Cinque (2010) treats prenominal adjectives in languages like Bosnian/Croatian/Serbian (BCS) and English either as APs in the specifiers of functional projections in the extended domain of N (direct modification) or as predicates in reduced relative clauses in the specifiers of functional projections (indirect modification), while postnominal adjectives in these languages are treated only as predicates in reduced relative clauses. In this model, the direct-modification source is associated with one set of interpretations and the indirect-modification source with a different set of interpretations. This proposed dual-source analysis raises the question whether there are any languages that mark this distinction with overt morphology. In this respect, Cinque considers languages, among them BCS, where adnominal modification has two distinct morphological shapes. Crucially, BCS adjectives have different morphological forms depending on their distribution: the “long form” and the “short form.” Cinque treats the two forms as being overt manifestations of the two sources of modification. In particular, he suggests that adnominal short-form adjectives are always reduced relative clauses (indirect modification), while long-form adjectives are ambiguous between the two sources. This proposal captures some distributional and ordering restrictions these two forms of BCS adjectives display.
In this squib, I provide several empirical arguments against the claim that BCS is a language that overtly distinguishes between two sources of modification. By closely inspecting the distribution and interpretations of BCS adjectives, the ordering between the two forms in the prenominal position, and extraction possibilities out of APs, I show that both long- and short-form adjectives represent challenges for Cinque’s model. I argue that treating BCS long adjectives, as well as prenominal short adjectives, as reduced relative clauses is problematic. More generally, my investigation of interpretations that adnominal adjectives in BCS receive, including several parallels between BCS and English, leads to the conclusion that, although there may be two syntactic sources of adnominal adjectives, these different sources are not identified by two distinct sets of interpretations associated with them. I also introduce facts that are problematic for the claim that the indirect-modification source involves reduced relative clauses.
The squib is organized as follows. In section 1, I introduce Cinque’s (2010) dual-modification-source model and its predictions. In section 2, I explore how the predictions of this model fare with respect to BCS adjectives. In sections 2.1 and 2.2, I discuss data regarding BCS long and short adjectives, respectively, that are problematic for the dual-source analysis. Section 3 is the conclusion.
1 The Dual-Modification-Source Hypothesis
Cinque (2010) proposes that adnominal adjectives have two sources—a proposal based on, among other things, the interpretation of adjectives within nominal phrases. Under his view, adnominal adjectives enter the structure either as APs in the specifiers of designated functional projections (FPs) in the extended domain of N or as predicates in reduced relative clauses in the specifiers of FPs.
Regarding the interpretation of adnominal adjectives, it has been known since Bolinger 1967 that prenominal and postnominal adjectives in English differ in their interpretation. Thus, the prenominal adjective in (1a) is ambiguous and can be understood as denoting a permanent (intrinsic) or temporary (episodic) property. (Svenonius (1994) and Larson and Marušič (2004) note that the distinction between permanent and temporary property corresponds to Carlson’s (1977) individual-level/stage-level distinction.) The postnominal adjective in (1b), on the other hand, is not ambiguous and can only receive the temporary-property reading.
Bolinger (1967) also notes that prenominal and postnominal adjectives differ regarding the availability of restrictive and nonrestrictive interpretations. The prenominal adjective in (2a) can be understood as either restrictive or nonrestrictive, while the postnominal one in (2b) receives only a restrictive reading.
Every unsuitable word was deleted.
‘Every word was deleted; they were all unsuitable.’
‘Every word that was unsuitable was deleted.’
Every word unsuitable was deleted.
#‘Every word was deleted; they were all unsuitable.’
‘Every word that was unsuitable was deleted.’
Further, Cinque (2010) notes that in English DPs with multiple prenominal adjectives, readings available only to prenominal adjectives (permanent, nonrestrictive) and the ones available to both postnominal and prenominal adjectives (temporary, restrictive) can cooccur. In such cases, adjectives with interpretations available to both prenominal and postnominal adjectives must precede adjectives with prenominal-only readings. This is shown in (3a), where the boldfaced adjective is interpreted as denoting a temporary property and the nonboldfaced one is interpreted as denoting a permanent property, and in (3b), where the boldfaced adjective is interpreted as restrictive and the nonboldfaced adjective as nonrestrictive.
