4. Supporting evidence
4.2 Typological evidence
4.2.3 Over-marking
making the few cases attested not real exceptions to our theory.
4.2.3 Over-marking
Finally, category 6-8 include the situation where over-making occurs, in other words the functions of countability marking and noun classification are represented more than once within the language. Following our hypothesis, we expect two levels: the first one is mutual exclusivity between numeral classifiers and plural marker, while the second is tendency not to co-occur between numeral classifiers and genders.
For category 6, plural marking is only attested once via numeral classifiers, but an overlap occurs on noun classification, since it is carried twice by numeral classifiers and genders. Nevertheless, the ratio is very low which is in accordance with our hypothesis: only 1% (2/155) of our data displays this phenomenon: Maybrat and Tidore. Moreover, as mentioned in the theoretical discussion: noun classification is a semantic function which allows a low level of stacking therefore it is not a real exception to our hypothesis. The existing literature also provides some possible explanations, as an example it is possible that these languages are in period of transformation from one system to another, as observed by Reid (1997) on an Australian language Ngan’gityemerri, in which generic classifiers developed into genders. From our point of view, this is similar to what is happening at the cross-linguistic level: the two systems fulfilling the same function of noun classification are in competition, resulting in a divided personality in some languages.
As an example for Maybrat, we may note that the numeral classifiers are present as in (34a) but they are optional (Gil, 2013), while the gender system is pronominal, as shown in (34b), being less constraining than conventional gender systems. As for Tidore, numeral classifiers are existent, as demonstrated in (34c), however they are also optional (Gil, 2013) while the gender system seems to be pronominal, as in (34d).
As a summary, in both languages the conflict between the two systems is thus not so strong, which partially supports our hypothesis.
(34) Occurrence of numeral classifier and gender but absence of plural marker a. Numeral classifiers in Maybrat (Reesink, 1996:10)
Raa m-abo trion tuuf m-jin tet abyo men 3P-human CL 3 3P-sleep cave above
‘Three men slept at the mouth of a cave.’
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
43
b. Genders in Maybrat (Waren, 2007:9)
Person 1SG 2SG 3SG(M) 3SG(F) 1P 2P 3P
Pronoun jio nyo ait au amu anu ana
c. Numeral classifiers in Tidore (Van Standen, 2001)
fayaa nga-rukange igo ngai-rora
woman CL-three coconut CL-six
‘three women’ ‘six coconuts’
hate ma-jaga ngai-moi igo futu-rora
three NM-branch CL-one coconut CL-six
‘one branch’ ‘six coconut trees’
d. Genders in Tidore (Reesink, 1996:3)
Person 1SG 2SG 3SG(M) 3SG(F) 1P-IN 1P-EX 2P 3P
Pronoun to- no- wo- mo- fo- mo- no- yo-
For category 7, the overlap is on countability marking since both numeral classifiers and plural markers are attested in the languages. The tendency is still low, only 6%
(9/155) in our data, nevertheless it still needs further analysis since we expected mutual exclusivity. The languages involved are Armenian, Garo, Hungarian, Indonesian, Mokilese, Teribe, Oriya, Turkish and Tuvaluan. Interestingly, a closer analysis shows that even if numeral classifiers and plural markers are both present in these languages, however they never co-occur on the same noun, as demonstrated in (35) for Hungarian. When the plural marker k is used in (35a), the numeral classifier cannot appear in the noun phrase, however in (35b) when the numeral classifier csik is present, the plural marker k cannot show up, displaying that the two elements are indeed mutually exclusive on the same noun within a language.
(35) Avoidance of numeral classifier and plural marker on the same noun, Hungarian (Csirmaz and Dekany, 2010:12)
a. rago-k
chewing.gum-PL
‘chewing gums’
b. ket csik rago
two CL-strip chewing.gum
‘two chewing gums’
‧
Armenian displays the same phenomenon: numeral classifiers are available in the language structure (36a), but we realize that when the plural marking -er is present in (36b) we do not add the numeral classifier had. Finally (36c) shows that the two elements cannot co-occur in the noun phrase without creating ungrammaticality.
(36) Avoidance of numeral classifier and plural marker on the same noun, Armenian (Bale and Khanjian, 2008:75)
We may observe the same phenomenon in (37) with Oriya, an Indo-Aryan language of India, in which number is realized either as a numeral or as a nominal inflection but not with both, e.g. in (37a) with numeral classifier while it is via plural marker in (37b), but the two elements cannot co-occur as demonstrated in (37c). Similar phenomenon are attested in Turkish (Her, 2013:12), Teribe (Quesada, 2000:52), Mokilese2 (Doetjes, 2012), Garo (Her, 2013:11) and Indonesian (Sato, 2009:201).
(37) Avoidance of numeral classifier and plural marker on the same noun, Oriya
2 For Mokilese, the plurality is attested to be marked on the determiner rather than the noun, therefore it does not present a real exception either.
‧
Another situation may be that the numeral classifiers attested are questionable, as an example in Tuvaluan, the numeral classifier inventory is actually not productive and it may be more appropriate to define it as classifier-like elements rather than real numeral classifiers (Besnier, 2000:367). As demonstrated in (38), the classifier-like element tau may be preceding ordinary numerals to refer to the number of coconuts in a bunch. However it behaves differently from numeral classifier and measure word: a numeral classifier represents the quantity of 1 rather than a group of things, while a measure word should carry the quantity of the bunch itself rather than using other numerals, e.g. in Chinese yi da ‘one M-dozen’ meaning ‘a dozen’ in English, the measure word da carries itself the information that the group of things contains twelve members. This is not the case in our example with tau in Tuvaluan, since the quantity of coconuts in the bunch is referred from the separately expressed numeral.
