Perfective Marking in the Breton Tense-Aspect System

Abstract

The tense-aspect system of Breton, a continental Celtic language, is largely under-described. This paper has two main goals. First, it gives an overview of the numerous verbal morphosyntactic constructions of Breton, with the aim of evaluating how they carve up the tense-aspect domain. The second goal is to zero in on one particular set of constructions, namely, perfect-like constructions. In particular, it investigates the use of the present perfect in narrative and oral discourse, compared to two other competing constructions, the simple past and the past perfect. In the spirit of de Swart and Le Bruyn’s Time in Translation project, we adopt a parallel corpus-based approach from Harry Potter and the Philosopher’s Stone and its Breton translation. We develop an account of the distinction between these temporal forms, in particular the present and past perfects, drawing on the interaction between rhetorical relations and temporal structure. Results show that in written narrative stretches, the simple past is the norm; however, in dialogues, the present perfect is required in cases of ‘weak’ narration, and if the past situation is somehow felt to be currently relevant, even if the situation refers to an explicit past time. However, the past perfect occurs in narrative stretches within the dialogue, in cases of ‘strong’ narration, especially if the situation described is anaphorically tied to a temporal antecedent.

1. Introduction

In this paper we pursue two main goals. First, we give an overview of the numerous verbal morphosyntactic constructions of Breton, with the aim of evaluating how they carve up the tense-aspect domain of that language. This is important as Breton remains an under-described language in that respect. Our second, and most important objective, is to zoom in on the perfective/perfect-like tenses of Breton, as four different morphosyntactic constructions are found to correspond to the English simple past vs. present perfect opposition: a simple past (henceforth abbreviated as SPST), which is limited to written language in present-day Breton; an imperfect (henceforth abbreviated as IPF), found in narrative and descriptive modes of discourse, respectively; a periphrastic present perfect (henceforth abbreviated as PRS.PRF) used mainly in dialogues with perfect as well as perfective functions; and a periphrastic past perfect (henceforth abbreviated as PST.PRF), which functions as an ordinary anterior tense, but also takes the place of the SPST in oral narrative discourse (dialogue). This alternation between the PRS.PRF and the PST.PRF has been noted by several linguists (Favereau 1987; Davalan 2017), but to the best of our knowledge, the present paper is the first linguistic account of the different conditions of use of these two forms. We show that the switch from one construction to another is straightforwardly sensitive to the mode of discourse (Smith 2003; Le Bruyn et al. 2019), and also to the type of rhetorical relations conveyed by these tenses (Lascarides and Asher 1993; Borillo et al. 2004).

Two concepts need to be defined at this stage, narrative discourse and dialogue, as they will be important for the rest of the study. We use the term narrative discourse to refer to a “sequence of consequentially related events and states” (Smith 2003, p. 196) and assume this is the standard mode we find in the narrated parts of novels. We use the term dialogue to refer to conversations between individuals and assume conversations between fictional characters also fall under this label. Narrative discourse and dialogue are not mutually exclusive, as dialogue participants can engage in narrative discourse. We see that past events in dialogue-based narrative discourse can be strictly ordered (what Borillo et al. (2004) call “strong narration”) or more loosely ordered (“weak narration”).

To give an idea of the distinction, this passage taken from the Breton translation of Harry Potter–The Philosopher’s Stone (Rowling 1997, 2012) illustrates first the use of the SPST in written narration, then of the PST.PRF, and finally the use of the PRS.PRF, in dialogues;1 in this example, and all the examples used throughout the chapter, the English text is the original and the Breton text is the translation. Depending on the tense-aspect construction and the language discussed, we sometimes chose to present the Breton translation first, as in excerpt (1):

1.	a	Tevel a	rankas	ur pennadig,	ha derc’hel	gant he frezegenn. (…)
		quieten VRP	must-SPST	a moment	and keep-INF	with his rant
	b	Er skol	he devoa kavet	ar Potter-se, emezi.
		in school	meet-PST.PRF.3SG	this Potter, she.said
	c	Goude e	oant dimezet,	bet	ur mab dezho,
		after VRP	marry-PST.PRF.3SG have-PST.PRF.3SG		a son to.them
		ha te	an hini ‘oa.	Gouzout a-walc’h	a raen
		and you	the one was.	Know well	VRP do- IPF.1SG
		e	vijes	bet evelto	ken iskis	ken amordinal.
		VRP	be-COND.2SG	be-PST.PTCP same	as strange as	abnormal
	d	A-benn ar fin ez		eus	be	unan bennak
		in the end	VRP	be-PRS.PRF.3SG someone
		o lakaat	he c’harr	da darzhañ
		making	her car	to crash
	e	ha	degaset out bet	amañ.
		and	bring-PRS.PRF.2SG	here

She stopped to draw a deep breath and then went ranting on. (…)

‘Then she met that Potter at school and they left and got married and had you, and of course I knew you’d be just the same, just as strange, just as–as–abnormal–and then, if you please, she went and got herself blown up and we got landed with you !’2

All the verbs in the English original appear in the SPST, as we are dealing with the narration of past events. The Breton version, however, makes a distinction between the past events that belong to written narration (in the SPST, in (1a)) and the narrated events that appear in the dialogue, which can be in the PST.PRF in (1b) and (1c) or the PRS.PRF in (1d) and (1e). This example shows that the behavior of the Breton PST.PRF is different from canonical past perfects: in temporal (Reichenbachian) terms, the English pluperfect, for instance, has its Reference time R distinct from the event time E, contrary to the SPST, whose Reference time coalesces with E. The discourse role of the Breton PST.PRF, as per Caenepeel (1995) and Lascarides and Asher (1993), is different as well: whereas the discourse relations allowed by the English pluperfect are mainly Elaboration and Explanation, never Narration, examples (1b) and (1c) show that the Breton PST.PRF is naturally used for Narration (see Section 4 for a complete discussion of rhetorical relations). In this respect, the Breton PST.PRF partly conforms to past perfect forms in Romance languages (Bertinetto 2010; Becker 2020) in having acquired aoristic functions. The difference is that in Breton, the PST.PRF takes the place of the SPST in dialogue, so it is normally used to create a narrative structure.3

In order to bring to light the differences in the use of perfect-like forms in Breton, we follow the same methodology as that of de Swart and Le Bruyn’s Time in Translation project4 (and of Le Bruyn et al. 2022; Fuchs and González 2022, ‘this issue’): we added the Breton translation of Harry Potter and the Philosopher’s Stone to the existing parallel corpus in the Time in Translation database. Therefore, our data consist of three chapters from that book: a more complete presentation of the corpus is found in Section 3. In order to minimize the bias that the use of translated language may induce, we have added a monolingual control corpus: we selected the Banque Sonore des Dialectes Bretons5, which is a collection of audio recordings of native Breton speakers who tell different kinds of stories. We used it essentially to check if the differences in use between the perfect-like forms in our main corpus (which is an example of standardized literary Breton) are found in the dialects of authentic native speakers of Breton (see Section 2.1 and Section 4 for examples and discussions based on this corpus). Theoretically, we adopt a semantic-temporal as well as a discourse pragmatic account of the distinction between perfect and non-perfect tenses (Reichenbach 1947; Benveniste 1966; McCoard 1978; Inoue 1979; Lascarides and Asher 1993; Portner 2003; Nishiyama and Koenig 2010).

