A contribution to the study of phonetic variation of /r/ in French and Italian
linguistic domains
Antonio Romano
Università di Torino (Italy) / Centre de Dialectologie de Grenoble (France)
Dip. di Scienze del Linguaggio - Università di Torino
via Sant’Ottavio, 20 – I-10124 Torino - Italy
[email protected]
Extended version (Draft)
Submitted to H. Van de Velde, R. van Hout & D. Demolin (eds.), ‘r-atics. Sociolinguistic,
phonetic and phonological characteristics of /r/.
Introduction
1. General variability : articulatory principles and acoustic cues
2. French r
2.1. Historical changes
2.2. Socio-geographical variation in common French
2.3. Double r
2.4. Devoicing and final deletion
2.5. A contribution to the phonetic description of r-sounds
2.5.1. Studies in laboratory speech make-up
2.5.2. Secondary articulations in r-sounds
2.5.3. Articulation manner and timing for r-sounds
2.5.4. Citation forms vs. connected speech
3. Italian r
3.1. Phonetic phenomena involving r
3.2. Back (and not only) r-sounds in Italian
3.3. Regional preferences and defective r-sounds
3.4. Instrumental phonetic analysis
3.5. A TV corpus: connected and spontaneous speech
4. Conclusions
Acknowledgements
References
1
Introduction
This paper aims to open a discussion around the phonetic variability of the different
realizations of r in Italy and France.
The work here presented summarizes a few experencies carried out on phonetic grounds
during the last 5 years in France, at the Centre de Dialectologie de Grenoble, and in Italy, at
the Universities of Lecce and Turin.
The paper discusses from a descriptive point of view a selection of phonetic phenomena
which have been observed in the current use of speakers from different areas and which could
be confirmed, in most cases, by experimental evidence.
The literature on the topic - for all the varieties here considered - shows how a simplified
view is still dominating in spite of progresses in the description of these sounds in other
linguistic domains (see Van de Velde & Van Hout 2001). The main r-variability described in
French and Italian domains is stylistic (and sometimes, for special pronunciations, also
diatopic) and, paradoxally, very few details are usually available for a phonetic approach and
for a better description of distributional properties.
About French, this lack is particularly marked. According to Demolin (2001:71), the
"phonetic variation of French rhotics needs to be described in much more detail". Behind the
traditional dichotomic representation of an apical r vs. a uvular r (sometimes masked by
general labels such as r grasseyé, which were used to describe, following the authors'
impressions, quite different classes of sounds) stands a typological vagueness which
obviously characterizes large diffusion books, but also part of the recent scientific literature.
The present contribution will be limited to an instrumental survey of a sample of about 1000
r-realizations from 8 French speakers (of different origins and with idiosyncratic phonetic
properties) and their description in terms of "narrow phonetics".
As for Italy, a symptom of the different considerations connected with r-pronunciation is the
disagreement on the sociolinguistic status accorded in the international literature to some rsounds. Even when the authors agree on their articulatory description, different opinions on
the prestige status of dialectal variants clearly reveal the uncomplete (and, more often, nonuniform) knowledge of the geographical and social variability of these units within the Italian
diasystem. In spite of the presence of a large number of interesting phenomena involving the
realizations of /r/ and /rr/, significant emphasis is given to what is usually called r moscia
('limp' or 'lifeless r') which is a simple way – as it has been underlined by some phoneticians
2
(see Canepari 1979 and Mioni 1987) – to class together, in the same popular category, more
than 10 phonetically different basic articulations.
For all these r-sounds, besides the supposed presence or not of uvular vibrations, we need a
better articulatory description, including details on the place and on the degree of constriction,
on the dynamics of vibrations (when really present), on voicing properties, and so on.
This paper summarizes the research I have been carrying out during the last 5 years in order to
give a contribution to this need. In particular, it includes the main outcomes of an
instrumental enquiry on more than 300 utterances of speakers from different regions of Italy
(and from TV speakers affected by non-standard pronunciations), showing as most of the
acoustic properties of the sounds described by various phoneticians for a number of languages
(e.g. Delattre 1944-71, Jakobson 1957, Fant 1968, Ladefoged et al. 1977, Recasens 1991, and
Ladefoged & Maddieson 1996) are also confirmed by Italian data. Besides a dependence on
general patterns of temporal organisation, this study shows how speakers have recourse to
different strategies to get non-apical r-sounds by using the acoustic (and perceptual) effects of
rapid changes in some formant frequencies patterns.
3
1. General variability : functional principles, articulatory strategies and acoustic cues
The aim of this secton is to introduce articulatory and acoustic properties of an enormous
class of speech sounds related to rhotic phonemes, whose realizations are renowned to be
particularly speaker- and language-dependent.1
Rhotics are usually associated with apical trills, which are the central members of the class,
but an enormous variety of other sounds are concerned in various languages and dialects.2
While phonetic modeling reveals that an efficient tongue-tip vibration depends on airflow,
impedance, and appropriate apical control (McGowan 1992:2903, Widdison 1997:191), apical
trills are also described as articulatory gestures needing narrower aerodynamic requirements
than other sounds (Recasens 1991, Solé 1999). That could be a valid reason explaining why
they usually undergo all kind of variation.
In the literature, trills are described in terms of extremely fine articulations:
Learning to make a trill involves placing the tongue, very loosely, in exactly the right
position so that it will be set in vibration by a current of air. [...] The problem
experienced by most people who fail to make trills is that the blade of the tongue is too
stiff (Ladefoged 1993:169).
Recently Barry (1997) and Catford (2001) reopened the classical debate on historical
evolution of r-sounds in different languages.
A well-known case study has been traditionally represented by French, whereas nowadays
many other languages, including Italian, show interesting social dynamics involving r-sounds.
The phonological starting assumptions are that in standard French there is synchronically only
one rhotic phoneme, whose common realization is described as a uvular trill, whereas in
1
According to Stevens (1989:40), in speech there are " articulatory states or configurations or gestures [...]
[which] give rise to well-defined patterns of auditory response in human listener, such that these patterns are not
strongly sensitive to small perturbations or inaccuracies in the articulation". Rhotics are perhaps the best example
of sounds categories representing such a kind of general mechanisms.
2
A clear description of the mechanism allowing to get apical trills is the following: "The tongue blade or tip is
brought into complete closure with the alveolar ridge (or very nearly so) often in a slightly cupped aspect.
Contact pressure is relatively light, and oral pressure builds up rapidly to the point of forcing its way through the
closure. A high-speed jet of air escapes through the gap, and a combination of elastic muscle forces and the
sucking action of the Bernoulli effect in the air-jet bring the tongue surface very quickly back into renewed
contact. The cycle is then repeated for as long as the stricture and the air supply are maintained" (Laver
1994:219).
4
standard Italian an apico-alveolar phoneme /r/ contrasts in postvocalic position, following the
consonant gemination pattern generalized in the whole system, with a geminate counterpart
/rr/.
As for French, I agree with the general Shibles' (1994:53) claim that "The phonetic
transcription of French is inconsistent and the descriptions of the articulations of a number of
sounds are controversial. A stereotyped or phonemic transcription is typically given instead of
the actual sounds heard."
This is generally true for a number of languages and transcription usages. It is especially true
for French rhotics whose real phonetic implementation is often disregarded. Ladefoged &
Maddieson (1996:225) observe that "Uvular trills occur in some conservative varieties of
Standard French and Standard German, although most speakers of these languages use uvular
fricatives or approximants rather than trills".3
On the other hand, very few details are given by articulatory research to support phoneticbased accounts (except the recent contributions by Autesserre & Chafcouloff 1998), and
uvular trills dominate the general linguistic literature on French r.
As for Italian, a simplified functional view allows to assume that the sounds realizing the two
phonological units are, in both cases, apical trills with a different number of contacts.4
3
A slightly different view is available in the description given by Canepari (1979:248). The author presents the
French /r/ phoneme as a uvular fricative, often realized as a weak uvular approximant, mainly after vocoid (and
followed by another vocoid, or a contoid, or final); moreover a fricative sound is described after a contoid or
silence (that is in initial position) and at the onset of a stressed syllable (even when following a vocoid).
Canepari (1979) transcribes it in the two following ways: uvular fricative (phoneme use) // and uvular
approximant [ ]. Note that some authors still use the non-conventional symbol [ ] (which has been retired from
the IPA chart) for the voiced uvular fricative instead of [] whereas Canepari uses [ ] as an additional symbol
for the approximant sound.
4
As a general reference, see Ladefoged & Maddieson (1996): "[...] Apical trills typically consist of two or three
periods of vibration [...] but may contain only one or have more than three. Each period consists of a closed
phase during which the articulators are in contact, succeeded by an open phase in which they are slightly apart."
(Ladefoged & Maddieson 1996: 218). Uvular trills are separately described at p. 225 (ibid.) following the
experimental evidences in Delattre (1971): they are considered as produced by an initial movement of the tongue
root backwards followed by an upward movement toward the uvula, which is also moved forward to a position
where trilling can occur. Aerodynamic properties of uvular pulses are now analysed with real time MRI by
Demolin (2001).
5
Ladefoged & Maddieson (1996:219) affirm that "[i]n Italian, single and geminate forms of
most consonants contrast in intervocalic position" and that "[t]he single/geminate opposition
also applies to trills". Later in the same section, they add:
In repetitions of the words karo and karro from a total of five speakers of Standard
Italian we found none of the intervocalic single trills to have more than two contacts. The
geminate trills showed no fewer than three contacts and up to seven. (Ladefoged &
Maddieson 1996:221)5
Acoustic cues associated with the articulatory and distributional properties of these sounds
have been extensively analysed for different languages (e.g. Meyer-Eppler 1959, Delattre
1944-71, Ladefoged et al. 1977, Recasens 1991, Hagiwara 1995, Alwan et al. 1997).
With regard to their description in terms of temporal structure, vibration frequencies and
dynamic properties as they appear at an acoustic insight I suggest considering the following
basic definitions given by Ladefoged & Maddieson (1996:218):
5
Phonologically driven descriptions tend to differentiate this situation from the one observable in Spanish-
speaking countries and described since Navarro Tomás (1916). Ladefoged & Maddieson (1996) highlight that
"In a number of languages in which, unlike Spanish, there is regularly a distinction between single and geminate
consonants, the single and geminate rhotics differ in just the way that the Spanish segments do. Single rhotics are
taps, and geminate are trills" (Ladefoged & Maddieson 1996:237). On the other side, one could note that trills
are not just series of taps in row: trills are quite different from taps in the respect that tongue body is subject to a
higher degree of constraint during the production of a trill than of a tap (Recasens 1991, Kavitskaya 1997). As it
is discussed in the present paper, in a number of Italian idiolects single rhotics are not trilled, in agreement with
the examples presented by Ladefoged & Maddieson (1996:220) and as opposed to their Italian counterexamples
described in the text, where the geminate rhotics are just considered longer trills, with a greater number of
contacts than single rhotics (for a detailed analyses of the phonetic distribution of /r/ and /rr/ allophones see
Canepari 1979, 1983, 1999; also see below). Besides the accounts in Ladefoged (1968) and in Ladefoged &
Maddieson (1996), taps and flaps are often not distinguished (cp. Barry 1997:38). In Romance languages, a
distinction is usually made between "polyvibrants" and "monovibrants" without further defining different
articulatory possibilities within these classes (this choice traditionally matches the two-way perceptual
distinction proposed in Barry 1997:40 accounting for single-strike vs. multi-strike r-sounds). In a recent work,
Canepari (1999) gave the following classification for monovibrant r-sounds as "vibrati" vs. "vibratili" (the latter
are not included in Canepari 1983 who, as compensation, describes lateral or - as it is underlined - "lateralized"
tap, cp. Canepari 1983:92): "definiamo vibratile [in inglese flap] questo tipo particolare di vibrato [in inglese
tap], nel quale c'è prima uno spostamento dell'apice della lingua piú indietro del punto d'articolazione effettivo
(quello postalveolare) che súbito l'apice tocca rapidamente, per poi avvicinarsi ai denti inferiori" (Canepari
1999:101). Another account is in Mioni (1986:45) who defines taps as battiti 'beats', and flaps as scatti 'triggers'.
6
On the spectrograms the closed phases appear as light areas, as the formant energy is
absent or weak. The open phase between these two closures, which is vowel-like in its
acoustic structure, appears as a dark area with concentrations of energy in
characteristics formant regions. The closed phases of these trills last on the order of 25
ms each. The open phases have roughly similar duration, so that each complete cycle
occupies about 50 ms.6
Even the traditional timing for a single vibration in intervocalic position is described in most
varieties of Spanish and Italian has having a mean closure duration around 20 or 25 ms
(Vagges et al. 1975, Quilis 1981, Contini 1983, Recasens 1991). In the examples given by
Ladefoged & Maddieson (1996:231) for two tap realizations the spectrograms show durations
of about 20 ms and 25 ms.7
On the consistent distinctions in the spectral domain between uvular and apical r-sounds, one
may find an interesting framework designed for a number of languages in some traditional
acoustic approach (see for instance Jakobson 1957, Fant 1960, 1968, Delattre 1966, 1970).8
Ladefoged & Maddieson (1996:231), present a spectrographic account for tap realizations. In
6
Specific temporal characteristics of trills are detailed in the same p. 218 with data mainly based on a study for
different languages (in Lindau 1985) and other studies surveyed at p. 226. A general measure for the mean
vibration rate for trills is in the range 26-30 Hz (Though anatomically very different, bilabial, apical and uvular
trills vibrate at similar frequencies. Ladefoged et al. 1977 proposed an explanation based on the compensation of
different masses involved by decreasing the articulators' tenseness).
7
In their Italian enquiry on Florentine students, Vagges et al. (1978) showed that 7 speakers out of 10 realized
[r] as a monovibrant, 2 as a flap allophone and 1 as a "multivibrant" (p. 3 of the Leeds preprint). Mean duration
and standard deviation are reported in Table IV for the r of parata and are of 25 ± 18 ms. Similar values are
reported by Contini (1983) in his acoustic analysis of Sardinian "constrictives à battements" whose realizations
are "avec un seul battement apical, produisant une brève occlusion d'une durée de 2-3 cs [...] Pour le [r] à
battements multiples on peut compter de deux à cinq interruptions du spectre d'une durée à peu près équivalente.
Leur nombre est généralement plus élevé en syllabe postaccentuelle, à l'intervocalique [...] En syllabe accentuée,
ils sont le plus souvent trois (rarement quatre), tout comme à l'initiale de mot; au contacte avec une consonne
occlusive (à spectre continu) leur nombre ne dépasse pas deux. [...] Chaque interruption du signal est suivie d'un
élément vocalique de faible durée (1,5 à 3,5 cs) dont la répartition spectrale se rapproche de celle d'une voyelle
centrale. [...] Le [r] à plusieurs battements se caractérise donc par une structure formantique interrompue par de
brèves occlusions intermittentes" (Contini 1983:414-415).
8
Also see Calliope (1989) for the description of a vowel-colouring of r-sounds as []-like when apical and [o]-
like for back articulations. According to Ladefoged & Maddieson (1996:226), "There is a consistent distinction
in the spectral domain between uvular and apical trills, with the uvular trills showing a much higher third
resonance (between 2500 and 3000 Hz [...])".
7
some cases, in the VC transition one may observe a rising F2 (from 1600 Hz of the preceding
vowel to over 2000 Hz) while other realizations show a clear anticipation of the formant
structure of the final vowel by decreasing down F2 to about 1400 Hz.9 However, in all cases
(as well as for the Italian r-sound shown at p. 220), F1 tends to stay around 500 Hz (similar
characteristics are discussed by Ladefoged et al. 1977), against the pharyngeal r described by
Delattre (1970), showing higher F1 (up to 700 Hz), low F2 (about 1000 Hz), and slightly
higher F3.10
Besides these basic evidences, studies in the articulatory-acoustic framework, usually account
for effects of lip-rounding and pharyngeal constriction getting F3 lowering as well. As shown
by Delattre & Freeman (1968), for American English speakers presenting the common
retroflex realization, all these three articulations contribute to lower F3. Fant's detailed
description (Fant 1968) relates F3-F2 convergences to constriction gestures in the palatal or
pharyngeal regions whereas it presents lip-rounding as getting a main effect of lowering both
F2 and F3.
In such a way, all these phenomena could be explained by the recourse to the Perturbation
Theory but, as recently shown by Alwan et al (1997) and Espy-Wilson et al. (1997) - who
also postulated that the low F3 of American English /r/ is accompanied by constrictions in the
pharyngeal, palatal and labial regions - additional mechanisms are required to explain the
acoustics of these articulations.11
9
Recasens (1991) showed that the tap in Catalan has similar formant transitions going into and out of the
consonant. Also see the Italian examples in Ladefoged & Maddieson (1996:220, Fig. 7.2) where F2 rises towards
(and lowers from) respectively 1500 and 2000 Hz. In the results of the research in Vagges et al. (1975) for Italian
consonants, the differences in the realizations of 10 informants prevented from averaging the formant patterns of
[r] whose formant frequencies are claimed to be conditioned by the terminal values of the transitions from the
preceding and following vowels. Anyway the values in Fig. 6 (Leeds preprint) are about of 550 Hz for F1, 1450
Hz for F2, and 2500 Hz for F3 (also see Ferrero et al. 1979:138).
10
More finegrained distinctions could be made between advanced alveolar trills vs. more retracted alveolar trills
(as shown by Ladefoged & Maddieson 1996:222 for Malayalam and Dravidian Languages) "the more forward
trill has a higher locus for the second formant (at approximately 1750 Hz in comparison with 1250 Hz). The
more retracted trill has a lower third formant, as is commonly found in apical post-alveolar sounds".
11
In fact, speakers of rhotic dialects of American English use a moltitude of different articulatory configurations
for /r/ which typically include constrictions in the pharynx, along the palatal vault, and some degree of
constriction at the lips, but they also show tip-up articulations of the retroflex kind. Furthermore, real palatal
constrictions seem to be longer, and appear to cover areas considerably forward of the constriction location
predicted by Perturbation Theory to have the maximal lowering effect on F3. According to Alwan et al. (1997),
the space formed below the tongue when the front of the tongue is lifted up may also contribute to the front
8
Anyway, back articulations - as it has been presented for French by Ladefoged & Maddieson
(1996:233) - present the effect of a constriction gesture in the palatal or pharyngeal region,
leading to a movement of F2 towards around 1000 Hz.12
cavity. They suggest therefore that the sublingual space may also act as a separate resonator, giving rise to
resonance/antiresonance pair.
12
Different values are described for female voices, raising F2 towards 1500 Hz. More details on acoustic
properties of r-sounds have been recently summarized in Catford (2001:173ff.) who reproduced formant
frequencies measured by various authors. Tongue tip r's are described for a female speaker (cp. Uldall 1958) as
having F1 about 550 Hz, F2 1500 Hz and F3 1900 Hz, against molar r's with F1 500 Hz, F2 1600 Hz and F3
1800 Hz (women's syllabic "bunched" r's presented by Hagiwara 1995:61 are in the same range, whereas men's
realizations are downscaled to lower typical values). The table in Catford (2001:176) shows a consistent
variation for molar r's formant patterns, in a number of studies for various languages, spanning into a range
where apical r-sounds would be included too. In the survey of formant values reported in Espy-Wilson et al.
(1997) for American English /r/, they found typical ranges across speakers of 250-550 Hz for F1, 900-1500 Hz
for F2, 1300-1950 Hz for F3, whereas Delattre & Freeman (1968) reported higher F3 ranges for uvular
articulations (about 2400 Hz).
9
2. French r
The literature on French r-sounds is impressive, but in several linguistic general surveys one
can found persistent considerations reflecting common places. The main r-variability
described for French refers to stylistic and sometimes geographical and/or historical
differences related to apical vs. uvular alternations - whereas very few details are available
about the distributional variability in present-day common French.13
Not surprisingly, phonological descriptions emphasize only apical vs. uvular distinctions, but
the traditional view, still dominating in the premises of large diffusion manuals, is sometimes
slightly improved.14
13
As usually admitted for phonological approaches, phonetic descriptions are simplified. Martinet
(1960*1991:62) underlines that at least a difference should be taken into account: "il serait bon de signaler
l'existence pour le phonème /r/ de deux variantes principales, l'une dite « grasseyé » propre aux milieux urbains,
et l'autre dite « roulée » encore très répandue dans les campagnes, bien qu'en régression". The variability in the
production of back r-sounds appeared only recently. An example is in Wioland (1991:23) where it is observed
that "la consonne [R] est l'articulation consonantique pour laquelle la langue est la plus en arrière dans la cavité
buccale. La luette peut toucher [...] ou ne pas toucher [...] le dos de la langue". Finer descriptions are only in
specific phonetic works like in Canepari (1979, 1983), Shibles (1994), Ladefoged & Maddieson (1996), and,
more recently, in Fougeron & Smith (1999) (where a simplified but interesting sketch is given). Lauret (2000)
and Demolin (2002) give examples of an effective usefulness of the distributional r-variability in transcriptions
whereas phonetic evidences are now available from studies like the one in Autesserre & Chafcouloff (1998).
14
At this purpose Duchet (1981:9) writes "Les deux "sons [r]" qui coexistent en français, à savoir le [ ]
"grasseyé", d'articulation uvulaire, et le [r] "roulé", d'articulation alvéolaire, sont phonétiquement très différents.
Mais on sent bien que ces deux sons constituent une seule et même réalité assurant la même fonction, et si
l'oreille les distingue bien, c'est que chacune de ces variantes a une connotation sociologique ou historique (le [r]
roulé étant le fait de gens plus âgés ou de sujets d'origine rurale), sans que pour autant on cesse d'identifier les
deux variantes comme une même unité du point de vue du fonctionnement de la langue française". This
description is completed by a more detailed picture given at p. 60: "[Le] /r/ français [...] n'a pas moins de trois
variantes, lesquelles sont phonétiquement très différentes (et en particulier fort éloignées sur le plan
articulatoire) :
[r] : vibrante apico-alvéolaire, dit r roulé
[] : vibrante dorso-uvulaire (r roulé du fond de la gorge)
[ ] : fricative dorso-uvulaire sonore (r grasseyé).
La première de ces variantes est caractéristique d'une prononciation rurale de Bourgogne ou de Berry, par
exemple, et elle est donc conditionnée par des facteurs géographiques et sociaux. La deuxième était naguère de
règle dans la chanson (la chanteuse Mireille Mathieu la pratiquait naguère encore), et elle est donc conditionnée
par un registre particulier de phonation. La troisième est celle qui se pratique le plus couramment dans tous les
contextes en français standard contemporain" (Duchet 1981:60).
10
Traditional phonetic descriptions, like the one given by Thomas et al. (1976:50), associate to
Burgundy French an apical pronunciation "à plusieurs battements à l'initiale de [r]", a uvular
flap in some Parisian models ("certains parisiens prononceront une uvulaire à un seul
battement []") where the normal realization is said to be a voiced uvular fricative ("la
réalisation normale en français cultivé de l'Ile-de-France sera une spirante uvulaire sonore
[]"). Thus, since a few decades, we are aware that nowadays the use of [r] is not only
regarded as stigmatized (Tranel 1987:140), but even perceived as very uncommon.
Furthermore, in a small selection of studies, one can read that even [R] is definitely
uncommon (Tranel 1987:141, Shibles 1994:65), though it is still frequently used in some
idiolects (see Duchet 1981).
