3.3 Level of modification and adjective ordering
3.3.3 Subsective adjectives
That subsective adjectives are gradable is evident from the examples in (46), where a size and a quality adjective are both modified by a degree adverb. In addition, subsective adjectives can take a complement as shown in the Greek examples in (47).
(46) a. Jad wrote an [[extremely long] letter].
b. Barb was wearing a [[very nice] skirt] on her graduation day.
(47) a. i the
konti short
Ja for
ti the
maria Maria
fusta skirt
‘the skirt short for Maria’
b. i the
kaces bad
Ja for
tin the
cikloforia circulation
trofes foodstuffs
‘foodstuffs bad for circulation’
These properties of subsective adjectives indicate that they have a phrasal na-ture. What is relevant to the current discussion, however, is that there seem to be restrictions in the ordering of subsective adjectives when these are modified, or when they appear with a complement. In particular, in languages with prenom-inal adjectives what we find is that adjectives which have a phrasal character
have to appear to the left of any non-modified adjectives.15 The examples in (49) show precisely that. The unmarked order of the two adjectives is big ≻ old, where the size adjective precedes the age adjective. If old is modified, however, then it needs to appear to the left of big.16 In order for very old to appear to the right of big there must be an obligatory intonational break between the two adjectives, something which is typical of adjectives in parallel modification. The break is signified by a vertical line in the examples. Moreover, big does not seem to scope over very old when it surfaces to its left, a property again associated with adjectives in parallel modification. The same effects are also observed in (50) and (51).
(49) a. the big old house b. #the old big house c. the very old big house d. the bigŠ very old house (50) a. a beautiful fat cat
b. #a fat beautiful cat c. a very fat beautiful cat d. a beautifulŠ very fat cat (51) Greek
a. to the
DiasceDastiko fun
paLo old
pexniDi game b. #to
the paLo old
DiasceDastiko fun
pexniDi game c. to
the poli very
paLo old
DiasceDastiko fun
pexniDi game
15The reason the discussion is focussed on languages with prenominal adjectives is because it is easier to identify the ordering patterns, as the prenominal order is the only order that is unique across languages. I also follow Cinque (2009) in assuming that the prenominal order corresponds to the underlying structure.
16The noncanonical ordering does not seem to be a consequence of phonological heaviness.
If this was the case, then we would expect Transylvanian in example (48) to always appear to the left of old as it is heavier, even when old is modified.
(48) a. an old Transylvanian cow b. #a Transylvanian old cow c. a very old Transylvanian cow d. *a Transylvanian very old cow
d. to
This phenomenon is not restricted to the ordering of two subsective adjectives. If we modify an intersective adjective with very, which according to what I claimed earlier would cause the adjective to have a relative reading, then the intersective adjective will have to surface to the left of the non-modified subsective adjective:
(52) a. Nee bought an expensive red shirt.
b.#?Nee bought a red expensive shirt.
c. Nee bought a very red expensive shirt.
d. Nee bought an expensive Š very red shirt.
Analogously, when an adjective comes with a complement, the A+Complement will again appear before any other adjectives in the phrase, even if this violates the canonical ordering of adjectives. In the Greek examples in (53) it is evident that even though the canonical order is ‘beautiful’ ≻ ‘tall’, when ‘tall’ appears with a complement, it has to appear to the left of ‘beautiful’. The reverse order, which is given in (53d), is degraded, even if there is an intonational break between
‘beautiful’ and ‘tall’.
A similar pattern is also witnessed in Polish, which also accepts adjectives with complements prenominally. While ‘big’ must appear to the left of ‘old’ in ac-cordance with the universal order, in (54) we see that when ‘old’ appears with a complement, the most natural order is ‘old’+Compl to the left of ‘big’. The orders in (54b) and (54c) are acceptable only with special intonation.
