The effect of the speech-to-song illusion
on the perception of speech rhythm
Master Brain and Cognitive Sciences
Research Project 1
Student name: Kanthida van Welzen Student number: 10797424
Supervisor: Makiko Sadakata
The effect of the speech-to-song illusion
on the perception of speech rhythm
Abstract
Studies comparing music and language processing often do not control for low-level acoustic differences, thereby making it uncertain whether differences in processing between music and language are due to domain-specific mechanisms or acoustic characteristics. A way to control for acoustic characteristics is by the speech-to-song illusion, which is a perceptual transformation from speech to song after a speech fragment is repeated. Music-specific processing arose when perceived as song, making listeners better at detecting pitch changes. Whether this also
accounted for rhythm changes was being examined by a same-different rhythm discrimination task at the first repetition and final repetition. Whenever the illusion occurred, heard from speech to song, participants were worse at detecting conforming rhythm changes compared to no changes. When the illusion was not elicited, heard as either speech or song, detection improved after repetition irrespective of rhythm changes. These results indicate that rhythm processing might not be music-specific but domain-general, thus playing a crucial role in the perception of music and language.
Keywords: speech, song, illusion, rhythm, music perception, speech perception
1. Introduction
Although music and language are both processed by the auditory system, our perceptual experience of music and language differ. Research has focused on what the similarities and differences are between music and language: more specifically, to what extent musical and linguistic stimuli are processed by shared or distinct mechanisms, respectively general or domain-specific processes (Newport, 2011; Peretz & Hyde, 2003). Domain-general mechanisms are thought to include attention and short-and long-term memory to be able to perceive, to understand, and to respond (Jackendoff & Lerdahl, 2006). Besides, language and music both consist of discrete elements in a hierarchical structure, enabling them to create an infinite amount of linguistic or musical phrases (Jackendoff & Lerdahl, 2006; Patel, 2003). Bounded by syntax, linguistic and musical phrases become meaningful. Furthermore, findings from neuroimaging studies indicate that syntax and semantics for language and music are processed in overlapping cerebral structures (Koelsch, 2005; Patel, 2003). Domain-specific mechanisms indicate whether an incoming sound is musical or linguistic as shown by neural and double dissociation studies. For example, whenever a stimulus is processed musically, brain activation patterns tend to be right-lateralized whereas it is more left-lateralized when processed in a linguistic way (e.g. Zatorre, Belin, & Penhune, 2002; Zatorre, Evans, Meyer, & Gjedde, 1992). Additionally, double dissociations studies support the notion of domain-specific mechanisms by showing that adults with congenital amusia have impaired pitch processing in music although the processing of language is unchanged (Peretz & Hyde, 2003). However, the evidence for pure double dissociations is little as adults with congenital amusia often have impaired emotional speech processing (Thompson, Marin, & Stewart, 2012). Furthermore, impairments in some aspects of music processing tends to occur in individuals with Specific Language Impairments (Jentschke, Koelsch, Sallat, & Friederici, n.d.; Mari, Scorpecci, Reali, & D’Alatri, 2016). As support has been found for both domain-general and domain-specific mechanisms, processing of language
and music are dependent on both mechanisms. Therefore, it is interesting to examine to what extent domain-general and domain-specific processes play a role in the perception of music and language.
One of the issues with studies comparing musical and linguistic stimuli is that music and language do not only differ in the way they are perceived, but also differ in their acoustic characteristics. Brain processes that are thought to be music or language-specific could just rather be differences in processing due to low-level acoustic differences such as pitch and rhythm. For pitch, music consists of discrete tones with a relatively stable pitch whereas speech is characterized by continuous pitch glides (Zatorre & Baum, 2012). Moreover, pitch plays a more crucial role in music than in speech. In music the pitch has to be produced and perceived in great accuracy, otherwise listeners will perceive it as an error (Warrier & Zatorre, 2002). In contrast, speech relies roughly on pitch relationships; a deviation will not be heard as a violation. As for rhythm, durational lengthening is a cue to group tones in music and words in language into phrases (Wightman, Shattuck-Hufnagel, Ostendorf, & Price, 1992). In music, this leads to groupings of strong and weak beats (known as meter) and results in highly regular temporal patterns. The rhythm raises periodic expectancies resulting in the ability to tap along to the beat (Dalla Bella, Białuńska, & Sowiński, 2013). In speech, these temporal patterns are not as regular and the perception of meter seems to play less of a role in ordinary speech (Jackendoff & Lerdahl, 2006; Patel, 2006).
