Linguistics Colloquium: Si Berrebi

11/01/2022 - 14:00 - 15:30

Si Berrebi, Tel Aviv University

Title: Hearing Hebrew pharyngeals: Experimental evidence for a covert phonemic distinction

Si Berrebi, Noa Bassel and Roey Gafter


Among Modern Hebrew speakers, the voiceless pharyngeal fricative /ħ/ is strongly associated with ethnicity (Matras and Schiff 2005). In the speech of Jews of European descent (Ashkenazis), /ħ/ has undergone a phonemic merger with its non-pharyngeal counterpart /x/. Among Jews of Middle Eastern and North African descent (Mizrahis), the phonemic distinction between /ħ/ and /x/ was traditionally retained, but due to an ongoing and advanced change in progress, the /ħ/–/x/ merger has become increasingly common among Mizrahi speakers as well (e.g. Bentolila 2002, Gafter 2019). Previous research on perception has demonstrated that speakers’ production of dialect features need not necessarily match their linguistic representations (Sumner and Samuel 2009). Building upon this insight, we show evidence from a perception experiment that “first generation” Mizrahi merged speakers, who grew up exposed to the unmerged variety, still maintain mental representations of a phonemic distinction between /ħ/and /x/.

The study included 75 participants, divided into three groups:
(i) Mizrahi speakers who distinguish between /ħ/and /x/ in production
(ii) Mizrahi speakers who merge /ħ/and /x/ in production, but were exposed to the unmerged variety (i.e. their parents’ dialect) during acquisition
(iii) Ashkenazi speakers who merge /ħ/and /x/, with no early exposure to the unmerged variety.

The participants performed a lexical decision task, in which they heard (in random order) a merged Ashkenazi speaker and an unmerged Mizrahi speaker. They then heard a manipulated block in the voice of the same Mizrahi speaker, consisting of words with:
a. /ħ/ realized as [ħ]                 b. /x/ realized as [x]
c. /ħ/ realized as [x]                 d. /x/ realized as [ħ].
Note that /x/→ [ħ] tokens, as in (d), do not occur in any variety of Hebrew (since merged realizations are always non-pharyngeal). /ħ/→ [x] tokens, on the other hand, as in (c), are the mainstream merged pronunciation, but may be perceived as incongruous from a speaker previously heard to be unmerged.

The results show a similarity in the patterns of participant groups (i) and (ii): both rejected /x/→ [ħ] in high rates as non-words, unlike group (iii), who accepted them as words. Furthermore, an analysis of reaction times shows that groups (i) and (ii) were significantly slower at recognizing /ħ/→[x] tokens in the Mizrahi voice. None of the groups, however, were slowed down by /ħ/→[x] tokens in the Ashkenazi voice (for whom the merger is expected).

These findings demonstrate that although group (ii) speakers merge /ħ/ and /x/ in production, their perception is consistent with speakers who maintain the phonemic distinction, suggesting that speakers of the same variety may have divergent lexical representations. Furthermore, since they were slowed down by /ħ/→[x] tokens only in the Mizrahi speaker’s voice, the results join a growing body of work advocating for a model of speech perception in which social information plays an integral role in word recognition (e.g., Sumner et al 2014, Cai et al. 2017). 


Subscribe to our Telegram channel to get notified about upcoming talks and events