(54) a. kupilam
c. kupilam bought.1sg
(ten) this
duzy big
dom house
Š stary old
jak as
na for
to this
miasto city
The above data suggest that subsective adjectives which surface in the strict uni-versal order do not have a phrasal character. In other words, subsective adjectives which are head-modifiers, are strictly direct modification adjectives. The question that arises is whether all adjectives in direct modification are head-modifiers, or if phrase-modifiers also have access to the direct source. The only adjectives that we have not considered in the discussion so far are intensional adjectives which, nevertheless, are often treated as heads in the literature of adjectival modification (Bernstein 1993; Zamparelli 1993; Alexiadou et al. 2007, and works cited there).
This is due to the fact that they can never be modified:
(55) a. *the very former president b. *the somewhat alleged thief
c. *the president former for her country d. *the thief alleged for his crimes
As a result, intensional adjectives also comply with the idea that when adjectives modify the noun as heads, they have a direct source as shown in (56).17
(56) DP
. . .
AP Deg+A
A+Complement A0intensional
A0subsective
A0intersective N direct modification
The question we need to answer at this point is whether the phrasal adjectives which appear above A0s can also be direct modifiers or if they only have an indirect source. Data from Mandarin Chinese suggest that phrasal adjectives are only found as indirect modifiers. In (57) we notice that Mandarin follows
17I am assuming that intensional adjectives are the highest in the direct modification hierar-chy. My reasons for this are given in section 3.5.
the same pattern as English, Greek and Polish; when adjectives do not have any phrasal properties (i.e. when they are not modified), they must follow the universal order where a subsective adjective appears to the left of an intersective one. When the intersective adjective is modified by a degree adverb, however, then it obligatorily surfaces to the left of the non-modified subsective adjective.
(57) Mandarin
Modif. Intersective ≻ Subsective d. *xiˇao
*Subsective ≻ Modif. Intersective
Where Mandarin differs from other languages with prenominal adjectives is that the noncanonical position is morphologically marked with the particle de. As we saw earlier in this chapter, Sproat and Shih (1991) argue that adjectives which appear with the de particle have an indirect modification source and are generated inside a RRC. The Mandarin data, consequently, suggest that phrasal adjectives only have an indirect source.18
An additional point worth noting is that the example in (57d) becomes gram-matical if xiˇao is also accompanied by de, as in (58a). In this case, however, the order between the adjectives is free, regardless of whether these are bare or not. This becomes evident when we compare (58a) to (58b). The freedom in this instance is unsurprising as both the subsective and intersective adjectives have an indirect source, and as has already been discussed, indirect modification does not obey any ordering restrictions.
(58) a. xiˇao
18Paul (2005) argues against analysing modifiers with de as instances of adjectives in a RRC, and he instead claims that de splits the nominal phrase into two syntactico-semantic domains.
A de-less modifier, is in the lower domain and is interpreted as a defining characteristic, while a modifier above de is interpreted as an additional, secondary property (Paul 2005:770). Even if Paul is right in assuming that adjectives with de are not merged inside an RRC, what is relevant to us is that he notes that de-less adjectives pattern with head-modifiers, while adjectives above de have a phrasal character.
b. hˇen
Earlier in the chapter we saw that Greek is another language that marks the direct vs. indirect distinction overtly. In contrast with Mandarin, in Greek it is possible to find a phrasal adjective both in direct and indirect modification. The former source is associated with monodefinite constructions and the latter with polydefinites, where the indirect modifier appears with its own definite article.
The fact that the adjective ‘bad’ has a phrasal character in the following exam-ples is confirmed by the fact that it can appear with the complement ‘for the circulation’:
(59) Monodefinites – Direct modification a. i
‘foodstuffs bad for circulation’
b. i
‘tasty foodstuffs bad for circulation’
(60) Polydefinites – Indirect modification a. i
‘foodstuffs bad for circulation’
b. i
‘tasty foodstuffs bad for circulation’
Drawing from the above discussion, I conclude that direct modification is not restricted to head modification. As the Greek data suggest, phrasal adjectives can also have a direct modification source. However, phrasal adjectives must appear higher than head adjectives and, as a consequence of this, adjective ordering restrictions are lifted when one adjective in the phrase modifies the noun as a head, and another as a phrase.19 An example of this was shown in (53), which is repeated below, and is structurally represented in (61).