Let us now provide a few semantic definitions. We call “perfective” (in the sense of Klein 1994) a tense-aspect construction for which the Situation (=Event) time is included in the Topic time, and the right boundary of the Situation is specified at the Speech time. In the same vein, for Hopper (1982) and Bybee et al. (1994, p. 55), with perfectives the “situation is viewed as bounded temporally. [It] is the aspect used for narrating sequences of discrete events in which the situation is reported for its own sake.” In English, the perfective tense is the simple past; in French, both the passé simple and the passé composé in its use as a “discourse aorist” (Benveniste 1966; Riegel et al. 1994; Apothéloz 2016) are perfective tenses. Breton has three candidates for perfective: the SPST, the PRS.PRF, and the PST.PRF, as illustrated in example (1).

As in most other Indo-European languages6, the perfect in Breton consists of a periphrastic construction formed by means of an auxiliary (be) and a past participle. Its general semantics are best captured by the Reichenbachian notation E-R,S, whereby the time of the Event E, which is past, is temporally distinct from the Speech time S, which coincides with the Reference time R: the latter is the perspective from which E is viewed, and it coincides with S; the situation is perceived as currently relevant. It is generally considered that the English PRS.PRF, with its four uses (Mc Cawley 1971; McCoard 1978), exemplifies this schema. The first type is the continuative (or universal) perfect, as in I’ve been here for two days, where the eventuality extends up to the Speech time; the second is the experiential (or existential), as in Mary has visited Italy before, where the eventuality occurred at some indefinite time prior to Speech time with consequences enduring at S; the third is the resultative (or perfect of result), as in John has broken his watch, in which the telic eventuality occurred in the past with present consequences (the watch is broken at S); the fourth is the ‘Hot News Perfect’, which we will consider as a variant of the latter type in this paper, following authors such as Brinton (1988) or Michaelis (1994), who have argued that the pragmatic nature of this category does not warrant a separate category. This classification relies on temporal parameters and says nothing about the discourse pragmatic use of the perfect. Breton has a PRS.PRF form with perfect meaning like the English one, but it has broadened its use to become a perfective in certain contexts (sentences 1d and e). This last point will be the bulk of the second part of our study.

Finally, we use the term “imperfective” as a general label for a series of constructions, following Comrie (1976) and Bybee et al. (1994): imperfective is the “contrast partner of perfective” (Bybee et al. 1994, pp. 125–26), and a general definition is that it views the situation “not as a bounded whole, but rather from within” (Comrie 1976, p. 125), with two main uses, ongoing and habitual situations. Many languages also have a specific construction to indicate that “an action [is viewed] as ongoing at Reference time”, and this construction may have other, non-aspectual characteristics: progressive constructions are often used for “locat[ing] an agent as in the midst of an activity” (ibid., p. 133). Breton has a rich imperfective system: it has an imperfect7, a progressive (present and past), and a special habitual construction. The use of perfective and imperfective tenses interacts with situation aspect: the imperfect is the normal tense-aspect for states and unbounded processes in descriptions, the SPST or the PST.PRF are used to move the action forward for eventive situations, and the PRS.PRF is polysemous.

The paper is organized as follows. In Section 2 we provide a description of the main features of the Breton verbal system, followed by a more detailed presentation of its tense-aspect system. Section 3 describes the method we used for this corpus-based study and the main results, in the aim of disentangling the distribution between the three perfective forms of Breton. Section 4 offers a discussion based on those results, and we sum up our main findings in Section 5 (the conclusion).

2. Description of the Breton Verbal System

The aim of this section is to describe the morphosyntax and the semantic functions of the inflectional classes of the Breton verb, what is referred to in traditional grammar as the different “conjugations” of Breton (displegadur in Breton; Kervella 1976; Desbordes 1983). All the illustrative examples used in this descriptive section are taken from the Harry Potter database. We begin with a short subsection on the dialectal status of Breton, and carry on with the description of the inflectional classes of standard Breton.

2.1. Note on the Status of Breton

Breton is the only continental Celtic language, and it was influenced by French as early as the Middle Ages. It is known for its important dialectal variation; two main groups of dialects are generally identified: the Western family (Cornouailles, Léon, and Trégor) and the Eastern dialect (Vannetais). The phonological and morphological differences are sometimes marked, but as far as our present study is concerned (perfect-like markers), the differences are not significant, as shown in the following examples taken from The Banque Sonore des Dialectes Bretons. The first line of each example shows a sentence in the spelling of the local dialect, the second line is the standard Breton version in the spelling system called peurunvan8:

2.	a	Town of Penmarch (West):
		local spelling: bar skol neus ket desket brezoneg he
		peurunvan: e-barzh ar skol n’hon eus ket desket brezhoneg heñ
		‘At school we didn’t learn Breton, eh ?’
	b	Town of Kervignac (East)
		local spelling: ouiañ ket petra e oè digoéheit.
		peurunvan: n’ouzon ket petra a oa degouezhet.
		‘I don’t know what had happened.’
	c	Town of Locmariaquer (South-East)
		Er hig en deoé débet e oé brain.
		Ar c’hig en doa/en devoa debret a oa brein.
		‘The meat he had eaten was bad.’

Sample (2a) is an example of the present perfect (hon eus desket, “we have not learned”, reduced to neus desket in the Penmarch dialect), and (2b) and (2c) are examples of the past perfect in the Eastern dialects: the standard oa becomes oè, the inflected en doa becomes en deoé. There are other phonological and morphological variants of these forms. However, our own examination of the recordings from different areas has shown that the differences in the use of these perfect forms are marginal. It must be added that in spite of this dialectal diversity of Breton, a rich literary tradition goes back to as early as the 17th century (religious writings, poetry, plays); a turning point occurred at the beginning of the 20th century, when a standardized literary language was established, notably to promote the use of Breton in literary works and translations in the context of a general decline of Breton and a rising nationalist sentiment. The standardized spelling was introduced, which was supposed to be a compromise between the local spelling of the dialects (but there exist other interdialectal spelling systems). Mark Kerrain’s Harry Potter translation follows this tradition: the author is a Breton speaker who has written several grammar methods and has performed a number of translations, and who has adopted the literary language (called “Neo-Breton” by some). Our investigation is based on this standard literary language.

2.2. The Different Inflectional Classes of the Breton Verb

Let us first give an idea of the verbal system of Breton. It is the only Celtic language that has a periphrastic have/be present perfect form, which originated under the influence of Old French (Hemon 1975). Typologically, Breton is both considered as a Verb-Subject-Object language, which comes from Insular Celtic, and a Verb 2 language, due to its contact with Old French; Hewitt (2002, p. 8) calls this latter property the Tense 2 constraint, to point to the fact that tense often appears in the second position in the sentence (this is illustrated below). The morphosyntax of the Breton verb is dictated by the element appearing in the initial position (the “focal”9, Bottineau 2010), which triggers the use of three different “conjugations”; we take examples from our Harry Potter corpus again for illustrative purposes.

2.2.1. Simple Verb Structure

-

First, there is the basic (or impersonal) conjugation, in which the verb remains uninflected and the subject must be expressed and precede the verb; a verbal particle (a) indicates an obligatory syntactic relation (S-V or O-V).

