A comparison of the temporal weighting of annoyance and loudness

Abstract

The influence of single temporal portions of a sound on global annoyance and loudness judgments was measured using perceptual weight analysis. The stimuli were 900-ms noise samples randomly changing in level every 100 ms. For loudness judgments, Pedersen and Ellermeier [J. Acoust. Soc. Am. 123, 963-972 (2008)] found that listeners attach greater weight to the beginning and ending than to the middle of a stimulus. Qualitatively similar weights were expected for annoyance. Annoyance and loudness judgments were obtained from 12 listeners in a two-interval forced-choice task. The results demonstrated a primacy effect for the temporal weighting of both annoyance and loudness. However, a significant recency effect was observed only for annoyance. Potential explanations of these weighting patterns are discussed. Goodness-of-fit analysis showed that the prediction of annoyance and loudness can be improved by allowing a non-uniform weighting of single temporal portions of the signal, rather than assuming a uniform weighting as in measures like the energy-equivalent level (L(eq)). A second experiment confirmed that the listeners were capable of separating annoyance and loudness of the stimuli. Noises with the same L(eq) but different amplitude modulation depths were judged to differ in annoyance but not in loudness.

The temporal weighting of loudness: effects of the level profile

“…Higher weights have been observed for the temporal portion at the beginning of the sound, showing that the first part contributes more strongly to the perceived loudness of the sound than does the middle portion of the sound. To a weaker extent, a recency effect has also been observed; that is, higher perceptual weights are placed on the ending portion of the sound than on the middle portion (Dittrich & Oberfeld, 2009;Pedersen & Ellermeier, 2008). This weighting pattern differs from that of an ideal observer, who would apply identical weights to all temporal portions of a sound (Berg, 1989) if each element provided the same amount of information concerning the correct response, as was the case in the experiments above.…”

“…Recent studies using level-fluctuating noise stimuli that remained constant in spectrum but changed in level every 100 ms or so showed, however, that this conjecture is not correct. Listeners' judgments of the global loudness 1 of a level-fluctuating noise with a duration of 1 s are more strongly influenced by the first 100-300 ms of the sound than by its middle portion (Dittrich & Oberfeld, 2009;Ellermeier & Schrödl, 2000;Pedersen & Ellermeier, 2008;. In other words, the temporal weighting of loudness shows a pattern akin to the primacy effect in short-term memory (e.g., Baddeley, 1966).…”

“…Temporal weights for sounds with a flat level profile (i.e., with no changes in mean level across the stimulus duration) were compared with weights for stimuli containing a gradual increase in level (fade-in) at the beginning. To estimate the temporal weights, perceptual weight analysis was used, just as in other recent studies on the temporal weighting of loudness (e.g., Dittrich & Oberfeld, 2009). The basic principle behind perceptual weight analysis is to present a stimulus consisting of several elements (in the present case, nonoverlapping temporal segments), to introduce trial-by-trial variation in the dimension of interest (here, sound intensity) on each of the elements, and to estimate the impact of the variation of each individual element on a behavioral or neural response.…”

“…However, because the initial spike in the neural response decays within a few milliseconds, such a mechanism should result in almost exclusive weight being assigned to the first temporal segment. Therefore, the observation that the second and third temporal segments (presented 100 and 200 ms, respectively, after stimulus onset) also receive a higher weight than do the following segments (e.g., Dittrich & Oberfeld, 2009) cannot be accounted for by the neural responses in the auditory periphery. Second, Plank (2005) obtained loudness judgments for a sequence of ten 20-ms noise bursts, separated by pauses of 5, 40, or 100 ms. For all three of these conditions, a primacy effect was observed, with higher weights being assigned to the first three segments of the sequence.…”

Annoyance of Time-Varying Road-Traffic Noise

“…How do we know that listeners, when they are asked to judge annoyance, are actually judging loudness? Fortunately, just recently, it has been proven that listeners are capable of separating annoyance and loudness of the same stimuli (Dittrich, Oberfeld, 2009).…”