EAT - acoustics and perceptual features of eating-places


Lindborg, PerMagnus (2013). “Acoustic and perceptual features of eating-places in Singapore.” Poster at Joint AESOP/ACSP Congress Dublin, July 2013. [poster] [slides]

Frid, Emma (2013). "Restaurant Soundscapes in Stockholm (A project carried out in collaboration with PerMagnus Lindborg, inspired by the EAT-survey, an investigation of the perceptual quality of servicescapes in Singapore." Project report, TMH, KTH, Stockholm, Sweden. Available upon request (contact PerMagnus).


The relatively few investigations of restaurant soundscapes, a.k.a. 'servicescapes', have focussed on how music style and ambient levels influence repeat patronage behaviour and appraisal (e.g. North & Hargreaves 1996, Novak et al. 2010). Dimensional emotion models (Russel & Mehrabian 1974) underpin both Novak’s work and the outdoors-oriented Swedish Soundscape Quality Protocol (SSQP; Axelsson et al. 2011). There is evidence that acoustic features other than A-weighted SPL are salient for urban soundscape quality perception (e.g. Irwin et al. 2010, de Counsel 2009, Berglund et al. 2006, 2007).

Keywords: soundscape quality, servicescape, perceptual features

The EAT project has sprung out from the ADM course Soundscape Design assignments during the past year. [autumn 2012] [spring 2013]

In study 1, we (the author and his students) investigated 116 ‘eating-places’, from hawker stalls via coffee bars to upmarket restaurants, noting Size, Priciness (estimated from price of typical F&B items), Occupancy, full SSQP, and estimation of on-site A- and C-weighted Leq. A set of computational acoustic features were extracted from audio recordings using MIRtoolbox (Lartillot 2010). Results indicate that eating-places categorised as ‘Chinese’ had significantly higher sound levels than ‘Western’, but lower than ‘OtherAsian’. Priciness was highest at ‘OtherAsian’ places, followed by ‘Western’, then ‘Chinese’ (all Tukey’s HSD). Somewhat surprisingly, there was no overall significant correlation between Priciness and ambient noise levels. SPLs were consistently higher than those indicated by Novak as eliciting the “highest levels of pleasure and approach behaviour”. This would indicate that there is a potential to increase servicescape quality in Singapore, with potential business profit gains as well as general health benefits.

Part of the data (S=15) was used in a perceptual rating study (in studio) with volunteers (N=30). An updated method from the SSER study was used, with some extensions (Weinstein's Noise Sensitivity scale, exlusion of POMS, more developed semantic associations, in-studio 'guesstimation' of price).

In study 2, we are investigating the congruency of quality perception between auditive and visual modalities, using photography gathered from 87 of the locations, in two experimental designs: 1) a forced-choice audiovisual matching task (“Which photo matches the soundscape you hear?”), and 2) a Priciness "guesstimation" task from a) soundscape alone b) photo alone c) soundscape and photo together. We expect to find higher-than-chance success rate in all matching tasks, with c > b > a.

In study 3, we are focussing on the question of 'priciness'. Is it really true that more expensive places have equally unpleasant sonic environment as cheap ones?

The experimental platforms have been programmed in MaxMSP and data collection is ongoing..

"Which image goes with this soundscape?" and "How pricy are the food & drinks at this eating-place?" Try these two experiments yourself! Download the MaxMSP patchers and all stimuli (172MB). Unzip and put the folder 'stimuli' inside the folder "EAT_experiments". Set Max file preferences to include the folder "EAT_experiments".

"Perceptual ratings by labels, colour, space". Download the MaxMSP patcher and soundfiles. Unzip and put the folder "soundfiles" inside the folder "EAT_percept". Set Max file preferences to include the folder "EAT_percept".

Max patchers are tested with Max 6.0.8 on Mac OS 10.5.8.

Response data [csv] Please contact me if you are interested in the Cran.R analysis scripts.

Lindborg, PerMagnus ()