3.	Buan e van-as		an itron kousket,
	He gwaz	avat	a chome	dihun.
	her husband	however	VRP remain-IPF.3SG	awake
	‘Mrs Dursley fell asleep quickly but Mr Dursley lay awake.’

The verb chom is uninflected for person and number (‘-e’ in chome is the invariable imperfect morpheme), its subject is he gwaz, ‘her husband’; this construction is chosen because a contrast is suggested between the lady (asleep) and her husband (awake); choice of he gwaz in the initial position (it is the focal) triggers this form of the verb.

-

Then, there is also a marked (or personal) conjugation, in which the verb bears a person and number inflection, but the pronominal subject must not be expressed preverbally (it may be cliticized after the inflection, for emphasis); the verb is preceded by the verbal particle e10 or by a conjunction (tra ma, ‘as long as’ in (4)):

4.	War greñvaat	ez ae	tra ma	choment
	on stronger	VRP go-IPF.3SG	as long as	remain-IPF.3PL
	da sellout war-du	an daou benn d’ar	straed.
	to look at	the two ends of the	street
	‘It grew steadily louder as they looked up and down the street.’

The subordinate status of the clause starting with tra ma, ‘as long as’, makes no other choice possible but to use the fully inflected form of the verb.

-

Finally, there is the auxiliary conjugation; the main lexical verb is in the first position, and the tense, person, and number inflections are carried by the dummy auxiliary ober, ‘do’. Hewitt (2002) refers to this construction as do-support; the particle a links the lexical verb and the auxiliary:

5.	Mousc’hoarzhin a reas Dumbledore o welout pegen saouzanet e oa Harry.(…)
	Chom	a	ra-e	Harry	difiñv	ha	digomz	war e	wele.
	remain VRP		do-IPF.3SG Harry		motionless and speechless on his bed.
	‘Dumbledore smiled at the look of amazement on Harry’s face. (…) Harry lay there, lost for words.’

In (5), an opposition is suggested between two situations: what Dumbledore did (smile) and Harry’s reaction (remain there). The verbal lexical predicate (the infinitive verb chom) is then the focal point (the question at this point is: ‘what was Harry’s reaction?’); in this case, the dummy auxiliary ober, ‘do’, must appear.

2.2.2. Compound Tenses

Breton has a present and a past perfect construction, formed by means of the existential form of ‘be’, eus in the present tense, oa in the past, with cliticized possessive pronouns (of the type ‘to-me is’) followed by the past participle of the verb. Perfect constructions obey the same constraints as the simple verb structure conjugations in terms of placement of the lexical verb (initial or not) and the corresponding verbal particle (a or e):

6.	Lavaret en deus:	« Ha Harry zo aet	war e lec’h, neketa? »
	say-PRS.PRF.3SG	and Harry is gone	after him, isn’t he
	‘He just said, “Harry’s gone after him, hasn’t he?”’

7.	Setu dres pezh	am eus lavaret	d’ar	c’helenner Dumbledore.
	here just thing	say-PRS.PRF.1SG to the			professor Dumbledore
	‘That’s what I said to Dumbledore.’

2.2.3. Periphrastic Structures

To these simple and compound conjugations we must add two periphrastic structures: the progressive (Hewitt 1990; Corre 2021) and the ober-activity construction (Hewitt 1990, 2002). These two constructions have specific properties which set them apart (syntactically and semantically) from the others presented above: they do not obey the Tense 2 constraint, generally have a fronted verbal particle, and are restricted to dynamic agentive verbs; they can naturally be combined with all the other tenses (present, past, perfect). In particular, the ober-activity construction is different from the ober-‘do’-support construction in example (5): the latter is available for all verbs (including state verbs), whereas the former is restricted to dynamic verbs.

The progressive is formed with the verb be in its situative form11, and a former locative preposition ouzh/o/oc’h, ‘at, on’, now a progressive particle with special mutation, followed by a verbal noun (which is the infinitive form of the verb).

8.	An holl	gelennerien all	a soñje ganto	edoSnape	o klask
	the whole teachers all		VRP think-IPF	be-IPF Snape	PROG try
	mirout ouzh ar Gripi-Aour		da c’hounit.
	stop	the Gryffindor	to win
	‘All the other teachers thought Snape was trying to stop Gryffindor winning.’

The interaction with the lexical aspect must be noted here. The progressive construction edo… o klask, ‘was trying’, is possible because the verb klask, ‘try’, is dynamic agentive; however, the verb soñj, ‘think’, a state verb, occurs in the imperfect. In analyzing our data, we will lump together the imperfect and the progressive past, as both encode either stative or stativized eventualities.

The ober-activity construction is found in the highlighted form in (9):

9.	Hejet em eus	Ron,	ur pennad brav on	bet	oc’h ober,
	shake-PRS.PRF.3SG Ron		a moment good	be-PRS.PRF.1SG PROG do
	Ha      disemplañ		en devoa graet.
	and      come.round-INF   do-PST.PRF.3SG
	‘I brought Ron round–that took a while.’

Sentence (9) reads literally: ‘I have shaken Ron, for a while I have been doing, and come round he had done.’ It contains a PRS.PRF in the progressive, which highlights the duration of the action, and then the ober-activity construction, with the main verb in the infinitive and the verb ober, ‘do’, in the PST.PRF. It is used to emphasize that the event did take place: ‘come round, that’s what he did then’. We will say no more about these constructions, which can be found in all tense-aspect forms; we mentioned them because we included them in our counts of PRS.PRT and PST.PRF forms.

There is one last construction that is often found in narrated (written) discourse and which acts as a substitute for the SPST (it is perfective): the subject of an infinitive verb introduced by the coordinating conjunction ha, ‘and’, followed by the infinitive particle da, ‘to’, and the verb itself. Excerpt (10) illustrates the alternation between the SPST and that construction:

10.	a	Lazh anezhañ	neuze, genaoueg			Echu an abadenn!”	a wic’has	Voldemort.
		kill him	then, fool			finished the game	VRP screech-SPST.3SG V.
	b	HaQuirrell da sevel	E zorn		da deurel mallozh	ar marv	warnañ,
		and Quirrell to raise-INF his hand			to perform curse	the death	on.him
	c	met Harry,	hep		gouzout dezhañ,	a lakaas
		But Harry	without	realizing		VRP throw-SPST.3SG
		E zaouarn	war dremm	ar c’helenner.
		his hands	on face		the teacher
		‘Then kill him, fool, and be done!’ screeched Voldemort.
		Quirrell raised his hand to perform a deadly curse, but Harry, by instinct, reached up and grabbed Quirrell’s face –‘

The non-finite construction is used in (10b): it suggests an automatic action from the character Quirrell, who seems to be acting on a script (he is Lord Voldemort’s puppet), contrary to Harry, who takes voluntary action against him. In (10a,c), the SPST is used. This ‘serial’ construction (Rouveret 1994) is also found in Welsh. It is present in the narration corpus, along with the SPST (see Section 3).

To conclude this section, the syntax and the range of constructions of the Breton verb are sensitive to discourse configurational parameters (what is the focal?) and to the lexical aspect of the main verb (the progressive is compatible with action verbs only, the imperfect is found with state verbs). Breton also developed several compound tenses (present and past perfects).

