Reading | rastlelab

Learning to Read & Skilled Reading

Reading is one of the most remarkable human achievements. In a relatively short space of time, most children go from painstakingly sounding out individual symbols, to the rapid and seemingly automatic access to meaning from these symbols that skilled readers experience. Understanding the mechanisms that underpin this learning process is critical.

Current Work.

Morphological Decomposition in the Ventral Stream.

Clare Lally, an ESRC-funded PhD student in the LLC Lab, is using searchlight representational similarity analysis (RSA) to localise and characterise neural representations of words within the ventral stream, which has been identified as a hierarchically-organised pathway during visual word recognition. Currently, Clare is investigating morphologically complex words (e.g. unlock, lockable, locker), which display one of the few systematic relationships between orthography (letters) and semantics (meaning). For example, morphologically complex words that share the same stem (e.g. lock, unlock, lockable) are related to the same meaning. Additionally, affixes appended to the stem modify the meaning of the word in a highly predictable manner (e.g. unlock, unpack, unplug). Thus, the ability to add orthographic units to modify the meaning of a word allows morphemes to act as ‘islands of regularity’ (Rastle, Davis, Marslen-Wilson & Tyler, 2000). An understanding of these morphological relationships could allow the reader to facilitate direct print-to-meaning mapping, and operate as a functional organisation principle of language.

Behavioural and neuro-anatomic evidence suggest that morphological decomposition is a sequential process featuring various orthographic and semantic constraints at different stages in word recognition, which appear to be hierarchical in nature (Gold & Rastle, 2007). In this way, morphological decomposition seems to align with the ventral stream proposal of word reading, in which the reader begins by parsing the visual input and then gradually maps larger orthographic units on to existing abstract linguistic representations in the left occipital temporal cortex (Dehaene et al., 2005). Therefore, Clare's research aimed to assess how the nature of representations for words are transformed at different levels in the processing hierarchy within the ventral stream.

A set of a priori prediction matrices were constructed based on the morphological properties of the words, which were expected to elicit different patterns of activation at different levels of the hierarchically-organised pathway. Words with similar visual features, such as shared letters, were expected to show similar patterns of activation in posterior regions, whereas words that were close in meaning were expected to show similar patterns of activation in anterior regions as representations became more abstract. Critically, there was expected to be an intervening shift in the representations of words based on their morphological properties, for example whether a word contained not only a plausible stem but also a viable affix, and further, whether this affix provided a semantic connection to the stem. Univariate analyses will be conducted to observe whether word reading elicits predicted activation within the ventral stream, and RSA will be used to detect differences in neural patterns of activation, and observe whether morphological representations follow a graded hierarchy of abstraction.

Related Reference:

Yablonski, M., Rastle, K., Taylor, J.S.H. & Ben Shachar, M (2018, in press). Structural properties of the ventral reading pathways are associated with morphological processing in adult English readers. Cortex. Epub ahead of print https://www.sciencedirect.com/science/article/pii/S0010945218301977

Rule-Learning and Generalisation.

As part of her PhD, Rebecca Lawrence is investigating rule-learning and generalisation in the context of reading acquisition. Particularly, Rebecca is looking at orthography-phonology correspondences, and exploring whether the Tolerance Principle (Yang, 2016) underlies the acquisition and representation of these correspondences in reading. Rebecca's work is exciting because the Tolerance Principle is a recently-proposed learning mechanism, which has previously only been considered in relation to grammar.

Meaningful Regularities in English Writing.

The English language is difficult to read and to spell, because words that sound similar are often represented differently in writing. While irregularities between sound and spelling have been studied extensively, an open question is why English spelling is so inconsistent, or has evolved to be so inconsistent. Recent linguistic work indicates that spelling-to-sound irregularity may support the formation of higher-level regularities between spelling and meaning (Berg & Aronoff, 2017).

Funded by the ESRC Future Research Leaders Fellowship, Dr Ana Ulicheva built on this idea under the supervision of Prof Kathy Rastle and Prof Mark Aronoff (Stony Brook University, US). Ana's project involved a comprehensive study of spelling-to-meaning regularities in English suffixation using a large-scale computational analysis of corpus data. In a series of behavioural experiments (nonword classification, sentence reading, and spelling), Ana and colleagues further showed that humans unconsciously make use of meaningful information that is encrypted in English spelling.

