The Realization of Information Focus in Catalan

Abstract

Answers to wh-questions are the most widespread method to elicit information focus. When studying the syntax of focus, however, this method is problematic because the most natural answer to a wh-question is often a fragment that only includes the focus. This problem has led to considerable controversy in the literature about the position in which information focus is realized, particularly for Spanish, but also for Catalan. In order to enhance the naturalness and reliability of the question-answer test, we designed an experiment with a new elicitation technique (i.e., questions with a delayed answer) in which some material is inserted between the question and the point in which the participant is asked to answer the question, so that the Catalan participants would spontaneously utter a full sentence instead of a fragment, without being explicitly instructed to do so. The material of this production experiment was then adapted in a rating experiment on the acceptability of preverbal and postverbal information foci in Catalan. The results of this second study confirm the findings of the production experiment: postverbal focus is always preferred over preverbal focus, both in the case of subjects and objects.

1. Introduction

In this paper we address the question of the syntactic position in which information focus is realized in Catalan. In this language, like in other null-subject Romance varieties, there are two possibilities for the syntactic realization of focus: the focal phrase can appear either in a postverbal or in a preverbal position, as shown in (1) and (2), where the focus is marked in bold:1

(1)	a.	Ha	pagat	el	Joan.
		has	paid	the	Joan
	b.	El	Joan	ha	pagat.
		the	Joan	has	paid
		‘Joan paid.’

 

(2)	a.	La	Maria	s’ha	comprat	un		Mercedes.
		the	Maria	refl=has	bought	a		Mercedes
	b.	Un	Mercedes	s’ha	comprat	la	Maria.
		a	Mercedes	refl=has	bought	the	Maria
		‘Maria bought a Mercedes (for herself).’

These two options, however, are not completely free and interchangeable. There is a general consensus that the unmarked position for narrow focus is the postverbal one, independently of its syntactic category and grammatical function (see Vallduví 2002; see also López 2009, sct. 2.4.4). The preverbal position, by contrast, is restricted to specific interpretations and pragmatic functions associated with focus. In particular, it is only available to foci that imply a correction of a given alternative (corrective or contrastive focus) or unexpectedness with respect to more likely alternatives (mirative focus) (see Cruschina 2012, 2019, 2021 and references therein on other Romance languages), as illustrated in (3) and (4):

(3)	a.	El	Joan	ha	pagat,	no	la	Maria. (corrective focus)
		the	Joan	has	paid	not	the	Maria
		‘Joan paid, not Maria.’
	b.	No	m’ho puc	creure!	El	Joan	ha	pagat! (mirative focus)
		not	me=it=can.1sg	believe.inf	the	Joan	has	paid
		‘I can’t believe it! Joan paid!’

 

(corrective focus)
(4)	a.	Un		Mercedes	s’ha	comprat	l’Eva,	no	un	Seat.
		a		Mercedes	refl=has	bought	the=Eva	not	a	Seat
		‘Eva bought a Mercedes (for herself), not a Seat.’
(mirative focus)
	b.	Increïble,			no	m’ho esperava!		Un	Mercedes	s’ha
		Unbelievable			not	me=it=expected.1sg		a	Mercedes	refl=has
		comprat	l’Eva!
		bought	the Eva
		‘Unbelievable! I was not expecting this! Eva bought a Mercedes!’

In the absence of these additional meanings, however, the data are more subtle and somehow controversial. This is the case of information focus, which we here define as the type of focus that introduces new information and gives rise to a contextually open set of alternatives, with no direct contrast with given alternatives, as in corrections, nor with (contextually) more likely alternatives, as in counter-expectational (mirative) readings.

A classical method to elicit information focus is by means of a wh-question, where the wh-phrase in the question corresponds to the focus constituent in the answer (cf. 5–8). When studying the syntax of focus, however, this method is problematic because the most natural answer to a wh-question is very often a fragment that only includes the focus, while the given background material is elided (cf. (6a) and (8a); see Merchant 2004). This problem makes it rather difficult to investigate the position that information focus occupies within the sentence, leading to considerable controversy and divergent views in the literature, particularly for Spanish, but also for Catalan. Indeed, when the answer is not a fragment, it is still a matter of dispute whether information focus should occur in the default postverbal position (cf. 6b, 8b), or can be fronted to a preverbal position (cf. 6c, 8c).2

(5)	Qui ha pagat?
	‘Who paid?’
(6)	a.	El Joan.
	b.	Ha pagat el Joan.
	c.	El Joan ha pagat.
		‘Joan (paid).’
(7)	Què s’ha comprat l’Eva?
	‘What did Eva buy for herself?’
(8)	a.	Un Mercedes.
	b.	(L’Eva) s’ha comprat un Mercedes.
	c.	Un Mercedes s’ha comprat (l’Eva).
		‘(Eva bought) a Mercedes (for herself).’

The goal of this study is to investigate the distribution of information focus in Catalan on the evidence of a new elicitation technique that overcomes the above-mentioned problem with fragment answers and that improves the ‘naturalness’ of the relevant methodology. After a thorough review of the question-answer test (cf. Section 2), the different views reported in the previous studies are discussed in Section 3, with a special attention to the methods adopted to elicit information focus. To overcome the previous potential methodological problems and weaknesses, a new elicitation technique is presented in Section 4. We implemented this technique both in a production and in an acceptability-judgement experiment, which will be described in Section 5 and Section 6, respectively. The experimental results are further discussed in Section 7, together with some final remarks and suggestions for future research.

2. The Question-Answer Test

In crosslinguistic research, answers to wh-questions are still “the most widespread and most widely used test for focus” (Van der Wal 2016, p. 265) and, in particular, for information focus.3 Despite its simplicity, the question-answer test can be easily misunderstood, and in fact misused, when the method is confused with the definition, that is, when any constituent addressing the wh-question is taken to be the focus of the answer or, if different pragmatic types of focus are distinguished, when the answering constituent is automatically analysed as an instance of information focus. In our view, it is fundamental to pay attention to the following aspects related to this test:

(9)	a.	Only congruent answers contain a narrow focus.
	b.	The most natural answers are fragments.
	c.	Not all foci in answers to wh-questions are information foci.

Let us consider these three aspects individually.

2.1. The Question-Answer Congruence

It is important to restrict the question-answer test to congruent answers. According to the question-answer congruence, the focus of an answer to a wh-question is the part of the answer that directly corresponds to the wh-phrase in the question (cf. (5)–(8) above). This congruence is generally described with reference to main prominence: the constituent that bears the main prominence of the answer (marked in bold in our examples) must correspond to the wh-question in the question (see Paul [1880] 1995; Halliday 1967; von Stechow 1990; Roberts [1996] 2012; Schwarzschild 1999; Krifka 2001, 2011; Reich 2002; and Zimmermann and Onea 2011, among others):

(10)	a.	#L’Eva s’ha comprat un Mercedes.
	b.	L’Eva s’ha comprat un Mercedes.

Sentence (10a) is not a congruent answer to the question in (7) because the focus constituent of the answer does not correspond to the object wh-phrase què ‘what’ in the question. In (10b), by contrast, this correspondence holds, so we have a congruent answer. Pretheoretically, congruent answers can be considered the most simple and straightforward answers to wh-questions. However, it is possible to respond to a wh-question with a variety of non-congruent, yet felicitous answers:

(11)	A:	Where is Diesel from?
	B:	Diesel is an Italian clothing company.
(12)	Who taught you to paint?
(13)	a.	I learned it by myself.
	b.	I had a private teacher.
(14)	A:	Who did Mary see yesterday on the street?
	B:	Mary saw that John went to the barber shop.
(15)	A:	Who went to Italy on a holiday?
	B:	John went to Siena.

(11B) clearly resolves the issue opened in question (11A), so it surely is a pragmatically felicitous answer to that question. It is nevertheless not a congruent answer, in that, the DP an Italian clothing company does not directly correspond to the wh-constituent of the question and thus, it does not count as a narrow focus. The same holds true of (13a) and (13b), which are both possible felicitous answers to the question (12). In these sentences, which lack a narrow focus, main prominence is by default assigned to the rightmost constituent. If the subject corresponding to the wh-constituent in the question were prosodically marked as focus (i.e., I learned it by myself, I had a private teacher), the pragmatic felicity would be overturned. The reply (14B) does contain a constituent (i.e., John) that roughly corresponds to the wh-constituent of the question (14A), providing the missing information. This is however not an exact correspondence, since the wh-phrase of the question is the direct object of the verb see, while in the answer, John is the subject of the complement clause. In this case too, the focal prosodic marking of this constituent (i.e., Mary saw that John went to the barber shop) would undermine the felicity of the response.

Let us finally consider the more subtle case in (15). Sentence (15B) is an appropriate and felicitous reply to the question in (15A); we also have a direct correspondence between the constituent of the question that provides the sought information and the wh-phrase in the question, which are both grammatical subjects. In fact, John in (15B) is a contrastive topic in the sense of Büring (1997, 2003, 2016) (see also Krifka 2007; Frascarelli 2017; Kamali and Krifka 2020). With contrastive topics, the question under discussion is divided into sub-questions, and each sentence that contains a contrastive topic answers one sub-question involving a predicate that applies separately to each member of a salient set. Like focus, a contrastive topic involves a set of alternatives, but unlike focus, it does not imply the exclusion of the other members of the set. Indeed, (15B) counts as a partial answer (Krifka 1999) and suggests a continuation with reference to the other members of the set (John went to Siena, Mary went to Padua, Paul went to Venice, etc.). Prosodically, a contrastive topic is distinctly marked by a specific pitch accent, which is different from the pitch accent associated with narrow focus (Büring 1997, 2003, 2016).

Therefore, when applying the question-answer test, it is essential to distinguish between pragmatically appropriate or felicitous answers that address the open issue introduced in the wh-question (i.e., the question under discussion) and congruent answers.4 Only the latter guarantee the presence of narrow focus in the answer. Following von Stechow (1990), congruence can be defined as a particular semantic relation between questions and answers, whereby the focus in the answer shares the same set of alternatives introduced in the question:

(16)	Who saw Ede yesterday?
(17)	a.	#Wolfgang saw Ede yesterday.
	b.	Wolfgang saw Ede yesterday.
		(adapted from von Stechow 1990, p. 68)

The alternatives determined by question (16) are the propositions of the form ‘x saw Ede yesterday’. Among the possible answers in (17), only (17b) shares the same set of alternatives (i.e., the same background, which stays constant in the set of alternatives) and is therefore a congruent answer. (17a) determines different alternatives of the form ‘Wolfgang saw x yesterday’. This semantic definition of the question-answer congruence allows us to characterize as non-congruent not only those answers in which the focus constituent does not directly correspond to the wh-phrase in the question (cf. 11B, 13b, 14B, 17a), but also those answers in which it is possible to identify a corresponding constituent that provides the missing information, yet this cannot be viewed as a narrow focus because it does not share the same alternatives as the question (cf. 13a, 15B).

2.2. Fragment Answers

Congruent answers allow for so-called short or fragment answers (see Merchant 2004; Krifka 2006). Since a congruent answer shares the same background as the question, it is not necessary to repeat these given parts: ellipsis of the given material would most naturally apply, yielding a fragment answer. Merchant (2004) argues that fragment answers are derived from full sentences by the fronting of the focal fragment constituent to a dedicated focus projection in the left periphery of the sentence and the ellipsis (i.e., PF-deletion) of the remnant sentence constituent in the response, corresponding to TP (see also Brunetti 2004):

(18)	A:	Who did she see?
	B:	[John] she saw.
	(Merchant 2004, pp. 673–75)

For ellipsis to apply, the deleted material must be given. More specifically, it must satisfy an identity condition by paralleling the propositional content of an antecedent in the immediate previous discourse. The elided material has to be truth-conditionally identical to the antecedent, and not necessarily morphosyntactically equivalent (Merchant 2001, 2004, 2018).

Fragment answers consist minimally of the focus phrase that satisfies the question-answer congruence. More importantly, in natural and spontaneous speech, fragments are more common and natural than full sentences. Vallduví (1992), among others, argues that non-fragment answers (what he calls ‘theme-containing utterances’) are most natural in contexts in which the question they address is not maximally salient at the time of utterance. The most natural answer to (19A) is the fragment fish, while the non-fragment answer We are having fish for dinner in (19B) would be rather marked. The same non-fragment answer, however, becomes necessary in (19B′) to indicate that this utterance is not addressing the immediately preceding question in (19A′) (i.e., Can’t we order pizza?), but rather the previous question in (19A), which was no longer maximally salient at the time of utterance.

(19)	A: What are we having for dinner?
	B: Fish. / We are having fish for dinner.
	A′: Oh, not again! We are having fish several days a week! Can’t we order pizza?
	B′: We are having fish for dinner.
	(Mayol and Vallduví 2020, p. 534)

Any linguistic investigation that makes use of the question-answer test must take into account the congruence requirement, the possibility of replying with fragments and the fact that non-fragment answers are most natural when they address a question that is not maximally salient.

2.3. The Type(s) of Focus in Answers to wh-Questions

Let us now consider the pragmatic types of focus that are found in answers. As already mentioned, information focus introduces new information, and is typical of answers to wh-questions. If the different pragmatic types of focus are defined according to the degree of contrast that they imply (Cruschina 2021), or that they establish in a specific discourse relation (Repp 2016), information focus is the least contrastive, since it only—and weakly—contrasts with the other possible alternatives in the (open) focal set; no contrast holds with alternatives that are given or salient in the context. In this sense, information focus can be defined as the neutral type of focus that does not involve any special implicature or pragmatic exploitation of the alternatives.

The idea that a wh-question triggers information focus is so widespread that the question-answer test is often assimilated to the definition: information focus is the type of focus found in answers to wh-questions. It is important to emphasize, however, that the use of a focus to answer a wh-question is not per se a sufficient condition for the definition of information focus, in that not all foci in answers to questions are necessarily information foci; other types can also occur in this context (Escandell Vidal and Leonetti 2019, p. 208). In particular, mirative focus can also appear in an answer to a question (Cruschina 2021, pp. 21–22):

(20)	A:	Quin nom li han posat al gos? (Catalan)
		‘How did they call the dog?’
	B:	Pepet li han posat! Quin nom més absurd per a un gos!
		‘They called it Pepet! What an absurd name for a dog!’
(21)	A:	¿Quién está tocando el piano? (Spanish)
		‘Who’s playing the piano?’
	B:	(¡No me lo puedo creer!) ¡Juan está tocando! (No tenía idea de que sabía tocar…)
		‘(I can’t believe it!) Juan is playing! (I had no idea he could play…)’
(22)	A:	Chi è entrato? (Italian)
		‘Who came in?’
	B:	Ma pensa un po’…. George Clooney è entrato! Chi l’avrebbe mai detto!?
		‘Guess what... George Clooney came in! Who would have thought so!?’
		(Belletti and Rizzi 2017, p. 45)

The focus in the replies in (20–22) provides the missing information requested in the corresponding question. The background in the answer is the same as in the question (at least truth-conditionally), thus meeting the question-answer congruence requirement. The focal alternative denoted in these replies, however, implies a contrast with other alternatives that in the speakers’ common ground, either as previous knowledge or as expectations, are considered more likely. This is the typical implicature associated with mirative focus (Bianchi et al. 2015, 2016; Cruschina 2021).5

To sum up this section, we have seen that the question-answer test provides a good tool for identifying focus. In order to make an appropriate use of this test, however, it is necessary that it complies with the question-answer congruence requirement, which very often yields fragment answers. In addition, the foci elicited by means of this test are not necessarily and automatically information foci. The role of the focal set of alternatives in the interpretation of the focus is crucially to identify the exact focus type. An inaccurate use of the question-answer test, or a misconception of information focus, may give rise to empirical challenges and controversial data, misleading scholars both on conceptual and on methodological grounds. This problem will be discussed in more detail in the following section.

3. Eliciting Information Focus: Methodological Questions

With respect to the question of the syntactic position in which information focus is realized, a certain tension exists in the literature. This tension seems to depend on the methodology adopted to elicit information focus (see Uth and García García 2018; Cruschina 2021). All studies make use of the question-answer test, but different data are reported according to whether introspection or a quantitative analysis is used. On the one hand, those scholars who based their data on their own intuitions and introspection claim that information focus, including subjects, must appear in a postverbal position, at the end of the sentence, as shown in (23) and (24) (see e.g., Zubizarreta 1998; Ordóñez 2000; Büring and Gutiérrez-Bravo 2001; Gutiérrez-Bravo 2006, 2008 on Spanish; López 2009; Cruschina 2022 on Romance, more generally). For example, Vallduví (2001) claims that (25A1) is much more neutral than (25A2), which conveys some added value. Therefore, narrow focus on the subject yields subject inversion (cf. also Belletti 2004).

(23)	Q:		¿Quién	te	regaló	la	botella	de	vino? (Spanish)
			who	to-you	gave	the	bottle	of	wine
			‘Who gave you the bottle of wine (as a present)?
	A1:	#	María	me	regaló	la	botella	de	vino.
			Maria	to-me	gave	the	bottle	of	wine
	A2:		Me	regaló	la	botella	de	vino	María.
			to-me	gave	the	bottle	of	wine	Maria
			‘Maria gave me the bottle of wine (as a present).’
			(Zubizarreta 1998, pp. 125–26)

 

(24)	Q:	¿Quién	compró	los	discos? (Spanish)
		who	bought	the	records
		‘Who bought the records?’
	A1:	Los	compró	una	muchacha.
		them.acc	bought	a	girl
	A2:	#Una muchacha		los	compró.
		a	girl	them.acc	bought
		‘A girl bought them.’
		(Gutiérrez-Bravo 2006, pp. 110–11)

 

(25)	Q:	Qui	va		venir? (Catalan)
		who	pst		come?
		‘Who came?’
	A1:	Va	venir	el	Pere.
		pst	come	the	Pere
	A2:	#El	Pere	va	venir.
		the	Pere	pst	come
		‘Pere came.’
		(Vallduví 2001)

