Content uploaded by Mariapaola D'Imperio
All content in this area was uploaded by Mariapaola D'Imperio on Oct 12, 2015
Content may be subject to copyright.
A multi-level approach to focus, phrasing and intonation
Mariapaola D’Imperio, James German, Amandine Michelas
To cite this version:
Mariapaola D’Imperio, James German, Amandine Michelas. A multi-level approach to focus,
phrasing and intonation in French. Prosody and Meaning, De Gruyter Mouton, pp.11-34, 2012.
HAL Id: hal-00743678
Submitted on 13 Dec 2012
HAL is a multi-disciplinary open access
archive for the deposit and dissemination of sci-
entiﬁc research documents, whether they are pub-
lished or not. The documents may come from
teaching and research institutions in France or
abroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, est
destin´ee au d´epˆot et `a la diﬀusion de documents
scientiﬁques de niveau recherche, publi´es ou non,
´emanant des ´etablissements d’enseignement et de
recherche fran¸cais ou ´etrangers, des laboratoires
publics ou priv´es.
A multi-level approach to focus, phrasing and intonation in French
Mariapaola D’Imperio1, James German1,2 and Amandine Michelas1
1Laboratoire Parole et Langage, CNRS and Aix-Marseille I – Aix-en-Provence, France
2Division of Linguistics and Multilingual Studies, Nanyang Technological University
firstname.lastname@example.org, email@example.com, firstname.lastname@example.org
Much recent work on German and English intonation has addressed the impact of information
structure on prosodic patterns in terms of the focus/background partition. In contrast with
stress-accent languages such as Italian, Spanish or English, French does not appear to signal
focus through pitch accent assignment, rather it appears to mainly exploit prosodic edge
marking for the same purposes. The fact that prosodic phrasing is highly sensitive to focus
structure is not only true for French, but also for pitch accent languages such as Japanese and
Basque (see Gussenhoven 2004 for a discussion), as well as for stress-accent languages
(Beckmann & Pierrehumbert 1986). A previous analysis (Féry 2001) has proposed that
French largely exploits phrasing in order to signal focus, and that narrow and contrastive
focus “lead to an initial boundary tone, usually high”. Here we attempt to build on Féry’s
insight by showing that, while phrasing is one of the strategies that French adopts in order to
signal focus, phrasing cues are different when either the left or the right edge of the focal
domain are taken into account. Our findings show that initial LHi rises are associated with the
left edge of contrastive focus regions in French, and may therefore serve an important
marking function. Crucially, phrase length also contributed to the distribution of LHi,
suggesting a probabilistic integration of factors from different levels.
Different languages use different methods to signal phrasing, mostly tonal demarcation but
also segmental or tonal sandhi rules (cf. Shih 1990 for Chinese), and most of them support the
existence of a higher level prosodic unit defined as the Intonation Phrase. This unit is
generally understood as corresponding to a sense unit or a syntactic clause, and in traditional
phonetic studies it has been linked to the notion of “breath group”. Despite the ongoing debate
about French intonation structure, most intonation models of French do generally include a
high level of constituency, the Intonation Phrase (Post 2000; Jun and Fougeron 2000, 2002),
corresponding to the intonation unit of more traditional descriptive work (cf. Di Cristo 2000).
Another basic constituent of the French prosodic hierarchy that is generally agreed
upon is a smaller unit, the Accentual Phrase (Jun and Fougeron 1995, 2000, 2002; Welby
2002, 2006), which roughly corresponds to, but does not always overlap with, the
Phonological Phrase (see Post 2000; Delais-Roussarie 2005; Astésano et al. 2007). This unit
can also be found in non-autosegmental descriptions with different labels, such as the
intonème mineur (Delattre 1966; Rossi 1985, 1999), the prosodic word (Vaissière 1992), the
rhythmic unit (Di Cristo and Hirst 1993) and the accentual group (Mertens 2004). In most
intonation descriptions we also find evidence for the existence of at least one smaller prosodic
unit, whose domain might correspond either to a Major or a Minor Phonological Phrase
(Selkirk 2000), either in the form of an intermediate phrase (Beckman and Pierrehumbert
1986) or of a smaller unit, the Accentual Phrase.
In this paper, we set out to address the complex interaction between initial rise
distribution in French and semantic and phrasing factors. Specifically, we argue that the initial
rise distribution and its relationship to information structure and syntax is likely to be
mediated by an extra level of phrasing, i.e., the intermediate phrase. Through a review of
existing proposals and experimental evidence, we further propose that neither a
straightforward Focus-to-Accent approach (see Ladd 2008) nor an approach based on phrase
edges (e.g. Féry 2001) are able to account for the full range of variability associated with
focus marking in French. We begin, however, with a review of some basic notions of prosodic
structure in French.
1.1. French prosodic structure: Basic notions
In our approach, we employ the notion of Accentual Phrase (AP) as defined by Jun and
Fougeron (1995, 2000), whose underlying tonal structure is a LHiLH* sequence (for a
detailed description of AP realizations, see Jun and Fougeron 2002). According to this model,
the AP is characterized by two tonal events, which in other frameworks may be referred to as
primary and secondary accents (Pasdeloup 1990; Di Cristo 2000). The primary accent is a
LH* phrase final accent marked by a noticeable f0 movement as well as by preboundary
lengthening. Moreover, this tonal sequence is associated with the last full syllable of the AP.
The secondary accent is better defined as an initial rise (see AM accounts such as those of Jun
and Fougeron 2000; Welby 2002), on the basis of both phonetic and phonological differences
between the two tonal rises (see 3.1 below for more detail).
les chattons sages
les chatons sages
Figure 1: Example of two types of Accentual Phrase for the sequence les chatons sages ‘the
good kittens’ realized with either only a final rise (left) or with an additional/optional early
rise (right). (Excerpt taken from the utterance Les chatons sages buvaient leur lait ‘The good
kittens were drinking their milk’ read by a native speaker of French).
Phonologically, only the final accent corresponds to a unique stress location (Dell 1984);
phonetically, rhyme lengthening only occurs within the final accent domain, while never
within the initial rise domain (Pasdeloup 1990; Astésano, 2001).