Cinque also observes that in some languages adjectives with permanent or nonrestrictive interpretations have more rigid ordering within the nominal phrase than those with temporary or restrictive interpretations (see also Sproat and Shih 1988, 1990). However, this ordering sometimes appears to be a mere preference or just an unmarked order, as is the case in English.
Cinque captures these interpretive properties of English adjectives and their order in the prenominal position by proposing a dual-source analysis for adnominal modification. One source is direct modification, which he assumes involves merger of APs in the specifiers of FPs in the extended domain of N. These adjectives are real attributive adjectives and receive interpretations that are not found in the postnominal position in English (permanent, nonrestrictive, nonintersective, etc.). The second source is indirect modification (Sproat and Shih 1988, 1990), which Cinque takes to be (reduced) relative clauses merged in the specifiers of projections higher than the ones associated with direct modification, as in (4). The indirect-modification adjectives receive interpretations that can occur in both prenominal and postnominal position in English (temporary, restrictive, intersective, etc.).
In short, in Cinque’s account two sources of adnominal modification correlate with two distinct sets of interpretations that are available to adjectives only in particular (distinct) positions within the nominal phrase.
2 Two Adjectival Forms in BCS
BCS has “long” and “short” forms of adjectives, which differ in the prosody of the adjectival stem and (in some dialects) the agreement endings they take. (Pitch accents are represented as follows: [`] – falling accent; [´] – rising accent; [:] – length. It is well-known that the prosody of adjectival stems usually differs in the long form and in the short form; see Aljović 2002 for discussion.) Some dialects use the so-called nominal declension endings for agreement in the short form (bijé:lakonja ‘white.ACC.SF horse’) and pronominal declension endings in the long form (bijè:logkonja ‘white.ACC.LF horse’), the distinction being present only in masculine and neuter singular contexts; other dialects productively use only pronominal declension endings in both the long and the short form (bijé:logkonja ‘white.ACC.SF horse’, bijè:logkonja ‘white.ACC.LF horse’), except in the nominative singular masculine and the accusative singular masculine inanimate, where the short form has the ending -Ø and the long form has the ending -i.1 What is relevant for our discussion is that in all dialects the long form is only used attributively (5a–b),2 while the short form typically occurs in the predicative position (5c) but can also be used attributively (5d).
(5) Long form
‘a/the famous poet’
*Ovaj pjesnik je poznati.
this poet is famous.LF
Intended: ‘This poet is famous.’
Ovaj pjesnik je poznat.
this poet is famous.SF
‘This poet is famous.’
‘a famous poet’
Given this distribution, Cinque (2010) argues that BCS marks the two sources of adjectival modification that he proposes through overt morphology. Further, he proposes that the short form can only be used as a predicate; that is, it can occur in the predicative position of the main clause (5c), or in a regular relative clause (6), or in a reduced relative clause where only the adjective is visible (5d).
Upoznala sam pjesnika koji je poznat.
met am poet who is famous.SF
‘I met a poet who is famous.’
On the other hand, Cinque claims that the long form is the morphologically different direct source of modification. This seems to be supported by Leko’s (1988, 1992) observation that when both short and long adjectives occur in the prenominal position, the short form generally precedes the long form.3
siromašan bolesni dječak
poor.SF sick.LF boy
‘a poor sick boy’
*bolesni siromašan dječak
sick.LF poor.SF boy
However, this treatment of BCS adnominal adjectives turns out to be problematic in a number of contexts with both the long and the short form of adjectives.
2.1 Long Adjectives
BCS long adjectives are challenging for Cinque’s model because they receive both the interpretations associated with direct modification and the interpretations associated with reduced relative clauses (i.e., indirect modification). This leads Cinque to assume that BCS long adjectives are ambiguous between direct and indirect modification. However, treating long adjectives as reduced relatives is problematic, as I will argue on the basis of their distribution, their interpretation, and the order between long and short adjectives in the prenominal position.
