Available online at www.sciencedirect.

com

Lingua 122 (2012) 979--1003

www.elsevier.com/locate/lingua

The syntax of simultaneity§

Joost Kremers *
University of Göttingen, German Department, Käte-Hamburger-Weg 3, 37073 Göttingen, Germany
Received 24 March 2011; received in revised form 21 March 2012; accepted 23 March 2012
Available online 9 May 2012

Abstract
Although language utterances primarily consist of sequences of phonemes, morphemes and words, language is not limited to
linearity. There are many linguistic phenomena that show a certain degree of simultaneity: cases in which two (or more) meaningful
elements are realized at the same time. Typical cases include intonation contours to mark e.g., focus, topic, questions, etc. While it may
be argued that such examples are marginal from the perspective of syntax and can therefore be safely ignored, the same cannot be said
for sign languages. Sign languages make extensive use of simultaneity, using it even for negation and adverbial modification,
phenomena that are generally considered to be syntactic. Trying to analyze simultaneous constructions syntactically, we run into
the problem that under standard assumptions, all terminal elements in a syntactic tree have a unique position in the linear string
(Exclusivity & Nontangling Conditions, Partee et al., 1993). The sign language examples obviously violate these conditions. In order to
account for these data, we need to extend ideas behind prosodic morphology to syntax: not only morphemes but also syntactic elements
can map onto prosodic forms in phonology. Standard principles of the syntax/phonology interface and of the phonological module then
apply to construct the simultaneous expression.
© 2012 Elsevier B.V. All rights reserved.

Keywords: Syntax/phonology interface; Simultaneity; Sign language; Prosodic morphology

1. Simultaneity

Language is linear. A linguistic utterance consists of a temporally ordered sequence of meaningful units (‘‘words’’),
which in turn consist of a temporally ordered sequence of non-meaningful units (phonemes).
Although this statement is not false, it is not entirely true either. Language is indeed linear, but it would be wrong to say
that meaningful elements are all sequentially ordered. Many languages use intonation contours to mark things such as
focus, topic, and questions. While the actual intonation contour is generally not considered to be a syntactic element, it
cannot be denied that it is meaningful in some sense.
Similarly, prosodic morphology shows non-linear aspects. Take the well-known Arabic example of kataba ‘he wrote’
(cf. McCarthy, 1981, 1986):

§
The research reported on in this paper was funded by the German Research Foundation (DFG), grant number GR 559/8-1. I thank Günther
Grewendorf for making this research possible. I would also like to thank Annika Herrmann, Jana Hosemann, Markus Steinbach and Eva
Waleschkowski for teaching me about sign language linguistics and for our discussions of the issues in this paper. My appreciation and thanks
also go to the Lingua reviewers, whose comments have been very beneficial. Of course, any and all remaining errors are mine.
* Tel.: þ49 551 39 4467.
E-mail address: [email protected].

0024-3841/$ -- see front matter © 2012 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.lingua.2012.03.008
980 J. Kremers / Lingua 122 (2012) 979--1003

(1) k t b

[TD$INLE]
× × × × ×

The two morphemes «ktb» (verb root) and «a» (perfective) are not sequentially ordered with respect to each other.
Obviously, the phonemes are sequentially ordered, but the phonemes are not the meaning-bearing units.
It may be argued that such examples are marginal from the perspective of syntax and can therefore be safely ignored,
because they may tell us something about the morphology module of language but not about the syntax module.1 There
are natural languages, however, for which this claim certainly cannot be made.
Bellugi and Fischer (1972) show that while individual (manual) signs in American Sign Language (ASL) generally take
longer to produce than words in spoken English,2 ASL utterances are generally about as long as spoken English
utterances. One way in which sign languages achieve this result is through what is often called simultaneity: the
simultaneous realization of two or more meaningful elements.
The following example from German Sign Language (DGS) illustrates this:

(2) face: with effort DGS

hands: STUDENT SIGN-LANGUAGE LEARN
‘The student learns sign language with difficulty.’
(Leuninger, 2005)

In this example, the manual sign LEARN is accompanied by a facial expression that expresses the adverbial with effort,
difficultly. That is, verb and adverb are realized simultaneously.
Note that there are in fact examples of simultaneity in spoken language. For example, Hayes and Lahiri (1991) treat
intonational patterns as morphemes, and Akinlabi's (2011) featural affixation can be considered a form of simultaneity as
well. Also, certain African languages use tone as a grammatical marker. In Tiv, for example, tone patterns mark tense on
verbs (see Goldsmith, 1976:4.2 and references there) and in Maasai, tone is used to mark case:

(3) a.
ε-d ɔ
l εmba
rt a

Maasai
3-see horse.ACC
‘He sees the horse.’
b.
ε-d ɔ
l εmbart a

3-see horse.NOM
‘The horse sees him’
(Tucker and Tompo Ole Mpaayei, 1955)

Maasai has four major tone classes in the nominal system, all with subclasses, making about 20 different noun tone
classes. The noun εmba
rt a

3 ‘horse’ belongs to a tone class that lowers the high tone on the penultimate vowel in the
nominative, as indicated in (3b).
The fact that such constructions exist obviously raises many questions, some of which are being addressed in the
literature (see especially Vermeerbergen et al., 2007). Some authors have made proposals for representing non-manual
elements syntactically (e.g., Neidle et al., 1998; Wilbur and Patschke, 1999; Cecchetto et al., 2009) but no unified proposal
exists that attempts to handle the phenomenon of simultaneity in general. The aim of this paper is to propose such an
analysis.
The most pressing question that these examples raise regarding syntax is how standard approaches to linearity in
syntax can deal with them. Theoretical approaches to syntax generally assume that a syntactic structure uniquely defines

1
Unless one assumes that word formation optionally or exclusively takes place in the syntactic module, as argued for by e.g., Halle and Marantz
(1993), Ackema (1995), Julien (2002) and others.
2
Emmorey (1995) makes a similar claim. Although I am not aware of similar studies on other sign/spoken language pairs, it seems safe to
assume that this is a general tendency.
3
Note that the accusative is the citation form of the noun in Maasai.
 J. Kremers / Lingua 122 (2012) 979--1003 981

a linear order, which they do either as an integral part of syntactic structure (e.g., Government & Binding) or through a
separate principle or module that derives linear order from a syntactic tree (GPSG, HPSG, minimalism).
This requirement is formalized as the Exclusivity Condition in Partee et al. (1993) (see section 3 below). Kayne (1994)
incorporates the same idea in his Linear Correspondence Axiom when he says that linear order is total. The example in
(2), however, does not seem to obey this principle: no linear order is defined between the adverbial and the verb.
While the adverb in (2) may be argued to be a morphological marker, the same cannot be claimed for the negation
in (4):

(4) head: neg

hands: JOHN BUY HOUSE

‘John is not buying a house’
(Neidle et al., 2000:44ff)

In this example, the negation is expressed through a head shake accompanying the entire VP. As such, it cannot be
considered a morphological marking of the verb (‘morphological’ in the sense of ‘relating to word structure’). Instead,
we seem to be forced to assume a syntactic analysis (also in the sense of ‘related to phrasal structure’). The point is,
as syntacticians, we cannot argue that simultaneity is something that we do not need to be concerned about as it only
occurs inside words. That being the case, we need to find a way to deal with these phenomena, because, as
mentioned, they seem to contradict standard syntactic views on the relation between syntactic structure and
phonological form.
The answer is clearly to be sought in the fact that the adverb and the negation in these examples are realized on
autosegmental phonological tiers. For this reason, I propose to extend the ideas behind McCarthy and Prince's (1996)
model of prosodic morphology to syntax: syntactic heads have the ability to map onto autosegmental phonological forms,
which are integrated with the segmental material.
This integration process is controlled by principles of phonology, syntax plays no direct role in it. Obviously, the
constituent structure that is constructed in syntax is relevant for the phonological form that is eventually created (if it were
not, there would be no such thing as language), but the role of syntax is indirect, because it is mediated by principles
governing the mapping from syntax to phonology.
The view of grammar that I adopt is inspired by Beard's (1988) Separation Hypothesis, according to which the
morphosyntactic and phonological features of an element should be considered separately, and by Jackendoff's (1997,
2002) parallel grammar architecture (as discussed in section 4.1). A central aspect of this model is that the phonological
form onto which a given morphosyntactic structure is mapped is essential for the analysis of linguistic phenomena.
Phonological structure plays an active role in the formation of linguistic utterances: the hierarchical constituent structure of
the negation example in (4) above does not fully determine the phonological form of the utterance, not even its linear order.
The phonological form onto which the Neg8 head is mapped is crucial for this.
In this paper, I discuss several simultaneity phenomena in sign language (section 2), focusing on the question how the
two (or possibly more) simultaneously realized elements relate to each other in terms of semantics, syntax and phonology.
The phenomena discussed by no means exhaust the range of possible simultaneity effects in sign language, but they
represent a good cross section of the relevant phenomena. Then, after briefly discussing the problems that simultaneity
poses for the Exclusivity and Nontangling Conditions (Partee et al., 1993), I turn to the model that I propose in section 4,
arguing that in order to account for the data discussed, we need to consider the whole process of what I term phonological
composition, which takes its input from morphosyntax and establishes a phonological structure on the basis of this input, a
process that is constrained by certain mapping principles of the syntax--phonology interface and by specific phonological
principles. In section 5, I discuss how the two mapping principles Linear and Input Correspondence (Ackema and
Neeleman, 2004), which are crucial for the analysis of simultaneity data, interact. Finally, in section 6, I discuss some
possible directions for future research.

2. Cases of simultaneity

Let us now turn to the data. I discuss adverbials, negation, interrogative marking and topic marking in turn. Note that it is
not the purpose of this paper to present an overview and classification of all types of simultaneity that occur in sign
languages. Rather, I limit myself to a number of cases that one would at first sight be inclined to analyze syntactically.
Admittedly, the inclination to treat these phenomena syntactically is due to a spoken-language bias and we obviously
should not expect sign language to express the relevant phenomena syntactically as well. However, choosing such
examples makes it somewhat easier to argue for the analysis that I propose, because we know that the relevant
phenomena at least can be syntactic (since they are so in spoken languages).
982 J. Kremers / Lingua 122 (2012) 979--1003

Note, however, that the analysis that I propose for the sign language phenomena to be discussed is not purely
syntactic. Rather, it is a combination of a syntactic and a prosodic analysis. The central point of this paper is that we should
strictly separate between morphosyntactic and phonological features but that we cannot achieve a descriptively and
explanatorily adequate theory unless we consider both. The sign language data discussed in this paper makes this point
quite clear.

2.1. Adverbials

I already discussed examples of simultaneous adverbs in the introduction. The example in (2) is repeated here as (5):

(5) face: with effort DGS

hands: STUDENT SIGN-LANGUAGE LEARN
‘The student learns sign language with difficulty.’
(Leuninger, 2005)

Non-manual adverbs also occur in other sign languages. The following examples are from ASL:

(6) a. face: /mm/ ASL

hands: HER HUSBAND COOK-[dur] DINNER
‘Her husband has been cooking the dinner with pleasure’
b. face: /th/

hands: HER HUSBAND COOK-[dur] DINNER

‘Her husband has been cooking the dinner inattentively’
(Corina et al., 1999:310)

/mm/ and /th/ represent different facial expressions, associated with the respective adverbial meanings indicated in the
translations. In fact, Anderson and Reilly (1998) identify eleven different facial adverbials for ASL. Similar examples can
be found in other sign languages: Kyle and Woll (1985:86) discuss a number of facial expressions that contribute adverbial
(or in some cases adjectival) meaning in British Sign Language (BSL), Meir and Sandler (2008:173--176) report non-
manual adjectives and adverbs for Israeli Sign Language (ISL) and Johnston and Schembri (2007:150) discuss facial
adverbials in Australian Sign Language (Auslan).
The phenomenon is therefore widespread in sign languages. Importantly, as most authors note, the forms and
meanings of these facial expressions are conventionalized, they differ from one sign language to the next, and they
contribute to the meaning of the utterance: leaving them out changes the meaning of the phrase. That is, as already
argued for by Liddell (1980), these facial expressions are not mere gestures, they are linguistic and we therefore expect
them to be represented in the syntactic structure underlying the utterance.4
Non-manual adverbs may also be expressed in other ways than through facial expressions, and furthermore, they may
be combined:

(7) body: exclusive DGS

face: reluctant

hands: APPROACH-EACH-OTHER.CluprightG.slow
‘two people approach each other slowly, reluctantly, and with hostility
(Leuninger et al., 2005:331)

This example contains two or possibly three simultaneous adverbs. The manual sign is a two-handed sign, with both
hands consisting of an upright G-handshape (stretched index finger). This handshape functions as a classifier for human
beings. The two hands move toward each other, expressing the meaning approach each other. Two non-manual
adverbials are expressed in (7): the upper body of the signer is leaned backwards, a body posture that expresses

4
Another interesting cross-linguistic characteristic of these facial adverbials, pointed out by a reviewer, is that they all involve the lower part of
the face, in particular the mouth and cheeks. The upper parts of the face tend to mark sentence types and properties such as topic and focus
(cf. Wilbur, 2000; Meir and Sandler, 2008:175).
 J. Kremers / Lingua 122 (2012) 979--1003 983

exclusivity, which is understood as hostility in the current case (cf. Wilbur and Patschke, 1998, who analyze backward tilt
body posture as expressing exclusivity). Furthermore, the signer's facial expression expresses the meaning reluctant.
Additionally, the movement of the hands is realized more slowly than usual, which expresses the adverbial slowly.
Because of the obvious iconic nature of this part of the utterance, it is difficult to tell whether it is merely gestural or whether
it is truly morphemic. For the present paper, however, this question is not crucial, as the other examples of adverbial
modification are clearly morphemic.
One might perhaps argue that these adverbials are morphological, not syntactic (in fact, Sandler and Lillo-Martin,
2006:61--63 suggest as much), especially considering the fact that they are realized simultaneously with the verb, not the
entire VP. However, although this may be a possible analysis for the non-manual adverbs demonstrated so far, there are
non-manual adverbials that cannot be analyzed morphologically. Happ and Vorköper (2006:363) give the following
examples from DGS:

(8) a. f/b: presumably DGS

h: (POSSIBLE) SVEN WORK:3 GO:Perf:3
‘Presumably, Sven has gone to work already.’
b. f/b: possible
h: (POSSIBLE) SVEN WORK:3 GO:Perf:3
‘Sven has possibly gone to work already.’

The examples in (8) contain sentential adverbs. Specifically, there are two components to the adverbs: a manual one,
5
POSSIBLE, which is optional, and a non-manual one, a combined facial expression and body position. The two components
are independent from each other: the manual component is optional and the two do not need to share the same meaning.6
The relevant aspect of these examples is the fact that the non-manual component spreads over the entire clause. As
with the VP adverbials above, these non-manual sentential adverbials contribute crucial information to the utterance and
their forms are conventionalized. We therefore expect them to be represented syntactically.
Similarly, there are adverbials in ASL that can spread over phrases. In (9), for example, from Sandler and Lillo-Martin
(2006:470), the non-manual adverbial obviously spreads over the phrase OBVIOUSLY SMART:

(9) face: whq obv whq ASL

hands: WHO YOU FEEL OBVIOUSLY SMART (WHO IX)
‘Who do you think is obviously smart?’

Since morphological analyses are by definition limited to the word/sign, these non-manuals cannot be morphological,
as they spread over more than one manual sign.
As regards their syntactic structure, for present purposes it suffices to concentrate on the general constituent structure
of the utterances. This means that I will remain agnostic as to the question whether adverbials are adjuncts or specifiers of
dedicated functional heads. The VP adverbs in examples such as (7) and (8) can then simply be represented as adjuncts
to VP:

(10) VP
VP
[TD$INLE]
AdvP
V ...

Similarly, sentential adverbs can simply be assumed to be adjoined to the CP. If we separate the manual and non-
manual components, we may be dealing with two adjuncts:

5
In fact, it is possible that the facial expression and body position contribute independently to the meaning. Such combinatoriality is often
assumed for different facial and eye brow configurations (e.g., Wilbur, 2000; Dachkovsky and Sandler, 2009). Note that even if there is a certain
combinatoriality in non-manual components, this system seems to lack the so-called duality of patterning of the segmental system, as the
independent components all have a specific meaning.
6
Possibly, the manual component expresses a general modality, which the non-manual component specifies further.
984 J. Kremers / Lingua 122 (2012) 979--1003

(11) CP

CP
Adv TP
[TD$INLE] Adv
C
presumably ...
POSSIBLE

Alternatively, we can adopt an analysis along cartographic lines. This provides a different way for dealing with the two
components, as it allows us to say that the manual component POSSIBLE is the phonological representation of an adverbial
head that has presumably in its specifier:

(12) CP
ModP
C TP
[TD$INLE] Adv
Mod
...
presumably

POSSIBLE

What is relevant here is the fact that the adverbials are part of the syntactic structure in a position c-commanding the
element(s) with which they are simultaneously realized. The LCA would therefore predict that they are realized sentence-
initially (in the case of sentential adverbs) and preverbally (in the case of VP adverbs) but obviously they are not.

2.2. Negation

Negation in sign language presents us with a similar picture. Different sign languages realize negation in different
ways, but most share the ability to realize negation through a simultaneous construction. Here, I will limit myself to
negation in DGS.
As Pfau and Pfau and Quer make clear in several papers,7 sentential negation in DGS is expressed through the
combination of a non-manual and a manual marker, the manual marker being optional:

(13) head: neg DGS

hands: MOTHER FLOWER BUY (NOT)
‘Mother does not buy a flower’
(Pfau and Quer, 2002)

The non-manual marking, here indicated as neg, consists of a headshake and is realized together with the verb and the
manual negation. If the manual negation is omitted, the headshake accompanies the verb alone. Note that the non-
manual negation must always be supported by manual material, it cannot appear on its own.
Pfau and Quer (2002) analyze the DGS data as follows: DGS is a language with split negation, i.e., it has two negative
elements that, when realized together, form a single negation, similar to French ne. . .pas. The manual negation NOT is one
element, the headshake associated with the verb8 is the other. NOT is in Spec,NegP and is lexically associated with
a headshake. The headshake is described by Pfau and Quer as a [þneg] feature sitting in the head position Neg8.
Because the [þneg] feature is affixal, the verb moves to Neg8 in order to support the headshake phonologically. Note
that DGS is an SOV language; Pfau and Quer assume that V and T are head-final, which means that this verb movement
is essentially string vacuous.

7
See Pfau (2001, 2002, 2008) and Pfau and Quer (2002, 2007) for details on negation in DGS, ASL and Catalan Sign Language (LSC); see
Pfau (2008) for some discussion of negation in a number of other sign languages; see also Zeshan (2006) for a typological overview of negation in
sign languages.
8
More accurately, with the predicate, as the non-manual negation can also be used to negate copular sentences, which do not contain a verb.
 J. Kremers / Lingua 122 (2012) 979--1003 985

As Pfau (2002, 2008) makes clear, it is also possible for the headshake to spread over the entire VP, even though this is
less common:

(14) head: neg DGS

hands: MOTHER FLOWER BUY
‘Mother does not buy a flower’
(Pfau and Quer, 2002)

What is important to note here is that both forms of negation in DGS are sentential, there is no difference in meaning
between (13) and (14). Pfau (2008) argues that the spreading of the headshake over the VP is phonological; that is, he
analyzes the headshake as an autosegment that may optionally spread over a phonological domain, specifically, the
phonological phrase. This analysis seems essentially correct, as indicated by the fact that the headshake may spread even
over the subject if this subject is a pronoun and therefore part of the same phonological phrase as the object and the verb:

(15) head: neg DGS

hands: IX:3 FLOWER BUY
‘She does not buy a flower’
(Pfau, 2008)

In this example, the subject of the clause is an indexical, that is, a pronominal element. Unstressed pronouns are cross-
linguistically phonologically weak, that is, they usually do not form their own prosodic word and criticize onto an adjacent
prosodic word. Assuming that this is true of the indexical, we predict that spreading of the headshake is possible with
indexical subjects. Pfau (2008:26) argues that this is indeed more common and provides the example in (15) as illustration.
This spreading property of negation in DGS contrasts with negation in certain other sign languages that do not allow the
headshake to spread. In Italian Sign Language (LIS), for example, the manual negation is obligatory and is combined with
a headshake, but this headshake cannot spread over the verb or the VP (Geraci, 2005). In Pfau's (2008) analysis, the
headshake in DGS is an autosegment, not a manual sign, which entails for him that it is not syntactic. The headshake that
accompanies the manual negation in LIS is lexically specified and therefore syntactic.9
Although I mostly concur with Pfau's (2008) analysis, I do not adopt this separation between a syntactic (lexical) and a
non-syntactic headshake, as it seems to entail that only manual signs are ‘‘syntactic’’ in his terms. I present my alternative
analysis in section 4.4; for the moment it suffices to note that negation, which under standard assumptions is represented in
syntax by a Neg8 head, can be realized simultaneously with the verb in DGS and other sign languages, violating the
Exclusivity Condition.

2.3. Interrogative marking

Another typical phenomenon that is usually expressed non-manually (and hence simultaneously) in sign language is
question marking. Interrogatives are normally marked by certain non-manual markers (which can take various forms,
depending on the type of interrogative and other factors; see Sandler and Lillo-Martin, 2006:459ff for details and further
references). In the case of polar questions, this is usually the only way the utterance is marked as interrogative:

(16) face: Q

hands: HE COME
‘Is he coming?’

The interrogative marking extends over the scope of the interrogative but excludes elements that are not in its scope:

(17) eyebrows: Q BSL

hands: K.I.L.B.Y BEFORE GOOD NOW GOOD
‘Kilby, who was good before, was he good now?’
(Kyle and Woll, 1985:156)

9
Note, however, that for Pfau the analysis must be either syntactic or prosodic. With the strict separation between morphosyntactic and
phonological features that I assume, however, there is no opposition between a syntactic and a prosodic analysis. Both aspects are relevant.
986 J. Kremers / Lingua 122 (2012) 979--1003

As has often been noted, such facial interrogative markers function in much the same way as interrogative intonation in
spoken languages:

(18) a. int: Q English

segm: Is he coming?
b. int: Q
segm: Kilby, who was so good before, was he good now?

In English, too, question intonation extends over the entire clause in (18a) but only over the second part in (18b). Such
intonational patterns are normally considered non-syntactic so that by analogy, a non-syntactic analysis seems preferable
for interrogative marking in sign languages as well.
This, however, raises the question in what way interrogative marking is different from negation and facial adverbials. All
three phenomena can be expressed non-manually in sign language, so would it not be preferable to have a similar
analysis for them? In fact, one thing standing in the way of treating interrogative marking as syntactic is our intuition that
syntax deals with the arrangement of words and that words consist of segments, an intuition that is formalized as the
Exclusivity Condition. For example, assuming that interrogatives are headed by a [þQ]-marked C head, the basic
structure of clauses such as (16) and (18a) is (19):

(19) CP
TP
[TD$INLE]
C
...
[+Q]

Given such a tree, the Exclusivity Condition would lead us to expect that if C8 is realized phonologically, it is a
segmental particle appearing either clause-initially or clause-finally. As I will argue in section 4.2, however, a different view
is possible: I assume that the intonational interrogative marking is a reflex of the [þQh] C8 head.10
In (17) and (18b), the interrogative marking does not extend over the entire utterance, but just over the actual
question. The element not marked for [Q] is a left-dislocated topic. In the English example, this topic is taken up by the
(resumptive) pronoun he. In the BSL example, no pronoun is visible but given that dropped subject and object
pronouns are not uncommon in many sign languages (similar to languages such as Chinese), a silent pronoun may be
present.
Let us assume that the left-dislocated phrase is adjoined to CP.11 Then the syntactic structure for (17) and (18b) is the
following:

(20) CP
CP
TP
[TD$INLE]DP
C
...
[+Q]

Based on this tree, it seems that the interrogative marking extends over the complement of the Q-marked head, i.e.,
over its c-command domain. In fact, Neidle et al. (2000) make such a claim for negation in ASL. However, as discussed
above, Pfau (2008) argues that the relevant spreading domain must be defined phonologically. Similarly, Sandler (2011)
argues on the basis of well-know mismatches between syntactic and phonological domains that spreading phenomena of

10
See Neidle et al. (2000), who make this assumption for ASL, and Cheng and Rooryck (2000) who propose that French has a similar
‘‘intonation morpheme’’ that licenses wh-in-situ and yes/no questions. Note also that I deliberately keep the analysis simple. There are still many
open questions regarding the analysis of interrogatives in sign language, cf. Sandler and Lillo-Martin (2006:431ff.) for discussion and further
references. See also Zeshan (2006) for a typological overview of interrogatives in sign language.
11
Nothing really hinges on this assumption, one may also argue that the left-dislocated phrase is in the specifier of some high functional head in
the C-domain.
 J. Kremers / Lingua 122 (2012) 979--1003 987

this type are constrained by phonological domains, not by syntactic ones. I adopt this analysis and assume that Q-marking
spreads in what is presumably an intonational phrase.12

2.4. Topics

Like negation and interrogatives, topics are often marked non-manually in sign language (cf. Sandler and Lillo-Martin,
2006:406--413 and references cited there). The phonological form of topic marking varies depending on the type of topic
marking (see Aarons, 1994:156 and Wilbur, 1994 for ASL) and (presumably) also varies form one sign language to the
next. A typical example from Danish Sign Language is (21):

(21) face: topic DSL

hands: NURSERY index:3 1:SEND: 3
‘I took (her) to the nursery’
(Engberg-Pedersen, 1994:75)

In (21), the signer first introduces the goal argument of the clause and uses an indexical to locate it in signing space.
Simultaneously, NURSERY and the indexical are marked non-manually as topic. Next, the signer signs the verb SEND, which
is a so-called agreeing verb13 that agrees with the subject (indicated with ‘1’ for first person) and the goal of the action,
which is the nursery. This locative agreement is indicated with the index ‘3’, which refers to the location in signing space
that is associated with the nursery.
Views on the syntactic position of topic markers differ. In syntactic approaches that are based on work by Rizzi (1997)
and Cinque (1999), topics appear in the specifier position of a dedicated Top8 head:

(22) CP
TopP
[TD$INLE]C
XP
Top
...

In the structure in (22), the XP would be topic. In languages that have an overt topic marker, the argument goes, this
marker is an overt realization of the Top8 head. However, if this is indeed the correct structure, the phonological marking of
the topic in sign languages (and in spoken languages that use intonation for this purpose) functions in a way that is very
different from interrogative marking and negation. Both interrogative marking and negation extend over their complements
(or, more precisely, over the phonological domain the complement is embedded in). The same is essentially true for non-
manual adverbs: VP adverbs are realized simultaneously with the verb, which is in the c-command domain of the adverb.
In the case of topics, however, the phonological marking associated with the Top8 head would extend over its specifier.14
However, there is data to suggest that this cannot be the only source of topic markers. If the topic marker is a Top-head
in the verbal projection line, we would expect that a topic marker can co-occur with a case marker on the DP in Spec,TopP.
However, this is not always the case. In Japanese, for example, the topic marker wa is in complementary distribution with
the case markers ga (nominative) and o (accusative):

(23) a.  ga
taiy o noboru Japanese
sun NOM rise
‘the sun rises’
Japanese
b.  (*ga) wa noboru
taiy o
sun NOM TOP rise
‘the sun rises’

12
It is standardly assumed that topics constitute separate intonational phrases (cf. Selkirk, 2005; Nespor et al., 2008). It then follows that the
domain in which the interrogative marking spreads in the examples in the text is also an intonational phrase.
13
Not all verbs in sign languages agree. Those that do can be divided into several different categories. See e.g., Sandler and Lillo-Martin (2006)
and references cited there for details.
14
As pointed out by a reviewer, the same would of course also be true of non-manual wh-marking that extends over a moved wh-phrase in Spec,
CP (Neidle et al., 2000).
988 J. Kremers / Lingua 122 (2012) 979--1003

The most straightforward analysis for these data is one in which the topic marker and the case marker (and presumably
also prepositions) are reflexes of the same functional head in the nominal projection line, as schematized in (24):

(24) CaseP
...
[TD$INLE] Case
N
[ TOP]

I will assume that the sign language examples under discussion have essentially this structure.15 On this analysis, the
non-manual topic marking would fall in line with the interrogative and negation markers: the topic marker is phonologically
realized simultaneously with the element in its scope, i.e., with its c-command domain. As above, I assume that the actual
domain is defined phonologically: the topic marking spreads over the phonological domain (presumably an intonational
phrase, cf. footnote 12) that contains the (phonological form associated with) the N8 head.

3. Exclusivity and Nontangling

As already mentioned in the introduction, for standard theories on the relation between hierarchical and linear structure
(i.e., in a sense between syntax and phonology), the simultaneity data are problematic. In most theories, syntactic
structures are represented as phrase structure trees, which represent the hierarchical relations between the elements in
the clause. Phrase structure trees define dominance relations, and in many versions also linear relations.16 Partee et al.
(1993) formulate two conditions on constituent structure trees that regulate the relation between hierarchical and linear
structure:

(25) a. The Exclusivity Condition:

In any well-formed constituent structure tree, for any nodes x and y, x and y stand in the precedence
relation P, i.e., either hx; yi 2 P or hy; xi 2 P, if and only if x and y do not stand in the dominance
relation D, i.e., neither hx; yi 2 D nor hy; xi 2 D.
b. The Nontangling Condition:
In any well-formed constituent structure tree, for any nodes x and y, if x precedes y, then all nodes
dominated by x precede all nodes dominated by y.
(Partee et al., 1993:442)

The Exclusivity Condition states that if two nodes do not dominate each other, one of them will precede the other. This
guarantees that a linear order is defined between any two terminal nodes. The Exclusivity Condition makes one implicit
assumption, which is that terminal nodes are associated with segmental material, or are phonologically null. To the extent
that this indeed applies in a given phrase structure tree, the Exclusivity Condition holds.
The Nontangling Condition ensures that if a node dominating a structure «abc» has a sister node dominating a
structure «d», the linear structure will be either «abcd» or «dabc», but not *«adbc» or *«abdc». Like the Exclusivity
Condition, the Nontangling Condition makes an implicit assumption, which is that the terminal (phonologically non-null)
elements are all realized on the same phonological tier. Again, to the extent that this assumption applies in a given phrase
structure tree, the Nontangling Condition holds.
However, in the simultaneity examples discussed in the previous section, it is not the case that every syntactic head
corresponds with segmental phonological material, nor is it the case that all phonologically non-null heads are realized on
the same tier. In such cases, the Exclusivity and Nontangling Conditions no longer hold. For example, when negation is
expressed as a headshake, Neg8 and VP are not linearly ordered, but at the same time neither dominates the other. The
same is true for the other simultaneity examples.
As already discussed, we cannot look to morphology for a way out. If it were the case that the non-manual negation, for
example, could only be associated with the verb, then morphology might have provided us with the means to account for

15
The actual category of the head carrying the [TOP]-feature may not be Case8. The label is inspired by the Japanese example, but since sign
languages generally do not have case marking, it is probably not the most appropriate label here.
16
Some theories, such as minimalism, HPSG and others, assume that linear order is defined by a principle or by rules that are independent from
phrase structure trees. For the remarks in this section, it is immaterial which view of the relation between dominance and linear order one
assumes.
 J. Kremers / Lingua 122 (2012) 979--1003 989

these data. As we have seen, however, the negation can be associated with the entire VP. A similar point can be made for
the sentential adverbs and in principle also for the interrogative marking.
Thus, the simultaneity data discussed above present us with a problem. There are certain aspects about the syntax--
phonology mapping that do not adhere to the standard view on phrase structure, as formalized in the Exclusivity and
Nontangling Conditions. The deeper reason for this problem is the common assumption that as far as syntax is concerned,
only a single phonological tier, the one with segmental material, is relevant. What the data in the previous section teach us
is that this assumption is not correct.

