Debunking Rhaeto-Romance: Synchronic Evidence from Two Peripheral Northern Italian Dialects

This paper explores two peripheral Northern Italian dialects (NIDs), namely Lamonat and Frignanese, with respect to their genealogical linguistic classification. The two NIDs exhibit morpho-phonological and morpho-syntactic features that do not fall neatly into the Gallo-Italic sub-classification of Northern Italo-Romance, but resemble some of the core characteristics of the putative Rhaeto-Romance language family. This analysis of Lamonat and Frignanese reveals that their con servative traits more closely relate to Rhaeto-Romance. The synchronic evidence from the two peripheral NIDs hence supports the argument against the unity and autonomy of Rhaeto-Romance as a language family, whereby the linguistic traits that distinguish Rhaeto-Romance within Northern Italo-Romance consist of shared retentions rather than shared innovations, which were once common to virtually all NIDs. In this light, Rhaeto-Romance can be regarded as an array of conservative Gallo-Italic varieties. The paper concludes with a discussion of the geo-sociolinguistic properties of the two peripheral dialect areas under investigation that lead to a conservative linguistic behaviour within the Lamonat and Frignanese speech communities. Given the relatively similar historical and geo-political background of these speech communities, we attempt the formulation of a geo-sociolinguistic model of linguistic innovation diffusion that captures the conservative behaviour of Lamonat and Frignanese. We propose that those dialect areas that, in Bartoli’s (1945) geo-spatial linguistic typology, are both “lateral” and “isolated” deflect linguistic innovations. This proposal must be interpreted within a more general “gravity” and “wave” sociolinguistic model of diffusion of linguistic innovations, whereby “lateral” and “isolated” dialect areas give rise to a mechanism that we call “the pond rock effect” and that renders such dialect areas resistant to language change.


Introduction
In Italian dialectology, an ongoing debate concerns the status of the putative Rhaeto-Romance language family. The name Rhaeto-Romance was the invention of Theodor Gartner in 1883, following the pioneering work of Ascoli in 1873. In his study of the Romance family, Ascoli This paper revisits the Questione Ladina by adding further evidence from two different, and relatively far apart, NIDs whose language family sub-classification would prove problematic if they were categorized as Gallo-Italic. In fact, they exhibit features of both Rhaeto-Romance and Gallo-Italic varieties, displaying de facto in-between characteristics. Our conclusions support Pellegrini's view in his seminal 1972 work Saggi sul ladino dolomitico e sul friulano ('Essays on Dolomitic Ladin and Friulian') that the linguistic features that most differentiate Rhaeto-Romance from the rest of the NIDs were once also common to most Gallo-Italic dialects and this is the reason they are partly found in isolated peripheral dialects like that of Zocca (Modena) and Lamon (Belluno). These features were lost due to common language innovations which, only relatively recently, have spread into the Gallo-Italian family.
We contend that the Rhaeto-Romance family can only be considered a single language group if we take into consideration the diachronic development of NIDs: the Rhaeto-Romance languages should be regarded as a group of conservative Romance languages. This view is supported by the observation that members of the putative Rhaeto-Romance language family are characterized by a greater deal of shared retentions from Medieval Romance (in the sense of Benincà 1: 3-19; Benincà 2: 53-86) than shared innovations. In genealogical terms, these shared retentions are not sufficient to make Rhaeto-Romance a full-fledged language family. Nevertheless, we do argue in favour of Rhaeto-Romance unity (but crucially not autonomy) based on the prominent conservative behaviour of its members in terms of linguistic retentions. Rhaeto-Romance must hence be conceived as a pre-Gallo-Italic stage common to most NIDs. By adopting this view, we can easily categorize borderline NIDs that do not fit neatly into either the Rhaeto-Romance or the Gallo-Italic classification. This is further confirmed by recent findings, which show that some Rhaeto-Romance features are commonly found outside the Rhaeto-Romance area and were once very productive across NIDs (Benincà,Schmid 3: 93). Diachronically speaking, these dialects represent an in-between stage: they are not as conservative as Rhaeto-Romance, but they have not yet developed linguistic innovations of the sort that characterize Gallo-Italic.
