Hesitation markers in Sign Language of the Netherlands

(1)

Laura Spijker 10175679 MA Thesis General Linguistics

Supervisor: dr. Marloes Oomen

Second reader: dr. Roland Pfau

June 2020

Hesitation markers in Sign

Language of the Netherlands

(2)

Laura Spijker – MA Thesis – University of Amsterdam Hesitation markers in Sign Language of the Netherlands

1

Inhoud

Abstract ... 3

1. Introduction ... 3

2. Hesitation markers in spoken and sign languages ... 4

2.1. Hesitation markers in spoken languages ... 4

2.1.1. The form and meaning of hesitation markers ... 7

2.1.2. The place in the discourse where hesitation markers appear ... 8

2.2. Hesitation markers in sign languages ... 10

2.2.1. Functions of PALM-UP in NZSL ... 10

2.2.2. A note on terminology and modality ... 11

2.3. Research questions and predictions ... 12

3. Methodology ... 14

3.1. Data ... 14

3.1.1. Corpus NGT ... 14

3.1.2. Data selection ... 14

3.2. Analysis ... 16

3.2.1. The annotation values ... 16

3.2.2. Determining what is a hesitation marker ... 17

4. Results ... 18

4.1. The form of hesitation markers ... 18

4.1.1. The manual markers ... 19

4.1.2. The non-manual markers ... 24

4.2. The place in the discourse of hesitation markers ... 25

4.3. The relation between the form of the marker and its position ... 27

4.4. Interim summary ... 30

5. Discussion ... 31

5.1. The form of hesitation markers ... 31

5.1.1. Manual markers ... 31

5.1.2. Non-manual markers ... 33

5.2. The place of hesitation markers in the discourse ... 34

5.3. The debate: symptoms or signals ... 35

5.4. Methodological shortcomings ... 36

5.5. The value of this research ... 36

5.6. Future research ... 36

6. Conclusion ... 38

(3)

2

8. Appendix ... 41 8.1. Data dialogues ... 41 8.2. Data monologues ... 42

(4)

3

Abstract

This thesis focusses on hesitation markers in Sign language of the Netherlands (NGT). In spoken languages there are filled pauses (uh/um) and unfilled pauses (silences). The aim of this study is to provide information on the form of hesitation markers in NGT and the position where they appear in the discourse. 20 monologue and 20 dialogue movie clips from Corpus NGT were analysed. It is found hesitation markers in NGT can take different forms. All hesitations involve a non-manual marker, namely the change in the direction of the eye gaze, in most instances combined with a manual marker. These manual markers can be signs, holds or breaks. In monologues most manual markers are holds, whereas PALM-UP is used most in the dialogues.

There are other non-manual markers used, but those are less frequent. The most frequent position in the discourse where the markers appeared is at the beginning of the sentence, as is common for filled pauses in spoken languages. Thus, this study shows that in NGT all hesitations include at least one non-manual marker, namely the change in the direction of the eye gaze, and that, in most instances, this non-manual marker is accompanied by a manual marker.

1. Introduction

This thesis presents a first study on hesitation markers in Sign Language of the Netherlands (NGT). I became interested in hesitation markers during my preparation for a webinar on hesitation markers in conversations and interpreting situations (De Boer, 2020). This preparation consisted of reading an article on hesitation markers in English, German, and Dutch, written by De Leeuw (2007), describing significant differences found between those spoken languages. I thought it would be worth investigating how hesitation markers would manifest themselves in sign languages, specifically in NGT. Hesitation markers have scarcely been studied for sign languages. Furthermore, for NGT interpreters, like myself, it would be useful to understand what the properties of hesitation markers in NGT are, so that they can be used in interpreting situations.

The following chapter (2) discusses hesitation markers in spoken and sign languages. Section 2.1 focusses on these markers in spoken languages, followed by section 2.2 focussing on these markers in sign languages. Section 2.3 describes the research question, followed by the method in chapter 3. In chapter 4 the results from the analysis are given, which are discussed in chapter 5. Finally a conclusion is given in chapter 6.

(5)

4

2. Hesitation markers in spoken and sign languages

Languages in their most natural form are spontaneous. Speakers build their utterances while producing language. Consequently, there might be moments of hesitation in their speech when speakers are planning what to say next. This hesitation can manifest itself in different ways, for example through repetitions or corrections. In addition, speakers can use unfilled pauses (silences), filled pauses (eh, um) or filler words (well, I mean, stuff) to gain more time to organise their utterance (Gilquin, 2008). These three categories can be subsumed under the label ‘hesitation marker’ and will be the focus of this thesis.

An unfilled pause, or silent/empty pause, is seen as a silence with an unusual length (Maclay and Osgood, 1959). A filled pause (uh) on the other hand, is defined as “a semantically empty element of speech which fits a language-specific conventional phonetic form and delays (either intentionally or not) the transfer of the speaker’s message” (Rose, 2007). Filler words, also known as smallwords, are multifunctional and can be used to mark hesitation (Guilquin, 2008). They can be used to fill a pause when a speaker thinks about what to say next, to fill knowledge or lexical gaps in the sentence, or to fulfill other pragmatic and non-pragmatic functions. These hesitation markers can occur on their own, but can also appear in combination with repairs. Levelt (1983), investigating monitoring and self-repair in speech, describes utterances with repairs as consisting of three phases: the original utterance, the editing phase and the repair. When detecting errors in their utterances, which can be lexical or syntactic, it causes speakers to rephrase their utterance. Before such a repair, speakers are likely to add filler words like uh to their utterances.

2.1. Hesitation markers in spoken languages

This section first describes the various theories on the function of hesitation markers in spoken languages discourse. Later in section 2.1.1, a more in-dept overview is given on the form and meaning of the markers, followed by a discussion of the place in the discourse where the hesitation markers appear (2.1.2).

A lot of research has been done on hesitation markers in spoken languages. There are various theories on the function of hesitation markers in spoken language discourses. As for filled pauses, Clark and Fox Tree (2002, p.75) describe three possible views on this phenomenon. Under the first view, these markers are seen as symptoms of a planning problem while speaking. Under the second and third view, these markers are non-linguistic and linguistic, respectively, and have a signaling function for dealing with problems in speaking.

(6)

5

The first view is first described in research from the late 50s by Maclay and Osgood (1959), who theorized that hesitation markers have the function of creating time for planning the message. This perspective is present in most earlier research and is called the symptom hypothesis. In this view, hesitation markers are linked to the speakers’ cognition, without the speaker being in control over their use. They are seen as an indication of an ongoing planning process, where the speaker gains time, while thinking about what to say next (Goldman-Eisler, 1968:26, cited in De Leeuw, 2007). Thus, the markers are seen as elements that mark disfluency (Wieling, Grieve, Bouma, Fruehwald, Colemand & Liberman, 2016, p. 201).

The signal hypothesis, which underlies the second and third view, states that fillers are signals with different functions. Under the second view, the filler is seen as a nonlinguistic signal. The best known function is that speakers can use fillers to show they are not finished speaking and want to keep their turn (Maclay & Osgood, 1959). However, speakers can also use these hesitation markers to give up their turn (De Leeuw, 2007, p. 86) Furthermore, it is said that these hesitation markers can help listeners predict upcoming discourse boundaries (De Leeuw, 2007).

The third view, originally proposed by James (1972), regards these fillers as linguistic interjections. He stated that fillers such as uh and um should be grouped with words like oh, well and say. Clark and Fox Tree (2002) also come to this conclusion in their article on uh and um in spontaneous speaking in English. They state speakers are able to control their use of hesitation markers, and can use them to signal an upcoming delay.

The article by Clark and Fox Tree (2002) consists of two parts, the first part focusses on uh and um as conventional English words and their meaning, while the second part focusses on the planning and production of these two words. Their evidence for the first proposal is based mainly on the London-Lund corpus (LL), a British corpus that consists of 170.000 English words in 50 face-to-face conversations, recorded between 1961 and 1976. They found 3904 fillers (uh, um, and their prologued versions u:h and u:m) in total (898 times uh, 1213 times u:h, 530 times um, and 1263 times u:m). Furthermore, they used a British answering machine corpus (AM), the North American switchboard corpus (SW) and the Pear stories (Pear) for auxiliary analysis. In the AM corpus 319 fillers were found, namely 69 uh, 166 u:h, 6 um, and 78 u:m. The SW corpus included 79.623 fillers, but only uh (67.065) and um (12.558) were found. The Pear corpus was based on narratives, containing only 282 fillers, 57 times uh, 68 times u:h, 54 times um, and 103 times u:m. Trained coders had to listen to the audio recordings of the LL, AM and Pear corpora and judge the pause length and prolongation of the fillers. The

(7)

6

LL corpus is used as main source of evidence because it includes enough data (fillers, pauses and prolongations) for a strong analysis.

Clark and Fox Tree (2002) found evidence which supports the claim that these fillers should be treated as conventional English words: they consist of segments present in standard English syllables, and they follow English prosody and syntax. As is found in their analysis, uh and um are added when speakers want to initiate an delay, and do not cooperate in syntactic structures, which is similar to the behaviour of other interjections. In addition, basic meanings of uh and um are conventional in British and North American English. They are signals, where uh is used initiating a minor delay in speaking and um a major delay. Also there might be variation across dialects of the English language, suggesting that speakers learn the form of the filler that fits their dialect.

This third view is criticized by O’Connell and Kowal (2005). They analysed media interviews of Hillary Clinton on the presence of uh and um, and silent pauses. They used PRAAT software (www.praat.org) to measure the fillers and silent pauses. In their analysis they found that in most instances neither uh or um was followed by a delay. Furthermore, in 85% of the cases when the filler was followed by a pause, the duration of the pause after both uh and um were similar. Clark and Fox Tree (2002) stated these fillers could be differentiated by the duration of the silent pauses following the filler. However, the results of the analysis by O’Connell and Kowal (2005) did not show this difference.

Furthermore, O’Connell and Kowal (2005) add that the means presented in Clark and Fox Tree’s Figure 2 (2002, p. 83) are shorter than the minimum Clark and Fox Tree give for the smallest possible perceptual unit of silent pauses. In addition, the results of Clark and Fox Tree are based on the perception of coders and not measured. Consequently, O’Connell and Kowal (2005) state that the duration of the silent pauses following the filler cannot be used to differentiate the fillers. Moreover, O’Connell and Kowal (2005) found that most of the fillers are not followed by a silent pause/delay, which leads them to conclude that the fillers are not used to signal a delay. However, there is a possibility that fillers have a signaling function for listeners, and are additionally a symptom of speakers’ cognitive processes (De Leeuw, 2007). This would explain the seemingly contradictive results found in diverse studies.

In another section of their research, again contradicting Clark and Fox Tree (2002), O’Connell and Kowal (2005) wrote that the hypothesis that ‘uh and um are interjections’, is unlikely, supporting their claim with the following arguments: fillers are not present in cited speech, they do not signal emotion, and they are not part of the turn itself. The authors studied interjections in interviews and found that these interjections themselves can be used to take

(8)

7

turns, which is not found with fillers. Therefore, the authors argue that uh and um cannot be seen as interjections, but have to be viewed as fillers.

Thus, there are three approaches on hesitation markers with contrasting views on the function and linguistic status of these markers. These approaches display the two main perspectives on the function of hesitation markers: they are said to be either symptoms or signals. However, there is a third possibility, where fillers could function as a signal for listeners, and simultaneously as a symptom for speakers (De Leeuw, 2007).

2.1.1. The form and meaning of hesitation markers

As mentioned earlier, speakers can use pauses, filled pauses or filler words to mark hesitation. This section focusses on the form of filled pauses across spoken languages. The hesitation markers that can be used as filled pauses are said to be language specific (Levelt, 1983; Clark & Fox Tree, 2002; De Leeuw, 2007). However, many of the markers are found to have similar properties. According to Clark and Fox Tree (2002, p. 92) the fillers that are most common “tend to be brief and built around central vowels in the language” and “are easy to prolong”. These properties are present in the fillers found in Table 1 below. This table shows the fillers found in several languages, with the spelling of the filler based on the that of the source.

Table 1. Fillers found in several languages, spelling from the sources is used (Clark & Fox Tree,

2002).

According to Levelt (1983), these variations of uh and um can be explained by the neutral position of the oral cavity that is assumed when these fillers are articulated. This is also shown in other research, for instance in Maclay and Osgood (1959), who found multiple alternatives for the pronunciation of the English filler, namely [ԑ, ӕ, r, ə, m], with [ə] being the most frequent. If we would argue that uh and um are conventional, these variations occurring in the English fillers is expected (Clark & Fox Tree, 2002). The filler being conventional would mean that speakers would learn the variant of the filler, fitting to their dialect.

(9)

8

There are various studies that have focussed on the semantic/pragmatic difference between uh and um. Both fillers have a general meaning of initiating a delay, but there is debate about whether these fillers have different specific meanings or not. Clark and Fox Tree (2002) say that uh shows a brief delay in English speech, while um shows a longer delay. This is contradicted by O’Connell and Kowal (2005), who did not find evidence for a meaning difference, as pointed out before. However, they did detect that the duration of the filler itself is longer, due to the fact that um consists of two phonemes and uh only of one.

