Syllabification and lowering

We noted in section 4.1.3. above that Lowering (a) determines the quality of unstable suffix-initial vowels and (b) interacts with syllabification (cf. (16)). We also saw that the source of lowering in the representation of a stem or a suffix may be an unpredictable ‘mark’ or one which is morphologically predictable (cf. (19)). We claim that, representationally, this mark consists of two distinct (though always coocurring) characteristics. A lowering stem/suffix has

1 d

a final floating [+ open ] feature and a morpheme-final defective vowel V as shown in (32):

(32) N

*

X X

* ]stem

root

ª

[+ open ]1

Thus, specially marked nouns and pronouns, all adjectives, and all inflectional suffixes end in a structure shown in (32). This means that some of the representations discussed above have to be modified because they are, or they contain, such formatives. For instance, szörny must be represented as /sörn V / instead of /sörn / because it is an (unpredictably) lowering noun^y d ^y

and bor-t must be /bortV / and not /bort/ because the final suffix is inflectional and therefored

lowering). Naturally, the syllabification of such forms is also different, but given our assumptions about the syllabification and the interpretation of defective vowels, these modified representations will not change the outcome of the derivations, i. e. the phonetically realized forms. In the two examples above, for instance, the final defective vowel does not syllabify because it cannot occur in an open syllable, so these forms are parsed as /{sörn }V / and^y d

/{bort}V /, and thus the correct surface forms ([sörn ] and [bort]) are derived. We shall discussd ^y

some more complex cases below.

We interpret lowering as a process that spreads the floating [+ open ] feature locally to1

a (full or defective) vowel which is licensed (i. e. incorporated into a syllable) and is at the edge of a morpheme. Lowering applies regardless whether the licensed vowel is morpheme-initial or morpheme-final. The spreading process is local and non-iterative, i.e. it targets a single V. If the target is a full vowel, the floating feature can spread to its aperture node. In the case of defective target vowels, we assume that the nodes necessary for preserving well-formedness (e. g. root, vocalic, aperture) are automatically created in the course of the spreading to the empty skeletal position (cf. Sagey 1986, Clements and Hume 1995). This is indicated by parentheses enclosing the relevant nodes.

(33) Lowering

F

*

N

*

X (root) (vocalic) (aperture) [+ open ]1

Condition: the target is peripheral in a morpheme

d f

Whether the target of Lowering is V or V , the output of the process is a structure shown in (34) (where irrelevant nodes are omitted). Crucially, this means that (in addition to its lowering effect) Lowering turns a defective vowel into a full one:

< September 12, 2007 (10:35am)> < DocChapter4_2_SOURCE_FINEW082.wpd> 174

For expository purposes we shall use the following special symbols in the representations

55

OP d 1 FOP

below: Subscripted ‘ ’ before V stands for the floating [+ open ] feature. V denotes the lowered full vowel that is the result of (33). It must be borne in mind, however, that the linear representations used are just shorthand for the corresponding non-linear ones in the same way as phonetic symbols are for the appropriate feature trees.

The placement of relative to a syllable boundary is irrelevant and is not meant toOP 56

indicate whether the floating feature is inside or outside a syllable.

(34) F

Spreading is a feature filling process, therefore Lowering is blocked if the target vowel has an aperture specification which is incompatible with the feature that is being spread.

Let us now examine how lowering interacts with syllabification. As pointed out⁵⁵ above, a word-final defective vowel is not realized phonetically because it cannot be syllabified. Therefore, in this position, a floating [+ open ] feature does not surface since (33)1

cannot apply to a syllabically unparsed V . (35) shows this with monomorphemic loweringd

stems and inflectional suffixes (Type A and Type B) attached to non-lowering stems: ⁵⁶

OP d OP d

Recall that Type B suffixes (such as the accusative) always show up with a (lowered) linking vowel after lowering stems regardless whether the stem final consonant can or cannot form a

licit coda with the suffixal -t. The reason is that lowering stems are vowel-final. In both cases the stem-final defective vowel can syllabify with the -t and, consequently, can be the target of Lowering, which turns it into a low full vowel (36a). This contrasts with the behaviour of -t after major stems where the defective vowel only appears as a result of overparsing by syllabification after stem-final consonants that cannot form a licit coda with the suffixal consonant (36b):

Thus, the generalization stated in (16) follows from the representation of lowering stems, the syllabification algorithm and the special constraints on the syllabification of defective vowels.