Historically, the Romansh-, Ladin-and Friulian-speaking areas have never enjoyed political unity and are usually not mutually intelligible, especially those varieties that are not in contact with one another. This, however, is true of most NIDs: the areas in which NIDs are spoken have never been united under the same political power and most NIDs are not mutually intelligible; nevertheless, we are able to classify them as Gallo-Italic. In the case of Rhaeto-Romance, more convincing evidence in favour of these languages' unity and autonomy as a language family is needed.  fittingly claim that, due to the lack of historical and external criteria, the unity of Rhaeto-Romance must base itself solely on structural considerations; nevertheless, the authors also note that such considerations are often ambiguous and, when it comes to linguistic levels other than phonology and morphology, there is no convincing evidence to put these languages together in a single language family. Regarding linguistic aspects, morphological and phonological features are, in our view, the most reliable for language reconstruction and the establishment of genetic links between languages. Syntax is not an ideal candidate for language reconstruction and genetic patterning due to its renowned high degree of volatility within a single language family (especially across NIDs) (see, for example, . Nonetheless, syntactic properties and parameters can be used to integrate morpho-phonological and morpho-syntactic clues. In our discussion, however, we omit purely syntactic notions and focus on some of the morphophonological and morpho-syntactic features of Lamonat and Frignanese, as we believe they suffice for this investigation.

Structure
The rest of this paper is structured as follows. In section 2, we present the features that distinguish Gallo-Italic from Rhaeto-Romance and introduce the dialectal areas under investigation. In section 3, we investigate the phonological features of Frignanese and Lamonat that are key for their categorization as either Gallo-Italic or Rhaeto-Romance. These include: (i) palatalization of the plosives [k] and [g] in front of [a]; (ii) the retention of the C + [l] cluster; (iii) the presence or absence of the front rounded vowels [y] and [ø]. In section 4, we focus on the morpho-syntactic traits of Frignanese and Lamonat that are ambiguous between Gallo-Italic and Rhaeto-Romance. Particular attention will be paid to the retention of the morpheme -s for the sigmatic plural and the second singular verb inflection. In section 5, we explore the extra-linguistic factors surrounding the characterization of what we define "Residual Rhaeto-Romance", in particular we try to identify those socio-cultural and sociohistorical factors that affected the spread of innovations across NIDs. We believe that similar models of communication density may have led to the same pattern of feature retention and innovation across relatively distant peripheral NIDs. Section 6 concludes the paper.

NIDs: Rhaeto-Romance vs. Gallo-Italic
Together with the varieties spoken in Liguria, Lombardy, Piedmont and some areas of Switzerland, the linguistic repertoire of Emilia Romagna belongs to the so-called Gallo-Italic group of the Romance languages (see Bossong 63). This linguistic group, in turn, is often referred to as part of Northern Italo-Romance, which also includes Venetan, Friulian and Ladin (Maiden & Parry 3). From a geographical perspective, the linguistic group comprising the dialects of northern Italy can be identified as the varieties spoken in the area contained between the Alps (north, east and west), the Adriatic Coast (east) and the Apennines (south). It is not, however, the mere geographical distribution that has rendered the dialects of northern Italy a linguistic sub-set. These varieties, though diverse, share and exhibit a coherent set of grammatical features.
Traditionally, the co-occurrence of the three phonological and morpho-syntactic features in (1) has been put forward as a key criterion by which to distinguish Rhaeto-Romance from Gallo-Italic within NIDs: i. The suffix -s also marks the second person singular of virtually all tenses and moods, except the imperative , as shown in (3) With the exception of the process of palatalization of velar plosives, the conservation of morphologically salient word-final -s and the C + [l] cluster represents shared retentions rather than shared innovations, challenging Rhaeto-Romance as an autonomous language family. Alongside the features in (1), Ascoli (337) proposes the following innovations as distinctive features of Rhaeto-Romance: (iv) the fronting of A to [e], (v) the diphthongization of tense Latin E and I to [ei], (vi) the fronting of tense Latin U and (vii) the velarization of [l] after A before a consonant. Haiman and Benincà in their seminal 1992 book exhaustively show that these features are not common to all members of the putative Rhaeto-Romance family. Therefore, they cannot be used as diagnostic traits of Rhaeto-Romance, despite, in genealogical terms, representing a much stronger set of features by virtue of being shared innovations. For the sake of completeness, Gartner (xxiii) adds two further morpho-syntactic features to those mentioned above, namely (viii) the preservation of reflexes of the Latin pronouns tu and ego, and (ix) the use of the pluperfect subjunctive in conditionals.  point out that this last trait is the only morpho-syntactic feature common to all members of the putative Rhaeto-Romance family; however, by itself, this is not sufficient evidence to claim in favour of the autonomy of Rhaeto-Romance as a discrete language family.
In sections 3 and 4, we discuss the core features in (1) for each dialect under investigation. We also briefly consider some of the additional features proposed by Ascoli (337) and Gartner (xxiii), establishing the level of conservation or innovation of each individual feature. The key assumption is that a continuum exists between Rhaeto-Romance and Gallo-Italic: Rhaeto-Romance is placed on the conservative side of the continuum, with Gallo-italic on side of innovation. Gallo-Italic has in fact lost the conservative traits typical of Rhaeto-Romance. For instance: the loss of word-final -s, palatalization of the C + [l], retention or reinstatement (see section 3) of the velar plosives [k] or [g] in front of [a]. In this respect, Lamonat and Frignanese constitute interesting cases as they are placed somewhere in the middle of the continuum. Most surprisingly, Frignanese seems to be closer to Rhaeto-Romance in the continuum of innovations than Lamonat, with the latter being unarguably closer to the Rhaeto-Romance geo-linguistic area. We hence make a synchronic case in the diachrony underlying the development of Rhaeto-Romance into Gallo-italic.
Specifically regarding Gallo-Italic, in his chapter on the issues surrounding the classification of Romance languages  questions whether the linguistic cohesion among the Gallo-Italic varieties is strong enough to be defined as a language in its own right. While he does not argue for re-classification, Bossong (68) does, however, note that the Gallo-Italic dialect group (or diasystem) 'forms a bridge between Gallo-and Italo-Romance'. Bossong's suggestion reflects the ongoing contention over the classification of Gallo-Italic as either a sub-set of Gallo-Romance or Italo-Romance. These two traditions are evident in the works of scholars who, whether explicitly or simply assumedly, refer to Gallo-Italic as belonging either to the same linguistic group (i.e. Italo-Romance) as all Romance varieties spoken across the peninsula, together with Corse and the Romance varieties of Canton Ticino (but excluding Occitan, Franco-Provençal, Sardinian, Ladin and Friulian) or as distinctive from the varieties spoken south of the Rimini-La Spezia line and therefore belonging to the Gallo-Romance family. In their literature review on the classification of Gallo-Italic,  critically evaluate the dominant Gallo-Romance tradition and argue in favour of Hull's (660) 'Padanian' proposal. In this genealogical approach, Hull recognizes a linguistic continuum, a unity, which spans across all Gallo-Italic varieties and includes the Rhaetic group. Although later reaffirmed in the works of Pellegrini (4: 1-13) and , this genealogical classification (which considers Gallo-Italic varieties as belonging to the same linguistic group of Rhaeto-Romance within the Gallo-Romance branch) has not helped the classificatory contention, de facto leaving the debate open: the evidence that we put forward in this paper supports this view.

Frignanese
The region of Emilia Romagna is split between two macro geo-linguistic areas: Emilian and Romagnol. The former is characterized by four main dialectal groups -namely, Western Emilian dialects, Central Emilian dialects, Western Romagnol dialects and Eastern Romagnol dialects (Hajek 271). At the heart of Emilia Romagna lies the province of Modena, whose Apennine inhabitants speak the so-called Frignanese varieties. It is possible to locate Frignano (the autonym attributed to most of the Modenese Apennines) as the portion of Modena belonging to the southernmost area of Emilia Romagna, geographically situated on the Tuscan-Emilian tract of the Italian Apennines. It is therefore a rural area whose linguistic varieties, as would be expected, still display conservative dialectal features. The Frignanese data presented in this paper were collected first-hand over several field trips in the village of Zocca. In addition, we also make use of data gathered from different sources for other varieties of Frignanese.

Lamonat
Lamonat is spoken in the south-western part of the province of Belluno, in the mountainous and fairly isolated municipality of Lamon. The speech community includes approximately 2,500 speakers. Lamonat is in close contact with Venetan varieties in the south-east, and Eastern Trentino varieties in the north-west. This NID exhibits phonological, morpho-syntactic and lexical characteristics of all main dialect areas from the region, namely Venetan, Ladin and Gallo-Italic (De Cia 43-55). The Lamonat data was collected first-hand during several field trips between 2015 and 2017.

Morpho-Phonological Features
Central to the classification of Rhaeto-Romance is the presence of two infrequent phonemes in the consonant inventory of its varieties, the palatal stops [c] and [Ɉ] (Schmid 1: 49)  Evidence of the once widespread presence of the palatal stops [c] and [Ɉ] across NIDs comes from isolated speech communities all over northern Italy, which demonstrate this trait synchronically: Alpine Lombard (Schmid 2: 480), Ligurian (Schmid 1: 50) and, crucially, in the Emilian dialects spoken in the Apennine area known as Frignano (Vitali 14). Schmid (80) puts forward the idea that this phonological phenomenon was widespread and, as well as being common to all Rhaeto-Romance languages, historically covered an area that included both Gallo-Italic and Gallo-Romance varieties. In his attempt to account for this diachronic explanation, which would unify these varieties under one linguistic group, Schmid (80) posits the label 'Northern Romance' in a narrower sense than its commonest use. 4 Vitali (39) supports this view, arguing that Proto-Emilian presented both palatal stops, [c, Ɉ], offering as evidence examples of conservative varieties (i.e. Lizzanese) where the phonemes are still retained (see also Loporcaro 85). Table 1 shows the different morpho-phonological environments in which the process of palatalization took place in two varieties of Frignanese.
The development of [c] and [Ɉ] did not occur in Latin and, as such, palatalization should be considered an innovation. There were two main morpho-phonological environments in which these processes took place: (i) before inherited vowel [a] and (ii) as reflex of Latin CL. underwent palatalization (and subsequent reintroduction), or whether they were conserved as such from Latin to present-day Lamonat. Given our discussion, the former seems the most plausible. With respect to this trait, Lamonat is in line with most NIDs.
The other main morpho-phonological feature that has been proposed as a distinctive trait of Rhaeto-Romance is the retention of the liquid in the cluster C + [l]. In Lamonat, along with most other NIDs, such a cluster has been palatalized. For instance, Latin CLAVE(M) 'key' retains the C + [l] cluster in Friulian [kla:f], but has undergone palatalization in Lamonat [ʧao]. The same is true for Latin GLACIE(M) 'ice' and UNGULA(M) 'fingernail', which have 5 With thanks to Arianna Uguzzoni for providing the data and confirming that these plosives can be heard in all word contexts, as illustrated in Table 1. Among the secondary features proposed to characterize Rhaeto-Romance, we note the presence of the two front rounded vowels [y] and [ø]. The vocalic system of both Lamonat and Frignanese comprise these two sounds -an addition that sets them apart from the common Romance vowel inventory (Schmid 2: 472). While it is attested that these vowels are found across NIDs, with the exception of Romagnol and Veneto in the southern part of the Northern Italian Dialect area (Benincà,Parry & Pescarini 190), they are only encountered in the Apennine varieties of Emilian (Loporcaro 106). These front rounded vowels were historically distributed across the Po plain, the so-called Taro-Secchia-Panaro isogloss (Savoia 226), and gradually lost as a result of the linguistic contact facilitated by the Via AEmilia. The loss, however, did not affect the varieties of Frignano until very recently, and some of the older speakers in the area can still be heard exhibiting both [y] and [ø], such as in [dy:r] 'hard' and [fø:g] 'fire'. As discussed in section 5, Vitali (6) argues that the reason for