Indirect discourse as mixed quotation? An experimental investigation

Sebastian Walter*

Abstract. The results of an experimental rating study are reported suggesting that
self-pointing gestures aligned with a third-person pronoun are acceptable in German
indirect discourse (ID) utterances. Following a proposal by Ebert & Hinterwimmer
(2022) for self-pointing gestures in free indirect discourse (FID), self-pointing gestures
in ID are interpreted as character viewpoint gestures (CVGs) quoted from the matrix
subject. Crucially, it is argued that in ID, a perspective shift to the matrix subject
can take place. It is proposed that ID is an instance of mixed quotation involving a
demonstration (cf. Clark & Gerrig 1990, Davidson 2015) where self-pointing is quoted
from the matrix subject’s original utterance.
Keywords. Perspective taking; gesture semantics; indirect discourse; mixed quotation

1. Introduction. Speakers have several options to report what someone else thought or said. For
example, they can do so by giving a verbatim repetition of the reported speaker’s thought or speech,
i.e., they directly quote the words the other person uttered. This is commonly referred to as direct
discourse (DD). Alternatively, they can simply paraphrase the propositional content of the reported
speaker’s original thought or utterance. This is called indirect discourse (ID) and is standardly
assumed to be non-quotational. A third option is what is called free indirect discourse (FID), a
way to report someone’s thoughts or speech without any overt marking, which has been argued
to be an instance of highly specialized mixed quotation (Maier 2015). Particularly interesting is
the behavior of so-called indexicals in speech reports, an example being the first-person pronoun
I. Broadly speaking, the reference of an indexical can vary from context to context, which sets
them apart from ordinary expressions. Coming back to the first-person pronoun I, this means
that it always refers to the current speaker. In DD utterances, however, it receives a so-called
shifted interpretation, that is, it is not interpreted from the current speaker’s (= reporting speaker’s)
point of view, but rather from the reported speaker’s point of view, as first-person pronouns in DD
utterances always refer to the reported speaker. In general, all indexicals shift in DD utterances. As
will be shown below, the picture is more complex for ID and FID because there, not all indexicals
have to shift.

Clark & Gerrig (1990) propose that quotations are demonstrational, meaning that they are de-
pictive rather than descriptive. Furthermore, in this approach not only words can be quoted, but, for
example, also gestures (including facial expressions), intonation, and even non-linguistic behavior.
These ideas have been formalized by Davidson (2015). In a study investigating Maier’s (2015)
claim that FID is an instance of mixed quotation with pronouns and tenses being systematically
unquoted, Ebert & Hinterwimmer (2022) found that a self-pointing gesture (= a pointing gesture

*Acknowledgments: This research was conducted in the DFG-funded project Visual and non-visual means of
perspective taking in language, which is part of the Priority Program 2392 (ViCom). I thankfully acknowledge the
financial support of the German Research Foundation (DFG). Furthermore, I would like to thank Cornelia Ebert and
Stefan Hinterwimmer for their valuable comments on this work. Finally, I thank Lennart Fritzsche for his help with
conducting the study reported in this paper. Authors: Sebastian Walter, Goethe University Frankfurt (s.walter@em.uni-
frankfurt.de).

Proceedings of ELM 3: 423-434, 2025

c©2025 Sebastian Walter
Published by the LSA with permission of the author(s) under a CC BY license. 423

https://journals.linguisticsociety.org/proceedings/index.php/ELM/issue/archive
https://www.elm-conference.net/


toward the speaker’s body) co-referent with the reported speaker is acceptable in FID. They found
the same for first-person pronouns in DD utterances. By unifying Davidson’s (2015) account of
quotation as demonstration with Ebert et al.’s (2020) approach to gesture semantics, they analyze
the results as follows: self-pointing gestures are interpreted as gestures quoted from the reported
speaker. Based on the assumption that ID is non-quotational, they hypothesized that self-pointing
in ID is unacceptable. However, they found that self-pointing was also surprisingly acceptable in
ID utterances. According to them, this suggests that a similar perspective-shifting strategy as in
FID is available in ID as well. However, it seems to be more constrained and is not necessarily
present. In other words, ID allows for the option of mixed quotation to be present. The study re-
ported in this paper further investigates this claim. Previous research has shown that temporal and
local adverbials, which are also indexicals, can shift in ID only if the reported speaker’s perspective
is prominent in the surrounding discourse (Plank 1986, Anderson 2019). Based on these findings,
it was hypothesized that self-pointing is acceptable in ID only if the perspective of the reported
speaker (= the matrix subject) is prominent, or, in other words, that mixed quotation is available
in ID if this perspective is prominent. To test this, a rating study was conducted. The results go
beyond the initial hypothesis, since they show that self-pointing in ID is acceptable regardless of
whether the reported speaker’s perspective is prominent on the speech level or not, thus suggesting
that ID in general allows for mixed quotation.

