Lexical and sublexical effects on visual word recognition in Greek: comparing human behavior to the dual route cascaded model

ABSTRACT

We evaluated the dual route cascaded (DRC) model of visual word recognition using Greek behavioural data on word and nonword naming and lexical decision, focusing on the effects of syllable and bigram frequency. DRC was modified to process polysyllabic Greek words and nonwords. The Greek DRC and native speakers of Greek were presented with the same sets of word and nonword stimuli, spanning a wide range on several psycholinguistic variables, and the sensitivity of the model to lexical and sublexical variables was compared to the effects of these factors on the behavioural data. DRC pronounced correctly all the stimuli and successfully simulated the effects of frequency in words, and of length and bigram frequency in nonwords. However, unlike native speakers of Greek, DRC failed to demonstrate sensitivity to word length and syllabic frequency. We discuss the significance of these findings in constraining models of visual word recognition.

KEYWORDS:

• Visual word recognition
• dual-route models
• syllable frequency effect
• polysyllabic words
• Greek

Disclosure statement

No potential conflict of interest was reported by the authors.

ORCID

Efthymia C. Kapnoula http://orcid.org/0000-0001-6640-1948

Athanassios Protopapas http://orcid.org/0000-0002-7285-8845

Notes

1. Since activation of the lexical route is delayed compared to the non-lexical one (see Figure 9 in Coltheart et al., 2001), this applies to words with more than two phonemes.

2. A list of word and nonword stimuli is available in Protopapas and Kapnoula (2016; Appendix).

3. In our data, positional syllable frequency was not a better measure, and this is why we have not chosen to base our report on it; in our preliminary analyses, positional first syllable frequency (equivalent to a syllable-based cohort frequency) accounted for slightly less variance in naming and lexical decision RT than frequency of the first syllable measured cumulatively over all word positions.

4. Given that this is the first investigation of its kind in Greek, we decided to use this metric of neighborhood density instead of the more recent OLD20 (Yarkoni, Balota, & Yap, 2008) in order to have data directly comparable to the majority of the literature.

5. This was done to minimise any effects on real word naming due to repeated nonword naming. As has been suggested in the literature (Coltheart & Rastle, 1994; Rastle & Coltheart, 1999), repeated exposure to nonwords can lead to a shift in the balance between the two routes, such that readers increase the contribution from the non-lexical route to phonemic activation and decrease the contribution from the lexical route. That is, as Coltheart (1978) suggested, when a reader is exposed to many nonwords in a row, they adjust their strategy either by turning down the lexical route, or turning up the non-lexical route, or both.

6. Details about model parameters and their values are listed in the Appendix.

7. Defining naming errors was straightforward because nonword stimuli had unambiguously correct pronunciations. Naming errors were mostly due to haste and inattention. For example, a nonword like δογορονήσω [ðoɣoronίso] could be misnamed as [ɣoðoronίso].

8. Even though we used Coltheart's N in our main analyses (because this measure of neighborhood was decorrelated from all other major predictors), we also performed the same set of analyses using OLD20 in its place. The results showed no difference between using Coltheart's N and OLD20: naming RT: β = .09781, p = .256; LD RT: β = .147, p = .106.

9. Given that the maximum number of cycles is reached for all nonwords, it is reasonable that the model latencies should be identical across nonword stimuli.

10. For example, ταχυδρόμος (taçiðɾˈomos), ταχυδρόμου (taçiðɾˈomu), ταχυδρόμο (taçiðɾˈomo), and ταχυδρόμε (taçiðɾˈomϵ) all correspond to the word postman in nominative, genitive, accusative, and vocative, respectively.

11. Despite it being a rule-based model, DRC could in principle exhibit sensitivity to lexical similarity when processing nonwords if the lexical route is strengthened. Keep in mind that for a word item to be named all that is required is for activation of its lexical entry to reach a given threshold. Critically, this may happen even if the input does not match the lexical entry 100%. For example, the nonword “signeficant” may be incorrectly named as “significant” if activation for the phonological entry of the word “significant” surpasses a given threshold.

12. Note that the DRC version with adjusted parameter settings mentioned above (i.e. strengthened non-lexical route) was still not able to simulate the syllable frequency effect: β = −0.029, p = .728.