Previous representations of secondary palatalization

68 t‚kaµt s ‘to change.’ When ojÚπıj precedes t‚kaµts, the final j of ojÚπıj palatalizes the t of t‚kaµts, which becomes voiced [ d j ], and the j is absorbed. Although this linear process results in consonants that are palatalized by secondary articulations of palatalization, it is different from the type of secondary palatalization that manifests a morpheme or is the result of final-syllable truncation by the fact that it spreads voicing to the consonant that it palatalizes, and it never affects a consonant cluster, since no consonant cluster occurs morpheme initial. Positionally, secondary palatalization that expresses a morpheme always occurs word initially or word finally, whereas, linear palatalization occurs only word medially. In the case of verb-final secondary palatalization, the consonants that precede secondary palatalization are palatalized, whereas, when a morpheme ends in a consonant and is followed by a morpheme beginning with j in a compound, the consonant preceding j is not palatalized. For example, when m ‚k ‘strong’ precedes jo πıj ‘to walk,’ it is pronounced [m‚kjoπıj] as in the expression j -ka π-m‚k-joπıj- j [k j a ÿπm‚kjoπıj] ‘he doesn’t walk well.’ A second source of palatalization in all the Oaxacan Mixe languages is related to historical processes of syllable truncation from which arise four categories of secondary palatalization: 1 the third person marker word initial i.e. Ði- has become secondary palatalization; 2 verb-final clause-type markers i.e. - hi or -ie have become secondary palatalization; 3 a deverbalizer; and 4 other word final palatalization. 88 See section 3.7.5 regarding the origin of secondary palatalization in the Oaxacan Mixe language for examples and further explanation. These historical processes have been completed, with the result that secondary palatalization has replaced these syllable affixes in the Oaxacan Mixe languages. In present-day Isthmus Mixe, there is no phonemic basis causing the secondary palatalization that represents the morphemes i.e. the third person, a verb-final clause-type marker, 89 and a deverbalizer. These morphemes are represented by secondary palatalization alone, which attaches to the word-initial consonant in the case of the third person marker, and the word-final consonant or consonant cluster in the case of the clause-type marker and the deverbalizer. 90 Since any consonant in the language may occur in the word-initial position, any model that is used to describe these abstract palatalization morphemes must be able to accommodate secondary palatalization of the complete inventory of Isthmus Mixe consonants.

5.3.1 Previous representations of secondary palatalization

The question may be asked as to why the affixes represented by secondary palatalization cannot be represented simply by a j prefix or suffix. This has been done in the past, especially by Crawford 1963 in describing Totontepec Mixe, in which he phonemicized the third-person prefix j- as Cj. Crawford includes a text and its analysis, where mention is made of y as third person: “{vyá ºhny} ‘he said’ has the person-prefix {y} ‘third person’ which is metathesized with the stem-initial consonant, …” 1963:164. In his phonetic description, Crawford indicates the Cj sequence patterns as one phonemic unit see section 3.7.2. Moreover, Crawford does not give any reason for preferring Cj i.e. the metathesis of the prefix over jC i.e. prefix plus initial consonant. As shown in section 5.3, in Isthmus Mixe there is a sequence of Cj word medial, in which the consonant that precedes j is not palatalized. Therefore, if one were to adopt the representation of the third- 88 There are possibly other categories of secondary palatalization related to syllable truncation not yet discovered, since an extensive study of related languages has not been undertaken. 89 Palatalization of the final consonant of the verb is characteristic of the three types of conjunct clauses, and it also occurs in conjunction with the verb-final suffix -p in the nonconjunct direct-transitive clause. These clauses are described in Dieterman 1995 and Dieterman 1998. 90 This is also true of Coatlán Mixe. Further study of the other Mixe languages is needed in this regard. 69 person morpheme as a linear segment j which metathesizes with the following consonant, palatalizes it, and is absorbed in the process, another rule would have to be made for Cj in word-medial position. In contrast to Crawford’s metathesized prefix j-, Wichmann 1995 does not metathesize the third person prefix j-; rather he indicates that j palatalizes a following consonant, with j absorbed in the process see chapter 4, section 4.4.7. This type of palatalization does not cause voicing of the following consonant. However, he does not mention the process in which word-medial j palatalizes and voices the following plosive, as happens in Isthmus Mixe see section 5.3. Wichmann 1995:22–23, 44–45, 167, 173 describes umlauting of vowels, as effected by palatalized consonants on adjacent vowels in some Mixe languages; however, neither Crawford nor Wichmann mention that [secondary] palatalization affects both consonants in a cluster. 91 The effects of palatalization have been shown to extend over multiple segments by Van Haitsma [Dieterman] and Van Haitsma 1976 and Hoogshagen 1984, as described previously see section 3.7.4. Van Haitsma [Dieterman] and Van Haitsma 1976:5–11 refer to [secondary] palatalization in the Mixe of San José El Paraíso as a suprasegmental phoneme, written as a tilde over one consonant or two consonants in a cluster. Distinct from palatalization is the phoneme j which does not palatalize the first person marker prefix n- or the second person prefix m- when they precede a word beginning with j. Hoogshagen 1984:4 describes the same distinctions in Coatlán Mixe. In Isthmus Mixe, the differences between m- ‘second person’ followed by a word initial j i.e. the sequence of mj and the third person morpheme as manifested by secondary palatalization attached to a word initial m i.e. m j are described and shown by waveform and spectrogram displays in section 4.4.4. Thus, a word-initial consonant cluster of a nasal person marker followed by a stem-initial j behaves differently than the word-initial secondary palatalization of a stem-initial nasal. In Isthmus Mixe, the morphemes manifested by secondary palatalization are morphologically prefixes and suffixes; however the phonological features of secondary palatalization are not well represented by the linear phoneme j plus consonant, because the secondary palatalization feature and the consonant to which it attaches does not behave as an ordered consonant cluster of either jC or Cj. Rather, it behaves as an unordered complex segment see section 5.4. The domain of secondary palatalization in Isthmus Mixe is not just one adjacent consonant, but includes a consonant cluster in both word-initial and word-final positions, and adjacent vowels, as shown in sections 4.4.2 and 4.4.6, and further described in section 5.5.1. Therefore, an autosegmental model better describes the process of secondary palatalization than is possible with a linear description. As an autosegmental feature, secondary palatalization in Isthmus Mixe is recognized as a mutation process, comparable to other mutation processes such as described by Lieber 1983, McCarthy 1983, and Mester and Ito  1989 see section 5.6.2. Before describing the autosegmental model, nonlinear feature geometry models that have been used to describe secondary articulations are briefly described in the following sections relative to their suitability for describing the Isthmus Mixe data. These are basically of two types: articulator models Sagey 1986; Halle, Vaux, and Wolfe 2000 and vowel-place models Hume 1994; Ní Chiosáin 1994.

5.4 Articulator models