4. Phonological composition

4.1. Theoretical background

Before I present my analysis of the data, it is necessary to discuss the theoretical background that I assume, because it
helps clarify and simplify the discussion. First, let me make explicit what I mean by the term ‘‘syntactic analysis’’ that I have
been using above. By this, I mean an analysis in terms of structures composed of morphosyntactic heads, which are
themselves bundles of morphosyntactic features. Phonological features are not part of this syntactic structure (cf. Beard's
(1988) Separation Hypothesis). Rather, they are part of the phonological structure, which is built in parallel with but
independently from the syntactic structure.
This division of labor is inspired by Jackendoff (2002) and is in fact crucial to the analysis I present here. Simultaneity
phenomena do not lend themselves to a purely syntactic analysis that only takes the morphosyntactic features into
account. Rather, in order to fully understand them, we need to look at the phonological structure of the relevant elements
as well.
For this reason, I adopt a version of the parallel architecture that Jackendoff (1997, 2002) develops. The central idea of
Jackendoff's model is that semantic, syntactic and phonological structures are built in parallel. This contrasts17 with the
standard assumption in minimalist theories, which is that syntactic structure is primary and that the phonological and
semantic structures are derived from it. Jackendoff proposes a grammar architecture in which each of the modules for
semantics, syntax and phonology generate structures independently, which are then co-indexed in order to establish the
necessary connections between them.
A lexical entry in this model is represented as illustrated in (26) for the word car:
 
(26) N; sg
lxðcarðxÞÞ $ $ /kaN/
count

In this entry, the associations are shown as double arrows rather than with indices. The representation is meant
to express the assumption that the semantic, syntactic and phonological features are distinct but linked to each
other.
An important part of the analysis of any phenomenon is the establishment of the relevant properties of (the class of)
lexical items involved. It will therefore be one of the goals of this paper to establish lexical entries for the elements involved
in simultaneity constructions along the lines of (26). However, since semantics is not relevant for our present purpose, the
semantic component of a lexical entry is often left underspecified or may be absent altogether.
Unlike common assumptions in minimalist theories (made explicit in e.g., Distributed Morphology, cf. Halle and
Marantz, 1993), syntactic and phonological features are not introduced independently into the derivation. A lexical
entry is a bundle of semantic, syntactic and phonological features (cf. Chomsky, 1965:214 fn. 15), which means that
when a syntactic head is introduced into the derivation, its semantic and phonological features become available as
well.
That is, when the generative component inserts a lexical item into the derivation, this includes not just the syntactic
features, but also the semantic and phonological features. All three systems will then attempt to integrate the new feature
bundle into the existing structure. In this sense, the grammar architecture is parallel.18

17
At least in Jackendoff's opinion, but see Freidin (2003) for a different view.
18
This in fact seems to differ somewhat from Jackendoff's architecture, in which semantic, syntactic and phonological structure are generated
separately and then matched through indices. I assume that the different modules can only generate structures in concert, so that there is no need
to match them: the necessary links are present from the start, they are in fact inherent in the derivation. It is not immediately clear, however, that
this is really a substantial difference; it may be merely cosmetic.
990 J. Kremers / Lingua 122 (2012) 979--1003

4.2. Autosegmental phonology

It is a standard assumption in phonological theory that phonological structure is best described as a collection of
independent but interrelated (autosegmental) tiers. This idea, introduced by Goldsmith (1976) to describe tone
phenomena, has been used very fruitfully by McCarthy and Prince (1990, 1996) to develop their theory of prosodic
morphology. What I propose here is that autosegmental phonology is not only relevant for morphology, but also for
syntax.
Prosodic morphology has not gone unnoticed among sign-language researchers. There are various suggestions that
certain phenomena in sign language morphology should be treated along these lines (cf. especially Sandler, 1989, 1990; see
also the discussion and references in Sandler and Lillo-Martin, 2006:139--142). Consider the example in (7), repeated here:
(7) body: exclusive

face: reluctant

hands: APPROACH-EACH-OTHER.CluprightG.slow
‘two people approach each other slowly, reluctantly, and with hostility
(Leuninger et al., 2005:331)

In this example, the adverbial slowly is expressed by realizing the path movement of the verb APPROACH more slowly
than normal. On the assumption that slowly is morphological, these facts obviously suggest an analysis in terms of
prosodic morphology, for example by treating the path movement as an autosegmental tier along the lines of (27)19:
(27) slow

[TD$INLE] σ

L M L

3 uprightG 2

Here, I adopt a representation of the structure of the syllable in sign languages proposed by Sandler (1986, 1989): a
sign consists of a movement (M) and two locations (L), the starting and end points of the movement.20 In the sign in (27),
the component slow is associated with the movement (M) part of the syllable. For adverbials realized on the face, we may
assume that it is associated with the entire syllable (although nothing in the current proposal really hinges on this):
(28) anxious

[TD$INLE] σ

L M L

3 flatC 2

The analysis is in fact similar to the rule of arc linking proposed by Sandler (1990:25):

(29) [+arc]
TD$INLE][ |
L M

The rule in (29) associates a feature [þarc] to the M part of the syllable, causing the sign to be realized with an arc
movement. We may similarly assume that e.g., a duration feature [þdur] is associated with the M in (27). These features

19
Note that the phonological representations of sign language data here and below are simplified: although technically not really correct, the
simplified representations are probably easier to understand than a system such as HamNoSys (basically an IPA for sign languages, cf. Hanke,
2004).
20
An earlier model of sign language syllables posits holds (H) instead of locations, cf. Liddell and Johnson (1989). See Sandler and Lillo-Martin
(2006:128--138) for a discussion of both models.
 J. Kremers / Lingua 122 (2012) 979--1003 991

are comparable to distinctive features in spoken language phonology, in that they determine the phonological shape of the
segment (sign). In the case of anxious in (28), it is not a phonological feature that is added. Rather, the sign is associated
with an autosegmental element.
From a theoretical point of view, however, there is no fundamental difference between the two: both the [þarc] or
the [þdur] feature on the on hand and the non-manual adverbial on the other are represented as autosegmental features.
The difference is solely their realization: on the hand(s) in the one case and on the lower part of the face in the other. The
analysis is the same in both cases: an autosegmental feature that is associated through standard phonological principles
(e.g., Left-to-Right Association) to elements on the segmental tier.

4.3. Arabic verbal nouns

In order to see how the forms in (27) and (28) are created, it may be best to look at a more familiar form, prosodic
morphology in Arabic. Consider the following Arabic verbal noun:

(30) ’aqlaqa-nı d-u

-ntiqa -l-rajul-i ‘-a
-l-mašru
annoy.3sg.m-1sg.OBJ criticizing-NOM DEF-man-GEN DEF-project-ACC
‘The man's criticizing the project annoyed me.’

The relevant word form is intiqa d ‘criticizing’, which is a verbal noun (masdar in Arabic) derived from the verb intaqada
‘to criticize’. The fact that this noun can assign accusative case suggests an _ analysis along the lines of Abney (1987), in
which the verbal noun starts out as a verbal root and is converted to a noun by a nominalizing affix. McCarthy and Prince
(1990) propose that verbal nouns in Arabic actually contain four different morphemes:

(31) Root: «nqd» Nominalizer: «i.a»

Stem VIII: (s)s m Non-finite: -s mm
|
t

Note that in Arabic, roots are category-neutral; that is, the root «nqd» occurs in both nouns and verbs. The element that
turns the root into a verb in this particular case is the stem VIII marker.21 The nominalizer is the morpheme that turns the verbal
projection into a noun; it performs the function that the suffix -ing performs in Abney's analysis of English gerunds. McCarthy
and Prince also assume that vowel length acts as a non-finiteness marker, but in Kremers (in press-a) I argue that this should
not be analyzed as a separate morpheme (among other reasons, because it does not appear in participles, which are the only
other category of non-finite verb forms in Arabic). Instead, the long vowel is part of the nominalizer:
(32) -σ μμ
[TD$INLE] |
/i.a/

I assume the following syntactic structure for the example in (30):

(33) D
N
D v
NOML
Subj
[TD$INLE] VIII √
Obj
/i.ā/
(σ μ )σ μ
/nqd/
|
/t/

The stem VIII marker verbalizes the root, which is why I assume it is a so-called little v head. The important point of this
tree is that the three morphemes---the nominalizer, the verbalizer and the root---do not form a distinct subtree. In standard

21
Each root in Arabic yields up to 15 different verb stems, which are numbered I--XV. Note, however, that no root exists for which all 15 stems are
instantiated.
992 J. Kremers / Lingua 122 (2012) 979--1003

views, this would be problematic, because they form a distinct (prosodic) word. What I argue for in Kremers (in press-a) is
that it is not actually necessary to combine the three morphemes in syntax. We only need to make sure their phonological
forms can be combined successfully in phonology.
What is required for this is that the phonology ‘knows’ that the three morphemes are to be combined into a single word
form. In order to achieve this, it is not necessary to establish a subtree that contains the three morphemes and nothing
else. We can instead use Ackema and Neeleman's (2004) principle of Input Correspondence22:

(34) Input Correspondence:

If A selects (a projection of) B, Φ(A) selects Φ(B).23

Input Correspondence crucially depends on Beard's (1988, 1995) Separation Hypothesis. If the morphosyntactic
structure does not contain phonological features, it is perfectly conceivable that in syntax, an affix attaches to a phrase
rather than to a head. The requirement that an affix attach to a head should, Ackema and Neeleman claim, be
construed as a phonological requirement: it is the phonological form of the affix that must attach to a prosodic word;
more specifically, it must attach to the prosodic word that corresponds to the head of the projection it attaches to in
syntax.
An example of a head attaching to a phrase is Abney's (1987) analysis of the English gerund. According to
Abney, the affix -ing may attach to V, creating a purely nominal construction in which the gerund's subject is
expressed as a possessor and the object with the preposition of. Alternatively, it may attach to VP, creating a
structure in which the object is assigned accusative case. Finally, -ing may also attach to IP, in which the subject is
assigned accusative:

(35) a. John's singing of the Marseillaise

b. John's singing the Marseillaise
c. John singing the Marseillaise

The crucial insight that Input Correspondence expresses is the fact that even if an affix attaches to a phrase in syntax,
its phonological form still attaches to the phonological form of the head of that projection. Occasionally, Ackema and
Neeleman argue, the phonological requirement imposed by Input Correspondence cannot be met, so that the derivation
crashes.
Applying Input Correspondence to the tree in (33), it is easy to see how we can derive the word form intiqa d without
resorting to syntactic head movement. The nominalizer selects a verbal category (v in the tree here). Thus, according to
Input Correspondence, Φ(NOML) combines with Φ(v), which is the verbal stem marker VIII. Since the stem marker itself
selects the root, all three morphemes combine into a single form. That is, Input Correspondence ensures that the three
morphemes are combined in phonology. It is therefore not necessary to combine them in syntax.
In phonology, the three morphemes are combined through standard principles of autosegmental phonology: the root is
realized on the segmental tier, the stem VIII and the nominalizer both combine elements realized on the segmental tier and
elements realized on the syllabic tier. The two tiers are associated with each other through phonological principles,
resulting in the following structure:

(36) (σ) σ σ (σ)

[TD$INLE] × × × × × × ×

n t i q a d

Note that the order of the stem marker and the nominalizer is determined by the fact that the latter is phonologically a suffix
(which is indicated in (32) by the hyphen). Furthermore, all Arabic stems end in an extrametrical syllable, so one is added to
the syllabic template in (36). Finally, the order in which the segments appear is determined in part by the order of the
segments in the morphemes (the root is «nqd» and any stem derived from this root must have the consonants in that order;
the same goes for the nominalizer). Additionally, basic phonological principles determine the order of the consonants and the
vowels: Left-to-Right Association puts the «n» in the first consonantal position, the coda position of the first (extrametrical)

22
See also Sadock's (1992) Constructional Integrity Constraint, which expresses the same idea.
23
Φ(X) here is to be read as ‘‘the phonological material onto which the syntactic element X is mapped’’.
 J. Kremers / Lingua 122 (2012) 979--1003 993

syllable. In the second syllable, the onset is filled by the «t»; the nucleus position is not available for consonants. As a result,
the second root consonant «q» appears in the onset of the third syllable. The third root consonant fills the onset position of the
fourth, extrametrical, syllable.24 The vowels of the nominalizer fill the nucleus positions from left to right.

4.4. Constructing a phonological form

The point of the Arabic example, complex as it may be phonologically, is to show how syntax and phonology cooperate
to create the word form intiqa d. Syntactic structures are built up out of syntactic heads, which are morphosyntactic feature
bundles without phonological content. These structures are mapped onto (or rather linked with) phonological structures,
which are then combined by the phonological component. As discussed in the previous section, the order and groupings
within the phonological structure are determined partially by phonology-internal constraints and principles, and partially by
lexical facts (e.g., whether an affix is a pre- or a suffix).25
Additionally, principles governing the mapping from syntax to phonology influence the ordering of elements in
phonology. One such principle, relevant for the present discussion, is Input Correspondence, as discussed above. It is this
principle that brings the three morphemes together in the analysis of Arabic verbal nouns sketched here.
And lastly, of course, the structure of the syntactic tree also has an influence on the phonological ordering of elements. The
Exclusivity and Nontangling Conditions form the basis of the mapping from syntax to phonology, but they hold only to the
extent that syntactic heads are mapped onto the same phonological tier.
In spoken languages, any structure above the syntactic head is usually mapped onto the segmental tier. Even heads
themselves almost always are, with the exception of those cases traditionally captured by prosodic morphology.
The result of this is that in spoken languages, there is little reason to assume that the Exclusivity and Nontangling
Conditions are not absolute conditions. What facts there are that might be problematic can be dealt with morphologically
(as in the case of prosodic morphology) or as non-syntactic, intonational effects (e.g., question intonation, so-called hat
patterns, etc.)
In sign languages, things are different. Simultaneity is much more pervasive and there are cases of simultaneity that
can hardly be claimed not to be syntactic. Negation and adverbs, especially sentential adverbs, are the best examples.
Negation and sentential adverbs, apart from the fact that they have sentential scope, are realized simultaneously with not
just a single manual sign, but with a sequence of manual signs. A morphological analysis (i.e., an analysis in terms of word
formation) is therefore not feasible.
Given the necessity of a syntactic analysis, we are faced with the fact that even syntactic structures can map onto
autosegmental tiers in phonology, resulting in configurations that do not conform to the Exclusivity and Nontangling
Conditions. The most straightforward way to describe these facts appears to be to extend the ideas behind prosodic
morphology to syntax.
To see how we may do this, consider again the example from Pfau (2008) given in (15), repeated here:

(15) head: neg DGS

hands: IX:3 FLOWER BUY
‘She does not buy a flower’

Pfau uses this example to show that the spreading of the negative headshake in DGS is prosodic in nature: the
headshake spreads over a prosodic domain. What this means is that the headshake, which is realized phonologically on a
separate autosegmental tier, is associated with a phonological phrase. This can be expressed as follows:

(37) NOT ↔ Neg ◦ ↔ hs

[TD$INLE] |
ϕ

The semantics in (37) is obviously oversimplified but is not at issue. On the phonological side, I have also made a
simplification: I make no attempt to provide an accurate description of the actual phonological form, I just indicate the
headshake with hs. What is important here, however, is the fact that the headshake is a prosodic morpheme associated

24
Note that this is a language-specific association, not determined by Left-to-Right Association: the coda position of the third syllable would in
principle allow a consonant. The need for an extrametrical syllable overrides Left-to-Right Association in this case. I suspect that this association
would be better analyzed in terms of alignment, but that would go beyond the scope of this paper.
25
I assume that information about the status of an affix as a pre- or suffix is in fact stored as a strong alignment preference, cf. footnote 24.
994 J. Kremers / Lingua 122 (2012) 979--1003

with the phonological phrase (w). It is this specification that causes the headshake to spread over (the phonological phrase
corresponding to) the VP in DGS.
As discussed in section 2, the negative headshake in DGS can also be limited to just the verb, as in (38) (cf. (13)):

(38) head: neg

hands: MOTHER FLOWER BUY

The phonological part of the negation's lexical entry therefore must be assumed to be variable. Apart from the form in
(37), there is a variant in which the headshake is associated with the prosodic word:

(39) NOT ↔ Neg◦ ↔ hs

[TD$INLE] |
ω

It is important to note that even though the object FLOWER is also a prosodic word, the negation is associated with the
verb---or, more precisely, with Φ(V). It is the principle of Input Correspondence that guarantees this. The relevant tree
structure for the sentence in (38) is (40):

(40) TP

NegP
N vP
T
Neg VP
[TD$INLE] tN
MOTHER v
N V
hs

FLOWER BUY

This structure is mapped onto a phonological representation, which is partially displayed in (41):

(41) hs (head tier)

IntP

[TD$INLE]
ϕ ϕ

MOTHER ω FLOWERω BUYω (segmental tier)

In principle, it seems that the headshake could be associated with any of the three prosodic words on the segmental
tier. In actual fact, only an association with BUY is grammatical:

(42) hs (head tier)

IntP

[TD$INLE]
ϕ ϕ

MOTHER ω FLOWERω BUYω (segmental tier)

We need no additional assumptions, however, in order to account for this fact. The principle of Input Correspondence
makes sure that the headshake can only be associated with the verb: Neg8 selects a verbal category (little v in my
 J. Kremers / Lingua 122 (2012) 979--1003 995

analysis), so that Input Correspondence requires that Φ(Neg) be associated with Φ(v), the head of the projection that Neg
selects, which in turn selects V.
Note that this analysis will not be significantly different when instead of the phonological form in (39), the form in (37) is
chosen. Input Correspondence still requires that the headshake be associated with the verb. What happens is that the
headshake is associated with the phonological phrase containing the verb, rather than with the prosodic word:

(43) hs (head tier)

IntP

[TD$INLE]
ϕ ϕ

MOTHER ω FLOWERω BUYω (segmental tier)

The other simultaneity examples discussed in section 2 can be treated in much the same way. VP adverbials, for
example, have a lexical entry of the following form:

(44) ⎡ ⎤
⎢Adv⎥
[TD$INLE]
WITH EFFORT ↔ ⎣ ⎦↔ w.e.
uV |
ω

Again, the phonological form here is a simplification: w.e. stands for the phonological features that make up the facial
expression that conveys the meaning with effort. As with the negation in (39), these phonological features are associated
with a prosodic word. Input Correspondence ensures that the adverb is associated with the prosodic word corresponding
to the verb, rather than with any other prosodic verb in the c-command domain of the adverb. Note that I assume that the
adverb actually selects an element of category V, as is often assumed in HPSG-style analyses.26
One important aspect of this analysis is the assumption that adverbial facial expressions and negation are aligned with
prosodic categories and not with (morpho)syntactic categories. It has been argued that non-manuals in sign language are
aligned with syntactic categories (e.g., Neidle et al., 2000), but it is by now quite clear that certain types of non-manuals,
especially interrogative and topic marking, are aligned with prosodic domains (see Sandler, 2011, who summarizes the
relevant discussion).
For non-manual negation and adverbials, matters are not as clear. One datum suggesting that at least the non-manual
negation in DGS aligns with prosodic categories and not with syntactic ones is the example in (15). For VP-adverbs,
however, the data are compatible with both types of analyses. They align with the verb; whether they align with the
syntactic head or with the prosodic word is not so easy to determine.
One fact that tentatively supports the assumption made here is that the current analysis makes one particular prediction.
Because phonological structure is organized along the prosodic hierarchy, and because the prosodic hierarchy obeys the
Strict Layer Hypothesis (Selkirk, 1984; Nespor and Vogel, 1986), we expect that two autosegmental morphemes are either
coextensive or that one completely contains the other. That is, we do not expect to find structures of the following type:

(45) tier 1: x
tier 2: y
segm: pword pword pword pword

In this hypothetical example, an autosegmental marker on tier 1 overlaps with a marker on tier 2, but only partially.
Given the assumption made in this paper that autosegmental markings coincide with prosodic constituents, the situation
in (45) could only obtain if the relevant prosodic constituents violated the Strict Layer Hypothesis.
Presumably, the same prediction would hold if all autosegmental elements would align with syntactic categories.
Crucially, the prediction would not hold if some autosegmental elements aligned with prosodic categories, while others

26
In (44), this selection has been formalized as an uninterpretable V-feature on the adverb, but obviously other methods are possible.
996 J. Kremers / Lingua 122 (2012) 979--1003

aligned with syntactic categories. Because of the fact that syntactic and phonological constituents do not map onto each
other one-to-one, cases such as in (45) might be possible.
In the simultaneity examples that I looked at, configurations of this type were conspicuously absent. Because at least
some autosegmental elements are aligned with prosodic categories, it would seem that all must do so. More research is
needed, however, to see if the prediction really holds.
Note, however, that under the assumptions made in this paper, alignment with a syntactic category is strictly speaking
not even an option. A syntactic head does not carry phonological information, it consists of morphosyntactic features only.
Therefore, it has no temporal extent, no left and right edges with which an autosegmental phonological element could be
aligned.
The relevant question is therefore whether non-manuals align with a prosodic category that is lexically associated with
a specific head or whether they can also spread over a prosodic domain that is not introduced by a lexical head. Take, for
example, the VP adverb in (2), repeated here:

(46) face: with effort DGS

hands: STUDENT SIGN-LANGUAGE LEARN
‘The student learns sign language with difficulty.’

