https://orcid.org/0000-0003-0391-6164
![Sample our Language & Literature journals, sign in here to start your FREE access for 14 days]({"type":"image" , "src" : "/sda/5941/TOC-Subject-Sample-2021-Language-Literature.jpg"})
ABSTRACT
Deaf individuals who learned a spoken language with the aid of a cochlear implant (CI) often experience difficulties with reading. In the present study, we investigate this issue by assessing the impact of lexical and orthographic predictors on visual word recognition in early and late deaf CI-users. Early deaf CI-users were comparable to age-matched hearing controls, for both response accuracy and latencies, whereas late deaf CI-users were slower albeit similarly accurate with respect to age-matched controls. Analyses of the impact of lexical and orthographic predictors, however, revealed that early deaf-CI users were slower than controls in recognising low frequency words and words with high proportion of consonants. In conclusion, early deaf individuals who developed their language using a CI show qualitative differences in visual word processing, above and beyond the overall performance. This suggests that typical overall performance may emerge also in the context of atypical lexical and word form representations.
Acknowledgements
We thank all participants in the study and their families. We thank Valentina Musella from Cooperativa Logogenia Milano, and Angela Chini, Giuseppina Tromballi and Francesca Bonfioli from the Presidio Ospedaliero Santa Maria del Carmine of Rovereto for their help in coordinating CI users recruitment. We are thankful to Gaia Fantoni from the centre “Mi prendo cura di te” in Milan for providing logistics for this study. We also thank Valeria Giannelli, Mara Dighero, Elena Ragusa and Vincenza-Claudia D'Avanzo for their help in data collection. Finally, we thank Fritz Günther for discussions and help in data visualisation. F.P. and F.V. are supported by a PRIN grant from the Italian Ministry for University and Research (Prot. 20177894ZH). F.P. is supported by a grant from Fondation Medisite (France), from Fondation Neurodis (France) and from Agence Nationale de la Recherche (France).
Disclosure statement
No potential conflict of interest was reported by the author(s).
Notes
1 In this study, we make use of Linear Mixed Effect Models (LME) to analyse our data. These models allow us to analyse the impact of continuous predictors at the same time on the dependent variable, and, to account for individual variability of items and participants. Differently from other studies of clinical tradition, but in keeping with the state of the art of psycholinguistic research, we have decided to maintain the continuous nature of our predictors instead of dichotomizing them (e.g., we did not operationalized frequency on two levels: high vs. low, but we used actual estimates of lexical frequency in our model). Indeed, it has been shown that when a variable is naturally continuous, such as lexical frequency, operationalizing it in a continuous way leads to better performance in comparison to its dichotomized counterpart, both in terms of accuracy of the estimated relations and statistical power (see, e.g., Cohen, Citation1983; DeCoster et al., Citation2009). Furthermore, it was shown that the dichotomization of continuous variables might inflate Type I error rates (Maxwell & Delaney, Citation1993). Furthermore, the use of mixed effect modelling, allows us to include trial level estimates of the dependent variables (both response times and accuracy estimates), without resorting to averaging across conditions. With these considerations in mind, we believe that LME design offers the better fit for our investigation (Baayen, Citation2004, Citation2010).
2 In line with the majority of studies of this type, also in this work we focus specifically on word items and do not provide a direct comparison between words and non-words. Besides, abiding to the standard in the literature, analyzing non-words was beside the scope of this work. However, we analysed responses to non-words to monitor the overall pattern in our data. For non-words, the only significant interaction emerges, in the RTs analyses, for early deaf CI users between group and length (t = 3.130; p = 0.00176). CI users are slower to respond to longer non-words in comparison to their controls (see Figure 4). No significant interactions emerged for late deaf CI users for either RTs or accuracy data.
3 We also included a measure of orthographic neighbourhood density as a covariate in our analysis. In fact, Carreiras et al. (Citation2009), suggested that the effect of CVP may be similar to the effect of orthographic neighbourhood density, so it was important to add this variable to our models in order to control for possible confounds. We operationalized neighbourhood density using the OLD20 measure (Yarkoni et al., Citation2008). In our dataset, CVP and OLD20 presented a very weak correlation (r = .063, p = 0.2), which allowed us including both predictors in our models. OLD20 did not prove to give significant contribute to any of the models fit, neither for lexical decision nor for accuracy, and was therefore removed from models prior to refitting.