A promising way to study domain-specific and domain-general processes while controlling for low-level acoustic differences is the speech-to-song (STS) illusion. When a short, spoken phrase is taken out of context and repeated multiple times, although initially heard as speech, both musicians and non-musicians can perceive it as a song after a few repetitions (Deutsch, Henthorn, & Lapidis, 2011; Falk, Rathcke, & Bella, 2014; Vanden Bosch Der Nederlanden, Hannon, & Snyder, 2015). If the perception of the STS illusion is dependent on domain-general mechanisms, then the neural and behavioural responses will not differ when the stimulus is held constant. However, if the neural and behavioural responses differ although the acoustic characteristics are constant, it indicates that domain-specific mechanisms are involved in eliciting the STS illusion.
Key components to elicit the illusion are repetition and pitch. Besides hearing speech excerpts as sung after repetition, environmental sounds being repeated can be heard as music as well (Rowland, Kasdan, & Poeppel, 2018). The repetition is necessary because it takes time to extract the melodic structure from the stimuli (Tierney, Patel, & Breen, 2018b). As for pitch, the illusion is facilitated when stable pitch portions occur in the speech fragments (Falk et al., 2014; Tierney, Patel, & Breen, 2018a). It facilitates the illusion because music consists of tones with discrete, stable pitches (Zatorre & Baum, 2012). Whether rhythm is a cue underlying the STS illusion remains uncertain. One option to facilitate the STS illusion rhythmically is by creating a steady beat, because high regular temporal patterns are crucial in music (Jackendoff & Lerdahl, 2006). When creating a steady beat by having syllables of equal duration, it does not facilitate the STS illusion (Tierney et al., 2018a), whereas combined with a regular accent distribution it does elicit the STS illusion faster (Falk et al., 2014). However, introducing variability in rhythm in every repetition of the speech excerpt, in other words the speech excerpt had a different rhythmic structure in every repetition, the illusion still occurs (Vanden Bosch der Nederlanden, 2013).
In the abovementioned studies the speech fragments were manipulated in a way that they sound more musical to facilitate the STS illusion (e.g. stable pitch because these occur often in music). This suggests that by making the speech fragments more musical, music-specific processes are recruited sooner. Therefore, this facilitates the STS illusion and this is reflected in the behavioural response as well. For instance, music-specific knowledge concerning pitch is applied more when a
speech excerpt is heard as sung than spoken. When altering the pitch of a syllable in a speech excerpt conforming or violating to the Western musical scale, participants are more sensitive to these pitch deviations when the speech excerpt was heard as sung than as spoken. Furthermore, the pitch deviations violating the Western musical scale were better detected than the conforming pitch deviations (Vanden Bosch der Nederlanden, 2013). Although shown that music-specific mechanisms relevant to pitch can be activated in speech fragments, it is not known whether sensitivity to rhythmic deviations also increases when perceived as song. When listening to musical utterances, listeners are able to detect rhythmic changes and this is reflected in an early negative deflection in event-related potential studies (N150: Geiser, Ziegler, Jancke, & Meyer, 2009; MMN: Näätänen, Paavilainen, Rinne, & Alho, 2007). Whether it is also true for transforming utterances from speech to song is what the current research is about to examine. It is thought that when the utterance is perceived as song, music-specific processes are more active, resulting in better detection of rhythm deviations when it is heard as sung rather than spoken. Furthermore, one’s sensitivity to rhythm deviations violating the musically expected rhythm would be higher than for musically conforming rhythm deviations. It is hypothesized that when the speech excerpts are heard as sung, the rhythm deviations that violate the musically expected rhythm are better detected than the rhythm deviations conforming to the expected rhythm. When the speech excerpts are heard as spoken, the conforming and violating rhythm deviations should be detected equally well. To test these hypotheses, speech excerpts that elicit the STS illusion will be repeated. At the beginning and the end of the trial, participants will rate the speech excerpts on how speech- or song-like it is and they will perform a discrimination task with the original speech excerpt and a conforming or non-conforming speech excerpt.
