The Lombard Effect as a Communicative Phenomenon

Abstract

UC Berkeley Phonology Lab Annual Report (2008) The Lombard Effect as a Communicative Phenomenon Priscilla Lau UC Berkeley 1. Introduction First described by Etienne Lombard in 1911, the Lombard effect is a phenomenon in which speakers alter their vocal production in noisy environments, such as loud parties or restaurants. Previous research examining the acoustic differences between Lombard speech and normal speech has found measurable differences in vowel duration and intensity (Summers et al. 1988, Junqua 1996). In addition to measuring vowel duration and amplitude, Summers et al. (1988) also measured formant frequencies, finding that F1 and F2 frequencies tend to show an increase in noise, thus causing a shift in the vowel space when the Lombard effect is produced. The results of a study examining the Lombard effect in Spanish (Castellanos et al. 1996) support the results of Summers et al. (1988) regarding the effects of noise on formant frequencies; Castellanos et al. similarly found a small increase in F1 and F2 frequencies in noise. In their paper “The Lombard Sign and the Role of Hearing in Speech,” Lane and Tranel (1971) stress the importance of intelligible communication as a factor to the Lombard effect. Specifically, they argue that “the speaker does not change his voice level to communicate better with himself, but rather with others” (Lane and Tranel 1971, p.692). In stating that the speaker produces the Lombard effect out of a desire to be more intelligible for a listener, there is an implication that the speaker may not be completely unconscious to his or her production of the Lombard effect. Much of the research done on the Lombard effect, however, has neglected this communicative aspect and has instead used experimental designs in which a subject is asked to read a list to no one in particular while hearing noise over headphones. This type of experimental design not only shows that the Lombard effect can be reproduced artificially, but also encourages the assumption that the Lombard effect is a “physiological effect” (Junqua 1996), implicitly suggesting that the speaker may not be aware of his or her production of the Lombard effect. Experiments such as these have also given rise to the interpretation that the Lombard effect is more of a reflex, arising automatically when speaking in noise. Although most research on the Lombard effect has neglected the communicative aspect of the phenomenon that was emphasized by Lane and Tranel (1971), it is important to study the Lombard effect as a communicative phenomenon, as most productions of the Lombard effect in the real world tend to be in communicative environments rather than isolated instances where a person is talking to him or herself. This paper examines the Lombard effect with a focus on its communicative aspect, presenting the results of an experiment designed with the hypothesis of intelligible communication in mind. Measurements of amplitude, duration, and formant frequencies will be examined to see if they correspond with the results of previous research. Additionally, a possible Lombard effect on voice onset time will be tested. The experiment designed for this study involves two subjects at once, with one functioning in the role of a speaker and the other as a listener. The experiment tests whether speakers will exhibit the Lombard effect even when not hearing noise themselves but knowing that the listener is hearing noise. At the beginning of this study, I hypothesized that speakers would display some indication of producing the Lombard effect even when not hearing noise