Residues and Extensions of Perfective Auxiliary be: Modal Conditioning

Abstract

This article provides both a diachronic and synchronic account of the generalization of perfective auxiliary be in specific irrealis modal contexts across numerous Romance varieties spoken in Italy and more widely within the Romània, which has essentially gone unnoticed in the descriptive and theoretical literature. In some cases (southern Calabrian, Latin American Spanish, Portuguese), the distribution of be is to be interpreted as a residue of an original unaccusative syntax which was exceptionally preserved under higher V-movement in irrealis contexts, whereas in others (person-driven dialects of central and southern Italy, southern peninsular Spanish, Romanian) this original unaccusative signal has been reanalysed as a specialized marker of irrealis (lexicalizing a high Mood head) and extended to all verb classes. In the case of Alguerès, by contrast, the generalization of irrealis be is argued to be the result of language contact with surrounding Sardinian dialects where a specific pattern of dedicated irrealis marking of Mood° has been replicated. Finally, the reverse pattern with generalization of irrealis have, the reanalysis of an aspectual distinction between resultative and experiential perfects found in early Romance varieties (Neapolitan, Sicilian, Spanish, Catalan), is shown to involve a similar pattern of dedicated irrealis marking in Mood°.

1. Introduction

An area of spectacular diachronic and synchronic microvariation in Italo-Romance and Romance more generally regards the numerous dimensions of variation characterizing the choice of auxiliary in the formation of various active perfective periphrases in conjunction with the past participle. Work over recent decades has brought to light a high degree of variation (for relevant bibliography, see Ledgeway 2012, pp. 292–99, 311–17; 2019; Loporcaro 2016), the precise empirical limits of which still remain to be defined (cf. Manzini and Savoia 2005, chs 5–6; Loporcaro and Pescarini 2022, §4.3). The principal dimensions of variation in Romance perfective auxiliaries are summarized in Ledgeway (2019), who reveals five broad dimensions of mesoparametric variation. The simplest option is represented by those varieties which generalize one auxiliary, either have (e.g., Sicilian) or be (e.g., some central-southern dialects of Italy, such as the Molisan variety of Pescolanciano). If, however, a dialect or language does present auxiliary alternation, this variation can, in order of complexity, be determined by mood (e.g., Romanian realis inflected have vs. irrealis invariable be; cf. Ledgeway 2014 and Section 3.3.2 below), tense (e.g., dialect of San Leucio del Sannio, where have obtains with the present perfect and be with the pluperfect; Iannace 1983, pp. 72–80, 88f.; Ledgeway 2012, p. 342f.), person (typically involving a binary [±discourse participant] split with be licensed by 1st/2nd persons and have by 3rd person; cf. Tuttle 1986, pp. 269–70; Kayne 1993; Manzini and Savoia 2005, II: §5.5; D’Alessandro and Roberts 2010; and Section 4.2 below), and argument structure (namely, verb class involving a binary active–stative split; cf. Ledgeway 2012, pp. 319–23).

One pattern, however, which appears to have gone entirely unnoticed is the restricted use of auxiliary be in specific irrealis modal contexts. Ledgeway (2000, p. 301, n. 22; 2001, 2002a, 2002b, 2003, 2009a, pp. 600–14) already noted that in (late) medieval Romance texts of southern Italy the first extensions of have to unaccusative syntax are licensed uniquely in irrealis modal contexts (cf. also Formentin 2001; Cennamo 2002), from where it gets a foothold in the system before progressively spreading to realis contexts yielding the generalized extension of have witnessed in the dialects today (cf. also Stolova 2006 for old Spanish and the discussion in Section 4.1 below). By contrast, a number of modern Romance varieties exhibit a quite different type of modally-determined auxiliary alternation involving be. For example, in the southern Calabrian dialect of Sant’Andrea Apostolo dello Ionio (cf. Voci 1994, p. XV) the auxiliary have has generalized to all verb classes, including unaccusatives, according to a pattern widespread in the dialects of southern Calabria and Sicily, as witnessed in (1a). However, be persists in this same dialect as a perfective auxiliary of unaccusatives as a relic of an original active-stative split uniquely in past counterfactual modal contexts such as (1b):

(1)	a.	Àiu	statu	fora	/	fattu. (S.Adr.)
		have.1sg	been	outside		done
		‘I have continuously been away / been doing.’ (Voci 1994. p. XV)
	b.	Si	fussa	statu		io. (S.Adr.)
		if	be.pst.sbjv.sg	been		I
		‘If it had been me.’ (Voci 1994, p. XVII)

Another pattern found in Italy, this time coming from the Sardinian variety of Catalan spoken in Alghero, is shown in (2), where we find a typical active–stative split, according to which transitives/unergatives license have (2a) and unaccusatives be (2b). However, this distribution is disrupted in counterfactual contexts, such as (2c), where auxiliary be generalizes, in this case in conjunction with a transitive.

(2)	a.	Pino	m’	ha	dat	una	cistella	de	pruna. (Alg.)
		Pino	me=	have.3sg	given	a	basket	of	plums
		‘Pino gave me a basket of plums.’ (http://prosodia.upf.edu/coalgueres/it/corpus/bosch/bo_et_15.html; 19 December 2021)
	b.	Són	arribats. (Alg.)
		be.3pl	arrived.mpl
		‘They have arrived.’ (http://prosodia.upf.edu/coalgueres/it/corpus/bosch/bo_et_15.html; 19/12/21)
	c.	No	la	fóra	casara	mai. (Alg.)
		neg	her=	be.cond.3sg	married.fsg	never
		‘He would never have married her.’ (https://www.algheroeco.com/la-rundalla-de-u-que-pugariva-essar-tambe-un-altru/; 19 December 2021)

These Andreolese and Alguerès patterns are reminiscent of a non-standard distribution found in Andalusian (3c) and Latin American (3d) Spanish as well as in European and Brazilian Portuguese (3e), where auxiliary have, otherwise generalized to all verb classes in the standard (3a–b), may exceptionally be replaced by auxiliary be in past counterfactuals (cf. Méndez García de Paredes 2011; Ledgeway 2012, p. 344f.).

(3)

a.

Habían

llorado

/

llegado

/

muerto. (Sp.)

b.

Tinham

chorado

/

chegado

/

morrido. (Pt.)

have.pst.3pl

cried

arrived

died

‘They had cried/arrived/died.’

c.

si

fueras

hecho

algo

desde

que

llegastes

a

if

be.pst.sbjv.2sg

done

something

since

that

arrive.pst.2sg

to

Córdoba (And.Sp.)

Córdoba

‘if you had done something since arriving in Córdoba’ (Ledgeway 2012, p. 345)

d.

¡No

actues

como

si

fueras

hecho

algo

not

do.prs.sbjv.2sg

as

if

be.pst.sbjv.2sg

done

something

malo! (Mex.Sp.)

bad

‘Don’t act as if you had done something wrong!’ (Ledgeway 2012, p. 345)

e.

ainda

que

a

implantação

dos

materiais

fosse

sido

feita

de

still

that

the

implantation

of.the

materials

be.pst.sbjv.3sg

been

done

of

boa

fé (EuPt., Lisbon)

good

faith

‘even if the introduction of the materials had been carried out in good faith’ (https://www.direitoemdia.pt/search/show/0365cba4a455186eec1732c3a22a17dedd0da0141ff56e2d8667c5609c9df49b; accessed on 1 December 2021)

Dialectal Romance data like these from the Italian peninsula and beyond raise a number of important empirical and theoretical questions about the licensing of perfective auxiliaries and, in particular, the irrealis modal features involved in the licensing of auxiliary be.1 In particular, it remains to be understood: (1) how such modal features can override the otherwise regular licensing of generalized have (Sant’Andrea, Andalusian/Latin American Spanish, Portuguese) or an active–stative have–be split (Alguerès); (2) whether the uses of counterfactual be in dialects, such as Andreolese, should be considered a ‘residue’ of an erstwhile more productive distribution of stative be in conjunction with unaccusative syntax, and, by the same token, whether the generalized counterfactual uses of be in such varieties as Alguerès and Andalusian Spanish effectively represent a novel extension of the auxiliary and, if so, how these diachronic patterns of conservation and innovation arise; (3) what parallels, if any, can be established with the historical and synchronic generalization of auxiliary have in irrealis contexts in (Italo-)Romance (Ledgeway 2003; 2019, §3.1) and whether these two cases can be unified by treating them as simply involving different lexicalizations; and (4) what similarities and differences exist between this modally-determined pattern of be licensing and the distribution of Romanian fi ‘be’ (Avram and Hill 2007; Ledgeway 2014) and central-southern Italo-Romance be (Ledgeway 2019, pp. 355–61) in irrealis contexts. These and other related questions will be investigated below, where I propose an analysis which unites in diachrony and synchrony all these irrealis patterns of auxiliary selection across Romance, irrespective of whether the auxiliary surfaces as have or be.

2. Irrealis be in Romance

2.1. have ⇒ be

2.1.1. Andreolese: Residues of be

On a par with most other dialects of central and southern Calabria, the central Calabrian dialect of Sant’Andrea Apostolo dello Ionio shows today a generalization of auxiliary have (Voci 1994, p. XV) in all persons, temporal and aspectual forms, and across all verb classes, as the following examples illustrate (cf. Ledgeway 2000, p. 205f.; 2012, p. 344 n. 31):

(4)	a.	Àiu	venutu	/	avutu	’a	friavi. (S.Adr.)
		have.1sg	come		had	the	influenza
		‘I’ve been coming / suffering from influenza (for some time).’ (Voci 1994, p. XV)
	b.	Avìa	statu	/	accattatu	/	vindutu. (S.Adr.)
		have.pst.1/3sg	been		bought		sold
		‘I/(s)he had been/bought/sold.’ (Voci 1994, p. XVI)
	c.	Mi	dissa	ca	io	l’	avìa	aiutatu. (S.Adr.)
		me=	say.pst.3sg	that	I	him=	have.pst.1sg	helped
		‘He told me that I had helped him.’ (Voci 1994, p. 174)
	d.	Si	avissa	avutu	tìampu,	avissa	jutu. (S.Adr.)
		if	have.pst.sbjv.1/3sg	had	time	have.pst.sbjv.1/3sg	gone
		‘If I/(s)he had had time, I/(s)he would have gone.’ (Voci 1994, p. XVII)
	e.	Si	avissa	statu	io,	l’	avissa	fattu. (S.Adr)
		if	have.pst.sbjv.1sg	been	I	it=	have.pst.sbjv.1sg	done
		‘If it had been me, I would have done it.’

On comparative grounds and given the strong precedents already in Latin (cf. Vincent 1982), there are good reasons to assume that in Andreolese as elsewhere in Italy (and indeed historically across the Romània) there was historically an auxiliary alternation driven by argument structure aligning have with active/transitive syntax and be with stative /unaccusative syntax. Although there are no early Andreolese texts which attest this stage, we do have early examples from other central-southern Calabrian dialects with generalized have today which formerly show the active–stative use of have (5a) and be (5b).

(5)	a.	Mai	v’	avesse	canosciuto (15th-c. Amendolea)
		never	you=	have.pst.sbjv.1sg	kown
		‘If I had never known you’ (Coletta di Amendolea, Grasso 1994, p. 113)
	b.	…	che	mai	al	mundo	nata	sia! (15th-c. Amendolea)
			that	ever	at.the	world	born.fsg	be.prs.sbjv.3sg
		‘…who was ever born in this world!’ (Coletta di Amendolea, Grasso 1994, p. 111)

As already noted, the sole exception to the generalized distribution of have in (4a–e) is in past counterfactual contexts where selection of have may optionally be overridden in favour of be in conjunction with unaccusative participles (Voci 1994, pp. XV, XVII).

(6)	a.	Si	avissa	/	fussa	statu	io… (S.Adr.)
		if	have.pst.sbjv.1sg		be.pst.sbjv.1sg	been	I
		‘If it had been me,…’ (Voci 1994, p. XVII)
	b.	Si	avissa	/	fussa	jutu	iddu… (S.Adr.)
		if	have.pst.sbjv.3sg		be.pst.sbjv.3sg	gone	he
		‘If he had gone,…’ (Voci 1994, p. XVII)

It is natural to assume therefore that this use of be, which, significantly, is restricted to unaccusative predicates in counterfactual contexts, represents an isolated residue of the erstwhile active–stative alternation which has otherwise been lost in realis contexts. If it were an innovation, then a priori we might expect it to also involve active syntax, contrary to fact. Rather, what we see in examples, such as (1b) and (6a–b), is a combination of contributing factors, namely unaccusativity and counterfactual modality, which together, but not individually, license be, the explanation for which we will come back to in Section 3.

2.1.2. Ibero-Romance Varieties

The facts just reviewed for Calabrian do not seem isolated within Romance but find a striking parallel in a number of substandard regional Spanish and Portuguese varieties where the generalization of have, which from around the 16th–17th centuries replaces an earlier active–stative (have–be) alternation (Mattos e Silva 1994, p. 62; Penny 2002, p. 166; Stolova 2006; Rosemeyer 2014, p. 18; Lopes and Brocardo 2016, p. 476; Loporcaro 2016, pp. 803, 815), is also frequently replaced by be in counterfactual contexts. Thus, before we turn to other varieties in Italy, it is instructive first to consider a number of regional varieties of Spanish and Portuguese spoken in Latin America, the Iberian Peninsula, and beyond.

2.1.2.1. Latin American Spanish: Residues of be

Beginning with Latin America, a number of scholars have observed, albeit only briefly in most cases, the optional but frequent use of auxiliary ser ‘be’ in the pluperfect subjunctive and/or the conditional perfect, henceforth referred to as the ‘counterfactual’. Such uses are reported for Latin America in general (Chumaceiro and Álvarez Muro 2004, p. 145; Méndez García de Paredes 2011, p. 1016f.) and in particular for Colombia (7a; Montes Giraldo 1974, p. 424; 1976, 1996, p. 138; Aleza Izquierdo 2010, p. 170; Mištinová 2012, p. 237; Bernal Chávez and Díaz Romero 2017, p. 28); Costa Rica (7b; Castillo Venegas 2013, p. 323); Ecuador (7c; Toscano Mateus 1953, p. 287), Mexico (7d), including among second- and third-generation migrants in the US (7e; Sánchez 1982, p. 26f.; Gutiérrez 1997); Nicaragua (7f; Pato 2018, pp. 1070f., 1086); Panama (7g; Pacheco et al. 2013, p. 368; Pato 2019, p. 1055); Paraguay (7h; de Granda 1988, p. 46f.; 1991, p. 87f.; Aleza Izquierdo 2010, p. 169) and Venezuela (7i; Montes Giraldo 1976, p. 561; Navarro Correa 1991, p. 306).

