Perception of European Portuguese [ɐ] and [ɨ] by Hungarian native speakers

Activity: Talk or presentationOral presentation

Description

Research addressing the perception and acquisition of non-native contrasts have identified the perceived distance between L1 and L2 categories as a key factor in learnability,
with studies showing that while some non-native contrasts are easy to discriminate, other pose significant difficulties (Best, 1995; Elvin et al., 2014; Escudero & Boersma, 2004; Flege, 1995; Flege & Bohn, 2021; Polka, 1995).
The present study focuses on the perception of the European Portuguese (EP) vowels [ɐ] and [ɨ] by Hungarian native speakers. Hungarian is a non-Indo-European language, with fourteen phonemic and phonetic monophthongs: /ɒ aː ɛ eː i iː o oː ø øː u uː y yː/ (Markó, 2017). Vowel length is contrastive (e.g., örül [ˈøryl] 'rejoiced' versus őrül [ˈøːryl] 'getting crazy'), but stress-induced vowel quality change has not been attested (Gósy, 1997).
Assuming that the perception of non-native sounds depends on the phonetic proximity of these sounds with L1 sounds, we predicted that Hungarian speakers categorize the EP
unstressed vowel [ɐ] into /ɛ/, /eː/ or /ø/, and [ɨ] into /y/, /eː/, /ɛ/ and /ø/, considering that these are the closest L1 categories to the EP vowels (Figure 1).
To test our predictions, we conducted a perceptual forced-choice identification task with goodness-rating, in which Hungarian participants were asked to match and evaluate EP vowels with Hungarian vowel categories (Bundgaard-Nielsen et al., 2011; Elvin et al., 2014; Faris et al., 2018; Polka, 1995). EP stimuli were presented in an auditory form and consisted of the nine oral vowels – [a ɐ ɛ e ɨ i o ɔ u] – inserted in a [ɡV] context and recorded by three female native speakers of Portuguese. Hungarian vowel categories were presented in real Hungarian words with a CVCVC structure, visually displayed in a 3x3 grid (Figure 2). The first syllable of these words contained the [ɡV] possible in Hungarian, with vowel length being disregarded whenever possible: [ɡɒ], [ɡaː], [ɡɛ], [ɡeː], [ɡi], [ɡo], [ɡø], [ɡu] and [ɡy]. After the identification task, participants had to evaluate the similarity between the EP vowel they heard and the chosen Hungarian category, in a 4-points Likert scale (1 = very bad example, 2 = bad example, 3 = good example, 4 = very good example). Each Hungarian speaker completed a total of 27 trials (9 EP vowels x 3 EP speakers), randomised across participants. The experiment was built in PsychoPy version 2020.2.3 (Peirce
et al., 2019) and hosted online by pavlovia.org.
Seventy-eight Hungarian native speakers with no previous contact with EP were recruited, as well as 30 Portuguese native speakers without previous contact with Hungarian, to
serve as a baseline condition. All participants were aged 18 to 45, and none reported hearing impairments.
Results partially confirmed our hypothesis: [ɐ] was categorized into /ɛ/ or /ø/, while [ɨ] was categorized into /y/ or /ø/ (Table 1). Regarding goodness-ratings, /y/ was evaluated as a
“good example” of [ɨ], while /ø/ had a negative rating. As for the perception of [ɐ], /ɛ/ and /ø/ were equally perceived as a “good example” of the EP vowel.
The results found can be grounded in different explanatory factors. First, the comparison of formant dispersion means between our stimuli and the values of Hungarian perceptual map was based in values from male Hungarian speakers. Therefore, the partial confirmation of our predictions can be explained by the fact that EP female production values were crossed with Hungarian male perception values. Second, differences in phoneme frequency of occurrence may have also biased the results: while [ɛ] has a frequency of 11,4%, [eː] only represents 3,1% (Gósy, 2004). Third, vowel length may have played a role in the results. Although EP stimuli were manipulated to equalize vowel length, Hungarian [e:] exists only as a long vowel and [ɛ] as a short vowel. Consequently, it is possible that the Portuguese [ɛ] did not meet the [e:] length criterion for Hungarian participants. Fourth, χ2 tests showed a speaker effect in the responses, in the perception of both [ɐ] and [ɨ] (χ2 = 152.19, p < 0.001 and χ2 = 99.601, p <0.001, respectively). Finally, we also look for effects of f0 in vowel openness’ perception (Fahey et al., 1996). Values of the tonotopic distances between the first formant and f0 confirmed that this may have also affected the perception of the two EP vowels by Hungarian listeners.
The present experiment is part of a longitudinal study in acquisition of EP Phonology by Hungarian speakers. Results from this study will allow us to observed changes in categorization of the EP vowels before and after an auditory perceptual training, designed for this purpose, thus contributing to the underexplored research field of L2 European Portuguese Phonology.
Period28 Oct 2023
Event titleXXXIX Encontro Nacional da Associação Portuguesa de Linguística
Event typeConference
LocationPortugalShow on map
Degree of RecognitionNational

Keywords

  • L2 phonology
  • Cross-language speech perception
  • Vowel categorization
  • European Portuguese