TONAL MELODIES AND MEEUSSEN’S RULE IN KHAYO * Michael R. Marlo 1. Introduction This article highlights some of the more theoretically interesting properties of the verbal tonal system of the Khayo dialect of Luyia, a Bantu language of western Kenya and eastern Uganda, on the basis of Marlo’s (to appear, b) comprehensive description.1 This paper has two main goals: (i) to demonstrate that the application of tonal rules in Khayo is tightly connected with the morphosyntactic context and (ii) to provide an analysis of some of the details of the realization of tonal melodies in Khayo, with special attention to a system-wide paradox in the application of Meeussen’s Rule—the common tonal rule that deletes a H tone that immediately follows another H tone (Goldsmith 1984a, b). The tonal properties of Khayo are most closely related to its southwestern Luyia neighbors Nyala-West (Ebarb et al. in prep, Marlo 2007, Onyango 2004, 2005, 2006), Saamia (Botne et al. 2006, Poletto 1998), Songa (Marlo Notes), and Tura (Marlo to appear, a). These dialects have two related tonal properties in common: (i) they are all “predictable” tone systems that have lost the historical lexical contrast between two tonal classes of verb roots (Odden 1989), and (ii) all verbal forms are inflected with a grammatical tone. As described in Marlo (2008), property (i) is unique to these Luyia dialects; property (ii) is shared with the central and northern Luyia dialects Bukusu (Austen 1974, Mutonyi 2000), Marachi (Marlo 2007), and Wanga (Marlo Notes); the eastern Luyia dialects Idakho (Marlo Notes), Logoori (Leung 1991), Tachoni (Odden 2009), and Tiriki (Kim & Paster 2007) do not have either of these properties. Like many other Bantu languages (Hyman 2001), Khayo has a privative tonal contrast between H and Ø (“toneless”). The tone-bearing unit is the mora (µ), and the following tonal contrasts are found on long and short vowels: vs. µµ vs. vs. vs. vs. µ. Thus, level H and toneless are found on short and long vowels, but falling and rising contour tones are found on long vowels only.

*

1

I would like to thank Alfred Anangwe and Moses Egesa for logistical support in the field, Meshack Kisa Nderema for providing the data, Lana Pierce and Ruth Rodriguez for their assistance in creating a database of the audio-recorded data, and Kris Ebarb, Philip Mutaka, David Odden, Sabine Zerbian, and the audience at the 2009 Mid America Linguistics Conference for discussion of the analysis. Research support was provided by a National Science Foundation Graduate Research Fellowship and National Science Foundation Doctoral Research Improvement Grant 0545246. The usual disclaimers apply. These data were elicited from a single male speaker of Khayo from Busia, Kenya, who was approximately 40 years old at the time.

Another property of Khayo shared with other Bantu languages (see, e.g., Kisseberth & Odden 2003, Marlo 2009, Myers 1998, and references therein) is that the verbal morphology plays an important role in the realization of tone in the language. A templatic view of Khayo verbal structure is provided in (1). The verb stem, whose boundaries are marked here by square brackets, is composed of the verb root plus other suffixes, up to a final marker of mood known as the “final vowel” (FV). The verb macrostem, whose boundaries are marked here by curly brackets when relevant, includes the stem and maximally two object prefixes preceding the stem. In constructions with two object prefixes, the second object prefix must be 1sg -.2 (1)

Simplified templatic structure of the Khayo verb H Ø H/Ø H H Ø Ø Ø Neg - Subj - Tns - { Obj1 - Obj2 [ Root - Deriv - FV ]stem }macrostem

The tonal properties of these morphemes are also indicated in (1). The negative prefix s - is H.3 Subject prefixes are Ø in all known contexts. Tense prefixes exhibit an underlying contrast between H and Ø. All object prefixes are underlyingly H, e.g., 3sg m -, 1sg -, except the toneless reflexive e-. As noted above, verb roots are underlyingly Ø, as are other morphemes of the stem.4 While the moras of the stem are underlyingly toneless, tonal melodies marking tense-aspect-mood-polarity and clause type (matrix vs. relative) distinctions result a complex array of surface tonal patterns on the stems. Following most work on Bantu tonal melodies within the autosegmental period (e.g., Bickmore 1997, 2000, 2007, Cassimjee 1986, Cheng & Kisseberth 1979, 1980, 1981, Clements 1984, Downing 1990, Hyman & Ngunga 1994, Kidima 1991, Liphola 2001, Marlo to appear, a, Mwita 2008, Odden 1981, 1987, 1996, 1998, 2009, Roberts-Kohno 2000), the inflectional tonal melodies are analyzed here as underlyingly floating H tones called “melodic Hs” that are assigned by rule to different positions of the verb stem and macrostem. A summary of the Khayo tone melodies is provided in (2). The different melodies are classified into different “Patterns”, which are given numeric and alphabetic labels.

2

3

4

Single-underlining indicates the position of an underlying H; double-underlining indicates the surface position of a melodic H. Negative s - is used in matrix clauses. In certain tenses, such as the Crastinal Future Negative and the Imperative Negative, and in relative clauses, the negative prefix -ra- is found in a position between the tense prefix slot and the macrostem (Marlo to appear, b). The causative and passive suffixes of the nearby Nyala-West and Marachi dialects are /H/ (Ebarb et al. in prep, Marlo 2007), a relatively undiscussed property in Bantu tonal systems that is also found in Ganda (Hyman & Katamba 1990). Khayo and Tura do not have this property (Marlo to appear, a, b).

(2)

Tonal melodies in Khayo Pattern 1a Present Pattern 1b Immediate Past Pattern 1c Near Future

H from σ2 to final a[βal

l- ]

y-a-x [lom l

- ]

a-n [fuun - ]

Pattern 2a Present Negative Pattern 2b Immediate Past Neg.

H on µ1 of [σ2]

Pattern 3a ImperativeSg

H on final

Pattern 3b Hodiernal Perfective

H on µ1 and final

Pattern 4a Crastinal Future

H on µ1 of {σ2}

Pattern 4b Remote Past

H on µ1

Pattern 5 Conditional

all toneless

sy -a[lom lom-a]

‘he is setting out to dry’ ‘he just talked’ ‘he will cover’ ‘he is not talking’

s -y-a-x [lom lom-a] ‘he did not just talk’ [lomalom- ] a[x laang-ir ]

n-aa{[lom lom-e]} y- [xalaang-a] a[βodoo an-a]

‘talk!’ ‘he fried’ ‘he will talk’ ‘he fried’ ‘if he goes around’

The primary basis for the classification of the melodic Patterns is the position of the melodic H on the verb. Pattern 1 melodies have a H from the second syllable of the stem to the final. Pattern 2 melodies have a H on the first mora of the second syllable of the stem. Pattern 3 melodies have a H on the stem-final mora, and Pattern 3b also has a H on the stem-initial mora, a property it shares with Pattern 4b (though the Pattern 4b melodic H is deleted following the H of the Htoned tense prefix -). The Pattern 4a melodic H is realized on the first mora of the second syllable of the macrostem, evidence for which will be provided in Section 4.3. The Pattern 5 melody is characterized by no Hs on the verb; the stem and all prefixes (even underlyingly H prefixes) surface toneless. In the sections that follow, we explore some aspects of the realization of these tonal melodies. In particular, we investigate tonal alternations that occur as a result of the presence of H-toned tense prefixes and object prefixes and differences in the prosody of the verb stem. In Section 2, we examine the Pattern 1 tonal melody and show that there is an unpredictable difference between the