Recall that under the dual-source analysis, reduced relative clauses are associated with temporary, restrictive, and intersective interpretation, while direct-modification adjectives are associated with permanent, nonrestrictive, and nonintersective interpretation. However, as Aljović (2000) observes, BCS long adjectives are ambiguous between the two types of readings. For example, the long adjective in (8a) can have either intersective or nonintersective interpretation, which is disambiguated in the situations given in (8c) and (8d), respectively. Given the context in (8b), in situation 1 (8c) the adjective bezbijedni ‘safe’ in (8a) is interpreted intersectively, and in situation 2 (8d) it is interpreted nonintersectively.
Stigao je bezbijedni vozač.
arrived is safe.LF driver
‘The safe driver arrived.’
A bus station employs five drivers. Bill is the only one who drives safely, while John, Jane, James, and Jack constantly cause accidents because of unsafe driving.
Situation 1: Intersective only
There was a fire in the station. Out of all the drivers, the firefighters have so far saved only James. News reporters ask for the driver who is now safe from fire (James) to come and give a statement about the incident.
Situation 2: Nonintersective only
News reporters are covering a story on the increased number of car accidents; they called the station and asked for their safest driver to come and talk about the importance of safe driving.
In addition, under the intersective reading of bezbijedni ‘safe’, (8a) entails ‘Someone who is safe arrived’ and ‘Someone who is a driver arrived’. If we know that this person is also a ‘man/worker/athlete . . .’, the inference in (9) would be valid.
Stigao je bezbijedni čovjek/radnik/sportista.
arrived is safe man/worker/athlete
‘The safe man/worker/athlete arrived.’
This inference is not valid when bezbijedni ‘safe’ yields a nonintersective reading, since in that case (8a) entails ‘Someone who is a driver arrived’, but it does not entail ‘Someone who is safe arrived’. Being safe is not a general property of the individual under this reading; rather, it depends on the property of being a driver.
Furthermore, the example in (10) shows that BCS long adjectives can have both a restrictive and a nonrestrictive interpretation. Under a restrictive reading, the adjective creates a subset of individuals denoted by the noun where each member has the property of being diligent; that is, sentence (10a) with a restrictive reading of the adjective is felicitous in a situation where a subset of our students love syntax, as in (10b). Under a nonrestrictive reading, such a subset is not created; that is, sentence (10a) with a nonrestrictive reading of the adjective is felicitous only if all of our students love syntax, as in (10c).
Naši vrijedni studenti vole sintaksu.
our diligent.LF students love syntax
‘Only our students who are diligent love syntax. The lazy ones hate it.’
‘All of our students are diligent and they love syntax.’ (Aljović 2000:107)
Cinque (2010:101) mentions this semantic ambiguity of long adjectives and concludes that they are ambiguous between the two sources of modification. That is, he assumes that BCS long adjectives are either direct-modification adjectives or reduced relative clauses, depending on the interpretation they have. However, if long adjectives can be reduced relative clauses, this leads to new predictions: (a) a long adjective should be able to occur in the predicative position inside or outside of relative clauses; (b) it should be possible to base-generate a long adjective (with a reading associated with reduced relative clauses) in front of a short adjective given that the order among reduced relative clauses is assumed to be free; and (c) a long adjective preceding a short adjective should receive a reading available only to reduced relative clauses, but a long adjective following a short adjective should be ambiguous.
Regarding the first prediction, long adjectives in BCS are problematic because they are always attributive; they can never function as predicates (5a–b). Crucially, long adjectives cannot occur in regular relative clauses (11a) or in postnominal reduced relative clauses (11b).4
*Pjesnik koji je poznati je došao.
poet who is famous.LF is come
Intended: ‘A poet who is famous came.’
*Pjesnik poznati po svojim sonetima je došao.
poet famous.LF by self’s sonnets is come
‘A poet who is famous for his sonnets came.’
Pjesnik poznat po svojim sonetima je došao.
poet famous.SF by self’s sonnets is come
‘A poet (who is) famous for his sonnets came.’
Given that the long form cannot occur in any of the contexts for predicative adjectives (5b)/(11a–b), there is no evidence that long adjectives can ever be predicates, as they would be in a prenominal reduced relative clause. Importantly, it is not clear why the long form should be permitted in a prenominal reduced relative clause, but not in a postnominal reduced relative clause (11b). All of this indicates that BCS long adjectives are never reduced relative clauses.
The second prediction (that it should be possible to base-generate long adjectives in front of short adjectives) is also not borne out: BCS does not allow this (7b)/(12) (see Leko 1992). The only time a long adjective is found in front of a short adjective is when it receives a discourse-linked or epistemic reading, a possibility limited to a subset of adjectives. In this case, a long adjective undergoes movement to the position preceding a short adjective, rather than being base-generated there (see Stanković 2015 and footnote 3).
*vrijedni pametan student
diligent.LF smart.SF student
Intended: ‘the diligent smart student’
*lijepi kamen most
beautiful.LF stone.SF bridge
Intended: ‘the beautiful stone bridge’
Finally, regarding the third prediction, since the order in (12) is not generally available for long adjectives (except when resulting from movement) that receive interpretations associated with reduced relative clauses, we also cannot test what interpretations long adjectives would receive in the context in (12).
Furthermore, certain adjectives in BCS (as well as English) occur exclusively in the attributive position. Under the dual-source analysis, these adjectives would be expected to occur in the nominal phrase only as direct-modification adjectives, meaning that they could have only permanent, nonrestrictive, nonintersective interpretations. In fact, in BCS attributive-only adjectives do not even have a short form. Some such adjectives are bivši ‘former’, budući ‘future’, pravi ‘real’, krivični ‘criminal’, mašinski ‘mechanical’, and električni ‘electrical’. Given that these adjectives never occur in the predicative position, they are not expected to occur as predicates in reduced relative clauses. Therefore, they should receive only the readings associated with a direct modification source (permanent, nonrestrictive). However, these adjectives are also ambiguous between the two types of readings. For example, the adjective električni ‘electrical’ in (13a) can be interpreted as either restrictive or nonrestrictive. With a restrictive reading of the adjective, (13a) is felicitous in a situation where electrical engineers are a subset of individuals who are engineers (13b); with a nonrestrictive reading of the adjective, it is felicitous in a situation where all engineers are electrical engineers (13c).
Dao je povišice električnim inžinjerima.
given is raises electrical.LF engineers
‘He gave raises to electrical engineers.’
The firm employs electrical and mechanical engineers.
All the engineers that the firm employs are electrical engineers.
For električni ‘electrical’ to receive a restrictive reading in (13a), it would need to be analyzed as a reduced relative clause under the dual-source analysis. However, it would then be surprising that this adjective can otherwise never occur in a predicative position.
*Ovaj inžinjer je električni.
this engineer is electrical.LF
Intended: ‘This engineer is electrical.’
*Ovaj inžinjer, koji je električni, dobio je
this engineer which is electrical.LF got is
Intended: ‘This engineer, who is electrical, got a raise.’
English also has exclusively attributive adjectives, which represent a challenge for the dual-source analysis as well. Some such adjectives in English are live, future, sole, and former. These adjectives never occur as predicates in copular constructions (15b,e) or in regular relative clauses (15c,f), and they are expected to be treated in terms of direct modification.5
our future prospects
*Our prospects are future.
*our prospects that were future
a live volcano
*The volcano is live.
*the volcano that is live
When a strictly attributive adjective like live is accompanied by another adjective in the prenominal position, we would expect the latter to necessarily be a direct-modification adjective because such adjectives are claimed to be in FPs closer to the noun and reduced relative clauses are more peripheral. We would also expect the adjective following live to receive only the interpretation associated with direct modification. On the other hand, if an adjective precedes a strictly attributive adjective like live, it could be either a direct-modification adjective or a reduced relative clause; hence, it is expected to be ambiguous. However, these predictions are challenged by the interpretations of the adjective visible in (16). In (16a), where the adjective visible follows the adjective live, visible can receive either a permanent- or a temporary-property interpretation (16c–d). Assuming live is a direct-modification adjective, the permanent-property reading associated with direct modification is expected, but not the temporary-property reading associated with reduced relative clauses. In (16b), where the adjective visible precedes the adjective live, it is much harder for visible to receive the permanent-property interpretation (16c), while the temporary-property interpretation (16d) is readily available.
(While combing through his hair,) He found live visible lice.
(While combing through his hair,) He found visible live lice.
