Skip to content
Publicly Available Published by De Gruyter Mouton March 16, 2023

Uptalk in L2 English: the phonetic identity and perception of final declarative rises in Serbian EFL

  • Tatjana Paunović ORCID logo EMAIL logo
From the journal Phonetica


Uptalk has been increasingly documented in different L1 English varieties and communicative contexts, but is rarely recognized in formal L2 educational contexts, where it is still attributed to learners’ inadequate mastery of intonation. This study is a cross-sectional corpus-based exploration of the phonetic realization of uptalk in Serbian EFL students’ semi-spontaneous expository speech, and its perception as a sentence-finality signal. The corpus comprised all rising intonation units (IU) produced by 14 female and 9 male participants, classified by structural clues as syntactic continuation, listing, polar questions, or uptalk, to explore the relatedness of the phonetic properties to structural position and gender. Next, 100 EFL students rated selected phrases, illustrating continuation rises, uptalk, and final falls, as possibly sentence final, on a 5-point scale. The findings showed that uptalk was consistently produced as a phonetically distinct signal, characterized by a larger pitch excursion, a steeper rise slope, a higher rise peak, and a longer post-IU pause. Females produced wider pitch excursions and steeper slopes. Uptalk examples were ranked high as possible sentence-finality signals. The study suggests that EFL speakers’ uptalk should be recognized as a novel socio-pragmatic prosodic device, deliberately used even in more formal academic contexts.

1 Introduction

The use of a final rising pitch contour on a declarative utterance, where, in accordance with descriptions of standard English intonation (e.g. Cruttenden 1997; Wells 2006), a default falling tone would be expected, is an increasingly observed socio-linguistic prosodic phenomenon of spoken English, referred to as High Rising Terminal (HRT, e.g. Fletcher and Harrington 2001; Warren 2005), upspeak (Bradford 1997), or uptalk (Gorman 1993; Liberman 2006; Lindsey 2019; Warren 2016). Originally viewed as peculiar only to some regional varieties – Urban North British (Cruttenden 1995; Wilhelm 2016), Australian (Fletcher 2005; Fletcher and Harrington 2001) and Southern Californian (Barry 2007; Ritchart and Arvaniti 2013), uptalk has been documented in a growing number of regions (cf. Grice et al. 2020): in New Zealand (Warren 2005, 2016; Warren and Fletcher 2016), Mid-western USA (Prechtel 2015; Prechtel and Clopper 2016), Massachusetts (Armstrong et al. 2015), in Canada (Di Gioacchino and Crook Jessop 2011; Sando 2009; Shokeir 2008), as well as in Southern British English (Arvaniti and Atkins 2016; Barry 2007; Lindsey 2019; Shobbrook and House 2003; Wilhelm 2016), including London (Levon 2016, 2020).

Uptalk seems to be spreading not only across regional varieties, but also across different communicative contexts, including academic settings (Lindsey 2019: 108). It is perceived as a “cross-cultural” phenomenon (Sando 2009: 1), as it has also been observed in languages other than English (Warren 2016), as a possible manifestation of a more universal intonational category – the ‘open meanings’ of rising intonation (Cruttenden 1981: 81 in Warren 2016: 15), such as interactive functions observed in Finnish final declarative rises (Ogden and Routarinne 2005), or socio-cultural and metapragmatic information, as in Japanese (Tanaka et al. 2021), particularly with younger speakers, such as teenagers in European Portuguese (Mata et al. 2016). In German, too, adding a slight rising tone to a globally falling contour at the end of a phrase was found to signal “reduction of dominance” and avoidance of “dominance or impoliteness” (Ambrazaitis 2005: 196), with possible communicative differences between final rises in terms of ‘activation versus restriction’ (Dombrowski and Niebuhr 2005), while a recent study of the German Conversation Database documented a frequent use of uptalk in non-questions by female L1 German speakers (Fey and Lewandowski 2020). Moreover, uptalk has been documented to be transferred from mother-tongue (L1) English into second or foreign languages (L2), e.g. Spanish (Buck 2016), and from L2 English into the speaker’s L1 (Tanaka et al. 2021), or into heritage languages (Kim and Repiso-Puigdelliura 2021).

However, although recognized as a developing and spreading sociolinguistic prosodic trait across English varieties, in English as a foreign language (EFL) contexts this sociolinguistic “change in progress” (Hughes et al. 2012: 6–7) has not been researched so far. Yet, observing EFL speakers’ use of a novel socio-pragmatic variable such as uptalk can offer valuable insights about its development and dissemination. EFL speakers represent a large and growing population around the world, which can influence the spread and reinforcement of the valuable, not only across the EFL varieties, but also, through a sort of “backwash” effect, in the L1 language use. In many “Expanding Circle” (Kachru 1985) contexts, such as Serbian, where English has no official status but is learnt widely as a foreign language, English “plays an increasing role in social and professional life” (Davies 2013: 55). EFL speakers use English in various spheres of life, in a variety of social and professional interactions with L2 and L1 English speakers – through the media, social networks, international mobility and study programs or professional projects (Davies 2013: 36). Such EFL students are likely to adopt some elements of language through use and exposure, bring them into the classroom “from the outside”, and use them in ways they feel appropriate, not only in English but also in their other languages, e.g. their mother-tongue (Tanaka et al. 2021). Therefore, observing the ways in which speakers of different levels of “functional nativism” in English (Kachru et al. 2006) may “appropriate” (Davies 2013: 150) a socio-pragmatic variable can contribute to a better understanding of its development and spreading.

1.1 The phonetic identity of uptalk in L1 English

The phonetic identity of uptalk documented in research studies varies in several respects. Firstly, while some researchers describe these rises as “moderate” (e.g. Guy and Vonwiller 1984; Ladd 1996), and state that HRT contours rise about 40%’ from the rise onset, others describe uptalk rises as more “dynamic” (e.g. Warren and Fletcher 2016). Authors also disagree on how exactly the uptalk rise is aligned with the nuclear accent, i.e. whether the nuclear accent preceding the uptalk rise is a rising tone (Allan 1986), a fall-rise (cf. Bradford 1997), or a fall. Lastly, there is conflicting evidence on whether uptalk declarative rises are similar to (cf. Britain 1992) or different from interrogative rises (Fletcher 2005; Shobbrook and House 2003; Warren 2016; Warren and Fletcher 2016).

Despite these differences, which may be attributed to regional variation in the phonetic realization of uptalk (cf. Prechtel and Clopper 2016), several phonetic parameters have been identified as relevant for defining the phonetic identity of uptalk: rise onset (F0 minimum) and peak (F0 maximum), the extent of pitch excursion (span from F0 minimum to F0 maximum), the duration of the rise, the steepness of the rise slope (pitch excursion divided by its duration), and rise alignment relative to the syllable bearing the nuclear accent.

For instance, comparing Australian and New Zealand English, Warren and Fletcher (2016) investigated both pitch measures (rise onset and peak in Hz and ERB; rise pitch range) and temporal measures (rise duration; the position of the rise onset; the pitch rate-of-change across the rise). The study showed that in both varieties the uptalk rise was characterized by being “more dynamic” than in questions, but that the varieties differed in how speakers chose to realize “the more dramatic rises” in uptalk (Warren and Fletcher 2016: 151) – by pitch effects (in Australian), or temporal effects (earlier rises in New Zealand English).

Armstrong et al. (2015) investigated the rise span, rise duration, and rise slope of uptalk in two American English varieties, South Californian and Massachusetts, either represented by 4 female and 4 male speakers. No differences were found in the frequency of uptalk use across region or gender, but there was some variation in its phonetic realization: female speakers from Southern California produced the longest rises, and female speakers from both regions produced rises with steeper slopes than respective male speakers.

The parameters of pitch range, rise excursion, and rise alignment were found to be relevant by Ritchart and Arvaniti (2013), who investigated uptalk in Southern Californian English relative to the speaker’s gender (12 female, 11 male speakers). The study showed that Southern Californian uptalk was characterized by a smaller pitch excursion and a later rise start compared to other rises, i.e. it showed approximately half the pitch rise of those in questions, confirmation requests, and floor holding.

An additional measure was proposed by Di Gioacchino and Crook Jessop (2011): the relationship between the uptalk pitch excursion and the pitch range in the surrounding intonational phrase. Their study examined three rising contours – plateau, uptalk, and question rises – used in free speech by 5 female speakers of Canadian English. Their findings (Di Gioacchino and Crook Jessop 2011: 12) showed that uptalk rises fell between the plateau phrases and questions with respect to the rise expressed as a ratio of the overall pitch range of the phrase in which the rise was produced.

The parameter of rise onset height was investigated by Shobbrook and House (2003) in Southern British English, comparing it with Fletcher and Harrington’s (2001) observation that in Australian English statement rises have a significantly lower rise onset F0 than question rises. Different speaking tasks were used with 18 speakers (12 male, 6 female), with rise onsets categorized relative to the speaker’s overall pitch range. However, the statistical analysis showed no significant differences in the rise onset height between statements and questions, and the authors conclude that other parameters, e.g. rise alignment, or the steepness of the slope, may be more relevant for this distinction.

In their study of uptalk use in Southern British English, Arvaniti and Atkins (2016) focus on the phonetic properties of rise duration, the rise excursion size (in semitones) and F0 velocity (in semitones/sec) to investigate the realization and functions of uptalk produced by four male and four female speakers in a map task and a board game task, i.e. in spontaneous and semi-controlled speech. The study found that compared to rises used in statements, questions and floor-holding, the ones used as confirmation requests had a specific phonetic identity, being both longer and slower, albeit not wider-spanned.

Investigating specifically London speech, Levon (2020) uses the corpus used for investigating the frequency, distribution and use of high-rising terminals (HRT) in the spontaneous conversations of young speakers (Levon 2016) to focus on the meanings and functions of uptalk as related to its phonetic realization, with three phonetic properties in particular: rise excursion (span), rise dynamism (slope), and rise alignment (the point at which the rise begins). Levon (2020: 56–57) points out that the study aimed to relate the acoustic properties of HRT to more general phonological categorizations of intonation contours (the Autosegmental Metrical theory), and to investigate if each of these phonetic properties varies across previously established social and pragmatic categories, i.e. gender, speech type, and information status. The study also included the phonetic properties of the rise duration, and the F0 values at the beginning and end of the rise. In difference to some previous studies with other speaker populations, here the mixed-model regression analyses did not show a clear correlation between the variations of the phonetic properties and the pragmatic differences in the final rise use.

Although there are important differences between research study findings, they highlight the phonetic parameters potentially relevant for defining the phonetic identity of uptalk in different communicative contexts and with different groups of speakers. Therefore, all these phonetic properties were investigated in the present study, aiming to explore the ones that may emerge as characteristic of uptalk in Serbian EFL students’ semi-spontaneous expository speech in the academic (classroom) context.

1.2 Aims and questions

1.2.1 Background – why these Serbian EFL speakers

The participants’ L1, Serbian, is a South Slavic language whose standard variety (Neo-Shtokavian Ekavian) is described as a pitch-accent language with lexical stress. The four lexical pitch accents (‘falling’ and ‘rising’, either ‘short’ or ‘long’) combine pitch and duration to signal lexical prominence (Godjevac 2000a, 2000b; Lehiste and Ivić 1986; Markovic and Sredojevic 2021; Smiljanić 2004, 2006; Sredojević 2017; Sredojevic and Markovic 2020; Zec and Zsiga 2012). That is why Standard Serbian intonation is characterized by a complex interaction between lexical pitch accents and sentence pitch contours (Lehiste and Ivić 1986; Smiljanić 2004; Sredojević 2011, 2017; Zec and Zsiga 2009).

However, unlike Standard Serbian, some older regional dialects, e.g. in the Prizren-Timok region, have preserved the old accentuation system which does not use lexical pitch accents, but is characterized by only one accent (Ivić 1956), often called ‘expiratory’ (Ivić 1985; Lončar Raičević and Sudimac 2017a) or ‘dynamic’ (Peco 1985; Pletikos 2008). This was also found in more recent acoustic analyses of the Prizren-Južna Morava sub-dialect of the Prizren-Timok dialect, specifically, the urban variety used by young educated speakers in the city of Niš (Lončar Raičević Sudimac 2017a, 2017b; Tomic 2020). In such dynamic-accent varieties of Serbian, the pitch contours result from sentence intonation (Ivić 1985: 112), like in intonation languages such as English.

Therefore, as the participants in both parts of the present study were L1 speakers of the urban Niš variety of the Prizren-Južna Morava sub-dialect, it was possible to observe their L2 English uptalk rises without the potential L1 influence of the complex interaction of lexical tone and intonation.

Secondly, the relevant intonation patterns in Serbian are not substantially different from the typical ones in English, except for uptalk. For instance, Godjevac (2000a, 2000b) states that within a phrase with “neutral intonation pattern” the “rightmost word is the strongest” (Godjevac 2000a: 102), and that the final position in declaratives shows a highly reduced pitch range, with the pitch close to the speaker’s baseline (Godjevac 2000b: 109), i.e. the final lowering (Inkelas and Zec 1988: 240) or laryngization (Lehiste and Ivić 1986: 186), similarly to English declaratives. The tune shapes which “occur at the final edge of a prosodic unit corresponding to a syntactic phrase or a sentence” (Godjevac 2000a: 109) are falling in “assertion, wh-question, imperatives”, and morphologically marked yes-no questions, while a rising tone is used for continuation, unmarked questions, and incredulity questions (Godjevac 2000a: 167). She points out that, just as in English, the final rising tone is used in question tags and “prompting intonation (utterances used as questions without a morphological marking)”, while there are “no prosodic differences between declaratives […] and morphologically marked questions” (Godjevac 2000b: 123), as their “final constituents do not exemplify a rise intonation” (Godjevac 2000b: 126). Therefore, taking into account that both more traditional (Lehiste and Ivić 1986) and more recent descriptions of Serbian intonation (Sredojević 2022) state that final declaratives are signalled by a falling contour (and downstep), Serbian EFL students’ use of uptalk could not be ascribed to L1 transfer.

For these reasons, this particular L1 background of our Serbian EFL participants made it possible to observe L2 English uptalk rises without the potential interference of L1 intonation, as an L2 intonational feature acquired spontaneously, out of the formal educational context. Therefore, the findings of this study can be a useful reference point for cross-linguistic comparisons of uptalk used by EFL speakers of different L1 backgrounds in different communicative contexts.

1.2.2 Previous research findings for Serbian EFL speakers

In a sense, this study explores further the occurrence of final declarative rises in Serbian EFL students’ spontaneous and semi-spontaneous speech observed in our previous research, which focused on other properties. In Paunović (2015), the findings showed a notable difference between the reading task, in which sentence finality was almost invariably signalled by falling contours, and the story-retelling task, i.e. semi-spontaneous speech, in which sentence-final rises were used rather frequently. In Paunović (2020), Serbian EFL students’ spontaneous speech was found to be characterized by quite different nuclear pitch contours in sentence-internal (continuing) and sentence-final intonation units (IUs).[1] IUs, although in both structural positions the participants used both falling and rising contours. The phonetic parameters of pitch height, span, slope, alignment, duration, and intensity all proved to be relevant for these distinctions, as continuative rises started with a moderate onset (F0 minimum), not as low as the onset of final rises, and reached a notably narrower pitch span (excursion) than final rises, with a moderate slope as well. On the other hand, the rises that were used in sentence-final IUs reached the highest F0 maximum, had a wide pitch range, and – due to their shortest duration – also had the steepest slope of all the nuclear contours, rising or falling.

1.2.3 Research aims and questions

Setting off from these initial observations, the present study aimed to investigate – as a prerequisite for investigating its usage, meaning, and interpretation – the phonetic identity of Serbian EFL students’ uptalk, that is, whether uptalk rises could be identified as distinct prosodic signals, with a distinct phonetic identity, different from other rising intonation contours produced under the same communicative circumstances, in semi-spontaneous expository speech. Further, the study aimed to explore the listener’s perception of uptalk rises as possible sentence-finality signals.

Uptalk was defined as the final rising pitch contour found at the end of a declarative sentence identified by independent criteria – the syntactic, semantic and discourse clues in its immediate context. The aim was to investigate whether uptalk rises would prove to be phonetically different from other rising intonation contours found in students’ speech, and whether the listeners would interpret such contours as legitimate sentence-finality signals. Another point of interest was a possible phonetic difference in the uptalk produced by female and male speakers, found in much previous research.

Therefore, the study focused on four main questions:

  1. Are any phonetic properties used consistently in final declarative rises (uptalk) independently identified as sentence-final? The hypothesis was that uptalk would prove to be a phonetically distinct signal of declarative-sentence ending, consistently linked to at least some of the relevant phonetic properties.

  2. Are uptalk rises consistently different from other kinds of rises produced by the participants in the same discourse? The hypothesis was that the differences between rises would be robust enough and consistent.

  3. Do EFL listeners perceive uptalk rises as a possible sentence-finality signal? The hypothesis was that they would, given the recent spread of uptalk across varieties and spoken contexts.

  4. Are there any differences between female and male participants’ uptalk? The hypothesis was that there would be some gender-related differences in the phonetic realization of uptalk, as in other researched speaker groups, but also that both male and female uptalk would show some consistency in terms of phonetic identity.

2 Materials and methods

2.1 Participants

The participants in the production part of the study were 23 BA students (14 female, 9 male; aged 20–25; CEFR level B2.2/B2+) of the English Department at the Faculty of Philosophy, University of Niš. The students were recorded delivering their individual expository oral presentations on a topic of their own choice, to their peers, as a classroom activity. The informed consent to use the recordings for this study was obtained from each participant after the sessions.

In the perception part of the study, the selected IUs extracted from the recordings were evaluated by 100 EFL students of the English Department, who had not participated in the production part of the study. The listeners were grouped by age and level of proficiency, as 1) first-year students (N = 75), aged 19–20 (CEFR B2.1 level), and 2) fourth-year students (N = 25), aged 22–24 (CEFR C1.2/C1+ level), who had already completed their TEFL Methodology training, as the BA and MA studies at the English Department lead to a degree in EFL teaching. Therefore, the lower-level group was illustrative of typical EFL learners with no specific linguistic, phonetic, or teaching training, while the higher-level group was illustrative of pre-service teachers with a sound linguistic and phonetic background.

As this study focused on the context of EFL learning and teaching as a community of practice, and the use of a prosodic socio-pragmatic variable as a stylistic move (Coupland 2007) in the language of that community, the listeners in the perception part of the study were not L1 English speakers, but EFL students, i.e. members of the same community of practice. This choice was based on Eckert and Labov’s view (Eckert and Labov 2017, p. 477) that it is through local communities of practice, which are intermediary between individuals and macro-social structures, that sociolinguistic novelties are adopted and implemented. Therefore, both the production and perception parts of the study focused on the “persona” (Coupland 2007) of an EFL speaker, as “a presentation that takes on meaning in, and with respect to, the social-semiotic landscape” (Eckert 2019a, in Eckert 2019b: 753), here the ‘landscape’ of the EFL classroom and the formal educational context.

This particular context was chosen because formal education, by definition, tends to promote “standard” and “correct” linguistic ideologies, in this case, discouraging the use of uptalk declarative rises (Wells 2006) as “non-standard”. In this sense, the lower-level group of listeners, with no theoretical linguistic training, was assumed not to be aware of the controversial status of uptalk in “correct” speech, while the higher-level group of listeners, who had already passed most of their linguistics and EFL Teaching Methodology courses as pre-service EFL teachers, were assumed to have been made aware of uptalk as an “undesirable” prosodic element of speech. Therefore, the hypothesis was that the two groups of listeners might perceive uptalk as a possible sentence-finality signal in different ways.

2.2 Expository speech corpus

In the production part of the study, students’ presentations, lasting from 3 to 4 min, were recorded digitally, in the classroom setting. The final selection of clear recordings comprised 73 min of students’ speech, which was transcribed verbatim and anonymised. The analysis of the transcripts (TextSTAT, v. 3.0) showed that the corpus comprised 9.232 words and 727 sentences, with the average of 12.9 words per sentence.

As the data-gathering task, students’ oral presentations were chosen because, compared to other elicitation tasks, e.g., interviews, they provided more control over some factors that can reflect on L2 students’ performance, such as relationships of power, or the amount of each speaker’s contribution. In this task, all students performed under the same conditions and for approximately the same time, in their peer group, on a pre-researched topic of their own choice. On the other hand, this type of task limited the students’ linguistic choices, since oral presentations do not provide opportunities for interaction, turn taking, and floor management. However, since this study did not investigate the functions of uptalk, but aimed to establish whether it would occur as a distinct phonetic signal in students’ continuous and relatively free speech, this limitation was not considered crucial.

2.3 Structural and positional types of rises

The preliminary analysis of the corpus was performed using the transcripts, the audio files, and the visual inspection of the pitch contours. Firstly, all intonation units (IU) with final rising contours were identified, and the tokens that had a clear pitch contour extracted for further analysis. The final corpus (outliers excluded) consisted of 316 IUs with a final rise, 193 (61.1%) produced by female and 123 (38.9%) by male participants.[2]

Secondly, based on the clues in their syntactic, discourse and semantic context, the selected IUs were categorized into six use types. Here are some examples for each type, with the transcription symbol (?) used to mark all rising pitch contours, and with the exemplified rise underlined:

Listing rises (20 IUs, 6.3%), characteristic of enumeration and giving examples, which occurred both sentence-medially, like continuation rises, and sentence-finally, like proper declarative (uptalk) rises:

Remove all unnecessary information like age(?), marital status(?), number of children(?) All those are things you don’t want in your resume.

Polar (yes/no) questions (3 IUs, 0.9%), used as rhetorical questions:

Are they aware of this?/Did a person beside you yawn?

Continuation rises, IU-, utterance-, or clause-final, but not sentence-final (144 IUs, 45.6%):

Being involved in a creative activity helps us to get to know ourselves better(?), and understanding ourselves better makes us happier.

Uptalk – the rising contours used on obviously sentence-final, even paragraph-final IUs (63 IUs, 19.9%):

It represented the peak of the large construction, monumental inscription, and the rise of the arts(?) The political landscape of the classic period was changed.

As some IUs were ambiguous and could not be classified beyond doubt, such IUs were categorised separately so as to ensure a more objective picture of their phonetic properties. Two ‘transitional’ categories were introduced for this purpose. One comprised uptalk-like rises, labelled Possible uptalk 1 or Quasi 1 (40 IUs, 12.7%), used on IU which were apparently not sentence final, because they were followed by an utterance beginning with a linker (and; so). However, a closer inspection of the context revealed no grammatical, semantic or discourse reason for the two utterances to be joined into a single sentence. In those cases, we hypothesised that the use of the linker might have been a performance “slip”, a pause-filler, or an expression of the speaker’s uncertainty about the syntactic structure that s/he wanted to use, but that nevertheless the speaker’s intention could have been to mark the end of the statement. For these ambiguous cases, we introduced a separate category for these IUs, in order to explore whether they would display phonetic properties similar to continuation rises or to uptalk rises:

Firstly, I’ll talk about comedies, then science fiction from the superhero genre, and third, horror movies(?) and the presentation will be short.

The other transitional category comprised uptalk-like rises, labelled Possible uptalk 2 or Quasi 2 (46 IUs, 14.6%), when there were no clear context clues as to whether the participant intended the utterance to be syntactically conjoined with the next one or not. Namely, there was neither a linker nor a grammatical, syntactic or discourse reason for the two utterances to be treated as part of the same sentence, but instead of a default sentence-final falling contour, a rising contour was used. When such sentences were not paragraph final, we hypothesized that either interpretation was possible (continuation within the sentence or sentence ending), and introduced a separate category for them, to compare their phonetic properties to both continuation and uptalk rises:

They were invented by a scienti […] by a teacher, French teacher known as Alfred Benet(?) An American scientist known as B. J. Carter took the test and […] the same test and tried to apply it to everybody.

This American heavy metal band was founded in Huntington Beach, California in 1999(?) This band consisted of […] five members, two of them…

For more transparent statistical comparisons, each IU was coded for the rise use type with variable values: 0 = listing rises, 1 = polar questions, 2 = continuation rises, 3 = Quasi-uptalk 1, 4 = Quasi-uptalk 2, 5 = Uptalk proper. In all the statistical analyses, an additional aggregate variable was also included, with all the uptalk-like IUs (Uptalk proper, and Possible or Quasi-uptalk 1 and 2) grouped together as ‘aggregate uptalk’, and all the procedures repeated with this aggregate variable as well. All uncertainties or conflicts in the IU categorization were resolved by an independent evaluator, an experienced EFL lecturer from the English Department.

2.4 Procedures

Acoustic analysis. The acoustic analysis was performed by Praat v.5.2.03 (Boersma and Weenink 2013), with the settings of 400 Hz pitch ceiling, 50 Hz pitch floor. The beginning of each rise was located manually; F0 contours were visually inspected for octave errors and disruptions. For each IU, the following measures were taken: the rise onset F0 (F0 minimum), the rise peak F0 (F0 maximum), the duration of the rise from the onset to the peak, and the duration of the pause after the rise. The F0 minimum and maximum were measured for the whole IU, and its pitch range computed in semitones (ST) and ERB, to facilitate comparisons of male and female speakers.

Lastly, when the nuclear syllable contained a detectable fall preceding the rise (in 73 IUs, 23.1%), the duration of the nuclear fall was measured as a separate variable. Also, when a plateau occurred during the rise (in 177 IUs, 56%), although included in the rise duration, the plateau duration was also measured as a separate variable.

From these measurements, the following variables were computed: the pitch excursion of the rise, i.e. the span from the F0 minimum to F0 maximum in Hz, ST and ERB; the slope of the rise, i.e. the pitch excursion divided by the duration of the rise, as Hz/ms, ST/ms, and ERB/ms; the ratio of the pitch excursion to the pitch range in its encompassing IU in Hz, ST, and ERB (cf. Di Gioacchino and Crook Jessop 2011); and the rise excursion expressed as the percentage of the rise F0 minimum. Lastly, the rise alignment was coded as a categorical variable, classifying each rise by its onset position relative to the IU’s nucleus (the last stressed syllable in all the IUs analysed): as early when it occurred during the nucleus (in 242 IUs, 76.6%); as immediate when it occurred on the post-nuclear syllable (67 IUs, 21.2%), and as late elsewhere (7 IUs, 2.2%). Although research has shown that vowels of different height may differ in intrinsic pitch up to 15 Hz (1.65 ST) (Mennen et al. 2014, p. 312), we could not control for vowel use in a free-speech task, so we could not include this factor in this study.

Statistical analysis. Since most variables did not show a normal distribution of the data (Shapiro-Wilk test), the statistical analysis (IBM SPSS v. 20) was performed by non-parametric tests: Spearman’s rank correlation for the relatedness of the rise use type and the measured phonetic parameters; Kruskal-Wallis tests to compare the means of the phonetic parameters (tested variables) in different rise use types (the grouping variable); and, for the variables that showed overall significance, Mann-Whitney U tests to compare pairs of rise use types and female and male participants. Standard deviations (SD), Confidence Intervals (CI) and medians (Md) were reported where relevant and the effect size calculated for the parameters that showed significance. The Two-way between-groups analysis of variance was used to explore the interaction of gender and the rise use type.

2.5 Procedures and materials for perception

In the perception part of the study, the listening materials consisted of 22 IUs excerpted from the recorded presentations. The listening materials were chosen after the acoustic and statistical analysis. The list of the selected phrases is given in Appendix 1. All the IUs were taken from declarative sentences undoubtedly intended by the speakers as statements. Five of these IUs had sentence-final falling contours; seven were sentence-medial IUs with continuation rises, and ten were sentence-final IUs with rising contours, i.e. examples of uptalk. Uptalk examples came from four different speakers, and continuation rises and final falls came from six speakers.

The participants were asked to listen to each of the IUs and state if what they had just heard was at the end of a sentence or not. All the IUs were heard in isolation, without any context. The items were randomized for each listening session. Allowing for different degrees of certainty in their interpretation, instead of a yes-no choice, the participants were offered a 5-degree scale (1 = certainly not at the end, 2 = probably not, 3 = not sure 4 = probably at the end, 5 = certainly at the end). They were instructed to answer quickly, without much thinking, i.e. to offer their first impression about each IU.

3 Results

3.1 Production

The variable for the type of the rise correlated with several phonetic parameters. Firstly, significant (p < 0.000) and moderately strong correlations were found between the use type and the rise excursion and rise slope (Table 1). Moderately strong correlations were also observed between the use type and the duration of the pause after the rise (r s = 0.349**, p < 0.000), and the pitch range of the whole IU in Hz (r s = 0.369**), ST (r s = 0.418**) and ERB (r s = 0.392**; all p < 0.000), as well as a weaker correlation with the rise/IU range ratio, in Hz/ms (r s = 0.185**, p < 0.052), ST/ms (r s = 0.183**, p < 0.053) and ERB/ms (r s = 0.185**, p < 0.054). The positive correlations suggested that the uptalk and quasi-uptalk rises, compared to continuation rises, lists, and questions, had higher values for the rise excursion and rise slope, that they occurred in IUs with a wider pitch range, and that all uptalk-like rises were followed by longer pauses.

Table 1:

The Spearman’s correlation coefficients of the variable rise use type, and the variables: rise excursion (in Hz, ST, and ERB) and the rise slope (in Hz/ms, ST/ms, and ERB/ms).

Spearman’s rho
1 2 3 4 5 6 7
1. Rise use type 1.000 0.423** 0.474** 0.447** 0.371** 0.402** 0.376**
2. Excursion Hz 1.000 0.899** 0.980** 0.870** 0.715** 0.789**
3. Excursion ST 1.000 0.948** 0.777** 0.787** 0.762**
4. Excursion ERB 1.000 0.847** 0.747** 0.803**
5. Slope Hz/ms 1.000 0.896** 0.941**
6. Slope ST/ms 1.000 0.921**
7. Slope ERB/ms 1.000

Therefore, we grouped together all uptalk-like rises in an aggregate variable value for the rise use type. This yielded a rather even distribution of two types of rises in the corpus, 144 (45.6%) continuation rises and 149 (47.2%) aggregate uptalk rises. The repeated correlation tests for the aggregate uptalk also showed moderately strong positive correlations between the rise use type and several variables: the rise excursion in HZ (r s = 0.447**), ST (r s = 0.491**), and ERB (r s = 0.471**), the rise slope in HZ (r s = 0.374**), ST/ms (r s = 0.390**) and ERB/ms (r s = 0.377**), the pause after the rise (r s = 0.303**), and the rise F0 maximum (r s = 0.231**), all significant (p < 0.000). The parameters of rise duration and rise alignment did not show correlation with the rise use type.

3.1.1 Differences between rise use types

The comparisons (Kruskal-Wallis) of the mean values showed statistically significant differences between the six rise use types for the duration of the pause after the rise (χ2 = 47.897, df = 5, p < 0.000; η 2 = 0.138; d Cohen = 0.802, large effect), with the highest mean rank for uptalk. Table 2 shows the means, standard deviations and medians for pause duration after the six rise types. The repeated tests with the aggregate uptalk confirmed the statistically significant difference (Sig. < 0.000; χ2 = 30.840; η 2 = 0.083; d Cohen = 0.603, medium effect), with the mean of the aggregate uptalk still higher than that in questions (Table 2).

Table 2:

Pause duration – means, standard deviations, and medians in the rise use types.

Rise use type Pause duration in ms
Uptalk rises 42.3 26.2 43.0
Questions 27.7 16.8 20.0
Quasi uptalk 1 26.7 20.7 27.5
Quasi uptalk 2 23.4 20.4 24.5
Continuation 18.7 17.3 15.5
Lists 13.2 15.7 1.5
Aggregate uptalk 32.3 24.6 31.0

Statistically significant differences between rise use types were also found for the rise excursion in Hz (p < 0.000; χ2 = 76.441; η 2 = 0.23; d Cohen = 1.094, large effect), ST (p < 0.000; χ2 = 85.037; η 2 = 0.258; d Cohen = 1.18, large effect) and ERB (p < 0.000; χ2 = 81.777; η 2 = 0.248; d Cohen = 1.148, large effect).

Uptalk and quasi-uptalk rises had the highest means, as shown in Table 3. Significant differences were also found for the rise slope in Hz/ms (p < 0.000; χ2 = 71.964; η 2 = 0.216; d Cohen = 1.05, large effect), ST/ms (p < 0.000; χ2 = 73.346; η 2 = 0.22; d Cohen = 1.064, large effect) and ERB/ms (p < 0.000; χ2 = 72.474; η 2 = 0.218; d Cohen = 1.055, large effect), with the means of uptalk-like rises higher than others, as shown in Table 3. The repeated tests with the aggregate uptalk confirmed these relationships. Additional statistics for these variables (SDs, medians, 95% Confidence Intervals) are displayed in Appendix 2.

Table 3:

Six rise use types (and the aggregate uptalk variable) – the means of the variables: rise excursion, rise slope, rise span as the percent of the F0 minimum, and the rise peak (F0 maximum).

Rise use type Exc Hz Exc ST Exc ERB Slope Hz Slope ST Slope ERB Span as % F0 min Rise F0 max
List rises 16.95 1.84 0.39 0.34 0.0445 0.0065 12.25 177.00
Polar questions 30.33 2.08 0.55 0.88 0.0633 0.0200 12.83 297.67
Continuation 64.39 5.80 1.39 2.02 0.1807 0.0447 41.63 225.40
Quasi uptalk 1 105.20 8.21 2.13 2.80 0.2178 0.0574 62.90 268.85
Quasi uptalk 2 81.02 7.56 1.77 2.02 0.1893 0.0449 56.80 229.20
Uptalk 97.57 8.33 2.04 2.92 0.2570 0.0610 63.16 250.06
Aggregate uptalk 94.51 8.06 1.98 2.6089 0.2255 0.0551 61.12 248.66

Uptalk rises also seemed to be characterized by a high peak (F0 maximum), as suggested by the statistically significant difference (p < 0.000; χ2 = 27.862) between the rise use types, supported by the mean rank comparison. As shown in Table 3 (and Table A1 – Appendix), the highest mean F0 peak value was found in questions, but uptalk rises followed closely. However, the analysis of the effect size of this difference (η 2, d Cohen) showed no effect, which could be explained by the fact that the CI for questions contained the zero (cf. Appendix 2). No statistically significant differences were found between the rise use types for the rise onset F0 (minimum).

To confirm the observed differences, the Mann-Whitney U tests were performed for pairs of rise use types, with the Boniferroni correction of α (α altered = 0.05/6 = 0.0083; α critical = 1 – (1–0.0083)6 = 0.049). Statistically significant differences were confirmed between the uptalk and continuation rises (Table 4) for the pitch excursion, pitch slope, pause duration and pitch peak (F0 maximum) variables, but not for the rise/IU pitch range ratio.

Table 4:

Mann-Whitney U test results – continuation rises compared to uptalk rises for the variables: rise excursion (in Hz, ST and ERB), rise slope (in Hz, ST and ERB), pause duration, rise peak F0 maximum, and rise/IU pitch range ratio (in Hz, ST and ERB).

Exc Hz Exc ST Exc ERB Slope Hz Slope ST Slope ERB
Mann-Whitney U 2,651.500 2,332.500 2,530.000 2,977.500 2,652.000 2,947.500
Wilcoxon W 13,091.500 12,772.500 12,970.000 13,417.500 13,092.000 13,387.500
Z −4.753 −5.557 −5.060 −3.930 −4.754 −4.033
Asymp. Sig. (2-tailed) 0.000 0.000 0.000 0.000 0.000 0.000
Effect size η 2 0.109 0.149 0.124 0.075 0.109 0.078
d Cohen 0.7 0.837 0.751 0.568 0.7 0.58
Pause F0 max Ratio Hz Ratio ST Ratio ERB
Mann-Whitney U 2049.500 3,427.500 4,426.000 4,430.000 4,430.500
Wilcoxon W 12,489.500 13,867.500 14,866.000 14,870.000 14,870.500
Z −6.338 −2.796 −0.519 −0.500 −0.497
Asymp. Sig. (2-tailed) 0.000 0.005 0.604 0.617 0.619
Effect size η 2 0.19 0.038 0.0 0.0 0.0
d Cohen 0.963 0.396 0.039 0.037 0.037
  1. Grouping Variable: Rise use type.

The repeated tests with the aggregate uptalk (Table 5) showed significant differences for the same set of variables, with a large effect for the rise excursion and moderate for the rise slope. The Mann-Whitney comparisons of uptalk and list rises showed significant differences for all these variables, too, including the pause duration (all p < 0.000). Furthermore, the comparison of uptalk and questions showed significant differences for the rise excursion in ST and ERB (both p < 0.004) and the rise slope in ST (p < 0.004) and ERB (p < 0.006), but not for pause duration.

Table 5:

Mann-Whitney U test results – continuation rises compared to aggregate uptalk for the variables: rise excursion (in Hz, ST and ERB), rise slope (in Hz, ST and ERB), pause duration, rise peak F0 maximum, and rise/IU pitch range ratio (in Hz, ST and ERB).

Exc Hz Exc ST Exc ERB Slope Hz Slope ST Slope ERB
Mann-Whitney U 6,712.500 6,010.000 6,378.500 8,110.500 7,754.500 8,094.000
Wilcoxon W 17,152.500 16,450.000 16,818.500 18,550.500 18,194.500 18,534.000
Z −5.539 −6.508 −6.000 −3.610 −4.104 −3.658
Asymp. Sig. (2-tailed) 0.000 0.000 0.000 0.000 0.000 0.000
Effect size η 2 0.105 0.145 0.123 0.044 0.057 0.045
d Cohen 0.684 0.822 0.748 0.432 0.494 0.434
Mann-Whitney U 7,170.500 8,555.000 9,998.000 10,001.000 10,003.500
Wilcoxon W 17,610.500 18,995.000 20,438.000 20,441.000 20,443.500
Z −4.953 −2.997 −2.159 −2.150 −2.143
Asymp. Sig. (2-tailed) 0.000 0.005 0.031 0.032 0.032
Effect size η 2 0.082 0.031 0.003 0.003 0.003
d Cohen 0.598 0.356 0.118 0.117 0.117
  1. Grouping Variable: Aggregate Uptalk.

Other parameters highlighted by research on L1 uptalk did not show clear effects in our study. For rise duration, no statistically significant differences were found, and for the proportion of the rise excursion and the surrounding IU pitch range the findings were inconclusive. Namely, although statistically significant differences between the rises were found by the Kruskal-Wallis test – but not confirmed by the Mann-Whitney U tests – for the ratio in Hz/ms (p < 0.001; χ2 = 22.077), ST/ms (p < 0.001; χ2 = 22.078) and ERB/ms (p < 0.001; χ2 = 22.048), with the medium effect size (all η 2 = 0.055; d Cohen = 0.483), the mean rank comparisons did not show robust differences, except in questions. Table 6 shows means, standard deviations and medians for this parameter in all rise use types. For illustration, Figure 1 shows the contours of different types of rises: continuation, uptalk and question rises.

Table 6:

Means, standard deviations, and medians for the ratio of the rise span and the encompassing IU (in Hz, ST and ERB) in the six rise use types.

Rise/IU ratio in Hz Rise/IU ratio in ST Rise/IU ratio in ERB
Rise type M SD Md M SD Md M SD Md
Questions 0.560 0.399 0.46 0.537 0.420 0.43 0.543 0.411 0.43
Lists 0.907 0.286 1 0.906 0.288 1 0.30 0.30 1
Continuation 0.955 0.14 1 0.956 0.14 1 0.957 0.14 1
Uptalk 0.984 0.06 1 0.983 0.07 1 0.985 0.06 1
Quasi uptalk 2 0.994 0.043 1 0.995 0.037 1 0.994 0.04 1
Quasi uptalk 1 1 0.01 1 1 0.01 1 1 0.01 1
Figure 1: 
Illustration of different types of rises: continuation (upper left), uptalk (upper right), and two question rises (lower left and right).
Figure 1:

Illustration of different types of rises: continuation (upper left), uptalk (upper right), and two question rises (lower left and right).

3.1.2 Gender differences

Significant differences were found between the gender groups regarding the observed phonetic parameters. Although all participants produced uptalk rises with a larger excursion and a steeper slope than other rises, the differences were more pronounced with female participants. Table 7 shows means, standard deviations and medians for the variables of pitch excursion, pitch slope, IU pitch range and pause duration for female and male participants, with higher means in females for pitch excursion, pitch slope, and IU pitch range, but higher means with males for pause duration.

Table 7:

Means, standard deviations, and medians for female and male participants for the variables: rise excursion (in Hz, ST, and ERB), rise slope (in Hz/ms, ST/ms, and ERB/ms), the pitch range of the encompassing IU in ST and ERB, and pause duration.

Rise excursion in Hz Rise excursion in ST Rise excursion in ERB
Female 94.00 45.32 94.0 7.16 3.32 7.2 1.90 0.88 1.9
Male 45.90 27.13 40.0 5.67 2.90 5.2 1.13 0.62 1.0
Rise slope in Hz/ms Rise slope in ST/ms Rise slope in ERB/ms
Female 2.80 1.69 2.20 0.22 0.13 0.19 0.057 0.034 0.05
Male 1.20 0.74 1.03 0.15 0.09 0.14 0.032 0.020 0.03
IU pitch range in ST IU pitch range in ERB Pause duration in ms Pause after uptalk/ms
Female 7.57 3.75 7.60 1.99 0.99 2.00 22.47 20.96 20.0 40.75
Male 5.84 2.80 5.40 1.15 0.60 1.02 28.54 23.37 28.0 45.95

These differences were further supported by the Mann-Whitney U tests. For pause duration, the difference proved to be significant (Z = −2.146, p < 0.032), with the male mean rank (=172.17) much higher than the female mean rank (=149.79). Significant differences between females and males were confirmed for rise excursion and rise slope, too (Table 8). The negative values of Z in all these cases, as well as mean rank comparisons, confirmed that female participants produced larger rise excursions and steeper rise slopes. The effect sizes (Table 8) for these differences were moderate or large.

Table 8:

Mann-Whitney U test results – female and male participants compared for the variables: rise excursion (in Hz, ST and ERB), and rise slope (in Hz, ST and ERB).

Exc Hz Exc ST Exc ERB Slope Hz Slope ST/ Slope ERB
Mann-Whitney U 4,221.000 8,493.000 5,652.000 4,340.000 8,379.500 6,115.500
Wilcoxon W 11,847.000 16,119.000 13,278.000 11,966.000 16,005.500 13,741.500
Z −9.659 −4.264 −7.853 −9.509 −4.410 −7.312
Asymp. Sig. (2-tailed) 0.000 0.000 0.000 0.000 0.000 0.000
Effect size η 2 0.295 0.058 0.195 0.286 0.061 0.167
d Cohen 1.294 0.494 0.985 1.266 0.512 0.896
  1. Grouping Variable: Gender.

To illustrate these differences, Figure 2 shows an uptalk rise produced by a female speaker, followed by a pause and the beginning of a new sentence. Figure 3 shows an uptalk rise produced by a male speaker, followed by a pause, a pause-filler, and a new sentence. The pitch tracks in the two spectrograms show that the female speaker’s uptalk rise had both a larger excursion and a steeper slope than that produced by the male speaker.

Figure 2: 
The pitch track of an uptalk rise produced by a female participant.
Figure 2:

The pitch track of an uptalk rise produced by a female participant.

Figure 3: 
The pitch track of an uptalk rise produced by a male participant.
Figure 3:

The pitch track of an uptalk rise produced by a male participant.

No statistically significant differences between the two gender groups were found for other variables, except for two details. Firstly, both a statistically significant difference between gender groups (Z = −5.435, p < 0.000), and the correlation direction established between the variables of rise use type and nuclear fall duration (r s = −0.233; p < 0.000) suggested that female participants used longer nuclear falls before the rise (in ms) than males did (female M = 4.94, SD = 7.7, Md = 0.0; male M = 0.92, SD = 3.43, Md = 0.0; for both, 95%CI > 0.0).

Secondly, a statistically significant difference was found between the two gender groups regarding the presence and duration of the plateau during the rise. Namely, male participants produced more plateaux during their rises, such as the one shown in Figure 4. Their plateaux were also much longer than females’, judging by the negative value of Z (=−4.332, p < 0.000) and the comparison of mean ranks (Males: mean rank = 185.22, M = 10.31, SD = 9.66, Md = 10; Females: mean rank = 141.47, M = 5.89, SD = 7.26, Md = 0.0; for both, 95%CI > 0.0).

Figure 4: 
The pitch track of an uptalk rise produced by a male participant, containing two plateaux: the shorter one on the IU nucleus, and the stepped-up and longer one on the post-nuclear unstressed syllable.
Figure 4:

The pitch track of an uptalk rise produced by a male participant, containing two plateaux: the shorter one on the IU nucleus, and the stepped-up and longer one on the post-nuclear unstressed syllable.

3.1.3 Gender similarities

However, despite these phonetic differences, female and male participants actually followed very similar patterns in their use of rises. Firstly, compared Spearman’s correlation coefficients for female and male participants showed that within both groups significant correlations existed, with small or no differences in the effect size. Specifically, as shown in Table 9, the correlations were found in both groups between the rise use types and the variables of rise excursion, rise slope, pause duration, and IU pitch range, with significance equal in both groups (p < 0.000 for all these variables).

Table 9:

Correlation coefficients compared – female versus male participants.

Rise excursion Hz Rise excursion ST Rise excursion ERB Rise slope Hz/ms Rise slope ST/ms Rise slope ERB/ms
Female r s 0.499** 0.519** 0.504** 0.355** 0.362** 0.356**
Male r s 0.326** 0.400** 0.354** 0.421** 0.465** 0.395**
q Cohen’s 0.21 small effect 0.151 small effect 0.185 small effect 0.078 no effect 0.124 small effect 0.045 no effect
IU pitch range in Hz IU pitch range in ST IU pitch range in ERB Pause duration
Female r s 0.393** 0.421** 0.405** 0.360**
Male r s 0.312** 0.397** 0.343** 0.367**;
q Cohen’s 0.093 no effect 0.028 no effect 0.072 no effect 0.008 no effect

Secondly, although female speakers produced IUs with wider pitch ranges, the means for the rise/IU pitch range ratio were very similar with female (Hz/ms M = 0.965, ST/ms M = 0.966, ERB/ms M = 0.967) and male participants (Hz/ms M = 0.965, ST/ms M = 0.965, ERB/ms M = 0.964), suggesting that the male speakers’ narrower-spanned rises were as proportional to their – narrower – IU pitch range as were female speakers’.

That both gender groups followed similar patterns in using these phonetic parameters was further confirmed by a Two-way between-groups (Univariate) analysis of variance. Since we tested 20 variables for the potential interaction of rise use type and gender, we applied the Boniferroni corrected α altered = 0.05/20 = 0.0025 (α critical  = 1 – (1–0.0025)20 = 0.049). With this correction, no significant interaction was found for most variables with established significance. When significant interaction was observed, the main influence of the rise use type showed both greater significance and a larger effect size than gender.

For instance, for pause duration, the interaction between gender and rise-use type was not statistically significant (F(5, 316) = 2.19, p < 0.056; partial η 2 = 0.035). The main influence of gender was not significant (F(1, 316) = 7.98, p < 0.005; partial η 2 = 0.026), while the main influence of the rise use type (F(5, 316) = 14.2, p < 0.000) showed a very large effect (partial η 2 = 0.189). Post-hoc Tukey HSD confirmed a significant difference (p < 0.000–0.001 for all) between uptalk rises and others, except questions. This is illustrated by the plot of estimated means for pause duration in Figure 5 (Left).

Figure 5: 
The plots of estimated marginal means for pause duration for female and male participants: Left: for the six rise use types (0 = lists, 1 = polar questions, 2 = continuation, 3 = quasi-uptalk 1, 4 = quasi-uptalk 2, and 5 = uptalk); Right: for the rise use types with all uptalk-like rises comprised as aggregate uptalk.
Figure 5:

The plots of estimated marginal means for pause duration for female and male participants: Left: for the six rise use types (0 = lists, 1 = polar questions, 2 = continuation, 3 = quasi-uptalk 1, 4 = quasi-uptalk 2, and 5 = uptalk); Right: for the rise use types with all uptalk-like rises comprised as aggregate uptalk.

It shows that although males produced remarkably longer pauses than females, both gender groups distinguished between rise use types by producing longer pauses after uptalk rises (except in questions for males). This was further supported by the repeated analysis with the aggregate uptalk, which also showed no significant interaction of gender and rise use type (F(3, 316) = 1.19, p < 0.313; partial η 2 = 0.011), with the main influence of the rise use type (F(3, 316) = 14.7, p < 0.000) showing a large effect (partial η 2 = 0.125). The main influence of gender showed no statistical significance (F(1, 316) = 2.66, p < 0.104; partial η 2 = 0.009), and the post-hoc Tukey HSD confirmed significant differences between aggregate uptalk and continuations and lists. This is illustrated in Figure 5 (Right).

As expected, the rise excursion in Hz indeed showed a significant interaction of gender and rise use type (F(5, 316) = 4.76, p < 0.000; partial η 2 = 0.072), though with a very large effect of the main influence of the rise use type (F(5, 316) = 18.48, p < 0.000; partial η 2 = 0.233), and medium effect for gender (F(1, 316) = 24.7, p < 0.000; partial η 2 = 0.075). When expressed in ST, the rise excursion did not show significant interaction between gender and rise use type (F(5, 316) = 1.97, p < 0.082; partial η 2 = 0.031), the main influence of rise use type (F(5, 316) = 20.66, p < 0.000) showed a very large effect (partial η 2 = 0.254), while the main influence of gender was not significant (F(1, 316) = 1.38, p < 0.241; partial η 2 = 0.005). The repeated test with the aggregate uptalk variable showed no interaction of gender and use type (F(3, 316) = 2.05, p < 0.107; partial η 2 = 0.020), and no main influence of gender (F(1, 316) = 0.07, p < 0.79; partial η 2 = 0.000). This is illustrated in Figure 6 (Left), where the estimated marginal means plots for the rise excursion expressed in ST show that both gender groups produced larger pitch excursions in uptalk rises.

Figure 6: 
The plots of estimated marginal means for male and female participants in the different rise use types, with the variable aggregate uptalk. Left: Rise excursion in ST; Right: Rise slope in ERB.
Figure 6:

The plots of estimated marginal means for male and female participants in the different rise use types, with the variable aggregate uptalk. Left: Rise excursion in ST; Right: Rise slope in ERB.

The results for the rise slope were quite similar. When expressed in Hz, it showed some interaction of use type and gender (F(5, 316) = 2.24, p < 0.05; partial η 2 = 0.036), but still the main influence of the rise use type had a large effect (F(5, 316) = 9.72, p < 0.000; partial η 2 = 0.138), and gender a medium effect (F(1, 316) = 14.96, p < 0.000; partial η 2 = 0.047). But when the rise slope was expressed in ST, there was no interaction of gender and the use type (F(5, 316) = 0.58, p < 0.72; partial η 2 = 0.009), and no significant main influence of gender (F(1, 316) = 1.67, p < 0.197; partial η 2 = 0.005), especially for the aggregate uptalk variable (interaction: F(3, 316) = 0.639, p < 0.59; partial η 2 = 0.006; rise use type: F(3, 316) = 15.7, p < 0.000; partial η 2 = 0.133; gender (F(1, 316) = 0.37, p < 0.54; partial η 2 = 0.001). Similarly, when the rise slope was expressed in ERB, there was no interaction (F(5, 316) = 1.58, p < 0.164; partial η 2 = 0.025), with the significant main influence of use type (F(5, 316) = 10.04, p < 0.000; partial η 2 = 0.142), and no main influence of gender (F(1, 316) = 6.76, p < 0.010; partial η 2 = 0.022). The same was found for the repeated test with the aggregate uptalk variable (interaction: F(3, 316) = 2.11, p < 0.099; partial η 2 = 0.02; rise use type: F(3, 316) = 15.96, p < 0.000; partial η 2 = 0.135; gender (F(1, 316) = 1.74, p < 0.188; partial η 2 = 0.006). This is illustrated by the estimated marginal means plots in Figure 6 (Right).

3.2 Perception

The selected IUs were ranked by their mean scores for being interpreted as sentence-final by the participants. Table 10 shows the means and variability measures for the ranked phrases, for the whole population (1), and then for the two groups of listeners (2) and (3).

Table 10:

Means and variability measures for the phrases ranked for being final.

(1) by all the participants
N = 100 Mean Std. Deviation Skewness Kurtosis
Statistic Std. Error Statistic Std. Error
Fall F6 2 4.59 1.120 −2.739 0.241 6.077 0.478
Fall F1 2 4.20 1.557 −1.537 0.241 0.458 0.478
Fall F4 3 4.18 1.592 −1.483 0.241 0.257 0.478
Fall F5 1 4.04 1.620 −1.244 0.241 −0.302 0.478
Uptalk F2 5 3.21 1.940 −0.227 0.241 1.934 0.478
► Fall F2 3 2.99 1.946 0.006 0.241 −1.978 0.478
Uptalk F2 8 2.96 1.938 0.032 0.241 −1.979 0.478
Continuation F1 3 2.92 1.926 0.081 0.241 −1.961 0.478
Continuation F3 5 2.89 1.938 0.124 0.241 −1.960 0.478
Continuation F6 2 2.70 1.829 0.284 0.241 −1.789 0.478
Continuation F4 1 2.69 1.862 0.331 0.241 −1.806 0.478
Uptalk F2 3 2.67 1.886 0.334 0.241 −1.832 0.478
Uptalk F5 2 2.60 1.886 0.395 0.241 −1.804 0.478

Uptalk F6 2 2.40 1.815 0.625 0.241 −1.527 0.478
Continuation F2 6 2.18 1.720 0.931 0.241 −1.026 0.478
Continuation F3 4 2.01 1.703 1.162 0.241 −0.610 0.478
Uptalk F3 2 1.96 1.620 1.259 0.241 0.282 0.478
Uptalk F5 3 1.94 1.613 1.292 0.241 0.192 0.478
Uptalk F3 3 1.85 1.592 1.432 0.241 0.144 0.478
Uptalk F6 5 1.75 1.459 1.582 0.241 0.720 0.478
Uptalk F6 3 1.47 1.202 2.443 0.241 4.394 0.478
Continuation F5 3 1.41 1.016 2.638 0.241 6.143 0.478
(2) by the first-year group (the items ranked above 2.5)
N = 75 Mean Std. Deviation Skewness Kurtosis
Statistic Std. Error Statistic Std. Error
Fall F6 2 4.67 0.963 −3.193 0.277 9.473 0.548
Fall F4 3 4.23 1.547 −1.583 0.277 0.601 0.548
Fall F1 2 4.23 1.539 −1.605 0.277 0.679 0.548
Fall F5 1 4.07 1.588 −1.297 0.277 −0.127 0.548
Uptalk F2 5 3.31 1.910 −0.334 0.277 −1.864 0.548
Uptalk F2 8 3.17 1.920 −0.195 0.277 −1.940 0.548
► Fall F2 3 3.04 1.941 −0.046 0.277 −1.982 0.548
Continuation F1 3 2.85 1.936 0.144 0.277 −1.975 0.548
Continuation F3 5 2.83 1.913 0.195 0.277 −1.925 0.548
Continuation F6 2 2.80 1.801 0.180 0.277 −1.797 0.548
Continuation F4 1 2.68 1.861 0.354 0.277 −1.801 0.548
Uptalk F2 3 2.59 1.839 0.421 0.277 −1.736 0.548
Uptalk F5 2 2.45 1.848 0.545 0.277 −1.666 0.548

Uptalk F6 2 2.40 1.808 0.617 0.277 −1.537 0.548
(3) by the fourth-year group (the items ranked above 2.5)
N = 25 Mean Std. Deviation Skewness Kurtosis
Statistic Std. Error Statistic Std. Error
Fall F6 2 4.36 1.497 −1.975 0.464 2.061 0.902
Fall F1 2 4.12 1.641 −1.438 0.464 0.177 0.902
Fall F4 3 4.04 1.744 −1.297 0.464 −0.354 0.902
Fall F5 1 3.96 1.744 −1.164 0.464 −0.621 0.902
Continuation F1 3 3.12 1.922 −0.108 0.464 −2.028 0.902
Continuation F3 5 3.08 2.040 −0.085 0.464 −2.174 0.902
Uptalk F5 2 3.04 1.968 −0.025 0.464 −2.099 0.902
Uptalk F2 3 2.92 2.040 0.085 0.464 −2.174 0.902
Uptalk F2 5 2.92 2.040 0.085 0.464 −2.174 0.902
► Fall F2 3 2.84 1.993 0.171 0.464 −2.098 0.902
Continuation F4 1 2.72 1.904 0.280 0.464 −1.951 0.902
Continuation F2 6 2.48 1.939 0.592 0.464 −1.768 0.902

Continuation F6 2 2.40 1.915 0.650 0.464 −1.689 0.902

As expected, of the five IUs with a final falling tone, four were ranked the highest for being sentence final. However, one fall was ranked lower than one of the uptalk IUs, and immediately followed by another uptalk IU. Moreover, two more uptalk IUs, as well as four of the sentence-medial continuation rises obtained mean scores higher than 2.5 (half the maximum score of 5).

No statistically significant differences were found (Mann-Whitney U-test) between the two groups of listeners in this respect, but with lower-level listeners two uptalk IUs ranked higher than one of the final-fall IUs, while in the higher-level group as many as three uptalk IUs and two continuation IUs were ranked higher than one phrase with a final fall, as shown in Table 10 (2) and (3).

For the two uptalk phrases ranked the highest, the participants’ decisions – in the whole corpus as well as in the two groups – were rather evenly split between ‘not final’ and ‘final’. For instance, the distribution of answers for highest-ranked uptalk IU (Uptalk F2 5) in the whole population was: 42% not final, with 1% probably not, only 2% undecided, and 51% final, with 4% probably final. Although the difference was not statistically significant, raw data showed that lower-level students classified this phrase as final somewhat more readily than higher-level students, as can be seen in Figure 7, which shows the distribution of answers for this IU (Uptalk F2 5) in the lower-level group (Left) – 39% not final, 1% probably not, 3% can’t tell, 5% probably final, 52% final, and higher-level group (Right) – 52% not final, 48% final. Most importantly, whether they interpreted it as final or not, students showed little hesitation and uncertainty about their interpretation, judging by few or no answers (in the higher-level group) in the middle part of the scale.

Figure 7: 
The frequency distribution of rankings for an uptalk phrase from 1 = certainly not final to 5 = certainly final. Left: first-year group. Right: fourth-year group.
Figure 7:

The frequency distribution of rankings for an uptalk phrase from 1 = certainly not final to 5 = certainly final. Left: first-year group. Right: fourth-year group.

The ten uptalk and seven continuation rises selected for the listening materials were rather representative of the production corpus. For instance, the mean values for all the continuation rises were lower than the mean values of all uptalk rises for pitch excursion (in ST: continuation = 7.81, uptalk = 9.58), and the rise F0 maximum (correlation with the rise type p = 0.037). In the listening task, of the rises ranked the highest (4 uptalk and 4 continuation rises, cf. Table 10), although both types were ranked relatively high as possibly final (over M = 2.50), the mean ranking value for all these continuations (M = 2.54) was lower than the mean value for all these uptalk rises (M = 2.77), in the whole population of listeners, as well in either group separately (First year: Cont M = 2.58, Uptalk = 2.79; Fourth year: Cont M = 2.44, Uptalk = 2.74).

However, no statistically significant correlations were found between the phonetic properties highlighted as relevant in the production corpus rises and the listeners’ ranking of IUs for being final.

4 Discussion and conclusion

The results of the production part of the study confirmed that the EFL students’ uptalk rises should be construed as distinct declarative-sentence finality signals, consistently different from other types of rises, and characterized by several relevant phonetic properties: a larger pitch excursion, a steeper rise slope, a longer pause after the uptalk rise, and a higher F0 peak than in other rises, except in questions.

Uptalk rises were found to be wider and steeper than those in questions, but both were followed by longer pauses than continuations or lists. Since the corpus consisted of expository speech, the number of polar questions investigated was too small to allow for generalizations, so the comparison of the phonetic realizations of uptalk and polar questions remains an issue to be studied further, particularly taking into consideration different communicative functions of polar questions, such as requests, confirmation- or information-seeking questions, or rhetorical questions, as the ones in this study. That is why, for instance, Warren (2016) points out that in the studies focussing on the interpretation of an utterance as a question or a statement, asking the listeners to state if what they heard was a ‘question’ may be a problem, since the participants may not share a common understanding of the term ‘question’ itself (Warren 2016: 180).

Although there were indications that uptalk rises may have often had a lower onset (F0 minimum), too, the findings were inconclusive, similarly to Shobbrook and House’s (2003), so this parameter calls for further investigation as well. Our findings did not support the claim that an uptalk contour involves a ‘moderate’ rise. Guy and Vonwiller (1984) and Ladd (1996: 123 in Sando 2009: 2) described the HRT contour as rising about 40%’ from the rise onset, but in our corpus, the uptalk contours rose higher, over 60% from the rise onset and in many cases even more, while the continuation rises were more ‘moderate’, rising about 40% from the onset.

