Soundscape experience of public spaces in different world regions: A comparison between the European and Chinese contexts via a large-scale on-site survey^a)

In soundscape studies, the concept of “context” is of utmost importance.^1,2 Although it is generally understood to refer to the space where the soundscape experience is taking place (see the soundscape definition of “acoustic environment as perceived […] by people, in context.”³), the ISO 12913-1:2014 standard and the soundscape research community at large tend to interpret it as a much broader concept that goes beyond the mere physical space to include social and relational constructs. Indeed, in Sec. 3.2 of the ISO document, context is defined as a notion that “includes the interrelationships between person and activity and place, in space and time” and can influence the soundscape interpretation through several factors such as “attitude to the sound source and to the producer of the sound, experience and expectations (including cultural background, intentions or reason for being at a place), as well as other sensory factors, like visual impression and odour.”³ Nevertheless, Tarlao and colleagues note that “the influence of context on soundscape is a broad and diverse question”⁴ and research on contextual factors is scarce and relatively unstructured at the moment with new frameworks and models still emerging.^5,6

The cultural background and geographical context of individuals and communities indeed appear to be key aspects in a soundscape investigation, those that we intuitively understand to have a potentially strong influence on how sound environments and events are interpreted by people. Yet, the ISO standard and subsequent technical specifications relating to data collection and analysis do not explicitly detail how to account for such differences.^7,8 Also, despite being generally acknowledged as an important topic by soundscape scholars around the world, the literature specifically addressing cross-cultural differences in soundscape assessments is, in fact, very limited.^9–11 

Yang and Kang¹² focused on the European context. In their soundscape study, 9200 interviews were conducted in 14 urban open public spaces in different European cities. The authors observed that while the subjective evaluation of the sound level generally related well with the mean, L_eq, such strong associations could not be determined instead between subjective evaluation of the sound level and the acoustic comfort assessment, suggesting that there may be other factors or dimensions that are not well captured by a generic “acoustic comfort” scale. Yu and Kang¹³ looked at the possible effects of cultural factors on the individual assessment of the acoustic quality in residential areas, comparing case studies in Sheffield [United Kingdom (UK)] and Taipei (Taiwan). One clear example of the effect of cultural difference on sound perception from this study concerns sound source preference: insect sounds were more preferred in Taiwan compared to the UK, whereas church bells were more preferred in the UK compared to Taiwan. The semiotics and semantics of sound sources are certainly likely to play an important role in soundscape appreciation.^14,15 This is the point that has found some confirmation in the literature, for instance, Hermida and colleagues¹⁶ stated that for the soundscape meaning formation of a place, the users' previous experiences, their activities, and the cultural and spatial aspects that influence the practices developed in the places should be considered. In a cross-national comparison (i.e., France, Korea, and Sweden) on the soundscape assessment of urban parks, Jeon and colleagues¹⁷ showed that across countries, similarities in the assessment of pleasantness of soundscapes can be observed. However, sociocultural effects are found in the soundscape eventfulness scores as are difference in individual assessments of specific sound sources. Deng et al.¹⁸ performed a cross-national comparison of soundscape assessment of public spaces between China and Croatia; their study found that the Chinese sample tended to give more importance to the eventfulness dimension as opposed to the Croatian sample, which gave more importance to the pleasantness dimension, and in terms of sound sources, natural sounds and children had mainly positive correlations to sonic and environmental satisfaction in China, whereas sounds of human activities had mainly negative correlations to sonic and environmental satisfaction in Croatia.

Overall, there is consensus that much of the soundscape experience outcome results from the semantic implications and cultural meanings that the listeners associate with being exposed to a specific sound source.³ Axelsson and colleagues¹⁹ have previously shown that with their participants' sample in Sweden, soundscape pleasantness was negatively correlated with technological sounds, such as traffic and industrial noise, but it was positively correlated with natural sounds such as birdsong or rustling leaves. Similarly, they observed a positive correlation between soundscape eventfulness and the presence of human sounds (e.g., voices, footsteps, etc.). This led them to theorize that introducing natural and human sounds while reducing technology sounds in any given acoustic environment may potentially enhance the quality of urban soundscapes. Such associations between sound sources and specific soundscape attributes/dimensions/descriptors²⁰ (e.g., calmness, vibrancy, etc.) have been confirmed by subsequent studies as well over the years,^6,21–23 although they have never been actually tested systematically across cultures and countries. Indeed, most studies thus far could not rely on both standardized (i.e., comparable) protocols for soundscape data collection and large datasets gathered on site in different regions of the world, which would allow for a deeper and more meaningful investigation of possible cultural differences across countries. There is an obvious and intrinsic complexity with “defining” let alone “measuring” the concept of “culture.” It is certainly challenging (if not impossible) to capture the entirety of a culture, particularly in studies relying on “quantitative” protocols. Although a country where data are collected may not fully capture the cultural nuances of its respondents, it is often used as a practical proxy for studying cultural differences due to its association with shared language, history, and societal norms. In this study, the focus on global macro-areas (Europe and China) acknowledges that the data collected represent the soundscape perceptions and experiences of individuals within specific geographical regions, which may not perfectly align with any single cultural group.

The International Soundscape Database (ISD),²⁴ generated within the context of the Soundscape Indices (SSID) project,²⁵ represents an opportunity for the soundscape research community to compare soundscape assessments that were effectively collected with the same (translated) survey instrument and general protocol in different countries as it comprises soundscape responses sourced in different European and Chinese cities. Therefore, this study aims to address the following research questions:

• Do soundscape descriptors and perceived dominance of sound sources categories relate similarly across world regions?

• Does “pleasantness” as a single dimension capture the “overall soundscape quality” of a place, and is there any difference in this relationship across world regions?

• What are the relationships between the perceived loudness of a soundscape and its perceived overall quality and appropriateness across world regions?

This study will look at individual responses to acoustic environments (i.e., soundscape assessments) gathered in situ at different locations around the world from general members of the public (adults), i.e., users of the public space. Data collection was performed in the context of the SSID project,²⁵ which generated the ISD.²⁴ Further details of the data collection protocol can be found in Mitchell et al.²⁶ Overall, the soundscape assessment section mainly refers to Method A of the ISO 12913 technical specifications, part 2,⁷ which was, in turn, adapted as a protocol based on the eight soundscape descriptors proposed by Axelsson and colleagues:¹⁹ pleasant, vibrant, eventful, chaotic, annoying, monotonous, uneventful, and calm. Furthermore, in this study, there are further questionnaire sections about perceived loudness of the acoustic environment, overall quality and appropriateness of the soundscape, and perceived dominance of some sound sources categories, namely, traffic noise (e.g., road traffic, railways noise, etc.), other sounds (e.g., sirens, loading of goods, etc.), sounds from human beings (e.g., chat, children playing, etc.), natural sound (e.g., birdsong, water feature, etc.). While no detailed instructions were given to participants about the interpretation of the idea of “dominance” of sound sources, it seems fair to assume that it may not necessarily be related to their “loudness” but rather to the duration of their (temporal) presence in the sound environment under consideration.²⁷ The items of the protocol with the corresponding scales are reported in Table I.

Part 3 of the ISO 12913 series on data analysis⁸ then specifies how the separate eight scores of the perceptual attributes category (i.e., the soundscape descriptors²⁰) should be converted into a set of x-y coordinates to pin any given soundscape assessment onto a pleasant-eventful circumplex space.¹⁹ This is because while the eight perceptual attributes scales are presented separately to the interviewees for them to score, they are not unrelated and together describe a circumplex space with two orthogonal dimensions, pleasant-annoying and eventful-uneventful. Thus, a soundscape that is both pleasant and eventful is vibrant; both annoying and eventful is chaotic; both annoying and uneventful is monotonous; and both pleasant and uneventful is calm.¹⁹ Mitchell and colleagues recently proposed a method²⁸ to next visualize the overall soundscape assessment of a location (rather than individual assessments or central tendencies for those) on the circumplex model, and an example is reported in Fig. 1, where all the assessments of a soundscape data collection session can be plotted and then statistical calculations can be applied. Such an approach is particularly helpful when looking at comparisons between different sets of soundscape assessments.²⁹ 

