The case of ( per ) addietro in Old Florentine.

The aim of this paper is to describe what prima facie seems to be a typological rarum in Old Floren-tine (i.e. Old Italian). Specifically, we address here the morphosyntax of the temporal adverbial (per) addie-tro (lit. for at-back), which was commonly used in Old Florentine texts to encode a meaning roughly corresponding to [BEFORE]. Thus, it seems to go against the accepted generalization that spatial relations of front and back regularly express, respectively, ante-riority and posteriority across languages when they are ‘shifted’ from space to time (Haspelmath, 1997). We will provide a simple morphosyntactic explanation of the seemingly ‘exceptionality’ of (per) addietro based on a finer-grained representation of temporal expressions and a locality constraint on hierarchical structure triggering kinda -suppletive patterns (Boba-ljik, 2012; Moskal, 2013).


Introduction
The aim of this paper is to illustrate what prima facie seems to be a (so far unnoticed) typological rarum in Old Florentine and to give a plausible morphosyntactic explanation for it 1 .Specifically, we address here the morphosyntax of the temporal adverbial (per) addietro (lit. (for) at-back), which was commonly used in Old Florentine texts to encode a meaning roughly corresponding to [before].Namely, when used spatially, (per) addietro expresses posterior location [behind/back], as shown in (1), whereas, when used temporally, it seems to express anteriority [before], as in (2).
(1) nel mare questo cotale correre in-the sea this such run-inf innanzi e addietro… in-front and at-back 'Such a run up and down into the see.' Bono Giamboni, Vegezio, a. 1292 (Fior.)(2) I servi che per addietro the slaves that for at-back in Roma si ribellaro… in Rome cl-refl rebel-3pl-pst 'The slaves who rebelled before in Rome.' Bono Giamboni, Orosio, a. 1292 (Fior.)From a typological point of view, this pattern seems to go against the fairly robust generalization of Haspelmath (1997), who stated that, when an adverb encoding the spatial meaning [behind / back] is used temporally, it consistently expresses the meaning [after].
The facts are summarized in table 1.
The observation that in many genetically diverse languages the linguistic items used to talk about the location of things in space are also used to talk about the orientation of events in time has fascinated many researchers (cf.Anderson, 1973;Clark, 1973;Traugott, 1978;Jackendoff, 1983Jackendoff, , 1996;;Geeraerts, 1993Geeraerts, , 2010;;Lakoff & Johnson, 1999;Tenbrink, 2007, among many others).
He affirmed that "almost all cases" (Haspelmath, 1997: 56) follow this path, but did not give any possible counterexample.In (3) we present data from Japanese, where the word mae (front) is used with a temporal characterization, both with (3a) and without (3b) a deictic anchoring (cf.also Franco, 2013).
(3) (a) Mae ni asonda front loc play.pstkoto ga aru.fact nom exist/have 'We have played before.' lit.''(We) have the fact that (we) played at front.''(Moore, 2011: 766) 1 A preliminary classificatory or labelling clarification concerns the term Old Florentine, by which we refer to what is commonly described as Old Italian in the literature (cf.Salvi & Renzi, 2010), basing on Florentine texts of the 13/14 th century. 2 Cross-linguistically, there are many different lexical sources that give rise to temporal items, other than anterior and posterior markers based on spatial anterior and posterior markers (cf.Haspelmath, 1997: 63-65, cf. also Barbiers, 2007;Franco, 2013).The crucial fact here is that when a language employs spatial markers to convey a (sequential) temporal meaning before is based on (in) front and after is based on back.Psycholinguistic researches (cf.Boroditsky, 2000 and subsequent works) support the conceptual 'closeness' and the univocal orientation of spatial and temporal relations in the lexicon3 .