Key Reference:

Ulicheva, A., Harvey, H., Aronoff, M. & Rastle, K. (2018). Skilled readers’ sensitivity to meaningful regularities in English writing. Cognition. Epub ahead of print https://www.sciencedirect.com/science/article/pii/S001002771830249X; data available on https://osf.io/hac5j/

Individual Variation in Non-word Reading.

What explains large individual differences in people's performance on language tasks? One task that reveals striking differences across individuals is reading aloud English non-words such as BETHOVE. Mousikou et al. (2017) found that disyllabic non-words elicited many different pronunciations across participants and identified item-level variables that contribute to this variation. Therefore, Dr Ana Ulicheva and Prof Kathy Rastle are conducting a project which explores variability across participants - which factors influence variability and via which mechanisms? It is suggested that the driving force behind subject-based variability is related to differences in individuals' prior reading experience.

Key References:

Coltheart, M. & Ulicheva, A. (2018). Why is nonword reading so variable in adult skilled readers? PeerJ. https://doi.org/10.7717/peerj.4879

Mousikou P., Sadat, J., Lucas, R., & Rastle K. (2017). Moving beyond the monosyllable in models of skilled reading: mega-study of disyllabic nonword reading. Journal of Memory & Language, 93, 169-192. [PDF]

Previous Work.

Whole-Word Pathway.

Mapping the printed word onto meaning can be achieved in two ways: (a) mapping from print to sound and then capitalising on the mapping between spoken forms and their meanings acquired in infancy, or (b) mapping directly from print to meaning. These two routes to reading are often referred to as the sub-word and whole word pathways. The acquisition of the sub-word pathway is critical for reading success in early years education (hence the focus on letter-to-sound phonic knowledge). However, less is known about what drives the development of the whole-word pathway crucial for efficient skilled reading.

In an ERSC funded project, Jo Taylor, Kathy Rastle, and Matt Davis (MRC CBU) looked at the development of this pathway in adult learners using an artificial orthography learning paradigm, taking advantage of both behavioural measures of reading skill, and functional magnetic resonance imaging (fMRI) to elucidate the brain structures that support skilled reading.

Key Reference:

Taylor, JSH, Davis, MH. & Rastle, K. (2017). Comparing and validating methods of reading instruction using behavioural and neural findings in an artificial orthography. Journal of Experimental Psychology: General. [PDF]

Related References:

Rastle, K. (2018). The Place of Morphology in Learning to Read in English. Cortex. https://www.sciencedirect.com/science/article/pii/S0010945218300571?via%3Dihub

Rastle, K. & Taylor, JSH (2018). Print-sound regularities are more important than print-meaning regularities in the initial stages of learning to read: Response to Bowers and Bowers (2018). To appear in Quarterly Journal of Experimental Psychology. [PDF]

Castles, A., Rastle, K., Nation, K. (2018). Ending the Reading Wars: Reading Acquisition From Novice to Expert. Psychological Science in the Public Interest. http://journals.sagepub.com/doi/pdf/10.1177/1529100618772271

Multi-syllable Words.

The most influential theories of reading have a very serious limitation, in that they only explain phenomena pertaining to words with a single syllable like cat or chair. In fact, in contrast to the hundreds of studies that have examined how people read words with just a single syllable, only a handful have examined the processes through which people read words comprising more than one syllable. One reason for this gap is that words with more than one syllable present special additional challenges – for example, in understanding how people decide which syllable to stress, as in camel versus canal. The result is that our understanding of reading is effectively limited to single-syllable words, which constitute less than 10% of English words, and far fewer in many of the world’s other languages.

In a Leverhulme Trust funded project, Maria Ktori, Betty Mousikou and Kathy Rastle exploited behavioural, neuropsychological, and computational modelling approaches to understanding how people read multi-syllable words. This involves examining how adults (dyslexics and non-dyslexics) read and pronounce multi-syllable words and nonwords, and the cues people use to assign stress in such words.

Key Reference:

Ktori, M., Mousikou, P. & Rastle, K. (2017). Cues to stress assignment in reading aloud. Journal of Experimental Psychology: General. [PDF] [DATA]

Related Reference: