4.1 Tonal neutralization in Taiwan Southern Min
4.1.2 Problematic grammatical words
with lower grammaticality degree (i.e. Fnc≤2) is improper for the excessive violations of SP-MAX-X0 incurred.
We have mentioned in the last section that there are a few grammatical words appear to be exceptional to the generalization which is complied with by most other members within the same class on the grammaticality scale. These exceptions are problematic because they would undermine the assumption that tonal neutralization correlates with the degree of grammaticality, and therefore they need to be fixed. The exceptions can be distinguished between two types, neutralization in low grammaticalized classes, and non-neutralization in highly grammaticalized classes, which are examined in order.
Two cases of neutralization in low grammaticalized classes are spotted in Be2 go7-tai5 hoo7-gua0 tioh8=bo0
To.buy five.CL give-me right=SFP
‘He bought me five, right?’
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
Taiwanese. The first case is the frequently used phasal complement tioh8 ‘attached, at, attained.’ Phasal complement is argued to be a degree 2 grammatical word in this dissertation, and as such -tioh8 should have resisted neutralization, in line with the other members within this class (cf. -ho2, -uan5, -liau2, etc. in (10)). Yet, it is attested to be neutral-toned in the final position, as the following examples show.
(16) Neutral-toned tioh8
For expository purposes, I assume that the complement -tioh8 is more grammaticalized than the other phasal complements in Taiwanese, as can be justified in terms of two of Lehmann’s (2015) parameters. The first parameter available is “the semantic integrity,”
which consists of the number and concreteness of semantic features of a word.
Reductions in semantic integrity (often involving semantic generalization and semantic bleaching) correlates with higher degree of grammaticalization. In terms of this, -tioh8 can be understood as more grammaticalized than the complements -ho2 , -kau3 and -tiau7 because it has a relatively more bleached, abstract, meaning ‘instant realization of some acquired state,’ compared to the meanings ‘well/properly done’ of -ho2,
‘arrival/attainment’ of -kau3, and ‘realization of loss’ of -tiau7. As for the remaining complements -uan5, -suah4, and -liau2, they can be regarded as less grammaticalized than -tioh8 based on the other parameter “paradigmatic variability,” which refers to the freedom with which a speaker can choose a word from a range of options.” A reduction a. Gua2 u7 khuann3-tioh0
I PRF.MODAL look-PHA
‘I have seen it.’
b. I1 khi3 khap4-tioh0 S/he go clash.with-PHA
‘He came striking against it.’
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
in paradigmatic variability, that is, having less competitors for conveying the same function, corresponds to a higher degree of grammaticalization. In this sense, -uan5, -suah4, and -liau2 are paradigmatically more variable than -tioh8, because they share the similar meaning ‘finishing/completion’ and thus competing with each other in the selection.
The discussion so far about tioh8 is only part of the story, however. Things get more complicated when we take into consideration such factors as the presence/absence of the elements expressing negation and/or modality. As shown in the table in (17), we can see that the modal and/or negation elements (i.e. -u7/-bo5 and -e7/-be7) may function as an infix blanked by the verb and the complement in question. What is particularly surprising is that, in this case, -tioh8 remains with its full lexical tone. It can be generalized from these examples that only when the complement -tioh8 is immediately preceded by the verb can it bear a neutral tone.
(17) Full-toned tioh8 with negation and/or modality
To account for this discrepancy of tonal behavior of -tioh8, I suggest treating the two constructions separately. Specifically, -tioh8 as a complement immediately preceded by the verb has a higher degree of grammaticalization, while in the construction with infixation of modal/negation elements, the complement -tioh8 is relatively lower-grammaticalized, presumably about the same degree of the other phasal complements a. Tshat8a2 liah8-beh7-tioh8
Thief catch-NEG-PHA
‘(They) cannot catch the thief.’
b. Gua2 khuann3-e7-tioh8 I see-be.able.to-PHA
‘I can see.’
‧
Lehmann’s parameters ― “bondedness,” which corresponds to the degree to which a word depends on, or attaches itself to, other words or phrases. Increase in bondedness means less autonomy and thus means a more grammaticalized element. A commonly used criterion for bondedness between elements is the possibility to insert something in between. Under this criterion, the bondedness between the complement -tioh8 and the main verb in the “V + modal/negation + -tioh8” construction is relatively weaker than in the construction with no insertion, which gives rise to the difference in grammaticalization degree and the ensuing discrepancy in tonal neutralization. The partial scale in (18) summarizes the grammaticalization degree of -tioh8 accordingly.(18) Grammaticalization degree of -tioh8
The other case of neutralization in low grammaticalized classes involves the numerals tsit8 (one) and nng7 (two), together with the classifiers attached to the end of them in the construction V + tsit8/nng7 + CL. In this case, the numerals do not denote real quantity, but a small amount, and the lexical tones carried by those numerals as well as the ensuing classifiers are neutralized altogether in the final position, as shown in the following examples. This occurrence of neutralization seems bizarre because numerals and classifiers are assumed as degree 1 and degree 2 grammatical words, respectively, which would have their lexical tones intact.
