

Dialogue and Discourse 4(2) (2013) 215-248 doi: 10.5087/dad.2013.210

Focus Triggers and Focus Types
from a Corpus Perspective

Arndt Riester ARNDT.RIESTER@IMS.UNI-STUTTGART.DE

Institut für Maschinelle Sprachverarbeitung
Universität Stuttgart
Pfaffenwaldring 5 B
D-70569 Stuttgart

Stefan Baumann STEFAN.BAUMANN@UNI-KOELN.DE

Institut für Linguistik – Phonetik
Universität zu Köln
Herbert-Lewin-Str. 6
D-50931 Köln

Editors: Stefanie Dipper, Heike Zinsmeister, Bonnie Webber

Abstract
The article discusses several issues relevant for the annotation of written and spoken corpus data
with information structure. We discuss ways to identify focus top-down (via Questions under Dis-
cussion) or bottom-up (starting from pitch accents). We introduce a two-dimensional labelling
scheme for information status and propose a way to distinguish between contrastive and non-
contrastive information. Moreover, we take side in a current debate, claiming that focus is triggered
by two sources: newness and elicited alternatives (contrast). This may lead to a high number of
semantic-pragmatic foci in a single sentence. In each prosodic phrase there can be one primary
focus (marked by a nuclear pitch accent) and several secondary foci (marked by weaker prosodic
prominence). Second occurrence focus is one instance of secondary focus.
Keywords: alternatives, contrast, corpus annotation, focus, givenness, information status, infor-
mation structure, prosody, Question under Discussion, secondary prominence

1. Focus in English and German: basic prosodic assumptions

The purpose of this article1 is to discuss the linguistic notion of focus – which has been intensively
investigated in the theoretical literature2 – in the light of corpus data. For this, it is necessary that
we take a stand on a number of issues about which the current theoretical debate is undecided,
for instance, whether one or several types of focussing should be assumed and how to delineate
contrastive from non-contrastive focus. We are convinced that precision with regard to conceptual

1. This is a completely revised and extended version of an article which appeared in 2011 under the title Information
Structure Annotation and Secondary Accents in the volume Beyond Semantics: Corpus-based Investigations of Prag-
matic and Discourse Phenomena in the series Bochumer Linguistische Arbeitsberichte, 3:111-127, ed. by Stefanie
Dipper and Heike Zinsmeister.

2. A few references: Halliday (1967); Chomsky (1971); Jackendoff (1972); Rooth (1992); Krifka (1992); Selkirk
(1995); Roberts (1996); Hajičová et al. (1998); É. Kiss (1998); Schwarzschild (1999); Büring (2003); Krifka (2008);
Beaver and Clark (2008)

c©2013 Arndt Riester and Stefan Baumann Submitted 04/2012; Accepted 02/2013; Published online 08/2013


RIESTER AND BAUMANN

issues is an indispensable prerequisite for corpus annotation of information structure. We will fur-
thermore argue in favour of a distinction between what we call primary focus and communicatively
less important kinds of secondary focus.

The first assumption which we shall make in this paper is that in English or German,3 and
presumably in other languages, there can be several foci in the same prosodic phrase but only one
of them can be marked by a nuclear pitch accent. We call this one the primary focus.

Definition 1 (Primary focus) A primary focus is a focus constituent which is marked by means of
a nuclear pitch accent.

Note that Definition 1 neither says what a focus is in general, nor why and when it may receive a
nuclear pitch accent, nor whether it has to be marked at all. This requires some clarification: in line
with most current theoretical treatments, we take focus to be a semantic-pragmatically – and not
prosodically or morpho-syntactically – defined notion. To investigate what the precise semantic-
pragmatic factors are which lead to focussing is one of the main subjects of the present article, and
we will soon turn to this issue (Section 2). For the beginning, let us settle for an intuitive focus
notion which says that focus indicates new or important parts of an utterance.

It seems as if the literature contains many examples in which there are two primary foci in the
same clause. Examples are given in (1) to (4), with relevant pitch accents marked by capital letters.

(1) Even JOHNF1 drank only WAterF2. (Krifka, 1992)

(2) A: Did Carl sue the company, or did the company sue Carl? (Büring, 2003)
B: CARLF1 sued the COMpanyF2.

(3) A: Who ate what? (Roberts, 1996; Büring, 2003)
B: FREDF1/CT ate the BEANSF2.

(4) [In MYF1 opinion]T , JOHNF2 stole the cookies. (Krifka, 2008)

The phenomenon in (1) has been called multiple focus. It is characterised by the presence of two
focus-sensitive particles, even and only, within the same clause. The focused constituents are in-
dexed by F1 and F2. In (2B), the two foci form a complex focus. They highlight a choice from
among two pairs of alternatives provided by question (2A). In addition to that, the (pairs of) foci
may also be called contrastive foci, due to the presence of overt alternatives. (3B) is a protoypical
example involving a contrastive topic (CT), which is assumed to share basic properties of focus, in
particular, the ability to highlight the availability of alternatives. In opposition to (2B), it addition-
ally signals that the question is not answered completely, and that people, rather than dishes, are the
“sortal key” (Büring, 2003: 530) along which (3A) is worked off. In using the contrastive topic, the
speaker of (3B) announces to continue and tell us what other persons ate. The first focus in (4) is
used within a frame-setting topic (T), which likewise divides the interpretational space into alterna-
tive partitions. Ignoring the fact that the pitch accents occurring on the two respective focal elements
in (1)-(4) might be of a different type, e.g. rising vs. falling, there is no clear observable difference
with respect to their prosodic strength or prominence.4 This holds true for all four examples.

3. Throughout the paper, we shall make the assumption that the prosodic marking of information structure is very similar
in English and German. This seems by and large justified with regard to pitch accent placement, but perhaps less so
with regard to some details of pitch accent types.

4. Note that Krifka (2008) does claim that the first accent in some cases of multiple focus is stronger than the second
one, and that this distinguishes multiple focus from complex focus. If this can be shown to be true in general, then (1)
is not an example with two primary foci.

216


FOCUS TRIGGERS AND FOCUS TYPES

All we have said so far is compatible with the requirement that there can be only one primary
focus per prosodic phrase, since we may assume that, in all of the above examples, the two respective
foci are confined to their own intermediate phrase (ip) – or even intonation phrase (IP) – defined by
Beckman et al. (2005) as the domain for a nuclear accent. In other words, the prosodic structure of
all sentences in (1) to (4) will look like (5).

(5) {{ (PN1 . . . PNn) N (pn) -}ip { (PN1 . . . PNm) N (pn) -}ip %}IP

Here, (PN) stands for an optional prenuclear accent, several of which may occur, N is the nuclear
pitch accent, and (pn) indicates an optional postnuclear prominence. The latter is not supposed to
carry the rank of a pitch accent and is signaled by means of increased duration and intensity but only
very little pitch movement. It is sometimes called phrase accent (Grice et al., 2000), or postlexical
stress (Beckman, 1986). Furthermore, ‘%’ indicates an intonation phrase break, and ‘-’ stands for
an intermediate phrase break.

However, it is possible and, in fact, widespread to have several (semantic-pragmatic) foci in
the same prosodic phrase. One particular case in point is second occurrence focus (SOF), which
has received a considerable amount of attention in recent years (Rooth, 1996; Partee, 1999; Bartels,
2004; Büring, 2008, ms.; Rooth, 2010; Beaver and Velleman, 2011), though not explicitly from the
perspective of prosodic phrasing. A well-known example from Partee is shown in (6), in which we
would assume that the first half of (6b) (until vegetables) can be realised as a single prosodic phrase.

(6) a. Everyone knew that Mary only eats VEgetablesF1.
b. If
{

even PAULF2

N
knew that Mary only eats VEgetablesSOF ,

pn -}
then he should have suggested a different restaurant.

Several experiments on both English (Rooth, 1996; Beaver et al., 2007) and German (Féry and
Ishihara, 2009; Baumann et al., 2010) have revealed that (semantically) focussed5 but given expres-
sions, like vegetables in (6b), can be realised with some sort of postnuclear prominence (pn), thus
differing from non-focal, given expressions, which are claimed to be completely deaccented. We
will return to the issue of second occurrence focus in Section 5. At this point, we merely note that
SOF is one instance from the class of secondary foci defined below.

Definition 2 (Secondary focus) A secondary focus is a focus constituent which is not marked by
means of a nuclear pitch accent but by some pre- oder postnuclear prominence.

Obviously, our definitions of primary and secondary focus are not fully pragmatic in nature since
they make reference to prosodic concepts. In the long run, one would like to replace Definitions 1
and 2 by purely pragmatic definitions which predict when exactly a focus receives a nuclear pitch
accent. We think that, seen from the perspective of corpus analysis, it is still too early for that.6

We would like to draw the reader’s attention to a related phenomenon which has received much
less attention than the role of postnuclear prominences in marking second occurrence focus: the

5. In both (6a) and (6b), vegetables is associated with the focus adverb only.
6. The question of predicting which focus receives more prosodic prominence is addressed in Selkirk (2008); Büring

(2008, ms.); Rooth (2010) or Beaver and Velleman (2011). Büring gives a purely pragmatic definition of primary
focus in terms of the sizes of focus domains. Of two foci, the one “whose domain contains the domain of the other”
is the primary one. Büring predicts that this focus will then receive the nuclear pitch accent. However, a definition
like that presupposes a solution to the problem of identifying focus domains.

217


RIESTER AND BAUMANN

information structural contribution of prenuclear accents. While it is easy to ignore these accents
when discussing “the” focus of constructed examples, they represent an important and ubiquitous
element of the prosody of almost every spoken utterance. Consider the phrase shown in Figure 1,
taken from the DIRNDL corpus of German radio news (Eckart et al., 2012), which is annotated for
pitch accents and prosodic boundaries following GToBI(S) (Mayer, 1995). The entire sentence is
given in (7).7

Figure 1: PRAAT (Boersma and Weenink, 1996) screenshot; DIRNDL s1222, 26-03-2007, 06:00,
7’12”: in the former slave trading post Elmina

(7) In
{{

GHAna
L*H -}

fand
{

ein FESTakt
L*H -}

im
{

Ehemaligen
H*

SKLAven-Handelsposten
L*H

ElMIna
H*L

statt.
%}}

In Ghana, a ceremonial act took place in the former slave trading post Elmina.

In the section shown in Figure 1 there are two prenuclear accents, H* on the adjective ehemaligen
(‘former’) and L*H on Sklaven-Handelsposten (‘slave trading post’). It is accents like these which
are usually ignored in formal discussions of focus. If they are mentioned at all, they are often
assumed to be optional or, at least, not meaning-related.8 We would like to raise doubts about the
optionality of prenuclear accents. In (7), it doesn’t seem possible to omit any of the prenuclear
accents. It is unclear, so far, whether this is because they play a role in indicating information
structure – perhaps marking a focus of their own – or for purely rhythmical reasons. It seems that
in order to settle this issue more theoretical background is necessary.

Analysing corpus data with regard to their focal properties raises a number of problems. We
may by now have sophisticated theories concerning the pragmatics of focus, as well as some good
experimental evidence about its prosodic marking. Nevertheless, all that knowledge still seems quite

7. Nuclear pitch accents are indicated in boldface, prenuclear accents in small capitals.
8. Büring (2007) calls these accents “ornamental”.

218


FOCUS TRIGGERS AND FOCUS TYPES

insufficient to analyse even a relatively simple and by no means untypical example like (7). In the
next section, we will try to sketch the problem from two different perspectives (top-down, i.e. from
the perspective of questions under discussion, and bottom-up, i.e. starting out from pitch accents).

In Section 3, we present our RefLex annotation scheme for given, accessible and new informa-
tion (information status), which accounts for a less controversial, though substantial, share of the
information structure of linguistic data. Section 4 discusses the distinction between non-contrastive
and contrastive focus and presents a method of how to spot contrastive (alternative-eliciting) features
in corpus data. In Section 5, we return to the issue of primary and secondary foci.

2. Determination of focus: top-down vs. bottom-up

Taking another look at Example (7), we may find it surprisingly difficult to tell how many foci it
contains. The simplest choice is to say that the entire sentence represents a single broad focus. An
obvious justification is that (7) contains only discourse-new information, and therefore serves to
answer the Big Question (Roberts, 1996) What is the way things are?, or simpler What happened?

For several reasons this cannot be a satisfactory solution. As it is known from the work of
Groenendijk and Stokhof (1984) and Roberts (1996), assertions need not be complete answers; in
fact, an unrestricted question like What happened? is never answered completely. But if an answer
is only partial with regard to the Big Question, as is the case with (7), it may carry additional
information about the structure of the conversation. According to Roberts, the prosody of assertions
reflects the (immediate) Question under Discussion (QUD), which is more specific than the Big
Question and usually implicit. Büring (2003) develops this idea further, elaborating on Example (8).

(8) FREDCT ate the BEANSF .

Consider (8) as a discourse-initial assertion. The contrastive topic in (8), marked by a rising accent
on FRED, signals the discourse strategy of the speaker to answer the question Who ate what? by
first providing an answer to the subquestion What did Fred eat? Our German corpus example (7)
shows – perhaps incidentally9 – a very similar pattern at the beginning. It is in accordance with our
interpretation of the sentence to assume that (7) has an analogous structure to (8), i.e. a contrastive
topic on the phrase in Ghana; in other words, (7) seems to signal a strategy – technically a stack
– consisting of the two implicit questions (i) What happened where? and (ii) What happened in
Ghana? We infer that the speaker intends to continue, at some stage, with news about other coun-
tries.10 If in Ghana is indeed a contrastive topic (and therefore, a special kind of focus), and if the
remainder of the clause is the main focus which actually answers question (ii), then the total number
of independent focal constituents in the sentence rises to at least two.

Determining the focal structure of some utterance in the way described – by reasoning about
implicitly asked questions – is what we might call a top-down process of focus identification: first,
determine what is being asked for, or under discussion, and what is the discourse strategy to which

9. Note that we do not propose to rely on the actually found pitch accents to guide us in the annotation process. We are
not claiming that a rising pitch accent necessarily signals a contrastive topic. Clearly, contrastive topics need to be
identified via pragmatic reasoning. What we hope to eventually find, however, is statistical evidence in the prosodic
domain that supports the chosen pragmatic analysis.

10. It must be noted though that intentions like this are often not more than vague promises, and that strategies may fall
prey to memory decay. This goes against Roberts’s assumption that questions have to remain on the QUD stack until
resolved or determined unanswerable. Another conceivable situation is that a person reading a text simply has a false
anticipation of what is to follow.

219


RIESTER AND BAUMANN

it belongs; second, identify the phrases which provide an answer to the current Question under Dis-
cussion; and third, predict or observe how these phrases are marked prosodically (and/or morpho-
syntactically).11 In other words, the marking itself should not be used to identify the focus or the
contrastive topic. We do not claim that we already possess a universal procedure which lives up to
these standards. Nevertheless, the top-down approach works straightforwardly in simple cases like
(9), in which the accented word JOHN is both the focus and the answer to the overt QUD.

(9) A: Who spilled the wine?
B: JOHNF spilled it.

The procedure can also be applied to more complex cases involving nested phrases and disjoint foci.
(10) is a German example from Höhle (1982), discussed in Gussenhoven (1999).

(10) A: Was hat das Kind erlebt?
What happened to the child?

B: KARLF hat dem Kind [einen FÜLler geschenkt]F .
Karl gave the child a fountain pen.

The phrase which provides the answer to the Question under Discussion – i.e. the focus – is split in
two parts, Karl and einen Füller geschenkt. Gussenhoven (1983, 1992, 1999) offers a general rule
for predicting the accent pattern of a focus that has been pragmatically determined, the so-called
Sentence Accent Assignment Rule (SAAR), which we reproduce here in a simplified form:

1. Place a pitch accent on every content word in focus.

2. Then, deaccent every focussed predicate which is adjacent to an accented argument.

The SAAR correctly predicts the pitch accents on Karl and on Füller (‘fountain pen’), the lack
of pitch accent on the unfocussed (backgrounded) word Kind (‘child’) and the deaccenting of the
predicate geschenkt (‘given as a present’).12 As for Example (7), we already assumed that the focus
encompasses the phrase in (11).

(11) [fand ein FESTakt im Ehemaligen SKLAvenhandelsposten ElMIna statt]F?

Again, the SAAR correctly accounts for the deaccentuation of fand. . . statt (‘took place’) which is
adjacent to its argument ein Festakt (‘a ceremonial act’). The remaining content words receive a
pitch accent because they do not stand in any predicate-argument relation.

What Gussenhoven’s rule cannot accomplish is to explain the intricate patterns of prosodic
phrasing. Neither does it tell us whether the focus in (11) is a single one or whether it actually
consists of several smaller foci. In order to answer this question, we return to Roberts’s (1996)
theory of Questions under Discussion. When answering question (12a) by means of (7), we are
actually behaving in an over-informative way. In some sense, it would have been enough to use the
simpler answer given in (12b).

11. As is well-known, some languages do not mark focus by means of prosodic prominence. Other devices found cross-
linguistically include moving the focal constituent to a particular syntactic position in the sentence, or attaching a
special focus morpheme.

12. Deaccentuation of a predicate with an accented argument is explained in a similar fashion within Selkirk’s
(1984,1995) focus projection framework.

220


FOCUS TRIGGERS AND FOCUS TYPES

(12) a. What happened in Ghana?
b. [In GHAna]CT [fand ein FESTakt statt]F .

(13) im ehemaligen Sklavenhandelsposten Elmina

In fact, the over-informative prepositional phrase (13) is a non-restrictive modifier and therefore
belongs to the class of so-called supplemental expressions, which Potts (2005: 6) defines as con-
ventional implicatures or as not-at-issue. Simons et al. (2010) define at-issueness by saying that a
proposition p is at-issue if and only if the question whether p is true or not is relevant to the QUD.
In our case, the proposition expressed by the PP, namely The location of the ceremonial act is in
Elmina, is not an answer to (12a), but rather to the supplemental question in (14).

(14) Where did the ceremonial act take place?

In fact, we can iterate this process once more by stating that the information that Elmina is a former
slave trading post – the meaning expressed by the non-restrictive modifier of Elmina – is not an
answer to (14) but to (15).

(15) What is Elmina?

This, admittedly complex, line of reasoning brings us to the following tentative conclusion: it is
very likely that our corpus sentence actually consist of four independent pragmatic foci (including
one contrastive topic), as shown in (16).

(16) [In
{{

GHAna]CT

L*H -}
[fand
{

ein FESTakt
L*H -}

[im
{

[Ehemaligen
H*

SKLAvenhandelsposten]F3

L*H
ElMIna]F2

H*L
statt]F1

%}}

We can furthermore state that CT, F1 and F2 are marked by nuclear pitch accents – in the terminol-
ogy defined in the previous section, they represent primary foci (see Definition 1 above). F3 is only
marked by (two) prenuclear accents; we therefore call it a secondary focus (Definition 2).

The top-down approach, as we have sketched it, has two important advantages: it is cross-
linguistically applicable, even to languages which do not mark information structure prosodically,
and it does not confuse form (pitch accent) and meaning (focus). It allows us to use the following
language-independent pragmatic focus definition.

Definition 3 (Focus) A focus is either an answer to the immediate (explicit or implicit) Question
under Discussion (at-issue focus) or to a supplemental question (not-at-issue focus).13

The problem is that the identification of QUDs in corpus data is still in its infancy, and the current
paper does not purport to propose a general procedure for it, although this is what clearly should be
envisaged in the future.

A competitor to the top-down approach, which is intrinsic to certain well-established theories
of focus, including Rooth (1992) and Selkirk (1995), is the assumption that pitch accents neces-
sarily indicate a focus.14 We call this the bottom-up approach to focus identification. Obviously,

13. In (16), we may call F1 an at-issue focus, and F2/F3 not-at-issue foci.
14. Quote Selkirk (1995: 555): “The Basic Focus Rule states that the assignment of a pitch accent to a word entails the

F-marking of the word[.]”

221


RIESTER AND BAUMANN

the bottom-up approach is not linguistically universal, and even for English or German we cannot
be entirely sure whether the assumption is empirically sound, especially if secondary prominence
comes into play. Note that, under the bottom-up approach, Example (16) would not contain four
focal expressions but five, since the prenuclear pitch accent on ehemaligen would count as an au-
tonomous marker of focus.15 Nevertheless, at the current stage, we cannot entirely do without the
bottom-up approach as long as the top-down approach is not fully worked out. This will become
clear in Section 4, when we will discuss whether an observed nuclear pitch accent is used to express
contrastive focus or not.

In the following sections, we will discuss those aspects of information structure which we think
can be annotated without too much controversy. This will enable us to create pragmatically en-
hanced corpus resources, which may subsequently serve as a basis for empirical investigation, even
though we do not yet caim to annotate all aspects of focus. Other proposals of annotating infor-
mation structure involve, for instance, Hajičová et al. (2000), Paggio (2006), Götze et al. (2007),
or Cook and Bildhauer (2013). In Paggio (2006), focus is defined, following Lambrecht (1994),
as “non-presupposed” information. Further rules given to the annotators include, for instance, the
requirement to identify at least one focus per sentence and at least one “main accent” per focus (for
Danish data). In Götze at al. (2007: 170), the definition for focus is “[that] part of an expression
which provides the most relevant information in a particular context [. . . ]”. Subcategories for new
information focus and contrastive focus are defined. We believe that these proposals capture some
important aspects of focus in corpus data but are less complete and less detailed than the picture that
we are developing in this article.

3. Annotating given-new information: the RefLex scheme

In accordance with Definition 3, we adopt a concept of focus in terms of answers to explicit or
implicit questions. This is not an uncontroversial decision. Often, what counts as the answer to an
implicit question can simply be called new information. However, not everybody seems to agree
that new information which is not otherwise marked (e.g. occurring in a contrastive constellation or
being associated with a focus-sensitive particle) deserves to be called focus. Consider the following
quote by Elisabeth Selkirk:

[. . . ] I am using the simple term “focus” to refer to “contrastive focus” [. . . ], as
involving Roothian alternatives. This should not be confused with the use of the term
“focus” to indicate newness in the discourse, a use which this paper argues should not
be made. (Selkirk, 2008: Section 4, Footnote 9)

We agree with Selkirk’s view that there is a linguistically relevant distinction between contrastive
focus and (purely) new information in terms of their pragmatic meaning (more on this in Section 4).
Besides, as Selkirk and numerous others have shown in their work, contrastive focus tends to re-
ceive higher prosodic prominence and sometimes uses syntactically less canonical structures than
new information, cf. Repp (2010) and references therein. However, we object against Selkirk’s con-
clusion that newness should not be called focus, since this is in conflict with the QUD approach.

15. Leaving out the H* accent on ehemaligen would presumably violate the givenness principle (Schwarzschild, 1999),
which says that an expression must either be given or F-marked. Focus projection rules (Selkirk, 1995) do not allow
an F-marker to project from Sklavenhandelsposten since it is not an argument of ehemaligen. Therefore, the adjective
needs its own pitch accent.

222


FOCUS TRIGGERS AND FOCUS TYPES

In Section 4, we will account for the distinction between contrastive and non-contrastive (novelty)
focus in a way that is compatible with both QUDs and Rooth’s Alternative Semantics. The common
denominator of new and contrastive constituents is their ability to answer questions – to reduce the
number of possibilities the world might be like – and this is what should be taken as the defining
characteristic of focus in general.

In Section 2, we have given a sketch of the top-down analysis of the focal structure of natural
discourse, not yet ready for the use in linguistic annotation. In this section, as a first step towards the
annotation of focus we are going to provide a scheme for annotating (different types of) given and
new (as well as accessible) information, what is also called information status, following the system
of Baumann and Riester (2012). The scheme combines earlier accounts of information status in
the aftermath of Prince (1981, 1992) – notably Chafe (1994); Lambrecht (1994); Eckert and Strube
(2000); Nissim et al. (2004); Götze et al. (2007); Riester et al. (2010) – with the givenness theory by
Schwarzschild (1999). To classify the constituents of a sentence into given and non-given ones is a
first move towards identifying its background-focus structure. Schwarzschild’s givenness principle
says that non-given constituents must be “F-marked” (and are therefore, in some sense, focal).16

The procedure described in this section will not tell us how many different foci a sentence contains.
But it already provides us with a fine-grained analysis with a view to its prosodic correlates; see
Baumann and Riester (in press).

The system distinguishes between a referential level and a lexical level (and is therefore called
the RefLex scheme). We will clarify why it is desirable to use such a fine-grained system rather than
just distinguishing between “given” and “new” constituents. Note well that we are not claiming
that the annotation labels presented below represent syntactic features of some kind, in the way
as, for instance, Selkirk (2008) treats her F and G markings. We will make no predictions as
regards the precise functioning of the syntax-phonology interface. The category descriptions below
are deliberately kept short, since we have introduced them in great detail in Baumann and Riester
(2012).

3.1 R-GIVEN and L-GIVEN

Givenness, loosely following Schwarzschild (1999: 151), can be interpreted as either synonymy /
hyponymy of lexemes (and the concepts they express), or as identity between referring expressions.
Likewise, Halliday and Hasan (1976: 288) distinguish between lexical cohesion and various refer-
ential relations. We call the two notions L-GIVENNESS and R-GIVENNESS, respectively. Interesting
constellations can be observed if the two notions are applied simultaneously, as shown below.

By use of the R-categories it is possible to classify referential determiner phrases and preposi-
tional phrases occurring in natural discourse; by use of the L-categories we can classify the infor-
mation status of content words and non-referential phrases. R-labels apply at the DP or PP level.
For instance, in Examples (17), (18) and (20) we find various kinds of coreferential expressions.
Lexical givenness, on the other hand, applies in (18) and (20) on the repeated words, and in (19) on
the hypernym man.

(17)
A colleague came in. The idiot dropped a vase.

R-GIVEN

16. A warning: Beaver and Clark (2008: 15) notice that not all F-markers used in the literature have the same function,
e.g. F-markers used in Rooth (1992) and in Selkirk (1995) are used for different purposes.

223


RIESTER AND BAUMANN

(18)
A student came in. Another student greeted him.

L-GIVEN R-GIVEN

(19)
A policeman came in. Another man left.

L-GIVEN

(20)
A man came in. The man coughed.

L-GIVEN

R-GIVEN

The most important take-home message is that neither is referential givenness a prerequisite for
lexical givenness, as shown in (17), nor vice versa, see (18) and (19), although the two sometimes
combine, as in (20).

3.2 R-NEW, L-NEW, R-UNUSED

Novelty is, on most treatments of information structure and discussions of the given/new distinction,
understood as “novelty in the discourse”. Remarkably however, Prince (1992) additionally distin-
guishes between discourse novelty and hearer novelty, the latter representing a stronger notion since
unmentioned (i.e. discourse-new) entities may nevertheless be familiar to the addressee (i.e. hearer-
old). In her earlier paper, Prince (1981) uses the labels unused (discourse-new, hearer-old) and
brand-new (discourse-new, hearer-new) for the same opposition. The labels R-NEW and R-UNUSED

that are employed on our account are defined in a slightly different way: both describe discourse-
new referential expressions but, while R-NEW is reserved for indefinites, R-UNUSED stands for
uniquely identifiable, definite, but not necessarily known, entities used on the first occasion in a
text. This decision, on the one hand, does justice to the long-standing semantic tradition to keep
indefinites and definites (for instance, proper names) apart, and, on the other hand, accounts for the
difficulty to decide with certainty whether, for instance, a named entity is hearer-known or not, cf.
Riester et al. (2010).

Independently of what has just been said, it is furthermore possible to separately describe the
discourse novelty of lexemes (L-NEW) and of the discourse referents (R-NEW, R-UNUSED) which
they introduce. Examples of the three categories in combination are given in (21) to (23).

(21)
A man came in. Another man left.

L-NEW L-GIVEN

R-NEW R-NEW

(22)
George came in. Mary likes George.

L-NEW L-NEW L-GIVEN

R-UNUSED R-UNUSED R-GIVEN

(23)

The man who stole my wallet yesterday is very tall.
L-NEW L-NEW

R-UNUSED

R-UNUSED

The complex subject phrase in Example (23) shows that information status needs to be assigned
recursively. This is an issue which is of particular relevance for the language of news, which contains
many expressions with several embeddings, cf. Riester et al. (2010).

224


FOCUS TRIGGERS AND FOCUS TYPES

3.3 R-BRIDGING, L-ACCESSIBLE

Prince (1981) and also Chafe (1994) have pointed out that it is desirable to not only distinguish
between given and new information but to take into account at least a third, intermediate, class:
expressions which have not been mentioned explicitly but are inferrable from material in the dis-
course. Chafe (1994) uses the term accessible for such information but does not distinguish between
different levels, as we would like to do. As far as discourse referents are concerned, a closely related
phenomenon has been discussed under the notion of bridging or associative anaphora (Clark, 1977;
Asher and Lascarides, 1998; Löbner, 1998; Poesio and Vieira, 1998), shown in Example (24).

(24)
Bill discovered a romantic house. The door was open.

L-NEW L-NEW L-ACCESSIBLE

R-UNUSED R-NEW R-BRIDGING

The label L-ACCESSIBLE is defined for words which are hyponyms or meronyms (part expressions)
of other words in the recent discourse context.17 The label R-BRIDGING, on the other hand, is
defined quite differently as a definite expression whose licensing depends on a previously introduced
scenario or frame. So, while in (24), house and door stand in a whole-part relation (door is lexically
accessible), no such relation exists between murdered and harpoon in (25). Since the harpoon is an
unusual murder instrument, it is labeled L-NEW. Nevertheless, we would still like to say that this is
a case of bridging, since the second sentence could not be uttered felicitously at the beginning of a
discourse.

(25)
John was murdered yesterday. The harpoon was lying nearby.

L-NEW L-NEW

R-UNUSED R-BRIDGING

Other than in the case of R-UNUSED expressions, the interpretation of items labeled R-BRIDGING

is context-dependent. In contrast to the label R-GIVEN, R-BRIDGING implies non-coreference. In-
definites never receive the label R-BRIDGING in the present system. In (26), lexical accessibility
combines with referential novelty.18

(26)
John lives in Italy and is married to a Neapolitan.

L-NEW L-NEW L-ACCESSIBLE

R-UNUSED R-UNUSED R-NEW

3.4 R-GENERIC

Definite or indefinite expressions which refer to a kind, see (27) and (28), receive the label
R-GENERIC.

(27)
The lion has a mane.

L-NEW L-NEW

R-GENERIC R-GENERIC

17. We assume a somewhat arbitrary window of five sentences. A more precise range in which word associations still
play a role needs to be determined experimentally.

18. Arguably in (26), in addition to the L-label on the word Italy, the DP Italy and the PP in Italy both should receive
separate R-UNUSED labels, since the former refers to the country itself and the latter to a location.

225


RIESTER AND BAUMANN

(28)
Mary likes vegetables. John likes vegetables, too.

L-NEW L-NEW L-NEW L-GIVEN

R-UNUSED R-GENERIC R-UNUSED R-GENERIC

As can be seen in (28), we do not treat the repeated mention of a generic expression (vegeta-
bles) as a case of coreference (R-GIVEN) but merely as repetition of the same concept (L-GIVEN,
R-GENERIC).

3.5 Overview and annotation of higher syntactic constituents

R-Level L-Level
Units: DP, PP, that-CP Units: A(P), Adv(P), N(P), V(P), S

Label Description Label Description
R-GIVEN coreferential L-GIVEN word identity /

anaphor synonym / hypernym /
holonym / superset

R-BRIDGING non-coreferential L-ACCESSIBLE hyponym / meronym /
context-dependent subset / co-hyponym /
expression related

R-UNUSED definite L-NEW unrelated expression
discourse-new (within last five
expression clauses)

R-NEW specific indefinite
R-GENERIC generic definite

or indefinite
OTHER e.g. cataphors

Table 1: Overview of basic RefLex scheme

Table 1 contains the most important labels of the RefLex annotation scheme. For a more detailed
list of labels consult Baumann and Riester (2012). In the following, we will turn to a number of
practical issues which arise when we apply the annotation scheme to corpus data. As we said at the
beginning of this section, we want to use our annotation system in order to arrive at a comprehen-
sive identification of the given and non-given parts of linguistic data since we consider the latter as
indicating focal material. In order to achieve this goal we cannot confine our analysis to referring
expressions, as it has been done in e.g. Prince (1981); Nissim et al. (2004); Götze et al. (2007);
Riester et al. (2010), but need to extend the annotations to other content expressions like adjectives,
verbs and adverbs as well as their syntactic projections. The account builds upon Schwarzschild’s
(1999) theory of focus and givenness. Of course, the question of what counts as a unit for anno-
tation is influenced by the choice of syntactic theory which the analysis is based on. A principled
distinction can be made between expressions which refer to some entity, like an individual, a place,

226


FOCUS TRIGGERS AND FOCUS TYPES

S [L-level]hhhhhhhhh
(((((((((

DP [R-level]
PPPPP

�����
D

The

NP [L-level]
PPPP

����
AP [L-level]

A [L-level]

tall

NP [L-level]

N [L-level]

man

VP [L-level]
aaaa

!!!!
V [L-level]

arrived

PP [R-level]
H
HH

�
��

P

in

DP [R-level]

NP [L-level]

N [L-level]

Tuscany

Figure 2: Basic target units for RefLex annotations

a fact etc. (DP, PP, that-CP),19 and expressions which denote a property / set of entities (NP, AP, VP,
AdvP, S) or a relation.20

As is shown in Figure 2, the former are assigned R-labels, the latter
L-labels. In practice, some of these labels will be redundant and can be left out. For instance,
one of the L-labels at either the N or the NP level above man can of course be dropped, since they
are identical. What we are proposing amounts to a practical explication – and further development –
of the approach taken by Schwarzschild (1999: 151), who distinguishes between categories of type e
(R-level) and of type 〈α, β〉 (L-level). Our definition of the L-level, however, is much simpler than
Schwarzschild’s since we completely abandon his notion of Existential F-closure. However, we
make use of his idea to generalise lexical relations to a notion of entailment.21

As we said, in corpus annotation practice, the system will have to be adapted to various con-
straining factors, such as the properties of the chosen parser with its specific syntactic tagset, as well
as features of the annotation tool. Figure 3 shows a sample annotation of a German sentence from
the DIRNDL corpus (Eckart et al., 2012).22

19. One reviewer criticised that PPs should count as properties rather than individual type entities, which is a common
assumption in semantics. What we are after, however, is the referent of an argument, which often comes in the form
of a PP. Sometimes PPs refer to a place or time; sometimes the preposition is subcategorised by the predicate and
semantically empty; sometimes, in German, preposition and determiner are amalgamated (im – ‘in the’, zum – ‘to
the’ etc.). In all those cases the simplest practical choice is to assign referential information status to the PP. A second
criticism pertained to allegedly non-referential quantifiers like few people, every dog. In corpus data, however, such
expressions almost always introduce or refer back to group entities, analogously to indefinites and definites.

20. We assume the DP hypothesis (Abney, 1987). Accordingly, we take NPs to denote properties, i.e. sets of individuals,
whereas DPs denote (or refer to) a single individual or group entity. We abstain from the debate whether a sentence
should be analysed as an IP (inflection phrase), as a CP (as in the German example shown in Fig. 3) or as a projection
of tense, aspect or voice, cf. Adger (2003).

21. According to this approach, the previous mention of chihuahua entails the successively mentioned hypernym dog,
as well as a successive mention of small dog, cf. Baumann and Riester (2012: 133ff.)

22. The sentence was parsed using XLE (Crouch et al., 1993-2011, ms.) and the German Lexical Functional Grammar
implementation by Rohrer and Forst (2006), and converted to be used with the SALTO tool (Burchardt et al., 2006),
which produces output in TIGER/SALSA-XML. In the rest of the paper, we shall abstract over such individual
choices, since it is our goal to provide the general annotation procedure and not one that is tied to a specific annotation
tool, format or syntactic theory.

227


RIESTER AND BAUMANN

Figure 3: Sentence annotated in SALTO, DIRNDL s165, 25-03-2007, 5:00: A strong earthquake
has hit central Japan.

In the following, we will briefly show how the extended annotations can be applied to news data.
Example (29) has been slightly adapted for ease of demonstration. The analysis is shown in (30) to
(32), using a simplified table notation. Note that in our envisaged annotation process of information
status, the labellers will have no access to prosodic information.

(29) a. Ein starkes Erdbeben hat Zentral-Japan erschüttert.
A strong earthquake has hit central Japan.

b. Die Behörden gaben eine Tsunami-Warnung für den Südwesten heraus.
The authorities have issued a tsunami warning for the Southwest.

c. Auch im Inselstaat Vanuatu im Südpazifik wurden zwei Beben registriert.
Also in the island state of Vanuatu in the South Pacific two earthquakes have been
registered.

(30)

Ein starkes Erdbeben hat Zentral-Japan erschüttert.
a strong earthquake has central Japan shaken

(AP ) L-NEW (N) L-NEW (NP ) L-NEW (V ) L-NEW

(NP ) L-NEW (DP ) R-UNUSED

(DP ) R-NEW (V P ) L-NEW

(S) L-NEW

(31)

Die Behörden gaben eine Tsunami-Warnung für den Südwesten heraus.
the authorities gave a tsunami warning for the Southwest out

(NP ) L-NEW (V ) L-NEW (NP ) L-NEW (NP ) L-NEW (V ) L-NEW

(PP ) R-BRIDGING

(DP ) R-BRIDGING (DP ) R-NEW

(V P ) L-NEW

(S)L-NEW

228


FOCUS TRIGGERS AND FOCUS TYPES

(32)

Auch im Inselstaat Vanuatu im Südpazifik wurden zwei Beben registriert.
also in the island state Vanuatu in the South Pacific were two quakes registered

(N)L-NEW (N)L-NEW (NP )L-NEW (NP )L-GIVEN (V )L-NEW

(NP ) L-NEW (PP ) R-UNUSED (DP ) R-NEW

(PP ) R-UNUSED

(V P ) L-NEW

(S) L-NEW

The annotation proceeds along the principles defined above and consists of the following steps:

1. All referring expressions (DPs and PPs) receive an R-label. The phrases the authorities and
(for) the Southwest are linked to (or anchored in) central Japan via bridging. In cases of
syntactic embedding, e.g. [in the island state of Vanuatu [in the South Pacific]], R-labels are
nested inside each other.

2. All content words (including verbal particles) receive an L-label (topmost line). In our exam-
ple, the only L-GIVEN word is quakes in sentence (32), which is a near-synonym of earth-
quake in sentence (30). (Still the earthquakes are referentially distinct and independent from
each other, thus the R-NEW label on two quakes.) It is debatable whether the phrase tsunami
warning is L-NEW or rather L-ACCESSIBLE due to its intuitive relation to earthquake. How-
ever, since it is neither a hyponym nor a meronym of the latter there is so far no clear criterion
which would license a classification of the expression as L-ACCESSIBLE.

3. Following the syntactic structure of the sentences, complex non-referential phrases are as-
signed L-labels as well. A complex phrase counts as L-GIVEN if it is entailed by another
phrase in the discourse. This does not occur in the present examples but see Section 5, Exam-
ple (55). A phrase is L-ACCESSIBLE if it entails an earlier phrase. At the sentence level, this
might happen with elaborations, e.g. in (33).

(33) a. Sandy bought a car.
b. [L-ACCESSIBLE She chose a hybrid model].

In order to test the reliability of some aspects of the RefLex scheme, we had two trained student
annotators independently assign R-labels to 3445 referring DPs/PPs in written news text from the
DIRNDL corpus (Eckart et al., 2012), as well as L-labels to 5045 content words.23 The anno-
tations were done in SALTO (Burchardt et al., 2006). Since the annotators themselves had to
identify the markables in pre-parsed syntactic representations, the spans of some of the mark-
ables had to be adjusted after annotation. For the label granularity defined in Table 1, an evalu-
ation of inter-annotator agreement following Cohen (1960) and Artstein and Poesio (2008) yields
κ = 0.75 for the R-level and κ = 0.64 for the L-level. The value for the L-level is lower than
in earlier annotation experiments conducted by the authors of this article. The main reason for the
relatively low score of the L-level seems to be a certain insecurity in classifying an expression as
L-ACCESSIBLE. Obviously, there are more ways in which two items can be lexically related than
the ones we list in Table 1.

23. For this evaluation we did not consider that-CPs and complex phrases.

229


RIESTER AND BAUMANN

4. Contrastive focus vs. novelty focus

A longstanding issue in information structure theory is the differentiation between so-called con-
trastive focus and novelty focus (information focus). When occurring in isolation, both types of
focus are marked by nuclear pitch accents in English and German. According to Definition 1, both
may be primary foci, although there is evidence that novelty focus sometimes receives compara-
tively weaker (or at least different) marking with regard to certain prosodic or acoustic parameters,
see e.g. Alter et al. (2001); Selkirk (2002); Hedberg and Sosa (2008); Hermes et al. (2008); Katz
and Selkirk (2011).

The notion of contrast has received various interpretations in the literature on information struc-
ture and discourse structure, see Umbach (2004); Repp (2010). The most straightforward, though
not tenable, definition is in terms of the explicit mention of alternatives. Example (34) contains a
pair of contrastive topics and a pair of contrastive foci in a parallel structure; the focus in (35) does
not have an overt alternative and simply represents new information.

(34) JOHNCT1 ordered WAterF1, and MAryCT2 ordered BEERF2.

(35) Mary went into a pub. She [ordered BEER]F .

(36) Only MAryF drank beer.

Example (36) shows that explicit mention is not the only criterion for contrast. The exhaustive
particle only in (36) requires a domain of individuals – an alternative set – who did not drink beer,
except for Mary. Rooth (1992) presents a uniform account of alternative-eliciting focus for cases
of association with focus-sensitive particles, overt contrast, scalar implicatures, question-answer
pairs, ellipsis and comparatives. Several researchers, including Selkirk (2008), have taken these
phenomena to establish the paradigm of contrastive focus.

Occasionally, contrastive focus has been assigned stricter definitions, especially in terms of
correction or of exhaustivity/identification. While corrective contrast is often used in the design of
minimal pairs of non-contrastive and contrastive contexts – see (37) vs. (38) – e.g. for the use in
experiments of laboratory phonology, we argue that it should be seen as an extreme case of contrast
(involving the rejection of a previous utterance) which might possess its distinct prosodic marking.

(37) A: What did Mary drink?
B: She drank BEERF . (non-contrastive / non-corrective)

(38) A: Mary drank water.
B: (No.) She drank BEERF . (contrastive / corrective)

Exhaustivity need not be expressed by means of a focus-sensitive particle like in (36) but also occurs
with it-clefts (Atlas and Levinson, 1981; Hedberg, 1990; Delin and Oberlander, 1995; Reeve, 2011),
as a default interpretation of certain syntactic positions, like the preverbal position in Hungarian
(Szabolcsi, 1981; É. Kiss, 1998; Kenesei, 2006; Horvath, 2010), or simply arises as a conversational
implicature (Schulz and van Rooij, 2006; Spector, 2006) like in (39). B’s answer is interpreted as
saying that Mary was in the pub but no one else of a certain group which the interlocutors have in
mind.

(39) A: Who was at the pub?
B: MAryF was there.

230


FOCUS TRIGGERS AND FOCUS TYPES

It can be assumed that in most cases in which an answer to a question is interpreted exhaustively,
an alternative set has been introduced or accommodated beforehand. Interestingly however, É. Kiss
(1998), who discusses exhaustivity in her account of so-called identificational focus, assumes it
to be an independent property from contrastivity. While exhaustivity requires the exclusion but not
necessarily the identification of the alternatives, contrastivity only requires that the alternatives form
“a closed set of entities whose members are known to the participants of the discourse” (É. Kiss,
1998: 267) – but not necessarily their exclusion. We believe that the latter definition is very ap-
pealing, and will demonstrate below that it can be nicely applied when annotating natural language
data.

Before doing so, however, we will return to Selkirk’s (2008) criticism mentioned at the be-
ginning of Section 3. Recall that Selkirk argues for a distinction between (contrastive) focus and
discourse-new information, and against calling the latter focus. In doing so, she cites the influential
work by Rooth (1992), who however – as far as we can tell – leaves it open whether his theory of
Alternative Semantics also applies to plainly new information. Against Selkirk, we argue that there
is a straightforward move to integrate new information into Alternative Semantics. This, however,
requires some degree of exegesis of Rooth (1992) as it has been undertaken by Riester and Kamp
(2010). The key to solving the problem lies in taking the semantics of focus as suggested in Rooth
(1992) more literally than has been done in parts of the contemporary literature on focus seman-
tics. Alternative Semantics, in a nutshell, provides us with two important semantic components
for a theory of focus: the first component is the F-feature, which, when applied to some syntactic
constituent, introduces a second meaning, also called the focus semantic value, which is a set of
elements of the same semantic type as the focussed constituent. The focus semantic value of the
focussed expression MAryF is De, the (unrestricted) set of individuals. Büring (2013) has used the
attribute “raw” for such unfiltered focus semantic values.

Riester and Kamp (2010) note that, for natural discourse, we have to assume that De contains
each and every individual on earth, since it has not undergone any restriction. Therefore, it certainly
does not meet the requirements for contrastiveness formulated by É. Kiss, namely that the alterna-
tives form a closed set and be known to the participants of the discourse. In fact, the semantics of the
F-marker can be seen as an operation which is completely blind to contextual information. Riester
and Kamp (2010) call the focus semantic value of a natural language expression an “anonymous”
alternative set because its elements are, at the outset, unidentified. We shall assume, however, that
such anonymous alternative sets, the result of F-assignment, are sufficent for the F-marked phrase
to be called a novelty focus. Consider Example (40), the beginning of a news feature, and its spoken
realization in Figure 4.

(40) Bundespräsident Köhler [HAT das GeSETZ zur GeSUNDheitsreform unterschrieben]F .
Federal President Köhler has signed the bill on the health care reform.

Let us assume that the VP of (40) is focussed – while ignoring other properties the sentence might
have. Alternative Semantics tells us that the focus semantic value will consist of other VP denota-
tions (properties) than signing the health bill. This means that the president might have “done other
things”, although we will have difficulties reaching an agreement what his specific other options
had been, since neither the discourse context nor world knowledge tell us. The VP focus simply
represents new information; the sentence is informative – a contingency24 – but nothing more.

24. We are grateful to Carla Umbach (p.c.) for suggesting the notion in this context.

231


RIESTER AND BAUMANN

Figure 4: Novelty focus; DIRNDL s1790, 26-03-2007, 18:00, 3’20”: has signed the bill on the
health care reform

The second important component of focus semantics which Rooth (1992) introduces is the so-
called focus interpretation operator ∼ (“squiggle”). Semantic-pragmatically, ∼ is defined as an
anaphoric operator, which minimally imposes the following constraints on the interpretation of the
constituent it attaches to:25

Identify at least one proper alternative in the context

1. which matches the pattern defined by the focus semantic value of the phrase, and

2. which is different from the ordinary meaning of the phrase.

The attachment site of the ∼ operator is sometimes called the focus domain (Büring, 2008; Rooth,
2010) or, perhaps, the focus phrase (Krifka, 2006). However, opinions differ where to attach ∼.

(41) Die Europäische Union hat den Sklavenhandel früherer Jahrhunderte bedauert und sich
gegen Formen ∼[NEUzeitlicherF Sklaverei] gewandt.
The European Union has expressed regret over the slave trade of earlier centuries and
turned against forms of modern slavery.
(DIRNDL s1618, 26-03-2007, 14:00, 5’47”)

Example (41) nicely illustrates the fulfilment of the constraints imposed by the ∼ operator. (For the
sake of simplicity we only highlight the nuclear pitch accent on the relevant phrase.) The phrase
neuzeitlicher Sklaverei (‘modern slavery’) has a focus semantic value of the form {[[X slavery ]] |X
is some intersective modifier}. The discourse context contains the phrase den Sklavenhandel früherer

25. For a formal specification in DRT and further explications see Riester and Kamp (2010). Note that there might be
uses of ∼ in the literature which are not compatible with our strict anaphoric semantics.

232


FOCUS TRIGGERS AND FOCUS TYPES

Jahrhunderte (‘the slave trade of earlier centuries’), whose ordinary meaning matches the template
provided by (is an element of) the just mentioned focus semantic value (assuming that slavery and
slave trade may be read as synonyms). The two complex expressions are distinct from each other
and represent proper alternatives.26 As we will clarify below, the identification of contrastive al-
ternatives is not limited to the actual discourse context but does often rely on situational or world
knowledge. This does not mean that identifying contrastive alternatives is always possible, as we
see in (40). We are now in a position to delimit (non-contrastive) novelty focus from contrastive
focus.

Definition 4 (Novelty focus) A novelty focus is a non-given constituent which serves as an answer
to the immediate QUD or to a supplemental question and whose alternatives remain unidentified.

Definition 5 (Contrastive focus) A contrastive focus is a constituent which serves as an answer
to the immediate QUD or to a supplemental question and whose alternatives can be unanimously
identified in the context (i.e. discourse context, encyclopaedic knowledge, lexicon etc.)

Let us briefly summarise our theoretical assumptions surrounding the focus notion. We postulate
that both new and contrastive information trigger (or represent) focus. Novelty focus merely car-
ries the F-marker generating a focus semantic value as defined in Rooth (1992: 76). Contrastive
focus, too, comes with an F-marker but is additionally interpreted by means of the anaphoric ∼
operator. The result of successful focus interpretation is the identification of at least one contrastive
alternative. We loosely follow É. Kiss (1998) in assuming that contrastive identification means that
the addressee is able to name with certainty who or what this contrastive alternative (or the set of
alternatives) in the respective context is (Definition 5). Novelty focus, on the other hand, is defined
as a focus with an anonymous – not clearly identifiable – alternative set (Definition 4).

Contrastive constituents may contain, or entirely consist of, given material. We see this in
Example (41), or in (42) – a selectional focus, which makes a choice from a given list.

(42) John and Mary were at the pub but ∼[JOHNF ] left early.

This means that, in order to identify contrastive focus in corpus data we cannot confine ourselves to
discourse-new constituents and decide whether their alternatives are identifiable or not. Rather, we
must look for specific alternative-eliciting features or constellations (ALT). A list of such features
is presented in Table 2.27

Note well that having spotted an alternative-eliciting constellation is not yet having identified
contrastive focus. Beaver and Velleman (2011: 1674) state that speakers are not obliged to mark
contrast. We agree with this view. The fact that e.g. two co-hyponymic expressions, say an elephant
and a lion, occur as the arguments of the same predicate (ALT-OVERT-ARG), as in (43), does not
mean that they are necessarily contrasted against each other, although the speaker might decide to
establish the contrast prosodically; for instance by forming two prosodic phrases rather than one.
This would then signal that the order of arguments could have been the other way around.

(43) [ALT-OVERT-ARG An elephant] chased [ALT-OVERT-ARG a lion].

26. Wagner (2006) shows that there has to be sortal compliance between two alternatives. For instance, HIGH-END
convertible is a proper alternative to CHEAP convertible but not e.g. to RED convertible. What we presumably want
is a kind of co-hyponymy, or what Lang and Umbach (2002) call the availability of a common integrator.

27. A different list of features for “contrastive focus” is provided in Götze et al. (2007: 178ff).

233


RIESTER AND BAUMANN

Sublabel of ALT Description
FSP Item is associated with a focus-sensitive particle.
OVERT Item is an element of a pair or list of overtly contrastive expressions
-ARG - Type-identical arguments of the same predicate.
-COMP - Items occur in a comparative construction.
-COORD - Items are coordinated.
-EXT - Items occur in different sentences – sentence-external contrast
SEL Item selects one element from a pair or list of previously

introduced alternatives.

Table 2: Alternative-eliciting features

On the other hand, there are also triggers, e.g. focus-sensitive particles, which seem to necessarily
associate with a contrastive focus (ALT-FSP).28 When we apply our set of features from Table 2 to
Example (29), we obtain an additional tier of elicited alternatives, shown in (44) and (45).

(44) Ein starkes Erdbeben hat Zentral-Japan erschüttert.
ALT-OVERT-EXT

(45)
Auch im Inselstaat Vanuatu im Südpazifik wurden zwei Beben registriert.

ALT-FSP / ALT-OVERT-EXT

The phrase im Inselstaat Vanuatu im Südpazifik associates with the additive particle auch. It can
furthermore be contrasted with Zentral-Japan. In the following, we give corpus examples for the
remaining alternative-eliciting features.

Exchangeable arguments of a predicate (ALT-OVERT-ARG):

(46) [ALT-OVERT-ARG Die Fluggesellschaft Air Berlin] übernimmt [ALT-OVERT-ARG den Düs-
seldorfer Konkurrenten LTU].
Airline company Air Berlin will absorb its Düsseldorf rival LTU.
(DIRNDL s2428, 27-03-2007, 08:00)

28. The following table shows the frequencies of the most common conventional FSPs in German, in a sample of 2484
sentences from the DIRNDL corpus. However, about 90% of the sentences do not contain an FSP from the list, which
shows the need to identify other features than just ALT-FSP.

Particle Translation n
auch too 107
nur only 62
wieder again 53
ebenfalls too 8
lediglich only 5
nicht einmal not even 1
weder . . . noch neither . . . nor 1
sogar even –
bloß only –

234


FOCUS TRIGGERS AND FOCUS TYPES

Comparative (ALT-OVERT-COMP):

(47) Huber meinte in Dortmund zur Begründung, [ALT-OVERT-COMP in der Metall-Branche]
sei die Lage stabiler als [ALT-OVERT-COMP in der Chemieindustrie].
Huber explained in Dortmund that the situation in the metal industry was more stable than
in the chemical industry.
(DIRNDL s1943, 26-03-2007, 21:00)

Coordination (ALT-OVERT-COORD):

(48) Falls sich [ALT-OVERT-COORD die protestantische Unionisten-Partei DUP] und
[ALT-OVERT-COORD die katholische Sinn Féin] bis 24 Uhr Ortszeit nicht auf ein Bündnis
einigen, [. . . ]
Unless the protestant Unionist Party DUP and the catholic Sinn Féin reach an agreement
on an alliance until 12 p.m. local time [. . . ]
(DIRNDL s1249, 26-03-2007, 07:00)

Parallelism, sentence-external contrast (ALT-OVERT-EXT):

(49) [ALT-OVERT-EXT Kevin Kuranyi] schoss in Prag beide Tore für die deutsche Elf.
[ALT-OVERT-EXT Milan Baros] erzielte den Anschlusstreffer.
In Prague, Kevin Kuranyi scored both goals for the German team. Milan Baros scored the
other goal.
(DIRNDL s59-s60, 25-03-2007, 01:00)

Selection (ALT-SEL):

(50) Darauf verständigten sich die Parteivorsitzenden Paisley und Adams bei ihrem ersten per-
sönlichen Treffen in Belfast. Das Amt des Ersten Ministers soll [ALT-SEL Paisley] über-
nehmen.
This was arranged by the party leaders Paisley and Adams at their first personal meeting
in Belfast. Paisley will become the First Minister.
(DIRNDL s1897, 26-03-2007, 20:00)

As we said, we are not claiming that speakers have to mark all of the above alternative-eliciting
constellations prosodically, although they often will do so. More research is necessary on each of
them. The fact that marking contrastive focus is sometimes optional is one reason why text allows
for intonational variation when being read aloud. From an annotation perspective, however, this
means that identifying all cases of contrastive focus requires some advanced (top-down) reasoning
about potential Questions under Discussion, including supplemental questions, as sketched in Sec-
tion 2. Alternatively, it requires reverting to the spoken signal – what we have called the bottom-up
approach – including the assumption that nuclear pitch accents always signal focus, which we can
then classify as either contrastive or non-contrastive.29

A final problem which we currently cannot solve without falling back on spoken language is
the identification of implicit contrast. By this we mean obvious cases of contrastive interpretation
which do not come with one of the above features and whose marking likewise seems optional.

29. As should have become clear by now, this will leave a number of problems unsolved, including the question which
other forms of prosodic prominence are also markers of focus.

235


RIESTER AND BAUMANN

Consider Figure 5, showing a section from the context given in (51), in which a nuclear pitch accent
occurs on the phrase Ersten Ministers (‘First Minister’).

Das Amt des Ersten Ministers soll Paisley übernehmen

L*H H* L*H % H*L %

50

200

100

0 2.70.6 1.2 1.8
time (s)

fu
nd

am
en

ta
l f

re
qu

en
cy

 (H
z)

Figure 5: Optional contrast; DIRNDL s1897, 26-03-2007, 20:00, 04’05”: First Minister

(51) In Nordirland soll am 8. Mai eine gemeinsame Regierung aus protestantischen Union-
isten und katholischer Sinn Féin die Arbeit aufnehmen. Darauf verständigten sich die
Parteivorsitzenden Paisley und Adams bei ihrem ersten persönlichen Treffen in Belfast.
∼[ALT-IMPLICIT Das AMT des [ERSten MiNISters]F %] soll PAISley übernehmen.
In Northern Ireland, a power-sharing government of the protestant unionists and the catholic
Sinn Féin will start its work on May 8. This was arranged by the party leaders Paisley and
Adams at their first personal meeting in Belfast. Paisley will become the First Minister.

Knowing that the nuclear pitch accent is on Minister intuitively evokes the contrastive interpretation
that the speaker assumed that other posts were also assigned. This is what we indicate by the ∼
operator. Clearly, the version in Figure 5 is not the only way the sentence can be pronounced.
Instead, the speaker might have chosen not to mark the contrast, thus simply producing a novelty
focus as in (52).

(52) [Das
the

AMT

post
des
of

Ersten
first

MiNISters
minister

soll
shall

PAISley
Paisley

übernehmen.]F
take

There are many conceivable ways of evaluating the proposals made in this section. For example,
it would have been possible to have annotators detect alternative-eliciting features. However, as a
first step we asked annotators to tell apart contrastive from non-contrastive focus. To this end, we
reverted to a bottom-up process: we selected 3842 nuclear pitch accents from the DIRNDL corpus
(only the final pitch accents of an intonation phrase), disregarding their accent type (rise, fall etc.).
We made the assumption that all of them mark some sort of focus. The question we were interested
in was whether these foci would be contrastive or merely new information. Two independent student

236


FOCUS TRIGGERS AND FOCUS TYPES

annotators were provided with the accented words as well as the containing sentence plus context.
Their task was to say whether they could identify against what an accented word was contrasted in
the given context. For validation purposes, they were asked to note down the explicit or implicit
alternative. Finally, they had to label the focus as either non-contrastive (NO-CONTRAST), implicitly
contrastive (ALT-IMPLICIT), or marked by an alternative-eliciting feature from Table 2. As a result,
we obtain a κ-value of 0.56 for the eight categories. The score slightly improves to κ = 0.58 if we
only consider the three categories non-contrastive, implicitly contrastive, and marked by an explicit
ALT-feature. Nevertheless, determining contrastive focus turned out to be a rather difficult task. In
a final consensus annotation we classify the nuclear pitch accents as marking the focus types shown
in Table 3. Note that the table does not allow for any conclusions with regard to the total ratio of
contrastive versus non-contrastive foci in the complete data, since it leaves aside foci marked by
nuclear pitch accents of non-final intermediate phrases as well as pre- and postnuclear prominences.

Label n

ALT-FSP 45
ALT-OVERT-ARG 315
ALT-OVERT-COMP 12
ALT-OVERT-COORD 484
ALT-OVERT-EXT 234
ALT-SEL 36
ALT-IMPLICIT 479
NO-CONTRAST 2237

Table 3: (Non-)contrastive focus on 3842 (intonation-phrase final) nuclear pitch accents

5. Second occurrence focus and secondary accents

In the remaining part of this article we return to the issue of second occurrence focus (SOF), already
discussed in Section 1. Describing the precise conditions which license second occurrence focus
is not straightforward. In particular, SOF is not sufficiently characterised by defining it as a (con-
trastively) focussed and given constituent. A counterexample is shown in (53b) (Büring, 2008, ms.)
in which the second occurrence of John is realised as a primary – not secondary – focus.

(53) a. Many people only drank JUICEF at John’s party.
b. Even JOHNF only drank JUICESOF at his party.

Beaver and Velleman (2011) discuss at length the licensing conditions for second occurrence fo-
cus. In particular, they reject proposals which rely on the size (or mutual embedding) of the focus
domains (Büring, 2008, ms.; Rooth, 2010) or which redefine the notion of givenness for focussed
constituents (Selkirk, 2008). We will not repeat their argumentation here. The characterization of an
SOF given by Beaver and Velleman (2011) is a constituent which is “important” (in our terminology
ALT-marked) as well as “predictable”, where predictability is defined as “givenness-all-the-way-up”
– a constituent is predictable if and only if it is given and its containing phrasal context is given as
well. From this it follows that an unpredictable constituent must either be new or occur in a new

237


RIESTER AND BAUMANN

argument slot. In (53b), the given phrase juice occurs in a likewise given VP – it is predictable –
while the given expression John newly occupies the agent role, and is therefore unpredictable.

In (54) and (55) we provide a complete RefLex and elicited-alternatives annotation of the SOF
example (6), which was introduced in Section 1.30 We observe that the second occurrence of veg-
etables in (55) is embedded under other given constituents.

(54)

Everyone knew that Mary only eats VEgetables.
(NP ) L-NEW (V ) L-NEW (NP ) L-NEW

(V ) L-NEW (DP ) R-UNUSED (DP ) R-GENERIC

(V P ) L-NEW

(S) L-NEW

(CP ) R-UNUSED

(V P ) L-NEW

(S) L-NEW

ALT-FSP

(55)

Even PAUL knew that Mary only eats VEgetables.
(NP ) L-NEW (NP ) L-GIVEN (V ) L-GIVEN (NP ) L-GIVEN

(V ) L-GIVEN (DP ) R-GIVEN (DP ) R-GENERIC

(DP ) R-UNUSED (V P ) L-GIVEN

(V P ) L-GIVEN

(CP ) R-GIVEN

(V P ) L-GIVEN

(S) L-GIVEN

ALT-FSP ALT-FSP

The reader is encouraged to verify that every constituent marked L-GIVEN is entailed by the context
(in this case, simply repeated), while R-GIVEN phrases are coreferential. We shall assume that the
that-CP behaves similar to a definite DP while referring to a fact.

The last issue which we are going to discuss are other types of secondary focus which are not
second occurrence foci. Consider that, instead of (55), the speaker would have chosen to say (56).

(56)

Even PAUL knew that Mary is picky.
(NP ) L-GIVEN (AP ) L-NEW

(DP ) R-GIVEN

(S) L-NEW

(CP ) R-GIVEN

The utterance in (56) is clearly subjective. The speaker does not merely tell us that even Paul
knew that Mary is a vegetarian but she also lets us know, en passant, that she considers vegetarians
picky. We only get this interpretation because the entire CP does not carry a nuclear pitch accent but

30. The definite subject everyone is left unannotated here because, strictly speaking, we are not dealing with an out-of-
the-blue utterance but a constructed example which presupposes some context. In normal discourse, speakers would
limit the use of everyone to contexts in which it is clear which group is being referred to. In other words, the domain
of the universal quantifier is restricted to an identifiable set in the context. Therefore, the most appropriate label for
everyone is probably R-GIVEN, which might strike some readers as odd in the absence of any explicit context.

238


FOCUS TRIGGERS AND FOCUS TYPES

instead, following our intuitions, comes with a compressed intonation contour. It is only because of
this postnuclear compression that we obtain the interpretation that (the speaker thinks that) the two
CPs boil down to the same thing.31 The second CP is therefore labelled as referentially given. The
question whether the CP in (56) is deaccented or whether it carries postnuclear prominence cannot
be solved by introspection and therefore requires experimental research. Telling from the RefLex
annotations, however, we are able to state that the embedded clause contains lexically new material,
and we might furthermore argue that the new information contained in the clause is not-at-issue
– compare Section 2 – and gives an answer to the speaker-oriented question What is Mary like,
according to the speaker? We think that this is sufficient reason to grant the embedded predicate
the status of a novelty focus, which, when indeed marked by postnuclear prominence would count
as a secondary focus.

In analogy to the case in (56), we can also construct examples involving a coreferential DP
consisting of lexically new material. Such expressions have sometimes been called epithets in the
literature (Clark, 1977; Schlenker, 2005; Potts, 2005; Riester, 2009). Expressives, as in (57B), are
one kind of epithets. Again they represent not-at-issue content and will be considered novelty foci.
(Implicit question: What does the speaker think about Fred?)

(57)

A: Do you know where Fred is?
B: I haven’t SEEN the goddam idiot.

(NP ) L-NEW

(DP ) R-GIVEN

In our corpus, we do find instances of epithets which are marked by postnuclear prominence (mainly
indicated by increased duration and intensity of lexically stressed syllables). Figure 6 shows an
example of the R-GIVEN, L-NEW DP der serbischen Provinz (‘of the Serbian province’), which
corefers with the phrase Kosovo in the context shown in (58).

(58) Der UNO-Sondergesandte Ahtisaari plädiert für eine Unabhängigkeit des Kosovo unter
internationaler Aufsicht. Dies sei die einzige politische und wirtschaftliche Option für die
ZUkunft der serbischen ProVINZ.
UN Special Envoy Ahtisaari is making the case for an independence of the Kosovo under
international control. According to him, this is the only political and economic option for
the future of the Serbian province.

So far, this is merely anecdotal evidence, and more elaborate statistical investigations are neces-
sary. We would like to point out, however, that Baumann and Riester (in press) did find – mainly
in a corpus of read speech – that expressions which possess a “hybrid” information status (e.g.
R-GIVEN, L-NEW), such as epithets, were marked as significantly less prominent than fully new
items and significantly more prominent than fully given items. Only some of them were encoded by
postnuclear prominences, others by prenuclear accents, or less prominent (e.g. low) accent types.
Actually, several kinds of secondary prominence have been proposed in the literature, e.g. duration

31. This is a behaviour well-known from definite DPs. Compare example (i) from Umbach (2002):

(i) {John has an old cottage.}
a. Last summer he reconstructed [the SHED]. (non-coreferential)
b. Last summer he reconSTRUCted [the shed]. (coreferential)

239


RIESTER AND BAUMANN

für die Zukunft der serbischen Provinz

H*L secondary prominence %

L-NEW

R-GIVEN

50

280

100

200

0 1.90.4 0.8 1.2
time (s)

fu
nd

am
en

ta
l f

re
qu

en
cy

 (H
z)

Figure 6: Realisation of an epithet (R-GIVEN, L-NEW); DIRNDL s1730, 26-03-2007, 17:00, 0’31”:
the Serbian province

accents (Kohler, 2005), ornamental accents (Büring, 2007), or phrase accents (Grice et al., 2000),
mentioned in Section 1 above. However, the various concepts refer to quite different phenomena
or levels: first, the presence or absence of a pitch movement, i.e. tonal vs. non-tonal prominence
(e.g. Kohler’s duration accents are non-tonal); second, accent position or the status of a prominence
in the prosodic hierarchy (e.g. ornamental accents are prenuclear, phrase accents postnuclear); and
third, accent type, i.e. the form of a pitch movement on a metrically strong syllable (high or rising
accents are more prominent than low or falling ones, see e.g. Baumann and Grice, 2006). While
there is a growing body of evidence that these three levels of prosodic prominence are used for
marking various aspects of information structure, cf. Baumann (2012), it is a matter of some debate
whether the current ToBI (and GToBI) systems are suited to represent the relevant distinct types
of information structural meaning – although at least the distiction between several types of pitch
accent has always been a central incentive for the definition of the ToBI labels, see for instance
Pierrehumbert and Hirschberg (1990); Steedman (1991).

6. Conclusions

In this article, we have discussed a number of problems which arise when issues from the theo-
retical and experimental tradition of information structure are brought together with corpus data,
such as spoken radio news. In parts of the literature, focus is discussed as a semantic-pragmatic
phenomenon which is related to the answering of Questions under Discussion. Since the latter
are mostly implicit in monological discourse, it seems, however, that their identification is almost
as subtle as the identification of focus itself. Corpus data can open our eyes to a number of is-
sues that have often been ignored in research on focus, for instance the fact that spoken utterances
contain prenuclear and nuclear pitch accents as well as various kinds of postnuclear prominence.

240


FOCUS TRIGGERS AND FOCUS TYPES

On the meaning side, this corresponds to the fact that the informative parts of a sentence have an
internal structure that can be revealed by asking (nested) questions. As a result, we can have sev-
eral semantic-pragmatic foci within one sentence, which have different degrees of communicative
significance and which are marked by different kinds of prosodic prominence.

Information structure has a long history of theory building which has produced highly complex
models, sometimes based on relatively thin empirical evidence. The precise mapping between the
semantic-pragmatic and the prosodic phenomena is still not sufficiently described. (The same can be
said about morpho-syntactic focus marking in cross-linguistic perspective.) We are convinced that
building and analysing annotated corpus resources can complement experimental and theoretical
research in this regard.

A big problem in focus theory is its non-standardised terminology. Are there one or several kinds
of focus, and is new information one of them? How should givenness and contrast be defined? Is
contrastive topic a kind of focus? What is the precise definition of the F-feature and the∼ operator?
What is a focus domain? Researchers have provided diverging answers to these problems, including
the authors of this paper. In principle, nothing speaks against different terminological choices as
long as they are made transparent. On the other hand, it is plain that certain terminological and
classificatory decisions can blur rather than clarify insights into linguistic phenomena and may
cause problems in view of larger theoretical structures. Linguistic annotation is an important testbed
for both theory and terminology: the ability to annotate a certain phenomenon is support for its
underlying theoretical conceptualisation.

In this article we have provided two proposals with respect to the annotation of information
structure: (i) the RefLex scheme, a framework for the annotation of information status, divided into
a referential and a lexical level, and (ii) a set of alternative-eliciting features, which represent the
basis for a speaker’s decision to mark contrastive focus. We furthermore postulate that plainly new
information represents the basic type of focus. While both types of focus involve alternatives, the
defining criterion for contrastive focus is the addressee’s ability to identify (and name) at least one
of these alternatives, in the respective context. While the annotation of information status can (and
should) be accomplished without access to prosody, the annotation of contrastive focus in English
and German will ultimately require the use of spoken information since it can only partially be
determined on the basis of written data alone. Finally, we discussed the phenomenon of second
occurrence focus in the light of general assumptions about prosodic structure and the pragmatics of
focus, and concluded that it is not the only kind of focus which is realised by means of secondary
prominence. Very likely, there are other focal phenomena, like not-at-issue content, or expressions
with a hybrid information status which are sometimes marked by means of prenuclear or postnuclear
prominence.

241


RIESTER AND BAUMANN

Acknowledgements

The authors would like to thank the editors of this volume Stefanie Dipper, Bonnie Webber and
Heike Zinsmeister for their support. We thank the three anonymous reviewers for their comprehen-
sive and thoughtful comments. Thanks to Jutta Hartmann, Fabienne Martin and Carla Umbach for
discussions, as well as Maria Drixler and Lisa Kienzle for their corpus annotation work. We would
also like to thank Kordula de Kuthy and Detmar Meurers for testing our work in their 2012/13
class on information structure annotation at Tübingen. Thanks go as well to the organisers of the
DFG Research Network Questions in Discourse, Edgar Onea and Malte Zimmermann for enabling
numerous opportunities for the exchange of thoughts about discourse and information structure.
Financial support by the German Science Foundation (DFG) is kindly acknowledged. (Project A1
of the Stuttgart Collaborative Research Center 732 Incremental Specification in Context, and the
project GR 1610/5 Degrees of Activation and Focus-Background Structure in Spontaneous Speech
in Cologne.)

References

Steven Paul Abney. The English Noun Phrase in its Sentential Aspect. PhD thesis, MIT, Cambridge,
MA, 1987.

David Adger. Core Syntax: A Minimalist Approach. Oxford University Press, 2003.

Kai Alter, Ina Mleinek, Tobias Rohe, Anita Steube, and Carla Umbach. Kontrastprosodie in Sprach-
produktion und -perzeption. Linguistische Arbeitsberichte, 77:59–79, 2001. In: Anita Steube and
Carla Umbach, editors, Kontrast – lexikalisch, semantisch, intonatorisch. Universität Leipzig.

Ron Artstein and Massimo Poesio. Inter-Coder Agreement for Computational Linguistics. Compu-
tational Linguistics, 34(4):556–596, 2008.

Nicholas Asher and Alex Lascarides. Bridging. Journal of Semantics, 15:83–113, 1998.

Jay Atlas and Stephen Levinson. It-clefts, Informativeness and Logical Form. In Peter Cole, editor,
Radical Pragmatics, pages 1–61. Academic Press, New York, 1981.

Christine Bartels. Acoustic Correlates of ‘Second Occurrence Focus’: Towards an Experimental
Investigation. In Hans Kamp and Barbara Partee, editors, Context-Dependence in the Analysis of
Linguistic Meaning, pages 354–361. Elsevier, Amsterdam, 2004.

Stefan Baumann. Types of Secondary Prominence and their Relation to Information Structure.
Talk at Satellite Workshop of GLOW 35 on Production and Perception of Prosodically-Encoded
Information Structure, University of Potsdam, March 27, 2012.

Stefan Baumann and Martine Grice. The Intonation of Accessibility. Journal of Pragmatics, 10
(38):1636–1657, 2006.

Stefan Baumann and Arndt Riester. Referential and Lexical Givenness: Semantic, Prosodic and
Cognitive Aspects. In Gorka Elordieta and Pilar Prieto, editors, Prosody and Meaning, volume 25
of Interface Explorations, pages 119–162. Mouton de Gruyter, Berlin, 2012.

242


FOCUS TRIGGERS AND FOCUS TYPES

Stefan Baumann and Arndt Riester. Coreference, Lexical Givenness and Prosody in German. Lin-
gua, in press. doi: http://dx.doi.org/10.1016/j.lingua.2013.07.012. Special Issue ‘Information
Structure Triggers’ ed. by Jutta Hartmann, Susanne Winkler and Janina Radó.

Stefan Baumann, Doris Mücke, and Johannes Becker. Expression of Second Occurrence Focus in
German. Linguistische Berichte, 221:61–78, 2010.

David Beaver and Brady Clark. Sense and Sensitivity. How Focus Determines Meaning. Wiley &
Sons, Chichester, 2008.

David Beaver and Dan Velleman. The Communicative Significance of Primary and Secondary
Accents. Lingua, 121:1671–1692, 2011.

David Beaver, Brady Clark, Edward Flemming, Florian Jaeger, and Maria Wolters. When Semantics
Meets Phonetics: Acoustical Studies of Second Occurrence Focus. Language, 83(2), 2007.

Mary Beckman. Stress and Non-Stress Accent. Foris, Dordrecht, 1986.

Mary Beckman, Julia Hirschberg, and Stefanie Shattuck-Hufnagel. The original ToBI system and
the evolution of the ToBI Framework. In Sun-Ah Jun, editor, Prosodic Typology – The Phonology
of Intonation and Phrasing, pages 9–54. Oxford University Press, 2005.

Paul Boersma and David Weenink. PRAAT – A System for Doing Phonetics by Computer, Version
3.4. Technical Report 132, Institute of Phonetic Sciences of the University of Amsterdam, 1996.
URL http://www.fon.hum.uva.nl/praat/.

Aljoscha Burchardt, Katrin Erk, Anette Frank, Andrea Kowalski, and Sebastian Padó. SALTO: A
Versatile Multi-Level Annotation Tool. In Proceedings of the Fifth International Conference on
Language Resources and Evaluation (LREC), Genoa, Italy, 2006.

Daniel Büring. On D-Trees, Beans, and B-Accents. Linguistics & Philosophy, 26(5):511–545,
2003.

Daniel Büring. Semantics, Intonation and Information Structure. In Gillian Catriona Ramchand and
Charles Reiss, editors, The Oxford Handbook of Linguistic Interfaces. Oxford University Press,
2007.

Daniel Büring. What’s New (and What’s Given) in the Theory of Focus? In Proceedings of the
34th Annual Meeting of the Berkeley Linguistics Society, pages 403–424, 2008.

Daniel Büring. Givenness, Contrast & Topic. Talk at Workshop on Prosody and Information Sta-
tus in Typological Perspective, 35th Annual Meeting of the German Linguistic Society (DGfS),
University of Potsdam, March 13, 2013.

Daniel Büring. Been There, Marked That – A Theory of Second Occurrence Focus. 2008, ms.

Wallace L. Chafe. Discourse, Consciousness, and Time. University of Chicago Press, 1994.

Noam Chomsky. Deep Structure, Surface Structure, and Semantic Interpretation. In Danny Stein-
berg and Leon Jakobovits, editors, Semantics. An Interdisciplinary Reader in Philosophy, Lin-
guistics and Psychology, pages 183–216. Cambridge University Press, 1971.

243


RIESTER AND BAUMANN

Herbert H. Clark. Bridging. In Philip Johnson-Laird and Peter Wason, editors, Thinking: Readings
in Cognitive Science, pages 169–174. Cambridge University Press, 1977.

Jacob Cohen. A Coefficient of Agreement for Nominal Scales. Educational and Psychological
Measurement, 1(20):37–46, 1960.

Philippa Cook and Felix Bildhauer. Identifying ‘Aboutness Topics’: Two Annotation Experiments.
Dialogue and Discourse, 4(2):118–141, 2013.

Dick Crouch, Mary Dalrymple, Ron Kaplan, Tracy King, John Maxwell, and Paula New-
man. XLE Documentation. Palo Alto Research Center, 1993-2011, ms. URL
http://www2.parc.com/isl/groups/nltt/xle/doc/xle toc.html.

Judy Delin and Jon Oberlander. Syntactic Constraints on Discourse Structure: the Case of it-Clefts.
Linguistics, 33(3):465–500, 1995.

Kerstin Eckart, Arndt Riester, and Katrin Schweitzer. A Discourse Information Radio News
Database for Linguistic Analysis. In Christian Chiarcos, Sebastian Nordhoff, and Sebastian Hell-
mann, editors, Linked Data in Linguistics. Representing and Connecting Language Data and
Language Metadata, pages 65–76. Springer, Heidelberg, 2012.

Miriam Eckert and Michael Strube. Dialogue Acts, Synchronising Units and Anaphora Resolution.
Journal of Semantics, 17(1):51–89, 2000.

Katalin É. Kiss. Identificational Focus versus Information Focus. Language, 74(2):245–273, 1998.

Caroline Féry and Shinichiro Ishihara. The Phonology of Second Occurrence Focus. Journal of
Linguistics, 45(2):285–313, 2009.

Michael Götze, Thomas Weskott, Cornelia Endriss, Ines Fiedler, Stefan Hinterwimmer, Svetlana
Petrova, Anne Schwarz, Stavros Skopeteas, and Ruben Stoel. Information Structure. In Stefanie
Dipper, Michael Götze, and Stavros Skopeteas, editors, Information Structure in Cross-Linguistic
Corpora: Annotation Guidelines for Phonology, Morphology, Syntax, Semantics and Informa-
tion Structure, volume 7 of Interdisciplinary Studies on Information Structure. Universitätsverlag
Potsdam, 2007. SFB 632.

Martine Grice, D. Robert Ladd, and Amalia Arvaniti. On the Place of Phrase Accents in Intonational
Phonology. Phonology, 17(2):143–185, 2000.

Jeroen Groenendijk and Martin Stokhof. Studies in the Semantics of Questions and the Pragmatics
of Answers. PhD thesis, Universiteit van Amsterdam, 1984.

Carlos Gussenhoven. Focus, Mode and Nucleus. Journal of Linguistics, 19:377–417, 1983.

Carlos Gussenhoven. Sentence Accents and Argument Structure. In Iggy Roca, editor, Thematic
Structure: Its Role in Grammar, pages 79–106. Foris, Berlin, 1992.

Carlos Gussenhoven. On the Limits of Focus Projection in English. In Peter Bosch and Rob
van der Sandt, editors, Focus. Linguistic, Cognitive and Computational Perspectives, pages 43–
55. Cambridge University Press, 1999.

244


FOCUS TRIGGERS AND FOCUS TYPES

Eva Hajičová, Barbara H. Partee, and Petr Sgall. Topic-Focus Articulation, Tripartite Structures,
and Semantic Content. Kluwer, Dordrecht, 1998.

Eva Hajičová, Jarmila Panevová, and Petr Sgall. A Manual for Tectogrammatical Tagging of the
Prague Dependency Treebank. Technical Report tr-2000-09, Charles University, Prague, 2000.
URL http://ufal.mff.cuni.cz/pdt/Corpora/PDT 1.0/Doc/tmanual/tmanen.pdf.

Michael Halliday. Notes on Transitivity and Theme in English. Part 2. Journal of Linguistics, 3:
199–244, 1967.

Michael Halliday and Ruqaiya Hasan. Cohesion in English. Longman, London, 1976.

Nancy Hedberg. The Referential Status of Clefts. PhD thesis, University of Minnesota, 1990.

Nancy Hedberg and Juan M. Sosa. The Prosody of Topic and Focus in Spontaneous English Dia-
logue. In Chungmin Lee, Matthew Gordon, and Daniel Büring, editors, Topic and Focus: Cross-
linguistic Perspectives on Meaning and Intonation, volume 82 of Studies in Linguistics and Phi-
losophy, pages 101–120. Springer, Dordrecht, 2008.

Anne Hermes, Johannes Becker, Doris Mücke, Stefan Baumann, and Martine Grice. Articula-
tory Gestures and Focus Marking in German. In Proceedings of the 4th Conference on Speech
Prosody, Campinas, Brazil, 2008.

Tilman Höhle. Explikationen für ‘Normale Betonung’ und ‘Normale Wortstellung’. In Werner
Abraham, editor, Satzglieder im Deutschen, pages 75–153. Narr, Tübingen, 1982.

Julia Horvath. ’Discourse Features’, Syntactic Displacement and the Status of Contrast. Lingua,
pages 1346–1369, 2010.

Ray Jackendoff. Semantic Interpretation in Generative Grammar. MIT Press, Cambridge, MA,
1972.

Jonah Katz and Elisabeth Selkirk. Contrastive Focus vs. Discourse-new: Evidence from Phonetic
Prominence in English. Language, 87(4):771–816, 2011.

István Kenesei. Focus as Identification. In Valéria Molnár and Susanne Winkler, editors, The Ar-
chitecture of Focus, Studies in Generative Grammar, pages 137–168. Mouton de Gruyter, Berlin,
2006.

Klaus J. Kohler. Form and Function of Non-Pitch Accents. Arbeitsberichte des Instituts für Phonetik
und digitale Sprachverarbeitung der Universität Kiel (AIPUK), 35a:97–123, 2005. Prosodic
Patterns of German Spontaneous Speech.

Manfred Krifka. Association with Focus Phrases. In Valéria Molnár and Susanne Winkler, editors,
The Architecture of Focus, Studies in Generative Grammar. Mouton de Gruyter, Berlin, 2006.

Manfred Krifka. Basic Notions of Information Structure. Acta Linguistica Hungarica, 55(3-4):
243–276, 2008.

245


RIESTER AND BAUMANN

Manfred Krifka. A Compositional Semantics for Multiple Focus Constructions. In Joachim Jacobs,
editor, Informationsstruktur und Grammatik, pages 17–53. Westdeutscher Verlag, Opladen, 1992.
Also in Proceedings of SALT 1. Cornell Working Papers in Linguistics 10. 1991.

Knud Lambrecht. Information Structure and Sentence Form. Cambridge University Press, 1994.

Ewald Lang and Carla Umbach. Kontrast in der Grammatik: spezifische Realisierungen und
übergreifender Konnex. Linguistische Arbeitsberichte, 79:145–186, 2002.

Sebastian Löbner. Definite Associative Anaphora. In Proceedings of the Discourse Anaphora and
Resolution Colloquium (DAARC), Lancaster, 1998.

Jörg Mayer. Transcription of German Intonation. The Stuttgart System. ms., 1995. URL
http://www.ims.uni-stuttgart.de/phonetik/joerg/labman/STGTsystem.html.

Malvina Nissim, Shipra Dingare, Jean Carletta, and Mark Steedman. An Annotation Scheme for
Information Status in Dialogue. In Proceedings of the Fourth International Conference on Lan-
guage Resources and Evaluation (LREC), Lisbon, 2004.

Patrizia Paggio. Annotating Information Structure in a Corpus of Spoken Danish. In Proceedings
of the Fifth International Conference on Language Resources and Evaluation (LREC), pages
1606–1609, Genoa, Italy, 2006.

Barbara H. Partee. Focus, Quantification and Semantics-Pragmatics Issues. In P. Bosch and
R. van der Sandt, editors, Focus: Linguistic, Cognitive, and Computational Perspectives, pages
213–231. Cambridge University Press, 1999.

Janet Pierrehumbert and Julia Hirschberg. The Meaning of Intonational Contours in the Interpre-
tation of Discourse. In Philip Cohen, Jerry Morgan, and Martha Pollack, editors, Intentions in
Communication, chapter 14. MIT Press, 1990.

Massimo Poesio and Renata Vieira. A Corpus-Based Investigation of Definite Description Use.
Computational Linguistics, 24(2):183–216, 1998.

Christopher Potts. The Logic of Conventional Implicatures. Oxford University Press, 2005.

Ellen F. Prince. Toward a Taxonomy of Given-New Information. In Peter Cole, editor, Radical
Pragmatics, pages 233–255. Academic Press, New York, 1981.

Ellen F. Prince. The ZPG Letter: Subjects, Definiteness and Information Status. In William C.
Mann and Sandra A. Thompson, editors, Discourse Description: Diverse Linguistic Analyses of
a Fund-Raising Text, pages 295–325. Benjamins, Amsterdam, 1992.

Matthew Reeve. The Syntactic Structure of English Clefts. Lingua, 121:142–171, 2011.

Sophie Repp. Defining ‘Contrast’ as an Information-Structural Notion in Grammar. Lingua, 120:
1333–1345, 2010.

Arndt Riester. Partial Accommodation and Activation in Definites. In Proceedings of the 18th
International Congress of Linguists, pages 134–152, Seoul, 2009. Linguistic Society of Korea.

246


FOCUS TRIGGERS AND FOCUS TYPES

Arndt Riester and Hans Kamp. Squiggly Issues: Alternative Sets, Complex DPs, and Intensionality.
In Maria Aloni et al., editors, Logic, Language and Meaning. Revised Selected Papers from the
17th Amsterdam Colloquium, pages 374–383. Springer, Berlin, 2010.

Arndt Riester, David Lorenz, and Nina Seemann. A Recursive Annotation Scheme for Referen-
tial Information Status. In Proceedings of the Seventh International Conference on Language
Resources and Evaluation (LREC), pages 717–722, Valletta, Malta, 2010.

Craige Roberts. Information Structure in Discourse: Towards an Integrated Formal Theory of Prag-
matics. OSU Working Papers in Linguistics, 49, 1996.

Christian Rohrer and Martin Forst. Improving Coverage and Parsing Quality of a Large-Scale LFG
for German. In Proceedings of the Fifth International Conference on Language Resources and
Evaluation (LREC), Genova, 2006.

Mats Rooth. Second Occurrence Focus and Relativized Stress F. In Caroline Féry and Malte
Zimmermann, editors, Information Structure: Theoretical, Typological, and Experimental Ap-
proaches. Oxford University Press, 2010.

Mats Rooth. A Theory of Focus Interpretation. Natural Language Semantics, 1(1):75–116, 1992.

Mats Rooth. On the Interface Principles for Intonational Focus. In Teresa Galloway and Justin
Spence, editors, Proceedings of Semantics and Linguistic Theory (SALT) 6, pages 202–226,
Ithaca, NY, 1996. CLC Publications.

Phillippe Schlenker. Minimize Restrictors! (Notes on Definite Descriptions, Condition C and Ep-
ithets). In Emar Maier, Corien Bary, and Janneke Huitink, editors, Proceedings of Sinn und
Bedeutung IX, pages 385–416, Nijmegen, 2005.

Katrin Schulz and Robert van Rooij. Pragmatic Meaning and Non-monotonic Reasoning: the Case
of Exhaustive Interpretation. Linguistics and Philosophy, 29(2):205–250, 2006.

Roger Schwarzschild. GIVENness, AvoidF, and Other Constraints on the Placement of Accent.
Natural Language Semantics, 7(2):141–177, 1999.

Elisabeth Selkirk. Contrastive FOCUS vs. Presentational Focus: Prosodic Evidence from Right
Node Raising in English. In Proceedings of the 1st International Conference on Speech Prosody,
pages 643–646, Aix-en-Provence, 2002.

Elisabeth Selkirk. Contrastive Focus, Givenness and the Unmarked Status of ‘Discourse-New’. Acta
Linguistica Hungarica, 55(3–4):331–346, 2008.

Elisabeth Selkirk. Phonology and Syntax. MIT Press, Cambridge, 1984.

Elisabeth Selkirk. Sentence Prosody. Intonation, Stress, and Phrasing. In J. Goldsmith, editor, The
Handbook of Phonological Theory, pages 550–569. Blackwell, 1995.

Mandy Simons, Judith Tonhauser, David Beaver, and Craige Roberts. What Projects and Why. In
Proceedings of SALT 20, pages 309–327, Vancouver, 2010.

247


RIESTER AND BAUMANN

Benjamin Spector. Scalar Implicatures: Exhaustivity and Gricean Reasoning. In Maria Aloni, Alis-
tair Butler, and Paul Dekker, editors, Questions in Dynamic Semantics, pages 229–254. Elsevier,
Amsterdam, 2006.

Mark Steedman. Structure and Intonation. Language, 67(2):207–264, 1991.

Anna Szabolcsi. The Semantics of Topic-Focus Articulation. In Jeroen Groenendijk, Theo Janssen,
and Martin Stokhof, editors, Formal Methods in the Study of Language. Part 2, volume 136 of
Mathematical Centre Tracts, pages 513–540. Mathematisch Centrum, Amsterdam, 1981.

Carla Umbach. (De)accenting Definite Descriptions. Theoretical Linguistics, 27(2/3), 2002.

Carla Umbach. On the Notion of Contrast in Information Structure and Discourse Structure. Journal
of Semantics, 21(2), 2004.

Michael Wagner. Givenness and Locality. In Proceedings of SALT XVI, pages 295–312, Ithaca, NY,
2006.

248