2. Method
2.1 Participants
Forty-four participants (31 females) were recruited through word-of-mouth communication. Participants’ average age was 25.6 years (±7.9 years). Eight participants were native English speakers (or learned another language simultaneously); others learned the English language between 4-14 years old (Dutch = 21, German = 3, Lithuanian = 2, Spanish = 2, Chinese = 1, Croatian = 1, Georgian = 1, Turkish = 1, Greek = 1, Bulgarian = 1, Polish = 1). On average, the participants had 3.4 years of musical training. None of the participants reported having impaired hearing.
2.2 Stimuli
Twenty-four short speech fragments (mean (sd) 6.4 (1.5) syllables, 1.31 (0.38) seconds) were taken from a corpus developed by Tierney, Dick, Deutsch, & Sereno (2013), consisting of 24 phrases that transform and 24 phrases that do not transform from speech to song.
Three forms of manipulation were performed, namely change in pitch, rhythm, or vowel quality. Six transforming speech fragments were used for the pitch and rhythm manipulation each, and eight transforming and four non-transforming speech fragments were selected for the vowel quality manipulation. Here, I will only focus on the results of the rhythm manipulation.
First, the underlying musical notation of each speech fragment was estimated. This allowed us to manipulate rhythm structure to conform or violate to the underlying musical structure. Three musically trained individuals were asked to annotate the 24 transforming speech segments from Tierney’s corpus (2013) when heard as sung. If the three had similar annotations or a fourth expert
had agreed with one of the annotations, this notation was chosen as the underlying rhythmic structure. In the case all three disagreed, a pilot listening study was performed to select the most appropriate notation. Seven participants heard 10 repetitions of each speech excerpt, followed by the three different notations (piano renditions). They had to state which notation was most similar to the speech excerpt. The notation with the most votes was chosen.
Second, the syllable onsets of each speech segment were annotated. Praat overlap-add method was used to create rhythm changes. For the conforming (rhythmically expected) changes, the rhythm of the segment was manipulated towards the perceived musical rhythm, whereas for the non-conforming (rhythmically unexpected) changes, the rhythm change moved away from the musical rhythm by the same increment (see Figure 1).
2.3 Procedure
The experiment started with instructions followed by 26 trials of which the first two were practice trials to get familiar with the task. The practice trials made use of Deutsch et al. (2011) stimulus “sometimes behave so strangely” and consisted of all three manipulations (pitch, rhythm, and vowel quality). The remaining 24 trials were presented within a single test block in which participants could continue to the next trial on their own pace. A trial consisted of the speech segment being repeated 10 times with pauses of 170 ms as this is suggested to be the optimal pause to elicit the illusion (Falk et al., 2014). After the first presentation of the speech segment, the participants were asked to rate the excerpt from 1 (‘very speech-like’) to 5 (‘very song-like’) via button press (STS rating). Next, participants performed a discrimination task: they had to compare the two speech segments whether they were the same (‘S’) or different (‘L’). Then a passive listening phase followed in which the same speech segment was repeated five times. After this phase, they gave an STS rating again. The last two repetitions made up another discrimination task. Thus, a trial consisted of 10 repetitions of the stimulus with a total of two iterations of the rating task and two iterations of the discrimination task (see Figure 2), in the same manner as Vanden Bosch der Nederlanden, Hannon, & Snyder (2015).
