Harmon 1 Phrase structure and Linearization: Adverbs in ASL

 Harmon 1 Phrase structure and Linearization: Adverbs in ASL Jessica Harmon In this paper, I address two issues of the syntax of adverbs: their phrase structure and the mechanism of their linearization. Adverbs can be analyzed as adjuncts or as specifiers of heads. The phrase structure and mechanism of linearization of adjuncts, can be further informed by looking at data from American Sign Language (ASL). Adverbs in ASL can be articulated either as manual signs or as non-­‐manual morphemes. I have discovered that the non-­‐
manual adverbs are manner adverbs, while higher adverbs have manual signs. This split in the distribution of manuals versus non-­‐manuals can inform the discussion of both phrase structure and linearization. While the manual adverbs can be analyzed using a fixed hierarchy of phrases (Cinque 1999), the non-­‐manual adverbs are better analyzed as unlabeled concatenation/adjunction, where the mechanism of linearization is unspecified. These non-­‐manual adverbs show us that strict precedence is not always required with adjunction. 1 Introduction First, I discuss the theoretical status of of adjunction with respect to adjunction by looking at the behavior of adjuncts and the operation Merge. In section 2 I examine both the structure and mechanism of linearization of adverbs described in Cinque (1999) and compared to Hornstein (2008). In section 3 I give a brief description of the grammar of American Sign Language. Section 4 presents the ASL adverb data, making a distinction between the manual and non-­‐manual adverbs. I section 5 I give an analysis of these adverbs, using Cinque’s (1999) fixed hierarchy analysis for the higher manual adverbs and Hornstein’s (2008) concatenation without labeling for the non-­‐manual adverbs. This allows a different account of linearization for the simultaneously-­‐articulated non-­‐manual adverbs. 1.1 The problem of adjunction for syntactic phrase structure A distinction between complements and adjuncts is central to theories of syntax. Complements are required by a predicate to check a categorical feature. When complements are merged with a head, as in (1), a different constituent is created. (1) Here, the complement YP merged with the head X, resulting in a new constituent, XP. Adjunction, on the other hand is an operation that adds a modifier B to a constituent A and creates another constituent A: Harmon 2 (2) a. b. Here, the adjunct B is attached to the constituent A. The resulting constituent is still A. Adjunction does not change the syntactic category of the modified constituent, so An NP with an AdjP adjoined remains an NP. This distinction between adjuncts and complements can be seen in the Chomsky (2000) separation of the operation Merge into “pair-­‐merge” and “set-­‐
merge”. The operation Merge both combines elements and assigns a label to the resulting structure. It is necessary to distinguish the two types of Merge with different labeling mechanisms because there is nothing else to distinguish the two structures otherwise. Merge is such a basic operation for combining smaller elements into a larger unit, that the only difference between attaching an adjunct and a complement is in the label. Chomsky (2000) makes a distinction between “set-­‐
merge” for arguments and “pair-­‐merge” for adjuncts. When an argument and a head are merged, a set is created and the category of the new constituent comes from the head; this is set-­‐merge. Pair-­‐merge is used when adjuncts are merged, forming an ordered pair. When the complement merged with a head (set-­‐merge), a new constituent with a new category is formed. When an adjunct merged, the new constituent retains the same category (pair-­‐merge). Below in (2) the adverb 'bravely' is adjoined to the VP 'fought a monster'. When the complement DP ‘the monster’ is attached to the V ‘fought’, a new constituent, ‘fought the monster’ with a new category, VP, is formed. However, when the adjunct AdvP ‘bravely’ is attached to the VP ‘fought the monster’, the new constituent remains a VP. (3) a. b. Harmon 3 Adjuncts, as well as complements, have linear variation, with some occurring to the left and others to the right of the constituent they modify. This can be seen in a structure like (3)b., where the adverb 'bravely' is attached to the right of the VP it modifies, whereas in (3)a., it is attached to the left. Within shell theory, the external argument is introduced by little v. So, the only arguments V can have are the complement and specifier (for double-­‐object constructions). If indirect objects are also introduced by a little v (different from the first), then there is only one argument per head. Unlike arguments, there is no apparent upper bound on the number of adjuncts a phrase can have. An adjunct does not change the syntactic category of the element it modifies, so adjuncts are allowed to be recursively added. An adjunct does not affect the distribution of an expression: if an XP can occur in some position, it can occur in that position with any number of its adjuncts. In (4) a., the VP ‘fight the dragon’ can be fronted to the beginning of its clause. In order for VP fronting to occur, the proposition must be salient to the immediately preceding discourse (thus the conjoined sentences here). So, in (4) a., ‘fight the dragon’ can be fronted to the beginning of the clause ‘ he did fight the dragon’. VP-­‐fronting can be a test of what kind of constituent a VP and an adjunct form. If a VP with an adjunct is also a VP, the adjuncts should also be able to occur before ‘he did’, as indeed they do in c. and d. (4) a. George could fight the dragon, and fight the dragon he did. b. George could fight the dragon with a lance, and fight the dragon he did with a lance. c. George