(7)

a.

si

el

oro

fuera

valido

más

hoy,

yo

fuera

if

the

gold

be.pst.sbjv.3sg

been.worth

more

today

I

be.pst.sbjv.1sg

sido

más

rico (Cmb.Sp.)

been

more

rich

‘if the gold have been worth more today, I would have been richer.’ (Bernal Chávez and Díaz Romero 2017, p. 28)

b.

si

fuera

ido

con

mi

silla

eléctrica

claro

que

si

if

be.pst.sbjv.1sg

gone

with

my

chair

electric

clear

that

yes

fuera

podido. (CRic.Sp.)

be.pst.sbjv.1sg

been.able

‘If I had gone with my electric wheelchair then of course I would have managed.’ (https://www.muniliberia.go.cr/muni/files/documents/73_583_actaextraord.142016.pdf; accessed on 29 November 2021)

c.

si yo

no

me

llamaría

así,

no

se

lo

juera

dicho (Ecd.Sp.)

if I

neg

me=

call.cond.1sg

thus

neg

dat.3=

it=

be.pst.sbjv.1sg

said

‘If that weren’t my name, I wouldn’t have told him.’ (Toscano Mateus 1953, p. 287)

d.

se

fuera

ido

directo

a

la

basura

si

ella

no

me

self=

be.pst.sbjv.3sg

gone

direct

to

the

rubbish

if

she

neg

me=

fuera

obligado

a

defender

a

Kanye (Mex.Sp.)

be.pst.sbjv.3sg

obliged

to

defend.inf

dom

Kanye

‘it would have gone straight into the bin if she hadn’t forced me to defend Kanye.’ (https://www.revistaclase.mx/gente-con-clase/se-reaviva-la-guerra-entre-kim-kardashian-y-taylor-swift; accessed on 29 November 2021)

e.

quizás

no

me

fuera

casado (Mex.Sp., 2nd generation USA)

perhaps

neg

me=

be.pst.sbjv.1sg

married

‘perhaps I wouldn’t have got married.’ (Gutiérrez 1997, p. 263)

f.

Me

encantaría

que

se

fuera

venido

pa

Nicaragua (Nic.Sp.)

me=

bewitch.cond.3sg

that

if

be.pst.sbjv.3sg

come

for

Nicaragua

‘I’d have loved it if she had come to Nicaragua.’ (Pato 2018, p. 1070)

g.

Si

hubiera

tenido

dinero

lo

fuera

comprado

(Pan.Sp.)

if

have.pst.sbjv.1sg

had

money

it=

be.pst.sbjv.1sg

bought

‘If I had had money, I would have bought it.’ (Pato 2019, p. 1055)

h.

si

el

fuera

venido

ayer (Pgy.Sp.)

if

he

be.pst.sbjv.3sg

come

yesterday

‘if he had come yesterday’ (de Granda 1988, p. 47)

i.

Si

Emiliano

fuera

perdido

por

nocaut […]

yo

fuera

if

Emiliano

be.pst.sbjv.3sg

lost

by

knockout

I

be.pst.sbjv.1sg

quedao

más

conforme (Vnz.Sp.)

stayed

more

satisfied

‘If Emiliano had lost by a knockout […] I would have been happier.’ (Montes Giraldo 1976, p. 561)

The phenomenon is not restricted to just these national varieties (see also Méndez García de Paredes 2011, pp. 1019–22) but is found widely across the continent, as can be readily verified by an internet search. Without making any claims to exhaustivity, it was not possible to identify any online examples for Belize, Puerto Rico and only a relatively small number for Bolivia (8a), El Salvador (8b), Guatemala (8c) and Honduras (8d), whereas for Argentina (8e), Chile (8f), Cuba (8g), the Dominican Republic (8h), Peru (8i) and Uruguay (8j) examples are more numerous.2

(8)

a.

menos

hubiera

avanzado

tanto,

si

fuera

sido

less

have.cond.3sg

advanced

so.much

if

be.pst.sbjv.3sg

been

gobernada

por

un

grupo

de

incompetentes (Bol.Sp)

governed

by

a

group

of

incompetents

‘[Bolivia] would have made much less progress, if it has been governed by a bunch of incompetents.’ (https://www.la-epoca.com.bo/2016/12/05/el-agua-y-sus-lecciones/; accessed on 29 November 2021)

b.

aun

cuando

fuera

sido

orientada

a

otro

centro

even

when

be.pst.sbjv.3sg

been

oriented

to

other

centre

hospitalario (Slv.Sp)

hospital

‘even if she had been sent to another hospital’ (https://www.transparencia.gob.sv; accessed on 29 November 2021)

c.

las

palabras

de

trump,

que

aunque

fueran

sido

hechas

en

the

words

of

Trump

that

although

be.pst.sbjv.3pl

been

done

in

estado

de

ebriedad (Gtm.Sp.)

state

of

drunkenness

‘Trump’s words which, although they were said in a state of intoxication’ (https://lahora.gt/posturas-sobre-el-matrimonio-gay/; accessed on 29 November 2021)

d.

si

desde

hace

años

los

centros

educativos

fueran

sido

mejorados,

if

since

ago

years

the

centres

educational

be.pst.sbjv.3pl

been

improved

creo

que

la

educación

no

se

hubiera

atrasado (Hon.Sp.)

believe.1sg

that

the

education

neg

self=

have.pst.sbjv.3sg

delayed

‘if years ago educational centres had been improved, I believe that education wouldn’t have fallen so behind.’ (https://www.elinformativo.hn/archivos/96169; accessed on 29 November 2021)

e.

si

fuera

existido

una

estrategia

adecuada

y

oportuna,

donde

if

be.pst.sbjv.3sg

existed

a

strategy

adequate

and

timely

where

[…]

el

personal

fuera

sido

capacitado

apropiadamente;

se

the

staff

be.pst.sbjv.3sg

been

trained

appropriately

self=

fuera

permitido

prevenir

esta

serie

de

acontecimientos. (Arg.Sp.)

be.pst.sbjv.3sg

permitted

prevent.inf

this

series

of

events

‘if there had been an adequate and timely strategy where […] the staff would have been appropriately trained; we would have been able to avoid this series of events.’ (https://codeinep.org/miembros/yohaflo/activity/3022/; accessed on 29 November 2021)

f.

gran

parte

de

las

baterías

almacenadas

presentan

filtraciones

de

ácido,

big

part

of

the

batteries

stored

present.3pl

leakages

of

acid

sin

que

fueran

sido

retiradas

y

almacenadas

en

bins (Ch.Sp.)

without

that

be.pst.sbjv.3pl

been

withdrawn

and

stored

in

bins

‘many stored batteries leak, but hadn’t been taken out and stored in bins.’ (08 Informe de Fiscalización Ambiental, https://snifa.sma.gob.cl; accessed on 29 November 2021)

g.

si

no

fuera

sido

por

EEUU

la

guerra

fuera

llegado

if

neg

be.pst.sbjv.3sg

been

for

USA

the

war

be.pst.sbjv.3sg

arrived

a

suramerica (Cub.Sp)

to

South.America

‘if it hadn’t been for the USA, there would have been war in South America.’

(http://www.cubadebate.cu/noticias/2010/11/26/china-advierte-a-estados-unidos-sobre-maniobras-militares-en-penisula-coreana/; accessed on 29 November 2021)

h.

gracias por

toda

su

colaboración

sin

ustedes

mi

pedido

no

thanks

for

all

your

collaboration

without

you

my

order

neg

fuera

llegado (Dmn.Sp.)

be.pst.sbjv.3sg

arrived

‘thanks for your help, without you my order wouldn’t have arrived.’ (https://www.amorossa.com/products/chivas-regal-18-anos; accessed on 29 November 2021)

i.

lo

han

pagado

como

si

fueran

comprado

chocolates

it=

have.3pl

paid

as

if

be.pst.sbjv.3pl

bought

chocolates

suizos. (Per.Sp)

Swiss

‘they paid for it as if they had been buying Swiss chocolates.’ (https://www2.congreso.gob.pe/Sicr/DiarioDebates/Publicad.nsf/SesinesPleno/05256D6E0073DFE905257EDF00575778/$FILE/PLO-2015-19.pdf; accessed on 29 November 2021)

j.

si

no

fuera

sido

por

que

mis

padres, […]

hubiera

if

neg

be.pst.sbjv.3sg

been

for

that

my

parents

have.pst.sbjv.3sg

sido

para

reírse (Urg.Sp.)

been

for

laugh.inf=self

‘if it hadn’t been for my parents, […] it would have been laughable.’ (https://www.mateamargo.org.uy/wp-content/uploads/2020/10/Cuestion.-No-14-17-02-1972-.pdf; accessed on 29 November 2021)

As the examples in (7)–(8) illustrate, be most readily and frequently surfaces in the protasis of conditional clauses (7h, 8a–b,d,j), but it may also surface in both protasis and apodosis (7a–b,d,i, 8e,g) and more rarely in just the latter (7c,g).3 More generally, it can also occur in past counterfactual clauses outside of conditional sentences where the pluperfect subjunctive is typically licensed (7e–f, 8c,f,h–i). Although most of the available examples come from oral sources (e.g., 7c,e,i, 8i) and highly colloquial and linguistically uncontrolled sources, such as online fora, blogs, discussion sites and social media, such as Instagram, Facebook and Twitter (7d,f, 8a,c,g–h,j; cf. Méndez García de Paredes 2011, pp. 1018–24), partially reflecting claims that this usage is typical of colloquial (Pato 2018, p. 1070) and above all uneducated and rural speech (de Granda 1988, p. 47; Aleza Izquierdo 2010, p. 170; Castillo Venegas 2013, p. 323), there are also a sizeable number of examples from more formal written sources, including official and governmental documents and reports (8b,d–f). This is further substantiated by Méndez García de Paredes (2011, pp. 1019–21) who reports many examples from the oral and written language of highly educated speakers coming from socioeconomically prestigious professions and backgrounds, including doctors, politicians and managers. This shows how for many speakers this feature of their grammar passes under the prescriptive radar and is perceived to be a core feature of the language.4 It is therefore surprising, given the frequency of such forms in spoken and even written varieties of Latin America, as well as their occasional brief acknowledgement in a number of studies since at least the 1970s, that they have gone unnoticed in the otherwise vast formal literature on Romance auxiliary selection and, in particular, on Spanish (cf. Rosemeyer 2014; Mackenzie 2006).

It is striking, however, that in the handful of studies that do briefly mention the distribution of ser ‘be’ in Latin American varieties of Spanish, it is typically referred to as an ‘archaism’ (Montes Giraldo 1976; Aleza Izquierdo 2010, p. 169; Mištinová 2012, p. 237; Pato 2018, p. 1071) and hence treated as a ‘retention’ (de Granda 1988, p. 1991). This follows in large part from their often explicit observation (cf. Montes Giraldo 1976, p. 562; de Granda 1988, p. 46f.; Pato 2018, p. 1070f., 1086f.; 2019, p. 1055 n.18) that this use of ser principally involves intransitive participles (viz. unaccusatives), the same class of verbs that regularly occurred with ser in both irrealis and realis contexts until the 16–17th centuries. Even where this is not explicitly acknowledged, most (if not all) of the examples reported involve unaccusative participles, as do the majority of examples above (7a–b,f,h, 8e,g–h,j), including passives (8a–f) and reflexives (7e). That is not to say, however, that we do not find examples of ser in conjunction with transitives and unergatives (7c, 8i), albeit inconsistently (7g) and often alongside unaccusatives (7b,d,i), but they are statistically much rarer, suggesting a more recent development, namely an extension of the residual use of ser with unaccusatives to transitives/unergatives.5

2.1.2.2. Spain: Extension of be

Turning now to Spain, the most extensive study of the phenomenon to date is by Méndez García de Paredes (2011, pp. 1012–16), who documents the widespread use of counterfactual be across Andalusia (9a; cf. also Ledgeway 2012, p. 344f.), including in and around the cities of Almería, Granada, Jaén, Málaga, Córdoba, Jerez de la Frontera, Seville and Huelva (9b–g),6 as well as in various areas of Murcia (10a–c; cf. also García Soriano 1932, p. XCVII; Guillén García 1974, p. 67; Gómez Ortín 2004, p. 20) and Extremadura (10d; cf. Flores del Manzano 1992).

(9)	a.	si	yo	fu[er]á	ehtao	allí,	eso	no	fu[er]á	pasao. (Andalusia)
		if	I	be.pst.sbjv.1sg	been	there	this	neg	be.pst.sbjv.3sg	happened
		‘If I had been there, this wouldn’t have happened.’ (Narbona Jiménez 2019, p. 560)
	b.	Si	fuéramos	veni(d)o	antes,	te	fuéramos
		if	be.pst.sbjv.1pl	come	before	you=	be.pst.sbjv.1pl
		ayuda(d)o. (Ubrique, Cádiz)
		helped
		‘If we had come earlier, we would have helped you.’ (Pérez Sánchez de Medina 2007, p. 35)
	c.	fwéra			abláo,	fwérä		abláo (Cúllar-Baza, Granada)
		be.pst.sbjv.1sg			spoken	be.pst.sbjv.2sg		spoken
		‘I, you would have spoken.’ (Salvador 1959, p. 58)
	d.	Si	lo	fuera		sabío. (Baez, Jaén)
		if	it=	be.pst.sbvj.1sg		known
		‘If I had known.’ (Carrasco Cantos 1981, p. 127)
	e.	ojala	lo	fuera	visto	antes! (Córdoba)
		if.only	it=	be.pst.sbjv.1sg	seen	before
		‘if only I had seen it before!’ (https://www.todocircuito.com/foro/compra-venta-equipacion-37/vendo-mono-berick-1pieza-oportunidad-17968/; accessed on 9 December 2021)
	f.	¿en	qué	me	fuera	gustado	trabajar? (Lucena)
		in	what	me=	be.pst.sbjv.3sg	pleased	work.inf
		‘What type of job would I have liked to do?’ (Méndez García de Paredes 2011, p. 1015)

(10)	a.	Es	lo	que […]	juera	dicho	to	ombre. (Murcia)
		be.3sg	the	that	be.pst.sbjv.1sg	said	your	man
		‘It’s what […] your husband would have said.’ (García Soriano 1932, p. XCVII)
	b.	Si	me	lo	fueras	dicho,	no	te	fuera	pasao
		if	me=	it=	be.pst.sbjv.2sg	said	neg	you=	be.pst.sbjv.3sg	happened
		eso. (Murcia)
		this
		‘If you had told me, this wouldn’t have happened to you.’ (Gómez Ortín 2004: 20)
	c.	si	yo	lo	fuera	cogío… (Orihuela, Murcia)
		if	I	it=	be.pst.sbvj.1sg	taken
		‘if I had taken it…’ (Guillén García 1974, p. 67)
	d.	Si	fuerah	venío. (Gredos, Extremadura)
		if	be.pst.sbjv.2sg	come
		‘If you had come.’ (Flores del Manzano 1992, p. 131)

The shared distribution of counterfactual be in both southern Spain and in Latin America is hardly surprising given the traditional view that the Spanish of Latin America historically represents in some sense a continuation of Andalusian Spanish, inasmuch as the first Spanish settlers in America predominantly came from Andalusia (Penny 2000, pp. 139–44). Also similar to the Latin American situation are frequent claims (see Méndez García de Paredes 2011, p. 1010f.) that this usage characterizes rural areas and the speech of the elderly, the poorly educated and the illiterate,7 although, once again, such traditional claims are contradicted by the observation (cf. Méndez García de Paredes 2011, p. 1010 n.2) that the distribution of irrealis be is not at all recessive but, rather, is widespread across Andalusia, including in the speech of younger speakers, as well as on the internet, where at least a degree of literacy is required (Méndez García de Paredes 2011, pp. 1019–24).