In the discrimination task, participants had to compare the standard to a comparison stimulus. This comparison stimulus could be one of the manipulation types (pitch, rhythm, vowel quality). In each trial, one of the manipulation types would be heard, so in both the initial and final discrimination task. Six of the 24 trials consisted of the rhythm manipulation, resulting in 12
Figure 1. An example of how a speech segment is rhythmically manipulated. The musical notation is agreed upon by three
musicians and verified with a pilot study. This musical notation was used to manipulate the rhythm. The three bars show the duration of the syllables in this speech fragment for the original fragment and for the conforming and non-conforming fragment. The conforming manipulation is identical to the rhythmic structure of the musical notation by shortening or lengthening syllables to match the duration of the musical notation. For the non-conforming manipulation, the syllables were shortened or lengthened by the same increment in the other direction than for the conforming manipulation.
discrimination tasks. The task consisted of ‘same’ trials (N = 4) and ‘different’ trials, in which the standard stimulus is followed by a comparison stimulus containing a conforming (musically expected) change (N = 4) or a non-conforming (musically unexpected) change (N = 4). In each trial, it was randomly selected which discrimination pair was presented and the discrimination pairs at the start and at the end of a trial were independent of each other. Each discrimination pair occurred twice in the initial discrimination task and twice in the final discrimination task allowing each type of discrimination task to be presented equally for each participant.
2.4 Questionnaires
Participants completed questionnaires about their language and musical background. They filled in a self-reported language questionnaire to assess their first language and second language (English) acquisition and competence skills (Roncaglia-Denissen, Schmidt-Kassow, Heine, Vuust, & Kotz, 2013). To determine whether participants had an adequate level of English perception, they had to fill in the lexTALE which stands for Lexical Test for Advanced Learners of English. As a valid and standardized test, it provides a fair indication of general English proficiency. It consists of 60 trials in each of which the participant has to decide whether the word on the screen is an existing English word or not (Lemhöfer & Broersma, 2011). For individual differences in musical sophistication, the Goldsmith’s Musical Sophistication Index (Gold-MSI) was assessed. This is a valid and reliable self-report inventory and concerns different aspects such as active musical engagement, perceptual abilities, musical training, singing abilities, and emotional engagement (Müllensiefen, Gingras, Musil, & Stewart, 2014).
2.5 Statistical Analysis
To examine sensitivity to rhythmic changes when heard as spoken or sung, first the excerpts had to be selected that elicited the STS illusion. A trial that was rated by a participant with 1 or 2 in the initial rating task and a rating of a 3, 4, or 5 in the final rating task was considered as a transforming trial. These trials were included in the analysis, whereas the non-transforming or so-called stable trials were analysed separately.
From the responses of the discrimination tasks, the correct response rates were calculated for each participant, each rhythmic manipulation (none/conforming/non-conforming) and each position (initial/final, referring to the first or last discrimination task in a trial). This resulted in 6 correct response rates per participant.
The correct response rates were entered in a 3 × 2 × covariate (Manipulation [none/conforming/non-conforming] × Position [initial/final] × MSI score) mixed-design ANCOVA
Figure 2. A representation of the order of tasks within a trial. Each trial consisted of 10 repetitions of the
same speech fragment with at the beginning and at the end of a trial a rating and a discrimination task in order to capture the participant’s response when the speech fragment is heard as spoken and heard as sung, respectively.
in R-studio. Interaction effects were added separately. The considered interaction effects were Position × Manipulation and MSI score × Manipulation. The ANCOVA models were compared by their corrected AIC.
3. Results
For the trials with manipulation in rhythm, 73% of the trials were perceived as stable resulting in 76 transforming trials and 178 stable trials for the analysis. Because the transformation of speech to song did not occur for all the participants, 12 participants were excluded from the analysis for the transforming trials. The analysis for stable trials included all 44 participants.
3.1 Questionnaires
The average score of the lexTALE was 86.3% (± 8.9%) calculated over the 44 participants. A t-test was run to examine whether native English speaker scores differed from English as their second language speakers. Native English speakers performed significantly better than other language speakers (t = 4.24, p < 0.01).
For the musical sophistication, the average general MSI score was 75.8 (±23.1). The sample group scores are similar to the scores found by Müllensiefen et al. (2014), so it is a good representation of the general population.
3.2 Transforming: from speech to song
Table 1 shows the different models with their corresponding corrected AIC values for the transforming trials (the STS illusion was elicited). The model including the interaction between position and manipulation was shown to be the best. When the interaction between position and manipulation was included in the ANCOVA model, no main effect of position was observed, F(1, 92) = 0.29, p = 0.59, ηp2 = 0.003, and the interaction was marginally significant, F(2, 92) = 2.39, p = 0.09, ηp2 = 0.05 (see Figure 3). However, the main effect of manipulation was significant (F(2, 92) = 3.49, p = 0.034, ηp2 = 0.07). Surprisingly, Tukey post-hoc analysis revealed that correct response rate for no manipulation stimuli was significantly higher than for conforming manipulation (difference = 30.21, t = 2.62, p = 0.03). There was no significant difference between non-conforming and no manipulation (difference = 12.26, t = 0.98, p = 0.59) and between conforming and non-conforming manipulation (difference = 17.95, t = 1.47, p = 0.31).
Model Explanatory variables AICc
T S
1 Position + Manipulation 74.05 16.74 2 Position + Manipulation + Position × Manipulation 73.67 20.80 3 Position + Manipulation + MSI 75.39 18.84 4 Position + Manipulation + MSI + Position × Manipulation 75.01 22.94 5 Position + Manipulation + MSI + Position × Manipulation + MSI × Manipulation 78.37 25.48
Table 1. Selection of ANCOVA models for transforming and stable trials. Shown are five ANCOVA models that were
considered plausible to explain the correct response rates. The correct AIC (AICc) was used as the measurement for best fitted model. T = transforming trials, S = stable trials.
3.3 Stable: only perceived as speech or song
For the analysis of the stable trials, both the trials that were perceived as only speech or as only song were included. In table 1, the considered models are shown. The best fitted model includes the main effects of position and manipulation on correct response rate. The main effect of position was significant, F(1, 213) = 6.33, p = 0.01, ηp2 = 0.03, showing that the correct response rate was significantly higher in the final than in the initial discrimination task. As expected, no main effect of manipulation was found, F(2, 213) = 1.80, p = 0.17, ηp2 = 0.02 (figure 3). Correct response rates were equal irrespective of the type of manipulation.
4. Discussion
The current study aimed to examine whether rhythmic deviations are better detected in speech excerpts when these are heard as sung. Moreover, once heard as song, one’s sensitivity to detect rhythm deviations that violate the musical rhythm was expected to be higher than conforming musical rhythms. When heard as spoken, no differences in sensitivity were expected for the deviations. However, our results did not confirm these hypotheses fully. It was found indeed that whenever the illusion did not occur, the rhythmic deviations were detected equally well. However, when the transformation was elicited, it did not matter whether the excerpt was perceived as speech or song, participants were better at detecting differences between no manipulation and conforming
Figure 3. The correct response rates for transforming and stable trials. The bar plots show the correct response rates for the main effects of position
and manipulation. For transforming trials, manipulation has a significant effect whereby the correct response rate is higher for no manipulation than conforming trials. For stable trials, position is significant; final correct response rates are higher than initial correct response rates. In the right plot, the interaction effect between position and manipulation is shown. *, p < 0.05. Error bars are within-subject standard error (Cousineau, 2005).
Position Manipulation Transforming Stable * * 40 50 60 70 80 90 100
Initial Final Initial Final
Transforming Stable Correc t re sp o n se ra te (in % )
deviations in both perception modes. From these results, it seems that rhythm processing can be recruited during speech and music.
As rhythm is an important component of speech and music, it might be that domain-general mechanisms rather than music-specific mechanisms were recruited when people perceived the speech excerpts as either speech or song. Actually, song is a combination of music and language. Supported by cognitive neuroscientific research, cerebral networks for language and music are similar and therefore the network involved in song is a combination of these two (Schön et al., 2010). Brain areas such as the superior temporal gyrus and superior temporal sulcus are involved in the perception of rhythm, for either speech rhythm and music rhythm (Alluri et al., 2012; Bengtsson et al., 2009; Zhang, Shu, Zhou, Wang, & Li, 2010). Overlapping brain areas suggest that rhythm is processed by domain-general mechanisms and this is shown in the behavioural responses too. As shown in the current research, whenever heard as only speech or song, sensitivity increased for all manipulation types over time. Perhaps the inconsistent findings when rhythm is manipulated to facilitate the STS illusion are as a result of domain-general mechanisms as well (Falk et al., 2014; Tierney et al., 2018a; Vanden Bosch der Nederlanden, 2013). It could be that some speech utterances are already highly rhythmical, so further manipulation does not facilitate the STS illusion for that utterance.