could fight the dragon cleverly with a lance, and fight the dragon cleverly he did with a lance. d. George could fight the dragon cleverly with a lance, and fight the dragon cleverly with a lance he did. The addition of various adjuncts does not affect the VP status of ‘fight the dragon’, as ‘fight the dragon’ can be moved without the adjunct as in (4)b. The various adjuncts plus VP are also VPs, and can also be fronted, but the complement ‘the dragon’ cannot be left behind: (5) *George could fight the dragon and fight with a lance he did the dragon. Because ‘fight’ by itself is not a VP, it cannot be fronted, even with the adjunct ‘with a lance’. Under Bare Phrase Structure, once an adjunct is merged, the maximal projection is no longer maximal. In (2) or (3), once the adjunct is merged into the sentence, it creates another segment of the VP. The VP ‘fought a monster’ is no longer a maximal projection, since there is another VP that dominates it. Another question about adjuncts is what level they are attached to. In (2) and (3) the adjuncts attach at the VP level. In X-­‐bar theory, adjuncts could also attach at the V’ level: Harmon 4 (6) In (6), the adverb is adjoined to the V’, and the resulting constituent is also a V’. Adjuncts have the property that they are not obligatory to the syntactic structure. (7) a. He killed [the monster] [on the bridge] b. He killed [the monster] c. He killed [on the bridge] In (7) a. and b. include the complement of the verb 'killed', [the monster], and the sentence is grammatical with (in a.) or without (in b.) the modifying PP [on the bridge]. However, if the complement DP [the monster] is removed, the sentence is not grammatical. This points to another property of adjuncts: they are not essential to the meaning of a constituent. Complements make up part of the inherent meaning of the head. They have a close semantic relationship to the head. 'Killed' above requires a DP complement to be interpreted as the patient. The relationship of a verb to its complements is idiosyncratic in the sense that it is unpredictable whether the complement is a patient (‘killed’) or a theme (‘see’) or whether the verb is intransitive and only has a subject (‘died’). It is not predictable in advance which verb will have a complement or how to interpret that complement. Adjuncts, which for verbs modify the time, locations, manner, etc, are not required, but can potentially be added. That is, it is not unpredictable that a verb can be modified for time, location, or manner. Adjuncts add extra information which can be interpreted just by virtue of being adjoined to the verb, not by any specific requirements of the verb itself. An adjunct can be defined as an element that has the same mother and sister. When an element is adjoined, it is not only that the categories are the same, it is that it is two segments of the same category itself. An ambiguous c-­‐command relationship is created when there are multiple segments of the same category. When DP is merged with V, a new category, VP is made, but when an AdvP is merged with a VP, the category stays the same: it is still VP. So, the mother and sister of AdvP are both VP. While an element does c-­‐command its sister, it does not c-­‐
Harmon 5 command its mother. This creates a structure where the relationship is not clear. Does an adjunct still c-­‐command its sister even if it is dominated by another segment of that same category? 1.2 The problem of adjunction for linearization Adjuncts are also a problem for linearization at the PF interface. Linearization of constituents relies upon asymmetrical syntactic relationships. While this is fine for most structures, the linearization of adjuncts is less clear since the c-­‐command relationships are ambiguous. If adverbs are analyzed as adjuncts, the analyses of linearization relying on asymmetrical relationships are not possible. Adverbials pose a problem for phrase structure because their mechanism of linearization is unspecified. A mechanism of adjunct linearization or an analysis of adverbs without adjunction is needed. This will be illustrated in greater detail below. 1.3 Why sign languages? Sign languages can inform the structure of adverbs because there is the possibility of having the adverb and the verb appear simultaneously. Manual signs are linearized with respect to each other, but non-­‐manual signs, like some adverbs, occur simultaneously with respect to other signs. These non-­‐manual adverbs are, in some sense, not linearized with respect to the verb they modify. This allows for the possibility of observing effects of hypothesized structures where linearization does not follow from an asymmetrical relationship. These effects would not be visible if the adverbs were forced to be linearized in precedence relationships with respect to the verb, as they are in spoken language. The possibility of simultaneous transmission from the syntactic system to the PF interface allows us to make a distinction between structures that are linearized because of an asymmetrical relationship and those that are linearized because it is only possible to say one word at a time. 1.4 Adverbs in American Sign Language Adverbs in American Sign Language (ASL) have at least two forms: independent manual signs and non-­‐manual facial expressions that occur simultaneously with other signs. High adverbs (evidential, speaker attitude, etc.) occur as manual signs and in a fixed order, corresponding to Cinque’s (1999) universal order of functional projections. This puts these adverbs into a Spec-­‐Head relationship in functional projections corresponding to the semantics of the adverb. This asymmetric relationship can be linearized using Kayne’s asymmetric c-­‐command or Hornstein’s (2008) Concatenate and Label mechanisms. These adverbs do not have to be categorized as adjuncts, so they can be analyzed with a theory of phrase structure that does not include adjuncts. Low (e.g. manner) adverbs can be non-­‐manual in ASL. They occur at the same time and for the duration of the verb. Manner adverbs have been analyzed as adjuncts in languages besides ASL, however, their place in phrase structure and linearization mechanism is unspecified. Hornstein (2008) analyzes adjuncts as Harmon 6 concatenation without labeling. From the distribution of manual vs. non-­‐manual adverbs in ASL, it can be said that if something is not labeled (is an adjunct), then it can be non-­‐manual. The fact that in ASL neither the verb nor its adjunct has precedence over the other shows that a precedence relationship is not necessary for grammaticality. This supports Hornstein’s concatenation without labeling analysis of adjuncts. 