Specifically, data considered in this study were collected for Europe: in the UK (London, population, 9.0 × 10⁶), in Italy (Venice, population, 261 900), in Spain (Granada, population, 232 200), in the Netherlands (Groningen, population, 200 300), and in China (Shenyang, population, 8.3 × 10⁶; Shenzen, population, 12.6 × 10⁶). From the perspective of the database and the Europe/China stratification, it was considered that in terms of territorial extension, it was reasonable to regard a specific country like China as a macro-area on a global scale, which is similar to how Europe is considered as another global region. When examining and presenting data, it is accepted to use “meaningful regional groupings” as a standard approach, which is also recommended in the United Nations (UN) guidelines.³⁰ These regional groupings should aim to “capture real-world characteristics that are of particular interest for analysis.”³¹ 

Table II reports a brief description of the locations selected for this study from the SSID database. The rationale for site selection was covering a rather broad range of public urban spaces where different sound sources (which would elicit a correspondingly broad range of soundscape assessments on the eight soundscape scales) could reasonably be expected. It is important to reiterate that the focus of the study lies on urban settings: extending the scope to consider comparisons between, for instance, “rural” China and rural Europe would likely pose significant challenges. The objective was not necessarily to find spaces in European and Chinese cities that share similar morphology, layout, or audio-visual features of the built environment. Instead, the emphasis was on identifying spaces that express similar functions and uses by people (e.g., a park that serves as a place of restoration, a bustling street or plaza for social activities, a commuting/transit area, etc.). By focusing on the “function” of the space, a wider range of sound sources profiles and compositions can be explored, resulting in a broader spectrum of scores obtained on the provided semantic scales from participants.^32,33 Yet, other physical elements of an urban public space, such as its morphology and geometrical configuration, also modulate its acoustic environment and, consequently, may potentially lead to different soundscape experiences,^34–36 such that these aspects should also be carefully considered.

While the protocol mentioned in Sec. II A is now relatively well established in the soundscape scientific community, collecting data in different countries and languages poses its own challenges when it comes to accurately translating the soundscape descriptors for use in non-English speaking contexts. This is an issue that deserves careful consideration from scholars and practitioners and is indeed attracting attention in recent research projects.^37–41 For this study, in each country, the local language was used to present the questionnaire to participants, using translations of the protocol as reported in Aletta et al.³⁷ 

Basic demographic data were sought from all participants, although providing the information was not mandatory to take part in the study. In the European sample (i.e., UK, Spain, Italy, and Netherlands combined), there were 2325 respondents, 53.9% female, 46.1% male, M_age = 34.3 years old, SD_age = 15.5 years old; in the Chinese sample, there were 1955 respondents, 48.5% female, 51.5% male, M_age = 40.3 years old, SD_age = 17.1 years old. The breakdown of the total database by language is English (ISO 639:eng), 45.7%; Dutch (ISO 639:nld), 2.1%; Spanish (ISO 639:spa), 5.7%; Italian (ISO 639:ita), 1.6%; and Mandarin Chinese (ISO 639:cmn), 44.9%. Occasionally, the survey may have been offered to a participant in one of the selected languages that did not match the “local” national language of the site (e.g., a native Spanish-speaking resident in the UK preferring to fill the questionnaire in Spanish rather than English), but these deviations were typically below the 2% of the sample across sites.

When comparing the European and Chinese datasets, visual inspection of the distributions of the individual scores assigned by participants to the scales revealed substantially similar trends across all the items of the questionnaire (these are available in the  Appendix), such that it was reasonable to exclude significant effects of sampling bias between the two world regions. Although the technical specifications in part 3 of the ISO 12913 recommend a trigonometric transformation of the median of the eight scores of the soundscape attributes into two coordinates (i.e., ISO pleasant and ISO eventful),⁸ in the context of this study, it was decided to work with untransformed data. The main reason for this is that the local language versions of the protocol are yet to be fully validated, therefore, applying the ISO part 3 transformations may result in essentially different soundscape circumplex spaces and comparing those directly may be misleading. To illustrate this issue, a principal component analysis (PCA) was conducted on the European and Chinese datasets, and the results can be found in the  Appendix. The European PCA yielded two primary components, which closely align with the existing literature, namely, pleasant-annoying and eventful-uneventful, with the secondary axes located approximately as expected for the soundscape circumplex. However, in contrast, the Chinese results do not clearly exhibit eventful-uneventful on the second axis, and the angular relationships between the attributes deviate significantly from the ideal circumplex. These findings suggest that the strict projection from the ISO cannot be currently applied to data from the Chinese translation. Thus, for all subsequent levels of analysis, it was decided to work directly on the five-point Likert scale values.

When looking at associations between items of the protocol, a Bonferroni correction has been applied to account for possible family-wise type I errors which may occur in case of multiple comparisons. The most conservative scenario of m = 8 hypothesis with a target α = 0.05 is taken as a baseline and it requires the Bonferroni correction to test each hypothesis at α = 0.05/8 = 0.00625 significance level.

The first question to address was whether different soundscape dimensions and the perceived dominance of sound sources categories would relate similarly in the European and Chinese samples. For this purpose, a set of Mantel-Haenszel tests of trend was run to determine if linear associations existed between the scores of the single soundscape descriptors (i.e., pleasant, calm, uneventful, monotonous, annoying, chaotic, eventful, and vibrant) and the scores of perceived dominance of the sound sources categories (i.e., traffic sounds, other sounds, human sounds, natural sounds): both sets of variables were scored from one to five with the verbal modifiers of the Likert scales ranging from “totally disagree” to “totally agree” for the former and from “not at all” to “dominates completely” for the latter. These tests were run separately for the European and Chinese datasets.

Table III reports all the abovementioned sound source/soundscape descriptor combinations and shows if the Mantel-Haenszel tests of trend revealed a statistically significant linear association between the items of the questionnaire, comparing China and Europe. A few general patterns can be identified, both within and between the Chinese and European samples, and to support the visualization of such trends, Figs. 2–5 also depict some of the investigated associations as examples.

Soundscape pleasantness was positively associated with the perceived dominance of natural sounds, but the correlation was much stronger for the European sample than the Chinese sample. Traffic sounds and other sounds, on the other hand, are negatively associated with pleasantness, but also in this case, the effect is much stronger for the European rather than the Chinese sample. Interestingly, human sounds did not appear to be correlated with soundscape pleasantness. Indeed, the interpolation lines in Fig. 2 show that for European participants, the soundscape pleasantness quickly decreases with increasing perceived traffic sounds while the trend stays much more uniform for the Chinese participants; likewise for increasing perceived dominance of Natural sounds, there is a quick increase in pleasantness for European participants but a relatively stable trend for Chinese participants. Annoying being the opposite soundscape descriptor to pleasant, as per the soundscape circumplex model,¹⁹ it was reasonable to expect an essentially symmetric pattern for the associations just described, as it can be observed in Table III and Fig. 3; also in this case, it is interesting to notice that human sounds do not correlate with annoyance in either region.

The associations of sound sources dominance with perceived soundscape calmness follow a somewhat similar pattern as per the pleasantness dimension (see Fig. 4); however, for the Chinese sample, no statistically significant association ( p > 0.05) was observed between calmness and perceived dominance of traffic sounds. On a different note, human sounds dominance is negatively correlated with calmness for the European and Chinese samples.

On the soundscape attribute “vibrant,” the European and Chinese samples revealed slightly divergent trends (see Table III and Fig. 5). For European participants, increasingly vibrant soundscapes were positively correlated with perceived dominance of human sounds and no statistically significant correlations with traffic sounds, natural sounds, and other sounds; while for Chinese participants, increasingly vibrant soundscapes where positively correlated with perceived dominance of natural sounds and negatively correlated with perceived dominance of traffic and other sounds (and no correlation with human sounds). Hence, it seems possible for the Chinese sample to experience vibrancy in a more natural sound environment rather than one characterized by human presence as per the European sample.

For the attribute chaotic, both samples show an expected pattern with positive correlations with traffic, human, and other sounds, and negative correlations with natural sounds. Essentially, the same applies to the attribute monotonous but with negative correlations with perceived dominance of human sounds for the European sample and no associations of these two sound source types for the Chinese sample (see Table III). Thus, for the Chinese sample, conventionally “positive” sound sources do not seem to help in “reducing” monotonous scores for a soundscape.