Domain-general mechanisms also explain why repetition is a key component of this illusion as proposed by the melodic structure hypothesis. It states that repetition is required to extract the melodic structure of the speech fragment (Tierney et al., 2018a). Since it takes time to decide the correct pitch and timing to each syllable, domain-general mechanisms as attention and short-term memory are required in order to compare the speech fragment to previous renditions and implicit knowledge. After hearing one repetition of the speech fragment, a poor representation of melodic structure is constructed. This might be good enough to discriminate the non-conforming changes from no manipulation and therefore lead to high correct response rates from the start, but it is still difficult to hear the conforming changes and therefore there is a low correct response. Although the interaction effect between manipulation and position was marginally significant, participants did improve on detecting the conforming changes after repetition. The repetition results in a greater sensitivity to rhythm deviations over time irrespective of the manipulation type (e.g. conforming or non-conforming). This is supported by the overall increased sensitivity in the final discrimination task compared to the initial one in the stable trials.
Overall, it was surprising that 73% of the trials were perceived as stable (speech-speech or song-song), whereas earlier research found that 44% of the trials was transforming (Vanden Bosch der Nederlanden et al., 2015). It was examined whether trials transform more often when participants have a higher MSI. A correlation analysis was performed, but turned out to be insignificant (rs = 0.13, p = 0.37). This supports the notion that everyday musical experience is sufficient to perceive this illusion (Vanden Bosch Der Nederlanden et al., 2015). Nevertheless, it does not explain why such a large number of trials did not transform. Therefore, a second correlation analysis was performed to examine whether a higher lexTALE score results in more transforming trials, since most participants had English as their second language. This could result in participants listening more attentively to spectral information of language and therefore it takes more repetition to hear the stimulus as a song. However, the correlation was not significant (r = 0.12, p = 0.44). Another reason why a small number of trials transformed could be as a result of individual differences in sensitivity to rhythmic cues. Some participants experienced the illusion often, whereas others never or occasionally perceived the illusion. Participants that are more sensitive to rhythm experienced the illusion more often than rhythmically insensitive participants (Falk et al., 2014).
Lastly, participants were allowed to take as much time as needed to respond on the rating and discrimination tasks. It could be that a long pause in between repetitions diminishes the perception of the illusion, as it is shown in earlier research that pauses longer than 530 ms reduces the amount of times the transformation is elicited (Falk et al., 2014; Castro and Vitevitch, 2018). Furthermore, participants that responded faster perceived the transformation from speech to song more often than slow-responders (Falk et al., 2014).
Usually in discrimination tasks, musicians perform better than non-musicians (Geiser et al., 2009; Näätänen et al., 2007). In our case, participants with a higher MSI score, so more musical experience, did not perform better. This contradicting result is because of the difference in definitions. In most research, musicians are identified by their years of formal training. The MSI score takes more aspects into account than just formal training, such as singing abilities and active engagement. Therefore, it might be that in this research musical experience did not play a role in the illusion. Furthermore, it supports the notion that implicit music knowledge is required to hear the illusion (Vanden Bosch Der Nederlanden et al., 2015).
In conclusion, people are good at detecting rhythm deviations in both speech and music. This might indicate that the processing of rhythm is a shared mechanism. However, a mix of domain-general and domain-specific might be more imaginable as a difference was found between manipulation types. It would be desirable to replicate this research with native English speakers and with more trials containing rhythmic manipulations. Moreover, to disentangle the role of domain-general and domain-specific mechanisms, this research could be performed with ERP or MRI. It could be that the same brain areas are activated in both perceptions (speech or song), but speech and music might elicit distinguishable patterns of activation in these areas. These patterns can be classified by multi-voxel pattern analysis in fMRI. Our findings indicate that more research is required to examine the neural and behavioural responses to language and music.