Note that both the initial and final (accentual) rises are important markers of phrasal
edges, and the location of such edges appears to interact with focus. The fact that prosodic
phrasing is highly sensitive to focus structure is not only true for French, but also for pitch
accent languages such as Japanese and Basque (see Gussenhoven 2004 for a discussion), as
well as for stress-accent languages (Beckmann and Pierrehumbert 1986). However, in contrast
with stress-accent languages such as Italian, Spanish or English, French does not appear to
signal focus through pitch accent assignment, rather it appears to generally exploit prosodic
edge marking for the same purposes. For instance, Féry (2001) has proposed that French
largely exploits phrasing in order to signal focus, and that narrow and contrastive focus “lead
to an initial boundary tone, usually high”. Here we shall attempt to build upon Féry’s insight
by showing that, while phrasing is one of the strategies that French adopts in order to signal
focus, phrasing cues are different when either the left or the right edge of the focal domain are
taken into account. Specifically, we propose that in French right edge marking can either
reflect the presence of an Accentual Phrase (AP) boundary (whenever a LH* is realized, cf.
Jun and Fougeron 2000), an Intonation Phrase (IP, signalled through a H%) or an intermediate
phrase (ip) break (Jun and Fougeron 2000; D’Imperio and Michelas 2010), signalled through
a H- tone (cf. Section 2 for a detailed definition).
Very recent work has in fact suggested that the emergence of an ip in French is not
simply linked to a specific focus pattern or marked syntactic structure, since a right ip
boundary can occur within broad focus utterances whenever the syntactic and prosodic
structure allow it (Michelas and D’Imperio 2010a, 2010b, 2010c; D’Imperio and Michelas
2010). While the intermediate phrase break is generally syntax-driven (in that it is found at
the NP/VP boundary when the NP is composed of at least 3 APs (cf. D’Imperio and Michelas
2010)), a right-edge intonation phrase break seems to be preferred when the focus domain is
very narrow (i.e. restricted to part of a DP, such as an adjective or a demonstrative).
Moreover, the placement of an initial boundary (the initial rise or LHi, cf. Jun and
Fougeron 2000; Welby 2002) does not appear to be restricted to the left edge of a Maximal
Projection, but can occur towards the left edge of an argument that is part of a complex
syntactic constituent, when focus is restricted to a single lexical item. Specifically, recent
evidence (German and D’Imperio 2010) suggests that initial LHi rises mark the left edge of
contrastive focus regions in French (see LHi on marron in upper panel of Figure 2), but that
the probability of LHi also increases with phrase length. In other words, both phrase length
and focus scope appear to be the relevant, additive factors for the appearance of an initial rise,
and thus it is unlikely that LHi is a focus marker in the traditional sense. The present paper is
intended to first touch upon the complexities of the distribution of phonetic and phonological
evidence in French that might account for all the variability found in previous and current
studies on focus and phrasing.
vendu la valise [marron]F
Figure 2 : F0 curve and spectrogram for a section of the sentence Mais à qui est-ce qu’Amélie
a vendu la valise marron dans la rue Mignet? ‘But to whom has Amelie sold the brown
suitcase on Mignet Street?’ uttered with narrow focus on marron ‘brown’.
2. The place of the intermediate phrase in the prosodic structure of French
2.1. The intermediate phrase in Jun and Fougeron’s model
Two of the most recent autosegmental models of intonation in French, Jun and Fougeron’s
and Post’s models, both agree on the existence of two levels of phrasing in French above the
prosodic word: the Intonation phrase (IP) and a lower ranked constituent, which is either
tonally defined (the Accentual Phrase or AP in Jun and Fougeron’s model) or rhythmically
defined (the Phonological Phrase in Post’s model). In contrast to Post’s model, Jun and
Fougeron (2000) postulate an additional third level of phrasing ranked between the Intonation
Phrase and the Accentual Phrase, i.e. the intermediate phrase (ip), for which we find evidence
in a number of stress-accent languages such as English and Italian. The intermediate phrase
(ip) is postulated in order to account for specific intonation structures observed in marked
syntactic constructions, such as tag-questions, in which a low pitch plateau is usually found
after an accentual H* rise (see Figure 3), as well as for high plateau found after focus in
yes/no questions (Jun and Fougeron 2000: 224). According to Jun and Fougeron’s first
proposal, these low and high plateaus are the result of the presence of either a L- or a H-
phrase accent controlling the pitch range from the last pitch accented word up to the final
syllable of the phrase (through spreading). In the example in Figure 3, the Intonation Phrase is
parsed into two ips. The first ip is right-demarcated by a L- phrase accent controlling the pitch
range from the last pitch accent to the end of the intermediate phrase.
Une bonne bouteille de champagne ça lui plairait?
Figure 3: F0 track of the utterance ([Une bonne bouteille de champagne]ip [ça lui
plairaît)ip)IP ‘A good bottle of Champagne, would he like it?’ read by native speaker of
Hence, Jun and Fougeron introduce the idea of an intermediate level of phrasing in French
primarily to account for similar tonal spreading phenomena. In other words, the authors link
the emergence of the ip to the presence of specific intonation contours and syntactic
structures. However, the authors’ later comment suggests that the phonetic and phonological
properties of the ip in French are mostly unknown: “It is possible that there is a phonetic
difference in the height or shape of f0 rise between H* and H*H-, and/or durational cues for
this intermediate phrase level” (Jun and Fougeron 2000: 237).
Recent studies (D’Imperio and Michelas 2010; Michelas and D’Imperio 2010a, 2010b,
2010c) offer additional evidence regarding (i) the factors affecting ip distribution and (ii) the
phonetic and phonological properties of ip. According to those studies, the intermediate
phrase is not restricted to specific intonation contours or syntactic structure but can appear
within broad focus (all-focus) utterances if both the syntactic and prosodic structures allow it.
We know that prosodic structure is independent of, but related to, both syntactic and
information structure. It is also generally assumed that phonological factors such as prosodic
weight or speech rate play an important role in prosodic phrasing. Taken together, this
evidence suggests that the emergence of an intermediate phrase level in French may be due to
two different constraints. First, a syntactic constraint enforces the alignment of the right edge
of a major syntactic break with the right edge of an ip. Secondly, a size constraint competes
with the syntactic constraint by requiring that an ip is made of minimally two APs. In the
example sentences in Figure 4, the break between the subject noun phrase (NP) and the verb
phrase (VP) is aligned with the ip right boundary, which is itself demarcated by a H- phrase
La mamie des amis de Rémy demandait l’institutrice
45( 4546( 45(
7( 7( 7(
La mamie de Rémy demandait l’institutrice
Figure 4: F0 track of the utterance La mamie de Rémy demandait l’institutrice ‘Remy’s
grandmother asked for the teacher’ where the subject noun phrase is made of 2 APs (4a) and
of the utterance La mamie des amis de Rémy demandait l’institutrice ‘The grandmother of
Remy’s friends asked for the teacher’ where the subject noun phrase is made of 3 APs (4b).
Note that the last syllable of the intermediate phrase is produced with final lengthening that is
significantly greater than that associated with an AP boundary (Michelas and D’Imperio
2010a). In stress-accent languages such as English or Italian, the ip is also the domain of
downstep, and it is delimited at its right edge by a phrase accent. For instance, Beckman and
Pierrehumbert (1986) showed that pitch reset is reinitialized after an ip boundary in American
English. D’Imperio and Michelas (2010) have recently shown that, as in English, the
intermediate phrase of French is the domain of downstep, but what is crucially different is that
the pitch is reset right at the intermediate phrase boundary and not after it. Specifically, a H-
phrase accent appears to also be responsible for blocking downstep of subsequent LH* within
an initial ip (see Figure 4b).
Figure 5: Scheme of ip-internal downstep and both total (at ip edge) and partial (after ip edge)
reset for sentences in which the target ip is made of 3 APs.
3. Information structure and the role of the initial rise
3.1. Phonetic and phonological properties of the initial rise
The initial rise LH rise (LHi) is an optional rise occurring on one of the first syllables of the
accentual phrase (Jun and Fougeron 2000). Specifically, Welby (2006) has proposed that the
initial rise tends to occur on one of the first syllables of the first content word of the AP, on
the basis of evidence of strong alignment between the L target of the rise and the segmental
region corresponding to the syllable onset of the content word. It has also been suggested that
LHi can be marked by strengthening and lengthening of the syllable onset (Mertens 1992;
Astésano 2001; Astésano et al. 2007), though strong empirical evidence is lacking.
Both the final LH* and LHi appear to be correlated with hyper-articulation (greater
tongue displacement, wider lip aperture, increase in duration and peak velocity) of the
associated syllable, though a stronger effect is found for LH* (Loevenbruck 1999; Dohen and
Loevenbruck 2009). Yet in spite of such accounts, the status of the initial rise is more
controversial than that of the final accent, primarily due to the much higher degree of
variability associated with its occurrence and realization. Neglected by most studies on French
prosody during the first part of the 20th century, more recent models treat the two rises either
as the same tonal event or as different events. For instance, in the first autosegmental metrical
model of French proposed by Hirst and Di Cristo (1984), the initial and final rise are not
distinguished since both correspond to the metrical head of their tonal unit (TU), defined as
the minimal unit of synchronisation of tones and segments. Several TUs combine to form a
larger unit called intonation unit (IU).
In a more recent and quite different autosegmental-metrical model proposed by Post
(2000), both rises are also described as similar, though in terms of two monotonal high pitch
accents (H*) having the same shape and the same accentual properties (see Figure 6).!!
Figure 6: Post’s model of French intonation. The intonation phrase IP is demarcated by two
boundary tones T% and includes one or more phonological phrases PP containing at least one
H*. Taken from D’Imperio et al. (2007).
In Post’s model, the final pitch accent in the intonational phrase can also take the form of a
bitonal H+H*, which is proposed to account for cases in which the f0 peak is aligned with the
penultimate syllable of the phonological phrase. Moreover, in that model, the occurrence of a
final LH* is accounted for in terms of phonetic implementation differences by means of a rule
leading to the optional insertion of a L tone between the initial and the final H* pitch accents
(“tone-linking rule”, cf. Gussenhoven 2004).
Other authors claim instead that the initial and final rises are structurally different, and
that only the final rise is an actual pitch accent. Jun and Fougeron (1995) first proposed that
the initial LHi rise is a bitonal phrase accent, associated to the left edge of the AP, while the
final accent is a bitonal LH* pitch accent whose H* tone is associated to the last full syllable
of the accentual phrase (while the L tone is unassociated). In Welby’s proposed revision of
this model (Welby 2002, 2006, “early L double association hypothesis”), the L tone of the
initial rise seeks a primary association with the first syllable of the first content word of the
AP and an optional secondary association with the edge of an earlier syllable, which is often,
but not always, the first syllable of the AP. Figure 7 below illustrates Jun and Fougeron’s
proposal including Welby’s revision.
Figure 7: Jun and Fougeron’s model with and Welby’s (2002, 2006) revision. Primary
association of the L tone is shown through a solid line, while the secondary association is
shown by the dotted line.
3.2. Factors Affecting the Distribution of LHi
The occurrence of LHi in an accentual phrase is highly variable, and the details of its
distribution and realization are poorly understood. While on the one hand LHi is generally
considered to be an “optional” feature of AP, its occurrence has been shown to be sensitive to
a wide range of factors, including phonological, syntactic, and discourse-level ones. Welby
(2006) shows, for example, that the length of an AP is a good predictor of whether LHi occurs
in it, such that longer APs are more likely to include LHi than shorter APs. This was true
whether length was measured in number of syllables or overall duration, though syllable
number was a slightly better predictor for most speakers. The same study showed that LHi
occurred more often for slower speaking rates than for faster ones.