2.3. The Tense-Aspect System of Breton

Let us now present in some detail the tense-aspect system of Breton, in particular the different perfective and perfect-like constructions. We do so in a separate subsection for two reasons: (1) to the best of our knowledge, teasing them apart has never been conducted before, and (2) we intend to raise the issues that receive a more comprehensive account in the second part of the paper.

Contrary to English, Breton has a strongly entrenched perfective/imperfective system. Concerning the reference to past events, we find perfective and perfect tenses—the SPST is perfective, as well as the PST.PRF in its use as a conversational substitute for the simple past, and the PRS.PRF is polysemous, both a perfect and a perfective. The following excerpt from Harry Potter and its translation into Breton illustrates the morphological richness of the Breton tense-aspect system in comparison with the English original; we feature the Breton translation before the English original because we focus on the tense-aspects of Breton:

11.	a	Gantañ e		oan en em gavet
		with.him VRP		meet-PST.PRF.1SG
		‘I met him…’
	b	Pa oan	o	vale	dre ar bed.	Un den diboell a	oan
		when I.was PROG travel.INF through the world. A man foolish VRP be-IPF.1SG
		neuze, leun a vennozhioù diot diwar-benn ar mad hag an droug.
		‘… when I travelled around the world. A foolish young man I was then, full of ridiculous ideas about good and evil.’
	C	Met Lord Voldemort an hini			en deus diskouezetdin pegen bras e oa		ma fazi.
		But Lord Voldemort the one show-PRS.PRF.3SG me how big VRP was my mistake
		‘Lord Voldemort showed me how wrong I was. (…)
	d	Abaoe an amzer-se em eus e			servijet	ez-leal,
		since that time		serve-PRS.PRF.3SG.1SG faithfully
		petra bennak ma		’m eus e zilezet		dre veur a wech.
		Although	let.down-PRS.PRF.3SG.1SG many times
		‘Since then, I have served him faithfully, although I have let him down many times.’
	E	Garv ouzhin-me en deus			ranket	bezañ.”
		Hard on.me	must-PRS.PRF.3SG be.INF
		A-greiz-holl e redas			ur gridienn	dre e gorf.
		Suddenly	VRP run-SPST.3SG a shiver through his body
		“Pardoniñ ar fazioù ne ra eta es.
		‘He has had to be very hard on me.’ Quirrell shivered suddenly. ‘He does not forgive mistakes easily.’

The difference between the two languages is striking. English uses the SPST whatever the discourse mode: in narrative (written) discourse, in a perfective (in (11e), Quirrell shivered suddenly) or an imperfective (in (11b), when I travelled around the world) construal, and in the dialogue to refer to a past event (11b,c), if the situation is understood to be severed from the Speech time.12 The English PRS.PRF is used when the reference to the Speech time S is explicit: the perfect forms of the excerpt (11d,e) are of the continuative sort, denoting a situation which extends up to and overlaps S. Breton also uses the PRS.PRF in these cases. However, the difference with English is that Breton has to use a different form, namely the PST.PRF, when in a dialogue the reference to an ordered sequence of past events is explicit, as in (11a). This alternation between the PRS.PRF and the PST.PRF in dialogue will be the main focus of what follows.

3. Materials, Methods, and Preliminary Results

Our main corpus consists of three chapters from Harry Potter (volume 1, The Philosopher’s Stone), which corresponds to about 15,000 words in all. Chapter 1 consists mainly of narration/description, and Chapters 4 and 17 comprise both dialogue and narrative passages within the dialogues. For this paper, we added the Breton data to the English data which were already available (Le Bruyn et al. 2019; De Swart 2021b), except for Chapter 4, which is of our own making. We adopted the same division of the corpus into three parts as Le Bruyn et al. (2019) in the investigation of the perfect in Dutch, in order to make more explicit the tense-aspect differences for Breton.

Corpus A gathers the narrated parts of the novel, as told by the narrator, outside of any dialogue: it consists of written narration and descriptions. Corpus B and corpus C comprise the dialogues, but with a difference: corpus B is made up of dialogues in which the characters refer to (a) past event(s) either in isolation, or at least when those are not sequentially ordered. One example is sentence (11c): the event of Lord Voldemort showing Quirrell how wrong he was is a past event having consequences in the present, but it is not part of a sequentially ordered narration. In contrast, we assigned to corpus C the excerpts that feature explicitly ordered narration within the dialogues, whenever the characters tell stories in a sequentially chronological fashion, as in (11a), where the character Quirrell starts recounting his first encounter with Voldemort at the (past) time when he was young. As we see later, it is sometimes a very fine line to draw between unordered narration and sequentially ordered narration, as in actual usage a character can mix his/her reference to past situations and allow for temporal gaps. Borillo et al. (2004), building on Lascarides and Asher (1993), suggest a difference between “strong” and “weak” narration: two sentences linked by “strong” narration describe two successive events ‘of the same story’ with what they call the ‘no-significant-(temporal)-gap’ constraint13, and the constituents in this case must have a common topic (319). That would correspond to our data in corpus C. However, Borillo et al. also allow for the existence of “weak” narration: in this case, there is no requirement on the topic, and sentences do not have the no-significant-gap constraint, only vague temporal precedence (329). That corresponds to our corpus B. Therefore, all the morphosyntactic constructions describing past situations (events, processes, and states) were selected and annotated. This division of the corpus should allow us to observe fine-grained tense-aspect distributions, in particular the distribution of the PRS.PRF vs. the PST.PRF.

In this section, we provide a quantitative and qualitative analysis of the results of our corpus data. The other corpus is the Banque Sonore des Dialectes Bretons, which we mentioned in Section 1 and Section 2. The aim of this collection of naturalistic data is to make “available on the Internet a wide range of recordings of traditional Breton-speakers”14. It is made up of audio extracts resulting from investigations led by specialists of Breton (linguists, students, etc.) starting in the year 2000; it consists of hundreds of hours of recordings of native speakers on various topics, across all areas of Brittany. The transcriptions that are provided adopt the local spelling as well as the peurunvan (the standard spelling), to make them more easily accessible; sometimes, the informants were asked to provide short translations, in order to further investigate the peculiarities of their dialect. As stated in Section 1, we used this corpus for control, to make sure that the perfect-like forms (PRS.PRF and PST.PRF) are used in the same way as in the literary translation of Harry Potter. A few examples have also been selected in the Discussion.

3.1. Corpus A: Narrative Discourse

In this sub-corpus, English has 1033 verb forms referring to a past situation, and Breton has 1014 (the difference in the number of forms lies in the differences in the translations, notably the frequent use of verbless sentences in Breton). The following graphs show the distribution of the forms for each language.

As Figure 1 shows, it comes as no surprise that English overwhelmingly uses the SPST (88% of all occurrences of past situations), for states and events alike; the small proportion of progressive pasts (6%) concerns processes or so-called ‘inactive actions’ (Croft 2012), i.e., verbs such as sit, stand, lie, wear, etc., that are semantically stative but still compatible with the progressive. The few PST.PRF forms (6%) all refer to events or processes that are anterior to the narrated situation. These data confirm the neutral aspectual function of the English SPST.