‘He found lice that are tiny but inherently visible with a naked eye.’
‘He found lice that are tiny but they are visible now because he moved the hair that was covering them.’
In sum, we have seen several cases where the interpretation, distribution, and ordering of long adjectives in BCS and adjectives in English is problematic for the dual-source analysis. The BCS facts are especially problematic for the claim that long adjectives can be reduced relative clauses.
2.2 Short Adjectives
BCS short adjectives typically occur as predicates in copular constructions; in prenominal and postnominal position, they are claimed to always have a reduced-relative-clause source. I argue below that BCS prenominal short adjectives should not be treated as reduced relative clauses, given their extraction possibilities out of APs. I also show that such a treatment of prenominal short adjectives is challenged in the context of exclusively predicative adjectives, and that short adjectives yield interpretations that are unexpected under a reduced-relative- clause analysis.
The first argument against treating prenominal short adjectives as reduced relative clauses comes from extraction possibilities. It is well-known that relative clauses are islands for extraction (Ross 1967); this is true both for English (17) and for BCS (18).
They met someone who knows Julia.
*[Which girl]i did they meet someone who knows ti?
Phineas knows a girl who is jealous of Maxine.
*Whoi does Phineas know a girl who is jealous of ti? (Ross 1967:124)
Upoznali su nekoga ko poznaje Kosaru.
met are someone who knows Kosara
‘They met someone who knows Kosara.’
*[Koju djevojku]i su upoznali nekoga ko
which girl are met someone who
Intended: ‘Which girl did they meet someone who knows?’
Crucially, Ross (1967) notes that even reduced relative clauses are islands; that is, extraction out of relative clauses is not possible even when they are reduced (see also, e.g., Chomsky 1986, Siloni 1997 (for French)).
*Whoi does Phineas know a girl jealous of ti?
*Which childreni did John write a book (for parents)
to read ti?
*D’ oùi Jean est-il le dernier revenu
from where Jean is-he the last.one returned
Intended: ‘From where is Jean the last one (who)
In this respect, the dual-source analysis of BCS adnominal adjectives predicts that APs with short adjectives should be islands for extraction, given that they always enter the structure as reduced relative clauses under this view. This prediction can be tested by using adverb extraction out of APs in different configurations. BCS allows extraction of adverbs out of predicative APs (20a). Such extraction is not possible if the predicative AP is embedded within a relative clause (20b).
Veomai je bio ti lijep.
very is been beautiful.SF
cf. Bio je veoma lijep.
‘It was very beautiful.’
*Veomai je vidjela kaput koji je bio ti lijep.
very is seen coat which is been beautiful.SF
cf. Vidjela je kaput koji je bio veoma lijep.
‘She saw a very beautiful coat.’
When a prenominal AP with a short adjective is embedded within a regular relative clause, adverb extraction out of the AP is also not possible (21b).
Vidjeli su djevojku koja je kupila [izuzetno
seen are girl which is bought extremely
‘They saw a girl who bought an extremely beautiful coat.’
*Izuzetnoi su vidjeli djevojku koja je kupila [ti
extremely are seen girl which is bought
Thus, regular relative clauses are islands for extraction of adverbs both from predicative APs and from prenominal APs with short adjectives. As shown in (19), reduced relative clauses are also islands for extraction. Now, if BCS prenominal and postnominal APs with short adjectives were reduced relative clauses, as the dual-source analysis holds, adverb extraction would be predicted to be blocked in both of these contexts. However, prenominal and postnominal APs with short adjectives behave differently in this respect. Crucially, adverb extraction out of a prenominal AP with a short adjective is possible in (22), contrary to what is predicted.6
Izuzetnoi je kupila ti lijep kaput.
extremely is bought beautiful.SF coat
‘She bought an extremely beautiful coat.’
Užasno su kupili ti ružan stan.
terribly are bought ugly.SF apartment
‘They bought a terribly ugly apartment.’
The data in (20)–(22) show that extraction of an adverb out of an AP is possible when the AP is not base-generated within an overt relative clause (20a)/(22), but is blocked when it is clear that the adverb is embedded within a relative clause (20b)/(21b). Hence, the lack of an island effect in (22) strongly suggests that prenominal short adjectives are not reduced relative clauses.