Jun and Fougeron (2000) show that position in a sentence matters. In that study, LHi
was more likely to occur in APs occurring sentence-initially than for APs in either sentence-
medial or sentence-final positions. The authors note additional stylistic and phonological
factors that have been suggested by various authors including speaking modality (e.g.
imperative vs. exclamation), speaking style (reading vs. spontaneous), segmental composition
of the syllable, and syllable structure (open vs. closed, onset vs. no onset) (Fónagy 1980;
Vaissière 1974; Lucci 1983).
Syntax can also be relevant for the occurrence of LHi. Astésano et al. (2007) provide
the most direct evidence of this, showing that the tendency for LHi (or IA in their
presentation) to occur in an AP is correlated with the degree of syntactic embedding of the
word boundary occurring at the left edge of that AP. Consider that for the sequence les
bagatelles et les balivernes saugrenues (‘the crazy trifles and nonsense’), the adjective
saugrenues (‘crazy’) may be interpreted as modifying only the second NP les balivernes (‘the
nonsense’) as in (1b), or as modifying the conjunction of the two NPs les bagatelles et les
balivernes (‘the trifles and the nonsense’) as in (1a). A typical phrasing pattern for this string
is shown in (2).
(1) Syntactic possibilities:
a. ([[les bagatelles] et [les balivernes]] saugrenues])
the trifles and the nonsense crazy
‘the crazy trifles and nonsense’
b. ([les bagatelles] et [[les balivernes] saugrenues])
(2) Typical phrasing pattern:
(les bagatelles)AP (et les balivernes)AP (saugrenues)AP
This predicts that the embedding level of the syntactic juncture between the two NPs
will be lower in (1a) and higher (1b), while the opposite is predicted for the juncture between
the second NP and the adjective. The authors show in a controlled production experiment that
LHi is more likely to occur when the left edge of AP1 corresponds to a syntactic juncture that
is less deeply embedded. In other words, LHi occurred more often at the left edge of the AP
comprising saugrenues for (1a) than for (1b), while it occurred more often at the left edge of
the AP comprising et les balivernes more often for (1b) than for (1a).
While this finding seems to point to a model in which LHi is recruited to disambiguate
syntactic alternatives, it is important to note that phrase length also had an effect on LHi
occurrence, such that longer content words were associated with higher rates of LHi. The
authors conclude from this finding that LHi is principally a marker of phonological structure,
and suggest that the syntactic effects might be indirect and possibly mediated by an
intermediate level of phrasing. That both effects were present in the same study suggests a
complex picture for the distribution of LHi.
3.3. LHi and Information Structure
The relationship of LHi to information structure presents a similarly complex picture. While
focus, contrast and certain types of emphatic meaning are often associated with accentuation
of non-final syllables (Di Cristo 1999a, 2000; Jun and Fougeron 2000; Féry 2001, inter alia),
typically these involve specialized contours that are distinct from LHi in several ways.
Perhaps most significantly, these accents are reported as occurring alone in their phrasal or
intonational unit. Additionally, they are typically assumed to be associated to a prominently
stressed syllable. LHi, by contrast, always occurs in an AP along with LH* and has no
necessary association to stress. Jun and Fougeron (2000) elicited such specialized non-final
accents in contexts like (3b), in which one word was singled out as carrying a corrective
(3) a. Marion ne mangera pas des ananas au petit déjeuner, mais…
Marion NEG eat.FUT NEG some pineapples at breakfast but
‘Marion will not eat pineapples for breakfast, but…’
b. Marion mangera des bananes au petit déjeuner.
Marion eat.FUT some bananas at breakfast
‘Marion will eat some bananas for breakfast.’
The authors report that these sentences were generally produced with a single
prominent rise (a focus accent, or Hf), which was higher than LH* in a corresponding neutral
focus pattern, and was followed by a low plateau extending to the end of the utterance (or a
high plateau for polar interrogatives). Although Hf occurred on both final and non-final
syllables, however, the resulting pattern was never LHLH. Such instances of non-final rises
therefore more closely resemble the various accents d’insistance or accents emphatiques
mentioned by Di Cristo (1999a, 1999b, 2000) and others, and they are distinct from the LHi
of the framework we are assuming here.
Di Cristo (2000) observes that non-final rises occur in conjunction with final accents
in cases of information focus2. In that proposal, constituency at various levels of the grammar
may be reflected in prosodic structure by a “bipolarization” effect, whereby “promotion of the
extremities” of some constituent results in accents being assigned to both its left and right
edges. More recently, Beyssade et al. (2010) present the results of two studies showing that
information focus is both marked in production and identified in perception by prosodic
highlighting, which in their proposal “involves an initial accentuation, (IA) which may form
an accentual arch with the following rising accent, or triggers a high plateau up to the
following accent.” To the extent that such non-final rises correspond to LHi, this suggests a
more complicated picture than even Astésano et al.’s (2007) study presents. It raises the
possibility, in other words, that LHi is sensitive not only to phonological, stylistic and
syntactic factors, but to discourse-level ones (i.e., focus) as well.
Overall, the relationship between prosody and information structure is not well-
understood for French. Of the wide range of effects that have been suggested, perhaps the
most well-documented is that the region between the end of a corrective or emphatic focus
and the end of the sentence (or post-focal region) may exhibit a marked reduction in f0 range
and f0 movement. In declaratives, this takes the form of a low plateau extending from the end
of the last major falling event, while for polar interrogatives this is a high plateau extending
more or less from the highest point of the last rise (Jun and Fougeron 2000). The phenomenon
has variously been formulated in terms of dephrasing or deaccenting, though Jun and
Fougeron (2000) show that such regions maintain their durational cues to phrasing, and thus
may not be appropriately characterized as “dephrased”.
On the view that the location and extent of the focused constituent form part of a
speaker’s intentions that must be recovered by the listener, then post-focal deaccenting
provides an effective cue to the location of the right edge of focus, since it marks a kind of
transition point between the focal region and the post-focal region. It is not clear whether such
cues exist for signaling the left edge of focus in French. For one thing, there does not seem to
be the equivalent of prefocal deaccenting. Instead, both the prefocal and focal regions are,
subject to conditions of rhythm and syntax, phrased into APs bearing final rises. The result is
a more or less continuous distribution of accentual features that span the boundary between
the prefocal and focal regions. Consider the sentence from (3b) shown in (4) with hypothetical
foci of two different sizes having different left edges.
(4) Marion mangera [des bananes]F au petit déjeuner
( )AP ( )AP ( )AP
LH* LH* LH*
Marion [mangera des bananes]F au petit déjeuner
( )AP ( )AP ( )AP
LH* LH* LH*
Notice that in both cases, regular phrasing into APs results in an identical distribution
of LH* accents. Thus, if only the distribution of LH* is considered, it is not possible for a
listener to recover the location of the left edge of focus in such a case. While there is limited
evidence that the pre-focal region is distinguished from the focal region by having a
compressed f0 range (Touati 1987; Jun and Fougeron 2000; Dohen and Loevenbruck 2004), it
has not yet been established whether this is a reliable marker that could be used by the listener
to recover the position of the left edge of focus.
In both Jun and Fougeron (2000) and other quantitative studies (Dohen and
Loevenbruck 2004), focus is conceptualized in terms of corrective or emphatic contrast. Féry
(2001) reports on a production study involving a notion of focus based on direct answers to
wh-questions, though the quantitative results are not presented. Other work has emphasized
the role of additional focus-like categories including intensification (Di Cristo 2000) and
information focus (Di Cristo 2000; Beyssade et al. 2010). Krifka tentatively proposes to unite
the various notions of focus cross-linguistically under the idea that it is used to “indicate the
presence of alternatives that are relevant for the interpretation of linguistic expressions”
(Kifka 2007: 18). What it means to be “relevant for interpretation” may vary depending on the
particular phenomenon being observed. In the case of certain focusing adverbs, for example,
the way in which focus gives rise to alternatives may have consequences for the truth
conditions of the sentence. In its more pragmatic uses, focus may be relevant for
interpretation because it establishes parallels between different linguistic units in the same
discourse. This latter notion is closely related to Rooth’s (1992) construal of Contrastive
Focus and forms the basis of a pair of studies conducted by German and D’Imperio (2009,
In the first of these studies, German and D’Imperio (2009) found preliminary evidence
that initial rises are more likely to occur in APs whose left edge coincides with the left edge of
a contrastive focus. In that study, a reading task was used to elicit contrastive focus in polar
interrogative clauses that were preceded by closely parallel clauses. This is illustrated in (5).
(5) a. Je sais qu’Amélie a vendu la valise jaune dans la rue Mignet, mais…
I know that Amélie has sold the suitcase yellow in the street Mignet but
“I know that Amélie sold the yellow suitcase on Mignet Street, but…”
b. …a qui est-ce qu’Amélie a vendu la valise [marron]F dans la rue Mignet?
to whom is it that Amélie has sold the suitcase brown in the street Mignet
“…to whom did Amélie sell the [brown]F suitcase on Mignet Street?”
Since the two complement clauses in (5a) and (5b) differ only in the value of the
adjective modifying the noun valise, the prediction is that marron (‘brown’) is established in
this context as a contrastive focus. Keeping the target sentence (5b) fixed, the context was
then varied in a way that systematically established foci of different sizes. In (6) and (7), for
example, the complement clause in the first sentence in each pair differs from that in the
second sentence in such a way as to establish the direct object (la valise marron) and the VP
(vendu la valise marron) as the contrastive foci in (6b) and (7b), respectively. Note that the
location of the right edge of the contrastive focus region is not predicted to differ across the
(6) a. Je sais qu’Amélie a vendu la caméra dans la rue Mignet, mais…
I know that Amélie has sold the camera in the street Mignet but
“I know that Amélie sold the camera on Mignet Street, but…”
b. …a qui est-ce qu’Amélie a vendu [la valise marron]F dans la rue Mignet?
to whom is it that Amélie has sold the suitcase brown in the street Mignet
“…to whom did Amélie sell [the brown suitcase]F on Mignet Street?”
(7) a. Je sais qu’Amélie a déjeuner dans la rue Mignet, mais…
I know that Amélie has eaten lunch in the street Mignet but
“I know that Amélie ate lunch on Mignet Street, but…”
b. …a qui est-ce qu’Amélie a [vendu la valise marron]F dans la rue Mignet?
to whom is-it that Amélie has sold the suitcase brown in the street Mignet
“…to whom did Amélie [sell the brown suitcase] on Mignet Street?”
In short, LHi showed a tendency to “track” the left edge of focus across three different
contexts by occurring on the AP whose left edge coincided with the left edge of the focus (see
Figure 2). While the overall number of items in the study was low, the observation echoes Di
Cristo’s (2000) suggestion that initial rises (or LHi) may be one structural feature that can be
recruited to mark the location of the left edge of focus.
In a follow-up study, German and D’Imperio (2010) more directly addressed the
question of whether LHi marks the left edge of focus. The targets in that case were wh-
questions including a direct object consisting of a 3-syllable noun phrase (i.e., le merlan) and
either a 3- or 5-syllable prepositional phrase (underlined), followed by a temporal modifier.
This is illustrated in (8).
(8) a. 3-syllable (short): Qui a commandé le merlan aux navets ce soir?
who has ordered the whiting with turnips this evening
“Who ordered the whiting with turnips this evening?”
b. 5-syllable (long): Qui a commandé le merlan aux macadamias ce soir?
who has ordered the whiting with macadamias this evening
“Who ordered the whiting with macadamias this evening?”