Breton offers a very different picture: Figure 2 shows that the SPST (46%) is used for perfective situations; imperfective situations (states and ongoing processes) are expressed by means of the imperfect and/or progressive past (38%). The PST.PRFs in this corpus refer to anterior situations. The main surprise comes from a special non-finite construction (10%), which is a substitute for the SPST (it is perfective). This non-finite construction, featuring the coordinating conjunction ha, ‘and’, followed by the verb in the infinitive, might correspond to a case of a non-aspectual construction being used as an aspectually perfective construction (see Koss et al. 2022, ‘this issue’).

3.2. Corpus B: Dialogue with Weak Narration

This corpus, made up of the past forms occurring in dialogue in cases of ‘weak’ narration (see the beginning of Section 3 for definitions), contains an equal number of forms for the two languages (302 for English, 301 for Breton).

Here again, English uses the SPST (73%) in dialogues that recount past situations; the PRS.PRF accounts for only 22% of all past-denoting constructions; the progressive past (3%) is marginal. The distribution of constructions in Breton is different: the SPST is almost excluded, as this is oral language. The number of imperfects/progressive pasts for states or ongoing processes is very similar to what we found in corpus A. Careful observation of the translational equivalences allows us to establish the following correspondences:

• English SPST → Breton PRS.PRF/imperfect (states)/PST.PRF
• English PRS.PRF → Breton PRS.PRF

These correspondences show that the PRS.PRF in Breton is the dominant form in this sub-corpus (47%), where English has a majority of SPST and fewer PRS.PRFs. In order to look at the differences between the English and the Breton PRS.PRF, we counted the occurrences of the types of perfects for each language. Figure 3 lists 67 occurrences of the English present perfect in corpus B vs. 140 in Breton in Figure 4; 8 of these occur in non-finite constructions (following a modal or occurring in the gerund) but were translated to finite PRS.PRFs in Breton.

Table 1. Breakdown of perfect and PRS.PRF types in English, in corpus B.

Perfect Type:	Number of Tokens:
Present—Continuative	26
Present—Experiential	5
Present—Resultative	28

classifies the other 59, following the breakdown into types of perfects as proposed by Brinton (1988) and Michaelis (1994).

In Breton, we find twice as many occurrences of the PRS.PRF as in English, thus confirming that the Breton PRS.PRF is a real perfect, but a much more “liberal” one than the “classical” English present perfect (De Swart 2021a). We note that it has also developed a ‘weak’ narrative use, not unlike the French passé composé, i.e., it is generally used to tell about past events considered in isolation or vaguely connected together temporally (see examples 15 and 16 below); one challenge is to distinguish this weak narrative use from the ‘strong’ narrative use of the PST.PRF in corpus C. The following table proposes a breakdown into the types of PRS.PRF uses in Breton.

First, concerning the continuative type, we note that in Breton the number of tokens is lower than in English: continuative meaning if most often rendered by means of the present tense, as we noted in Section 1. However, the experiential and resultative Breton PRS.PRF outnumber those in English. This requires an explanation.

The main reason why we counted more experiential uses of the PRS.PRF in Breton than in English is that in the latter, the SPST frequently has experiential uses, especially in dialogues. The English SPST can host a lot of different meanings: it can refer to a single recent event as well as an indefinite past event having relevance at the Speech time, as has been frequently observed in the literature (Partee 1973; Schaden 2008; among others, Zhao, ‘this issue’)15. Here are a few examples of this quasi-experiential use of the English SPST, with their Breton translations using a PRS.PRF:

12.	‘Did you never wonder where yer parents learnt it all ?’
	‘Feusket klasket	biskoazh	goût
	try-PRS.PRF.2SG NEG	never	know
	p’lec’h ’oa bet	da dud		‘tiskiñ	‘n traoù-se ?
	where be-PST.PRF.3SG	your parents		PROG learn	those things

13.	‘An’ it’s your bad luck you grew up in a family o’ the biggest Muggles I ever laid eyes on.’
	Feus	ket a chañs	‘vezañ degouezhet e tiad	ar gwashañ Mougouled
	you.have NEG the luck		be raised in family	the biggest Muggles
	zo bet	biskoazh	er vro.
	be-PRS.PRF never		in.the country

In these examples, the SPST is truth-conditionally equivalent to a PRS.PRF (Have you never wondered, I’ve ever laid eyes on…). Breton uses the PRS.PRF, which denotes an event that occurred at some indefinite past time but somehow is felt to have relevance at Speech time; the SPST is not possible because this is dialogue, and this automatically increases the number of PRS.PRFs in the Breton corpus.

Let us now turn to the resultative type. Breton has a lot of resultative PRS.PRFs (81 vs. 28 for English). In many cases where English uses a resultative PRS.PRF, Breton does so too:

14.	‘He’s going to Stonewall High and he’ll be grateful for it. I’ve read those letters and he needs all sorts of rubbish’
	Mont a raio da skolaj ar Poull-Fank, ha gwelloc’h dezhañ bezañ anaoudek.
	Lennet em eus	ho lizhiri	ha		gwelet em eus
	read-PRS.PRF.1SG their letters		and		see-PRS.PRF.1SG
	peseurt garzaj	en dije	da brenañ.
	What rubbish	he would.have		to buy

The English and Breton PRS.PRFs have a similar function; for Uncle Vernon, knowing the content of the letters helps him make his point: Harry is not going to the school of wizardry. The English PRS.PRF has an explanatory function (having read the letter explains why it is better for Harry to go to a normal school). However, in English, the simple mention of an isolated past event, as in (15), even if it has present consequences, triggers the SPST, contrary to Breton, in which the SPST is impossible for reasons of register again (it is restricted to written narration). The Breton PRS.PRF functions like the French passé composé as a discourse aorist: the pre-hodiernal past events have present consequences, and these are considered as loosely relevant in the discussion of the present situation.

15.	‘It was on their news.’ She jerked her head back at the Dursleys’ dark living-room window. ‘I heard it.’
	Er c’heleier zoken ez eus bet kaoz eus se.” Gant he fenn e tiskouezas prenestr saloñs an tiegezh Dursley, a oa en deñvalijenn.
	“Kement-se	am eus klevet	ma-unan.
	that.much	hear-PRS.PRF.1SG	myself

This discourse aorist function of the Breton PRS.PRF explains why we counted 20 occurrences of the ‘weak’ narrative type (see

Table 2. Breakdown of perfects and PRS.PRF types in Breton, in corpus B.

Perfect Type:	Number of Tokens:
Continuative	13
Experiential	19
Resultative	81
’weak’ narration	20
Other (infinitive, conditional, future)	5

), which can be illustrated by the following two examples; this time, we provide the translation before the original for ease of exposition:

16.	a	Ha	pelec’h hoc’h eus	kavetar marc’h-tan-se?”
		and	where get-PRS.PRF.2PL this motorbike
		“Amprestet eo,	Kelenner	Dumbledore, Aotrou,”
		borrowed it.be-PRS.3SG	Professor	Dumbledore sir
		a respontas ar ramz, en ur ziskenn goustadik, gant evezh, diwar ar marc’h-tan.
	b	Digant ar skoliad Sirius Black,	a zo	bet tapetganin	hiriv.”
		from the pupil Sirius Black VRP obtain-PASS.PRS.PRF.3SG by.me			today
		‘And where did you get that motorbike?’
		‘Borrowed it, Professor Dumbledore, sir,’ said the giant, climbing carefully off the motorbike as he spoke. ‘Young Sirius Black lent it me. I’ve got him, sir.’

