Chapter 3 Methodology
3.2 Syntactic and Semantic Tests
國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
55
1) It could directly be attached to a numeral and noun.
2) It is a profiler which denotes the property of a noun.
3) The number of the object it refers to is x1.
then this proves that “本” is a true C.
As raw data is collected from the corpora for the syntactic tests, most of the elements other than Cs could be ruled out.
For some cases in which it is difficult to determine whether the number of the object is equal to one, sometimes the pictures in Google search are used. For instance, when the number of flowers in the phrase “一鉢花” is not defined, the pictures of flowers in one pot will be helpful to identify “鉢” as M. For Japanese native speakers,
“一鉢花” does not necessarily equal one flower but a pot of flowers.
3.2 Syntactic and Semantic Tests
3.2.1 General syntactic and semantic tests
There are three general syntactic tests – [Num CL N] and [N Num CL]
formations, generic C-replacement, and adjective insertion – to differentiate Cs and other word classes. The first one is to test whether the possible Cs could construct [Num CL N] and [N Num CL] constituents, combined with both numerals and nouns.
The second “generic C-replacement” is revised from Her’s “ge-replacement test” in
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
56
Chinese, stating that a true C could be replaced by generic C. The “adjective insertion”
test proposed by Her tests whether the possible C could be preceded by an adjective.
In Japanese, only nouns and Ms could be modified, Cs could not be modified by adjectives.
The semantic test is the “C in Numeral Quantification Scope” test provided by Her (2012). Her claimed that in the [Num C N] structure, numeral scoped over a noun, so the C could be deleted without changing the meaning. However, the deletion of Ms (or other word classes in this thesis) changes the meaning. Thus true Cs could only refer to one object (x1 multiplicand).
3.2.2 Distinguishing true classifiers from measure words
The tests eliminating Ms from 673 possible Cs are “adjective insertion” and “C in Numeral Quantification Scope”.
Cs and Ms could occupy the same syntactic position since their syntactic behaviors are similar. Consider the following example (25), the insertion of “の” is acceptable for both C and M. The tests are based on Her (2012).
(25) a. 一 個 の リンゴ ik ko no ringo
1 C GEN apple 1x1=1 apple “one apple”
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
57
b. 一 袋 の リンゴ hito hukuro no ringo 1 M-bag GEN apple
“one bag of apples” 1xM=not necessarily 1 apple Also, Her proposed that C/M are mutually exclusive. The example (26) shows that the C “個” cannot co-occur with the M “袋”, even though they can be used with the noun (apple) and numeral (one).
(26) a. *一 個 袋 の リンゴ ik ko hukuro no ringo 1 C M GEN apple “*one, bag apple ”
b. *一 袋 個 の リンゴ hito hukuro ko no ringo 1 M C GEN apple “*one bag, one apple”
The “adjective insertion” test can identify Cs and Ms. It is revised by Her’s
“C/M Distinction in Adjectival Modification Scope” mentioned in the example (15) (on page 32). He claimed that the adjective preceding the C could modify the
following N in Chinese. This could not be used to test Japanese, since an adjective is
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
58
not favored by Japanese Cs. However, Lareau (2003) suggested that Japanese Ms could be modified by adjectives, while modification to Cs is not favored. The following example (27) is quoted from Lareau (2003; 5). He claimed that modification to Japanese Cs is relatively ungrammatical as compared to Ms.
(27) a. 大きな 三 樽 の 水 ookina san tatu no mizu big 3 M-barrel GEN water
“three big barrel of water”
b. *大きな 三 匹 の 猫 ookina san biki no neko big 3 C GEN cat “*three big cats.”
For “C in Numeral Quantification Scope”, the scope of a numeral includes the noun because the C is a x1 multiplicand. So the deletion of C not change the number of object it refers to. As example (28) displays.
(28) 二 個 ルーター = 二 ルーター ni ko ruutaa ni ruutaa 2 C router 2 router “two routers” “two routers”
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
59
These tests can delete Ms from the list of possible Cs.
3.2.3 Distinguishing true classifiers from nouns
The tests eliminating nouns from 673 possible Cs are “[Num CL N] and [N Num CL] formation”, “generic C-replacement,” and “adjective insertion”.
First of all, two nouns cannot co-occur. Comparing examples (32a) and (32b) below, “台” and “車” were both listed as possible Cs. In (29a), “台” is a true C so it fits in the [Num C N] structure. In (32b), “車” is a pure noun, so it cannot be followed by another noun.