These sentences were elicited as the second of a series of three information-seeking
questions3. Thus, specific patterns of focus were induced by manipulating the size of
correspondence between the target and the questions occurring before and after it. In (9b), for
example, the focused element is predicted to be the entire direct object [le merlan aux navets],
since the surrounding context (i.e., (9a) and (9c)) suggests that that constituent is the relevant
point of contrast. In the PP-focus condition shown in (10b) on the other hand, the focus is
limited to the prepositional phrase aux navets for similar reasons. Thus, each target item
occurred in each of four conditions: DO-Focus/short, DO-Focus/long, PP-Focus/short and PP-
a. Qui a commandé l’entrecôte ce soir?
who has ordered the steak this evening
“Who ordered the steak this evening?”
b. Qui a commandé [le merlan aux navets]F ce soir?
who has ordered the whiting with turnips this evening
“Who ordered the whiting with turnips this evening?”
c. Qui a commandé les gambas ce soir?
who has ordered the prawns this evening
“Who ordered the prawns this evening?”
a. Qui a commandé le merlan à la sauce citron ce soir?
who has ordered the whiting with the sauce lemon this evening
“Who ordered the whiting with lemon sauce this evening?”
b. Qui a commandé le merlan [aux navets]F ce soir?
who has ordered the whiting with turnips this evening
“Who ordered the whiting with turnips this evening?”
c. Qui a commandé le merlan aux câpres ce soir?
who has ordered the whiting with capers this evening
“Who ordered the whiting with capers this evening?”
If LHi marks the left edge of focus, then the prediction is that LHi should be more
likely to occur at the left edge of an AP comprising the noun phrase in the case of (9b) (since
that is also predicted to be the left edge of the contrastive focus), while it should be more
likely to occur on an AP comprising the prepositional phrase in a case like (10b). Recall,
however, Welby’s (2006) finding that longer APs are more likely to include LHi. On the
assumption that the prepositional phrase comprises a single AP in both (8a) and (8b), this
leads to a second, independent prediction that LHi should be more likely to occur on the
prepositional phrase when it is long than when it is short.
In fact, the results of 192 tokens taken from eight speakers showed a fairly strong
correlation between focus and LHi. For the position corresponding to the left edge of the
prepositional phrase, LHi was more likely when that position coincided with the focus left
edge (51% versus 33%), as illustrated in Figure 8. Similarly for the left edge of the noun
phrase, LHi was significantly more likely when that position coincided with a focus edge
(34% versus 21%). Importantly, the length of the prepositional phrase also mattered. As
predicted, LHi was more likely to occur on the prepositional phrase when it was long (54%)
than when it was short (30%).
Qui a noté les aveux aux végétariens ce matin
!"#$ !"#$ !"%$ !"#$ "&$
Figure 8: LHi realized at the left edge of the prepositional phrase in a PP-focus context (long)
for one speaker in the utterance Qui a noté les aveux [aux végétariens]F ce matin ‘Who wrote
down the confessions to the vegetarians this morning’.
Statistical analysis4 revealed that both factors (length and focus size) were significant for the
occurrence of LHi for the position corresponding to the left edge of the prepositional phrase.
However, the analysis revealed no interaction between these factors. In other words, the effect
of focus size was constant whether the prepositional phrase was long or short. Conversely,
prepositional phrase length had a constant effect whether the focus included just the
prepositional phrase or the entire direct object.
This latter finding is somewhat surprising. Consider that if the realization of LHi is
directly related to AP length, due, for example, to some (possibly probabilistic) constraint on
the minimum size of AP that will support both LHi and LH*5, then LHi should be so unlikely
in very short APs that its occurrence is insensitive to the locations of focus boundaries. In the
limit, one-syllable APs cannot support LHi at all6. Similarly, if the relationship between LHi
and phrase length is due to a rhythmic constraint requiring longer APs to include LHi, then
LHi should be so common in very long APs (e.g., seven syllables) that the effects of focus are
negligible. Instead, the data show that the effects of both factors are independent and additive,
at least for the particular range of AP length that was addressed.
Consider also the implications of the fact that LHi was sensitive to both AP length and
focus. It suggests, among other things, that LHi is not a marker of focus in the traditional
sense. Given that LHi is sensitive to AP length at all, in other words, the association between
LHi and focus can only ever be partial or imperfect at best. Setting aside for a moment the
issue of precision in the study, a follow-up analysis showed that a listener who uses LHi as
the sole indicator of focus size would be able to make the correct prediction only 60% of the
time7. A similar remark applies to the findings of Astésano et al.’s (2007) study.
The findings presented above suggest that the relationship between LHi and each of the
factors influencing its distribution is indirect in one or more ways. But what type of model
could capture such effects without assuming a direct relationship? Astésano et al. (2007)
conclude from their results that LHi is first and foremost a reflex of phonological structure,
and on that basis suggest that it is linked to an intermediate level of phrasing. While the
authors do not propose a specific mechanism along these lines, there are certain
characteristics that such a model would have to include in order to account for the observed
correlations. In what follows, we outline some of those characteristics and suggest promising
approaches for incorporating them into existing models. However, in the absence of further
data, for example, linking the effects of syntax with the effects of focus and phrase length, or
a more direct metric for the identification of ip, we stop short of making an explicit proposal.
One promising approach to indirect relationships like those suggested by the data
above is to assume that an additional level of description, such as ip, mediates between the
two levels of primary interest. To explain the association between focus and LHi, for
example, we might assume that the operative principle involves a rule or constraint requiring
that a focus constituent project this intermediate level. In short, a focus constituent must be
contained in a single ip8. A very similar proposal would apply to the syntactic effects: all else
being equal, a less embedded syntactic constituent has a greater tendency to project an iP than
a more embedded one.
Such a principle might itself be weak, probabilistic, or even optional. We know from
the D’Imperio and Michelas (2010) studies, for example, that ip boundaries appear at the
subject-VP juncture in broad focus utterances, so any focus-based rule or constraint must at a
minimum accommodate the role of syntax. Pierrehumbert (1994) proposes that for English,
there is a variable constraint on the size of the prosodic domain that gets promoted to a level
of stress that would be sufficient for accentuation. This process of prosodic promotion is not
random in that account, but may “reflect discourse factors” (Pierrehumbert 1994: 15) and is
“generally available to strengthen prosodically weak elements if the speaker for any reason
wishes to accent them” (Pierrehumbert 1994: 13). In the case of French, the process would
apply to promote a given constituent, not to a higher degree of stress, but to a higher level unit
in the prosodic hierarchy. Thus, a string that is initially parsed as an AP within some larger ip,
would by way of prosodic promotion be reparsed as its own ip. In effect, the suggestion is that
speakers select, with some variability, the size of the unit to which this process applies based
on discourse needs such as signalling the scope of focus or disambiguating between possible
Note that it is not sufficient to establish a relationship between focus or syntax and ip.