On the other hand, some researches have more recently questioned this claim on the basis of experimental and quantitative data. In particular, both for Spanish and for Catalan, it has been argued that information focus can occur sentence-initially, especially when it corresponds to the subject (see Gabriel 2010; Hoot 2012, 2016; Vanrell and Fernández-Soriano 2013, 2018; Jiménez-Fernández 2015a, 2015b; Leal et al. 2018; see Heidinger 2021 for a detailed overview on Spanish; and Frascarelli and Stortini 2019 on Italian).6 Gabriel (2007, 2010), for example, reports data on Argentinian dialects of Spanish based on semi-spontaneous elicitation, which show that sentence-initial narrow focus (SVO) is rather frequent in Spanish. In particular, he shows that the availability of this order depends on the sentence structure: final focal subjects are preferred in intransitive structures or in transitive sentences with a cliticized object, while a preverbal information-focus subject is preferred in the case of transitive structures with a full object, as shown in (26) and in (27). Based on this evidence, Gabriel (2010, p. 189) concludes that “it can be stated that the mechanisms of syntactic focus marking, i.e., the use of a particular syntactic construction in a given pragmatic context, is governed by strict rules to a lesser degree than suggested in much of the literature”.

(26)	A:	¿Quién	secuestra	a	Tarzán? (Argentinian Spanish)
		who	kidnaps	acc	Tarzan
		‘Who kidnaps Tarzan?’
	B:	Blancanieves	secuestra	a	Tarzán.
		Snow White	kidnaps	acc	Tarzan
		‘Snow White kidnaps Tarzan.’
		(Gabriel 2007, p. 67)

 

(Argentinian Spanish)
(27)	A:	¿Quién	le	da	el	diario	a	su	hermano?
		who	to-her	gives	the	newspaper	to	her/his	brother
		‘Who gives the newspaper to her/his brother?’
	B:	María le	está	dando	el	diario	a	su	hermano.
		María to-him	is	giving	the	newspaper	to	her	brother
		‘María is giving the newspaper to her brother.’
		(Gabriel 2010, p. 202)

Similar data are reported for European Spanish and Catalan dialects in Vanrell and Fernández-Soriano (2013, 2018), for constituents other than subjects (cf. (28) and (29)):

(Basque-Country Spanish)
(28)	a.	A:	¿Qué le		dio		el	marinero	al	viejo?
			what		dat=give.pst.3sg		the	sailor	to-the	old-man
			‘What did the sailor give to the old man?’
		B:	La	carta	le	dio	el	marinero	al	viejo.
			the	letter	dat=give.pst.3sg		the	sailor	to-the	old-man
			‘The sailor gave the letter to the old man.’
	b.	A:	¿Qué		sacó		María	sin	problemas?
			what		take-out.pst.3sg		María	without	problems
			‘What did Maria take out without any difficulties?’
		B:	El		coche		sacó	María	sin	problemas.
			the		car		take-out.pst.3sg	María	without	problems
			‘Maria took out the car without any difficulties?’
			(Vanrell and Fernández-Soriano 2013, p. 61)

 

(Balearic Catalan)
(29)	a.	A:	Què	va entrar,		na	Maria,	amb		dificultat?
			what	put-in.pst.3sg		the	Maria	with		difficulty
			‘What did Maria take in with some difficulty?’
		B:	Es	cotxe	va entrar	na	Maria	amb		dificultat.
			the	car	put-in.pst.3sg	the	Maria	with		difficulty.
			‘Maria took in the car with some difficulty.’
(Central Catalan)
	b.	A:	A	qui	va enviar	la	carta,	el		mariner?
			to	whom	send.pst.3sg	the	letter	the		sailor
			‘Who did the sailor send the letter to?
		B:	A	la	dama	va enviar	la	carta	el	mariner.
			to	the	lady	send.pst.3sg	the	letter	the	sailor
			‘The sailor sent the letter to the lady.’
			(Vanrell and Fernández-Soriano 2013, p. 61)

It could well be that dialectal variation or language contact are responsible for these contrasting views and data; however, similar findings are reported for other varieties of Spanish, especially with respect to preverbal focal subjects (see, e.g., Muntendam 2009, 2013 for Andean Spanish; Hoot 2012, 2016; and Uth 2018 for Mexican Spanish; Leal et al. 2018 for Mexican and Chilean Spanish; Jiménez-Fernández 2015a, 2015b for Southern Peninsular Spanish; and Sánchez-Alvarado 2018 for Asturian Spanish). Thus, these factors cannot alone be the explanation for the contrasting views and the growing challenge to the consensus view on the distribution of focus. Here we discuss some methodological issues.

Most of the quantitative studies were interested in the prosody of focus constructions and in syntactic variation, so they included production tasks, as well as acceptability judgments. The different focus configurations were elicited by means of picture-based tasks (see Gabriel 2007, 2010; Muntendam 2009; Vanrell and Fernández-Soriano 2013; Feldhausen and Vanrell 2014, 2015, among others; see also Leal et al. 2018 for the use of video clips). In particular, in the prosodic studies on Romance languages, a Discourse Completion Task (DCT) was used to elicit narrow focus. DCT consists of a questionnaire containing different scenarios designed to elicit a specific speech act or structure; the participants complete the task by each taking a turn of dialogue (Kasper and Dahl 1991; Brown 2001). In the studies on information focus, the DCT involved the question-answer test: a question was inserted at the end of a contextualized scenario and the participants had to answer that question (see Vanrell et al. 2018 for a detailed overview of the DCT method in this area of research).