this gap is due to the role played by the Via AEmilia and that, while the two front rounded vowels did make their way into Modena and Bologna, they regressed once in contact with Romagnol. It is indeed thought that the absence of the vowels from the inventory of Romagnol acted as a deterrent and pushed their assimilation into remission, leaving traces of their existence only in the periphery. Along the same lines, the isolated location of the Lamonat speech community has favoured the retention of the front rounded vowels [y] and [ø], 6 of which there is remarkably no trace in the neighbouring Venetan and Eastern Trentino dialects. For instance, Lamonat [skøla] 'school', [sʧøs] 'snail', [dyr] 'hard', [myr] 'wall'. As previously mentioned, this feature is typical of western Rhaeto-Romance  and also some western Gallo-Italic varieties but is unattested in the linguistic area where Lamonat is spoken, namely among North-Eastern Italian Dialects. In a similar fashion, the Frignanese variety spoken in the mountainous town of Sestola exhibits both [ø], [y] and [ø ø], [yy] as contrastive phonemes, and although the synchronic presence of the front rounded vowels is no longer found in the variety of Zocca, it can still be heard in some of its hamlets, such as Montalbano, Monte Ombraro and Rosola (cf. Vitali 6). The scattered distribution of this feature across NIDs touches virtually all dialectal areas of northern Italy, and the marked nature of the feature itself suggests that, diachronically, the presence of front rounded vowels was much more widespread. It is no coincidence that varieties of Rhaeto-Romance, being highly conservative themselves, display this feature. The evidence points towards a redefinition of Rhaeto-Romance as a group of highly conservative Northern Italo-Romance varieties, whose traits are also found in peripheral isolated NIDs outside the traditional Rhaeto-Romance geographic Lamonat and Frignanese therefore appear to have some secondary morpho-phonological features of Rhaeto-Romance, but synchronically lack the main features of palatalization, namely [k] and [g] in front of inherited A (with the exception of Alto Frignanese) and the retention of the cluster C + [l]. In this respect, secondary traits are not accurate criteria for the classification of Rhaeto-Romance, whose existence in various peripheral NIDs seems to point to the conclusion that Rhaeto-Romance varieties are conservative forms of Gallo-Italic and lack proper autonomy. For instance, the process of velarization of [l] and the presence of the vowels [ø] and [y] are absent from Friulian , which exhibits the main morpho-phonological and morpho-syntactic traits of Rhaeto-Romance, and yet fails to conform to other varieties of Rhaeto-Romance.

Morpho-Syntactic Features
At the morpho-syntactic level, the feature that sets Rhaeto-Romance languages apart from the other NIDs is retention of the morphologically salient suffix -s for plural and second person singular verb endings. As far as the sigmatic plural is concerned, across NIDs wordfinal -s has been lost. Frignanese and Lamonat have also lost the suffix -s, but have not yet developed an alternative plural-marking strategy, showing invariant singular and plural forms for singular and plural number. This is especially true of masculine nouns. In Lamonat, the vast majority of masculine plural nouns have the same form as their singular counterparts: Lamonat singular [ranʧ] and plural [ranʧ] 'spiders' or singular [pom] and plural [pom] ' apples'. Unlike the neighbouring Venetan varieties, Lamonat does not differentiate the plural and the singular forms of masculine nouns after the loss of word-final -s. In this respect, it reflects a frozen stage between Rhaeto-Romance and Gallo-Italic. Unsurprisingly, Lamonat's plural nominal morphology resembles that of Friulian, but without word-final -s, as shown in Table 2. In this example, the masculine noun is ' dog' and the feminine noun is ' chair'. In order to better inform our discussion, we attempted the reconstruction of Proto-Lamonat's plural morphology prior to the loss of word-final -s. Furthermore, zero-marking is shown with the empty set symbol -ø.