It appears that the preference for uh versus um may be language-specific. De Leeuw (2007) and Wieling et al. (2016) conducted research on fillers in Germanic languages. Uh is articulated as a neutral vowel in an open syllable, whereas um is articulated as a neutral vowel followed by a final labial nasal (Wieling et al., 2016). Both studies found that Dutch people prefer the vocalic uh, whereas English and German speakers show a preference for a vocalic-nasal marker, like um. When comparing speakers from these three languages, German speakers seem to use nasal markers more frequently than speakers of the other two languages. Wieling et al. (2016) even found there is an ongoing language change of um rising in frequency over uh. At the moment um is mostly used by younger speakers and women, with other groups prefering uh, but there is a change currently taking place in the usage of these markers across different Germanic languages. Evidence for this claim becomes visible over time, as is common in sociolinguistic research. Nevertheless, in different studies that do longitudinal analysis on corpora of other languages (Philadelphian English, Norwegian and to a lesser extent, Dutch) there was found evidence on the rise of um.

Most spoken languages have two or more fillers that can mark hesitation. These markers are language specific, but often have similar properties: they have a tendency of being brief, are built around central vowels in the language, and can easily be prolonged.

2.1.2. The place in the discourse where hesitation markers appear

This section focusses on the place in the discourse where we can find hesitation markers. Clark and Fox Tree (2002) describe three possible locations for filled pauses, based on intonation units. An intonation unit is defined as “a stretch of speech under a single intonation contour” (Clark & Fox Tree, 2002, p. 94). These units could encompass a single word, but also cover whole sentences. The first possible location for filled pauses is at the boundary of the intonation unit, the second is after the first word and the third possibility is at any other later location, with the planning process decreasing in difficulty with each location. In the London-Lund corpus (LL) they used for their research, Clark and Fox Tree (2002) computed the frequency of these

(10)

9

three possible locations for filled pauses. Their results show that the greater part of the attested filled pauses occurred at the boundaries of intonation units. Specifically, for each 1000 possible locations for filled pauses, they found that most filled pauses (43 per 1000 possibilities) occurred at the boundary of the intonation unit. The next most frequent location was after the first word (27 per 1000), while instances at any position later than the second word were least frequent (13 per 1000).

Earlier research by Maclay and Osgood (1959) focused on filled vs unfilled pauses and their places of occurrence. They found that both types of pauses occur more frequently before lexical words than before function words although filled pauses are more frequently placed before function words. According to the findings, filled pauses are often used inbetween larger syntactical units and at phrase boundaries. At these places, the speaker makes decisions that can both be structural and content-related, therefore these filled pauses occur both before lexical and function words. Unfilled pauses occur more frequently at word boundaries within phrases (Maclay & Osgood, 1959).

Finally, De Leeuw (2007) also examined the relation between the form of the hesitation marker and its position in the discourse for spoken English, Dutch and German. In her article, she mentions an earlier finding by Swerts et al. (1996) that in Dutch, the vocalic-nasal hesitation marker, um, is used more frequently at major discourse boundaries and that these markers often involve unfilled pauses at both sides. These hesitation markers surrounded by silences are seen as reflecting major delays in speech. Hesitation markers surrounded by words on the other hand, are seen as minor delays (De Leeuw, 2007). For English, De Leeuw (2007) found that there is a preference for the vocalic-nasal um in all positions. Dutch speakers show a preference for using a vocalic filler (eh) directly after words, without a silence between the word and the filler (van Donzel et al, 1996 as cited in De Leeuw, 2007). In sentences where the filler is placed between silences, they show a preference for vocalic-nasal fillers like um. German speakers have nasal dominance for fillers surrounded by silences. Thus, these Germanic languages show different preferences for the two articulations of the fillers and their position in the discourse.

In conclusion, filled pauses tend to occur at the boundaries of the phrases or the intonation units, while unfilled pauses are found more often at word boundaries within phrases. Research on the manifestation of hesitation markers in three Germanic languages (Dutch, English and German) shows that the preference for a particular form of the filled pause can be language-specific.

(11)

10 2.2. Hesitation markers in sign languages

The form and function of hesitation markers is not a subject that has been researched much for sign languages. Nevertheless, in an article by McKee and Wallingford (2011), which describes the different forms and functions of PALM-UP in New Zealand Sign Language (NZSL), the

marking of hesitation is named as one of various discourse functions of PALM-UP in this sign language. This section focusses on hesitation markers in sign language. In section 2.2.1, the functions of PALM-UP in NZSL are described, followed by a note on the terminology and modality in section 2.2.2.

2.2.1. Functions of PALM-UP in NZSL

A PALM-UP (Figure 1a) can be used to signal the beginning or the end of a conversational turn in NZSL (McKee & Wallingford, 2011). When articulated more fiercely at the beginning of a sentence, it is claimed to be similar to a speaker starting his sentence with um, ah or well. When

PALM-UP is followed by the hesitation g:finger-wiggle (turning the hand with the palm facing up, and wiggling the fingers, figure 1b), these movements are argued to have the function to hold the signing floor while speakers are formulating their message. This way, the marker serves a discourse management function (McKee & Wallingford, 2011).

(a) (b)

Figure 1. PALM-UP (a) on the left and PALM-UP+g:finger-wiggle (b) on the right.

Both for American Sign Language (ASL) and NGT, it has previously been claimed that the one or two-handed PALM-UP can also be used to indicate the start or end of a conversational turn

(Cokely & Baker-Shenk, 1991, as cited in McKee & Wallingford, 2011; Van Loon, 2012; Oomen & Pfau, 2017). Therefore, it could be possible that PALM-UP can function as a hesitation

(12)

11

The article by McKee and Wallingford (2011) contains a table with mouthings that co-occur with PALM-UP. Only in six out of 104 tokens represented in the table, PALM-UP was found to be combined with the mouthings and, um, or well, which are analyzed as filled pauses in spoken languages (McKee & Wallingford 2011, p. 238). In such cases, it is evident that PALM

-UP marks a hesitation. However, that does not necessarily mean PALM-UP can only function as a hesitation marker when it is combined with a filler word, but it can also take multiple functions at the same time.

Overall, the study by McKee and Wallingford (2011) shows PALM-UP can be used with

a lot of different functions, of which the function of hesitation marker is one. Whether or not

PALM-UP may also fulfil this function in NGT, is one of the questions this thesis aims to answer.

2.2.2. A note on terminology and modality

In most research on hesitation markers in spoken languages, the terms filled and unfilled pauses are used. These terms describe pauses with or without a verbal utterance: unfilled pauses are silent intervals within the speakers’ utterances, and filled pauses are intervals, occupied with a non-lexical speech element (Esposito, McCullough and Quek, 2002). However, in order to describe hesitation markers in sign languages, I argue other terminology should be used. Motivation for this decision comes from previous research on disfluencies in gesture.

Esposito, McCullough and Quek (2002) investigated whether augmented holds and simple holds in gesture are gestural correlations to filled and unfilled pauses in speech. With hold, the abrupt end of the movement of the hand is meant. This usually occurs at the end of the sign, when the signer holds his/her hand at the same location (De Nobel & Harder, 1991).

Augmented holds are defined as holds with a detectable motion in the holding hand, while simple holds are defined as having “an active configured gestural state without a detected intended motion” (Esposito et al., 2002, p.1). The authors analysed speech and co-speech gestures of speakers of English, and studied the use and locations of filled and unfilled pauses. They found that speakers use pauses equally often in gesture and in speech, with equally common use of unfilled pauses and simple holds, and filled pauses and augmented holds. In addition, the duration of filled pauses and augmented holds, as well as the duration of unfilled pauses and simple holds, were comparable. Moreover, unfilled pauses and simple holds show similar behaviour in terms of them being mostly used to signal the end of a sentence or to give up the turn. Based on these observations, it might seem that filled pauses in speech are indeed comparable in function to augmented holds in gesture. However, they show a difference in patterning, since both filled pauses and augmented holds appear at the location in the sentence

(13)

12

where there is a word/sign-search problem or a conceptualization difficulty (Esposito et al., 2002).

For the signed modality, the terms filled and unfilled pauses seem inappropriate when considering non-manual elements present in sign. Unfilled pauses in speech might seem ‘unfilled’ when only focussing on the vocal articulators, but these pauses in spoken languages may also involve a manual or nonmanual marking. This is why a distinction between filled and unfilled pauses in sign languages might be difficult to make. Furthermore, in the manual-visual modality of sign languages, unfilled pauses could also include nonmanual markers.

Sign languages consist of both manual and non-manual elements. To be able to describe hesitation markers in Sign Language of the Netherlands, there is a need for terms that can describe both manual and non-manual articulations. Therefore, the terms I will use to describe hesitation markers in sign languages are simply manual and non-manual hesitation markers. Manual hesitation markers include all articulations of manual signs or gestures to show hesitation. Non-manual hesitation markers cover all non-manual expressions, for instance a change in the signer’s eye gaze, the use of mouthings and a change in the position of the head or upper body.

2.3. Research questions and predictions

Almost all research previously conducted related to hesitation markers is based on spoken languages. For NGT interpreters, there is no information on how to translate hesitations from a spoken language to NGT in the most natural way. Furthermore, for linguists trying to describe the NGT, there is no information on the form of hesitations and where they occur in the discourse. Therefore, it is important to investigate where hesitation markers in a sign language appear and what the forms of these hesitation markers are. Research on spoken language hesitation markers focusses mainly on phonological variants of uh and um, which do not have a one-to-one equivalent in sign languages. Therefore, we may expect that differences in the form of hesitation markers in sign languages arise due to modality. Another possibility is that the form of the markers found in monologues and dialogues in Sign Language of the Netherlands are influenced by spoken languages, for instance in the use of a mouthing.

(14)

13

In order to learn more about the marking of hesitation in the signed modality, the aim of this study is to describe hesitation markers in (semi-)spontaneous corpus data in Sign Language of the Netherlands. The main research question is therefore: What are the properties of hesitation markers in NGT? To answer this main research question, three subquestions are asked:

• What are the forms of hesitation markers in NGT?

• At which point(s) in signed discourse do hesitation markers appear?

• Is there a relation between the form of hesitation markers and their position in the discourse?

(15)

14

3. Methodology

In order to provide an answer to the main question of what properties hesitation markers have in NGT, this study analyses naturalistic corpus data from the Corpus NGT. Section 3.1 focusses on how the data was selected. Section 3.2 focusses on the procedure for the analysis of the data.

3.1. Data

In this section the selection of the data is described. Section 3.1.1 focusses on the description of the Corpus NGT, from which the data were extracted. The following section 3.1.2 focusses on the data selection procedure.

3.1.1. Corpus NGT

In order to investigate the phenomenon of hesitation in NGT, data was needed of spontaneous signing. This study on hesitation markers in NGT therefore uses movie clips from the Corpus NGT, which includes (semi-)spontaneous signing. This corpus is an open access database consisting of over 2000 video clips, with a total of 92 different deaf signers, aged between 17 and 84 years old, who have Sign Language of the Netherlands as their native language (Crasborn, Zwitserlood, & Ros, 2008; Crasborn & Zwitserlood, 2008). The database consists of monologues and dialogues in NGT, in which participants talk about different subjects, for instance by reciting picture books or comics in monologues or by discussing deaf issues in dialogues.

3.1.2. Data selection

In this section the selection of the participants is explained first (3.1.2.1), followed by a description of the selection of the movie clips (3.1.2.2).

3.1.2.1. Participants

This research is based on 40 movie clips from the Corpus NGT with 20 different signers, with each signer participating in two different settings, once in a monologue and once in a dialogue. To ensure a diverse and representative selection of participants, gender, age, and region of the signers was taken into account. Therefore, two groups were formed, one with ten males and one with ten females. Within these groups an equal distribution of age was chosen. Furthermore, the regional variation of the signers NGT was taken into account. Before there was a standardised NGT, there were five dialects, arisen around the five schools for the deaf in the Netherlands (Amsterdam, Groningen, Rotterdam, Sint-Michielsgestel and Voorburg). These dialect were taken into account by selecting participants from all five dialects and the types

(16)

15

mixed and other. The participants had to select their own dialect/region when participating in the Corpus project. The difference between these two categories is not clear, however I think the category mixed means that the signers selected multiple regions, whereas the category other means they selected the option ‘other’, meaning they do not think they use one distinct category.

3.1.2.2. Movie clips

In the selection of the sessions, only monologues with content type ‘Tweety & Sylvester’ and dialogues with content type ‘discussions’ were considered. Tweety and Sylvester are two cartoon characters, Tweety being a bird and Sylvester a cat. The characters appear in a cartoon called ‘Canary Row’, a 1949 Warner Bros short movie. This cartoon is widely used in linguistic research as a test stimulus and therefore included in this corpus. Signers have to watch the cartoon and are then asked to retell the story to their signing partners.

For each chosen participant, the first clip of the content type in which the participant signs for at least 55 seconds was selected. In total, 68 minutes of material was analysed. This led to the data selection listed in Table 2.

Table 2. Selected data

Nr. Participant Sex Age-group Signing region Monologue Dialogue

1 S073 Female 17 – 39 YO Amsterdam 1775 1801 2 S008 Female 17 – 39 YO Other 0122 0128 3 S065 Female 17 – 39 YO St. Michielsgestel 1538 1551 4 S059 Female 40 – 62 YO Rotterdam 1366 1383 5 S088 Female 40 – 62 YO Voorburg 2196 2220 6 S041 Female 40 – 62 YO Groningen 0912 0922 7 S003 Female 40 – 62 YO Amsterdam 0024 0015 8 S060 Female 63 – 84 YO Rotterdam 1365 1389 9 S061 Female 63 – 84 YO Voorburg 1421 1435 10 S018 Female 63 – 84 YO Groningen 0289 0294 11 S029 Male 17 – 39 YO Groningen 0605 0608 12 S077 Male 17 – 39 YO Mixed 1899 1914 13 S080 Male 17 – 39 YO Voorburg 1953 1991 14 S069 Male 40 – 62 YO Mixed 1664 1682 15 S085 Male 40 – 62 YO Amsterdam 2145 2164 16 S071 Male 40 – 62 YO Amsterdam 1716 1731 17 S037 Male 40 – 62 YO Groningen 0759 0766 18 S042 Male 63 – 84 YO Groningen 0913 0921 19 S043 Male 63 – 84 YO St. Michielsgestel 0966 0970 20 S045 Male 63 – 84 YO Voorburg 1016 1019

(17)

16 3.2. Analysis

This section focusses on the analysis of the data in the movie clips. Section 3.2.1 describes the annotation values used for the analysis and in section 3.2.2, the guidelines for determining hesitations are explained.

For the analysis of the data, the chosen sessions have been annotated in ELAN, an

annotation tool which can be used for audio and video recordings

(https://archive.mpi.nl/tla/elan). Thirteen annotation tiers were newly created for the purposes of this study. One tier is for general comments (Remarks) and the other twelve are divided into six for signer 1 and six for signer 2. More specifically for each participant, there is one tier for manual hesitation markers, and five for nonmanual hesitation markers, namely the mouth, the eyes, the head, the eyebrows and the upper body.

3.2.1. The annotation values

For the manual hesitation markers, the manual movements of the signers are annotated on the manual tier (Manual S1/Manual S2). When a sign is produced during the hesitation, this is glossed on the tier. When there is a hold, this is annotated as hold, combined with the specification of the type of hold, i.e. augmented, simple or reduplication.

On the tiers for the mouth (Mouth S1/Mouth S2), the used values are closed, open, open-closed, eh, ehm, of and in one instance the mouthing ‘Aap. Poes’ (‘Monkey. Cat’) was used. The annotations used on the tiers for the eyes (Eyes S1/Eyes S2) have the values closed, up, down, away from addressee, left, right and gaze (a stare at a location besides or behind the addressee). The head movements are written on the tiers for the head (Head S1/Head S2) and may have any of the following values: nod, shake, tilt, down, left and right. The glosses on the tiers for the eyebrows (Eyebrows S1/Eyebrows S2) only have the value frown. For the movements of the upper body there is also a tier for both speakers (Upper body S1/Upper body S2). The glosses that are used are shrug, backwards lean and left-right. When no nonmanual action could be observed, no annotation was made on the relevant tier(s).

All video clips were watched in full, and each instance of a hesitation marker was annotated on the tiers. The annotations, including the time-mark, were then compiled into an excel sheet.

In a second round of analysis, the position in the discourse at which the marker occurs in the discourse was determined and added to the excel sheet. There are nine possible values. The first two are at the beginning of a sentence unit, namely: beginning of the sentence and start of turn. Then there are four for a position in the middle of the sentence, namely: after first

(18)

17

word, after second word, during enumeration and following other hesitation. The category ‘during an enumeration’ means that the signer hesitates while signing about something and using an enumeration. Right before or after touching the hand marked for enumeration, the signer hesitates. The category ‘boundary intonation unit’ encompasses the hesitations that arise at the boundary of intonation units. An intonation unit is: ‘a stretch of speech under a single intonation contour’ and it can include a small constituent like a single word, but also whole sentences or clauses (Clark & Fox Tree, 2002, p. 93-94). For a position at the end of the sentence, there is just one value: end of sentence.

3.2.2. Determining what is a hesitation marker

When looking at the movie clips and deciding whether something is a hesitation marker or not, there are some factors that need to be taken into account. Generally, all moments where ongoing utterances were interrupted were considered to be hesitations. At the beginning of a turn, signers often look up or down, followed by a nod. This has been interpreted as a hesitation at the start of the turn. In the middle of sentences, it often happens that the utterance is obstructed by a pause and an eye movement, sometimes combined with a manual sign or hold. Such instances were also viewed as hesitation markers.

Sometimes, it is difficult to decide whether a hesitation comes at the beginning or end of the sentence. In such cases, the time between the last sign of the sentence and the hesitation as well as the time between the hesitation and the first sign of the following sentence were measured. The position of the hesitation was then determined relative to the sentence closest to the hesitation.

(19)

18

4. Results

A total of 40 videoclips were analysed, of which 20 are monologues and 20 are dialogues. As can be observed in Table 3, the total signing time of the monologues is 25 minutes and 30 seconds. The total signing time of the dialogues is 42 minutes and 21 seconds. The total number of hesitations is 40 for the monologues, with a mean of 1,6 hesitation per minute, and for the dialogues it is 49 hesitations, with a mean of 1,2 hesitation per minute.

Table 3. Signing time, number of hesitations and the mean of hesitations per minute in the

analysed monologues and dialogues.

Type Signing time Number of hesitations Hesitations per minute

Monologue 25 min, 30 sec 40 1,6

Dialogue 42 min, 21sec 49 1,2

As can be seen in Table 4, not all movie clips included hesitations. In the movie clips with monologues, twelve out of 20 movie clips included hesitations, while for the dialogues eleven out of 20 movie clips included hesitations.

Table 4. Movie clips with and without hesitations.

Type Total With hesitations Without hesitations

Monologue 20 12 8

Dialogue 20 11 9

In section 4.1, the form of the hesitation markers is described. The results regarding the place in the discourse where the hesitation markers appear are described in section 4.2, followed by a section on the relation between the form of the marker and its position in the discourse in 4.3. This chapter ends with an interim summary in section 4.4.

4.1. The form of hesitation markers

This paragraph describes the results relating to the form of the hesitation markers. The hesitations in the analysed movie clips are all marked by a change in the direction of eye gaze. This change is sometimes accompanied by other non-manuals such as a mouthing or a movement of the head. In most instances, the hesitations are also accompanied by manual markers. In section 4.1.1 the form of the manual markers is discussed, followed by a discussion of the non-manual markers in section 4.1.2.

(20)

19 4.1.1. The manual markers

As can be seen in Table 5 below, manual markers can take different forms. The manual markers found in the analysed data can be classified into three types, namely sign, hold and break. The category sign includes signs and gestures. The category hold means that the movement of the hand comes to an abrupt end, usually at the end of the sign (De Nobel & Harder, 1991). The category break is similar to hold, meaning there is no intended movement. Instead of holding a sign, as the signer does during the hold, the signer holds the hands during the break in a resting position. The hands can be lifted during the break, but in most cases, they are folded or placed on the speakers’ lap. An example of a break with the hands lifted can be seen in Figure 2.

Figure 2. Break with the hands lifted.

In monologues, most of the manual markers (N=25) are holds, nine are signs or gestures, and four manual markers are breaks. In the dialogues, most of the manual markers are signs (N=26), 18 are holds and four are breaks. In the monologues, there are two instances of a hesitation without a manual marker and in dialogues there is just one hesitation that does not involve a manual marker.

Table 5. Frequencies of different types of manual markers found in monologues and dialogues.