The second argument against treating prenominal APs with short adjectives as reduced relative clauses comes from a contrast between prenominal and postnominal APs with short adjectives. Postnominal APs are also often treated as reduced relative clauses (Sadler and Arnold 1994, Larson 1998, Larson and Marušič 2004, Cinque 2010), although this claim has not been explicitly made for BCS postnominal APs (23).
Upoznali su roditelje izuzetno pó:nosne na svoju
met are parents extremely proud.PL.SF of self’s
‘They met parents extremely proud of their children.’
Posjetili su zemlju izuzetno bógatu rijekama.
visited are country extremely rich.PL.SF rivers.INSTR
‘They visited a country extremely rich in rivers.’
Uslikala je lice strašno ruméno od hladnoće.
photographed is face terribly red.SG.SF from cold
‘She photographed a face terribly red from cold.’
Evidence that postnominal BCS adjectives may be reduced relative clauses comes from the fact that only the short adjectival form can occur in this position. The long form, which is never allowed in a predicative position, is excluded in the postnominal position, as illustrated with the contrasts between (23a) and (24a), (23b) and (24b), and (23c) and (24c). Crucially, it is not possible to use the long forms in (24) in the postnominal position (they differ in the length of the final vowel from their short counterparts in (23a–b); moreover, in (23c) the rising accent is on the second syllable, but in (24c) it is on the initial syllable).
If BCS postnominal APs are reduced relative clauses, then adverb extraction out of the sentences in (23) should be impossible. Interestingly, this is exactly what we find. Adverb extraction out of postnominal reduced relative clauses is disallowed.
*Izuzetnoi su upoznali roditelje ti pó:nosne na svoju
extremely are met parents proud.SF of self’s
Intended: ‘Extremely, they met parents proud of their children.’
*Izuzetnoi su posjetili zemlju ti bógatu rijekama.
extremely are visited country rich.SF rivers.INSTR
Intended: ‘Extremely, they visited a country rich in rivers.’
*Strašno je uslikala lice crveno od hladnoće.
terribly is photographed face red from cold
Intended: ‘Terribly, she photographed a face red from cold.’
The contrast between the availability of adverb extraction out of prenominal APs with short adjectives (22) and the unavailability of such extraction out of nonreduced (21b) and reduced (25) relative clauses further suggests that BCS prenominal short adjectives should not be treated as reduced relative clauses.
Regarding adverb extraction out of direct-modification adjectives, the prediction seems to be that if a language in principle allows adverb extraction from prenominal APs, extraction from direct-modification APs should be possible since they are not reduced relative clauses. Since BCS long adjectives have been analyzed as either direct-modification adjectives or reduced relative clauses in this model, we might expect that at least in some cases adverb extraction should be allowed out of APs with long adjectives. However, this is not the case. Compare (22) with (26).7
*Izuzetnoi su kupili [ti skupi] automobil.
extremely are bought expensive.LF car
cf. Kupili su izuzetno skupi automobil.
‘They bought the extremely expensive car/one of the extremely expensive cars.’
*Užasnoi su kupili [ti skupi/ružni] stan.
terribly are bought expensive/ugly.LF apartment
cf. Kupili su užasno skupi/ružni stan.
‘They bought the terribly expensive/ugly apartment.’
This indicates that APs with long adjectives are islands for adverb extraction. Nevertheless, this does not necessarily indicate that they are reduced relative clauses. In the previous section, I have provided evidence from the distribution of the long form that this form is never found in a predicative position in copular constructions (5b)/(11b), a regular relative clause (11c), or a postnominal reduced relative clause (11d). Therefore, it is unlikely to occur in the predicative position only with prenominal reduced relative clauses.
Further evidence that reduced relative clauses are not available in the prenominal position in BCS and English comes from adjectives that are strictly predicative, like those in (27). In BCS, such adjectives only have a short form, which is the form typically used as a predicate.
On je voljan da pomogne.
he is willing.SF that helps
‘He is willing to help.’
On je sam.
he is alone.SF
‘He is alone.’
The boy is present (spatial sense) /alone /asleep.
