European Commission logo
English English
CORDIS - EU research results

Constructing a theory of phonotactic processing during speaking

Periodic Reporting for period 1 - PhonPred (Constructing a theory of phonotactic processing during speaking)

Reporting period: 2018-10-01 to 2020-09-30

World languages are characterised by various types of phonological regularities. Some of those are encoded in the mappings between spelling, meaning, and sound, others are present within sequences of sounds. Existing theoretical models of reading that are built to mimic readers’ minds are remarkably underspecified in the level of phonological detail. This project was designed to advance our understanding of the nature of phonological representations that readers use when they produce a vocal response.

A. Our work has shed light on aspects of phonological representations used for reading. We have shown that fine-grain phonological information beyond a single sound (for example, that vocal folds vibrate when producing the sound /b/ vs /p/) exerts influence on reading aloud in two languages with very different phonological systems, English and Russian. In a series experiments using different paradigms (masked priming, word and nonword reading aloud, lexical decision), we have shown that this influence is associated with several types of phonological information (e.g. where in the vocal tract the sound is produced, how long the sound lasts), and may depend on the how language’s phonological and orthographic spaces are structured.

B. Our work has also advanced our understanding of how phonological information is exploited. A common assumption in the field is that an individual’s phonological knowledge is stable and can be characterised precisely by administering the nonword reading task. Our findings challenge this assumption: we show that the exploitation of phonological information is dependent on a variety of factors, including characteristics of individuals, structure of nonwords, and testing conditions. Our findings raise important methodological questions about the use and interpretation of the nonword reading task when assessing people’s phonological knowledge. Further, they have implication for situations when the reading aloud task is used to shed light on our linguistic knowledge (in science), to diagnose language impairments (in the clinic), or to assess learning (in schools).

C. Using cutting-edge tools from distributional semantic modelling, we characterised interactions between sound, letter, and meaning regularities, and measured individual differences in people’s sensitivity to these regularities. Our findings suggest that prior experience with language influences the ability to assimilate these from exposure.

Overall, our results challenge and inform existing theories of reading aloud that do not take fine-grained phonological information into account and do not describe how this information interacts with regularities on other levels. Our results also bridge reading aloud and speech production that are traditionally considered separate research domains.
Work Package 1. We conducted an extensive analysis of the phonological structure of English, as well as several additional languages (German, Dutch, and French). This work has laid the foundation for all subsequent packages. These linguistic analyses of English provided new insights relevant to related questions regarding skilled readers’ use of morphological and orthographic information in reading (Ulicheva, Marelli & Rastle, 2020, published in Morphology). We are finalising the full set of cross-linguistic analyses and expect to submit a manuscript in the new year.

Work Package 2. Four masked priming experiments were conducted in England and Russia. A manuscript reporting these experiments is in the second round of review at Journal of Experimental Psychology: Learning, Memory, and Cognition. Initially, we envisioned only 4 experiments in WP2, however, it became clear to us early in the project that our research focus needs to be broadened to include a number of other paradigms for this project to have the impact that we had planned. Therefore, WP2 was expanded to include an additional line of enquiry. In the second set of experiments, we studied how individuals exploit phonological knowledge that they possess when tested repeatedly across a number of occasions. The manuscript reporting those findings was submitted to Psychonomic Bulletin and Review, and an invited revision is in progress.

Work Package 3. In this series of four experiments, we tested whether phonological representations include mechanistic (articulatory) information about word durations in addition to feature-level information. Four experiments using word reading aloud and lexical decision tasks were conducted. The manuscript reporting those findings is in preparation for Journal of Experimental Psychology: Human Perception and Performance.

Work Package 4. This package explored how phonological structure of artificial languages influenced the learning of these languages and their change over time, depending on whether communication between participants was involved or not. This work package is innovative in that it draws on both artificial language learning and language evolution paradigms. Data collection has been completed, and we were invited to present these results the Evolution of Linguistic Complexity workshop, Edinburgh, UK. We expect to submit the corresponding paper to one of the top journals in the field.
The objectives of the actions were fully met. In addition to the planned objectives, we were able to substantially broaden the scope of the proposed work by
1. Studying range of phonological regularities as opposed to a narrower set of regularities, including the interactions between sound, spelling, and meaning regularities; using a wider range of paradigms. This improvement makes our work relevant to a wider range of audiences.
2. Extending the analyses and the experiments to other languages as opposed to an exclusive focus on English. This cross-linguistic aspect of the action makes our findings generalisable to populations that speak languages other than English.
3. Including an extensive training in computational modelling for the ER.
4. Developing a novel proposal that builds on these findings and looks deeper into the interaction between phonological and orthographic structures of languages.

Wider implications
Firstly, our findings have important implications for how individual’s knowledge of spoken language is assessed in clinical settings (e.g. for diagnosing language impairments), in schools (for measuring pupils’ progress), and in science (e.g. when characterising individual knowledge of spelling-sound regularities). We have shown that what individuals know is often different from what individuals choose to use on a specific occasion, which imposes limitations on the use and interpretation of assessment tools that are available at present. Secondly, we have shown that readers differ greatly with regards to what regularities they have assimilated in their lifetime. Further research needs to establish how language structure interacts with language experience in order to develop enhanced tailor-made teaching approaches. Thirdly, our work in Work Package 1 represents the first steps in tying phonological regularities across languages with variations in spelling. Understanding the relationship between phonological structure of a language and its writing system would allow us to inform the development of writing systems for unwritten languages and guiding spelling reforms.