Pattern 1a and Pattern 1b tenses, on the one hand, and the Pattern 1c tenses, on the other. All the Pattern 1 tenses that have the same general tonal melody, but the Pattern 1a and Pattern 1b tenses are subject to an additional tonal rule that the Pattern 1c tenses are not. In Section 3, we look to Pattern 3b and Pattern Pattern 4b to explore the inner workings of Meeussen’s Rule, which deletes H tones after H. We see that there is an important difference between sequences of H-toned prefixes and sequences of H tones terminated by a melodic H. Section 4 explores the realization of other tonal melodies (Pattern 2a, Pattern 2b, and Pattern 4a), which provide further examples of tense-specific tonology and further evidence for the analysis of Meeussen’s Rule. Section 5 provides concluding remarks.

2. Tense-specific rules: Pattern 1 There are three sub-types of Pattern 1 tenses, classified as Pattern 1a, Pattern 1b, and Pattern 1c. Pattern 1a tenses have a toneless tense prefix. Pattern 1b and Pattern 1c have H-toned tense prefixes. All three tenses have the same basic rules of Melodic H Assignment and Unbounded Spread, which result in a melodic H from the second syllable of the stem to the final, but Pattern 1a and Pattern 1b have an additional spreading rule affecting disyllabic stems that Pattern 1c does not. 2.1 Pattern 1a: Present

Pattern 1a forms from the Present tense are provided in (3). The general Pattern 1a tonal properties are demonstrated in trisyllabic and longer stems whose peninitial syllable is short. In these forms, the melodic H surfaces from the second syllable of the stem to the final; note that the melodic H does not surface on the stem-initial syllable. In trisyllabic and longer stems with a long peninitial syllable, the beginning of the H tone span begins on the second mora of the second stem syllable. In disyllabic stems, the melodic H surfaces on the final two moras of the stem, producing a rising tone on the stem-initial syllable of CVVCV stems and a level H on the stem-initial syllable of CVCV stems. In monosyllabic stems, the melodic H is realized on the lone stem syllable. (3)

Present CV CVCV CVVCV CV(V)CVCV+

a[ly- ] a[β - ] a[re β- ] a[βu l- ] a[siind - ] a[βal l- ]

‘he is eating’ ‘he is shaving’ ‘he is asking’ ‘he is taking’ ‘he is pushing’ ‘he is setting out to dry’

CV(V)CVVCV+ a[xala n - ]

‘he is frying’ ‘he is watching’ ‘he is going around’

a[liinga l- ] a[βodo n- ]

Pattern 1a forms always realize the melodic H on the stem-final mora. This is captured by a rule of Melodic H Assignment in (4), which assigns the melodic H to the stem-final mora.5 In disyllabic stems, the melodic H extends one mora to the left. This is expressed by the rule of Minimal Spread, which spreads the melodic H one mora to the left within the stem. A second spreading rule, Unbounded Spread, spreads the melodic H left across the stem, but not into the stem-initial syllable. The fact that the H tone span begins on the second mora of the second syllable of trisyllabic and longer stems with a long peninitial syllable is captured by the rule of Rightward Delay, which creates a rising tone when a H is linked across multiple syllables and begins on a long syllable. (4)

MHA - Final H’ µ ]stem

(applies only in Pattern 1)

Unbounded Spread H (applies only in Pattern 1, iterative, σ µ µ domain: stem)

Minimal Spread H (applies only in Pattern 1a-b, domain: µ2 and µ3 µ’ 1 µ2 µ3 are in stem) Rightward Delay H x µ µ µ σ

Derivations for disyllabic and longer stems are provided in (5) – (8), showing the effects of Minimal Spread in disyllabic stems, Unbounded Spread in longer stems, and Rightward Delay in long stems with a long peninitial syllable. Minimal Spread does not apply in monosyllabic stems because the target mora must be within the stem. (5)

a[β - ] ‘he is shaving’

MHA - Final H a[βe -a]

5

Minimal Spread H  a[βe -a]

Unbounded Spread  Does Not Apply

In these rules and others below, floating tones (tones not linked to a mora) and toneless moras (moras not linked to a tone) are indicated by an apostrophe, as appropriate. Floating melodic Hs are set off in derivations with a circle around the H.

(6)

a[re β- ] ‘he is asking’

MHA - Final H a[reeβ-a]

(7)

(8)

a[βal

Minimal Spread H  a[reeβ-a]

 Does Not Apply

l- ] ‘he is setting out to dry’

MHA - Final H

Minimal Spread H

a[βala al-a] 

a[βala al-a]

a[βodo

Unbounded Spread

Unbounded Spread H  a[βala al-a]

- ] ‘he is going around’

MHA - Final H

Minimal Spread H

a[βodoo an-a] 

a[βodoo an-a]

Unbounded Spread H  a[βodoo an-a]



Rightward Delay H x a[βodoo an-a]

In combination with a H-toned OP, the Pattern 1 melodic H occupies the same position on the stem, except in disyllabic stems. In CVCV stems, the melodic H surfaces on the final vowel but not on the stem-initial syllable, cf. a[β - ]. In CVVCV stems, the melodic H is realized as a level H from the stem-initial syllable to the final vowel, cf. a[re β- ]. (9)

Present + OP (/m -/) CV a-m [ry- ] CVCV a-m [βek- ] CVVCV a-m [r β- ] CVCVCV+ a-m [βu l- ]

a-m [siind - ] a-m [liinga l- ] a-m [xala n - r- ]

‘he fears him’ ‘he is shaving him’ ‘he is asking him’ ‘he is taking him’ ‘he is pushing him’ ‘he is watching him’ ‘he is frying for him’

The fact that the melodic H does not surface on the stem-initial syllable in CVCV stems with an object prefix is accounted for by the formulation of Minimal Spread, which requires a toneless mora (µ’) before the target mora. Minimal Spread does not apply in a-m [βe - ] because the H-toned prefix m precedes the stem-initial mora. In CVVCV stems, Minimal Spread applies because the target mora is preceded by a toneless mora. An additional rule of

Plateau, which levels out a rising tone to H after H, applies to the intermediate representation created by Minimal Spread. The formulation of Plateau is given in (10), and the derivation in (11) shows its application, following Minimal Spread. CVCV stems with an object prefix do not show effects of Plateau because the ØH sequence must be within a syllable, not across syllables. (10) Plateau H H µ µ µ σ

σ

(11) a-m [r β- ] ‘he is asking him’ MHA - Final Minimal Spread H H H H a- u[reeβ-a] 

a- u[reeβ-a]

Plateau H

H

 a- u[reeβ-a]

The data in (12) show that the 1sg object prefix - is tonally identical to the 3sg object prefix m -: it is H-toned and has the same effects on the stem tone patterns. The only difference is that the nasal lengthens the preceding vowel, and the tone of - is realized on the immediately preceding mora.6 Since the immediately preceding mora is the second half of a long vowel, the ØH sequence is realized as a rising tone. -/) [ry- ] [mbek- ] [r β- ] [mbuk l- ] a- [ndeex- r- ] a- [niinga l- ] a- [xala n - r- ]

(12) Present + OP1sg (/ CV aCVCV aCVVCV aCVCVCV+ a-

‘he fears me’ ‘he is shaving me’ ‘he is asking me’ ‘he is taking me’ ‘he is cooking for me’ ‘he is watching me’ ‘he is frying for me’

Two object prefixes are possible in Khayo when a CV- object prefix like 3sg m - combines with the 1sg - since the combination fits into a single syllable (see Marlo 2009). As we see in (13), 1sg - triggers lengthening of the 6

The nasal prefix is deleted before voiceless fricatives, nasals, and (variably) /r/. Other NC effects in Khayo—common within Bantu (see Hyman 2003)—include the hardening of /β, l/ and (variably) /r/, the voicing of voiceless stops and affricates, and Meinhof’s Law, which renders /N+lVN(C)/ into [nVN(C)]. Meinhof’s Law effects are also found in Vinitial stems: the nasal prefix is realized as ny- ([ -]) before VN(C) stems and as - ([ -]) before other V-initial stems.

preceding mora, and its H is deleted following the H of the first OP, so /m - -/ is realized as [m -u-]. Deletion is accomplished via the common cross-Bantu rule, Meeussen’s Rule (Goldsmith 1984a, b), formalized in (14), which deletes H after H. As discussed further in Section 2.2, Meeussen’s Rule applies iteratively from right to left. It also applies early in the derivation, so Plateau does not apply after Minimal Spread in the CVVCV stem a-m -u[r-i - ]. (13) Present + OPx2 (/m - -/) a-m a-m a-m a-m a-m

-u[r-i r- ] -u[mbon- r- ] -u[reeβ- r- ] -u[mbuk l- r- ] -u[nyandi - r- ]

‘he fears him for me’ ‘he is seeing him for me’ ‘he is asking him for me’ ‘he is taking him for me’ ‘he is employing him for me’

(14) Meeussen’s Rule I HHØ (Iterative, R-to-L, µ µ domain: word) 2.2 Pattern 1b: Immediate Past

The Pattern 1b Immediate Past has a H-toned tense prefix x - and the same rules of Melodic H Assignment and spreading as Pattern 1a. As shown by the data in (15), Pattern 1b forms lacking an object prefix are tonally equivalent to Pattern 1a forms with an object prefix. We thus find the blocking of Minimal Spread in CVCV stems and the triggering of Plateau in CVVCV stems. (15) Immediate Past (/x -/) CV y-a-x [fw- ] CVCV y-a-x [βek- ] CVVCV y-a-x [r β- ] CVCVCV+ y-a-x [βu l- ]

y-a-x [fuund - ] y-a-x [lexu l- ] y-a-x [lom l - ]

‘he just died’ ‘he just shaved’ ‘he just asked’ ‘he just took’ ‘he just knotted’ ‘he just released’ ‘he just talked’

When Pattern 1b forms have an object prefix, the H of the object prefix is deleted after the H of the tense prefix, and the usual Pattern 1a stem tone patterns apply. In particular, note that Minimal Spread applies in disyllabic stems, spreading the melodic H into the stem-initial syllable.

(16) Immediate Past + OP (/x -m -/) CV y-a-x -mu[ry- ] CVCV y-a-x -mu[β k- ] CVVCV y-a-x -mu[re β- ] CVCVCV+ y-a-x -mu[βu l- ]

‘he just feared him’ ‘he just shaved him’ ‘he just asked him’ ‘he just took him’ y-a-x -mu[deex- r- ] ‘he just cooked for him’ y-a-x -mu[liinga l- ] ‘he just watched for him’ y-a- -mu[xala n - r- ] ‘he just fried for him’

The application of Meeussen’s Rule removes the H tone immediately preceding the stem that blocks Minimal Spread in CVCV stems and that creates the environment for Plateau to apply in CVVCV stems. Thus, as shown in the derivations in (17) – (18), the melodic H surfaces on the same positions of the stem as in Pattern 1a forms that lack an object prefix. A final step in these derivations, which does not affect the tonal outputs, is that the underlying sequence of the subject marker a- plus the tense prefix a- becomes y-a-; the various processes in Khayo that resolve underlying vowel hiatus are labeled “Resyllabification” in the derivations here. (17) y-a-x -mu[β - ] ‘he just took him’ Meeussen’s Rule I MHA - Final H HØ H H H a-a-xa-mu [βe -a] 

a-a-xa- u[βe -a] 

Unbounded Spread

Resyllabification H H

Does Not Apply

Minimal Spread H H a-a-xa- u[βe -a]



 y-a-xa- u[βe -a]

(18) y-a-x -mu[re β- ] ‘he just took him’ Meeussen’s Rule I MHA - Final H HØ H H H

Minimal Spread H H

a-a-xa-mu [reeβ-a] 

a-a-xa- u[reeβ-a]  a-a-xa- u[reeβ-a] 

Unbounded Spread

Resyllabification H H

Does Not Apply

Plateau

 y-a-xa- u[reeβ-a]  Does Not Apply

In Pattern 1b forms with two object prefixes, shown in (19), the Hs of both OPs are deleted, so that only the leftmost H in the underlying H-H-H sequence

surfaces. This is the result of the iterative, right-to-left application of Meeussen’s Rule, a topic that is discussed further in Section 3. (19) Immediate Past + OPx2 (/x -m - -/) y-a-x -mu-u[mbon- r- ] ‘he just saw him for me’ y-a-x -mu-u[ndeeβ- r- ] ‘he just asked him for me’ y-a-x -mu-u[mbuk l- r- ] ‘he just took him for me’ y-a-x -mu-u[nyandi - r- ] ‘he just employed him for me’ 2.3 Pattern 1c: Near Future

The Pattern 1b and Pattern 1c tenses both have H-toned tense prefixes, e.g., Pattern 1b Immediate Past x - and Pattern 1c Near Future n -, and both have Melodic H Assignment to the final vowel plus leftward Unbounded Spread. Unlike Patterns 1a and 1b, however, Pattern 1c tenses are not subject to Minimal Spread. As we see in the examples in (20) and (21), disyllabic stems do not show spreading into the stem-initial syllable. This is true even when the mora immediately preceding the stem is toneless, as is the case in forms with an object prefix, whose H is deleted by Meeussen’s Rule. (20) Near Future (/n -/) CV a-n CVCV a-n CVVCV a-n CVCVCV+ a-n

[ly- ] [βek- ] [reeβ- ] [βu l- ] a-n [fuun - ] a-n [lexu l- ] a-n [liinga l- ]

