract and purely Referential composition strategies regarding familiarity, but were overall perceived as being more varied. This suggests that the impact of the added sounds on familiarity was more pronounced with deeper sonic abstractions and was somehow proportional to the ratio of Abstract/Referential sounds within the compositions, while hybrid composition strategies were overall perceived as more varied, likely due to a greater diversity of sound sources. Together with the qualitative feedback, this confirms our prior intuition that more processed, artificial sounds can be considered more oppositional as they were perceived as more varied and less familiar than recorded sounds, hinting to a non-energetic masking that is stronger when sounds are least expected. Still, hybrid compositions received the highest scores on the varied components, highlighting the role of soundscapes' diversity on perceived variety. Some of the participants also reflected on the ability of the added sounds to transport them elsewhere, evidencing the potential for sound art interventions to improve the restorativeness of a space by allowing its users to experience a sense of being-away.³³ Otherwise, we could not observe the impact of the composition strategies on soundscape appreciation except for Synthesized excerpts that were rated significantly lower than Referential excerpts on the Pleasant component. Incidentally, we acknowledge that person-related factors—specifically whether participants were local residents or workers, but also variables such as age or sensitivity⁷⁷—may influence people's evaluation of the forecourt's soundscape in the presence of added sounds. Further analysis is required to investigate the relationship between these person-related factors and the impact of sound installations.

Follow-up interviews provided more nuanced feedback and showed that participants' opinion on Referential excerpts was generally positive especially when they referred to natural sounds (in line with recent works; see, for instance, Refs. 26–29) while it was more nuanced regarding Abstract sounds. This suggests that participants more easily reached a consensus when evaluating least liked excerpts than preferred ones. Incidentally, we could not observe significant results in situations where the effect size was small to moderate. This indicates a probable lack of statistical power due to the small sample size, constrained because of the inclusion criterion requiring that participants are familiar with the forecourt of the Judicial Court. This could explain why we did not detect significant differences between conditions on the Pleasant component when qualitative results suggested otherwise. This advocates for further research on the impact of sound art interventions on soundscape appreciation.

The proposed methodology was the result of shared reflections between the scientific team and the sound artist (the second author) to help inform the composition of a public space sound installation prior to its deployment while investigating our research questions [RQ1] and [RQ2]. This research-creation methodology was fruitful both for the sound artist by providing perceptual feedback on sketches of the sound installation and for the scientific team in generating theoretical knowledge on the relationship between sound art composition strategies and their impact on soundscape evaluation.

In this study, we validated the soundscape simulation tool⁵⁶ in a context of comparison between sound art interventions: participants recognized the Baseline sound environment as familiar and realistic and we observed significant differences between composition strategies across all components. However, the sound environment was only representative of the forecourt's average level of activity, and the removal of disruptive sounds and especially of sirens during the edition of HOA recordings tempered its validity, as those sounds are part of the identity of the Judicial Court: for instance, five participants mentioned that sirens were missing from the soundscape during the follow-up interviews (e.g., “I expected to hear a lot more sirens”). Further work is required to evaluate the impact of the Niche Acoustiques sound installation over different scenarios representative of the diversity of sound environments that can be heard within the forecourt. Conversely, the sound installation sketches were made up of elementary building blocks to allow the comparison of well-delineated composition strategies. Another study is required to evaluate compositions that will be closer to the final installation and consider different soundscape scenarios related to different times and days of the week. Otherwise, we could not evaluate the influence of prolonged or repeated exposure of public space users to the sound installation, a key feature as it will be permanently integrated into the public space. This was also a constraint for the second author in their creation process as the final composition is intended to evolve over long periods of time. It should also be noted that although our participants were seated in a fixed location, users will experience the installation on-site while moving through space, which will modulate their exposure to added sounds and create variations across space. Other sensory modalities, including visual cues⁹¹ may also influence the reception of the sound installation.