The lexical entry for the verb LEARN is (47):

(46) lelxlyðlearnðx; y; eÞÞ $ [V, uN] $ LEARN

The phonological component, which I have indicated simply as LEARN, is a prosodic word. The non-manual adverb with
effort is aligned with this p-word by virtue of its lexical specification. The relevant question is now what happens with the
adverb when the verb's p-word is modified in some way. The problem is that there are few modifications that one can think of
that may provide a clue. In larger prosodic domains, e.g., IntPs or Utterances, it is not uncommon for additional boundaries to
be introduced (e.g., when a parenthetical is added to the structure). At the level of the p-word, there is no equivalent process
(at least, not that I am aware of). At best, one may glean something from processes that extend the p-word, for example, by
incorporating a clitic. But in such cases, it would first need to be established on independent grounds that verb þ clitic really to
form a (single) p-word and not a larger prosodic category such as a clitic group or a recursive p-word.
Summarizing, the grammar architecture basically forces the conclusion upon us that non-manuals align with prosodic
categories: the component of the non-manual that must be aligned is the phonological one, not the syntactic or semantic
one, and the only type of element it can align with is also phonological. The question that needs to be answered in future
research is whether non-manual adverbs align with the prosodic constituent that the verb introduces or whether they can
also align with prosodic constituents that are not introduced lexically.
The lexical entries that I assume for the [þQ] C8 head and the Top8 head are the following:

(48) ⎡ ⎤
C
⎢ ⎥
a. INT ↔ ⎣ ⎦ ↔ rb
+Q
|
IntP
[TD$INLE]
⎡ ⎤
⎢Top⎥
b. TOP ↔ ⎣ ⎦ ↔ rb
uN
|
IntP

Here, as before, I have simplified the representations. The rb in the interrogative and topic markers stands for raised
brows.27 Crucial for the current analysis is that these phonological forms are autosegmental: they are realized on an

27
A brow raise is an essential part of both the interrogative and the topic marker, but the representation in (48) is not intended to be exhaustive.
Note, by the way, that in DGS, constituent questions are usually associated with furrowed brows. Brow raise as question marking occurs mainly
with yes/no questions.
 J. Kremers / Lingua 122 (2012) 979--1003 997

autosegmental tier and they are associated with a specific prosodic constituent. The autosegmental tier that is relevant for
interrogative and topic marking may be called an upper face/head tier, as it involves the upper part of the face and the
position of the head. It is different from the tier on which facial adverbials are realized, because, as Meir and Sandler
(2008:175) remark, these are usually associated with the mouth.28
The lexical entries in (48) are built up in the by now familiar way: a piece of meaning is associated with a syntactic head,
which is associated with a chuck of phonology. That is, there is no fundamental distinction between these entries and the
entries for negation and adverbials above. They all follow the pattern that Jackendoff (2002) proposes for lexical entries in
general. This analysis thus assigns syntactic status to intonational patterns, which is markedly different from the usual
treatment of such phenomena.29
The fact that questions may have atypical non-manuals or intonation can simply be accounted for by assuming that
there are in fact different C[þwh] heads in the lexicon, in the same way that there are different N heads in the lexicon.
Syntactically, the nouns dog and cat are not differentiated, they both consist of a head N with features such as [þsg,
þcount, þanimate]. The differences lie in the semantic and phonological features associated lexically with both heads.
The same can be assumed for C[þwh]: there are several such heads, which are not differentiated syntactically, just
semantically and phonologically. That is, the existence of atypical interrogative non-manuals and intonation does not
argue against a syntactic analysis.
Similarly, I do not regard the fact that interrogative and topic non-manuals in sign language are realized on the upper
part of the face, contrary to other facial expressions, as an indication that these elements are fundamentally different. It is
obviously an interesting question why it should be the case that interrogative and topic non-manuals are cross-
linguistically realized in this manner, but claiming that they are non-syntactic does not provide an explanation of this fact. It
would still need to be explained why the former follows from the latter.
As a last point, the lexical entry that I propose for sentential adverbs in DGS is the following:

(49) ⎡ ⎤
⎢Adv⎥
POSSIBLE↔ ⎣ ⎦ ↔ possible
[TD$INLE] uC
|
U

Here again the actual phonological form is not represented properly, instead I have just used the indication possible. It
is associated with the phonological Utterance, although Happ and Vorköper's (2006) data is not really sufficient to
determine whether sentential adverbials are associated with the Utterance or with the Intonational Phrase.

5. Input and Linear Correspondence interaction

Ackema and Neeleman (2004) propose not only Input Correspondence, discussed above in section 4.3, as a mapping
principle. They also propose a second principle, called Linear Correspondence.30 Because the analyses proposed here
for the simultaneity data appear to violate Linear Correspondence, it is necessary to discuss this principle and how it
interacts with Input Correspondence in some more detail. What I will argue is that Linear Correspondence is actually the
primary principle of the two, but it may be violated under certain conditions. It is exactly in such cases that Input
Correspondence plays a role: an element that violates Linear Correspondence must obey Input Correspondence.
Linear Correspondence is formulated as in (50):

(50) Linear Correspondence:

If a node A is structurally external to B, then Φ(A) is linearly external to Φ(B).

Linear Correspondence states that if a node A is not contained in (dominated by) a node B, the phonological strings that
A is mapped onto is outside the string onto which B is mapped. In essence, Linear Correspondence has the same effect as
the Nontangling Condition: it establishes that association lines between non-terminals and terminals do not cross. The

28
See also Wilbur (2000) for an overview of the different types of non-manuals found in ASL. Wilbur also discusses the different grammatical
functions that different non-manual articulators have.
29
But see Neidle et al. (2000) and Cheng and Rooryck (2000) for similar proposals; cf. also footnote 10.
30
This principle also has a precursor in Sadock (1992). Note, by the way, that Ackema and Neeleman actually propose a third mapping principle,
Quantity Correspondence. For the present discussion, this principle is not relevant.
998 J. Kremers / Lingua 122 (2012) 979--1003

main difference is that Linear Correspondence presupposes the Separation Hypothesis, so that the association lines are
in fact mapping links between morphosyntactic and phonological structure.
For Ackema and Neeleman, both Input and Linear Correspondence are principles governing the mapping of word-
syntactic structures onto word-phonological structures. That is, they are morphological principles. Obviously, in the
analysis proposed here, I assume that they also govern the mapping from syntax to phonology.31 There is no real reason
to restrict these principles to morphology: like morphology, syntax generates tree structures that must be mapped onto
linear phonological structures. Besides, as just stated, Linear Correspondence is closely related to the Nontangling
Condition, which is obviously a condition on syntactic structures.
When we examine both mapping principles a bit closer, however, it becomes clear that they are not absolute. For
instance, the example of the Arabic verbal noun violates Linear Correspondence: in the tree in (33) in section 4.3, the
nominalizer is structurally higher than the subject, which should prevent it from combining with the verbalizer. The same is
true of the sign language negation: if the negative headshake is realized on the verb alone, as in (38/40), it appears linearly
inside the VP, although it is structurally external to it.
Ackema and Neeleman (2004) discuss such configurations and they claim that when Linear Correspondence cannot
be met, an element that by Input Correspondence should combine with the head of its sister cannot do so. Consider the
three in (51):

(51) AP

[TD$INLE]A BP

CP B

Suppose that A in (51) selects category B. Input Correspondence then requires Φ(A) to take Φ(B) as its host. If B's
complement CP is linearized between A and B, because both A and CP precede or follow B, this is not possible: A is
linearly separated from B by CP, so that A and B cannot combine into a single form. A possible solution to such a conflict is
to linearize A and CP on opposite sides of B, if the language allows this. In this way, A and B end up linearly adjacent,
which is a configuration that allows affixation of A to B.32 Another option is to realize A as a null affix, which according to
Ackema and Neeleman is not subject to Linear Correspondence.
The tree given in (33) for the Arabic verbal noun displays the configuration in (51), since the nominalizer is structurally
external to the projection of the verbalizer VIII (little v), so that it should also appear linearly external to it. Since the subject is
contained in the projection of VIII, Linear Correspondence requires that the nominalizer be linearized linearly external to the
projection of the non-finiteness marker, which is not the case.
Note that the way out that Ackema and Neeleman offer does not really help us here. It would hardly be satisfying to
assume that the nominalizer is a suffix, given that it is obviously a prosodic morpheme that cannot be classified as a pre- or
suffix. Besides, it would require that the nominalizer be head-final, which is quite unlikely, given that Arabic is a head-initial
language throughout.
However, there is another reason why such a proposal would be unsatisfactory, which has to do with the fact that the
Arabic verbal noun is a head-initial structure (cf. the example in (30)). The proposed tree structure does not reflect this fact,
however. In fact, it is not at all clear from this tree where the verbal noun should materialize. There are three heads
involved in its formation, in three different structural positions. Which of those is the position in which the form is eventually
realized? The example itself leaves no doubt as to which position it is: because the verbal noun is in initial position and
because Arabic is generally head-initial, it must be the position of the nominalizing head. In other words, it is not so much
the nominalizer that is violating Linear Correspondence, but rather the other two heads: the root and the verbalizer.
Note that all the heads involved have one property in common: they map onto phonological structures that are
incomplete; i.e., forms that cannot be realized on their own. Without wishing to go into details,33 I propose that a
phonologically incomplete form cannot be realized until it is combined with sufficient phonological material to allow a licit

31
In actual fact, I assume there is no distinction between phrasal syntax and ‘‘word syntax’’, i.e., morphology, contrary to what Ackema and
Neeleman (2004) assume. Arguing this point would go well beyond this paper, but see Kremers (in preparation) for discussion.
32
Ackema and Neeleman use this to account for the fact that in Quechua, both OV and VO main clause orders exist, while subclauses are
strictly OV: the complementizer, which is suffixed on the verb, can only take the verb as its host in the OV order. (But see Myler, 2009 for an
analysis of Quechua subclauses, contradicting some of Ackema and Neeleman's claims.)
33
See Kremers (in press-a) for discussion.
 J. Kremers / Lingua 122 (2012) 979--1003 999

phonological form to be realized. Specifically, there must be enough segmental material to produce a prosodic word,34
and all prosodic requirements must be satisfied. An element for which these conditions are not met may violate Linear
Correspondence in order to meet them.
The root of the verbal noun in Arabic cannot be realized, because it only contains three consonants. It selects the
object, but since this complement is phonologically complete, the root cannot combine with it. The only way in which the
root can be saved is to combine it with a higher head. The verb stem marker selects the root, but combining with it does not
yield a licit structure either. Another selecting element is needed to rescue the combination of VIII and V. Only when the
nominalizer becomes available can a licit form be produced. I assume that it is for this reason that the word form intiqa d
appears in the position of the nominalizer: it is the nominalizer that introduces the material that allows the phonology to
finally produce a licit form.
A slightly different situation exists in the case of DGS negation. To repeat the relevant tree:

(40) TP

NegP
N vP
T
Neg VP
[TD$INLE] tN
MOTHER v
N V
hs

FLOWER BUY

Here, the verb BUY is a licit phonological form: it is a prosodic word with no additional prosodic requirements. Therefore it
is realized phonologically in the position of Φ(V). Things are different for the negation, however. The negation is an
autosegmental morpheme, which cannot form a prosodic word on its own and which additionally has a prosodic
requirement, namely that it be associated with a prosodic word. Because of these requirements, Φ(Neg8) can violate
Linear Correspondence and be realized together with Φ(V).
What this means is that Linear Correspondence is a violable principle. The syntax--phonology mapping adheres to it
only to the extent that the phonology allows it. If a phonological form would arise that violates basic phonological
wellformedness principles, Linear Correspondence gives way in order to allow the phonology to construct a licit form.
Still, when it does so, it adheres to the principle of Input Correspondence. In principle, the requirements of the DGS
negation could also be met by associating it with the first prosodic word contained in the object, i.e., with FLOWER in our
example. Because of Input Correspondence, this does not happen. This analysis suggests that Input and Linear
Correspondence do not have the same status as mapping principles. Linear Correspondence is the primary principle,
while Input Correspondence is secondary, only applying when the phonological form of the element(s) involved violates
Linear Correspondence.
For Ackema and Neeleman (2004), Input Correspondence is relevant for quite a number of cases involving affixes.
This does not contradict the conclusion reached here, however, because an affix is an element that carries a prosodic
requirement: it has an alignment requirement with the left or right edge of a prosodic word (cf. McCarthy and Prince, 1993;
Hayes and Lahiri, 1991). This requirement may result in a situation in which Linear Correspondence is violated, so that
Input Correspondence applies.35
Summarizing, Linear Correspondence is the primary principle: it applies to any two elements that are realized on the
same autosegmental tier. When an element cannot obey Linear Correspondence, however, because it is phonologically
not a licit form, it must still obey Input Correspondence.36 The deeper reason for these two principles seems clear.

