From N to Num
3.5 The Spurious NP Ellipsis
3.5.6 The SNPE and the Exhaustive Lexicalization Principle
Let me conclude this section with a few remarks on how the Exhaustive Lexicalization Principle bears on genuine ellipsis and the SNPE. The Exhaustive Lexicalization Principle requires that every feature and terminal in the output of syntax must receive a spellout, even if that spellout is zero. In light of this constraint, it is not possible to treat ellipsis as failure of vocabulary insertion.
This still leaves a variety of options on the table: syntactic or PF-deletion, deployment ofproor another form of null spellout, non-generation of the ellipted constituent, and no doubt others. But at the end of the day, all features in the syntactic output must receive a lexicalization.
For the SNPE, the Exhaustive Lexicalization Principle raises the question of what spells out Cl and Num if the classifier is pushed down ton and N. The null classifier I defended in Section 3.2.2 provides the natural answer here. Suppose that the null classifier can spell out both Cl and Num, like I suggested fordarab in a footnote. Then (145) is lexicalized as in (146).
(145) egy one
(silent cl) z¨old green
fej clhead
‘one green one’
31That classifiers in noun-classifier compounds do not exercise their divisor function receives support from the distribution of such compounds in Mandarin, too. As Zhang (2011, ch. 6.) shows, these compounds "behave the same as regular nouns in the language" (p. 111). They have the same distribution and readings that bare nouns do, and they need a unit-word (classifier, massifier, group classifier) to combine with a numeral. This is unexpected if classifiers deploy their apportioning potential in these compounds.
(146) NumP egy
Num ClP
Cl AdjP
Adj z¨old
nP
n NP
N fej silent classifier
Now I have suggested that the silent classifier has the same feature content as the general classifier darab, both are specified for Cl and Num. Accordingly,darab should also be able to spell out Cl and Num in the SNPE. This prediction is borne out.
(147) egy one
darab clgeneric
z¨old green
fej clhead
‘one green one’
(148) NumP
egy
Num ClP
Cl AdjP
Adj z¨old
nP
n NP
N fej
darab
It is, however, not possible to spell out Cl with the same specific classifier that is pushed down to the noun position. That is, we do not get doubling of the form specific classifier > low adjective
> specific classifier.
(149) *egy one
[ClP fej clhead
[AjdP z¨old green
[N P fej clhead
]]]
‘one green one’
Theoretically, this should be possible. I argued that in non-elliptical DPs classifiers spell out Cl and underassociate theirn and N features. This is what the first instance of the fej is doing in (149). The second instance offej does what classifiers in the SNPE regularly do: underassociates the Cl feature and spells outnand N. No terminal is left without a spellout, and the configurations for the first and the second instance of the classifier are independently attested.
What goes wrong in (149), I suggest, is that it is not as economical a spellout as (145) or (147). Darab and the silent classifier are both a prefect match for the structure in (148), while the first instance offej has two underassociated features and on top of that it cannot spell out Num.
Therefore (149) is ruled out because both the silent classifier anddarab are a better fit.
In this analysis, it needs to be explained why classifiers can appear in the classifier position in the first place. That is, if (150-a) and (150-b) are more economical spellouts than (150-c), then why is (150-c) possible at all?
(150) a. numeral > null Cl > N b. numeral > darab > N c. numeral > specific Cl > N
In Chapter 2 I argued that competition between lexical items arises only if we keep the meaning constant, and discussed how this accounts for the existence of both regular and irregular plural and past tense forms. All the different lexicalization patterns depicted in (150) produce a different shade of meaning, therefore these lexicalizations are not in competition with each other.
The lexicalization of the Cl position is different in the SNPE because N andnare spelled out by a specific classifier. Whatever lexical, conceptual or semantic difference or‘plus’ specific classifiers have with respect to the general or the null classifier, they already contribute that to the meaning in N andn. Therefore when it comes to the lexicalization of Cl, using a specific classifier for this position will not give anything different or ‘extra’. Therefore in this case specific classifiers do compete with and lose to the general classifier and the null classifier.
Let us summarize the results of this section. I argued that classifiers are specified for the features N,nand Div/Sort. They are always spelled out in the highest head they lexicalize in a given structure. In non-elliptical DPs, this is Div/Sort. In the SNPE, on the other hand, classifiers are squeezed down to and spelled out in the noun position. This accounts for why they occur so low, why their co-occurrence restrictions disappear and why the noun cannot be restored in this apparent ellipsis construction.
3.6 Conclusions
This chapter mapped out the functional sequence in the the lower portion of the Hungarian DP, between N and Num. Building on Borer (2005); Svenonius (2008a) and D´ek´any and Csirmaz (2010), I proposed the decomposition in (151).
(151) NumP > AP > ClP > AP > nP > NP
The way these positions are lexicalized in non-elliptical, plain DPs is depicted in (152).
(152) NumP
numeral
Num darab
AdjP
high adjective
Adj ClP
Cl specific Cls
null Cl
AdjP low adjective
Adj nP
n NP
I have also shown, however, that parts of the functional sequence in (151) can be lexicalized in a different way. This leads to a different word order than what is shown in (152). Such changes in word order require a careful investigation to establish what exactly has changed with respect to (152), that is, what remains in the same position and what ends up in a different node.
I proposed that an example of such a different lexicalization happens with classifiers in the SNPE. Classifiers can spell out the range of functional projections from N to Cl, but in the SNPE
they are pushed down tonand N. This has an effect both on the word order and the distribution of classifiers. The analysis sheds light on how the Superset Principle can derive that one lexical item can occur in various positions (when those positions are unrelated by movement) and how the same lexical item can have different properties in the different positions. The Superset Principle thus provides an insight into the lexicalization problem.
CHAPTER3.FROMNTONUM
3.7 Appendix I
classifier nominal gloss shapes/types of objects referred to nouns modified bokor ‘bush’ plants with the shape of a shrub potato, raspberry, rose
cikk ‘article/item’ 3-D, crescent shaped garlic, orange, grapefruit, mandarin
cser´ep ‘pot’ flower, plant plant
cs´ık ‘strip, line’ chewing gum chewing gum
cs˝o ‘tube’ long, hollow/unusable middle sweet corn, green/red pepper or paprika
darab ‘piece’ neutral w.r.t shape or size any count noun
fej ‘head’ big spherical cabbage, onion, lettuce, kohlrabi, cauliflower
f˝o ‘head’ people people in regimented situations, eg. gyalogos‘infantryman’,
leg´enys´eg ‘crew’
gerezd ‘clove’ 3-D, crescent shaped garlic, orange, grapefruit, mandarin
karika ‘ring’ flat round sausage
k¨otet ‘volume’ bound paper book or other bound volume
´ıv ‘sheet’ paper paper
r´ozsa ‘rose’ floret cauliflower, broccoli
r´ud ‘stick/rod’ long cylindrical chitterlings and salami, vanilla
(in part synonym ofsz´al)
sz´al ‘thread’ long cylindrical hair, fur, match, sausage, frankfurter, salami, welt, flower,
candle, (green) onion, carrot, cigarette, cigar, grass, plank szem ‘eye’ small spherical grape, tomato, berry (and all types of berries), nuts (all
types), corn, sweet corn, potato, pepper, biscuit, pill, pearl, sand
t˝o ‘stem’ plant with a stem grape, rose, nursling/plant, any specific type of plant that
has a nursling
vekni ‘loaf’ bread bread
Table 3.1: Hungarian classifiers