One important respect, however, in which the peninsular data differ from those of Latin America is the distribution of ser ‘be’ beyond unaccusative syntax.8 Above we saw how in Latin America the counterfactual ser was principally, though not exclusively, found with unaccusative participles, whereas in the peninsular data ser is found equally with unaccusatives (9a–b,f, 10b,d) and transitives/unergatives (9b–e, 10a–c); again internet searches readily confirm this, witness the following selection of representative examples:

(11)	a.	¿Si	tu	lo	fueras	sabido	lo	fueras	denunciado? […]
		if	you	it=	be.pst.sbjv.2sg	known	him=	be.pst.sbjv.2sg	reported
		si	lo	fueran	acusado	el	que	lo	fuera	hecho […]	no
		if	him=	be.pst.sbjv.3pl	accused	he	that	it=	be.pst.sbjv.3sg	done	neg
		sale	de rositas,	aunque	lo	fuera	demostrado	todo	y
		exit.3sg	for free	although	it=	be.pst.sbjv.3sg	shown	all	and
		aunque	lo	fueran	metido	en	la	cárcel (Seville)
		although	him=	be.pst.sbjv.3pl	put	in	the	prison
		‘If you had known, would you have reported him? […] if they had accused him, the one that allegedly did it […] won’t get off scot-free, although he had proven it entirely and had been thrown in jail.’ (https://www.foro-ciudad.com/sevilla/paradas/mensaje-11446001.html; accessed on 9 December 2021)
	b.	si	fueran	podido	se	la	fueran		cargado. (Olvera, Cádiz)
		if	be.pst.sbjv.3pl	been.able	self=	it=	be.pst.sbjv.3pl		loaded
		‘if they had been able to, they would have got rid of it.’ (https://www.foro-ciudad.com/cadiz/olvera/mensaje-12190957.html; accessed on 9 December 2021)
	c.	si se	fuera	parado,	si	no	fuera		respondido. (Antequera)
		If self=	be.pst.sbjv.3sg	stopped	if	neg	be.pst.sbjv.3sg		replied
		‘if the heart had stopped, if the baby hadn’t shown any response’ (https://www.elsoldeantequera.com/antequera/32120-una-enfermera-salva-la-vida-de-un-bebe-en-plena-calle-cuando-su-madre-iba-en-coche-al-colegio-en-antequera.html; accessed on 9 December 2021)

As noted by Méndez García de Paredes (2011, p. 1010), claims like those for Latin America that counterfactual be represents a residual archaicism (cf. Narbona Jiménez et al. 2003, p. 237) simply do not hold for southern Spain, inasmuch as perfective be was historically never an option with transitive and unergative syntax. Rather, examples such as (11a–c) must be interpreted as an analogical extension of the distribution of be, which was originally restricted to unaccusatives (Méndez García de Paredes 2011, p. 1012). Indeed, this interpretation of the facts finds further support in Reixac’s (1749) Instruccions per la ensenyansa de minyons, a manual written in Fontcoberta (province of Girona) regarding some basic principles in the teaching of Catalan-speaking children through the medium of Spanish. Significantly, in the relevant sections on the conjugation of verbs Rexiac provides for transitive amar ‘love’ (pp. 315–18) active compound perfective paradigms formed with auxiliary haber ‘have’ (e.g., he/habia [sic]/habré/haya/hubiesse [sic] /hubiera amado ‘have.prs.ind.1sg/pst.ipfv.1sg/fut.1sg/prs.sbjv.1sg/pst.sbjv.1sg/pst.sbjv-cond.1sg loved’), whereas in the case of ser/estar ‘be’ (pp. 312–14) he gives auxiliary haber ‘have’ for the present and past indicative and the past subjunctive/conditional in -ra (e.g., he/habia [sic] /hubiera sido/estado ‘have.prs.ind.1sg/pst.ipfv.1sg/pst.sbjv-cond.1sg been’), but both haber and ser ‘be’ for the past subjunctive in -se and the future (e.g., hubiesse [sic]/habré sido/estado ‘have.pst.sbjv.1sg/fut.1sg been’ or fuese/seré sido/estado ‘be.pst.sbjv.1sg/fut.1sg been’). Although the distribution of be with the future is an aberration from the patterns witnessed above (though crucially still involving an irrealis modal form), we nonetheless see some quite compelling mid-18th evidence from Spain for a once more restrictive distribution of irrealis be limited to unaccusatives, a stage still largely preserved in Latin American Spanish and, as we shall see, in substandard Portuguese.

2.1.2.3. Portuguese: Residues of be

As in the case of Spanish, the formal literature on Romance auxiliary selection simply assumes that in modern Portuguese the auxiliary have—today typically ter < tenere ‘hold, keep’, more rarely haver < habere ‘have’, now principally restricted to formal written registers—has generalized to all verbs and contexts, replacing a previous active–stative alignment of the auxiliaries (haver>/)ter ‘have’ and ser ‘be’ (Huber 1933, p. 221; Mackenzie 2006; Loporcaro 2007, pp. 177, 179f.; 2016, p. 815; Rosemeyer 2014, p. 32 n.8; Ledgeway 2019, p. 349). Even specific studies examining the progressive retraction of Portuguese ser fail to recognize anything different and variously conclude that auxiliary ser with unaccusative participles was lost during the course of the 16th (cf. Guilherme 2009, p. 83), 17th (Hricsina 2017, p. 176) or 18th (Carasco González 2020, p. 86) century. However, this characterization is simply incorrect, in that ser is still widely employed today, albeit in non-standard usage, an observation which continues to fall under the radar of those working on Portuguese morphosyntactic variation (cf. Scherre and Duarte 2016). In particular, my data come from the internet since, as already noted, the use of ser in modern Portuguese is simply not reported, not even in passing, in either standard grammars and manuals or specific linguistic studies. On a par with what was observed for Andreolese (Section 2.1.1) and Latin American Spanish (Section 2.1.2.1), auxiliary ser continues to occur in the past subjunctive in past counterfactual contexts with unaccusative participles, but not with transitives or unergatives, in both European (12a–e) and Brazilian (13a–e) Portuguese.

(12)

a.

mesmo

que

o

cheque

fosse

sido

devolvido […]

não

even

that

the

cheque

be.pst.sbjv.3sg

been

returned

neg

teria

igualmente

sido

pago (Coimbra)

have.cond.3sg

equally

been

paid

‘even if the cheque had been returned […] it still wouldn’t have been paid.’ (http://www.dgsi.pt/jtrc.nsf/c3fb530030ea1c61802568d9005cd5bb/1a95735e95e912ec80257a9b00579083?OpenDocument; accessed on 1 December 2021)

b.

se

fosse

sido

menino

não

teria

sido

fácil. (EuPt.)

if

be.pst.sbjv.3sg

been

boy

neg

have.cond.3sg

been

easy

‘if she had been a boy it would not have been easy.’ (https://demaeparamae.pt/forum/mamas-dexembro2010?page=36; accessed on 1 December 2021)

c.

mesmo

que

tivesse

acabado

de

cometer

algum

crime

ou

even

that

have.pst.sbjv.3sg

finished

of

commit.inf

some

crime

or

fosse

fugido

da

cadeia. (Murgido, northern Portugal)

be.pst.sbjv.3sg

fled

from.the

prison

‘even if he had finished committing some crime or had escaped from prison’ (http://files.murgido-aldeia-serrana-candemil.webnode.pt/200001084-bc9even86bd91d/LENDAS%20DE%20MURGIDO.pdf; accessed on 1 December 2021)

d.

ele

poderia

muito

bem

receber,

se

o

caso

avnancasse

he

can.cond.3sg

very

well

receive.inf

if

the

case

advance.pst.sbjv.3sg

e

não

fosse

cadído

em

esquecimento...

18

a

36

milhoes

de

and

neg

be.pst.sbjv.3sg

fallen

in

oblivion

18

to

36

millions

of

dólares. (EuPt.)

dollars

‘he might well receive 18 to 16 million dollars, if the case were to go ahead and if it hadn’t sunk into oblivion.’(http://apipocamaisdoce.sapo.pt/2016/01/o-making-murderer-deu-comigo-em-doida.html; accessed on 2 December 2021)

e.

Lembre-se

disso,

se

fosse

ficado,

eles

ainda

estariam

remind=self

of.this

if

be.pst.sbjv.3sg

stayed

they

still

be.cond.3pl

juntos. (EuPt.)

together

‘Just remember this, if you had stayed, they would still be together.’ (https://ebstomasborba.pt/sera-que-o-silk-touch-funciona-nos-jogadores/; accessed on 2 December 2021)

(13)

a.

Se

fosse

sido

ao

contrário,

teríamos

semanas

de

manifestações (BtPt.)

if

be.pst.sbjv.3sg

been

to.the

contrary

have.cond.1pl

weeks

of

protests

‘If it had been the other way round, then we would have (had) weeks of protests.’ (https://twitter.com/melccs; accessed on 2 December 2021)

b.

desde

que,

num

prazo

não

superior

a

duas

horas,

o

detido

since

that

in.a

period

neg

more

to

two

hours

the

detainee

fosse

sido

entregue

a

uma

autoridade

judiciária (BrPt.)

be.pst.sbjv.3sg

been

delivered

to

a

authority

judiciary

‘on condition that, within a period of no longer than two hours, the detainee had been handed over to a judiciary authority.’ (https://ruicastro.jusbrasil.com.br/artigos/152544827/codigo-de-processo-penal-portugues-justica-celere-mas-nao-tanto; accessed on 2 December 2021)

c.

se

não

fosse

ido

ao

médico,

poderia

ter

tido

if

neg

be.pst.sbjv.3sg

gone

to.the

doctor

can.cond.3sg

have.inf

had

problemas

gravíssimos (BrPt.)

problems

very.serious

‘if he hadn’t gone to the doctor’s, he could have had very serious problems.’ (https://pt-br.facebook.com/pg/ImportsLMoficial/posts/?ref=page_internal; accessed on 2 December 2021)

d.

é

como

se

nada

fosse

acontecido (BrPt.)

be.3sg

as

if

nothing

be.pst.sbjv.3sg

happened

‘it’s as if nothing had happened.’ (https://patriciateixeiraadvogada.jusbrasil.com.br/noticias/536387584/carmen-lucia-suspende-a-posse-de-cristiane-brasil-no-ministerio-do-trabalho; accessed on 2 December 2021)

e.

Era

até

necessário

que

as

famílias

fossem

saído

do

meio

urbano (BrPt.)

be.pst.3sg

even

necessary

that

the

families

be.pst.sbjv.3pl

exited

from.the

centre

urban

‘It was even necessary for families to move out of the city centre.’ (https://clubedeautores.com.br/livro/o-decimo-quinto; accessed on 2 December 2021)

Although further detailed research is required, an internet search for similar examples in other Portuguese-speaking countries (Mozambique, Cape Verde, Guinea Bissau, and São Tomé and Príncipe) returned no results,9 save the now lexicalized use of chegar ‘arrive’ with a temporal subject (cf. Hricsina 2017, p. 178; Carasco González 2020, p. 82f.) in the isolated Angolan example Quando fosse chegado o momento ‘When the moment/time had come’. The exceptions are Angola (14a) and Macao (14b–c), where the use of counterfactual ser was found in three examples with morrido ‘died’ and chegado ‘arrived’.

(14)

a.

E

vivem

como

se

nunca

fossem

morrido

e

morrem

como

and

live.3pl

as

if

never

be.pst.sbjv.3pl

died

and

die.3pl

as

se

nunca

tivessem

vivido (Balombo, Angola)10

if

never

have.pst.sbjv.3pl

lived

‘And they live as if they had never died and they die as if they had never lived.’ (https://www.findglocal.com/AO/Lobito/1632309003714149/Colectivo-NdingaNzol; accessed on 6 December 2021)

b.

Lancelote

nunca

negou

que

não

fosse

chegado

ao

Lancelote

never

deny.pst.3sg

that

neg

be.pst.sbjv.3sg

arrived

at.the

governo

americano

government

American

‘Lancelote never denied that he hadn’t reached the American government.’ (https://jtm.com.mo/local/um-homem-grandioso-na-sua-simplicidade/; accessed on 2 December 2021)

c.

talvez

fosse

chegado

a

um

acordo

entre

K

e

a

1ª

perhaps

be.pst.sbjv.3sg

arrived

at

an

agreement

between

K

and

the

first

ré (Macao)

defendant

‘perhaps he had come to an agreement between K and the first defendant.’ (https://www.court.gov.mo/sentence/pt/20993; accessed on 2 December 2021)

As with Spanish (cf. notes 3, 8 above), alongside the past subjunctive (e.g., fosse ‘be.pst.sbjv.1/3sg’), we also find occasional examples of counterfactual ser in the conditional (e.g., seria ‘be.cond.3sg’), such as (15a–b).

(15)

a.

Estes

retratos […]

nunca

seriam

sido

queimados,

mesmo

que

o

these

portraits

never

be.cond.3pl

been

burned

even

that

the

artista

não

tivesse

sido

presidente

dos

Estados

Unidos (EuPt.)

artista

neg

have.pst.sbjv.3sg

been

president

of.the

states

united

‘These portraits […] would never have been burned, even if the artist hadn’t been the president of the United States.’ (https://expresso.pt/cultura/2017-04-22-A-expiacao-de-Bush; accessed on 2 December 2021)

b.

acredito

que

seria

ficado

melhor

como

dois

livros

believe.1sg

that

be.cond.3sg

stayed

better

as

two

books

separados. (BrPt.)

separate

‘I think that it would have been better as two separate books.’ (https://www.skoob.com.br/livro/resenhas/513195/mais-gostaram; accessed on 2 December 2021)

While some of the examples above (and many others not reported here) clearly reflect a casual, relaxed register, inasmuch as they are taken from public discussion sites (12b, 13b,d,e, 15b), blogs (12d–e) and social media (13a,c), in many other cases the examples occur in controlled formal spoken and above all written registers, including, for example, online newspapers (14a, 15a), published short stories (12c) and official court proceedings and judgements (12a, 14b). Indeed, the most frequent examples in my corpus of online examples occur in the passive (cf. 12a, 13b, 15a), a construction which is independently known to be occur most frequently in formal and especially written registers (Ledgeway 2021, §2.1, §2.3). This conclusion is further supported by the observation that the counterfactual use of ser ‘be’ can also occur in a form continuing the Latin pluperfect indicative, e.g., fueram/-t > fora ‘I/(s)he had been’. In modern Portuguese, this paradigm today functions mainly as a pluperfect indicative and is limited to written and above all literary registers, both in Portugal (Cunha and Cintra 1985, p. 329) and Brazil (Thomas 1969, p. 133). In the past, and still today in some set expressions and in the literary language (Thomas 1969, p. 136; Teyssier 1984, p. 212; Cunha and Cintra 1985, p. 329f.), this same form may also function as a past subjunctive and a conditional (cf. related Spanish form fuera above). Significantly, this otherwise marked, formal paradigm of ser occurs with a surprisingly high degree of frequency as an alternative subjunctive or conditional form of the auxiliary alongside the more usual fosse (and seria) forms, both in the passive (16a-b) and more generally with other unaccusative predicates (16c–f).