Under the dual-source analysis, these adjectives are expected to occur in nominal phrases only as predicates in regular or reduced relative clauses, but not in the position of direct modification. Therefore, they are expected to occur in the postnominal position and in the prenominal position preceding direct-modification adjectives. However, consider the following distribution.
In both BCS and English, strictly predicative adjectives can occur in regular relative clauses.
Došao je svaki čovjek koji je bio voljan da
come is every man who is been willing.SF that
‘Every man who was willing to help came.’
Every boy who was present was singing.
They can also be used in the postnominal position, which is expected if postnominal adjectives are reduced relative clauses.
Došao je svaki čovjek voljan da pomogne.
come is every man willing.SF that helps
‘Every man willing to help came.’
Every boy present was singing.
Given that such adjectives can occur in nonreduced relative clauses and postnominal reduced relative clauses, we would expect them to be able to occur in prenominal reduced relative clauses preceding direct-modification adjectives as well. This is not the case, however, as (30) shows.8
*Jedan sam (mali) dječak sjedi na stepenicama.
one alone (little.LF) boy sits on stairs
Intended: ‘A little boy who is alone is sitting on the stairs.’
*On je uvijek bio jedan da pomogne voljan
he is always been one that helps willing.SF
Intended: ‘He has always been an honest man willing to help.’
*An alone little boy is sitting on the stairs.
*Every present boy was singing.
These data suggest that a reduced-relative-clause source is not available in the prenominal position in BCS and English.
Finally, the interpretive properties of short adjectives also indicate that they cannot be (only) analyzed as reduced relative clauses. Recall that under the dual-source analysis, reduced relative clauses are associated with temporary, restrictive, and intersective interpretation. However, the short adjective in (31a) is unambiguously nonintersective (i.e., this sentence is felicitous in a situation where all students (not just a subset) love syntax), and the one in (31b) is interpreted as a permanent property (see also Despić and Sharvit 2011 for more discussion of contexts where short adjectives are both intersective and nonintersective).
??Naši vrijedni studenti vole sintaksu.
our diligent.SF students love syntax
‘All of our students are diligent and they love syntax.’
Trenutno su nepristupačne neke inače
at.the.moment are inaccessible some otherwise
‘Some otherwise navigable rivers are inaccessible at the moment.’
Given (31a–b), which indicate that the short form doesn’t just yield interpretations associated with reduced relative clauses, the only way to capture this semantic ambiguity of short adjectives in the dual-source analysis seems to be to assume that short adjectives are ambiguous between the two sources of modification, parallel to Cinque’s proposal for long adjectives. However, treating both short and long adjectives as ambiguous between indirect modification (reduced relative clauses) and direct modification would completely undermine the claim that BCS overtly distinguishes between the two sources of adjectives because neither the long form nor the short form would then be reserved exclusively for marking one source.9
In summary, we have seen that the dual-source analysis of prenominal adjectives in BCS proposed by Cinque (2010) is problematic, given the interpretation, distribution, and extraction possibilities with adjectives. Crucially, interpretations associated with reduced relative clauses under this analysis occur with adjectives that cannot be analyzed as reduced relative clauses on the basis of their distribution (BCS long adjectives, BCS and English exclusively attributive adjectives). Similarly, interpretations associated with direct modification occur with adjectives that can only be reduced relative clauses in this system (BCS short adjectives). Finally, analyzing BCS prenominal short adjectives as reduced relative clauses would lead to the wrong prediction that extraction out of such phrases is not available. I have shown that BCS disallows adverb extraction out of APs with short adjectives that are embedded in a nonreduced relative clause or a postnominal reduced relative clause. However, such extraction is possible out of APs with short adjectives that are not embedded in regular relative clauses, including APs in the prenominal position. This indicates that prenominal APs with short adjectives are not reduced relative clauses. BCS then turns out not to be a language that overtly distinguishes between direct- and indirect-modification sources. Although there may be two syntactic sources of modification, these two sources are not associated with two different sets of interpretations of adjectives.
For helpful comments, I am very grateful to Željko Bošković, Jonathan Bobaljik, Nadira Aljović, Susi Wurmbrand, Jon Gajewski, Magdalena Kaufman, Nedžad Leko, two anonymous LI reviewers, and the Squibs and Discussion editors.