(29) a. 一 台 車 ichi dai kuruma 1 C car “one car”
b. *一 車 車 ichi sha kuruma7 1 car car “*one car car”
Next we describe the “generic C-replacement” test. Due to the fact that Cs are semantically redundant, they are all x1 multiplicand. The substitution of generic C
7 Some Japanese morphemes have different spellings (音讀 and 訓讀). The noun “車” mostly spells as くるま(kuruma) , and it spells as しゃ (sha) when preceded by numeral to count “the load of cars”, or act as “suffix” to noun to form words like “人力車”, “汽車”, “機関車”, etc.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
60
“個” or “つ” is grammatical. As the following example (30) shows, (30a) and (30b) are both grammatical, and there is no possible case in which one is true and the other is false.
(30)a. 一 匹 犬 ip piki inu 1 C dog “one dog”
b. 一 つ 犬 hito tsu inu one C-generic dog “one dog”
Since nouns are not semantically redundant, they cannot be substituted by the generic C “個” or “つ”. Consider example (31), “案” and “本” are both possible Cs, but “一案” is not equal to “一つ”, so it cannot be a true C. However, “本” cannot be substituted by the generic C “つ.” Thus, “本” is a true C and “案” is actually a noun.
(31) a. 一 案 ≠ 一 つ ichi an ≠ hito tsu 1 case ≠ 1 C ？ “one case” “one ？”
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
61
b. 一 本 線 = 一 つ 線 ip pon sen = hito tsu sen 1 C string = 1 C-generic string “one string”
For the third test “adjective insertion”, modifications to a true C is not preferred in Japanese, but there could be modifiers indicating the kinds of nouns. It is
reasonable because the meaning of a C is relatively empty (represents x1 multiplicand), while the meaning of a pure noun is more concrete.
For instance, the morpheme “樹” (meaning “tree”, the spelling is き ki /ki/), which is listed as one of the 673 possible Cs, is not a true C but a noun. As example (32) shows, “樹” can be modified by “松” to denote the type of tree.
(32) 一 松 樹 ichi matsu ki one pine tree
“A pine tree”
These tests could be useful in distinguishing true Cs from pure nouns in Japanese.
3.2.4 Distinguishing true classifiers from units of measurement
The syntactic behaviors of Cs and units of measurement are similar, as shown in
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
62
the following example (33). Both the true C and the unit of measurement fit in the [Num X N] constituent.
(33) a. true C
一 羽 鳥 ichi wa tori
1 C bird 1x1 bird “one bird”
b. unit of measurement 一 斤 鳥
hito kin tori 1x0.6 kilograms birds 1 catty bird
“one catty of birds” (but not necessarily 1 bird)
Still, there is one method to distinguish them, the “C in Numeral Quantification Scope” principle. In example (36b), the unit of measurement “斤” means 0.6
kilograms in weight, so “一斤鳥” may not refer to only one bird., the deletion of “斤”
would change the meaning. This is the main method to distinguish true Cs and units of measurement.
3.2.5 Distinguishing true classifiers from echo classifiers
For echo Cs listed from previous studies, some echo Cs form the [Num C N] [N
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
63
Num C] constituents, while some cannot.
Most of the echo Cs do not pass syntactic tests because they cannot directly attach to both numeral and noun, the only possible combination of using echo Cs should be concordant with genitive marker “の”. See the difference between examples (34) and (35) shows below. Most of the echo Cs can only form a grammatical
combination with “の”, as shown in (34a), but the [Num C N] [N Num C]
constituents in (34b) and (34c) are ungrammatical combination in Japanese.
(34) a. 一 色 の 色 i shoku no iro 1 echo C GEN color “one color”
b. *色 一 色 iro i shoku color 1 echo C c. *一 色 色 i shoku iro 1 echo C color
However, the [Num C N] [N Num C] constituents are grammatical for some echo C, as in (35b) and (35c).
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
64
(35) a. 一 校 の 学校 ik ko no gakko 1 echo C GEN school “one school”
b. 学校 一 校 gakko ik ko school 1 echo C c. 一 校 学校 ik ko gakko 1 echo C school
As the results, the echo C “校” in (35) is a true C while “色” in (34) is not. It is possible that an echo C refers to “kind” reading of the noun. As the following example (36) shows, the echo C “色” does not refer to one color but a kind of color, thus echo Cs that failed the test are actually nouns.
(36) 一 色 の 色 = 一 種 色 の 色 i shoku no iro = i shu iro no iro 1 echo C GEN color = 1 kind color GEN color “one kind of color”
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
65
3.2.6 Classifiers with different spellings
After the syntactic tests have eliminated most of the elements in the possible C list, one 28-year-old female Japanese native was recruited as a reviewer. She has lived in Japan until she was 26, and then she came to Taiwan as an exchange student for two years.
In this step, she was asked to review the list. The reviewer was helpful because she suggested that some Cs repeated in the list due to different spellings. For example, the C “頭” was listed as two Cs “頭/to/” and “頭/kaʃila/”, but “頭” did not spell as /kaʃila/ anymore, so it was is not required to list this C twice. Wydell, Butterworth &
Patterson (1995) suggested that the difference in the spellings of a Japanese character (kangi, 漢字) is due to the influence of both Chinese and Japanese origins. It was the phonological factors that result in the distinct spellings of 頭(tou) and 頭(kashira), but the two possible Cs are actually the same at the semantic and syntax levels.