(16)

a.

embora

tivesse

deixado

claro

que

fora

sido

although

have.pst.sbjv.3sg

left

clear

that

be.cond.3sg

been

confrontado

com

algumas

propostas

bem

interessantes

de

outros

confronted

with

some

proposals

well

interesting

from

other

clubs (EuPt.)

clubs

‘although he had made it clear that he would have been confronted with some very interesting proposals from other clubs’ (https://www.record.pt/futebol/futebol-nacional/liga-bwin/v--guimaraes/detalhe/transferencia-de-alex-pode-resolver-se-hoje; accessed on 2 December 2021)

b.

a

vítima

já

fora

sido

alvo

de

seis

tiros,

mas

havia

the

victim

already

be.cond.3sg

been

target

of

six

shots

but

have.pst.3sg

sobrevivido (Ipatinga, State of Minas Gerais)

survived

‘the victim had allegedly already been the target of six shootings, but had survived.’ (https://www.diariodoaco.com.br/noticia/0009911-vitima-de-tentativa-de-homicidio-morre-apos-um-novo-atentado; accessed on 2 December 2021)

c.

James

Franco

já

fora

sido

notícia

quando

se

soube

James

Franco

already

be.cond.3sg

been

news

when

self=

know.pst.3sg

que

o

filme

já

havia

estreado

na

Rússia (EuPt.)

that

the

film

already

have.pst.3sg

premiered

in.the

Russia

‘James Franco had allegedly already become the news story when it was discovered that the film had already premiered in Russia.’ (https://www.insider.pt/2019/09/24/san-sebastian-em-zeroville-um-james-franco-punk-sonha-com-new-hollywood-mas-volta-a-ser-um-disaster-artist/; accessed on 2 December 2021)

d.

Ao

longo

destes

anos,

entraram

elementos

novos,

foram

saído

to.the

length

of.these

years

enter.pst.3pl

elements

new

be.cond.3pl

exited

outros

tantos,

mas

o

importante

era

avançar (Lisbon)

other

so.many

but

the

important

be.pst.3sg

advance.inf

‘Throughout these years, new elements came in, many others had allegedly been lost, but the important thing was to move forward.’ (https://www.medicina.ulisboa.pt/newsfmul-artigo/100/os-principais-equipas-que-foram-construindo-newsletter; accessed on 2 December 2021)

e.

Se

o

cellista

fora

saído

da

orquestra,

eu

seria

if

the

cellist

be.pst.sbjv.3sg

exited from.the

orchestra

I

be.cond.1sg

no

mínimo

degolada

em

cena (São Paulo)

in.the

minimum

beheaded

on

scene

‘If the cellist had left the orchestra, I would at the very least be destroyed on the scene.’ (http://www.alexandresanttos.com.br/RitaLeeAutobiografia.pdf; accessed on 2 December 2021)

f.

Tinha

o

rosto

sulcado

de

rugas,

como

se

já

have.pst.3sg

the

face

furrowed

of

wrinkles

as

if

already

fora

entrado

em

anos (BrPt.)

be.pst.sbjv.3sg

entered

in

years

‘His face was covered in wrinkles, as if he had already aged considerably.’ (http://www.educadores.diaadia.pr.gov.br/arquivos/File/2010/literatura/obras_completas_literatura_brasileira_e_portuguesa/VISCONDEDETAUNAY/INOCENCIA/INOCENCIA_TEXTO.HTML; accessed on 2 December 2021)

However, as these examples show, fora and related forms occur once again in both informal and formal registers and a variety of styles: a sports webpage (16a), online newspaper (16b), entertainment website (16c), online newsletter (16d), autobiography (16e) and a novel (16f). The most natural explanation of these facts, and the distribution of counterfactual be in general in Portuguese, is to interpret them as a residue of a once more generalized use of be with unaccusatives on par with what was argued for Latin American Spanish. In this light, the otherwise exceptional retention of fora and related forms with subjunctive and conditional value across a variety of registers and styles now finds a natural explanation, inasmuch as they have been preserved as part and parcel of the retention of the archaicising distribution of unaccusative be when these subjunctive and conditional values of fora were still very much productive.

2.1.3. Summary

In summary, we have seen how the erstwhile distribution of be with unaccusatives has been exceptionally retained as a frequent option exclusively in counterfactual contexts in the Calabrian variety of Sant’Andrea and in non-standard varieties of Latin American Spanish and European and Brazilian Portuguese. In these latter varieties the distribution of counterfactual be in stative syntax neither appears to be recessive nor restricted to informal registers, despite not being reported in the formal literature on Romance auxiliary selection. In Latin American varieties of Spanish, we also noted some rarer cases of extension of counterfactual be to transitive and unergative participles, a development which has been taken to its ultimate conclusion in southern peninsular Spanish varieties (particularly those of Andalusia but also Murcia and Extremadura), where counterfactual be now surfaces indiscriminately with all verb classes. The relevant developmental stages of residual retention and progressive extension of counterfactual be can be plotted as in (17).

(17)	Stative --------------------------------------------------------------------------Active
	Sant’Andrea, Portuguese > LA Spanish > southern peninsular Spanish

2.2. Alguerès: have/be ⇒ be

We now turn to the Catalan variety spoken in Alghero in northwestern Sardinia. The few available descriptions of Alguerès auxiliary selection usually present a situation identical to that of Italian and Sardinian with haver ‘have’ and ésser ‘be’ distributed according to an active–stative alignment (Pais 1970, pp. 58f., 64, 77f.; Blasco Ferrer 1984, pp. 157f.; Palomba 2000, pp. 155f., 159f.; Moll 2006, p. 290), continuing an archaic pattern found in old Catalan (Sendra i Molió 2013, pp. 54, 56; GLC 2016, pp. 249, 847f.; Loporcaro 2016, p. 813). The most detailed and accurate description is found in Loporcaro (1998, pp. 117–24) who notes the selection of haver with transitives (18a) and unergatives (18b) in contrast to the use of ésser with unaccusatives (18c), inherent retroherent reflexives (18d), direct transitive reflexives (18e) and indirect unergative reflexives (18f). By contrast, indirect transitive reflexives oscillate between haver (18g) and, less frequently, ésser (18h). This same active-stative alignment is further evidenced in the distribution of participle agreement which can be controlled by undergoers, such as clitic direct objects (18a,g), and unaccusative subjects (18c-f), but not by agents, such as transitive (18a,g) and unergative (18b) subjects.

(18)

a.

la

pera […]

Si

l‘

ha

presa

vostè,

dongui-me-la (Alg.)

the.fsg

apple.f

if

it.fsg=

have.3sg

taken.fsg

you

give.imp=me=it.fsg

‘the pear […] If you’ve taken it, give it to me!’ (Bosch i Rodoreda and Sanna 1996, Les dues peres i el carretoner)

b.

després

que

ella

ha

ballat (Alg.)

after

that

she

have.3sg

danced.msg

‘after she has been dancing.’ (Armangué and Scala 1997, Conte del Sidaru de la dona del carrer de Sant Francesc)

c.

Vengut

és

a

l’

Alguer (Alg.)

come.msg

be.3sg

to

the

Alghero

‘He came to Alghero.’ (Armangué and Scala 1997, Al país de l’Alguer hi habita una minyona)

d.

No

me

só

mai

pentida. (Alg.)

neg

me=

be.1sg

never

repented.fsg

‘I’ve never repented.’ (Armangué and Scala 1997, La rondalla del cafiter del rei)

e.

Mos

sem

coneixuts (Alg.)

us=

be.1pl

known.mpl

‘We made each other’s acquaintance.’ (Corbera and Chessa 2009, Intervista 1)

f.

um

problema

que

no

me

só

posat (Alg.)

a

problema

that

neg

me=

be.1sg

posed.msg

‘a problem which I haven’t considered’ (Sari Bozzolo 1996, p. 194)

g.

a

ella

se

l‘

ha

esposada (Alg.)

dom

her

self=

her=

have.3sg

married.fsg

‘He married her.’ (Armangué and Scala 1997, La rondalla del cafiter del rei)

h.

se

són

esballats

lo

camí (Alg.)

self=

be.3pl

mistaken.mpl

the

way

‘they took the wrong road.’ (Prieto and Cabré 2010, Task 2)

As already noted in (2a–c), this is not, however, the full story, in that be may surface across the board in past counterfactual contexts,11 extending from unaccusative to transitive/unergative syntax. The only recognition of this situation is found in the two studies by Scala: he explicitly notes that the conditional and past subjunctive forms of have, e.g., hauria ‘have.cond.1sg’ and hagués ‘have.pst.sbjv.1sg’, are not generally used today in the spoken language in the formation of the conditional perfect and pluperfect subjunctive but are replaced by the corresponding forms of be, e.g., sigueriva/fóra ‘be.cond.1sg’ and fossi ‘be.pst.sbjv.1sg’ (Scala 2003b, pp. 41–43).12 An identical picture emerges from Scala’s (2003a) study, where he provides, without further comment, from p. 87 onwards, the complete paradigms of 131 regular and 72 irregular verbs, which highlight the preferred use of be over have in the future perfect, conditional perfect and pluperfect subjunctive. This is illustrated in

Table 1. Alguerès active compound paradigms of admitir ‘admit’ and anar ‘go’13.

	admitit ‘admitted’	anat(s)/-da/-des ‘gone.msg(pl)/fsg/fpl’
prs.pfv.ind.	he/has/ha/havem/haveu/han	só/sés/és/sem/séu/sónt
prs.pfv.sbjv.	hagi/hagis/hagi/hàgim/hàgiu/hagin	sigui/siguis/sigui/siguem/sigueu/siguin
plpf.ind.	havia/havies/havia/havíem/havíeu/havien	era/eres/era/érem/éreu/eren
fut.pfv.	sigueré/siguerés/sigueré/siguerem/sigueres/sigueran
plpf.sbjv.	fossi/fossis/fossi/fóssim/fóssiu/fossin
cond.pfv.	fora/fores/fora/fórem/fóreu/foren

with the compound paradigms for transitive admitir ‘admit’ (p. 92f.) and unaccusative anar ‘go’ (p. 96f.).

Although a phenomenon which once again has failed to attract the attention of both descriptive and theoretical linguists, the incidence of counterfactual be in Alguerès is pervasive. An examination of a range of sources, including oral copora (Bosch i Rodoreda and Sanna 1996; Armangué and Scala 1997; Ballone 2000–2008; Viaplana and Perea 2003; Corbera and Chessa 2009), digital atlases (Martínez Celdrán et al. 2003–2020; Prieto and Cabré 2010), online newspapers (Alguer.cat 13 December 2019–20 September 2021; ‘Narracions an algherés de Saldeyna’ 19 November 2014–19 June 2019 of Alghero Eco), online magazines (Revista de l’Alguer December 2019–April 2021) and modern plays (Ceccotti 2006; Sari Bozzolo 1996; Sari 2006), has brought to light 159 examples of counterfactual ésser ‘be’ in conjunction with transitives, unergatives and indirect transitive reflexives, verb classes which otherwise license haver ‘have’ outside of this irrealis context. This number rises even higher (by a further 110 examples) if we include unaccusative verb classes, but because they do not involve a detectable switch in auxiliary, their incidence is less significant to the outside observer. In (19) I illustrate a selection of representative examples of counterfactual be (namely, conditional forms in fora and past subjunctive forms in fossi) from a range of sources.

(19)

a.

No

el

fossi

mantovat,

jo,

ara (Alg.)

neg

it=

be.pst.sbjv.1sg

mentioned

I

now

‘I wouldn’t have mentioned it now.’ (Bosch i Rodoreda and Sanna 1996, La dona del minudu)

b.

Mai

me

fossi

dit

aqueeta

cosa! (Alg.)

never

me=

be.pst.sbjv.1sg

said

this

thing

‘I never would have imagined such a thing!’ (Ballone 2000–2008, Etnotesto 4)

c.

si

al

mes

d’

abril

m’

havessin

dit

que

forém

if

to.the

month

of

April

m=

have.pst.sbjv.3pl

said

that

be.cond.1pl

reseixits

a

organitzar

les

Festes

31

de

agost […]

no

li

succeeded.mpl

to

organize.inf

the

feasts

31

of

August

neg

dat.3sg=

fóra

escomitit (Alg.)

be.cond.1sg

bet.ptcp

‘if back in April they had told me that we would have succeeded in organising the celebrations for 31 August […] I wouldn’t have bet on it.’ (Alguer.cat., http://cat.alguer.it/noticies/n.php?id=154086; accessed on 2 December 2021)

d.

Si

jo

no

fossi

tengut

menester

de

advocats

jo

no

fóra

if

I

neg

be.pst.sbjv.1sg

had

need

of

lawyers

I

neg

be.cond.1sg

fet

lo

tràfic. (Alg.)

done

the

traffic

‘If I hadn’t needed (to pay) lawyers, I wouldn’t have got involved in money laundering.’ (Sari Bozzolo 1996, p. 216)

e.

ningú

fora

mai

sabut

arrés,

ningú

fora

mai

pansat

no.one

be.cond.3sg

ever

known

nothing

no.one

be.cond.3sg

ever

thought

de

rubà

lu

fíl

de

Misipel.li […],

ningú

fóra

mai

anat

an

of

steal.inf

the

son

of

Misipel.li

no.one

be.cond.3sg

ever

gone

in

garera,

ningú

fóra

molt

i,

aspetxalment,

ningú

fora

paldut

jail

no.one

be.con3sg

died

and

especially

no.one

be.cond.3sg lost

lu

bon

trabal […], i

no

fóra

mai

ragallat

a

ma

la

mullé. (Alg.)

the

good

work   and

neg

be.cond.3sg

ever

gifted

to

me

the

wife

‘nobody would ever have known anything, nobody would ever have thought about kidnapping Misipel.li’s son […], nobody would have ever gone to prison, nobody would have died and, above all, nobody would have lost their job […], and nobody would ever have given me their wife.’ (Alghero Eco, https://www.algheroeco.com/llao-cuntent-rundalla-de-capalla-7/; accessed on 2 December 2021)

f.

primé

no’

l

fóran

pistat,

sagon

lu

fóran

ajurat,

talcé

first

neg

him=

be.cond3pl

beaten

second

him= be.cond.3pl helped

perhaps

lu

fóran

salvat

acumpanyant-lu

a

l’

hospital. (Alg.)

him=

be.cond.3pl

saved

accompanying=him

to

the

hospital

‘first they wouldn’t have beaten him up, second they would have helped him, perhaps they would have saved him by taking him to hospital.’ (Alghero Eco, https://www.algheroeco.com/rundalla-de-capalla-24/; accessed on 2 December 2021)

g.

