Effects of word frequency and contextual predictability on sentence reading in aphasia: an eye movement analysis

ABSTRACT

Background: Mild reading difficulties are a pervasive symptom of aphasia. While much research in aphasia has been devoted to the study of single word reading, little is known about the process of (silent) sentence reading. Reading research in the non-brain-damaged population has benefited from the use of eye-tracking methodology, allowing inferences on cognitive processing without participants making an articulatory response. This body of research identified two factors, which strongly influence reading at the sentence level: word frequency and contextual predictability (influence of context).

Aims: The main aim of this study was to investigate whether word frequency and contextual predictability influence sentence reading by people with aphasia (PWA), in parallel to that of neurologically healthy individuals (NHI). A second aim was to examine whether readers with aphasia show individual differences in the effects, and whether these are related to their underlying language profile.

Methods & Procedures: Seventeen PWA and associated mild reading difficulties and 20 NHI took part in this study. Individuals with aphasia completed a range of language assessments. For the eye-tracking experiment, participants silently read sentences that included target words varying in word frequency and predictability while their eye movements were recorded. Comprehension accuracy, fixation durations, and the probability of first-pass fixations and first-pass regressions were measured.

Outcomes & Results: Eye movements by both groups were significantly influenced by word frequency and predictability, but the predictability effect was stronger for the PWA than the neurologically healthy participants. Additionally, effects of word frequency and predictability were independent for the NHI, but the individuals with aphasia showed a more interactive pattern. Correlational analyses revealed (i) a significant relationship between lexical-semantic impairments and the word frequency effect score and (ii) a marginally significant association between the sentence comprehension skills and the predictability effect score.

Conclusions: Consistent with compensatory processing theories, these findings indicate that decreased reading efficiency may trigger a more interactive reading strategy that aims to compensate for poorer reading by putting more emphasis on a sentence context, particularly for low-frequency words. For those individuals who have difficulties applying the strategy automatically, using a sentence context could be a beneficial strategy to focus on in reading intervention.

KEYWORDS:

• Reading
• aphasia
• eye movements
• word frequency
• predictability

Acknowledgements

We would like to express our gratitude to all individuals who took part in this study. We also want to thank Penny Roy for her advice in data analysis, Catherine Prior for reading aloud comprehension questions, and an anonymous reviewer for helpful comments. This work was supported by a PhD studentship from City, University of London, granted to the first author.

Disclosure statement

No potential conflict of interest was reported by the authors.

Supplemental data

Supplemental data for this article can be accessed here.

Notes

1. Notably, some research involving eye tracking in aphasia has been published earlier, but these studies were limited to more global parameters of eye movements such as saccade behaviour and number of fixations (e.g., Huber, Lüer, & Lass, 1983; Klingelhöfer & Conrad, 1984).

2. Nineteen PWA took part in the study and completed the background assessments, but one participant’s eye-tracking file was corrupt, and another participant represented as an outlier in the eye-tracking experiment. Hence, both participants were excluded from the whole data set, and are not reported.

3. Written word frequencies were obtained from the SUBTLEX database (Brysbaert & New, 2009, http://www.ugent.be/pp/experimentele-psychologie/en/research/documents/ Accessed 07/07/2016). These frequency norms were shown to predict human processing latencies much better than existing norms so far (Brysbaert & New, 2009). Naming and lexical decision latencies based on British English are available through the British Lexicon Project (Keuleers, Lacey, Rastle, & Brysbaert, 2012).

4. There was one exception: The word pair doctor/explorer differed in length by two letters.

5. Another measure that captures the earliest moment of processing is first fixation duration, which refers to the duration of the first fixation on the critical word. This measure was not chosen for the present analysis, because people with aphasia tend to make multiple fixations on a word even in first-pass reading. Hence, gaze duration was thought to be a more critical measure in this experiment.

6. For the purpose of readability, Table 3 represents raw, i.e., untransformed data.

7. Since PWA and NHI spanned a large age range, the contributory role of age was checked with an additional analysis, which is presented in supplementary materials.

8. Figures represent transformed data to match the results from data analysis.

9. Simple difference scores (calculated by subtracting fixation durations in the high-frequency items from fixation durations in low-frequency items) ran the risk of confounding any group differences.

Additional information

Funding

This work was supported by a City, University of London PhD studentship.