3.2.7 Frequency in Contemporary Japanese
The true Cs identified are theoretical true Cs, however, the reviewer also
mentioned that some of the Cs are borrowed from ancient Chinese and they are rarely used now. Therefore, one questionnaire is to evaluate the frequency in contemporary Japanese.
However, the purpose of the questionnaire is to do a preliminary analysis on
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
66
frequency, to provide future recommendations in teaching of Japanese as a second language in Taiwan. The main focus of this thesis is to identify the true Cs.
There are 121 questions in the questionnaire. The first four questions are
background information, included their native language, gender, age, and the number years they have lived in Japan. The other questions involve 117 Cs (114 true Cs) with an [Num C N] constituent example. For example, the first C is “尾” with an example of “海老（えび）一尾” (“one shrimp”). There are 5 frequency options for each Cs:
“quite often”, “sometimes”, “ordinary”, “seldom” and “never.” Each reviewer required to scale and select an option for 117 Cs in the questionnaire.
The questionnaire is translated into Japanese (included in the appendix). It was sent to the international center of National ChengChi and TsingHua University, to request their help to transmit to exchange students. Also, it was posted to a Japanese forum called “2ch”, and Taiwan BBS site the “JapanStudy” and “Japan_Living”
forums requesting Japanese informants.
In total, 36 native Japanese informants completed the questionnaire. 28 of them are aged between 21-40 years old, 6 of them are over 40 years old and 2 of them are under 20 years old. Among the informants, 19 are male and 17 are female. Almost all of the informants have lived in Japan for over 20 years (except for the two informants under 20 years old). After the questionnaire is completed by the Japanese native
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
67
speakers, the frequency of each C is compiled.
3.3 Remarks
This chapter describes the methods for identifying true Cs out of the 673 possible Cs. First, corpora and dictionaries are needed to obtain the raw data of natural Japanese sentences, so that grammatical and ungrammatical sentence combinations are collected.
Second, to identify true Cs and eliminate other morphemes, syntactic tests (and sometimes semantics) are included. Pure nouns, units in measurement, measure words, and some echo classifiers can be deleted based on the different syntactic behavior from true Cs. Moreover, the list of Cs is checked by one native speaker to remove repeated Cs with different spellings.
Finally, a questionnaire written in Japanese is created for native speakers to determine whether those Cs are still commonly used. Thus the frequency of each Cs can be preliminarily analyzed.
‧
Results and Discussions
4.1 True Classifiers
There are originally 673 possible Cs collected from previous works, and each of the 673 possible Cs is checked. Based on the previous methodology and syntactic tests, there are only 105 true Cs remaining in the list of Japanese Cs, as shown in Table 8.
‧
Table 8. 105 true Cs that passed the syntactic test and corpus examination Additionally, there are 10 elements that are a mixture of Cs and Ms. So these elements could be a C or a M depending on the context.
Classifier readings Measure word readings 部 bu one book or document a part of
丁 chou a canon a lump of object like tofu
具 gu a brush a set of cloth or a suite
杯 hai a cuttlefish a cup of something
巻 kan rolled-up paper rolled-up movies or documents 口 kuchi a share or bank account a mouthful of something
行 kou a bank lines of words
面 men one flat object an extent of something 垂 tare a hanging object one drop
点 ten a work of art one point drawn paintbrush Table 9. 10 Elements that could be both classifiers and measure words
As shown in the following example (37), “杯” is an M in (37a) , it means “cup of” or “cupful of something”, but in (37b), it is a C denoting “one ship, cuttlefish, or squid”. It could be a C or M depending on the context and the number of NP.
(37)a. コーヒー 一 杯 coohii ichi pai coffee 1 M-cup “one cup of coffee”
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
70
b. 船 一 杯 fune ichi pai ship 1 C “one ship”
After examining each of the 673 morphemes, deleting the syntactic behaviors and semantic, these 115 morphemes could be true Cs in Japanese.
4.1.2 Deleted items
As mentioned in the previous chapter, the 558 deleted morphemes from the 673 possible Cs list are Ms (measure words), pure nouns, units of measurement, and echo Cs.
In addition, some of the morphemes are deleted not due to syntactic behaviors.
First, some of the possible Cs in the list are actually a combination of two or more morphemes; they are phrases but not true Cs. For example, “ヶ所” with the spelling /kasho/ is a combination of true C “ヶ”(generic C) and NP “所”(place), so it would not be counted as another C instead of “ヶ”(generic C). Take “第 日” as another example. “第”(/dai/) denotes the ordinal number, and “日”(/nichi/) is a noun that means “day”, “第 X 日” means “the Xst day.” Thus, it is not a true C either.
