Periodic Reporting for period 3 - SPRINT (Speech Prosody in Interaction: The form and function of intonation in human communication)
Periodo di rendicontazione: 2022-10-01 al 2024-03-31
An important but neglected part of language is intonation, the melody of speech. Intonation is critical for communication: it helps us organize our speech into phrases and is also used to highlight important words and convey nuances about our intended meaning; e.g. one cue that someone is sarcastic is that what they say and how they say it, their intonation, do not match. We also know that intonation transfer from one’s native language to a second language can create miscommunication problems, and that conditions such as being on the autism spectrum can lead to difficulties with the use and interpretation of intonation. Despite intonation’s importance for communication, we still know relatively little about it: we do not know exactly how it is learned by speakers and interpreted by listeners, or how it varies from language to language. Intonation has been challenging to study because it can vary significantly from utterance to utterance, depending on utterance length and related factors, such as the number and location of stressed syllables.
The aim of SPRINT is to document and understand this variability in speech production and examine how it is treated by listeners during comprehension. To tackle production, we use novel mathematical and statistical modelling techniques, which allow us to narrow down variability to a handful of dimensions and understand at which points in time it occurs. Our production results are further tested by means of perception experiments the aim of which is to examine how relevant our findings are for speech comprehension. We finally aim to link our findings to characteristics of the speakers, such as their level of empathy, their musicality, and the hearing acuity with which they can detect frequency changes related to intonation. The role of these characteristics will help us understand why individuals vary in how successfully they produce and interpret intonation. We test the above with speakers from two locations in the UK and Greece, so that we can examine both dialect differences within each language and differences between Greek and English. Our ultimate aim is to fully understand how intonation is structured, used, and interpreted.
The aim of SPRINT is to document and understand this variability in speech production and examine how it is treated by listeners during comprehension. To tackle production, we use novel mathematical and statistical modelling techniques, which allow us to narrow down variability to a handful of dimensions and understand at which points in time it occurs. Our production results are further tested by means of perception experiments the aim of which is to examine how relevant our findings are for speech comprehension. We finally aim to link our findings to characteristics of the speakers, such as their level of empathy, their musicality, and the hearing acuity with which they can detect frequency changes related to intonation. The role of these characteristics will help us understand why individuals vary in how successfully they produce and interpret intonation. We test the above with speakers from two locations in the UK and Greece, so that we can examine both dialect differences within each language and differences between Greek and English. Our ultimate aim is to fully understand how intonation is structured, used, and interpreted.
Since the beginning of the project, spoken data have been collected in the UK’s South East and in Athens. These data have been analysed using dimension-reduction techniques. Doing so allow us to filter out random variability and understand essential differences that reflect the linguistic categories that are the building blocks of intonation. We have done this for both English and Greek data in order to understand how these two languages use intonation to indicate emphasis and highlight particular words in an utterance.
Our findings show that Greek and English differ with respect to highlighting but not emphasis. With respect to emphasis, we found that both English and Greek speakers raise the pitch of their voice when they want to emphasize a particular word, e.g. when they want to make their speech sound livelier. When it comes to highlighting particular words, however, the two languages differ. Greek speakers make a distinction between new and contrastive information: they use just high pitch when they provide new information to a speaker, but use a rising tune when they want to highlight a word that is in contrast with others. English speakers, at least speakers in the UK’s South East, do not make this distinction. For example, in response to a question such as “what colour is your new coat?” a Greek speaker would say “red” with high pitch but if they were to use the same phrase in response to “is your new coat red or orange?” , which contrasts orange and red, they would use a pitch rise instead. English speakers would use high pitch for both. We have also found that this feature of English production matches perception: listeners judge both types of responses described here for "red" to be equally emphatic or prominent, contrary to theoretical expectations.
Overall, our findings indicate that there are both individual differences in how intonation is produced and perceived by speakers of the same variety, and systematic differences between languages. The methodologies we use, which we have adapted from other domains to intonation research, have helped us reach these conclusions, but they have also allowed us to understand systematic variability in our data. By documenting cross-linguistic differences, and separating systematic from random variability, we are making progress in understanding how intonation is structured and used. Finally, our techniques, which we openly share, provide a blueprint for a more insightful study of intonation not only by the SPRINT team but by other scholars as well.
Our findings show that Greek and English differ with respect to highlighting but not emphasis. With respect to emphasis, we found that both English and Greek speakers raise the pitch of their voice when they want to emphasize a particular word, e.g. when they want to make their speech sound livelier. When it comes to highlighting particular words, however, the two languages differ. Greek speakers make a distinction between new and contrastive information: they use just high pitch when they provide new information to a speaker, but use a rising tune when they want to highlight a word that is in contrast with others. English speakers, at least speakers in the UK’s South East, do not make this distinction. For example, in response to a question such as “what colour is your new coat?” a Greek speaker would say “red” with high pitch but if they were to use the same phrase in response to “is your new coat red or orange?” , which contrasts orange and red, they would use a pitch rise instead. English speakers would use high pitch for both. We have also found that this feature of English production matches perception: listeners judge both types of responses described here for "red" to be equally emphatic or prominent, contrary to theoretical expectations.
Overall, our findings indicate that there are both individual differences in how intonation is produced and perceived by speakers of the same variety, and systematic differences between languages. The methodologies we use, which we have adapted from other domains to intonation research, have helped us reach these conclusions, but they have also allowed us to understand systematic variability in our data. By documenting cross-linguistic differences, and separating systematic from random variability, we are making progress in understanding how intonation is structured and used. Finally, our techniques, which we openly share, provide a blueprint for a more insightful study of intonation not only by the SPRINT team but by other scholars as well.
In SPRINT we have used methods for collecting and analysing intonation data that are not new but have rarely been used for the study of intonation. This includes the extensive use of natural speech, which is challenging to analyse, and dimension-reduction methodologies that allow us to make sense of our highly variable data. Further, this is the first time that these methodologies have been used to address the nature of intonation’s abstract structure. The results are promising and by being popularized by SPRINT research they could become part of the standard toolkit for research in intonation. In the rest of the project, we hope to analyse larger amounts of spoken data (something that due to COVID-19 we were not able to do until very recently) and connect our findings to (i) experiments investigating how intonation is perceived, and (ii) speaker characteristics, such as empathy. In addition, we intend to probe further the many technical and methodological challenges we faced while conducting this research, with a view to sharing methodological best practices with the field. In short, our results and our methodological recommendations are likely to shape how we understand and study intonation.