The link still needs to be made between LHi and ip. Importantly, however, the results of the
Michelas and D’Imperio (2010b) studies are not suggestive of a one-to-one association
between these two levels either (though no conclusive analysis has been conducted). More
data is needed to determine whether this relationship is itself probabilistic, or whether there
are additional factors or levels of description that are still unaccounted for.
Finally, any viable model needs to account for the effects of length on LHi. One fairly
straightforward approach is to assume that longer APs are more likely to be parsed as their
own ip. Similar to the case of focus, this in turn could be explained by a probabilistic or
stylistically driven constraint on the domain of ip assignment. Thus, with the single
assumption that LHi is a feature of an ip (as opposed to an AP), it is possible to explain why
longer APs are more likely to include LHi without invoking either rhythmic or tonal crowding
constraints, which are not suggested by the data in any case.
The model elements sketched above are merely suggestive, and we wish to be clear
that there is currently is no data that straightforwardly supports one particular model over any
other. However, the fact that the data may be explained by assuming a level of prosodic
structure for which there is independent evidence is, we believe, significant. The purpose of
presenting it in this fashion is therefore to highlight the potential for complex interactions in a
problem that is typically approached from only one perspective. On the one hand, a traditional
focus-to-accent approach seeks to identify robust markers of information structure in the
traditional sense, leaving little role for structural effects at the level of phrasing, for example.
Crucially, it cannot explain why the left edge of a focus region would be so unreliably marked
given that it is marked at all, or why LHi is sensitive to so many different factors. On the other
hand, a phrasing or edge-based approach that emphasizes the comparative lack of intonational
“marking” in French as compared with Germanic languages, for example, has too little to say
about the role of post-focal deaccenting on the one hand, and the complex distribution of LHi
on the other. The point we wish to highlight here is that the type of data that will eventually
lead to an adequate model of prosody and information structure in French will need to take
into account both issues simultaneously by, for example, establishing the relative effect size
of factors from different levels of description within the same study, following the precedent
of Astésano et al. (2007), German and D’Imperio (2010), and others. Moving forward, then,
our approach seeks an integrated model that takes into account both types of descriptions, as
well as any additional interdepencies that they bring to the problem.
We would like to thank two anonymous reviewers and the editors of this volume for helpful
comments on this paper. This research was partly funded by a grant from the ANR (Agence
Nationale de la Recherche), Pro-Gram "La Prosodie dans la Grammaire", from the French
Ministry of Research.
" Note that AP is reserved solely for reference to the Accentual Phrase throughout our
discussion, and not to the Adjective Phrase.
# In direct answers to wh-questions, an information focus is the portion of the utterance that
corresponds to the wh-element of the question that it resolves.
$ The participants used the questions to retrieve specific pieces of information from a
speaking partner in a cooperative task.
% The data were modeled using mixed effects linear regression treating PP-length and focus
size as fixed effects, and items as random effects.
& In Post’s (2000) proposal, non-final accents essentially represent a strategy for stress-clash
avoidance after restructuring (i.e., boundary erasure). Since restructuring typically leads to
larger phrases, non-final accents are predicted to be unnecessary in very small phrases.
' In Jun and Fougeron’s (2000) proposal, each tone must be realized on its own syllable, and
every AP must include at least one low tone. This predicts that LHi should be impossible for
even a two-syllable AP, though Welby (2006) shows experimentally that this prediction is not
born out. In that study, initial rise occurred on 13% of all two-syllable words.
( A chi-square showed that this rate differed from chance (51%) at the p<0.1 level.
) Notice the similarity to Féry’s (2001) proposal, which suggested that a focus is constraint to
project its own phonological phrase.
Astésano, Corinne 2001 Rythme et accentuation en français : Invariance et variabilité
stylistique. Paris: L’Harmattan.
Astésano, Corinne, Ellen Gurman Bard and Alice Turk 2007 Structural influences on
initial accent placement in French. Language and Speech 50(3): 423-446.
Beckman, Mary and Janet Pierrehumbert 1986 Intonational Structure in Japanese and
English. Phonology Yearbook III, 15-70.
Beyssade, Claire, Barbara Hemforth, Jean-Marie Marandin and Crisetl Portes 2010
Information focus in French. Proceedings of Speech Prosody 2010, 100109:1-4.
Chicago, United States.
Delattre, Pierre 1966 Les dix intonations du français. The French Review 40(1): 1-14.
Dell, François (1984) L’accentuation dans les phrases en français. In : François Dell,
Daniel Hirst and Jean-Roger Vergnaud (eds.), Forme Sonore du Langage: Structure des
Représentations en Phonologie, 65-122. Paris: Hermann.
Di Cristo, Albert and Daniel Hirst 1993 Rythme syllabique, rythme mélodique et
représentation hiérarchique de la prosodie du français. Travaux de l’Institut de Phonétique
Di Cristo, Albert 1999a Vers une modélisation de l’accentuation en français. Première
partie: la problématique. Journal of French Language Studies 9: 143-179.
Di Cristo, Albert 1999b Le cadre accentual du français contemporain: Essai de
modélisation. Première partie. Langues 2 (3): 184-205 and Langues 2 (4): 258-267.
Di Cristo, Albert 2000 Vers une modélisation de l’accentuation
en français. Deuxième partie: Le modèle. Journal of French Language Studies 10 :27-44.
Delais-Roussarie, Elizabeth 2005 Interface Phonologie / Syntaxe : des
domaines phonologiques à l’organisation de la Grammaire. In : Jacques Durand, Noël
Nguyen, Véronique Rey and Sophie Wauquier-Gravelines (eds.), Phonologie et Phonétique:
Approches actuelles, 159-184. Paris: Hermès.
D’Imperio, Mariapaola, Roxane Bertrand, Albert Di Cristo and Cristel Portes 2007
Investigating phrasing levels in French: Is there a difference between nuclear and prenuclear
accents?. In: José Camacho, Viviane Déprez, Nydia Flores and Lilian Sanchez (eds.), Selected
Papers from the 36th Linguistic Symposium on Romance Languages (LSRL), 97-110. New
Brunswick: John Benjamins Publishing Company.