The DCT method has the advantage of eliciting (semi-)spontaneous speech, allowing the informants to freely utter the answer that they consider most appropriate in the corresponding scenario. Moreover, it is possible to control both the context (pragmatic factors) and desired aspects of the target sentences (e.g., syntactic properties). The use of pictures in the above-mentioned studies, however, might have affected—or in fact compromised—the results. On the one hand, images may help the participants to better understand the scenario or the story, and may also guide them to produce the desired utterances. On the other hand, the use of pictures to answer a question can give rise to a sense of obviousness: if the participants are asked to reply to questions about what is evident in the images, they might express that the answers are obvious. This problem was already pointed out by Uth (2014, p. 95) with reference to the task used in Gabriel (2007, 2010): “the informants very probably end up with the most likely pragmatic enrichment, i.e., the encoding of the obviousness and/or the certainty of the assertions they make when answering the corresponding Wh-questions. […] The participants are questioned about what they see on the pictures, and even if they manage to properly reenact the ‘game’, the answers are still more than obvious” (see also Vanrell et al. 2018). What happens in these situations is that an important default assumption is missing in this task: speaker ignorance, that is, the assumption that a speaker asks a question because they do not know the answer (see Farkas 2020). The fact that the pictures were already illustrating the answer to the question presumably favoured a non-canonical interpretation of the question. As we know (cf. Section 2.3), answers to wh-questions do not necessarily and exclusively involve information focus, but may involve other focus types. The motivation for a preverbal occurrence of the focus (presumably, fronted) may then well lie on the special interpretation associated with the focus constituent.

A further potential weakness of the DCT methods used in these studies has to do with the way the problem of fragment answers was tried to be avoided (cf. Section 2.2). The elicitation task often included precise instructions given to the participants, who were explicitly asked to repeat in their answer the predicate or all the constituents appearing in the question (cf., e.g., Gabriel 2007, 2010; Vanrell and Fernández-Soriano 2013, 2018) or, at least, to use complete sentences, “including as many details as possible” (cf. Leal et al. 2018, pp. 230–231; see Sánchez-Alvarado 2018 for a different solution to this problem). This instruction not only limits the freedom of the participants in their replies, but forces them to avoid the most natural and spontaneous tendency to reply with a fragment answer—or even to refrain from the cliticization of given objects.

In sum, most of the experiment and quantitative studies on narrow focus in Romance neglected important characteristics of the question-answer test, which are crucial for a full and accurate functionality of this test (cf. Section 2). We assume that only congruent answers were analysed in these studies; however, no specific measures had been taken to control for two important aspects: the naturalness of fragment answers, which goes hand in hand with the fact that full sentences are less natural answers, and the specific interpretation of the focus. In the next section, we propose some changes to the DCT method that can be seen as improvements to remove—or at least ameliorate—the weaknesses that we have just discussed.

4. A New Elicitation Technique: Questions with a Delayed Answer (QDA)

In order to enhance the naturalness and reliability of the question-answer test, we designed an improved version of the DCT, giving careful consideration to the characteristics discussed in (9) (cf. Section 2) and repeated here below:

(9′)	a.	Only congruent answers contain a narrow focus.
	b.	The most natural answers are fragments.
	c.	Not all foci in answers to wh-questions are information foci.

Let us start with the problem of fragment answers (9b). Any syntactic study of focus needs full sentences in order to determine its syntactic distribution within the sentence. However, instead of explicitly instructing the informants to produce full sentences in contexts in which fragments would be more natural answers, we introduced a special elicitation technique that overturns the preference for fragments over full sentences. This technique consists in the insertion of some lexical material between the question asked in the first turn of dialogue, within the imaginary scenario and the point in which the participant is asked to answer the question, as shown in (30) and (31), where the question is marked in bold. We label this technique Questions with a Delayed Answer (QDA). As typical of DCT, the participants were placed in a role-playing situation with the experimenter in an online session: the experimenter presented a scenario in which one of the characters asked a question, and the participant was asked to provide a natural answer to such a question:

(30)	You go to your parents’ place. You show your mum a watercolour portrait of yourself. She asks “Who drew it?”. At that point you get a phone call. Somebody got the wrong number. You hang up and, to answer your mum, you say:
(31)	You are watching a film with your roommate. Since she wakes up really early every day, she falls asleep and misses the ending. When you switch off the TV, she wakes up and asks you: “What did they find? I don’t think I’ll watch this movie again. I’m sure I would fall asleep again.” To reply you say:

This simple strategy favours the spontaneous production of full sentences. Given the lack of proximity between the answer and the question, the ellipsis of the background material (cf. Section 2.2) does not apply or proves less natural, while the use of a non-fragment answer becomes preferred to address the non-maximally salient explicit question. As a result, the participants spontaneously uttered full sentences instead of fragments, without being explicitly instructed to do so. In addition, they freely made use of the cliticization of given objects, as well as of dislocation of given arguments, without having to repeat all constituents appearing in the question.

Once we were able to ensure that the problem in (9b) did not hold, it was easier to control the other two characteristics of the question-answer test. Firstly, only congruent answers were considered in the analysis of the results, in order to make sure that we were dealing with sentences featuring a narrow focus (9a). Felicitous answers that did not meet the congruence requirement (cf. Section 2.1) were excluded, even when they addressed the question under discussion or indeed resolved the issue opened in the question. Secondly, to control the interpretation of the focus (9c), no images were used: the informants were thus free to produce the answer they deemed most fitting and appropriate in the given context. Moreover, the scenarios were designed in order to guarantee a neutral interpretation of the focus (i.e., to elicit information focus), without any contextual cues that could hint at or trigger special meanings. To encourage participation, we used proper names of relatives, friends, and specific places familiar to the informants (see Billmyer and Varghese 2000; cited in Vanrell et al. 2018).

We used the QDA-technique in a production experiment, which will be described in the next section. An additional rating experiment based on acceptability judgements was later carried out to explore the degree of acceptability in the perception of patterns that were absent or very marginal in production. The latter will be described and discussed in Section 6.

5. The Production Experiment

We created 16 scenarios in Catalan featuring the QDA-technique. The grammatical category of the wh-phrase in the question was manipulated in order to have subject and object questions—and hence, have subject and object foci in the answers.7 In particular, 8 scenarios featured a question about the subject (e.g., ‘Who drew it?’ in (32)) and 8 contained a question on the object (e.g., ‘What did they find?’ in (33)).

(32)	Entres a casa dels teus pares. Li ensenyes a la teva mare un retrat que t’han fet amb aquarel·la. Ella et pregunta: “Qui ho ha dibuixat?”. En aquell moment reps una trucada. Era algú que s’equivocava. Penges i per respondre a la teva mare dius:
	‘You go to your parents’ place. You show your mum a watercolour portrait of yourself. She asks “Who drew it?”. At that point you get a phone call. Somebody got the wrong number. You hang up and, to answer your mum, you say:’
(33)	Estàs veient una peli amb la teva companya de pis. Com que cada dia es lleva molt d’hora, s’adorm i es perd el final. Quan tu apagues la tele, ella es desperta i et pregunta: “Què han trobat al final? No crec que torni a veure aquesta peli. Total, al final sempre m’adormo.” Per respondre dius:
	‘You are watching a film with your roommate. Since she wakes up really early every day, she falls asleep and misses the ending. When you switch off the TV, she wakes up and asks you: “What did they find? I don’t think I’ll watch this movie again. I’m sure I would fall asleep again.” To reply you say:’

Two types of delay were created. The lexical material between the question and the answer was either part of the narration or description of the scenario (32), or it was a continuation by the same speakers who uttered the question (33). This difference did not affect the results (cf. below). In both cases, an explicit question is posed, which is not maximally salient by the time the speaker can reply to it. The experiment also included 16 fillers, without delay and with other types of questions (polar or alternative questions).

A total of 20 participants (15 females, 5 males, mean age = 22) took part in the online elicitation sessions with the experimenter and received a compensation of 15€ for their time. They were all native speakers of Catalan. We elicited 320 target answers (16 stimuli × 20 participants) which were transcribed and annotated. Only 24 (7.5%) answers were fragments, meaning that the experimental stimuli with QDAs were largely successful in eliciting full sentences.8 We performed the analysis on a subset of 216 answers, which had been annotated as congruent: the participants directly addressed the relevant question and did so by repeating the same verb used in the scenario; in the answer there was therefore a focus constituent (prosodically marked as such) that directly corresponded to the wh-phrase in the question, and the background in the answer paralleled the background in the question, thus fully meeting the congruence requirement (cf. Section 2.1). As mentioned in the previous section, this was necessary to ensure that the information focus in the answer was an instance of narrow focus, either on the subject or the object, with a given background. Examples of non-congruent answers that were excluded from the analysis are given in (34) and (35). The sentences in (34a) and (34b) are felicitous answers to the questions in (32) and (33), respectively, which in fact address the question under discussion and resolve the issues opened in the corresponding question (i.e., they provide an informative answer to the question). However, they do not meet the congruence requirement, in that they do not feature a focus structure with a narrow focus and a given background.9 In (35), in particular, we can see examples of different ways of reacting to a question, leading to non-congruent answers to (33) in which the speaker confesses their inability (35a) or lack of willingness (35b) to resolve the issue introduced in the question:

(34)	a.	Era una persona que feia retrats amb qui ens hem trobat.
		‘It was a person who made portraits and who I have met.’
	b.	Era la típica pel·lícula que sempre acaben trobant el tresor.
		‘It was the typical film where they always end up finding a treasure.’
(35)	a.	Realment no m’hi estava fixant gaire, així que realment no ho he vist.
		‘Actually I was not paying attention at all, so I didn’t actually see it.’
	b.	Si vols saber-ho, l’hauràs de veure.
		‘If you want to know it, you have to watch it.’

Figure 1 shows the results of the analysis of the 216 congruent answers, which were annotated according to the grammatical category (subject vs. object), the position of the focus (postverbal vs. preverbal), or whether a cleft was used to mark the focus in the answer:10

Participants overwhelmingly produced postverbal foci both for subjects (90.9%) and object (94.8%), as exemplified in (36) and (37), respectively, along with a marginal number of preverbal (subject = 5.1%; object = 2.6%) and clefted foci (subject = 4.0%; object = 2.6%), as in (38).

(36)	a.	M’ho ha	dibuixat	la	Marta.		(VS)
		me=it=has	drawn	the	Marta
		‘Marta has drawn it for me.’
	b.	L’ha	dibuixat	la	meva	amiga.	(VS)
		it=has drawn		the	my	friend
		‘My friend has drawn it.’

 

(37)	a.	Han	trobat	el	tresor	perdut.		(VO)
		have.3pl	found	the	treasure	lost
		‘They have found the lost treasure.’
	b.	Han	trobat	una	caixa	amb	or.	(VO)
		have.3pl	found	a	trunk	with	gold
		‘They have found a trunk chest with gold.’

 

(38)	a.	La	Cristina	ho ha	dibuixat.			(SV)
		the	Cristina	it=has	drawn
		‘Cristina has drawn it.’
	b.	He	estat	jo	que	ho he	dibuixat.	(subject cleft)
		have.1sg	been	I	that	it=have.1sg	drawn
		‘It was me who has drawn it.’
	c.	En	Pere	és	qui	ho ha	dibuixat.	(subject cleft)
		the	Pere	is	that	it=has	drawn
		‘It’s Pere who has drawn it.’

As for the non-focal, given argument, in 85% of the instances it was realized as a null (i.e., 37a,b) or clitic (i.e., 36a,b) pronoun, as expected. That is, despite the delay, the given arguments were still highly active in the participants’ discourse model. In the other 15% of cases, the non-focal argument was realized as a strong pronoun or a DP, doubled with a clitic, in the case of objects. This type of realization was particularly frequent in those contexts that naturally lend themselves to a contrastive interpretation. For example, in (39), the context establishes a clear contrast between what the speaker wants to drink and what the addressee wants to drink, which is why a strong subject pronoun is used as a contrastive topic. Something similar happens in (40), where the speaker may want to draw a contrast between ‘these pills’ and some other pills, which is achieved by means of a left-dislocation. In this example, left-dislocation also serves the function of restating the topic of the sentence after the delay created in the description of the scenario:

(39)	a.	Vas al bar amb un amic que fa temps que no veus. Seieu i el teu amic et pregunta: “Què vols prendre?”. Abans que puguis contestar, el teu amic rep una trucada. Parla per telèfon durant un minut i després penja. Per respondre la seva pregunta dius:
		‘You to go a bar with a friend you haven’t seen in a long time. You sit and your friend asks: “What do you want to drink?”. Before you can answer, you friend gets a call. She speaks on the phone for a minute and hangs up. To answer her question, you say:’
	b.	Jo	vull	una	cervesa,	i	tu?
		I	want.1st	a	beer	and	you?
		‘I want a beer, and you?’

