Metathesis in Judeo-Spanish Consonant Clusters

Judeo-Spanish (JS) denotes those varieties of Spanish preserved by the Sephardic Jews who were expelled from Spain in 1492 and have emigrated throughout Europe, North Africa, the Middle East, and the United States. Some descriptions of JS emphasize its conservative, archaic nature, highlighting its similarities to Old Spanish (OS), while other research points up the novel characteristics that distinguish JS from both OS and other varieties of Modern Spanish (MS). Despite its status as an endangered language due to the lack of monolingual speakers and to the ever decreasing number of fluent speakers under the age of fifty-five (T. Harris 1994), modern JS remains underrepresented in the most recent work on Ibero-Romance phonetics and phonology (although see Bradley and Delforge 2006 on sibilant voicing and rhotics in Istanbul JS). One phonological phenomenon of JS in which both retention and innovation can be observed is consonant metathesis. This paper examines two types of productive metathesis in JS consonant clusters, e.g., daldo < dadlo ‘give (PL.) it’ and tadre < tarde ‘late, afternoon’. Holt (2004) proposes an account in Optimality Theory of dl and dn metathesis in OS as a strategy for repairing bad syllable contact. I show how the analysis straightforwardly accounts for dl metathesis in JS imperative-clitic sequences. In OS, metathesis did not affect heteromorphemic dm clusters, and the same restriction is found in modern-day JS, e.g., dadme vs. *damde/dande ‘give (PL.) me’. I extend Holt’s account, through nasal place assimilation and positional faithfulness constraints, in order to explain the failure of dm metathesis across morpheme boundaries. Unlike dl metathesis, transposition of rd clusters does not result from syllable contact optimization. I analyze this innovation as an effect of the Obligatory Contour Principle (Leben 1973, McCarthy 1986), whereby adjacent segments identical in place, manner, and voicing specifications are prohibited. I hypothesize that /ɾ/ in JS came to be realized phonetically as approximant [ɾ]̞ in coda versus noncontinuant [ɾ] in onset, due to syllable-position effects (Kochetov 2006, Krakow 1999). Heterosyllabic [ɾ.̞ð] contains adjacent coronal voiced approximants, whereas coronal clusters of coda [ɾ]̞ followed by [t], [n], [l], or [s], as well as tautosyllabic [.ðɾ], all have non-identical featural specifications. The ranking of OCP(place, manner, voice) above constraints on segment order explains why only rd clusters underwent metathesis in JS. Finally, I consider some of the difficulties posed by the OS and JS data with respect to perceptuallybased approaches to consonant metathesis (Blevins and Garrett 2004, Hume 2004). This paper is organized as follows. Section 2 presents the lateral and nasal metathesis data and reviews the analysis of Holt (2004). Section 3 extends the analysis to account for the restriction involving dm. Section 4 examines rd metathesis and develops an account in terms of the Obligatory Contour Principle. Section 5 considers perceptual approaches to consonant metathesis. Section 6 concludes with a summary of the analyses proposed here.