The syntax of confirmationals: Form and function of extra-clausal constituents

Full text

 
ECCs: a grammar of their own?

doi 10.1075/slcs.178.wil
© 2016 John Benjamins Publishing Company
e syntax of conrmationals
A neo-performative analysis
Martina Wiltschko & Johannes Heim
University of British Columbia
is paper explores the form, function and distribution of certain discourse
markers which seem to occur outside traditional clause boundaries and are
used to request conrmation. ese ‘conrmationals’ dier according to what
is expected to be conrmed. Some conrmationals trigger a response from the
addressee to conrm that the proposition is true; others require a response to
conrm that the addressee knows that the proposition is true. is variation is
reminiscent of scope eects and suggests that conrmationals should be analysed
in syntactic terms despite their peripheral position. For our analysis, we adopt the
Universal Spine Hypothesis (Wiltschko 2014), which promotes a hierarchically
organized series of core functional projections. We propose that the highest
functional projection of a clause is dedicated to a ‘grounding’ layer, which in turn
consists of a speaker-oriented and an addressee-oriented structure. e topmost
layer is dedicated toregulate response and consists of a position that encodes
thecall on the addressee. Our analysis of speech act structure is an updated
version of Ross’ (1970) performative hypothesis. We explore the predictions and
implications of this hypothesis for the syntax-pragmatics interface.
1. Introduction
In this paper we explore the syntax of a class of sentence-peripheral particles, namely
those that are sometimes classied as invariant tags (Columbus 2010). A typical exam-
ple of these particles is eh as in (1), which has similar functions as the full tag in (2).1
1. eh is oen characterized as a genuinely Canadian particle (Avis 1957; Woods 1980).
However, it occurs in other dialects of English as well (e.g., Guernsey and New Zealand
English) though its prosodic and distributional properties vary across different dialects
(within and outside of Canada; Gold 2008). e function of eh reported in this paper reflects
its use by speakers who grew up in Western Canada (Alberta and BC). is study is intended
as a case-study and is meant to show what a possible system of confirmationals may look like

 Martina Wiltschko & Johannes Heim
(1) You have a new dog, eh?
(2) You have a new dog, don’t you?
We show that with the use of eh in (1) the speaker (henceforth S) requests that the
addressee (henceforth A) conrm that the proposition expressed in the host sentence
is true. To reect this function of tags such as eh, we refer to them as conrmationals.
e main goal of this paper is to explore the form and function of conrmationals
and to propose a syntactic analysis of such particles within a generative framework. In
particular, our goal is to illustrate how a generative syntactic approach can serve as a
valuable heuristic in the exploration of discourse markers. It allows for in-depth analy-
sis of the multi-functionality of conrmationals, which in turn can serve as the basis
for the comparison of dierent conrmationals within and across languages.
It may be surprising to approach the exploration of conformational particles from
a syntactic point of view. Aer all, as sentence-peripheral particles, they do not seem to
display much in the way of visible syntactic eects (they cannot be modied, moved,
or coordinated, for example). However, there are several reasons why we think that
a syntactic approach is justied. First, there is a long tradition of developing syntac-
tic analyses for full tags (Ross, 1970; Culicover 1992; Sailor 2009) and, crucially, they
have a similar function and distribution as particle (invariant) tags. On the assumption
that particular hierarchical structures are associated with particular functions (as in
Wiltschko’s 2014 Universal Spine Hypothesis) the functional similarity between full
tags and invariant tags is an indicator of their syntactic similarity despite their dier-
ence in form. Second, it is generally acknowledged that full phrases (syntactic objects)
can be replaced by simple words, namely in the form of pro-forms, which however
are still analysable as syntactically complex (e.g. Postal 1969; Déchaine & Wiltschko
2002). While pro-forms are mostly explored in the nominal domain, pro-forms that
replace clausal constituents have recently received attention. In particular, in Kria
(2013), response particles such as yes and no are analysed as propositional anaphors. In
this context it is interesting to note that response particles such as yes and no in(3), just
like conrmationals, have a syntactically complex counterpart in the form of echo-
verb constructions, as illustrated in (4) on the basis of Finnish (Holmberg 2016).
(3) Q: Do you have a new dog.
A: a. Ye s (I do).
b. No (I don’t)
(4) Q: Tul-i-vat-ko lapset kotiin? F
come--3- children home
‘Did the children come home?’
A: Tul-i-vat.
come--3
‘Yes.’ (lit.: ‘ey came.’) (Holmberg 2016, example 5)

e syntax of conrmationals 
We thus assume that a generative syntactic approach is justied as an analytical tool.
In particular, given that syntax serves as the module that mediates between form and
interpretation, we view it as an ideal analytical tool: it allows for an interesting way to
understand the relation between the prosodic, semantic, and pragmatic properties of
conrmationals.
However, while the general tenets of generativism are readily adaptable to the
exploration of conrmationals, we have to make some specic assumptions: namely
that syntactic structure is not restricted to propositional sentences (henceforth
p-structure) of the familiar type, but instead includes speech-act structure (hence-
forth SA-structure). is is consistent with the observation that speakers have clear
judgments about the context of use of conrmationals suggesting that just like
p-structure is part of a grammatical competence, so is SA-structure. We thus assume
that speakers have knowledge about their use, which may be viewed as a conversa-
tion competence similar in nature to the competence readily assumed for standard
syntactic phenomena.
Hence, we argue for an updated version of Ross’ (1970) performative hypoth-
esis, according to which p-structures (such as the sentence in Example5) are domi-
nated by a SA-structure. As illustrated in Figure 1, under Ross’ original performative
hypothesis, SA-structure is comprised of a representation of S, A and a performative
verb such as tell. To account for the fact that the sentence in (5) does not always mani-
fest traces of SA-structure, Ross further argues that SA-structure undergoes a process
of deletion (performative deletion). However, crucially as we will see, not all elements
of SA-structure are deleted. Given the model that Ross assumes (Transformational
Grammar), this analysis accounts for the fact that at the level of interpretation (Deep
Structure) a sentence is complex in that it contains SA-structure (which is itself a
higher order p-structure). is additional structure is responsible for associating the
clause-type expressed in p-structure with its primary illocutionary force (assertion
in Example5).
(5) I have a new dog.
S
SI tell you that
SA-
structure
I have a dog
p-
structure
Surface Structure (form)
Performative
Deletion
Deep Structure (interpretation)
S
I have a dog
p-
structure
Figure 1. Ross’ (1970) performative hypothesis

 Martina Wiltschko & Johannes Heim
We argue here for an updated version of Ross’ original insight. In particular, follow-
ing insights by Speas and Tenny (2003), we propose that the SA-structure consists
of a functional architecture above the functional structure assumed in contemporary
generative syntactic theories of the clause-structure. In particular, we argue that the
function associated with this functional architecture is grounding (in the sense of Clark
& Brennan 2009) and the Call on the Addressee (in the sense of Beyssade & Marandin
2006). Moreover, we argue that sentence-peripheral particles, such as conrmationals,
are associated with this layer of structure and hence provide us with an ideal empirical
domain to explore SA-structure. On this analysis, then, sentence-peripheral linguistic
markers are regulated by the same type of syntactic operations as other elements that
are uncontroversially assumed to appear inside the clause.
is paper is structured as follows. In Section 2, we introduce the core set of data
that forms that basis of our analysis. We then turn to the question as to whether con-
rmationals are best viewed as occurring inside or outside the sentence (Section 3).
In Section 4, we introduce the framework within which our analysis is couched. In
Section 5, we introduce in more detail the analysis for conrmationals, as well as its
predictions. In Section 6, we conclude.
2. e form, function, and distribution of conrmationals
Conrmationals are discourse markers (in the sense of Blakemore 2004; Fraser 2006)
sometimes referred to as invariant tags (Columbus 2010). e latter term reects the
fact that they serve a similar function as tag questions such as those in (6).
(6) a. I have a new dog, don’t I?
b. You have a new dog, don’t you?
c. She has a new dog, doesn’t she?
In terms of their form, the dening feature of the tags we explore here is that they are
invariant particles. at is, in contrast to the tag questions in (6), which vary according
to the linguistic context, conrmationals such as eh in (7) do not change depending on
the linguistic context.
(7) a. I have a new dog, eh?
b. You have a new dog, eh?
c. She has a new dog, eh?
Next we turn to the discourse function of conrmationals. What they all have in com-
mon is that they turn the host sentence into a request for conrmation. e particular
conrmational illustrated in (7) may, however, dier in terms of what S is requesting
A to conrm. Eh may be used to request conrmation for the truth of the proposition

