Jo-Anne Bachorowski, Ph.D., j.a.bachorowski@vanderbilt.edu
Michael J. Owren, Ph.D., mjo9@cornell.edu
Department of Psychology
Wilson Hall
Vanderbilt University
Nashville, TN 37240
Bachorowski and Owren will be attending the conference from Wed., Nov 11, through Friday, Nov. 13. Bachorowski can be reached at the Hyatt Regency Columbus, 614-463-1234.
Popular version of paper 4pSC35
Presented Thursday afternoon, November 4, 1999
138th ASA Meeting, Columbus, Ohio
Studies we have conducted indicate that laughter is in fact an acoustically complex and highly variable vocal signal. Individual laughers routinely produce a panoply of laugh sounds, even within a single laugh episode. Given this variety, we reasoned that rather than being a uniform signal, laughter might be comprised of distinct acoustic subtypes with differentiated effects on listeners. If so, one would expect that humans should emphasize different acoustic subtypes when producing laughter in particular social contexts, in accordance with the function that the sound has on that particular occasion. In turn, variation in laugh acoustics is expected to be predictably associated with differences in listeners' emotional response when hearing such sounds. Thus, an important aspect of our thinking is that at least some forms of laughter evolved precisely because they allow individuals to influence the emotional responses of others.
Everyday experience shows that simply hearing laughter can induce positive feelings, and that laughter can readily capture our attention, for instance, even if produced from across a crowded room. The commercial success of laugh-box gadgets and the "Tickle-Me Elmo" doll that was popular a few years ago are further evidence of the power of this vocal signal. Yet despite convergence between anecdotal observations and more scientifically grounded accounts, no one knows for sure why particular laughs seem to be acoustically special to listeners. However, a potentially important distinguishing feature seems to be whether laughter has perceptibly salient vocal pitch, and perhaps significant pitch variability.
As one step towards investigating these hypotheses, we conducted a series of five listening experiments to test whether particular kinds of laugh sounds are acoustically "special" in eliciting positive emotional responses in listeners-whereas others might not be. In each study, we explicitly tested whether laughs that were comparatively high in pitch, with clear tonal and song-like qualities, had an especially strong impact on listeners' emotional responding. One distinguishing feature of these laughs is that they are voiced (i.e., characterized in part by regular vibration of the vocal folds). Laugh samples 1 and 2, produced by a female and male respectively, illustrate these kinds of laughs. In contrast, we expected that laughs without voicing and discernible pitch would be far less likely to elicit positive emotions in listeners-even though they are produced at very high rates as well. These unvoiced laughs are noisier and include sounds that sound more like pants and cackles. A female and male example of these are included as laugh samples 3 and 4, respectively. The laughs we tested with listeners were selected from a very large sample recorded from college students as they watched funny film clips with a same-sex or other-sex friend or stranger in a comfortable laboratory setting. We know from self-report ratings used to measure emotional responding to these clips that all of these laughs were produced when participants were in a positive emotional state.
In each of the five listening experiments, male and female participants were seated at library-style desks in a quiet room and listened to a series of laughs over headphones. A set of 70 laughs were presented twice in random order, and listeners rated each sound by pressing one of four buttons on a response box. Fifty of the laughs (25 produced by males and 25 by females) were voiced and were intentionally selected to vary in duration, with many of the longer laughs showing notable pitch fluctuations. The remaining 20 laughs (10 produced by males and 10 by females) were unvoiced, sounding more like pants or cackles. These unvoiced laughs were selected because while many like them were produced in response to the film clips, they did not contain the pitch characteristics that might be important for eliciting strong positive emotional responses in listeners.
It is inherently difficult to measure emotional responses, and there was no guarantee that we could reliably do so simply by asking listeners to rate their own responses to these sounds. Therefore, the five groups of listeners were tested with the same set of laugh stimuli, but rated the sounds in response to a somewhat different question in each case. The strategy was to select questions where ratings would be higher if listeners were experiencing a positive emotional response to the laugh they had just heard, and lower if they were not. Thus, if ratings were consistent across these five independent experiments, we could conclude that listeners were responding based on common emotional reactions rather than more narrowly to the particular question they were being asked. In one experiment, listeners provided a direct rating of the extent to which they had a positive emotional response to each laugh. In the other four studies, listeners rated each sound based on whether they would include it in a laugh-track to enhance viewer enjoyment of a funny video-tape, if the person laughing was friendly sounding, if the laugh was sexy-sounding, and whether they would actually like to meet the individual who produced the laugh.
Listeners were remarkably consistent in rating the 70 laughs, invariably scoring voiced laughs more positively than unvoiced laughs. Thus, not only were voiced laughs more likely to elicit positive emotions, they were also deemed to be better for use in a laugh track, to be more friendly and sexy-sounding, and to come from someone the listener was more interested in meeting. While the exact ratings made in response to voiced laughs varied somewhat among individual listeners, the lower ratings given to unvoiced laughs were almost identical in every case. In other words, while individuals varied in just how positively they responded to the voiced laughs, the unvoiced laughs evoked more uniformly negative responses. Listeners were virtually unanimous rating unvoiced laughs produced by females as being the most negative. Not only did female unvoiced laughs draw the lowest average scores, these ratings were also the most consistent.
These findings provide strong evidence that voiced laughs are indeed acoustically special. Whether they were produced by males or females, and regardless of whether they were heard by male or female listeners, each of these five experiments showed that voiced laughs were the most likely to elicit positive emotions in listeners. No one knows why these particular kinds of sounds are inherently acoustically special, but one implication of these findings is that laughter does not just function as an expression of positive emotional states. All the laughs tested had been produced when the laugher was in a positive emotional state. Nonetheless, our listeners showed significant differences in their responses to these sounds. Instead, these results suggest that the impact a laugh has on a listener depends on its particular acoustic properties. Individuals that laugh may be producing one kind or another precisely because of the different effects they can have. These ideas are being examined through detailed acoustic analysis of the laugh stimuli used in these listening experiments as well as in ongoing laugh-production studies.