Type Sign Hold Break No manual marker

Monologue 9 25 4 2

Dialogue 26 18 4 1

Signs used as manual markers of hesitation are PALM-UP, PALM-UP:wiggle, INDEX and DAARNA

(AFTER). The sign DAARNA is a time-related sign, here used as a marker for hesitation. In

monologues, most variation is found: there is one PALM-UP, three instances of PALM-UP with a finger wiggle, four times a variation of INDEX and once the sign DAARNA is used. In the example

(21)

20

with DAARNA,the signer tries to recall an event occurring in the cartoon she is retelling. While thinking, she used this sign, which is therefore labelled as hesitation marker. In the dialogues there is less variation. Only two types of signs are used, namely 24 instances of PALM-UP and two variations of the INDEX sign.

As has been noted before, different forms of PALM-UP are found in the data. Some instances of PALM-UP are articulated as a two-handed sign, some only with one hand. Most

instances are articulated without a movement of the fingers, but finger wiggle has also been found. Another attested form of PALM-UP found in the movie clips, is a sign where a handshape

change occurred from an open B-hand, to a pointy handshape, with the index finger and thumb extended and the other three fingers closed. However, most signs attested were articulated with lax handshapes.

In Figure 3, two different forms of PALM-UP can be found. The signer in the left picture

(3a) articulates the sign with two hands and combines this manual marker with a shoulder shrug, closed mouth and downwards eye gaze. The signer on the right (3b) signs a one handed PALM -UP with a lax handshape. She closes her mouth and her eyes look up, and combines these

(non)manual markers with a head nod.

(a) (b)

Figure 3. Two forms of PALM-UP. On the left (a) the sign is two-handed and on the right (b) it

(22)

21

For holds, three types are distinguished: augmented holds, simple holds and reduplications (Esposito, McCullough and Quek, 2002). Augmented holds involve a detectable motion in the ‘holding hand’, whereas simple holds do not have a detectable movement. The difference between these holds can be seen in Figure 4 and 5. Figure 4 shows an augmented hold (total duration of the hold is 242ms). The first frame is taken 10ms before the hesitation starts and the last frame 10ms after the hesitation ended. As can be seen in this figure, there is a motion in both hands of the signer.

Figure 4. Example of an augmented hold, with a detectable motion in the holding hand.

Figure 5 shows an example of a simple hold. Again, the first frame is taken at 10ms before the start of the hesitation and the last frame at 10ms after the hesitation. Here, both hands stay at the same location during the hold (total duration of the sign is 520ms). In the first frame, the signer looks to the addressee and in frame two the start of the change in the direction of the eye gaze away from the addressee, to the right side of the signer can be seen. In frame three to five, a movement of the head to the right can be seen.

Figure 5. Example of a simple hold, without a motion in the hands.

Reduplicated holds involve reduplication of the held sign. In the data, the reduplicated hold only occurs twice. Both instances occur in dialogues, at the beginning of the sentence. One instance is barely visible, since the movement is repeated fast. In this case the signer makes the sign INDEX1 and touches his chest twice. The other instance is better visible in the video stills.

(23)

22

movement from open to closed fingers occurs twice. This example of a reduplicated hold can be seen in Figure 6, again with the first frame 10ms before the hesitation and the last frame 10ms after. In Frame one, three and five, the fingers of the right hand are open, while they are closed in frame two and four.

Figure 6. Example of a reduplicated hold.

In Table 6, the distribution of the different types of holds attested in the data is shown. More holds were found in the monologues (N=25) than in the dialogues (N=18). The number of augmented and simple holds found in the monologues are almost the same, with twelve simple holds and only one augmented hold more, (N=13). In the dialogues the number of augmented and simple holds is equal, they both occur eight times in dialogues. The reduplicated holds are only found in the dialogues (N=2).

Table 6. Distribution of the three types of holds (augmented holds, simple holds and

reduplicated signs) in the monologues and dialogues.

Type Augmented Simple Reduplication

Monologue 13 12 0

Dialogue 8 8 2

In the movie clips, there are not many instances found in which a hesitation combines with a break. In both monologues and dialogues, there are only four instances where there is a pause of the hands to mark a hesitation. In one instance, the hands are lifted during the break and seven times the hands are folded or placed on the speakers’ lap. An example of a break with the hands on the speakers’ lap can be found in Figure 7 below. During this break, the speaker uses this moment of thinking what to say next for giving his arms a rest.

(24)

23

Figure 7. A break with the hands on the speakers’ lap.

Finally, in the monologues there are also two instances where the non-manual hesitation markers are not accompanied by a manual marker. For both monologues, the hand of the signer is in transition from the handshape and place of articulation of one sign to another during the hesitation, therefore I did not annotate them as manual markers. For the dialogues, only one instance was found, but in this case the speaker touches his/her own nose.

In conclusion, there are three types of manual markers possible for marking hesitation. Signers can use a sign to mark a hesitation, in most cases this is a form of PALM-UP, which can be one or two-handed, and could be articulated with or without a movement. Especially in dialogues, PALM-UP is used frequently. Another possible sign that can be used, and is more frequently found in the monologue video clips, is the INDEX sign. Furthermore there was one instance found of the sign DAARNA,which is a sign used in relation to time marking.Signers

can also make use of a hold to mark hesitation. Holds are more frequent in the monologues than in the dialogues and can take the form of augmented hold, simple holds and reduplications. The first two occur equally often, while the latter are not as frequently found in the data. The last type of manual marker signers can use for marking hesitation is a break. However, in the this is not used as frequently as signs or holds. During a break the hands can be lifted, folded or placed on the speakers’ lap.

(25)

24 4.1.2. The non-manual markers

This section focusses on the form of the non-manual markers. As has been noted before, all hesitations involved a change in eye gaze direction. As can be seen in Table 7, the movement of the eyes went to different directions during the hesitation. There are striking differences in the direction of the eye gaze used in monologues and dialogues. In dialogues the downwards eye gaze is used more than twice as often than in monologues. In contrast, in monologues the eyes are closed more than ten times more often than in dialogues. Furthermore, movements of the eyes toward the left or right are used in dialogues nearly twice as many times as in monologues.

Table 7. Direction of eye movement during a hesitation.

Type Up Down Left/right Closed Total

Monologue 14 8 7 11 40

Dialogue 13 21 13 2 49

An example of eyes up can be found in Figure 8. This figure shows the eyes before the hesitation in (a) and the eyes during the hesitation in (b). In (b) a change can also be seen in the position of the head (tilted). Furthermore, there is a slight backwards lean of the upper body and the right hand of the signer is in transition upwards. Figure 8c shows a change in the direction of the eyes downwards. Here there is only one non-manual hesitation marker, namely the direction of the eye gaze. The manual marker this eye movement is combined with, is a PALM-UP with a

pointy handshape.