1 As one of the reviewers notes, there are speakers who use both nominal declension and pronominal declension endings on short-form adjectives for whom this distinction is stylistic.
2 As a reviewer points out, examples like (5b) are acceptable only in the case of elliptical definite nominals that establish an identity relation (specificational or equative predicates). The required context where this sentence would be acceptable is the one where there are two poets established in the previous discourse, one famous and the other not famous. For some BCS speakers, there is a strong preference to use a demonstrative before the adjective in such contexts (see Cinque 2010:144). Cinque (2010) also suggests that these contexts involve NP-ellipsis; that is, the long adjective is in the attributive position, but the NP it modifies is elided (see also Babby 1970, 1973, 1975, Siegel 1976a,b, and Bailyn 1994 for Russian).
3 While this observation holds in most cases where short and long adjectives cooccur, Stanković (2015) notes that there is a closed set of discourse-related adjectives such as pomenuti ‘mentioned’ whose long form can precede a short-form adjective (i), which at first seems problematic for Cinque’s account.
(i) već pomenuti studiozan pregled njegovog lingvisti?kog opusa already mentioned studious overview his linguistics work ‘already-mentioned studious overview of his work in linguistics’ (Stanković 2015:241)
However, the structure of such cases may involve the presence of additional projections and additional movement operations, as Stanković himself suggests, so these examples are not parallel to the long-short adjective combination in (7b), which is meant to illustrate that long adjectives cannot be base-generated before short adjectives. Thus, (7b) needs a separate account from (i) (see footnote 9).
5 In the context of such adjectives, Cinque (2010:51–52) discusses a long tradition in generative grammar of deriving all adjectives from postnominal relative clauses, pointing out that such a derivation is only possible with adjectives like former if we assume they are derived from adverbs in relative clauses, since they never function as predicates. However, considering a larger set of strictly attributive adjectives, he notes that such an analysis would be highly unconstrained and suggests that attributive-only adjectives have a direct-modification source.
6 There appears to be some variation among speakers regarding when an adverb can be separated from a short adjective. In particular, some speakers do not accept examples where an adverb is separated from a short adjective modifying a noun in an argumental NP, but they accept the separation from a predicative adjective as in (20). Regarding the source of this variation, the speakers in question may differ from the ones who accept (22) in that, for them, attributive adjective phrases may always have more structure than predicative adjective phrases, regardless of whether they are headed by long or short adjectives. For these speakers, this additional structure would then block adverb extraction with both long and short adjectives (see Talić 2015, to appear, for why additional structure matters).
7 Some speakers do not find the examples in (26) acceptable even without extraction (Kupili su izuzetno skupi automobil /Kupili su užasno skupi/ružni stan) in out-of-the-blue situations. However, such interpretations are more salient in the following contexts:
(i) There are three cars in a dealer’s showroom. One of them is cheap, one of them is a little expensive, and one of them is extremely expensive. (cf. (26a))
(ii) They looked at two apartments. One of them was cheap/beautiful and the other one was extremely expensive/ugly. (cf. (26b))
8Cinque (2010:59) notes that predicative adjectives with complements obligatorily extrapose just like full relative clauses. However, adjectives like sam ‘alone’ do not take complements, so it is less clear why their extraposition should be obligatory.
9 The question arises again why the base-generated order *long A+ short A + N is impossible. I conjecture that in this configuration, the long adjective fails to agree with the noun owing to the presence of the intervening short adjective. Both long and short adjectives agree with the noun in number, gender, and case, but only long adjectives are restricted to specific contexts. Thus, I assume that BCS long adjectives have one feature more than short adjectives; that is, while long adjectives have unvalued number, gender, case, and specificity features (see Talić 2015 for a more detailed discussion), short adjectives have only unvalued number, gender, and case features, being unmarked for specificity. If a short adjective enters the derivation before a long adjective and agrees with the noun, getting its features valued, when the long adjective enters the derivation, there are two potential goals with valued features, the noun and the short adjective. The short adjective is closer to the long adjective than the noun, so the long adjective probes the short adjective. However, this can only result in valuation of the number, gender, and case features on the long adjective, the specificity feature remaining unlicensed.