Abstract:

This study focuses on how speakers of Polish, English, and Norwegian perceive retroflex consonants. It explores how the similarity between a learner's native language (L1) and second language (L2) affects their ability to distinguish L2 sounds (Flege & Bohn 2021; Cebrian 2022). The research investigates how the presence or absence of retroflexion influences the perceived similarity of consonants in Norwegian, Polish, and English.

Across different languages, retroflex consonants are considered marked (Greenberg 1966). In Norwegian, there is a group of coronal consonants that are distinguished by retroflexion: alveolar sounds such as /t, d, s, l, n/ and retroflex sounds such as /ʈ, ɖ, ʃ, ɭ, ɳ/. On the other hand, English only has the retroflex /ɽ/, while the retroflex status of sibilants and stops in Polish is a subject of debate. Some argue that cues to retroflexion are present in sounds such as /ʂ/, /ʐ/, /t͡ʂ/, and /d͡ʐ/, while cues to allophonic retroflexion are observed in /ʈ/ and /ɖ/ (Żygis 2005; Żygis, Pape & Jesus 2012).

34 L1 Polish L2 English L3 Norwegian students of Scandinavian studies (all classroom setting English and Norwegian learners) and 35 L1 Polish L2 English undergraduate students of English Philology (all classroom setting English learners) participated in two perceptual tasks: a rated similarity task (RST) and an oddity categorial discrimination (OCD). Additionally, 13 L1 Norwegian L2 English speakers participated in OCD as a control group. In RST, 180 triads in the L3 Norwegian discrimination task were made up of tokens of consonant categories that contained both retroflexes (i.e., /ʈ ɖ ʂ ɭ ɳ/) and non-retroflexes (i.e., /t d s l n/) in inter-vocalic position. In the cross-linguistic RST, participants evaluated the similarity between retroflex and non-retroflex sounds in Norwegian and their counterparts in English and Polish in 160 diads, on a scale from 1 to 7. In OCD, participants discriminated between tokens of consonant categories that contained both retroflexes and non-retroflexes in the L3 Norwegian 180 triads.

We studied how Polish learners of Norwegian, at naïve or advanced proficiency levels, perceive differences in retroflexion and articulation. We hypothesized gradience in the salience of retroflex sounds based on matching in retroflexion and articulation. Results confirmed that dissimilarity ratings aligned with this hierarchy for both naïve and advanced learners.

Results from an OCD task showed high discrimination accuracy (97%) among both naïve and advanced Norwegian learners for /ʂ/-/s/, but surprisingly, native Norwegians scored lower (84%). Advanced learners excelled in discriminating /ʈ/-/t/ and /ɖ/-/d/ pairs (85%), as well as /ɳ/-/n/ (80%). /ɭ/-/l/ discrimination was below chance level (36%). Naïve learners performed slightly above for /ʈ/-/t/, /ɖ/-/d/, and /ɳ/-/n/ (50-60%) and below for /ɭ/-/l/ (36%). Norwegian natives scored highest for /ʈ/-/t/ (87%) and accurately for /ɖ/-/d/ and /ɳ/-/n/ (70-80%), but poorly for /ɭ/-/l/ (35%), possibly due to infrequent occurrence of /ɭ/ in their phonemic inventories. The study suggests that the importance of phonemic inventories in discrimination abilities is less significant than previously thought, and the absence of a sound doesn't necessarily imply inability to perceive the contrast, warranting further investigation.

We hope to have brought some insights to the understanding of non-native speech perception from the perspective of multilingual acquisition.

Keywords:

Crosslinguistic similarity, perception, discrimination, rated similarity, Polish, Norwegian, English, retroflexes.