 

(40)	a.	En els últims anys has tingut alguns problemes de salut i t’has visitat amb diversos especialistes. Avui t’està visitant la teva metgessa de capçalera i esteu parlant de la medicació que prens. Et pregunta: “Qui et va receptar aquestes pastilles?” Quan estàs a punt de contestar, una enfermera entra a la consulta i ella i la doctora parlen durant un parell de minuts. Després l’enfermera marxa. Per respondre la pregunta de la doctora, dius:
		‘You’ve had some health issues during the last years and are seeing several specialists. Today your family doctor is seeing you and you are discussing the medication you are currently taking. She asks “who prescribed these pills?”. As you are about to answer, a nurse enters the office. The doctor and the nurse talk for a couple of minutes and then the nurse leaves. To answer the question the doctor had asked you say:’
	b.	Aquestes	pastilles	me les va receptar	el	dermatòleg.
		These	pills	me=them=prescribed	the	dermatologist
		‘The dermatologist prescribed these pills to me.’

To test for the statistical significance of the results in Figure 1, mixed-effect logistic regressions were performed, using R (R Core Team 2013) and lme4 (Bates et al. 2015). A mixed-effect logistic regression, with grammatical function of focus as a fixed effect, and items and participants as random effects, shows that the realization of preverbal subjects is not significantly higher than preverbal objects (b = 0.84, p = 0.38)11. Our production experiment therefore shows that information foci in Catalan are predominantly produced in a postverbal, sentence-final position, irrespective of the grammatical category, and in line with the previous studies based on introspection. However, in order to gather further information about native speakers’ intuitions on the syntax of focus, and to compare the results of our QDA-production experiment with a different task, we additionally conducted a rating experiment on the acceptability of preverbal and postverbal information foci in Catalan. The results of this second experiment are consistent with those of the production experiment, but are not as clear-cut.

6. The Acceptability Judgement Experiment

To test the acceptability of preverbal focus in Catalan, in the context of an answer to a wh-question (i.e., with information focus), we conducted an experiment with 390 native speakers of Catalan. The same QDA-scenarios were used as in the production experiment (cf. (30) and (31)), but an answer to the question was additionally provided. The answer was taken from the elicited data in the production task and was the target of the rating task. The experiment had a (2 × 2) factorial design that involved two independent factors with two levels each: (i) syntactic category (subject vs. object) and (ii) focus position (preverbal vs. postverbal).

We tested a total of 32 stimuli, that is, 16 stimuli (8 with subject questions, 8 with object questions) in two conditions (preverbal vs. postverbal foci). The two factors were manipulated between items and within subjects. Each item was shown in two versions, which differed from one another with respect to the position of the focus (with a preverbal and with a postverbal focus), but the context was kept the same in both versions; the stimuli were then divided into two lists, so that each participant was only presented with one stimulus per item (i.e., each item under one condition only), so that the same lexical material of the target sentence was only judged once. Each list consisted of 16 experimental stimuli plus 16 fillers, and each subject was randomly assigned a list. The order of the trials was pseudo-randomized, each trial was presented individually, and the participants were not able to modify their answers. Each session started with a brief sociolinguistic questionnaire and a short familiarization practise.

In (41) we can see an example of the answer to the subject question contained in context in (32) under the two conditions, while (42) exemplifies the answers to the object question in (33):

(41)	a.	Ho ha dibuixat una amiga.
	b.	Una amiga ho ha dibuixat.
		‘A friend drew it.’
(42)	a.	Han trobat un tresor.
	b.	Un tresor han trobat.
		‘They found a treasure.’

The experiment was presented to the Catalan native speakers through an online interface.12 The subjects were asked to rate the degree of acceptability of each sentence in the relevant context using a 7-point Likert scale. The results of this study confirm the findings of the production experiment: postverbal focus is always preferred over preverbal focus, both in the case of subjects and objects. Crucially, as we can see in Figure 2, the mean score assigned to postverbal subjects is almost the same as that of postverbal objects (postverbal subjects = 5.30, postverbal objects = 5.67).

In the preverbal position, subjects and objects received the same rating (preverbal subjects = 4.06, preverbal objects = 4.06).13 To test for the statistical significance of the results, a mixed-effect logistic regressions was performed, with grammatical function, position of the focus, and their interaction as fixed effects, as well as items and participants as random effects14. The role of the position of focus is significant (b = 3.12, p < 0.001), while grammatical function (b = −0.79, p = 0.24) or the interaction between the two fixed effects (b = 0.73, p = 0.49) is not.

Interestingly, these results show that preverbal foci are not judged as completely ungrammatical and are still (marginally) acceptable, even though to a lesser extent than postverbal foci. We will go back to this issue in the next section.

To account for individual variation and for the possible different use of the Likert-scale by the participants, the responses were converted into z-scores across focus category and position, as shown in

Table 1. Average normalized scores by category and position.

	Mean Normalized Rating
Preverbal subject	−0.36
Postverbal subject	0.27
Preverbal object	−0.35
Postverbal object	0.44

. While postverbal foci received mean positive ratings (i.e., above the mean), the mean ratings for preverbal foci are negative (i.e., below the mean). In the postverbal position, objects receive higher ratings than subjects, presumably because this coincides with their default position. In the preverbal position, there is no difference between subject and object.

We close this section with a note on the adjacency between the verb and the postverbal focus in Catalan. A further problem with the previous studies concerns the use of answers that were less natural, not only because they consisted of full sentences (as immediate answers to wh-questions), but because the postverbal focus constituent was not adjacent to the verb. Several scholars have pointed out that narrow focus needs to be sentence-final in Spanish, reporting examples with an intervening object between the verb and a focal subject (cf. (23) above from Zubizarreta 1998, pp. 125–26). Adjacency with the verb seems to be required in Catalan, which does not readily accept either VOS or VSO (see Ordóñez 2000; Vallduví 2002; Vanrell and Fernández-Soriano 2013; Leonetti 2017; Cruschina 2022 and references therein).15

In this respect, it is therefore important to point out that VOS and VSO were never realized in our Catalan production experiment, and we therefore avoided them in our rating experiment. Given arguments are always cliticized and dislocated in Catalan (Vallduví 1992, 1995; Mayol 2007; Villalba 2009, 2011; Leonetti 2017), thus guaranteeing the adjacency between the verb and the postverbal focus.

