1 Introduction

1.1Tuscan Phonology applies across word boundaries

3We explore the expansion of Florentine Gorgia into Western Tuscany, where lenition/Voicing seems much more established since Medieval time, the diachronic difference between the outcomes of lenition in Central and Western Tuscan.

4Moreover, we set the position of Tuscany within the Italo–Romance and Western Romance lenition (Voicing), establishing the relation between Tuscan lenition (Voicing and Gorgia), Western Romance and Northern Gallo–Italian lenition.

5In this study postvocalic lenition is treated along with the concept of locality, usually defined in phonology as a strict adjacency condition on related units (Piggott and van der Hulst 1997, van der Hulst 2018), i.e., phonological processes tend to apply locally.

7We study the relation of segmental lenition to the more complex objects in which voiceless stops and their strong/weak variants are syllabified and prosodified (syllables and the feet), in order to settle a locality domain on lenition. We describe the lenition phenomena within the framework of Government Phonology (= GP offspring and principles).

8We model lenition in the framework of the Coda Mirror Theory (Ségéral and Scheer 2001; Scheer and Zikova 2010) based on the linear Strict–CV theory (Lowenstamm 1996; Sheer 1994) as well as in the framework of a –linear models (i.e. structural–based models of the syllable), based on a recursive conception and a ‘maximal’ hypothesis (Sauzet 1994). The latter aims at the relation between syntactic locality and phonological locality in a cyclic Syntax/Phonology interface, while the former places at its core, as its essence, the term of lateral relations. We use the a–linear model to check if a structural–based phonology is relevant to explain the locality for Tuscan lenition. We aim through these models to set the conditions and predictions for Tuscan lenition/fortition (also depending on sandhi phenomena), based on the observation of diachronic and synchronic data.

9Finally, we also ask within a foot–based analysis if the locality conditions are only dependent on the segmental phonological substance or Tuscan lenition is foot–sensitive (within a quantity–sensitive, left–dominant, trimoraic trochee (QS–ld).

10Structural patterns of lenition have already been observed in different languages. In Franconian dialects for instance, consonant Voicing depends on foot–structure and it is foot–sensitive (see Smith 2020; Kohnlein and Smith 2021; especially in foot–medial onsets, see Honeybone 2012; 2019; Iosad 2021 for North Germanic). Thus, in these languages the locality of lenition is defined at a trochaic structured templatic level. Katz (2021) also pointed out the interactions between the consonantal lenition and prosodic structure in lenition.

11In Tuscan stress is trochaic, syllables interact with trochaic feet. In Section 3, we analyze the Tuscan foot to determine if the trochaic prosodic forms are relevant for lenition and locality at segmental level in order to establish prosodic properties for consonantal Voicing and spirantization.

12For Medieval Tuscan and Italo–Romance varieties the data shown below and in the following sections (where not otherwise indicated) have been directly selected by the author from the OVI corpus http://tlio.ovi.cnr.it/​ (dir. Paolo Squillacioti and Pär Larson, based in Florence at the OVI CNR Institute), the digital database of the Medieval vernacular Italian dialects. We illustrate on a dataset of the OVI corpus the process of lenition/Voicing of Latin stops /p t k/. We also discuss through the OVI dataset the alleged Medieval origin of Gorgia/spirantization (Section 5).

1.2 The evolution of voiceless stops in Romance and the Tuscan puzzle. The basic pattern

15Therefore, non–Voicing it is considered normal in Standard Italian words like: lupo ‘wolf’ (Lt lupu), cipolla ‘onion’ (late Lt cepulla), vita ‘life’ (Lt vita), dito ‘finger’ (Lt digitu), aceto ‘vinegar’ (Lt acetu), and so it should be in Tuscan dialects (see Rohlfs 1979, 1966 § 151, 155, 162, 163, 198, 208, 202, 260).

17The La Spezia–Rimini line, as Romance and Italian isogloss, is an important point of the dialectal partition, however this line should not be understood rigidly (Bruni 1984: 290; Fanciullo 2007; Vignuzzi 2010b; Russo 2015; 2022a).

(1) Intervocalic Stop Lenition in Medieval Tuscan:

• 8 <fuogo> (Andrea da Grosseto (+9 occ.)), <fuoco> (Fiori di filosafi).

Bifurcated outcomes (13^th c.)8

V_V [ v́__v ] (Foot–Medial)

Lt focu /–k–/ > [–g–] It. fuoco ‘fire’

(a) Tuscan <fuogo> [ˈfwɔ:go] Voiced

(b) Florentine <fuoco> [ˈfwɔ:ko] Voiceless

2 Tuscan Lenition (Onset Voicing) in Medieval dialects: principles, parameters and scales

2.1 The position disjunction {#,C}__ and __{#,C}. Strength– and Weakening–type environments

22This lenition/Voicing is also encountered in Tuscan dialects allophonically in phonotactics (across word boundaries) already according to Franceschi (1965), thus it is not limited to cases of lexicalized Voicing in a small number of words.

23For Formentin (2010), various instances of Medieval Florentine Voicing, inherited from Standard Italian (lago ‘lake’, luogo ‘place’, riva ‘shore’, padre ‘father’, madre ‘mother’, strada ‘street’, etc.) are to be explained because of a 5^th–6^th centuries influence, from the Northern Italian linguistic type.

24The idea of capturing Tuscany in the Northern Italian Voicing domain was first challenged by Rohlfs (1966, § 151, 194–212; 1979), then especially by Giannelli and Savoia (1978–80), Giannelli (1976/2000, 1978, 1983, 1988; Wanner and Cravens (1980); Franceschini (1983), Maiden (1995).

26We emphasize the existence of two lenitions corresponding to two phases in the diachrony of Tuscany characterized by the co–presence of a Voicing–defined lenition and a lenition Gorgia/Spirantization–defined after the 16^th century. Nevertheless, we show below that it existed a more extensive postvocalic Voicing in the Medieval period set back by Florentinization through Gorgia aspiration in the present days.

27As we show below based on first–hand data collected, Tuscan lenition (Voicing) in weak position ((V_{ r l j w}V) word internal and across boundaries) was extended into the Middle Ages even in Florence, when the Florentine Gorgia was not yet attested. Later, it was especially kept in Western Tuscany (mainly in Lucca and Pisa).

28An Element Theory study of Tuscan lenition/Gorgia in GP theoretical framework is done by Bafile (1997), who uses the loss of the privative closure edge element |ʔ| to represent lenition (see Russo and Ulfsbjorninn (2020) on Neapolitan lenition, who also used the occlusivity feature ‘edge’|ʔ| to express the loss of phonological primes in lenition processes according to positional strength).

30The position known as Strong Position is called by Ségéral and Scheer (2001) the Coda Mirror within the framework of Strict–CV theory as they support the idea of a shared identity among strong positions ({#, C}) mirrored in the phonological representation by the coda lenition {#, C}: the consonant located in the two strong positions is preceded in Strict–CV representation by an empty governed nucleus ({#, C}__ = ø__):

Coda Mirror (Strong Position): (+ Lic, –Gvt) Fortition (ø__)

Postvocalic (Weak Position): (+Gvt, –Lic) Lenition (vCv)

31The Coda Mirror considers that there exist two weak positions, the intervocalic position (vCv) and the coda position (__{#, C}) vs. two strong positions: initial and post–consonantal ({#, C}__ = ø__).

32The two mechanisms of Government (Gvt) and Licensing (Lic) should discriminate the Coda Mirror environments ({#, C}__ = ø__) from its opposite distribution, the coda (neither licensed nor governed: (–Gvt, –Lic). This theory of lenition is called the Mirror Coda precisely because it reduces the two disjunctions of the Strong Position and the Weak coda Position ({#, C}__ vs __{#, C}) to a single symmetrical phonological object (ø__ vs __ø). According to this theory, this is a natural consequence of the lateral relations. The intervocalic (V_V) position is governed and not licensed: (+Gvt, –Lic) (according to Scheer and Zikova 2010).

33Nevertheless, an onset can be found in strong position (see the a–linear representation in (2) with syntactic bracketing)

34(2) Positional onsets (syllable-initial):

but it can also occur in weak (postvocalic) position:

Yet, lenition can modify a syllabic onset when it is in postvocalic position (… V [_σ C V …) while a syllabic onset is not lenited in initial position (#[_σ C V …) or after a coda… C [_σ C V ...

2.2 Medieval Tuscan lenition–Voicing in postvocalic position: a shift towards reduction

36In Section § 1.2, we focused on the fact that postvocalic lenition–Voicing is generally believed do not occur below La Spezia–Rimini line, used as a proposal for the subdivision of the Italo–Romance linguistic domain (see Table 1):

Table 1 Non–Voicing /p t k/ Tuscan Italian (expected below La Spezia–Rimini line)

37We also emphasized in Section 1.2 that the Italian descendants of the Latin focu, locu or ripa (with an intervocalic singleton Latin voiceless stops) offer examples with a striking development: Tuscan Medieval dialects, and in some cases modern Italo–Tuscan (SI = Standard Italian), have bifurcated outcomes of original Latin intervocalic stps /p t k/ (Table 2):

Table 2 Medieval Tuscan bifurcated outcomes in V [_σC V or vCv position

Foot–Medial	[v́__v]	HL (Parse–ft)
	Voiceless		Voiced
<fuoco>	[ˈfwɔ:ko]	<fuogo>	[ˈfwɔ:go]	‘fire’
<ripa>	[ˈri:pa]	<riva>	[ˈri:va]	‘shore’
<spica>	[sˈpi:ka]	<spiga>	[sˈpi:ga]	‘ear’
<laco>	[ˈla:ko]	<lago>	[ˈla:go]	‘lake’
	[v́__v]	LLL
<lacrima>	[ˈla·krima]	<lagrima>	[ˈla·grima]	‘tear’

We are dealing here with one of the longest–running debates in historical Italian and Romance phonology.

39One must notice on this map that Siena and Pistoia belong today to Central Tuscany.

42As we have seen in (1) and Tab. 1, 2, for the Italian word <fuoco> ‘fire’ [ˈfwɔ:ko], we find in Medieval Tuscan also occurrences with the voiced stop [ˈfwɔ:go] from Lt focu (see Tab. 3):

Table 3 Lenition–Voicing /k/ in vCv Medieval Tuscan [ˈfwɔ:go] Lt focu

(σμμ: σμ)UnFt	[ v́__v ]	Foot–Medial
Tuscan	fuogo	[1268 Andrea da Grosseto]
Tuscan	fuogo	[1294 Guittone Rime]
Florentine	fuogo	[13th c. Cronica fiorentina (+1)]
Pisano	fuogo	[1327 Statuti Pisani]

In terms of CVCV or Strict–CV theory (Scheer 2004, see Section 2.1)), the lenition forms in Tab. 3 can be represented as in (3):

(3) Weak position vCv – Intervocalic consonant

Lt focu [ˈfwɔ:go]

43In (3), the intervocalic consonant is not licensed (according to the Coda Mirror v2 principles), the strength of government damages its target /k/, something that eliminates the possibility for /k/ to be realized as such, i.e., it causes lenition. In Strict–CV theory, Government (Gvt) and Licensing (Lic) are two antagonistic forces, Gvt inhibits the full melodic realization of the consonant. The consonant (the voiceless stop) is attached to a templatic position between two realized nuclei and thus, it is weakened intervocalically.

44In some cases, we have in SI a lexical restructuring of the Tuscan phonological outcome, as in the following examples of voiced/spirantized stops (captured into Italian lexicon) from Latin voiceless stops (see Fanciullo 2007: 168); the /v/ is the second stage of lenition sonority of voiceless Latin /p/ in V_V position in (4), see also Tab. 4–6:

The historical restructurings through lenition are attested quite early in Medieval Tuscan texts.

45This is the case for instance of the word list in (5). In these items, the lenited/voiced consonant has entered the Italian lexicon:

46With regard to It. <podere> (Noun in (5l)) and It. <potére> (Verb) from Lt *potere ‘can’ from potens), it is to be noted that in SI the verb (<potére>) survived with the voiceless consonant. However, in the meaning of ‘farm, land ownership’ (Noun), in SI survived the voiced consonant (It. <podere>). We also find some derivative lexemes from Lt *potere ‘can’ with the voiced consonant: podestà, poderoso ‘powerful’, see Tab. 4:

Table 4 It. potére / podére [N /V] from Lt *potere ‘can’ (from potens)

Foot–Initial		[ v__ v́ ]	(po[dé:re]UnFt
			σμ(σμμ σμ)UnFt
		It. <podére> [N]	‘farm’
Senese		(nel) podére	1235 Lira 3
Florentine		(se ‘l) podére	1279 Testamento Capraia
Pratese		(lo) podére	1285–86 Conti Sinibaldo
Florentine		poderétto	1308–30 Libro Giotto
		It. potére [V]	<podere> ‘can’
Tuscan		(lo) podére	13th c. Jacopo Mostacci
Tuscan		(e l’) podére	1268 Andrea da Grosseto
Florentine		(in) podére (del)	1260–61 Brunetto Latini (+9)
Lucchese		non podére	1295 Guidiccioni (+9)
Senese		podere	1262 Andrea Tolomei (+1)
But also <potere> ‘can’ [V] with voiceless /t/:
Tuscan	(non) potére		1230–50 Giacomo da Lentini
Senese	potére (andare/’go’)		1262, Andrea Tolomei

47On the irregular intervocalic Voicing found in the modern Italian lexicon, especially for cases such as ‘podere’ [N] ‘agricultural fund’, flanked by ‘potere’ [V] ‘can’, where the consonants /t d/ are distributed in an unmotivated way to different syntactic categories, see the hypothesis of reactions to Voicing as an anti–lenition process made by Carlucci (2015); on the interpretation of some bifurcated outcomes also Contini (1961); Manni (1994); Zamboni (1987, 2000).

Table 5 Lenition–Voicing of /p/ lexical restructurings scaled /v/ (vCv)

Table 6 Lenition–Voicing of /p/ It. ‘póvero/poor’ (Lt pauperu) – scaled /v/ (vCv)

[ v́__v ]	[ˈpɔ·vero]	Foot–Medial (LLL)
		(σμ σμ σμ)UnFt
Tuscan	póvera	1268 Andrea da Grosseto
Lucchese	póvero	1252–53 Alta maiestà
Senese	póvero	13th c. Ruggieri Apugliese
Florentine	povertádi	1260–61 Brunetto Latini
Florentine	povertáde	1275 Albertano volg (+33)

48Such results illustrated in the Tab. (4), (5) and (6) take us back to the development of voiceless Latin stops in Western Romance languages, to Castilian for example ‘riba/shore’ < Lt ripa, ‘escudo/shield’ < Lt scutu (moreover Italian ‘scudo’) French ‘ami/friend’ < Lt amicu (Castilian ‘amigo’), etc. (see Carlucci 2015).

