When speech sounds like music.

Abstract

Repetition can boost memory and perception. However, repeating the same stimulus several times in immediate succession also induces intriguing perceptual transformations and illusions. Here, we investigate the Speech to Song Transformation (S2ST), a massed repetition effect in the auditory modality, which crosses the boundaries between language and music. In the S2ST, a phrase repeated several times shifts to being heard as sung. To better understand this unique cross-domain transformation, we examined the perceptual determinants of the S2ST, in particular the role of acoustics. In 2 Experiments, the effects of 2 pitch properties and 3 rhythmic properties on the probability and speed of occurrence of the transformation were examined. Results showed that both pitch and rhythmic properties are key features fostering the transformation. However, some properties proved to be more conducive to the S2ST than others. Stable tonal targets that allowed for the perception of a musical melody led more often and quickly to the S2ST than scalar intervals. Recurring durational contrasts arising from segmental grouping favoring a metrical interpretation of the stimulus also facilitated the S2ST. This was, however, not the case for a regular beat structure within and across repetitions. In addition, individual perceptual abilities allowed to predict the likelihood of the S2ST. Overall, the study demonstrated that repetition enables listeners to reinterpret specific prosodic features of spoken utterances in terms of musical structures. The findings underline a tight link between language and music, but they also reveal important differences in communicative functions of prosodic structure in the 2 domains.

Vocal Features of Song and Speech: Insights from Schoenberg's Pierrot Lunaire

“…By listening to a spoken phrase several times, listeners perceive the phrase as sung. This simple and compelling experiment suggests that the categorization of a performance as speech or song does not rely on the acoustical characteristics (always the same) but rather on the repetition effect leading to the reinterpretation of the material as musical (Falk et al, 2014). However, when listening to a single sound or isolated phrase (i.e., not repeated and without previous exposition to the phrase), listeners are able to detect a clear spoken or sung utterance (Merrill et al, 2016) and therefore rely on acoustical parameters of the signal itself or on perceptual impressions.…”

Acoustic foundations of the speech-to-song illusion.

“…The presence of flat pitches, therefore, may be one of the cues which primes the speech-to-song illusion (henceforth, "song illusion"). Recent work by Falk, Rathcke, & Dalla Bella (2014) has presented evidence that pitch contour variability has a causal effect on the magnitude of the song illusion: flattening the pitch contour between tonal targets--the pitch targets hypothesized by autosegmental approaches to intonation (Ladd, 2008)--led to earlier and more frequent reports of song transformations for two German sentences. However, the resulting artificially flat pitch contours are uncharacteristic of natural speech, which leaves open the question of whether within-syllable pitch variability can explain variance in the song illusion among naturalistic stimuli, as well as whether a more subtle manipulation could modulate the song illusion.…”

“…This finding suggests that the song illusion causes listeners to hear syllable pitches in terms of musical scale tones, a process which may be facilitated if a stimulus' pitches are an easier fit to a musical key template. Yet Falk et al (2014) found that changing two pitch intervals into perfect fifths (a pitch interval which forms a backbone of scale structure in many cultures) had only a trending effect on the frequency of reports of the song illusion, and thus the role of melodic structure in driving the song illusion remains in question. Here we investigate this possibility by comparing the magnitude of the song illusion with the extent to which each phrase's sequence of pitches fits a computational model of melodic structure (Temperley, 2007).…”

“…In particular, the existence of a steady beat may be an important cue supporting the song illusion, as the presence of a steady pulse is one of the most robust features useful for computational discrimination of speech and music (Scheirer & Slaney, 1997). Falk et al (2014) found that sentences with both a regular distribution of accents and isochronous inter-vowel and inter-accent intervals led to earlier and more frequent song transformations. This suggests that pulse regularity may be an important cue driving the song illusion in naturalistic stimuli as well.…”

“…First, it occurs very rapidly: Falk et al (2014), for example, reported that the song transformation commonly takes place by the third stimulus repetition. This is far more rapid than any demonstrated semantic satiation effects; indeed, most reports of semantic satiation have contrasted the effects of three repetitions of a stimulus with thirty repetitions of a stimulus (Smith, 1984;Smith & Klein, 1990;Pilotti et al, 1997;Pilotti & Khurshid, 2004).…”