Various authors point out the general variability but most of the time they choose, for didactic
reasons and/or for typographic constraints, not to take it into account. Among a list of various
titles, a witness of this tendency is in Derivery (1997:24-25):
Les dorso-uvulaires [...] sont réalisées par contact ou rapprochement de l'extrémité du
voile du palais (ou luette) contre la partie postérieure du dos de la langue, avec un léger
battement [R] ou sans battement [ ]. Dans la transcription phonétique, le signe [R] sera
utilisé même s'il ne représente pas l'usage le plus fréquent.15
Similar details are given in his phonetic description by B. Malmberg (1954*1994:53): "[...] en France comme
dans quelques autres pays européens le r apical a été remplacé à l'époque moderne par une prononciation
uvulaire du phonème. Ce n'est plus la pointe de la langue, mais la luette, qui vibre et qui forme les contacts
répétés avec la partie postérieure du dos de la langue (fig. 41). Cet r postérieur vibrant (appelé parfois en français
r "grasseyé") est assez répandu aussi bien en France qu'ailleurs". Another variant is described farther in the same
text, in relation with a process of r-weakening: "une transformation analogue se constate aussi pour le r
postérieur qui est très souvent fricatif. La partie postérieure du dos de la langue forme un rétrécissement du
passage de l'air contre le palais mou ou la luette mais sans qu'il se produise de vibrations. C'est souvent le cas du
r parisien (appelé aussi r "dorsal")" (Malmberg 1954, *1994:54).
15
A similar choice is in Léon & Léon (1997). Shibles (1994:65) points out that the symbol suggested by the
Principles of the IPA for French was already [] in 1949. He observes that "[t]exts and dictionaries typically give
the generic /r/ as [r], in place of the actual French value . This is a category mistake as it treats a generic
phonemic symbol as if it were a phonetic symbol". The 1949 edition of The Principles of the IPA doesn't
mention explicitly the need for narrower r-transcriptions for French, whose specimen at p. 21, transcribing a
form of Parisian, shows only the symbol /r/. The foreword note states that r =  (rather than , still extensively
used) but no specific phonetic detail is added in the text, exactly as the phonemic principle suggests to do ([r] is
written in English dictionaries instead of the more common [] and appropriate allophones: [] after alveolar
stop, [] after dental fricative, [] after voiceless stop in stressed syllable, and so on). What is strange is that,
11
For different purposes, other authors have recourse to the opposite choice. For instance,
referring to Carduner & Hagiwara (1982), B. Lauret (2000:90) transcribes three different kind
of r in "liberté, égalité, fraternité" [libte ealite fatnite].
Similar examples for Belgian French (e.g. "elle prendra le train" transcribed as
[lpdalœt]) can be found in Demolin (2001:71) who underlines as nowadays various
authors are, at last, distinguishing between [], [] and [] as realizations of French /R/
(Demolin 2001:64).16
Another recent account giving simple but clear explanations on distributional r-variety in
French is now available in Fougeron & Smith (1999) who also present an example of
coherent usage in transcriptions (p.81):
French has one rhotic, whose production varies considerably among speakers and
phonetic contexts. The speaker presented here [a young Parisian female speaker (p. 78)]
uses a uvular fricative [] that is sometimes reduced to an approximant [], particularly
in final position; it may also be devoiced (for example see the transcribed text), and can
be reduced to zero in some word-final positions (Fougeron & Smith 1999:80).
despite the attempt to get finer notation criteria for other languages and for different purposes, this criterium is
still applied in widespread present-day French transcriptions.
16
Following Canepari (1979:248, see footnote 1), I believe that, as it is discussed below, in present days the
voiced uvular fricative // is the more common realization (with unvoiced and approximant allophones). Tranel
(1987:24) prefers the approximant, as well as Mioni (1986:43) who describes as approximant "la realizzazione
più frequente di /r/ in francese e tedesco soprattutto in posizione postnucleare" (and also "una delle più frequenti
realizzazioni « non normali » di /r/ in italiano"): the transcription  is given for French rare, with the onset r
transcribed as a fricative and the coda realization as an approximant (Mioni 1986:45). Good arguments in favour
of voiced fricative/approximants as intervocalic more common sounds come from well-known misundertandings
between hispanophones and francophones often recalled in the literature (such as the pronunciation of Spanish
perro by Haïtian French-speaking people giving place to phonetic forms more similar to realizations of Spanish
pego [peo], cp. Calvet 1987; a similar description is also in Delattre 1944*1966:207).
12
2.1. Historical changes
Diachronic data on the historical passage from an alleged Latin and Romance apical r to the
present-day French uvular r have been recently discussed in Demolin (2001).17
Fundamental contributions have been given by Straka (1979) and by Bonnard (1982) who
collected elements to show that the back r is a creation of the high society and dates back to
the second half of the XVIIth century (according to the former author) or from XVth to
XVIIth c. (following the latter). The change took place as a consequence of the raising of the
tongue dorsum towards the velum (with or without flapping of the uvula).18
In reality, the lingual r has always been described as endangered in popular French where it
frequently underwent phonetic changes towards [z] or [l] or towards its deletion.
In modern French, there are a few traces of lexicalized r spirantization: those of chaise <
chaire < lat. CATEDRA and besicles f. pl. 'lunettes' < béryl <lat. BERYLLIU(M) ('beryl', a mineral
used for lens production) well described since Bloch (1927). We found a detailed description
of the phenomenon in Meunier (1928:60) who deals with s-rhotacism in Nivernais (where
urage was attested instead of 'usage', [r] instead of 'chose', and where it was a systematic
phenomenon in liaison contexts, les hommes > [lar um]).19
17
Several descriptions report details on the gradual shift taking place for French. In 1910, Roudet considers that
"En français et en allemand, r vélaire tend, de plus en plus, à remplacer r lingual qui a été dans les deux langues
le son primitif" (Roudet 1910:136). A critical review of such a kind of claims is now in Catford (2001). Very
poor phonetic details are given in general for this phenomenon in historical phonetics surveys (e.g. Bourciez &
Bourciez 1982, where in addition, the conventional use of [r] and [] is uselessly inverted).
18
This kind of explanation is in Delattre (1944): "The history of modern French r must have taken place in two
phases. In the first phase, uvular trilled r coexisted with a gradually replaced apical trilled r [...] In the second
phase, the uvular trilled r became a fricative r without changing its point of articulation. The vibrating of the
uvula failed to get a start, and the only sound produced was a voiced friction caused by the running of the breath
through the constriction" (Delattre 1944*1966:207).
19
The s-rhotacism involving alternations in Latin declination (like GENUS > *GENEZIS > GENERIS), is extensively
described by Pisani (1964, 1970). It is often recalled in surveys on various languages: see for instance Meunier
(1928:58), Delattre (1944*1966:208), Carton (1974:164) and Catford (2001:179, with examples such as mus muris etc.). According to Meunier (1928), "Le changement de s intervocalique en r remonte donc aussi dans le
Nivernais à près de quatre ou cinq siècles. On sait d'ailleurs que ce changement a eu une grande amplitude au
XVIe et XVIIe siècles dans la région parisienne [ref. to Bloch 1927]. Marot, dans une épître connue, se moque
des courtisans qui affectent de dire Pazis pour Paris, et Mazie pour Marie. De même, au XVIe siècle aussi, les
dames de Paris prononçaient r comme z et disaient : «Mon mazi est à Pazis», pour «Mon mari est à Paris». On
s'est tant moqué d'elles, que les Parisiennes, pour se corriger, au lieu d'élever un peu le bout de la langue pour
13
We are sure that the "rolled" r existed and was widespread all over the country until the
XVIIIth century (and, according to some sources, it would be still living in regional varieties
of Eastern Pyreneans and in Burgundy, in small parts of Midi, and of course in Québec and in
Wallonie; see below). It is still considered normal at the beginning of the XXth c.
We know that Rousselot's r was lingual; as he gives notice in his Principes "les variétés d'r
sont nombreuses, je ne prendrais pour exemple que celle qui m'est familière, l'r linguale"
(Rousselot 1901:427).20 Anyway, according to Demolin (2001:63), Rousselot would have
commented, in previous works, that its r was not Parisian (grasseyé), neither Lorrain
(fricative as in German ch), nor a Limousin (strongly rolled), nor a Blaisois (close to z), nor
the southern one (apical r) (Rousselot 1891).21
prononcer r, ont élevé le dos de la langue, et l'r est devenu grasseyé : «Mon mari est à Paris» (Meunier 1928:60).
The reverse passage z > r is also attested in the XVIth c. (see Carton 1974:164 who also quotes examples from
Cl. Marot, as doux ami > doux rami) and seems to be at the origin of the argot expression "les Ritals" used to
refer to Italian people (deriving from a snobbish pronunciation of the apocopated form les Itals > le[z] Itals,
followed by a false reconstruction of an r from z probably influenced by a r-shibboleth). Other phenomena are
attested in the XVIth-XVIIth centuries: r > l in borrowings and in recent words, leading to a weakening of the
apical articulation (in the XVII c. people speaking in a snobbish way pronounce celtes 'certes', et pal 'par').
According to Demolin (2001:64), in 1689, Andry de Boisregard suggested to pronounce r without scratching the
ears as people from Paris were doing. Another famous testimony of a different r-pronunciation is in Furetière
(1670) "«Grasseyer». Parler gras et ne pouvoir pas bien prononcer certaines lettres et entre autres l'r" (cp. Millet
1926:20). In the "Bourgeois Gentilhomme", Molière writes that the "maître de philosophie" taught to Monsieur
Jourdain, the apical articulation of r: "Et l'R [se prononce] en portant le bout de la langue jusqu'en haut du palais,
de sorte qu'étant frôlée par l'air qui sort avec force, elle lui cède et revient toujours au même endroit, faisant une
marque de tremblement: R, RA" [acte II, sc. 6]. Last, I found in Roudet (1910) and Nyrop (1914) a reference to
final (and not only final) r-deletion, attested since the XVIth-XVIIth centuries.
20
The apical r is traditional for M. de Saint-Genès whereas a back r is considered the normal Parisian realization
at the beginnings of the XXth century: "L'r traditionnelle est prononcée du bout de la langue. Les Parisiens ont
une r qu'ils prononcent, la pointe de la langue appuyée derrière les dents inférieures, en faisant vibrer les parties
molles de l'arrière-bouche" (Saint-Genès 1912:303).
21
In contrast with these descriptions, Carton (1974) described as pharyngeal the sound heard in Lorrainese and
for the Lille variety of French (see below). Such a kind of articulations is surely attested in Parisian as well (also
cp. Germi 1968:6 quoting the r-variants of Parisian examined in Durand M. 1960 on the base of a 14-feature list
described by Borel-Maisonny). Another reference is in Rousselot & Laclotte (1903:56) describing r-sounds in
their numerous variants (see Demolin 2001).
14
Uvular sounds are not taken into account among the back r-sounds defined by Roudet
(1910:136): (1) a "r vélaire vibré [r  ]", (2) a more frequent "r vélaire non vibré [...] parisien",
(3) a pharyngeal r occurring in foreign languages.22
Velar articulations are analyzed by M. Durand (1929:253) in her palatographic study on back
Parisian phonemes, where she mainly observes for /r/ weakened fricative articulations.
On the contrary, uvular sounds are included in Grammont's surveys and are mentioned by
Nyrop (1914:42):
ce phonème est dû à la vibration rapide de la luette : c'est pourquoi on l'appelle r
uvulaire, par opposition au [r] linguo-dental ou apical [...]. On sait que le premier a
triomphé du second dans le français de Paris et des villes de la France septentrionale.23
Grammont (1933*1971:72-77) distinguishes several varieties of back rhotics i.e. a uvular
(grasseyé, with vibrations), a dorsal (Parisian, without vibrations), a pharyngeal (retracted
variant, also see Grammont 1941), but, once again, the distinction only accounts for dialectal
variations.24
22
Following Roudet (1910): "En français, r vélaire a d'abord été considéré comme un défaut de prononciation
appelé grasseyement. Puis, lorsque cette manière de prononcer r s'est généralisée, on n'a plus su quel était le sens
exact du mot grasseyement et on l'a appliqué soit à r pharyngal [...], soit à toute autre prononciation anormale de
la consonne. r vélaire non vibré est une consonne qui disparaît facilement. Il suffit que la pression de l'air expiré
diminue pour qu'elle se fonde avec la voyelle qui précède ou celle qui suit [...]. C'est ainsi qu'on peut expliquer le
langage des élégants du Directoire : «Ma pa'ole d'honneu' c'est inc'oyable»" (Roudet 1910:136). Several other
French authors until our days quote the r (non-)pronunciation of the "Incroyables" as a renowned case of
"amuïssement du r" (e.g. Carton 1974:164). On the other side, pharyngeal sounds were excluded from common
French pronunciation: "Lorsque la racine de la langue se rapproche de la parois postérieure du pharynx, il se
produit un son fricatif analogue à r. Il se peut que dans ce cas les piliers postérieurs du voile du palais entrent en
vibration. D'après M. Jespersen c'est ainsi que s'articule r en danois. On a vu plus haut [...] que le rhaïn arabe est
probablement produit d'une façon analogue" (Roudet 1910:136).
23
Other considerations in this sense are available in Millet (1926): "Il y a 40 ans, le grasseyement alors peu
répandu prêtait encore à rire [...] il n'en est plus ainsi parmi la génération des jeunes, où l' rׂ grasseyée est
devenue commune ; on ne songe plus même à la considérer, à l'exemple des anciens, comme un défaut ou un
accident" (Millet 1926:19).
24
It is strange that only a few references support a pharyngeal articulation for the common French r-sound (see
Delattre 1969): the alternations only concern dorsal vs. uvular (that is, fricative vs. trill, in real terms). In the IPA
symbols description opening his book, Carton (1974:16) defines the usage of [] for a "r dorsal" and a [ ] for an
untrilled r ("r dévibré"): the possibility of a trilled dorsal is introduced (defining a brand new definition as
regards to Grammont's dorsal without vibrations). Yet in the IPA Principles things seem to be clear since 1905,
where the general northern French r is a uvular  (defined as "r grasseyé" at p. 6, against Duchet's definition in
15
Since then, besides the terminological confusion, a dichotomic view has been dominating. As
introduced above, the main feature observed, all over the century, until our days, was the
substitution of a front r with a back r even though it was already clear, since the beginnings of
the XXth century, that the terms involved in the change were more than two.25
Nonetheless, throughout the century, from different reasons, several authors keep
distinguishing only trills, a uvular [R] and a lingual [r] (e.g. Passy 1932:97-98, Falc'hun
1972:112, and many others).
Describing his 4 r-variants (rolled, velar fricative, grasseyé, Parisian), Bonnard (1982) defines
grasseyé the uvular [R] which is assigned to regional or popular usages whereas a weakened
uvular is said to be Parisian and suggested as a reference for official French pronunciation. As
it is highlighted also by Demolin (2001:64) a terminological ambiguity marks the literature on
the topic, where Parisian r is sometimes called grasseyé whereas other times it is described as
a weakened sound, just opposed to it.26
note 14),  being the symbol for Danish r (and Arabic gh, p. 8) whereas the Parisian r is said to be intermediate
between  and  (ibid.). In the "Note sur la transcription", it appears that "r peut valoir r ou  [...] [ou]  à Paris"
(p. 11). More precisely, in the 1912 Principles (PIPA 1912:11-12),  is said to be often heard as a variety of
Parisian r and in the specimen at p. 22 for (Northern) French it is explicitely stated that "r may be r or  (or even
 in Parisian speech)". Similar considerations apply to subsequent editions through the century (also see Jones &
Camilli 1933, PIPA 1949) where the  sound is acknowledged as the more common pronunciation.
25
Another example is offered by Millet (1926) "Aujourd'hui l'oreille indigène la moins prévenue réunit sous le
nom de r trois sonorités distinctes, inégalement représentées dans le parler local : celle de r roulée, celle dite de
rׂ grasseyée, et une autre, propre à un groupe de familles voisines, que nous appellerons r grasseyée spirante :
rׂ" (Millet 1926:17). And then, defining the "rׂ grasseyée" as Parisian, he distinguishes a "r grasseyée spirante
(rׂ)" as a sound which is "ni roulé ni franchement grasseyé. La constriction est opéré par le dos de la langue
devenue incapable de vibrer : c'est pourquoi nous la considérons comme une spirante. Devant une dentale,
comme dans le mot « partir », l'r nous a paru quelquefois relâchée au point d'imiter la prononciation d'un Anglais
du Sud" (Millet 1926: 20-21). As above mentioned, well-established dominant fricative realizations are also
described in Delattre (1944) and Thomas et al. (1976) (see above).
26
For Léon & Léon (1997:23) the r "grasseyé" is a fricative dorso-vélaire as opposed to the "dorso-uvulaire, non
vibré [ ] (le dos postérieur de la langue contre la luette qui ne vibre pas[)]" which is said to be "le R qui se
répand le plus actuellement dans toute la francophonie (dans nos transcriptions, nous noterons le R dorso-vélaire
par R majuscule, pour plus de simplicité)". Far in the text they distinguish a "uvulaire vibré [...] produit par les
vibrations de la luette" (ibid.:24) and two trills (vibrantes) "le [r] apical ou le R uvulaire roulé (celui d'Edith Piaf
dans ses chansons)" (Léon & Léon 1997:72). All things considered, they take into account therefore three
different types of back r, leaving the first one (grasseyé), without any example, just determined by a pretended
16
The misuse of the term grasseyé is the starting point for this puzzle. In 1914 Nyrop already
pointed out that this term presented some arbitrary usage:
Grasseyer : Terme, généralement méprisant, qu'on applique à la prononciation d'autrui ;
les gens qui articulent l'r d'une certaine façon, se servent de ce mot pour caractériser
toute manière différente d'articuler la dite consonne (Nyrop 1914:48).27
difference in its articulation place. A lack of precision in the r-definition is particularly disappointing in works
dealing with foreign languages learning of pronunciation. Renard (1971:100) acknowledges the specific
difficulties of this sound class ("la correction de [R] [...] s'avérera de toute manière relativement délicate", p.
108) and perhaps refers to a distributional r-variation, but he never defines it explicitly. A velar articulation
(transcribed []) is also acknowledged by Callamand (1982:100, 161) as the common French realization as
opposed to a uvular articulation [] common in the South (sic, p. 148). She attempts to make finer distinctions
but confuses the terms. Dealing with the [R] acquisition by foreign learners, she points out how "Plusieurs
réalisation de l'entité [R] sont attestées sur le territoire français. Le plus répandu est sans doute le son [R], qui se
caractérise pour une articulation vélaire [...]. Il s'agit du [R] dit «parisien»". A terminological contradiction then
appears: "Un son proche de celui-ci, mais qui paraît être vélarisé avec plus de force, est localisé dans le Sud et le
Sud-Est de la France. Ce [] est uvulaire ; il est connu sous l'appellation de "r" grasseyé, et doit son articulation à
la rencontre intermittente du dos de la langue avec la luette. Quelquefois, les locuteurs étrangers qui s'efforcent
d'atteindre les caractéristiques du [R] vélaire tombent sur ce [] ; moindre mal, puisque des millions de Français
l'utilisent !" (Callamand 1982:161). Similar contradictions are particularly deceiving and really encourage a
forthcoming radical renewal of r-description in French linguistic literature.
27
"On entend souvent parler d'"r grasseyé", de "grasseyement", etc., et ces termes apparaissent non seulement
dans la conversation et dans la littérature, mais aussi dans les savants traités de phonétique. [...] En présence des
définitions variées et contradictoires de la prononciation "grasseyante", P. Passy a fait une petite tentative de
classement historique : d'après lui l'expression "r grasseyée" aurait commencé par signifier : r prononcée avec la
gorge, mais actuellement elle s'appliquerait à une certaine variété parisienne de l'r qu'il note [ ] (fricative
vélaire). Cette théorie est précise, mais elle l'est beaucoup trop : nous voyons les Français d'aujourd'hui appliquer
à tort et à travers les mots "grasseyer" et "grasseyement" à toutes les formes possibles de l'r. [...] le dictionnaire
français-allemand de Sachs [...] accumule trois interprétations différentes du mot "grasseyer" : "1o das R
schnarren beim sprechen (=parler gras); 2o auch l statt r sprechen; 3o das R nicht aussprechen (paole statt
parole)." Comme on le voit, il y en a pour tous les goûts, et le même terme désigne tour à tour le [R] uvulaire
moderne (schnarren), l'l de la précieuse Alphée, et enfin la suppression totale de l'r par les "incoyables" du
Directoire [...]. D'après Littré, « ceux qui grasseyent ont de la peine à prononcer la lettre r, et ils lui substituent
souvent la lettre l ». Les exemples cités ne prouvent d'ailleurs absolument rien, ni pour ni contre cette définition.
Quant au "véritable grasseyement", il consiste, - toujours d'après Littré, - à « faire entendre une sorte de
roulement guttural. Le grasseyement affecté consiste à ne prononcer nullement la lettre r »." (Nyrop 1914:4648). The same references are in Millet (1926) who objects "Ce n'est assurément pas son caractère guttural qui
mérite à l'rׂ son appellation, sans quoi l'r des Alsaciens, des Lorrains, devrait passer pour « grasseyée » ; elle
n'est que gutturale et aspirée de l'avis général" (Millet 1926:20).
17
I believe the Millet's and Grammont's dialectal definitions should be taken as a departure for a
better definition of the link between popular usage and scientific terms.
Anyway, whatever the historical oscillations, it is very interesting to see how, in general parts
of southern France, two r-phonemes survived together and are described as phonological in
Provençal varieties. In Haute-Loire [r] and [] appear in fact in minimal pairs such as [ura]
'heure' and [ua] 'marmite' (see Coustenoble quoted in Jones 1967, with the examples [sro]
'soir' and [so] 'scie', [ari] 'guérir' and [ai] 'chêne').
That is a point surely not to be missed in the history of the apical-to-uvular passage. I suggest
below to consider revisiting this contrast in terms of /r/ vs. /rr/.28
28
In the specimen for Provençal (dialect of Arles) of the 1949 Principles of the IPA (p. 23) one may find that "r
and  are distinctive" and are both used in the text (r always between vowels,  everywhere - including the
intervocalic position where the two sounds are supposed to contrast). Note that this is not valid everywhere in
the Midi. In Vinay's (1936:62, 1937:48) specimina for Provençal dialects of the Landes and of the Béarn a
unique /r/ is mentioned (even if different dialectal realizations are described: in the Landes, a uvular , in the
Béarn, a strongly rolled lingual r).
18
2.2. Socio-geographical variation in common French
Another common place in the literature is the attempt to explain the front-to-back r-sound
change in terms of articulatory bases and coarticulation. For instance, the French r shift is
interpreted by Delattre (1944*1966:207) as the consequence of a language-dependent
articulatory constraint
[d]uring the French articulation of the consonant, the tongue always tries to take the
position of the following vowel, thereby eliminating diphthongization. This vocalic
anticipation requires as much freedom of the tongue as possible. With the apical r
keeping the tip of the tongue occupied, the vowel position could not be anticipated and
a transitory movement from r to the following vowel was inevitable. The change from
apical to uvular r was the first step to liberate the tongue [...] allowing it to articulate
the r while holding in advance the position of the following vowel (Delattre 1944:564).
Carton (1974) seems to be going in the same direction when he writes
Pourquoi [r] apical (roulé) est-il devenu R dorsal [...]? Sans doute ce passage a-t-il été
favorisé par la tendance à l'anticipation vocalique. La place de /r/ dans le système
phonologique lui permet une grande latitude de réalisation (Carton 1974:164).
Despite his initial charming by this kind of explanation, Carton (1974) concludes following
the social explanation which in my opinion is the better reason to investigate if we want to
account for the same shift happening in linguistic communities without " vocalic
anticipation": "On peut penser que le [R] uvulaire est le modèle que les diverses régions
imitent par souci de prestige culturel et social" (ibid.).
The first "naive" attempt I found in this direction is in Passy (1937:14) who takes care in
avoiding the core problem ("quant à , c'est un cas douteux qu'il vaut mieux laisser de côté")
but gives us anyway a hint:
"r s'apprend généralement tard. Le plus souvent on commence par l'omettre tout à fait,
puis on le remplace par l ou  ; mais dans ce dernier cas je crois que l'enfant n'imite pas
ses parents, mais fait un emprunt au parler des personnes qui emploient " (Passy
1937:14).
Besides the importance of the social factor (it's useless to draw attention to Labovian
approaches), the change could be supported by evidences in terms of sound patterns
19
acquisition and contrasts' reinterpretations, such the one discussed in Demolin (2001:72)
referring to perceptual factors associated with an undershoot of the apex.29
It would be extremely difficult to sketch a complete synchronic snapshot of the geographical
r-variability in France and French-speaking countries.30
Surveys are encouraged in micro-areas (as Demolin 2001 is doing for Belgium, Sankoff et al.
2001 for Montreal French and hopefully studies like the one in Durand J. et al. and other
forthcoming researches will do for France) without disregarding social factors and language
acquisition dynamics.
In fact, a geographical variation in the pronunciation of r-sounds is still evident even though
nowadays it seems dominated by socially-conditioned idiosyncratic preferences: back rsounds are clearly widespread everywhere and present an interesting differentiation in
subclasses needing to be reviewed.
Carton et al. (1983) discuss r-variation like any other geolinguistically relevant phenomenon
in French pronunciation. Notwithstanding, their description is lesser detailed than for other
sounds: four sounds are presented in the foreword as r-realizations ([r], [], [ ], [x]), but one
29
Besides the accounts above surveyed, various other tentatives have been made to explain the frequent r > R
shift from an articulatory viewpoint. Widdison (1997:188) summarized some concerns, commenting that
acoustics alone cannot explain the nature of the relationship between the various forms of /r/ and considering
physiological and aerodynamics factors as equally important. As introduced above, the weakening of the apical
tension is often considered as the starting point for the shift: a lighter contact of the tongue-tip against the
alveolar ridge can be obtained through a lingual cupping gesture (Laver 1994:219) or a lowering of the dorsum
jointly with a tongue retraction bringing the postdorsum closer to the uvula (and hence increasing the potential
for a uvular trill). A pharyngeal narrowing is also described as taking place and creating a threshold condition for
the high airflow to trigger uvular excitement. Widdison's (1997:191) arguments (the physiological
predisposition, the pretended acoustic equivalence, and the common place of the relatively simple /R/) are
probably involved. They could explain by themselves this shift, but I believe they are secondary factors taking
part to a more complex shift which is instead strongly conditioned by language acquisition and social dynamics.
Possible mechanisms explaining the back-r spread in some varieties spoken in the Netherlands are discussed in
van Bezooijen (2002, this volume). Some of them could be at the origin of the "pioneering" French shift (at least
as reinforcing mechanisms) as well as to more recent evolutions which interested German, Portuguese, Swedish
and so on. Reminding a selection of references describing as the r > R shift was vigorously in progress in 1971
in Québec French, Sankoff et al. (2001) and Sankoff & Blondeau (in this volume) show an interesting fact: the
variability in social change in progress is paralleled by intra-individual variability in later language acquisition
(also see Passy 1937 quoted above).
30
Among various interesting sociolinguistic descriptions accounting for Québec r-situation in different periods,
see Vinay (1950), Santerre (1979), Tousignant (1987) and Sankoff et al. (2001).
20
of them ([ ]) is never used in the descriptions of spoken French in the different regions,
whereas other symbols are employed (without any explicit mention, see below). They give
several examples of r-deletion for various regions and extensively use only the following
variants (their transcriptions, passim):
[] as in [tuu] 'toujours', mentioned for Francoprovençal (but similar examples are
commonly used in the book for other varieties);
[r] as in [rd] 'raide' (Touraine), also used for various other regions;
[h] as in [h] 'raie' (Normandie);
[] as in [alo] 'alors' (Lorraine);
[] as in [fini] 'finir' (Picardie);
[x] as in [bux] 'bourg' (Bretagne).31
In such a description, "strong" (unvoiced and fricative) r-varieties would be features of a
peripheric northern areas, against the common [] and the "traditional" [r] here mentioned for
a central area. That does partially collimate with Millet's (1926) assumptions on Haut-Berry
and Lorraine r-sounds, but the author points out that:
Ce n'est assurément pas son caractère guttural qui mérite à l'rׂ [Parisian grasseyé] son
appellation, sans quoi l'r des Alsaciens, des Lorrains, devrait passer pour
« grasseyée » ; elle n'est que gutturale et aspirée de l'avis général (Millet 1926:20).
Linguists, and common people in general, are aware of a certain r-variability and, following
their impression, are inclined to acknowledge different ways of pronouncing these sounds.
The main difficulty stays in finding a way to describe them coherently in a specifically
designed framework.32
31
Corse is a separate domain where completely different phenomena involving r-outcomes are described: (1) tr
> t as in [dekalit] 'décalitre' and (2) l > r as in [mur] 'moulin' (see below the short description of these
phenomena I give for southern and northern Italy respectively). The l-rhotacism (followed by a passage > t) in
guascon dialects is well-known (with reference to Rohlfs' studies on Guascon, Tagliavini 1982*1949:423 quotes
the examples *BELLA > bero, *BELLU > bet, and so on).
32
Considering the introduction in §2., a simplified representation of the common tendency can be summarized.
In a linguistic communities like the French-speaking one, r-pronunciation is easily avalaible as a shibboleth. At
the beginning, people (and linguists) observed just a difference between a "normal r" and a "back, velar r"
popularly called grasseyé. Subsequently, they needed to tell apart a Parisian r from the grasseyé, and uvular
articulations were opportunely described. In the last decades, a way for distinguishing a brand new r-usage
21
In this sense, French r pronunciation shows very interesting dynamics often recalled in recent
linguistic analyses (e.g. Binisti & Gasquet-Cyrus 2001, forthcoming). Common r-sounds,
above all in the North but also in southern towns and in young groups, include pharyngeal rsounds (especially when final) almost becoming a vowel sound (like [a]) similar to the one
described by Carton (1974) for central and northern varieties of French.33 On the contrary,
very well marked articulations seems to characterize banlieue young speakers' productions
(Billiez et al. 2002, see below).
By the way, as introduced above, the scope of the geographical variation usually discussed
should be enlarged including distributional properties of r-sounds sometimes a little neglected
in current researches (landmarks are offered by the pioneering studies on Quebec French, cp.
Santerre 1979, Tousignant 1983).
After this insight in the literature of the beginning of the past century, two examples are
highlighted in this paper: (a) the contrast /r/ vs. /rr/ (which has been disregarded in last
decades), (b) a distinction between different allophones of the uvular trill or of the back
fricative/approximant.
commonly heard and easily identified has been to call it pharyngeal. This seems happening nowadays without a
real proof of its articulatory features but simply on the bases of subjective suspicions or by matching it within a
partial representation as the one available from IPA charts, originally conceived for other purposes, but
distributional details are necessary for a better geographical account.
33
The term pharyngeal could be considered at least not misused when referred to [a]-like r-sounds. Carton
(1974:27) defines a "[R] pharyngal lillois ([fait] pour 'frite') ou du Massif Central" and a "[R] pharyngal lorrain
«accompagné de battements uvulaires»" but pharyngeal articulations are often mentioned by Delattre (1962,
1969, 1971) even for common French.
22
2.3. Double r
Lenght contrast are synchronically not admitted for the consonant system of French. Anyway,
a contrast between r and r double which is suggested by the writing system (like other purely
orthographic single-double contrasts) was claimed in linguistic literature of the first half of
the XIXth c., often dealing with examples like: guère vs. guerre, taire vs. terre, mourais
(imperf.) vs. mourrais (cond.).
An experimental phonetic analysis of these sounds is set apart by Rousselot (1913:52-59) who
already raised the problem in his Principes, without mentioning any detail, when he reported
results from the analysis of the word irresponsable "prononcé la Ire fois (A) normalement, la
2e fois (B) avec une seule r" (Rousselot 1897-1901:352). In a subsequent research, the same
author, after having underlined that the contrast was not maintained anymore in his own
pronunciation, concludes: "Il y a 1 ou 2 [battements] pour mon r, de 2 à 4 pour rr" (Rousselot
1913:52).34
Yet in 1959, a pretended surviving contrast pushes Fouché (1959:480) to suggest different
pronunciations (and transcriptions with a different r-symbol [Ro]) for two words spelled with
-rr- (horrible, larron) against words spelled with -r- (transcribed instead with the symbol [R]).
At p. 319, he defines "[Ro] = [R] légèrement plus long et plus fort".
A different opinion is expressed by Martinet (1971:196) who excludes the possibility to hold
this contrast.35
34
The same Rousselot was in fact aware that the constrast was not well based: "Le redoublement de l'r dans les
mots d'origine savante est un signe d'éducation latine. Les personnes du peuple simplifient cette consonne,
comme en général les consonnes doubles de la langue savante. La différence entre r et rr est très sensible à
l'oreille [...]" (Rousselot 1913:52). The past conditions are clearly explained by referring to the undangered
contrast observed by Vaugelas (1647) at its time, and quoting an older passage by De Bèze (1584) who wrote
about the r double: "« Lorsque cette lettre, dit-il, est redoublée, elle doit être émise avec force, l'une finissant la
première syllabe, et l'autre annonçant la suivante : quum geminatur, fortiter est efferenda, una quidem priorem
syllabam finiente, altera vero sequentem inchoante » (p. 37)" (Rousselot 1913:57-58). Farther in the text, I found
again "J'ai, moi même, bien étonné M. Jean Barès […][en lui disant] qu'il était bien inutile de chercher le moyen
de distinguer l'r et l'rr dans les mots ou elles se prononçaient de même. « Comment, reprit-il vivement, vous
prononcez la terre […] et se taire de la même manière ? […] »" (Rousselot 1913:58-59).
35
Introducing the topic, Martinet states that the contrast was lost far back in the past: "Thurot (De la
prononciation française II, p. 374) estime que la distinction entre -r- et -rr- est assurée jusqu'à 1660 environ"
(Martinet 1971:196). The conclusion is then in favour of a progressive reduction caused by a weak functional
contrast.
23
Later analyses, such as the experimental work by Germi (1968), refer to a "disparition de
l'opposition [r] roulé fort et [R] vélaire faible qui distinguait les deux graphies guerre et
guère" (Germi 1968:10), thus eliminating the doubt that a length contrast could be still
functioning.36
What is interesting to find in these considerations is mainly that two distinctive sounds (in
place and manner) were at least supposed to appear in the oppositions.
This seems to be in perfect analogy with the /r/ vs. /rr/ contrast in Spanish and Portuguese
who even lack geminate contrasts and yet keep /r/ distinctive from /rr/ by the recourse to
different articulations (// vs. /r/ in Spanish and // vs. // in Portuguese, see - among others respectively Quilis 1981 and Morais Barbosa 1983).37
This let imagine that also the [r] vs. [] contrast claimed for Provençal (see §2.1., [sro] 'soir'
vs. [so] 'scie', [ari] 'guérir' vs. [ai] 'chêne') derives from an original /r/ vs. /rr/
contrast.38
36
Anyway, since numerous French informants (even Parisian) assured me to be keeping a distinctive use, I
decided to carry out an informal pilot-study on a few examples of two of them (see §2.5.3).
37
For recent discussions of this kind of contrast see Inouye (1995) who proposes to have recourse - for some
languages - to a length feature. An instrumental contribution is also in Machuca-Ayuso (1997). A number of
references are made to Hammond (1998) who examined the acoustics of the Spanish rhotic system, criticizing
the traditional descriptions based on laboratory speech as not accurate for read text (where trills are said to be not
produced in a great majority of the expected positions). Measurements of segment durations and number of
pulses are proposed for different speech production modes (isolated words, read text and spontaneous
conversation) and in comparison with previous studies, concluding in favour of a well-grounded persisting
contrast.
38
As described above, in the specimen for dialect of Arles in the 1949 Principles of the IPA (p. 23) one may
observe r to  alternations such as in [faire] vs. [aibaje] which confirm the impression that  is attested
intervocalically (besides as a general outcome of an original /r/ elsewhere) in places where other Romance
languages and dialects present a /rr/ phoneme (see for instance French "garrigue" spelled with -rr- and so on). I
encourage the comparison of Provençal cases with some conservative r-contrasts in Ligurian dialects (see
Canepari 1999:376: "un tempo in dialetto c'era opposizione fonematica fra [le erre di] c āru «carro» [...] e [...] di
c ārׂu «caro»"; the rhotic in the first word being described as an alveolar tap against the uvular tap of the second
word).
24
2.4. Devoicing and final deletion
As already partially introduced, a common research topic in French phonetics and phonology
is word-final postconsonantal r-deletion which in my view should be better accounted for in
connection with a general final r-weakening.
A distinction between voiced and unvoiced r-sounds is usually disregarded in literature but is
already aknowledged by Nyrop (1914:42) who defines two allophones ("le [R] sonore et le
[R] sourd correspondant") and discusses the context in which the devoicing occurs:
L'assourdissement du [R] se produit dans les mêmes conditions que celui de l'l, de l'm.
Le [R] (ou le [r]) final après une consonnes a, comme le [l] final [...], une tendance à
devenir sourd, quelle que soit la consonne qui précède, et il disparaît aussi dans le
parler populaire ou négligé. Les Français cultivés disent dartre [daRtR], mais le peuple
prononce [daRt]. Cet amuïssement du [R] a lieu surtout devant un mot à consonne
initiale.39
A common rule of final devoicing or deletion is traditionally discussed in final CC clusters
(Carton 1974; Laks 1977; Chevrot et al. 2000) where deletion is a well-known phenomenon
([kat] for /katr/ quatre 'four' or [liv] for /livr/ livre 'book').40
Postvocalic weakening and/or devoicing are instead less investigated. Yet the phonetic
reduction is already highlighted by Classe (1934:47):
 est une fricative qui s'affaiblit de plus en plus et me semble bien être en voie de
disparition. À la finale ce n'est déjà plus qu'une sorte de voyelle accompagnée d'un
léger bruit fricatif. On peut lui prédire un sort semblable à celui de l'r anglaise.
39
With an example, Nyrop (1914) also underlines the existence of postvocalic deletion conditions "Quelle
différence y a-t-il entre le pacha d'Egypte et un tanneur?" Réponse : "Il n'y en a pas, car quand le tanneur
travaille, il est sur le tan (=[sylt])". This phenomenon is obviously related to postconsonantal final devoicing
but it should be also associated to intervocalic reduction highlighted in Millet (1926:18) "En position
intervocalique et finale, le roulement est attenué, au point de présenter une certaine douceur".
40
Chevrot et al. (2000) define this phenomenon as a perduring variation without change. Their historical sources
describe it in popular French along three centuries (also see Bourciez 1982:185 reporting a pronunciation suc(re)
'sugar' since the XVIIth c.). Grammont (1933:75-76) observes r devoicing after voiceless stops and whispered r
appearing in final positions. Laks (1977) studied the process in terms of social stratification and reported
different rates of deletion depending on the contexts and the speaker. He writes "Dans notre corpus, nous
trouvons en place du /r/ attendu (norme) quatre réalisations phonétiques possibles : [...] sonore, [...] sourde, zéro
et un bruit résiduel proche de l'aspiration" (Laks 1977:116). Anyway, more recent enquiries summarized in
Chevrot et al. (2000) report higher deletion rates than in the past.
25
As presented above, Fougeron & Smith (1999:80) also account for devoicing in final position:
French has one rhotic [...] that is [...] reduced to an approximant [], particularly in final
position; it may also be devoiced [...] and can be reduced to zero in some word-final
positions.
Barolo et al. (2001) recently analyzed more than 200 items containing postvocalic prepausal
final r (uttered by two French speakers - a male and a female from southern regions – in formal
speech). They showed as the male speaker pronounced unvoiced sounds in 81% of his r's in
this position, whereas the female speaker had only 36% of voiceless postvocalic final Rs (see
Table 1). The two speakers' realizations mainly diverge for front vowels (the male speaker
usually devoices all the r-sounds in these contexts, indipendently on vowel quality, whereas the
female speaker tends to preserve final r-voicing after front vowels [i], [y] and []).
As for r-deletion in final consonantal clusters the same authors found that after voiceless
consonants such as /p, t/, all of the 87 words analyzed were realized with a voiceless sound
tending to the deletion, whereas after voiced consonants /b, d, v/ (on 38 words), // was found
unvoiced only in the 8% of cases.41
Table 1. Postvocalic final // devoicing as observed by Barolo et al. (2001)
Male speaker
Female speaker
(from Toulouse)
(Aix-en-Provence)
/_#/
93% (54/58)
17% (9/52)
/a_#/
83% (40/48)
56% (31/55)
/_#/
83% (19/23)
52% (11/21)
/u_#/
75% (15/20)
61% (11/18)
/i_#/
73% (16/22)
12% (2/17)
/œ_#/
73% (32/44)
29% (10/34)
/y_#/
69% (11/16)
14% (2/14)
Total
81% (187/231)
36% (76/211)
/r/ -> []
Once again, outside phonological contexts, social and situational factors would condition the
application of phonetic reduction rules to various degrees often determined by individual styles
and inconscious adaptation dynamics (or choices) of the speaker.
41
Similar tendencies are described by Germi (1968:22). In her study, final r-sounds are unvoiced in about 60%
vs. 40% (most of all, after back vowels). The author observes the absence of pulses in 45% of the cases
(especially in unrounded context, thus indirectly hinting at a negative effect of lip-rounding on uvular trills).
26
A very common outcome of instrumental enquiries is that, when not followed by final vocalic
appendices, // following a vowel at the end of a tonal (or intonational) unit often presents
voiceless realizations [] in careful speech, anticipated by effects on the preceding vowels
allowing the transition to be perceptually reanalyzed as a gliding vowel (such the ones described
for High German realizations of final r; see Barry 1995, Schiller & Mooshammer 1995 and
many others). Sometimes, the fricative phase, furtherly weakened, may disappear just leaving
the vocalic phase.
Such a mechanism is not the result of a recent phenomenon since it was already commonly
spread in popular (mainly southern) French of the first half of the XXth c. as it is evident, for
instance, from the recordings available in the first Marcel Pagnol's movies where approximant
realizations and r-weakening phenomena are everywhere in the actors' speeches.
In connection with the general prosodic properties of southern French varieties allowing final
- and showing a significant occurrence of pseudo-paroxytonic 'CVCV structures (that is
trochaic feet), two different r-allophones (a final weak r, often realized as a weak vowel
sound, and an internal, intervocalic, approximant/fricative r) potentially contribute to the
maintenance of a general lexical distinction in pairs generally reduced to homophones in
common French such as mer 'sea' and mère 'mother' both phonologically corresponding to
/m/. These two words in the traditional Provençal pronunciation of French could be
respectively assigned to different phonetic (and probably also phonological for southern
French) forms: [m] and [mœ] (or [mœ] with more care). Of the two realizations, the
former agrees with the female speaker tendencies, in the Barolo et al.'s study, preserving final
r-voicing after front vowels. 42
42
This mechanism is evident in a clip of the Pagnol's movie Marius (1931) where César, a character played by
the Provençal Raimu, seems not understanding the current pronunciation of "mère" by Marius (the actor Pierre
Fresnay, who makes it collide with the pronunciation of "mer"), in the following passage:
- (Fresnay) Tu en as parlé à sa mère ?
- (Raimu) Comment ? (doucement)
- (Fresnay) Tu en as parlé à sa mère ? (plus fort)
- (Raimu) Sa mer ?! (étonnement) Qu'est-ce que c'est que ça, "sa mer" ? (pause) Ah, sa mère ! Ah !
[- (F) Have you spoken about it to her mother? - (R) What? (gently) - (F) Have you spoken about it to her
mother? (louder) - (R) Her sea?! (amazement) What's that, "her sea"? (pause) Ah, her mother! Ah!].
27
2.5. A contribution to the phonetic description of r-sounds
At present my contribution consists of an instrumental survey of r-realizations from some
standard and non-standard speakers, their exact description in terms of "narrow phonetics"
and some results of extensive acoustic measurements.
The speech make-up is structured in different sub-corpora:
C1) a sample of about 600 acoustic items related to 60 words uttered by 8 speakers of
different origins and with idiosyncratic phonetic properties;43
C2) a sample for the pilot-study on double r's based on /r/-/rr/ pairs for two female speakers;44
C3) two free recordings of spontaneous speech collected in the framework of the DGLFLF
research in Billiez et al. (2002) (about 30 min. each).45
43
The items are of the following type: garage, nourriture, partir, terrible etc. and are reiterated 3 to 5 times
resulting in a sample of quite different speech styles, for some speakers ranging from hypercontrolled to
hypoarticulated.The sample is extracted from a larger corpus of 475 general items (words, collected in citation
forms and/or in frame sentences, and more complexe expressions such as des ordres, quatre chevreaux,
fédération de chasse etc.) giving more than 1000 sound-samples collected in 1999-2000 thanks to the following
speakers (recorded in Grenoble, at the CDG – Univ. Stendhal; the name-coding uses acronyms in which the first
two characters stand for surname-name initials followed by two digits corresponding to the age of the speaker):
CT28 (male=M), a researcher coming from Besançon but with a ten years spent in the Dijon area and in the Midi
(Toulouse-Marseille-Grenoble); FD28 (M), a young engineer, visiting Grenoble, born and living in Paris; four
students: AL20 (female=F, Gap, Hautes-Alpes); AM20 (F, Valence, Drôme); AP23 (F, Yonne); LB25 (M,
Voiron, Isère, but born in Paris) + two francophone not-French control speakers LM27 (F, Montréal, Québec);
AT38 (F, Charleroi, Belgium) + a few isolated items recorded from JP32 (Aubagne, Bouches-du-Rhône). Only
samples from speakers CT28, FD28, AP23 and AL20 are discussed in this paper. CT28 is proposed here as a
reference speaker, showing a very relaxed and spontaneous style and a near-normative pronunciation. The other
French speakers frequently tend to hypoarticulate (even in careful speech), especially FD28, and are labelled as
"reducing" speakers.
44
The speakers are personnels of the Université de Grenoble: CC33 (F, Grenoble, father from Le Mans, Sarthe,
mother from the Landes) and CH42 (F, Sisteron, Provence). Potentially contrasting r-sounds are from the
following sentences: Si je courais je mourrais, Je mourrais si je courais, Si je mourais ce serait très triste, Je
courrais si je mourais, Si je courais j'arriverais plus tôt.
45
Anonymous speakers are two male and two female teen-agers of the Grenoble banlieue. Additional speech
recordings have been considered, based on the productions of four young male speakers collected by Cyril
Trimaille.
28
2.5.1. Studies in laboratory speech make-up
All the r-realizations I collected and described in distributional terms have been compared to
phonetic tendencies observed for other French-speaking areas. It is important to underline that
control speakers from Belgium and Quebec I have recorded presented excellent
correspondances with previous studies (see Tousignant 1983, Demolin 2001, and others).
For France French speakers in the samples here considered I observed quite clear uvular
articulations (see formant transition patterns in §3.3.) undergoing important coarticulatory
processes, slightly modifying the articulation place, and involving lip-protrusion gestures.
On the bases of the analysis of the corpus in C1), I summarized the distributional tendencies
illustrated in Table 2 which mainly agree with broad descriptions in other studies (Fougeron
& Smith 1999) and - mutata mutandis - with Belgian French phonetic patterns described by
Demolin (2001).
Table 2. Distributional tendencies for phonetic realizations of // which are dominant in mainstream French.
Phonetic contexts
#()RV-
Subcases*
f(V)?
-VR#
-V()RV-:
f(V)?
f(V)?
Outcomes
→ [] / []
→ [] / []
→ [] / []
→ []
-VR.CV- / -VRC#:
(-V.)CRV-:
-VCR#:
-VRC{V, #}-
→ []
-VRC{V, #}-
→ [] / []
-{V, #}CRV-{V, #}CRV-VCR#-VCR#-
→ [] (/ [])
→ [] / [] (/ [])
→ deletion or []
→ deletion or
V
[](V) / V[]V !
Examples
relier:
[ølje]/[ølje]
bord: [b]
heureux:
[øø]/[øø]
garage: [aa];
arriver: [aive]
parti: [pati],
carte: [kat]
gardé:
[ade]/[ade];
garde:
[ad]/[ad]
crabe: [kab]
grave: [av].
quatre: /kat/ [kat]
cadre: /kad/
[kadœ]/[kadœœ]
* f(V)? stands for a suspicion of vowel quality conditioning not yet evaluated.
The speech described in this section is mainly that of CT28 (male, researcher, from Besançon)
and of FD28 (male, engineer, from Paris) (see note 43).
From these examples, it is quite clear that the unmarked most common French r-sound should
be considered a [], undergoing the following distributional "rules" (cp. Table 2).
29
In most cases, people preserving in France a vibrant pronunciation [] tend to present it
mainly in syllable onset position and usually as a single-strike sound. For those speakers,
uvular flaps seem dominating also in tautosyllabic clusters with stops (mainly voiced) but the
presence of a flap seem conditioned by vowel quality (back vowels disfavour trilling).
Very frequently final r's, whose lenghtening effect on the preceding vowel are well known,
undergo at least partial devoicing, even in prepausal position. Devoicing seems to apply more
frequently when // is followed by a voiceless sound. In any case, devoicing is accompanied
by weakening. That is the reason why I chose not to transcribe these sounds as [] (the
symbol [] standing for a weakened devoiced uvular fricative).
Followed by a consonant in postvocalic position (heterosyllabic or final cluster) the r-voicing
always agrees with the consonant voicing status, voiced consonant frequently allowing a
weakening to approximant articulation. That applies also for initial clusters (see above) where
the flap variant follows the voicing rule too (so that a [] becomes possible).46
Final postconsonantal r-sounds determine quite different outcomes (affecting the
phonological analysis of the entire word structure) as a function of the voicing status of the
preceding consonant. Even though depending on a number of variables, these r-sounds - when
not deleted - tend to be realized as weakened and devoiced after a voiceless consonant, and
reduced to approximants with the vanishing support of one or two surrounding vocalic traces.
An example of this tendency is available in Fig. 1 illustrating a selection of representative
realizations by the reference speaker CT28.
Intervocalic // is often reduced to an approximant [], especially when back rounded vowels
are involved (which usually causes lip-rounding all along its pronunciation).
In the upper row are displayed four spectrograms reproducing different fricative realizations
of postvocalic final r-sounds. In the second row the voiceless onset in crabe determines a
devoicing of the fricative realization with a residual flapping trace, whereas the following
spectrogram (le cadre) shows a weakened approximant realization appearing in final position
with the support of two short vowels before and after (attempting to create the ideal condition
where an approximant can be realized).
46
Similar distributional tendencies are widespread in the productions of the Belgian speaker here considered
who showed dominant trilling realizations, in agreement with Canepari's (1999:101) description and according to
the recent instrumental account in Demolin (2001).
30
[ y nw b a
[ k  a b
œ
 ]
]
[ l a b  
[
l ø k a
 ]
d    ]
[ l ø b 
[l a l  t


]
]
[ l a b u
[ p a  t


]
i  ]
Fig. 1. Speaker CT28: (up) final r-sounds in the words une barre 'a bar', la bière 'the beer', le bord 'the edge', la
bourre 'the hard labour'; (down) r-clusters in different positions in the words crabe 'crab', le cadre 'the frame', la
lettre 'the letter', partir 'to leave, go away' [All the spectrograms here presented are obtained with Wasp SFS - UCL]
[l a  a d  j o ]
[l a  a d  o ]
[
l ø c  a d œ  ]
Fig. 2. (left to right) Speakers CT28 and FD28: intervocalic r-sounds at word boundary in la radio 'the radio';
FD28: final r-cluster in le cadre 'the frame' (cp. the same word realized by CT28 in Fig. 1).
31
Final devoicing is evident from the spectrograms of la lettre, after a voiceless stop, and of
partir, where two coda realizations (internal and prepausal) are presented. Note the complete
devoicing of the fricative r in the heterosyllabic cluster with the following voiceless stop, and
the evident effects onto the foregoing vocalic formant pattern before the final r, raising a
perceivable transition to a different vowel quality.
All these considerations would seem to apply as well to phonosyntactic context (that is at
word boundaries). The first two spectrograms in Fig. 2 present a comparison between two
productions of speakers CT28 and FD28 of an intervocalic // in the same context.
Realizations are clearly both weakened even though in the second case, the sound - probably
showing the effects of a tongue-dorsum gesture towards the uvula (and mid-upper pharynx)
associated with a general decrease of the vocal energy - could hardly be described as an
approximant. The third spectrogram in Fig. 2 accounts for a different realization of the final
postconsonantal r in cadre. Acoustic cues of a vocalic gesture appear only before more
fricative r-sound (in comparison with the realization reproduced in Fig. 1).
32
2.5.2. Secondary articulations in r-sounds
Going back to the first row spectrograms in Fig. 1, one could observe various noise-band
patterns for final fricatives in the different vocalic contexts, hence arguing that a contextual
conditioning may be functioning (similar to the one applying for velar stops which usually
present advanced allophones when followed by prevelar vocalic sounds, acquiring liprounding before rounded vowels and so on).
These considerations can be extended to approximant intervocalic realizations and their
formant patterns. In Fig. 3 are displayed formant measurements for approximant r-sounds, as
realized by the two male speakers here taken into account, for those words presenting the
more frequent contexts. The measures have been obtained taking care to select similar
stylistic reiterations in order to avoid effects on the formant structure related to undershoot.
To do that, only segments with similar durations (and similar timing for surrounding sounds)
have been measured (in their exact middle-point).
Fig. 3. Formant measurements for approximant r-realizations in different V_V contexts for speakers CT28 (up)
and FD28 (down). Appreciable differences in their "vocalic" quality affects r-sounds.
33
The graphs display significant differences in the "vocalic" quality affecting r-sounds in
various contexts (the phenomenon is audibly appreciable in isolation).
Both speakers show uvular formant patterns with a general higher F3 (at least 2400 Hz, see
footnote 12 in §1; see Delattre 1967 below) but front unrounded vowel contexts tend to raise
F2 (indicating an advancement in the tongue position), whereas back vowel contexts lead to
some velarization.
In the context e_o FD28 normally anticipates lip-rounding more than CT28 thus obtaining a
more rounded r-articulation: in fact r formant patterns in this context tend to be influenced by
back vowels (especially [o]) ; on the contrary the general relatively low F3 (also when before
back vowel) shown by FD28, jointly with an F1-F2 convergence in the mid-low range, could
be a cue of more pharyngealized articulation. This is also clearly shown by the formant
pattern in the a_a context where it reveals an evident pharyngeal pattern (see below).
On the one hand, a partial conclusion, also confirmed by evidences farther presented in this
text, is that labialization and velarization affect most of these // realizations, confirming data
in Calliope (1989) where a generally [o]-like r-sound is presented for French.
Notwithstanding, better accounts for different tongue body shapes associated to back r-sounds
are needed, especially in comparison with similar articulatory effects appearing for other
sounds.47
On the other hand, the main conclusion is that uvular fricative sounds can be obtained with
different 'vocalic base' hence defining a set of characterizing sounds.48
That holds also for [], as I proved by articulating myself a series of vowel-coloured trilled
uvular [] shown in Fig. 4. A vibration of the uvula, depending on opportune aerodynamic
arrangements, may be triggered in quite different articulatory general postures involving
secondary articulations which influence the acoustic resonances characterizing the sound.
The first-four-formant values of these trills have been measured and reported on the diagrams
presented in Fig. 5 (see below).
Provided that a language-dependent preference for one of this kind of uvular r could be
generalized in different vocalic contexts, dealing with uvular r's would therefore always need
to specify this feature.
47
Recasens (2002, this volume) describes an [o]-like tongue body shape during the articulation of Catalan [l].
48
On the notion of possible speech sounds see Lindblom (1990a).
34
Fig. 4. Uvular trills with different 'vocalic base' (vowel-coloured []); left to right: u-based, o-based, and back abased; a-based, low e-based and mid e-based uvular trills.
Dispersion of vowel-coloured [R]
vowel coloured [R]
F1 [Hz]
3000
4000
2200
1800
1400
1000
600
200
3000
F1
2500
F1split
2000
F2
1500
F3
1000
F4
300
400
u-base
500
600
500
700
0
u-base
o-base
back abase
a-base
e-base
i-base
o-base
F2 [Hz]
3500
[Hz]
2600
back a-base
a-base
e-base
i-base
800
900
1000
Fig. 5. (Left) Diagram showing different formant patterns possible for uvular r-sounds (see Fig. 4.). (Right)
Dispersion diagram for these "vowel-coloured []".
35
2.5.3. Articulation manner and timing for r-sounds
In 1913, Rousselot (1913:53) claimed that the double r was necessary in French in order to
maintain a number of contrasts between imperfect and conditional verbal forms (for verbs
such as courir, mourir which would have presented it as in courrais/mourrais vs.
courais/mourais). The following literature during the second half of the XXth c., surveyed in
§2.3., considered the contrast nearly neutralized by that time.
The present day total absence of references to this kind of contrast makes one imagine that
nobody takes care of the topic anymore because he/she disregards the possibility. Yet a few
informants of mines, probably influenced by their literary vision of the language, assured to
be keeping the oppositions. I decided to carry out a pilot-study by having recourse to a small
number of sentences containing these pretended contrasting units (described as in C2, §2.5.).
The sample was based on /r/-/rr/ pairs realized by two female speakers (CC33 and CH42) in
two iterations of 5 sentences (like Si je courais je mourrais, see note 39).49
Durations of the intervocalic r-segments were measured by means of CoolEdit, and a
statistical evaluation carried out by keeping separate /r/ from /rr/ occurrences in the first or in
the second part of the IF-sentence (here referred to as 1st phrase and 2nd phrase).
Table 3 summarizes means and standard deviations of these measures.
Table 3. Durations (means and standard deviations) for different intervocalic r's in sentences uttered by two
female French speakers (words mourais/courais vs. mourrais/courrais).
CC33
1st phrase /r/
2nd phrase /r/
1st phrase /rr/
2nd phrase /rr/
µ [ms]
61,8
65,6
73,2
81,0
σ [ms]
13,8
10,9
13,4
12,0
µ [ms]
69,8
55,6
61,2
80,0
σ [ms]
22,0
5,4
11,3
20,8
CH42
49
Interviewed after the recordings, speaker CC33 revealed that, in her judgement, during the recordings, she
always pronounced a double r for the conditional tense verbs. Anyway, she asked me to repeat once again the
recordings, getting quite different durations with an evident emphasis on constrasts involving r-doubling.
CC33
1st phrase /r/
2nd phrase /r/ 1st phrase /rr/ 2nd phrase /rr/
µ [ms]
51,8
55,3
108,0
97,5
σ [ms]
15,9
10,2
14,2
3,5
36
The sample is certainly not enough representative to generalize the results (only 20 measures
per speakers are considered here, + 10). Some variables show a sensible variation (important
standard deviation) as a consequence of the presence of one or two extra-long realizations
(probably once having achieved the awareness of the experiment aims). In any case,
differences statistically resulted as weakly significant (except those of the second session for
CC33).
In spontaneous realizations, for both speakers a slight lengthening is apparent for the 2nd
phrase /rr/ more than elsewhere.
The expected general relative lengthening of the 2nd phrase segment as compared to the
corresponding one in the 1st phrase is not always respected (see /r/ for CH42).
In the 1st phrase, /r/ may be occasionally longer than /rr/ (the apparent increased duration for
/rr/ with regards to /r/ for CC33 is not significant: the two segments have mean durations of
the same magnitude order).
No general conclusion can be drawn from this experiment, but the suspicion is raised that
some speakers are still maintaining the contrast in some stylistic registers (or feel to be
keeping it, that is linguistically relevant).50
In compensation, as works on different languages have recently shown, the realization of
various r-allophones more suitably depends on the articulatory rate.51
In particular, without making any distinction between -r- and -rr-, the shortest intervocalic
passages were always characterized by approximant realizations.
On the contrary, the longer the time taken on to articulate the r-sound, more likely the sound
contained vibrations (single but even multiple for speaker CC33, in this hypercontrolled
style), indifferently for -r- and -rr-.
Taking a look to the typical duration values, in relationship with the articulation type, it seems
that if the temporal slot is about 50 ms the only articulations allowed are the approximant or,
at a lesser extent, voiced fricatives; with a slightly longer slot (say about 65 ms) a middle
range of reduction is active and definite fricative sounds are more likely; with a time of 80 ms
to 120 ms uvular flaps or trills are allowed respectively as far as the duration increases.
50
That's particularly important if one thinks that the same speaker above do not show any intention to make
contrasting other consonants on length bases (in spite of the visual reminder provided by orthography).
Therefore dégoûtant and dégouttant, Monet and monnaie, palier and pallier are nowadays homophones.
51
Many of these researches have been stimulated in the light of the H&H theory in Lindblom (1968, 1990, and
many other references).
37
Hence, being different r-variants influenced by the timing, whatever the maintenance of /r/ vs.
/rr/ contrastiveness in a situation of hypercontrolled speech, the reductive nature of
spontaneous speech would probably give way to neutralization.
As it could be the case for other languages, the production mode seems playing a more
important role in the phonetic implementation of French rhotic system than a weak
phonological distinctiveness.
That seems going in the opposite direction as compared to the situation of other languages
with more stable contrasts (as it has been shown, for instance, by Machuca-Ayuso 1997 for
Spanish).
38
2.5.4. Citation forms vs. connected speech
As described in §2.5.1., an extensive study - based on formal speech - has been carried out on
French //. The sound distribution has been analyzed in terms of common tendencies in
productions of a reference speaker and of three "reducing" speakers.
Formant patterns have been analyzed in common words like the ones presented in Fig. 6,
reproducing samples from the most reducing speaker FD28.
Intervocalic r-sounds (mainly approximants) showed allophonic processes determining
special formant transitions in different vowel-contexts, but the general pattern would follow
Delattre's descriptions (see Fig. 6).52
For this speaker, I found vowel-dependent formant patterns with F1 typically over 300 Hz (up
to 800 Hz in a_a) and F2 reaching the highest values (>1500 Hz) in front high vowel
contexts, middle values (1100 Hz) in open vowel contexts, and low values in back vowel
contexts (about 1000 Hz). F3 usually presented a lowering in a_a context towards 2000 Hz
(as it happens for pharyngeal articulations), remaining mainly stable for back rounded
entourages (around 2400 Hz) and raising up to more than 2600 Hz for front unrounded
contexts (thus revealing velar/uvular dominance).
Different the cases of the spontaneous speech analyzed from corpus C3 in the framework of a
projet on French banlieue languages where I contributed for the phonetic section (see Billiez
et al. 2002).
The main outcomes of this enquiry generalize the distributional scheme in Table 2 to wordboundary contexts in spontaneous speech (e.g. tomates rondes 'round tomatoes' > -t#-).
52
The acoustic characteristics of a French pharyngeal r are described in Delattre (1967:36) and Delattre
(1962:281), where its formant pattern is compared with the one necessary for an apical trill, and the movement
of F2 is acknowledged as similar to the one characterizing labial sounds ([w], [v], [b]; note that the same author,
in Delattre 1951:234, considers a velar activity in its production not mentioned anymore in further studies). My
reference French r-sounds are in most part uvular sounds with frequent pharyngeal secondary articulations. I
often find exactly the same transitions discussed in Delattre (1966, even if in an extremely simplified view), but as it is presented in Fig. 18 below - I made different spectrogram interpretations excluding a velar concern in
some contexts for common speakers (it is the general F1 rising that suggests a uvular/pharyngeal articulation
place).
39
[ t e  i b
l
]
[j a  a  ]
[ m a l ø  ø ]
Fig. 6. Spectrograms of intervocalic r-realization by speaker FD28;
(left to right) words: terrible 'terrible', garage 'garage', malheureux 'unhappy, unlucky'.
In the intervocalic positions "on observe grosso modo la tendance normale au relâchement
articulatoire, réduction à l'approximante, et peut-être reculement de l'articulation associé à une
réduction du temps de l'approximation" (Romano in Billiez et al. 2002:42-43).
The common velarizing effect that usually appears in consonant clusters involving palatal or
velar consonants, it is generalized in many other contexts in socially marked r-usages in
spontaneous speech (as in §2.5.; cp. Fig. 7).53
53
There I found a matching case in the description given by Germi (1968:93) stating that "précédés des voyelles
i e  a les formants de /R/ ont une valeur voisine de celle des formants de la constrictive vélaire russe [x]" (cp.
Fant 1960:175-179). Quoting from the DGLFLF report (Billiez et al. 2002:42-43): "Les réalisations ici analysées
respectent une certaine régularité distributionnelle concernant les phénomènes de désonorisation et de réduction
à approximantes [...] mais s'éloignent de la "norme" pour plusieurs raisons. L'articulation de certains "r" non
intervocaliques [...] est particulièrement énergétique, donnant lieu peut-être à un allongement, associé à la
présence de vibrations irrégulières (souvent en condition d'assourdissement, cf. planches [...]). Cette tendance est
particulièrement marquée pour les garçons [...] il pourrait s'agir d'une [...] sorte de fricative/vibrante dorsouvulaire (pas toujours sonore et pas forcément vibrante) avec vraisemblablement un soulèvement de la partie
prédorsale de la langue dans une région avancée, prévélaire ou palatale. Une autre particularité réside dans le fait
que cette articulation (qui semble naturelle en contexte de voyelles palatales, en français commun et en situation
informelle) se manifeste aussi dans des contextes non palataux, et donc en correspondance de voyelles
postérieures ou centrales. Les filles semblent aussi "frotter" les "r" et présentent moins souvent un autre
paramètre important chez les garçons : la présence de fréquentes pharyngalisations des voyelles, liées à des
phénomènes pour lesquels la scansion rythmique des syllabes se fait à l'aide d'un ictus intensif très marqué"
(Romano in Billiez et al. 2002:42-43).
40
[p a  t

i  ]
[zm a  a t  i z]
Fig. 7. Spectrograms comparing two postvocalic r-sounds in the same following consonant context (left to right):
CT28 saying partir 'to leave, go away' (reproduced from Fig. 1.) and K (anonymous male speaker) saying
smarties 'smarties' (in a noisy environment).
In Fig. 7, two spectrograms allow the comparison between postvocalic r-sounds in the normal
pronunciation of the reference speaker CT28 with the marked pronunciation of an anonymous
young speaker (here labelled K) collected as in corpus C3 (as in §2.5).
CT28's postvocalic r-sounds in the word "partir" are two voiceless (weak) fricatives (about
the final realization see comments to Fig. 1). The first one, before a palatalized consonant,
acquires a more advanced and raised tongue body thus conditioning the formant pattern so
that it shows velar features. K's realization in the word "smarties" shows a similar pattern but
clearly (despite the bad quality of the recording) associated to a devoiced uvular trill, here
transcribed as [] accounting for the velarization.54
Many other examples (as the ones presented in Fig. 8.) I analyzed in the DGLFLF framework
showed similar tendencies. The realization from a banlieue young female speaker recorded in
spontaneous conversation (cp. "planche 4. M232 - ta mère" in Billiez 2002) may be
compared, for instance, with an old pronunciation of the same word (e.g. the one given by the
actor Raimu in 1931 while somehow imitating a pretended northern accent).55
54
In this case, it would be useful to extend an interesting transcription listed by possible r-sounds in Czech by
Simackova (2002, this volume) [] defining a velar fricative with uvular vibrations (in the example above a
devoicing diacritic would be added, or a definite voiceless fricative could be instead taken as a base).
55
These speech sounds are available thanks to the Enregistrement sonore procédé "Western Electric" adopted by
M. Pagnol in Marius (1931) and are captured from the movie soundtrack in a commercial VHS tape.
41
[s a m e 

]
[t
a
m
e æ

a ]
Fig. 8. Spectrograms comparing two "spontaneous" realizations (in different situational context) of final rsounds (left to right): the actor Raimu saying sa mère 'her mother' and M (anonymous female speaker) saying ta
mère 'your mother (used as excl.)'.
I would point out the particular vocalized realization of the final r in the banlieue speech here
reproduced, which seems associated with a rapid change in spectral energy distribution (here
transcribed, on auditory analysis bases, as a transition through uvular fricative conditions).
That represents a very frequent case in which rapid gestures and changes in aerodynamical
conditions result in stop-like or fricative-like configurations (without the acoustic cues of real
closures or frictions).56
On the other hand, these examples give me the opportunity to highlight as, most of the time,
the original character of idiolectal realizations is not necessarily in different place or manner
of articulation but in many cases only in a different distribution from the normal scheme.
For instance, banlieue young speakers (see Figg. 7 and 8), even though presenting the general
more pharyngeal articulation in final codas, seem preferring "hard" raised uvular trills (with
joint devoicing and frequent velar frictions bound to anticipatory effects) in strong positions
where usually common speakers have just weakened devoiced uvular fricatives.57
Another example is in the same Raimu's Provençal-accented pronunciation which could arise,
as regards to r-realizations, from the generalization of a pharyngealized "soft" uvular
approximant articulation in unexpected positions (such as in prendre 'to take' pronounced
56
For a glimpse on the reduction possibilities driven by pharyngeal articulations see the literature on German
postvocalic r-sounds raising diphtongization (cp. Barry 1995; also see examples Kohler 1999:88).
57
I use here two shortcut terms "hard" and "soft"; "hard" meaning "presence of high tension articulation arising
friction noises" and "soft" referring to "weakened articulations reduced to vocalic glides".
42
[pad] vs. the present day "normal" and "reduced" French forms [pdœœ]/[pdœœ]/
[pd] and many other examples).
Having discussed a set of topics related to common French rhotics with a few instrumental
arguments (several other studies are needed before to have a rigourous representation of
various phenomena), and having observed a significant number of reduced approximant
realizations for French r-sounds, I conclude here with a word in favour of a further distinction
proposed in recent "rhotics" literature (cp. Simackova 2002, this volume) between trillingvarieties as opposed to waving-varieties representing (as it is also shown in the Italian
examples) quite different ways to implement this class of sounds.
43
3. Italian r
Concerning the actual status of Italian r-sounds, the literature is relatively poor. As partially
introduced in §1., the main interest is devoted to r-variability in some geographical varieties
and to the diffusion of defective variants known as r moscia to whom, as far as I know, no
instrumental study has been explicitly dedicated.
