The "end effect" in retrospective sound quality evaluation

Abstract

Chalmers Room Acoustics Group, Department of Applied Acoustics, Chalmers University of Technology,SE-412 96, Go¨teborg, Sweden(Received 8 August 2003, Accepted for publication 8 September 2003)Keywords: Subjective annoyance, Retrospective evaluation, Time-variation, Overall judgment, Product sound qualityPACS number: 43.50.Ba, 43.66.Lj [DOI: 10.1250/ast.25.170]1. IntroductionTake a moment and think back at the last time youtraveled in an aircraft. Now answer the question: Howannoying was the sound during this flight? In order to answerretrospective questions of this type, a person need to integratethe temporal sound variation causing annoyance into a singleglobal estimate. It seems unlikely that a memory representa-tion of every instance of a several hours long aircraft ride isreadily available to help form a judgment. Nevertheless,people are often able and willing to provide retrospective orglobal judgments of their experiences. In fact, many compa-nies rely on that a memory of a prior positive experience withtheir product will lead to continued use of the company’sservices. But how do people form global judgments ofexperiences that varies in hedonic intensity? This articledescribes some research trying to relate retrospective, globaljudgments of sound experience to instantaneous or ‘‘on-line’’hedonic experiences. It is believed that such research willenable an deeper understanding of human sound perceptionbut also help in optimizing sound experiences varying overtime (for instance products such as cars, aircraft, hand-heldtools etc.).Previous research in the domain of decision-making andemotion psychology has shown that on-line or momentaryexperiences are not reflected in a straightforward way whenglobal judgments are requested. This research typically relateoverall hedonic profiles, obtained by having participantsrating their momentary affective experience by some ‘‘on-line’’ measurement scale, with overall retrospective ratings[1]. A recurrent finding is that when reporting globaljudgments, people heavily weight the peak moment duringan episode, and also strongly focus on how the episode ends[2–4]. Fredrickson and Kahneman [2] found that the un-weighted average of the peak and end could well explainretrospective judgments. The ‘‘peak+end rule’’ therefore alsoimplies that duration of the experience should have arelatively small weight in retrospective judgments. Studieson duration have all been retrospective, meaning thatparticipants evaluate sequences they have been previouslyexposed to. By examining retrospective evaluation of sequen-ces, this research focuses on how people remember andencode past experiences rather than on how they chooseexperiences [4]. Nevertheless, in choosing too repeat anexperience (i.e. flight with a certain aircraft type) or not,people may consult their memories of prior experiences.Figure 1 shows a schematic illustration of the peak-end ruleand its’ application for noise annoyance.2. ExperimentAnother consequence of the peak+end rule is that addingmoderate pleasure to a pleasurable experience could result ina lower retrospective evaluation. In a similar vein, extendingdispleasure by some moderate discomfort could make theexperience less aversive in retrospective evaluation. The latterimplication of the peak+end rule seems to be especiallyinteresting for product sound quality and sound design. Morespecifically, it suggest that a) noise control and sound designshould primarily be concerned with the experienced peak andend of the sound and b) that adding a better end may improvethe overall judgment of sound quality. In partial agreementwith these predictions, Schreiber and Kahneman [5] foundthat adding an short section of mildly aversive sound (66dBsine wave) at the end of a aversive sound (78dB sine wave)resulted in more favorable ratings for the longer, but betterending sound, than for the shorter sound. In the present articlethese findings are extended using complex stimuli of longerduration.More specifically, the following predictions are tested: a)adding a better end will make the sound more favorableretrospectively (compared to the sound without an endingadded), and b) adding a worse end will make the sound lessfavorable retrospectively (compared to the sound without anending added). Note that a preference of a longer exposure toan aversive experience over a shorter exposure runs counter tohuman judgment theories and theories on noise perception.In addition, prediction that both the ‘‘better-ending’’ and‘‘worse-ending’’ sounds will retrospectively be judged aslouder than the ‘‘no-ending added’’ sound was tested. Thisprediction is derived from the idea that the better- and worse-ending sounds include a comparison or reference processwhere participants can compare the first part (i.e. louder) withthe latter part (i.e. less loud). For the no-ending added soundsno such comparisons will arise since the sound is constant inintensity.3. Method3.1. Participants and designTwo experiments were performed. In total 142 partic-ipants (88 females) from Chalmers University took part (70 inexperiment 1 and 72 in experiment 2). Their mean age was