Second, a few morphemes do not appear in any search results in the Japanese corpus. Their usage of Cs cannot be determined. Thus, they are deleted as well.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
71
In addition, 73 morphemes are counted more than once in previous studies due to the different spellings. For instance, the character ”具” has three spellings, /gu/(ぐ), /sonae/(そなえ), and /yosoi/(よそい), but only /gu/ is commonly used, /sonae/ and /yosoi/ are seldom used in contemporary Japanese. Also, “本” has two spellings, /hon/(ほん) and /moto/(もと), and only /hon/ is used for C reading. Therefore, /moto/
is not counted as another C. For cases like “通”, the spellings denote different word classes. The spelling of /toori/ (とおり) is for kind reading (M), while the spelling of /tsuu/ (つう) is for C counting letters or documents, so “通” should be counted as
“C/M,” not two separate Cs. Table 10 displays the number of morphemes deleted.
Word Class or Reason for Deletion Deleted Number
Measure Words 116
Nouns 189
Units of Measurement 125
Echo Classifiers 11
Phrases, Combination of different Morphemes 28
Not Used in Japanese Anymore 16
Different Spellings 73
Table 10. 558 Deleted morphemes and the reasons for deletion
Among all the deleted morphemes, nouns and units of measurement can both be
‧
viewed as nouns, and the third is Ms, so a majority of the morphemes that are treated as Cs in previous studies are nouns, and Ms are the second essential word class that could be mistakenly identified as Cs in previous Japanese studies.
4.1.3 Some difficult cases
While performing the tests on identifying true Cs, there are some difficult cases in which much time is spent analyzing whether they are true Cs or not. They are echo Cs and Ms of “pair”.
There are 14 echo Cs listed in previous studies.
echo C 部屋 錠 校 社 食 色 店 字 芸 言 声 誌 舗 園
Table 11. 14 echo Cs listed in previous works
These echo Cs are listed by duplicating a noun itself, so they easily co-occur with the noun they originated from. All of the 14 echo Cs are tested carefully. Based on the results, only 3 echo Cs could combine with numerals and nouns without the
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
73
help of the genitive marker “の”, 11 of them are excluded because they cannot exist alone without the genitive marker “の”.
The other case that is not easy to test during the procedure is “Ms that mean pair of something”. The three items are “足,” “膳,” and “耳”. At first, they are mistakenly identified as true Cs, but they are finally ruled out from the C list.
Their syntactic behaviors are similar to true Cs. The special characteristic is that they count “a pair of something”, and “one pair” is possibly “one integral unit” in Japanese. In example (41), the possible C “足” is used to count how many pairs of shoes. In (38a), “靴一足” means one pair of shoes, that is to say two shoes, though the numeral preceding the possible C “足” is “一” (one), it still counted one integral set of “two shoes”, and “two shoes” is viewed as “one single object”.
(38)a. 靴 一 足 kutsu ichi ashi
shoe 1 possible C-pair “a pair of shoes”
b. 靴 片 足 kutsu kata ashi
shoe part possible C-pair “a shoe (a part of one pair of shoes)”
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
74
Example (38b) supports to the preceding viewpoint. If there is only one shoe, and the element is limited to only the possible C “足 (pair of shoe) ”, then native Japanese speakers would tend to say “one part of ”(片) the two-shoe set. As a result, it seemed that “足” is a true C because it counts a unit of “two shoes.” It is possible that
“one pair of shoes” is counted as “one unit” in Japanese.
Japanese expression Graphical representation of the meaning
“靴一足”
“one pair of shoes”
complete
“靴片足”
“one shoe”
incomplete
Table 12. Concept of “one pair of shoes” and “one shoe” with Japanese “足”
Example (39) shows the usage of “膳” and “耳”, which are similar to “足.”
(39)a. 箸 一 膳 hashi ichi zen
chopstick 1 possible C-pair “a pair of chopsticks (two chopsticks)”
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
75
b. 箸 片 膳 hashi kata zen
chopstick part echo C-pair
“part of two chopsticks (one chopstick)”
c. 兎 一 耳 usagi ichi mimi
rabbit 1 echo C-pair “a pair of rabbits (two rabbits)”
d. 兎 片 耳 8 usagi kata mimi rabbit part echo C-pair “part of two rabbits (one rabbit)”
However, “足,” “膳,” and “耳” are excluded in the final true Japanese C list, because at last one evidence that proves the Japanese concepts in Table 12 is worth considering. The evidence is that: shoes, chopsticks, and rabbits are actually dividable in Japanese.
When it comes to the generic C “つ,” it can be applied to shoes, chopsticks, and
When it comes to the generic C “つ,” it can be applied to shoes, chopsticks, and