Acknowledgements
I would like to thank Makiko Sadakata for the supervision, Eylül Turan and Maud Zweers for their support and collaboration, and Adam Tierney for the use of the corpus with auditory illusions.
References
Alluri, V., Toiviainen, P., Jääskeläinen, I. P., Glerean, E., Sams, M., & Brattico, E. (2012). Large-scale brain networks emerge from dynamic processing of musical timbre, key and rhythm. NeuroImage, 59(4), 3677–3689. https://doi.org/10.1016/J.NEUROIMAGE.2011.11.019
Bengtsson, S. L., Ullén, F., Henrik Ehrsson, H., Hashimoto, T., Kito, T., Naito, E., … Sadato, N. (2009). Listening to rhythms activates motor and premotor cortices. Cortex, 45(1), 62–71.
https://doi.org/10.1016/J.CORTEX.2008.07.002
Dalla Bella, S., Białuńska, A., & Sowiński, J. (2013). Why movement is captured by music, but less by speech: Role of temporal regularity. PLoS ONE, 8(8), e71945. https://doi.org/10.1371/journal.pone.0071945
Deutsch, D., Henthorn, T., & Lapidis, R. (2011). Illusory transformation from speech to song. The Journal of the Acoustical Society of America, 129, 2245-2252. https://doi.org/10.1121/1.3562174
Falk, S., Rathcke, T., & Bella, S. D. (2014). When speech sounds like music. Journal of Experimental Psychology: Human Perception and Performance, 40(4), 1491-1506. https://doi.org/10.1037/a0036858
Geiser, E., Ziegler, E., Jancke, L., & Meyer, M. (2009). Early electrophysiological correlates of meter and rhythm processing in music perception. Cortex, 45(1), 93–102. https://doi.org/10.1016/J.CORTEX.2007.09.010 Jackendoff, R., & Lerdahl, F. (2006). The capacity for music: What is it, and what’s special about it? Cognition,
100(1), 33–72. https://doi.org/10.1016/J.COGNITION.2005.11.005
Jentschke, S., Koelsch, S., Sallat, S., & Friederici, A. D. (2008). Children with Specific Language Impairment also show impairment of music-syntactic processing. Journal of Cognitive Neuroscience, 20(11), 1940-1951. Koelsch, S. (2005). Neural substrates of processing syntax and semantics in music. Current Opinion in
Lemhöfer, K., & Broersma, M. (2011). Introducing LexTALE: A quick and valid Lexical Test for advanced learners of English. Behavior research methods, 44(2), 325-343. https://doi.org/10.3758/s13428-011-0146-0
Mari, G., Scorpecci, A., Reali, L., & D’Alatri, L. (2016). Music identification skills of children with specific language impairment. International Journal of Language & Communication Disorders, 51(2), 203–211.
https://doi.org/10.1111/1460-6984.12200
Müllensiefen, D., Gingras, B., Musil, J., & Stewart, L. (2014). The musicality of non-musicians: An index for assessing musical sophistication in the general population. PLoS ONE, 9(2), e89642.
https://doi.org/10.1371/journal.pone.0089642
Näätänen, R., Paavilainen, P., Rinne, T., & Alho, K. (2007). The mismatch negativity (MMN) in basic research of central auditory processing: A review. Clinical Neurophysiology, 118(12), 2544–2590.
https://doi.org/10.1016/j.clinph.2007.04.026
Newport, E. L. (2011). The Modularity Issue in Language Acquisition: A Rapprochement? Comments on Gallistel and Chomsky. Language Learning and Development, 7(4), 279–286.
https://doi.org/10.1080/15475441.2011.605309
Patel, A. D. (2003). Language, music, syntax and the brain. Nature Neuroscience, 6(7), 674–681. https://doi.org/10.1038/nn1082
Patel, A. D. (2006). Musical rhythm, linguistic rhythm, and human evolution. Music Perception, 24(1), 99–104. Peretz, I., & Hyde, K. L. (2003). What is specific to music processing? Insights from congenital amusia. Trends in
Cognitive Sciences, 7(8), 362–367. https://doi.org/10.1016/S1364-6613(03)00150-5
Roncaglia-Denissen, M. P., Schmidt-Kassow, M., Heine, A., Vuust, P., & Kotz, S. A. (2013). Enhanced musical rhythmic perception in Turkish early and late learners of German. Frontiers in Psychology, 4.
