Editing Chinese classifier (section)

===Categories and prototypes===
While mass-classifiers do not necessarily bear any semantic relationship to the noun with which they are used (e.g. ''box'' and ''book'' are not related in meaning, but one can still say "a box of books"), count-classifiers do.<ref name=ChengSybesma/> The precise nature of that relationship, however, is not certain, since there is so much variability in how objects may be organized and categorized by classifiers. Accounts of the semantic relationship may be grouped loosely into ''categorical'' theories, which propose that count-classifiers are matched to objects solely on the basis of inherent features of those objects (such as length or size), and ''prototypical'' theories, which propose that people learn to match a count-classifier to a specific prototypical object and to other objects that are like that prototype.<ref>{{Harvnb|Tai|1994|pp=3–5}}; {{Harvnb|Ahrens|1994|pp=208–12}}</ref>

The categorical, "[[Categories (Aristotle)|classical]]"<ref name=Tai3Ahrens209-10>{{Harvnb|Tai|1994|p=3}}; {{Harvnb|Ahrens|1994|pp=209–10}}</ref> view of classifiers was that each classifier represents a category with a set of conditions; for example, the classifier {{lang|zh|条}}&nbsp;({{lang|zh-Hant-TW|2=條}})&nbsp;''tiáo'' would represent a category defined as all objects that meet the conditions of being long, thin, and one-dimensional—and nouns using that classifier must fit all the conditions with which the category is associated.  Some common semantic categories into which count-classifiers have been claimed to organize nouns include the categories of shape (long, flat, or round), size (large or small), consistency (soft or hard), [[animacy]] (human, animal, or object),<ref name=Tai5Allan>{{Harvnb|Tai|1994|p=5}}; {{Harvnb|Allan|1977}}</ref> and function (tools, vehicles, machines, etc.).<ref name=Hu1>{{Harvnb|Hu|1993|p=1}}</ref>

{{multiple image
 | footer    = James Tai and Wang Lianqing found that the horse classifier {{lang|zh|匹}}&nbsp;''pǐ'' is sometimes used for [[mule]]s and [[camel]]s, but rarely for the less "horse-like" [[donkey]]s, suggesting that the choice of classifiers is influenced by prototypal closeness.<ref name=Tai12/>
 | image1    = Frecklesmule.jpg
 | width1    = 143
 | caption1  = A mule<br>骡子, ''luózi''
 | image2    = Donkey 1 arp 750px.jpg
 | width2    = 150
 | caption2  = A donkey<br>驴子, lǘzi
 | align     = right
 | caption_align = center
}}

On the other hand, proponents of [[prototype theory]] propose that count-classifiers may not have innate definitions, but are associated with a noun that is prototypical of that category, and nouns that have a "family resemblance" with the prototype noun will want to use the same classifier.<ref group=note>The theory described in {{Harvtxt|Ahrens|1994}} and {{Harvtxt|Wang|1994}} is also referred to within those works as a "prototype" theory, but differs somewhat from the version of prototype theory described here; rather than claiming that individual prototypes are the ''source'' for classifier meanings, these authors believe that classifiers still are based on categories with features, but that the categories have many features, and "prototypes" are words that have all the features of that category whereas other words in the category only have some features.  In other words, "there are core and marginal members of a category.... a member of a category does not necessarily possess all the properties of that category" {{Harv|Wang|1994|p=8}}. For instance, the classifier {{lang|zh|棵}}&nbsp;''kē'' is used for the category of trees, which may have features such as "has a trunk", "has leaves", and "has branches", "is deciduous"; maple trees would be prototypes of the category, since they have all these features, whereas palm trees only have a trunk and leaves and thus are not prototypical {{Harv|Ahrens|1994|pp=211–12}}.</ref> For example, ''horse'' in Chinese uses the classifier {{lang|zh|匹}}&nbsp;''pǐ'', as in:
{{fs interlinear|lang=zh|indent=2|三 {{uline|匹}} 马|sān {{uline|pǐ}} mǎ|"three horses"}}
In modern Chinese the word {{lang|zh|匹}} has no meaning. Nevertheless, nouns denoting animals that look like horses will often also use this same classifier, and native speakers have been found to be more likely to use the classifier {{lang|zh|匹}} the closer an animal looks to a horse.<ref name=Tai12>{{Harvnb|Tai|1994|p=12}}</ref> Furthermore, words that do not meet the "criteria" of a semantic category may still use that category because of their [[association (psychology)|association]] with a prototype. For example, the classifier {{lang|zh|颗}}&nbsp;({{lang|zh-Hant-TW|2=顆}})&nbsp;''kē'' is used for small round items, as in:
{{fs interlinear|lang=zh|indent=2|一 {{uline|颗}} 子弹|yì {{uline|kē}} zǐdàn|"one bullet"}}
When words like {{lang|zh|原子弹}}&nbsp;(''yuánzǐdàn'', "atomic bomb") were later introduced into the language they also used this classifier (颗 [顆] kē), even though they are not small and round—therefore, their classifier must have been assigned because of the words' association with the word for bullet, which acted as a "prototype".<ref name=Zhang46-47>{{Harvnb|Zhang|2007|pp=46–47}}</ref> This is an example of "generalization" from prototypes: Erbaugh has proposed that when children learn count-classifiers, they go through stages, first learning a classifier-noun pair only, such as
{{fs interlinear|lang=zh|indent=2|{{uline|条}} 鱼|{{uline|tiáo}} yú|{{uline|CL}} fish|}}
then using that classifier with multiple nouns that are similar to the prototype (such as other types of fish), then finally using that set of nouns to generalize a semantic feature associated with the classifier (such as length and flexibility) so that the classifier can then be used with new words that the person encounters.<ref name=Erbaugh415Hu>{{Harvnb|Erbaugh|1986|p=415}}</ref>