The rest of the paper is organized as follows.In section 2, I present the relevant empirical data concerning the distribution of (per) addietro in Old Florentine.In section 3, I highlight a 'transitivity' constraint on the distribution of the temporal adverbial under consideration.Section 4 provides a functionalist explanation of the Old Florentine facts, based on a finer-grained representation of temporal expressions.Section 5 attempts a characterization of the 'suppletive' behaviour of temporal expression in the terms of an internalist perspective, given a locality constraint and a hierarchical organization of 'morphology' into 'syntax'.The conclusions follow.

The data: (per) addietro in Old Florentine
As far as the case of (per) addietro (and its lexical variants per adietro, per adrieto, etc.) is concerned, interestingly, the phenomenon seems to be almost exclusively circumscribed to Old Florentine (with rare occurrences from Old Tuscan), which is standardly considered (cf. Salvi & Renzi, 2010) the direct ancestor of Modern Italian, as already pointed out (cf. fn. 1).The present study is based on a survey performed on the OVI [Opera del Vocabolario Italiano] database (cf.http://gattoweb.ovi.cnr.it/).We have found 273 tokens of per addietro and its lexical variants.
In (4) we provide a set of examples of constructions with per addietro.
(4) (a) The correspondence between Old Florentine per addietro and Modern Italian prima is attested in Italian etymological dictionaries (see e.g.Pianigiani, 1907), which confirm that the 'archaic' expression per addietro stands for prima or innanzi ('before', a word derived from Latin in ante, 'in front').The preposition per seems to encode here an 'atelic-extent marker' in the past (cf. Haspelmath, 1997: 120ff.),whereas its most common use in contemporary Italian is with a cause or benefactive meaning.Notice however that the use of the preposition per to encode an 'extension' in time or space is quite common in Modern Italian (e.g.sono rimasto lì per due mesi, 'I stayed there for two months'; ho continuato per due km, 'I continued for two kilometers').Obverse also that, in Old Florentine, expressions in which a determiner introduces the temporal adverbial are attested as shown in (4e) [there are 5/273 occurrences of (per) l'addietro and lexical variants in the OVI database]4 .
The temporal adverbial prima, with the meaning 'before', was widely attested in Old Florentine (hence, it was coexistent with addietro) and in the early stages of the language (XIII century) prima was commonly introduced by the stative preposition in, as shown in ( 6 Moreover, the puzzling nature of the linguistic encoding of the concept of temporal anteriority in Old Florentine can be well-represented by the presence of temporal expressions, which rely on the 'right' spatial source Tuvaluan, Welsh, Kabardian, Basque, among many other languages, provide evidence for a stative-like PP (or Case affix) that takes a temporal expression as its complement (Franco, 2011).Just to give an example of this pattern, consider data from Tuvaluan (Besnier, 2000), an Austronesian language, spoken in Tuvalu: (i)

A transitivity constraint on (per) addietro
In the examples above, avanti (in front) optionally preceded by an analogous (per, for) adpositional item 'correctly' encodes a relation of temporal anteriority and it is somewhat symmetrical to (per) addietro (hugely more diffuse in the OVI database).The relevant question now is: what triggers the unstable encoding of temporal expressions in Old Florentine?In particular, how can an item that means [back] apparently against Haspelmath's (1997) generalization sketched in table 1 encode a flavour of anteriority?
The first thing to be noted is that constructions with (in) prima and (per / in) addietro in Old Florentine differ in one crucial aspect: the former can 'take complements', while the latter is strictly intransitive (cf.Kurzon, 2008).I have detected 316 occurrences of per addietro (273) and in addietro (43) and their lexical variants, and in no case do these forms have an overt complement.On the other hand, with prima, as shown in ( 10 In this respect, (per) addietro shares some similarities with deictic items such as ago (fa in Modern Italian), precisely analysed by Williams (1994) as intransitive prepositions, namely prepositions that do not license a complement.Indeed, many languages have specific items to encode deictic temporal distance, namely the distance related to the time of speech 7 .Haspelmath (1997: 36-37) dubbed the semantic functions expressed by these items {distance-past} and {distance-future}.Examples from Italian are given in ( 11) and ( 12): (11) Gianni è arrivato un mese fa 'Gianni arrived a month ago.' (12) Gianni arriverà tra/in un'ora 'Gianni will arrive in a hour' In Old Italian {distance-past} was expressed with the same addietro or with a bi-clausal con-struction involving the verb fare (do) or essere (be) in the matrix clause, as shown below in ( 13) and ( 14 Hence, we may claim that the expression (per) addietro has solely an intransitive interpretation signalling the temporal distance of an event from the time of elocution.Nevertheless, as we have seen in ( 4), ( 5), (per) addietro in Old Florentine would be glossed/translated with prima ('before') and not with fa/addietro ('ago') in Modern Italian 8 .
