Professional Documents
Culture Documents
Squartini, M. (2008) - Lexical vs. Grammatical Evidentiality in French and Italian PDF
Squartini, M. (2008) - Lexical vs. Grammatical Evidentiality in French and Italian PDF
grammatical evidentiality
in French and Italian*
MARIO SQUARTINI
Abstract
Since the publication of Franz Boas’ work (especially Boas 1938: 133)
it has been generally acknowledged that some languages have morpho-
logical means for obligatory dedicated expression of the information
source specifying how the information has been acquired, e.g., distin-
guishing whether it has been directly witnessed (visually, auditorily, etc.)
or indirectly known either as reported information or as the result of
the speaker’s own reasoning (inferences and conjectures). With respect
to this or similar consensus definitions Aikhenvald (2003a: 19) recently
stigmatized an increasing and ‘‘gratuitous extension’’ of evidentiality out-
side its proper domain, a reminder of the misuse of ergativity in the past
decades. According to Aikhenvald (2003a) the limits of evidentiality have
been misconceived, dangerously blurring the distinction between lan-
guages where it is obligatorily expressed by dedicated grammatical mor-
phemes and languages where it sparsely appears as a secondary mean-
ing or a pragmatic extension (‘‘evidentiality strategies’’ in Aikhenvald
derived is not always a straightforward task, the most crucial point being
that in several cases the basic meaning encompasses di¤erent domains
such as tense and aspect, tense and modality, making it impossible to dis-
entangle them. The solution adopted in Dahl (1985) and pursued in Dahl
(2000: 7) is based on providing definitions of the semantic content of sin-
gle grammatical markers in a way that includes the interplay of di¤erent
semantic domains. This is not tantamount to denying the existence of a
primary meaning and the possibility of distinguishing it from a secondary
and derived one, but requires transcategorial definitions of the grammati-
cal meaning, admitting that a primary meaning can also be formed of ele-
ments derived from di¤erent semantic domains. Pursuing this methodol-
ogy further, it should not be surprising that evidentiality can be conceived
as a part of the primary meaning of a form that also encompasses other
semantic domains, especially epistemic modality. As already noted in Mi-
thun (1986: 90), the evidential ‘‘source’’ and the epistemic ‘‘probability’’
coexist in one and the same marker in di¤erent languages (in English as
well in Iroquoioan languages), which might be much more widespread
than so far assumed. In fact, the distinction between evidentiality and
epistemic modality, although in principle possible and desirable (Comrie
2000: 2), can be di‰cult to apply (De Haan 1997: 6). In this perspective
the typology of evidentiality cannot be reduced to a confrontation be-
tween languages having grammatical expression of evidentiality and lan-
guages lacking it. It seems more reasonable to imagine a set of inter-
mediate possibilities in which evidentiality blends in di¤erent ways with
epistemic modality (Plungian 2001: 354; Pietrandrea 2004: 200–203,
2005), on a parallel with what occurs with past tense and perfective aspect
(Dahl 1985: 23) that are lumped together in the semantics of a form such
as the French passé simple.
The case of the French and Italian modals devoir/dovere is particularly
interesting in this perspective since their evidential meaning as inferential
markers and their epistemic meaning as means of downgrading the speak-
er’s commitment with regard to the factuality of the situation coexist in
most cases, making it di‰cult to ascertain which is a pragmatic extension
and which is part of the basic meaning. This makes a transcategorial so-
lution admitting both evidentiality and epistemic modality as part of the
basic meaning especially viable (on the interplay of evidentiality and epis-
temic modality in the semantics of modals see De Haan 1999a, 1999b,
2001a; Mortelmans 2000; Cornillie 2004; and Squartini 2004).
The present article can thus be considered as an attempt to interpret
evidentiality as part of the basic meaning of the modals devoir/dovere þ
infinitive, which elaborates on previous research on French (especially
Dendale 1994) and Italian (Squartini 2004). Such a treatment of modals
Evidentiality in French and Italian 921
Future þ þ
DOVERE þ þ
In this section the functional distribution of two lexical items, the adverb
evidentemente ‘evidently’ and the adverbial construction a quanto pare
‘apparently’, will be presented elaborating on their inferential meaning.
