Perceptual evidence of Modern Greek voiced stops as phonological categories

Mark Antoniou1, Catherine T. Best1,2, Michael, D. Tyler1

MARCS Auditory Laboratories, University of Western Sydney, Australia 2 Haskins Laboratories, New Haven, CT, USA,,

Despite much academic debate, the phonological status of voiced stops in Modern Greek (MG) remains unclear. Native listeners typically discriminate phonological contrasts in their native language without difficulty. MG listeners showed excellent discrimination of the MG bilabial [p]-[b] and coronal contrasts [t]-[d], significantly better than nonnative Australian English (AE) [ph]-[p] and [th]-[t]. These results are interpreted as evidence of the phonological status of MG voiced stops /b, d/. Index Terms: speech perception, Greek, stop consonant, voicing.

3.Results and Discussion
MGs showed consistent ID of all MG stops. MG [p, b, t, d] were consistently identified as /p, b, t, d/, respectively. All MG stops were rated above ‘6,’ near the native ideal, as would be expected for native phonological categories. As predicted, both AE voiced and voiceless stops were assimilated to MG voiceless categories (Table 1). AE [t] was identified as ‘t’ in 53.75% of responses, as ‘d’ 37.50% of the time and as prenasalised ‘nd’ and ‘nt’ for the remainder. Table 1. Identification category label, % identification and mean category-goodness rating for each phone (1 “very unusual” to 7 “native”).
Stimulus AE /p/ AE /b/ ID p p % 100 95.00 Rating 3.73 4.36 Stimulus MG /p/ MG /b/ ID p b % 98.75 98.33 Rating 6.29 6.14 AE /t/ AE /d/ t t 100 53.75 3.51 3.97 MG /t/ MG /d/ t d 95.00 88.75 6.32 6.13

Voiced stops, such as [b, d], occur phonetically in MG, and may be prenasalised [mb, nd] in some contexts. Despite their occurrence, the phonological status of MG voiced stops remains unclear [1]. No one disputes the phonological status of MG voiceless stops, which are typically described as short-lag voice onset time (VOT) [2], e.g. [p, t], although their production may vary from partially voiced to voiceless [3]. The variability in MG is of interest to phonetic and phonological theories. Despite longstanding phonological debate, MG listeners’ perception of stop voicing distinctions has not been sufficiently investigated to address the unusual phonetic and phonological status of MG stop voicing, nor has cross-language perception been examined. AE differs from MG in the phonetic settings of VOT used for stop voicing distinctions. AE voiced stops have short-lag VOT in initial position, e.g. [p, t], while voiceless stops are long-lag aspirated in initial position, e.g. [ph, th]. If MG voiced stops have full phonological status as contrastive with voiceless stops, MGs should easily perceive MG stops produced with voicing lead [b, d] and short-lag [p, t] and discrimination will be excellent. The Perceptual Assimilation Model (PAM) [4] predicts that MGs will perceptually assimilate AE [p, t] (short-lag) as good exemplars of MG /p, t/ and AE [ph, th] (long-lag) as poor exemplars of MG /p, t/. This will result in a categorygoodness (CG) assimilation, and discrimination of the AE contrasts should be moderate.

The MG contrasts were discriminated significantly better than the AE contrasts (F(1, 20) = 24.277, p < .001, ηp2 = .548). Discrimination of MG contrasts was excellent (both >90%), while for AE it was moderate, AE /p/-/b/ = 73.21% and AE /t/-/d/ = 75% (Figure 1). This is consistent with what would be expected when comparing discrimination of native phonological contrasts, which belong to separate phonological categories, compared to nonnative contrasts, which in this case reflect a CG assimilation.
M % correct
1 00 90 80 70 60 50

AE [ph]-[p]

AE [th]-[t]

M G [p]-[b]

M G [t]-[d]

Figure 1. Mean % correct discriminations per contrast. Standard error bars are displayed.

Two AE and two MG speakers produced AE and MG bilabial and coronal tokens in /Ca/ context. Native MGs (N = 21) completed an AXB discrimination task (16 randomised trials per contrast) and a forced-choice identification (ID) task (‘p,’ ‘b,’ ‘t,’ ‘d,’ ‘mp,’ ‘mb,’ ‘nt,’ ‘nd’) with ratings (1 “very unusual” to 7 “native”). Tokens were presented once for identification and a second time for rating.

The results provide perceptual evidence that MGs perceive MG [b, d] as phonological categories, i.e. /b, d/, and that native MG categories constrain discrimination of nonnative contrasts, such as AE /p/-/b/ and /t/-/d/. Future research should investigate MG listeners’ perception of stops in wordmedial contexts to assess the effect of prenasalisation on MG perception.

[1] Arvaniti, A., & Joseph, B. D. (2000). Variation in voiced stop prenasalization in Greek. Glossologia, 11-12, 131-166.

[2] [3] [4]

Lisker, L., & Abramson, A. (1964). A cross-linguistic study of voicing in initial stops: Acoustical measurements, Word, 20, 384– 422. Botinis, A., Fourakis, M., & Prinou, I. (2000). Acoustic structure of the Greek stop consonants, Glossologia, 11-1, 167-199. Best, C. (1995). A direct realist view of cross-language speech perception, Speech perception and linguistic experience: Issues in cross-language research, 171-204.