A simplified description of the phonological Italian system basically assumes a phoneme /r/
and its geminate counterpart /rr/ whose phonetic realizations, as already seen in §1, are both
apical trills with a different number of contacts (cp. Ladefoged & Maddieson 1996:219-221).
In Canepari's traditional finer analysis, recently summarized in Canepari (1999:97-102), the
phoneme /r/ is associated to both [r] and []:
L'italiano ha un solo fonema vibrante /r/ (alveolare sonoro [...]), caratterizzato da una
serie di leggère e rapide occlusioni tra gli alveoli e l'apice della lingua [...] Si ha pure
una sola rapida occlusione, che produce un fono monovibrante, o meglio vibrato, []
[...], che ricorre in pronuncia neutra perlopiú in sillaba non-accentata (Canepari
1999:97-98).
A very detailed distributional analysis is given in the following passage:
[N]ella pronuncia neutra odierna effettiva abbiamo, normalmente [r] in sillaba
accentata: [(C/V)'rV-, 'CrV-, 'Vr:C(V), 'V(:)r#] (oppure, solo come variante occasionale,
non sistematica, e non enfatica, []). Mentre negli altri casi si ha []: ['V:V, (V/C)()V-,
V-, -(')C-] (oppure come variante possibile, specie per enfasi, [r]). Per /rr/ si ha:
['Vr:V, V'rV, ()VV, V()V] (oppure anche [r:r, rr], soprattutto per enfasi). (ibid.).
I checked with an instrumental approach the examples proposed by Canepari (1999:328)
(raro [rao] < /raro/ 'rare', parlare [parlae] < /parlare/ 'to speak', Mario [majo] <
/marjo/, carro [karo] < /karro/ 'cart, wagon', Enrico [enriko] < /enriko/,) which
presented various phonotactic possibilities, determining different phonetic realizations for the
/r/ phoneme. The speech sample came from the tape associated with the handbook and the
speaker was a professional male speaker without particularly evident regional marks. The
results are in the spectrograms in Fig. 9. Taps appear only intervocalically in unstressed
44
position or as the explosive phase in the realization of /rr/.58 They usually have a closing
phase of 54÷67 ms which is slightly (and suspiciously) higher than taps usually measured, for
instance, for Spanish (see Recasens 1991), thus requiring a better articulatory account. Other
/r/ realizations (such as the coda /r/ in the first unstressed syllable of parlare and the onset of
the stressed syllable of raro and Enrico) are not single-strike sounds and are realized with a
2÷3-strike trill against the longer 5-strike trill realizing /rr/.
As already introduced in note 3., like the standard representation above, a number of regional
Italian idiolects follow this distribution, hence presenting intervocalic single rhotics realized
as single-strike sounds.59
On the bases of a number of items observed in spontaneous dialogues and for different
regional Italian varieties (variously disposed to tap spreading in many other phonotactic
contexts, see §3), I observed that single-strike r-sounds tend to preserve a higher tension in
the vowel-to-consonant transition than usually shown by speakers of languages traditionally
described as tap-languages.
58
According to Rousselot (1913:53): "L'r double se comporte donc comme les autres consonne redoublées qui,
doubles pour l'oreille, ne sont, au point de vue articulatoire, que des consonnes simples fortes et longues", but
these sounds lead to a phonetic distinction: "La 1re r entendue est une r implosive ; la 2e, une r explosive" (ibid.).
In agreement with Canepari's distributional scheme, as it is shown farther in this work, this assumption for
Italian, does not contrast in principles, with Inouye's (1995) generalization of a phonetic length feature as for the
relationship between trills and taps, which remains valid for languages without geminate/singleton contrasts.
59
This agrees with Inouye (1995) who demonstrated, with researches on various languages, that intervocalic
tapping of trills is widespread crosslinguistically (in this case only as realizations of a single, not-geminate,
consonant).
45
[ r
a
 o ] [p a r l
[k
a
r  o
]
a
 e
] [ m
[ e n r i
a
 j
o ]
k o ]
Fig. 9. Spectrograms showing standard Italian pronunciation for /r/ in the words: (upper row) raro 'rare', parlare
'to speak', Mario; (lower row) carro 'cart, wagon', Enrico (professional speaker AV (male), see Canepari 1999).
46
3.1. Phonetic phenomena involving r
In the evolution of the Italian language and of Romance dialects spoken in Italy, a significant
number of phenomena, related to sound changes and derivational processes, involved rhotics.
Besides the alternations that were present in Latin described in §2., and general properties
related to liquids surviving in Romance modern dialects, I would like to mention a selection
of phenomena particularly marking, excluding from this section those dynamics which are
still active on a sociophonetic ground.60
Consonant assimilation involving r in infinitive forms with clitics has been often described in
various period of the history of the literary language and in different places. Migliorini (1937)
recorded that in the XIVth c.:
Davanti alle particelle enclitiche –lo –la ecc. la r dell'infinito [...] si [può] assimilare:
coprilla (Bisticci), pensalle (Pulci; ma anche trovarlo in rima con Carlo), perdonalli
(Piov. Arlotto), trovalla (Poliziano) [...].
Resta[no] sempre mal digeribil[i] [...] i gruppi di consonante seguita da l: cripeato
"clipeato" (Gherardi, Paradiso degli Alberti), compressione "complessione" (Alberti),
Prinio, exempri (in una lettera del Bisticci), frutto "flutto", pepro "peplo" (nel
Vocabolista del Pulci), fragello, obrivione (Leonardo); così è nato sopperire da
sopplire, supplire (Morelli, ecc.) (nota 163: Per reazione, si ha invece splimere per
esprimere (M. Franco), refligerio, plecipitare (Leonardo); il Pulci, definendo fleto nel
Vocabolista "pianto ed il mormorio del mare", mostra di confondere fletus con fretus)
(Migliorini 1937*1992:261)61
60
In this kind of processes r-sounds took part to phonological changes traditionally listed as assimilation,
dissimilation and so on. For instance it is latent in standard Italian (even if rarely signalled by handbooks) the
common assimilation at the boundary between the article il and a following noun beginning with r- e.g. il ragno
-> ir ragno (cp. with Roman article er, and common pronunciation in Tuscanian ir mi babbo < It. il mio babbo).
Liquids played an important role in determining some Latin morphological alternations (still living in Romance
languages and even in English) such as the suffix -ALIS/-ARIS which gives It. culturale, generale etc. vs. solare,
popolare, particolare etc. In a number of Latin forms evolved in Romance languages (often entwined with
Greek roots or other languages' lexical forms) we found common phenomena related to svarabhaktic vowels
evolved from consonant cluster including r-sounds or, on the contrary, to original vowels absorbed by r-sounds.
E.g. Lat. morphological alternation -TOR / -TRIX, AGER - AGRI, NIGER - NIGRU(M), ALACER - ALACRIS (> it. alacre,
allegro) etc., also see Lat. scient. prefix hydro- < gr. hýdor, hýdatos, gastero- in gasteropoda, but gastro- in
gastronomia. More specifically for Romance languages, It. altro < lat. ALTERU(M), cfr. ALIUS, and the special
cases of Italian cetra, chitarra < CITHARA(M) (Gr. kithára), temprare < var. of temperare, tirare, trarre < lat.
TRAHERE,
61
comprare, comperare < lat. COMPARARE, sgombrare, sgomberare < der. A. Fr. 'combre'.
Canepari (1999:419) describes similar phenomena in Umbrian dialects.
47
A very common phenomenon in Latium dialects, around Rome, is /rr/-degemination (see
Canepari 1999:102; but sometimes degeminated r-sounds are also heard in north-eastern
Sicily).62 This pronunciation often extends to regional Italian in careless speech.
Piedmontese dialects, as well as other gallo-roman patois, sometimes present an l-rhotacism
in intervocalic position which rises forms like surei (at Frabosa) < Lat. SOLICULU (cfr. fr.
soleil) and in fact a number of speakers use lateralized taps when they speak Italian.63
Several other dialects, including Roman, present a very common l-rhotacism only in internal
heterosyllabic clusters such as -L+consonant- : sordi where It. has soldi, corpa where It. has
colpa and so on, Sallentinian (Sall.) and southernmost dialects have carèndula < KALENDULA
and many others including examples of svarabhaktic vowels appearing as in Sall. càncaru <
CANCRU(M),
cancarena < CANCRENA, putàraca < PODAGRA and so on.
Well described, even if disregarded in world languages' surveys, is the presence of retroflex
sounds characterizing southernmost varieties.64 In some places where cacuminalization is
widespread, -tr- undergoes retroflexion yielding to a retroflex stop or affricate (Sic. ṭ r  e nu
cp. It. treno). Furthermore -tr- affrication after s- gives place to assimilation ( -str- > -- > -) as in Sall. noscia < NOSTRA, finescia cp. It. finestra, Sic. ata < STRATA and so on.65
62
I heard a number Sicilian speakers from the area of Milazzo presenting this phenomenon. I am not dealing
here with general degemination processes interesting northern varieties and particularly evident in prestressed
position (very common e.g. in Veneto): all the other singleton/geminate contrasts are preserved in these areas.
"In molte zone del Centro, compresa Roma e la Toscana occidentale, /rr/ nell'accento marcato si confonde con /r/
semplice" (Canepari 1999:102). Moreover, in the common intervocalic position, the dominant realizations could
be single-strike sounds (cp. Canepari 1999:430), whereas weakened sounds are sometimes heard in other
positions (such as consonant clusters) perhaps as a further effect of common defective pronunciations.
63
Similar examples are witnessed by Contini (1983:396-397) for northern Sardinian "Dans les parlers sassariens
-r- continue aussi -l- intervocalique originaire zè  r u "ciel" sòri "soleil" [...] Ce -r-, par son articulation très
relâchée et dévibré, se différencie nettement du -r- du domaine sarde en général".
64
An historical cacuminalisation is attested in Sicily, parts of Calabria and the Sallentine Peninsula, but also in
Sardinia: -LL-> -d  d  - (and sometimes evolved through > -tr- > -r-): cad d  i na < GALLINA, ad  d  u ,
aut  r  u , auru < ALIUS, ALTERU(M), puricinu < PULLICINU(M). The phenomenon is well-described in Italian
linguistic literature (an abridged selection of references should include Millardet 1933, Bianco 1981, Caracausi
1986). A number of articulatory possibilities are listed for Calabrian dialects by Romito & Belluscio (1996),
Sorianello & Mancuso (1998) and Romito & Sorianello (1998).
65
Interestingly non-etymological -r- appear in t  r  o nu < TONU 'thunder', where other dialects (and standard
Italian) show diphtongization tuono [twno], and in trullo 'typical cone-shaped stone hut' < ?Gr. θ ό λ ο ς
48
In Sicilian (Sic.) and southern Calabrian (Cal.) an initial fricative (retroflex) r undergoes a
lengthening process (getting an allophonic variation like in Spanish varieties with
"rehilamiento"; see, among many others, Navarro Tomás 1934, and Gomez 2002, this
volume). Most part of these pronunciations are still common for conservative speakers when
they speak their regional Italian.66
'dome', whereas r and s are interlaced. Similarly Sicilian retroflex r raises interesting backwards assimilation
phenomena too: Palemmo instead of Palermo, giunnu instead of giurnu It. giorno 'day' etc. (cp. Rohlfs
1949*1966:376 who gives the examples babba, ussu, funnu etc., see below). Inversely, it is not difficult to find
examples of intrusive velar consonants in initial r environment (as in Sall. croffulare 'to snore' cp. Fr. ronfler,
probably related to RHONCUS, based on a Greek stem with ‛ ρ-), perhaps related to aspirated r-sounds which are
described for ancient Greek.
66
According to Canepari (1999:102), in these regions (plus Sardinia), word-initial /#r/ is replaced by /#rr/. In
Sicily and southern Calabria, this is then realized, in the more conservative accents, as a voiced alveolar or
postalveolar fricative/approximants which are neither [z] nor [] and not even their weakened counterparts
(Canepari transcribes them respectively [ ] and [ ]). Missing fundamental information on tongue sulcalization, I
usually simplify the transcription of these sounds, assuming postalveolar (retroflex) fricatives and approximants
as basic sounds (as in Mandarin Chinese) and then having recourse to diacritics, thus obtaining alternations such
as []/[] or []/[] (or even []/[] I heard from speakers of the Agrigento area). The principle I follow is
choosing the well-defined IPA-symbol as a reference pronunciation and then describing the main sources of
variation in relation to it. This basic assumption allows me to get [z] as a lowered articulation of [z] not reaching
the necessary conditions to get [] (as usual, I transcribe [] the southern English /r/ realization after alveolar
stops which is a sound different from [] and - of course - from [z] or [z]; a similar example of diacritics use is in
Ladefoged & Maddieson 1996:223, when they transcribe [] for a French uvular fricative further raising it). The
variety of possible slightly different sounds in this articulatory region is so developped that one cannot explain
the missed change in a given linguistic system from [r] to [], or to fricative variants of it, with the claim that a
/z/ already existed in that system. In the light of the enormous variability directions, and of the slight nuances,
we cannot share opinions such the one referred to, for instance, in Schiller & Mooshammer (1995) and Barry
(1997) where it is said that a variant may not be viable in a language purely because of the danger of a
widespread confusion with other phoneme realizations. As it is proved by many examples, some changes could
be simply phonetic and be limited to appear in some hypo-/hyper-articulation conditions. In connection with the
examples in the case-study, Navarro Tomás (1934), for instance, always distinguishes a fricative r from an
assibilated r. Describing the fricative apical r in Mexico, Delattre (1944*1966:206) wrote: "The intended sound
was a trilled (multiple-vibration) r, but the vibration of the tip of the tongue, functioning as an elastic organ,
failed to get a start, and the result was a mere constriction between the tongue tip and the alveolar ridge, at a
point between those of [z] and []. Naturally, the fricative sound resembles both [z] and [], and would simply be
an intermediary sound of these, were it not for the fact that the aperture is greater, the tongue tip more raised and
49
Another very common process involves r-metathesis which is still idiosyncratically
functioning for speakers preferring e.g. interpetrare to interpretare (and other examples).
The process is very active in Sardinian dialects where various dialects regularly have sradu <
prokku < PORCU, lragu < LARGU etc. (see Contini 1983:401-402).67
SARDU,
A particular case on which I propose to open a discussion (see below) is the simplification of
the Lat. -ARIU(M), in which a cluster formed by r+j was simplified with r-deletion. This is
nowadays the normal Italian outcome by the way of Central varieties including Tuscanian
(notaio < NOTARIU, marinaio < MARINARIU, calamaio < CALAMARIU etc.) but the common
outcome in peripheral varieties is –aro/u, as in notaru, marinaru, calamaru (and other
dialectal forms eventually promoted into the standard - often creating allotropes - such as
zampognaro, carbonaro etc.) with r-preservation and i-absorption. According to Migliorini
(1937*1992:69) "la differenziazione del tipo toscano e umbro –aio, -oio dal settentrionale e
meridionale –aro, -oro si ha almeno fin dall'ottavo secolo" (ibid.:261) and still in the XIVth c.
"L'esito fonetico normale –aio –ari (danaio –danari, scolaio –scolari) compare ancora non di
rado, quantunque indebolito dall'analogia".68
much more tense, and the sound is held longer". After Cárdenas (1958), a survey on slightly differing r-sounds
in Spanish America is now avalaible in the section 3.2.6 (430-434) of Miotti (1998).
67
Sardinian varieties have a pharyngeal r as an outcome of Lat. -L- (Contini 1983, see below) and a frequent
passage s > r of the kind already discussed for French, as in the pl. det. sos > sor before a consonants like /b d /
or /f m dz j/ (see Contini 1983 and many others). Examples of a r > s reverse change are available for IE
languages (also see footnote 19) within the linguistic domain here surveyed, e.g. in Oscan (*kús[ú]l-s < Lat.
CORYLUS
'hazel', see Trumper & Chiodo 1997:200-201), and in Messapian alternations karp-/kasp- (surviving in
Sallentinian càrparu/càsparu 'type of stone', see Ribezzo 1950). Examples of r-metathesis are instead from
Sallentinian dialects, who also have some lexicalized metathetic forms like terracata 'root' (< RADICATA × terra
'earth'), palora 'word' (< It. parola), scalora 'salad type' (< It. scarola); crapa 'goat' (<
FEBRE),
fràbbica/fràbbaca 'factory' (<
PULVERE),
68
FABRICA),
CAPRA),
freve 'fever' (<
crasta 'vase, flowerpot' (< Gr. γαστρα), prule 'dust' (<
craune/craone 'charcoal' (< CARBONE), ṭ r  u bbu 'trouble' (< TORBIDU), fràulu 'flute' (< FLAUTU).
Accounts are also available for the XIIth c. Florentine (Migliorini 1937*1992:418). Tagliavini
(1949*1982:411) describes the passage ri > i (such as in area > aia; -ario > -aio) as a characteristic of Tuscanian
dialects, but at p. 413 (n. 119) he extends the area where the process took place to Umbria and parts of the
Sabina. Note that the passage is less concerned in –ARI where i is nuclear. The phenomenon is well accounted for
in Castellani (1950) quoted by Tagliavini (1949*1982:128). This author has recourse to the substratum theory to
state a supposed phonetic tendency of some Ligurian dialects (as well as Provençal dialects) to the weakening of
intervocalic -r- which tends to be palatal (cp. Migliorini 1937*1992:100, n. 45). In my view, mainly based on
present day pronunciations, one should distinguish the lambdacism (r > l) found in Piedmontese, Ligurian and
Provençal dialects (and also described by Rohlfs for Tuscanian), as an example of r-lateralization connected with
50
On Romance r-pronunciation in Italy we found plenty of details in Rohlfs (1949*1966:376),
where various dialectal r-sounds are listed.
A passage r > l is described as taking place in Tuscany "Nel dialetto popolare di Pisa e del
suo 'contado', come a Livorno e a Lucca": final-syllable r takes a lateral sound as in Pisan
calta instead of carta 'paper'.69 Similar cases are also taken into account for a few
Piedmontese, Calabrian and Sardinian dialects [...] whereas a velar pronunciation in almost all
phonetic contexts is recorded in the North, in Valtellina (Isolaccia) for which he notes mòrt
cp. It. morto 'dead', porter cp. It. portare 'to bring' and so on (ibid.:377).
Traditional dialects described as having back rhotics are in northern areas (almost exclusively
northwerstern dialects or in the bilingual areas in North-East, at the boundary with Germanspeaking countries) but descriptions do not give any quantitative detail of this dominance.70
Very interesting is also the case of the i-vocalization of some r-sounds in the Ligurian area
before a labial or a velar consonant: in Sassello he recorded baiba, aicu, coipu instead of
barba 'beard', arco 'bow', corpo 'body' etc. which he makes matching with Sicilian northern
dialects East and West of Palermo with gemination effects on the following consonant (pal.
baibba, cuòippu, [...] Palièimmu, juòinnu cp. It. giorno 'day' and so on).
A well-known phenomenon for Romance dialects south of Naples and in northern Calabrian
is the passage d > r, similar to the tapping of alveolar stops in English varieties.71
the l-rhotacism (cp. data from ALEPO, see below), from the passage r > j which I feel as a condition still active
by means of a retraction and a weakening of the r-articulation in some Ligurian and Tuscanian speakers.
69
A number of examples are given "in pisano calta, pòlta, pèldere, sóldo, pòlco, còlpo, èlba, félmo, vèlso,
melcato [...]; in livornese mèlda, aldo, schèlzo, giolno, soffelsi, gualdo, selvizio, permette, solte; in lucchese
qualto, taldo, salto, invelso, toldo, stalna, pelso, felmo, celto [...]" (Rohlfs 1949*1966:376).
70
No back r-sounds are taken into account by Invrea (1936) in his description of northern Italian. Describing the
common dialect of Turin (Piedmont, NW) in 1914, Rivetta (1914:73-74) always transcribes r without particular
comments, but at p. 74 he writes kdy (sic) for Piedm. crëdù It. 'creduto' better accounting I guess for a
syllabic r-allophone, usually realized nowadays as a soft alveolar tap preceded by a short svarabaktic schwa as it
happens in Slavic languages with syllabic rhotics (I heard a number of Piedmontese speakers pronouncing in
Italian something like perparo instead of preparo, porgetto instead of progetto, particamente instead of
praticamente and so on, cp. in Portuguese). A prosthetic vowel is often lexicalized in Piedmontese dialects in
words beginning with r.
71
See Tagliavini (1982*1949:105): "nei dialetti italiani centro-meridionali non è raro il passaggio d > r (tipo
maronna = Madonna) [...] L'area attuale del fenomeno -d- > -r- secondo i dati dell'AIS, comprende la Campania
meridionale [...] a cominciare da Napoli (pere "piede" [...]), ricorrre in Lucania [...]. In Sicilia ricorre nella zona
costiera settentrionale [...] e a Giarratana (Siracusa)".
51
One can find many examples of it in Rohlfs (1949*1966:204) for Cal. ricía It. diceva 'said',
rúrici It. dodici 'twelve' etc., tu rici It. tu dici 'you say' vs. chi ddici? (RF) It. che dici? 'what
do you say?', Cilento li rienti It. i denti 'the teeth', Neapolitan o rit It. il dito 'the finger', o
rènde It. il dente 'the tooth' etc.72
As it can be seen, little space is reserved to r-varieties in the north-eastern areas of Italy which
in my experience are also very interesting (and in strong connection with the cases of rlateralization seen above) and to r-variants usually described for north-western varieties by
local dialectologists and sociolinguists.73
As shown by a few examples, I think that a large part of the dialectal phenomena widespread
in Romance dialects spoken in Italy are influencing present day Italian r-pronunciation.
Various factors are involved in the sociolinguistic dynamics and the phonetic conditioning
which are determining r-variation throughout the community. Dialectal data must be collected
in a narrow phonetic perspective and local preferences analyzed in order to give a more
finegrained sketch of the present linguistic situation outside the stereotypical description of
general "defective" or "retroflex" pronunciations which represent the more frequent common
places.
72
The same would stand for some Corsican varieties too, as in the Ajaccio dialect tu rici It. tu dici 'you say', le
reci dite It. le dieci dita 'the ten fingers' against the common l outcome in Guagno u lènte It. il dente 'the tooth'
aggu lèttu It. ho detto 'I said' (Rohlfs 1949*1966:205).
73
Examples of r-deletion and r-variants are described for Piedmontese dialects (see below). Discussing
examples of r-deletion in Sardinian varieties, Contini (1983:393) writes "En gallurien [...] [r] s'est effacé (n. 2
C'est la situation actuelle dans les parlers de la Ligurie, du Piémont et de la Lombardie)". In the Occitanian
dialect of the Germanasca Valley, Pons (1973) recorded infinitive verbal forms with final r such as pilhar
corresponding to It. prendere, pigliare 'to take' (e.g. Pons 1973:LXIII) but two decades later Pons & Genre
(1997) prefer to write pilhâ for the same verb: "Fino a qualche decennio fa, r finale dell'infinito, ora caduta
allungando la vocale precedente, in fonosintassi veniva ancora pronunciata da alcune persone anziane, es. anaramount oggi anâ amount 'andare a monte'" (Pons & Genre 1997:XIX).
52
3.2. Back (and not only) r-sounds in Italian
As in French, the term "grasseyé" is nowadays used to refer to a varieties of sounds, the
general category for Italian r-sounds differing from the apico-alveolar vibrant sounds
described as standard in §3. is traditionally under the label r moscia which could be translated
with "limp r" / "lifeless r".
As shown by some phoneticians (see Canepari 1979, Mioni 1987), limp or lifeless r-sounds
are in reality quite different articulations gathered by an impressionistic principle seeking for
simple general way to mark defective or snobbish pronunciations.
People using a different kind of r are euphemisticly said to have a French r (r alla francese or,
shortly, r francese) even when these sounds have nothing to do with French r-pronunciation.
Other common expressions to speak about a burrer are just saying ha la erre '(s)he has the r'
or, in some cases, non ha la erre '(s)he has not the r'. In other cases the r-pronunciation is not
lifeless at all (e.g. the case of long uvular trills) but the label r moscia is extended to them by
some informants. On the contrary, I heard the definition of r pizzicata 'pinched r' also used in
some regions with regards to this kind of more sustained but even "different" pronunciation.
In the phonetic literature on this topic a disagreement appears on the reference to the
sociolinguistic status of such r-sounds between various authors because different opinions are
expressed on the prestige status of idiolects showing these r-sounds. That reveals an
incomplete (more often, non-uniform) knowledge of the geographical and social variability of
these phenomena in Italy.
As it is also recalled in Ladefoged & Maddieson (1996:226), Ladefoged et al. (1977)
described a uvular trills appearing in Italian in a prestige dialect, and it is clear that the authors
were dealing with social dialects (but, since there is no clearcut social differentiation, they
were properly considering only idiolects).74
Chambers & Trudgill (1980:191) write about a « uvular /r/ only in some educated speech »
but that description could not even fit to the real Italian situation where the usage of this kind
of r-sounds is still considered (as it was for French in the past centuries, see above) a
pronunciation defect or a symptom of snobbery and affectation, more than "education".
74
As it could be clear from the survey in §3.1, traditional dialects described as having uvular r-sounds are in
northern areas (almost exclusively north-western dialects or in the bilingual areas in North-East, at the boundary
with German-speaking countries). Even if one accepts that the dialects of this domain are presenting since a time
these r-variants, there are no reason to consider them as prestige dialects. Individual burrers are known
everywhere in villages from South to North where specific burring styles are widespread and are sometimes
promoted as markers of local socio-geographical identity.
53
In most cases the sounds labelled as r moscia are even considered as "pathological".75
Canepari (1999) includes them, among pronunciation defects, in a detailed articulatory
classification (sometimes needing the introduction of non-conventional phonetic symbols):
[C]'è una certa varietà d'«erre mosce» usate in italiano per caratteristiche individuali. Ci
sono quattro tipi uvulari sonori, rispettivamente: vibrante [...], costrittivo [...],
approssimante [...] e vibrato [...]. [] è il tipo normale in lingue come il francese belga,
[] in tedesco, [ ] in francese; [ ] è un suono piú debole, che può ricorrere come
variante occasionale. [...] Altrove, comunque, possono essere piú o meno diffuse in
tutte le regioni [...]. Un altro tipo piuttosto frequente d'«erre moscia» è l'approssimante
sonoro labiodentale [] [...], che, nella variante uvularizzata [], suona rivoltantemente
snobistico in italiano (Canepari 1999:98).76
In fact , more than considering them prestige variants, different types of r moscia appear
everywhere, even in rural areas and in a social unfavoured milieu, and are often considered a
pronunciation defect. Barry (1997) highlights as the apical r-pronunciation is simply
something that a number of speakers in any country just cannot do.
In Italy and Spain, and Bulgaria, where trilled and/or flapped lingual "Rs" are de
rigueur, efforts are made at primary school level to help children with problems. A
good proportion do indeed achieve the goal, but there are always "pathological" cases
which have to resort to e.g. a "labial R" (Barry 1997:38).
Some concessions are made by Mioni (1986) who gives a reduced list of possible r-variants
and writes "tutti questi foni sono possibili sostituti di /r/ in patologia anche se l'uvulare [] è
così ampiamente diffusa tra gli italiani, che ci si può domandare se debba ancora essere
considerata come deviante" (Mioni 1986:46, n. 27).
75
A better case is perhaps presented by Widdison (1997:189) who includes Italian back r-sounds among the
cases of "deviation from the norm" (of course this applies not only to northern Italian).
76
At p. 99 (F 3.9.3-6), [ ] is defined "vibrato uvulare" whereas [] is the conventional symbol for a uvular trill.
Referring to r moscia, the author gives very useful phonetic details when he observes that these sounds "In
italiano di solito si accompagnano anche a una struttura sillabica caudata più «strascicata» /'VC/ ['VC] (invece di
['VC])". Furthermore, a better account on the conditions in which these pronunciations appear is in the following
excerpt "Non raramente alcuni tipi d'«erre moscia» sono usati volontariamente come degli xenofonemi stilistici,
parlando in italiano, anche se spesso i risultati sono ridicoli e insopportabili. Di solito, l'erre moscia dà
un'impressione d'affettazione [...] [M]olti intellettuali, veri o sedicenti, e «personaggi famosi», come per esempio
«l'avvocato Gianni Agnelli», [...] esibiscono l'erre moscia" (Canepari 1999:99-100).
54
A more tolerant opinion is also expressed by the same Canepari (1999):
in alcune zone d'Italia la realizzazione piú diffusa per /r/ è uvulare [  ], che
localmente può essere considerata quasi il tipo «normale», mentre l'articolazione
alveolare diviene minoritaria; si tratta dell'Alto Adige, della Val d'Aosta e di buona
parte della provincia di Parma. (Canepari 1999:101).77
Anyway, if I should give an estimate of the quantity of r moscia pronunciations in (urban
mainly) northern Italian, I would probably say that surely less than 10% of speakers show a
systematic recourse to this kind of (various) pronunciations (and perhaps more than 10% only
in Piedmont and in the Parma province). On the contrary, I would probably establish a
definetely upper threshold for French back-r pronunciation standing everywhere over 90%.
That in order to give an idea of the difference between the two situations.
As for the Italian back r-sounds, the origins of the irregular presence of these pronunciation
styles are rarely investigated. Migliorini (1937*1992:485) reported a source of the XVIIth c.:
La moda franceseggiante fa che parecchi accolgano la pronunzia uvulare della r alla
francese [...] (nota 146: "Bello è sentire [...] la lettera r pronunciata nell'ugola, ch'era
difetto d'organo viziato, divenuti grazie e vezzi di pronuncia in Italia" [per influenza
francese]: Gozzi, Chiacchiera intorno alla lingua litterale italiana, p. 65 Vaccalluzzo
(cfr. Lingua nostra, XVII, 1956, pp. 80-81)).
High society French models have traditionally been described as the origin of the diffusion of
back r-sounds in various central and northern European languages (see among others
Chambers & Trudgill 1980), but several authors quoted in Van de Velde & van Hout (2001),
in various contributions, claimed an older and indipendent origin for different areas (e.g.
Holland and Rhineland). The theory of the French back-r spreading could be valid for some
Italian areas but we cannot exclude other hypotheses (see above note 27).
The real problems are here the same already discussed for French. A clear classification of the
different articulations gathered in this general category (as Canepari does, see above) is
necessary, jointly with an attentive evaluation of the phonetic contexts in which different
allophones are realized by various speakers.78
77
For a "normal" diffusion of uvular r-sounds in the area of Parma see Canepari (1999:387; also see a few
comments at p. 381, raising suspicions of an important diffusion in northern Lombard provinces).
78
It is quite deceiving to read about defective r's described without any explicite mention of variants and
allophones observed for a given speaker and representing his/her own individual divergences from the normal
distributional scheme (cp. §3., Fig. 9).
55
3.3. Regional preferences and defective r-sounds
According to Canepari (1999:101-102), besides the abovementioned traditional phenomena
(see §3), one should take into account the following r-variants as typical realizations in some
regions, even though some speakers can reasonably have recourse to other choices.79
Among the most interesting regional r-sounds there are north-eastern alveolar approximants
and taps which are generally lateralized. In Venice the most common r-realization is a
postalveolar tap (something like a retroflex tap) tending to show lateralization. These sounds
are used as realizations of /r/ in almost all the positions often violating the general scheme
illustrated in §3.1.80
As far as I have been able to observe, similar sounds are also occasionally found in idiolectal
usages in Apulia and in other places on the Adriatic coast, probably in relationship with
Albanian origins (historical attested).81
Other places where liquid-r's (as we may call them on the grounds of a popular impressionist
usage) are de rigueur, as already introduced, are south-western Piedmont (with the usual []
around Frabosa and the Pamparato [] both determining varieties of those r-sounds known as
r monferrina). The same [] is a typical sound for some conservative patois speaker from
Salbertrand (in the Turin province).82
79
A single-strike articulation is widespread in northern areas in almost all the contexts (even as a /rr/ realization
in conservative accents) but Piedmont, Aosta Valley and Emilia-Romagna, have an apical trill usually uvularized
[r] [...] whereas in Liguria an alveolar uvularized tap [] seems more frequent. Also see the aforementioned
Piedmontese varieties (§3.1).
80
I use to transcribe these sounds with [], [] and [] respectively. Canepari's definitions are often more
finegrained and need additional special symbols: "Nel Veneto c'è anche un sonoro lateralizzato [ ] [...], piú
debole di [] perché senza nessun contatto (e che assomiglia abbastanza all'r inglese, che è approssimante
postalveolare lateralizzato sonoro [...]). Inoltre a Venezia c'è pure il vibratile postalveolare sonoro" (Canepari
1999:101; also cp. p. 401). Useful descriptions of r-variants are also available in Canepari (1983) provided with
two audiotapes (cp. the list at p. 260).
81
Canepari (1999:102) gives more detailed descriptions by distinguishing a voiced alveolar approximant (not
lateralized) [ ] common in Apulia (and in Albanian) from the voiced alveolar fricative tap common in the
northern Calabrian area around Cosenza.
82
According to some speakers, it does contrast with /r/ in minimal pairs such as /mar/ > [mr] 'mother' vs.
/ma/ > [m] 'Tuesday'. Western varieties in the same valley are renowned as using a different r-sound
known as r valsusina whose realizations are somewhat oscillating between [] and [z] (see below).
56
I also heard slightly different varieties of these sounds in coastal areas of Tuscany (on the
Tyrrhenian sea) which I consider in relation with what is presented in §3.1. In particular I
would like to emphasize that these r-variants are rarely perceived as marked and are usually
attributed to a regional "accent". These sounds could be described as a kind of rhotic palatal
approximant ([j] or, if it is possible, something like a retracted retroflex approximant []) and
occur as realization of /r/ in internal coda position or as the implosive phase of /rr/. They are
particularly evident in stressed syllable in casual speech of people living in the area of Pisa
and Viareggio (see an example in the next section).83
Other apparently standard pronunciations are those connected with apical trills devoicing and
with increase in the tension of the constriction (with slight retraction of the articulation place).
Central and southern Italian speakers frequently hyperarticulate r-sounds in these ways above
all in coda position before voiceless consonants by producing [r] and/or [r] (see Canepari
1999, p. 447 for Campanian Italian and pp. 440 and 445 for Abruzzo and northern Apulia
accents; also see below).
All the sounds described above are not generally considered as r moscia. In compensation,
many people belonging to these community are said to have the r moscia when not
pronouncing r's in this way or following the standard scheme.
In spite of the common idea that r moscia is a uvular r, the most common defective r-sounds
are labiodental approximants [] (often velarized []). The Neapolitan r moscia may be a
labiodental or a back r-sound but the most common (the stereotype given by the actor Totò
when playing a snobbish character) is a dental approximant (something like [] or [z] cp. the r
valsusina in footnote 82).84
83
This is not in disagreement with the examples given by Rohlfs (1949*1966:376) (see above Pisan calta, pòlta,
Ligurian (Sassello) baiba and so on) who probably represent a simple orthographic way to represent these
"liquid" sounds presenting an [i]-like vocalic component (cp. with Rossi 1974:413 describing, for the dialect of
Rossano (province of Massa), a postalveolar [r]; also cp. with Toda trills in Ladefoged & Maddieson 1996:224).
At this purpose it would be also interesting to carry out a contrastive analysis with the caipira r presented by
Head (2002, this volume) for some varieties of Brazilian Portuguese.
84
These articulations naturally appears for speakers who had problems during r acquisition and who often also
reinterpreted consonantal clusters involving r-sounds (as well as vowel coarticulation which sometimes yields to
rhotic vowels and so on). In a few Piedmontese speakers I have observed, the following cases have been verified.
Speakers presenting the labiodental approximants [], when they do not suppress the sound, they tend to
articulate the clusters /pr-/ and /br-/, in particularly prominent positions, respectively as [()] and [()] (maybe
only single-strike sounds – the former IPA-symbol for the voiceless bilabial trill was [P] and has now been
retired). That seldom happens even for Piedmontese speakers with uvular trills. These sounds are present in the
pronunciation adopted for the Italian voice of the Warner Bros' cartoon Roger Rabbit, for instance, when he says
ti prego! 'I beg you!'). As for the phonotactic properties of dental approximants I mention here just the example
57
Northern Italian speakers using a back r do not have all recourse to the same kind of
articulation but use a quite significant varieties of them. Here is a simplified list of
possibilities (also possible everywhere in Italy):
(1) speakers using a velar fricative [] present as well the unvoiced variant [x] and the
approximant variant [] in the appropriate contexts (mainly the unvoiced in voiceless
consonant context and the approximant between vowels);
(2) speakers preferring a uvular articulation may present trilled variants [] with one or more
strikes (weakened forms of these sounds are the fricative or approximant variants [] or [])
and with unvoiced allophones in voiceless contexts ([] and []; cp. §2. distributional rules
for French);85
(3) speakers having occasionally recourse to less controlled post-uvular articulations (the
same speakers of the other points above may be subject to these alternations) could give rise
to [], [], [] and many variants, often appearing as simple []-like sounds in positions where
a weakening is likely to take place (generally in coda) or where a reduction gives rise to
vocalic glides (between vowels);
(4) speakers presenting labialization and/or multiple articulation places give rise to many
other variants for velar and uvular r-sounds (see above);
(5) people affected by r moscia (that is a more or less velarized/uvularized labiodental
approximants) tend to let occasionally prevail the back articulation or to realize simple
wavings between vowels even yielding to no gesture traces at all.
of a professional speaker of the regional TV News of the National Broadcast RAI (who has been nicknamed
linguetta 'little tongue/tonguy' by local comic showmen). He frequently let the tip of its tongue come out from
the mouth while speaking (occasionally showing linguolabial contacts). I observed this phenomenon
systematically taking place during the production of the clusters -rt-, -rd-, -rl- and -rn- all normally including
apico-alveolar contacts replaced by predorsoalveolar contacts induced by a preceding interdental approximant
gestures (something like []) which is the common r-sound for this speaker.
85
A number of other possibilities arise for speakers not respecting this "natural" distribution, then generalizing
for instance [x] in all the positions or extending the allophones to both /r/ and /rr/ thus creating further
neutralizations. I would recall here the case of a southern area (northern Apulia) where among a number of
speakers using [] and [x] or [] and [], as common variants of pinched r, one may often hear some of them
only using the voiceless variants in phonetic contexts where they are not usual, thus distinguishing them from the
rest of the community.
58
3.4. Instrumental phonetic analysis
As several studies have shown (see, for instance, Vagges et al. 1978, Ferrero et al. 1979, cp.
footnote 7), Italian apical r are stable and resistant to coarticulatory effects of neighboring
sounds (this is generally true for languages with apical trills, cp. Lindau 1985, Recasens
1991:277).
I extensively studied my own pronunciation of /r/ and /rr/ which is mostly based on apicoalveolar articulations (trills and taps, following Canepari's distributional scheme for Italian see
Fig. 9), as it is shown in the static MRI in Fig. 10. Only lip-rounding seems visibly affecting
r-sounds in rounded vowel contexts (see the image of urru at the right; the uvular contact,
also appearing in the middle image, is a parasite effect of the velum position and it is not
related to any uvular activity).86
In reference to the description proposed by Recasens (1991:278), I found a slight
mediodorsum lowering associated with a significant postdorsum retraction (or rising) towards
the soft palate.87
Comparing these images with those referring to cacuminal sounds, common in my dialect and
usual in my own pronunciation, I detected an apical contact in the (post)alveolar region with
reduced retroflexion (more apparent for subapical contacts, see Ladefoged & Maddieson
1996:15, 25-27).88 For these sounds, a lowering of the predorsum is observable as well as a
raising of the dorsum associated with a general advancement of the tongue root, thus showing
a visible compression in a front region and avoiding any possible constriction in the soft
palate - uvula region (also see the velum lift associated with the pressure conditions of the
occlusion articulation, as compared to apical trills).
86
Anyway the suspicion is that, with appropriate aerodynamic and articolatory conditions, it could vibrate too.
87
Recasens (1991) describes what follows: "The multiple vibrations of the trill are easier to produce if apical
closure take place about the corner of the alveolar ridge, with the tongue predorsum and mediodorsum actively
lowered, the sides of the tongue fixed to assist the tip and the blade in the making of the alveolar closure, and the
postdorsum retracted towards the soft palate and/or the rear wall of the upper pharynx" (Recasens 1991:278).
88
Voiced cacuminal stops (with possible affrication) are an outcome of Lat. -LL- (see above) and are generalized
as realizations of -(d)dr- clusters, but voiceless stops arise only as realization of -(t)tr- clusters (here transcribed
as -t  t  - ). MRI images here reproduced refer to the holding phase of the occlusion.
59
Fig. 10. Static MRI during the holding of my own r-articulation in logatoms such as (left to right) irri, arra and
urru.[Images in Figg. 10, 11, 12 are courtesy of CHRU of Grenoble and of P. Badin - ICP].
Fig. 11. Static MRI during the holding of consonant articulations in logatoms such as (left to right) it  t  i ,
at  t  a and ut t  u .
Fig. 12. Static MRI during the holding of consonant articulations in logatoms such as (left to right) itti, atta and
uttu.
60
Cacuminal articulations are far different from the ones observable for the corresponding
dental stops in the same contexts (see Fig. 12). Acoustic data related to these sounds have
been discussed in Romano (1999) observing lower F2 targets than for dental stops (about
1700 Hz vs. 1800 Hz or more) with F3 vaguely converging towards F2. Similar transitions
have been observed in the corresponding r-variants analyzed in the present contribution and
are summarized in Fig. 18 in comparison with formant transition patterns observed for a
significant set of other Italian r-variants.
Parts of the audio data commented in this section come from samples available from projects
on spoken Italian whose audio supports have been already published (CD-ROM/DVD AVIPAPI, coord. P.M. Bertinetto and F. Albano-Leoni) or are under development (CLIPS database,
coord. F. Albano-Leoni). Other data on Piedmontese dialects come from the unpublished
audio collection of the Atlante Linguistico e Etnografico del Piemonte Orientale (ALEPO,
resp. S. Canobbio), whereas back-r sounds have been collected by myself for 6 Piedmontese
Italian speakers (four of them with back-r articulations, one with normal apical trill, and one
with a uvularized apical trill common in the region, see Canepari 1999).89
The third corpus analyzed in §3.5. is a set of sound samples from a TV corpus on defective rsounds I am collecting since 2002.
From the first, I have been able to detect at least a sample for many of the examples discussed
in the preceding sections.90 I discuss here only a few examples. Fig. 13 illustrates one
example of a rhotate postalveolar glide realizing a postvocalic implosive r (appearing in /rr/
realizations too) of a Pisan speaker (note the F3 trajectory) and one example from a Venitian
speaker presenting various retroflex sounds, among which a lateralized retroflex flap (see how
retroflex realizations affect surrounding vowels).91
89
Back-r speakers are MB35 and MT30, both male speakers of Turinese and of the Turinese variety of regional
Italian, GA38 and CL28 both female speakers; the former living and having spent most of her past life in Turin,
the latter born in Alessandria and living in Turin. The two female speakers have a passive and somewhat active
knowledge of a Piedmontese dialect but use to speak almost only Italian. Apical trills are realized by MR28
(male, living in Rorà, in the province, but working in Turin) and OR26 (female, from the province of Cuneo)
who realized apical trills with secondary uvular articulation.
90
These r-realizations have been compared with samples collected for French, German, Swedish and other
languages (e.g. Albanian, useful for comparison with retroflex realizations) and described in contrastive terms.
91
I found interesting realizations of this type in the ALEPO archive. In particular I analyzed, a speaker from
Fontane di Frabosa (Cuneo), recorded in 1985 when he was on his 60, pronouncing in patois [sui] 'sun' and a
speaker from Pamparato (Cuneo) recorded in 1987, pronouncing different r-sounds, among which, a lateralized
retroflex flap in [paple] 'eyelid' while his young interviewer reproduced [parpre].
61
In Fig. 14 are examples of different flap realizations in the same word by two female speakers
from Venice; in the second case the speaker EM21 associates a creaky voice register shift
during the r-realization.
[f
e 


i
 ]
[ f  i 
o l i
f
  o
]
Fig. 13. Examples of supposed retroflex realizations of r: (left) retracted postalveolar approximant by an Italian
speaker from Pisa (corpus AVIP, speaker p1) saying ferrovia 'railway', (right) approximant and (lateralized) flaps
by an Italian speaker from Venice (EM21, corpus CLIPS) saying frigorifero 'fridge' .
[
m
a

e
]
[
m
a
a

e e ]
Fig. 14. Examples of single-strike realizations of r by Italian speakers from Venice saying mare 'sea' (corpus
CLIPS): (left) speaker GD22, weak vowel rhotacism and alveolar flap, (right) speaker EM21, retroflex flap with
creaky voice.
62
Typically retroflexion affects F3 and F4 tightly converging towards about 2000 Hz or more,
tracking F2 which instead avoids them, staying around a frequency of 1500÷1600 Hz.92
With retracted retroflex (if they do exist, see Hamann 2001) or rhotate palatal glide, this
convergence involves even F2 which is raised to more than 2000 Hz while F3 and F4 move to
about 2400 Hz (palatal approximants without rhotacism usually have higher F3 and F4,
staying at [i]-like values). In this example the converging F3-F4 level is higher, bounded
around 2400 Hz (with a slow F4 convergence onto F3 only reached at the end of the sound
duration, cp. data for Malayalam in Ladefoged & Maddieson 1996:222).
At this purpose it is interesting to compare these data with the measures given by Rossi
(1974) for a postalveolar r described in the dialect of Rossano (in the Massa province) that
shows mean formant values for F1, F2, F3 of about 459 Hz, 1438 Hz, and 2094 Hz (before
front vowels) and 507 Hz, 1320 Hz, and 2008 Hz (before back vowels; cp. Rossi 1974:413).
At present no phonological representation has been attempted in order to discuss these
phonetic characteristics in reference to traditional arguments on retroflexion described in
literature (see Bhat 1973, 1974, revisited by recent works such the ones in Kvale & Foldvik
1992, 1995 and Hamann 2001).
The only conclusions are here addressed only in order to give a contribution to the
explanation of the historical change Lat. -ARIU/-ARIA > It. -aio/-aia through Tuscanian
conditions (whereas in other peripheral dialectal areas r has prevailed on -j-). As shown
above, a factor contributing to this change could be the retracted r-pronunciation still
surviving in some Tuscany areas.
The other realizations analyzed belong to a small corpus obtained from a list of words and
sentences containing 84 r-sounds in different positions. 6 Piedmontese Italian speakers
pronounced each a variable number of words, giving an overall number of 372 r-items (see
examples in Fig. 15).93
92
Note that in the present study, retroflex approximants have shown lower F2 targets than dental stops (1800 Hz
or more). In a survey of formant values reported in the literature for American English /r/, Espy-Wilson et al.
