LINGUIST List 12.891

Thu Mar 29 2001

Disc: New: Review of Packard

Editor for this issue: Karen Milligan <>


  1. Jerry Packard, Author's response to LINGUIST List 12.844 Review of Packard

Message 1: Author's response to LINGUIST List 12.844 Review of Packard

Date: Wed, 28 Mar 2001 18:35:43 -0600
From: Jerry Packard <>
Subject: Author's response to LINGUIST List 12.844 Review of Packard

Author's response to
LINGUIST List 12.844 Review of Packard 2000 The Morphology of Chinese (K.

> Packard's terminology overall is clear and functional.
>There are, however, a few frequently-used terms which are not
>explicitly defined, the most important one being 'gestalt word'
>(Chapter 4 is entitled "Gestalt Chinese Words"). It becomes
>clear in the course of the book that this is essentially what a
>Chinese would call a _ci2_, a (possibly polysyllabic) 'word',
>as opposed to a _zi4_, or single-syllable morpheme
>corresponding to a written Chinese character. I have not seen
>the term 'gestalt word' used elsewhere, and found it a bit hard
>to get used to. A clear definition of what it is early in the book
>would have been appreciated.

Many readers have reacted the same way; by 'gestalt word' I just meant
to focus on the complete, compiled word. Perhaps I should have just
said 'compiled word', or simply left off the word 'gestalt', since it
seems to have been a source of confusion.

> The distinction between free vs. bound morphemes features
>prominently in Packard's classifications.I wonder if it is necessary to
>up so many categories based merely on whether a morpheme
>is bound or not. It seems to me that a high price has been paid in
>terms of numbers of categories if the main goal of distinguishing
>free 'root words' from 'bound roots', plus word-forming affixes
>and grammatical affixes, is to state that a free root word is the
>only element in the system that can expand and allows recursion.

To me, the free-bound distinction is a critical system property. Root
words can be used as free words and as 'expansion nodes' to build
larger words. Bound roots are listed in the lexicon, but can't be
retrieved as free forms for use in sentences, and can't serve as
'expansion nodes' for the construction of larger words.

> Furthermore, it is often difficult to say definitively whether a
>morpheme is free or bound, since there is a good deal of overlap
>and numerous ambiguous cases. Packard addresses this issue in
>part on p. 68, saying that a morpheme is classified as 'free' or
>'bound' *according to a specific usage*, and not all possible
>meanings of a morpheme. However, the 'free' or 'bound' nature
>of a morpheme simply is not that clear-cut in Chinese, and it also
>varies over time and according to geographical area.

I spend a lot of time defending this position (pp. 67-69) : basically
I say that if what appears to be a single morpheme has both bound and
free usages, that there is a separate mental lexicon entry for each
one---you can call it 'two morphemes'---quite independent of whether
they are associated with the same orthographic form (Chinese
character) or not. In other words, native speakers know whether an
entry in the mental lexicon can be used as a free form or not. I agree
that the bound-free status of Chinese morphemes varies according to
time, geography, dialect, register and among individuals. That is
precisely the point. My claim is that for a given morpheme in the
mental lexicon of a given individual, there is no bound-free
ambiguity, and the Chinese language is not less clear-cut than other
languages in the bound-free identity of its morphemes. 

> The large number of categories based on 'free' and 'bound'
>properties seems to have come at the expense of leaving out
>other important information about the words. As regards verbs,
>for example, something I expected to find in Packard's system
>and did not was a classification of action, copular and purely
>stative verbs; and, for the action verbs, whether they are
>transitive or intransitive - all of which seem to me to be more
>fundamental and useful distinctions than whether the morpheme
>elements are free or bound.

I did not discuss general properties of verbs or other lexical items
in the book, except to the extent that those properties are derived
from the properties of word constituents.

> This brings us to Packard's Headedness Principle, which states
>that '(bisyllabic) noun words have nominal constituents on the
>right and verb words have verbal constituents on the left.' (p. 39).
>I find here a discrepancy between the name of the principle and
>its content. The first problem is that we do not seem to agree on
>what a 'head' is. Packard defines a 'canonical head' as "a function
>of the form class of a word...: verbs have their canonical head
>on the left, and nouns have their canonical head on the right."
>(p. 194). I propose, rather, that a 'head' (1) is the morpheme
>which corresponds to the form class (i.e. part of speech) of the
>word or construction as a whole; (2) is the superordinate
>element modified by the subordinate element(s) in the formation,
>if there is/are such; and (3) may or may not appear in the surface
>form of a word or construction.'

The definition provided by the reviewer is a time-honored traditional
one; I wanted to provide a broader definition that might help explain
some of the other properties of complex Chinese words.

> Saying, as Packard does, that a verbal formation tends to
>have a verbal element on the left is not the same as saying
>that it is left-headed. If the verb on the left is followed by a
>noun, the relationship is normally verb + object and not
>verb + modifying material, thus no head can be identified.

Why not? Especially if the V-O form is lexicalized, if there is a
category match between the word and the left-hand constituent, this is
the classic definition of head.

>If the second element is also a verb, the relationship is likely
>to be (though not always) a _coordinate_ one, in which both
>elements perform the same verbal function on a basis of
>equality, as in the case of coordinate formations like
>_xi3huan1_ [to feel joy + to feel pleasure] 'to like', and neither
>can be said to modify or be subordinate to the other. This kind
>of formation is usually considered 'headless'; or one could say
>_both_ elements are heads.