2 Adverbs in theories of phrase structure. Adverbs have been traditionally analyzed as adjuncts (Pollack 1989, Chomsky 1986, 1995, Hornstein 2008 among others) and as specifiers to functional projections (Cinque 1999) depending on what data is emphasized. 2.1 Hornstein’s Concatenate and Label Phrase Structure Hornstein (2008) lays out a theory of syntax based on an analysis of Merge as consisting of the combination of the operations Concatenate and Label. Concatenate is not a specifically linguistic concept, but is a basic step for many cognitive operations. Concatenate creates longer and longer strings without any nesting. Label is the mechanism that creates nested syntactic constituents that can be used higher up in the tree. Higher constituents can only ‘see’ the lower parts of the tree if they are properly labeled. When an element is concatenated, a label is projected up the tree to label the new constituent. This creates a phrase structure where nodes are labeled in order to be interpreted at the next level of structure. An adverb in this structure would be concatenated to whatever level of structure is necessary for the scope of that adverb, and the resulting constituent would be labeled based on the element that the adverb is modifying. (8) ‘Gandalf often survives.’ Here, the adverb ‘often is concatenated to the V and the resulting constituent is labeled, according to Bare Phrase Structure, as V. The label allows the structure to be seen by higher elements. However, Hornstein also allows for the possibility that an element can be concatenated without being labeled. Adjuncts are independent enough that they can be interpreted with just concatenation. There is no need to label them so they can be “seen” and interpreted further up in the structure, as they do not need to enter any Harmon 7 further dependency. This creates a multidominance structure such that the concatenated element “dangles off” the sentence. (9) [vP V opened ^DP ‘the door’]^Adv ‘quickly’ T’
(no label)
vP
AdvP
V
opened
DP
quickly
the door
‘opened the door quickly’ This allows for multiple adjuncts to be concatenated and reordered without changing the structure of the rest of the sentence, as seen in (10). (10) a. [vP V opened ^DP ‘the door’]^PP with magic ^Adv ‘quickly’ ‘opened the door with magic quickly” b. [vP V opened ^DP ‘the door’]^Adv ‘quickly’ ^PP with magic “opened the door quickly with magic” Adverbs then come in two varieties of structure: those that are concatenated and labeled, and those that are just concatenated. For Hornstein (2008) the distinction comes from decomposing Merge into its two steps, and the phrase structure differs depending on the occurrence of labeled or unlabeled constituents. If the adverbs are just concatenated, without labeling, then they will not be able to be fronted as in (4) above. Only if the adjuncts are integrated into the V-­‐projection by concatenation and labeling are they able to be fronted. Dangling, unlabeled adjuncts are interpreted without being labeled because of their relationship to the V, but because they do not enter into any further dependency, they are not able to be moved as a labeled structure would. 2.2 Cinque’s universal adverb order Adverbs have ordering restrictions such that certain adverbs always precede others. This can be seen in ordering restrictions of English adverbs like ‘evidently’ and ‘probably’. (11) a. Gandalf evidently probably survived. b. *Gandalf probably evidently survived. Harmon 8 Cinque (1999) looks at adverbs in several languages and finds that adverbs have a fixed order across languages. This order reflects the order of functional projections of the universal hierarchy. In Cinque’s (1999) analysis, adverbs are in the specifiers of their respective functional projections. So ‘evidently’ is in the specifier of the evidential mood phrase, and ‘probably’ is in the specifier of the slightly lower epistemic modality phrase. The ordering restrictions stem from the universal order of the functional projections. The higher phrases have scope over the lower phrases. (12) In (12), the evaluative mood phrase is higher than the epistemic modality phrase, so the respective adverbs in the specifiers of these phrases are ordered linearly based on the universal ordering of these phrases. Each adverb is in the specifier of a functional projection. Some lower adverbs do not seem to have these same ordering restrictions. Cinque (1999) refers to these adverbs as ‘circumstantial’. Manner adverbs belong to this class and can be reordered. (13) a. Gandalf quietly quickly opened the door. b. Gandalf quickly quietly opened the door. Cinque (1999) says that these adverbs are not inherently affiliated with any specific functional projection, but are in a relationship with vP. Cinque retains a Spec-­‐head relationship for these adverbs, but allows that there may be multiple vP heads and that they can be reordered. Harmon 9 (14) In (14) the manner adverbs are in the specifiers of little v. The phrases could be reordered in into to obtain the various adverbs orders. 2.3 Linearization 2.3.1 Kayne and the LCA Kayne (1994) proposes that phrase structure completely determines linear order. If phrases are linearized differently, there must be a difference in structure. The mechanism of linearization relies on c-­‐command, particularly, asymmetrical c-­‐