https://doi.org/10.3389/fpsyg.2013.00645
Rowland, J., Kasdan, A., & Poeppel, D. (2018). There is music in repetition: Looped segments of speech and nonspeech induce the perception of music in a time-dependent manner. Psychonomic Bulletin and Review. https://doi.org/10.3758/s13423-018-1527-5
Schön, D., Gordon, R., Campagne, A., Magne, C., Astésano, C., Anton, J.-L., & Besson, M. (2010). Similar cerebral networks in language, music and song perception. NeuroImage, 51(1), 450–461.
https://doi.org/10.1016/J.NEUROIMAGE.2010.02.023
Thompson, W. F., Marin, M. M., & Stewart, L. (2012). Reduced sensitivity to emotional prosody in congenital amusia rekindles the musical protolanguage hypothesis. Proceedings of the National Academy of Sciences of the United States of America, 109(46), 19027–19032. https://doi.org/10.1073/pnas.1210344109
Tierney, A., Dick, F., Deutsch, D., & Sereno, M. (2013). Speech versus Song: Multiple Pitch-Sensitive Areas Revealed by a Naturally Occurring Musical Illusion. Cerebral Cortex, 23(2), 249–254.
https://doi.org/10.1093/cercor/bhs003
Tierney, A., Patel, A. D., & Breen, M. (2018a). Acoustic foundations of the speech-to-song illusion. Journal of Experimental Psychology: General, 147(6), 888. https://doi.org/10.1037/xge0000455
Tierney, A., Patel, A. D., & Breen, M. (2018b). Repetition Enhances the Musicality of Speech and Tone Stimuli to Similar Degrees. Music Perception: An Interdisciplinary Journal. https://doi.org/10.1525/mp.2018.35.5.573 Vanden Bosch der Nederlanden, C. M. (2013). The role of music-specific representations when processing speech:
using a musical illusion to elucidate domain-specific and -general processes.
Vanden Bosch der Nederlanden, C. M., Hannon, E. E., & Snyder, J. S. (2015). Finding the music of speech: Musical knowledge influences pitch processing in speech. Cognition, 143, 135–140.
https://doi.org/10.1016/j.cognition.2015.06.015
Vanden Bosch Der Nederlanden, C. M., Hannon, E. E., & Snyder, J. S. (2015). Everyday musical experience is sufficient to perceive the speech-to-song illusion. Journal of Experimental Psychology: General, 144(2), e43-e49. https://doi.org/10.1037/xge0000056
Warrier, C. M., & Zatorre, R. J. (2002). Influence of tonal context and timbral variation on perception of pitch. Perception & Psychophysics, 64(2), 198–207.
Wightman, C. W., Shattuck-Hufnagel, S., Ostendorf, M., & Price, P. J. (1992). Segmental durations in the vicinity of prosodic phrase boundaries. Citation: The Journal of the Acoustical Society of America, 91, 1707.
https://doi.org/10.1121/1.402450
Zatorre, R. J., & Baum, S. R. (2012). Musical melody and speech intonation: singing a different tune. PLoS Biology, 10(7), e1001372. https://doi.org/10.1371/journal.pbio.1001372
Zatorre, R. J., Belin, P., & Penhune, V. B. (2002). Structure and function of auditory cortex: music and speech. Trends in Cognitive Sciences, 6(1), 37–46. https://doi.org/10.1016/S1364-6613(00)01816-7
Zatorre, R. J., Evans, A. C., Meyer, E., & Gjedde, A. (1992). Lateralization of Phonetic and Pitch Discrimination in Speech Processing. In Source: Science, New Series (Vol. 256). Retrieved from
Zhang, L., Shu, H., Zhou, F., Wang, X., & Li, P. (2010). Common and distinct neural substrates for the perception of speech rhythm and intonation. Human Brain Mapping, 31(7), 1106–1116. https://doi.org/10.1002/hbm.20922
