INTER-NOISE 2006
3-6 DECEMBER 2006 HONOLULU, HAWAII, USA
From language and concepts to acoustics: How do people cognitively process soundscapes?
Catherine Guastavinoa
McGill University, GSLIS / CIRMMT 3459 McTavish street, Montreal, QC, H3A 1Y1 Canada
Daniele Duboisb
Centre National de la Recherche Scientifique, LCPE/LAM 11 rue de Lourmel, 75 015 Paris France
ABSTRACT We present converging results from studies using a subject-centred approach to soundscape research. First, we report the linguistic exploration of free-format verbal description of soundscapes, contrasting field surveys and experimental settings. Results indicate that the meanings attributed to sounds act as a determinant for sound quality evaluations. Second, we report the main results of two categorisation tasks conducted to identify categorical structures elicited using soundscape recordings and the underlying principles of organisation derived from the analysis of verbal comments. In agreement with the linguistic exploration, people categorised sound recordings on the basis of semantic features. Results converge to show that acoustic stimulations are cognitively processed as semiotic cues pointing to meaningful events. Therefore, these events need to be identified before measuring physical properties relevant for each category of event. We conclude on the productivity of such methods grounded in human and social sciences to explore cognitive processes in the first place, and later establish relationships between sound quality judgements and physical indicators to best reflect human cognitive processes. 1 INTRODUCTION
Presented in this article is an investigation of cognitive representations of soundscapes as elaborated over the years by people living in large French cities. We report the main findings of a series of studies consisting of : - 4 questionnaires studies, including a mail survey, a field survey (in-situ interviews) and 2 listening tests, - free sorting tasks of urban soundscapes recordings The studies reported here is part of a larger project which focuses on semantic representations of environmental sounds in a attempt to identify perceptually relevant acoustic parameters to improve the sound quality of urban public spaces [2][5][17]. There is converging evidence that sound quality judgments do not rely exclusively on “bottom-up” processing of incoming stimulations but rather integrate previous knowledge and expectations (e.g.[10][19]). Specifically, semantic features play a determinant role in quality assessment through “top-down” knowledge-driven involved in selecting relevant information for the task at hand in a given context [9]. Gaver [3] introduced the distinction between musical listening and everyday listening. Musical listening focuses on perceptual attributes of the sound itself (e.g. pitch, loudness), whereas everyday listening focuses on events to gather relevant information about our environment (e.g. car approaching), that is, not about the sound itself but rather about the sound source and the actions producing sound. Our research question is therefore to identify relevant
a b
Email address: Catherine.Guastavino@mcgill.ca Email address: ddubois@ccr.jussieu.fr
categories of sound sources in urban soundscapes and semantic features associated with each of these categories. Acoustics descriptors can then be addressed within relevant cognitive categories.
2 SOUNDSCAPES VERBAL DESCRIPTIONS
Using a methodology based on the linguistic analysis of spontaneous verbal descriptions of urban soundscapes, four studies were conducted to investigate how people cognitively process familiar urban soundscapes. A mail survey was conducted to identify cognitive representations of urban soundscapes with a mail survey. Eighty subjects were asked to freely describe familiar urban sonic environments, in the absence of direct auditory stimulation. The same open questionnaire was used in the second experiment for a field survey in actual outdoors environments. A second study consisted of 42 interviews carried out in-situ in locations identified as representative of city noises in previous studies. The verbal data collected in reallife situations were then compared with that collected during two listening tests to test the ecological validity of experimental settings in an acoustically damped room. The third study used stereophonic reproduction, with 29 participants listening to six different soundscapes. The fourth study used multichannel Ambisonics reproduction with six channels regularly spaced in the horizontal plane (see [7] or [9] for a full description of the set-up). A new set of 27 participants were questioned on five different soundscapes. The same semi-structured questionnaire was used in all four studies. It was constructed in a progressive way, starting with very general questions about the appraisal of urban soundscapes and ending with more specific ones about transportation noise. Ten open questions were formulated in French with very general terms (’feel, be affected’) in order not to influence respondents’ judgments or confine their answers in predefined categories. Participants in all four studies had been living in a large French city (Paris, Lyon or Nantes) for at least 5 years. In this article, we will not report all the results of each study but rather highlight the findings that were common to all four studies as well as the main difference observed. For a detailed presentation of the mail survey, the reader should refer to [6] and [5] for a detailed description of the field survey and the laboratory conditions, to [9].
