Indexical and Linguistic Channels in Speech Perception: Some Effects of Voiceovers on Advertising Outcomes

This article examines the effects that voice features have on advertising. Previous research in neurolinguistics and psycholinguistics shows that linguistic and extralinguistic (“indexical”) properties of speech are closely coupled in speech perception and spoken language processing. We review research from the advertising and marketing literature that examines which voices are the most suitable for voiceovers, whether speech rate compression is advisable, and under what circumstances voice selection is most important. We integrate these two bodies of literature and conclude that the voices used in advertising should be familiar and consistent across the campaign and the speaking rate may be increased without deleterious effects. Components of Speech Marshall McLuhan wrote “the medium is the message.” That is, not only is the content of the message itself important in conveying information, but so too is the medium, or the way in which the intended message is conveyed to an audience. When people perceive spoken language, information about the content of the message is transmitted to the listener, along with information about the specific person who produced the message. Because these two sources of information are ineluctably bound together in the speech stream, both channels of information contribute to the final product of perception and both should be considered by advertisers when developing voiceovers. Speech is a complex, multimodal time-varying pattern. Although both auditory and visual cues function in speech perception, we will focus only on the auditory portion. Spoken language encodes two different sources of information. First, it carries linguistic information about the symbolic content of the talker’s intended message. This content contains several levels of linguistic information: phonological (sounds), morphological (units which form words), syntactic (combining words into sentences), and semantic (meaning of an utterance). Taken together, this linguistic information provides the content of an utterance. The second type of information that is carried in the speech stream is often termed paralinguistic, extralinguistic, or indexical. Indexical information can be thought of as the “medium” through which the message is conveyed. Abercrombie (1967) wrote that “[s]uch ‘extra-linguistic’ properties of the medium... may fulfill other functions which may sometimes even be more important than linguistic communication, and which can never be completely ignored” (p. 5). Abercrombie divided the indexical properties of speech into three sets: (1) those properties that indicate group membership (e.g., regional, dialectal, and social aspects of speech), (2) those that characterize the individual (e.g., age, gender, and size and shape of the vocal tract), and (3) those that reveal changing states of the speaker (e.g., affective properties such as fatigue, excitement, amusement, anger, suspicion, health, speaking rate). Indexical and linguistic information in speech correspond to what cognitive psychologists often refer to as source and item information, respectively (see Hilford, Glanzer, Kim, & DeCarlo, 2002). What makes speech a complex signal is that these two properties are carried simultaneously in a single acoustic waveform that is at first produced by an individual speaker and then perceived by a listener who can extract both sources of information. Speech is generated by a speaker’s larynx and supralaryngeal vocal tract. The vocal tract which extends from the larynx through the throat and mouth to VOICEOVERS AND SPEECH PERCEPTION 67 the lips acts as an acoustic filter, enhancing certain resonance frequencies (formants) and attenuating others. When speakers produce different sounds in a language, they constrict their vocal tract at different locations. Which frequencies are enhanced or attenuated in the vocal tract is determined both by its length and by the location of the constriction. In the productions of sounds, the relative frequencies provide the linguistic information about the place of constriction of sounds. In contrast, the absolute frequencies that resonate in a particular person’s vocal tract are dependent on the length of that person’s vocal tract and thus provide talker-specific information. The sound spectrogram in Figure 1 provides a specific example of the integration of linguistic and indexical properties of speech in the production of speech. The formant values produced by the female speaker (first author) are higher than those produced by the male speaker (second author), showing one indexical difference resulting from differences in vocal tract length. The overall movement and relative locations of the formants, on the other hand, provide linguistic information and indicate that the speakers are saying the same utterance. Thus, the same vocal mechanisms produce both linguistic and indexical information simultaneously and both sources of information are encoded and carried in the same signal. Figure 1. Waveform (a) and spectrogram (b) of the word “psychology” produced by the first author (SVL) and the second author (DBP). Dark lines in the spectrogram represent the first formant (lower curve) and second formant