If both elements can be defined as head, it kind of reduces the
utility of the notion. 

> Packard seems to be partly influenced by the occurrence 
>of verbs on the left in VO type constructions, such as 
>_chu1ban3_ [to put out + edition] 'to publish'. We will set 
>aside for the moment whether these constructions are 'words' 
>(and I think they are); the problem with these items is that 
>they require a separate analysis from pure transitive verbs
>with no object element, because they are derived from phrases complete
>with objects, and are naturally going to have the verb 
>on the left; Chinese is basically an SVO language. (3) This does 
>_not_ mean that objectless transitive verbs usually have their 
>head on the left - and indeed they do not; they are generally
>right-headed or headless. VO formations are a red herring 
>that detracts away from the true nature of headedness in

Although there are a significant number of exceptions (which the
language tends to treat exceptionally), there is little grammatical
motivation for saying the head is on the right in a two-syllable
Mandarin verb, and plenty of motivation for saying it is on the
left. For example (1) resultative verbs 'inflect' on the left (eg,
zuo-wan � zuode-wan (2) A-not-A question inflection 'inflects' on
the left (xihuan � xi-bu-xihuan (3) aspect marking occurs on the
left, even when the left member is a verbalized nominal element (eg,
qiangbi gun-kill 'to execute by gunfire' � qiang-le-by"
gun-ASP-kill 'to have executed by gunfire'; or even a novel
code-switching example I heard a couple of days ago: good-le bye 'to
have said good-bye')y"20

> The 'verbs have their head on the left' rule also means that
>Packard needs to treat adverb + verb formations, of which there
>are an enormous number in Chinese, as 'exceptions' to his
>Headedness Principle. The verbal head of adverb + verb
>formations such as _mo4du2_ [silent + to read] 'to read
>silently', _gong1ren4_ [publicly + to admit] 'to acknowledge
>publicly', and _lian2shu3_ [jointly + to sign] 'to jointly sign
>(e.g. a petition)' , is on the _right_; and this is a widespread
>pattern rather than a 'relatively small' set of exceptions (p. 63).
>Chinese is a dyed-in-the-wool right-headed language in all
>parts of speech, with few exceptions - when, that is, there is
>a head at all. It would seem that Packard's 'Headedness Principle'
>requires rethinking.

ok, but I think that calling Chinese right-headed because of its
general modifier-modified structure might cause us to miss important
generalizations we can make about the function of the left-hand
constituent in the case of complex verbs.

>p. 196-215 give numerous examples - many rather creative -
>of X-bar analysis of English words, which I found very
>interesting in its own right. But is this amount of space for
>English examples justified in a book on Chinese morphology?

I just wanted to make the point that the proposed system for Mandarin
also happens to account for English complex word formation data.

> Occasional 'words from our sponsor' were a bit jarring
>to this non-formalist, e.g. on p. 19: '...the word is
>biologically hard-wired and psychologically real, and
>has a tendency in natural language to 'weaken' the status
>of individual component morphemes, undermining their
>ability to function as free forms.' There are those of us
>who think that the extension of the 'hard-wiring'
>computer metaphor brings inappropriate analogies into
>our understanding of how the brain acquires, stores,
>and processes linguistic and other information.
>Fortunately remarks like these are only made in passing
>and one's personal views on UG theory don't need to
>enter into one's evaluation of the study as a whole.

Though some may see the analogy as inappropriate, I think we can
mostly agree that one of the things we are interested in knowing about
language is how much of it is nature/inherited/'hard-wired' and how
much is nurture/learned/'not hard-wired'

>(2) In my view, _cai3_ is in fact the structural or
>syntactic head of this compound, giving a breakdown
>of something like 'colors (made) of clouds'. But users
>of Chinese may vary in their analysis of formations
>like this.

I agree that it's the structural head; what I'm saying is that it's
not the semantic head. Although it's easy enough to think of it as
'color' metaphorically, I think most native Mandarin speakers would
say if asked that _yuncai_ cloud-color 'cloud' means _yun_ 'cloud'
more than it means _cai_ 'color'.

>(3) In the case of verbs like _dong4yuan2_
>[move + personnel] 'to mobilize', the N object has been
>reanalyzed as a verb element, and the verb can take a
>further object; but this in my opinion does not make
>_dong4_ the head of construction. In its original
>incarnation, _dong4yuan2_ (similar formations likewise)
>is a verb + noun object, and thus does not have a head. It
>is a fossilized expression in its reanalysis as a verb; and
>therefore it does not make sense to say that the first
>element is now the head. The whole has become verbal;
>_yuan2_ is denominalized and now part of the verbal
>whole, i.e. it is no longer a true object or noun. This may
>be in conflict with Packard's principle (p. 27) that 'the
>identity of a gestalt word as a verb never results in the
>_right_-hand member being reanalyzed as a verb.'

But there is no basis for saying that 'the N object has been
reanalyzed as a verb element', except by fiat since it is occurring
within a verb word: _yuan2_ 'personnel' never behaves in verb-like
fashion, and is never treated like a verb at all by native

I do appreciate the close, careful reading and insightful comments
given both by Karen Steffen Chung and by the earlier reviewer Richard

Jerry Packard
University of Illinois
Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue