2.1 Echo questions versus information seeking questions

Although wh-echo and information seeking questions can be string-identical, it is clear from the cross-linguistic literature that they display different properties in several components of the grammar.

From a syntactic point of view, the most obvious property of echo questions is that their wh-phrase can be left in-situ in many languages that obligatorily front the wh-phrase in information seeking questions, e.g. John bought what? (Reis Reference Reis and Rosengren1992, Artstein Reference Artstein2002). Another syntactic difference is the possibility to replace only part of a constituent by a wh-phrase, e.g. John bought a what? (Reis Reference Reis and Rosengren1992, Reference Reis2012; Artstein Reference Artstein2002; Sobin Reference Sobin2010).Footnote ^[4]

Pragmatically, the use of an echo rather than an information seeking question signals that the speaker of the echo question does not yet accept a previous discourse move (Ginzburg & Sag Reference Ginzburg and Sag2000, Engdahl Reference Engdahl, Bonami and Hofherr2006, Poschmann Reference Poschmann2015, Biezma Reference Biezma2018). If the speaker of the echo question did not understand or perceive part of the previous utterance, s/he is not yet in a position to accept it; if the speaker is surprised by part of the previous utterance or does not believe it, s/he refuses to accept (part of) it for that reason.

Semantically, an information seeking wh-question is generally taken to be the set of propositions that constitute possible (true) answers to it, as depicted in (3) (Hamblin Reference Hamblin1973, Karttunen Reference Karttunen1977). An echo question raises a question regarding the meaning of a previous utterance, e.g. ‘What did you assert just now that John bought?’. Hence, the meaning of an echo question can be analysed as the set of propositions expressing the potential content of the preceding utterance, as in (4) (Ginzburg & Sag Reference Ginzburg and Sag2000, Engdahl Reference Engdahl, Bonami and Hofherr2006, Poschmann Reference Poschmann2015, Biezma Reference Biezma2018). (In accordance with standard notation, the denotation of an utterance is placed between and a set of propositions is indicated by { }.)

As echo questions clearly differ from information seeking questions regarding their syntax and their semantico-pragmatic properties, the question rises whether they display different prosodic properties as well.

2.2 The prosody of French wh-in-situ echo and information seeking questions

It is unknown whether speakers distinguish echo and information seeking questions by prosodic means in French. Previous research investigating the prosody of information seeking wh-in-situ questions has yielded conflicting results. French wh-in-situ echo questions have not yet been studied in much detail.

Most of the debate surrounding the prosody of French wh-in-situ information seeking questions has centred on the question whether the utterances obligatorily end in a large rise (i.e. a rise with a large pitch excursion), as proposed by Cheng & Rooryck (Reference Cheng and Rooryck2000). In contrast, some authors claim that they standardly end in a fall (Di Cristo Reference Di Cristo, Hirst and Di Cristo1998; Starke Reference Starke2001: 23; Mathieu Reference Mathieu2002, Reference Mathieu, Eguren, Fernandez-Soriano and Mendikoetxea2016). A third position maintains that a sentence-final rise is possible, but optional rather than mandatory (Wunderli Reference Wunderli1978, Reference Wunderli, Winkelmann and Braisch1982, Reference Wunderli1983; Wunderli & Braselmann Reference Wunderli and Braselmann1980, Adli Reference Adli, Meisenburg and Selig2004, Reference Adli2006; Di Cristo Reference Di Cristo2016). Testing Cheng & Rooryck’s (Reference Cheng and Rooryck2000) claim experimentally, Déprez, Syrett & Kawahara (Reference Déprez, Syrett and Kawahara2013) concluded that the majority of speakers uttered information seeking wh-in-situ questions with a small final rise, although this result was not replicated by Tual (Reference Tual2017). Reinhardt (Reference Reinhardt2019) shows in two corpus studies that while French wh-in-situ questions have a tendency to display a sentence-final rise, this is not a strict constraint. Her studies settle the debate to some extent, but do not yet explain the observed variation.

Another line of research has focussed on the prosody associated with the wh-phrase. According to Baunaz & Patin (Reference Baunaz and Patin2009) and Baunaz (Reference Baunaz2016), the wh-phrase does not bear any accent when a wh-in-situ question is uttered out of the blue (see also Wunderli & Braselmann Reference Wunderli and Braselmann1980, Wunderli Reference Wunderli, Winkelmann and Braisch1982). In contrast, Gryllia, Cheng & Doetjes Gryllia et al. (Reference Gryllia, Cheng and Doetjes2016) systematically found an emphatic pitch accent in questions uttered out of the blue as well as a rise at the end of the wh-phrase (see also Wunderli Reference Wunderli1983, Zubizarreta Reference Zubizarreta1998, Mathieu Reference Mathieu2002, Engdahl Reference Engdahl, Bonami and Hofherr2006, Hamlaoui Reference Hamlaoui2011).

So far, no studies have been designed to investigate the prosody of French wh-in-situ echo questions. There have been some brief descriptions, some of which describe echo questions as displaying a higher pitch overall (Di Cristo Reference Di Cristo, Hirst and Di Cristo1998, Boeckx Reference Boeckx1999) or a sentence-final rise (Di Cristo Reference Di Cristo, Hirst and Di Cristo1998, Reference Di Cristo2016; Boeckx Reference Boeckx1999; Mathieu Reference Mathieu2002, Reference Mathieu, Eguren, Fernandez-Soriano and Mendikoetxea2016; Adli Reference Adli2006; Engdahl Reference Engdahl, Bonami and Hofherr2006; Déprez et al. Reference Déprez, Syrett and Kawahara2013). Also, echo questions are described as exhibiting a prominent accent on the wh-word (Chang Reference Chang1997: 17; Mathieu Reference Mathieu2002; Engdahl Reference Engdahl, Bonami and Hofherr2006). Engdahl (Reference Engdahl, Bonami and Hofherr2006) mentions that the wh-word may also be lengthened. Déprez et al. (Reference Déprez, Syrett and Kawahara2013) investigated the prosody of echo questions in relation to information seeking questions experimentally, focussing exclusively on the final part of the utterance. This study offers some first evidence suggesting that the final rise in echo questions may be present more consistently and may display a somewhat larger pitch excursion. However, the methodology of the study does not allow for a statistical comparison of the question types, nor for a mapping of the F0 (i.e. pitch) movements to individual syllables.

In summary, it is not clear (i) what the prosodic properties are of wh-in-situ information seeking questions and (ii) whether these properties differ from the ones of wh-in situ echo questions, and if so, how: a final rise and an accent on the wh-phrase have been claimed for both question types.

2.3 Cross-linguistic comparison

For some languages other than French, there have been more detailed comparisons between wh-in-situ echo questions expressing auditory failure and information seeking questions. Although the reported prosodic properties of echo questions differ cross-linguistically, echo questions seem to be uttered with a distinct prosody in several languages. The following, tentative, generalisation seems to hold within the small sample of languages for which relevant descriptions could be found:

Brazilian Portuguese (Kato Reference Kato, Camacho-Taboada, Fernández, Martín-González and Reyes-Tejedor2013), Farsi (Esposito & Barjam Reference Esposito and Barjam2007, Sadat-Tehrani Reference Sadat-Tehrani2011) and Manado Malay (Stoel Reference Stoel, van Heuven and van Zanten2007) are examples of pattern (A); pattern (B) is exemplified by North-Central Peninsular Spanish (González & Reglero Reference González, Reglero, Repetti and Ordóñez2018), Greek (Roussou, Vlachos & Papazachariou Reference Roussou, Vlachos, Papazachariou, Lavidas, Alexiou and Sougari2014) and Shingazidja, a Bantu language spoken on Comoros (Patin Reference Patin2011). German also follows pattern (B), but the difference between the question types is very small (Repp & Rosin Reference Repp and Rosin2015), possibly because information seeking wh-in-situ is very restricted in this language. Mandarin Chinese seems to be the only language for which the two question types have been compared, but no distinct prosody for echo questions was consistently found (Hu Reference Hu2002).

2.4 Focus in wh-questions

Focus has not been taken into account in the studies surveyed in the previous section. We will now motivate why we include focus in our study.

For declaratives, it is well-known that there is an interaction between the focus of a sentence and its prosodic properties (see for instance Büring Reference Büring2016 for a recent overview). Although discussed to a lesser extent, the interaction between context and prosody has also been observed in wh-questions (Jacobs Reference Jacobs1984, Erteschik-Shir Reference Erteschik-Shir1986, Jacobs Reference Jacobs1991, Lambrecht & Michaelis Reference Lambrecht and Michaelis1998, Engdahl Reference Engdahl, Bonami and Hofherr2006, Eckardt Reference Eckardt2007, Truckenbrodt Reference Truckenbrodt, Elordieta and Prieto2012, Büring Reference Büring2016). Whereas some languages mark the wh-phrase prosodically as the focus of the question, this is not the case in all languages (Ladd Reference Ladd2009: 226–227). In many languages, including English, German and also French, the neutral location for the main accent is not on the wh-word and context may affect the prosody of a wh-question. In a wh-question such as (5), adapted from Erteschik-Shir (Reference Erteschik-Shir1986: 118), who cites Gunter (Reference Gunter1966: 172), the accentuation represented in a. rather than the one in b. or c. is the most neutral one, while b. or c. impose specific restrictions on the context. For instance, (5b) might be uttered if the preceding context specifies that John ate the beans, but not at what point in time this happened. Similarly, (5c) could be uttered if the preceding context indicates the time at which John prepared the beans, but not when he ate them.

The responses of researchers to this type of data fall into two categories. Some authors stress the grammatical character of the wh-phrase as the focus of the sentence and attempt to explain the prosodic properties of wh-questions without reference to focus (Culicover & Rochemont Reference Culicover and Rochemont1983, Lambrecht & Michaelis Reference Lambrecht and Michaelis1998). Others, viewing focus from the perspective of pragmatics, propose that the relation between context and prosody illustrated in (5) can be explained in terms of focus marking, similar to in declaratives (Jacobs Reference Jacobs1984; Erteschik-Shir Reference Erteschik-Shir1986; Jacobs Reference Jacobs1991; Rosengren Reference Rosengren1991; Kadmon Reference Kadmon2001: 388–397; Reich Reference Reich2002; Beyssade Reference Beyssade2006; Engdahl Reference Engdahl, Bonami and Hofherr2006; Eckardt Reference Eckardt2007; Truckenbrodt Reference Truckenbrodt, Elordieta and Prieto2012; Büring Reference Büring2016; Di Cristo Reference Di Cristo2016). Here we adopt the second approach, following the implementation of Jacobs (Reference Jacobs1984, Reference Jacobs1991), which has been applied to French by Beyssade et al. (Reference Beyssade, Delais-Roussarie, Doetjes, Marandin and Rialland2004a, Reference Beyssade, Delais-Roussarie, Doetjes, Marandin and Riallandb).

In this approach, focus is defined as the part of the content that is ‘inhaltlich besonders betroffen’, i.e. specifically affected, by the illocutionary operator of the sentence (Jacobs Reference Jacobs1984: 30). The partition of an utterance into focus and ground takes place within the scope of an illocutionary operator. In declaratives, the focus contributes the part of the content that is specifically affected by the assertive operator, i.e. the part that is asserted in particular and that is informative. In interrogatives, the focus is specifically affected by the question operator and provides information that is not given in the discourse situation in which the question is uttered. As Jacobs points out, it is important to distinguish the existential implicature that arises from all the wh-questions in (5), namely that there is a time at which John ate the beans, from the notion of ground. In (5a), the discourse situation does not need to provide information about the contents of the question as in this case the whole question is marked as focus. In (5b) and (5c), however, the context that directly precedes the question needs to specify the information that is marked as ground by the prosody of the wh-question. As a result, (5b) can only be uttered felicitously if the information that John ate the beans is given in the context directly preceding it, as the accent on when introduces a narrow focus on the wh-word. Similarly, the contextual restrictions on the use of (5c) follow from the narrow focus on the verb eat.

For wh-echo questions, it is commonly assumed that there is a narrow focus on the wh-phrase (Jacobs Reference Jacobs1991, Bartels Reference Bartels1997, Reis Reference Reis2012): the whole non-wh part of the utterance content must belong to the ground, since it is ‘echoed’ from the previous utterance. Assuming the approach to focus outlined above, the focus structure of information seeking wh-questions may be very different, depending on the preceding context. Therefore, a comparison between the two question types should involve information seeking questions with a focus structure resembling the one in echo questions, i.e. information seeking questions with narrow focus on the wh-phrase, as well as information seeking questions in a context that is maximally different from the echo context in the sense that it gives hardly any information about the content of the question.

2.5 Prosodic focus marking in French

In languages like English, focus is marked by the presence of the nuclear pitch accent (Pierrehumbert Reference Pierrehumbert1980, Selkirk Reference Selkirk, Aronoff and Oehrle1984, Pierrehumbert & Hirschberg Reference Pierrehumbert, Hirschberg, Cohen, Morgan and Pollack1990); for French, this is less clear. The right edge of a focus is preferably aligned with (i.e. situated at) the right edge of a prosodic constituent (Clech-Darbon, Rebuschi & Rialland Reference Clech-Darbon, Rebuschi, Rialland, Rebuschi and Tuller1999; Féry Reference Féry, Féry and Sternefeld2001, Reference Féry2013; Beyssade et al. Reference Beyssade, Delais-Roussarie, Doetjes, Marandin and Rialland2004a, Reference Beyssade, Delais-Roussarie, Doetjes, Marandin and Riallandb; Hamlaoui Reference Hamlaoui2008; Delais-Roussarie et al. Reference Delais-Roussarie, Post, Avanzi, Buthke, Di Cristo, Feldhausen, Jun, Martin, Meisenburg, Rialland, Frota and Prieto2015). A focus is marked by a tone at its right edge (Martin Reference Martin1981, Rossi Reference Rossi1981); in this paper, we assume that this is the boundary tone associated with the right edge of the prosodic constituent (Intonation Phrase) with which the focus is aligned (Beyssade et al. Reference Beyssade, Delais-Roussarie, Doetjes, Marandin and Rialland2004b).Footnote ^[5]

The boundary tone at the right edge of a focus tends to be low in declaratives and high in interrogatives, reflecting the illocutionary force of the utterance (Martin Reference Martin1981, Clech-Darbon et al. Reference Clech-Darbon, Rebuschi, Rialland, Rebuschi and Tuller1999, Beyssade et al. Reference Beyssade, Delais-Roussarie, Doetjes, Marandin and Rialland2004b, Delais-Roussarie, Doetjes & Sleeman Reference Delais-Roussarie, Doetjes and Sleeman2004, Doetjes, Rebuschi & Rialland Reference Doetjes, Rebuschi and Rialland2004; Delais-Roussarie et al. Reference Delais-Roussarie, Post, Avanzi, Buthke, Di Cristo, Feldhausen, Jun, Martin, Meisenburg, Rialland, Frota and Prieto2015). In all-focus utterances and sentences in which a narrow focus occurs sentence-finally, the tone is located at the end of the utterance. In other narrow focus utterances, the tone usually occurs twice, both at the end of the focus and at the end of the utterance: this phenomenon is referred to as tone copying (Martin Reference Martin1981). Specifically, Martin claims that the F0 minimum (in declaratives) or maximum (in interrogatives) of the final syllable of the focus is copied to the final syllable of the utterance.

Whereas it is generally accepted that a tone or accent marks the right edge of a focus, it is less clear whether the pre-focus domain is marked as well. Some authors have observed a compression of the pitch (Touati Reference Touati1987, Jun & Fougeron Reference Jun, Fougeron and Botinis2000, Dohen & Lœvenbruck Reference Dohen and Lœvenbruck2004) and a reduced amplitude (Jun & Fougeron Reference Jun, Fougeron and Botinis2000). In contrast, Beyssade et al. (Reference Beyssade, Delais-Roussarie, Doetjes, Marandin and Rialland2004b) state that there is no pitch compression in the pre-focus domain.

Previous work on focus marking in French interrogatives rarely addresses wh-in-situ questions. According to Di Cristo (Reference Di Cristo2016), French wh-in-situ questions can have a narrow focus on the wh-phrase or a broad focus reading, but this difference is not reflected in the prosody. Di Cristo claims that the right edge of the wh-phrase is marked by an accent in all cases. However, in the examples he discusses, the wh-phrase is situated at the end of the sentence. This means that the right edge of the focus co-occurs with the right edge of the sentence and tone copying cannot apply.

3.1 Elicitation paradigm: Scripted Simulated Dialogue

To elicit the three conditions, we designed a paradigm that we will refer to as Scripted Simulated Dialogue. This elicitation paradigm simulates a series of short dialogues, in which the participant’s interlocutor is a recorded voice. The participant’s half of the dialogue is scripted: s/he reads his/her speech turns from a computer screen. Every dialogue has one target sentence or filler embedded in it, always at the same position in the dialogue. As this position is almost at the end of the dialogue, the preceding discourse can be used to manipulate a particular reading of the sentence. The participant does not know which is the target sentence.

Every dialogue is preceded by a description of the conversational setting, which contains information about who the interlocutors are and where the conversation takes place. The context manipulation thus has two elements: the description of the conversational setting and the preceding speech turns.

Following a dialogue, the participant receives a question about the information supplied by the recorded ‘interlocutor’. The purpose of this is to direct the participants’ attention to the content of the dialogue, rather than the form of the utterances.

3.2 Materials

Twelve target stimuli were created. As each of these was presented in three conditions, this yielded a total of thirty-six target utterances. The stimuli had twelve syllables; an example is shown in (8).

All stimuli contained the pronoun tu ‘you’ as a subject (to avoid differences in the information status of the subject), followed by a verb composed of the auxiliary as ‘has’ and a three-syllable past participle. Next came the wh-phrase, which was the direct object of the utterance. It contained the wh-word quel ‘which’ and a disyllabic noun. We chose to use complex wh-phrases (rather than, for instance, quoi ‘what’) to keep the prosody associated with the wh-word distinct from the prosodic correlates of a possible phrase boundary at the end of the wh-phrase. A PP, composed of a one-syllable preposition and a three-syllable DP, followed the wh-phrase. Its purpose was to separate the prosody associated with the wh-phrase from the prosody associated with the end of the utterance.Footnote ^[6] Sonorants were used as much as possible to facilitate F0 measurements.

The target sentences were intermingled with thirty-six fillers. Twelve of these were declaratives that resembled the discourse–pragmatic function of echo questions, such as Désolé, je n’ai pas bien entendu ‘Sorry, I didn’t hear that well’. The remaining 24 fillers were sentences that fitted naturally in the context and that were not wh-in-situ questions.

Each stimulus or filler was embedded in a dialogue. The dialogues had six speech turns in total (three for the participant and three for the recorded ‘interlocutor’). The stimulus or filler was part of the participant’s last speech turn, with the ‘interlocutor’s’ last speech turn following it. The dialogues, which were written and checked by three native speakers of French, were constructed to be natural and informal. The voice that represented the ‘interlocutor’ was a female native speaker of French, while the description of the conversational setting that preceded a dialogue was read by a male speaker (to make the distinction clear). Both were recorded in the phonetics lab at Leiden University.