e syntax of conrmationals 
(henceforth p) as shown in (8). In addition, eh may also be used to conrm S’s assump-
tion that A knows that p is true, as shown in (9). Interestingly, in this context, huh, right,
as well as full tags cannot be used. ey can only be used to conrm the truth of p.
(8) John knows that Mary would like to have a new dog. He hasn’t seen her in a
long time. And he keeps wondering whether she got a new dog. One day he
runs into her while she’s walking a new puppy. John utters:
You have a new dog, {eh/huh/right}?
= Conrm that p is true
(9) Mary is walking her new dog when she runs into John. She is expecting that
he would congratulate her on the new dog, but he’s not mentioning it. She
isn’t sure anymore whether he actually realizes that she has a new dog. So
she utters:
I have a new dog, {eh/*huh/*right}?
= Conrm that you know that p is true
Finally, in terms of their distribution, we observe that conrmationals occur in sentence-
peripheral positions. In particular, in English, conrmationals have to appear at the right
periphery, i.e., they are found at the end of a sentence (see the preceding examples). ey
cannot be used sentence-internally (10) or sentence-initially (10), unless they are associ-
ated with the intonational contour of an independent utterance. at is, conrmationals
can also be used as stand-alone utterances, in which case they are associated with dier-
ent discourse functions, as indicated by the follow-up utterances in (11).
(10) a. *You {eh, huh, right} have a new dog2
b. *{Eh, huh, right} you have a new dog
(11) A: I have a new dog.
B: a. Eh? – Can you repeat what you just said? I didn’t understand you.
b. Huh? – I thought you were allergic to dogs.
c. Right! – You were telling me that you would get one.
2. An anonymous reviewer points out that Dutch may have an instance of a sentence-
internal conformational in the form of the particle toch.
i) Jij heb-t toch een nieuwe hond?
2 have-2 part indef new dog
‘You have a new dog, don’t you?’
However, the use of toch does not itself trigger the confirmational reading. Instead, it indicates
that the Addressee should know this. e request for confirmation is triggered by the rising
intonation on the sentence (i.e., it functions as a rising declarative in the sense of Gunlogson
2003). See oma (forthc.) for an analysis of sentence-medial discourse particles in terms of
the syntactic framework developed here.

1 Martina Wiltschko & Johannes Heim
. Are conrmationals inside the clause?
We propose a syntactic analysis for conrmationals. Assuming that the domain of
analysis of syntax is the sentence, this would imply that conrmationals are in fact
part of the sentence and hence are governed by familiar syntactic operations. In other
words, we suggest that conrmationals are in fact inside the clause.3
But how do we diagnose clause-internal elements? In this section we address this
issue. First we discuss some empirical issues pertaining to this question (Section 3.1)
and then we move to some more theoretical considerations (Section 3.2).
.1 Empirical considerations
Based on the distributional properties of conrmationals alone, it is in fact hard to tell
whether conrmationals appear inside or outside the clause. Firstly, consider word
order. In English, conrmationals can only appear sentence-nally, but this in itself
does not really tell us much about the question as to whether they are part of the sen-
tence proper. e fact that conrmationals occur sentence-nally is not a sucient
criterion to suggest that they occur outside the clause.
.1.1 Prosodic considerations
e intonational contours associated with conrmationals suggest that they are tightly
integrated with the sentence proper. at is, conrmationals form a prosodic unit with
the sentence. Unlike sentence-initial conrmationals, which have to form an indepen-
dent prosodic unit if they are to be well-formed, not all sentence-nal conrmationals
are well-formed as independent prosodic units.4
. For the purpose of this paper, we use the terms sentence and clause interchangeably. is
is justified because confirmationals are restricted to root-clauses and hence the distinction
between sentence and clause is irrelevant: to say that they appear inside the clause amounts to
saying that they appear inside the sentence.
. An anonymous reviewer points out that the following dialogue is well-formed despite the
fact that right here is separated from the host clause by the grumbling of the interlocutor and
hence cannot be prosodically integrated.
Context: A and B are staring at a computer screen, scanning through a spreadsheet of newly
acquired data.
A: e results look promising.
[B grumbles ambiguously]
A: Right?
B: ey could be better. e standard deviation’s a little worrying…
However, right is multi-functional in ways that other confirmationals are not. at is, we
would not classify right as a conformational here.

e syntax of conrmationals 11
(12) You have a new dog. {*Eh?/*Huh?/Right?}?
Moreover, note that the ill-formedness of prosodically integrated sentence-initial con-
rmationals is a language-specic property. In particular, some German conrma-
tionals can appear sentence-initially, and in this context they can either be prosodically
integrated or form an independent prosodic unit. Note that in the latter case, the con-
formational obligatorily relates to a preceding sentence. e examples below are from
Swabian German, spoken in the South of Germany (see Heim 2015, for a detailed
discussion of this particle and its intonational patterns).
(13) a. Gell du hoscht an naia Hond.
Conrm you have a new dog.
‘You have a new dog, eh?’
b. Gell? Du hoscht an naia Hond.
Conrm you have a new dog.
‘Conrm! You have a new dog.’
Assuming that sentences dene independent prosodic units, this pattern suggests that
conrmationals can be part of the sentence. However, the distributional properties of
German conrmationals suggest that conrmationals may indeed be outside of the
sentence.
.1.2 Conrmationals occur outside the German Verbalklammer
German word-order diers signicantly from word-order in English. While English
is a strict Subj(ect)-(Aux[iliary])-V(erb)-Obje(ect) language (at least in declarative
clauses), German is not. Descriptively, German declarative sentences may be Subj-V-
Obj as in example (14)a, but this is not always the case. In the presence of an auxiliary
as in example (14)b, word order has to switch from V-Obj to Obj-V; and given the
right discourse context, German also allows for Obj-Aux-Subj-V as in example (14)c.
(14) German word order
a. Du hast einen Hund. Subj V Obj
‘You have a dog.’
b. Du hast einen Hund gehabt Subj Aux Obj V
you aux a dog have.part
‘You have had a new dog.’
c. Einen Hund hast du gehabt. Obj Aux Subj V
A dog aux you have.part
‘You have had a new dog.’
e descriptive generalization that captures this pattern of seemingly free word order
in German is as follows: in matrix clauses the inected verb always appears in second
position; it is only preceded by a single constituent, but this constituent can either be
the subject or the object (or any other constituent for that matter). us in the absence

12 Martina Wiltschko & Johannes Heim
of an auxiliary, the main verb inects and appears in 2nd position (hence German is
known as a V2nd language); however, in the presence of an auxiliary, it is the auxiliary
that inects and hence appears in 2nd position. e occurrence of the main verb in
nal position in this case suggests that German is underlyingly Obj-V. is is con-
rmed by the fact that the Obj-V order is also observed independently of the absence
or presence of an auxiliary.
(15) a. weil du einen Hund hast COMP Subj Obj V
since you a dog have
b. weil du einen Hund gehabt hast COMP Subj Obj V Aux
since you a dog have.part aux
e word order patterns of German dene the so called Verbalklammer (‘verbal
bracket’), as shown in (16).
(16) XP [Vn … O… V]
us, the distributional properties of German are such that we can identify clear
indicators of sentence boundaries: at the le edge (i.e., sentence-initially), there can
only be one constituent (XP) preceding the nite verb, while at the right edge (i.e.,
sentence-nally) the nal verb denes a sentence boundary. Crucially, the distribu-
tional properties of conrmationals in German suggest that they appear outside of
the sentential domain. At the le edge, they appear to the le of the initial constituent,
resulting in what would appear to be a V3rd constellation, while at the right edge they
follow the verb in nal position.
(17) a. Gell, du hast einen Hund gehabt.5 Conf Subj Aux Obj V
conf you had a dog have.part
You had a dog, right.’
b. Du hast einen Hund gehabt, gell Subj Aux Obj V Conf
you aux a dog have.part conf
‘You have had a new dog.’
Crucially, elements that are clearly part of the p-structure (such as adverbs for exam-
ple) cannot occur outside of the verbal bracket.6
. e confirmational gell is a spoken language phenomenon and hence different dialects use
different variants. Gell is meant to be an approximation in Standard German and is here used
for reasons of exposition.
. ere is an exception to the generalization that the verb has to appear in S-final position:
clausal constituents may appear at the right edge. is phenomenon is known as extraposition.

e syntax of conrmationals 1
(18) a. *Gestern, du hast einen Hund gehabt Adv Subj Aux Obj V
‘Yesterday, you had a dog.’
b. *Du hast einen Hund gehabt, gestern Subj Aux Obj V Adv
you aux a dog have.part yesterday
‘You have had a dog yesterday.’
Regarding the question as to whether conrmationals are best analysed as being inside
or outside the clause, we have now encountered two contradictory diagnostics. Sen-
tence prosody suggests that conrmationals may appear inside the clause, while word
order considerations in German suggest that they appear outside the clause. In the
next subsection we discuss the problem from a theoretical point of view, and we argue
that the notion of a clause has to be relativized.
.2 eoretical considerations
To answer the question as to whether conrmationals occur inside or outside the
clause, it is essential to dene the very notion of a clause. While we have an intuitive
understanding of what it means to be a clause, it is not fully straightforward to provide
a precise denition. As we will see, what constitutes a complete clause depends in part
on the linguistic context.
As a rst approximation we could dene a clause as the linguistic unit that mini-
mally contains a subject and a predicate and expresses a proposition. is denition
is necessary to capture the properties of all sentences, but it is not sucient. Consider
the example in (19). We observe that in some contexts the subject precedes a bare
verb which in turn precedes the direct object. is is known as a small clause precisely
because it constitutes a minimal sentence (19). In structural terms, small clauses can
be analysed as bare VPs as in Figure 2.
(19) I saw [John walk his dog]
VP
N
PSubj
John
V′
V
walk
NPObj
his dog
Figure 2. Small clause structure
In terms of their distribution, small clauses are limited to certain embedded
contexts, however. For example, in (19) it occurs with a verb of perception. In
most contexts, the verb has to be inflected for tense and agreement, as shown in