34
The requirement that a p-word can be produced is probably language-dependent. In languages such as Chinese a syllable may be sufficient.
35
Ackema and Neeleman do not discuss clitics, which take phrases as host, not words. We may tentatively assume that clitics are a lot like
affixes, in that they are elements with a prosodic alignment requirement. Contrary to affixes, however, their alignment requirement is not with a
prosodic word but with a phonological phrase. Because they are as small as or smaller than syllables, they will still incorporate into the adjacent
prosodic word, but the alignment requirement is with the larger category.
36
As pointed out by a reviewer, there are a few cases in which Linear and Input Correspondence seem to interact in a more complex way:
although *pass by-er is ungrammatical, come-outer is not. Sometimes the suffix -er even appears twice: fixer-upper. I assume these are lexically
stored violations of the mapping principles, which do not undermine the general argument. See also the discussion in Ackema and Neeleman
(2004:160--161).
1000 J. Kremers / Lingua 122 (2012) 979--1003

Basically, both principles state that elements that belong together syntactically must be realized together phonologically.
This is obviously a necessary requirement for any mapping that must be reversible: the parser must be able to reconstruct
the syntactic tree on the basis of the phonological form and it is aided in this task by the fact that it can assume that things
that appear together in the phonological input belong together in the syntax as well.

6. Outlook

The main claim I have presented and tried to support in this paper is that simultaneity can be analyzed as a form of
prosodic morphology at the syntax level. Specific syntactic heads map onto autosegmental phonological elements, which
are integrated into the phonological structure by aligning with prosodic categories. However, I believe that the theoretical
import of simultaneity does not end there. The phenomenon also raises important questions about linearization and the
relation between syntax and phonology in general.
Often, the syntax/phonology relation is seen as in some sense temporally sequential: a syntactic (partial) structure is
built, which is then linearized and subsequently interpreted by the phonological component in order to construct a
phonological representation.37
However, if the analysis presented here is on the right track and there are indeed syntactic heads that map onto
autosegmental tiers in phonology, we may need to rethink our conception of the relation between syntax and phonology.
First of all, there is little point in having linearization take place before all other phonological operations: linearization is not
total, some heads are not linearized with respect to the rest of the clause. Only when the phonological form itself is taken
into account, can it be determined which heads need to be linearized and which do not. It follows that linearization takes
place in phonology, as part of a more general process that composes the entire phonological form, not just the linear order
of the segmental elements.
As one reviewer points out, we may argue that syntax specifies a partial linear order. This raises the question how
syntax determines which heads to linearize and which not to. That would seem to require a feature of some sort, but such a
feature would obviously be redundant: it specifies information that is implicit in the phonological representation.
Furthermore, it does not suffice to assume a feature [linearize], because two autosegmental elements that are realized
on the same tier do need to be linearized with respect to each other. The feature would need to be a multi-valued feature
specifying the exact phonological tier on which an element is realized. Such a move, however, just makes the redundancy
of the feature more obvious.
To avoid such redundancy, we need to assume that linearization takes place in phonology. More precisely, linear order
emerges when the phonology tries to combine various chunks of phonological structure into a larger structure. In a way,
linear order is not even the goal that the phonological system is striving for. Rather, it is an emergent property, the
inevitable outcome of a phonological system trying to combine the phonological parts of the elements in the structure (see
also Kremers, in press-b for some discussion).
If linearization is indeed a phonological process, we need new ways to deal with such basic things as the head
parameter, basic clause ordering, etc. There appears to be a correlation between the head parameter and the alignment of
stress within the phonological phrase (Truckenbrodt, 1995; Nespor et al., 2008), and there are indications that children
acquire the setting of the head parameter through stress patterns, a process known as prosodic bootstrapping
(Christophe et al., 2008). It may therefore be possible to reduce the head parameter to an alignment requirement of
p-phrase stress. Obviously, though, much research will be needed in order to determine to what extent this assumption
holds and how it can be implemented theoretically.
All in all, simultaneity suggests that there is a more direct relation between syntax and phonology than usually assumed
in generative approaches. One cannot consider the syntactic structure of an utterance without taking its phonology into
account. In the analysis developed here, such syntax--phonology interaction is accounted for by assuming a parallel
grammar architecture: the introduction of a syntactic head into the derivation also makes its phonological features
available, so that syntax--phonology interaction can be modeled.
This conclusion actually finds corroboration in data that suggests an even stronger link between syntax and phonology:
there are syntactic structures that are influenced by the phonological structure of the elements involved. A well-known
example is Heavy-NP Shift, which, according to Zec and Inkelas (1990) requires that the shifted NP consists of at least two
phonological phrases. More recently, Erteschik-Shir and Rochman (2010) collect a number of papers that argue for such a
phonology-to-syntax connection.
Furthermore, Richards (2010) argues extensively that certain syntactic phenomena are triggered by phonological
requirements of the elements involved. One of the phenomena that Richards discusses in some detail is what appears to

37
Depending on the framework, linearization is trivial (Kayne, 1994) or it is a distinct process involving linearization parameters, as in HPSG but
also in some theories within G&B and minimalism (e.g., Neeleman and Weerman, 1999; Abels and Neeleman, 2009; Kremers, 2009).
 J. Kremers / Lingua 122 (2012) 979--1003 1001

be a cross-linguistic requirement for wh-words to appear in a single prosodic domain with a (possibly covert) interrogative
complementizer. If the prosodic structure of the language allows this domain to be created with the wh-word in situ, no
wh-movement takes place. If the prosodic structure does not allow creation of this domain, syntactic operations can be
triggered that improve the situation.38
Phonology--syntax interactions of this kind make a parallel architecture virtually unavoidable. It is not clear how else we
could describe the fact that syntax and phonology influence each other: because the interactions between the two
modules are bidirectional, it cannot be the case that the two modules operate sequentially.
In a grammar architecture of this type, it is especially important to specify the principles that underlie the syntax--
phonology interface, whereby we must distinguish the principles that govern the mapping from syntax to phonology from
the principles that govern the feedback from phonology to syntax. For the moment, there seems relatively little that we can
say about the latter; the relevant phenomena need further study in order for us to come to meaningful generalizations
about them. We can say more about the former. The principles of Linear and Input Correspondence are two fundamental
principles governing the mapping from syntax to phonology. We also know quite a bit about the relation between syntactic
and prosodic categories (cf. Selkirk, 1984; Nespor and Vogel, 1986; Truckenbrodt, 1995 and much related work).
Note, by the way, that the bidirectional interactions are not entirely equivalent. This is evident in the formulations that I
use: I speak of mapping from syntax to phonology, but of feedback from phonology to syntax. We expect all parts of
syntactic structure to be mapped onto phonology, at least in principle, otherwise the parser could not reconstruct them.
This expectation does not hold in the other direction: we do not expect all parts of phonological structure to have some
effect on syntax. It generally does not matter for syntax whether a head N is realized phonologically as /kæt/ or /dɔg/. Only
in some cases does phonology have an effect on syntax, namely when an initial syntax--phonology mapping fails due to
some phonological constraint being violated.39

7. Summary and Conclusions

Sign languages have the ability to express two or even more meaningful units simultaneously. These units are part of
the syntactic structure and are realized phonologically on different tiers. Analyzing such phenomena requires that we
extend the principles behind prosodic morphology to syntax. Syntactic heads may be associated with an autosegmental
phonological form. This form is not restricted to prosodic words or syllables: it may also be associated with higher levels in
the prosodic hierarchy, such as the phonological or the intonational phrase.
The mapping from syntax to phonology is governed by two important principles: Input Correspondence and Linear
Correspondence. Crucially, however, Linear Correspondence only holds for elements that are phonologically ‘‘complete’’.
Forms that are not complete, i.e., that do not constitute a licit phonological form on their own, are subject to Input
Correspondence.
The implications of the simultaneity data (and of Richards's wh-data) for theories of the syntax--phonology interface are
possibly far-reaching. Unlike common assumptions, the relation between syntax and phonology is not unidirectional.
Rather, it is a bidirectional relation, with syntactic structure influencing, but also being influenced by, phonological
structure. Future research will need to shed more light on this topic.

References

Aarons, D., 1994. Aspects of the syntax of American Sign Language. Ph.D. Thesis, Boston University.
Abels, K., Neeleman, A., 2009. Universal 20 without the LCA. In: Brucart, J.M., Gavarro, A., Sola, J. (Eds.), Merging Features. Oxford University
Press, Oxford, (Chapter 4), pp. 60--79.
Abney, S., 1987. The English noun phrase in its sentential aspect. Ph.D. Thesis, MIT, mIT Working Papers in Linguistics.
Ackema, P., 1995. Syntax below zero. Ph.D. Thesis, Utrecht Research Institute for Language and Speech OTS.
Ackema, P., Neeleman, A., 2004. Beyond Morphology: Interface Conditions on Word Formation. Oxford University Press, Oxford.
Akinlabi, A., 2011. Featural affixation. In: van Oostendorp, M., Ewen, C.J., Hume, E.V., Rice, K. (Eds.), The Blackwell Companion to Phonology.
Wiley-Blackwell, pp. 1945--1971.
Anderson, D.E., Reilly, J.S., 1998. Pah! the acquisition of adverbials in ASL Sign Language and Linguistics 1--2, 117--142.
Beard, R., 1988. On the separation of derivation from morphology: toward a lexeme/morpheme-based morphology. Quaderni di Semantica 9,
3--59.
Beard, R., 1995. Lexeme-Morpheme Base Morphology. SUNY Press, Albany.
Bellugi, U., Fischer, S., 1972. A comparison of sign language and spoken language. Cognition 1, 173--200.