In Modern Italian, prima has both an intransitive value (namely, anchored to the time/ context of elocution) as in ( 16) (thus matching Old Florentine addietro) and a transitive value, as in (17).Time measures (e.g.ora, now; allora, then; un anno, a year, etc.) can be (and usually {distance-past} {distance-future} 7 Still, this is not a general property which holds cross-linguistically.In fact, many languages (e.g.Turkish, Armenian, Ancient Greek, Persian, Punjabi, among others) mark the distance from the moment of speech with the same item which marks the distance with respect to an explicitly indicated point of time (i.e. when the temporal item takes a complement).ONOMÁZEIN 31 (junio de 2015): 282 -301 Ludovico Franco The case of (per) addietro in Old Florentine.When before was (apparently) based on back are) unexpressed and retrieved from the context/discourse (cf.Kayne, 2004).On the contrary, a temporal measure is obligatory with so-called {distance-past} markers of the ago type, as shown in ( 15 We clearly need a finer-grained system of temporal expressions to account for the data presented above and we will see that such a finer-grained characterization is crucial to account for the behaviour of potential counterexamples to Haspelmath's generalization such as the Old Florentine time adverbial (per) addietro, where at first sight [before] ≈ [back] 9 .

A finer-grained characterization of temporal distance: introducing retrospective (and prospective) values
A solution to the puzzles and difficulties outlined above may be found thanks to the following observation: in many languages, a different morpheme must be used when the reference point (i.e.somewhere in the future or in the past) is not the moment of speech.Haspelmath (1997: 36-38) employed the terms {distance-retrospective} and {distance-prospective} for these cases.In particular, the difference between retrospective and past distance expressions, and between prospective and future distance expressions can be represented by the English sentences in ( 18) and ( 19), taken from Haspelmath (1997: 98) 10 .
Further notice that, interestingly, many languages employ only clausal adverbials in order to express {distant past} functions (Haspelmath, 1997).An example of these bi-clausal constructions, which match the Old Florentine type in (15), is given here in (ii), for Babungo, a West African language.
(ii) ŋw kû.ndw lùu ŋú's bc`c Babungo (Schaub, 1985: 169) [he die] [now be years two] 'He died two years ago.' (lit.'He died.It's now two years.')9 A few words are in order on the preposition per, widely attested in Old Florentine texts together with the temporal marker addietro.Possibly per is generated in the temporal counterpart of PP path (as a marker of atelic extent, cf.Haspelmath, 1997) in a layered adpositional phrase structure (cf.Koopman, 2000;Schweikert, 2005;Svenonius, 2006;Cinque, 2010;Pantcheva, 2010;Den Dikken, 2010, among others).For instance, in recent work, Cinque (2010: 10) has proposed a very fine-grained configuration for spatial adpositions, on the basis of cross-linguistic data, as shown in (ii): ] ] ] Note indeed that in Old Florentine it can be separated to the temporal expression addietro and this is a hint of the fact that it is processed in a higher node and surfaces attached to addietro when the temporal expression (as common for retrospective items) is implicit/silent (cf.also Kayne, 2003;Cinque, 2011) il quale per due anni addietro era stato Capitano the-m rel for two years at.backbe-3sg-ipfv be-pst-ptcp Captain 'who had been Captain two years before.'Marchionne, Cronaca Fior., 1378-85 That this idea is on the right track can be confirmed by the fact that posteriority in time is commonly expressed (compositionally) in Modern Italian in a way which seems quite symmetrical to the construction with per addietro, namely by the strings da ora in avanti, from now on, lit.'from now to in front', or da allora in avanti, from then on (in these cases, again, temporal posteriority is markedly based on 'in front').