(38) Numeral classifier-like elements in Tuvaluan (Besnier, 2000:571) a. E tau lima te fui pii teelaa.
Nps N-in-bunch five the bunch drinking-coconut that
‘That bunch of drinking coconuts has five [nuts].’
b. Tii mai aka i te tau tolu teenaa ! bring-down Dxs please at the N-in-bunch three that
‘Pitch down [some nuts] from that threesome over there!’
For category 8, we have the most extreme case of over-marking occurring, since the languages involved are attested to have numeral classifiers, genders and plural markers. Numeral classifiers carry both countability marking and noun classification, genders carry noun classification while plural markers carry countability marking, resulting in double presentations for each of the two functions. The tendency is still low, as predicted by our hypothesis: only 6% (9/155) of the data are included here (Russian, German, Marathi, Awadhi, Polish, Bhojpuri, Nepali, Irish, Kolami).
Nevertheless, this category is estimated to be on the extreme of the continuum between economy and expressiveness but it seems that its ratio is much higher than other dispreferred categories such as 6. A more detailed analysis reveal that these exceptions are actually explicable: similar to category 7 we may find two explanations regarding the overlap of countability marking between numeral classifiers and plural markers. First, even though the languages are attested to have numeral classifiers, their inventory is in fact not productive, sometimes being even more extreme than poor classifier languages defined by Tang (2004). As an example, detailed analysis of a language may reveal that it is attested as a classifier language, however it just has one or few numeral classifiers which are restrictedly used only with specific nouns.
‧
Samples are demonstrated in (39a) with German, which is only attested to have one numeral classifier used in restricted contexts. The similar situation is found in (39b) with Polish by the classifier sztuk borrowed from German stuch.
(39) Unproductive numeral classifiers (Sussex and Cubberley, 2006:314-315) a. German
er hat 5 Stück Brötchen gekauft
‘He bought five CL bread rolls.’
b. Polish
Jan kupil dziesiec sztuk probowek
‘Jan bought five CL test-tubes.’
Nevertheless, there are also languages attested to have their numeral classifier system in a more evolved stage. An example would be Russian, which has a more productive numeral classifier inventory compared to German and Polish, as demonstrated in (40).
(40) Developing numeral classifiers in Russian (Sussex and Cubberley, 2006:314) a. u nego bylo pjat stuk karandsej
‘He had five CL-inanimate pencils.’
b. u nego bylo pjat golov skota
‘He had five CL-head cattle.’
c. v klasse sidelo pjat celovek studentov
‘In the classroom sat five CL-person students.’
In (40a) the classifier stuk occurs generally with non-animate referents, while in (40b) golov usually combines with nouns referring to countable nouns of herd animals such as horses, and in (40c) celovek refers to human nouns (Goto, 2012: 15-18). It is necessary to point out that the usage of numeral classifiers is still defined as not productive since it is not obligatory and their usage is ongoing changes. An example would be that in Russian some elements such as stuck and celovek are developing as numeral classifiers therefore becoming applicable only on count nouns while some others e.g. golov are still used with collectives (Goto, 2012: 26).
Similar situations are found for both cases with the other languages in this category:
Marathi is attested to have only one numeral classifier (Emeneau, 1956:11), two in Bhojpuri (Verma, 2007:258), four with Awadhi (Barz and Diller, 1985:162). On the other hand, Irish is only attested to have classifier-like elements (Toyota, 2009:125) and Kolami is attested to have classifier-like elements with human nouns when using
‧
numerals borrowed from Marathi (Emeneau, 1956:12). Since numeral classifiers are not productive or in the process of mutation in these languages, the co-occurrence of two systems representing countability marking or noun classification per se does not really exist. If applying more strict terms they should be attributed to category 1 with only genders and plural markers. In this case, the double marking of noun classification would also be solved.
An opposite extreme is Nepali, which the literature attests as having a change of systems among classifiers and genders (Pokharel, 2010:40), resulting in co-occurrence. This phenomenon is attested within various languages of South-America, Western Asia and in the Northern-Australian region, as an example with Arawak, Dravidan, Iranian and Oceanic languages (Aikhenvald, 2000: 185-186), which as mentioned previously is exactly on the meeting line between the two systems of numeral classifiers and genders, partially explaining our observation.
4.2.4 Summary
provide concrete evidence. Regarding the categories of under-marking (12%, 19/155) and over-marking (13%, 20/155) they have different levels of dispreference but both occur in low frequency in our data, as we imagined. Moreover, explanation is provided: regarding under-marking, the lack of numeral classifiers or genders fulfilling noun classification is speculated to be caused by the occurrence of other elements carrying this function, e.g. case marking, noun classifiers, verb classifiers.On the other hand, we propose that the absence of countability marking via numeral classifiers or grammatical plurals is due to the restricted numeral systems in the languages involved. Regarding over-marking we hypothesize that the languages attested are in the middle of a competition between different systems fulfilling the same function such as genders and numeral classifiers or numeral classifiers and plural markers. Since lexical diffusion (Wang, 1969, 1977) is occurring gradually rather than abruptly, we do expect to witness a co-occurrence of two different systems, until a balance is reached or one of them takes advantage over the other.
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
48