We have seen that Type A suffixes show up with a low linking vowel after lowering stems. Since Type A suffixes underlyingly begin with a full vowel, the stem-final V ofd

lowering stems is deleted by Hiatus and the floating [+ open ] feature of the stem can spread1

to the licensed suffix-initial V .f

(37) fog-ak ‘tooth’ (pl.). fog-atok ‘your tooth’

OP d f OP d OP d f OP d

syllabification {fo}{g V k }V {fo}{g V }{tok }V

FOP OP d FOP OP d

Lowering {fo}{gV k }V {fo}{gV }{tok }V

[fog]k] [fog]tok]

As pointed out above, a suffix-initial V does not lower (i.e. it cannot receive the spreadingf

feature), if it has an aperture feature which is incompatible with the feature spread by Lowering. That is the reason why the suffix-initial vowels of two Type A suffixes,

< September 12, 2007 (10:35am)> < DocChapter4_2_SOURCE_FINEW082.wpd> 176 superessive -on/-en/-ön and possessive 1pl -unk/-ünk, do not lower after lowering stems: e.g.

fal-on ‘on the wall’ and fal-unk ‘our wall’. The initial V s of both these suffixes aref

underlyingly specified as [!open ] (cf. Siptár & Törkenczy 2000) and thus cannot receive the1

spreading [+ open ] feature.1

Lowering may be unordered with respect to Hiatus as both possible orderings yield

1 d

the correct results. Note that Lowering can spread [+ open ] past an unlicensed V onto the closest potential target (the suffix-initial V ) because defective vowels have no melodicf

structure. Compare the two ways of ordering Hiatus and Lowering in (37) and (38):

(38) fog-ak ‘tooth’ (pl.)

OP d f OP d

/fog V -V k V /

OP d f OP d

syllabification {fog }V {V k }V

d FOP OP d

Multiply suffixed forms of lowering stems are also derived in a straightforward manner. (39) shows how the accusative plural of a lowering major stem (fog ‘tooth’) and an ‘epenthetic’

lowering stem (sátor ‘tent’) is derived.

(39) fog-ak-at ‘tooth’ (pl. acc.) sátr-ak-at ‘tent’ (pl. acc.)

OP d f OP d d d OP d f OP d OP d

As can be seen in the derivations, syllabification is continuous, i. e. potentially, it reapplies after each phonological rule. Both Hiatus and Lowering show derived environment effects: the

Similar rules have been identified in a number of other languages, e.g. Finnish and

57

Ondarroan Basque, cf. Hualde (1989) and Cole (1995).

Jensen and Stong-Jensen (1989a) argue for cyclic epenthesis in Hungarian to account for

58

the behaviour of ‘epenthetic’ stems. However, their arguments do not contradict our claim because (i) essentially, they are based on their epenthesis process blocking in a non-derived context; and (ii) the arguments do not carry over to the present analysis because they are crucially dependent on the assumption that it is epenthesis that is responsible for the vowel-zero alternation in ‘epenthetic’ stems, Type A, and Type B suffixes alike. This is a view that we reject for the reasons discussed in 4.1.1.

former can only apply if its target is in another morpheme and the latter at the edge of a morpheme in the environment of another one. Neither applies intramorphemically. However, there is no evidence that their application is cyclic: they only ever need to apply once in the course of the derivation (naturally, they may have multiple targets). Rules of this kind challenge the traditional claim in Lexical Phonology that only cyclic rules are subject to the derived environment constraint on rule application. In fact, we know of no phonological rules⁵⁷ in Hungarian that must be considered cyclic on grounds other than the derived environment constraint. Therefore—⁵⁸ although the phonological rules belong to blocks (Block 1 and Block 2) and each suffix is marked according to whether it is analytic or synthetic—we assume that

< September 12, 2007 (10:35am)> < DocChapter4_2_SOURCE_FINEW082.wpd> 178

It must be pointed out that as long as only synthetic suffixes are attached to the stem, it

59

makes no difference if we assume that the whole ‘preassembled’ suffixed stem is subjected to the relevant phonological rules, or that the rules are (re)applied gradually (i. e. ‘cyclically’) as each suffix is considered, over the suffixed form. This can be seen in the derivation below, in which we adopt the Halle and Vergnaud (1987) approach to cyclicity:

OP d f OP d OP d

fog V -V k V -t V cycle1

OP d f OP d OP d

syllabification {fog }V -V k V -t V

Hiatus n.a.

Lowering n.a.

cycle2

OP d f OP d OP d

syllabification {fog }V {V k }V -t V

OP f OP d OP d

Hiatus {fog }V k V -t V

OP f OP d OP d

syllabification {fo}{g V k }V -t V

FOP OP d OP d

Lowering {fo}{gV k }V -t V

cycle3

FOP OP d OP d

syllabification {fo}{gV }{k V t }V

Hiatus n.a.