2.1 Analysis
Using linguistic discourse analysis methods, the free descriptions were analyzed to identify shared conceptual cognitive representations from individual representations of auditory experiences. A complete linguistic analysis was performed regarding morphology, syntax and semantics. Morphology concerns the internal structure of word constructions. Relying on the linguistic forms available in a given language, the choices that speakers make can be seen as indicating the conceptualization they intend to convey with their utterances. For example, simple adjectival forms (e.g. clear, loud) refer to abstract and consensual (shared) knowledge in contrast to constructed forms (e.g. unbearable, unpleasant) express a subjective (individual) effect that the speakers intend to communicate and are negatively marked. Syntax concerns the construction of sentences and the arrangement of words in utterances. It is used in the present context to evaluate the implication of the participants’s in their discourse : syntactic cues indicates whether they are talking about some state of the world or about their feeling about effect(s) from the world. Semantics concerns the meaning of linguistic expressions. Their interpretations have to be related to cognitive theories of mental representations. The main results that can be drawn from
the psychological exploration of cognitive categories related to urban soundscapes and their representations in language are discussed in the following sections (for a detailed presentation of the linguistic analysis see [1] and [6]).
2.2 How do people talk about urban soundscapes?
From previous studies concerned with everyday sounds representations, the lack of basic simple terms to describe soundscapes as complex sounding environments was confirmed. Complex syntactic constructions were observed, mostly based on nouns referring to sound sources (e.g. cars or people) or verbs referring to actions (e.g. accelerating or talking). Respondents described sound sources, actions producing sound and social activities including notions of time and location (“sounds like a place where I would go out for lunch” or “for a walk”). Urban soundscapes reflect city life as interactions between human activities producing and perceiving sounding effects. The diversity of expressions and the lack of consensus further confirm that soundscapes are conceived as a subjective (individual) evaluation of the sounding environment and not as an objective stimulation.
2.3 How do people sort out soundscapes into cognitive categories?
Two broad categories were derived from the linguistic analysis: SOURCE EVENTS, which can be attributed to an identified source and agent (truck, bus), and BACKGROUND NOISE OF THE CITY, considered as collective noise, where no specific events can be discriminated. As the category name suggests, SOURCE EVENTS tended to be described with reference to specific sources or part of sources (more than 70% of the descriptions in all four studies), such as cars or motors. An important use of constructed adjectives was also observed, namely deverbal adjectives, derived from verbs referring to the psychological effects the stimuli are assumed to cause, such as annoying or jarring. Thus, SOURCE EVENTS appear to be processed as effects that the sound produced by an object has on the subject. In the descriptions of the BACKGROUND NOISE, however, there are fewer references to the source itself (less than 20% of the descriptions in all four studies), and a majority of physical descriptions (around 60% of descriptions) referring to temporal structure and the timbre by means of simple adjectives (e.g. continuous). This result suggests a more abstract conceptualization of the BACKGROUND NOISE as a sound object in itself.
2.4 How are people affected by low frequencies in everyday life situations?
The judgments about SOURCE EVENTS appeared to be closely linked to the appraisal of the source itself, resulting in annoyance judgments beyond the auditory field. For example, the sounds produced by exhaust pipes are associated with bad smells, since they usually co-occur in everyday life [6]. Semantic features played an important role in auditory judgments. In the case of traffic noise, public transportation noise (train, tram) appears to be better accepted than noise produced by private vehicles [5][11]. More generally, traffic noise evaluations are influenced by the value given sound source with generally negative judgements for road traffic noise and less negative ones for railway noise [18]. It can be concluded that the social value (individual vs. collective) of noise acts as a semantic cue and not as an intrinsic property of the acoustic signal (see [2] for further discussion). The BACKGROUND NOISE was generally rather well accepted as it was interpreted as characteristic of city life. Different qualitative evaluations were observed depending on interactions between the perceived temporality, the spatial impression and the integration of human sounds as part of the BACKGROUND NOISE (for more details, see [6]).
2.5
How do people process soundscape in laboratory conditions?
To measure the ecological validity of the reproduction system in experimental settings, the verbal data collected during the two listening tests were compared to those collected in the field survey, which served as the reference context of expression. Similar descriptions of the SOURCE EVENTS were obtained independently of each context. They were primarily described in terms of sources and psychological effects caused by the sources generating the sounding events. Hence, both laboratory conditions can be considered ecologically valid in terms of source identification. However, the descriptions of the BACKGROUND NOISE differed significantly in the first laboratory condition (with stereo reproduction), suggesting that different cognitive processes were involved [9]. In the stereophonic listening test, the BACKGROUND NOISE was attributed to technical considerations of the representation given by the reproduction system rather than to the (represented) ambient urban noise. It was described in terms of physical properties as an acoustic signal rather than as meaningful collective noise characteristic of city life. Furthermore, it has been shown that differences between listening tests and everyday listening situations could not be imputed only to the artificial laboratory conditions and the required processes of abstraction, but also to the “quality” of the recreated listening situation. Indeed, descriptions similar to the field survey were collected in the second listening test, which used an immersive multichannel reproduction method (Ambisonics). Hence, it is necessary to provide a good sense of spatial immersion in the recreated (represented) acoustic environments to ensure ecological validity of an experimental protocol to test psychological effects of urban BACKGROUND NOISE in urban soundscapes as if they where processed in actual environments.