Ja

el

sabiva

que

vos

fóra

ofesa

‘Espia

dels

already

it=

know.pst.1sg

that

you=

be.cond.3sg

offended.fsg

spy

of.the

morts. (Alg.)

dead.pl

‘I knew that “Spy of the dead” would have offended you.’ (Ceccotti 2006, scene 2a)

Note that the active–stative alignment in the distribution of participle agreement reviewed in (18a–h) remains unaffected by the licensing of counterfactual be, witness the agreement of transitive participle ofesa with the 2sg feminine reference of the object clitic vos in (19g) and of unaccusative reseixits with the understood 1pl null subject in (19c). However, in line with the other varieties reviewed in Section 2.1, the generalization of be in counterfactual contexts, although today predominant (cf. Scala 2003a, p. 41f.), is ultimately optional, inasmuch as have is still found (for explanation, see Section 3.3.4). Illustrative in this respect are examples such as (20): in (20a), for instance, have (viz. havessi; cf. also 19c) occurs in the protasis rather than be (cf. fossi in 19d) but not in the apodosis, where the expected counterfactual be surfaces. Similarly, (20b) offers us a near minimal pair, where counterfactual have in the first sentence is replaced by be in the second.

(20)	a.	havessi		tangut	la	pussibiritat,	ja	hi	fóra
		have.pst.sbjv.1sg		had	the	possibility	already	dat.3=	be.cond.1sg
		dunara	la	munera	pé	sa	ubrí	un’	ativitat. (Alg.)
		given.fsg	the	money	for	self=	open.inf	an	activity
		‘had I had the possibility, I certainly would have given him the money to start up a company.’ (Alghero Eco, https://www.algheroeco.com/antoni-arca-lu-raso-rundalla-de-capalla-39/; accessed on 2 December 2021)
	b.	an	vira	d’	ella	havariva	pugut	fé	qual sa sia cosa.	Tot
		in	life	of	her	have.cond.3sg	been.able	do.inf	whatever	all
		fóra	pugut	fé	an	vira	d’	ella. (Alg.)
		be.cond.3sg	been.able	do.inf	in	life	of	her
		‘in her life she could have done anything. She could have done absolutely anything in her life.’ (Alghero Eco, https://www.algheroeco.com/rundalla-del-capalla-28/; accessed on 2 December 2021)

In summary, we see that, in contrast to southern Calabrian, Spanish and Portuguese, the distribution of counterfactual be in Alguerès does not have its roots in an archaicizing residue of a once generalized use of the auxiliary with unaccusatives but, rather, represents the extension and concomitant generalization of the still fully productive stative auxiliary of an active–stative split, which is exceptionally suspended under marked modal conditions.

3. Marking the Irrealis

3.1. Verb Movement in Realis and Irrealis Contexts

A major dimension of variation between northern and southern Romance concerns the ability of T to probe V (Ledgeway 2009a; 2012, pp. 140–50; Ledgeway 2020; forthcoming; Ledgeway and Lombardi 2005, pp. 103–6, 2014; Schifano 2015, 2018; Ledgeway and Schifano, forthcoming: §21.2.3). For example, in Gallo-Romance varieties, such as Milanese (21a) and in Alguerès (21b), the finite verb raises to a clause-medial position within the T-domain from where it precedes all lower adverbs, such as always, and many higher adverbs. In southern varieties, such as southern Calabrian (21c), Spanish (21d) and Portuguese (21e), by contrast, T does not probe the finite verb, which remains low within the v-domain, as shown by its position to the right of higher adverbs and, in particular, many lower adverbs such as always.

(21)	a.	La	sua	miè	la	cuzina	semper. (Mil.)
		the	his	wife	scl.3fsg	cook.3sg	always
		‘His wife always cooks.’ (Schifano 2018, p. 257)
	b.	jo	tenc	semper	fam. (Alg.)
		I	have.1sg	always	hunger
		‘I’m always hungry.’ (Armangué and Scala 1997, La rondala del cafiter del rei)
	c.	Francu	sempa	fumava. (SCal.)
		Franco	always	smoke.pst.ipfv.3sg
		‘Franco always used to smoke.’
	d.	El	niño	siempre	llora. (Sp.)
		the	child	always	cry.3sg
		‘The child always cries.’
	e.	O	Pedro	sempre	vê	muitos	filmes. (BrPt.)
		the	Pedro	always	see.3sg	many	films
		‘Pedro always watches lots of films.’ (Schifano 2018, p. 72)

Turning now to irrealis contexts, we begin by noting that across Romance the finite verb targets a very high position within the clause. Evidence to this effect from a range of Romance varieties is adduced in Ledgeway (2009b; 2013; 2015; 2020, pp. 38–40), D’Alessandro and Ledgeway (2010a, pp. 2053–56), Ledgeway and Lombardi (2014), Taylor (2016, pp. 96–101) and Schifano (2018, pp. 42–51, 96–113, 237f.), where we witness, even in southern low V-movement varieties, a higher position of the verb in irrealis clauses than in realis clauses, as the examples in (22)–(25) illustrate, where the verb precedes the relevant adverb in the former context but follows it in the latter.

(22)	a	(Dicianu	ca)	Lello	sempe	fatica. (Cos.)
		say.3pl	that	Lello	always	work.3sg
		‘(They say that) Lello always works.’ (Ledgeway and Lombardi 2014, p. 37)
	b	Vuonnu	ca	Lello	fatica	sempe. (Cos.)
		want.3pl	that	Lello	work.3sg	always
		‘They want Lello to always work.’ (Ledgeway and Lombardi 2014, p. 37)

(23)	a.	Tice	ca	l’	Anna	già	u	sape. (NSal.)
		say.3sg	thatrealis	the	Anna	already	it=	know.3sg
		‘He says that Anna already knows.’ (Ledgeway 2020, p. 38)
	b.	Speru	cu	u	sape	già. (NSal.)
		hope.1sg	thatirrealis	it=	know.3sg	already
		‘I hope that she already knows.’ (Ledgeway 2020, p. 38)

(24)	a	(Spun	că)	mereu	munceşte. (Ro.)
		say.3pl	that	always	work.3sg
		‘(They say that) he always works.’ (Ledgeway and Lombardi 2014, p. 37)
	b	Vor	să	muncească	mereu. (Ro.)
		want.3pl	that	work.sbjv.3	always
		‘They want him to always work.’ (Ledgeway and Lombardi 2014, p. 37)

(25)

a.

Piedro

me

ha

dicho

que

su

mujer

siempre

hace

el

postre. (Sp.)

Piedro

me=

have.3sg

said

that

his

wife

always

make.3sg

the

dessert

‘Piedro has told me that his wife always makes the dessert.’ (Schifano 2018, p. 99)

b.

Piedro

quiere

que

su

mujer

prepare

siempre

el

postre. (Sp.)

Piedro

want.3sg

that

his

wife

prepare.sbjv.3sg

always

the

dessert

‘Piedro wants his wife to prepare the dessert.’ (Schifano 2018, p. 99)

As Schifano (2018, p. 237f.) notes, this higher position of the verb in irrealis contexts is particularly characteristic of perfective auxiliaries in many Romance varieties, the verbal class which most interests us here. In what follows I assume that in irrealis contexts the perfective auxiliary verb targets one of two tense-related functional heads in Cinque’s (1999) hierarchy (for a different cartographic implementation of a neo-Reichenbachian analysis of tense in terms of the heads T1 and T2, see also Giorgi and Pianesi 1997), although we shall revise this view slightly in Section 3.3.4. In particular, following Cinque’s (1999, ch. 4) claims about the fine structure of the sentential core, the highest portion of the IP-/T-domain includes projections specialized for past and future tenses, irrealis mood (viz. the indicative/subjunctive opposition; Cinque 1999, pp. 78, 88), root modality and various aspectual categories, as sketched in (26).

(26)	[TPPast > TPFuture > MPIrrealis > MPRoot > AspP… [v-VP V ]]

The three core paradigms we need to account for, the conditional (> conditional perfect, e.g., 2c), the past subjunctive (> pluperfect subjunctive, e.g., 1b) and the future (> future perfect, e.g., note 11 i.a-b), can then be said to all target one of the two T-related projections situated in the highest layer of the sentential core. Within a compositional analysis, Cinque (1999, pp. 190 n. 27) proposes that the conditional be interpreted as a result of the relevant verb raising from T°_Future to T°_Past, thereby also transparently deriving the future-in-the-past reading (27a; cf. also Iatridou’s 2000 diachronically-inspired account of the French conditional as a past imperfective attached to a future stem). By the same token, I assume that the past subjunctive form of the auxiliary raises from the head of MP_Irrealis to T°_Past (27b) and, similarly, the future auxiliary is licensed by raising to T°_Future (27c). The result is that in all three cases the licensing of the relevant irrealis auxiliary form, whether conditional, subjunctive or future, requires movement to (one of) the highest positions of the sentential core, either T°_Past or T°_Future.

(27)	a.	[TPPast Auxcond	[TPFut Auxcond	[MPIrrealis	[MPRoot…[AspP… [v-VP Auxcond]]]]]]
	b.	[TPPast Auxsbjv	[TPFut	[MPIrrealis Auxsbjv	[MPRoot…[AspP… [v-VP Auxsbjv ]]]]]]
	c.	[TPPast	[TPFut Auxfut	[MPIrrealis	[MPRoot…[AspP… [v-VP Auxfut ]]]]]]

3.2. Romance Auxiliary Selection

As argued at length in Ledgeway (2020, pp. 45–47; forthcoming, §2.4), a correlation deriving from the variability of V-movement surfaces in perfective auxiliary selection (cf. Manzini and Savoia 2005, II–III; Ledgeway 2012, ch. 7; 2019; Loporcaro 2016). In most northern Romance varieties, including what we have observed in Alguerès (18a–h), we see the continuation of an inherited active–stative split, whereby predicates with agent subjects select auxiliary have and predicates with undergoer subjects select auxiliary be. Simplifying somewhat, in southern Romance this active–stative distribution has, in most cases, been replaced by a nominative–accusative alignment variously involving the generalization of a single auxiliary, whether have (cf. 1a, 3a–b) or be (28a), or, alternatively, by a person-based system as in (28b), which generally contrasts be in the first and second persons with have in the third persons (though other patterns are attested).

(28)	a.	sɔŋgə /	si	/	ɛ…	maɲˈɲɛɐtə	/	məˈnuːtə. (Pescolanciano, Molise)
		be.1sg	be.2sg		be.3sg	eaten		come
	b.	So  /	si  /		a…	magnate /		minute. (Arielli, eastern Abruzzo)
		be.1sg	be.2sg		have.3	eaten		come
		‘I have/you have/(s)he has… eaten/come.’

In traditional work on the active–stative patterns of auxiliary selection within the Unaccusativity Hypothesis (cf. Perlmutter 1978; Burzio 1986), it has generally been assumed that auxiliary be represents the superficial reflex of a co-indexation relation between T and V, in accordance with the idea that unaccusative structures involve raising of the object to the surface subject position, as formalized in (29).

(29)	Auxiliary be is selected whenever (Spec)T is indexed with V(,DP)

Now, we have seen that in northern Romance and in Alguerès verbs overtly raise to the T-domain, an operation that automatically results in the co-indexation of V and T which, in accordance with (29), produces the observed sensitivity of the perfective auxiliary to the active–stative distinction. By the same token, we now also have a natural and principled explanation for the typical absence of active–stative auxiliation patterns in southern Romance varieties, such as Calabrian, Spanish and Portuguese, since verbs do not raise to T in the south but, rather, remain within the v-domain (cf. 21c–e). It follows from the PIC that T and V will never be co-indexed in these southern varieties and auxiliary be will never therefore surface as the result of an unaccusative structure. At the same time, this conclusion also explains why, alongside the generalization of one of the two auxiliaries as in (1a, 3a–b, 28a), many dialects of central and southern Italy (cf. Section 4.2) and some northern Catalan dialects display a person-driven auxiliary pattern as in (28b): given that in these varieties the auxiliary fails to raise to T but remains in the v-domain, the auxiliary under v finds itself in a local Spec-Head configuration with the subject externally or internally merged in SpecvP (cf. D’Alessandro and Roberts 2010), whose person feature it spells out in the PF-lexicalization of the auxiliary.

3.3. Irrealis be

3.3.1. Southern Calabrian, Latin American Spanish and Portuguese

Given our arguments above regarding the independently observed higher movement of irrealis verbs in Romance and the precondition on the availability of V-to-T movement for the licensing of an active–stative alignment in the perfective auxilaries, we now have a principled diachronic and synchronic explanation for the occurrence of irrealis be in southern Calabrian, Spanish and Portuguese. In particular, auxiliary be with unaccusatives is predicted to be licensed in these low V-movement varieties only if the verb can exceptionally raise to the T-domain (cf. 29). This is indeed what we find in these varieties with generalized have where the auxiliary switches to be uniquely in irrealis contexts where the verb exceptionally targets a high position within the sentential core. Focusing on southern Calabrian, Latin American Spanish and Portuguese, we have seen how in these varieties that irrealis be is restricted (almost) entirely to unaccusatives, a distribution which faithfully preserves an earlier synchronic stage when the former active–stative have–be alternation had given way to generalized have, following the establishment of low V-movement, but where be could still exceptionally surface whenever the finite auxiliary was able to target a position within the T-domain. In short, when the varieties shifted to low V-movement grammars yielding generalized have, the condition on be licensing in (29) could only be met in irrealis contexts, such as counterfactual conditionals, when the verb was forced to raise to the highest position within the T-domain (cf. 27a–c) to license its marked modal interpretation. This explains why it is often claimed that the distribution of irrealis be in these varieties represents an archaic feature, in essence the last residue of a former active–stative alignment.

3.3.2. Spanish of Southern Spain and Romanian

We have also seen that there are a small number of examples of irrealis be outside of unaccusative syntax in Latin American Spanish, in accordance with a development which has been taken to its extreme in the Spanish of southern Spain. In these latter Spanish varieties irrealis be is no longer a residual signal of an active–stative alignment but, rather, has been reanalysed as a distinctive marker of irrealis modality, which has been extended and generalized to all verb classes irrespective of the transitive–unaccusative distinction. Indeed, this diachronic explanation also carries over straightforwardly to modern Romanian, another low V-movement language (cf. Ledgeway, forthcoming), where the former active–stative alignment in the auxiliaries (Dragomirescu and Nicolae 2009; 2013; Ledgeway 2019, p. 376) has been replaced by generalized have (30a), except in the future (30b) and conditional perfect (30c) and in the subjunctive (30d), where the higher position of the irrealis auxiliary (cf. 24a–b) correlates with the selection of be (Avram and Hill 2007; Ledgeway 2014).

(30)	a.	Am   /	ai     /	a    /	am   /	ați   /		au	mâncat	/	venit. (Ro.)
		have.1sg	have.2sg	have.3sg	have.1pl	have.2pl		have.1pl	eaten		come
		‘I/you/(s)he/we/you/they have (/has) eaten/come.’
	b.	Voi   /	vei   /	va   /	vom /	veți  /		vor	fi	mâncat /	venit. (Ro.)
		fut.1sg	fut.2sg	fut.3sg	fut.1pl	fut.2pl		fut.3pl	be	eaten	come
		‘I/you/(s)he/we/you/they will have eaten/come.’
	c.	Aș     /	ai     /	ar     /	am   /	ați    /		ar	fi	mâncat /	venit. (Ro.)
		cond.1sg	cond.2sg	cond.3sg	cond.1pl	cond.2pl		cond.3pl	be	eaten	come
		‘I/you/(s)he/we/you/they would have eaten/come.’
	d.	Nu	cred	să	fi	mâncat	/	venit. (Ro.)
		neg	believe.3pl	that	be	eaten		come
		‘They don’t believe that I/you/(s)he/we/you/they have (/has) eaten/come.’

