The role of orthography in auditory word learning
Learning to read affects spoken language processing. It has been shown for instance, that the way words are written can facilitate or hinder their auditory perception: words containing rhymes with unique spellings (e.g., /oʊb/ in globe can only be written as <obe> in English) are recognised faster than those containing rhymes with multiple possible orthographic representations (e.g., /eɪm/ in name can be written as either <ame> or <aim>; see Ziegler & Ferrand, 1998). The present thesis expands on the relationship between spoken and written language by investigating whether orthography plays a role in spoken word learning. Recent findings have shown that both children and adults use their knowledge of sound-to-letter correspondences to generate preliminary orthographic representations (hereafter orthographic skeletons) for aurally acquired words (Wegener et al., 2018; Wegener, Wang, Nation, & Castles, 2020). Consequently, we set out to further investigate the mechanism(s) by which sounds are mapped onto letters during spoken word learning.
In Experiment 1, we tested whether orthographic skeletons are generated for all aurally acquired words (regardless of the number of possible spellings) or only for words with unique, and hence completely predictable spellings. Forty-eight Spanish adult speakers first acquired a set of novel Spanish words through aural training, and then saw their spellings in a self-paced reading task. Crucially, words were shown in their unique (consistent words) or one of their two possible spellings (preferred and unpreferred inconsistent words). Participants’ spelling preferences for all novel words were collected two weeks before the experiment took place. The results showed no differences in reading times between preferred and unique spellings. However, words with unpreferred spellings yielded significantly longer reading times as compared to words with unique spellings. This indicates that participants indeed generated orthographic skeletons for novel spoken words even when there was uncertainty regarding the correct spelling. In Experiment 2 we further investigated the leniency of the process underlying the creation of orthographic skeletons by testing speakers of an opaque language. Compared to Spanish, in which most sounds map onto only one grapheme thus making Spanish highly consistent for spelling, French is highly inconsistent for spelling. In French one sound is usually associated with more than one letter (e.g., the sound /o/ has at least three orthographic representations <o>, <au> and <eau>). Due to the overall higher probability of generating incorrect orthographic representations, French speakers may not engage in the process of generating orthographic skeletons when multiple spelling options are possible. Nevertheless, the pattern of results was similar to the one observed with Spanish speakers. This suggests that generating orthographic skeletons during spoken word learning may be beneficial independently of language opaqueness. Finally, to test whether orthographic skeletons are generated every time a new spoken word is acquired (i.e., automatically) or only when generating them is beneficial for the task (i.e., strategically), two groups of Spanish speakers completed the same aural training task, but received different instructions. While active learners knew they were supposed to learn the words, passive learners were naive regarding the aim of the experimental task. As in Experiment 1, we observed longer reading times for previously acquired spoken words with unpreferred spellings, but only in the group of active learners. This suggests that generating orthographic skeletons during spoken word learning may not be automatic, but can actually represent a strategy participants employ with the aim to facilitate the learning process.
Overall, the findings of the present thesis suggest that skilled readers can use their knowledge of sound-to-letter mappings to generate preliminary orthographic representations for aurally acquired words before encountering them in writing. Importantly, we show that orthographic skeletons are generated even when there is uncertainty regarding the correct spelling, both in languages with overall high, as well as those with overall low probability of generating an incorrect representation. These findings raise a possibility that generating orthographic skeletons may be beneficial. Finally, our data suggest that the process by which orthographic skeletons are generated might not be automatic and inherent to the cognitive system, but can actually be driven by strategies participants employ in order to facilitate the word learning process. Taken altogether, the present findings reveal new ways in which orthography affects spoken language processing.
This PhD was supervised by Prof Clara D. Martin and Dr. Efthymia C. Kapnoula at the Basque Center on Cognition, Brain and Language and the University of Basque Country.
Research conducted in the course of the thesis was supported by a Predoctoral fellowship (associated to the project PSI2017 82941-P; grant number PRE-2018-083946) from the Spanish Ministry of Science, Innovation and Universities and the Fondo Social Europeo.
A six month research stay supported by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) – Project-ID 317633480 – SFB 1287 was carried out at the University of Potsdam, Germany