19Semantically, it could be that adjectives found below phrasal modification denote a prop-erty, and it is at the phrasal level that a property turns into a predicate. Sadler and Arnold (1994:195) also suggest this, although for them all direct modification adjectives are heads, while indirect modification adjectives are phrasal. I will not be going into this in any more detail, but if this assumption holds then it is unsurprising that we find differences in the interpretation of adjectives depending on whether they modify the noun as heads or phrases.
(53) a. to
psilo Ja tin ilicia tu
‘tall for his age’
A0
To summarise this section, I proposed that adjectives in direct modification can either be head-modifiers or phrasal-modifiers. While ordering restrictions are lifted if one adjective in direct modification is an AP and another just an A0, I assume that the ordering restrictions are still observed within head-modifiers and phrasal-modifiers. A question we have not touched upon is what the implications are for languages that allow postnominal adjectives if we are to assume two types of direct adjectival modification. For languages which follow the mirror image order we expect to observe the same constraints that hold in prenominal position, but in the mirror image order. The data from Hebrew and Lebanese Arabic in (62) and (63) corroborate this idea; the modified adjective, which is marked in bold, is found further away from the noun than the non-modified adjective. That the adjectives are found in the mirror image order is evident from the English translations.
(62) Hebrew
‘I bought the very old big house.’
b. ??kaniti
‘I bought the very old big house.’
c. kaniti
‘I bought the very big old house.’
d. ??kaniti
‘I bought the very big old house.’
(63) Lebanese Arabic
‘I bought the very old big house.’
b. ??Street
‘I bought the very old big house.’
c. Street
‘I bought the very big old house.’
d. ??Street
‘I bought the very big old house.’
Under an analysis where the mirror image order is base generated as in Abels and Neeleman (2012), the adjectives will observe the hierarchy found prenominally, i.e. AP > A0 > N. The difference is that the structure is right branched as shown below:20
(64) DP
N A0 AP D0
If the mirror image is derived via movement, which is what Cinque (2010) as-sumes, then the order will have to be derived via head-movement and subsequent
20Abels and Neeleman’s analysis is discussed in more detail in section 5.6.1.2, chapter 5.
roll-up movement. In the structure in (65) we see the first part of the derivation;
the noun adjoins to its closest A0creating a new complex head, and the new com-plex head subsequently moves to the next A0 in the structure. This is repeated if there are more head-adjectives in the structure. If, as Kayne (1994) proposes, adjunction of the moved head is to the left of the target head, then the order of the adjectives will be the mirror image of the prenominal order as shown below:
(65) DP
AP4
AP3 AP2
A02 A01
N0 A01 A02
AP1 A01 N0 A01
NP N0
The next step involves roll-up movement. The phrase that contains the noun and the head adjectives (i.e. AP2) moves above the nearest phrasal adjective to the Spec of some XP as shown in (66a).21 Subsequently, the whole XP moves above the next AP in the structure as shown in (66b). Again, if there are more APs in the structure, then the whole phrase will cyclically roll-up above each AP.
(66) a. DP
AP4 XP
AP2
N0A01 A02
AP3 AP2
N0A01 A02
21The XP is an AgrP in Cinque’s (2010) analysis. More details of this are given in section 5.6.1.1, chapter 5.
b. DP
YP
XP N0A01 A02AP3
AP4 XP
N0A01 A02 AP3
If adjunction of a moved head is always to the left of another head, then head movement cannot derive the postnominal order where the adjectives stay in the universal order. As a result, the prediction would be that postnominal adjectives in languages like Welsh and Irish, which follow the universal order, modify the noun only as phrases. However, more research needs to be done in order to establish whether this prediction holds, and to determine whether there are any further ramifications that follow from the mirror image derivations I have sketched out above.22