7. Discussion and Conclusions

Our findings provide experimental evidence in support of the theoretical view, mostly based on native speakers’ intuitions and introspection, that informational foci are realized in a postverbal position in Catalan. From a methodological viewpoint, we believe that our QDA-technique offers significant advantages in striving for natural and neutral answers in the elicitation of informational focus, making the most of the question-answer test. These results are neat and clear-cut in production. The acceptability experiment further confirms this view, but at the same time shows that preverbal foci are not judged as ungrammatical by native speakers.

To understand the reasons for the marginal acceptability of preverbal foci in the rating experiment is not an easy task; several possible explanations come to mind. Firstly, it could well be that the production experiment was more successful in maximizing the question-answer test because it involved a direct interaction with the experimenter, which in turn guarantees more control over possible extraneous variables that could have interfered with the experiment, which might have affected the pragmatic interpretation of either the question or the answer. In addition, in the rating task, the experimental stimuli were presented in a written format. As pointed out by Escandell Vidal and Leonetti (2019), in the absence of prosodic information it is not possible to ascertain whether speakers were interpretating an SVO order with a marked initial focus subject (SVO) or rather a sentence with an unmarked SVO order. This is indeed a shortcoming of written stimuli, but at the same time it cannot be the full explanation of our results because the OVS order was also judged as marginally acceptable.

We did consider the possibility of using audio-stimuli in the acceptability experiment; however, we decided not to do it because it would expose the study to other types of problems and complications. The experimental subjects may pay too much attention not only to the position of the main stress (i.e., to the nuclear pitch accent, NPA), but also to the type of pitch accent (PA). We know that the prosodic realization of a pitch accent depends on the type of focus: the PA associated with corrective focus is different from the PA of an information focus, and it is presumably also from that of a mirative focus.16 The specific type of NPA, moreover, may also be subject to dialectical variation. Since our main goal was the syntax of focus, we wanted judgements on the syntactic position of focus rather than on its prosodic realization. Furthermore, from the production experiment we know that information focus is barely realized in a preverbal position; thus, the prosodic pattern associated with a preverbal focus may have sounded artificial and unnatural. Indeed, we would have had to ask a native speaker to produce a sentence with a prosodic pattern that is hardly natural and use these recordings as audio-stimuli. For these reasons, we decided to go for written stimuli. We also ruled out the possibility of using a forced choice task, since we would have expected a predominant choice of postverbal foci, without any information about the other option.

Overall, despite the less straightforward results, we believe that the QDA-technique is able to elicit more reliable data in the case of a rating task, and that our results just reflect a real scenario with respect to the mapping between syntax/prosody and information structure. From an information-structure viewpoint, some syntactic configurations are more natural than others. In Catalan, postverbal foci satisfy all prosodic and syntactic requirements, which lead to the realization of a prosodically marked focus in a syntactic marked position, that is, the rightmost position within the sentence. Deviations from this order, however, do not give rise to ungrammaticality, but are perceived by native speakers as more marginal or deteriorated. This different “weight” of the divergent configurations in terms of grammaticality could also (at least partially) be at the origins of the contrasting views discussed in Section 3.

This hypothesis is confirmed by the comparison of the rating-task scores with the scores assigned to some fillers. In our experimental design, some fillers were intentionally ungrammatical (cf. 43) or pragmatically odd (cf. 44):

(43)	A:	Estàs passejant amb el teu pare pel centre de la ciutat. Passeu davant d’una llibreria i t’atures davant de l’aparador. Et quedes mirant fixament un llibre. El teu pare et pregunta “Coneixes aquest llibre?”. Per respondre dius:
		‘You are walking in the city centre with your father. You walk past a bookshop and you stop in front of the shop window. You stare at a book. Your father asks you: “Do you know this book?”. To reply you say:
	B:	A	mi	interessant	sembla.
		to	me	interesting	looks
	Grammatical version:			(A mi) em sembla interessant
				‘It looks interesting to me.’

 

(44)	A:	És dissabte al matí. Estàs al sofà, mirant el mòbil. Ve la teva mare i et diu: “Vaig a comprar una mica de roba. Vols venir?”. Per respondre dius:
		It’s Saturday morning. You are on the sofa, looking at your mobile phone. You mother comes and tells you: “I am going to buy some clothes. Do you want to come along?”. To reply you say:
	B:	No	m’agrada	el	tenis.
		not	me=please.3sg	the	tennis
		‘No, I don’t like tennis.’

Such fillers were included in order to encourage the participants to use all different values of the Likert-scale. As expected, they received lower scores: the mean score for the ungrammatical fillers was 1.10, while the mean score for the pragmatically odd fillers was 2.12. If we compare these scores with those assigned with preverbal foci, it seems clear that the participants deliberately chose middle values of the scale to signal marginality or deterioration (or else, lower preferences) rather than an ungrammaticality or pragmatic oddity.

It is already well known from the literature that Catalan specific mapping rules operate at the interface between syntax and prosody, on the one hand, and information structure, on the other. These rules are responsible for the alignment between focus and the postverbal position, where the focus constituent is assigned the NPA, as well as the dislocation of the non-focal given material that facilitates the focus–NPA alignment (see Vallduví 1992; Mayol 2007; Villalba 2009, 2011). In light of our experimental findings, we hypothesize that these rules are stricter in production, while violations of these rules in perception yield marginality and lower preferences. Focus fronting of the information focus in answers to wh-questions, therefore, is certainly dispreferred over postverbal focalization, but it does not give rise to full ungrammaticality.

Further research is certainly needed to verify this hypothesis. It would be interesting to test whether the marginality of preverbal foci depends on the type of focus, proving more acceptable with corrective and/or mirative focus. The comparison with other languages could also be revealing. Spanish, for example, makes less consistent use of information-structure operations than Catalan (see Villalba 2011; Leonetti 2017). We would thus expect violations of the mapping rules between syntax and information structure to be more tolerated in Spanish than in Catalan. We hope to be able to address and answer these open questions in future work.