ONOMÁZEIN 31 (junio de 2015): 282 -301 Ludovico Franco The case of (per) addietro in Old Florentine.When before was (apparently) based on back The crucial fact is that in (18b)-(19b) the elapsed time is not (necessarily) related to the time of speech/elocution.{distance prospective} and {distance-retrospective} values have in common with {distance future} and {distance past} ones the properties of being 'intransitive' and of being somewhat anchored to the 'context' of elocution (vs.'transitive' markers of 'general' anteriority and posteriority which, as we have seen in ( 10) with the case of prima, are able to introduce arguments and may encode events detached from either the time or the context of elocution).
Given this finer-grained representation, it is easy to see that the Old Florentine item (per) addietro can express a {distance-retrospective} value, as clearly exemplified below in ( 20 Haspelmath (1997: 98-100) has found that, about in half (n. 13) of the languages of a sample of 27, the {distance-prospective} morpheme was identical to the {distance future} morpheme.Hence, an overt distinction {distance future} vs. {distance-prospective} is found in many languages of Haspelmath's sample, but there are also a number of languages (n.14) in which this distinction is lacking.Furthermore, Haspelmath (1997: 100) has shown that {distance-prospective} morphemes are also frequently identical to {posterior} morphemes.Thus, he proposed the implicational generalization reported in (21) (cf.also Franco, 2013): (21) If in a language the posterior and the distance-future markers are identical, then the distance-prospective marker also takes the same form.
In    20) , if in a given language the {anterior} and the {distance past} morphemes are lexicalized with the same item, then the {distance-retrospective} morpheme also takes the same form.This pattern, namely the representation of all the three values with the same morpheme, seems to be quite frequent (actually, the most common) in worlds' languages.
Comparing table 2 and table 3, we can also see that there are other specular patterns in the representation of posteriority and anteriority across languages (cf. Franco, 2013).The other attested patterns are indeed the following: (a) There are languages in which {distance future} = {distance-prospective} ≠ {posterior} morphemes (e.g.Hungarian, Lezgian) and, symmetrically, there are languages in which {distance past} = {distance-retrospective} ≠ {anterior} morphemes (e.g.Kolyma Yukaghir or, precisely, Old Florentine).
Given these empirical facts, we have to consider a further descriptive issue, crucial for the present discussion.In a number of languages, the spatial directional adverb 'back' is used for the {distance past} function (cf. Haspelmath, 1997: 93).Just to give an example, consider the data in (22) from Evenki.
(22) (a) Tar beje ilan-ma tyrgani-1-va that man three-acc day-pl-acc amaski suru-che-n.back go.away-pst-3sg 'That man left three days ago.'(b) Esikeken erne-re-p, and.now come-nfut-1pl.incsi-de suru-mu-d' e-nni amaski.you-clt go.away-vol-prs-2sg back 'We have just come, and/but you (already) want to go back.' Evenki (Nedjalkov, 1997: 186) Interestingly, also English in ( 23) and Modern Italian in (24) can lexicalize {distance past} in the same way as Evenki.Modern Italian can use for {distance past} purposes the same lexical item, addietro, which prima facie seems to encode the meaning before in Old Florentine.Kuteva, 2002: 49), can be a potential problem for Haspelmath's generalization (i.e. an item meaning back/behind acquires a +deictic before flavour) and Haspelmath (1997: 93) himself argued that the most interesting point about this use of 'back' is that it contrasts strikingly with the use of 'before' or 'in front' for expressing the same semantic function.Clearly, in this case the image of the observer moving forward in stationary time is predominant.