1 Martina Wiltschko & Johannes Heim
(20)a. A small clause cannot serve as a matrix clause, as shown in (20)b, which is
ill-formed.7
(20) a. John walks his dog.
b. *John walk his dog
In terms of their structure, matrix clauses with an inected verb are analysed as clauses
where the bare predicate argument structure (the VP) is dominated by functional
architecture. (e amount of structure assumed for such clauses diers across dierent
frameworks; see for example Cinque [1996] for the most articulated structure above
VP). is functional architecture is responsible for hosting the required inectional
categories (Pollock 1989).
IP
N
PSubj
John
I′
I TENSE
walks
VP
NPSubj
John
V′
V
walk
NPObj
his dog
Figure 3. Matrix clause with inected verb
Finally, in embedded contexts, some verbs require a sentence introduced by a comple-
mentizer, as shown in (21)a. In this context, neither a small clause structure nor a
clause with an inected verb but without a complementizer are well-formed (21)b/c,
at least not for all speakers of English.
(21) a. I regret that John walks his dog.
b. *I regret John walk his dog.
c. *I regret John walks his dog.
In terms of their structure, clauses introduced by a complementizer are assumed to
be dominated by even more functional architecture, namely a CP (where C stands for
complementizer; Chomsky 1986).
. We use the term ‘ill-formed’ to refer to examples that are unattested and judged as bad. It
is important to note that in some cases, ill-formedness is relativized to a particular context and
hence may rather be labelled as infelicitous. However, since the line between ungrammatical
and infelicitous is not always easy to draw, we will consistently use the term ill-formed.

e syntax of conrmationals 1
IP
NPSubj
John
I′
C′
C
that
CP
I TENSE
walks
VP
NPSubj
John
V′
V
walk
NPObj
his dog
Figure 4. Embedded clause preceded by complementizer
is establishes that sentences may grow depending on the immediate linguistic con-
text. Within the generative tradition, the size of a sentence correlates with the complex-
ity of the functional architecture dominating the bare predicate-argument-structure
(VP). Consequently, the denition of a sentence is not as straightforward as it may
seem at rst sight. e denition of a sentence as consisting of a subject and a predicate
is not sucient. Rather we can approximate the denition of a clause as in (22).
(22) A clause is the maximal projection of the highest functional category
associated with a small clause.
Under verbs of perception, a VP suces; an IP is required for matrix declarative
clauses; and certain verbs embed a CP. is is illustrated in Figure 5.
IP
VP
CP
complement
clause
nite
clause
small
clause
Figure 5. Growing clause-structure
What is relevant for our purposes then is that the size of a clause may vary: it dif-
fers depending on the linguistic context. If so, it is not straightforward to determine
whether or not conrmationals are inside or outside the clause. If we assume that

1 Martina Wiltschko & Johannes Heim
clauses can grow with the linguistic context, then the conclusion we have to draw is
as follows: given what we have seen above, conrmationals denitely occur outside of
small clauses or matrix nite clauses, but at the same time, we now know that clauses
can grow. So in this way we can say that there might be contexts in which the clause
grows to include structure that hosts conrmationals. at is, in certain contexts,
most notably in conversations, conrmationals (or other forms that serve to modify
the speech act) are obligatory elements of the clauses in which they are contained.
Consider the examples in (23): in this context it would be awkward for Mary to sim-
ply utter an assertion in the form of a declarative clause with falling intonation as
in example(23)a. But the sentence can be modied in various ways to become well-
formed. It can either be uttered with a dierent intonational contour (a surprise into-
nation 23b), it can be followed by an epistemic marker (23)c, an evidential marker (23)
d, or it can be followed by the conformational eh (23)e.8
(23) Mary runs into John who is walking his new dog. She didn’t know that John
has a new dog, so she utters:
a. #You have a new dog.
b. You have a new dog!
c. You have a new dog, it seems.
d. You have a new dog, I see.
e. You have a new dog, eh?
us, in certain contexts, speech act modiers such as conrmationals are obligatory
and hence we may conclude that there may be further functional structure dominating
the clause.
is is precisely the line of analysis we pursue in this paper. In particular, we side
with Ross (1970) in assuming that p-structure is dominated by a functional architec-
ture which is responsible for encoding the way the speaker relates to the utterance,
how the speaker thinks the addressee relates to the utterance, and nally what the
speaker wants the addressee to do with the utterance. For now, we simply label this
structure FP (for functional phrase) and suggest that this type of clause corresponds
to a conversational clause as in Figure 6. A more detailed analysis will be presented in
Sections 4 and 5.
In sum, there is no a priori reason to think that units of language (UoLs) that
express these relations should be considered outside of the sentence. is is eectively
expressed in the following quote: “What if we make the prototype sentence one in
which the bulk of the information is about the relationship between the interlocutors?
(Richard A. Rhodes, facebook, emphasis added).
. We are grateful to an anonymous reviewer for pointing out the relevance of the evidential
and epistemic markers in this context.

e syntax of conrmationals 1
What about the assumption that elements that are taken to be outside of the clause
proper are outside of ordinary sentence grammar? In this context it is important to
observe that even with what is considered to be the sentence proper, dierent domains
have dierent grammatical properties. e VP is the domain of argument-structure;
this is where predicates are saturated by their arguments. e IP is the domain of case-
and grammatical role assignment and the target domain for A-movement. CP is the
domain of topic/focus structure and the target for A′-movement.
IP
VP
CP
A′-syntax
focus structure
A-syntax
case structure
theta-syntax
argument-structure
Figure 7. Syntax across dierent domains
Since each of these syntactic domains that are considered to be inside the sentence proper
is associated with dierent syntactic properties, it should not come as surprise that
another (higher) domain should also be associated with dierent grammatical properties.
In the next section, we discuss a framework that will give us the means to include
the grammatical properties of those sentences that are embed in conversations.
. A framework for analysis
e goal of this paper is to develop an explanatory model for conrmationals which
not only accounts for their empirical properties, but which also allows us to set the
stage for cross-linguistic comparison.
CP
IP
FP
conversation
clause
complement
clause
nite
clause
VP small
clause
Figure 6. Growing clause-structure

1 Martina Wiltschko & Johannes Heim
e existing literature on these elements focuses mainly on socio-linguistic
aspects (Meyerho 1992; Stubbe & Holmes 1995; Cheshire & Williams 2002; Gregg
2004), but has nothing to say about the range of variation of conrmationals within
a language, letalone across languages. Most treatments of conrmationals describe
their dierent functions without attempting to correlate formal, functional, and dis-
tributional properties. As a consequence, we nd analyses that postulate up to sixteen
functions of eh (e.g., Columbus 2010). Formal syntactic approaches towards eh are
almost non-existent (with the notable exception of Gibson 1976). In this paper, we
adopt the Universal Spine Hypothesis (Wiltschko 2014) to develop a formal syntactic
analyses of conrmationals. One of the key assumptions of this framework is that the
syntactic domains identied in Figure 7 are associated with core abstract functions
( Section4.1). Given that we assume that conrmationals are associated with the high-
est functional architecture, which denes the conversation clause, we have to iden-
tify the core functions of this domain. In order to do so, we introduce some recent
advances in the analysis of speech acts (Section 4.2).
.1 e Universal Spine Hypothesis
e Universal Spine Hypothesis (henceforth USH) was developed to provide a frame-
work for discovery and comparison of language-specic categories. e motivation
for the USH is rooted in the tension between generative assumptions and typologi-
cal ndings. at is, the main tenet of generative grammar (Chomsky 1965) has it
that the human language faculty is innate and, consequently, that the languages of
the world share some core properties. is is known as Universal Grammar (UG).
Relevant for our purpose is the assumption that the building blocks of any sentence,
no matter which language, can be characterized in terms of the functional categories
that comprise the architecture in Figure 7. Under this conceptualization of UG, at least
some linguistic categories are assumed to be innate. According to many typologists,
this assumption is in stark contrast with their ndings as illustrated by the following
quote from Haspelmath (2007: 119): “almost every newly described language presents
us with some ‘crazy’ new category that hardly ts existing taxonomies.”
e USH is designed to be a framework that comes to terms with this tension in
that it seeks to provide a tool to analyze the types of “‘crazy’ new categories” Haspel-
math refers to. Conrmationals may be classied as such a “‘crazy’ new category”
because their properties do not t into existing taxonomies. First, there are no existing
taxonomies for word-classes that seem to occur outside the clause. And second, con-
rmationals even display some unexpected formal properties that are independent of
their distribution. For example, conrmationals might be classied as particles, if only
because they do not t the prole of any other type of word class. e problem with this
classication is, however, that particles are usually considered to be not inectable. But

e syntax of conrmationals 1
conrmationals in Austrian German may be inected. In particular, the form of the
conrmational goi diers depending on whether the relation between the speaker and
the addressee is informal (the proper address is du) or formal (the proper address is
Sie).9 In the former case the conrmational appears in its bare form goi (24)a whereas
in the latter case the conrmational is suxed by -ns (24)b. Finally, some speakers of
Upper Austrian German also make use of a special form for plural addressees (24)c.
(24) a. Ea hot an neichn Hund, goi
He has a new dog, .2informal
b. Ea hot an neichn Hund, goi-ns
He has a new dog, -2formal
c. Ea hot an neichn Hund, goi-ts
He has a new dog, -2
us conrmationals comprise a word-class that is not typically considered in formal
syntactic theories. Given their properties, it is not immediately clear how they should be
treated. is is precisely the type of methodological problem the USH seeks to address.
e core assumption of the USH is that the functional architecture dominating
small clauses does not consist of pre-dened categories (such as  or ) but
rather that these categories are constructed on a language-specic basis. In particular,
Wiltschko (2014) argues that grammatical categories (c) are constructed by means of
language-specic UoLs and a universal syntactic spine, which is itself comprised of a
series of abstract categories (κ).
(25) c = κ + UoL
us UG is not to be considered as a repository of universal categories. Instead
under this conceptualization, UG restricts categorization patterns. us, the way
to approach “‘crazy’ new categories” (such as conrmationals) in a way that makes
them amenable to comparative analyses is to determine where and how they associ-
ate with the spine. e crucial assumption about the spine that is relevant for our
purposes is that each layer comes with a particular abstract function, which may in
turn be based on general cognitive functions (see Ramchand & Svenonius 2014, for
a similar view). is contrasts with most current analyses of clausal architecture,
. For this reason, an anonymous reviewer suggests that the Upper Austrian German con-
firmationals are more appropriately termed variant tags. However, variant tags typically refer
to tag questions of the form ‘doesn’t he; isn’t it,’ etc. which involve the copying (or addition) of
an auxiliary and the subject. is differs from the inflected confirmationals of Upper Austrian
German in that we are still dealing with a particle, however, one that can be suffixed with an
appropriate agreement ending. Crucially, this agreement suffix does not agree with an argu-
ment of the host clause, but instead with a speech-act participant.

2 Martina Wiltschko & Johannes Heim
which assume that the labels of the categories are dened by their substantive con-
tent (such as Inner Aspect, Outer Aspect, and Tense). According to the USH, such
categories are dened by the core function of the spine and language specic UoLs
which provide the substantive content. As schematized in Figure 8, the rst layer
serves to classify the event; if this classication is based on temporal content like
telicity, then the resulting category is inner Aspect (IAsp; also known as Aktionsart).
e second layer serves to introduce a point of view, and thus to add a perspective
relative to which the event is viewed; if this point of view is based on time (i.e., a ref-
erence time), then the resulting category is (outer) Aspect (OAsp). e third layer
serves to anchor the event to the utterance; if anchoring is based on time (when
relative to the utterance did the event take place?), then the resulting category is
tense (TP).
OAspP
IAspP
TP
Anchoring

 


Perspectivizing
Classication
Figure 8. Correlating the spine with language specic categories
Analysing a given UoL in terms of the USH involves (among other things) determin-
ing where on the spine it associates with. is involves not only determining its rel-
ative hierarchical position within sentence structure, but also its absolute position.
Assuming that the linear order of UoLs tells us something about their hierarchical
position (Kayne, 1995) we can (in part) glean their relative position from linear order-
ing eects. However, to determine their absolute position, the function of UoLs has to
be taken into consideration.
In this way the USH serves to solve the tension between universalist theories
(according to which languages share a core inventory) and typological claims (accord-
ing to which the inventory of languages diers beyond comparison). On the one hand,
according to the USH, grammatical categories are always language-specic, precisely
because they are construed with language-specic UoLs; but at the same time, these
categories have much in common because their construction is constrained by the
universal spine. e USH is thus a well-suited framework to investigate “‘crazy’ new
categories” including conrmationals.

e syntax of conrmationals 21
As we have seen in Section 2, the distribution of conrmationals suggests that they
occur above the conventionally assumed clausal architecture, i.e., above the anchoring
catego r y. 10 is suggests that the spine must be extended; and given the discourse
function of conrmationals, this function relates to how the proposition encoded in
the bare sentence is integrated into the conversation. at is, if we assume with Hinzen
(2014) that the spine corresponds to our language of thought, it doesn’t come as a sur-
prise that part of the spine is dedicated to the communicative aspect of language, i.e.,
how we package our thoughts to relate them to others.
In Heim etal. (2014), it is argued that there are in fact two functions that play a
role here. One of these layers is dedicated to relating the speaker’s attitude towards the
proposition. We refer to this as the grounding layer. e second (higher) layer is dedi-
cated to letting the addressee know what the speaker wants them to do with the utter-
ance, i.e., the Call on the Addressee (in the sense of Beyssade & Marandin 2006). We
refer to this as the responding layer. is is illustrated in Figure 9, where S represents
the traditional clausal architecture.
RespP
GroundP
S
CoA
responding
propositional
attitude
“sentence”
Figure 9. Extending the spine
In the next subsection, we provide some independent motivation for the postulation
of these two functions based on recent work within speech-act theory as well as con-
versation- and discourse-analyses.
1. is structure differs from the fine-grained structure of the le periphery as proposed in
Rizzi 1997. e latter is meant to capture the distribution of force, and information-structural
notions such as topic and focus whereas the SA-structure we explore here is meant to encode
conversational structure.

22 Martina Wiltschko & Johannes Heim
.2 e complexity of speech acts
e goal of this section is to establish that in the context of a conversation, speech acts
are more complex than assumed in the traditional speech act analysis (Searle 1969).
e speech act structure assumed in Ross’ (1970) performative hypothesis reects
the following discourse conditions, which have been commonly assumed to be associ-
ated with assertions. ese are given in (26):
(26) Discourse conditions for assertions
i) S believes p
ii) S wants A to believe p (adapted from Bach & Harnish 1979)
Under normal conditions, the utterance of a declarative clause-type is associated with
the primary illocutionary force of assertion, which is only well-formed if the condi-
tions in (26) hold. Hence, Ross’ (1970) performative hypothesis has it that the SA-
structure above the p-structure can be translated into another p-structure of the form
S gives p to A. is can be schematized as in Figure 10 where S believes p (Bel [p]),
asserts p, and as a consequence A believes p.
Speaker
Bel (p) Bel (p)
p
Addressee
Figure 10. What happens when S asserts p (preliminary version)
More recent analyses of speech acts, however, highlight the fact that their dialogi-
cal aspect is more complex. In particular, in current theories of speech acts and
conversations, researchers have introduced the concept of a ‘table’ (Farkas & Bruce
2009) – an (imaginary) space in the dialogical eld on which one can place proposed
additions to the common ground (Stalnaker 1978). Introducing the table is a way
to model the idea that assertions do not change the common ground, they merely
propose a change (see also Clark & Schaefer 1989; Clark 1992; Ginzburg 1996, 2012;
Malamud& Stephenson 2006).
Accordingly, an assertion doesn’t simply transfer p from S to A, but instead the
assumption is that S puts p on the table. It can be viewed as an intermediate stage in
which S request that A adopts p into their set of beliefs. Clark and Schaefer (1989)
refer to this stage in the conversation as the presentation phase. We can illustrate this
as in Figure 11a. S believes p and by asserting p communicates that s/he believes p. At
this point S can at least assume that A believes that S believes p. But S does not know
whether A also adopts p into their set of beliefs. is is where the acceptance phase,

e syntax of conrmationals 2
illustrated in Figure 11b comes in. In this phase, A indicates that s/he believes p. Only
at this point will S know that A believes p, and hence that the communicative goal of
the assertion is achieved.
a. Presentation phase
Speaker
Bel (p) Bel (S,p)
Bel (S,p)
Addressee
b. Acceptance phase
Speaker
Bel (p)
Bel (A,p)
Bel (S,p)
Bel (p)
Bel (A,p)
Addressee
Figure 11. What happens when S asserts p (still preliminary)
us, any version of the performative hypothesis that wants to do justice to the com-
plexity of speech acts embedded in conversations has to encode more than “S gives p
toA.” By putting the assertion that S believes p on the table, S requests A to also believe
p. us, conversational moves consist of putting a proposition on the table AND ask-
ing A to do something with it. e latter aspect of the conversation situation is known
as the Call on Addressee (CoA; Beyssade & Marandin 2006). Accordingly, an assertion
is a complex conversational move: it consists of a proposition (27)a, the speaker’s atti-
tude towards it (27)b and some instructions for A about what to do with p (27)c. In
(27) pa stands for propositional attitude.
(27) Ingredients of an Assertion
a. p
b. (pa [S, p])
c. Req (pa [A,p])
e claim that speech acts may be complex is explicitly put forth in Beyssade & Maran-
din (2006: 1): “We claim that utterances impact dialogue in two ways that we describe
in terms of update. On the one hand, Speaker commits herself to some content: utter-
ing amounts to update Speaker’s commitments. On the other hand, Speaker calls on
Addressee for him to change his own commitments.” Hence we have to modify the
presentation phase of a typical assertion to include the CoA. What is on the table is not
only S’s attitude towards p, but also what S wants A to do with p. us, the presentation
phase of a well-formed utterance can be represented as in Figure 12.

2 Martina Wiltschko & Johannes Heim
Presentation phase
Speaker
Bel (p) Bel (S,p)
Req (Bel(A,p))
Bel (S,p)
Addressee
Figure 12. What happens when S asserts p (nal version)
. Matching complex speech acts onto complex sentences
e complexity of dialogical sentence structure proposed in Section4.1 matches the
complexity of speech acts. p is expressed in the familiar p-structure. Above p-structure,
we postulate a layer dedicated to encoding the speaker’s attitude towards p.11 We
refer to this as the grounding layer (see also oma, forthc.). e term grounding is
used in Clark & Brennan (1991) to capture the conversational moves introduced in
Section4.2: “For whatever we say, our goal is to reach the grounding criterion: that we
and our addressees mutually believe that they have understood what we meant well
enough for current purposes. is is the process we have called grounding” Clark &
Brennan (1991: 147). And nally, the highest layer, which we refer to as the response
layer is dedicated to encoding what S wants A to do with p (CoA).12 us, to incorpo-
rate the complex conversational moves discussed thus far, we assume that the spine is
extended as in Figure 13.
In sum, the proposed framework is an updated version of the performative hypoth-
esis. It has in common with Ross’ original proposal that p-structure is dominated
by SA-structure, it diers in that the SA-structure is not itself a form of p-structure
(which it is in Ross’ original analysis). In particular, S and A are not represented as
participant roles but instead they are indirectly encoded by virtue of representing the
commitment towards the utterance in GroundP. is commitment can be S-oriented
and A-oriented, though the latter is always mediated through S’s perspective (Heim
etal. 2014, oma forthcoming.)
11. As we shall see in Section5, there is evidence that the grounding layer is in fact more
articulated such that it can be divided into two separate layers: one for encoding S’s attitude
towards p, and one for encoding (what S believes to be) A’s attitude towards p (Lam 2015;
Wiltschko 2015; oma, forthc.).
12. Another related function associated with this layer is to encode that the utterance serves
as a response to a previous conversational move. us, the UoLs that serve to acknowledge
that the tabled proposition has made it into A’s set of beliefs, UoLs that serve to answer ques-
tions or to backchannel are assumed to associate to this domain (Wiltschko, forthc.).

e syntax of conrmationals 2
e same holds for RespP, which may also be relativized to either A (requesting that
A respond) or to S (asserting that S’s utterance is to be interpreted as a response; cf.
Wiltschko in press). As for Ross’ (1970) claim that SA-structure contains a feature
bundle corresponding to a performative verb, this claim is recast in terms of a complex
functional structure representing grounding and responding as the main spinal func-
tions (in the sense of Wiltschko 2014).
In what follows we show that conrmationals lend support to the claim that
speech acts are complex and that their complexity has a structural correlate. at is,
given that there are UoLs that target the dierent functions we have identied (ground-
ing and CoA), we may conclude – following Ginzburg (2012) – that these interactive
functions are built into the grammar.
. Towards a syntax of conrmationals
A structural account of conrmationals needs to integrate their complex conversa-
tional properties, which come with dierent levels of speech act modication. is
section discusses the role of syntax as a mediator between form and interpretation in
order to reect the place of speech acts within the universal spine. e complex struc-
ture resulting from this discussion will host a layer dedicated to the dierent speech
act roles and the CoA to reect that conrmationals always come with a request about
the believe, and not only about the truth of p.
.1 e discourse function of conrmationals
As we saw in Section 2, conrmationals are used to modify speech acts. To see this,
consider the data in (28). A declarative sentence like You have a new dog is typically
RespP
GroundP
S
Req (pa (A,p))
pa (S,p)
p
Figure 13. Matching the ingredients of speech acts onto the spine

2 Martina Wiltschko & Johannes Heim
associated with the primary illocutionary force of assertion. As we have seen above,
typical assertions are complex conversational moves that involve presenting S’s atti-
tude towards p: Bel (S,p); and a second move involves S’s request for A to adopt the
same attitude towards p: Req (Bel[A,p]). For this to be a well-formed conversational
move, S has to assume that A does not already believe p, otherwise S’s request would
be redundant. e context in (28) provides the appropriate discourse conditions for
the assertion to be well-formed. Crucially, in this context the use of a conrmational
is ill-formed.
(28) Mary has long known that John wants a new dog, but that he needs a little
push to actually make the decision. Mary has a friend who works at the ani-
mal shelter and one day she asks Mary whether she knows anyone who may
be able to take an abandoned dog who is in desperate need of a good home.
Without hesitating, Mary picks up the dog, and brings him over to John’s
place. As he opens the door, she tells him:
a. Surprise! You have a new dog.
b. #Surprise! You have a new dog, {eh/huh/right}?
ere is however nothing intrinsic about the declarative ‘You have a new dog.’ that
would make it incompatible with a conrmational. It is just that the particular context
in (28) is incompatible with the discourse function of conrmationals. To see this,
consider the data in (29). Here the declarative can be modied with a conrmational.
If the declarative is not followed by a conrmational, it has to be uttered with a dier-
ent intonational contour than the declarative in (28). is is indicated by the exclama-
tion mark.
(29) Mary has long known that John wants a new dog, but she doesn’t know
whether he has gotten one yet. One day, she runs into John while he is walk-
ing a puppy. She concludes that he must have nally gotten a new dog and
utters:
a. You have a new dog!
b. You have a new dog, {eh/huh/right}?
Let us start by exploring the discourse function of the conrmationals in (29)b. ey
modify the typical assertion in two ways. First, they modify S’s commitment towardsp.
In this context, S is not certain that the proposition expressed in the declarative is in
fact true, though they have some good reason to believe that it might be true. We can
express this attitude of uncertainty towards p by placing both Bel (p) and Bel (¬p)
into S’s set of beliefs joined by a disjunction operator, as in Figure 14. We indicate the
bias towards Bel(p) by underlining it. Aside from modifying S’s attitude towards p,
conrmationals also change the CoA. Whereas in regular assertions S requests for A
to believe p, a conrmational requests for A to respond to S’s utterance. In the context
given in (29), where S is unsure about the truth of p, and where S knows that A is the

e syntax of conrmationals 2
source of the truth of p (in the sense of Gunlogson, 2003) the use of the conrmational
results in a speech act whereby S requests conrmation for the truth of p. As illustrated
in Figure 14 the presentation phase includes a biased attitude towards p and a request
for response. In the subsequent acceptance phase A conrms the truth of p, by provid-
ing an appropriate response (such as yes).13 Once A has conrmed the truth of p, S can
replace the disjunctive belief with the positive belief.
Speaker
Bel (p)
∨
Bel (¬p)
Bel (p)
Bel (S,p)
∨
Bel (S,¬p)
Req (Resp (A,p))
Bel (p)
∨
Bel (¬p)
Addressee
Speaker
Bel (p)
Bel (A,p) Bel (p)
Bel (p)
Bel (S,p)
∨
Bel (S,¬p)
Addressee
Figure 14. Modifying declaratives with a conrmational
us, what dierentiates the two contexts in (28) and (29) has to do with the propo-
sitional attitude of S towards p (belief or uncertainty) as well as with what S wants A
to do with the utterance (adopting the same pa or responding to clarify which pa is
appropriate). us, by adding a conrmational to a declarative clause, the speech act
changes from an assertion to a request for conrmation. is is summarized in Table1 .
Table 1. Modifying speech acts
Unmarked declarative Declarative+conrmational
What S believes Bel (p) Bel (p) ∨ Bel (¬p)
What S wants A to do Req (Bel [A,p]) Req (Resp [A,p])
Speech act Informative assertion Request for conrmation
1. is is not to say that A has to respond with an assertion; it merely constitutes one of the
possible responses that will make this a felicitous conversation. If, on the other hand, Adoesn’t
have the relevant information to confirm, s/he may respond with a statement such as Ihave no
idea. While this will not resolve the Question under Discussion tabled by S, it will be a well-
formed response nevertheless.

2 Martina Wiltschko & Johannes Heim
With this analysis, we can account for the distribution of speech acts in (28)–(29). e
context in (28) is such that S wants A to inform about p. Hence only assertions are
well-formed; the conrmational is ill-formed. In contrast, the context in (29) is such
that S is not quite sure about the truth of p, but has reason to belief that A will be able to
verify. us, in this context, the conrmational is well-formed. e question remains,
however, why the declarative is also well-formed. As mentioned above, in this context,
the declarative is marked by a special intonational contour, like for example a surprise
intonation.14 us, a regular declarative can but need not be interpreted as an infor-
mative assertion. As we shall see, the syntactic analysis for conrmationals we develop
below is able to capture the ambiguity of unmarked declaratives.
.2 Associating conrmationals with the spine
In this subsection, we argue that conrmationals associate with the spine. We show
that there is evidence that conrmationals are in fact complex, consisting of the form
‘eh’ and the intonational contour associated with it. Each of these UoLs serves a dif-
ferent function. We rst introduce the syntactic analysis in more detail (5.2.1). en
we provide independent evidence for the decomposition of ‘eh’ into its components
(5.2.2); and nally we discuss the issue of non-informative assertions.
.2.1 e analysis
As we saw above, with the addition of a conrmational to a declarative, two things are
being modied: S’s attitude towards p and what S wants A to do with p. e result is a
request for conrmation, which can be viewed as a biased question. Recall that one of
the core assumptions of the syntactic framework we adopt (the USH) is that syntactic
layers can be identied by their core function. us, if a particular UoL fullls a par-
ticular function, we can conclude that it associates with the corresponding syntactic
layer. Hence we are lead to conclude that conrmationals simultaneously associate
with the grounding layer, where S’s propositional attitude is encoded, and the response
layer, where CoA is encoded.15 is is schematized in Figure 15.16
1. e nature of this contour, the range of contours possible in this context, and the ques-
tion as to whether a regular assertive contour is felicitous has to be further explored.
1. An anonymous reviewer asks whether confirmationals may be analysed as bi-clausal
structures involving both an ASSERT operator and a QUEST(ion) operator. is would cer-
tainly capture the interpretation of confirmationals. It is however less economic than the
present proposal in that it would require an obligatory process of deleting the complement of
one of the operators because only one of the two p-structures that would have to be postulated
can be spelled out.
1. Confirmationals typically appear in sentence peripheral position. Sometimes right-
peripheral (as in English), sometimes le-peripheral (possible in German). e peripheral

e syntax of conrmationals 2
RespP
GroundP
conrmational
S
CoA
propositional
attitude
p
Figure 15. Associating conrmationals with the spine
However, there is evidence that conrmationals are in fact complex: they consist of
the lexical form (e.g., eh) and a rising contour. Heim etal. (2014) show that each of
these components of the conrmational eh is associated with a distinct position.17 In
particular, the lexical form (eh) associates with the grounding layer, whereas the ris-
ing contour is associated with the response layer, as illustrated in Figure 16 where ‘/’
represents a rising intonation.
RespP
GroundP
/
Seh
CoA
propositional
attitude
p
Figure 16. Decomposing conrmationals
position indicates that they associate with the spine in a hierarchically high position. Whether
this position is to the right (head-final) or to the le (head-initial with subsequent movement)
is a question we do not address in this paper.
1. For a discussion of the differences between eh and other confirmationals see Heim etal.
(2014).

 Martina Wiltschko & Johannes Heim
A few remarks are in order in support of this analysis. First, the assumption that intona-
tional contours can serve as UoLs goes back to Gussenhoven (1984; see also Pierrehumbert
& Hirschberg 1990). Moreover, the claim that these intonational morphemes associate
with syntactic structures, just like other UoLs has been proposed in Truckenbrodt (2012).
At the same time however, it has been argued that intonational contours operate over
speech acts (Trinh & Crnič, 2011). Our analysis of conrmationals in Figure16 is able
to capture these insights: speech act structure is part of the syntactic spine, and hence
intonational contours operate over speech act structure by virtue of associating with the
syntactic spine. On this view, syntax is a module that serves as the interface between form
and interpretation – a view that is at the heart of the generative model. Ever since the
Principles and Parameters framework initiated in Lectures on Government and Binding
(Chomsky 1981), syntactic computation has been viewed as mediating the form/meaning
relation without a direct link between the module that computes sound (read o of pho-
netic form – PF) and the module that computes meaning (read o of logical form – LF).
is is illustrated in Figure 17, which represents the standard model of grammar within
the generative tradition (Chomksy 1995, and subsequent work).
syntax
LF
(interpretation)
PF
(form)
Figure 17. e Y-model of grammar
In what follows, we present evidence for the syntactic decomposition of conrmation-
als that our analysis rests upon.
.2.2 Evidence for the decomposition (intonation)
If indeed conrmationals consist of two UoLs (the lexical form and the intonational
contour), which correlate with the two functions they have (modifying S’s commit-
ment and CoA), we predict that each of these UoLs may also have a life of its own. is
is indeed the case, as we now show. First, eh can associate with a dierent intonation,
which correlates with a dierence in CoA. Consider the following example taken from
a popular website on the use of Canadian eh.

e syntax of conrmationals 1
(30) A dierent type of eh
“So I go to this shrink, eh, and he goes like I don’t have no condence, eh. I go,
‘No way, man.’ He goes I should take assertiveness training. Weird, eh? Like
I’m always supposed to be seeking approval, eh, from, you know, other people?
I felt like he could kiss my Royal Canadian, eh? But, sayin’ it wouldabeen too
pushy. Dyuh think?”
〈http://www.billcasselman.com/casselmania/mania_eh.htm〉
ere are several examples of eh which are not followed by a question mark, which is a
way of indicating that there is no rising intonation and consequently these instances of
eh do not require A to respond. is instance of eh is known as narrative eh.
First consider its intonational properties. According to Dombrowski & Niebuhr
(2005, 2010) there are two dierent sentence nal rises, each associated with a dier-
ent communicative function. A convex nal rise is interpreted as questioning, while a
concave nal rise is interpreted as progredient (i.e., signaling continuation).
is is consistent with our ndings: conrmational eh (associated with a convex
rise [Figure 18; le hand side]) requires a response (i.e., is interpreted as questioning)
while narrative eh (with a concave rise; Figure 18 right hand side) does not require a
response, in fact its characteristic function is to be used in story-telling where it can
serve as an indicator of continuing the story telling (i.e., it is interpreted as progredient).
Figure 18. Two dierent intonational contours
Moreover, according to our analysis, the lexical form (without the intonation) is
associated with the grounding layer, and this should not be aected by the change in
intonation. is assumption is consistent with the data: with the use of narrative eh
S indicates that s/he assumes that p is in the common ground aer the time of utter-
ance. If so, this suggests that eh by itself simply indicates that S believes p; it is only the
requesting function of the rising intonation associated with conrmational eh that is
responsible for the eect that S is not sure about the truth of p.
e second piece of evidence for the decomposition of eh into a lexical form and
an intonational contour stems from the fact that the intonational contour can be used
without the conrmational, i.e., a (convex) rise can be associated with unmarked

2 Martina Wiltschko & Johannes Heim
declaratives as well, giving rise to so called ‘rising declaratives’ as in (31) (see Gunlogson
2008, for an extensive discussion).
(31) John runs into his friend Mary. Mary’s old dog has recently passed away and
she was so sad that she told John that she would never get a dog again. It’s
just too hard to lose a pet. One day, John runs into Mary who is walking a
young puppy. So he utters:
You have a new dog?
According to our analysis, rising declaratives then would be characterized by rising
intonation alone, without any material associating with the grounding layer, as sche-
matized in Figure 19.
RespP
S
/
CoA
p
Figure 19. e syntax of rising declaratives
Hence our syntactic analysis captures Gunlogson’s (2003, 2008) insight that rising
declaratives, like interrogatives, fail to commit S to the content of their utterance. As
a consequence, rising declaratives are compatible with two dierent discourse condi-
tions as schematized in Figure 20. At the time of utterance, S may either consider p
without associating an explicit propositional attitude with it (as in i) or else S may
actively believe ¬p (as in ii).
Speaker
i) p
ii) Bel(¬p) Req (Resp(A,p))p
Addressee
Figure 20. Two discourse conditions for rising declaratives
us what characterizes a rising declarative is that something in the immediate con-
text (verbal or non-verbal) triggers S to consider at least the possibility that p, even
though p was not in her previous set of beliefs or even contradicts one of her beliefs.

e syntax of conrmationals 
Conversely, rising declaratives are not compatible with a discourse context in which S
believes p (Bel[p]).
A nal piece of evidence for the syntactic decomposition of conrmational eh into
its lexical form and its intonation comes from the fact that in some languages these
two functions are associated with two separate particles (see Heim etal. 2014 for a
more extensive discussion). For instance, in Medumba (Grasselds’ Bamiléké Bantu)
the sentence-initial particle kʉla serves to mark the propositional attitude of S towards
p (hence we assume it associates with the grounding layer) while the sentence-nal
particle a serves to mark the CoA: S wants A to respond (hence we assume it associates
with the Response layer). us the example in (32) can be analysed as in Figure 21.
(32) k
ʉ
la u ɣʉ
ʙ
ʉ swə a?
part 2 have dog new q
‘You have a new dog, eh?’
RespP
GroundP
a
Skula
CoA
propositional
attitude
p
Figure 21. 2 particles to express a request for conrmation
We have now seen that eh can associate with a dierent type of intonation (i.e., progre-
dient intonation) and furthermore that the questioning intonation on conrmational
eh may also directly associate with p-structure, giving rise to rising declaratives. is
supports the syntactic decomposition of eh we assume here. Furthermore, the fact that
the same type of intonation may associate with unmarked declarative sentences as well
as with sentence-peripheral conrmationals is consistent with our claim that conr-
mationals are “inside the clause.” In particular, sentence-intonation is clearly part of
a sentence and hence has to be analysed as being part of the clause. Consequently, we
can conclude that conrmationals, which can be associated with sentence-intonation
have to also be part of the clause. We model this assumption by associating both con-
rmationals and sentence-intonation with the syntactic spine.

 Martina Wiltschko & Johannes Heim
. Conrmationals that target A’s set of Beliefs
According to the analysis developed thus far, conrmationals modify the speech act at
the layer of the commitment: they activate the function of the grounding layer. us,
we argue that conrmationals do not simply seek to conrm the truth of p, but instead
they seek to conrm the propositional attitude towards p. ese two alternatives
(conrming truth versus conrming belief) cannot be distinguished from each other.
However, there is clear evidence that (at least some) conrmationals do indeed target
the grounding layer. In particular, consider again the contrast between (8) and (9),
repeated below as (33) and (34). It indicates that conrmationals dier in the target of
conrmation. While all three conrmationals considered (eh, huh, right) are compat-
ible with conrming that p is true, only eh can be used to conrm that A knows p.18
(33) John knows that Mary would like to have a new dog. He hasn’t seen her in a
long time. And he keeps wondering whether she got a new dog. One day he
runs into her while she’s walking a new puppy. John utters:
You have a new dog, {eh/huh/right}?
= Conrm that p is true
(34) Mary is walking her new dog when she runs into John. She is expecting that
he would congratulate her on the new dog, but he’s not mentioning it. She
isn’t sure anymore whether he actually realizes that she got a new dog. So
she utters:
I have a new dog, {eh/*huh/*right}?
= Conrm that you know that p is true
us the discourse conditions for the use of eh in (34) dier from those of the exam-
ples analysed thus far. S doesn’t need conrmation for the truth of p, or for her belief
that p because in this context S is the expert. Since Mary is the dog-owner, she knows
that p is true. Nevertheless, the use of eh introduces a request for conrmation, but
in this case it requests conrmation that A knows p. us, the proposition that is
under discussion is not the truth of p (and thus S’s belief that p) but instead S’s belief
that Abelieves p (Bel [A,p]). us, we have on the table the biased, but uncertain
belief that A believes p (modeled like above by means of a disjunction with the nega-
tive belief).
is suggests that at least the conrmational eh does indeed target the grounding
layer, rather than the propositional structure. Following Lam (2015), oma (forthc.),
1. is generalization holds for speakers which use eh. However, there are (esp. younger)
speakers for which right has taken on all of the functions of eh (see Wiltschko & d’Arcy 2015;
D’Arcy, Denis & Wiltschko, forthcoming for discussion). For such speakers, right can also
perform the narrative function.

e syntax of conrmationals 
and Wiltschko (2015), we suggest that the grounding layer is best analysed as a more
articulated layer of structure, as schematized in Figure 23:. In particular, the rst layer
(GroundP) is dedicated to encode S’s propositional attitude towards p (Ground-S)
whereas the second layer (groundP) is dedicated to encode A’s propositional attitude
towards p (Ground-A). For evidence that A-oriented groundP is structurally higher
than S-oriented GroundP see Lam 2015 and oma, forthc.19
RespP
GroundP
GroundP
CoA
responding
Ground-A
Ground-S
CP
propositional
structure
Figure 23. e articulated grounding layer
1. e fact that there is syntactic evidence to the effect that GroundA is higher than
GroundS may come as a surprise: one might expect that S’s attitude towards p would outscope
A’s attitude towards p because how A relates to p must be conceived of via S’s point of view.
However, one can make sense of this state of affairs if we assume that the mere act of assertion
itself will suffice to relativize GrondA to the speaker’s belief. e direct encoding of S’s attitude
towards p is independent of S’s actual assertion of p.
Speaker
Bel (S(Bel (A,p)))
∨
Bel (S(¬Bel (A,p))) Req (Resp(Bel(A,p)))
∨
¬Bel (A,p)
Addressee
Figure 22. Two discourse conditions for “conrm-that-you-know-eh”

 Martina Wiltschko & Johannes Heim
Hence we assume that eh may associate with A-oriented groundP, while right and huh
can only associate with S-oriented GroundP.20
. Conclusion
e goal of this paper was to show that there is a correlation between the complexity
of speech acts and their structural representation. We have explored a class of seem-
ingly sentence-peripheral particles, conrmationals, which serve to modify speech
acts. at is, they turn a formally declarative clause, which typically is interpreted with
assertive force, into a request for conrmation. We have based the analysis on the Uni-
versal Spine Hypothesis (Wiltschko 2014), according to which each structural layer
that makes up sentences is dened by its function. Since conrmationals modify two
separate functions (grounding and the call on addressee), we proposed that senten-
tial structure includes two layers of structure that are dedicated to these functions
(GroundP and RespP).
According to our analysis, then ‘conrmationals’ are constructed: they consist of
two UoLs (eh and rising intonation). In addition, the syntactic context into which they
are inserted also contributes to their interpretation.21
On our analysis, conrmationals are thus considered to be inside the clause, but we
have also shown that the notion of a ‘clause’ has to be relativized because clauses come
in dierent sizes depending on their context. e advantage of considering these modi-
ers to be clause-internal is that properties that correspond to the properties of constit-
uents within the traditional sentence boundaries can be analyzed with existing means.
2. An anonymous reviewer points out that with the right prosody (not provided by the
reviewer), confirmationals can co-occur with tag questions, as in (i) and (ii)
(i) You’re coming back, arent’ you, right?
(ii) You’re coming back, right, aren’t you?
ese are interesting examples, which need to be explored. An informal survey suggests that
the right prosody is such that the second tag serves as an in dependent utterance and that
the discourse context for these utterances is one in which the speaker is second guessing
themselves aer uttering the first sentence. ough the precise prosody and discourse
conditions have to be explored further.
21. An anonymous reviewer suggests that the term confirmational is misleading due to
their complexity. We are reluctant to do so for two reasons. First, we find the term useful for
cross-linguistic comparison of such forms (note that not all sentence-final particles serve the
conformational function). And second, the complexity of confirmationals is not in itself an
argument against labelling them as such. ough a more accurate descriptor would be to refer
to the confirmational function of ‘eh’ (as opposed to its narrative function).

e syntax of conrmationals 
Assuming that syntax serves to mediate between form and interpretation, it allows us
to capture the relation between the proposition and the speech act. In this way, the
revival of Ross’ (1970) performative hypothesis couched in a framework that assumes
a universal spine presents us with a structural model that is ideal for a cross-linguistic
comparison. As the general function of the grounding phrase has been established, dif-
ferent UoLs from dierent languages can be associated with this level. Across languages,
we now have the tools to establish a link between a proposition and the interlocutors
that contribute their knowledge to the conversation to which that proposition matters.
e newly-established category of conrmationals invites us to further explore
the formal properties of this category. In particular, we are interested in investigat-
ing the relationship between conrmationals and their host clauses. Possible areas of
further research are selectional relations between particle and clause type, potential
restrictions that hold between the grounding expression and a clause-internal con-
stituent, and potential co-occurrence restrictions among dierent types of grounding
elements. Beyond that, we still need to investigate linear ordering restrictions that help
us to further establish the structural properties of conrmationals, prosodic properties
that help us to tease apart the precise contribution of the intonation and the lexicaliza-
tion patterns in order to understand the historical development of conrmationals and
their usage. In brief, the research of conrmationals is still very much in its infancy,
but the potential of breaking grounds in the structural analysis of speech act modica-
tion in this context is real and promising.
References
Avis, Walter. 1957. Canadian English merits a dictionary. Culture XVIII: 245–256.
Bach, Kent & Harnish, Robert. 1979. Linguistic communication and speech acts. Cambridge
MA: e MIT Press.
Beyssade, Claire & Marandin, Jean-Marie. 2006. e speech act assignment problem revisited:
Disentangling speaker’s commitment from speaker’s call on addressee. In Selected papers
of CSSP 2005, Olivier Bonami & Patricia Cabredo Hoerr (eds), 37–68. 〈http://www.cssp.
cnrs.fr/eiss6/index_en.html〉
Blakemore, Diane. 2004. Discourse markers. In e Handbook of Pragmatics, Laurence Horn &
Gregory Ward (eds), 221–240. Oxford: Blackwell.
Cheshire, Jenny & Williams, Ann. 2002. Information structure in male and female adolescent
talk. Journal of English Linguistics 30: 217–238. doi: 10.1177/007242030002008
Chomsky, Noam. 1965. Aspects of the eory of Syntax. Cambridge MA: e MIT Press.
Chomsky, Noam. 1981. Lectures on Government and Binding. Dordrecht: Foris.
Chomsky, Noam. 1986 Barriers. Cambridge MA: e MIT Press.
Chomsky, Noam. 1995. e Minimalist Program. Cambridge MA: e MIT Press.
Cinque, Guglielmo. 1996. Adverbs and Functional Heads. A Cross-linguistic Perspective. Oxford:
O U P.

 Martina Wiltschko & Johannes Heim
Clark, Herbert H. & Brennan, Susan E. 1991. Grounding in communication. Perspectives on
Socially Shared Cognition 13: 127–149. doi: 10.1037/10096-006
Clark, Herbert H. & Schaefer, Edward F. 1989. Contributing to discourse. Cognitive Science
13(2): 259–294. doi: 10.1207/s15516709cog1302_7
Clark, Herbert H. 1992. Arenas of Language Use. Chicago IL: University of Chicago Press.
Columbus, Georgie. 2010. A comparative analysis of invariant tags in three varieties of English.
English World-Wide31(3): 288–310. doi: 10.1075/eww.31.3.03col
Culicover, Peter W. 1992. English tag questions in universal grammar. Lingua 88(3):
193–226. doi: 10.1016/0024-3841(92)90042-H
Déchaine, Rose Marie & Wiltschko, Martina. 2002. Decomposing pronouns. Linguistic Inquiry
33(3): 409–442. doi: 10.1162/002438902760168554
De Wolf, Galean Dodds, Fee, Margery, Gregg, Robert J. & McAlpine, Janice. 2004. e Survey
of Vancouver English: A Sociolinguistic Study of Urban Canadian English (No. 5). Strathy
Language Unit, Queen’s University.
Dombrowski, Ernst & Niebuhr, Oiver. 2005. Acoustic patterns and communicative functions
of phrase-nal f0 rises in German: Activating and restricting contours. Phonetica 62:
176–195. doi: 10.1159/000090097
Dombrowski, Ernst & Niebuhr, Oliver. 2010. Shaping phrase-nal rising intonation in German.
Proceedings of the 5th International Conference on Speech Prosody, Chicago, USA.
Farkas, Donka F. & Bruce, Kim B. 2009. On reacting to assertions and polar questions. Journal
of Semantics 27: 81–118. doi: 10.1093/jos/p010
Fraser, Bruce. 2006. Towards a theory of discourse markers. Approaches to Discourse Particles
1: 189–204.
Ginzburg, Jonathan. 2012. e Interactive Stance. Oxford: OUP.
doi: 10.1093/acprof:oso/9780199697922.001.0001
Ginzburg, Jonathan. 1996. Interrogatives: Questions, facts and dialogue. In e Handbook of
Contemporary Semantic eory, Shalom Lappin (ed.), 385–422. Oxford: Blackwell.
Gibson, Deborah Jean. 1976. A esis on eh. Ph.D. dissertation, University of British Columbia.
Gold, Elaine. 2008. Which eh is the Canadian eh? Toronto Working Papers in Linguistics 27:
73–85.
Gunlogson Christine. 2008. A question of commitment. Belgian Journal of Linguistics 22:
101–136. doi: 10.1075/bjl.22.06gun
Gunlogson, Christine. 2003. True to Form: Rising and Falling Declaratives as Questions in
English. London: Routledge.
Gussenhoven, Carlos. 1984. On the Grammar and Semantics of Sentence Accents. Berlin: Walter
de Gruyter. doi: 10.1515/9783110859263
Haspelmath, Martin. 2007. Pre-established categories don’t exist: Consequences for language
description and typology. Linguistic Typology 11: 119–132. doi: 10.1515/LINGT Y.2007.011
Heim, Johannes, Keupdjio, Hermann, Lam, Zoe Wai-Man, Osa-Gomez, Adriana & Wiltschko,
Martina. To appear in Proceedings of CLA. 〈http://syntaxofspeechacts.linguistics.ubc.ca/
wp-content/uploads/2014/02/heim-et-al-cla-2014-proceedings.pdf〉
Heim, Johannes. 2015. Expertise and common ground. Ms. UBC.
Hinzen, Wolfram. 2014. Recursion and truth. In Recursion: Complexity in Cognition, Margaret
Speas & omas Roeper (eds), 113–137. Dordrecht: Springer.
Hirschberg, Julia & Ward, Gregory. 1995. e interpretation of the high-rise question contour in
English. Journal of Pragmatics 24: 407–412. doi: 10.1016/0378-2166(94)00056-K

e syntax of conrmationals 
Lam, Zoe Wai-Man. 2014. A complex ForceP for speaker- and addressee-oriented discourse
particles in Cantonese. Studies in Chinese Linguistics 35: 61–80.
Kayne, Richard S. 1994. e antisymmetry of syntax. Cambridge MA: e MIT Press.
Malamud, Sophia A. & Stephenson, Tamina C. 2011. ree ways to avoid commitments: Declar-
ative force modiers in the conversational scoreboard. Journal of Semantics 31: 1–37.
Meyerho, Miriam. 1992. ‘We’ve all got to go one day, eh?’: Powerlessness and solidarity in the
functions of a New Zealand tag. In Locating Power: Proceedings of the Second Annual Berke-
ley Women and Language Conference, Kira Hall, Mary Bucholtz & Birch Moonwomon
(eds), Berkeley CA: Berkeley Women and Language Group.
Pierrehumbert, Janet, & Hirschberg, Julia. 1990. e meaning of intonational contours in the
interpretation of discourse. In Intentions in Communication, Philip R. Cohen, Jerry Morgan
& Martha E. Pollack (eds), 271–311. Cambridge MA: e MIT Press.
Pollock, Jean Yves. 1989. Verb movement, universal grammar and the structure of IP. Linguistic
Inquiry 20: 365–425.
Postal, Paul. 1969. Anaphoric islands. Chicago Linguistic Society 5: 205–239.
Ramchand, Gillian & Svenonius, Peter. 2014. Deriving the functional hierarchy. Language
Sciences 46: 152–174. doi: 10.1016/j.langsci.2014.06.013
Ross, John R. 1970. On declarative sentences.In Readings in English Transformational Grammar,
Roderick A. Jacobs & Peter S. Rosenbaum (eds),222–272. Waltham MA: Ginn & Co.
Sailor, Craig. 2009. Tagged for Deletion: A Typological Approach to VP Ellipsis in Tag Ques-
tions. MA thesis, UCLA.
Speas, Penny & Tenny, Carol. 2003. Congurational properties of point of view roles. Asymme-
try in Grammar 1: 315–345. doi: 10.1075/la.57.15spe
Stalnaker, Robert 1978. Assertion. In Syntax and Semantics 9: Pragmatics, Peter Cole (ed.), 315–
332. New York NY: Academic Press.
Stubbe, Maria & Holmes, Janet. 1995. You know, eh and other exasperating ‘expressions’: An anal-
ysis of social and stylistic variation in the use of pragmatic devices in a sample of New Zealand
English. Language and Communication 15: 63–88. doi: 10.1016/0271-5309(94)00016-6
oma, Sonja. Forthcoming. Discourse particles in Bavarian German. Ph.D. dissertation, UBC.
Trinh, Tue & Crnic, Luka. 2011. On the rise and fall of declaratives. Proceedings of Sinn and
Bedeutung (SuB) 15: 645–660.
Truckenbrodt, Harald. 2013. Satztyp, Prosodie und Intonation. In Satztypen des Deutschen, Jörg
Meibauer, Markus Steinbach & Hans Altmann (eds), 570–601. Berlin: De Gruyter.
Truckenbrodt, Harald. 2012. Semantics of intonation. In Semantics: An International Handbook
of Natural Language Meaning, Vol. 3, Claudia Maienborn, Klaus von Heusinger & Paul
Portner (eds), 2039–2969. Berlin: De Gruyter.
Wiltschko, Martina. Forthcoming. Ergative constellations in the structure of speech acts.
Wiltschko, Martina. 2014. e Universal Structure of Categories: Towards a Formal Typology.
Cambridge: CUP. doi: 10.1017/CBO9781139833899
Woods, Howard B. 1980. e Ottawa Survey of Canadian English. Kingston: Strathy Language
Unit, Queen’s University.