Introduction
Repetition musicalizes speech sounds, a perceptual transformation known as the speech-to-song illusion (Deutsch et al., 2011). Earlier works showed that certain acoustic properties of speech enhance this transformation, while less is known about the effect of different types of repetitions. Previously, exact repetition was thought to be necessary for the speech-to-song transformation, but recent findings challenge this assumption. For example, the songlikeness rating of every repeat on a zig-zag pitch shift stimulus (alternating up and down shifts) follows the same zig-zag pattern – higher ratings for ascending and lower ratings for descending shifts, suggesting that the transformation still occurs with repetition with (higher) pitch shifts (Vanden Bosch Der Nederlanden et al., 2015). This observation is hypothesized to reflect general pitch properties of song and speech: because song tends to include higher-pitched sounds than speech, modulating to a higher pitch in a repetition facilitates the transformation effect, and vice versa (Mean Fundamental Frequency Hypothesis, MFFH). We aim to show that slight pitch shifts over repetitions still induce perceptual transformation and to test the validity of the MFFH.
Methods
Sixteen individuals participated in the study. The stimuli consisted of 24 transforming speech samples from Tierney et al. (2013) and 1 matched sample from Cornelissen et al. (2016). Participants rated the songlikeness of speech samples – first in isolation, then following nine successive repetitions. We employed a within-subject design to test the effect of 5 repetition types: Ascending, Descending, Zig-zag, and Exact repeats with a high or low pitch.
Results
The Ascending condition induced a reliable transformation effect. The Descending condition induced a weaker transformation effect than the Ascending and Exact conditions. Importantly, exact repetitions at both high and low pitch induced comparable significant transformation effects, suggesting that MFFH is not the primary mechanism underlying the observed result pattern. However, we found that the first songlikeness rating of speech samples (upon hearing the initial presentation of stimuli) was higher for samples with higher mean F0 than those with lower mean F0. Thus, the general perception of the musicality of speech fragments may be affected by their mean fundamental frequency.
Discussion & Conclusion
We show that exact repetition is not required for the speech-to-song transformation: A pitch shift over repetitions does not counteract the effect. Also, the transformation is more strongly influenced by the direction of pitch changes than by the mean fundamental frequency of the fragment. Potential causes of the asymmetry of pitch shifts will be discussed.
Repetition musicalizes speech sounds, a perceptual transformation known as the speech-to-song illusion (Deutsch et al., 2011). Earlier works showed that certain acoustic properties of speech enhance this transformation, while less is known about the effect of different types of repetitions. Previously, exact repetition was thought to be necessary for the speech-to-song transformation, but recent findings challenge this assumption. For example, the songlikeness rating of every repeat on a zig-zag pitch shift stimulus (alternating up and down shifts) follows the same zig-zag pattern – higher ratings for ascending and lower ratings for descending shifts, suggesting that the transformation still occurs with repetition with (higher) pitch shifts (Vanden Bosch Der Nederlanden et al., 2015). This observation is hypothesized to reflect general pitch properties of song and speech: because song tends to include higher-pitched sounds than speech, modulating to a higher pitch in a repetition facilitates the transformation effect, and vice versa (Mean Fundamental Frequency Hypothesis, MFFH). We aim to show that slight pitch shifts over repetitions still induce perceptual transformation and to test the validity of the MFFH.
Methods
Sixteen individuals participated in the study. The stimuli consisted of 24 transforming speech samples from Tierney et al. (2013) and 1 matched sample from Cornelissen et al. (2016). Participants rated the songlikeness of speech samples – first in isolation, then following nine successive repetitions. We employed a within-subject design to test the effect of 5 repetition types: Ascending, Descending, Zig-zag, and Exact repeats with a high or low pitch.
Results
The Ascending condition induced a reliable transformation effect. The Descending condition induced a weaker transformation effect than the Ascending and Exact conditions. Importantly, exact repetitions at both high and low pitch induced comparable significant transformation effects, suggesting that MFFH is not the primary mechanism underlying the observed result pattern. However, we found that the first songlikeness rating of speech samples (upon hearing the initial presentation of stimuli) was higher for samples with higher mean F0 than those with lower mean F0. Thus, the general perception of the musicality of speech fragments may be affected by their mean fundamental frequency.
Discussion & Conclusion
We show that exact repetition is not required for the speech-to-song transformation: A pitch shift over repetitions does not counteract the effect. Also, the transformation is more strongly influenced by the direction of pitch changes than by the mean fundamental frequency of the fragment. Potential causes of the asymmetry of pitch shifts will be discussed.