With respect to gender, the findings showed that both female and male speakers produced phonetically distinct uptalk contours, but with some gender-specific differences in their phonetic realization. Female participants, in addition to speaking with an overall wider pitch range, produced uptalk contours with a wider excursion and a steeper slope, while male participants produced longer pauses after the uptalk rises. Our findings are in this respect similar to Warren and Daly’s (2005), as their study also found steeper rise slopes for female than for male speakers in New Zealand English, as also found for Southern British English by Barry (2007), where female uptalk rises had larger excursions than in male uptalk. Our participants were in this respect also similar to South Californian speakers, where females showed greater uptalk pitch excursion (Ritchart and Arvaniti 2013), and produced steeper uptalk rises than males (Armstrong et al. 2015). However, South Californian speakers (Ritchart and Arvaniti 2013) generally produced uptalk statements with a significantly later alignment than in questions, and female speakers showed even later alignment than males, while in the present study this distinction was not found to be significant, although female participants indeed used more nuclear falls and fall-rises. As observed by Armstrong et al. (2015), females produced very long rises, notably longer than those by male speakers, whereas this gender distinction was not found for Massachusetts L1 speakers. In our study, the duration of the rise was not found to be a significant uptalk property for either gender.

Echoing Prechtel and Clopper (2016), who investigated possible gender effects on uptalk usage in northern and southern Midwest dialects in the USA, we can also conclude that uptalk usage may be “less gendered” than suggested by previous research (Prechtel and Clopper 2016: 136), since in the present study, despite the phonetic differences in the realization of prosodic cues, both male and female participants followed the same patterns in distinguishing between different rise types, and both produced distinct uptalk contours.

The rise span expressed as the ratio of the IU pitch range, proposed by Di Gioacchino and Crook Jessop (2011), did not show conclusive findings. In question rises, the ratio was the lowest, suggesting that the final rise was just a part (about a half) of the overall IU pitch range. In all other rises, however, the ratio was much higher, close to one, indicating that, irrespective of its use type, the rise was often the widest part of the IU pitch range. This resonates with the findings reported by Mennen and colleagues, who investigated the pitch range of German learners of L2 English (Mennen et al. 2014), and found that, unlike L1 English speakers, German speakers expanded the pitch range in later parts of IUs (Mennen et al. 2014: 309), similar to Serbian EFL learners in the present study.

Another point of interest for further research is the finding that 56% of the rising contours investigated in this study contained a plateau. On the one hand, some research suggests that the use of plateaux might be a rather general characteristic of EFL intonation. Investigating pitch peak alignment in Japanese and Spanish learners of English, Graham and Post (2018) found that about 14% of their tokens were produced with an “extended plateau”, significantly more frequently than by L1 American English speakers (Graham and Post 2018: 8). On the other hand, the use of stepped-up plateaux instead of rising contours observed in the present study could also be ascribed to the participants’ L1 transfer. Namely, as shown by Lončar Raičević and Sudimac (2017a, 2017b), in the participants’ native variety of Serbian, the Prizren-Juzna Morava dialect, lexical nuclear accents tend to be phonetically realised with little pitch movement on the nuclear syllable, and with a slight step-up between the accented and post-accented syllable. With EFL speakers of this particular L1 Serbian background the presence of plateaux in the realization of pitch movement has also been documented in falling intonation contours (Paunović 2019). Therefore, this phonetic feature of EFL students’ uptalk requires further investigation, particularly bearing in mind that no other evidence of L1 transfer was observed in this study.

That uptalk should be recognized as a legitimate sentence-finality signal in EFL speech was confirmed in the perception part of the study, albeit not directly. The lack of statistically significant correlations between the distinct phonetic properties and the finality ranking of IUs suggests that listeners’ interpretation of the speaker’s communicative intents is complex, and that phonetic cues do not translate into perception in a simple and direct way, as shown in previous research, too (cf. Warren 2017). Therefore, the lack of statistical correlations between the listeners’ ranking of rising IUs as possibly final and the phonetic properties undoubtedly highlighted as characteristic of uptalk in the production part of the study can be explained by the methodological design of the listening task, i.e. the lack of context for the ranked IUs. The context – both syntactic, prosodic, and broader discourse – proved to be necessary for the correct syntactic interpretation of uptalk, as it is for other prosodic features in speech. But with respect to our research question, two findings – the high ranking as sentence-final of several uptalk phrases and even some continuation rises, as well as the participants’ confidence about their interpretations – indicate that for the participants a rising contour at the end of a declarative statement was considered a legitimate and quite possible sentence-ending prosodic signal. That is why the Cronbach’s Alpha (=0.459) for the listening questionnaire, although rather low, was considered satisfactory for this part of the research, as our aim was not to check the correctness of the participants’ interpretation – which would be impossible with phrases taken out of their immediate context – but to explore whether the listeners would allow for the possibility that declarative statements may end in a rising intonation contour at all, which they did.

The fact that the lower-level participants were more prone to this interpretation could be attributed to the background of the study, i.e. the fact that higher-level participants could have listened to the samples with a sort of EFL-teaching bias, relying on their formal textbook knowledge about English intonation (e.g. Cruttenden 1995; Wells 2006) which made them aware that uptalk was “incorrect”. However, this observation requires further research, particularly with qualitative methods that could explore the reasons and rationales motivating the participants’ decisions in interpreting the prosodic signal of uptalk. Also, while the present study focused on the phonetic identity of uptalk as a distinct prosodic signal used and recognized by EFL students, further research is needed to explore a wider range of contexts it is used and interpreted in, as well as the attitudes uptalk triggers in the listeners (e.g. Paunović, unpublished), particularly those listeners most likely to be reluctant to this intonational innovation – foreign language teachers.

Despite its limitations, the present study pointed to some relevant implications. Firstly, it showed that in EFL specifically, and L2 more generally, the notions of “standard” and “correct” promoted through the formal educational contexts should be recognized as dynamic, flexible and open to contemporary changes. Unlike new developments in grammar and vocabulary, which are readily accepted through the use of current authentic materials, novel prosodic and socio-pragmatic aspects of L2 seem to be recognized less willingly, and allowed into foreign language classrooms much less readily, even when L2 learners acquire them outside the classroom through L2 exposure and authentic use.

The example of EFL students’ use of uptalk illustrates this clearly. The results of this study showed that in EFL students’ speech uptalk was indeed a novel prosodic device, distinctively and consistently produced, but in the typical EFL classroom it would most likely still be either ignored or misinterpreted. In the L1 classroom setting, Shepherd (2011) found that students used rising intonation on declarative answers “when they anticipated rejection”, i.e. as “a pragmatic face-saving strategy” (Shepherd 2011: 3), with a specific socio-pragmatic meaning and function. But in L2 classrooms, the question of whether uptalk may be used as a distinct and deliberate prosodic signal – with specific pragmatic, semantic, and communicative functions – is barely even asked yet. Uptalk would still be seen as the “error” of the “replacement of rises with falls and vice versa” (cf. Mennen 2007: 55), attributed to students’ inadequate knowledge of English intonation (e.g. Widagsa 2014), and usually advised to be avoided or at least not be “overused” (Wells 2006: 38). However, the findings of this study show that it may indeed be time for EFL teachers and curriculum makers to “make linguistic peace” with their ‘uptalking’ students (Liberman 2008), to recognize uptalk as a legitimate prosodic device even in the formal setting of the L2 classroom, and show more sensitivity for ongoing prosodic and socio-pragmatic changes.

More generally, research suggests that uptalk seems to be spreading more globally, not only across different English varieties and communicative contexts, but, most importantly, into other languages. This is particularly (but not exclusively) prominent among younger generations of speakers, who seem to be “appropriating” this prosodic device as a general socio-pragmatic variable indexing meanings and functions they find important and current. Therefore, this socio-pragmatic prosodic change in progress seems to be gaining a wider relevance in language learning and teaching, as an important current issue for further research in languages other than English, as well.

Corresponding author: Tatjana Paunović, English Department, Faculty of Philosophy, University of Niš, Ćirila i Metodija 2, 18 000 Niš, Serbia, E-mail:

Funding source: University of Niš, Faculty of Philosophy, Serbia

Award Identifier / Grant number: Project No. 455/1-1-1-01, English Studies in the D


This study was part of the project (No. 455/1-1-1-01) English Studies in the Digital Age, conducted at the University of Niš, Faculty of Philosophy, Serbia.

  1. Research funding: This work was financially supported by the University of Niš, Faculty of Philosophy, Serbia (Project No. 455/1-1-1-01, English Studies in the D).

  2. Statement of Ethics: The participants gave their written informed consent after the recording sessions, because informing them about the research in advance could influence their oral production. The explicit approval of the Faculty’s Committee on research ethics was not required, as the data collection procedure involved only common classroom tasks regularly performed by students as part of their classwork.

  3. Disclosure Statement: The author has no conflict of interest to declare.

Appendix 1

The phrases (IUs) selected for the perception part of the study

Fall F1 2 …than in any other part of our body. ↘
Fall F2 3 …it was in the classic period. ↘
Fall F4 3 …can lead to their extinction. ↘
Fall F5 1 …in everyday communication. ↘
Fall F6 2 …than to talk face to face. ↘
Continuation F1 3 …and the third is about our body posture, … ↗
Continuation F2 6 …on the Pacific coast, … ↗
Continuation F3 4 …in a creative activity, … ↗
Continuation F3 5 …to get focused, … ↗
Continuation F4 1 …cutting down trees indiscriminately, … ↗
Continuation F5 3 …sounds have cycles, … ↗
Continuation F6 2 …leads to depression, … ↗
UPTALK F2 3 …Maya was divided into. ↗
UPTALK F2 5 …handing you out the booklets. ↗
UPTALK F2 8 …and chili peppers. ↗
UPTALK F3 2 …effects of creativity. ↗
UPTALK F3 3 …everything I would like to say. ↗
UPTALK F5 2 Thank you all for coming. ↗
UPTALK F5 3 …and that it was interesting. ↗
UPTALK F6 2 …don’t have to be shared. ↗
UPTALK F6 3 …identifiable information on Facebook. ↗
UPTALK F6 5 …to the end of my presentation. ↗

Appendix 2

Means, Standard Deviations, Medians and 95% Confidence Intervals for 12 of the variables

Mean 95% Confidence interval Std. Deviation Median
Lower bound Upper bound
Rise EXCURSION in HZ Lists 16.95 8.5717 25.3283 17.90 9.0000
Questions 30.3333 3.5399 57.1267 10.78579 35.0000
Continuation 64.3889 58.6132 70.1646 35.06280 63.5000
Quasi-uptalk 1 105.2000 88.2381 122.1619 53.03661 98.5000
Quasi-uptalk 2 81.0217 69.6166 92.4268 38.40572 77.500
Uptalk 97.5714 85.5972 109.5457 47.54581 99.0000
Aggregate uptalk 94.5101 86.8683 102.1518 47.20321 91.0000
Rise EXCURSION in ST Lists 1.8440 0.7948 2.8932 2.24180 0.7600
Questions 2.0800 0.5308 3.6292 0.62362 2.0800
Continuation 5.8001 5.3516 6.2486 2.72272 5.3600
Quasi-uptalk 1 8.2070 7.2879 9.1261 2.87369 7.8000
Quasi-uptalk 2 7.5633 6.6898 8.4368 2.94146 7.3400
Uptalk 8.3289 7.6119 9.0459 2.84690 8.0000
Aggregate uptalk 8.0598 7.5930 8.5266 2.88365 7.8000
Rise EXCURSION in ERB Lists 0.3855 0.1673 0.6037 0.46619 0.1550
Questions 0.5467 0.4432 0.6501 0.04163 0.5600
Continuation 1.3893 1.2754 1.5032 0.69169 1.3000
Quasi-uptalk 1 2.1308 1.8388 2.4227 0.91285 2.0000
Quasi-uptalk 2 1.7672 1.5480 1.9863 0.73795 1.8200
Uptalk 2.0411 1.8248 2.2574 0.85874 2.0000
Aggregate uptalk 1.9806 1.8436 2.1176 0.84611 2.0000
Rise SLOPE Hz/ms Lists 0.3435 0.2126 0.4744 0.27963 0.2650
Questions 0.8767 0.2803 1.4730 0.24007 1.0000
Continuation 2.0203 1.7812 2.2595 1.45209 1.5700
Quasi-uptalk 1 2.7950 2.2054 3.3846 1.84341 2.1500
Quasi-uptalk 2 2.0191 1.6611 2.3771 1.20549 1.7000
Uptalk 2.9213 2.5005 3.3420 1.67074 2.3400
Aggregate uptalk 2.6089 2.3445 2.8732 1.63269 2.0700
Rise SLOPE ST/ms Lists 0.0445 0.0233 0.0657 0.04536 0.0250
Questions 0.0633 0.0060 0.1207 0.02309 0.0500
Continuation 0.1807 0.1616 0.1999 0.11607 0.1400
Quasi-uptalk 1 0.2178 0.1836 0.2519 0.10681 0.1950
Quasi-uptalk 2 0.1893 0.1586 0.2199 0.10313 0.1600
Uptalk 0.2570 0.2258 0.2882 0.12400 0.2300
Aggregate uptalk 0.2255 0.2067 0.2444 0.11631 0.2000
Rise SLOPE ERB/ms Lists 0.0065 0.0021 0.0109 0.00933 0.0000
Questionsa 0.0000 0.0000 0.0000 0.0000 0.0000
Continuation 0.0447 0.0396 0.0497 0.03070 0.0315
Quasi-uptalk 1 0.0574 0.0468 0.0680 0.03314 0.0450
Quasi-uptalk 2 0.0449 0.0372 0.0526 0.02581 0.0400
Uptalk 0.0610 0.0531 0.0689 0.03136 0.0500
Aggregate uptalk 0.0551 0.0501 0.0601 0.03087 0.0400
Rise span %

F0 minimum
Lists 12.2508 4.7229 19.7787 16.08470 4.4949
Questions 12.8331 2.4752 23.1910 4.16961 10.4972
Continuation 41.6281 37.6672 45.5891 24.04574 36.2534
Quasi-uptalk 1 62.8996 53.9624 71.8368 27.94481 57.0145
Quasi-uptalk 2 56.7950 48.7986 64.7913 26.92710 52.7300
Uptalk 63.1581 56.2777 70.0385 27.31980 58.4795
Aggregate uptalk 61.1242 56.6985 65.5499 27.33767 56.7251
Rise F0 maximum Lists 177.1000 157.5440 196.6560 41.78504 173.5000
Questions 297.6667 -86.0207 681.3540 154.45496 373.0000
Continuation 225.3958 215.1949 235.5968 61.92735 236.0000
Quasi-uptalk 1 268.8500 241.9891 295.7109 83.98887 268.5000
Quasi-uptalk 2 229.1957 208.7042 249.6871 69.00326 243.0000
Uptalk 250.0635 231.7605 268.3665 72.67512 270.0000
Aggregate uptalk 248.6644 236.3907 260.9382 75.81526 265.0000
Rise F0 minimum Lists 160.1500 140.3506 179.9494 42.30500 152.5000
Questions 267.3333 -89.5996 624.2663 143.68484 338.0000
Continuation 160.9514 153.6801 168.2227 44.14238 166.5000
Quasi-uptalk 1 163.6500 150.6291 176.6709 40.71385 171.0000
Quasi-uptalk 2 148.1087 134.8668 161.3506 44.59109 159.0000
Uptalk 152.4762 143.4325 161.5198 35.90932 161.0000
Aggregate uptalk 154.1275 147.6146 160.6404 40.23047 166.0000
Rise DURATION in ms Lists 42.6500 33.6216 51.6784 19.29078 40.0000
Questions 34.0000 25.0433 42.9567 3.60555 35.0000
Continuation 36.8056 34.5974 39.0137 13.40508 36.0000
Quasi-uptalk 1 41.6500 37.0956 46.2044 14.24070 38.5000
Quasi-uptalk 2 43.2174 39.5105 46.9243 12.48272 43.0000
Uptalk 36.0635 33.1171 39.0098 11.69901 33.0000
Aggregate uptalk 39.7718 37.6694 41.8742 12.98654 39.0000
Nuclear fall duration in ms Lists 0.6000 -0.6558 1.8558 2.68328 0.0000
Questionsa 0.0000 0.0000 0.0000 0.0000 0.0000
Continuation 3.9583 2.7743 5.1423 7.18767 0.0000
Quasi-uptalk 1 3.0750 1.0415 5.1085 6.35847 0.0000
Quasi-uptalk 2 3.6304 1.4123 5.8486 7.46952 0.0000
Uptalk 3.0794 1.5708 4.5879 5.99005 0.0000
Aggregate uptalk 3.2483 2.1903 4.3064 6.53567 0.0000
Plateau duration in ms Lists 23.0500 17.5448 28.5552 11.76290 23.5000
Questions 6.6667 -22.0177 35.3510 11.54701 0.0000
Continuation 6.1875 4.9282 7.4468 7.64507 0.0000
Quasi-uptalk 1 6.1750 4.0636 8.2864 6.60182 7.0000
Quasi-uptalk 2 8.5435 6.2282 10.8588 7.79660 9.0000
Uptalk 6.2222 4.7987 7.6458 5.65242 7.0000
Aggregate uptalk 6.9262 5.8459 8.0065 6.67295 8.0000
  1. aRise SLOPE ERB/dur and Nuclear fall duration are constant for the rise use type = 1.00 (yes/no questions).


Allan, Keith. 1986. Linguistic meaning. Volume two. London: Routledge & Kegan Paul.Search in Google Scholar

Ambrazaitis, Gilbert. 2005. Between fall and fall-rise: Substance-function relations in German phrase-final intonation contours. Phonetica 62(2–4). 196–214. in Google Scholar

Armstrong, Meghan, Page Piccinini & Amanda Ritchart. 2015. The phonetics and distribution of non-question rises in two varieties of American English. In Maria Wolters, Pauline Campbell, Christine DePlacido, Amy Liddell & David Owens (eds.), Proceedings of the 18th International Congress of Phonetic Sciences (paper number 0927), 1–5. Glasgow: University of Glasgow.Search in Google Scholar

Arvaniti, Amalia & Madeleine Atkins. 2016. Uptalk in Southern British English. In Jon Barnes, Alejna Brugos, Stefanie Shattuck-Hufnagel & Nanette Veilleux (eds.), Proceedings of speech prosody, vol. 8, 153–157. Boston: Boston University.Search in Google Scholar

Barry, Angela. 2007. The form, function and distribution of high rising intonation in Southern California and Southern British English. Sheffield, UK: University of Sheffield Doctoral dissertation.Search in Google Scholar

Boersma, Paul & David Weenink. 2013. Praat: Doing phonetics by computer (Version 5.2.03) [Computer software]. Available at: in Google Scholar

Bradford, Barbara. 1997. Upspeak in British English. English Today 13(3). 29–36. in Google Scholar

Britain, David. 1992. Linguistic change in intonation: The use of high rising terminals in New Zealand English. Language Variation and Change 4. 77–104. in Google Scholar

Brazil, David. 1997. The communicative value of intonation in English. Cambridge: Cambridge University Press.Search in Google Scholar

Buck, Gavin. 2016. Systematic transfer of ‘uptalk’ from English (L1) to Spanish (L2) for native English speakers. Ann Arbor: University of Michigan Honors Senior Thesis.;sequence=1 (accessed 11 April 2018).Search in Google Scholar

Chafe, Wallace. 1992. Intonation units and prominences in English natural discourses. In Proceedings of the IRCS workshop on prosody in natural speech, 41–52. Philadelphia: University of Pennsylvania, Institute for Research in Cognitive Science.Search in Google Scholar

Chafe, Wallace. 1994. Discourse, consciousness, and time. The flow and displacement of conscious experience in speaking and writing. Chicago: University of Chicago Press.Search in Google Scholar

Couper-Kuhlen, Elizabeth. 2001. Intonation and discourse: current views from within. In Deborah Schiffrin, Deborah Tannen & Heidi Hamilton (eds.), The handbook of discourse analysis, 13–34. Oxford and Malden, MA: Blackwell.Search in Google Scholar

Coupland, Nikolas. 2007. Style: Language variation and identity. Cambridge, UK: Cambridge University Press.Search in Google Scholar

Cruttenden, Alan. 1981. Falls and rises: Meanings and universals. Journal of Linguistics 17. 77–91. in Google Scholar

Cruttenden, Alan. 1995. Rises in English. In Jack W. Lewis (ed.), Studies in general and English phonetics. Essays in honour of Professor J. D. O’Connor, 155–173. London: Routledge.Search in Google Scholar

Cruttenden, Alan. 1997 [1986]. Intonation, 2nd edn. Cambridge: Cambridge University Press.Search in Google Scholar

Davies, Diane. 2013 [2005]. Varieties of Modern English: An introduction. London: Routledge.Search in Google Scholar

Di Gioacchino, Martina & Lorena Crook Jessop. 2011. Uptalk: Towards a quantitative analysis. Toronto Working Papers in Linguistics 33. 1–16.Search in Google Scholar

Dombrowski, Ernst & Oliver Niebuhr. 2005. Acoustic patterns and communicative functions of phrase-final F0 rises in German: Activating and restricting contours. Phonetica 62(2–4). 176–195. in Google Scholar

Eckert, Penelope. 2019a. The limits of meaning: Social indexicality, variation, and the cline of interiority. Language 95(4). 751–776. in Google Scholar

Eckert, Penelope. 2019b. The individual in the semiotic landscape. Glossa: A Journal of General Linguistics 4(1). 1–15. in Google Scholar

Eckert, Penelope & William Labov. 2017. Phonetics, phonology and social meaning. Journal of Sociolinguistics 21(4). 467–496. in Google Scholar

Fey, Fabian & Natalie Lewandowski. 2020. Uptalk in German: Investigation on the GECO database. In Elissa Pustka, Markus Pöchtrager, Alexandra Lenz, Johanna Fanta-Jende, Julia Horvath, Luise Jansen, Julia Kamerhuber, Nicola Klingler, Hannah Leykum & John Rennison (eds.), Proceedings of the Conference Phonetics and Phonology in the German Language Area (P&P14), 54–57. Wien: University of Wien, Institut für Romanistik.Search in Google Scholar

Fletcher, Janet. 2005. Compound rises and ‘uptalk’ in spoken English dialogues. In Proceedings of the Ninth European Conference on Speech Communication and Technology – Interspeech, 1381–1384. Lisabon, Portugal.Search in Google Scholar

Fletcher, Janet & Jonathan Harrington. 2001. High-rising terminals and fall-rise tunes in Australian English. Phonetica 58(4). 215–229. in Google Scholar

Godjevac, Svetlana. 2000a. Intonation, word order, and focus projection in Serbo-Croatian. Ohio State University Doctoral dissertation.Search in Google Scholar

Godjevac, Svetlana. 2000b. An autosegmental/metrical analysis of Serbo-Croatian intonation. Ohio State Working Papers in Linguistics: Varia 54. 79–142. [Craige Roberts, Jeniffer Muller, & Tsan Huand (eds.). Columbus, OH: Ohio State University.].Search in Google Scholar

Gorman, James. 1993. On language, like, uptalk? The New York Times Magazine, August 15th. (accessed 18 April 2017).Search in Google Scholar

Graham, Calbert & Brechtje Post. 2018. Second language acquisition of intonation: Peak alignment in American English. Journal of Phonetics 66. 1–14. in Google Scholar

Grice, Martine & Stefan Bauman. 2007. An introduction to intonation – functions and models. In Jürgen Trouvain & Ulrike Gut (eds.), Non-native prosody: Phonetic description and teaching practice, 25–52. Berlin & New York: Mouton de Gruyter.Search in Google Scholar

Grice, Martine, James Sneed German & Paul Warren. 2020. Intonation systems across varieties of English. In The oxford handbook of language prosody, 285–302. Oxford: Oxford University Press.Search in Google Scholar

Guy, Gregory & Julia Vonwiller. 1984. The meaning of an intonation in Australian English. Australian Journal of Linguistics 4(1). 1–17. in Google Scholar

Hirst, Daniel. 1998. Intonation in British English. In Daniel Hirst & Albert Di Cristo (eds.), Intonation systems: a survey of twenty languages, 56–77. Cambridge: Cambridge University Press.Search in Google Scholar

Hughes, Arthur, Peter Trudgill & Dominic Watt. 2012. English accents and dialects, 5th edn. London: Hodder Education.Search in Google Scholar