The PRS.PRF in Breton is used because at the Speech time, Hagrid has obtained the motorbike16. This example is of the resultative type: the sentence (I) borrowed it is rendered by means of a present resultative construction in Breton (16a) and the translator added the deictic adverb hiriv, ‘today’, in the last line (16b) with the PRS.PRF. We judged that this was an occurrence of a ‘weak’ narration because the three events (obtaining the motorbike, borrowing it, and lending it), although illustrating a common topic, are not sequentially ordered. Example (17), in which the PRS.PRF is compatible with an adverbial clause headed by pa, ‘when’, illustrates a past perfective use and is more difficult to handle.

17.	Touet hon eus,	pa hon eus en kemeret	ganimp,
	swear-PRS.PRF.1PL when take-PRS.PRF.1PL		with.us
	lakaat fin	d’an drocherezh-se,
	put end to	that rubbish,
	eme Donton Vernon. Touet		disober eus an holl draoù-se.
	said Uncle Vernon. swear-PRS.PRF to		undo of all these things
	‘We swore when we took him we’d put a stop to that rubbish,’ said Uncle Vernon, ‘swore we’d stamp it out of him.’

Here, Uncle Vernon explains what happened when Harry was entrusted to them, and he insists on the fact that they vowed to stop him from doing magic. In spite of the presence of the temporal conjunction pa, ‘when’, we note that the two events (we swore… took him) are not ordered temporally, they are part of a set of connected situations that obtained at the time; moreover, the situations described still have present consequences (Uncle Vernon still refuses to allow magic for Harry)17. It is the reason why we assigned examples like these to corpus B and not corpus C, but we are fully aware that it is a very fine distinction to make.

We now turn to our last sub-corpus.

3.3. Corpus C: Mini-Discourse in Dialogue

This dataset, which gathers the passages of ‘strong’ narration within the dialogues (i.e., when the characters tell stories to each other in a sequentially ordered manner), contains 80 forms for English and 81 for Breton.

Once again, Figure 5 shows that English uses the SPST (90%) in dialogues that narrate past situations; the PRS.PRF is absent, and the progressive past (7%) is marginal. The distribution of constructions in Breton is as follows (in Figure 6): contrary to what we found in corpus B, the PST.PRF is dominant (44%), the SPST is extremely marginal (just one occurrence), the imperfect/progressive past is found for states or ongoing processes (27%), and the PRS.PRF is used in 24% of the cases. In short, the proportions (PR.PRF vs. PST.PRF) are reversed in comparison to corpus B. What must be noted, though, is that the PRS.PRF is still used in cases of ‘weak’ narration. The following excerpt illustrates this alternation between the two constructions in an oral narration of past events:

18.	a	“Hervez a gonter,	e oa en em gavet	Voldemort	dec’h	e Godric’s	Hollow
		as they say	VRP turn.up-PST.PRF.3SG	Voldemort	yesterday	in Godric’s	Hollow
	b	Aet e oa	da-gaout	ar Bottered.	Hervez	ar vrud
		go-PST.PRF.3SG	to get	the Potters.	According.to	the rumour
		eo...	marv...	Lily ha	James Potter.
		be-PRS.3SG	dead	Lily and	James Potter.
	c	“Gwashoc’h zo	c’hoazh. Lavaret e vez	en devoa klasket	lazhañ o mab		Harry
		Worse there.is	still said it.is	try-PST.PRF.3SG	kill	their son	Harry
	d	met ne oa ket		betgouest	d’ober.
		but NEG be-PST.PRF.3SG		able	to do.
		Ne	oaket deuet a-benn	da lazhañ	ar paotrig-se.
		NEG	manage-PST.PRF.3SG	to kill	that little.boy
	e	“Ha gwir eo se?” a valbouzas ar gelennerez McGonagall.
		“Goude an holl	daolioù en deus graet...	An holl dud	en deus lazhet...
		after the whole things do-PRS.PRF.3SG the whole people kill-PRS.PRF.3SG
	f	n’ eo ket	betgouest	da lazhañ	ur paotrig?
		NEG be-PRS.PRF.3SG		able to kill	a little.boy
		Sed a zo saouzanus... netra ne c’halle mirout outañ a-raok...
	g	Met penaos	ma Doue en deus gallet Harry chom bev?”
		But how	my God can-PRS.PRF.3SG Harry stay alive

‘What they’re saying,’ she pressed on, ‘is that last night Voldemort turned up in Godric’s Hollow. He went to find the Potters. The rumour is that Lily and James Potter are–are–that they’re–dead.’ (…) ‘That’s not all. They’re saying he tried to kill the Potters’ son, Harry. But–he couldn’t. He couldn’t kill that little boy. (…)

‘It’s–it’s true?’ faltered Professor McGonagall. ‘After all he’s done … all the people he’s killed … he couldn’t kill a little boy? It’s just astounding … of all the things to stop him … but how in the name of heaven did Harry survive?’

When Prof. McGonagall tries to reconstruct the events that supposedly happened on the fateful night Voldemort killed Harry’s parents, she uses the PST.PRF, in (18a–d). The temporal localization of those past events is explicit: the specific temporal adverbial dec’h, ‘yesterday’, appears in (18a), and we understand that event to be pragmatically presupposed (cf. the introductory phrase Hervez a gonter, ‘what they’re saying’). However, afterwards, starting from sentence (18e), she shifts to the PRS.PRF: she goes from a narration of last night’s events to a present evaluation of those events; the two PRS.PRFs in (18e) are real perfects of result, and the one in (18f) is a repetition of the PST.PRF in (18d). The passage from (18a) to (18d) illustrates what we mean by ‘strong’ narration: the events follow each other sequentially (Voldemort turning up > going to find the Potters > trying to kill Harry > failing to do so). Sentences (18e) to (18g) illustrate ‘weak’ narration: Voldemort doing many things <> killing many people <> not being able to kill Harry <> Harry surviving.

In order to account for the difference between the PST.PRF and the PRS.PRF in this dataset, we hypothesized that if the PST.PRF is indicative of a ‘strong’ narrative function, it should be more compatible with specific time adverbials and when-clauses than the PRS.PRF18. Therefore, we counted the number of passages/sentences that contain those elements in corpora B and C. Out of the 30 sentences that contain either a past time adverbial (of the yesterday type) and/or a when-clause, 24 occur with a PST.PRF and 6 with a PRS.PRF, thus showing a clear preference for the former.

➢

With the PST.PRF:

-

The main locating specific past time adverbials found are: dec’h, ‘yesterday’; ur wech e oa, ‘there was once’; en noz-se, ‘that night’.

-

The when-clause types are: (un deiz) pa, ‘(one day) when’; abaoe an deiz ma, ‘since the day when’; diwezhañ ma, ‘the last time that’; en noz end-eeun ma, ‘the same night when’.

-

The connectives are: a-raok, ‘before’; (ha) goude, ‘(and) afterwards’; neuze, ‘then’.