This paper is structured as follows: Section 2 provides the relevant theoretical background on
perspective in speech (Section 2.1), speech-accompanying gestures (Section 2.2), and quotation
(Section 2.3). The experimental study is reported in Section 3. Section 4 concludes the paper.

2. Background.

2.1. PERSPECTIVE IN SPEECH. Many expressions depend on a perspective in their interpreta-
tion. Among these expressions are predicates of personal taste ((be) tasty), epithets (that idiot),
relational expressions (this, that, left, right), and evaluative expressions (fun). The default for them
is to be interpreted from the perspective of the current speaker (Harris 2012). However, this default
changes in instances of speech reports. As will be shown below, depending on the type of speech
report, they can receive what is commonly referred to as a shifted interpretation, i.e., an interpre-
tation not from the current speaker’s perspective, but from the perspective of the speaker whose
thoughts or speech is currently being reported. Examples of different types of speech reports are
given in (1).

(1) On her way home, Mary heard a song by Kendrick Lamar that she liked on the radio.
a. She thought: “I will buy his new album tomorrow.”
b. She thought that she would buy his new album on the following day.
c. She would buy his new album tomorrow.

(Hinterwimmer 2017, p. 284)

The utterance given in (1a) is an instance of DD. In DD, the words or thoughts of the reported
speaker are directly quoted. Therefore, a complete perspective shift toward the reported speaker
takes place and all perspective-dependent expressions are evaluated from their perspective. Thus,
the first-person personal pronoun I refers to Mary in (1a) rather than to the current speaker. More-
over, temporal adverbials, such as tomorrow in (1a), are also evaluated in the context where the

Proceedings of ELM 3: 423-434, 2025

Sebastian Walter:
Indirect discourse as mixed quotation? An experimental investigation. 424

https://journals.linguisticsociety.org/proceedings/index.php/ELM/issue/archive
https://www.elm-conference.net/


reported speech event took place, meaning that tomorrow interpreted as referring to the day fol-
lowing Mary’s thought.

In (1b), where the same thought as in (1a) is rendered as an utterance in ID, perspective-
dependent expressions behave differently than in DD. ID is generally analyzed as a non-quotational
report of some other individual’s speech or thoughts. Consequently, the systematic shifting be-
havior of perspective-dependent expressions attested for DD does not fully generalize to ID. For
example, personal pronouns (at least in German, English, and many other languages) never shift
in ID (but see, e.g., Anand & Nevins 2004 for data from other languages where personal pronouns
shift in ID). This is illustrated by (1b), where a third-person pronoun is needed to refer to Mary.
Using a first-person pronoun would result in a different interpretation of the sentence.

By contrast, predicates of personal taste and evaluative expressions receive a shifted interpre-
tation in ID. Temporal and local deictic expressions are normally interpreted from the reporting
speaker’s perspective, thus explaining why on the following day is used to refer to the day follow-
ing the thinking event in (1b). However, these expressions can also receive a shifted interpretation
if the reported speaker’s perspective (i.e., Mary’s in (1b)) is prominent in the discourse (Plank
1986, Anderson 2019).

Finally, the utterance in (1c) is an instance of FID—a way to report another individual’s
thoughts or speech without overt marking. In FID, all perspective-dependent expressions shift
toward the reported speaker, the only exceptions being pronouns and tense markings. FID thus
shares properties with both ID and DD and is often seen as a blend of the two. In formal seman-
tics, the most popular treatment of FID is the so-called double context analysis (e.g., Schlenker
2004, Eckardt 2014). There is, however, also a proposal by Maier (2015), which treats FID as an
instance of highly specialized mixed quotation, where a FID utterance is seen as a verbatim quote
of an individual’s thoughts or speech with pronouns and tense markings being systematically un-
quoted. A similar proposal will be made for ID in this paper.

2.2. SPEECH-ACCOMPANYING GESTURES. Gestures are defined as communicative movements
of the hands, arms, and even the face which transport a speaker’s thoughts, intentions, or emotions.
Therefore, gestures that occur alongside speech add meaning to an utterance, or, more precisely,
speech and gesture work together to convey a multimodal message (e.g., McNeill 1992, Kendon
2004, de Ruiter 2000, Ebert 2024). The following examples illustrate this:

(2) a. I brought a bottle of water to the talk. + BIG1

b. I brought this bottle to the talk. + POINTING TO BOTTLE

The iconic BIG-gesture in (2a) contributes size information of the bottle to the verbal part of the
utterance, i.e., that it is big. The pointing gesture in (2b), by contrast, selects its referent directly
and adds to the verbal part of the utterance that the bottle pointed at is the bottle the speaker
brought to the talk. Crucially, in the semantic theory modeling the meaning contributions of
speech-accompanying gestures by Ebert et al. (2020), it is argued that this is the only difference
between pointing and co-speech gestures: while iconic gestures refer to an abstract individual,
pointing gestures make reference to an individual in the world. Besides this, they behave alike

1Underlining in examples indicates gesture-speech alignment, small caps are used to indicate a speech-
accompanying gesture.

Proceedings of ELM 3: 423-434, 2025

Sebastian Walter:
Indirect discourse as mixed quotation? An experimental investigation. 425

https://journals.linguisticsociety.org/proceedings/index.php/ELM/issue/archive
https://www.elm-conference.net/


from a semantic point of view in Ebert et al.’s (2020) account and are therefore both analyzed as
making meaning contributions similar to that of appositives (cf. Potts 2005). They thus, unless
aligned with a demonstrative as in (2b), make not-at-issue contributions by default (for experimen-
tal validation, see Ebert et al. 2020), meaning that they cannot be directly denied in discourse and
project through semantic operators such as negation. Especially this first property has often been
ascribed to not-at-issue content not being on the table for discussion (Farkas & Bruce 2010). The
short discourse in (3) illustrates this.

(3) A: There is a window in Peter’s living room. + ROUND

B1: # That’s not true! The window is not round.
B2: Yes, true, but the window isn’t round.

When trying to directly object the gesture’s contribution of A’s utterance as B1 does, this results in
infelicity. What one can do instead to target the not-at-issue content conveyed by the gesture is, for
example, to assent with the main clause proposition followed by an adversative continuation (cf.
diagnostic #1c in Tonhauser 2012). This explains why the response of B2 is felicitous.

It has been shown in Section 2.1 already that perspective plays a decisive role when it comes
to the interpretation of many lexical expressions. However, perspective can also be encoded in
gesture (McNeill 1992), the most important distinction being the one between so-called character
viewpoint gestures (CVGs) and observer viewpoint gestures (OVGs). When performing a CVG, the
whole body is often involved in the gesture production as it depicts an event from an internal, first-
person perspective. This contrasts with OVGs where normally only hands and arms are involved
in the production of the gesture as it depicts an event from an outside, third-person perspective.
The examples given in Figure 2a and Figure 2b illustrate this. The pictures are taken from a study
reported in Parrill (2010) where participants described short cartoon clips to a friend they brought
to the study. The two persons shown in Figure 2 describe the scene shown in Figure 1 where a
skunk is hopping across a room. The person in Figure 2a produces a clear instance of a CVG as
they mimic the skunk’s body posture and also its facial expressions, thus adopting a first-person
perspective. The person in Figure 2b, by contrast, describes the same hopping event by means
of an OVG because they only use their right index finger to trace the hopping movement of the
skunk. Thus, they adopt the external, third-person perspective that is representative of OVGs.
From these examples, another interesting can be made: the two types of viewpoint gestures differ
with respect to what they typically express. For example, OVGs encode trajectory more often than
CVGs do (Parrill 2010). Another interesting observation is that CVGs have been shown to be more
informative than OVGs (Beattie & Shovelton 2002). Several other differences between CVGs and
OVGs have been attested. However, they will not be discussed here, as they would exceed the
scope of the present paper.

2.3. QUOTATION AS DEMONSTRATION. Quotations, such as the DD utterance in (1a), have tra-
ditionally been argued to be verbatim repetitions of what someone else said (e.g., Quine 1940,
Davidson 1979). The quoted words were assumed to be mentioned, but not used, suggesting they
cannot interact with surrounding linguistic material. More recent research, however, has shifted
away from a strict verbatim condition (e.g., Davidson 2015), and it is more common to assume
that quoted elements are often simultaneously used and mentioned. This is evidenced by examples

Proceedings of ELM 3: 423-434, 2025

Sebastian Walter:
Indirect discourse as mixed quotation? An experimental investigation. 426

https://journals.linguisticsociety.org/proceedings/index.php/ELM/issue/archive
https://www.elm-conference.net/


Figure 1: Cartoon scene of a skunk hopping across a room. Taken from Parrill (2010)

(a) CVG used to depict the skunk in Figure 1
(Figure 3 in Parrill 2010, p. 652)

(b) OVG used to depict the skunk in Figure 1
(Figure 2 in Parrill 2010, p. 651)

Figure 2: Examples of a CVG and an OVG to depict the event shown in Figure 1

such as the following:

(4) Stiviano’s lawyer has not denied the part about the gifts, although he says there is not a
“peppercorn of a fact” that any fraud was involved.

(NYT, May 1, 2014, cited from Davidson 2015, p. 483)

Here, the quoted constituent a peppercorn of a fact is used and mentioned at the same time. It is
mentioned because the author of the sentence in (4) indicates by use of the quotation marks that
the words quoted are a verbatim repetition of what Stiviano’s lawyer said. However, the phrase
is also used as it needs to be integrated with the surrounding linguistic material to derive a full
proposition. The content of the quotation thus also plays a role in this example. This means that
the underlying report paraphrasing the original proposition expressed in the reported speech act
can be retrieved when ignoring the quotation marks. The quotation marks add another level of
meaning—namely that whatever stands in the quotation marks was part of the original utterance
(Potts 2005, Maier 2015). This is commonly known as mixed quotation.

For FID utterances (cf. (1c)), an analysis has been proposed treating it as a highly specialized
form of mixed quotation where the original thought or utterance of the reported speaker with
pronouns and tense markings being systematically unquoted, thus accounting for their shifting
behavior (Maier 2015). The unquotation mechanism is conceptualized as a pragmatic process.
The quotational structure of (1c) is thus as in (5). Square brackets indicate unquotation.

Proceedings of ELM 3: 423-434, 2025

Sebastian Walter:
Indirect discourse as mixed quotation? An experimental investigation. 427

https://journals.linguisticsociety.org/proceedings/index.php/ELM/issue/archive
https://www.elm-conference.net/


(5) On her way home, Mary heard a song by Kendrick Lamar that she liked on the radio. “[She]
[would] buy [his] new album tomorrow.”

Note that the quotation marks and the square brackets are just used to illustrate (un)quotation.
Normally, they are omitted in FID utterances.

Before turning to experimental evidence in favor of the mixed-quotational account of FID
proposed by Maier (2015), let’s briefly introduce the idea that quotation involves demonstration
(Clark & Gerrig 1990). This means that quotations in general serve a depictive rather than a
descriptive purpose and are thus iconic in their nature. Additionally, the idea that quotation involves
demonstration implies that speakers can quote aspects of an utterance relating to its form and not
only to its content.

(6) And so she said “[whispering] What are we going to do?”

(Clark & Gerrig 1990, p. 487)

Here, not only words are quoted, but also the manner in which they were uttered, that is, with a
whispering voice.

The predictions of the demonstrational account to quotation go even further, as evidenced by
(7). The be like-construction is also an instance of quotation (e.g., Davidson 2015) where special
emphasis is put on quoting the form of an utterance. The quotation marks are again only inserted
to highlight the quoted part of the utterance and are normally omitted.

(7) My cat was like “feed me!”

(Davidson 2015, p. 485)

By virtue of uttering (7), an event is demonstrated, i.e., quoted, where the speaker’s cat behaves in
a certain manner leading them to infer that it is hungry. This illustrates that when assuming that
quotation is inherently demonstrational, even non-linguistic behavior can be quoted.

Davidson (2015) proposes a formal analysis for these observations and explicitly claims that
also speech-accompanying gestures can be part of a demonstration, or, in other words, quoted.
This leads Ebert & Hinterwimmer (2022) to combine Davidson’s (2015) proposal with the formal
proposal for the semantic contribution of gestures as proposed by Ebert et al. (2020) (cf. Section
2.2 of the present article). By means of a rating study, they test their claims by testing DD, ID,
and FID speech reports in which a pronoun occurs that is co-referent with the reported speaker
(first-person pronoun in DD, third-person pronoun in ID and FID). A sample item (translated from
German) is given below. Capitalization indicates focalization.

(8) a. FID: Leona was extremely annoyed. Again, SHE had paid the bill for the entire group.
+ SELF-POINTING

b. ID: Leona was extremely annoyed. She was angry that SHE had paid the bill for the
entire group again. + SELF-POINTING

c. DD: Leona was extremely annoyed. Angrily, she thought: “Now, I have paid the bill for
the entire group again.” + SELF-POINTING

(Ebert & Hinterwimmer 2022, p. 345)

Proceedings of ELM 3: 423-434, 2025

Sebastian Walter:
Indirect discourse as mixed quotation? An experimental investigation. 428

https://journals.linguisticsociety.org/proceedings/index.php/ELM/issue/archive
https://www.elm-conference.net/


Following Davidson (2015) in the assumption that quotation is demonstrational and therefore also
gestures can be quoted, Ebert & Hinterwimmer (2022) hypothesize that self-pointing aligned with
the first-person pronoun should be acceptable in the DD example (8c) because it is then interpreted
as a CVG quoted from the reported speaker. If Maier’s (2015) theory is correct and FID in fact
involves mixed quotation, the self-pointing should also be acceptable in examples like (8a). Finally,
for ID utterances as in (8b) they hypothesize that self-pointing should not be acceptable when
aligned with a third-person pronoun as ID does not involve quotation and therefore the gesture
cannot be interpreted as a quoted CVG. They thus predict that self-pointing should be judged
significantly worse in ID than in FID and DD.

Their results were only partly in line with their hypothesis. Descriptively, the results show
that ID was the condition that received the worst ratings. An ANOVA revealed, however, that this
difference was only significant in the F1 and F2 analysis in the contrast between DD and ID. The
contrast between FID and ID only reached significance in the F1 analysis, which they attribute to
the surprisingly good ratings obtained for the ID condition. Overall, they interpret the results as
follows: The good ratings for the FID condition show that it is indeed quotational, thus presenting
evidence in favor of the approach proposed by Maier (2015). For ID, they argue, the results imply
that a similar perspective-shifting strategy as in FID is available there, as well, meaning that self-
pointing in ID can also be interpreted as a quoted CVG.

Ebert & Hinterwimmer’s (2022) interpretation of the findings for ID thus suggests that it also
is an instance of mixed quotation. It should thus receive a similar treatment as FID in the account
proposed by Maier (2015). This aligns with the observation that some indexical expressions can
shift or obligatorily shift toward the reported speaker in ID. However, the quotation mechanism
should arguably be more constrained than in FID as more expressions are interpreted from the
reporting speaker’s perspective in ID than in FID. The study reported in this paper aims to test this
hypothesis by adapting the study reported in Ebert & Hinterwimmer (2022). It tested whether alter-
ing the perspective prominent at the speech level in ID could affect the availability of interpreting
a self-pointing gesture as a CVG quoted from the matrix subject. Following work suggesting that
some indexicals can shift if the matrix subject’s perspective is prominent (Plank 1986, Anderson
2019), it is hypothesized that self-pointing gestures aligned with a third-person pronoun referring
to the matrix subject are acceptable in ID only if the matrix subject’s perspective is prominent on
the speech level. When the reporting speaker’s perspective is prominent in the ID utterance, by
contrast, self-pointing should not be acceptable as it then cannot be interpreted as a CVG quoted
from the matrix subject.

3. Experimental study.

3.1. METHOD.

3.1.1. PARTICIPANTS. Self-reported native speakers of German (n = 60) were recruited via Pro-
lific. All of them were naive with respect to the research question.

3.1.2. MATERIALS. Sixteen videotaped experimental items were constructed. Each experimen-
tal item consisted of an opening statement, followed by the target sentence rendered in ID. The
first sentence described how the protagonist, i.e., the matrix subject of the ID sentence, felt. The
subsequent ID utterance explained why they felt this way. Crucially, in the ID sentence, either

Proceedings of ELM 3: 423-434, 2025

Sebastian Walter:
Indirect discourse as mixed quotation? An experimental investigation. 429

https://journals.linguisticsociety.org/proceedings/index.php/ELM/issue/archive
https://www.elm-conference.net/


the matrix subject’s (cf. (9a)) or the speaker’s perspective (cf. (9b)) was made prominent (factor
PERSPECTIVE: matrix subject vs. speaker). This was tested in a pilot study. In addition, each
ID sentence contained a focalized third-person pronoun which was co-referent with the matrix
subject. This pronoun was either aligned with a self-pointing or a beat gesture (factor GESTURE:
self-pointing vs. beat). Thus, the study was of a 2x2 design. An example is given in (9):

(9) a. Matrix subject’s perspective: Pia ging es erbärmlich. Sie fragte sich, warum ihre beste
Freundin Anna, diese gottverdammte Saufziege, gestern Abend mal wieder IHR zu viel
Wein nachgeschüttet hat, obwohl sie doch so wenig verträgt. + SELF-POINTING/BEAT

‘Pia was feeling miserable. She wondered why her best friend Anna, that damned lush,
had poured HER too much wine again last night, even though she couldn’t handle it.’

b. Speaker’s perspective: Pia ging es erbärmlich. Sie fragte sich, warum ihre beste Fre-
undin Anna, die aber eigentlich nur die letzte Pfütze aus der Weinflasche loswerden
wollte, gestern Abend mal wieder IHR zu viel Wein nachgeschüttet hat, obwohl sie
doch so wenig verträgt. + SELF-POINTING/BEAT

‘Pia was feeling miserable. She wondered why her best friend Anna, who was just try-
ing to get rid of the last drops in the wine bottle, had poured HER too much wine again
last night, even though she couldn’t handle it.’

In (9a), the matrix subject’s perspective is prominent due to the appositive diese gottverdammte
Saufziege (‘that damned lush’). Under its most salient interpretation, the appositive conveys an
attitude toward Anna that is most likely Pia’s. It seems implausible to assume that this is the
speaker’s attitude in this context. By contrast, in (9b), the appositive die aber eigentlich nur die
letzte Pfütze aus der Weinflasche loswerden wollte (‘who was just trying to get rid of the last drops
in the wine bottle’) is most plausibly interpreted as additional information about the wine-pouring
scenario to which only the speaker has access, not Pia. This makes the speaker’s perspective
more prominent in this example. To distract participants from the experimental manipulations, the
experimental items were interspersed with 30 unrelated fillers.

3.1.3. PROCEDURE. Participants were first made familiar with the task by means of an intro-
ductory text. In this text, they were explicitly instructed to always pay attention to the audio and
the videotape. In addition, they were informed about their data protection rights and had to give
informed consent before starting to complete the questionnaire, which started with a training ses-
sion consisting of two items. SoSci Survey (Leiner 2022) was used to create the questionnaire, an
online platform which can be used free of charge for academic purposes. The questionnaire was
distributed via Prolific using its pre-filtering function to exclude participants from the aforemen-
tioned pilot study and to ensure that only native speakers of German participate in the experiment.
Each list was run independently in order to prevent participants from participating multiple times.
Again, this was achieved by using Prolific’s pre-filtering function.

The items were split up according to a Latin square design onto four lists. The 30 fillers and
the two training items were included on every list. Experimental items and fillers were randomized
for each participant. Their task was to rate the items on a 7-point Likert scale for acceptability (1
= completely unacceptable; 7 = completely acceptable). To check whether the participants paid
attention to the stimuli, they were asked attention questions about the videos (e.g., they were asked

Proceedings of ELM 3: 423-434, 2025

Sebastian Walter:
Indirect discourse as mixed quotation? An experimental investigation. 430

https://journals.linguisticsociety.org/proceedings/index.php/ELM/issue/archive
https://www.elm-conference.net/


what the color of the background in the videos was) at the end of the questionnaire.

3.2. PREDICTIONS. Based on the hypothesis that self-pointing gestures in ID can be interpreted
as CVGs quoted from the matrix subject when their perspective is prominent, it is assumed that
self-pointing will be rated significantly higher in the matrix subject condition of the factor PER-
SPECTIVE. Since beat gestures synchronize with the speech rhythm in general (McNeill 1992),
they are assumed to be highly acceptable regardless of the manipulation of the factor PERSPEC-
TIVE. Therefore, an interaction between the factors PERSPECTIVE and GESTURE is predicted.

3.3. RESULTS. Statistical analysis was done using the programming language R (R Core Team
2022) inside the integrated development environment RStudio (RStudio Posit Team 2023). For
data processing and visualization, the package ‘tidyverse’ (Wickham et al. 2019) was used. To test
for significant effects, the results were analyzed using a cumulative link mixed effects model with
the clmm() function in the R package ‘ordinal’ (Christensen 2023). Using effect coding, the two
factors GESTURE and PERSPECTIVE and all their interactions were entered as fixed effects into
the model, meaning that the intercept represents the unweighted grand mean and the fixed effects
compare the factor levels to each other. The analysis script as well as the materials are available
on osf: https://osf.io/h6rt5/.

Figure 3: Mean values and standard deviations for each condition. (Abbreviations: SP = Self-
pointing, MS = matrix subject’s perspective prominent, Speaker = speaker’s perspective prominent)

Figure 3 shows the mean values and standard deviations. The self-pointing condition was
rated equally well in both conditions of the factor PERSPECTIVE (matrix subject: M = 4.47, SD
= 1.95; speaker: M = 4.54, SD = 1.96). The same can be observed for the beat condition for
GESTURE (matrix subject: M = 5.04, SD = 1.68; speaker: M = 5.24, SD = 1.67). In general, it can
be observed that beat gestures were overall rated better than self-pointing gestures.

The output of the ordinal mixed-effects model is given in Table 1. It shows a main effect for
the factor GESTURE, confirming the tendency observed in the descriptive data shown in Figure
3: Beat gestures were rated significantly higher than self-pointing gestures, irrespective of the
manipulation of the factor PERSPECTIVE.

3.4. DISCUSSION. Contrary to the predicted interaction, the results only show a main effect for
GESTURE. Despite this contradiction, the results are still in line with the hypothesis: as evidenced

Proceedings of ELM 3: 423-434, 2025

Sebastian Walter:
Indirect discourse as mixed quotation? An experimental investigation. 431

https://journals.linguisticsociety.org/proceedings/index.php/ELM/issue/archive
https://www.elm-conference.net/


Estimate Std. Error z value Pr(> |z|)
Gesture -.757 .12 -6.212 5.24e-10 ***
Perspective .162 .118 1.375 .169
Gesture:Perspective -.181 .235 -.767 .443

Table 1: Ordinal mixed-effects model with Mode and Match as fixed effects and participants and
items as random intercepts. Formula: choiceO ∼ Gesture * Perspective + (1|CASE) + (1|item)
Significance codes: *** 0.001 | ** 0.01 | * 0.05 | . 0.1

by the descriptive data in Figure 3, self-pointing gestures are acceptable in ID. In a way, the results
of the study reported here even go beyond the initial hypothesis because not only are self-pointing
gestures in ID acceptable when the matrix subject’s perspective is prominent on the speech level,
but also when the speaker’s perspective is made prominent. As will be elaborated below, the results
are in line with the assumption that—in a similar vein as in FID (Ebert & Hinterwimmer 2022)—
a perspective shift can also take place in ID utterances and consequently, self-pointing gestures
aligned with third-person pronouns can be interpreted as CVGs quoted from the matrix subject.

4. General discussion and conclusion. The study reported by Ebert & Hinterwimmer (2022)
tested whether FID utterances are an instance of mixed quotation (cf. Maier 2015). Under the
assumption that quotation involves demonstration (Clark & Gerrig 1990, Davidson 2015), they
argue that a self-pointing gesture aligned with a third-person pronoun co-referent with the reported
speaker should be acceptable as it is then interpreted as a CVG quoted from them. The results
supported the mixed-quotational account of FID put forth in Maier (2015). As a control condition,
they also included ID utterances, hypothesizing that self-pointing should not be acceptable here,
since ID utterances are normally argued to not involve quotation. Surprisingly, however, the ID
condition was also rated surprisingly well, thus suggesting that some form of quotation is—or at
least can be—present in ID, as well.

The study reported in this paper aimed to further investigate this. Previous research has shown
that some indexical expressions can receive a shifted interpretation if the perspective of the ma-
trix subject is prominent in the surrounding discourse (Plank 1986, Anderson 2019). Given the
findings of Ebert & Hinterwimmer’s (2022), this shifted interpretation could be re-analyzed as an
instance of these expressions being quoted from the matrix subject, in line with their suggestions
for the self-pointing data. Therefore, the present study manipulated the perspective prominent on
the speech level (matrix subject’s perspective vs. speaker’s perspective), hypothesizing that self-
pointing interpreted as a quoted CVG is only acceptable in ID if the perspective of the matrix
subject is prominent on the speech level. The results go beyond this hypothesis, as self-pointing in
ID was found to be acceptable regardless of the perspective prominent on the speech level. This
suggests that mixed quotation is available in ID. This means that ID should receive a formal treat-
ment along the lines of the one proposed for FID by Maier (2015). However, as not all elements
can be quoted in ID, the quotation mechanism should be more constrained in ID in comparison to
FID, as it remains unclear for now which elements in ID can be quoted and under which circum-
stances. The results of a study reported elsewhere (Walter 2024) suggest that face emoji, which
normally receive an author-oriented interpretation by default (Grosz et al. 2023), can also receive

Proceedings of ELM 3: 423-434, 2025

Sebastian Walter:
Indirect discourse as mixed quotation? An experimental investigation. 432

https://journals.linguisticsociety.org/proceedings/index.php/ELM/issue/archive
https://www.elm-conference.net/


a shifted interpretation in ID. I argued that they should be analyzed as quoted facial expressions
from the matrix subject. For emoji to shift, however, the matrix subject’s perspective has to be
prominent, highlighting the importance for future research to further investigate the influence per-
spective prominence in ID has on the availability of shifted interpretations and thus the availability
of mixed quotation in ID.

References

Anand, Pranav & Andrew Nevins. 2004. Shifty operators in changing contexts. In Robert B.
Young (ed.), Proceedings of Semantics and Linguistic Theory (SALT) 14, 20–37. Ithaca, NY:
CLC Publications. 10.3765/salt.v14i0.2913.

Anderson, Carolyn J. 2019. Tomorrow isn’t always a day away. In M. Teresa Espinal, Elena Cas-
troviejo, Manuel Leonetti, Louise McNally & Cristina Real-Puigdollers (eds.), Proceedings
of Sinn und Bedeutung 23, 37–56. Barcelona, Spain: Universitat Autònoma de Barcelona.

Beattie, Geoffrey & Heather Shovelton. 2002. An experimental investigation of some properties
of individual iconic gestures that mediate their communicative power. British Journal of Psy-
chology 93(2). 179–192. 10.1075/gest.1.2.03bea.

Christensen, Rune H. B. 2023. ordinal—Regression Models for Ordinal Data. https://CRAN.
R-project.org/package=ordinal. R package version 2023.12-4.

Clark, Herbert H. & Richard J. Gerrig. 1990. Quotations as demonstrations. Language 66(4).
764–805. 10.2307/414729.

Davidson, Donald. 1979. Quotation. Theory and Decision 11(1). 27. 10.1007/bf00126690.
Davidson, Kathryn. 2015. Quotation, demonstration, and iconicity. Linguistics and Philosophy

38(6). 477–520. 10.1007/s10988-015-9180-1.
Ebert, Christian, Cornelia Ebert & Robin Hörnig. 2020. Demonstratives as dimension shifters. In

Proceedings of Sinn und Bedeutung 24, 161–178. Osnabrück: University of Osnabrück.
Ebert, Cornelia. 2024. Semantics of gesture. Annual Review of Linguistics 10(1). 169–189.
Ebert, Cornelia & Stefan Hinterwimmer. 2022. Free indirect discourse meets character viewpoint

gestures. In Sam Featherston, Robin Hörnig, Andreas Konietzko & Sophie von Wietersheim
(eds.), Proceedings of Linguistic Evidence 2020: Linguistic theory enriched by experimental
data, 333–349. Tübingen, Germany: University of Tübingen.