3 SOUNDSCAPE CATEGORIZATION
In this section, we summarize the results of 2 studies investigating soundscape categorization in free sorting tasks using recordings of urban soundscapes. In both studies, participants were asked to listen to 16 recordings of typical urban sounsdcapes and freely organize them into as many groups as they wished according to perceived similarity. They were then asked to freely describe the groupings in their own words. The analysis is two-folded. First a dissimilarity matrix can be derived from the free sorting task and similarity and classes (categories) were inferred from distance evaluations processed by to additive tree algorithms and multidimensional scaling. Second, the verbal descriptions were analyzed as described in section 2.
3.1 How do people categorize soundscapes?
Different sets of recorded sounds of environmental urban noises were processed with the same free sorting task. Maffiolo [12] first studied soundscape categorization using 16 stereophonic recordings from locations identified as representative of Paris. Results indicated that participants organized recordings into two main categories - EVENT SOUNDSCAPE including specific source events (‘starting a car’, ‘bus braking’, ‘people talking’), - BACKGROUND NOISE i.e. amorphous ambient noise in which no specific event could be isolated. Recordings were not clustered according to the sound level with larger differences within certain categories (more than 7 dBA) than across categories [13]. Furthermore, participants spontaneously described characteristics of the object producing sound rather that the sound
produced in agreement with the findings reported in section 2. A finer grain categorisation distinguished categories of sequences within each of these two main categories. EVENT SOUNDSCAPES were subcategorized according to either the type of sound sources (e.g. ‘car’, ‘buses’, ‘people’) or to qualitative judgments (‘pleasant’, ‘aggressive’, ‘unbearable’). Meanwhile, BACKGROUND NOISE recordings were subcategorized on the basis of qualitative judgments as well as acoustic parameters (spectral content and temporal structure ‘continuous’ vs. ‘discontinuous’). In each category, recordings were structured at different distances, suggesting that some may be more representative of the category than others. For example, among the different exemplars of “very unpleasant” traffic soundscapes, a specific sequence seems to possess the greatest number of properties defining this category, and could therefore be considered as the most typical exemplar of this class [1], as a prototype of “unpleasant noise”. The linguistic analysis of verbal descriptions of categories confirm that soundscapes are structured into complex semantic categories integrating not only acoustic parameters but also notions of time, location and activities. These notions are reflected in discourse by complex prepositional phrases with multiple complements such as “riding motorcycles at Bastille on Saturday night”
3.2 How do people categorize urban ambient noise?
These distinctions between categories of soundscapes were further explored with a more precise psycholinguistic analysis of free descriptions of urban environments [5][6] and free sorting tasks of 6.1 Ambisonics recordings [4]. Specifically, a free sorting experiment was carried out on sixteen sequences of ambient BACKGROUND NOISE to further investigate the categorization process when the process of source identification is hindered. Two main categories of soundscapes emerged on the basis of absence or presence of human activity, in relation to perceived pleasantness. A finer grain categorization distinguishes subcategories within each of these two categories. Half of the participants subcategorized traffic sequences on the basis of the presence of human sounds in relation to the judgments of pleasantness: just "unbearable" in the absence of human sounds vs. “only” "unpleasant" when a few human sounds can be heard. Meanwhile, human activity sequences were subcategorized by 75% of the participants on the basis of the type of activities performed ("do the groceries", "have a drink", "take a walk"), ranging on a second dimension related to the degree of tranquility (from "busy" open markets to "quiet" parks). The activity sequences were described mostly by nouns referring to the type of locations ("market", "café" or "restaurant", "park") and identified sound sources ("vendors", "music", "birds"). The human activity category was spontaneously described as pleasant, whereas the traffic noise category was described as unpleasant. At this subordinate level, subcategories were mediated by the subject's interactions with his or her environment through socialized activities. The verbal descriptors were very successful for discriminating categories, but no interaction between acoustic parameters (even loudness) and category membership was observed. These findings indicate that complex environmental sounds are processed as meaningful events providing relevant information about possible interactions with the environment. Furthermore spatial attributes played an important role at a subordinate level, although all recordings were carried out in outdoors environments. One third of the subjects used urban morphology criteria to distinguished reverberating spaces (described as "reverberating", "semi-closed", "hall", "shopping mall") and open spaces ("open", "large squares", "outdoors") within the human activity category.