command. A distinction between categories and segments (maximal/minimal projections) is also made. C-­‐command is defined: X c-­‐commands Y iff Y and Y are categories and X excludes Y and every category that dominates X dominates Y. Kayne proposes the Linear Correspondence Axiom, which applies to categories: Linear Correspondence Axiom: α precedes β if α asymmetrically c-­‐commands β The LCA leads to the specific hypotheses that complements follow their heads and specifiers precede their phrases. Kayne’s proposal does not allow for unclear hierarchical relationships. Harmon 10 (15) In this tree, A asymmetrically c-­‐commands D, so A (and thus maximal projection a) precedes D (and d). Kayne (1994) allows for adjunction in the case of specifiers, using the segment/category distinction to create the asymmetrical relationship necessary for linearization. (16) Here, G is adjoined to F, but F does not dominate G (only one of its segments does). Since the lower F is a segment, not a category, it cannot c-­‐command. So, the c-­‐
command relationships in (16) are: G asymmetrically c-­‐commands I, J, and K, and I asymmetrically c-­‐commands K. This gives the complete linear order of the terminal nodes: h precedes I, h precedes k, and I precedes k., for a final order of h, I, k. Other adjunction structures are ruled out because the mechanism of linearization does not provide a single complete order. Kayne’s phrase structure predicts that adverbs will be in an asymmetric relationship with the phrase they modify. The c-­‐command relationships between all the elements must be clear. That is the only way they can be linearized, as adjunction creates ambiguity in ordering. 2.3.2 Hornstein For Hornstein, linearization does not happen because of a c-­‐command relationship (since sisters also c-­‐command each other), but because of an asymmetrical relationship. When a concatenated structure is labeled, it creates an Harmon 11 asymmetrical relationship between the constituents. One element projects its label onto the new structure. This asymmetry allows for precedence relationships to be determined. Like Kayne’s linearization by c-­‐command, an asymmetrical relationship determined by labeling can be linearized based on the structure alone. However, while the linear order of a labeled structure can be derived from the asymmetrical relationship, the linear order of the unlabeled structure needs a different mechanism. The unlabeled adjunction structure can be ordered with respect to the other elements in the sentence. However, because there is no label on the structure it directly concatenates with, the order with respect to the element it modifies is unclear. There is no asymmetrical relationship that determines the order. A different mechanism of linearization is needed for unlabeled structures which is not discussed in Hornstein (2008). 2.3.3 Cinque The adverbs asymmetrically c-­‐command the heads of the functional projections. Because everything is in an asymmetrical relationship, the linear order of the elements is simple to determine. Using Kayne’s asymmetrical c-­‐command or Hornstein’s labeling, it is possible to order all of the adverbs with respect to the phrasal heads as well as with respect to each other. For the lower adverbs as well the c-­‐command relationships are clear. While the lower adverbs may be reordered, once the syntactic structure is determined, asymmetrical relationships are created. This makes linear order possible to determine once the syntactic structure of a given sentence is decided. While the adverbs can be reordered from sentence to sentence, in any specific case, the precedence relationships are clear from the asymmetrical syntactic structures. 2.4 Predictions Hornstein’s separation of Merge into Concatenate and Label makes the prediction that adverbs can behave differently with respect to linearization based on whether the structures are labeled or not. Labeled structures should have clear precedence relationships based on the asymmetries created by labeling (i.e. one element projects its label). Unlabeled structures cannot be linearized in this way, so a difference in behavior of unlabeled structures in expected. Cinque’s (1999) analysis makes the prediction that the higher adverbs (mood, modality, aspect, etc.) will be in a strict order based on the universal order of the corresponding functional projections. The lower adverbs (manner, means, purpose, etc.) can be reordered from sentence to sentence, but their linear order in any given utterance is based on a spec-­‐head relationship. Both the fixed-­‐order adverbs and the lower reorderable ones are in spec-­‐head relationships. Their linear order with respect to the non-­‐adverbs in the structure and each other is clear from the c-­‐command relationships based on Kayne’s LCA. There should not be a difference in the linearization behavior of higher adverbs and lower adverbs once the syntactic phrase structure is determined. Harmon 12 3 About ASL 3.1 Data Collection ASL is the language used by native signers belonging to the United States Deaf community. The data discussed below comes from published sources as well as native signers. The majority of ASL signers are nonnative since only a small percentage of deaf individuals have deaf parents who have exposed them to sign language from birth. Elicitation of the ASL sentences was accomplished through an interpreter as well as through the signer’s knowledge of English. One possible concern is that elicitation in this manner could have induced sentences that were influenced by English input rather than ASL grammar. I am aware of this risk, but have attempted to minimize this by working with a speaker who has a strong Deaf identity and a desire to promote ASL. I have also confirmed my data with previously published sources. I am confident in my data also for the reason that the ASL sentences turn out to be quite different from the English inputs in the adverbial structures I am investigating. 3.2 Grammar of ASL Here I will present some aspects of ASL grammar that will be essential to the discussion that follows. For more extensive discussion of ASL syntax, see Neidle et al (2000) and Lillo and Sandler Martin (2006). Although surface word order can be variable though such mechanisms as topicalization, the normal word order for ASL declarative sentences is SVO (small capital English words are used here for ASL signs); as in example (17). (17) ZACH READ BOOK ‘Zach read the book.’ Like a word in a spoken language, signs can be inflected for person, number, tense, etc. A sign is defined by a movement, a location, and a handshape (Perlmutter 1992). Inflection changes or adds to the movement, location, or handshape. Every sign has a movement, and it is analogous to the nucleus of a spoken language syllable. A movement may be either “primary” or “secondary”; primary movement involves proximal joint movement, and secondary movement is defined by distal movement. For example, the movement of the sign CANDLE (18)1 is only the wiggling of the fingers, the most distal part of the arm. Many other signs have only primary movement, such as THINK (19), which uses shoulder/elbow movement to touch the index finger to the forehead. A monomorphemic sign can only have one primary movement (Sandler and Lillo-­‐Martin 2006). However, a sign can have both a primary and secondary movement, as in the sign fire, where the primary movement is a circular movement in front of the torso, and the secondary movement is wiggling the fingers for the duration of the sign. 1
Photos used by permission of Bill Vicars and www.lifeprint.com © 2004. All forms are given as the
(18) ‘candle’ (19) ‘think’ Harmon 13 (20) ‘fire’ Changes in movement such as reduplication or changes in speed or size can add aspectual or perhaps adverbial meaning to a verb. For example, a habitual aspect or the adverb ‘often’ can be accomplished with reduplication of the verb movement. While non-­‐manual elements of manner adverbs will be discussed in 4.4 below, it is important to note that changes in verb movement can also accompany a manner adverb. ‘Quickly’ is accomplished with a “clenched teeth” facial expression and is accompanied with a faster-­‐than-­‐normal articulation of the verb movement. 3.3 Iconicity This kind of change in verb movement brings up a discussion of iconicity that is outside the scope of this paper, but is interesting to consider. A discussion of the visio-­‐spatial modality is important to the study of ASL, but it should not be a great concern for the structures here. While the articulation of some signs may be iconically motivated, the combination of these signs into syntactic structures is not significantly different from the combination of spoken morphemes and words into syntactic structures. In fact, language produced with the hands and body is strikingly similar to spoken language as such phenomena as recursive embedding and syntactic movement are readily observable. 4 The ASL Data 4.1 Strictly ordered high adverbs Many adverbs in ASL follow a strict ordering as Cinque (1999) predicts. These adverbs, like ‘probably’, ‘fortunately’, and ‘apparently’, can appear in more than one position in a sentence, however, their order with respect to each other is fixed. These adverbs can occur at the beginning of the sentence, at the end of the sentence, or directly before the verb. (21) a. SEEM ZACH READ BOOK b. ZACH READ BOOK SEEM c. ZACH SEEM READ BOOK ‘Apparently, Zach read the book.’ The position where the adverb is not on the edge of the utterance, c., has a slightly different meaning from a. or b. According to the speakers interviewed, when adverbs appear at the beginning or end of the sentence, there is a “pause” between Harmon 14 the adverb and the rest of the sentence; they are “separate comments”. This may be an indication that the initial and final adverb positions can be analyzed as separate utterances. The meaning difference is subtle, and is perhaps an indication of emphasis where putting the adverb in “a separate comment” highlights the adverb. When more than one adverb occurs in a sentence, the universal ordering of adverbs as shown in Cinque (1999) holds. (22) a. SADLY SEEM EMMA ACCIDENT b. *SEEM SADLY EMMA ACCIDENT Unfortunately, apparently Emma crashed her car. In (22), the evaluative mood adverb SADLY ‘unfortunately’ must precede the evidential mood adverb SEEM ‘apparently’. This agrees with the ordering of adverbs and their respective functional projections given in Cinque (1999). (23) a. LUCKY MAYBE ZACH READ BOOK b. *MAYBE LUCKY ZACH READ BOOK Fortunately, Zach probably read the book. In (23), the evaluative mood adverb LUCKY ‘fortunately’ must precede the epistemic modality adverb MAYBE ‘probably’. This also agrees with the universal order of adverbs described by Cinque (1999). The adverb ordering restrictions also hold when the adverbs appear after the subject: (24) a. ZACH LUCKY MAYBE READ BOOK b. *ZACH MAYBE LUCKY READ BOOK Fortunately, Zach probably read the book. (24) shows that the evaluative mood adverb LUCKY ‘fortunately’ must precede the epistemic modality adverb MAYBE ‘probably’, even when the adverbs must be analyzed as being a part of the main sentence. In (23) it