Table 7 Lenition–Voicing of Lt –t– → [–d–] Lt. litu – spata

It. lido ‘beach’ and spada ‘sword’
[ v́__v ] [ˈli:do]/ [sˈpa:da] Foot–Medial (HL) (σμμ σμ)UnFt
Florentine	lído	1313 Ovidio Volg
Florentine	sído	1316 Lancia
Florentine	(Quido dela) Spáda	1211 Libro conti di banchieri
Lucchese	spáde	1213 Ritmo lucchese
Tuscan	spáda	1268 Andrea da Grosseto
	[v__v́] σ(da: ri)UnFt	Foot–Medial (HL) (σμμ σμ)UnFt
Florentine	spadári ‘swordsman’	13th c. Cronica Fiorentina

49Some of these Tuscan bifurcated lexical items have entered in Standard Italian with a voiced lexicalized stop, however in Medieval Tuscan this process appears to be of wider dimensions. Many lexemes show the voiced and the voiceless pattern. This is a case for instance of It. ‘spiga/ear (of wheat)’ Lt. spica <spiga> (as in SI) or <spica> and the derivative verb ‘spigolare/to edge’ (/k/ in V_V position):

Table 8 Lenition–Voicing of Lt –c– /–k–/ → [k/g] <spíga>/<spíca> ‘ear (of wheat)’

Foot–Medial	[ v́__v ] [sˈpi:ga]	(HL) (σμμ σμ)UnFt
Aretino	spíga	1282 Restoro
Pratese	(Bonacorso dala) Spíga	1285–86 Libro Sinibaldo
Florentine	spíghe	1292 Bono Giamboni
Florentine	spíga	13th c. Antidotarium Nicolai
Tuscan	spíghe	1333 Simintendi
Florentine	spigoláre ‘to edge’	1333 Ottimo
Tuscan	spíca	1314 Francesco da Barberino
Tuscan	spíca	1345–67 Fazio degli Uberti
Tuscan	spíca	14th–15th Bibbia

50Note that some words with the lenited variant accepted by Florentine are not always accepted by all Tuscany, what we say applies to Florentine <lago> vs. the voiceless outcome <laco> in Medieval Aretino (Tab. 9), and <aco> in Medieval Senese (Tab. 29), etc. However, there is no evidence to consider <lago> ‘lake’ (Tab. 10), or <ago> ‘needle’ (Tab. 27, 28) and <luogo> ‘place’ (Tab. 10), etc. Northern Italian loans.

Table 9 Lenition/Voicing of Lt –c– /–k–/ → [k/g] <lago> / <laco> ‘lake’

Table 11 Non–Voicing of Lt –c– /–k–/ → [–k–] <luogo> ‘place’ < Lt locu <ch> = [k]

51In all the examples above and below, Medieval Tuscan lenition–Voicing responds to distributional criteria (V_V) and fits the Coda Mirror theory as in (6), all weak consonants onsets are found in postvocalic position (after a realized nucleus, V_V) and are represented in CVCV lateral phonology with no recursive properties. The targeted consonant is governed, the nuclei are contentful:

(6) Intervocalic Consonant / Tuscan lenition–Voicing

Coda Mirror v2 general Template vCv

53For Lt acutus (Tab. 12), the data from OVI show numerous cases of Voicing also for final intervocalic /–t–/ besides /–k–/ in Northern Italy (in Veneto <agudi>), but not for Tuscan. The case of <aguto/aguti> with lenition/Voicing of the velar but not the coronal /t/ is a further confirmation that the Tuscan lenition does not have the same conditions as that on Northern Italian (Gallo–Italian) lenition. In Tuscany the lenition of voiceless Latin stops is asymmetrical, the velar stops are the ones most involved in the intervocalic Voicing process, hence the difference between Tuscan <aguti> (only [g]) and Veneto <agudi> ([g] and [d]), see Franceschini (1983: 138) pointing to the same pattern for Lt ficātu ‘liver’; Lüdtke 1961: 66).

Table 12 It. acuto [aˈku:to] ‘sharp’ Lenition/Voicing of Lt –c– /–k–/ → [g] Lt acutus

Foot–Initial	[ v́__v ]	(HL) (σμμ σμ)UnFt
	[aˈgu:to]	a(guμμ toμ)UnFt
Pisano	aguti	Doc. Pis. 12th c. Conto navale (+1)
Senese	aguti	1233–43 Mattasalà (+4 occ.)
Tuscan	aguto	1268 Andrea da Grosseto
Florentine	aguta	1274 Brunetto Latini
Florentine	aguti	Doc. Flor. 1286–90 Registro Cafaggio (+16)
Florentine	agute	1292 Bono Giamboni
Florentine	agutissimi	1292 Bono Giamboni
Florentine	aguti	1292 Bono Giamboni (+ 5)

54For Tuscan <aguto> in (7) the governing relation identifies the weak position vCv as in (6). The lateral relations match the empirical facts:

(7) Lenition: intervocalic onset /k/

55Our data show an advanced lenition–Voicing also in Central Tuscan. Florentine therefore seems to have experienced a stage of Voicing (and further Weakening) that we do not find anymore today. Regarding Western Tuscan, and especially Pisano, the treatment of intervocalic occlusives has been studied by Franceschini (1983). He examined Latin documents from Lucca dating back to the Lombard period and notes some cases of intervocalic Voicing coeval with documents from Northern Italy (for /k/ <exeguta> ‘executed’, for /t/ <memedipsum> ‘It. stesso/himself’, etc.).

57Serianni (1998: 68–69) also observed that in the dialects of Western Tuscany, during the Middle Ages, a few lenited consonants were unknown in Florentine and in literary Italian. In addition to <pogo> ‘little’ < Latin paucu (Prato, Pistoia, Lucca, Pisa), <oga> ‘goose’ < Latin *avica (Pistoia, Lucca, Pisa), <Mighele> (Pisa), <duga> ‘duke’ (Lucca, Pisa) and for the labial voiceless stop <cavestro> ‘thick rope’ (Lucca, Pisa) < Latin capĭstrum.

58As for the labial voiceless stop, for Western Tuscan Franceschini (1983: 136) adds <savore> ‘flavor’ (Lt sapore), <savere> ‘knowledge’ (Lt *sapēre), <cavretto> ‘little goat’ (Lt capra), <abrile> ‘April’ < Lt aprile, as forms widespread anciently in Tuscany (but not in Northern Italian dialects), /p/ → [b] Pisano <gubbie> < Lt cop(u)lae (also with initial Voicing /k/ → [g]), Lucchese <tiebbito> < Lt tepidu, etc.

Table 13 Voicing of /–t–/ → [–d–] D (Demontrative) It. codesto ‘this’ < Lt eccu tibi

Foot–Initial		[v__ v́ ] (HL) (σμμ σμ)UnFt
Florentine	codésto	1335 Boccaccio Filocolo
Florentine	codésto	14th c. Tavola ritonda
Florentine	codésto	1347–94 Giovanni dalle Celle
Tuscan	codésto	1345–67 Fazio degli Uberti
Pisano	codésto	1385–95 Francesco da Buti
BUT <cotesto>
Florentine	cotésto	1292 Bono Giamboni (+7)
Florentine	cotésto	1321 Dante Commedia
Senese	cotésto	13th c. Cecco Angiolieri

60The case of ‘codesto’ is also of crucial importance, in that, it is not Northern Voicing for certain, as there are no matches outside of Tuscany. The same phenomenon is found as we have shown, with Voicing after the Latin diphthong au as in paucu ‘little’, a form that is typically Tuscan.

2.3 Voicing Pattern in Phrase–Initial Tuscan Position. A real lenition in strong environments?

62Conversely, the phonological lenition of Northern Italian dialects does not concern the initial/strong position, whereas already in the Medieval Tuscan dialects (in the Pisano, Lucchese and Senese vernaculars) lenition–Voicing in phrase–initial position (#_ or ø__ according to the Coda Mirror) is widespread for Latin /k/ → /g/ in #_ (this initial Voicing is also common in Central–Southern Italian vernaculars), see the lateral representation in (8):

(8)

Lenition in initial onset /k/ → /g/ <gattivo> Pisano

How is government damaging? Lt captivu

63In (8), there is a problem with the voiced onset consonant and Government [+Gvt] relation, the Coda Mirror Theory here predicts [+Gvt] on V₁ from V₂, after an initial empty nucleus (ø__), since the Strong sentence–initial position should express the relation [+ Lic, –Gvt] on the voiceless stop /k/ as for fortition environments, and voiceless /k/ should be kept in an undamaged status. Conversely, the onset consonant gets harmed as it were in a weak context, we get lenition–Voicing in (5) in phrase–initial position #_, this is prohibited by the symmetry of the Coda Mirror Theory: ø__ vs. __ø.

Table 14 Lenition/Voicing of –t– It. fatica Lt *fatiga, der. fatigare/ ‘to tire’ > Medieval Tuscan [–d–] in vCv

Exactly as we expect for instance in Western/Gallo–Italic Milanese or in Northern Veneziano:

Table 15 Lenition/Voicing of –t– Northern dialects (Western/Gallo–Italic)

Foot– Initial	[ v́__v ] = [d]	(HL) (σμμ σμ)UnFt
Milanese	(tute) fadíge	f.h. 14th c. Elucidario
Veneziano	fadíga	1250 Panfilo
Veneziano	fadíga	1308 Cedola Zulian

Table 16 Lenition/Voicing of –t– Verbal forms It. ‘(af)faticare/to tire’ Lt fatigare

Senese	s’affadíga, s’affadigáro	13th c. Fatti di Cesare
Senese	affadigáti, affadigáto	1340 Ugurgieri
Senese	s’affadíga	1367–77 Caterina da Siena (+6)
Senese	affadigár(si)	1367–77 Caterina da Siena (+2)
Senese	affadigándo, affadigáti	14th Neri Pagliaresi
Senese	affadigárci	1309–10 Stat.Sen.Gangalandi
Senese	fadíghino	1343 Stat.Sen. Arte della Mercanzia
Senese	si fadígano	1309–10 Gangalardi
Tuscan	affadíga, affadigáti	1333 Simintendi

Table 17 Voicing of /–k–/ → [–g–] It. loríca ‘armour’ < Lt lorīca

Foot–Medial	[v́__ v]	HL ( σμμ σμ)UnFt
Senese	loríga	1340 Ugurgieri

69As for It. fatica/ fatigue [N] < * Lt fatiga also according to Guazzelli (1996) a lexical surviving with the voiceless non etymological variant (Lt g > /k/) in modern Italian can be interpreted diachronically as a hypercorrection. Based on the reversal functioning of pseudo–suffixes –iga (<fatica>) and –ica (<loriga>), it is quite clear that such pseudo–suffixes get underspecified for Voicing in Tuscan diachrony of lenition.

70Franceschini (1983: 133) points out hypercorrections from /g/ to [k] in loanwords such as <macagna> ‘shortcoming, flaw’ < Occitan maganhar and <macone> < Germanic <mago> ‘stomach’, present in Medieval Pisano and Lucchese, then preserved in Western Tuscan dialects. There is also retention of /k/ in initial position (despite widespread Voicing phrase–initial in #_, see below): <cabbia> ‘It. gabbia/cage’ < Lt cavea (in Medieval Lucchese, Pisano and Senese) or <callina> ‘It. gallina/hen’ < Lt gallina in Medieval Senese (ib. 134).

71We find also the expected /t/ and the devoiced [k] in many occurrences as bifurcated outcomes, most of them in Florentine, Pisano, Aretino, only a very few in Senese (Tab. 18):

Table 18 Non–Voicing < Lt *fatiga Medieval Tuscan /–t–/ (It. fatíca ‘effort’)

Tuscan	fatíca	1268 Andrea da Grosseto
Tuscan	fatíca	13th c. Tomaso di Sasso
Florentine	fatícha	1273 Doc. Flor. Lasciti
Florentine	faticóso	1274 Brunetto latini
Florentine	fatíche	1271–75 Fiori di filosafi
Florentine	fatíca	1282 Restoro d’Arezzo
Pisano	fatíca	1287–88 Trattati Albertano
Pisano	fatícha	13th c. Bestiario toscano
Pisano	faticósa	1309 Giordano da Pisa
Pistoiese	fatíca	1320–22 Carteggio Lazzari
Pratese	fatíca	1333 Simintendi
Senese	fatíca, fatíche	1288 Egidio Romano
Senese	fatíca	1340 Ugurgieri (+1)
Florentine	si fatíchino	14th c. Libro Pietre Preziose
Florentine	faticáre, faticáto	14th c. Andrea Cappellano

72In the case of <fatica> (Tab. 18) with regard to Western Tuscan, we must note that the voiceless stop /–t–/ is preserved in Pisano vs. <fadiga> widespread in other Tuscan areas including Western Tuscany (Lucchese).

73It seems that in the modern period, Pisa is one of the three buffer areas where lenition is curbed (with Arezzo and Siena), at least for /p t/ (see also Giannelli and Savoia 1979–80: 92).

74As it has been shown by Marotta (2008: 241), in Pisa we find today intermediate outcomes of Gorgia–Aspiration in comparison to today Florentine/Central Gorgia outcomes. According to us, this is the consequence of the interaction between the indigenous Voicing and Gorgia–weakening Florentinization, which makes Pisa a ‘peripheral’ (Western) area in relation to Gorgia (see below).

Table 19 Voicing of /–t–/ → [–d–] < Lt + atore

Foot– Initial	[ v́__v ] = [d]	(HL) (σμμ σμ)UnFt
	…[ˈdo:ri]	(doμμ riμ)UnFt
Senese	amministratori	1309–10 Stat. Sen. Gangalardi
Senese	amministradori	1318 Stat. Sen. Statuto Spedale
Tuscan	inperadori	1268 Andrea da Grosseto
Tuscan	pagadori	1268 Andrea da Grosseto
Tuscan	troffadori	1268 Andrea da Grosseto
Pisano	corridore	1337 Guido da Pisa
Florentine	corritore	1341 Libri Alfonso X

The Voicing of the stops seems to concentrate not only in Western Tuscany (especially Lucchese) but during Medieval period it is widespread in Tuscany, also in Senese and in Florentine areas (before the arrival of the Gorgia/aspiration):

Foot– Initial	[ v́__v ] = [d]	(HL) (σμμ σμ)UnFt
		Sta (duμμ:toμ)UnFt
Senese	staduto	1298–1309 Stat.sen. Addizioni
Lucchese	staduti	1376 Stat.lucch. Mercanti Lucca (+6)
Lucchese	staduto	1376 Stat.lucch. Mercanti Lucca (+1)

Table 21 Voicing of /–t–/ → [–d–] It. privato [A] < Lt. privatus, PP of privare ‘to privatize’

Foot–Medial	[ v́__v ] = [d]	(HL) (σμμ σμ)UnFt
	[priˈva:da]
Tuscan	privada	1268 Andrea da Grosseto (+2)
Tuscan	privadi	1294 Guittone
Florentine	privado	1284–87 Brunetto Latini (+1)
Florentine	privado, privade	1310 Bencivenni
Senese	privada	1298 Stat.sen. Arte della Lana
Senese	privadamente	1298 Stat.sen. Arte della Lana
Senese	privadi	1340 Doc. sen. Arte senese
Pisano	privadamente	1300 Palamedés (+ 5)

77The form with a voiced lenited onset consonant <privado> (Tab. 21) has a generally ancient Tuscan spread in Pisano, Lucchese, Senese, as well as <sollecido> (Tab. 22) in Pisano and in Senese (see Fransceschini 1983: 134).

78The paradigmatic exchanges between voiceless and voiced stops in the same (pseudo)suffixal series (see also –ido/–ito, Tab. 22, –ada/ata, Tab. 21, etc., and other forms such as strata/strada, where –ata is the root, Tab. 25–27) triggered hypercorrective mechanisms that in some cases promoted phonologization in the lexicon of one variant (see Fanciullo 2009, 2007). Thus, such correspondences facilitate the implementation of hypercorrective drives.

Table 22 Voicing of /–t–/ → [–d–] It. sollecito ‘responsive’ PP of Lt sollicitare

Foot–Medial	[ v́__v ] = [d]	(LμLμLμ)UnFt
Senese	solécido	1260 Lett. Sen. Aldobrandino
Pisano	solécido	1323 Lett. Pis. Bonagiunta

79We find a very few cases with /t/ for the It. strata < *strata ‘road’ which also gives ‘strada’ in SI but most cases have the voiced stop: as today in SI: ‘strada/road’ and It. ‘stradella’ ‘small road’ (Tab. 23):

Table 23 No Voicing of /–t–/ → [–t–] It. strata < *strata ‘road’

Foot–Medial	[ v́__v ]	(HL) (σμμ σμ)UnFt
	[sˈtra:ta]
Florentine	strata	1274 Brunetto Latini
Tusc. sud–or.	strata	1298 Questioni filosofiche
Pistoiese	strata	1353 Doc. Pist.
Pistoiese	strata	1353 Doc. pist.

80The voiced stop is already well established in the lexicon (also in derived forms: ‘stradella/small road’ stratella, stratariu, ‘stradicciuola/small road’ *strata + –iciu + –eolu), see Table 23, 24, 25) :

Table 24 Voicing of /–t–/ → [–d–] It. strata < *strata ‘road’

Table 25 Voicing of /–t–/ → [–d–] < stratella < stratariu < *strata + –iciu + –eolu

Table 26 It. pregare ‘to pray’ Lenition/Voicing of Lt –c– /–k–/ → [g] Lt precari

Foot–Initial	[v__v́]	(HL) (σμμ σμ)UnFt
	[preˈga:re]	pre(gaμμ reμ)UnFt
Tuscan	pregare	13th c. Rinaldo d’Aquino
Senese	pregare	Lett. Sen. 1260, Aldobrandino
Florentine	pregare	1260–61 Brunetti Latini
Pisano	pregare	1287–88 Trattato di Albertano
Senese	pregare	1288 Egidio Romano
Tuscan	(n’agio) precato	13th c. Rime Panvini
Florentine	(avea) precato	13th c. Rinuccino

2.4 Some more Voicing distinctions in the reflexes from Latin to Tuscan

82The lenition–Voicing pattern is increasingly pronounced for the velar stop, in Western Tuscan more than in Florentine, although some forms with the voiced stop have a generally ancient diffusion in Tuscany, as for the It. ago ‘needle’ from < Lt. acus and <figo> ancient variant of It. ‘fico/fig’. There are numerous forms with the voiced velar stop <ago>, more than the ones with voiceless stops <aco>, see Tab. 27–29:

Table 27 Voicing of /–k–/ → [–g–] <ago> < Lt. acus

Foot–Medial	[ˈa:go]	[v́__ v] (HL) (σμμ σμ)UnFt
Tuscan	ago	1276 Guinizzelli
Tuscan	ago	13th c. Rime Panvini
Tuscan	ago	14th c. Uccelli (+1)
Tuscan	ago	14th c. Trattato Falconi
Tuscan	ago	1314 Fr. Barberino
Tuscan	ago	1333 Simintendi
Senese	ago	1288 Egidio Romano (+1)
Senese	ago	1340 Ugurgieri
Florentine	ago	1292 Bono Giamboni
Florentine	ago	13th c. Tenzone Davanzati
Florentine	ago	1310 Bencivenni
Florentine	ago	1315 Novellino
Florentine	ago	1321 Dante Commedia
Florentine	ago	1370 Boccaccio Decameron
Pisano	ago	1304 [Stat. Pis.] Arte della lana
Pisano	ago	1321 Arte mercatanti
Pisano	ago	14th c. Barlaam
Pisano	ago	1342 Cavalca (+1)
Pisano	ago	14th c. Ovidio
Pisano	ago	1385–95 Francesco da Buti (+13)
Pratese	ago	1330–40 Lucano Volg.
Lucchese	ago	1376 Stat. Lucch. Mercanti

Table 28 Voicing of /–k–/ → [–g–] <aghi> ‘needles’ PL < Lt. acus

	<gh> = <g> [–g–]
Florentine	aghi	1338 Valerio Massimo
Tusc. Occ.	aghi	14th c. Itinerarium Volg (+4)
Tuscan	aghi	14th c. Contemptu mundi
Tuscan	aghi	14th c. – 15th c. BibbiaVolg

Table 29 No–Voicing of /–k–/: <aco> < Lt. acus

Florentine	aco	14th c. Tesoro Volg.
Senese	aco	1340 Ugurgieri (+1)
Senese	aco	14th c. Bestiario
Senese	aco	1378 Caterina da Siena (+1)
Tuscan	aco	14th c. – 15th c. Bibbia
Aretino	aco	1282 Restoro

83For It. fico ‘fig’<figo> and <fico> Lat. ficu, we find the voiced forms early in anthroponyms, see Tab. 29; this lexeme with voiced variant is already found in Medieval Italian (1321, Dante, Commedia, Inf. 33, v. 120 [che qui riprendo dattero per figo/ Who here a date am getting for fig]). However, Dante’s example is less interesting, since it could depend on the literary text in rhyming triplets. In fact, the <figo> example in Dante (Tab. 30) is rhyming with Alberigo:

Tercets, three lines stanzas: aba 1321, Dante (Inf. 33)

v. 118 Rispuose adunque: I’ son frate Alberigo

Then he replied: I am Friar Alberigo

v. 120 che qui riprendo dattero per figo

 Who here a date am getting for fig

Table 30 Voicing of /–k–/ → [–g–] Lt. ficu <figo>

84The form <miga> < Lt mica is also found by Franceschini (1983: 132–133, n. 4) in Medieval Lucca texts. It is a Western Tuscan form as well as generally Old Tuscan. However, Franceschini reports (ib.) the retention of the voiceless consonant in <mica> and <micca> (see Figure 2) for Medieval Pisano (Western Tuscan), which is opposed to <miga> Lucchese and Boccaccio’s <miga> (Tab. 31). But he points out inversely <botteca> Lucchese vs. <bottega> Pisano and generally Tuscan (Lt apothēca).

Table 31 Voicing of /–k–/ → [–g–] <miga> ‘crumb’ [N] and [Neg] = emphatic negation

Foot–Medial		[v́__ v] (HL) (σμμ σμ)UnFt
[N] and [Neg] [ˈmi:ga]		< Lt mica
Florentine	miga	1339–4 Boccaccio Teseida
Florentine	(non) miga	1334 Boccaccio Diana
Florentine	miga	14th c. Tavola Ritonda (+3)
Florentine	miga	1362 Pucci
Florentine	miga	1370 Boccaccio Decameron (+3)
Lucchese	miga	13th c. Luoghi Santi
Pisano	mica	13th c. Nocco di Cenni
Pisano	mica	13th c. Leggenda aurea

85The form <mica> penetrates standard Italian as a negation, and it is widespread in other Northern Italian dialects during the Middle Ages already as an emphatic negation with the expected Gallo–Italian lenition–Voicing (Tab. 32):

Table 32 Voicing of /–k–/ → [–g–] <miga> ‘crumb’ in Medieval Northern dialects

[Neg] = emphatic negation		< Lt mica
Veneziano	(no m’alegro) míga [Neg]	12th c. Proverbia que dicuntur
Veneziano	no fe’ míga ben [Neg]	1250 Pamphilus
Lombardo	non è míga ga... [Neg]	1274 Bescapè
Genovese	laxero’ míga [Neg]	1311 Cocito

Table 33 Voicing of /–k–/ → [–g–] It. bigóncio [biˈgont͡ʃo] ‘container’ Lt. /k/ *bicongiu

(9) Lt bi ‘two’ + congius – Initial Voicing #_not really Phrase–Initial

87The ancient environment of the early initial Voicing in phonotactics appears crystallized here in (9). Furthermore, in <bigóncio> the lenition–sonority is Foot–Initial with the prosodic trochaic pattern (HL) (σ_μμ σ_μ)_UnFt as in [skoˈdɛlla], which allows us to state that lenition is needed both Foot–Initial [ v́__v ] or in Foot–Medial onset [v__v́ ], see Tab. 34.

Table 34 Lenition/Voicing of Lt –t– scutĕlla → [–d–] It. ‘scodella/bowl’ [skoˈdɛlla] (vCv)

Foot–Initial		[ v́__v ]	(HL) (σμμ σμ)UnFt
	[skoˈdɛlla]		sco(dèlμμ laμ)UnFt
Flor.	scodèlla		1282 Restoro
Flor.	scodèlle		1286–90 Registro Cafaggio (+7)
Flor.	scodelláio		1286–90 Registro Cafaggio
Flor.	scodellétta		1321–30 Cavalca
Flor.	scodellíno		1370 Boccaccio Decameron
Flor.	scodelliére		1390 Sacchetti Pataffio
Flor.	scodellína		14th Sacchetti
Flor.	scodellíne		13th c. Muscia

88A lenition in onset positions word– or foot–initial seems a counter–argument to the pattern where foot–medial position is considered the main target for strengthening. Foot–initial onsets may be weak in vCv position even if the trochaic foot acts as licensor of syllabic structure in Tuscan (see Section 3).

2.5 Tuscan Lenition–Voicing in Branching Onset. Branching constituents?

89Tuscan lenition–sonority of voiceless Latin stops is regular in branching onset C + r/l (Tab. 35, 36, 37) also in phonotactic environments. This also leads to the loss of supralaryngeal place and manner features, i.e., loss of stricture (or loss of elements and phonological primes), such as in Lucchese or general Tuscan <saramento> It. ‘sacramento/sacrament’ Lt  sacramentu, where /k/ → Ø, with the entire delinking of melodic material, also enhanced by the lack of licensing see Tab. 35:

Table 35 Lenition of Branching Onset /k/ → Ø It. ‘sacramento/sacrament’ Lt  sacramentu

90The Coda Mirror Theory v2 (2010) only covers/models the lenition of single onsets, lenition of branching onset has been dealt with in Strict–CV afterwards, especially by Brun–Trigaud and Scheer (2010), Scheer and Brun–Trigaud (2012a). In Strict–CV Theory a branching onset (obstruent + liquids) exhibits a lateral relation between melodic segments, called the Infrasegmental Government (= IG, Scheer 1999, 2004: § 14), see in (10a/b):

(10) Lenition Branhing Onset in V_V position

(a) Cr – No prediction (non–local)

91The IG circumscribes the empty nucleus that it spans and it is responsible of its emptiness, however, IG does not trigger a segmental effect. Moreover, in terms of the Coda Mirror Theory, when the branching onset is in intervocalic position, the /r/ is governed and licensed by V, which makes impossible to get a prediction within this theory about the lenition of the voiceless stops (see Brun–Trigaud and Scheer 2010; Scheer and Brun–Trigaud 2012a). This happens because in this configuration /k/ is not the local target of the lateral relations, government and licensing.

92This situation makes it hopeless to account for lenition of branching onsets in the Coda Mirror Theory as such. This fact leads Scheer and Brun–Trigaud (2012a) to consider an amended structure to justify lenition of branching onset, in which the nucleus enclosed in it governs the consonant in onset position even if this nucleus is empty:

(10) Lenition Branhing Onset in V_V position

(b) Branching Onset (amended structure, i.e. local)

94Outside the CVCV theory and in prosodic terms, on the other hand in Tab. 34 we have a recursive foot, with a derivational suffix added (–mentu) and the main stress shifted during derivation from root on H penult (men_σto)_UnFt. Lenition of /kr/ affects the more embedded foot (HL : see It. [ˈsa:kro] ‘holy’), which carried the prederivational stress muted in a secondary stress in the complex word It. [ˌsakraˈmento], Tuscan lenited [ˌseraˈmento].

Table 36 Lenition/Voicing of Branching Onset /–kr–/ It. lacrima ‘tear’ Lt. lacrima

Foot–Medial	[v́__ v]	LLL (σμ σμ σμ)UnFt
Florentine	lágrime	1260–61 Brunetto Latini
Tuscan	lágrima / lágrime (+5)	1268 Andrea da Grosseto
Tuscan	lagrimáre (+5)	1268 Andrea da Grosseto
Pisano	lágrime	13th c. Leggenda aurea
	But also <lácrima/–e>:
Lucchese	lácrime	13th c. Bonagiunta
Tuscan	lácrime	1230–50 Giacomo da Lentini
Senese	lácrime	13th c. Pietro Jacomi

Table 37 Lenition/Voicing of Branching Onset /–tr–/ Lt. patre, matre, latro

Foot–Medial	[v́__ v]	(HL) (σμμ σμ)UnFt
Tuscan	pádre It. padre ‘father’	13th c. Rinaldo d’Aquino
Senese	mádre It. madre ‘mather’	1235 Lira 3
Florentine	mádre	1260–61 Brunetto Latini
Florentine	ládro It. ladro ‘thief’	1260–61 Brunetto Latini
Tuscan	ládro ladróne, ladróni	1268 Andrea da Grosseto
Florentine	ladronéccio	13th c. Tesoro volg
Florentine	ladronéggio	13th c. Tesoro volg
Florentine	ladronézzo	13th c. Tesoro volg
	but also <latro> /–t–/
Tuscan	látro	13th c. Laude cortonesi

95In Tab. 36 and 37 the lenited onsets /–tr–/ and /–kr–/ are not at the end of the scale as in Tab. 35, we do not get loss of manner articulation features and the shift moves up to the sonority scales to get the voiced lenited variants.

96In an a–linear model where all segments are hierarchized, a process of syllabification structures segmental sequences, including branching onsets. In these, the sonorant in considered as a second vowel, a syllabic consonant [r̩] (see Sauzet and Brun–Trigaud 2012). The syllabification algorithm is based on the phonological sonority interpreted as a phonological prime. This approach gives to the syllabicity of the sonorant a phonological role:

(11) Branching Onset in an a–linear model

Bracketing in Syntax–Phonology interface → Structural Recursion

(la)((dr̩)o) Lt latro

97In (11) the phonological syllabification shows that /r/ is syllabic and a syllabic head of a recursive syllable. The abstract syllabification of /r/ in Branching onset is ((CV)V), which means a syllable embedded recursively in a larger syllable with a higher projection node in the X–bar type of arborescent structure. In this approach, the lenition of Branching Onset is not modular as it is in Strict–CV phonology, but transversal to the grammar components.

98This approach seems more logical because it is based on the cerebrum loquens (as pointed out by Sauzet and Brun–Trigaud (2012: 162).

99Summing up, the picture we gave with Tuscan dialects up to now points out one thing: in Tuscan dialects the lenition seems an indigenous development; in other terms, the voiced consonants from Latin voiceless stops in linear vCv or a–linear V [_σCV) weak intervocalic environments are indigenous.

101Contini (1961) had already pointed out the existence of a ring along peripheral Tuscany functioning in terms of lenition similar to what is defined Italia Mediana (Vignuzzi 2010, see above). He based his arguments particularly on initial lenition (lenited/voiced variants in #_ contexts) by establishing a relationship between the process of lenition in initial position with examples of /k–/ > [g–] in ancient Tuscan and Central–Southern texts.

3 Parsed trochaic feet, the Stress-To-Weight principle and the lenition/fortition scale

103We explain in this section lenition (Voicing) and fortition patterns within prosodic and metrical theory and how Tuscan syllables are parsed into feet. This is to check if the trochaic Italo–Romance foot built on word stress has an influence on the lenition/fortition outcomes and scales.