Actually, recalling again the fact that (per) addietro is necessarily anchored to the context/discourse but not necessarily connected to the time of elocution, with the fine-grained (tripartite) interpretation of temporal expressions reported above, we can argue that when endeavoured with a before 'flavour' it is nothing else than a {distance-retrospective} marker (see table 3, cf.example (20)).Now, if we assume that Haspelmath's generalization (cf.table 1) holds only for markers of general {anteriority/posteriority} (cf.again table 2ONOMÁZEIN 31 (junio de 2015): 282 -301 Ludovico Franco The case of (per) addietro in Old Florentine.When before was (apparently) based on back and table 3), possibly as a reflex of a universal cognitive constraint, we may therefore argue that nothing prevents a {distance-retrospective} morpheme from being expressed by means of a word meaning [back], just as happens for {distance past} morphemes in a number of languages (possibly due to a 'context sensitive' interpretation).In other words, the ban [back] ≈ [before] is crosslinguistically restricted to items of general anteriority only.
(Per) addietro merely represents an instance of a pattern in which the {anterior} morpheme is different from the {distance retrospective} morpheme, which is in turn equal to the {distance past} morpheme.This is an attested pattern in the temporal domain of 'anteriority' (as shown in table 3) and an analogous pattern is also attested in the domain of 'posteriority', as shown, for instance, by the cases of Hungarian and Chechen, where {posterior} is different from {distance prospec-tive} which is the same as {distance future} (cf.Franco, 2013).
With such an explanation, Haspelmath's generalization now circumscribed to the domain of general anteriory and posteriority is safe.
Crucially, in Franco's sample, there are no items signalling [back], which are able to lexicalize a marker of general anteriority (i.e. an item with a transitive value).
Hence, (per) addietro is only an apparent counterexample to Haspelmath's generalization, made more exceptional by the fact that Modern Italian takes a 'specular' path and expresses {distance-retrospective} with the item prima, which is the specific morpheme for general anteriority both in Old Florentine and in Modern Italian.
Once we accept that Haspelmath Generalization involves only markers of general anteriority/ posteriority, per addietro is perfectly licit as a marker of {distance-retrospective}.Support to our claim is also given by the fact that its use was subject to a transitivity constraint (i.e. it could not take an unrestricted DP complement, contrary to markers of general anteriority, as shown above in section 3).This functionalist explanation is possibly adequate enough to account for the Old Florentine facts, but I think that theoretical advancements on lexicalization allow us to provide a formal (i.e.internalist) characterization of the phenomena illustrated above.
Such a formal interpretation is presented in the following section.

Suppletion and locality in the temporal domain
The use of 'more specific' items to express {retrospective/past} (and/or {prospective/ future}) features in some languages may be interpreted as an instance of suppletion in the temporal domain.Suppletion is standardly defined as the association of a single lexical item with two (or more) phonologically unrelated forms (e.g.go vs. went), where the choice of forms is dependent on the morpho-syntactic context (Hippisley et al., 2004;Veselinova, 2006;Bobaljik, 2012;Moskal, 2013).As an instance of a tripartite suppletion pattern, consider the comparative / superlative morphology from Middle Persian and Italian in ( 25), where we have three distinct forms for three distinct functions (labelled here default, comparative and superlative, cf.Bobaljik, 2012).( 25 In order to formally explain the temporal patterns in table 2 and table 3, I will assume, following the recent literature (cf. Bobalijik, 2012;Moskal, 2013), that suppletion is governed by hierarchical structure and 'restricted' by locality.
Evidence that a kinda suppletive mecha-nism may be at work in the temporal domain in a number of languages is given by those patterns in which anterior/posterior bases are instead 'regularly' suffixed/prefixed by specific, usually deictic, morphemes11 .Consider, for instance, the case of Hungarian in the past (cf.Kenesei et al., 1998) Ezelo tt can only mean 'ago', due to the use of the proximal demonstrative ez.Crucially, both retrospective and distance past have a deictic/context sensitive value overtly expressed by the (distal/ proximal) demonstrative: they can be decomposed respectively into {that-before} (with the use of the distal demonstrative az), and into {this-before}.A similar regular pattern is at work for [anterior] values in Mosetén, an isolate language spoken in Bolivia, where the word poroma ('before') can be combined with the particle jike, a past-tense marker, to obtain a retrospective value or with the completive marker win, to obtain a distance-past 'ago' value (Sakel, 2004: 364;cf. table 3).