Eckardt, Regine. 2014. The semantics of free indirect disocurse: How texts allow to mind-read
and eavesdrop. Leiden, The Netherlands: Brill. 10.1163/9789004266735.

Farkas, Donka F. & Kim B. Bruce. 2010. On reacting to assertions and polar questions. Journal of
Semantics 27(1). 81–118. 10.1093/jos/ffp010.

Grosz, Patrick G., Gabriel Greenberg, Christian De Leon & Elsi Kaiser. 2023. A semantics of face
emoji in discourse. Linguistics and Philosophy 46(4). 905–957.

Harris, Jesse A. 2012. Processing perspectives. Amherst, MA: University of Massachusetts
Amherst dissertation.

Hinterwimmer, Stefan. 2017. Two kinds of perspective taking in narrative texts. In Dan Burgdorf,
Jacob Collard, Sireemas Maspong & Brynhildur Stefánsdóttir (eds.), Proccedings of Seman-
tics and Linguistic Theory (SALT) 27, 282–301. University Park, MD: University of Maryland.
10.3765/salt.v27i0.4153.

Proceedings of ELM 3: 423-434, 2025

Sebastian Walter:
Indirect discourse as mixed quotation? An experimental investigation. 433

https://journals.linguisticsociety.org/proceedings/index.php/ELM/issue/archive
https://www.elm-conference.net/


Kendon, Adam. 2004. Gesture: Visible action as utterance. Cambridge, UK: Cambridge Univer-
sity Press. 10.1017/CBO9780511807572.

Leiner, Daniel J. 2022. SoSci Survey (Version 3.3.14a). Available at https://www.soscisurvey.de.
Maier, Emar. 2015. Quotation and unquotation in free indirect discourse. Mind & Language 30(3).

345–373. 10.1111/mila.12083.
McNeill, David. 1992. Hand and Mind: What Gestures Reveal about Thought. Chicago, IL:

University of Chicago Press.
Parrill, Fey. 2010. Viewpoint in speech-gesture integration: Linguistic structure, discourse

structure, and event structure. Language and Cognitive Processes 25(5). 650–668.
10.1080/01690960903424248.

Plank, Frans. 1986. Über den Personenwechsel und den anderer deiktischer Kate-
gorien in indirekter Rede. Zeitschrift für germanistische Linguistik 14(3). 284–308.
10.1515/zfgl.1986.14.3.284.

Potts, Christopher. 2005. The logic of conventional implicatures. Oxford, UK: Oxford University
Press. 10.1093/acprof:oso/9780199273829.001.0001.

Quine, Willard V. 1940. Mathematical logic, vol. 4. Cambridge, MA: Haravrd University Press.
R Core Team. 2022. R: A Language and Environment for Statistical Computing. R Foundation for

Statistical Computing Vienna, Austria. https://www.R-project.org/.
RStudio Posit Team. 2023. RStudio: Integrated development environment for R. Posit Software,

PBC Boston, MA. http://www.posit.co/.
de Ruiter, Jan P. 2000. The production of gesture and speech. In David McNeill (ed.), Language

and gesture, Cambridge, UK: Cambridge University Press. 10.1017/cbo9780511620850.018.
Schlenker, Philippe. 2004. Context of thought and context of utterance: A note on free indirect

discourse and the historical present. Mind & Language 19(3). 279–304. 10.1111/j.1468-
0017.2004.00259.x.

Tonhauser, Judith. 2012. Diagnosing (not-)at-issue content. In Elizabeth Bogal-Allbritten (ed.),
Proceedings of Semantics of Under-represented Languages of the Americas (SULA) 6, 239–
254.

Walter, Sebastian. 2024. Can face emoji receive a shifted interpretation in indirect discourse?
Poster presented at Architechtures and Mechanisms of Language Processing (AMLaP) 30,
Edinburgh, UK.

Wickham, Hadley, Mara Averick, Jennifer Bryan, Winston Chang, Lucy D’Agostino McGowan,
Romain François, Garrett Grolemund, Alex Hayes, Lionel Henry, Jim Hester, Max Kuhn,
Thomas Lin Pedersen, Evan Miller, Stephan Milton Bache, Kirill Müller, Jeroen Ooms, David
Robinson, Dana Paige Seidel, Vitalie Spinu, Kohske Takahashi, Davis Vaughan, Claus Wilke,
Kara Woo & Hiroaki Yutani. 2019. Welcome to the tidyverse. Journal of Open Source Soft-
ware 4(43). 1686. 10.21105/joss.01686.

Proceedings of ELM 3: 423-434, 2025

Sebastian Walter:
Indirect discourse as mixed quotation? An experimental investigation. 434

https://journals.linguisticsociety.org/proceedings/index.php/ELM/issue/archive
https://www.elm-conference.net/