4
CONCLUSIONS: FROM COGNITIVE CATEGORIES TO ACOUSTIC DESCRIPTORS
Results of linguistic explorations and free sorting tasks converge to show that people process environmental sounds in urban soundscapes as indicating the presence of objects or events affecting people. Since acoustic stimulations are cognitively processed as semiotic cues pointing to meaningful events, one needs to identify these events before measuring physical properties for each category of events. Cognitive processes must be investigated in the first place to identify relevant categories of soundscapes. Then within each category of events, one can establish relationships between sound quality judgments and physical indicators that best reflect human cognitive processes. The productivity of this approach has been demonstrated for indoor environments [14] as well as urban outdoor environments [16][17]. Specifically, Raimbault [17] showed the limitation of matching a unique acoustic descriptor for all urban locations. For example, if the sound level is appropriate to describe main thoroughfares, it is not relevant for describing similar sound level locations including activities, such as squares, markets, or playgrounds. Since soundscape evaluations vary across situations (e.g. a specific activity at a given time in an given environment), acoustic measurements must be processed in relation to the categorical structure of soundscapes identified from a subject-centered analysis. Further research on significant physical descriptors is therefore needed to contribute to the quality of urban environments.
5 ACKNOWLEDGEMENTS
The writing of this article was supported by FQRNT and CFI grants to C. Guastavino.
6 REFERENCES
[1] [2] [3] [4] [5] [6] [7] [8] [9]
Dubois D. “Categories as Acts of Meaning: The Case of Categories in Olfaction and Audition,” Cognitive Science Quarterly, 1, 35-68 (2000). Dubois D., Guastavino C. and Raimbault M. “A cognitive approach to urban soundscapes: Using verbal data to access everyday life auditory categories,” Acta Acustica united with Acustica (in press). Gaver, W. “What in the world do we hear? An ecological approach to auditory event perception,” Ecological Psychology, 5(1), 1-29 (1993). Guastavino, C. “Categorization of environmental sounds,” Canadian Journal of Experimental Psychology / Revue canadienne de Psychologie Experimentale (in press). Guastavino, C. “The ideal urban soundscape: Investigating the sound quality of French cities,” Acta Acustica united with Acustica (in press). Guastavino, C. and Cheminée, P. “Conceptualisations en langue, représentations cognitives et validité écologique,” Psychologie Francaise, 48(4) (2003). Guastavino, C. and Katz, B. “Perceptual evaluation of multi-dimensional spatial audio reproduction”, Journal of the Acoustical Society of America, 116(2), 1105-115 (2004). Guastavino, C. and Cheminée, P. “A psycholinguistic approach to the ecological validity of experimental settings,” Food Quality and Preference, 15, 884-886 (2004). Guastavino, C., Katz, B., Polack, J-D., Levitin, D.J. and Dubois, D. “Ecological validity of soundscape reproduction” Acta Acustica united with Acustica, 91(2), 333-341 (2005).
[10] Jekosch, U. “Basic Concepts and Terms of 'Quality', Reconsidered in the Context of Product-Sound Quality,” Acta Acustica united with Acustica, 90(6), 999-1006 (2004). [11] Lercher, P. and Shulte-Fortkamp, B. “The relevance of soundscape research to the assessment of noise annoyance at the community level,” Proceedings of the 8th International congress on noise as a public health problem, Rotterdam: The Netherlands (2003). [12] Maffiolo V. “De la caractérisation sémantique et acoustique de la qualité sonore de l’environnement sonore urbain.” Unpublished doctoral dissertation, Université du Maine, Le Mans (1999). [13] Maffiolo, V., David, S., Dubois, D., Vogel, C., Castellengo, M., and Polack, J.-D. “Sound characterization of urban environment,” Internoise 97, Budapest (1997). [14] Mzali, M., Dubois, D., Polack, J-D., Letourneaux, F. and Poisson, F. “Mental representation of auditory comfort inside trains: methodological and theoretical issues,” Proceedings of Internoise 2001, The Hague (2001), pp. 1691-1696. [15] Raimbault M. “Simulation des ambiances sonores urbaines: intégration des aspects qualitatifs ” Unpublished Ph.D. dissertation. Univ. Nantes, 268 (2002). [16] Raimbault M., Lavandier C. and Bérengier M. “Ambient sound assessment of urban environments: Field studies in two French cities,” Applied Acoustics, 64, 1241-1256 (2003). [17] Raimbault M. and Dubois D. “Urban soundscapes: Experiences and knowledge”, Cities, 22(5), 339–350 (2005). [18] Schreckenberg, R., Schuemer, B., Griefhan, B., and Moehler, U. “Attitudes toward noise sources as determinants of annoyance,” Proceedings of Euro-noise 1998, München (1998). [19] Västfall, D. “Contextual Influences on Sound Quality Evaluation, Acta Acustica united with Acustica, 90(6), 10297-1037 (2004).