(21) Near Future + OP (/n -m -/) CV a-n -mu[ry- ] CVCV a-n -mu[βek- ] CVVCV a-n -mu[reeβ- ] CVCVCV+ a-n -mu[βu l- ] a-n a-n a-n a-n

-mu[deex- r- ] -mu[lexu l- ] -mu[liinga l- ] -mu[xala n - r- ]

‘he will eat’ ‘he will shave’ ‘he will ask’ ‘he will take’ ‘he will cover’ ‘he will release’ ‘he will watch’ ‘he will fear him’ ‘he will shave him’ ‘he will ask him’ ‘he will take him’ ‘he will cook for him’ ‘he will release him’ ‘he will watch him’ ‘he will fry for him’

The analysis of the Pattern 1c forms is simply that they lack Minimal Spread; thus there is no impetus to spread into the stem-initial syllable in disyllabic stems. It is not possible to predict on phonological grounds the tenses in which Minimal Spread applies. That Minimal Spread applies in Pattern 1a and Pattern

1b—and not Pattern 1c or other Patterns—must be stipulated as part of the context of the rule. 3. H-H-H sequences: two versions of Meeussen’s Rule We saw in Section 2.2 that sequences of two or three H-toned prefixes in a row are simplified such that only the leftmost H surfaces: H-H-H  H-Ø-Ø. This has been analyzed as the result of iterative, right-to-left application of Meeussen’s Rule. In this section, we explore cases where the melodic H is subject to deletion following H. As with H-toned prefixes, the melodic H is systematically deleted when it immediately follows a single H-toned prefix. However, in sequences of three or four Hs where the final H is the melodic H, the output is different from a series of H-toned prefixes. In these H-H-H and H-H-H-H sequences, the leftmost H-toned prefix and the melodic H (the rightmost H) remain, i.e., H-Ø-H and H-Ø-Ø-H. These patterns are analyzed with two versions of Meeussen’s Rule. First is the early rule, Meeussen’s Rule I, which we have already seen. Meeussen’s Rule I applies iteratively from right to left, deleting all but the leftmost H-toned prefix. Meeussen’s Rule I precedes all rules of Melodic H Assignment. The later rule, Meeussen’s Rule II, applies after all rules of Melodic H Assignment (except the Pattern 1 rule) and thus deletes any melodic H (except the Pattern 1 melodic H) that immediately follows a H-toned prefix.7 In the remainder of this section we see how this plays out in Pattern 3b and Pattern 4b—melodies that assign the melodic H to the stem-initial mora. 3.1 Pattern 3b

Pattern 3b has two melodic Hs: one assigned to the stem-final vowel and one assigned to the stem-initial mora. This is shown by the data in (22). (22) Pattern 3b Hodiernal Perfective a[l- ir ] ‘he ate’ a[r eβ-er ] ‘he asked’ a[x laang-ir ] ‘he fried’

a[β k-er ] a[β kuul- ] a[f uniix- ]

‘he shaved’ ‘he took’ ‘he covered’

The rule assigning H to the stem-initial syllable is given in (23). This rule applies in Pattern 3b and Pattern 4b. We see in Section 3.2 that this rule does not apply in monosyllabic stems, and thus requires a syllable following the stem7

The Pattern 1 melodic H does not delete after a H-toned tense prefix in monosyllabic stems: Pattern 1a a-m [ry- ] ‘he fears him’, Pattern 1b y-a-x [fw- ] ‘he just died’, Pattern 1c a-n [ly- ] ‘he will eat’. Marlo (to appear, b) accounts for this by ordering the Pattern 1 rule of Melodic H Assignment after Meeussen’s Rule II.

initial mora. (Monosyllabic stems are not possible in Hodiernal Perfective forms as the perfective suffix -ire creates minimally disyllabic stems.) (23) MHA - µ1 H’ stem[

µ σ

(applies only in Pattern 3b and Pattern 4b, domain: stem)

When Hodiernal Perfective forms are preceded by a H-toned object prefix, the melodic H on the stem-initial mora does not surface. (24) Hodiernal Perfective + OP (/m -/) a-m [r-iir ] ‘he feared him’ a-m [βe -er ] ‘he shaved him’ a-m [reeβ-er ] ‘he asked him’ a-m [βu uul- ] ‘he took him’ a-m [fuuniix- ] ‘he covered him’ a-m [lexuul-ir ] ‘he released him’ a-m [liingaal-ir ] ‘he watched him’ Deletion of the melodic H is accomplished by the second, later version of Meeussen’s Rule, Meeussen’s Rule II, shown in (25).8 (25) Meeussen’s Rule II HHØ µ µ

(domain: word)

The derivation in (26) shows that Meeussen’s Rule I is ordered before the rules of Melodic H Assignment. The melodic H on the stem-initial mora is deleted by Meeussen’s Rule II following the H of the object prefix. (26) a-m [βe -er ] ‘he shaved him’ Input Meeussen’s Rule I MHA - Final H H H H H H a- u[βe -ere]



MHA - µ1 H H H a- u[βe -ere]

8

Does Not Apply  a- u[βe -ere]



Meeussen’s Rule II H HØ H 

a- u[βe

-ere]

It is not possible to determine if Meeussen’s Rule II is identical to Meeussen’s Rule I in all ways except its position in the derivation since there are no contexts in which iterativity or directionality of Meeussen’s Rule II are at stake.

With two H-toned OPs, it becomes clear that it is not the early iterative Meeussen’s Rule I that deletes the melodic H in the Hodiernal Perfective, since in these forms, the melodic H surfaces on the stem-initial mora. (27) Hodiernal Perfective + OPx2 (/m - -/) a-m -u[ny mb-iir- ] ‘he sang to him for me’ a-m -u[r eβ-eer- ] ‘he asked him for me’ a-m -u[mb kul-iir- ] ‘he took him for me’ a-m -u[ny ndiik-iir- ] ‘he employed him for me’ As shown in the derivation in (28), Meeussen’s Rule I first deletes the H of the second OP. The melodic H is then assigned to the stem and is not deleted by Meeussen’s Rule II since the preceding mora is toneless, having first been deleted by Meeussen’s Rule I. (28) a-m -u[r eβ-ee - ‘he asked him for me’ Meeussen’s Rule I MHA - Final HHØ H H H H H

MHA - µ1 H H

H

a-mu-N[reeβ-eer-e]  a-mu-N[reeβ-eer-e]  a-mu-N[reeβ-eer-e]

Meeussen’s Rule II

Does Not Apply

NC Effects H H



H

 a-mu-u[reeβ-eer-e]

3.2 Pattern 4b

Another context in which a melodic H targets the stem-initial mora and shows effects of Meeussen’s Rule II is Pattern 4b. As shown by the examples in (29), the Pattern 4b Remote Past has a H-toned tense prefix -, which immediately precedes the stem. The H of the tense prefix causes the deletion of the melodic H from the stem-initial mora. (29) Remote Past (/ -/) y- [fw-a] ‘he died’ y- [reeβ-a] ‘he asked’ y- [xalaang-a] ‘he fried’ y- [saangaal-a] ‘he was happy’

y- [βe -a] y- [βu ul-a] y- [fuunix-a]

‘he shaved’ ‘he took’ ‘he covered’

In combination with an object prefix, the melodic H is realized on the steminitial mora in disyllabic and longer stems.