The rest of the associations between sound sources dominance and the attributes eventful and uneventful are mainly non-statistically significant, but an interesting exception is a positive correlation for the European sample of the attribute eventful with the perceived dominance of human sounds (and the “symmetrical” negative correlation with the attribute uneventful), which is not observed for the Chinese sample. Therefore, having in mid the pleasant-eventful space, for the European sample, the perceived dominance of human sounds seems to discriminate soundscapes along the vertical eventfulness dimension but not its orthogonal pleasantness dimension, whereas the Chinese sample is mostly indifferent to human sounds.

To address the second research question, a further set of Mantel-Haenszel tests of trend was run to determine if linear associations existed between scores of the overall soundscape quality item (see Table I) and the soundscape descriptors pleasant and annoying, with the rationale for testing the latter being that pleasant and annoying are opposite attributes on the soundscape circumplex model, thus, the associations may be stronger or weaker depending on the “direction” from which the relationship is viewed. Again, tests were run separately for the European and Chinese datasets. Overall soundscape quality was positively correlated with soundscape pleasantness for the European and Chinese samples with small variations in terms of strength of the association, as shown in Table IV. An essentially symmetrical trend applies for the correlation between overall soundscape quality and soundscape annoyance, as visible also in Fig. 6, with the strengths of the associations for these pairs of items (Europe and China) being slightly weaker than the former ones (see Table IV).

For the third research question, a final set of Mantel-Haenszel tests was performed to explore linear associations between the scores of perceived loudness (see Table I) and the items related to a general assessment/appreciation of a soundscape—i.e., overall soundscape quality and soundscape appropriateness. For the European and Chinese samples, there were statistically significant negative correlations between perceived loudness and overall soundscape quality and soundscape appropriateness (see Table V); however, for soundscape appropriateness, the correlation was much weaker for the European sample than for the Chinese sample, which can also be observed in Fig. 7, where soundscape appropriateness is decreasing with increasing perceived loudness, much more quickly for China than for Europe.

Following the approach adopted for data analysis in this study, it is worth highlighting a few points about its theoretical assumptions. Considering the overall pleasantness-eventfulness soundscape model as proposed by Axelsson and colleagues¹⁹ and adapted from Russell's circumplex model of affect,⁴² it is fair to assume that there is a difference between its “conceptual” space and “mathematical” space. The conceptual space is underpinned by the “circularity” assumption, i.e., a spatial model in which the eight soundscape dimensions are arranged in a circular fashion, in the following order: pleasant, vibrant, eventful, chaotic, annoying, monotonous, uneventful, and calm (see Browne⁴³ for the mathematical distinctions between circumplex, quasi-circumplex, and circular factors). Following from Russell, this theory holds that valence and arousal as the principal components of affective emotion explain the highest degrees of variance in people's perception. It is this general relationship in the theory that defines conceptual space. Although Russell (and subsequently Axelsson) did claim strict correspondences between affective dimensions and angles on the circumplex, such structure has never been properly confirmed in statistical terms for the soundscape circumplex specifically.^43,44 On the other hand, the trigonometric transformations imposed by the ISO/TS 12913-3:2019 formulas to reduce the eight individual scores to a pair of Pleasant-Eventful coordinates define a very “rigid” mathematical space, where it is essential that the soundscape dimensions strictly retain their mutual positions and relative angles for the model to be valid.

The statistical approach adopted in this study does not deny the circumplex theoretical basis, but the use of projection specified in the ISO/TS 12913-3:2019 would require that the attribute translations closely match the relative angles between the (English) soundscape attributes. Nothing suggests that the translations of the individual soundscape attributes in languages other than English are wrong (i.e., pleasant and the Mandarin translation of pleasant are equivalent), and indeed this study builds on the work currently being performed within the Soundscape Attributes Translation Project (SATP) to address this issue by using the “best possible” translations.³⁷ However, there is not enough evidence that different languages express the mathematical relationships among the attributes (i.e., the angles) in the same way, therefore, the mathematical space cannot really be confirmed for now. Hence, the strategy of analysing associations/correlations on an attribute-by-attribute basis is possibly a sensible compromise to work on the available dataset as it does not invalidate the conceptual space and investigates the perceptions on equivalent attributes individually and, to some extent, it avoids “potentially invalidating” the mathematical space.