Except for the dialogues containing fillers, each dialogue had certain properties that were intended to trigger either an echo question expressing auditory failure (Condition A), an information seeking question with broad focus (Condition B) or an information seeking question with narrow focus on the wh-phrase (Condition C). Figures 1, 2 and 3 present examples of dialogues used in the three conditions; in each case, the description of the conversational setting precedes the dialogue. Speaker A represents the participant and Speaker B the ‘interlocutor’. The target sentence is underlined, although this was not the case in the actual experiment.

Figure 1 Scripted Simulated Dialogue used in Condition A (echo question).

In the example in Figure 1, pink noise would cover the word Monette (represented as struck through text), causing a need to ask for repetition. An episode of pink noise was also present in all other contexts (pertaining to the other conditions and the fillers), but in a position where it would not hinder the conversation, for instance on the final syllable of a long word.

Figure 2 Scripted Simulated Dialogue used in Condition B (broad focus).

Condition B was designed to elicit ordinary information seeking wh-in-situ questions with broad focus. Although the target sentence was preceded by context, it provided little information about the content of the question. Whereas the interlocutors would have to share certain assumptions to make it natural to ask an (in-situ) wh-question, no part of the content of the question would be mentioned in the preceding context. To signal this departure from the topic of the preceding conversation, a topic change marker was used (Fraser Reference Fraser1999), like d’ailleurs tu ne m’as pas dit ‘by the way, you didn’t tell me’ in Figure 2.

Figure 3 Scripted Simulated Dialogue used in Condition C (narrow focus).

In Condition C, the context was designed to force a reading as an information seeking question with narrow focus on the wh-phrase. The context would mention all elements of the content of the question except the wh-word, i.e. ‘booking a restaurant for Thursday evening’ in Figure 3. Hence, everything but the wh-word would belong to the ground of the focus domain. In order to implement this restriction on the context, while keeping the flow of the discourse natural, we used wh-in-situ questions with a contrastive topic, as in Engdahl (Reference Engdahl, Bonami and Hofherr2006: 100). Subject pronouns in French are clitics and cannot be contrastively stressed (Kayne Reference Kayne1975). To express contrastive topichood, French uses another, ‘strong’ pronoun, which may be coreferential with a clitic (Lambrecht Reference Lambrecht1994: 115–116). We used et toi ‘and you’, which was taken up by the resumptive clitic tu ‘you’ in the clause proper. Hence, the sentence following the contrastive topic et toi ‘and you’ was string-identical to the target stimuli used in Conditions A and B.

3.3 Recording procedure

Recordings took place in a soundproof booth at Pôle Audiovisuel et Multimédia (PAM) at the University of Nantes. Participants were seated in front of a computer screen at an approximate distance of 50cm. They wore AKG K 44 perception headphones. The speech was recorded onto digital audio tape (DAT) at a sampling rate of 44.1 kHz, using a TASCAM DR-100 recorder and a TRAM TR50 clip-on microphone.

Participants were informed that they would be taking part in a series of short dialogues with a recorded ‘interlocutor’ that they would hear through their headphones and that their side of the dialogue would appear on the screen in front of them. They were encouraged to project themselves into the situation represented by the dialogue, speaking naturally, ‘as if they were just talking to someone’, and to repeat their utterance in case of a lapse. These instructions were presented visually on the computer screen and reiterated orally at the beginning of the experiment. After that, the experimenter did not intervene.

Participants pressed a key once they were ready to start the experiment. This would prompt the recording of the first conversational setting to be played through the headphones (in a male voice), while the screen was blank. Every conversational setting ended with Tu dis: ‘You say:’. Then the participant’s first speech turn would appear on the screen. The participant would utter his/her speech turn, after which s/he would press a key for the ‘interlocutor’s’ speech turn to start playing through the headphones (in a female voice) while the screen was blank again. Then the participant’s next speech turn would appear on the screen. This process would be repeated until the participant had uttered the question that formed the target sentence in his/her third speech turn (or a filler) and received an answer in the ‘interlocutor’s’ third and last speech turn. The alternation of speaking, then pressing a key and listening to the ‘interlocutor’ very soon became an automatic process. After the last speech turn of every dialogue (items or filler), a multiple choice sentence completion task would appear on the screen, asking about information supplied by the ‘interlocutor’. The participant would answer the question by pressing 1, 2 or 3. Feedback on the answer would appear on the screen: it was usually correct, since the task was designed to be easy if the participant paid attention to the ‘interlocutor’s’ speech turns. The participant would then press a key to move on to the next trial. The whole paradigm was programmed in E-Prime (Psychology Software Tools Inc. 2012).

The dialogues were randomized and presented to participants in three blocks, with breaks in between. Three practice trials were used for familiarisation purposes. The experiment lasted approximately an hour.

3.4 Participants

Twenty graduate and postgraduate students at the University of Nantes, monolingual native speakers of French, were reimbursed to participate in the experiment (13 female and 7 male, age range 18–29 years old). None of them reported any speech or hearing disorders.

3.5 Acoustic analysis

We obtained a total of 720 utterances. After inspection of the data for speech errors, hesitations or unnatural pausing, we excluded 98 utterances from further analysis. The remaining 622 utterances were segmented into phones, syllables and words using EasyAlign (Goldman Reference Goldman, Cosi, De Mori, Di Fabbrizio and Pieraccini2011). The segmental boundaries were then checked and adjusted manually where necessary.

We inspected the utterances again to uncover any patterns in the data, such as the occurrence of different prosodic tunes within the data elicited in one condition. We marked the number of occurrences of each prosodic tune.

Based on this inspection, we selected the utterances for the statistical analyses. We followed the reasoning that if a. a prosodic tune occurred in the majority of cases elicited in a particular condition and b. none of the other tunes came close to its frequency, then this prosodic tune might be considered to be the characteristic prosodic tune of utterances elicited in that condition. We included the items uttered with these characteristic prosodic tunes in the statistical analyses, with the exception of cases that exhibited the characteristic prosodic tune, but with a variation (see also Section 4.1 below). This was done to achieve a sample that was as homogeneous as possible.

We selected the following seven F0 measurement points, which are visualised in Figure 4:

Figure 4 Waveform and F0 tune of the question Tu as réservé quel resto pour jeudi soir? ‘You booked which restaurant for Thursday evening?’ elicited in Condition A (echo question), together with a textgrid illustrating the seven measurement points.

We also obtained the following F0 measurements:

We extracted the F0 values with the help of a Praat script (Boersma & Weenink Reference Boersma and Weenink2017), which took the values from the voiced part of the respective syllables. As rises in French have been shown to continue onto a sonorant syllable coda, and even (in rare cases) on a voiced obstruent coda (Welby & Lœvenbruck Reference Welby and Lœvenbruck2005), we included voiced consonants in our analysis. The F0 values in Hertz were subsequently converted into semitones (st) to reduce variation. (We used the formulas st = 12 log2 (Hz/100) for female speakers and st = 12 log2 (Hz/50) for male speakers respectively, following Li & Chen Reference Li and Chen2012.) In addition, we extracted the duration and the mean intensity in decibel (dB) of every syllable, using two more Praat scripts.

3.6 Statistical analysis

We ran a series of linear mixed-effects models using the lmer function of the lme4 package (Bates et al. Reference Bates, Mächler, Bolker and Walker2015) in R (R Core Team 2017). P-values were obtained using the package lmerTest (Kuznetsova, Brockhoff & Christensen Reference Kuznetsova, Brockhoff and Christensen2017). Specifically, we ran a model with the relevant measurement as the dependent variable, question type as a fixed factor, and items and participants as random factors for every measurement. To obtain all relevant comparisons we ran the analyses for each reference category (Echo, Broad focus, Narrow focus). The reader is referred to the statistical appendix for more details.