(a) (b) (c)

(26)

25

Thus, all hesitations are marked non-manually through a change in the direction of eye gaze, but those are not the only non-manuals that are used during hesitations, even though they are the most consistent marker. In some instances mouth actions or mouthing occur, mostly in the form of an opening (N=11) or closing (N=14) of the mouth, and in one case by a change from open mouth to closed mouth. In some cases, the signer used mouthings during the hesitation, mostly eh or ehm (N=11), but twice the mouthings were other words, namely of (‘or’) and aap-poes (‘monkey-cat’).

Another non-manual that was found to sometimes accompany eye movements is a change in the position of the head. Multiple variations were found, including a head tilt (see Figure 8b), a sideward movement (see Figure 5), a headshake, a nod or a downward head movement. These head movements occurred 18 times in monologues, and in ten out of 18 times they cooccurred with a mouth movement. In dialogues the head movements were found seven times, of which four were combined with a mouth movement. The two downwards movements in the dialogues were both combined with the mouthing eh, while none of the movements in the monologues were combined with this mouthing.

In four instances, all occurring in monologues, the signer was found to frown during a hesitation, twice combined with a upwards change in the eye gaze and twice with an eye gaze away from the addressee. In the dialogues, the frowns were not used during hesitations.

In dialogues, but not in monologues, movements of the upper body occurred in four instances. Once the signer moved from left to right, once the signer leaned backwards (Figure 8b), and twice the signer shrugged their shoulders (Figure 3a).

4.2. The place in the discourse of hesitation markers

This section focusses on the place in the discourse where the hesitations are found. Figure 9 below shows the distribution of the hesitations that occurred in the monologues and dialogues in percentages. As can be seen, hesitations occurred at the beginning of the sentence more than half of the time both in monologues (67.5%) and dialogues (53%). A position inside the sentence is the second-most common location of hesitation markers in both monologues (20%) and dialogues (35%). Finally, the smallest amount of hesitations are found at the end of the sentence, namely just 12,5% of the hesitations in the monologues, and 12% of the hesitations in the dialogues.

(27)

26

Figure 9. Percentage of occurrences at different places in the discourse in monologues and

dialogues.

The three categories in the figure above can be divided in smaller categories. Table 8 below presents a more detailed description of these subdivisions. The category ‘beginning of sentence’ can be divided into ‘beginning of sentence’ and ‘beginning of turn’. This distinction is made because the start of the turn and start of the sentence do not necessarily coincide. In most cases (five out of seven instances in monologues, and four out of eight instances in dialogues) the hesitation marker at the start of the turn is a hold. For the hesitations at the beginning of the sentence, this is the case in 14 of 20 times for monologues, and eight out of 18 times for dialogues.

Table 8. Place in the discourse of the different hesitations found in monologues and dialogues.

Place in discourse Monologue Dialogue

Beginning of sentence 20 18

Start of turn 7 8

After first word 2 4

Any position after second word 6 8

During enumeration 0 4

Following other hesitation 0 1

End of sentence 5 6 67,5 20 12,5 53 35 12 0 10 20 30 40 50 60 70 80

Beginning of sentence Middle of sentence End of sentence

Per

ce

nt

ag

e

Place in the discourse

(28)

27

The category ‘middle of the sentence’ can be divided into four different subcategories, of which two were only found in dialogues. The category ‘after first word’ is found for two hesitations in monologues and four in dialogues. The category ‘any position after the second word’ includes all instances of hesitations that occur at any place in the sentence later than the second word, except from the occurrences at the end, following other hesitations or during enumerations which are mentioned in other distinct subcategories. For the category ‘any position after the second word’, six hesitations are found in monologues, and eight in the dialogues. In dialogues there are also four hesitations attested combined with an ‘enumeration’. In these enumerations, the hesitation occurred before or after the dominant hand touches the non-dominant hand used for the enumeration. Finally, one hesitation was found ‘following another hesitation’.

The last category ‘end of sentence’ cannot be divided in other categories, because it only encompasses the hesitations at the end of the sentence. At this location only five instances were found in monologues, and in the dialogues six were found.

In summary, most of the hesitation markers were found at the beginning of the sentence. The middle of the sentence was the second favourite position for hesitation markers, and the position where the fewest hesitations are found, is the end of the sentence.

4.3. The relation between the form of the marker and its position

This section focusses on the relation between the form of the hesitation marker and the position in the discourse where the marker occurs. In monologues, the position at the beginning of the sentence is occupied by a hold 19 times, twice by a break, five times by a sign and only once by a hesitation without a manual marker. In the middle of the sentence, the manual hesitation markers are four times a hold, once a break, once a sign, and once there is no manual marker. At the end of the sentence, a hold is used as manual hesitation marker two times, a break is used once and a sign twice. A more detailed overview of the manual hesitation markers and their position in the discourse is given in Table 9. This table shows that for monologues most hesitation markers occur at the beginning of sentences and little are found at the middle or end of the sentences.

(29)

28

Table 9. Overview of the hesitation markers and their position in the discourse in monologues.

Monologues Start turn Beginning of sentence After first word Other position after second word Following other hesitation During enumeration End of sentence Hold Augmented 2 8 1 2 0 0 0 Simple 3 6 0 1 0 0 2 Reduplication 0 0 0 0 0 0 0 Sign PALM-UP 0 1 0 2 0 0 1 INDEX 1 2 0 0 0 0 1 Other 0 1 0 0 0 0 0 Break ₁ ₁ ₁ ₀ ₀ ₀ ₁ No manual ₀ ₁ ₀ ₁ ₀ ₀ ₀

In the dialogues, the position at the beginning is 13 times occupied by a hold, twice by a break and eleven times by a sign. At the middle of the sentence, four of the manual hesitation markers are holds, one is a break, eleven are signs and once no manual marker is used. For the end of the sentence only once a hold is used, once a break and in four times a sign is used to mark the hesitation manually. A detailed overview of these manual markers in the dialogues and their position in the discourse can be seen in Table 10. In the dialogues, signs are more often used to mark the hesitations than holds. At the beginning of the sentence for instance, the difference in number is small (13 holds and eleven signs are found). However, in the middle of the sentence, we found a bigger difference in number (four holds and eleven signs) in the dialogues. In the monologues we found only one sign in the middle of the sentence.

Table 10. Overview of the hesitation markers and their position in the discourse in dialogues.

Dialogues Start turn Beginning of sentence After first word Other position after second word Following other hesitation During enumeration End of sentence Hold Augmented 0 5 0 1 0 1 1 Simple 4 2 0 2 0 0 0 Reduplication 0 2 0 0 0 0 0 Sign PALM-UP 3 7 1 5 1 3 4 INDEX 0 1 1 0 0 0 0 Other 0 0 0 0 0 0 0 Break 1 1 1 0 0 0 1 No manual ₀ ₀ ₁ ₀ ₀ ₀ ₀

(30)

29

When looking at the position of PALM-UP,there is a difference. There are many instances of PALM-UP found in the dialogues, while there are relatively few instances found in the monologues. In monologues, the PALM-UP sign is used only once at the beginning of the sentence and once at the end, and it is used twice in the middle of the sentence. In the dialogues,

PALM-UP is used ten times at the beginning, four times at the end, and ten times in the middle. For changes in the position of the head in monologues, it is found that when the head tilts left or right, it almost always occurs at the beginning of the sentence. Backward head tilts and nods are more often used in the middle of the sentence. In dialogues the tilts and nods often occur at the boundaries of the sentence.