Once again, we can interpret the presence of be (viz. fi) in (30b–d) as a residue of an original unaccusative syntax which was exceptionally preserved under higher V-movement in irrealis contexts, but which was subsequently reanalysed as a specialized marker of irrealis modality and extended to all verb classes.

3.3.3. Alguerès

Like southern Calabrian, Spanish, Portuguese and now also Romanian, Alguerès also shows the generalization of be in irrealis contexts but, differently from these, we cannot interepret irrealis be as a residue of an original unaccusative syntax exceptionally preserved under high V-movement in irrealis contexts, inasmuch as Alguerès still has a fully productive active–stative (have–be) alternation (cf. 18a–h) with generalized high V-movement (cf. 21b). Rather, the origins of the distribution of irrealis be in Alguerès must be sought elsewhere and, in particular, I argue in PAT(tern) borrowing (cf. Matras and Sakel 2007) from the surrounding indigenous Sardinian dialects which also exhibit a marked strategy of irrealis—and more specifically counterfactual—marking. However, before exploring this assumption, we must first outline some core assumptions about counterfactuality.

3.3.3.1. Ingredients of Counterfactuality

Crosslinguistically there is a strong tendency for languages to develop specialized or dedicated markers of irrealis and, in particular, counterfactual modality (for an overview, see Karawani 2014, ch. 1; Sansò 2020), including specialized verbal morphemes (e.g., Hungarian), temporal morphemes (e.g., Romance), spatial morphemes (e.g., Burmese) and person morphemes (e.g., Blackfoot). As Karawani (2014, pp. 6, 42) notes, the temporal morphemes typically employed in counterfactuals are those borrowed from the past tense (cf. also Steele 1975; James 1982; Fleischman 1989; Dahl 1997; Sansò 2020, p. 410), to which we can also add the conditional, the so-called future-in-the-past, although in some languages past tense morphology is a necessary though not sufficent ingredient of counterfactuality which must be further combined with additional temporal, aspectual or modal morphemes (cf. Givón 1990). Consequently, while the imperfective past indicative is sufficient to license a past counterfactual reading of the auxiliaries in conjunction with the participle in the Calabrian example in (31a), in Italian the past auxiliaries must variously be bundled with subjunctive and future (> conditional) in order to convey counterfactuality with the participle in (31b).

(31)	a.	Si	nd avia	avutu	kju	ssuardi	m	avia	akkatthatu	na
		if	have.pst.ipfv.1sg	had	more	money	me=	have.pst.ipfv.1sg	bought	a
		makkina	nuova. (Polistena, SCal.)
		car	new (https://www.asica2.gwi.uni-muenchen.de; accessed on 15 December 2021)
	b.	Si	avessi	avuto	più	soldi	mi	sarei	comprato	una
		if	have.pst.sbjv.1sg	had	more	money	me=	be.cond.1sg	bought	a
		macchina	nuova. (It.)
		car	new
		‘If I had had more money I would have bought a new car’.

Following Iatridou’s (2000, p. 244) seminal study of counterfactuality, if the pluperfect is taken to include two past morphemes (cf. Steedman 1997), viz. the past auxiliary (marking the relation between event and speech time) and the past participle (marking the relation between event and reference time), then the participle can be taken to express a genuine temporal past interpretation (encoding an anterior event argument by virtue of its perfect aspect; cf. Karawani 2014, p. 108), whereas the past temporal form of the auxiliary represents a ‘fake’ tense,14 a key ingredient (together with the specialized if complementizer in the protasis) in the semantic composition of counterfactuality. Accordingly, this ‘fake’ tense serves as a marker of modal remoteness, not of past time semantics (cf. Palmer 1986a; Schlenker 2005), denoting exclusion from the actual world/time, i.e., the here and now (Iatridou 2000), or a non-actual veridicality presupposition that specifies that the world–time pair in which the proposition is true is different from the actual world and time of utterance (Karawani 2014). Now, in some languages the structure of counterfactual sentences shows that there can even be an additional layer of the ‘fake’ tense, resulting in the stacking of several past temporal morphemes. Illustrative in this respect are substandard varieties of British and American English where more than two layers of ‘fake’ past may co-occur in the protasis of past counterfactuals (Lambert 1986; Palmer 1986b; Fillmore 1990; Mittwoch et al. 2002, p. 752; Dancygier and Sweetser 2005, pp. 63–65; Ippolito 2013, p. 98; Zencak 2018, p. 30),15 witness the contrast in (32a–b).

(32)	a.	If he had told me, I’d have done it.
	b.	If he had’ve (‘d have/’d’ve/’d of/hadda/’da) told me, I’d have done it.

Whereas the protasis of the standard past counterfactual in (32a) contains just two layers of past, a ‘fake’ past had on top of a real past told, the substandard variant in (32b) presents three layers of past, a genuine past encoded by the participle told preceded by two layers of ‘fake’ past instantiated by various (reduced) forms of have (themselves subject to considerable orthographic variation as non-standard forms), e.g., (had >) ‘d + (> have) ’ve.16 Data like these beg the question of how to accommodate the third layer of past (cf. Ippolito 2013, 146 n.19) but arguably can be interpreted to suggest the existence of an additional functional projection within the highest layer of the sentential core to host this additional auxiliary morpheme. Indeed, based on data from languages such as Palestinian Arabic and Hebrew, Karawani (2014) argues for an additional functional projection (viz. TP2) to host such dedicated irrealis markers. This higher position needed to accommodate an additional ingredient of counterfactuality can be considered a marked option since, in contrast to the standard option in (32a), substandard examples such as (32b) ‘seem to convey both strong counterfactuality, and strong alternativity: that is to say, they highlight not only the irrealis nature of the situation referred to, but also the contrast between that situation and the absolutely opposite one which is assumed to actually obtain’ (Dancygier and Sweetser 2005, p. 64). Similarly, Karawani (2014, p. 86) observes, ‘[t]he addition of optional markers results in an emphatic effect hence a semantic/pragmatic effect that we are familiar with in other areas of grammar where the employment of optional, and thus redundant, markers produces emphasis.’ In short, I take substandard strings, such as (32b), to instantiate a marked structure, which includes an additional dedicated irrealis marker that is responsible for an emphatic effect which licenses a strengthening of the counterfactual inference and the impossibility of cancelling it.

Henceforth I take examples with marked irrealis morphemes such as (32b) to involve the activation of a specialized higher functional head lexicalized by dedicated irrealis markers which is not activated and hence unavailable in unmarked structures such as (32a). In particular, I follow Cinque (1999, pp. 84–86) in assuming that the highest portion of the sentential core includes projections specialized for speech act mood (e.g., hypothetical, optative, jussive, exclamative), evidential (quotative) mood and epistemic modality which, for expositional convenience, I conflate here into a single projection MoodP (cf. also Karawani and Zeijlstra 2010) generated on top of the fine structure already identified in (27) above. On this view, the fine structure of the clause can now be sketched as in (33).

(33)	[MoodP > TPPast > TPFuture > MPIrrealis > MPRoot > AspP… [v-VP V ]]

The three core values conflated within MoodP correspond precisely to the key uses of irrealis be reviewed above in Section 2.1 and Section 2.2 for southern Calabrian, Spanish, Portuguese, Alguerès and Romanian, namely (i) counterfactual, including the protases and apodoses of hypothetical clauses, by far the most frequent in our corpus (e.g., 1b, 3c,e, 7a, 9a, 12b, 19d), optatives (e.g., 9e), dubitatives (e.g., 7e, 14c, 19b,f), various types of dependent clause such as concessives and as if comparatives with the verb in the subjunctive (e.g., 3d–e, 7f, 8b–c,f,i, 11a, 12a,c, 13b,e, 14a-b, 16f, 30d) and embedded uses of the future-in-the-past (e.g., note 2 i.b, 16a, 19g),17 and root/embedded counterfactual uses of the conditional auxiliary (e.g., 2c, 8h, 9f, 10a, note 7 ii.a–b, 15a, 19a, 20b,e–f, 30c); (ii) evidential and quotative functions of the conditional (e.g., 11a, 16b–d); and (iii) epistemic uses of the future (e.g., note 11 i.a–b, 30b). In the representations in (27) above, we noted how the conditional can be interpreted as a result of the auxiliary raising from T°_Future to T°_Past (cf. 27a) and, by the same token, the past subjunctive form of the auxiliary involves raising from the head of MP_Irrealis to T°_Past (cf. 27b) and the future auxiliary raising to T°_Future (cf. 27c), such that licensing of the relevant irrealis auxiliary involves movement to either T°_Past or T°_Future. I take this to represent the unmarked situation which obtains under default licensing of irrealis auxiliary forms, such as was seen in (32a), which raises to T°_Past (34a), whereas marked structures, such as (32b), involve the additional lexicalization of Mood° (34b).

(34)	a.	…[MoodP	[TPPast had	[TPFut	[MPIrrealis had …	[v-VP had	…]]]]]
	b.	…[MoodP ‘d	[TPPast ‘ve	[TPFut	[MPIrrealis ‘ve …	[v-VP ‘ve	…]]]]]

3.3.3.2. Sardinian and Alguerès: Dedicated Irrealis Markers

Above I suggested that the generalized use of irrealis be in Alguerès can be interpreted as an example of PAT(tern) borrowing under language contact with the surrounding indigenous Sardinian dialects. More specifically, Nuorese and, more importantly for the discussion here, (northern) Logudorese dialects show a remarkable parallelism with the English data in (32a–b), as shown in (35a–b).

(35)	a.	Si	fis	vénnitu	pruskitho,	aíamos	mandicatu
		if	be.pst.ipfv.2sg	come	earlier	have.pst.ipfv.1pl	eaten
		imparis. (Lula, Srd.)
		together
	b.	Si	fis	istatu	vénnitu	pruskitho,	aíamos	áppitu
		if	be.pst.ipfv.2sg	been	come	earlier	have.pst.ipfv.1pl	had
		mandicatu	imparis. (Lula, Srd.)
		eaten	together
		‘If you had come earlier, we would have eaten together.’ (Jones 1993, p. 308)

Alongside the unmarked structure in (35a) in which the counterfactual reading is expressed by a single layer of ‘fake’ tense in the form of the imperfect past form of the auxiliary alone, we also find the marked strategy in (35b) with two layers of ‘fake’ past realized through a surcomposé (viz. doubly compound) form in which the finite auxiliaries fis ‘were’ and aíamos ‘had’ are reinforced by corresponding participial forms istatu ‘been’ and áppitu ‘had’. As noted in the literature (Pittau 1972, pp. 112, 156f.; Jones 1993, p. 83; Pisano 2010; 2016; Loporcaro 2016, p. 818; Mensching and Remberger 2016, pp. 285, 288), these forms are limited to past counterfactual conditionals such as (35a–b) and other past irrealis contexts such as unrealized wishes (36a) and main-/embedded-clause conditional perfects (36b; cf. Pisano 2010, p. 129f.), but are excluded from realis contexts and compound paradigms outside of the pluperfect (cf. Pisano 2010, p. 125).18

(36)	a.	tamˈbɛne	si	aˈiað	ˈapːi̯u	ˈɸropːi̯u. (Dorgali, Srd.)
		if.only	if	have.pst.ipfv.3sg	had	rained
		‘If only it had rained!’ (Pisano 2010, p. 130)
	b.	no	aˈi	ˈapːi̯u	maj	ˈkretːi̯u	k	ˈɛre
		neg	have.pst.ipfv.1sg	had	ever	believed	that	be.pst.ipfv.3sg
		ˈrːikːu. (Nuoro, Srd.)
		rich
		‘I never would have believed he was rich!’ (Pisano 2010, p. 130)

Although Jones (1993, p. 308) considers these surcomposé forms ‘as purely stylistic, having no effect on meaning’, something of an odd observation for a variety which functions almost predominantly as a spoken code, Pisano (2010, p. 130f.) provides compelling evidence from minimal pairs, such as (37a–b), to show that these forms have an emphatic strengthening effect very close to that noted above for substandard English counterfactual examples, such as (32b).

(37)	a.	si	vir	ˈbenːi̯u	ˈtue	nom	bi	vipː	aɳˈɖau
		if	be.pst.ipfv.2sg	come	you	neg	there=	be.pst.ipfv.1sg	gone
		ˈðɛo. (Nuoro, Srd.)
		I
		‘If you had come, I wouldn’t have gone.’ / ‘If you came, I wouldn’t go.’
	b.	si	viz	isˈtau	ˈβenːi̯u	ˈtue	nom	bi	vipː	isˈtau
		if	be.pst.ipfv.2sg	been	come	you	neg	there=	be.pst.ipfv.1sg	been
		aɳˈɖau	ˈðɛo
		gone	I
		‘If you had come, I wouldn’t have gone.’/*’If you came, I wouldn’t go.’

While at least for some speakers the unmarked structure in (37a) proves ambiguous between a past and simple counterfactual interpretation, this ambiguity is absent in the marked structure in (37b), where the additional layer of ‘fake’ tense forces an emphatic past counterfactual interpretation. We therefore take this as evidence for treating these latter structures as exhibiting a dedicated emphatic marker of irrealis, which involves the activation of MoodP, whose head is lexicalized by the higher of the two ‘fake’ past tense morphemes.19 On this view, the difference between (35a–b) can be sketched as in (38a–b) on a par with the representation in (34a–b) for English.20

(38)	a.	…[MoodP	[TPPast fis/aíamos	[TPFut	[MPIrrealis fis/aíamos…	[v-VP	fis/aíamos… ]]]]]
	b.	…[MoodP fis/aíamos	[TPPast istatu/appitu	[TPFut	[MPIrrealis istatu/appitu…	[v-VP	istatu/appitu…]]]]]

Turning now to Alguerès, there is considerable evidence to suggest that it has been in contact with Sardinian since at least as early as the 15th century. Such has been the influence from the surrounding Sardinian dialects that today Alguerès is said to be ‘full of Sardinianisms’ (Corbera 2003, p. 321) in all areas of the language, including its morphosyntax (Corbera 2003, pp. 325–28; Dessì Schmid 2017, pp. 466–68), witness its retention of an active–stative alignment in the perfective auxiliaries (cf. 18a–h), otherwise lost in most other Catalan varieties (Wheeler et al. 1999, pp. 355, 410; Moll 2006, p. 290; GLC 2016, pp. 249, 847f.; Loporcaro 2016, p. 813) but apparently preserved in Alghero under the pressure of the Sardinian model (Corbera 2003, p. 325; Dessì Schmid 2017, p. 467). During the course of the 16th and 17th centuries there was also a notable Sardinianization of Alguerès following the repopulation of the city by Sardinian-speaking communities from the surrounding countryside (Dessì Schmid 2017, p. 462). Against this background I would like to suggest that the structural PAT(tern), but not the MAT(ter), of the Sardinian dedicated irrealis marker was transferred to Alguerès. By way of illustration, consider again the minimal contrast in (20b), repeated here as (39).