‧
(19) tsit8/nng7 denoting small quantity
This problem can be solved by the same token. That is, the numerals together with the classifiers are deemed relatively more grammaticalized as the meaning has been bleached, no longer referring to the original descriptively specific quantity. On the other hand, the higher degree of grammaticalization is also manifested by the strong bondedness between the numerals in question and the preceding verb, given the fact that tsit8 and nng7 in this usage are always attached to the preceding verb. These arguments are cast in the revised scale (20), where numerals and classifiers in the construction V + tsit8/nng7 + CL with unspecific reference are categorized in the degree 3 class.
(20) Grammaticalization degree of tsit8/nng7
NUM PHA
Let us turn to the exceptional non-neutralization in highly grammaticalized classes.
There is only one case of this type of exception: the possessive marker e5. As a subtype of modifier marker, possessive marker generally shares the same phonological status as the marker generally suffixed to other types of modifiers. However, this is not the case for Taiwanese. Although indeed sharing the same segments and citation tone as the
‧
modifier marker in other usages, the possessive marker e5 differs crucially in preserving its lexical tone even in the final position, as shown in (21). It is this preservation that strikes a discordant note in comparison with the general modifier marker, which as a degree 5 marker has its lexical tone neutralized.
(21) Adjectival modifier marker vs. possessive marker
One may assume that this is also a case where seemingly identical grammatical word have different degrees of grammaticalization in different constructions/usages and thus can be settled in the same way that we have operated before. Indeed, the grammaticalization degree of possessive marker may be a little less grammaticalized than the general modifier marker. In terms of the parameter with respect to desemanticization, the general modifier marker denotes nothing, only serving as a plain grammatical connector in a modifying construction, while a possessive marker conveys possession/ownership in addition. Accordingly, we may distinguish between the two types of modifier marker on the scale of grammaticalization degree, with possessive marker at degree 5 and the general version at degree 5.5, as depicted in (22).
(22) Grammaticalization degree of modifier marker vs. possessive marker DIR
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
This move alone, however, fails to solve the paradox that we are confronted with. This is because the threshold of neutralization is attested to be at degree 3 in Taiwanese, and therefore the assumed degree 5 possessive marker would still be in the range where tonal neutralization occurs.
To get rid of this dilemma, I venture to propose that the non-neutralization of tone in possessive marker is driven by the avoidance of having categories with different degrees of grammaticalization surface as homophonous. This proposal extends the notion of “contrast preservation” in Lubowicz (2003) that it is imperative to preserve the distinction between contrastive elements, to the domain of grammaticalization. That is, with what was originally one category being reanalyzed as two in the course of grammaticalization, distinction would be drawn or maximized between the new, more grammaticalized, category on the one hand, and the old, less grammaticalized, category on the other. Such a distinction can be made by syntactic means, for example, the structural difference between auxiliaries (such as do, have, etc.) and their lexical source verbs (Hopper & Traugott 2003), or by morphophonological means, for instance, the different lexical tones assigned to the moving verb do4 and the phasal complement do3 that is grammaticalized from the verb (M. Chiang 2018). This conception can be cast in OT by adopting the anti-merger constraint family, the systemic faithfulness constraint in Padgett (1997, 2003) and Ito & Mester (2003), which are in turn based on Dispersion Theory developed in Flemming (1995, 2002). The definition of an anti-merger constraint that is relevant to the current problem is given in (23), which is manipulated to refer to specific words or categories, in this case, the two types of modifier markers.
The reason why we do not use a more general formulation here is because it is unclear and beyond the scope of this dissertation whether there are other grammatical words/categories characteristic of anti-merger of this sort.
‧
Assign one violation mark for every instance of homophonous modifier marker and possessive marker.
With this constraint, we can account for the non-neutralization of possessive marker by ranking the anti-merger constraint above ALIGN-R(ω, Fnc≤2), as illustrated in (24).
(24) NOMERGE(MOD/POSS)≫ ALIGN-R(ω, Fnc≤2) ≫ SP-MAX-X0