The phenomenon of coronal palatalization in Modern Greek dialects was first described and phonologically treated by Newton (1972). Newton reports three levels of palatalization for /l/ and /n/: strongest before yod, intermediate before /i/, and weak before /e/, although he acknowledges that some varieties may have strong palatalization before /i/ as well. Papazachariou (2004) and Pappas (2006, 2008) report the existence of both palatal and palatalized /l/ and /n/ before /i/ in the city of Patras, and on the island of Cephalonia, where they also found expressions of strong and negative evaluative force against their usage, especially by young speakers. The stereotype appears to have emerged in the late 1990s. However, more recently Baltazani et al. (2016) argue that non-standard Modern Greek dialects do not have palatalized variants of /ni/ and /li/, only palatal ones.
Our research, which focuses on lateral palatalization, provides some crucial insights into the development of this phenomenon. The data come from the corpus of sociolinguistic interviews conducted for the purposes of the Immigrec project (Anastassiadis et al. 2017) which comprises 400 participants and over 200 hours of high quality recordings of Greeks who immigrated to Canada as adults. The particular dataset is constructed from the interviews of 66 participants (male and female) with basic education from 8 different dialect areas of Greece. Since these speakers emigrated at least two decades before palatalization emerged as a stereotype, an examination of their speech pattern allows us to determine the nature and distribution of this phenomenon across the varieties represented in the sample.
We analyzed 3274 tokens of /l/ in three different environments: before a glide (/lj/), before the low vowel (/la/) and before the high front vowel (/li/) with F1 and F2 values extracted at the midpoint in Praat, and the difference F2-F1 calculated in order to determine the degree of palatalization (Sproat and Fujimura 1993). These measurements were entered as dependent variables in linear mixed effects models with post-lateral segment (/j/, /i/ or /a/), home region in Greece and gender as fixed effects, and speakers and word as random ones. The analysis shows that there is significant interaction between region and post-lateral segment. As shown in Table 1 (after the references section), while the alveolar variants from all regions have a F2-F1(/la/) average between 818Hz and 937Hz, the F2-F1 (/li/) averages range from 1311Hz to 1530Hz, and the F2-F1(/lj/) ones range from 1368Hz to 1590Hz. Post hoc (Tukey) tests show that Athens is the only region which has a clear distinction among all three environments, whereas in all other regions the F2-F1 difference for (/lj/) and (/li/) is almost indistinguishable. In our discussion, we present a hypothesis of how the differentiation between palatal and palatalized variants of /l/ may have emerged in Athens during the period of rapid urbanization of the 1950s that brought speakers from all regions of Greece into close contact.