D’Imperio, Mariapaola and Amandine Michelas 2010 Embedded register levels and
prosodic phrasing in French. Proceedings of the Speech Prosody 2010 Conference,!100879:1-
4. Chicago, United-States.
Dohen, Marion and Hélène Loevenbruck 2004 Pre-focal rephrasing, focal enhancement
and post-focal deaccentuation in French. Proceedings of the 8th International Conference on
Spoken Language Processing, 785-788. Jeju, Korea.
Dohen, Marion and Hélène Lœvenbruck 2009 Interaction of audition and vision for the
perception of prosodic contrastive focus. Language & Speech 52(2-3): 177-206.
Feldhausen, Ingo 2008 The Prosody-Syntax Interface in Catalan. PhD thesis.
University of Postdam, Germany.
Féry, Caroline 2001 Intonation of focus in French. In: Caroline Féry and Wolfgang
Sternefeld (eds.), Audiatur Vox Sapientes: A Festschrift for Arnim von Stechow, 153-181.
Berlin: Akademi Verlag.
German, James and Mariapaola D’Imperio 2009 Information structure and the alignment of
phrasal features in French qu-interrogatives. Poster presented at the international conference
Interfaces Discours et Prosodie, Paris, France.
German, James and Mariapaola D’Imperio 2010 Focus, phrase length, and the distribution
of phrase-initial rises in French. Proceedings of International Conference on Speech Prosody
2010, 100207:1-4. Chicago, United-States.
Gussenhoven, Carlos 2004 The Phonology of Tone and Intonation. Cambridge: Cambridge
Hirst, Daniel and Albert Di Cristo 1984 French intonation: a prametric approach. Die
NEuren Sprache 83(5): 554-569.
Jun Sun-Ah and Cécile Fougeron 1995 The Accentual Phrase and the Prosodic Structure
of French. Proceedings of the 13th International Congress of Phonetic Sciences 2: 722-725.
Jun Sun-Ah and Cécile Fougeron 2000 A Phonological model of French intonation. In:
Antonis Botinis (eds.), Intonation: Analysis, Modeling and Technology, 209-242. Dordrecht :
Kluwer Academic Publishers.
Jun Sun-Ah and Cécile Fougeron 2002 The realizations of the Accentual Phrase in
French intonation*! Probus 14 (special issue on Intonation in the Romance Languages): 147-
Krifka, Manfred 2007 Basic notions of information structure. In: Caroline Féry and
Manfred Krifka (eds.), Interdisciplinary Studies of Information Structure 6, Potsdam.
Ladd, Robert Dwight. 2006. Intonational Phonology. Cambridge University Press (first
Loevenbruck, Hélène 1999 Articulatory effects of contrastive emphasis on the Accentual
Phrase in French. Journal of Acoustic Society of America, 106(4) : 1553-1573.
Lucci, Vincent 1983 Etude Phonétique du Français Comtemporain à travers la
Variation Situationnelle. Grenoble: Publications de l’Université des Langues et des Lettres de
Mertens, Piet 1992 L’accentuation de syllabes contiguës. International Journal of Applied
Linguistics, 95-96: 145-165.
Mertens, Piet 2004 Quelques allers-retours entre la prosodie et son traitement automatique.
Le français moderne, 72(1): 39-57.
Michelas, Amandine and Mariapaola D’Imperio 2010a Durational cues and prosodic
phrasing in French: Evidence for the intermediate phrase. Proceedings of International
Conference on Speech Prosody 2010, 100881:1-4. Chicago, United-States.
Michelas, Amandine and Mariapaola D’Imperio, M 2010b Indices phonétiques et
contraintes phonologiques: Caractérisation du syntagme intermédiaire en français. Actes des
XXVIIIème Journées d’Etude sur la Parole, Mons, Belgique, 77-81.
Michelas, Amandine and Mariapaola D’Imperio 2010c Rôle des indices globaux dans la
définition des constituants prosodiques en parole spontanée en français. Paper presented at the
Journées Rhapsodie de juillet 2010. Toulouse. France
Pasdeloup, Valérie 1990 Modèle de règles rythmiques du français appliqué à la synthèse
de parole. PhD Dissertation, Université de Provence, Aix-en-Provence, France.
Pierrehumbert, Janet 1994. Alignment and prosodic heads. In: Proceedings of the Eastern
States Conference on Formal Linguistics 10: 268-286.
Post, Brechtje 2000 Tonal and Phrasal Structures in French Intonation. The Hague:
Holland Academic Graphics.
Rooth, Mats 1992 A theory of focus interpretation. Natural Language Semantics 1: 75-
Rossi, Mario 1985 L’intonation et l’organisation de l’énoncé. Phonetica 42 :135-153.
Rossi, Mario 1999 L’intonation: Le Système du Français. Gap: Orpheus.
Selkirk, Elisabeth 2000 The interaction of Constraints on Prosodic Phrasing”. In: Merle
Horne, Prosody: Theory and Experiment (eds.), 231-262. Dordrecht: Kluwer Academic
Shih, Chilin 1986 The Prosodic Domain of Tone Sandhi in Chinese. PhD dissertation.
University of California San Diego, La Jolla, CA, USA.
Touati, Paul 1987 Structures Prosodiques du Suédois et du Français. Lund : Lund
Vaissière, Jacqueline 1974 On French prososdy. Research Laboratory of Electronics
Progress Report 115: 212-223.
Vaissière, Jacqueline 1992 Rhythm, accentuation and final lengthening in French. In: Johan
Sundberg, Lennard Nord and Rolf Carlson (eds.), Music, Language, Speech and Brain.
Houndmills: MacMillan Press.
Welby, Pauline 2002 The realization of early and late rises in French intonation: A
production study. Proceedings of the Speech Prosody Conference 2002, 695-698. Aix-en-
Welby, Pauline 2006 French intonational structure: Evidence from tonal alignment.
Journal of Phonetics 34(3): 343-371.