Assuming broadly Distributed Morphology (Halle & Marantz, 1993 and subsequent works) as a framework, I argue here that syntactic structure is the input to morphology, which in turn has to provide phonological material via Vocabulary Insertion.Then, a crucial assumption is that Vocabu-lary Insertion proceeds cyclically and locally, from the lowest element in the structure outwards (cf. Embick, 2010;Bobaljik, 2012;Moskal, 2013).We will see that such a model allows a principled explanation of the suppletive behaviour of the temporal items consider in this work 12 .In Distributed Morphology, suppletion is treated as contextual allomorphy, namely (bundles of) features have a context-free default exponent, but more specific contexts may trigger the insertion of a different element (Chung, 2009;Bobaljik, 2012;Moskal, 2013).In other words, Vocabulary Insertion may be underspecified, and thus may compete to realize a given node.Such competition is resolved by the Elsewhere Condition (Kiparsky, 1973), with the effect that more specific elements take precedence over more general ones.Furthermore morphological operations are constrained to act under strict conditions of locality13 .An example of the rough mechanism discussed above is given in ( 27), for comparatives and superlatives suppletive items (Bobaljik, 2012).
ONOMÁZEIN 31 (junio de 2015): 282 -301 Ludovico Franco The case of (per) addietro in Old Florentine.When before was (apparently) based on back Bobaljik (2012) has shown that such representation (in which comparative and superlative morphologies hierarchically 'contains' the adjectival root) correctly predicts that a pattern in which only the comparative (e.g.*good -better -goodest) is suppletive is virtually unattested in his sample of ca.300 languages.Bobaljik labelled this impossible pattern of lexicalization an *ABA pattern.
Indeed, the combination of a nested structure, locality and the elsewhere logic to root allomorph selection easily explains the nonexistence of *ABA lexicalizations: given that the comparative is 'contained' in the superlative, the comparative suppletive allomorph of an adjectival root will necessarily block the default allomorph of that root in a higher node, due to locality constraints (cf.Moskal, 2013 for similar assumptions for suppletive nouns and pronouns, and Caha, 2009 for an analogous interpretation of (im)possible case patterns).Hence, the *ABA pattern, namely the recycle of default root in the superlative is ruled out.
The crucial fact here is that the *ABA pattern described by Bobalijk can be easily translated in the domain of temporal expressions.Indeed, Haspelmath's implicational generalization and the facts reported in table 2 and table 3 can be interpreted in the terms of an *ABA constraint.A possible representation is given below 14 . ( With the representation in (28) we assume a configuration in which {retrospective / prospective} values are sandwiched between {distance past / distance future} and {anterior/posterior} slots.This interpretation is justified by the fact that {retrospective/prospective} elements have in common some features with both deictic and general posterior/ anterior ones, as shown by the possible nested structure represented in (29).In particular {retros- ONOMÁZEIN 31 (junio de 2015): 282 -301 Ludovico Franco The case of (per) addietro in Old Florentine.When before was (apparently) based on back (29) distance past/future {+ deictic }  retrospective/prospective {+ context sensitive / -anchored to time of speech } α anterior/posterior {-anchored to time of speech } √ anterior/posterior Without entering into technical details not strictly relevant for the present discussion at every (terminal) node dominating the root (α, ) a more specific (features' coopted) item can be inserted.
Locality prevents the use of the same item for {anteriority/posteriority} and {distance past/distance future} ones, once that a 'suppletive' form has been employed for {retrospective/prospective} items, while nothing present double suppletive patterns (see e.g.Modern Greek in table 2) with three distinct items employed in three distinct environments (just as it happens with comparative/superlative suppletion, as shown in (25)).Hence, the sensitivity to deictic features (and to the moment/context of speech) seems to be crucial to trigger suppletive patterns in the domain of temporal (distance) expressions.