(30) Remote Past + OP (/ -m -/) CV - -mu[ry-a] CVCV+ y- -mu[β k-a] yy-

-mu[r eβ-a] -mu[f unix-a] -mu[l xuul-a] -mu[l ingaal-a]

‘he feared him’ ‘he shaved him’ ‘he asked him’ ‘he covered him’ ‘he released him’ ‘he watched him’

These data show the combined effects of Meeussen’s Rule I, Melodic H Assignment to the stem-initial mora, and Meeussen’s Rule II. As illustrated in the derivation in (31), Meeussen’s Rule I does not apply in forms lacking an object prefix because the floating melodic H is not adjacent to the H-toned prefix. After Melodic H Assignment, the melodic H on the stem-initial mora immediately follows the H-toned tense prefix and is deleted by Meeussen’s Rule II. (31) y- [ eeβ-a] ‘he asked’ Input H H a-a[reeβ-a]

Meeussen’s Rule II H HØ a-a[reeβ-a]

Meeussen’s Rule I

 Does Not Apply

MHA - µ1 H H  a-a[reeβ-a]



Resyllabification H  y-a[reeβ-a]

The derivation in (32) shows that in forms with an object prefix, Meeussen’s Rule I first deletes the H of the object prefix following the H of the tense prefix. The melodic H, which is assigned to the stem-initial mora, is not subject to Meeussen’s Rule II, since the preceding mora is toneless. (32) y- -mu[r eβ-a] ‘he asked him’ Meeussen’s Rule I MHA - µ1 H HØ H H H a-a-mu [reeβ-a]

Resyllabification H H y-a- u[reeβ-a]

Meeussen’s Rule II

 a-a- u[reeβ-a]  Does Not Apply



As shown by the data in (33) with two object prefixes, only the first and last underlying Hs surface: the leftmost H-toned prefix, the tense prefix -, and the melodic H on the stem-initial mora. (33) Remote Past + OPx2 (/ -m - -/) y- -mu-u[r- ir-a] ‘he feared him for me’ y- -mu-u[mb n-er-a] ‘he saw him for me’ y- -mu-u[r eβ-er-a] ‘he asked him for me’ y- -mu-u[mb kul-ir-a] ‘he took him for me’ y- -mu-u[ny ndiik-ir-a] ‘he employed him for me’ The derivation in (34) shows that the Hs of both object prefixes are first deleted by the iterative, right-to-left application of Meeussen’s Rule I, which allows the melodic H to be assigned to the stem-initial mora without subsequently being deleted by Meeussen’s Rule II. (34) y- -mu-u[mb n-er-a] ‘he saw him for me’ Meeussen’s Rule I Meeussen’s Rule I H HHØ H H HØ H

MHA - µ1 H H

a-a-mu-N[βon-er-a]  a-a-mu-N[βon-er-a]  a-a-mu-N[βon-er-a] 

Meeussen’s Rule II

Does Not Apply

NC Effects & Resyllabification H H  y-a-mu-u[mbon-er-a]

4. Other melodies In this section we provide further examples of tense-specific tonal rules and further evidence for the analysis with two versions of Meeussen’s Rule. 4.1 Pattern 2a

Pattern 2 tenses are characterized by a melodic H that surfaces on the first mora of the second syllable of trisyllabic and longer stems. In disyllabic stems, the melodic H surfaces on the penultimate mora, i.e., on the initial mora of CVCV stems, and as a rise on the stem-initial syllable of CVVCV stems. The melodic H does not surface in monosyllabic stems.

(35) Present Negative CV sy CVCV sy CVVCV sy CVCVCV+ sy sy sy sy sy

-a[ly-a] -a[β k-a] -a[re β-a] -a[βu l-a] -a[siind x-a] -a[xal ang-a] -a[liing al-a] -a[lom lom-a]

‘he is not eating’ ‘he is not shaving’ ‘he is not asking’ ‘he is not taking’ ‘he is not pushing’ ‘he is not frying’ ‘he is not watching’ ‘he is not talking’

These patterns are captured by the two rules in (36)—Melodic H Assignment, which assigns the melodic H to the first mora of the second syllable of the stem, and Throwback, which shifts the melodic H from the final mora to the penult in disyllabic stems.9 The application of Pattern 2 Melodic H Assignment and Throwback is illustrated in the derivation in (37). (36) MHA – [σ2] H’

Throwback H x µ µ]stem (domain: stem)

µ ( µ) stem[

σ σ

(applies only in Pattern 2)

(37) sy -a[re β-a] ‘he is not asking’ MHA - [σ2] Throwback Resyllabification H H H H H H x si-a[reeβ-a]  si-a[reeβ-a]  sy-aa[reeβ-a] In combination with an object prefix, the melodic H occupies the expected position on the stem, with the exception of disyllabic stems. (38) Present Negative + OP (/m -/) CV s -βa-m [xw-a] CVCV s -βa-m [βek-a] CVVCV s -βa-m [r β-a] CVCVCV+ sy -a-m [βu l-a] s -βa-m s -βa-m s -βa-m sy -a-m

9

‘they are not paying her dowry’ ‘they are not shaving him’ ‘they are not asking him’ ‘he is not taking him’ [siind x-a] ‘they are not pushing him’ [liing al-a] ‘they are not watching him’ [xal ang-ir-a] ‘they are not frying for him’ [lom lom-er-a] ‘they are not talking for him’

Throwback applies in Pattern 2 and Pattern 4a but not in Pattern 1 or Pattern 3. Marlo (to appear, b) captures this fact through rule-ordering: the Pattern 2 and Pattern 4a MHA rules precede Throwback; the Pattern 1 and Pattern 3 MHA rules follow Throwback.

The melodic H is deleted in CVCV stems due to Meeussen’s Rule II applying after the melodic H undergoes Throwback to the penultimate mora. (39) s -βa-m [βe -a] ‘they are not shaving him’ MHA - [σ2] Throwback Meeussen’s Rule II H H H H H H H H HØ x si-βa- u[βe -a]  si-βa- u[βe -a]  si-βa- u[βe -a] The melodic H undergoes Plateau in CVVCV stems after having undergone Throwback to the penultimate mora. (40) s -βa-m [r β-a] ‘they are not asking him’ MHA - [σ2] Throwback Meeussen’s Rule II H H H H H H x si-βa- u[reeβ-a]  si-βa- u[reeβ-a]  Does Not Apply  Plateau H H

H

si-βa- u[reeβ-a]