(39)	an	vira	d’	ella	havariva	pugut	fé	qual sa sia cosa.	Tot	fóra
	in	life	of	her	have.cond.3sg	been.able	do.inf	whatever	all	be.cond.3sg
	pugut	fé	an	vira	d’	ella. (Alg.)
	been.able	do.inf	in	life	of	her
	‘in her life she could have done anything. She could have done absolutely anything in her life.’ (Alghero Eco, https://www.algheroeco.com/rundalla-del-capalla-28/; accessed on 2 December 2021)

In the first clause the past counterfactual is marked by the conditional form of the have auxiliary but in the second clause the auxiliary switches to be. Although the alternation could in principle be taken to represent a case of pure optionality given the minimally distinct nature of the two clauses, there is nonetheless a small but important difference between the two: the second involves a pragmatically marked word order in which the bare quantifier tot ‘everything’ is fronted to the left periphery, hence the English emphatic rendering ‘absolutely anything’, whereas the quantifier qual sa sia cosa ‘whatever’ in the first clause occurs in the unmarked postverbal object position and receives a more neutral reading. Significantly, this difference in information structure correlates with a differential selection of the two auxiliaries, ultimately suggesting, as we saw in relation to Sardinian, that there are two patterns of irrealis marking: (i) an unmarked strategy as exemplified in the first clause of (39), in which the irrealis auxiliary targets one of the two T-related heads (40a; cf. 37a), surfacing as have or be in accordance with the regular active-stative alignment; and (ii) a marked strategy as in the second clause of (39) where the irrealis auxiliary targets the head of MoodP (40b; cf. 37b), where it is invariably spelt out as be in the guise of a marked dedicated irrealis formative.21

(40)	a.	…[MoodP	[TPPast havariva	[TPFut	havariva [MPIrrealis …	[v-VP havariva…]]]]]
	b.	…[MoodP fóra	[TPPast Aux	[TPFut	Aux   [MPIrrealis …	[v-VP Aux…]]]]]

At the appropriate level of abstraction, the relevant difference between Sardinian (together with substandard English; cf. 34b) and Alguerès is that the dedicated irrealis marker under Mood° in the former instantiates the first-merge option, hence the appearance of two layers of ‘fake’ tense, whereas in the latter case the dedicated irrealis marker represents the overt spell-out of an additional movement operation which raises the verb from a T-related head to Mood°. The relevant feature bundle (viz (M_Irrealis) + T_Past/Fut + Mood) produced by this latter head movement operation is lexicalized as be, superficially yielding a single layer of ‘fake’ tense and overriding the underlying have-be auxiliary alternation. We thus see the extension and transferral of a Sardinian PAT(tern) of dedicated irrealis marking to the Catalan variety of Alghero, not a surprising result given the presence of the marked Sardinian irrealis structure in the localities of northern Logudorese and Anglona in close vicinity to the city (Pisano 2010, p. 125). Further evidence for the role of language contact in this development comes from the observation that Alguerès is the only variety of Catalan to display the generalized licensing of auxiliary be in irrealis contexts,22 thereby making the influence of Sardinian on this aspect of Alguerès grammar so much more plausible.

3.3.4. Southern Calabrian, Spanish, Portuguese and Romanian Revisited

In light of the above discussion about the structural and pragmatic distinction between unmarked and marked strategies of irrealis marking, we can now return to the irrealis uses of auxiliary be in southern Calabrian, Spanish, Portuguese and Romanian, whose synchronic distributions can be arranged in terms of the residual and extensional patterns of diachronic change in (41).

(41)	a.	[MoodP	[TPPast/Fut (be/)have	[MPIrrealis (Aux) [v-VP Aux PtPUnacc/Tr]]]] (S.Adr., Pt.)
	b.	[MoodP (be)	[TPPast/Fut (be/)have	[MPIrrealis (Aux) [v-VP Aux PtPUnacc/Tr]]]] (LASp.)
	c.	[MoodP	[TPPast/Fut have	[MPIrrealis (Aux) [v-VP Aux PtPUnacc/Tr]]]] (And.Sp.)
	c‘.	[MoodP be	[TPPast/Fut Aux	[MPIrrealis (Aux) [v-VP Aux PtPUnacc/Tr]]]] (And.Sp.)
	d.	[MoodP	[TPPast/Fut be	[MPIrrealis (Aux) [v-VP Aux PtPUnacc/Tr]]]] (Ro.)

In Andreolese and Portuguese low V-movement guarantees generalized have in accordance with (29), except in irrealis contexts where verbs raise to a high T-related position (41a), the sole configuration in which auxiliary be can still exceptionally surface as a residue of an unaccusative syntax, albeit alongside have whose generalization in realis contexts may also extend to irrealis contexts. Consequently, in Andreolese and Portuguese irrealis be represents nothing more than the optional output of an unaccusativity trigger residually licensed in contexts of modally-driven high V-movement. This same state of affairs is broadly attested in Latin American varieties of Spanish (41b) but with the difference that irrealis be is occasionally extended beyond unaccusative syntax to include transitive and unergative participles, the first signs of an incipient stage of reanalysis in which unaccusative auxiliary be progressively regrammaticalizes as a dedicated marker of irrealis lexicalizing the head of MoodP. Latin American Spanish therefore represents an intermediate diachronic stage between Andreolese and Portuguese on the one hand and Spanish varieties of southern Spain (in particular Andalusia) on the other. In the latter the extension of erstwhile unaccusative be has now been taken to its logical conclusion and reanalysed as a full-fledged dedicated marker of irrealis, completely divorced from its original stative syntax and semantics (41c’), which now functions as a marked emphatic strategy for strengthening non-actual veridicality in contrast to the unmarked strategy with generalized have (41c). The final stage in this development is represented by Romanian where the previous alternation between an unmarked strategy with have and a marked strategy with be has been lost in favour of the generalization of the latter, weakening its semantic force in the process. Consequently, irrealis be in Romanian no longer functions as an emphatic marker of non-actual veridicality since it no longer enters into a contrast with a non-emphatic have but, rather, is now an unmarked generalized marker of irrealis, presumably lexicalizing the head of one of the two highest T-related projections rather than Mood°, a case of downward regrammaticalization (cf. Roberts and Roussou 2003, ch. 5; Quinn 2009).

4. Other Patterns of Irrealis have and be in Romance

4.1. be ⇒ have

In a number of early dialects of southern Italy (Ledgeway 2000, p. 301 n. 22; 2003; 2009a, pp. 600–14; Formentin 2001; Cennamo 2002), as well as in old Spanish (Stolova 2006) and old Catalan (Massanell i Messalles and Mateu 2018, pp. 106–8; Massanell i Messalles 2020, p. 158; Salvà i Puig 2021, pp. 320f.),23 there operates a traditional active/stative split in the system of perfective auxiliation, whereby transitives/unergatives typically align with have and unaccusatives with be (42a–d). Nonetheless, the same texts show a progressive extension of the active auxiliary have to unaccusative syntax, replacing, in part, the traditional stative auxiliary be. However, the replacement is not random but shows a gradual encroachment of have on be in accordance with a sensitivity to a realis/irrealis modal distinction (for a discussion of similar Germanic facts, see Shannon 1995; McFadden and Alexiadou 2006a, 2006b, 2006c; 2010, pp. 394–99; Alexiadou 2015). In particular, the initial spread of have with unaccusatives in early texts appears, with very few exceptions, quite consistently to affect only those clauses marked as [–realis], typically containing a verb in the subjunctive or conditional (42a’–d’).

(42)

a.

erano

fuyute

a

li

templi (ONap., LDT 74.27–8)

be.pst.ipfv.3pl

fled.mpl

to

the

temples

‘they had run to the temples.’

a‘.

ben che

avessero

puro

foyuto

(ONap., LDT 238.29–30)

although

have.pst.sbjv.3pl

even

fled.msg

‘although they had even fled.’

b.

li

pili

ià

li

eranu

caduti (OSic., LDSG 117.25.2)

the

hairs

already

dat.3=

be.pst.ipfv.3pl

fallen.mpl

‘he had already lost his hair.’

b‘.

si

killa

dirrupa

avissi

caduta (OSic., LDSG 178.16.25)

if

that

cliff

have.pst.sbjv.3sg

fallen.fsg

‘if that cliff had collapsed’

c.

Si

el

sieruo

que

es

fuydo

mora

mucho (OSp., FJ)

if

the

servant

that

be.3sg

fled.msg

remain.3sg

much

‘If the servant who has fled stays a long time’

c‘.

si

ladrones

que

furtan

de

dia

&

de

noche

ouissen

if

thieves

that

steal.3pl

of

day

and

of

night

have.pst.sbjv.3pl

entrado (OSp., GE IV)

entered.msg

‘if thieves who steal by day and night had entered’

d.

totz

aquels

qui

eren

vengutz

per   él (OCat., Vides, 159)

all

those

who

be.pst.ipfv.3pl

come.mpl

for   him

‘all those who had come for him’ (Massanell i Messalles and Mateu 2018, p. 107)

d’.

si

vós,

sényer,

vos

agésets

vengut (OCat., Sereneta, Cartes II, c.15)

if

you

lord

you

have.pst.sbjv.2pl

come.msg

‘if you, sire, had come’ (Massanell i Messalles and Mateu 2018, p. 107)

Such uses of have in place of be in irrealis contexts have typically been explained as a way of cancelling the unmarked implication associated with unaccusative auxiliary be, which generally entails achievement of the resultant state, a reading tendentially incompatible with the non-actualized nature of irrealis situations and events. By contrast, auxiliary have is taken to mark a genuine experiential perfect and hence the only available form to express the perfect in irrealis contexts without forcing a resultant state interpretation. However, that is not to say that we do not find be in in irrealis contexts; rather, what we find is a degree of competition between the two auxiliaries, witness such minimal pairs as (43).

(43)

Declarasse

ancora

in

questa

ystoria

quali

ri

e

quali

duca

de

declare.3sg=self

still

in

this

history

which

kings

and

which

dukes

of

Grecia […]

e

altra

gente

avessero

andate

co

lloro

exercito

Greece

and

other

people

have.pst.sbjv.3pl

gone.pl

with

their

army

contra

Troyani […]

e

quali

ri

e

quali

duca

de

parte

de

Troyani

against

Trojans

and

which

kings

and

which

dukes

of

part

of

Trojans

fosseronce

andati

in

loro

defensa (ONap. LTD 48.11-5)

be.pst.sbjv.3pl

gone.mpl

in

their

defence

‘It is claimed in this (hi)story which kings and dukes of Greece […] and other people had (been said to have) gone with their army against the Trojans […] and which kings and which dukes of Troy had gone to their defence.’

As Shannon (1995, p. 143) puts it, ‘have is the more marked auxiliary that specifically denies—or at least calls into question, defocuses […]—the result. In this way there was a possible incipient semantic split here, with have indicating that the change was not attained, and be indicating that it was.’ Assuming Shannon’s interpretation of early Germanic to equally hold of early Romance, this would suggest that, at least for unaccusatives, we are once again dealing with an alternation between an unmarked strategy with auxiliary be, signalling a weak degree of counterfactuality (cf. fossero-) and a marked strategy with auxiliary have (cf. avessero), which functions as a dedicated ‘strong’ marker of irrealis modality. Following our analysis above in (41) for the extension of irrealis be, this generalization can be informally modelled in structural terms as in (44), where we take the switch from irrealis be to have with unaccusatives to represent the spell-out of a movement from the head TP_Past/Fut to the head of MoodP.

(44)

[MoodP (have)

[TPPast/Fut (be/)have

[MPIrrealis (Aux) [v-VP Aux PtPUnacc/Tr]]]] (ONap.)

Significantly, the account developed here is able to accommodate both the residue and extensions of irrealis be, seen above for southern Calabrian, Portuguese, varieties of Latin American and southern peninsular Spanish and Romanian on the one hand and the extension of irrealis have to unaccusatives in the early dialects of southern Italy, old Spanish and old Catalan on the other. As argued in Ledgeway (2003, 2009a), once have begins to extend to unaccusative syntax in irrealis modal contexts, it can then gain a foothold in the system from where it can progressively spread to realis contexts, yielding the generalized extension of have witnessed in the relevant Romance varieties today. Indeed, the data considered in this study highlight how the unidrectionality of the so-called irrealis effect (Shannon 1995, pp. 138–44), as formulated in (45), simply cannot be upheld.

(45)	If a language had a choice between have vs. be as a perfect auxiliary, in modal contexts have replaced be; the switch is unidirectional and is from be to have. (Alexiadou 2015, p. 123)

Rather, we have seen that the replacement (or encroachment) can proceed in both directions and that there are (at least) three ways in which be and have can emerge as specialized emphatic markers of irrealis modality lexicalizing the head of MoodP: (i) through the reanalysis of a residual unaccusative trigger preserved under high V-movement (cf. 41a–d); (ii) under language contact, as in the case of Sardinian influence on Alguerès (cf. 38a–b, 40a–b); and (iii) through the reanalysis of an aspectual distinction between resultative and experiential perfects (cf. 44). Indeed, not only are extensions in both directions found in different Romance varieties but even within varieties of the same language. This is the case in Spanish, where the reanalysis of an original aspectual avoidance strategy brought about the extension of have into the realm of unaccusative syntax in irrealis contexts (cf. 42c’,) from where it was able to spread subsequently to realis contexts. This is the situation found in the standard and in more northerly varieties of peninsular Spanish (cf. 3a). By contrast, in more southerly varieties of peninsular Spanish, in turn imported into Latin America, the isolated preservation of unaccusative be under the exceptional high V-movement associated with irrealis contexts was open to reanalysis as a dedicated emphatic marker of irrealis and extended beyond unaccusative syntax (cf. Section 2.1.2.2). Ultimately, which of the two auxiliaries is extended beyond its original realm of use and reanalysed as a dedicated emphatic marker of irrealis (lexicalizing the head of MoodP) proves irrelevant; rather what is crucial is that the original paradigmatic alternation between the two auxiliaries is (partially) overridden in favour of the generalization of a single auxiliary, whether have or be.

4.2. Person-Based Systems

As noted in Section 1, a common pattern of auxiliary alternation in the dialects of central and southern Italy involves a person split typically structured according to subdivisions based around the discourse participants. This most frequently surfaces as a simple binary split between the discourse participants (1st/2nd persons) marked with be and the non-discourse participants (3rd persons) marked by have (46a, 50a), although variations on this distribution and other patterns are possible (47a–b, 48a–b, 49a–b; for an overview see Ledgeway 2019, pp. 354–62). In principle, we might expect such person splits to cut across all temporal and modal specifications so that they occur not only in the present perfect, but also in the pluperfect and the counterfactual perfect (= conditional perfect/pluperfect subjunctive). However, contrary to the claims in Legendre (2010, p. 190), such a person-based distribution across all three paradigms (cf. Pattern 1 in

Table 2. Patterns of auxiliary distribution in central and southern dialects with person-based auxiliation.