➢

With the PRS.PRF:

-

Specific past time adverbials: n’eus ket pell zo, ‘not, long ago’.

-

A few connectives: da gentañ, ‘at first’; neuze, ‘then’; a-benn ar fin, ‘in the end’.

With all these results, we can now turn to the Discussion and offer a more refined discourse analysis of the Breton PRS.PRF and the PST.PRF. For this purpose, we rely on several works in SDRT, starting from Lascarides and Asher (1993) to more recent developments (Borillo et al. 2004), to zoom in on the interaction between rhetorical relations and temporal structure.

4. Discussion

Our results confirm that a narrative presentation of past (pre-hodiernal) events is ensured preferably by means of the PST.PRF. However, as excerpt (18) has shown, the PRS.PRF is also possible with some of the adverbials indicated above, notably ago. PRS.PRF forms with dec’h, ‘yesterday’, are actually frequent in spontaneous speech. Here are two further examples from our corpus of naturalistic data: (19) and (20) are the recordings of a Breton-speaking woman talking about her everyday life.

19.	a	Me a	wele	dec’h	war ar	journal,
		I VRP	see-IPF.3SG	yesterday	in the	paper
		hiziv ne	‘m eus ket bet	amzer	da lenn,
		today NEG have-PRS.PRF.1SG time			to read
	b	mes dec’h	em eus lennet		un tammig ha	neuze...
		but yesterday read-PRS.PRF.1SG			a bit and	then
		beñ	int	en em glemm	dija,
		well	they	complain-PRS	already…
		‘I saw that yesterday in the paper, today I didn’t have time to read, but yesterday I read a bit, and then… they’re already complaining…’

20.	a	Ma niz eo,	hag … a	oa dle dezhañ	dont
		my nephew it.is and VRP		must-IPF.3SG come
		abalamour	eñ en deus komañset	troc’hañ	an han-,
		because	he begin-PRS.PRF.3SG	cut	the hedge
		ma hae neuze evit serriñ tout an traoù…
	b	Mes eñ n’ en doa ket		telefonet	din	an deiz,
		but he NEG		telephone-PST.PRF.3SG to.me that day
		heu, ma, pas	dec’h,	an deiz e-raok…
		Er but NEG	yesterday	the day before

‘Itssss’s my nephew, and… He was supposed to come because he’s started cutting the… my hedge, to collect the whole thing… But he didn’t call me on that day, eh, not yesterday, but the day before.’

In (19b), the PRS.PRF is fine with dec’h because the event of reading is still vivid in the woman’s mind. The time when she read the paper does not matter per se: what is salient at the Speech time is the content of what she (has) read. In (20b), however, the woman is trying to recall on which day she got a call from her nephew (yesterday or the day before), which means that the time when this occurred matters and is clearly dissociated from now, so the same woman switches to the PST.PRF. Another interesting example from Harry Potter helps us zero in on the semantics of the Breton PRS.PRF when it occurs in what corresponds to a when-clause in the English original:

21.	a	Neuze e	oamp aet	d’an daoulamm da borzh ar c’haouenned
		then VRP        go-PST.PRF.1PL		at a run	to port of owls
		da gas un tamm lizher d’ar c’helenner Dumbledore,
	b	Ha kavet hon eus	anezhañ	war an hent, en Trepas Bras.
		and find-PRS.PRF.1PL	him	on the road in Entrance Hall
		‘… and we were dashing up to the owlery to contact Dumbledore when we met him in the Entrance Hall.’

The English when-clause (when we met him) is not translated by means of a when-clause in Breton, but by a coordinated clause: the sentence literally reads: ‘Then we had gone to the owlery to contact Dumbledore, and we have found him in the Entrance Hall.’ In this excerpt, the character (Hermione) seems to be summing up a list of recent events that she finds relevant. In the Breton translation, the whole passage can be read like a list (a juxtaposition of significant events), not a real story19. De Swart (2007)’s use of the Continuation relation for the English PRS.PRF gives rise to an unordered list of events with no temporal ordering involved, which is exactly what we see with the Breton PRS.PREF in (21b).

We can now suggest a discourse and temporal function for the Breton PST.PRF vs. PRS.PRF distinction, using Lascarides and Asher (1993)’s theoretical framework, supplemented by Borillo et al. (2004). In their 1993 paper, Lascarides and Asher study the discourse structure of tensed clauses in English by resorting to several rhetorical relations (namely, Narration, Explanation, Elaboration, Background, and Result). They argue that the English SPST and PST.PRF have a similar temporal structure at the sentential level (contrary to Reichenbach, who accounts for the difference by positing different relations holding between Event time and Reference time). The main differences between these two tenses lie in the discourse role of the events, and how they are connected using temporal structure and world knowledge. The function of the PST.PRF in English consists in restricting the discourse connections that can hold (ibid., p. 29). Furthermore, the discourse structure is related to topic structure20. The authors provide a text that illustrates the two discourse relations Elaboration and Narration:

22.	[7]
	a-Guy experienced a lovely evening last night. b-He had a fantastic meal. c-He ate salmon. d-He devoured lots of cheese.

The topic for this text is provided by sentence a (“the experience of a lovely evening”). Then, sentence b stands in an Elaboration relation to a, and both c and d Elaborate sentence b, but sentences c and d stand in a Narration relation. Elaboration is a rather loose relation between two events: it indicates that a given event y is part of a preceding event x. Narration is stricter: an event y is a consequence of an event x (Lascarides and Asher 1993, p. 4).

Now, concerning the difference between the Breton PRS.PRF and the PST.PRF in the expression of perfective events, we can refine the proposal. Both tense-aspects set up an Elaboration relation with the Topic but each one of them implies two types of rhetorical relations. With the PRS.PRF, we generally have what Borillo et al. (2004) call ‘weak narration’, i.e., the sentences are linked by a relation of Continuation, which is a weak discourse relation that implies just temporal precedence (338). Continuation implies the loss of a strong temporal link, with temporal gaps allowed. The PST.PRF, however, corresponds more to what they call the relation of Occasion: the idea behind Occasion is that a sequence of sentences reporting a series of events is coherent only if this sequence reflects a ‘natural’ order of events (343). It implies strong narration. This squares with our preceding observation that verbs in the PRS.PRF maintain a deictic reference to Speech time (see example (21) with ‘out of the blue’ past events), contrary to PST.PRF sentences, in which a Breton speaker signals that the situations described are squarely in the past and form a story. It is the reason why, in our data, most PST.PRF sentences occur with time adverbials and are anaphorically linked to a temporal antecedent in the discourse. We believe that the temporal/deictic-anaphoric properties of each construction can be read off the rhetorical relations. To illustrate, excerpt (23) is a typical configuration of the alternation between the PRS.PRF and the PST.PRF:

23.	a	Me	zo bet savet	gant ma mamm-gozh…
		1SG	raise-PASS.PRS.PRF.	by my grandmother
	b	Hag ur wech en devoa skoet ac’hanon		er mor diwar-lein beg ar c’hae, e Blackpool
		and one time push-PST.PRF.3SG me		in sea off end of pier	in Blackpool
	c	Darbet e oa bet		din bezañ beuzet.
		be on the verge-PST.PRF.3SG		to.me be drowned
	d	Netra ne	c’hoarvezas	a-raok ma’m boa tapet	ma eizh vloaz.
		nothing NEG   happen-SPST.3SG before that get-PST.PRF.1SG my eight years
	e	Un deiz e	oa deuet	ma eontr-kozh	Algie…
		one day VRP	come-PST.PRF.3SGF my great-uncle		Algie

