Does the mean adequately represent reading performance? Evidence from a cross-linguistic study

Chiara V. Marinelli, Joanna K. Horne, Sarah P. McGeown, Pierluigi Zoccolotti, Marialuisa Martelli

Reading models are largely based on the interpretation of average data from normal or impaired readers, mainly drawn from English-speaking individuals. In the present study we evaluated the possible contribution of orthographic consistency in generating individual differences in reading behavior. We compared the reading performance of young adults speaking English (one of the most irregular orthographies) and Italian (a very regular orthography). In the 1st experiment we presented 22 English and 30 Italian readers with 5-letter words using the Rapid Serial Visual Presentation (RSVP) paradigm. In a 2nd experiment, we evaluated a new group of 26 English and 32 Italian proficient readers through the RSVP procedure and lists matched in the two languages for both number of phonemes and letters. The results of the two experiments indicate that English participants read at a similar rate but with much greater individual differences than the Italian participants. In a 3rd experiment, we extended these results to a vocal reaction time (vRT) task, examining the effect of word frequency. An ex-Gaussian distribution analysis revealed differences between languages in the size of the exponential parameter (tau) and in the variance (sigma), but not the mean, of the Gaussian component. Notably, English readers were more variable for both tau and sigma than Italian readers. The pattern of performance in English individuals runs counter to models of performance in timed tasks (Faust et al., 1999; Myerson et al., 2003) which envisage a general relationship between mean performance and variability; indeed, this relationship does not hold in the case of the English participants. The present data highlight the importance of developing reading models that not only capture mean level performance, but also variability across individuals, especially in order to account for cross-linguistic differences in reading behavior.