In most instances when there is a change in the direction of the eye gaze, the movement goes up or down (22 times in the monologues and 34 times in the dialogues). Some changes are towards the left or right of the addressee (seven times in the monologues and 13 times in the dialogues, and only a few are done by closing the eyes (eleven times in the monologues and twice in the dialogues). Closing the eyes is in almost all instances done at the beginning of the sentence. For the dialogues closing the eyes only occurs at the beginning, whereas for the monologues this is also done once after the first word and once at the end. Changes towards the left or right of the addressee do not occur at the position after the first word for both monologues and dialogues. However, these changes do occur at other positions in the middle of the sentence, but only after the second word.

(31)

30 4.4. Interim summary

To summarize, 40 hesitations are found in the analysed monologues and 49 in the dialogues. All hesitations are marked by a change in the direction of the eye gaze, and most (38 resp. 48) are additionally accompanied by a manual marker. For monologues, the most common manual markers are holds, and for the dialogues the most frequently used markers are signs, such as

PALM-UP. In addition to eye gaze, other non-manuals that were attested were for instance the mouthings eh and ehm.

There is a difference in distribution of hesitation markers between the monologues and the dialogues. For both monologues and dialogues, more than half of the hesitations are found at the beginning of the sentence, namely 67,5% of the hesitations in the monologues and 53% in the dialogues. The middle and end of the sentence are resp. the second and third most common positions. At the middle of the sentence, more hesitations were found in the dialogues (35%) than in the monologues (20%). Hesitations at the end of the sentence were almost equally distributed for monologues and dialogues, namely 12,5% in the monologues and 12% of the hesitations in the dialogues were found at this position. A more detailed overview of all hesitation markers can be found in the appendix in chapter 8. The data of the dialogues can be found in section 8.1 and the data of the monologues in section 8.2.

(32)

31

5. Discussion

The discussion is divided into six sections, starting with a discussion of the findings related to the form of the hesitation markers (5.1). First, remarkable findings concerning the manual hesitation markers are discussed (5.1.1), followed by a discussion on non-manual hesitation markers (5.1.2). In the second section (5.2), the data concerning the place in the discourse where the hesitation markers appear will be discussed. The third section (5.3) focusses on the debate wheter hesitations are symptoms or signals, followed by a section on methodological shortcomings (5.4). Then there will be a section on the value of this research (5.5) and this chapter ends with a section with suggestions for future research (5.6).

5.1. The form of hesitation markers

An analysis of the data revealed that hesitation markers in NGT can take different forms, although all hesitations involve a change in the direction of the eye gaze. This is combined in most instances with a manual marker which can be a sign (mostly an INDEX or PALM-UP), a hold

(simple, augmented or reduplication) or, in a few cases, a break. 5.1.1. Manual markers

When comparing the monologues and dialogues, the most striking result is the difference between the frequencies of signs vs holds. In monologues only nine out of 38 of the manual markers are signs (or gestures), of which only one is a PALM-UP, and three are a PALM-UP

combined with a wiggle. In contrast, in dialogues there are 24 instances of a PALM-UP found out of 48 manual markers in total. This means that in monologues only 24% of the hesitations is marked by a PALM-UP, compared to 50% in dialogues. A possible explanation for this difference can be found in the function of PALM-UP. McKee and Wallingford (2011) found that PALM-UP

in NZSL can serve for a lot of different functions. It might be that this marker in some instances in the data is not used as a hesitation marker, but as a discourse marker, for instance for holding a turn. Another possibility is that the marker in some instances has a dual function, where it can serve as a marker for both hesitations and other discourse related functions. When PALM-UP

would take up the role of holding-a-turn, it would more likely appear in a setting with multiple speakers, like a dialogue. We would then expect that the PALM-UP would either be used at the beginning of the turn to take the turn, or at the end of the sentence to hold the turn. However, when in the dialogues a signer started his/her turn with a hesitation, only three of out of eight times the signer used a PALM-UP. In addition, in four out of six times a signer used PALM-UP at

the end of the sentence. Therefore it might be that PALM-UP could have taken other functions

(33)

32

Holds, on the other hand were more frequently found in monologues, although they did occur in the dialogues too. In their research, Esposito et al. (2002) found that unfilled pauses and simple holds, as well as filled pauses and augmented holds, are comparable in function. In their data unfilled pauses and simple holds are said to mostly be used by speakers to signal the end of a sentence or to give up the turn. Filled pauses and augmented holds, on the other hand, are claimed to be used in the discourse when speakers have a problem with finding a word/sign or with conceptualising their sentence. Moreover, according to Maclay and Osgood (1959), filled pauses can be found at locations in the discourse where speakers make decisions, such as phrase boundaries and inbetween larger syntactical units.

In the analysis of the movie clips, monologues and dialogues combined, 44 holds were found in total, encompassing 22 augmented holds, two reduplications and 20 simple holds. The reduplicated holds are left out of this comparison due to their low frequency in the data, although if we were to take them into account, I would argue they should be viewed as augmented holds, since they also involve a movement. Furthermore, it is noteworthy they only appeared in the dialogues settings.

If the augmented holds and simple holds found in hesitations in sign language are comparable to filled and unfilled pauses in speech, we would expect to find augmented holds at phrase boundaries (the beginning or end of the sentence), and the simple holds in the middle of the sentence (Maclay & Osgood, 1959) or at the end of the sentence (Esposito et al., 2002). However, both augmented and simple holds were most often found at the beginning of the sentence. Of the augmented holds, 15 were attested at the beginning of the sentence, only one was found at the end and five occurred within the sentence. This is in line with the expectation that augmented holds are mostly found at the borders of the sentence, which happened in 16 out of 21 cases. However, most instances of simple holds were also found at sentence boundaries, with 15 occurring at the beginning of the sentence, three in the middle and two at the end. The prediction for simple holds is that most should be located in the middle or end of the sentence, which is not the case as only three out of 20 simple holds occur at the middle and only two at the end of the sentence. Thus, even if augmented holds might be comparable to filled pauses in terms of their preferred places of occurrence, simple holds do not pattern similarly to unfilled pauses. The findings of this research are not in line with Esposito et al. (2002), nor Maclay and Osgood (1959) and should be studied further. Speech and sign are two different domains with their own articulators, while gesture can be used while using both speech or sign. It is possible that the findings of my research are not in line with Maclay and Osgood (1959), nor Esposito et al. (2002) because the research focusses on other domains. Maclay and