‘Well, my gran brought me up and she’s a witch… he [a great-uncle] pushed me off the end of Blackpool pier once, I nearly drowned–but nothing happened until I was eight. Great-uncle Algie came round for tea…’

The first sentence, (23a), constitutes the topic for the whole passage (the character Neville telling about his past): it is deictically introduced into the discourse, and it is in the PRS.PRF because bringing up the topic at this stage is currently relevant. It illustrates the Continuation relation (the event simply precedes the Speech time). The events that follow (23b–e), however, are all in the PST.PRF: they are in an Occasion relation to the first event, with the third event (nearly drowning) in a Result relation to the third (pushing him off a pier). The last event (coming round for tea) is in an Elaboration relation to the first. All these PST.PRF verbs illustrate the relation of Occasion as defined above: the PST.PRF events are anaphorically linked to past temporal adverbials (ur wech, un deiz) that disconnect them to the Speech time. They all temporally and anaphorically Elaborate the topic provided by the initial PRS.PRF sentence; in other words, they tell a sequentially ordered story. This pattern is almost systematic in our data: each time a past time adverbial serves as a temporal anchor (an anaphor) in the Breton sentence, the PST.PRF is used (example (24)), whereas a past time adverbial which is embedded in a sentence which talks about the present is in the PRS.PRF (example (25)):

24.	a	Da noz Kala-goañv en devoa klasket tremen dirak		ar c’hi			e dri fenn!
		at Halloween    try-PST.PRF.3SG pass		before the dog			of three heads
	b	Setu da belec’h ‘oa ‘vont				pa	hor boa gwelet	anezhañ.
		this to where go-PST.PROG.3SG				when see-PST.PRF.1PL him
		‘He tried to get past that three-headed dog at Halloween. That’s where he was going when we saw him.’

25.	a	Snape n’	en devo nemet lavarout ne		oar ket
		Snape NEG have.to only say			NEG know-PRS.3SG
	b	Penaos eo deuet		an troll e-barzh	ar skol	da Gala-goañv
		how	come-PRS.PRF.3SG the troll into		the school	at Halloween
	c	ha	n’	en deusket taolet troad ebet en trede estaj.
		and	NEG	set.foot-PRS.PRF.3SG     none in third floor

‘Snape’s only got to say he doesn’t know how the troll got in at Hallowe’en and that he was nowhere near the third floor…’

In (24), da noz Kala-goañv (‘at Halloween’) constitutes a temporal antecedent to the “trying to get past” and “seeing him” events which form a temporally coherent story (the sentence is a reminder of what happened on that night in that order); in (25), the same adverbial da Kala-goañv does not constitute a temporal antecedent for the event but is just an addition in a paragraph that enumerates the events as the speaker recalls them (in this excerpt, the character seeks a justification for Snape, he does not seek to reconstruct the order in which the events unfolded).

5. Conclusions

This corpus-based study of the distribution of perfect-like tense-aspects in Breton has helped us address the thorny question of the alternation between three different forms, where English uses mainly the SPST in narration and the PRS.PRF in dialogues which report currently relevant situations. The distinction between the SPST and the other two periphrastic forms (PRS.PRF and PST.PRF) of Breton is straightforward: a strict written narrative mode exclusively uses the SPST to move the action forward. In dialogue, however, a choice is possible between the PRS.PRF and the PST.PRF, and the reasons for the choice were the main goal of our investigation.

We were able to obtain some results by breaking down our corpus into three sub-corpora. We found that the Breton PRS.PRF is much more liberal than its English counterpart: in addition to functioning as a ‘real’ perfect, with experiential and resultative (occasionally, also continuative) uses connected to Speech time, it has also developed aoristic uses, like the French passé composé, thus confirming De Swart (2021 a)’s conclusion that the Breton perfect is an intermediate type of perfect, not unlike the Dutch one. Now, all three perfect-like forms (SPST, PRS.PRF, and PST.PRF) can be assigned similar rhetorical relations as per Lascarides and Asher, except for the PRS.PRF, which resists pure Narration. However, subsequent work in SDRT allowed us to refine the picture, and introduce a distinction between the relations of Continuation, giving rise to ‘weak’ narration, and Occasion, which creates ‘strong’ narration; we showed that this distinction was helpful in trying to tease apart the two perfect-like tenses uses in dialogues, namely the PRS.PRF and the PST.PRF. This account in terms of rhetorical relations also established that whereas the PRS.PRF is felicitous if a situation is deictically introduced ‘out of the blue’ into the discourse (it is non-presuppositional, in the sense of Michaelis 1994), the PST.PRF is preferred to express temporally pre-hodiernal events which are linked to an anaphoric anchor, most frequently in the form of a past temporal adverbial (though not obligatorily). This was confirmed in our study of the favored interaction of specific temporal adverbials of the yesterday type and when-clauses with the PST.PRF. The PRS.PRF retains its characteristics of a perfect: it often deictically introduces a currently relevant event, but contrary to the English PRS.PRF, there is no prohibition against specific past time adverbials (Klein’s 1994 present perfect puzzle). More work needs to be conducted, though, to refine the narrative function of the PRS.PRF, notably by analyzing in more detail the range of adverbials (the ago type) that are compatible with it.

Another important feature of the Breton system, which sets it apart from the English (and Dutch) one, is the existence of a strong imperfective system, which is reflected in the use of the (descriptive and habitual) imperfect and/or the progressive past for unbounded situations (state and activity verbs), where in English the aspectually neutral SPST is massively used. Contrary to the SPST, the imperfect is widely used in Breton, in narration and dialogue alike, showing that the Breton tense-aspect system is sensitive to the lexical aspect.

We would like to finish by stressing the cross-linguistic relevance of these findings: our study confirms that taking into account both the temporal and discourse structure for the use of tense-aspect constructions is crucial. This was particularly visible in the types of datasets that we provided: many of our examples showed that the same past event in dialogue can be reported in the PRS.PRF, then in the past PST.PRF, or the reverse. When that happens, the shift is clearly due to discourse pragmatic reasons: the PST.PRF is often a clear indication directed at the addressee that a narrative mode of discourse is in force, whereas the PRS.PRF points to a currently relevant past event that is topical at the Speech time. The fact that the Breton PST.PRF is normally used for Narration points to a necessary revision of its Reichenbachian characterization; in Breton, the PST.PRF is compatible with the two notations, R, E_S and R_E_S. In a recent paper, following a suggestion by Didier Bottineau21, we hypothesized that the oa auxiliary form of the PST.PRF, which has traditionally been labeled as the imperfect form of be, has partially replaced the simple past form (v)oe, which is totally unavailable in oral speech, thereby becoming a generalized past form of be, alongside the habitual form veze, which has developed simple durative uses. This language shift could explain why the PST.PRF of Breton has become a strict equivalent of the perfective SPST in spoken language, alongside its other, more traditional use as an anterior tense.