could be possible to analyze one or both of the adverbs as being “a separate comment”, as the speaker described it. The correspondence to the universal order is again confirmed in (25), where the evidential mood adverb SEEM ‘apparently’ precedes the epistemic modality adverb MAYBE ‘probably’. (25) a. EMMA SEEM MAYBE CAR CRASH b. *EMMA MAYBE SEEM CAR CRASH Apparently, Emma probably crashed her car. The order of the adverbs corresponds to the order of functional projections given by Cinque (1999): (26) [Moodevaluative [Moodevidential [Modepistemic Harmon 15 4.2 Temporal adverbs Adverbs that place an utterance in time do not necessarily follow strict ordering relationships. Temporal adverbs, like YESTERDAY, can occur in the same places as other adverbs (i.e. initially, finally, between subject and verb). (27) a. YESTERDAY ZACH READ BOOK b. ZACH YESTERDAY READ BOOK c. ZACH READ BOOK YESTERDAY Yesterday, Zach read the book. The initial and final orders have the same ‘pause’ between the adverb and the rest of the utterance. The use of YESTERDAY places the sentence in the relative past (c.f. Zucchi 2009 for Italian Sign Language), but there can be other tense marking as well (Neidle et al 1995). When temporal adverbs occur with other adverbs, their order is variable. (28) a. YESTERDAY LUCKY ZACH READ BOOK b. LUCKY YESTERDAY ZACH READ BOOK Fortunately, yesterday Zach read the book. In (28), it is grammatical for YESTERDAY to appear either before or after the evaluative mood adverb LUCKY ‘fortunately’. The perfect aspect adverb ALWAYS must follow the higher adverbs of mood and modality. (29) a. ZACH MAYBE ALWAYS READ BOOK b. *ZACH ALWAYS MAYBE READ BOOK Zach probably always reads the book. Here the epistemic modality adverb MAYBE ‘probably’ must precede the perfect aspect adverb ALWAYS. This also corresponds to the universal hierarchy of functional projections in Cinque (1999). The modality phrase is higher than the aspect phrase. In (30) below, the temporal aspect adverb ALWAYS must precede the generic aspect adverb DAILY. (30) a. ZACH ALWAYS DAILY READ BOOK b. *ZACH DAILY ALWAYS READ BOOK Zach always daily reads the book. This ordering also reflects the ordering of functional projections. For the adverbs described above, Cinque’s (1999) universal order holds: (31) [Moodevaluative [Moodevidential [Modepistemic[Aspperfect [Aspgeneric Harmon 16 4.3 Adverb or aspect marker? Many of the meanings expressed by various adverbs in Cinque (1999) are not expressed with separate lexical signs in ASL. In the universal hierarchy of functional projections, Cinque (1999) lists several types of Aspect (habitual, frequentative, etc.) that occur in a strict order with corresponding adverbs. Instead of using an adverb to express these aspectual meanings, ASL can use aspectual marking directly on the verb. This marking is expressed by modifying the movement of the verb sign itself. This can be seen when the movement of the verb is reduplicated and the resulting sentence gains the meaning ‘often’, a frequentative aspect. (32) ZACH READ-­‐red BOOK ‘Zach often read the book.’ In section 4.4 below, there are adverbs that are marked simultaneously with the verb. Reduplicating the verb movement for the meaning ‘often’ could be analyzed as such an adverb, but it is perhaps better analyzed as morphological aspect marking. It has been shown that tense can be marked simultaneously with the verb in ASL and Italian Sign Language (LIS) (Neidle et al 1995, Zucchi 2009). A similar mechanism could be employed for the verb-­‐movement modifications found in these aspect-­‐marking constructions. This kind of aspect marking/adverb can co-­‐occur with other adverbs, even if those other adverbs are also at the aspect level. (33) ZACH ALWAYS READ-­‐red BOOK Zach always often read the book. Here the aspect adverb ALWAYS precedes the verb, which is modified by reduplicating the movement adding the aspect ‘often’. This can tell us that the adverb is not in the aspectual head position, since that position is presumably already filled by the reduplication aspect marking. 4.4 Non-­‐manual manner adverbs Manner adverbs can occur as non-­‐manual facial movements or as changes to the verb movement itself. Changes to the verb movement can be analyzed as aspect marking and not as adverbs, as with ‘often’ above. There are some adverbs that are expressed solely with facial movements. These facial movements have meanings that are not easily attributable to a functional projection like aspect or mood. The adverb ‘carelessly’, glossed “th” in the literature (Baker Shenk and Cokey 1980), is produced with the mouth slightly open and the tongue slightly protruding between the teeth, the cheeks are also puffed. This adverb occurs simultaneously with the verb. (34) th EMMA ACCIDENT ‘Emma carelessly crashed her car’ Harmon 17 ‘Carelessly’ is not the meaning that would be expected from aspect marking. It is a contentful manner adverb. However, it is not marked with a manual sign, it is marked with a facial expression. Facial adverbs like the ‘clenched teeth’ expression used with the adverb ‘quickly’ can also occur with changes in verb movement. This facial expression is articulated with spread lips and clenched teeth. This expression can be used to convey the meaning ‘quickly’. However, the verb movement is also articulated relatively more quickly. The faster verb movement may be analyzed as aspect marking. (35) clench ZACH READ-­‐ fast BOOK ‘Zach quickly read the book.’ In (35), the non-­‐manual ‘clench’ occurs for the duration of the (quickly-­‐articulated) verb, and the resulting meaning is ‘quickly read’. In (34) and (35) the non-­‐manual adverbs occur for the duration of the verb they modify. There is some evidence that the non-­‐manual adverbs can occur over the object as well. (36) furrowed brow HE WROTE COMPLAINT ‘He furiously wrote the complaint.’ This non-­‐manual is articulated with the eyebrows lowered and the lips pressed together. It has the meaning ‘furiously’ or ‘with concentration’. In this utterance, the same non-­‐manual occurs simultaneously with the verb and with the object, COMPLAINT. This could be analyzed as the adverb having scope over both and marking the entire duration of the VP, or the ‘furrowed brow’ non-­‐manual could be modifying COMPLAINT separately. One finding of the paper is that while manner adverbs can occur non-­‐
