Izvorni znanstveni članak
https://doi.org/10.31724/rihjj.51.1.10
Bilingual Reading Corpus Analysis: A Comparison Between Frequentist and Bayesian Approach
Marijan Palmović
orcid.org/0000-0002-4587-4348
; Edukacijsko-rehabilitacijski fakultet Sveučilišta u Zagrebu, Odsjek za logopediju, Laboratorij za psiholingvistička istraživanja
Kristina Cergol
; Učiteljski fakultet Sveučilišta u Zagrebu, Katedra za obrazovanje učitelja engleskog jezika
Sažetak
The reading corpus constitutes a new methodology in psycholinguistic research, a methodology where eyetracking data are collected from experimentally non-manipulated stimuli, i.e. texts. This paper discusses the appropriate statistical analysis method for this type of research. A logistic regression model, either frequentist or Bayesian, is proposed, as such a model is used for predicting categorical target variables, which most linguistically relevant categories are. The superiority of this analysis over previous analyses in confirming the Implicit Prosody Hypothesis has been demonstrated and the logistic regression models confirmed that we unconsciously follow the stress structure of the text even in silent reading. For this analysis the existing parallel Croatian-English corpus has been used. No substantial differences were found between the frequentist and Bayesian models, although Bayesian model has shown some advantages.
Ključne riječi
reading corpus; bilingualism; Implicit Prosody Hypothesis; statistical models
Hrčak ID:
340105
URI
Datum izdavanja:
30.12.2025.
Posjeta: 655 *