3.1 Oxytonic words: Lenition/Fortition and Apocope

104Lenition is at the source of a nominal oxytonic word–class in Medieval Tuscan Italian (and afterwards in SI), since it introduces the new Italo–Romance prosodic class stressed on last syllable (e.g. città ‘town’ < Lt civita(te)). This is displayed in Medieval texts from Tuscany where we note that lenition/apocope affect the serial suffixes –ate and –ute from the Latin nominal 3^th and 5^th declension inflectional paradigms attached to a radical.

105From Medieval Tuscan we get evidence of three orthographic representations for these inflectional series Lt –ate/–ute, two of them indicate lenition. In the first spelling the stressed <–à> or <–ú–> become final when the lenition triggers the apocope of the last syllable /–te/ (and /–tu/); in the second spelling, lenition is just represented through the Voicing of the coronal /–t–/ without syllabic apocope <–àde> and <–úde> (e.g. see It. ‘città/town’, ‘bontà/goodness’ < Lt bonita(te), ‘eredità/inheritance’ < heredita(te), ‘virtù/virtue’ virtu(te) > <virtù> and <virtude>), see Tab. 38 and 39.

106In the former case (<virtù>), lenition and apocope create a restricted class of words with a final stress prosodic pattern in Italo–Tuscan (then in SI), e.g. stress on the final syllable, which is a marked pattern in comparison to the unmarked prosodic trochee, on which Italo–Romance and SI are considered metrically based (den Os and Kager 1986; Repetti 1991, 1992, 1998; Russo 2011, 2013a/b). Indeed, in Italian the main stress usually falls on the penult or antepenult syllable.

107Prosodically, at a glance oxytonic forms create degenerate feet, inconsistent with the metrical structure of Italian and Tuscan where trochaic feet are supposed to be at least binary, bounded and leftheaded (Marotta 2000: 192; Russo 2011, 2013a), see the s/w head–marked metrical notation developed in Lerdahl and Jackendoff (1983):

108(12) s/w head-marked metrical notation (Lerdahl and Jackendoff 1983)

The constituent node of the foot is placed geometrically above the head (on the stressed rhyme).

109In standard metrical theory (Halle and Vergaud 1987) the structure of Italian and Tuscan oxytones (such as città ‘town’, virtù ‘virtue’) creates a degenerate (monosyllabic) foot (see c): a single foot consisting of a strong syllable.

(13) Oxytone words – Degenerate Foot?

(a) (tʃit.) (tá)_unary-ft ‘town’ degenerate foot: (tá)

(b) ((láa.)_σ (dro)_σ)_ft–bin ‘thief’ bisyllabic trochee

110In Italian and in Medieval Tuscan the final syllable of this class is stressed, and it allows to form a degenerate foot (12c), assuming that this parsing has been exhaustive. Since the degenerate foot is stressed, the requirement that any feet have a stressed head is still satisfied.

111However, stress–final words in a supposedly trochaic system is a problem for the metrical theory: degenerate feet are not acceptable templates, repair strategies frequently restore the binary division (see Kager 1995).

112In (12a) rhymes are bipositional for stress assignment ([ˈla:dro]), i.e. there is a bimoraic constraint on the stressed syllable. Syllables receiving the primary stress are bimoraic, with the tonic syllable taking the form V: or VC. As a consequence of this, we find long vowels in stressed syllables (such as 12a) in words stressed on penult syllables. 

113If the surface–stressed open penult is obligatory foot–headed, light–headed unary feet are excluded.

Thus, words like (t͡ʃit.)(tà) ‘città’ in (13) at a glance have a structure which does not fit with the metrical structure of Italian (Nespor 1993), where a minimal binary foot size condition is expected to hold and for which a binary bound foot left–headed (i.e., left dominant) is usually postulated (Roca 1999: 666):

*

…(σ σ)<σ>]_mot

114The simplest way to test for degenerate constituents is to examine monosyllables diachronically. It appears that in many cases monomoraic monosyllabic words are simply absent in Latin because this type of underlying representation is augmented by the addition of another mora/syllable.

116The left dominant binary foot of Latin is still at work in Italo–Tuscan (Jacobs 1994a; D’Imperio and Rosenthall 1999; Russo 2013a/b, 2019, 2021). The accent on the penult word is the unmarked pattern. The extrametrical nature of word–final syllables confines the Latin stress to antepenult and penult syllables. Latin quantity–sensitive system is a system that obeys the Weight–to–Stress principle (if heavy then stressed, Hyde 2011a: 1061; Prince 1990): if the penult is heavy, it is stressed; otherwise, the antepenult is stressed. Tuscan also seems a quantity–sensitive left dominant foot (QS–ld), the only innovation between the Latin and Tuscan stress systems is the oxytonic pattern.

117Binary feet can be identified as a general weight–stress parameter on the right edge of the metrical domain.

118To account for oxytonic and paroxytonic stress, Nespor (1993) proposes that Italian allows degenerate monosyllabic feet (= unary feet) in oxytonic words next to binary feet (bisyllabic trochees), which are stressed on their leftmost syllable in words like (tʃit.)(tà) ‘town’.

Table 38 Lenition, Apocope and oxytonic stress: <–à(te)>/<–à(de)>/<–à> an iamb, unary feet or a trochee?

/t/ → Ø	[ˈ ta:de]		(HL) (σμμ σμ)UnFt
No –Apocope			< Lt –áte
Florentine	hereditàde		1275 Albertano volg (+2)
Florentine	hereditàte		1275 Albertano volg (+2)
Pisano	hereditàde/ heredità		1309 Giordano da Pisa
Apocope	[ˈtaμ]/ [ˈtaøC]		H (σμμ)FtBin
Florentine	heredità		1284–87 Brunetto Latini
Pisano	heredità		1287–88 Albertano Volg
Senese	heredità		1309–10 Stat. Sen. Gangalandi (+22)
Lucchese	heredità		14th c. Stat.Lucch. Altopascio
Tuscan/Volt.	heredità		1348–53 Lett. Volt. Belforti (+8)
Tusc.Sud–Or.	heredità		1298 Questioni filosofiche

Table 39 Lenition, Apocope and oxytonic stress: <–à(te)>/<–à(de)>/<–à>

Moraic FtBin = μμ		[ˈtaμ]/ [ˈtaøC] H (σμμ)FtBin
–á(te) Apocope		Lt < sēcūrĭtāte (te) → Ø
Pisano	sigurtàμ	1287–88 Trattati Albertano (+3)
Pisano	sigurtàμ	1309 Giordano da Pisa
Pisano	sigurtàμ	1321 [Stat. Pis.] Ordine Mercatanti (+8)
Pisano	sigurtàμ	1327 [Stat. Pis.] Chiesa Sigerro (+2)
Pisano	sigurtàμ	1330 [Stat. Pis.] Breve Popolo (+4)
Pisano	sigurtàμ	14th c. [Stat. Pis.] Compagnia Crocione (+1)
Pisano	sigurtàμ	1322–51 [Stat. Pis.] Breve Ordine del mare di Pisa (+13)
Volterra	sigurtàμ	1348–53 [Lett. Volt.] (+4)
Senese	sigurtàμ	1367 Colombini
Tuscan	sigurtàμ	1294 Guittone (+2)

It must be noticed that in the oxytonic formation [sigurˈta] ‘safety’ Tab. 39 and 40, also /–k–/ is lenited to [–g–] in the second unstressed syllable.

Table 40 /–t–/ Lenition with No–Apocope (Stage 2) Lt < sēcūrĭtāte [sigurˈta:de] ‘safety’

120In Medieval Italo–Tuscan (and SI) the rarest ‘iambic’ context also becomes the most productive trigger for fortition, since it triggers the initial phonological gemination called Raddoppiamento Sintattico (henceforth RS). In other words, in a W₁W₂ sequence, the initial consonant of W₂ is a geminate if W₁ ends with a stressed vowel like in (13) (tʃit.) (tá), i.e., RS to apply requires that stressed syllables have branching rhymes.

122The RS consonantal position is metrically predictable, in Tuscan words like città its insertion is licensed by the final stress, following the Stress–to–Weight constraint.

123Indeed, in oxytonic words città–like which trigger the RS conditioned by final stress, an empty consonantal position responsible for initial gemination is inserted to satisfy the stress algorithm of Italo–Tuscan; in other words, the gemination process is linked to the assignment of the metrical structure In Medieval Tuscan, the RS rule triggered by stress to fulfil the foot requirement was already active, as you can see in (14):

(14) RS licensed by final stress (→ Clitic Onset Gemination (μ μ)_FtBin )

(a) Florentine pagolli ‘he payed them’ (1211, 1256, 1348–50...)

(b) Senese pagolli (1277–82)

(c) Pistoiese pagolli (1297–1303)

(d) Florentine rekolle ‘he brought them’ (1211)

• 75 <andollo a raccomandare> = It. lo andò a raccomandare ‘went to recommend him’.

(e) Florentine andollo (2^nd h. 14^th )75

(f) an(døl)(lo) a–linear model (_øl)(lo) Chain

(g) an(do_cv)(lo) linear model CVCV 2 Units

124In (14f) we give a representation of RS geminate in the a–linear model based on a Syntax–Phonology interface, which predicts an asymetric geminate, and in (14g) we give the representation in the linear Strict–CV model, in which an empty CV triggers a full geminate. Both models share the idea of an empty final position on W₁ available in Strong initial context to produce RS.

125We must notice that RS, also when it is stress–conditioned, always corresponds to a Strong Position, the post–coda [C.__ ], a context where lenition is disfavoured (hence gemination/fortition). Thus, RS functions as a context of lenition inhibition (see for inhibitions/lenition factors Honeybone 2019, 2005).

126We assume here that the final stressed syllable of oxytonic words città–like is a binary moraic trochee left–headed (this is for all words with a final stress), with 2 moras: H= μμ FtBin (=binary foot) (Russo 2021, 2019). Moraic feet are otherwise uneven trochees in Italo–Tuscan, an UnFt (= Uneven Foot) is a foot with three moras (μμμ). Thus, if we consider Tab. 40 and (15):

(15) Oxytones H= μμ FtBin

(a) heredità → (ta_μμ)_FtBin (Final stress)

(b) hereditàde → (ta_μμ:de_μ)_UnFt (Penult stress)

127In both cases (15a/b) feet are left–headed bounded trochees (no more than two moras in FtBin, no more than three moras in an UnFt). In both feet, stress licenses a position, vowel length in (12a) is derived to satisfy the bimoraic minimal foot, in (12b) consonantal length in the form of initial gemination of word₂ onset (RS) is also required to satisfy Foot Binarity in phonotactic conditions. In Italo–Tuscan the stress–weight mapping operates both in word–internal position and across word boundaries.

129(16) Classical Latin stress algorithm based on the Weigt_to_Stress Principle :

by changing it to a Stress–to–Weight Principle, however it still provides a left prominent trochaic foot which is quantity sensitive, i.e., responsive to syllable structure (Russo 2013a/b, 2019, 2021). This happens with the final stress (Tab. 38 and 39), which triggers the insertion of an underspecified (empty) consonantal position and determines its integration into the metrical structure via RS fortition = initial gemination (Russo 2013a).

131In standard metrical theory Halle and Vergnaud (1987) proposed a treatment of bounded stress in Classical Latin based on extrametricality and on the Bounded parameter (= BND) (see also Halle and Idsardi (1995). The system for the computation of antepenult and penult stress is the following:

(17) Treatment of bounded stress in Classical Latin (Standard Metrical Theory)

1. The final syllable is extrametrical

2. The settings for line 0 are [+ HT, +BND, right to left, left] HT = Head Terminal

3. Place the heads of the constituents of line 0 on line 1, etc.

132In this view, the extrametrical nature of word–final syllables confines the Latin stress to antepenult and penult syllables. In an optimality theoretic analysis of the stress system applied to the metrical standard theory (Halle and Vergnaud 1987), we would say that Italian has preserved Latin extrametricality of the final syllable (NonFinality, see Hyde 2007, 2011b; Kager 1999).

133In our development of the standard theory replacing the syllabic trochee with a moraic trochee, the word–final syllable is not always extrametrical, rather it is footed in a stress–window bearing penult and antepenult stress (see the uneven moraic trochee in (15b): (ta_μμ:de_μ)_UnFt). The extrametrical nature of word–final syllables/moras in Italo–Tuscan only applies when the prosodic window does not respect the cardinality of 3 from the right–edge to the left (in words of more than three syllables: the trisyllabic stress window from Latin (see Roca 1999, 2006; Russo 2019, 2021). In this case the final one/two moras/syllables are left unparsed on the right edge. This grid–based non–finality approach produces a similar pattern in word layers. It operates by inserting a boundary to the right of the rightmost element of the line 0 grid. x# line 0 when the stress is not in the final position, according to the maximum constituent construction parameter. The function of ICC (= Iterative Constituent Construction) is used here to characterize maximum ternary measures (the antepenult stress pattern), without generalization of the extrametricality as a prosodic edge parameter (Liberman and Prince 1977; Roca 2005; Hyde 2011b).

Head–based and non–finality approach does not prohibit stress on syllable–final moras. Head moras of a trochaic foot can be final.

Without extrametricality, the Tuscan pattern requires the quantity–sensitive ternary uneven moraic trochee parsed at the trisyllabic domain right edge (15b): HL (ta_μμ:de_μ)_UnFt.

When the penult is light, the ternary uneven moraic foot is constructed at the right edge of the word, resulting in antepenultimate stress (LLL proparoxytonic words).

135For Tuscan Italian with this window (= 3), one must therefore admit a persistence of the Latin binary or ternary moraic foot with left–head, where heavy syllables systematically attract the stress, according to a mirrored/symmetrical Stress–to–Weight Principle (stress occurs on the first mora (18), see Kager (1993), Hyde (2011a):

136(18) acts as a s/w syllable contour, the second mora being subject to initial gemination in RS contexts (W₂ with a consonantal onset in a left–to–right trochaic system). The internal prominence contrast is responsible for the behavior after the heavy syllable within the parsing window and parsing iterative algorithm.

137For quantity–sensitive metrical parsing we have postulated a specific rule, which assigns heavy syllables an asterisk on line 1; the analysis places a stressed syllable in the head position. This maximizes the idea that the stress reflects the projection of headed constituents.

138In Tuscan stress is also sensitive to the weight of prosodic word–final syllables; this confirms the trochee stress system. In oxytones <città>–like, stress occupies a heavy final syllable: we have argued (Russo 2013a/b, 2019, 2021) that an empty position is lexically present in the metrical structure of the oxytonic words of Tuscan Italian /<città>_C/ and that the quantity–sensitive stress generates its realization to satisfy the binarity of the final foot. Rhythmic lengthening applies to increase the duration of stressed syllable. In Medieval and Modern Tuscan (as well as in SI), this causes vowel lengthening word–internally and gemination of the following onset in RS contexts W₁W₂ within the prosodic word (φ). When an underlyingly light final syllable is stressed (<tà>), the syllable becomes heavy by RS–lengthening. RS is a prosodic licenser for otherwise stray position; it protects the latent consonantal position <tà>_C from deletion through Stray Erasure.

139We could also argue, following Kiparsky (1991), that a catalectic mora is assigned to the right edge of the radical. This catalexis would be motivated by the need to eliminate degenerate monosyllabic feet (see Jacobs 1994b; Kager 1995; Meinschaefer 2011, Russo 2013a): through a FtBin compensatory lengthening strategy (a ‘rhythmic’ lengthening) the RS is placed after the oxytones because of a catalectic mora licensed by stress, which allows its realization through initial gemination in RS contexts. This analysis eliminates the idea of monomoraic feet in the <città>–like and produces the correct stress pattern which is predictable in the case of word–final stress. The stressed vowels of underlyingly light syllables lengthen in the Tuscan trochaic system. Rhythmic lengthening indicates weight–sensitivity since stress avoids light syllables (Hyde 2006, 2011a/b).