manually, all of the higher adverbs have lexical manual signs. While a difference in the behavior of high adverbs and low adverbs has been described before, as in Cinque (1999) for his “circumstantial” adverbs, the data presented here show a clear difference in articulation. 5 Analysis of ASL data There is a difference in the articulation of higher adverbs, which have manual signs, and the lower manner adverbs, which can be expressed non-­‐manually. 5.1 Higher adverbs The higher adverbs, like ‘fortunately’ and ‘probably’ show precedence relationships. They occur in a fixed order within the sentence, and the manual signs must be linearized with respect to other manual signs. I adopt Cinque’s (1999) analysis of Harmon 18 these adverbs. So, these adverbs are in the specifiers of functional projections, and the functional projections are in a fixed hierarchy. (37) ‘fortunately, probably read the book’ So here, the adverb LUCKY ‘fortunately’ is in the Spec of the evaluative mood phrase, and the adverb MAYBE ‘probably’ is in the Spec of the lower epistemic mood phrase. Linearization of these elements is clear as the asymmetrical relationships are clear from c-­‐command. The Specifier asymmetrically c-­‐commands the head of the phrase, and the different phrases are in asymmetrical relationships with each other. Following Hornstein’s Concatenate and Label analysis, the adverb SEEM is concatenated to the sentence and the structure is labeled as an evidential mood phrase. The evidential mood phrase is concatenated to the higher evaluative mood and the structure is labeled as the evaluative mood phrase. The higher adverb LUCKY is concatenated to the evaluative mood phrase and the structure is labeled as the evaluative mood phrase. In (33) above, aspect was marked by reduplication on the verb, and there was also a perfect aspect adverb ALWAYS. This can be analyzed as having the reduplication morpheme being in the head of the aspect phrase, and the adverb ALWAYS in the specifier of the aspect phrase. Harmon 19 (38) This corresponds to Cinque’s (1999) structures, were adverbs are in the specifier of functional projections, and the functional heads can be filled separately. Linearization happens with these structures since everything is in asymmetrical relationships. The adverb ALWAYS asymmetrically c-­‐commands the aspect head, so, using Kayne’s LCA, the adverbs precedes the head (and the verb). What is important here is that the higher adverbs are in a fixed order based on their semantic scope. These adverbs appear in the same order as adverbs in other languages. They fit into the framework of Cinque’s (1999) universal order of adverbs, and are linearized by that fixed order of functional projections. High manual adverbs in ASL fit into phrase structure in Spec-­‐head relationships with functional projections. Linearization of these functional projections and their corresponding adverbs is determined by the asymmetries created by the hierarchical structure. 5.2 Lower Adverbs Lower manner adverbs show no precedence relationship with the verbs they modify. They are linearized with respect to the rest of the sentence. That is, they are linearized with respect to the rest of the sentence as if they and the verb are in the same place in the asymmetrical hierarchy of functional projections. If these adverbs were in Spec-­‐head relationships, say with vP as Cinque (1999) has, then there should be a clear linearization between the elements. They can be reordered with respect to other such low adverbs in the specifiers of other vPs, but because asymmetrical relationships between the vPs are established for a given structure, linear order should be clear. Because there is not a clear linear order between the verb and the adverb when they are simultaneous, the asymmetrical structure for the lower adverbs in Cinque (1999) shown in (14) above does not adequately describe what is happening with non-­‐manual manner adverbs in ASL. If there is a clear asymmetrical hierarchy, why is there no clear linear order? However, if the non-­‐manual adverbs are analyzed as Hornstein’s unlabeled adjuncts, the mechanism of linear order is not asymmetrical structure. Hornstein’s (2008) concatenation without labeling is a mechanism that allows for an adverb to attach to a sentence but not be given a linear order with respect to what it modifies. (24)
Harmon 20 T’
vP
V
CRASHED
AdvP
DP
th
CAR