What is most significant in the spirit of the present article is that the dis-
tributional patterns of these lexical items correspond to none of the cases
discussed above, neither to the modal dovere nor to the inflectional future.
This raises new issues in the classification of inferentiality that will be dis-
cussed at the end of this section as well as in Section 5.
928 M. Squartini
future þ þ
dovere þ þ
evidentemente þ
a quanto pare þ
the category self (as inferential processes produced by the speaker) and
the other (as based on external sensory evidence). In such a deadlock,
the solution here proposed envisages a combination of Willett’s type of
evidence and Frawley’s source of evidence, which can account for the pe-
culiarities of circumstantial inferences marked as [þ other] and at the
same time expressing an indirect mode of knowing. In Section 5 addi-
tional empirical arguments will be provided confirming the role of the
feature [þ other] in the functional distribution of one of the lexical items
here considered (a quanto pare) and corroborating the hypothesis of a
multifarious account in which the source of evidence and the mode of
knowing can be conceived as two interacting dimensions.
The data presented in Section 4 demonstrate that the two lexical items
considered here, a quanto pare and evidentemente, cover the same func-
tional stretch of the inferential gradient, being both restricted to circum-
stantial inferences. However, if the whole evidential domain is taken into
account, the functional distribution of the two adverbs turns out to be
distinct. In particular, a quanto pare is also compatible with reports (24),
while evidentemente disallows them (25).11 This property is tested in (24–
25) using the collocation of the two adverbs with the inflectional form of
the conditional, whose evidential reportative function has been repeatedly
pointed out in Romance (see Dendale 1993 among others and the discus-
sion in Section 6 below):
(24) A quanto pare il presidente sarebbe scomparso: lo dicono i giornali
‘Apparently the president has disappeared (be:cond disappeared):
that is what the newspapers say’
(25) ??Evidentemente il presidente sarebbe scomparso: lo dicono i
giornali
‘Evidently the president has disappeared (be:cond disappeared):
that is what the newspapers say’
The behavior of a quanto pare, neutralizing the distinction between in-
ferences and reports, is not surprising, if one considers that some of the
corresponding adverbials in other languages show the same neutraliza-
tion pattern. The French adverb apparemment, mentioned by Lazard
(2000: 214) in his list of French evidential lexemes, has been proved in
Ramat (1996: 293) as occurring as a reportative (corresponding to En-
glish allegedly) as well as an inferential marker (English evidently), and
the same neutralization is also demonstrated in Ramat’s data by the Irish
Evidentiality in French and Italian 933
construction is cosúil. However, apart from Dendale (1994: 37), who inci-
dentally notes the di¤erent behavior of the French adverb apparemment
as opposed to the modal devoir, what has not been focused on is that the
distributional patterns of a quanto pare and similar adverbs in other lan-
guages do not cover the whole domain of inferentiality, being restricted
to circumstantial inferences. Now, the distributional pattern of a quanto
pare can be connected to the conclusions reached in Section 4 with respect
to the interaction of the source of evidence and the mode of knowing.
Being restricted to circumstantial inferences, a quanto pare can be con-
sidered as marked with the feature [þ other]. Thus, it is not surprising
that it also expresses reports, which, being based on external evidence,
can also be defined as [þ other]. Such a characterization of reports as
[þ other] is obviously compatible with Frawley’s (1992: 413) original ac-
count, in which they are primary representatives of his category other.
What Frawley does not recognize is the role of the category other as a
way of connecting reports and circumstantial inferences, provided that in
his model the latter depend on the category self. Although a di¤erent ter-
minology and a di¤erent model are used, such a connection is recognized
in Plungian (2001: 352), who suggests treating reports as a special case of
inferences, which he describes as ‘‘synchronic inference’’: instead of hav-
ing access to sensory evidence, as is the case in ‘‘retrospective inferences’’,
the speaker reports a synchronic indication gathered from other people.
While in Plungian (2001) this is only a working hypothesis, his final
model preferring a solution in which reports are separated from infer-
ences (Plungian 2001: 353), the solution proposed here maintains the con-
nection throughout, considering it as another instance of the interaction
between the mode of knowing and the source of evidence. In this perspec-
tive the neutralization pattern shown by a quanto pare can be described
as a spread of the feature [þ other] across di¤erent modes of knowing,
including all cases based on an external source of evidence, i.e., circum-
stantial inferences and reports, and excluding those lacking it, i.e., ge-
neric inferences and conjectures. If compared to a quanto pare, the non-
neutralizing behavior of evidentemente can be deemed as confirming
the independence of the source of evidence and the mode of knowing.
While a quanto pare can be described exclusively by means of the feature
[þ other], irrespective of the mode of knowing, evidentemente requires a
double descriptive label, including reference to the mode of knowing (in-
ference) and the source (other).
The neutralization of inferences and reports characterizing a quanto
pare also confirms the internal flaws of most current classificational mod-
els of evidentiality, making their deficiencies even more apparent than
in the conclusions of Section 4. The problems noted in Section 4 with
934 M. Squartini
The basic result of the discussion on a quanto pare is that the source of
evidence and the mode of knowing are two independent dimensions, pos-
sibly interacting one with the other. This implies that none of the recog-
nized modes of knowing (inferences and reports) is totally equivalent to
one given source of evidence, as postulated in Frawley’s (1992: 413) clas-
sification, where all inferential processes are derived from the self source,
while all reports are derived from the category other. This result has
been made apparent considering an evidential lexical marker crosscutting
the distinction between inferences and reports. In order to demonstrate
that similar conclusions can also be extended to grammatical markers,
936 M. Squartini
the analysis will now focus on a form that also shows a neutralization
pattern of inferences and reports. This is the case of the French condi-
tional (26) whose reportative evidential function has often been pointed
out, either as a proper evidential marker (Dendale 1993) or as a ‘‘mixed
marker’’ expressing both evidential and modal meaning (Kronning 2003).
(26) Aux dernières informations, les concurrents auraient franchi le Cap
Horn (Charaudeau 1992: 464)
‘According to the latest information, the competitors have rounded
(have:cond rounded) Cape Horn’
The French conditional also has an inferential function, whose evidential
meaning has been explicitly noted in Tasmowski (2001). A significant
point is that the inferential conditional is restricted to questions (27),
which makes its distribution di¤erent from other French inferential
markers mentioned above:
(27) On sonne; serait-ce Géraldine?
‘[The bell rings] Is (be:cond) it G.?’
(Melis 2001: 75)
Interestingly this neutralizes the distinction between inferences and re-
ports in questions, given that the conditional also occurs in questions
with the speaker reporting an opinion attributed to the addressee, who is
asked for confirmation:
(28) — Ce que vous dites est terrible: l’angoisse serait le prix à payer
pour l’émancipation de l’individu?
‘What you are saying is terrible: is (be:cond) anguish the price to
pay for the individual’s emancipation?’
(Le Nouvel Observateur 1634, 1996: 14, Haillet 2001: 319)
The di¤erent function of the conditional in (27) and (28) is demonstrated
by a comparison with Italian in which the two contexts are not neutral-
ized by the same form, a conditional being required in the reportative
question (30), while a di¤erent inferential marker (the future) occurs in
(29). As shown in (29–30), the two forms have a complementary distribu-
tion in these contexts (see also Jensen 2000: 1279), which makes them
uninterchangeable.
(29) Suonano. Sarà (*sarebbe) G.?
‘[The bell rings] Is (be:fut/*be:cond) it G.?’
(30) — Ciò che lei dice è terribile: l’angoscia sarebbe (*sarà) il prezzo
da pagare per l’emancipazione dell’individuo?
‘What you are saying is terrible: is (be:cond/*be:fut)13 anguish
the price to pay for the individual’s emancipation?’
Evidentiality in French and Italian 937
followed in an attempt to find a solution that may account for the distri-
bution of the French conditional.
It must be admitted that the results reached on the behavior of a quanto
pare cannot be extended as such to the French conditional, the way the
latter crosscuts the inferential gradient being di¤erent from the former.
As shown in Section 4, the adverbial a quanto pare is restricted to circum-
stantial inferences, which are distinguished by the feature [þ other]. On
the contrary, the French conditional occurs in generic inferential (33)
and purely conjectural contexts (34):
(33) On sonne; serait-ce Géraldine?
‘[The bell rings] Is (be:cond) it G.?’
(Melis 2001: 75)
(34) On sonne. Je n’attendais personne pourtant. Serait-ce Jean
peut-être?
[The bell rings] I was not expecting anybody. Is (be:cond) it J,
perhaps?
(Tasmowski 2001: 339)
However, as repeatedly pointed out (Diller 1977; Haillet 2001) the condi-
tional is also compatible with inferential processes based on sensory evi-
dence, thus spanning the whole inferential gradient.
(35) Aurait-il froid, que je le vois si transi?
‘Is (have:cond) he cold, for I see him so numbed’
(Tasmowski 2001: 338)
Due to the absence of restrictions, it would make no sense to use the fea-
ture [þ other], which in connection with inferences accounts for the re-
striction to external sensory evidence, as is the case with a quanto pare.
However, what the French conditional has in common with a quanto
pare is the connection with the source of evidence. As noted by Aikhen-
vald (2003b: 144–145), ‘‘The use of an evidential in a question presup-
poses the questioner’s assumption about the answerer’s source of infor-
mation.’’ Therefore, if an inferential marker occurs in a question, the
questioner assumes that the answerer will be able to produce an inferen-
tial statement. More generally speaking, a question involves a switch of
the evidential responsibility from the questioner to the answerer, the for-
mer explicitly negating its role as the source of the information. This is a
marked configuration in which the speaker ceases to be the source of the
information, as is the default case, attributing it to the addressee. Using
Frawley’s features [other]/[self] such a configuration can be described
as [ self], indicating that the speaker has given up his/her role as the
source of the information. In this perspective, the divergent behavior of
Evidentiality in French and Italian 939
source from the speaker to the addressee, thus making the label [ self]
reasonable.
7. Conclusion
This article has demonstrated that the study of the lexical expression of
evidentiality is not only significant from a diachronic point of view, i.e.,
with respect to the grammaticalization path of evidential markers (Ai-
khenvald 2003a: 26), but it is also synchronically relevant in determining
a consistent classificational model of evidentiality as a linguistic category.
Crucially elaborating on the unitary treatment of lexical and grammatical
markings, the above discussion has focused mainly on the inferential sub-
domain of evidentiality, investigating its internal structure. The data pre-
sented have demonstrated the internal partition of the inferential gradient
ranging from circumstantial inferences to conjectures and including ge-
neric inferences as an intermediate step. More significantly, a preferential
connection between circumstantial inferences and reports was demon-
strated, challenging most current classifications of evidentiality and pro-
viding further evidence in favor of a distinction between the mode of
knowing and the source of evidence. More research is now needed in or-
der to test the explanatory adequacy of such a bidimensional model with
respect to the internal structure of reports. The behavior of the Romance
conditional suggests that some subdistinctions in the reportative area, in-
cluding preferential neutralization patterns with other evidential domains,
can also be accounted for by postulating an interaction between the
source of evidence and the mode of knowing.
Notes
* I am indebted to two anonymous referees for their most valuable comments. This re-
search was funded by the FIRB 2003–2005 grant L’italiano nella varietà dei testi
directed by Carla Marello (Università di Torino). Correspondence address: Mario
Squartini, Dipartimento di Scienze del Linguaggio e Letterature moderne e compa-
rate, Università di Torino, Via S. Ottavio 20, I-10124 Torino, Italy. E-mail: mario.
squartini@unito.it.
1. Although in this article attention will be paid mainly to French and Italian, some
generalizations on the grammatical expression of evidentiality can also be extended to
942 M. Squartini
Ibero-Romance languages (see Squartini 2001). As for Ibero-Romance, see also Costa
Campos (2003), Cornillie (2004), Haßler (2002, 2003).
2. See Squartini (2005) for a comparison of French and Italian with Rumanian, the only
Romance language with clear grammatical markers dedicated to evidentiality (Fried-
man 2003: 212–213).
3. Abbreviations used: fut ¼ Future, cond ¼ Conditional.
4. This inferential gradient expands on the current typological distinction between generic
and circumstantial inferences, adding conjectures as an opposite pole with respect to
circumstantial inferences. Nonetheless, this does not exclude that a wider typological
application of the gradient might require subtler distinctions in the inferential domain
(see the case of Kashaya in Oswalt 1986: 38–39, 41–42), especially in languages that
mark the type of sensory evidence (visual, auditory, etc.) for directly perceived situa-
tions (see the discussion in Anderson 1986: 279–280).
5. Pietrandrea (2004: 188) mentions some marginal cases showing collocations of dovere
with forse ‘perhaps’, which require further research.
6. Elaborating on similar French data Tasmowski and Dendale (1998: 330) suggest that
the epistemic notion of certainty can be interpreted either with respect to the speaker’s
confidence or to the communicated message, the French future expressing the former
kind of certainty and devoir þ infinitive the latter. However, referring to the speaker
as the locus where the confidence expressed by the future is formed is equivalent to ad-
mitting a di¤erent source as opposed to the external, and potentially objective, sensory
evidence on which the judgment expressed by devoir þ infinitive is based.
7. Other evidence showing that the distribution of the epistemic adverb sicuramente is
independent from the evidential interpretation of the future can be found in (i–ii),
which demonstrate that sicuramente, unlike the future, is not sensitive to the evidential
source available, being compatible both with circumstantial (i) and generic inferences
(ii):
(i) [Indicando un ragno] Attento, è sicuramente ancora vivo, perché ho visto che si
muove
‘[Pointing to a spider] Be careful! It is certainly still alive, for I saw it moving’
(ii) [Suonano alla porta] È sicuramente il postino.
‘[The doorbell rings] It is certainly the postman’.
11. Apart from the adverb asseritamente, which only occurs in juristic specialized texts
(Venier 1991: 34), in standard colloquial Italian there is no adverb restricted to repor-
tative contexts such as English allegedly, reportedly or French prétendument (Nøjgaard
1992–1995: 235–236).
12. As noted above, a way of singling out sensory evidence inferences connecting them to
reports (described as ‘synchronic inferences’) is tentatively proposed in Plungian (2001:
353), but the suggestion is ultimately dropped in favor of his alternative final model.
13. The future becomes grammatical in Italian if temporal reference is shifted to the deictic
future, while the conditional in (30) can also refer to the deictic present.
14. Note, however, that this is only a possible extension of the distinction: as also noted by
Aikhenvald (2003a: 6, 2004: 58–59) subtypes of reportivity can be distinguished on a
di¤erent basis (see also Anderson 1986: 289), as is the case of the opposition between
reported (1) and reported (2) in Uto-Aztecan (‘‘previously known to the hearer’’ vs.
‘‘previously unknown to the hearer’’, Willett 1991: 161–166). This presupposes a prag-
matic interpretation (old vs. new information) of reportative evidentials, on which see
also Blass (1989). Furthermore, two di¤erent reportative markers in one language may
also be distinguished on a textual and genre-sensitive basis, one of the two forms only
occurring in narratives (myths or stories), as is the case in Tonkawa (Aikhenvald 2004:
51).
15. Palmer’s (1986: 70, 73) distinction between quotative and report, based on data from
Hidatsa, partially overlaps Aikhenvald’s (2003a: 15) opposition, even though in Palmer
the main focus is on the ‘‘intersubjective’’ nature of quotatives (‘‘the speaker regards
what he has said to be something that everyone knows’’, i.e., including the speaker) as
opposed to the totally external character of reports (information given by someone
else, i.e., originally not shared by the speaker). For a distinction between subjective
and intersubjective evidentiality see Nuyts (2001a: 34–35, 2001b).
References
Boas, Franz (1938). Language. In General Anthropology, Franz Boas (ed.), 124–145. Boston
and New York: Heath.
Botne, Robert (1997). Evidentiality and epistemic modality in Lega. Studies in Language 21,
509–532.
Bybee, Joan L.; Perkins, Revere; and Pagliuca, William (1994). The Evolution of Grammar.
Tense, Aspect, and Modality in the Languages of the World. Chicago and London: Univer-
sity of Chicago Press.
Charaudeau, Patrick (1992). Grammaire du sens et de l’expression. Paris: Hachette.
Cinque, Guglielmo (2004). Restructuring and functional structure. In Structures and Beyond:
The Cartography of Syntactic Structures, Vol. 3, Adriana Belletti (ed.), 132–191. Oxford:
Oxford University Press.
Comrie, Bernard (2000). Evidentials: semantics and history. In Evidentials: Turkic, Iranian
and Neighbouring Languages, Lars Johanson and Bo Utas (eds.), 1–12. Berlin and New
York: Mouton de Gruyter.
Cordin, Patrizia and Calabrese, Andrea (1988). I pronomi personali. In Grande grammatica
italiana di consultazione, Vol. 1, Lorenzo Renzi (ed.), 535–592. Bologna: Il Mulino.
Cornillie, Bert (2004). Evidentiality and epistemic modality in Spanish (semi-) auxiliaries: a
functional-pragmatic and cognitive-linguistic account. Unpublished doctoral dissertation,
University of Louvain.
Costa Campos, Maria Henriqueta (2003). Les verbes modaux dans l’expression de la catégo-
rie du médiatif. In Actas del XXIII Congreso internacional de Lingüı́stica y Filologı́a ro-
mánica, Volume 2, part 1, Fernando Sánchez Miret (ed.), 185–192. Tübingen: Niemeyer.
Dahl, Östen (1985). Tense and Aspect Systems. Oxford: Blackwell.
Dahl, Östen (2000). The tense-aspect systems of European languages in a typological per-
spective. In Tense and Aspect in the Languages of Europe, Östen Dahl (ed.), 3–25. Berlin
and New York: Mouton de Gruyter.
De Haan, Ferdinand (1997). The Interaction of Modality and Negation: A Typological
Study. New York and London: Garland.
De Haan, Ferdinand (1999a). Evidentiality and epistemic modality: setting boundaries.
Southwest Journal of Linguistics 18, 83–101.
De Haan, Ferdinand (1999b). Evidentiality in Dutch. Proceedings of the 25th Annual Meet-
ing of the Berkeley Linguistics Society, Steve S. Chang, Lily Liaw, and Josef Ruppenhofer
(eds.), 74–85. Berkeley, CA: Berkeley Linguistics Society.
De Haan, Ferdinand (2001a). The relation between modality and evidentiality. Linguistische
Berichte, Special issue 9, 201–216.
De Haan, Ferdinand (2001b). The place of inference within the evidential system. Interna-
tional Journal of American Linguistics 67: 193–219.
DeLancey, Scott (1997). Mirativity: the grammatical marking of unexpected information.
Linguistic Typology 1, 33–52.
DeLancey, Scott (2001). The mirative and evidentiality. Journal of Pragmatics 33, 339–464.
Dendale, Patrick (1993). Le conditionnel de l’information incertaine: marqueur modal ou
marqueur évidentiel? In Actes du XX e Congrès international de linguistique et philologie
romanes, Volume 1, Gerold Hilty (ed.), 163–176. Tübingen and Basel: Francke.
Dendale, Patrick (1994). Devoir épistémique, marqueur modal ou évidentiel? Langue fran-
çaise 102, 24–40.
Dendale, Patrick (2001). Le futur conjectural versus devoir épistémique: di¤érences de valeur
et de restrictions d’emploi. Le français moderne 69, 1–20.
Dendale, Patrick and De Mulder, Walter (1996). Déduction ou abduction: le cas de devoir
inférentiel. In L’énonciation médiatisée, Zlatka Guentchéva (ed.), 305–318. Louvain and
Paris: Peeters.
Evidentiality in French and Italian 945