3.2 Lenition, Stress and Foot. Evidence for prosodic conditions of lenition in Tuscan dialects?

140The hypothesis that stress triggers voicing/lenition of the following consonantal onset was made by Graziadio Isaia Ascoli (1873: 88ff., see discussion in Clark 1903, and recently Canalis 2014), and already discussed by Izzo (1972, 1980), who excluded the stress conditioning i.e., that stress play a role on lenition processes. He ascribed devoicing to a spontaneous development due to the effect of the prosodic and/or vocalic context.

141The degree of stress of adjacent syllables could affect the segmental position. Ascoli’s idea was that the tonic /a/ in the penult syllable triggers Voicing of the following onset, giving opposite pairs in SI such as (19a/b):

(19) Does tonic /a/ in the penult trigger Voicing?

UnFt: HL ((μμ)σ σ(μ)) = [ˈpa:dre] and [sˈtra:da]

(a) padre ‘father’ Voicing

strada ‘road’

UnFt: HL ((μμ)σ σ(μ)) = [maˈri:to]= ma[(ri:)(to)]_UnFt and [ˈkre:ta]

(b) maríto ‘husband’ No–Voicing

creta ‘clay’

144As we have seen in Section 1 and 2, this does not completely fit our (qualitative) data found in Medieval Tuscany where lenition involves such contexts. In each Table, in the previous sections, we have indicated the stress and foot structure explained in Section 3.1. The same applies looking at Voicing in Northern Italian dialects, see (20), where lenition is also represented with orthographic <dh> or with <d> instead of /–t–/ (<dh> indicates a fricative stage); in Medieval texts both graphemes are used to indicate lenition:

(20) Lenition in weaker Foot–internal (Voicing and loss of material Stop → Fricative)

Northern Italian dialects after a stressed high vowel (< Ascoli)

Lt maritus ‘It. mariti/husbands’ [maˈri:ti]

UnFt: HL ((μμ)σ σ(μ)) = [maˈri:ði] = ma[(ri:)(ði)]_UnFt

Veneziano <maridhi> (14^th c. Tristano Veneto)

Veneto <maridi> (1325 Armannino)

146Meyer–Lübke develops even more the stress theory, and interprets the stress role in determining patterns of lenition stating that the voiceless Latin stops remain voiceless when they occur after a tonic syllable (however see for instance (21):

(21) No–Voicing in posttonic syllables (Meyer–Lübke § 205)

voiceless consonant after the stress (It. amíco ‘friend’)?

UnFt: HL ((μμ)σ σ(μ)) = [aˈmi:ko] = a[(mi:)(ko)]_UnFt

Lt amicus

(a) It. amíco [aˈmi:ko] ‘friend’

Lt caput

(b) It. cápo [ˈka:po] ‘head’

When the onset precedes the tonic vowel, we should find instead the voiced consonant as in Tab. 41.

147According to this theory this would be the reason why we get the Voicing for the examples like: It. pagáre ‘to pay’ Lt pacare or It. padélla ‘pan’ Lt. patella:

Table 41 Lenition–Voicing of /–t–/ before the stress? Lt. patĕlla ‘pan’ [ v__v́ ]

pa(dèμlμlaμ)	[ˈdɛlla]	(HμμLμ)UnFt
[v__ v́ ]		Foot–Initial
Pratese	padèμlμlaμ	1288–90 [Doc. Prat.] Ragionato Cepperello
Senese	padèμlμlaμ	1301–1303 [Stat. Sen.] Statuto della gabella
Senese	padèμlμlaμ	1321–37 Chiose Selmiane
Florentine	padèμlμlaμ	14th c. Bencivenni
Tuscan	padèμlμlaμ	1333 Simintendi

148Again, it is hard to accept the idea if we have a look at data in Tuscany or in Northern Italy, where we find intervocalic Voicing in such positions (contradictions to this stress theory come even from Florentine), see Tab. 42:

Table 42 Stress does not prevent Voicing!

[v́ __v ]	(HL)UnFt	Foot–Medial
Cremonese	amíμμgoμ	13th c. Patecchio
Ligure	amíμμgoμ	14th c. San Gregorio
Florentine	amíμμgoμ	14th c. Sacchetti Rime

The same happens in words with antepenult stress: we find the Voicing of the onset following the stressed vowel (Tab. 43):

Table 43 < Lt pícula It. pece ‘tar’ Voicing in posttonic position

Foot–Medial	[v́__v]	(LμLμLμ)UnFt
Florentine	péμgoμlaμ	1321 Dante Commedia
Florentine	péμgoμlaμ	1321–22 JacopoAlighieri
Florentine	péμgoμlaμ	14th Pegolotti
Florentine	péμgoμlaμ	1366–72 Boccaccio Rubriche
Bolognese	péμgoμlaμ	1324–28 Jacopo della lana
Padovano	péμgoμlaμ	1390 Jacobus

In that case we have to notice that the rule fits in Tuscan Italian and in Northern Italian dialects even for prosodically similar words (see Izzo 1972), such as:

(22) Posttonic Consonantal Voicing (LLL)_UnFt [ˈredina] It. redina ‘bridle’

< Lt. retinere

It. (ré_μdi_μna_μ)

Pisano <rédina> (1287–88, Trattati Albertano volg)

149As we have seen in the previous section Italo–Tuscan feet are moraic (bimoraic or uneven) trochees. Lenition seems to occur in all types of intervocalic position, weaker foot–medial and stronger foot–initial. Medieval data show that lenition in intervocalic vCv can occur in multiple prosodic position: in foot–internal position after a stressed long vowel VV (li_μμdo_μ)_UnFt (HL), after a single V (ré_μdi_μna_μ)_UnFt (LLL), (ˈla_μgri_μma_μ)_UnFt (LLL), as well as Foot Initial [po_μ(ˈde_μμre_μ)_UnFt] (HL), see (23):

(23) Consonant weakening (lenition) and Prosodic environment

150(a) Foot–medial weaker foot–internal [ v́__v ] (HL)

[ˈri:va] (ri_μμva_μ)_Unft Lt ripa

[ˈla:go] (la_μμgo_μ)_Unft Lt lacu

[ˈli:do] (li_μμdo_μ)_Unft Lt litu

[ˈspi:ga] (spi_μμga_μ)_Unft Lt spica

[ˈpa:dre] (ˈpa_μμdre_μ)_Unft Lt patre

[ˈma:dre] (ˈma_μμdre_μ)_Unft Lt matre

[ˈla:dro] (ˈla_μμdro_μ)_Unft Lt latro

151(b) As well as after unlengthened/unstressed V:

Foot–initial [ v__ v́ ]

[sco_μ(ˈdɛ_μl_μla_μ) _ft]) Lt. scutélla

[po_μ(ˈde_μμre_μ)_ft] Lt. *potere

Foot–Medial [ v́__v ]

[(ˈlagri_μma_μ) _ft] Lt. lacrima

152We did not find in Tuscan two types of intervocalic environment, a stronger [v__v́] and a weaker [ v́__v ] (see Honeybone 2012 for a typological survey) : both environments are lenis, [v__v́] (padè_μl_μla_μ) and [ v́__v ] (amí_μμgo_μ). Singleton /p t k/ do not lenite only when the onset is a weak branch of a trochaic foot ((14) amis(tà_μμde_μ)_UnFt) but also foot initial [sco(ˈdɛ_μl_μla_μ)] or medial [(ˈla_μgri_μma_μ)] ‘tear’. Conversely, strong Position/variants can be stress–driven through Syntactic Doubling (RS) after final stress.

153Concerning oxytonic words in Medieval Tuscan, lenition could be understood diachronically as foot shortening or pre–deletion, see (24); apocope is conditioned by segmental lenition, the diachronic loss of coronal in weaker foot–medial [ v́__v ]) from Lt áte and úte (see Section 3.1):

154(24) Apocope conditioned by lention in [ v́__v ]

155In the process of Lenition/Apocope (24) a prosodic constraint is active: Head Foot Right (Hdft–R), see McCarthy (2002), Wheeler (2005): (σˈσ) (ˈσ), where ˈσ is a BinFt trochee (with a prominent mora aligned with its left edge) subject to RS gemination in phonotactic environments W₁W₂, on the base of the Stress–to–Weight principle (Swp).

4 The Florentine throat/Gorgia. Different patterns of lenition

156The Lenition–Voicing trend (Section 1–4) was opposed especially in the Florentine area, to the trend of maintaining voiceless stops, which have conversely developed through Gorgia/Weakening. In this section we examine the lenition–Gorgia and study its interaction with lenition–Voicing from two perspectives: to study the relationship between Western Tuscan lenition (Voicing) and its Florentinization through lenition–Gorgia and to put the pieces together; to compare at phonological level the two lenition environments where occurs the consonantal strength shift from the voiceless Latin stops / p t k/ to Tuscan lenited variants.

Let us take a closer look at this difference.

4.1 Lenition–Gorgia. Distributional patterns correlated with strong and weak positions

158It is a weakening/spirantization of /p b t d k g/, realized as spirantized variants (→ [ɸ β θ ð x/h ɣ]) in intervocalic position V_V and across word boundaries. To spirantize, stops are pronounced with a more open articulation between vowels, the segmental reduction can go up to laryngeal fricatives (debuccalization: h/ɦ, i.e. loss of supralaryngeal place features). This pronunciation is typical of the Florentine area.

(25) Positional Gorgia. Florentine Spirantization/Aspiration in Phonotactics

Strong Position / p t k/

(+ Lic, – Gvt) : Coda Mirror

(a) ##_ [ˈka:sa] ‘the house’ (Phrase–Initial)

(b) C._ [ŋˈka:sa] ‘at home’ (Post–Consonant)

(c) ø_C_ or cv_ [ˈtre ˈk:a:se] ‘three houses’ (Prosodic RS)

(– Lic, + Gvt) Weak Position V_V

(d) #_ [la ˈha:sa] ‘the house’ (Postvocalic)

[di ˈɸe: ɸe] ‘of pepper’

[la ˈθu: θa] ‘the jumpsuit’

160According to the Coda Mirror v2 Theory, in (25a/b/c) the stop is licensed but ungoverned (= undamaged); this predicts a strong consonant in 3 Strong Positions, RS, phrase–initial, postconsonantal. In V_V position the voiceless stop is non licensed but governed (– Lic, + Gvt); this predicts the weak consonant.

161In (25c) [ˈtre ˈk:a:se], the RS latent consonant (ø_c) is licensed by the stressed monosyllable W₁ (‘three’) in headed DPs made up of W₁W₂ in phonotactics, since the trochaic Tuscan foot need an extra mora (given by the latent consonantal position) to be fulfilled in its prominent/tonic left position. This can also be represented within the Coda Mirror Theory assuming that the trigger W₁ [ˈtre_CV]_RS has a final [CV]_RS licensed by word stress (in Strict–CV stress in ‘space’). This [CV]_RS is initial in W₂ triggering gemination (for such analysis see for instance Russo 2013a; Russo and Ulfsbjorninn 2020; Russo 2022b). The initial consonant of W₂ (/k/ in ˈk:a:se]) is in strong position (# ___ ^or ø __) if a latent object [CV]_RS /(ø_c) is realized in surface, see (26a/b):

(26) Prosodic RS trigger/ Fortition

(a) [ˈtre ˈk:a:se] RS[CV]_RS

162(b) Coda Mirror ø _

163[ˈtre ˈk:a:se] ‘three houses’

165In autosegmental terms, as in Strict–CV phonology, the empty latent consonant in W₁ is treated as a floating consonant, the occurrence of the strong consonant thus meets distributional criteria, after floating consonant (25c), after a stable consonant (25b), and phrase–initial (25a). In these three cases the initial consonant is protected by the lenition. The UR for (25b) is given in (27a/b):

(27) Fortition trajectory: Strong variant after a stable consonant

(a) Clitic P (Locative) [ŋ̩] < /Vn/

(b) Clitic P + N [ŋˈka:sa] ‘at home’

166The initial consonants of W₂ phrase–initial behave like those in C._V or C#_V position (with fixed C), see (26a) and surface as strong variant [k] but there is not gemination since the final consonantal position of the (locative P) clitic is occupied by a lexical nasal. In the coda Mirror theory, the consonants in these two positions are licensed but escape the government, defined according to the CVCV theory (Ségéral and Scheer 2001, 2008; Sheer 2012a).

This CV houses the floating consonant in [ˈtre]_cv. In this theory the initial consonants are strong (# ___ = ø __) because of an initial CV see (28):

(28) Strong consonant Phrase–Initial (ø __):

##_ [ˈka:sa] ‘the house’ (Phrase–Initial)

167In contrast, W₁ ending in a vowel triggers Gorgia–Aspiration in cases such as (25d), see (29a/b) [la ˈha:sa], where the outcome of lenition is Debuccalization with laryngeal opening:

168(29) Gorgia–Lenition (Debuccalization)

169(a) W₁ = la ‘the’ D = Definite article

170[la ˈha:sa] V_V Weak context

171(b) Gorgia Debuccalization in V_V

172The complementarity between Weak and Strong contexts, as for lenition/Voicing, shows for Syntactic Doubling that the initial CV is parametric since it depends on the kind of Word₁ that precedes Word₂ (whether or not its UR has a latent consonant due to stress or lexical structure).

173In intervocalic position, across word boundaries or in word–internal position, within the Coda Mirror Theory the Gorgia/weakening (as lenition–Voicing) is represented by the presence of two realized nuclei, moreover the initial consonant is governed by the vowel, thus weakened.

178Tolomei (1525) with his Gorgia examples referred to the town where he was born, Siena. In Siena during the 16^th century it was too early to talk about a Florentine influence (Castellani 1960–80: 205). Moreover, Rhiys in De italica pronunciatione et ortographia libellus (Padua 1569), referred to a pronunciation [x] rather than [h] of voiceless velar stops (without mentioning spirantization of dental or labials). He affirmed that such “aspiratio familiaris admodum pistoriensibus, ac Senensibus est, omnium autem familiarissima Florentinis/ The aspiration is familiar among the Pistoiese speakers and Sienese speakers, but it is most familiar among the Florentine speakers” (Izzo 1972: 20; Franceschini 1983: 142). From the Rhiys we receive information that Gorgia is attested in Florence, Pistoia and Siena, but none of these records tell us that in the 16^th century it was attested in Western Tuscany, that is, in Lucca or in Pisa.