Pattern	*1	*2	3	4
				a			b
				i	ii	iii
present perfect	H/B	H/B	H/B	H/B	H/B	H/B	H/B
pluperfect	H/B	H or B	H/B	B	H	H~B	B
counterfactual perfect	H/B	H/B	B	B	H	H~B	H

) is not attested. A priori that therefore leaves three other possible distributional patterns, as outlined in Table 2:

In practice, what we find are two principal patterns: (i) Pattern 3, where, in accordance with a simple [±realis] distinction, the person split surfaces in the present (46a) and pluperfect (46b) but not in the counterfactual perfect, which shows the generalization of be (46c), as in the Marchigiano dialect of S.Benedetto del Tronto; and (ii) Pattern 4, according to which the person split is restricted to just the present perfect, with different patterns of auxiliation in the pluperfect and the counterfactual perfect. In this latter case we can recognize two subtypes: in the first, which is the most frequent in Manzini and Savoia’s (2005, II–III: §5.5, §5.9) survey, pluperfect and counterfactual perfect display the generalization of the same auxiliary (Tuttle 1986, p. 268; Manzini and Savoia 2005, II:, p. 729; D’Alessandro and Ledgeway 2010b, §4; Ledgeway 2019, p. 357), typically be (Pattern 4a.i) in western and central dialects (47c–d) and have (Pattern 4a.ii) in eastern dialects (48c–d),24 and, much more rarely, free variation between the two auxiliaries in both paradigms (Pattern 4a.iii), as in the Abruzzese dialect of Castelvecchio Subequo (49c–d) and the Campanian varieties of Giffoni and Montecorvino (Manzini and Savoia 2005, III, pp. 25f.). In the second subtype (Pattern 4b), which is much rarer, we find distinct auxiliation patterns in the pluperfect (> be; 50c) and counterfactual perfect (> have; 50d), a distribution found in the Abruzzese dialects of Pescocostanzo and Popoli and the Campanian dialect of Morcone (Manzini and Savoia 2005, II, pp. 688–90; III, pp. 22f.).

(46)

a.

sɔ

/

ʃi

/

a   /

ʃɛmə /

ʃɛtə /

a

dərˈmiːtə /

be.1sg

be.2sg

have.3

be.1pl

be.2pl

have.3

slept

vəˈnuːtə. (S.Benedetto del Tronto)

come

b.

sɔvə   /

ʃivə   /

aˈvi    /

ʃaˈvamə /

ʃaˈvatə /

aˈvi

dərˈmiːtə /

vəˈnuːtə. (S.Benedetto del Tronto)

be.pst.1g

be.pst.2g

have.pst.3

be.pst.1pl

be.pst.2pl

have.pst.3

slept

come

c.

sarˈri    /

sarˈriʃʃə  /

sarˈri    /

sarˈrɛssəmə /

sarˈrɛʃʃə  /

sarˈri

dərˈmiːtə /

be.cond.1sg

be.cond.2sg

be.cond.3sg

be.cond.1pl

be.cond.2pl

be.cond.3pl

slept

vəˈnuːtə. (S.Benedetto del Tronto)

come

‘I/you/(s)he/we/you/they have/has // had // would have slept/come.’ (Manzini and Savoia 2005, II, p. 681)

(47)

a.

so  /

si   /

a   /

simo /

sete /

a

(p)par‘lato. (Amandola)

be.1sg

be.2sg

have.3

be.1pl

be.2pl

have.3

spoken.msg

b.

so  /

si   /

ɛ  /

simo /

sete /

ɛ

(v)viˈnutu/a/i. (Amandola)

be.1sg

be.2sg

be.3

be.1pl

be.2pl

be.3

come.msg/fsg/mpl

c.

ɛro    /

ɛri     /

ɛra   /

sɛmo   /

sɛte    /

ɛra

par‘lato /

vinutu/i. (Amandola)

be.pst.1sg

be.pst.2sg

be.pst.3

be.pst.1pl

be.pst.2pl

be.pst.3

spoken.msg

come.msg/mpl

d.

sarˈrio

/

sarˈriʃti

/

sarˈria

/

sarˈrimmo

/

sarˈrete   /

sarˈria

be.cond.1sg

be.cond.2sg

be.cond.3

be.cond.1pl

be.cond.2pl

be.cond.3

par‘lato

/

viˈnutu/-i. (Amandola)

spoken.msg

come.msg/mpl

‘I/you/(s)/he/we/you/they have/has//had//would have spoken/come.’ (Manzini and Savoia 2005, II, pp. 684f.)

(48)

a.

sɔ/ajə

/

ʃi

/

a

/

sɛːmə/aˈvɛːmə

/

sɛːtə/aˈvɛːtə

/

annə

be/have.1sg

be.2sg

have.3

be/have.1pl

be/have.2pl

have.3

(p)parˈlaːtə. (Secinaro)

spoken

b.

sɔ/ajə

/

ʃi

/

ɛ

/

sɛːmə/aˈvɛːmə

/

sɛːtə/aˈvɛːtə

/

annə

be/have.1sg

be.2sg

be.3

be/have.1pl

be/have.2PL

have.3pl

(m)məˈnuːtə. (Secinaro)

come

c.

aˈvɛvə

/

aˈviːvə

/

aˈvɛvə

/

ɛvəˈvaːmə

/

ɛvəˈvaːtə

/

have.pst.1sg

have.pst.2sg

have.pst.3sg

have.pst.1pl

have.pst.2pl

aˈvɛvənə

parˈlaːtə /

məˈnuːtə. (Secinaro)

have.pst.3pl

spoken

come

d.

aˈvɛssə       /

ɛˈvıʃʃə

/

aˈvɛssə

/

avasˈsammə  /

have.pst.sbjv.1sg

have.pst.sbjv.2sg

have.pst.sbjv.3sg

have.pst.sbjv.1pl

avasˈsatə     /

aˈvıssənə

parˈlaːtə

/

məˈnuːtə. (Secinaro)

have.pst.sbjv.2pl

have.sbjv.3pl

spoken

come

‘I/you/(s)/he/we/you/they have/has // had // would have spoken/come.’ (Manzini and Savoia 2005, II, pp. 691f.)

(49)

a.

ajə/sɔ

/

ʃi   /

a     /

emə   /

etə   /

annə/avə

have/be.1sg

be.2sg

have.3sg

have.1pl

have.2pl

have.3pl

(p)parˈlɛːtə. (Castelvecchio Subequo)

spoken

b.

ajə/sɔ

/

ʃi   /

e   /

emə   /

etə   /

annə/avə

have/be.1sg

be.2sg

be.3sg

have.1pl

have.2pl

have.3pl

(v)vəˈnɛutə. (Castelvecchio Subequo)

come.

c.

fevə/aˈvevə

/

fivə/aˈvivə.   /

fevə/aˈvevə …

parˈlɛːtə

/

be/have.pst.1sg

be/have.pst.2sg

be/have.pst.3sg

spoken

vəˈnɛutə. (Castelvecchio Subequo)

come

d.

fossə/aˈvessə. /

fuʃʃə/ɛˈviʃʃə   /

fossə/avessə …

parˈlɛːtə /

vəˈnɛutə. (Castelvecchio Subequo)

be/have.sbjv.1sg

be/have.sbjv.2sg

be/have.sbjv.3sg

spoken

come

‘I/you/(s)/he/we/you/they have/has//had//would have spoken/come.’ (Manzini and Savoia 2005, sec. II: 692f.)

(50)

a.

sɔ

/

ʃi

/

a

/

semmə /

seːtə /

ianə

dərˈmiːtə /

be.1sg

be.2sg

have.3sg

be.1pl

be.2pl

be.3pl

slept

məˈnuːtə. (Pescocostanzo)

come

b.

ɛra    /

irə    /

ɛra    /

ɛraˈvammə /

ɛraˈvaːtə /

ˈɛranə

dərˈmiːtə /

məˈnuːtə. (Pescocostanzo)

be.pst.1sg

be.pst.2sg

be.pst.3sg

be.pst.1pl    be.pst.2pl

be.pst.3pl

slept

come

c.

aˈvɛssə

/

ɛˈvıʃʃə

/

aˈvɛssə

/

avasˈsassəmə /

have.pst.sbjv.1sg

have.pst.sbjv.2sg

have.pst.sbjv.3sg

have.sbjv.1pl

avasˈsaʃtə

/

aˈvıssərə

dərˈmiːtə

/

məˈnuːtə. (Pescocostanzo)

have.pst.sbjv.2pl

have.pst.sbjv.2pl

slept

come

‘I/you/(s)/he/we/you/they have/has // had // would have slept/come.’ (Manzini and Savoia 2005, II, pp. 698f.)

Once again, we observe how the tendency to generalize one of the two auxiliaries as a dedicated marker of irrealis (grammaticalized as the expression of Mood°), which we have seen to be a common development across Romance, is also robustly attested in the modern dialects of central and southern Italy, displaying person-driven auxiliation. In all cases, the choice of the non-alternating auxiliary is ultimately arbitrary; what is relevant is that a classic person split in the present perfect (and sometimes in the pluperfect) invariably contrasts with a single auxiliary in the counterfactual perfect. Significantly, however, Pattern 2 in Table 2, consisting of the generalization of the person split to both the present perfect and counterfactual perfect to the exclusion of pluperfect, is not attested. Given, however, the presence of Pattern 3, in which the person split ranges over both the present perfect and the pluperfect to the exclusion of the counterfactual perfect, we can deduce that the generalization of be (4ai) or have (4aii), or free variation between the two (4aiii), in the pluperfect in Pattern 4 must represent an innovation based on an analogical extension of the relevant auxiliary from the counterfactual perfect. Presumably, this extension from the counterfactual to the pluperfect represents the initial step in a process of generalization which can lead to the eventual extension of the auxiliary across all paradigms, as was argued in Section 4.1 to have happened in the history of varieties such as Neapolitan, Sicilian, Spanish and Catalan. Indeed, it is not by chance, as originally observed by Tuttle (1986), that many of these dialects with person-driven auxiliation are flanked by more innovative neighbouring dialects with generalized be (51; Manzini and Savoia 2005, II, §5.7) or have (52; Manzini and Savoia 2005, II, §5.8), where we can assume that the original dedicated irrealis auxiliary has now extended its distribution to penetrate all realis paradigms.

(51)

a.

sɔŋgə

/

sï

/

ɛ

/

sammə

/

saːtə

/

sɔ

parˈlaːtə

be.1sg

be.2sg

be.3sg

be.1pl

be.2pl

be.3pl

spoken

mməˈnïutə. (Miranda)

come

b.

jɛva

/ jïvə    /

jɛva  /

jaˈvammə /

jaˈvatə /

ˈjɛvanə

parˈlaːtə

/

mməˈnïutə. (Miranda)

be.pst.1sg

be.pst.2sg

be.pst.3

be.pst.1pl

be.pst.2pl

be.pst.3

spoken

come

c.

sera

/

sera

/

sera

/

sarˈrïmmə

/

sarˈriːtə

/

ˈsɛranə

be.cond.1sg

be.cond.2sg

be.cond.3sg

be.cond.1pl

be.cond.2pl

be.cond.3pl

parˈlaːtə

/

mməˈnïutə. (Miranda)

spoken

come

‘I/you/(s)/he/we/you/they have/has//had//would have spoken/come.’ (Manzini and Savoia 2005, II, p. 761)

(52)

a.

adˈʤu

/

a

/

a

/

amˈmu /

aˈlitə

/

anˈnu

r(/dd-)urˈmutə /

(v)vəˈnutə. (S.Maria a Vico)

have.1sg

have.2sg

have.3sg

have.1pl

have.2pl

have.3pl

slept

come

b.

aˈlevə

/

aˈlivə

/

aˈleva

/

aˈlemwə

/

aˈlevwə

/

aˈlevənə

rurˈmutə /

vəˈnutə. (S.Maria a Vico)

have.pst.1sg

have.pst.2sg

have.pst.3sg

have.pst.1pl

have.pst.2pl

have.pst.3pl

slept

come

c.

aˈlessə

/

aˈlissə

/

aˈlessə

/

aˈlessəmə /

have.pst.sbjv.1sg

have.pst.sbjv.2sg

have.pst.sbjv.3sg

have.sbjv.1pl

aˈlesswə

/

aˈlessənə

rurˈmutə / vəˈnutə.

(S.Maria a Vico).

have.pst.sbjv.2pl

have.pst.sbjv.2pl

slept

come

‘I/you/(s)/he/we/you/they have/has//had//would have slept/come.’ (Manzini and Savoia 2005, sec. II: 779f.)

In summary, deviations from the classic person split in the counterfactual and, in many cases, also in the pluperfect appear to support the idea that person splits, possibly once systematic across all three perfective paradigms (cf. Pattern 1),25 have progressively been eradicated from the counterfactual in favour of the generalization of a single auxiliary (cf. Pattern 3). The latter represents a dedicated marker of irrealis lexicalizing the head of MoodP, which, in many cases, has subsequently been analogically extended to the pluperfect (cf. Pattern 4a), leaving the person split intact only in the present perfect. In turn, this development can provide the catalyst for an additional extension of that same auxiliary to the present perfect, giving rise to consistent single auxiliary systems, such as (51)–(52), in the same areas of central and southern Italy. Furthermore, note that the analogical extension of the irrealis auxiliary first to the pluperfect (cf. Pattern 4a), rather than to the present perfect (cf. absence of Pattern 2), can be explained by the fact that the imperfect, and hence also the pluperfect, whose auxiliary occurs in the imperfect in these varieties (and in Romance more generally), is a very frequent competitor to the formal irrealis paradigms of the conditional/future-in-the-past. As such, the pluperfect too is frequently employed with irrealis value and hence presumably finds itself subject to the same mechanisms of dedicated irrealis marking.

5. Conclusions

This article has examined the irrealis-conditioned distribution of auxiliary be across Italo-Romance and Romance more generally, a phenomenon which surprisingly has received little or no attention in either the descriptive or the vast theoretical literature on Romance perfective auxiliation. In particular, irrealis contexts have been shown to be especially productive in licensing auxiliary be, either as a residue of a former unaccusative syntax exceptionally retained under high V-movement (Andreolese, Latin American Spanish, Portuguese) or as an extension of this same residual unaccusative distribution to all verb classes (southern peninsular Spanish, Romanian), leading to its reanalysis as a dedicated marker of irrealis modality. Such specialized markers of irrealis modality lexicalizing the head of MoodP are known to be widespread crosslinguistically (e.g., substandard English, Palestinian Arabic, Hebrew), including in many Sardinian dialects which have grammaticalized a surcomposé construction involving two layers of ‘fake’ past as a dedicated emphatic marker of irrealis marking. Significantly, we have seen how extensive centuries-old contact between Sardinian and Alguerès has led to the transferal of this Sardinian pattern of specialized irrealis marking to Alguerès, which, uniquely among Catalan dialects, has generalized auxiliary be to all verb classes as an emphatic marker of irrealis. Finally, it was argued that the apparently distinct cases of generalization of irrealis have exhibiting the so-called ‘irrealis effect’ (Shannon 1995; Alexiadou 2015), where the resultative aspectual reading associated with auxiliary be is cancelled by the use of auxiliary have, are ultimately not unrelated to the cases of irrealis be: in both cases the auxiliary lexicalizes a specialized high functional head (Mood°) dedicated to the licensing of irrealis mood. This same line of reasoning can also be extended to many dialects of central and southern Italy where a person-driven alternation in the present perfect, and sometimes preserved in the pluperfect, contrasts once again with the generalization of one of the two auxiliaries in the counterfactual perfect which functions as a dedicated marker of irrealis.