The Pattern 2 data show that the melodic H is deleted after a H-toned prefix in Khayo not only from the position to which the melodic H is assigned, as in Pattern 4b, but also from a position that the melodic H reaches via a second rule (Throwback) after Melodic H Assignment. This is important because it suggests a uniform analysis of the non-surfacing of the melodic H via Meeussen’s Rule, and not, e.g., the result of restrictions on the various rules of Melodic H Assignment such that they require a toneless mora before the target mora. An alternative to positing two versions of Meeussen’s Rule is to assume that there is only a single late version of Meeussen’s Rule, modifying our assumptions that each object prefix contributes a H tone that is underlyingly linked. The effects of Meeussen’s Rule I could be accounted for without an additional deletion rule by claiming that the Hs of object prefixes are underlyingly floating and assigned to the macrostem-initial mora by a rule that requires the preceding mora to be toneless. An object prefix would therefore not realize a H following a H-toned prefix because of the restriction on object prefix H assignment. To account for the data with multiple object prefixes—in which the H of the rightmost object prefix systematically does not surface—we would have to make

some further assumptions. One possibility is that the combination of the object prefixes contributes only a single underlying H. A second possibility is that there is only one macrostem constituent beginning at the left edge of the leftmost object prefix, irrespective of the number of object prefixes, so at most one underlying object prefix H can be assigned to the macrostem. The problem with these alternatives is that they preclude the possibility that each of the object prefixes could contribute a surface H, yet the closely related Luyia dialect Nyala-West has contexts in which each of the Hs of two object prefixes surfaces (Ebarb et al. in prep, Marlo 2007).10 4.2 Pattern 2b

The Pattern 2b Immediate Past Negative has a H-toned tense prefix x -, and thus is tonally identical to Patterna 2a Present Negative forms with an object prefix. (41) Immediate Past Negative (/x -/) CV s -y-a-x [fw-a] CVCV s -y-a-x [βek-a] CVVCV s -y-a-x [r β-a] CVCVCV+ s -y-a-x [βu l-a] s s s s

-y-a-x -y-a-x -y-a-x -y-a-x

[fuun l-a] [xal ang-a] [liing al-a] [lom lom-a]

‘he did not just die’ ‘he did not just shave’ ‘he did not just ask’ ‘he did not just take’ ‘he did not just uncover’ ‘he did not just fry’ ‘he did not just watch’ ‘he did not just talk’

In combination with an object prefix, the H of the object prefix is deleted by Meeussen’s Rule I. As we saw above, the early application of Meeussen’s Rule I allows the melodic H to occupy the normal stem positions—on the penultimate mora of disyllabic stems and on the first mora of the second syllable of the stem in longer stems. A sample derivation for CVCV stems is provided in (43). (42) Immediate Past Negative + OP (/x -m -/) CV s -y-a-x -mu[ry-a] CVCV s -y-a-x -mu[β k-a] CVVCV s -y-a-x -mu[re β-a]

10

‘he did not just fear him’ ‘he did not just shave him’ ‘he did not just ask him’

The second approach might be salvageable given the possibility that the representation of the macrostem is subject to cross-linguistic variation such that Khayo has only one macrostem that includes all object prefixes, while Nyala-West has recursive macrostems—one macrostem per object prefix (see Hyman & Ngunga 1994, Marlo 2009). Thus, at most one object prefix H could surface in Khayo, but each object prefix H could potentially surface in Nyala-West. The full implications of this type of analysis are not known. Note that under this alternative, the surface tonal properties of the prefixes are established before the melodic H is assigned, as in the adopted analysis.

CVCVCV+

s s s s

-y-a-x -y-a-x -y-a-x -y-a-x

-mu[βu l-a] -mu[fuun l-a] -mu[liing al-a] -mu[xal ang-ir-a]

‘he did not just take him’ ‘he did not just uncover him’ ‘he did not just watch him’ ‘he did not just fry for him’

(43) s -y-a-x -mu[β k-a] ‘he did not just shave him’ Meeussen’s Rule I MHA - [σ2] H H HØ H H H H si-a-xa-mu [βe -a]  si-a-xa- u[βe -a]

Meeussen’s Rule II

Does Not Apply

Throwback H H

H x

 si-a-xa- u[βe -a]



Resyllabification H H H  si-y-a-xa- u[βe -a]

Data with two object prefixes, shown in (44), provide a further example illustrating the iterative, right-to-left application of Meeussen’s Rule I. As we have seen, only the leftmost of the sequence of three underlying H-toned prefixes surfaces, and the melodic H surfaces on the usual position of the stem. (44) Immediate Past Negative + OPx2 (/x -m - -/) s -y-a-x -mu-u[xw-e r-a] ‘he did not just pay her dowry for me’ s -y-a-x -mu-u[mbon- r-a] ‘he did not just see him for me’ s -y-a-x -mu-u[ndeeβ- r-a] ‘he did not just ask him for me’ s -y-a-x -mu-u[mbuk l-ir-a] ‘he did not just take him for me’ s -y-a-x -mu-u[nyand ik-ir-a] ‘he did not just employ him for me’ 4.3 Pattern 4a

This final section, describing the Pattern 4a melody, provides a final example of a tense-specific rule of Melodic H Assignment as well as the two versions of Meeussen’s Rule. Pattern 4a forms without an object prefix are tonally similar to, but not identical to, Pattern 2a forms: the melodic H surfaces on the lone mora of monosyllabic stems, on the penultimate mora of disyllabic stems, and on the first mora of the second syllable of trisyllabic and longer stems. On the surface, these forms differ from Pattern 2a forms only in that the melodic H surfaces on monosyllabic stems, cf. sy -a[ly-a] ‘he is not eating’. Because it is relevant to the analysis of these forms, the macrostem is indicated in curly brackets.

(45) Crastinal Future CV n-aa{[ly- ]} CVCV n-aa{[β k-e]} CVVCV n-aa{[re β-e]} CVCVCV+ n-aa{[βu l-e]}

‘he will eat’ ‘he will shave’ ‘he will ask’ ‘he will take’ n-aa{[xal ang-e]} ‘he will fry’ n-aa{[fuun l-e]} ‘he will uncover’ n-aa{[liing al-e]} ‘he will watch’ n-aa{[lom lom-e]} ‘he will talk’ n-aa{[βod oxan-e]} ‘he will go around’

In combination with an object prefix, a more dramatic difference with the Pattern 2 forms emerges: the melodic H does not surface after the H of the object prefix. (46) Crastinal Future + OP (/m -/) n-aa{m [ry-e]} ‘he will fear him’ n-aa{m [βe -e]} ‘he will shave him’ n-aa{m [reeβ-e]} ‘he will ask him’ n-aa{m [βu ul-e]} ‘he will take him’ n-aa{m [siindix-e]} ‘he will push him’ n-aa{m [lexuul-e]} ‘he will release him’ n-aa{m [xalaang-ir-e]} ‘he will fry for him’ n-aa{m [lomalom-er-e]} ‘he will talk for him’ In combination with two object prefixes, the first object prefix retains its H, and the melodic H surfaces on the stem-initial mora. (47) Crastinal Future + OPx2 (/m - -/) n-aa{m -u[xw- er-e]} ‘he will pay her dowry for me’ n-aa{m -u[mb n-er-e]} ‘he will see him for me’ n-aa{m -u[nd eβ-er-e]} ‘he will bring him for me’ n-aa{m -u[mb kul-ir-e]} ‘he will take him for me’ n-aa{m -u[ny ndiik-ir-e]} ‘he will employ him for me’ These data can be given a uniform analysis by positing the rule of Melodic H Assignment in (48), which targets the first mora of the second syllable of the macrostem. In verbs without an object prefix, the second syllable of the macrostem is the second syllable of the stem, so the surface tone patterns of disyllabic and longer are the same as in Pattern 2. The fact that monosyllabic stems are assigned a H is accounted for by placing the macrostem-initial syllable in parentheses—a further difference with the Pattern 2 rule of Melodic H Assignment.