180As we have seen already in (29b), Gorgia sound change can reach up to a laryngeal fricative /h/, as for the voiceless stops /p t k/ we have the whole sketch of intervocalic weak outcomes in (30) (see also Giannelli and Savoia 1991):

(30) Florentine Gorgia (Lenition/Spirantization)

/p t k/ → [ɸ θ/h x/h] Lenited voiceless variants

(a) /t/–lenition trajectory

[ v́__v ]

[ˈdi:θo] ‘finger’

[v__ v́ ]

[la ˈθɛr:a] ‘the earth’

[v__ v́ ]

[la ˈθɾap:ola] ‘the trap’

(b) /p/–lenition trajectory

[ v́__v ]

[ˈpe:ɸe] ‘pepper’

[v__ v́] and [ v́__v ]

[la ˈɸi:ɸa] ‘the pipe’

[v__ v́]

[la ˈɸɾo:da] ‘the bow’

(c) /k/–lenition trajectory

[ v́__v ]

[aˈmi:xo]

[aˈmi:ho] ‘friend’

[v__ v́]

[la ˈxa:sa]

[la ˈha:sa] ‘the house’

[v__ v́]

[la ˈxɾo:t͡ʃe]

[la ˈhɾo:t͡ʃe] ‘the cross’

181Branching onset as in (30a/b/c) are also represented within the Coda Mirror Theory as we have done in Section 2.5 as separated by an empty nucleus (/tr/ (a) [la ˈθɾap:ola] (b) /br/ [la ˈɸɾo:da] or (d) /kr/ [la ˈhɾo:t͡ʃe]): C_øR, R governs C (on the basis of IG = Infrasegmental Government, Scheer (2004): C is a governee, R is a governor), thus C is in a lenition position, even if it is in a complex onset.

(31) /k/ – Glottalizing in Branching Onset (Gorgia) V#_rV

183Voiced stops are affected by similar processes as voiceless stops. In vCv position /b d g/ spirantize (β ð ɣ/ɦ 0); to /g/ always corresponds [ɣ ɦ]. The lenition/spirantization of voiceless stops also affects the past participle (PP) category as in (32):

/t/ Lenition trajectory → Glottalness

/t/ → [h] t–Glottalization

[ v́__v ] HL (σ_μμσ_μ)_UnFt = (da_μμθo_μ)_UnFt

(a) [anˈda:θo] ‘gone’

[anˈda:ho] ‘gone’

(b) [vesˈti:θo] ‘dressed’

[vesˈti:ho] ‘dressed’

188However, we found the deletion of the weak–Gorgia variant (as predicted by Kingston 2008). The deletion of the fricative is a sentence–level process, it is more frequent in phrase–internal position (before last foot) than in phrase–final position.

(33) Gorgia Glottalization and Deletion in Phrase–Internal Position:

(a) [ɛra ˈʃɛ:a | ðɑ̃ ũn ɔ̃c:o]‖ [ˈʃɛ:a] ‘blind’ k→ Ø /_ [+back]V

(she) was blind in one eye

(b) [ e ˈl a b:uˈaθo | ˈlui]‖ [b:uˈaθo] ‘pierced’ k→ Ø /_ [+back]V

it pierced he

he pierced it

When the following vowel is [–back] /i e ɛ/, the realisation is more likely to be [x] or [h].

192Different variants are also conditioned by the kind of intervocalic context V_{l r}{j w}V); for instance to get ‘0’ from /k/ (complete debuccalization) with branching onsets (tautosyllabic clusters), it is also necessary for the item to be in phrase–internal position as in (34):

(34) Gorgia Plosive Deletion in Phrase–Internal Position for Branching Onsets

V_{r}V

[ ʃi ˈreðe ˈɸ̞ɔx̞o] ‘It. ci crede poco ([ˈkre:de]) / (he) believes in it little’

‘In it believes little’ /k/ → Ø

Concerning a palatal variant of /k/ [ç] triggered by the prevocalic palatal glide /j/ as in [ˈçja:ve]/ [ˈça:ve]/[ˈhja:ve] ‘It. chiave [ˈca:ve]/key’, it can be also deleted in phrase–internal position, giving loss ‘0’: [la ˈja:ve di ˈhasa] ‘It. la chiave di casa/the house key’.

193Furthermore, in phrase–internal position the weak variants [h/x] can also be the spirantized/glottalized/deleted realisations of /kw/ (Giannelli and Savoia 1978: 41; Giannelli 2000: 32):

(35) Gorgia Plosive Deletion in Phrase–Internal Position

V_V Across Boundaries

/kw/ ‘It. due quintali [ˈdue kwinˈtali]/two quintals’

(a) du hĩnˈtali] [h]

(b) [du ˈxĩntali] [x]

‘It. e torna qua [ˈkwa]/and come back here’

(c) ˈθornã ˈxa] [x]

(d) [e ˈθornã ˈa] [0]

In (35) it is a development that speakers from the urban classes feel as rustic, the urban usage of the middle classes involves the maintenance of [w] (Giannelli 2000: 32).

194Thus, variants may also change depending on sociolinguistic factors. The velar approximant, for example, is a conservative variant linked to age (elderly speakers) and rural groups, whereas the laryngeal variant is more common in middle–class and young speakers.

4.2 Tuscan Syntactic Doubling promotes fortition

195In fortition/lenition–Gorgia contexts, consonantal onsets at the right–edge of the trochaic foot undergo strengthening. The voiceless stops /p t k/ are pronounced as such only after a consonant, in phrase–initial position (except cases of initial Voicing discussed in Section 2), or as a geminate under Syntactic Doubling (RS) at the right edge of the foot (and of the phonological word W₁W₂=PWd). As we have seen in Sections 3, 4, 4.1, RS applies in almost all Tuscan dialects and is triggered by oxytonic words; besides tonic monosyllables such as ‘three’ (see (26a/b)), also by the tonic pronouns ‘you’ It. tu/voi, Tuscan [ˈtu/ˈvu]: [ˈvu vveˈni:te] ‘you come’, and apocopated infinitives [veˈde lˈluj] ‘It. vedere lui/to see him’ (see Giannelli 2000: 56; Marotta 1995, 2000; Rohlfs 1968 § 612):

(36) RS after W₁ Tuscan oxytonic apocopated infinives → [ll]

196[veˈde lˈluj] ‘It. vedere lui/to see him’

(37) RS after W₁ Tuscan tonic pronouns ‘you’ [vu]

197[ˈvu vveˈni:te] ‘you come’

198So far, we have discussed RS stress–triggered, and not yet RS lexical/syntax–triggered, where W₁ are atonic, with a latent lexical consonant in their UR licensed in precisely syntactic positions to surface as geminate, such as [X1 ...[N....] → [X_øc [N , √C...]. We refer for this type of RS to Russo (2013a, 2019, 2021), Russo and Ulfsbjorninn (2017). Within this kind of RS, the RS (pro–)clitics W₁ cliticise on a contiguous host and must have a syntactic fixed position in the XP where they c–command W₂. We mention it to address a specific example, that of the definite article MSG that unlike Italian has a doubling effect in Tuscan and promotes fortition mirroring Gorgia–lenition in weak symmetrical environments, see below.

199Tuscan RS is also triggered by atonic cliticized (pro–)clitics (on W₂) with a floating C in the UR, e.g. atonic monosyllables ‘da/from’, ‘a/to’ ([amˈme] ‘It. a me/to me’), ‘se/if’ etc.

200In Tuscan this kind of RS involved in a recent period also ‘il/the’ (Lt ille), the MSG definite article (see Giannelli 2000: 33–35), which shows a RS geminate in a DP sequence of W₁W_2: [i sˈsale] ‘il sale/the salt’, [ikˈka:ne] ‘It. il cane/the dog’ (RS ‘i’ Def MSG) vs. [i ˈha:ni] ‘It. i cani/the dog’ (Gorgia ‘i’ Def MPL). This would be represented in our a–linear hierarchical format as follows within Syntax–Phonology approach as in (38):

(38) Def D = /iø_l/ MSG ‘It.il’ Lt ille RS trigger (weak atonic)

(a) [D_øk [N , √C...]

(b) (i(_øk))(k(a))(n(e))

(c) [ikˈka:ne]

In terms of the linear CVCV phonology and Coda Mirror Theory, we get the configuration in (39), where the template shows a delinking of the lateral from the first CV unit, in order to let the dorsal to spread on a Coda Mirror Position with an empty CV available (ø__)

(39) RS after Def D MSG

201[ikˈka:ne] ‘It. il cane/the dog’ [i]_CV Def MSG (= /i_øl/)

(40) Def D MPL Gorgia /Aspiration (Weakening in Initial Position)

202[i ˈha:ni] ‘It. i cani/the dog’ [i] Def MPL (= /i/)

203It must be noted in (39) and (40) that the Italo–Romance definite article always has two CV units in its UR potentially available for its spell–out (see Faust et al. 2018; Russo and Ulfsbjorninn 2021; Russo 2022b). In (40), the initial consonant is preceded and followed by a filled nucleus, thus it is in intervocalic position and undergoes lenition/aspiration – Gorgia.

204In Tuscan dialects the lenited forms derived from Gorgia (fricativized or glottalized) in same cases can also be considered morpho–phonemic or morpho–syntactic markers (see Gess 2009), if we regard the reverse realizations of such segments in strong contexts, phrase–initial and RS (at the left–edge of the prosodic constituent), where the closure is often associated with strong consonants triggered by the Syntactic Doubling.

4.3 Debated cases. Weakening/Aspiration Gorgia in Strong Position

205In Florence and in the near areas stops can get aspirated in strong position (this is first reported by Giacomelli 1934, confirmed later by Giannelli and Savoia 1978), i.e., Gorgia does not seem to be blocked by geminates or post–coda positions are not strong. In this section we look at the behavior of those cases with consonants occurring in the three strong positions, phrase–initial #__, after a floating ø_C__ (RS), post–Coda position, where the strength scale does not respect the distribution expected from the strong/weak mirror contexts. We aim to account for this unexpected disjunctive phonological identity of the alleged Coda Mirror object.

206We also connect to these reverse distributions what Castellani (1960–80: 191 [Pronuncia enfatica fiorentina] calls Emphatic Gorgia from the AIS maps (n°357 and n° 669) discussed in Section 6, where spirantized weak variants occur in the same strong contexts (post–Coda position and phrase–initial #__).

4.4 The reverse distribution: when Gorgia Post–coda is not strong

207Post–coda consonants may not be strong after sonorants. The Coda Mirror Theory predicts a strong variant in C.__ position, while a governed empty nucleus occurs between the two CøC in the sequence coda–onset. The Coda Mirror object (ø__) should not be disjunctive in initial and coda position.

208We expect lenition–Gorgia to apply only postvocalic in word internal position and across boundaries. In (41) you can see such special cases:

(41) Lenited Gorgia variants in Strong Position /p t/

Disjunctive context {C,#}__? Coda Mirror violated

(a) C._ [col ˈɸe: ɸe] ‘with pepper’ (Post–Coda)

[in ˈθu: θa] ‘in jumpsuit’

(b) ø_C_ [ˈt͡ʃɛ ˈɸ:e: ɸe] ‘there is pepper’ (Prosodic RS/geminate)

[ˈtre ˈθu: θe] ‘three jumpsuit’

209Still unexpectedly, a spirantization–Gorgia process in a position other than intervocalic is also observed for the voiced stops /b d g/. The weak variant of the voiced stop (/d b g/ → [ð β ɣ]) is also found in post–coda position C._ (Giannelli and Savoia 1978: 46):

(42) C._ / d b g/ → *[ð β ɣ]! (After the Latin prefix ex > [z])

Coda Mirror violated

(a) [zðɛ̃nˈtaθo] ‘It. sdentato/ toothless’

(b) [zβuˈdɛlla] ‘It. sbudella/(he) guts’

(c) [zˈɣot͡ʃ:ola] ‘It. sgocciola/(it) drips’

C._ (After the rhotic r.C = [r.ð]

(d) [si sˈkɔrða ði ˈθutto] ‘It. si scorda di tutto/(he) he forgets everything’

(e) [ˈmɔrtho] ‘It. morto/dead’

C._ (After the lateral l.C = [l.ð/t^h]

(f) [ˈmolt^ho] ‘It. molto/a lot’

C._ (After nasals N.C = [n/.C] Across the boundary

(g) [i sˈsantho] ‘It. il santo/the saint’

(h) [m phɔ ˈmɛʎʎɔ] ‘It. un pò meglio/a little better’

(43) Strong Position (Phrase–initial):

#_ *[ð β ɣ]! Coda Mirror violated

(a) [ˈβe:vilo ˈθe] ‘It. bevilo tu / drink it yourself’,

(b) [ˈðaʎʎelo] ‘It. daglielo/give it to him’

(c) [ˈɣrattalo] ‘It. grattalo/ scratch it’

210Lenited–Gorgia variants are also found in Strong positions from voiced Latin stops /b d g/, even in Syntactic Doubling positions after RS triggers such as the tonic monosyllable /ˈtre_cv/ < Lt tres (Giannelli and Savoia 1978: 46):

(44) W₁ /ˈtre_cv/ triggers RS Geminate after ø__ Lt tres

UR / d g/ → [ðð ɣɣ ]! Coda Mirror violated

(a) [ˈtre ˈð:iθi] ‘three fingers’

(b) [ˈtre ˈd:iθi]

(c) [ˈtre ɣ:alˈletti] ‘three cockerels’

(d) [ˈtre g:alˈletti]

In all these contexts ((41)–(44)) the Gorgia is not predictable based on the Coda Mirror Theory, since even the voiced obstruents do not seem positional sensitive.

211Normally we should encounter the weak variants only in postvocalic position, and the strong variants in all other contexts, phrase–initial, after a latent RS consonant (ø_C_) triggered by stress or by syntactic licensing, in post–coda position).

212In Strict–CV theory, according to the Coda Mirror Theory in C._ context, a consonant is licensed but not governed, the following nucleus is expected to govern the empty nucleus between the two consonants. This should produce the strong consonant as an outcome, but in some cases, as in (41), (42) and (43), we find the weak consonant.

213In post–consonantal positions based on the theory of Strict–CV, the consonants are preceded by a governed empty nucleus and a filled onset.

214We find the same unexpected spirantized cases in word internal strong post–coda position /r._V/ (such as in [la ˈθɔrθa] ‘It. la torta/the cake’) *[θ ɸ ɣ], in (45) the obstruent in this position should be protected by lenition, while the post–consonantal position is not always strong in Tuscan and we get the lenited–Gorgia outcome:

(45) Weak/Spirantized variant in post–coda position

215*[θ] in C._V A real Gorgia?

216In (45), a consonant after a coda is placed in the Strict–CV after an empty nucleus, as it happens with consonants in strong initial position. Its strength should come from the fact that the consonant is licensed while escaping from government. However, in the context (r._V) we have the weak variant *[θ] instead of [t], and it could be inferred that the disjunction of the Strong Position {#, C}__ is not reduced to a single context, the unique object (ø __).

219Moreover, the weakening of geminate is doubly problematic since geminates have always been defined as one feature bundle or root node attached to two structural units (C–slots) symmetrical one–to–two relationships, see (46) and (47) (but see discussion in Russo and Ulfsbjorninn 2017):

(46) True Geminates – Integrity/Inalterability?

(47) Fake or bogus geminates?