In (24) we have the “dangling off” adverb structure due to the adverb th ‘carelessly’ concatenating to the verb ‘crashed’ without being labeled. The asymmetrical relationships between the adverb+verb structure and the rest of the sentence allow it to be linearized with respect to the rest of the sentence. Because a higher-­‐up constituent (TP) is labeled, creating an asymmetrical structure, the adverb+verb structure can be linearized with respect to the subject and object and other elements in the sentence. However, because there is no label on the adverb+verb structure, the asymmetrical relationship mechanism of linearization cannot occur. Lower adverbs fit into phrase structure as constituents that are concatenated but not labeled. They are allowed to do this because adjunct adverbs do not need to be labeled to be interpreted. Hornstein (2008) says that labeled structures can be linearized based on asymmetries in the structure, however a different mechanism is needed to linearize the unlabeled structures. The ASL data show a difference in linearization between the higher adverbs and the manner adverbs. The difference in linearization points to a difference in phrase structure. Why are the manner adverbs not linearized like the higher adverbs? What is the difference in structure that allows the manner adverbs to occur simultaneously? The higher adverbs are linearized based on the Spec-­‐head relationships, while the manner adverbs cannot be linearized with respect to the verbs based on asymmetrical relationships alone. There is no asymmetrical relationship. Hornstein’s (2008) “concatenate without labeling” mechanism allows adverbs to attach to the sentence and be interpreted, but it does not provide a mechanism of linearization. In ASL it is possible to see this difference in structure because the simultaneous adverb+verb structure shows no linearization of the adverb with respect to the verb. 6 Other simultaneous expressions While a claim about the structure of non-­‐manual manner adverbs has been made, there are other non-­‐manual expressions that should be accounted for in futher studies. These other non-­‐manual expressions are presumably in asymmetrical hierarchical relationships, such as the relationship of NegP to the rest of the TP, or the relationship of TP to VP. Why can some things be in a non-­‐linear relationship, even when they have a clear asymmetrical (and thus easily linearized) relationship? Harmon 21 What cannot be said, then is the strong claim that if something is non-­‐manual, then it must be an adjunct (be non-­‐labeled). 6.1 Negation Negation can also occur non-­‐manually. There is some discussion in the literature of non-­‐manual negation occurring alone (Neidle et al 2000), however, according to the speakers interviewed, non-­‐manual negation must always occur with manual negation. Manual signs like NOT or NO occur before the negated constituent, and the non-­‐manual negation marking occurs for the duration of the negation. The NegP is fairly high in the clause structure, appearing above AgrP (Neidle et al 2000). The non-­‐manual negation marks the scope of negation, much like the ne pas construction in French (Neidle et al 2000). Negation could also perhaps be analyzed as a VP adjunct. Under this analysis, the negation sign would be an adverbial, and it would be adjoined at the VP or vP level (Zeijlstra 2004). This would allow the analyses of adjunction given above to be used. 6.2 Tense Tense has been analyzed as a simultaneous morpheme in Italian Sign Language (LIS). The non-­‐manual tense occurs at the same time as the verb. Can this be an adjunction relationship? Tense is most commonly analyzed as a functional projection above the verb. What other mechanisms could account for simultaneity? Perhaps the tense marking is a simultaneous affix, and the verb moves to the T head to be marked for Tense. 7 Conclusion The fact that adverbs in ASL can be articulated either as manual signs or as non-­‐
manual morphemes reveals a difference in the phrase structure of adverbs that is not readily visible in spoken language. Analyzing the higher manual adverbs using a fixed hierarchy of phrases (Cinque 1999) captures their fixed order and further shows the universality of such adverbs. The non-­‐manual adverbs, when analyzed as unlabeled concatenation/adjunction, provide evidence for a different mechanism of linearization. These non-­‐manual adverbs show us that strict precedence is not always required with adjunction. While the higher adverbs clearly have a linear order based on asymmetrical structural relationships, the linearization of the adjoined adverbs is not specified in the structure. The ASL data shows this difference in linearization: the higher manual adverbs are in a fixed linear order, while the precedence relationship between the verb and non-­‐manual adverbs is unclear. Harmon 22 References BAKER-SHENK, CHARLOTTE LEE. 1991 American Sign Language: A teacher's resource text on grammar
and culture. Gallaudet University Press.
BICKFORD, ALBERT 2006 Mouth Morphemes in ASL. Theoretical Issues in Sign Language Research
Conference, Florianopolis, Brazil
BRIDGES, BYRON AND MELANIE METZGER. 1996. Deaf tend you: Non-manual signals in ASL. Silver Spring
MD: Calliope Press.
CHAMBERS, DIANE P. 1998. Communicating in Sign. Brooklyn, NY. Fireside.
CHOMSKY, NOAM 1986. Barriers. Cambridge: MIT Press.
CHOMSKY, NOAM 1995. The Minimalist Program. Cambridge: MIT Press.
CHOMSKY, N. 2000. “Minimalist inquiries: The framework,” in R. Martin, D. Michaels and J. Uriagereka
(eds.), Step by step. Cambridge, MA: MIT Press, 91–155.
LIDDELL, SCOTT K. 1980. American Sign Language Syntax. The Hague: Mouton.
NEIDLE, C. 2000. The Syntax of American Sign Language : Functional Categories and Hierarchical Structure. Cambridge, Ma: MIT Press. PERLMUTTER, DAVID 1992. Sonority and syllable structure in American Sign Language. Linguistic Inquiry
23:407-442. Reprinted as: Sonority and Syllable structure in American Sign Language. In Phonetics and
Phonology, Volume 3 Current Issues in ASL Phonology, ed. G Coulter. San Diego, CA: Academic Press
(1993).
POLLOCK, JEAN-YVES. 1989. “Verb Movement, Universal Grammar, and the Structure of IP”. Linguistic
Inquiry 20:365-424.
SANDLER, WENDY AND LILLO-MARTIN, DIANE. 2006. Sign language and linguistic universals. Cambridge
University Press.
STRUXNESS, KEVIN. 1996. Mouth morphemes in American Sign Language. Video. DeBee
Communications.
VICARS, BILL 2004. American Sign Language University. Lifeprint.com.
ZEIJLSTRA, HEDDE H. 2004 Sentential Negation and Negative Concord. PhD Dissertation. Univesity of Amsterdam. Utrecht. LOT Publications.