38
See also Cecchetto et al. (2009), who show that the analysis proposed by Richards can also be used to account for wh-word placement in
Italian Sign Language (LIS).
39
It is perhaps for this reason that the feedback from phonology to syntax is not understood as well: it is more arbitrary and probably more
idiosyncratic than the mapping from syntax to phonology.
1002 J. Kremers / Lingua 122 (2012) 979--1003

Cecchetto, C., Geraci, C., Zucchi, S., 2009. Another way to mark syntactic dependencies: the case for right-peripheral specifiers in sign
languages. Language 85 (2), 278--320.
Cheng, L., Rooryck, J., 2000. Licensing wh-in-situ. Syntax 3 (1), 1--19.
Chomsky, N., 1965. Aspects of the Theory of Syntax. The MIT Press, Cambridge, MA.
Christophe, A., Millotte, S., Bernal, S., Lidz, J., 2008. Bootstrapping lexical and syntactic acquisition. Language and Speech 51 (1), 61--75.
Cinque, G., 1999. Adverbs and Functional Heads: A Cross-Linguistic Perspective. Oxford University Press, New York.
Corina, D., Bellugi, U., Reilly, J., 1999. Neuropsychological studies of linguistic and affective facial expressions in deaf signers. Language and
Speech 42 (2--3), 307--331.
Dachkovsky, S., Sandler, W., 2009. Visual intonation in the prosody of a sign language. Language and Speech 52 (2--3), 287--314.
Emmorey, K., 1995. Processing the dynamic visual-spatial morphology of signed languages. In: Feldman, L.B. (Ed.), Morphological Aspects of
Language Processing. Lawrence Erlbaum, Hillsdale, NJ, pp. 29--53.
Engberg-Pedersen, E., 1994. Some simultaneous constructions in Danish Sign Language. In: Brennan, M., Turner, G.H. (Eds.), Word-order
Issues in Sign Language. University of Durham, The Deaf Studies Research Unit, Papers Presented at a Workshop in Durham, England,
18--22 September 1991, pp. 71--88.
Erteschik-Shir, N., Rochman, L. (Eds.), 2010. The Sound Patterns of Syntax. Oxford University Press, Oxford.
Freidin, R., 2003. Shaky foundations -- review of Ray Jackendoff's Foundations of Language: Brain, Meaning, Grammar, Evolution. Glot
International 7 (7/8), 218--221.
Geraci, C., 2005. Negation in LIS (Italian Sign Language). In: Bateman, L., Ussery, C. (Eds.), Proceedings of NELS 35. GLSA, Amherst, MA,
pp. 217--229.
Goldsmith, J., 1976. Autosegmental phonology. Ph.D. Thesis. The MIT Press, Cambridge, MA.
Halle, M., Marantz, A., 1993. Distributed morphology. In: Hale, K., Keyser, S.J. (Eds.), The View from Building 20: Essays in Linguistics in Honor of
Sylvain Bromberger. Vol. 24 of Current Studies in Linguistics. The MIT Press, Cambridge, MA, (Chapter 3), pp. 111--176.
Hanke, T., 2004. HamNoSys -- representing sign language data in language resources and language processing contexts. In: Streiter, O., Vettori,
C. (Eds.), LREC 2004, Workshop Proceedings: Representation and Processing of Sign Languages, ELRA, Paris, pp. 1--6.
Happ, D., Vorköper, M.-O., 2006. Deutsche Gebärdensprache: ein Lehr- und Arbeitsbuch. Fachhochschulverlag, Frankfurt am Main.
Hayes, B., Lahiri, A., 1991. Bengali intonational phonology. Natural Language and Linguistic Theory 9 (1), 47--96.
Jackendoff, R., 1997. The Architecture of the Language Faculty, Vol. 28 of Linguistic Inquiry Monograph. The MIT Press, Cambridge, MA.
Jackendoff, R., 2002. Foundations of Language: Brain, Meaning, Grammar, Evolution. Oxford University Press, Oxford.
Johnston, T.A., Schembri, A., 2007. Australian Sign Language (Auslan): An Introduction to Sign Language Linguistics. Cambridge University
Press, Cambridge.
Julien, M., 2002. Syntactic Heads and Word Formation. Oxford University Press, Oxford.
Kayne, R., 1994. The Antisymmetry of Syntax. The MIT Press, Cambridge, MA.
Kremers, J., 2009. Recursive linearization. The Linguistic Review 26 (1).
Kremers, J. Morphology -- like syntax -- is in the eye of the beholder, MS. University of Göttingen, in preparation.
Kremers, J. Arabic verbal nouns as phonological head movement, to appear in: Working Papers of the SFB 732 ’Incremental Specification in
Context’, University of Stuttgart, in press-a.
Kremers, J. Linearization as repair, to appear in: Patrick Brandt & Eric Fuß, Repair, in press-b.
Kyle, J.G., Woll, B., 1985. Sign Language: The Study of Deaf People and Their Language. Cambridge University Press, Cambridge.
Leuninger, H., 2005. Sign languages: Representation, Processing, and Interface Conditions. MS. University of Frankfurt.
Leuninger, H., Hohenberger, A., Menges, E., 2005. Zur Verarbeitung morphologischer Informationen in der Deutschen Gebärdensprache (DGS).
In: Leuninger, H., Happ, D. (Eds.), Gebärdensprachen: Struktur, Erwerb, Verwendung. Helmut Buske Verlag, Hamburg, pp. 325--358.
Liddell, S.K., 1980. American Sign Language Syntax. Mouton de Gruyter, Berlin.
Liddell, S.K., Johnson, R., 1989. American Sign Language: the phonological base. Sign Language Studies 64, 197--277.
McCarthy, J., 1981. A prosodic theory of nonconcatenative morphology. Linguistic Inquiry 12 (3), 373--418.
McCarthy, J., 1986. OCP effects: gemination and antigemination. Linguistic Inquiry 17 (2), 207--263.
McCarthy, J., Prince, A., 1990. Prosodic morphology and templatic morphology. In: Eid, M., McCarthy, J. (Eds.), Perspectives on Arabic
Linguistics II. Vol. 72 of Current Issues in Linguistic Theory. John Benjamins, Amsterdam/Philadelphia, pp. 1--54.
McCarthy, J., Prince, A., 1993. Generalized alignment. Yearbook of Morphology 79--154.
McCarthy, J., Prince, A., 1996. Prosodic Morphology 1986. MS. University of Massachusetts, Rutgers University.
Meir, I., Sandler, W., 2008. A language in Space: The Story of Israeli Sign Language. Lawrence Erlbaum Associates, New York.
Myler, N., 2009. Mixed categories and the syntax--morphology interface: The case of Quechua subordinate clauses, MS. NYU.
Neeleman, A., Weerman, F., 1999. Flexible Syntax: A Theory of Case and Arguments. Kluwer Academic Publishers, Dordrecht.
Neidle, C., Bahan, B., MacLaughlin, D., Lee, R.G., Kegl, J., 1998. Realizations of syntactic agreement in American Sign Language: similatities
between the clause and the noun phrase. Studia Linguistica 52 (3), 191--226.
Neidle, C., Kegl, J., MacLaughlin, D., Bahan, B., Lee, R.G., 2000. The Syntax of American Sign Language: Functional Categories and Hierarchical
Structure. The MIT Press, Cambridge, MA.
Nespor, M., Vogel, I., 1986. Prosodic Phonology. Foris, Dordrecht.
Nespor, M., Mohinish, S., van de Vijver, R., Avesani, C., Schraudolf, H., Donati, C., 2008. Different phrasal prominence realizations in VO and OV
languages. Lingue e Linguaggio 7 (2), 139--168.
Partee, B., ter Meulen, A., Wall, R., 1993. Mathematical Methods in Linguistics. Kluwer, Dordrecht.
Pfau, R., 2001. Typologische und strukturelle Aspekte der Negation in Deutscher Gebärdensprache. In: Leuninger, H., Wempe, K. (Eds.),
Gebärdensprachlinguistik 2000: Theorie und Anwendung. Signum, Hamburg, pp. 13--31.
Pfau, R., 2002. Applying morphosyntactic and phonological readjustment rules in natural language negation. In: Meier, R., Cormier, K.,
Quinto-Pozos, D. (Eds.), Modality and Structure in Signed and Spoken Languages. Cambridge University Press, Cambridge, (Chapter 11),
pp. 263--295.
Pfau, R., 2008. The grammar of headshake: a typological perspective on German Sign Language negation. Linguistics in Amsterdam 1, 34--71.
 J. Kremers / Lingua 122 (2012) 979--1003 1003

Pfau, R., Quer, J., 2002. V-to-Neg raising and negative concord in three sign languages. Rivista di Grammatica Generativa 27, 73--86.
Pfau, R., Quer, J., 2007. On the syntax of negation and modals in Catalan Sign Language and German Sign Language. In: Perniss, P.M., Pfau, R.,
Steinbach, M. (Eds.), Visible Variation, Comparative Studies on Sign Language Structure. Trends in Linguistics. Mouton de Gruyter, Berlin,
pp. 129--161.
Richards, N., 2010. Uttering Trees. The MIT Press, Cambridge, MA.
Rizzi, L., 1997. On the fine structure of the left-periphery. In: Haegeman, L. (Ed.), Elements of Grammar. Kluwer Academic Publishers, Dordrecht,
pp. 281--337.
Sadock, J.M., 1992. Autolexical Syntax. University of Chicago Press, Chicago.
Sandler, W., 1986. The spreading hand autosegment of American Sign Language. Sign Language Studies 50, 1--28.
Sandler, W., 1989. Phonological Representation of the Sign: Linearity and Nonlinearity in American Sign Language. Foris, Dordrecht.
Sandler, W., 1990. Temporal aspects and ASL phonology. In: Fischer, S.D., Siple, P. (Eds.), Theoretical Issues in Sign Language Research.
Vol. I. University of Chicago Press, Chicago, pp. 7--36.
Sandler, W., 2011. Prosody and syntax in sign languages. Transactions of the Philological Society 108 (3), 298--328.
Sandler, W., Lillo-Martin, D., 2006. Sign Language and Linguistic Universals. Cambridge University Press, Cambridge.
Selkirk, E., 1984. Phonology and syntax: The relation between sound and structure. Ph.D. Thesis. The MIT Press, Cambridge, MA.
Selkirk, E., 2005. Comments on intonational phrasing in English. In: Frota, S., Vigário, M.C., Freitas, M.J. (Eds.), Prosodies. Mouton de Gruyter,
Berlin, pp. 11--58.
Truckenbrodt, H., 1995. Phonological phrases: Their relation to syntax, focus and prominence. Ph.D. Thesis. The MIT Press, Cambridge, MA.
Tucker, A., Tompo Ole Mpaayei, J., 1955. A Maasai Grammar with Vocabulary. Longmans, Green and Co., London.
Vermeerbergen, M., Leeson, L., Crasborn, O. (Eds.), 2007. Simultaneity in Signed Languages: Form and Function. John Benjamins, Amsterdam/
Philadelphia.
Wilbur, R.B., 1994. Foregrounding structures in American Sign Language. Journal of Pragmatics 22 (6), 647--672.
Wilbur, R., 2000. Phonological and prosodic layering of nonmanuals in American Sign Language. In: Emmorey, K., Lane, H. (Eds.), The Signs or
Language Revisited. Lawrence Erlbaum Publishers, Mahwah, NJ, pp. 215--244.
Wilbur, R., Patschke, C., 1998. Body leans and the marking of contrast in American Sign Language. Journal of Pragmatics 30, 275--303.
Wilbur, R.B., Patschke, C., 1999. Syntactic correlates of brow raise in ASL. Sign Language & Linguistics 2 (1), 3--41.
Zec, D., Inkelas, S., 1990. Prosodically constrained syntax. In: Inkelas, S., Zec, D. (Eds.), The Phonology--Syntax Connection. University of
Chicago Press, Chicago, pp. 365--378.
Zeshan, U. (Ed.), 2006. Interrogative and Negative Constructions in Sign Languages.. Ishara Press, Nijmegen.