(1997) found typical ranges across speakers of 250-550 Hz for F1, 900-1500 Hz for F2, 1300-1950 Hz for F3
(cp. Delattre & Freeman 1968).
93
Sentences are structured in order to present in various positions phrases like quaranta lire, trenta lire and so
on, and have been only partially analyzed, whereas the wordlist, submitted to an instrumental accurate analysis,
contains the following words: bar, baracca, barra, barrata, parlata, rara, brava, barca, barchetta, apra, labbra,
ladra, carta, scartata, concerto, coperta, regolare, dottore, nervature, ricondurre, sera, serata, serranda, cura,
curato, cure, scuro, scuri, scurire, caro, carro, carri, carretta, corra, corro, corre, correte, ferro, ferroso, erre,
terrete, sotterra, sotterrare, burro, birra, sbirri, sbirrino, arriva, arrivare, provare, proprio, prendo, prendete.
63
Fig. 15. Trilled r's pronounced in the word barra 'bar' by Piedmontese speakers: (upper row) male speakers,
MB35, MT30, MR28; (lower row) female speakers, GA38, CL28, OR26. The first and the second realizations in
each row correspond to uvular trilled articulations whereas the third ones present apico-alveolar trills.
Pulse shapes vary significantly from apical to uvular trills and are even very different among
back-r speakers, probably according to uvula mass and general aerodynamics individual
conditions, with considerable effects on duration and energy closure-aperture ratio.
Though presenting an apical vibration, OR26 clearly shows a formant deviation typical of
uvular articulations. This agrees with Canepari's (1999:101-102) claim that single-/multistrike articulations widespread in northern areas (partially including Piedmont) have apico-
64
alveolar taps or trills that are usually uvularized ([] and [r], also see the aforementioned
Piedmontese varieties in §3.2).94
a   
[
i
v
a
] [
a iv
a

 e
]
Fig. 16. Trilled uvular r's pronounced in the words arriva '(s)he arrives' and arrivare 'to arrive' by GA38. The
second word also presents a uvularized labiodental approximant.
Table 4. Number of strikes for uvular trills of Italian back-r speakers fordifferent positions
'VrV
V'rV ' 'VrrV Vr'rV
'VCrV V'CrV 'VrCV Vr'CV
MB35
1,43
1,17
4,40
4,83
2,-
1,67
2,50
2,67
MT30
1,-
1,-
2,80
3,33
2,-
1,50
2,50
2,17
GA38
1,-
1,33
4,25
4,-
1,83
2,-
2,63
2,-
CL28
1,17
1,83
2,50
4,17
1,-
1,83
2,13
2,50
µ
1,15
1,33
3,49
4,08
1,71
1,75
2,44
2,33
All the back-r speakers presented at least one uvular pulse, except GA38 (who realized a
uvularized labiodental approximant in 'VrV, see Fig. 16) and CL28 (who realized a uvular
fricative in 'VCrV and 3 times in 'VrV). As already seen above with creaky voice, note that
sometimes r-sounds may also affect the general amplitude of surrounding vowels (see the
second stressed vowels in Fig. 16).
94
These articulations are auditively easily recognized for a latent diphthongization of the preceding vowel
towards an []-like sound (this appeared a number of times in the production of the sentences containing lire).
The observed conditions of velarization or pharyngealization pushed me to adopt the transcriptions I have seen
already used for Czech defective r-variants by Simackova (2002, this volume), that is [r] and [r].
65
In order to give more space to general trends, I skip here the description of individual
tendencies in Table 4 (some columns of the original table are omitted).95
There is a clear opposition between /r/ and /rr/. Most of the time /r/ is realized a single-strike
sound with a more likely trill triggering in heterosyllabic clusters, the number of contacts
becoming definitely greater in coda and word-initial positions.
Interesting phenomena are the consistent increase in the number of pulses from unstressed to
stressed and from coda to onset positions for intervocalic /r/ and /rr/. Similar behaviours are
presented in average by /r/ in heterosyllabic clusters (not for speaker GA38) and in wordmedial onset clusters (where male and female tendencies seem diverging) while initial onset
clusters (mainly available in stressed syllables in the sample) presented a number of trills
slightly higher.
Paying attention to vowel qualities, I underline the detrilling effect of a high front vowel in
the surroundings (especially if preceding) and dissimilatory effect of adjacent back-rounded
vowels on lip-rounding during the r-production. That seems particularly important
considering typical cases of reduction of some speaker.
Like many other back-r speakers, GA38 presented an allophonic rule reducing postvocalic /r/
in unstressed position to a labiodental approximant (more or less uvularized). That is very
frequent, for instance, in the infinitive suffix -are but would more rarely happen in liprounding contexts (as in nervature, dottore).
It is very interesting to underline that those speakers usually present the same kind of
reduction on the onset phase of /rr/ in the -urre suffix (as in condurre): they hold a uvular
unrounded articulation during the implosive phase (occupying the coda of the preceding
stressed syllable) and then let prevail a labial gesture yielding to a labiodental approximant
articulation starting the following syllable (/'VrrV/ -> [-ue]). This mechanism exactly
reproduces the way in which normal speakers (according to Canepari 1999:328, see §3 and
Fig. 9) realize /rr/ in /'VrrV/: a trilling phase is realized before, during the stressed coda,
followed by a tap (and often slow transitions) on the onset of the following unstressed syllable
(['VrV]). That accounts for a double (implosive-explosive) nature of these r-sounds as
described by Rousselot (1913:53, see footnote 58) which, nevertheless, can be further reduced
to a rapid single waving gesture by losing the length distinctiveness. This happens to many
Italian waving burrers, reproducing the conditions in which French speakers in §2 realized a
simple, rapid approximant in words containing -rr-.
95
Vowel qualities are ignored even though presenting significant conditionings on surrounding rhotics.
66
Other interesting comments may come from a comparison between French (presented in §2.)
and Italian back-r realizations whose transition patterns have been compared with other rvariants analyzed in the present contribution and summarized in Fig. 18.
In both cases, there are uvular articulations, but intervocalic French r's seem to be more
pharyngeal than the Piedmontese speakers here considered which have true uvular trills
(perhaps more advanced in MT30) with a vocalic-base of [a]/[æ]. French pronunciations
seems more []-like and lip-rounding during r production may appear also in not protruding
contexts.96
Once again, besides the fundamental difference in the articulatory manner (trills vs.
fricatives/approximants) back-r Italian speakers present an overall different tongue shape and
vocal tract configuration raising different vowel-coloured uvular sounds (see Fig. 17 for a
comparison of uvular r's dispersions for different speakers).97
Fig. 17. Dispersion of typical uvular r-sounds for different speakers and for different vowel contexts (e_e front
high vowels, a_a open vowels, o_o back rounded vowels). Italian speakers (empty symbols for speakers MB35,
MT30, GA38, CL28) are all Piedmontese speakers (two male and two female) with a uvular trill. French
speakers (full symbols, for speakers CT28, FD28, AP23, AL20, described in §2.5) presented always uvular
fricative/approximant articulations without occasional vibrations (larger symbols, with vertical stripes, reproduce
the ones in Fig. 5.).
96
Such as in [kadœ]/[kadœœ], cp. the realization of FD28 in Fig. 2. A rounded r-allophone must be supposed
unless admitting a final /ø/-like vowel target in standard French too, which does not seem the case.
97
As discussed in §2.5, a context-dependent vowel-colouring seems respected in French in normal styles but
could be stylistically untriggered in common "harder" r-pronunciations in young groups whose members do not
produce it when expected.
67
I reproduced all the observed formant transition patterns in Fig. 18, commented by a few
considerations in agreement with the traditional acoustic theory of standing waves (also
known as Perturbation Theory, see Fant 1960-1973, Delattre 1966, Alwan et al. 1997).
Fig. 18. Formant transitions for r-variants (VCV, with V=[a])
68
In the case of //, straight line or rising-lowering F2 transitions signal a tendency to a
pharyngealization (in this cases I also generally observed a straight line for F3 or, in any case,
an F4-to-F3 convergence towards lower values) whereas lip-rounding (common in protrusion
contexts and for speakers generalizing everywhere [o]-like uvular fricatives) seems affecting
aboveall F3 (usually significantly lowered) and F1-F2 (also tendentially lowered, especially
F2). According to the Perturbation Theory, a laryngeal stricture affects all formants by raising
them significantly.
[]-like fricatives (or approximants) usually affect the vocal folds vibration arrangement
triggering phases of creaky voice status before or during the transition onto the following
vowel often laryngealized.
In this sample, pharyngealization has appeared always very similar to velarization were it not
for the significant F1 raising of pharyngeal articulations against a mirrored lowering for
velars.98
98
Pharyngeal r-sounds are described in Jakobson (1957): "there is a tendency to emit the pharyngealized
phonemes with a protrusion and slight rounding of the lips; on the other hand, the rounded phonemes occur with
a slight narrowing of the phaynx to reinforce the acoustic effect of labialization. Whatever orifice is contracted,
there appears a concomitant velarization: it pertains not only to the pharyngealized, but also to the labialized
phonemes" (Jakobson 1957:513). An extensive acoustic account is in Contini (1983:416-419) for southern
Sardinian varieties (e.g. Quartu) as realizations of a Lat. -L-: "Caractéristique de nombreux parlers de la moitié
sud de l'île où il continue un ancien -l- intervocalique, le r pharyngal [...] connaît plusieurs variantes, y compris
chez un même locuteur. La réalisation la plus fréquente, d'une durée moyenne de 9 cs en syllabe postaccentuelle,
et de 5 cs en syllabe accentuée, présente une structure spectrale typiquement vocalique." Mean formant values
are measured before front vowels (585 Hz, 1260 Hz, 2450 Hz) and central vowels (645 Hz, 1180 Hz, 2410 Hz)
and are said to be continuing "les formants correspondants des voyelles contigües desquels ils se distinguent
généralement par une intensité plus faible". These sounds present: "(a) [...] fréquence relativement plus élevée de
F1 (entre 510 et 700 Hz) [...] (b) [...] un F2 relativement bas, pouvant descendre jusqu'aux environs de 1000 Hz".
Furthermore "Les transitions de F2 et F3 des voyelles suivantes peuvent avoir des schémas différents. Nous
avons observé les trois suivants: (a) F2 et F3 sont reliés en continuité aux formants correspondants de [ρ] [Pl. 91
sa βàρia 'la pelle'...] (b) Le F2 se divise en deux branches [n. 91: Cette transititon double apparaît uniquement
lorsque [ρ] est précédè d'une voyelle labialisée]: la partie supérieure converge vers la transition de F3. Avec [a]
les deux formants se rejoignent vers 1800 - 1900 Hz (donc au-dessous du F2 de [ρ]); avec [i] la jonction a lieu à
peu près à la hauteur du F2 de [ρ] [v. Planche 92 boρai 'voler' et s uρìa 'les olives'] (c) T2 et T3 sont convergents
et peuvent se rejoindre à la hauteur du F2 de la consonne [v. Planche 91 mèρa 'pomme']" (ibid.:419).
69
3.5. A TV corpus: connected and spontaneous speech
The corpus analyzed in this section is a set of sound samples from a TV corpus on defective rsounds I have been collecting since 2002. The aim of this section is to verify the applicability
of the analysis strategies presented in preceding sections, to a sample of connected speech and
spontaneous productions.99
More than presenting regional characteristics, often censured, TV speakers offer a significant
variability of defective r-sounds spontaneously realized in different structural positions.
The speakers presented in the selection of samples I propose in this section come from northwestern regions (MF and GL live in Milan whereas ED and SR are from Turin), but as it is
pointed out, one of them (CF, working in Bari) comes from a southern region.
Two r-sounds pronounced by the show-woman MF (about 40, working for a private TV), are
shown in Fig. 19 presenting acoustic cues of recurrent back-tongue gestures (rapidly
approximating the pharynx rather than the soft palate) she uses to realize intervocalic r
singletons (cp. the corresponding standard realization in Fig. 9).
Completely different realizations are shown by the Italian journalist ED (about 60, male
speaker of the public TV) who usually presents uvular constrictions (with tight trilling
involving a saliva fry) sometimes weakened by a further retraction who causes the lowering
of the main constriction point down to the larynx: a voiced glottal approximant frequently
appears in intervocalic position, by causing steep decreases and increases in the energy at
vowel boundaries (see Fig. 20; I observed the same mechanism for other Turinese speakers).
Another speaker having recourse to back r-articulation is GL (about 50, male speaker, now
mainly working for a private channel) who mostly uses strong single-strike uvular flaps as
intervocalic /r/ realizations (but, depending on the surrounding sounds, he can even articulate
velar approximants [] as in terremoto 'earthquake' /terremto/ realized as [teemto]
sounding almost as it was spelled *teghemoto). Like in the example in Fig. 20, depending on
the context, he sometimes let the aerodynamic process used to get the uvular vibration
generating multiple trill possibility.
Examples of true r moscia are offered in Fig. 21 by one Piedmont regional TV newsspeaker
SR (about 40, female) who has recourse to vanishing labiodental approximants (often
uvularized but never trilled) realizing r. These realizations are frequently so weakened that
they spread into the surrounding sounds (see §3.3 and r-deletion in French in footnote 23) and
99
At present, I am attending to a full description of the database by the recourse to complete measurements
including all the acoustic variables here introduced.
70
they represent a quite different way to implement r-sounds, just waving more than trilling. On
the contrary, Fig. 22 displays a selection of examples of r-realizations by CF (about 50, male,
newsspeaker of the Apulia regional public TV) who uses to produce strongly fricative (mostly
unvoiced) sounds. This is a clear example of a defective r that could hardly be defined r
moscia. Even though its realizations are mainly uvular fricatives, like other cases discussed
above, the property is here connected to a widespread unexpected devoicing jointly with noisy
velarization and pharyngealization secondary gestures.
[ ma j
a
 
i ]
[ p a  l
a  e
e
]
Fig. 19. Two examples of pharyngealized velar approximant gestures realizing r in the Italian words magari
'excl.' and parlare 'to speak' (cp. Fig. 18.) pronounced by the private TV show-woman MF (about 40).
[ t
  m i n i
i m e 
e s
e
]
[p w oun t
a
t a ]
Fig. 20. Examples of back r-sounds from two Italian journalists. Left: ED (about 60, male speaker, public TV)
saying Termini Imerese 'city name' with an unvoiced uvular trill (with salival noise) in internal coda position and
a voiced glottal approximant in intervocalic prestressed position. Right: GL (about 50, male speaker, private
channel) saying pronunciata 'uttered (f.)' with a labialized uvular (and possibly bilabial) trill.
71
[deni t

[ fu t u  o d i m i a  f j  
 i
]
i
]
Fig. 21. Two examples of vanishing labiodental approximant gestures realizing r in the word genitori 'parents'
(left) and in the phrase futuro di Mirafiori 'future of Mirafiori (placename)' (right) pronounced by SR (about 40,
female, newsspeaker of the Piedmont regional public TV).
[ ilp a de di
j
 a t s j e
l a ]
[ko mo l t i p
a

t i k o l a
 i]
Fig. 22. Examples of (mainly unvoiced) velarized/pharyngealized noisy uvular fricatives realizing r in the
phrases il padre di Graziella 'Graziella's father' (left), and con molti particolari 'with many details' (right)
pronounced by CF (about 50, male, newsspeaker of the Apulia regional public TV).
In this section, after having highlighted the difficulties of a complete and well-based
description of r-variants in this linguistic domain, I have just proposed a reference framework
for future researches. In conclusion of this paragraph, I underline once again the need to
ground these studies on solid articulatory and acoustic bases, especially referring to
distributional coarticulatory preferences of each speaker whose realizations are observed.
72
4. Conclusions
In the present study, general terminological topics have been discussed in reference to
historical and present-day representations of r-sounds in French and Italian linguistic domains
which are interested by quite different sociophonetic dynamics.
A fricative/approximant // has been described as the more common French r-sound, but a
distributional scheme has been sketched accounting for a reliable prediction of a number of
allophones. Furthermore, the need of better accounts for different tongue body shapes
associated to back r-sounds has been raised, given that different uvular fricative sounds can
be obtained with different 'vocalic bases'.
Particularly promising seems the study of different articulatory degree of reduction associated
with the timing of these segments, which could relate various r-allophones to the
aerodynamic-muscular variables and to the overall articulatory rate.
For the weakened back articulations characterizing contemporary common French, the only
articulations allowed with short temporal slots are the approximants or, at a lesser extent, the
voiced fricatives; with slightly longer slots, fricative sounds begin to dominate whereas flaps
or trills are more likely as far as the articulation time increases.
Thus, even though /r/ vs. /rr/ contrasts would seem still possible in situations of
hypercontrolled speech, the reductive nature of spontaneous speech would probably give way
to neutralization.
In Italian, where singleton/geminate contrasts are generally widespread in the phonological
system, /r/ and /rr/ are associated to different phonetic realizations often reinterpreted in
different regional varieties on the grounds of the underlying dialectal systems. Nonetheless,
the main source of r-variability is in social preferences and on pathological and first-language
acquisition difficulties.
Several varieties of unusual r-sounds are surveyed, spanning from limp r's and lifeless r's to
pinched r's and liquid r's.
On the one hand, the variety of possible slightly different sounds in the articulatory regions
interested by rhotics has been presented as significantly developped. Theories explaining the
missed passage in a given linguistic system from [r] to [], or to fricative variants of it, or to
back sounds, with the claim that a fricative sound already existed in that system are not
supported by the apparent enormous variation possibilities (for instance, vocalized
realizations of French final r have been observed as just minimal changes in the spectral
73
energy distribution probably related to rapid gestures and changes in aerodynamical
conditions which are very difficultly transcribed).
On the other hand, common r-variants used by speakers violating the normal distribution are
candidate to be stigmatized, as well as speakers with normal /r/ basic realizations ( [], [r] and
[r] for Italian, and []/[] for French), could be described as having defective r's, just
retaining their basic articulations from undergoing "normal" reduction effects.
A relevant contribution of this enquiry is to propose a workbench where to compare various rsounds for different idiolects.
Several varieties have been analyzed with an instrumental approach, showing as the Acoustic
Perturbation Theory could be appropriate for a number of examples, provided that detailed
articulatory descriptions are available and that context-dependent allophones have been
previously reviewed within the idiophonic system of each speaker observed.
An interesting verification suggested in the study of specific rhotics variants of different
languages would be testing the presence vs. absence of lip-rounding and secondary
articulations, as well as of other concomitant gestures.
With regard to the Italian situation, several characteristics have been highlighted, helping to
determine different kind of r moscia some of which encouraging the phonetic distinction
proposed in recent rhotics' literature between trilling-variants as opposed to waving- variants.
Besides the traditional parameters such as articulatory place and manner, voicing and liprounding, the role of the allophonic distributional individual scheme has been underlined,
jointly with timing specific/general properties, internal dynamics (during the sound
articulation), secondary articulations, vowel rhotacism and other effects conditioning
surrounding sounds.
74
Acknowledgements
I would like to thank Zsuzsanna Fagyal, who first encouraged me in carry on my research on
r-sounds, also accepting to discuss with me of r-representation problems, and Hans van de
Velde who first gave me the opportunity to publish my results. Part of the work reached the
suitable maturity during the researches I carried out in collaboration with Cyril Trimaille and
Patricia Lambert of the LIDILEM in the framework of the DGLFLF project on banlieue
languages. I am grateful to them for their helpful comments and suggestions. I acknowledge
people of the two laboratories where I have been working during the past years: the Institut de
la Communication Parlée (ICP) - Grenoble and the Centre de Dialectologie de Grenoble
(CDG, in particular Michel Contini) for having exposed me to so many r-arguments. I am
also indebted with Jean Pierre Lai, for having let me discover M. Pagnol's movies, especially
Marius with the Raimu's anecdote on mer/mère. I would like to thank Manuel Barbera &
Marco Tomatis of the Faculty of Foreign Languages of Turin, and the staffs of the Linguistic
Atlases ALI, ATPM and ALEPO (in particular Sabina Canobbio) for giving me the access to
some audio materials on Piedmontese r's. I am also indebted with the Synthesis team of
Loquendo Technologies Ltd. to have allowed me to have a look to their databases for
different languages.
References
Alwan A., Narayanan S. & Haker K. (1997). Toward Articulatory-acoustic Models for Liquid
Approximants based on MRI and EPG data. Part II: The Rhotics. Journal of the
Acoustical Society of America, 101, 1078-1089.
Autesserre D. & Chafcouloff M. (1998). L'extraction et la mise en relation des indices
acoustiques de /R/ en français : première étape d'une analyse de phonétique intégrée.
Travaux de l'Inst. de Ph. d'Aix-en-P., vol. 18 (cit. in Demolin 2001).
Barolo Cl., De Sio L., Mancini F., Rossi I. & Sanfaçon A. (2002). Fenomeni di
devocalizzazione in alcune lingue indoeuropee. Il fonema /r/. In: A. Regnicoli (ed.), La
fonetica acustica come strumento di analisi della variazione linguistica in Italia, Atti
delle XII Giornate di Studio del GFS (Macerata 2001), Roma, Il Calamo, 171-176.
Barry W.J. (1995). Variation in schwa + /r/ in German. Proc. of ICPhS 95 (Stockholm 1995),
2, 214-217.
75
Barry W.J. (1997). Another R-tickle. Journal of the International Phonetic Association, 27,
35-45.
Bhat D.N.S. (1973). Retroflexion: an areal feature. Working Papers on Language Universals,
13, 27-67.
Bhat D.N.S. (1974). Retroflexion and retraction. Journal of Phonetics, 2, 233-237.
Bianco E. (1981). Gli sviluppi di LL in Calabria. La Ricerca Dialettale, III, 12, 121-137.
Billiez J., Krief K., Lambert P., Romano A. & Trimaille C. (2002). "Pratiques et
représentation langagières de groupes de pairs en milieu urbain". Rapport pour
l'Observatoire des pratiques linguistiques en France – DGLFLF (Délégation Générale à
la Langue Française et aux Langues de France), Ministère de la Culture et de la
Communication, en cours de publication.
Binisti N. & Gasquet-Cyrus M. (2001). Le français de Marseille : description
sociolinguistique. Rapport pour l'Observatoire des pratiques linguistiques en France –
DGLFLF (Délégation Générale à la Langue Française et aux Langues de France),
Ministère de la Culture et de la Communication, manuscript.
Binisti N. & Gasquet-Cyrus M. (2002). Les accents de Marseille. Cahiers du français
contemporain (eds: Billiez & De Robillard), à paraître.
Bonnard E. (1982). Synopsis de phonétique historique. Paris, Sedes.
Boucher F. (1975) Aspects phonétiques de la consonne /r/ dans la langue parlée de la région
de Joliette, P.Q. Thèse de M.A. (linguistique), Université de Montréal.
Bourciez E. & Bourciez J. (1982). Phonétique française. Étude historique. Paris, Klincksieck,
(5ème éd.).
Brun A. (1931). Le français de Marseille. Marseille, Institut Historique de Provence.
Callamand M. (1982). Méthodologie de l'enseignement de la prononciation : organisation de
la matière phonique du français et correction phonétique. Paris, Clé international.
Calliope (Tubach J.P. ed.). (1989). La parole et son traitement automatique. Paris, Masson.
Calvet L.-J. (1987). La guerre des langues. Paris, Payot.
Canepari L. (1979). Introduzione alla Fonetica. Torino, Einaudi.
Canepari L. (1983). La Notazione Fonetica / Phonetic Notation. Venezia, Cafoscarina.