The results show that there is a difference in the way that European and Chinese participants assess soundscape pleasantness. The correlation between pleasantness (and calmness) and the perceived dominance of natural sounds is stronger for European participants than for Chinese participants. Additionally, the negative association between pleasantness and traffic sounds is also stronger for European participants. European and Chinese cultures may have different values and beliefs about what constitutes a pleasant soundscape, and these cultural differences may influence how people perceive sounds in their environments.^12,13 Being exposed to different types of environmental sounds during daily life possibly affects how different populations experience soundscapes (e.g., Chinese participants may be more accustomed to and, therefore, less sensitive to traffic sounds than European participants). This is also connected to situational factors, such as sampling in populations living in different levels of urbanization and different access and exposure to natural acoustic environments (e.g., in urban parks).⁴⁵ It is worth noting that all such hypotheses are not mutually exclusive and there might be interactions at play and/or other factors that could influence the results.⁴⁶ 

The concept of vibrancy is receiving increasing attention in urban soundscape studies.^47–50 It has been previously reported that vibrant soundscapes tend to be associated with the presence of human sounds (e.g., footsteps, laughter, music, and activity);^6,21 however, results from this study suggest that this may have been merely a European understanding of the concept, whereas for the Chinese context, vibrant soundscapes seem to relate more generally to acoustic environments with an overall perceived richness and diversity of sounds and, as such, potentially applicable to urban and natural spaces. These kinds of discrepancies and ambiguities about the meaning of vibrancy were somehow also reflected in the translation process of the term, where several iterations were needed between the UK-based and China-based working groups originally developing the protocol to find consensus.³⁷ 

A general consideration is that for the Chinese sample, the correlations between perceived dominance of sound sources and soundscape attributes were always less extreme—i.e., the Chinese sample seems to be more indifferent to semantic content of sound sources when it comes to assessing soundscapes dimensions. In more speculative terms, it may also indicate that Chinese participants are less likely to be influenced by the specific sounds that compose a sound environment and more likely to focus on the overall experience of the soundscape—i.e., they listen more “holistically.”^51,52

In this sense, Fig. 8 provides a summary schematic representation of all the associations discussed above and previously presented in Table III, plotted on an adapted version of the soundscape circumplex model, to map the relationships between soundscape descriptors and sound sources for the European and Chinese samples.

Results showed that in the Chinese and European contexts, the valence-related dimension of the soundscape circumplex model (i.e., the annoying-pleasant horizontal axis) was the underlying construct that the participants had been using to form their own assessment of the overall soundscape experience of a given public space. This is somewhat in contrast with earlier findings in the literature,^17,18 where it had been suggested that significant differences existed in soundscapes pleasantness appreciation between Western and Eastern cultures within the soundscape framework. However, previous studies were more limited in terms of samples sizes and distributions, hence, it could be argued that with increased coverage of possible soundscapes in the bidimensional pleasant-eventful space, such differences in soundscape valence tend to decrease and assessments converge. Furthermore, there is still an ongoing debate about what the overall soundscape quality item, as currently phrased in the ISO protocol, is actually capturing when presented to layperson participants (“Overall, how would you describe the present surrounding sound environment?”); it has been previously argued that by focusing on the “overall” element of the question, participants may be more inclined to bring in more non-acoustic aspects (e.g., preconception, expectations, visual context, etc.) into the soundscape assessment task, making the item, again, strongly dependent on the context where the survey is conducted.²⁰ Further disentangling this relationship in future studies would also be relevant from a practical perspective as this poses the question about whether shortened/simplified soundscape data collection protocols may be implemented when there are limited resources and/or other constraints on surveying members of the public using the full list of attributes of the ISO/TS 12913-2:2018. However, evaluating any part of the environment, including its soundscape, in isolation is not feasible. Humans inherently integrate various factors, including preconceptions, expectations, and visual context, into their judgments whether consciously or unconsciously. Previous studies on urban spaces show that isolating a specific sensory aspect, such as visual elements or density/densification, may lead to assessment issues in other domains.^53,54