(48) MHA - {σ2}

H’ µ ( µ)

macrostem{

(σ) σ

(applies only in Pattern 4a and ImperativeSg + OP)

When there is an object prefix, the second syllable of the macrostem is the steminitial syllable. The melodic H therefore immediately follows the H of the object prefix and is deleted by Meeussen’s Rule II, as shown in (49). (49) n-aa-m [ eeβ-e] ‘he will ask him’ MHA - {σ2} Meeussen’s Rule II H H H HØ ni-a{ u[reeβ-a]} 

ni-a{ u[reeβ-a]}

Resyllabification H  n-aa{ u[reeβ-a]}

When there are two object prefixes, the second syllable of the macrostem is also the stem-initial syllable, since 1sg - does not constitute its own syllable. As shown in (50), Meeussen’s Rule I deletes the H of the second object prefix, which prevents the subsequent application of Meeussen’s Rule II after the melodic H is assigned. (50) n-aa-m -u[mb n-er-e] ‘he will see him for me’ Meeussen’s Rule I MHA - {σ2} HHØ H H H ni-a{mu-N[βon-er-e]} 

ni-a{mu-N[βon-er-e]}

Meeussen’s Rule II

NC Effects & Resyllabification H H

Does Not Apply



 ni-a{mu-u[mbon-er-e]}

5. Summary and discussion The realization of the Khayo tonal melodies is inextricably intertwined with the morpho-syntactic context. There are several distinct tense- and constructionspecific rules of Melodic H Assignment targeting different positions of the stem and macrostem, summarized in (51).

(51) Positions targeted by the melodic H in Khayo a. the final mora of the stem (two rules; Pattern 1, Pattern 3) b. the first mora of the stem (Pattern 3b, Pattern 4b) c. the first mora of the second syllable of the stem (Pattern 2) d. the first mora of the second syllable of the macrostem (Pattern 4a) Additionally, there are other rules with tense-specific effects, including Minimal Spread, which applies in Patterns 1a and 1b but not in Pattern 1c (or other Patterns).11 Other tense-specific effects noted in footnotes 7 and 9 above are due to rule-ordering. Ordering the Pattern 1 rule of Melodic H Assignment after Meeussen’s Rule II prevents the Patten 1 melodic H from being deleted after H. Ordering Throwback after the Pattern 2 and Pattern 4a rules of Melodic H Assignment but before the Pattern 1 and Pattern 3 rules of Melodic H Assignment prevents Throwback from applying in Patterns 1 and 3. There is an important asymmetry among Hs in Khayo in their participation in the common cross-linguistic rule that deletes H after H, Meeussen’s Rule. With the exception of the Pattern 1 melodic H, any H tone deletes after a single H, but there is a difference between sequences of three or more Hs ending in a prefixal H vs. those ending in a melodic H.12 Sequences of H-toned prefixes are simplified to just the leftmost H, while the melodic H is deleted only in two-tone sequences; in three- or four-tone sequences, there is always a sequence of Htoned prefixes that simplifies first, so the melodic H also surfaces. This is handled here with an early version of Meeussen’s Rule I which applies iteratively, deleting all but the leftmost H-toned prefix, followed by most rules of Melodic H Assignment and then Meeussen’s Rule II, which deletes any melodic H (except the Pattern 1 melodic H, which has not yet been assigned to the stem) after H. These patterns may be of general interest because they involve a split between the tonology of the prefixes and the stem. What is surprising is that the split does not involve a typical inside-out derivation in which the stem-level phonology 11

12

Marlo (to appear, b) argues that the rule of Unbounded Spread, which applies in Pattern 1 but no other Patterns, is also a tense-specific rule, whose effects cannot be accounted for through rule-ordering alone. A surface exception to these generalizations is found in Remote Past Negative forms such as - - [l aal-a] ‘he did not watch’, - - - [l aal-a] ‘he did not watch him’, in which the H-toned negative marker - is immediately followed on the surface by the Htoned tense prefix - without any trace of Meeussen’s Rule (Marlo to appear, b: Appendix).These forms derive from underlying sequences of three vowels / -a- -/. Importantly, the toneless 3sg subject prefix intervenes between the H-toned prefixes. Presumably Meeussen’s Rule does not apply because of this intervening toneless mora. The surface forms do not preserve the underlying moraic weight of the input, which is probably the result of a vowel-shortening phenomenon common in Luyia (Dalgish 1986), which is not fully understood in Khayo.

precedes the word-level phonology. As in Khayo, the tonology of prefixes also precedes the stem-level tonology in closely related Luyia dialects such as NyalaWest (Ebarb et al. in prep, Marlo 2007) and other Bantu languages such as Gusii (Bickmore 1997, 1999). Moreover, in those data, there is evidence that the stem is visible at the time the rules affecting prefixal tones apply. A further, related issue these data and their analysis raise is the separation of morphology and phonology. One type of architecture that would have a greater separation of the morphological and phonological aspects of tonal processes than in the present approach is that morphologically specific rules of Melodic H Assignment precede general tone rules.13 The data presented here suggest that this hypothesis is not tenable, as the general rule, Meeussen’s Rule I, must apply to prefixal Hs before the Melodic H is assigned to the stem; otherwise, it would be incorrectly predicted that the melodic H would be deleted after a sequence of two or more H toned prefixes. Indeed, other analyses of Luyia tonal systems such as Ebarb et al. (in prep) and Marlo (to appear, a, b) crucially rely on the assumption that morphologically specific tonal rules can be ordered differently with respect to general tonal rules to account for the diverse tonal patterns of the many inflectional melodies found in the Luyia dialects—in particular, to account for the fact that some tonal rules required for one or more melodies do not apply in some other melodies (see footnotes 7 and 9).14 These kinds of effects call for continued careful investigation of Bantu tonal systems, which potentially lend important insights into phonology-morphology-syntax interactions.

REFERENCES Austen, C. 1974. Anatomy of the tonal system of a Bantu language. Studies in African Linguistics, Supplement 5: 21-33. Bickmore, L. 1997. Problems in constraining High tone spread in Ekegusii Lingua 102: 265-290.

13

14

Cassimjee & Kisseberth’s (1998) Optimal Domains Theory assumes that the position of melodic Hs is already set in the input of the phonology. However, this is not a well developed component of their approach, without explicit arguments for or against. In much contemporary work in phonology, particularly in Optimality Theory, it is often assumed without question that phonological rules/constraints can refer to morphological and/or syntactic information; see Scheer 2009 for a critical review and discussion of de Lacy’s 2007 Cambridge Handbook of Phonology. My work on Luyia tone acknowledges that its approach does not follow some classic views on phonology-morphology interaction and does so, not without question, but because reference to tense-aspect-mood features in the context of tonal rules and to the stem boundary in word-level and phraselevel tonal phonology seems to be required by the data.