Ivić, Pavle. 1956. O nekim problemima naše istorijske dijalektologije [On some problems of our historic dijalectology]. Južnoslovenski filolog XXI. 97–129.Search in Google Scholar

Ivić, Pavle. 1985. Dijalektologija srpskohrvatskog jezika: Uvod i štokavsko narečje [Serbo-Croatian Dialectology: Introduction to Stokavian], 2nd edn. Novi Sad: Matica srpska.Search in Google Scholar

Inkelas, Sharon & Draga Zec. 1988. Serbo-Croatian pitch accent: The interaction of tone, stress, and intonation. Language 64(2). 227–248. in Google Scholar

Kachru, Braj B. 1985. Standards, codification and sociolinguistic realism: The English language in the Outer Circle. In Randolph Quirk & Henry George Widdowson (eds.), English in the world: Teaching and learning the language and literatures, 11–30. Cambridge: Cambridge University Press.Search in Google Scholar

Kachru, Braj B., Yamuna Kachru & Cecil L. Nelson (eds.). 2006. The handbook of world Englishes. Malden, MA & Oxford: Blackwell.Search in Google Scholar

Kim, Ji Young & Gemma Repiso-Puigdelliura. 2021. Keeping a critical eye on majority language influence: The case of uptalk in heritage Spanish. Languages 6(1). 13–40. in Google Scholar

Ladd, Robert D. 1996. Intonational phonology. Cambridge: Cambridge University Press.Search in Google Scholar

Lehiste, Ilse & Pavle Ivić. 1986. Word and sentence prosody in Serbocroatian, vol. 899. Cambridge, MA: MIT Press.Search in Google Scholar

Levon, Erez. 2016. Gender, interaction and intonational variation: The discourse functions of High Rising Terminals in London. Journal of Sociolinguistics 20(2). 133–163. in Google Scholar

Levon, Erez. 2020. Same difference: The phonetic shape of high rising terminals in London. English Language and Linguistics 24(1). 49–73. in Google Scholar

Liberman, Mark. 2006. Uptalk is not HRT. Language Log March 28th.∼myl/languagelog/archives/002967.html (accessed 16 July 2016).Search in Google Scholar

Liberman, Mark. 2008. Uptalk anxiety. Language Log September 7th. (accessed 16 July 2016).Search in Google Scholar

Lindsey, Geoff. 2019. English after RP: Standard British pronunciation today. Cham: Palgrave Macmillan.Search in Google Scholar

Lončar Raičević, Aleksandra & Nina Sudimac. 2017a. Akustički opis naglaska u govorima prizrensko-južnomoravskog dijalekta [Acoustic description of lexical accents in the varieties of the Prizren-Južna Morava dialect of Serbian]. Zbornik Matice srpske za filologiju i lingvistiku LX/2. 209–225.Search in Google Scholar

Lončar Raičević, Aleksandra & Nina Sudimac. 2017b. Acoustic parameters of speech as forensic markers for speakers of Prizren-Timok dialect. In Biljana Simeunović-Patić (ed.), Thematic proceedings of the International scientific conference Archibald Reiss Days, vol. I, 465–475. Belgrade: Academy of Criminalistic and Police Studies.Search in Google Scholar

Marković, Maja & Dejan Sredojević. 2021. Spectral characteristics of stressed vowels in Standard Serbian: Quantity and quality relations. Zbornik Matice srpske za filologiju i lingvistiku 64(2). 53–76. in Google Scholar

Mata, Ana Isabel, Helena Moniz & Fernando Batista. 2016. Stylistic variation in the intonation of European Portuguese teenagers and adults. In Meghan Armstrong, Nicholas Henriksen & Maria del Mar Vanrell (eds.), Intonational Grammar in Ibero-Romance: Approaches across linguistic subfields, 45–68. Amsterdam: John Benjamins Publishing Company.Search in Google Scholar

Mennen, Ineke. 2007. Phonological and phonetic influences in non-native Intonation. In Jürgen Trouvain & Ulrike Gut (eds.), Non-native prosody: Phonetic description and teaching practice, 53–76. Berlin & New York: Mouton de Gruyter.Search in Google Scholar

Mennen, Ineke, Felix Schaeffler & Catherine Dickie. 2014. Second language acquisition of pitch range in German learners of English. Studies in Second Language Acquisition 36. 303–329. in Google Scholar

Ogden, Richard & Sara Routarinne. 2005. The communicative functions of final rises in Finnish intonation. Phonetica 62(2–4). 160–175. in Google Scholar

Paunović, Tatjana. 2015. Pitch height and pitch range in Serbian EFL students’ reading and speaking tasks. Nasleđe, časopis za književnost, jezik, umetnost i kulturu XII/32. 73–94.Search in Google Scholar

Paunović, Tatjana. 2019. Focus on Focus: Prosodic signals of utterance-level information structure in L1 Serbian, L1 English, and Serbian L2 English. Zbornik Matice srpske za filologiju i lingvistiku LXII/2. 213–238.Search in Google Scholar

Paunović, Tatjana. 2020. EFL students’ spontaneous speech: Tonality, tonicity, and tone. Belgrade English Language and Literature Studies – BELLS XII. 51–76. in Google Scholar

Peco, Asim. 1985. Osnovi akcentologije srpskohrvatskog jezika [Preliminaries to Serbo-Croatian Accentology], 3rd edn. Beograd: Naučna knjiga.Search in Google Scholar

Pierrehumbert, Janet. 1980. The phonology and phonetics of English intonation. MIT Doctoral dissertation.Search in Google Scholar

Pierrehumbert, Janet & Julia Hirschberg. 1990. The meaning of intonational contours in discourse. In Phil Cohen, Jerry Morgan & Martha Pollack (eds.), Intentions in communication, 271–311. Cambridge, Mass: MIT Press.Search in Google Scholar

Pletikos, Elenmari. 2008. Akustički opis hrvatske prozodije riječi [Acoustic description of word prosody in Croatian]. Sveučilište u Zagrebu: Filozofski fakultet Doctoral dissertation.Search in Google Scholar

Prechtel, Christine. 2015. Effects of gender and regional dialect on uptalk in the American Midwest. Columbus: Ohio State University undergraduate research thesis for Research Distinction in Linguistics.Search in Google Scholar

Prechtel, Christine & Cynthia Clopper. 2016. Uptalk in Midwestern American English. In Jon Barnes, Alejna Burgos, Stefanie Shattuck-Hufnagel & Nanette Veilleux (eds.), Proceedings of the 8th International Conference on Speech Prosody. International Speech Communication Association (ISCA) Online Archive, 133–137. Boston, MA: Boston University.Search in Google Scholar

Ritchart, Amanda & Amalia Arvaniti. 2013. The use of high rise terminals in Southern Californian English. In Proceedings of Meetings on Acoustics – The 166th meeting of the Acoustical Society of America, vol. 20, 060001.Search in Google Scholar

Sando, Yves Talla. 2009. Upspeak across Canadian English accents: Acoustic and sociophonetic evidence. In Proceedings of the 2009 annual conference of the Canadian Linguistic Association, 1–12. Ottawa, Ontario. (accessed 28 September 2016).Search in Google Scholar

Shepherd, Michael. 2011. Functional significance of rising-intonation declaratives in settings with special discursive norms. Linguistic Society of America Annual Meeting Extended Abstracts 2. 10–11. in Google Scholar

Shobbrook, Katherine & Jill House. 2003. High rising tones in southern British English. In Proceedings of the 15th international congress of phonetic sciences, 1273–1276. Barcelona.Search in Google Scholar

Shokeir, Vanessa. 2008. Evidence for the stable use of uptalk in Ontario English. University of Pennsylvania Working Papers in Linguistics 14(2). 16–24.Search in Google Scholar

Smiljanic, Rajka. 2004. Lexical, pragmatic and positional effects on prosody in two dialects of Croatian and Serbian: An acoustic study. New York & London: Routledge.Search in Google Scholar

Smiljanić, Rajka. 2006. Early vs. late focus: Pitch-peak alignment in two dialects of Serbian and Croatian. Laboratory Phonology 8. 495–518.Search in Google Scholar

Sredojević, Dejan. 2011. Eksperimentаlno-fonetsko ispitivаnje prozodijskih kаrаkteristikа novosаdskog govorа [Experimental-phonetic investigation of the prosodic characteristics of the Novi Sad speech]. Faculty of Philosophy, University of Novi Sad Doctoral dissertation.Search in Google Scholar

Sredojević, Dejan. 2017. Fonetsko-fonološki opis akcenata u standardnom srpskom jeziku – od specifičnog ka opštem [Phonetic and phonological description of accents in standard Serbian – from the specific to the general]. Novi Sad: Filozofski fakultet.Search in Google Scholar

Sredojević, Dejan & Maja Marković. 2020. Qualitative and quantitative distinctions between lexical accents of standard Serbian in the speech of Novi Sad (minimal pair corpus). Zbornik Matice srpske za filologiju i lingvistiku LXIII/2. 33–53. in Google Scholar

Sredojević, Dejan. 2022. Dikcija. O glasu, glasovima, akcentu i prozodiji u srpskom jeziku [Diction: On voice, sounds, accents and prosody in Serbian]. Novi Sad: Filozofski fakultet.Search in Google Scholar

Tanaka, Lidia, Kaori Okano, Ikuko Nakane, Claire Maree, Shimako Iwasaki & Chie Takagi. 2021. Japanese women’s speech through life‐transitions (1989‐2000): An analysis of youth language features. Journal of Linguistic Anthropology 31(1). 119–143. in Google Scholar

TextSTAT. v. 3.0. [Computer program]. Available at: in Google Scholar

Tomic, Kristina. 2020. Acoustic analysis of pitch accent as a regional forensic marker in Serbian. Facta Universitatis: Linguistics and Literature 18(2). 235–256. in Google Scholar

Warren, Paul. 2005. Patterns of late rising in New Zealand: Intonational variation or intonational change? Language Variation and Change 17(2). 209–230.Search in Google Scholar

Warren, Paul. 2016. Uptalk: The phenomenon of rising intonation. Cambridge & New York: Cambridge University Press.Search in Google Scholar

Warren, Paul. 2017. The interpretation of prosodic variability in the context of accompanying sociophonetic cues. Laboratory Phonology: Journal of the Association for Laboratory Phonology 8(1). 1–21. in Google Scholar

Warren, Paul & Nicola Daly. 2005. Characterizing New Zealand English intonation: Broad and narrow analysis. In Allan Bell, Ray Harlow & Donna Sparks (eds.), Languages of New Zealand. Wellington: Victoria University Press.Search in Google Scholar

Warren, Paul & Janet Fletcher. 2016. Phonetic differences between uptalk and question rises in two Antipodean English varieties. In Proceedings of the International Conference on Speech Prosody 2016, 148–152. Boston, MA: Boston University.Search in Google Scholar

Wells, John C. 2006. English intonation: An introduction. Cambridge: Cambridge University Press.Search in Google Scholar

Wichmann, Anne. 2013 [2001]. Intonation in text and discourse: Beginnings, middles and ends. Abingdon & New York: Routledge.Search in Google Scholar

Widagsa, Rudha. 2014. Intonation of English declarative sentence produced by English Department students. ELTICS Journal 1(1). 35–43. in Google Scholar

Wilhelm, Stephan. 2016. Towards a typological classification and description n of HRTs in a multidialectal corpus of contemporary English. In Proceedings of Speech Prosody 2016, 138–142. Boston, MA: Boston University.Search in Google Scholar

Zec, Draga & Elizabeth Zsiga. 2009. Interaction of tone and stress in Standard Serbian: Phonological and phonetic evidence. In Wayles Browne, Adam Cooper, Alison Fisher, Esra Kesici, Nikola Predolac & Draga Zec (eds.), Formal approaches to Slavic linguistics 18: The Cornell Meeting 2008, 537–555. Ann Arbor: Michigan Slavic Publications.Search in Google Scholar

Zec, Draga & Elizabeth Zsiga. 2012. Contextual evidence for the representation of pitch accents in standard Serbian. Language and Speech 56(1). 69–104. in Google Scholar

Received: 2022-08-06
Accepted: 2023-02-05
Published Online: 2023-03-16
Published in Print: 2022-12-16

© 2023 Walter de Gruyter GmbH, Berlin/Boston

Downloaded on 4.10.2023 from
Scroll to top button