Generally, the perceived loudness of the acoustic environments was still an important factor of the overall soundscape quality and appropriateness assessment. Interestingly, it was less so for the Chinese sample and its correlations between loudness and appropriateness, which were weaker than those for the European sample. That is, the Chinese sample was less tolerant toward louder acoustic environments and considered them to be less appropriate. Whereas cultural difference may certainly play a role in this context, other person-related factors may be affecting this outcome, such as demographic factors⁵⁵ (as in Erfanian et al.⁵⁵) or noise sensitivity aspects.^56,57

A general limitation is, of course, related to a potential sampling bias. One could argue that the deviations observed in soundscape ratings are simply due to objective physical differences in the acoustic environments at the different locations where soundscape perceptions data were surveyed. A similar potential sampling bias certainly goes beyond the simple site/location scale and applies equally to a higher scale of city/region/country. On this note, it is worth highlighting that this study does not claim to establish causal relationships between specific acoustic parameters and soundscape assessments and test them between countries—it rather suggests that when large enough samples or participants are surveyed and diverse enough urban locations that are representative of a community/society are included, there are still some residual differences that are likely to be ascribable to cultural meanings associated with sound sources and how these form different soundscape dimensions.

In terms of further limitations, the translations aspects (especially for the eight soundscape descriptors of the circumplex model) are still an issue and potential source of uncertainty. Some differences may be accounted for by translation-related issues rather than actual cultural differences. However, based on concurring work going on within the SATP, every effort was made in this study to use the best possible translation of the soundscape descriptors in the local languages (English, Spanish, Italian, Dutch, and Chinese) and contain this potentially confounding factor.³⁷ 

Despite efforts made in translation, there is a possibility that some perceptual aspects that contribute to the formation of one's soundscape experience may have been overlooked or inadequately represented in the questionnaire, as the closed-questions configuration of ISO's Method A itself limits the opportunities for respondents to fully express their thoughts and feelings. Indeed, one of the core principles of soundscape research is the “triangulation” of different methodologies (e.g., structured questionnaires and narrative interviews) to better capture the nuances of people's perception of sound environments.⁵⁸ Yet, in the context of this study, structured questionnaires and protocols such as those proposed in the ISO 12913 series are more likely to offer practical data comparability across different regions, facilitating quantitative analysis, and allow for meaningful cross-cultural comparisons. Furthermore, it may be argued that the evolutionary significance of soundscapes could be more fundamental than culture in addressing this environmental assessment as previously proposed in the literature.⁵⁹ 

At participant's level, it is important to acknowledge that individual differences and person-related factors may influence how individuals experience and evaluate urban soundscapes.⁴⁸ Previous research has indeed highlighted that, among other factors, demographic variables and socioeconomic status can exert a statistically significant impact on soundscape perception; however, it is worth noting that the magnitude of these effects is often relatively small, potentially rendering them negligible in large-scale surveys or broader contexts.⁵⁵ 

Regarding the statistical analysis of the associations, it should be noted that the Mantel-Haenszel test of trend cannot discriminate between independent and dependent variables;⁶⁰ furthermore, it is worth noticing that the Mantel-Haenszel test of trend assesses whether the linear component of the association between two ordinal variables is statistically significant and not that the association is linear per se,⁶⁰ therefore, it cannot be excluded that for some cases, a potentially stronger and statistically significant curvilinear association could exist.⁶¹ However, in the context of this study, determining if soundscape descriptor assessments determine sound sources perceived dominance or vice versa is outside the scope as the main focus was on whether the associations would hold between the European and Chinese datasets. Furthermore, this specific test would require that the variables involved be intervally scaled; this is something that has already been discussed in soundscape literature⁴⁴ and remains an open question. Yet, with the focus on the Europe-China differences, it seems fair to assume that the verbal modifiers of the Likert scales can be straightforwardly translated without introducing further errors as it has been achieved for other socio-acoustic surveys in the past.⁶² 

Cross-cultural differences in soundscape assessment have historically been considered to be an important topic by the scientific community; however, attempts at investigating those have been limited and rely on relatively small samples of participants or range of locations. By looking at data sourced from diverse world regions, the present study aimed to offer some insights into potential culturally driven differences in how people react to individual dimensions of the soundscape circumplex model of affect by using standardized protocols for data collection across different countries and involving relatively large samples of participants in Europe and China. The main conclusions of this study are