FOP FOP OP d

Lowering {fo}{gV }{kV t }V

[fog]k]t]

In the representations that follow the irrelevant suffix-final floating [+ open ] feature and1 60

V are disregarded.d

Note that Hiatus must be non-iterative, because it would delete the suffix-initial V if itf 61

could apply to its own output.

the derivation proceeds in a non-cyclic way. ⁵⁹

If lowering stems that are phonetically vowel-final in isolation are represented on a par with the lowering stems discussed above, then they must end in a sequence of a full

f OP d

vowel and a defective vowel underlyingly (CV V ). The syllabification algorithm and the rules discussed predict that both Type A and Type B suffixes attach to these stems with a phonetically expressed lowered linking vowel. Type A suffixes are underlyingly vowel-initial.

When they are added to these stems, an underlying sequence of three vowels is created:

f OP d f d

-CV V + V C. Hiatus deletes the stem-final V , and the two full vowels syllabify in the⁶⁰

f OP f f

following way: -{CV }{V C}. Lowering can apply to the suffix-initial V giving -⁶¹

f FOP

{CV }{V C}. Type B suffixes are consonant-initial. Thus, suffixation by a Type B suffix

Modulo the OCP effect and the optionality of the linking vowel after mid vowels as

62

discussed in 4.1.3.

I deliberately avoid using the word ‘regular’ here, since lowering stems are marked

63

compared to non-lowering ones. ‘Normal’ is intended to mean ‘representing the norm for lowering stems’.

We have no explanation why all surface vowel-final lowering stems belong to this set. It

64

must be pointed out, however, that the set contains some surface consonant-final lowering stems as well, cf. note 23. It is an interesting fact that, in contrast to consonant-final inflectional suffixes, vowel-final ones do not lower: lány-ai-m ‘my daughters’ and not *lány-ai-am. The latter is predicted if Hiatus and Lowering apply to the underlying representation

OP d f OP d f OP d

*/la+n -ai V -V m V /, whereas the correct output is derived if the UR is /la^y +n -ai-V m V /.^y It is as if a hiatus consisting of a full vowel and a defective one were dispreferred within a morpheme.

f OP d d

creates the string -CV V + C. Note that Hiatus cannot delete the stem-final V because the

f OP d

vowel sequence is not derived. Syllabification yields -{CV }{V C} and via Lowering the

f FOP

derived representation is the same as in the case of Type A suffixes: -{CV }{V C}. Thus, the prediction is that Type A and Type B suffixes behave in the same way when added to surface vowel-final lowering stems: a lowered linking vowel shows up before both types of suffixes.

This prediction is not borne out. Some suffixes never have a linking vowel after a surface vowel-final lowering stem (e.g. accusative, superessive, comparative), others do (e. g.

plural, modal). This difference in behaviour only partially correlates with the distinction⁶² between Type A and Type B suffixes. In 4.1.3. above we pointed out that the unpredictability of behaviour indicates that lexical marking must be involved. We suggest that the source of this idiosyncratic behaviour is allomorphy. Surface vowel-final lowering stems have two lexical allomorphs: a ‘normal’ one that (like all lowering stems) ends in a defective vowel,⁶³ and another one whose final defective vowel is missing. By default, suffixes select the

‘normal’ allomorph. Some suffixes, however, are marked to select the other allomorph. When subjected to phonology, the former concatenations will surface with a lowered linking vowel while the latter ones will not have a linking vowel. Under this interpretation, a linking vowel is phonologically required after all lowering stems. Some lowering stems are special in that they have non-lowering lexical allomorphs as well, and some suffixes are morphologically⁶⁴ irregular because they select the non-lowering allomorphs of these stems.

< September 12, 2007 (10:36am)> < DocChapter4_3_SOURCE_FINEW083.wpd> 180

In the following discussion we abstract away from the effects of postlexical Degemination

65

(cf. Siptár & Törkenczy 2000, Siptár 2000, 2001b, 2002), which may shorten a geminate past -tt, compare Eve[t] körtét. ‘(S)he ate some pears’ and Eve[t+] epret. ‘(S)he ate some strawberries’.

(i) Recall that only some licit codas are available for the past tense suffix to syllabify.

66

I disregard this complication here (cf. 4.1.2.2.) and assume that there must be a stipulation specific to the past suffix that disallows its syllabification into a complex coda whose first term is an obstruent. It would be desirable to derive this effect from the representation of the past suffix and/or (more) general conditions on syllabification (cf. 3.4.3). At present, I do not see how this could be done. (ii) Note that, similarly to the accusative (cf. 4.1.2.2.), after t-final verbs a linking vowel appears ever though geminate /tt/ is a well-formed coda: üt-ött ‘hit’ (3sg past indef) and not *üt-t. For a discussion of this problem cf. 4.1.4.6.

This consonant would degeminate postconsonantally later. Since the past suffix must be

67

distinguished from the accusative (whose suffixal consonant never shows up as a geminate), the former cannot be a single consonant underlyingly.

<

In document The Phonotactics of Hungarian (Pldal 172-180)