Regarding the questionnaire itself, the three components not only helped understand how public space users evaluate familiar soundscapes in the presence of sound art [RQ1] but facilitated the comparison of the sonic contents [RQ2] and provided useful, easy to understand feedback to the second author in their creative process: reduced into these three components, the results helped assess how the different composition strategies affected soundscapes in relation to the installation's design goals. In addition, follow-up interviews provided interpretative guidance by revealing the multifaceted nature of the components and evidencing the presence of non-energetic masking, advocating the use of methodological triangulation.^36,92 To the sound artist, these interviews were considered insightful as they provided an in-depth understanding of participants' perceptual and emotional feedback and a contextualization of the quantitative results. Moreover, outputs from the interviews allowed the sound artist to better identify compositional outcomes pointed out with the quantitative results. Open-ended responses from participants in relation to familiarity and appropriateness also highlighted the importance of previous experience of a specific site and of their relationship to the site. It is therefore important to collect participants' experience in their own words and beyond closed-ended questions, which is an oft-ignored recommendation from the ISO 12913 series.

This experiment was part of a greater art-science collaboration to inform the composition of the Niches Acoustiques sound installation with soundscape evaluations. Future work includes the laboratory evaluation of more elaborate composition sketches in different usage scenarios of the forecourt, and a comparison between laboratory results and the in situ evaluation of the sound installation once it is deployed.

Overall, this study showed the potential and feasibility for soundscape simulation to inform the composition of public space sound art prior to its deployment in situ. If it is possible to adjust the content of a sound installation once it is deployed or during its prototyping, the changes must be done under strong constraints (see, for instance, Ref. 51) In contrast, a laboratory setting provides the sound artist the opportunity to freely explore and anticipate the impact of composition strategies relevant to their artistic intention so that they can implement modifications before deployment, although the laboratory situation considerably restrains the compositional aspects to be evaluated (absence of characteristic multimodal and kinesthetic aspects that come with in situ outdoors experience, time constraints, etc.). To find an answer to our research questions, we reported here on relations between broad composition strategies and their impact on soundscape measurement.

In the context of the composition of the sound installation Niches Acoustiques, the experiment yielded valuable perceptual feedback on the effect of elementary compositional materials on soundscape evaluation, a first step towards their combination into finite compositions and their mapping with data collected by the future installation using environmental sensors. On this matter, an excerpt-to-excerpt analysis was also useful as it provided detailed feedback on the sound installation sketches. Ultimately, the experiment showed that the sound installation could provide novelty and variety in the forecourt's soundscape and advocated the use of hybrid compositions to achieve such a goal without reducing soundscape appreciation. Furthermore, participants' feedback implicitly indicated a potential that the present composition samples haven't explored yet with regard to such a hybrid composition strategy; the combination of the two different referential sound types (narrative scenes and natural tones) as a composition/sound production basis, with their different temporal and spectral implications.

In terms of design and planning, this experiment highlights the potential of sound installation to affect soundscape familiarity and variety and the relationship between the Abstract/Referential nature of added sounds and their impact on these components. While we believe that some of these outcomes might transfer to other sound installations, sound art in public spaces is closely related to site-specific characteristics, which should be accounted for. We advocate for the use of similar methodologies for the design and evaluation of sound installations throughout the creation process, to better understand the complex and crucial role of sound in everyday experiences of public spaces.

We would like to thank Coralie Vincent, who performed acoustic measurements in the laboratory; Elise Nicolas, who contributed to the measurement campaign; Antoine Le Dreff, who assisted with some of the listening sessions; Benoit Alary, who made it possible for EVERTims to be integrated in the simulator patch; Cynthia Tarlao, who provided guidance regarding the statistical analysis. We would also like to thank Olivier Warusfel for his technical and scientific guidance. We would finally like to thank all the participants in the study.

The authors have no conflict to disclose.

This research has been approved by McGill Research Ethics Board (REB #22-01-033) and by Sorbonne University Research Ethics Committee (CER-2021-109). Informed consent was obtained from all participants.

Due to the nature of consent obtained from participants, data cannot be made available outside of the research team, as per the ethics approval obtained from the McGill Research Ethics Board and the Sorbonne University Research Ethics Committee. For future data requests, please contact the McGill Ethics Officer at lynda.mcneil@mcgill.ca or the Sorbonne University Research Ethics Committee at CER@sorbonne-universite.fr. However, the analysis code as well as the Python script used to generate the randomized baseline audio are available here: https://github.com/valerianF/analysis_codes/blob/main/JASA_Using_Soundscape_Simulation.