• A positive correlation between soundscape pleasantness and perceived dominance of natural sounds was observed: this association was stronger for European participants (r = 0.451) than for Chinese participants (r = 0.081). On the other hand, the negative correlation between pleasantness and traffic sounds was also stronger for European participants (r =−0.450) compared with the Chinese sample (r = −0.143); the calmness construct followed a very similar pattern for both samples;

• soundscape vibrancy was positively correlated with perceived dominance human sounds (r = 0.129) for European participants; however, for Chinese participants, vibrant soundscapes were positively correlated with perceived dominance of natural sounds (r = 0.077) and negatively correlated with perceived dominance of traffic sounds (r =−0.102) and other sounds (r =−0.098);

• soundscape pleasantness is an important dimension in the formation of an overall soundscape experience assessment for the European (r = 0.595) and Chinese (r = 0.472) samples; and

• the perceived loudness of a sound environment does not considerably affect the perceived soundscape appropriateness for the European sample (r = −0.109), but it does for the Chinese sample (r =−0.342; i.e., the louder, the less appropriate in this case).

These findings reinforce the idea that different communities (and potentially cultures) across world regions have indeed different perceptions of what constitutes an acceptable or desirable soundscape and acoustic environment. Understanding these differences is crucial for developing effective soundscape management strategies and design interventions that consider the diverse needs and values of different communities and individuals, including aurally diverse people.⁶³ Additionally, it will also help to deepen our understanding of how different communities interact with and relate to their environment, which can have important implications for environmental conservation and sustainability.⁶⁴ Soundscapes are shaped by cultural norms, beliefs, and values—these, of course, are not “eternal” either and will changeover time and evolve as societies do. Acknowledging these differences is the first step toward creating more inclusive and supporting living environments.

This research was funded through the European Research Council (ERC) Advanced Grant (No. 740696) on SSID. We want to thank all the researchers and international SSID collaborators for their support with data collection, in particular, the research team in Shenyang led by Professor Yuan Zhang, and the research team in Shenzhen led by Professor Lei Yu. The authors also would like to thank Xiang Fang for his support in generating some of the figures in the paper.

The authors declare that they have no conflicts to disclose.

Ethical approval for the study was obtained via the low-risk research departmental route at UCL Bartlett School of Environment, Energy and Resources – Research Ethics Committee (UCL BSEER REC) (approval letter, 14 October 2019), and all participants approached on site at the different public spaces provided informed consent before taking part in the survey.

The data that support the findings of this study are openly available in Ref. 24. The ISD: An integrated multimedia database of urban soundscape surveys—questionnaires with acoustical and contextual information (0.2.2) [dataset], available at Zenodo (Ref. 65).

Distributions for the European and Chinese sample groups are reported in Figs. 9–23 for the scores on the eight soundscape attributes, the overall soundscape quality, the overall soundscape appropriateness, the perceived dominance of sound source types, and the perceived loudness.

A PCA was run separately for the European and Chinese datasets on the soundscape questionnaire items in Table I. The suitability of PCA was assessed prior to analysis. Inspection of the correlation matrix showed that all variables had at least one correlation coefficient greater than 0.3. The overall Kaiser-Meyer-Olkin (KMO) measures were 0.865 for Europe and 0.789 for China. Bartlett's test of sphericity was statistically significant for Europe and China (p < 0.001), indicating that the data were likely factorizable.

Visual inspection of the scree plot and eigenvalues greater than one indicated that two components should be retained for Europe and China.⁶⁶ A varimax orthogonal rotation was employed to aid interpretability. The rotated solution exhibited “simple structure.”⁶⁷ The interpretation of the data was consistent with soundscape literature about the circumplex model for the European dataset with more deviations for the Chinese dataset along the eventful-uneventful dimension.

See Tables VI–XII and Figs. 24 and 25 for information related to the Principal Component Analysis (PCA) performed on the European dataset.

See Tables XIII–XIX and Figs. 26 and 27 for information related to the Principal Component Analysis (PCA) performed on the Chinese dataset.