220We expect geminate segmental melodies symmetrically distributed across two syllabic positions as in (46), and geminate length should resist lenition (integrity is a consequence of their representation, see Hayes 1986). The symmetry of the two positions composing the geminate under an autosegmental approach stems from an identical relation to specifications. However, in these Tuscan cases we get aspirated geminates as outcomes and in such examples positions and specifications are not in a bijective relationship. This can corroborate the insight that the representation of geminates is not even symmetrical, like that of long vowels, as suggested already by Russo and Ulfsbjorninn (2017). Aspiration on geminates has been reported in the phonological literature, see for instance the false geminates’ of Tigrinya (Kenstowicz 1994), which undergoes spirantization on their left part: /mərak + ka/ → [məraxka] ‘your calf’ (Russo and Ulfsbjorninn 2017: 167). In (47) false geminates or bogus geminates are represented within Strict–CV phonology as two root nodes specified for the same segment. Another possible account is proposed by Russo and Ulfsbjorninn (2017) where geminates are not associated to two C positions in their underlying representation, but a segmental /k/ is followed by an empty CV unit in the UR of the geminate, which guarantees its inalterability, while still being available to replace at melodic level the historical geminate with post–aspiration, see (48):

(48) Asymmetrical geminates F– ø_F from a Strict–CV perspective

221C₄ is ungoverned, thus it can be realized phonetically, but without (fully) copying its melody from C₃. The only possible interpretation from the UR is of a consonantal geminate CVCV. In this configuration consonantal geminates have an asymmetrical interpretation, and their representation gets closer to that of a–linear hierarchical models based on syntax (Sauzet 1994; Sauzet–Brun Trigaud 2012; Russo 2013a), where consonantal geminates should be represented by asymmetrical chains of two positions based on a partial features identification: F– ø_F.

222It should be noted that the geminates obtained through RS have the same format however with the specified position (the controller) on the left ø_F–F (in the CVCV template this would be expressed also positing the length CV on the left). However, geminates with this format, having a controller on the right, exist in languages, the controller is on the left for example in Berber, as suggested by Dell and Elmedlaoui (1985).

223All these weakenings (spirantization and aspiration) in strong positions ((41)–(48), after a heterosyllabic consonant or word–initial {C,#}__ (= ø__) are not expected on the basis of a lenition theory grounded on the positional distribution of strong and weak symmetrical contexts.

For such contexts, it could also be assumed for some areas that a lexical restructuring of the Gorgia outcomes generalizes the weak variant beyond the weak contexts.

It is a reminiscent situation of the initial Voicing in Tuscan in strong position (Pisano <gattivo> < Lt captivu or Italian ‘gabbia/cage’ < Lt cavea, etc. see above in Section 2).

One might think that [ɸ β θ ð x/h ɣ] are lexically stored phonemes by Tuscan speakers that reconstruct words with the weak variant independent on the contexts in such cases.

4. 5 The mirage of Medieval Gorgia

« In quolibet idiomate sunt aliqua turpia, sed pre ceteris tuscum est turpissimum/ in every language there are turpish things, but of all of them the most turpish is Tuscan ». (Dante in the De vulgari eloquentia)

(49) No Gorgia! <h> = <ch> = [k] in Medieval Tuscan

# [_σC V …

<hasa> next to <chasa> It. ‘casa/house’ (Senese)

<harta> next to <carta> It. ‘carta/paper’ (Pratese)

229Therefore, it is hard to think that in Medieval texts the graphical distinction between <h> and <ch> indicates a laryngeal sound. In V_V position, graphical <h> associated to a coronal or labial stops could indicate the lenition of stops intervocalically (see above in Northern Italian dialects <maridhi> <dh> = [ð]. However, <h> in Tuscan Italian represents graphically [k]:

230(50) No Gorgia! v<h>v = vkv

… V [_σC V … *[aˈmi_μμ:ho_μ] (<ho> =L )

Weak branch of the Foot HL

(It is not an enough condition!)

<amico> = <amiho> (Pratese)

Correct transcription : [aˈmi_μμ:ko_μ] <h> = <c> = [k]

No lenition at all here!

If we look at the OVI data, we find the words quoted by Hirsch (1885: 562) as Gorgia/weakening examples: e.g., diano ‘decano/dean’ which occurs essentially in Medieval Senese from 1260:

Table 44 Gallicism – Deletion of /k/ < Western Lenition

/k/ <c> → Ø	diano It. ‘decano’
Senese		1260 [Lett. Sen.]Vincenti di Aldobrandino
Senese	Diano	1262 Lettera Tolomei
Senese	1263 [Doc. Sen.] Compagnia Ugolini

But the word diano (Lt decanus) is a Gallicism (see Castellani 1960–80: 201) and the deletion of <c> is the result of the Western lenition.

231For the other two forms reported by Hirsh, <dihiarare> (and the derived forms) ‘declare’ and <hontenere> ‘to contain’, they look like suspicious forms. We have not found them in the corpus of Medieval dialects (the OVI database). For those two forms OVI does not show any occurrences with deleted <c>, not even for <havaliere>, we have always <ch>, <chavaliere> from Aretino, Pisano, Senese and Florentine, which also corresponds to <k>. For <contenere> only graphical <ch> is frequent in Tuscan area:

Table 45 <contenere> NO Gorgia!

hontenere ‘con–’?
Doc. Sen.		1263 Compagnia Ugolini
Stat. Flor.	chontengha	1320 Notaro della Signoria

also in many other lexical forms where <ch> = [k]:

Table 46 <ch> NO Gorgia! → [k]

Lucchese	sono cho(n)tenti	1296 Guidiccioni
Tusc. Sud–Or.	la chontenzione	1298 Questioni filosofiche

232These forms are spread everywhere in Tuscany, in Pistoiese, Senese, Florentine, Lucchese, Aretino and are just graphic forms for singleton <c> = [k]. <ch> is normal before back vowels /a o u/ to differentiate graphically a velar from a palatal consonant. We did not find the form <havaliere> quoted by Hirsch (1885: 563) in OVI, but in any case, <h> would correspond to <c> = [k], and it cannot represent aspiration.

233Also, the other forms put forward by Hirsch, <hosì> for <così> ‘so’ and <halende> for <calende> ‘calends’, quoted for Medieval Senese, are not in OVI, however Castellani (1960–80: 203) checked the context and found that <hosì> is in initial position, which is not a context of weakening (Gorgia):

(51a) #C_ (Strong initial position) <h> = [k]

Hosì (fummo d’accordo) (Siena 1475)

The form <halende> is not either in intervocalic position, but in post–coda position:

(51b) C.C_ (Post–coda position) <h> = [k]

in halende di dicembre (Siena 1359)

234Therefore, we can think that in the texts there is no graphical distinction between <h> and <ch>. In the OVI corpus we do not find the same examples quoted by Hirsch albeit we find others very similar. The following two examples also seem to contradict lenition, since they occur again in initial position, one Phrase–Initial (Figure 4), the other in RS context W₁W₂, where W₁ (<è> ‘is’) is a RS trigger, a tonic monosyllable (Lt est) with a final latent consonant licensed by stress already in Medieval Tuscan:

In Pratese initial position <harta> = <carta>: <Harta per Raneri> (Doc. Prat. 1245).

Again, we can assess that graphic <h> = <ch>

236We also found in the OVI database <chasa> along with <hasa> for ‘casa/home’, thus Phrase–Initial as well as in intervocalic position V_V:

Table 47 <ch> in V_V = [k]

Senese 1233–43	<hasa> <chasa>	Libro di Mattasala di Spinello
xxviij d. p(er) dispesa dela chasa. Ite(m) iij s. m. iij d. p(er) dispesa dela hasa.

237Other examples, such as <a haso> for <a caso> ‘randomly’, prove that <h> = [k] is a generalised spelling for all contexts, both fortition and lenition. As we know for the syntactic doubling conditions (RS), this is also a fortition context (<a> is a W₁ RS trigger) where we can think the graphical <h> = graphical <ch>. In the anthroponym <Bonahorso> < bonu + adcursu, we find again the RS trigger /a(d)/ Lt ad /a_cv/, thus we expect the strong consonant. However, we have seen that <h> is always [k] or geminate [kk] in strong and weak contexts:

As we see in Table 48, <h> is found in fortition contexts (Phrase–Initial or RS), as well as in intervocalic contexts V_V (<Iahopo>, <di he(r)monese>):

Table 48 <h> = always [k] or [kk] in Strong and Weak Contexts

239As we have seen in Tab. 48, the graphical <h> can even correspond to an underlying geminate /kk/ normally transcribed with = <cc> and this <h> can also be used in the Northern Italian scripta after a consonant (see also Castellani 1960–80: 201 and n. 37). Also, intervocalically several examples show that the graphical <h> indicate certainly not the aspiration but only the /k/ <c>:

240(52) Pratese <h> <cc> = /kk/ Lexical geminate (Strong Context)

(a) <Riho> = <Ricco> ‘rich’

Pratese <ahomandamolo a Riho> (1245 libro di conti)

(b) Pratese <h> = <c> = /k/ <amiho> = <amico> V_V (Weak context)

Pratese <amiho badareli> (1245 libro di conti)

Pratese <Domeniho> (1245 libro di conti)

In all these examples listed above, <h> corresponds to [k].

241A use of <h> instead of a coronal should be noted in the Northern Italian scripta <h>, where we have lenition/Voicing as part of Western lenition. One might think that this <h> is a development of Lenition/Voicing → Spirantization, and that <h> could represent the Romance lenition of a singleton coronal voiceless stop /–t–/:

242(53) Lenition Voicing Northern Italian Dialects <h> = /t/ or /d/ ?

243(1) Lenition/Voicing

244(2) <h> Anti–hiatus!

• 118 1300 Doc. Venez. C. Navagero.

Veneziano <viha> = <vita>118 < Lt. vita ‘life’

• 119 1274 Pietro da Bescapè.

Lombardo <exultaho> = <esultato> ‘rejoiced’119

• 120 1315 Cedola anonima.

Veneziano <ordenaho> = <ordinato> ‘ordered’120

• 121 1353 Passione genovese.

Genovese <verifficaho> <verificato> ‘verified’121

• 122 14^th c. Codice Servi.

Ferrarese <staha> = <stata> ‘been’122

Voiced Stop Lenition Voicing /d/ + <h> Hiatus Resolution

• 123 1353 Passione Genovese.

Genovese <seha>123 < Lt. sedere

245The anti–hiatus tendency of the orthographic <h> is evident also in the case of other voiced and voiceless obstruents subject to lenition (Voicing, Spiratization and Deletion) in V_V intervocalic position, such as for a voiced consonant as /v/ in <aveha> = <aveva> ‘(he) has’ or /d/:

(54) Lenition of voiced obstruents→ Deletion + <h> Anti–hiatus!

Genovese <aveha> (1353, Passione Genovese) Hiatus Resolution → <h>

246We do not find this orthographical usage in Tuscany since here <h> corresponds graphically to a dorsal voiceless stop. In Northern Italian dialects <h> can be used to represent graphically the lenition of the stop associated with a coronal stop, as it is evident from other Medieval examples localized in Northern Italy where the stops undergo lenition as we expect according to Western lenition. This is notated with <th/dh> or <d>:

(55) Lenition/Voicing: graphical <d> instead of /t/: /–utu/

• 124 1207–08 Aleppo.

Veneziano <reçevudo> ‘received’, <saipudo> ‘known’, vedudo ‘seen’124

• 125 13^th c. Patecchio.

Cremonese <cognosudo> ‘known’, <tegnudo> ‘held’, <veg[n]udo> ‘come’125

247Therefore, this graphical <h> has got a ‘ratio’. It does not show Gorgia aspiration at all in Tuscany nor Western lenition (Voicing, Spirantization and Deletion) in Northern dialects. Rather it represents in Northern Italian dialects a Hiatus Resolution after the Western Lenition at last stage: deletion of obstruents to ‘0’. We can also find this result sometimes in Tuscany, where <h> does not only represent a dorsal stop <c> = /k/, but a voiced coronal /d/ lenited graphically <h>:

(56) <h> Hiatus Resolution after Western lenition in Tuscany /d/ → [0]

Tuscan 1361 Piero da Brescia

<ahoperala> = <adoperala>

<ahoperalo > = <adoperalo>

<ahoperai> = <adoperai >

<h> is a letter that can also be used to replace the consonant which underwent lenition if it is a voiced or a voiceless obstruent in Tuscany, and especially in Northern Italian dialects, like in Genovese <verifficaho> ‘verified’. However, this <h> does not represent a spirantization of the consonant lenited but only an anti–hiatic segment, used for Hiatus Resolution of VV sequences.

248In the OVI corpus we can also find <ahoperato> next to <aoperato>, both represent lenition of /d/ in V_V, one with hiatus resolution (56), the other maintains the hiatus (57):

(57) Hiatus VV /d/ → ‘0’

Lucchese <aoperato> (1298 Lett. Lucch.)

Florentine <aoperato> (1310–12 Compagni Cronica)

Amiatino <aoperato> (1376, Doc. Amiat.)

Tuscan <aoperato> (1388 Esopo)

249Furthermore, there is no way, <h> could represent Gorgia in Tuscany during the Medieval period.

The anti–hiatic <h> can also intervene in Medieval texts to split a hiatus after other type of phonological lenition as the Neapolitan lenition (see Russo and Ulfsbjorninn 2020):

(58) Neapolitan: Hiatus Resolution after /g/ deletion in V_V position

<drahone> ‘dragon’ (16 occ.) (14^th c. HistTroya)

(59) Genovese: Hiatus Resolution after /p/ deletion in V_V position

<aseniho> ‘example’ (1311, Anonimo Cocito)

250This graphic letter <h> also comes to split two vowels in hiatus even if the separation of the two vocalic segments is in the lexicon:

(60) Hiatus Resolution <h> (No lenition at all here)

(a) Romagnolo Uno <paho> da chuorpi (1362, Doc. Imol.) = <paio> ‘a couple of body’

(b) Tuscan <aholore> (14^th c. Diretano Bando) = <aulore> ‘scent’

251This graphic habit comes probably from the orthographic way to write the onset of the syllable with a consonant whenever there are two vocalic segments next to each other as it is evident in <aholore>. This graphic <h> can be also completely unmotivated again between two vowels as in: Flor. <pagahare> (= <pagare> ‘to pay’) (1348–50 Libro arancio).

252It is important to notice that the use of anti–hiatic <h> also occurs in Latin without a consonantal deletion between the two vowels in hiatus: e.g., Michael (ecclesiastical Latin). It is therefore an already accepted usage in Medieval Latin scripta.

253Although <h> for /k/ in the Medieval Senese texts or in the fifteenth and sixteenth century Florentine texts are not interpretable as representations of spirantization, however, it cannot be ruled out that the phenomenon did not develop in Medieval times.

6 Lenition Gorgia in AIS - Sprach– und Sachatlas Italiens und der Südschweiz

254In this section we complete information given in Section 4 on the areal distribution of Gorgia in today’s Tuscan dialects from the linguistic atlantography records.

(61) List of AIS the maps analysed Lt

AIS 18cp ‘il nipote, la nipote/the nephew, the niece’ nepote

AIS 21 ‘il vostro nipote, i vostri nipoti/your grandchildren’

AIS 22 ‘la vostra nipote/your granddaughter’

AIS 23 ‘le vostre nipoti/your grandchildren’

AIS 91 ‘la pelle/the skin’ pelle

AIS 92 ‘il pelo, i peli/the hairs’ pilu

AIS 97 ‘capelli/hair’ capillu

AIS 163 ‘il piede, i piedi/the foot, the feet’ pede

AIS 247 ‘ammazzatelo/kill him’ * mactiate

• 127 Emphatic Florentine Gorgia according to Arrigo Castellani.

AIS 357 ‘lontano/far’127 * longitanu

• 128 Emphatic Florentine Gorgia according to Arrigo Castellani.

AIS 669 ‘spogliarsi/undress’128 spoliare

AIS 1097 ‘(il cane) i cani ‘dogs’ cane

AIS 1383 ‘lupini/lupin beans’ lupinu

(62) Symbolic sounds [th] and [ph] are used to transcribe /t/ and /p/ spirantized

• 131 Similarly in AIS 21 ‘il vostro nipote, i vostri nipoti/your grandchildren’ (no Gorgia is represente (...)

