Theoretical issues in the devoicing of /ʒ/
The devoicing of /3/ has been approached from both generative-related perspectives and usage-based perspectives. The generative approach to phonology informed sociolinguistic work on the topic, particularly regarding the underlying representation posited by the researcher. Section 11.2.1 addresses the nature of the associated theoretical concerns. Diaz-Campos and Gradoville (2011) looked at the phenomenon from a usage-based standpoint. Section 11.2.2 describes the various usage-based theoretical constructs that are relevant for the devoicing of /3/.
From a generative perspective, a source of considerable discussion has been the underlying representation of the palato-alveolar fricative in Argentine Spanish. Within traditional transformational frameworks, a phoneme is considered to be an “underlying sound” that is stored in the lexicon and suffers rule-based transformations to produce the allophonic output (see Kenstowicz 1994: 66). Optimality Theory similarly requires a process to arrive at the allophonic output from the phonemic specification in the lexicon (Kager 1999). While most of the discussion about the underlying representation of the  and [(] phones has focused on whether they are underlyingly voiced, Harris and Kaisse (1999) are notable for arguing that they are underlyingly a vowel due to alternations such as those in example (1), where a word-final glide alternates with the palato-alveolar fricative, or example (2), where the first segment of several verbal morphemes alternates between a glide after a syllable onset and the palato-alveolar fricative when it must serve as the syllable onset. Harris and Kaisse (1999) propose that the palato-alveolar fricative is derived through a rule-based process, highlighting certain perceived problems that the data present for the then-emerging Optimality Theory. Since it is not the goal of this chapter to exhaustively discuss the proposals by Harris and Kaisse (1999), readers interested in their approach should consult the article in question.
(1) le[jj “law,” le[z]es “laws”
Urugua[ “Uruguay,” urugua$[o “Uruguayan”
(2) cre.cj]e.ron “they grew,” cre.[^]e.ron “they believed”
cre.c$en.do “growing,” cre.[^]en.do “believing”
cre.c[j]6 “s/he grew,” cre.[ з]о “s/he believed”
While the most common approach is to assume that the phoneme is /3/, represented by symbol /3/, it may be considered to either represent an underlying voiced palato-alveolar fricative or the reflex of an ultimately underlying vowel that manifests itself as a fricative, depending on the individual’s theoretical position within generative linguistics. More germane to the topic of the devoicing of the palato-alveolar fricative is work that addresses the voicing status of voiced and voiceless palato-alveolar fricatives. Fontanella de Weinberg (1979: 12) posited the need to model two separate phonemes for speakers that voice the fricative in native words. The additional phoneme /// accounts for recent loan words such as shopping “shopping mall,” kosher “kosher,” and Chenaut “surname of French origin,” which would be pronounced as the voiceless palato-alveolar fricative [f], regardless of the voicing status of the individual’s native Spanish words. The voiceless variants produced by such speakers in native words are considered to be a consequence of a variable rule. As the devoicing process approaches completion, the opposition between /3/ and /J7 is then said to neutralize in favor of the latter.
Rohena-Madrazo also addressed the issue of underlying representation. Specifically, this author points to many recent acoustic studies that show evidence of partial or complete voicing of phonologically voiceless stops and fricatives in contemporary Spanish in some instances (2015: 292). Consequently, the threshold at which the devoicing of /3/ should be argued to be complete should not be when all tokens are 100% voiceless, but rather when the distribution is no longer significantly different from another phonologically voiceless sound. In the case of Rohena-Madrazo’s (2015) study, realizations of /3/ were compared to the sibilant Isl and, for a given speaker, the sound was considered to be underlyingly /J7 when it did not differ significantly from Isl. Speakers with an underlying /3/ are considered to be voicers while speakers with an underlying /// are considered to be devoicers.