Some classifier-noun pairings are arbitrary, or at least appear to modern speakers to have no semantic motivation.<ref name=Hu1Tai13Zhang55-6>{{Harvnb|Hu|1993|p=1}}; {{Harvnb|Tai|1994|p=13}}; {{Harvnb|Zhang|2007|pp=55–56}}</ref> For instance, the classifier {{lang|zh|部}}&nbsp;''bù'' may be used for movies and novels, but also for cars<ref name=Zhang55-56>{{Harvnb|Zhang|2007|pp=55–56}}</ref> and telephones.<ref name=GaoMalt1134>{{Harvnb|Gao|Malt|2009|p=1134}}</ref> Some of this arbitrariness may be due to what linguist [[James Tai]] refers to as "fossilization", whereby a count-classifier loses its meaning through historical changes but remains paired with some nouns. For example, the classifier {{lang|zh|匹}}&nbsp;''pǐ'' used for horses is meaningless today, but in [[Classical Chinese]] may have referred to a "team of two horses",<ref name=Morev79>{{Harvnb|Morev|2000|p=79}}</ref> a pair of horse skeletons,<ref name=Wang172-3>{{Harvnb|Wang|1994|pp=172–73}}</ref> or the pairing between man and horse.<ref name=Tai15>{{Harvnb|Tai|1994|p=15, note 7}}</ref><ref group=note>The apparent disagreement between the definitions provided by different authors may reflect different uses of these words in different time periods. It is well-attested that many classifiers underwent frequent changes of meaning throughout history ({{Harvnb|Wang|1994}}; {{Harvnb|Erbaugh|1986|pp=426–31}}; {{Harvnb|Ahrens|1994|pp=205–206}}), so {{lang|zh|匹}}&nbsp;''pǐ'' may have had all these meanings at different points in history.</ref> Arbitrariness may also arise when a classifier is [[Loanword|borrowed]], along with its noun, from a dialect in which it has a clear meaning to one in which it does not.<ref name=Tai13>{{Harvnb|Tai|1994|p=13}}</ref> In both these cases, the use of the classifier is remembered more by association with certain "prototypical" nouns (such as ''horse'') rather than by understanding of semantic categories, and thus arbitrariness has been used as an argument in favor of the prototype theory of classifiers.<ref name=Tai13/> Gao and Malt propose that both the category and prototype theories are correct: in their conception, some classifiers constitute "well-defined categories", others make "prototype categories", and still others are relatively arbitrary.<ref name=GaoMalt1133-1134>{{Harvnb|Gao|Malt|2009|pp=1133–4}}</ref>