Bickmore, L. 1999. High tone spread in Ekegusii revisited: an optimality theoretic account. Lingua 109: 109-153. Bickmore, L. 2000. Downstep and fusion in Namwanga. Phonology 17: 297-333. Bickmore, L. 2007. Chilungu Phonology. Stanford, CA: CSLI Publications. Botne, R., H. Ochwada & M. Marlo. 2006. A Grammatical Sketch of the Lusaamia Verb. Grammatical Analyses of African Languages 30. Cologne: Rüdiger Köppe. Cassimjee, F. 1986. An autosegmental analysis of Venda tone. Ph.D. dissertation, University of Illinois. Cassimjee, F. & C. Kisseberth. 1998. Optimal Domains Theory and Bantu Tonology: A Case Study from Isixhosa and Shingazidja. In L. Hyman & C. Kisseberth (eds), Theoretical Aspects of Bantu Tone, 33-132. Stanford, CA: CSLI Publications. Cheng, C.-C. & C. Kisseberth. 1979. Ikorovere Makua tonology (part 1). Studies in the Linguistic Sciences 9: 31-63. Cheng, C.-C. & C. Kisseberth. 1980. Ikorovere Makua tonology (part 2). Studies in the Linguistic Sciences 10: 15-44. Cheng, C.-C. & C. Kisseberth. 1981. Ikorovere Makua tonology (part 3). Studies in the Linguistic Sciences 11: 181-202. Clements, G.N. 1984. Principles of tone assignment in Kikuyu. In G.N. Clements & J. Goldsmith (eds), Autosegmental studies in Bantu tone, 281-399. Dordrecht: Foris Publications. Dalgish, G. 1986. /-a-/ reduction phenomena in Luyia. Studies in African Linguistics 17: 155-76. de Lacy, P. (ed). 2007. The Cambridge Handbook of Phonology. Cambridge: Cambridge University Press. Downing, L. 1996. Problems in Jita tonology. München: LINCOM EUROPA. Ebarb, K., M. Marlo, & P. Otiato. in prep. Nyala-West verbal tonology. Indiana University, University of Maryland, & University of Illinois. Goldsmith, J. 1984a. Tone and accent in Tonga. In G. N. Clements & J. Goldsmith (eds), Autosegmental Studies in Bantu Tone, 19-51. Dordrecht: Foris Publications.

Goldsmith, J. 1984b. Meeussen’s rule. In M. Aronoff, R. Oehrle, F. Kelley, & B. Stephens (eds), Language sound and structure, 245-259. Cambridge, MA: MIT Press. Hyman, L. 2001. Privative tone in Bantu. In S. Kaji (ed), Cross-linguistic studies of tonal phenomena, 237-257. Tokyo: ILCAA. Hyman, L. 2003. Segmental phonology. In D. Nurse & G. Philippson (eds), The Bantu languages, 42-58. London: Routledge/Curzon. Hyman, L. & A. Ngunga. 1994. On the non-universality of tonal association conventions: evidence from Ciyao. Phonology 11: 25-68. Hyman, L. & F. Katamba. 1990. Spurious high-tone extensions in Luganda. South African Journal of African Languages 10: 142-158. Kidima, L. 1991. Tone and accent in Kiyaka. Ph.D. dissertation, UCLA. Kim, Y. & M. Paster. 2007. Downstep in Tiriki. Ms, UC-Berkeley & Pomona College. Kisseberth, C. & D. Odden. 2003. Tone. In D. Nurse & G. Philippson (eds), The Bantu Languages, 59-70. London: Routledge. Leung, E. 1991. The tonal phonology of Llogoori: a study of Llogoori verbs. Ithaca: CLC Publications. Liphola, M. 2001. Aspects of the phonology and morphology of Shimakonde. Ph.D. dissertation, Ohio State University. Marlo, M. 2007. The verbal tonology of Lumarachi and Lunyala: two dialects of Luluyia (Bantu, J.30, Kenya). Ph.D. dissertation, University of Michigan. Marlo, M. 2008. Reversal and predictability in Luyia verbal tone systems. Paper presented at the Annual Conference on African Linguistics 39. Marlo, M. 2009. Prefix-stem fusion in Bantu. ms. University of Maryland. Marlo, M. to appear, a. Tura verbal tonology Studies in African Linguistics 37: 153243. Marlo, M. to appear, b. Khayo verbal tonology. Africana Linguistica 15: 77-129. Mutonyi, N. 2000. Aspects of Bukusu phonology and morphology. Ph.D. dissertation, Ohio State University.

Mwita, C. 2008. Verbal tone in Kuria. Ph.D. dissertation, UCLA. Myers, S. 1988. AUX in Bantu morphology and phonology. In L. Hyman & C. Kisseberth (eds), Theoretical aspects of Bantu tone, 231-264. Stanford, CA: CSLI Publications. Odden, D. 1981. Problems in tone assignment in Shona. Ph.D. dissertation, University of Illinois. Odden, D. 1987. Predicting tone in Kikuria. In D. Odden (ed.), Current approaches to African linguistics 4. Dordrecht: Foris, 311-326. Odden, D. 1989. Predictable tone systems in Bantu. In H. van der Hulst & N. Smith (eds), Autosegmental studies on pitch accent systems, 225-251. Dordrecht: Foris Publications. Odden, D. 1996. The phonology and morphology of Kimatuumbi. Oxford: Oxford University Press. Odden, D. 1998. Verbal tone melodies in Kikerewe. In T. Hinnebusch & I. Maddieson (eds), Theoretical Approaches to African Linguistics, vol. II, 177-184. New Brunswick: Red Sea Press. Odden, D. 2009. Tachoni verbal tonology. Language Sciences 31: 305-324. Onyango, J. 2004. Tone mapping and tense in Olunyala. Malilime: Malawian Journal of Linguistics 4: 65-81. Onyango, J. 2005. Universal properties of tone: the interaction between verbal mood and tone patterns in Olunyala. In R. Bett, C. Etzhhold, & E. Müller (eds), Proceedings of the Conference “Across Borders: Benefiting from Cultural Differences”, 132-149. Nairobi: DAAD, Regional Office for Africa. Onyango, J. 2006. Verb tonology in Olunyala. Ph.D. dissertation, Kenyatta University. Poletto, R. 1998. Constraints on tonal association in Olusamia: An optimality theoretic account. In L. Hyman & C. Kisseberth (eds), Theoretical aspects of Bantu tone, 331364. Stanford, CA: CSLI Publications. Roberts-Kohno, R. 2000. Kikamba phonology and morphology, Ph.D. dissertation, Ohio State University. Scheer, T. 2009. What OT is, and what it is not, Review of de Lacy (ed), 2007, The Cambridge Handbook of Phonology. Journal of Linguistics.