Canepari L. (1999). MaPI. Manuale di Pronuncia Italiana. Bologna, Zanichelli.
Caracausi G. (1986). "Cacuminalizzazione di -LL-". In: Lingue in contatto nell'estremo
mezzogiorno d'Italia. Influssi e conflitti fonetici. Suppl. Boll. del Centro di Studi Fil. e
Ling. Sic., 8, Palermo, 121-143.
76
Cárdenas D. (1958). "The geographic distribution of the assibilated R, RR in Spanish
America". Orbis, 7, 407-410.
Carduner & Hagiwara (1982). D'accord : la prononciation du français international,
acquisition et perfectionnement. New York, John Wiley & Sons.
Carter P. (1999). Abstractness in Phonology and Extrinsic Phonetic Interpretation: the case of
liquids in English. Proc. of the 14th ICPhS (San Francisco 1999), paper 0528.
Carton F. (1974). Introduction à la phonétique du français. Paris, Bordas.
Carton F., Rossi M., Autesserre D. & Léon P. (1983). Les accents des Français. Paris,
Hachette.
Castellani A. (1950). L'area di riduzione di ri intervocalico a i nell'Italia mediana. Archivio
Glottologico Italiano, XXXV, 141-166.
Catford J.C. (2001). Rs, rhotacism and paleophony. Journal of the International Phonetic
Association, 31, 171-185.
Chambers J. & Trudgill P. (1980*1998). Dialectology. Cambridge, Cambridge Univ. Press
(2nd ed.).
Chevrot J.-P., Beaud L. & Varga R. (2000). "Developmental data on a french sociolinguistic
variable : Post-consonantal word-final /R/". Language variation and change, 12,
Cambridge Univ. Press, 295-319.
Classe A. (1934). Les tendances du Parisien moderne. Le Maître Phonétique, 46, 46-47.
Contini M. (1983*1987). Etude de Géographie Phonétique et de Phonétique instrumentale du
sarde. Thèse de Doctorat d'Etat, Université de Strasbourg, 1983, 4 voll. (partiellement
publiée avec le même titre, Alessandria, Ed. Dell'Orso, 1987, 2 voll.).
Coustenoble H.N. (1945). La phonétique du provençal moderne en terre d'Arles. Hertford,
Austin.
Delattre P. (1944*1966). A Contribution to the History of "R grasseyé". Modern Language
Notes, December 1944, 562-564 (republ. in Studies in French and Comparative
Phonetics. Selected papers in French and English, The Hague, Mouton, 1966, 206-207).
Delattre P. (1962). Le jeu de transitions des formants et la perception des consonnes. Proc. of
the 4th ICPhS (Helsinki 1961), The Hague, Mouton, 407-417 (republ. in Studies in
French and Comparative Phonetics. Selected papers in French and English, The Hague,
Mouton, 1966, 276-286).
Delattre P. (1966). The Physiological Interpretation of Sound Spectrograms, PMLA, LXVI, 5
(09/1951), 864-875 (republ. in Studies in French and Comparative Phonetics. Selected
77
papers in French and English, The Hague, Mouton, 1966, 225-235).
Delattre P. (1967*1970). Des indices acoustiques aux traits pertinents. Proc. of the 6th ICPhS
(Prague 1967), Academia Publishing House of the Czechoslovak Academy of Sciences,
1970, 35-47.
Delattre P. (1969). L'/R/ parisien et autres sons du pharynx. The French Review
Delattre P. (1971). Pharyngeal features in the consonants of Arabic, German, Spanish, French
and American English. Phonetica, 54, 93-108.
Delattre P. & Freeman D.C. (1968). A dialect study of American r's by x-ray motion picture.
Linguistics, 44, 29-68.
Delattre P., Liberman A.M. & Cooper F.S. (1955). Acoustic Loci and Transitional Cues for
Consonants. Journal of the Acoustical Society of America, 27, 769-773.
Demolin D. (2001). Some phonetic and phonological observations concerning /R/ in Belgian
French. In: H. Van de Velde & R. van Hout (eds.), ‘r-atics. Sociolinguistic, phonetic
and phonological characteristics of /r/. Bruxelles, Univ. Libre de Bruxelles, ILVP, 6373.
Derivery N. (1997). La phonétique du français. Paris, Seuil.
Duchet J.-L. (1981). La phonologie. Que sais-je ?, Paris, PUF, (4ème éd. corrigée 1995).
Durand J., Laks B. and Lyche Ch. (forthcoming). La phonologie du fançais contemporain
(PFC): présentation et premiers résultats.
[http://www.univ-tlse2.fr/erss/textes/operations/OPERATIONphono/Projetfrancais.html;
http://www.up.univ-mrs.fr/veronis/Atala/jecoprus/Durand.html]
Durand M. (1929). Etude sur les phonèmes postérieurs dans une articulation parisienne.
Revue de Phonétique, 6, 241-256.
Durand M. (1960). "La spécificité du phonème : cas de /R/ et /l/". Journal de Psychologie, -.
Espy-Wilson C.Y. (1991). Consistency in /r/ Trajectories in American English. Proc. of
ICPhS 91 (Aix-en-Provence 1991), 4, 370-373.
Espy-Wilson C.Y., Narayanan S., Boyce S.E. & Alwan A. (1997). Acoustic Modelling of
American English /r/. Proc. of Eurospeech '97 (Rhodes, 1997).
Falc'hun F. (1972). de l'R gaulois à l'R français: In: A. Valdman (ed.), Papers in Linguistics to
the Memory of Pierre Delattre, The Hague, Mouton, 109-115.
Fant G. (1960). Acoustic Theory of Speech Production. The Hague, Mouton.
Fant G. (1968). "Analysis and synthesis of speech processes". In: B. Malmberg (ed.), Manual
of Phonetics, Amsterdam, North-Holland, 171-272.
Fant G. (1973). Speech Sounds and Features. Cambridge, Mass., MIT Press.
78
Ferrero F., Genre A., Boë L.J. & Contini M. (1979). Nozioni di Fonetica Acustica. Torino,
Ed. Omega.
Fouché P. (1959). Traité de prononciation française. Paris, Klincksieck.
Fougeron C. & Smith C.L. (1999). "French". In: Handbook of the International Phonetic
Association. A Guide to the Use of the International Phonetic Alphabet. Cambridge,
Cambridge Univ. Press, 78-81.
Germi C. (1968). "Etude spectrographique de /R/ et /l/ français dans la prononciation d'un
sujet Parisien". Travail d'Etude et de Recherche, Université de Grenoble (dir. Prof.
Gsell), 1968 (manuscript).
Gomez R. (2002). Sociolinguistic characteristics of rhotics in Ecuadorean Andean Spanish.
This volume.
Grammont M. (1933*1971). Traité de phonétique. Paris, Delagrave.
Grammont M. (1941). Traité pratique de prononciation française. Paris, Delagrave.
Hagiwara R. (1995). Acoustic Realizations of American /R/ as produced by Women and Men.
UCLA Working Papers in Phonetics, 90.
Hamann S. (2002). Against the articulation of palatalized retroflex rhotics. This volume.
Hamann S. (2002). Retroflexion and Retraction revised. In: T.A. Hall, B. Pompino-Marschall
& M. Rochon, ZAS papers in linguistics, 28 (10/2002),13-25.
Hammond R.M. (1998). "On the non-Occurrence of the Phone [r] in the Spanish Sound
System". Manuscript paper presented at the 2rd Hispanic Linguistic Symposium at The
Ohio State University.
Harris J. (2002). R-effects and feet in English. This volume.
Head B.F. (2002). Retroflex /r/ in Brazilian Portuguese. This volume.
Inouye S. (1995). "Trills, Taps and Stops in Contrast and Variation". Diss. at the Dept. of
Linguistics University of California, Los Angeles (on-line abstract).
Invrea R. (1936). La pronuncia settentrionale dell'italiano. Le Maître Phonétique, 53, 6-9.
IPA (1905). Exposé des Principes de l'Association Phonétique Internationale, Bourg-la-Reine
(suppl. boll. Fil. mod. 1905).
IPA (1912). The Principles of the IPA. Suppl. to the Maître Phonétique 09/10 1912 (eds.
P.Passy & D. Jones).
IPA (1949). The Principles of the IPA. London, Univ. College (reprint 1966).
Jakobson R. (1957*1962). “Mufaxxama: the ‘emphatic’ phonemes in Arabic”. In: Studies
Presented to Joshua Whatmough , The Hague, Mouton (republ. in Selected writings.
The Hague, Mouton, 1962, 510-522).
79
Jones D. & Camilli A. (1933). Fondamenti di Grafia Fonetica secondo il sistema
dell'Associazione Fonetica Internazionale, Hertford, Austin.
Jones D. (1967). The Phoneme: its Nature and Use. Cambridge, Heffer.
Jorge Morais Barbosa (1983). Etudes de Phonologie Portugaise. Évora, Universisade de
Évora, 2a ed. (1a ed. Lisbona, Junta de Investigações Científicas do Ultramar, 1965).
Kavitskaya D. (1997). Aerodynamic constraints on the production of palatalized trills: the
case of the slavic trilled [R]. Proc. of Eurospeech '97 (Rhodes, 1997).
Kohler K. (1999). "German". In: Handbook of the International Phonetic Association. A
Guide to the Use of the International Phonetic Alphabet. Cambridge, Cambridge Univ.
Press, 86-89.
Kvale K. & Foldvik A.K. (1992). The multifarious r-sound. Proc. of ICSLP 92 - 2nd
International Conference on Spoken Language Processing (Banff 1992), 1259-1262.
Kvale K. & Foldvik A.K. (1995). An Acoustic Analysis of the Retroflex Flap. Proc. of ICPhS
95 (Stockholm 1995), 2, 454-457.
Labov W. (1972). Sociolinguistic Patterns. Philadelphia, Pennsylvania Univ. Press.
Labov W. (1994). Principles of Linguistic Change. Internal Factors. Cambridge Mass.,
Blackwell.
Ladefoged P. (1968). A Phonetic Study of West African Languages. Cambridge, Cambridge
Univ. Press (2nd ed.).
Ladefoged P. & Maddieson I. (1996). The Sounds of the World's Languages. Oxford,
Blackwell.
Ladefoged P., Cochran A. & Disner S. F. (1977). Laterals and trills. Journal of the
International Phonetic Association, 7, 46-54.
Laks B. (1977). "Contribution empirique à l'analyse socio-différentielle de la chute de /r/ dans
les groupes consonantiques finals". Langue Française – Linguistique et socioling., 34,
Paris, Larousse, 109-125.
Lauret B. (2000). Interaction des aspects segmentaux et suprasegmentaux en phonétique
expérimentale contrastive et implications en enseignement / apprentissage de la
prononciation. In: E. Guimbretière (ed.), Apprendre, enseigner, acquérir : la prosodie
au cœur du débat. Rouen, Publications de l'Univ. de Rouen (CNRS), 85-133.
Laver J. (1994). Principles of Phonetics. Cambridge, Cambridge University Press.
Léon M. & Léon P. (1997). La prononciation du français. Paris, Nathan.
Lindau M. (1985). "The story of r". In: V.A. Fromkin (ed.), Phonetic Linguistics, Orlando,
Academic Press, 157-168.
80
Lindblom Bj. (1968). Temporal Organisation of Syllable Production. Speech Transmission
Laboratory (Stockholm) Quarterly Progress and Status Report, 2, 1-5.
Lindblom Bj. (1983). Economy of Speech gestures. In: P. MacNeilage (ed.), The Production
of Speech, New York, Springer, 217-245.
Lindblom Bj. (1990a). On the notion of "possible speech sound". Journal of Phonetics, 18,
135-152.
Lindblom Bj. (1990b). "Explaining Phonetic Variation: a Sketch of the H&H Theory". In:
W.J. Hardcastle & A. Marchal (eds.), Speech Production and Speech Modelling.
Dordrecht, Kluwer, 403-439.
Machuca-Ayuso M.J. (1997). "Las obstruyentes no continuas del español: relación entre las
categorias fonéticas y fonológicas en habla espontánea". Ph.D. Thesis, Universitat
Autònoma de Barcelona.
Malmberg B. (1954). Manuel de phonétique. Que sais-je ?, Paris, PUF (17ème éd. 1994).
Martinet A. (1945*1971). La prononciation du français contemporain. Genève, Droz (1ère
éd. Paris, Droz, 1945).
Martinet A. (1960*1991). Éléments de linguistique générale. Paris, Colin.
McGowan R. (1992). Tongue-tip trills and vocal tract wall compliance. Journal of the
Acoustical Society of America, 91, 2903-2910.
Meunier J.-M. Chanoine (1928). Le rhotacisme indo-européen dans les parlers du Nivernais.
Revue de Phonétique, 5, 57-69 (endnote by H. Pernot on rhotacisme in Tsakonian).
Meyer-Eppler W. (1959). Zur Spektralstruktur der /r/-Allophone des Deutschen. Acustica, 9,
247-250.
Migliorini B. (1937*1992). Storia della lingua italiana. Firenze, Sansoni.
Millardet G. (1933). Sur un ancien substrat commun à la Sicile, la Corse et la Sardaigne.
Revue de Linguistique Romane, IX, 346-369.
Millet l'Abbé (1926). Recherches expérimentales et historiques sur l'R d'une région du Haut
Berry. Paris, Librairie Philosophique J. Vrin.
Mioni A.M. (1987). "Fonetica articolatoria: descrizione e trascrizione degli atteggiamenti
articolatori". In: L. Croatto (ed.), Trattato di foniatria e logopedia. Aspetti fonetici della
comunicazione (vol. II), Padova, La Garàngola, 15-88.
Miotti R. (1998). Descrizione fono-tonetica delle varietà regionali dello spagnolo d'America e
di Spagna. Annali di Ca' Foscari, XXXVII/1-2, 393-440.
Mota Gorriz C., de la (1991). A Study of [r] and [] in Spontaneous Speech. Proc. of ICPhS
81
91 (Aix-en-Provence 1991), 4, 386-389.
Navarro Tomás T. (1916). Las vibraciones de la RR española. Revista de Filología Española,
3, 166-168.
Navarro Tomás T. (1932*1980). Manual de prononciación española. Madrid, CSIC (20th ed.)
Navarro Tomás T. (1934). "Rehilamiento". Revista de Filología Española, 21, 274-276.
Nyrop K.R. (1914). Manuel phonétique du français parlé. Copenhague, Gyldendalske
Boghandel Nordisk Forlag (Paris, Picard).
Ohala J. (1990a). Respiratory Activity in Speech. In: W.J. Hardcastle & A. Marchal (eds.),
Speech Production and Speech Modelling, Dordrecht, Kluwer, 23-53.
Ohala J. (1990b). The origin of sound patterns in vocal tract constraints. In: P. MacNeilage
(ed.), The Production of Speech, New York, Springer, 23-53.
Öhman S.E.G. (1966). Coarticulation in VCV Utterances: Spectrographic Measurements.
Journal of the Acoustical Society of America, 39/1, 151-168.
Passy P. (1932). Les sons du français. Paris, Didier.
Passy P. (1937). Une question embarrassante : quand les petits enfants apprennent les sons de
leur langue maternelle, qu'est-ce qu'ils imitent, les sons qu'ils entendent, ou les
mouvements qu'ils voyent ? Le Maître Phonétique, 58, 14-15.
Pisani V. (1964). Le lingue dell'Italia antica oltre il latino. Torino, Rosenberg & Sellier (2nd
ed.).
Pisani V. (1970). Glottologia indoeuropea. Torino, Rosenberg & Sellier.
Pons T.G. (1973). Dizionario del dialetto valdese della val Germanasca. Torre Pellice
(Torino), Collana della Società di Studi Valdesi.
Pons T.G. & Genre A. (1997). Dizionario del dialetto occitano della val Germanasca: con un
glossario Italiano-Dialetto. Alessandria, Dell'Orso.
Prairie M. (1980) La rétroflexion du /r/ en français de Montréal. Communication inédite, 48e
Congrès de l'ACFAS.
Quilis A. (1981) Fonética acústica de la lengua española. Madrid, Gredos.
Recasens D. (1991). On the production characteristics of apicoalveolar taps and trills. Journal
of Phonetics, 19, 267-280.
Recasens D. (2002). Relating vocalic assimilatory processes to patterns of C-to-C
coarticulatory direction. Evidence for the alveolar trill and dark /l/. This volume.
Renard R. (1971). Introduction à la méthode verbo-tonale de correction phonétique. Paris,
Didier.
82
Ribezzo F. (1950). Sulla originaria unità linguistica e culturale dell'Europa mediterranea. Atti
del I Congr. Int. di preistoria e protostoria mediterranea, 185-194.
Rivetta P.S. (1914). Specimen: Common Dialect of Turin (Torino). Le Maître Phonétique,
07/08 1914, 73-74.
Rohlfs G. (1949*1966). Historische Grammatik der Italienischen Sprache und ihrer
Mundarten. Lautlehre. Bern, A. Francke AG, 1949 (it. ed. Grammatica storica
dell'italiano e dei suoi dialetti. Fonetica. Torino, Einaudi, 1966).
Romano A. (1999). A phonetic study of a Sallentinian variety (southern Italy). Proc. of the
XIVth International Congress of Phonetic Sciences (San Francisco 1999), 1051-1054.
Romito L. & Belluscio G.M.G. (1996). Studio elettropalatografico dell'opposizione
fonematica [ll], [dd], [] nel dialetto di Catanzaro e [], [], [d], [] nella parlata
albanese di San Basile. Proc. of the XXIV Conv. Naz. dell'AIA (Trento 1996), 141-144.
Romito L. & Sorianello P. (1998). Ridefinizione delle consonanti retroflesse nei dialetti
calabresi. Proc. of the SILFI (Catania 1998), 12 pp, forthcoming.
Rossi M. (1974). "Description phonétique et phonologique du parler de Rossano (Province de
Massa, Italie)". Thèse d'État, Paris, 3 voll. (Contribution à la méthodologie de l'analyse
linguistique avec application à la d.escription phonétique et phonologique du parler de
Rossano. Paris, Champion, 1976).
Roudet L. (1910). Eléments de Phonétique Générale. Paris, Welter.
Rousselot l'Abbé (1891). Les modifications phonétiques du langage. Paris, Welter.
Rousselot l'Abbé (1897-1901). Principes de Phonétique Expérimentale. I, Paris-Leipzig,
Welter.
Rousselot l'Abbé (1901-1908). Principes de Phonétique Expérimentale. II, Paris-Leipzig,
Welter.
Rousselot l'Abbé (1913). Dictionnaire de la prononciation française (suite) : r double –
aspiration – éléments atones des diphtongues françaises. Revue de Phonétique, 3, 50-83.
Rousselot l'Abbé & Laclotte G. (1903). Précis de prononciation française. Paris, Welter.
Saint-Genès, de M. (1912). Tableau sommaire des sons du français. Revue de Phonétique, 2,
286-311.
Sankoff G. & Blondeau H. (2002). Solving the puzzles of the [r] -> [R] change in Montreal
French. This volume.
Sankoff G., Blondeau H. & Charity A. (2001). Individual roles in a real-time change:
Montreal (r->R) 1947-1995. In: H. Van de Velde & R. van Hout (eds.), ‘r-atics.
83
Sociolinguistic, phonetic and phonological characteristics of /r/. Bruxelles, Univ. Libre
de Bruxelles, ILVP, 141-157.
Santerre L. (1979). Le R apical montréalais en régression rapide. Prothée, 7(2), 117-131.
Santerre L. (1982) Des /r/ montréalais imprévisibles et inouïs. Revue québécoise de
linguistique, 12(1), 77-96.
Schiller N.O. & Mooshammer C. (1995). The character of /r/-sounds: articulatory evidence
for different reduction processes with special reference to German. Proc. of ICPhS 95
(Stockholm 1995), 3, 452-455.
Séguy J. (1951). Le français parlé à Toulouse. Toulouse, Privat.
Shibles W.A. (1994). The phonetics of French: Real Phonetik. Revue de Phonétique Appl.,
110, 53-85.
Simackova S. (2002). Retracted pronunciation of /r/ by Czech speakers This volume.
Solé M.J. (1999). Production requirements of Apical Trills and Assimilatory Behavior. Proc.
of the 14th ICPhS (San Francisco 1999), 487-489.
Sorianello P. & Mancuso A. (1998). Le consonanti retroflesse del cosentino: un'analisi
preliminare. Proc. of the VIIIth workshop of the "Gruppo di Fonetica Sperimentale" of
the AIA (Pisa 1997), -.
Stevens K.N. (1989). On the quantal nature of speech. Journal of Phonetics, 17, 3-45.
Straka G. (1979). Les sons et les mots. Paris, Klinksieck.
Tagliavini C. (1982*1949). Le Origini delle Lingue Neolatine. Bologna, Pàtron, 1949 (6a ed.
1972).
Thomas J.M.C., Bouquiaux L. & Cloarec-Heiss F. (1976). Initiation à la phonétique. Paris,
Presses Universitaires de France.
Tousignant C. (1983) Le /r/ en français montréalais : un champ de la variation
sociophonétique québécoise. Thèse de Ph.D. (linguistique), Université de Montréal.
Tousignant C (1987a). Relations sociales entre le /r/ vélaire et le /r/ rétroflexe anglais de
Montréal. Revue québécoise de linguistique théorique et appliquée, 6/2, 39-55.
Tousignant C. (1987b) Les variantes du /R/ montréalais : contextes phonologiques favorisant
leur apparition. Revue québécoise de linguistique théorique et appliquée, 6(3), 73-113.
Tousignant C., Sankoff D. & Santerre L. (1983) New results on Montreal French /r/.
Université de Montréal, Centre de recherche en mathématiques appliquées.
Tranel B. (1987). The sounds of French. Cambridge, Cambridge Univ. Press
Trumper J.B. & Chiodo G. (1997). A changing Europe: the presence vs. absence of drastic
events provoking or blocking internal migration and their possible contribution to
84
linguistic change or conservation. Part I: Reggio Calabria. In: K. Mattheier (ed.), Dialect
and Migration in a Changing Europe. Frankfurt, Peter Lang, 195-244.
Uldall E. (1958). American 'Molar' R and 'Flapped' T. Bolletim do Laboratório de Fonética
Experimental da Fac. de Letras da Univ. de Coimbra, 4, 103-106.
Vagges K., Ferrero F.E., Magno Caldognetto E. & Lavagnoli C. (1975*1978). Some Acoustic
Characteristics of Italian Consonants. Journal of Italian Linguistics, 3, 69-85 (paper
presented at the 8th ICPhS, Leeds 1975, preprint 23 pp.).
van Bezooijen R. (2002). Back /r/ in the urban dialect of Leeuwarden. This volume.
van de Velde H. & van Hout R. (eds.) (2001). ‘r-atics. Sociolinguistic, phonetic and
phonological characteristics of /r/. In: Etudes & Travaux, 4, Bruxelles, Univ. Libre de
Bruxelles.
Vinay J.P. (1936). Specimen: dialecte provençal des Landes et du Béarn. Le Maître
Phonétique, 56, p. 62.
Vinay J.P. (1937). Specimen: corse d'Ajaccio. Le Maître Phonétique, 59, p. 48.
Vinay J.P. (1950). Bout de la langue ou fond de la gorge? French Review, 23/6, 489-498.
Widdison K.A. (1997). Variability in Lingual Vibrants: Changes in the History of /r/.
Language and Communication, 17, 187-193.
Wioland F. (1991). Prononcer les mots du français : des sons et des rythmes. Paris, Hachette
FLE.
85