For what concerns specifically Old Florentine, addietro would be inserted in the node α (instead of the default anteriority root prima) due to the pressure of a +deictic environment.The same +deictic (or, more precisely, context sensitive) environment is responsible in a number of language for the insertion in retrospective and/or distant past nodes of items originally conceived with the meaning [back] in the spatial domain.On the contrary, bare anteriority cannot be subject to such a lexicalization pattern.

Conclusion
Given the data reviewed in section 2 and the theoretical discussion provided in section 4, we have shown that (per) addietro in Old Flo-rentine is only an apparent counterexample to Haspelmath's generalization illustrated in section 3 (once we accept that it involves only markers of general anteriority/posteriority), because its use is subject to a deictic constraint (i.e. it cannot take an unrestricted DP complement, contrary to markers of general anteriority).Then, the 'locality condition' on Vocabulary Insertion introduced in section 5 is the innermost reason of Haspelmath's implicational generalization in (21).
We have provided a simple morphosyntactic explanation of the seemingly 'exceptionality' of (per) addietro based on a finer-grained representation of temporal expressions and a locality constraint on hierarchical structure triggering kinda-suppletive patterns along the lines of Bobaljik (2012) and Moskal (2013).

Acknowledgements
I wish to thank Rita Manzini, Leonardo Savoia and the anonymous reviewers of Onomázein for their helpful comments and suggestions.Thanks are also due to Molly McIlwrath, who proofread the manuscript.All errors are my own.I gratefully acknowledge the Portuguese National Science Foundation, Fundação para a Ciência e a Tecnologia (FCT), for supporting this work with the research grant IF/00846/2013.

References
anderson, John, M., 1973: An essay concerning aspect: some consideration of ageneral character arising from the Abbé Darrigol's analysis of the Basque verb, The Hague: Mouton.
), I have found many instances of transitivity.Consider also the examples from Modern Italian in (10c-d-sbjv-prs before of you '… that your sins die before you.' Fiori di filosafi, 1271-75 (Fior.)ONOMÁZEIN 31 (junio de 2015): 282 -301Ludovico Franco The case of (per) addietro in Old Florentine.When before was (apparently) based on back (c) Non ho mai visto NEG have-1sg-prs never see-pst-

(
18) (a) {distance past} Our son returned from the army two weeks ago.(b) {distance-retrospective} Do you remember when your brother paid us an unexpected visit, coming all the way from Chile?Fortunately, all of us were at home.Our son had returned from the army two weeks earlier/before (*ago).(19) (a) {distance future} Will Switzerland exist in fifty years' time?(b) {distance prospective} Tito died in 1980.Ten years later/after (*in ten years) Yugoslavia began to crumble.
Franco (2012)ausal expression for 'ago' resulted from the compositional meaning of the single elements which formed it (in the bi-clausal construction), in Modern Italian the syntactic transparence is lost and the modern form has become "lexically deictic".See alsoFranco (2012)for a syntactic analysis of the reconstruction[yp [xp]] > [xp] that derived the particle fa from the verb fare (to do).Notice also that addietro retains a temporal {distance-past} deictic meaning in contemporary Italian, sharing approximately the same distribution as particles like fa and or sono(lit.'noware').See the examples in (i) below: (i) Ho incontrato Gianni due anni fa/or sono/addietro have-1sg-prs meet-pst-ptcp Gianni two years ago 'I met him two years ago'.

TABLE 3
The case of (per) addietro in Old Florentine.When before was (apparently) based on back The case of (per) addietro in Old Florentine.When before was (apparently) based on backIn table 3 we can see that matching Haspelmath's implicational generalization in ( The case of (per) addietro in Old Florentine.When before was (apparently) based on back . ez means 'this', while az means 'that'.