AIS 18cp ‘il nipote, la nipote/the nephew, the niece’131

Florence (523) [–th–]

Montespèrtoli ([Flor.] prov. Florence) (532) [–th–]

Barberino di Mugello ([Flor.] prov. Florence) (515) [–th–]

Carmignano ([Flor.] prov. Florence) (522/1) [–th–]

Radda in Chianti Centr.Tusc. (prov. Siena) (543) [–ph–]

Chiusdino Volt. (prov. Siena) (551) [–th–]

Siena (552) [–th–] and [–ph–]

Faùglia ([Pis.] prov. Pisa) (541) [–th–]

257From AIS 22 ‘la vostra nipote/your granddaughter’ transcriptions, it is also evident the absence of spirantization/aspiration in Arezzo (pt 544) and in Arezzo area: (pt 545) Chiavaretto, Cortona (pt 554), but also in Pisano (pt 541, Faùglia, pt 542, Montecatini Val di Cècina), near Lucca (pt 520, Camaiore [Lucch. –Vers.].

259In AIS 23 ‘le vostre nipoti/your grandchildren’, in addition to the Arezzo area, there is not Gorgia in Western Tuscany: No–Spirantization/Aspiration in pt 520 Camaiore ([Lucch. –Vers.] prov. Lucca), pt 541 Faùglia (prov. Pisa), pt 542 Montecatini Val di Cècina Montecatini Val di Cècina (prov. Pisa).

262The same problem can be noticed in other AIS maps, as AIS 1383 ‘lupini/ lupin beans’, where the same transcription is used to represent aspiration but also other types of lenition as in Arezzo or in Pisa (59), where [–ph–] can just note lenition of /p/ but not spirantization:

(63) AIS 1383 ‘lupini/ lupin beans’ (problems with Scheuermeier’s transcriptions?)

All [–ph–]!

Firenze (523)

Siena (552)

Montecatini Val di Cècina [Volt. prov. Pisa] (542)

Castagneto Carducci (prov. Livorno) (550)

Gavorrano, Maremma Massetana (prov. Grosseto) (571)

Arezzo (544)

264Furthermore, in the first two volumes of the AIS spirantization is ignored, see for instance the AIS map 163 ‘il piede, i piedi/ the foot, the feet’ (there is no Gorgia represented at all on this map).

266As well as on map AIS 669 ‘spogliarsi/undress’, only in Florence (pt 523 [spʰoˈʎassi]) in complex initial onset sC_, after a coda (C._V), therefore a context once again not normally suited to a positional weakening, as we discussed in Section 4.3, 4.4.

267This aspiration was observed by Giacomelli (1934: 196 based on AIS),

268As we said in Section 4.3, this aspiration was observed by Giacomelli (1934: 196) in Florence for positions other than intervocalic, in strong position and even as a secondary articulation of a geminate. We have shown the challenge of these data for the Coda Mirror Theory and for any theoretical approach to consonantal lenition/fortition, since they violate cross–linguistically validated positional symmetries, see (64):

(64) Aspirated Geminate F– ø_F (see (44)) AIS

(a) [t͡ʃikˈkʰare] ‘to smoke a fag’ (Map 759)

(b) [ˈt͡ʃeppʰo] ‘strain’ (Map 781)

(c) [i pˈpɛtt^hine] ‘It il pettine/the comb’ (Map 673)

(d) [i kˈk^hɔrpo] ‘It. il corpo/the body’ (Map 677)

(e) [gratˈt^hati] ‘scratched’ (Map 679)

(f) [ˈmatthɔ] ‘mad’ (Map 723)

(g) [aˈve pphaˈura] ‘be afraid’ (Map 727)

In phrase–initial position (while in (60g) geminate is RS–driven, licensed by W₁ stress):

(h) [spʰoˈʎassi] ‘undress’ (Map 669)

(i) [p^hɛttiˈnassi] ‘to comb’ (Map 672)

269Gorgia/aspiration on geminate is also reported by Giannelli and Savoia (1978: 41), see cases such as in (63), where the first full geminate with no–aspiration (foot–initial) is triggered through RS licensed by W₁ stress, while the second lexical geminate is aspirated (foot–medial):

(65) Aspirated Geminate F– ø_F (see (44)) Foot–medial

[ˈe] /e_cv/ = RS trigger C.C Stress licensed in #_

(a) [l: ˈe b:ruθ:o] ‘It. è brutto [ˈɛ b:rut:o]/ (he) is ugly’

(b) [l: ˈe s:ex:o] ‘It. è secco [ˈɛ s:ekko]/(it) is dry’

270In (65) in the initial position we have full intervocalic geminates ([bb ss…]) triggered by RS (through the auxiliary verb [ˈɛ/e] W₁ that acts a strong/stressed monosyllable in phonotactics), while the intervocalic geminates foot–medial are aspirated with the typical result expected for single lenited Gorgia consonants ([θ x ɸ]). This also happens foot–medial for /pp/, as in [|ˈtaɸ:alo] ‘It. tappalo [ˈtappalo]/cup it’ (ib.).

271Giannelli and Savoia (1978: 41) also confirm an unexpected Gorgia in initial position [ˈx/hu:cilo] ‘It. cucilo/sew it’, [ˈɸɔrtalo] ‘It. portalo/bring it’, as for (60h) and (60i), see (39) in Section 4.3.

(66) Post–coda (After l/r/N.C) See Section 4.3 (42)

(a) [ˈmoltho] ‘a lot’ (Map 703)

(b) [i sˈsantho] ‘the saint’ (Map 808)

(c) [m phɔ ˈmɛʎʎɔ] ‘a little better’ (Map 696)

(d) [ˈmɔrtho] ‘dead’ (Map 708)

273This corresponds to the examples of lenition/Gorgia of voiceless stops in post–coda reported by Giannelli and Savoia (1978: 41) especially for Florentine, as in (67):

(67) Post–coda (after r/s.C) Lenited/Spirantized *[θ] in Strong Position

(a) [la ˈθɔrθa] *[r.θ] ‘It. la torta [la ˈtɔrta]/the cake’ (See Section 4.3, (41))

(b) [la ˈʃesθa] *[s.θ] ‘It. la cesta [la ˈt͡ʃesta]/the basket

274Giacomelli (1934: 198) confirmed AIS data (gathered by Scheuermeier) with numerous control surveys. The aspiration of voiceless stop on geminate or after a coda liquid (/l r/) is only found in Florence or only rarely in other places of Tuscany.

276Summing up the AIS results, the major problems are, especially with regard to the symbols [th]/[ph], that those transcriptions indicate aspiration but sometimes other types of lenition (e.g. Voicing) as we can notice in Aretino data which is an area without aspiration according to Castellani. However, according to Giannelli and Savoia (1978–80), in the Arezzo area there should be spirantization with Voicing [θ̬ ɸ̬ x̬] (not the Florentine Gorgia but still related to it). The transcription [th] is although unsatisfactory, since it does not distinguish between the Florentine aspiration (Gorgia) without Voicing and this particular outcome that associates Voicing with aspiration unlike Florentine Gorgia. Furthermore, in the AIS maps cases of aspiration/weakening are not captured (evidence is on maps like AIS 163 ‘il piede, i piedi/the foot, the feet’ and others).

7 Where is Locality in Lenition?

279Returning to the a–linear model and the question of locality in phonology, let us represent the positional contexts of the Mirror Coda ø__ and vCv as follows, Tab. 49:

Table 49 Coda Mirror ø__ and vCv in a–linear model. Structural Locality (Bracketing)

Onset
Strong Position		Weak Position
# [σ C V …	… C [σ C V …	… V [σ C V …

280The weak position has always been described as a unitary object: the intervocalic position V_V (for us V [_σ C V). Conversely, in the linear model of Coda Mirror, it is possible to refer to the anti–coda (= Coda Mirror) as a unitary object in a specified way: iff (if and only if) a pre–initial CV is given. Thus, based on this linear theory, to make out the strong position as unitary object a change of the Underlying Representation UR is required, which allows to characterize {# / C}__ as the unitary object øCv (with an empty nucleus before the onset). Nonetheless, it is the reverse disjunctive context __{# / C} that has motivated the introduction of the syllable and the coda in phonology.

281One could assume that the weak position, defined without having recourse to a disjunctive context (thus defined just as a postvocalic onset), is sufficient to describe the complementary contexts of the strong position (see Sauzet 1994).

282Lenition requires that phonology refers to locality other than through linear syllabic constituents (___ {C / #} and {C / #}__). We assume that locality is structural, see (68):

(68) Postvocalic context in Structural Locality (a–linear constituents)

283Lenition (Voicing and Gorgia) are spread in a licensing chain (nucleus onset), in the case of Tuscan lenition : (_v' v° __ ) C) = (_a' a°) g)… = <ago> = [ˈa:go] (Lt acu).

Conversely, the strong position is the onset insofar as it is only an onset (licensed without conditions on its content).

We have two types of structural adjacency in phonological constituents following a hierarchical model, the first one applies to lenition of postvocalic onset (the weak position) … V [_σC V …:

284We proposed a novel solution to the locality problem by advancing a model in which lenition always spreads locally at a non–oriented (structural) segmental adjacency upstream of syllabic constituents’ construction.

8 Conclusions

285In this paper, we studied the relation between lenition–Voicing and Gorgia in Tuscany, precisely the relation between Tuscan lenition (Voicing) and Florentine lenition Gorgia (Spirantization/Glottalization/loss). We focused on the differences between Central and Western Tuscan lenition and on the relation between Western Tuscan lenition (Voicing) and its Florentinization through Gorgia. The typology of lenition–Voicing in the Middle Ages supports the hypothesis that lenition–Voicing was first extended throughout Tuscany, including central Tuscany, but much more settled in the Western Tuscany. This hypothesis is also supported by the remnant of lenition–Voicing found in the Central area where Gorgia/aspiration dominates today (and also from the remnants of the lenition in SI). The lenition is thus for us an endogenous process not imported by an external lenition (Gallo–Italian or Western Romance), i.e., it is not borrowed from the outside. This also catches up with Contini’s (1961) idea that spirantization developed for structural reasons as opposed to widespread lenition (Voicing) that threatened the structural balance of voiceless Latin consonants.

286This hypothesis is strengthened by instances of initial Voicing in Tuscan dialects and SI conceivable as early as Late Latin, ‘gatto/cat’, ‘grasso/fat’, ‘gamba/leg’, ‘gambero’shrimp, ‘golfo/gulf, etc.

287This trend finds further development in the Pisano, Lucchese, Sienese, and Arezzo vernaculars. There is a Tuscan internal lenition/Voicing; as argued by Contini (1961) Gorgia/spirantization is an innovation, and we still find lenition around the spirantization (Florentinized) area (see also Giannelli and Cravens 1997). Lenition–Voicing has affected almost all Tuscany, although with asymmetrical lines of development and spatial differentiation that we emphasize. Lenition/Voicing in V_V has been dominant since the Middle Ages in Western Tuscany within words and across boundaries. It extended mainly with the velar stop in the area of Lucca and Pisa to Pistoia and Prato. The original condition of this lenition is phonotactics.

288We also used positional arguments to prove it. We show the allophonic variation (strong/weak) according to phonotactic conditions and the phonological input of lenition. The lenition/fortition process are almost always symmetrical, lenition (Voicing and Gorgia) are mostly found in weak contexts (V_V). There are exceptions that we discussed: aspiration (Gorgia) can also be found in strong contexts, #_, C._ (= ø__ in Strict CV), as well as in strong geminates C.C word–internally (as it was pointed out since Giacomelli 1932), in word internal positions and in phonotactics, while in a number of cases of lenition–Voicing can also be initial (Pisano, Senese #_ <gammello> ’camel’, <gattivo> ‘bad or prisoner’, It. gabbia ‘cage’, etc.). This also shows that the two lenition systems (Voicing and Gorgia) were coexistent and competitive (as it was also pointed out by Weinrich 1969²). The external influences have perhaps strengthened the lenition (Voicing) zones, while the Florentine influence pushed towards the aspiration (Gorgia).

289The area of Western Tuscany is an area where the prevailing lenition is Voicing, disrupted in the recent phase by the advancing Florentine spirantization from central Tuscany. But it is evident that Voicing persists in the Northern part and in some areas such as those of Lucca.

290We have also questioned the symmetry between the Strong/Weak position ({#, C}__ vs __{#, C}) and the symmetry/unicity of the Coda Mirror (ø__ vs __ø) within the Gorgia process and Tuscan lenition/Voicing of voiceless Latin stops. We were particularly concerned with the seemingly asymmetrical cases, since we have shown in a few areas Gorgia/weakened outcomes after a coda (such as C. _ [in ˈθu: θa] ‘in jumpsuit’, see Section 4.3). These cases do not let us to fully reduce straightforward in Tuscan the disjunctions of the coda and of the strong position in a unique symmetrical phonological object (ø__ vs __ø), the Coda Mirror. The symmetry also fails in several cases phrase–initial #– , in the phonotactic strong position triggered by Syntactic Doubling ([ˈt͡ʃɛ ˈɸ:e: ɸe] ‘there is pepper’): after a RS trigger and a latent consonant (C._ in W₁W₂). The double symmetry fails in a number of cases since we get the lenited geminate variant which replaced the voiceless stop in the UR. We found parallel examples of spirantized geminates also word–internally. This goes with the fact that Florentine middle class shows aspiration sometimes even in geminates word–internally [t͡ʃikˈkʰare] ‘to smoke a fag’. Based on Russo and Ulfsbjourninn (2017), we proposed to treat spirantized geminates as asymetrical geminates CVCV units in the templatic UR (Section 4.3), not fully specified underlyingly or as a chain F– ø_F. and ø_F –F in an a–linear model.

291We have observed in Sections 3 and 4 that the strength of consonants across word boundaries is a phonotactic and/or prosodic property, as it depends on W₁, which is either vowel–final or consonant–final; a latent consonant attached to it is realized under specific conditions with a morpho–syntactic and prosodic variability, this empty position is syllabified with the onset of W₂ under Syntactic Doubling conditions. The strong initial position is triggered either by syntax (syntactical constituency, the initial empty CV is triggered by a syntactic phase) or by prosody (word–final accent, i.e. footed constituency). This positional strength is symmetrical to the weak context V_V where we find the lenited variants under lenition–Voicing or Gorgia.

292We observed that Italo–Romance lenition/fortition is foot–sensitive regarding RS gemination (fortition) licensed by stress; for the lenition/Voicing (and Spirantization), the lenited segments are in a preferred foot–medial position but also foot–initial (in word internal position e.g. the lenited onset of penult stressed syllable), while the left edge of the foot is reserved to the strong position in phonotactics W₁W₂ under RS conditions. Furthermore, a few sub–types of lenition are also foot–sensitive, for instance, Gorgia is more likely to debuccalize in posttonic syllable in the weak right branch of the trochaic foot (to get the laryngeal sound [h]). The prosodic RS fortition/gemination is foot–sensitive (see Section 3), since gemination depends on the correct parsing of an uneven trochee or a binary trochaic foot, which incorporates a latent positional consonant stress–licensed across word boundaries (the QS–ld stress system licenses an empty C).

293We have concluded saying that locality in phonology refers to something other than syllabic constituents alone, this other thing is a structural adjacency in syllable–based and foot–based constituency. This structural adjacency is a minimal relationship that pre–exists the constituents (it defines the differentiated lexical accessibility of the phonic material).