Without written records of Lamonat, it is virtually impossible to reconstruct the plural morphology of the NIDs under scrutiny before the loss of plural -s. For instance, we cannot say whether or not the feminine plural ending in -e developed before or after the loss of -s. Invariant singular and plural forms of masculine nouns are common in the Bellunese area  (Benincà,, suggesting that the loss of the sigmatic plural occurred relatively recently in Bellunese as opposed to other NIDs. The same can be said for Frignanese, in which both singular and plural forms of the masculine are invariant, while the plural form of feminine nouns exhibits, in some varieties of Frignanese, the ending -i and the absence of -s. Another conservative trait that both Lamonat and Frignanese have maintained apropos marking the plural is palatalization of stem-final coronal consonants. This morpho-phonological process was once very productive across NIDs (Benincà,. In Lamonat, this is now limited to nouns ending in -l and -on. For example, Lamonat singular [kaval] 'horse' becomes plural [kavaj] and [porton] 'gate' becomes plural [portuj]. In Frignanese the same pattern is also exhibited in words such as [kaval] 'horse' > [kavaj] 'horses', but also [fradel] 'brother' and [fradej] 'brothers', while absent in plurals of nouns ending in -on. The retention of this trait in both NIDs may not signal direct affiliation with Rhaeto-Romance; however, it clearly shows the conservative nature of Lamonat and Frignanese as peripheral NIDs. In this respect, one striking feature of Lamonat concerns the formation of a restricted set of plural nouns. Possibly due to the later contact with Venetan and the relative distance between the two varieties at the time, Lamonat has borrowed from Venetan its entire plural forms (as opposed to just the plural morphology) of a limited number of nouns such as singular [ . Another interesting phenomenon in Lamonat, possibly contact-induced with Venetan, is the double plural marking -i on some singular nouns ending -l whose plural form is already marked by the palatal sound -j. For instance, we see Lamonat singular form [porθel] 'pig' and plural form [porθjei] 'pigs' or Lamonat singular form [fraðel] 'brother' and plural form [fraðjei] 'brothers'. These intriguing exceptions in the plural-marking system of Lamonat reveal that the language is still somehow transitioning from a Rhaeto-Romance system to a more innovating Gallo-Italic system, after a prolonged period of geo-political isolation and, allegedly, a relatively recent period of sustained contact with Venetan, which triggered such innovations.
As for Frignanese, we previously mentioned that invariant forms are found between masculine singular nouns and their plural counterparts. Nevertheless, in a limited set of nouns ending in a vowel, the quality of such vowels changes between the singular and plural form: some varieties of Frignanese thus employ word-final vowel raising and lengthening to distinguish the singular from the plural form; for example, [pɛ] 'foot' vs. [pe:] 'feet'. Crucially, a comparable, yet not identical, pattern is exhibited in present-day Oriental Andalucian. Hualde et al. (88) point out that certain Spanish varieties lower the middle vowels [o] and [e] as a plural marking strategy to differentiate the plural form from the singular form, which would otherwise be homophonous as a consequence of aspiration or elision of the sigmatic plural marker -s. Given that the loss of the plural marking -s is an ongoing innovation in Oriental Andalucian that affects word-final vowel quality, it is reasonable to consider vowel lengthening in Frignanese as direct synchronic evidence signalling the relatively recent loss of plural -s diachronically. 7 7 To a lesser extent, vowel lengthening in Frignanese resembles Friulian vowel lengthening, which encodes masculine vs. feminine gender distinction e.g.
[love] 'wolf' FEM. In Friulian, the devoicing of an originally intervocalic consonant along with the apocope of the final unstressed vowel (all vowels except feminine -a) determine the lengthening of the preceding vowel allegedly through a process of compensatory lengthening . Friulian long vowels are hence only found in final closed syllables with a single consonant coda . In Frignanese, the lengthening distinction is between singular vs. plural number. In both Friulian and Frignanese, the lengthening process marks a grammatical feature distinction and is the result of phonological erosion.
Identical forms for singular and plural nouns are also found in some peripheral Lombard and Piedmontese varieties (Benincà,. This implies again that the loss of the sigmatic plural has gradually affected all NIDs in a fashion that, crucially, is not simultaneous. Unavoidably, with respect to this morpho-syntactic feature some NIDs will be closer to Rhaeto-Romance and others to Gallo-Italic. The sigmatic plural found in Rhaeto-Romance languages must therefore be considered a conservative feature once common to both Gallo-Italic and Italo-Romance varieties, and later lost in the latter two sub-groups. In peripheral NIDs, this loss seems to be relatively recent, giving rise to the phenomena mentioned earlier and thus weakening the autonomy of Rhaeto-Romance languages as a language family. In this respect, the loss of the sigmatic plural is also gradually affecting Rhaeto-Romance. For example, the loss of final plural -s is evident within a restricted set of lexical items, including some determiners and numerals bearing masculine gender features as in (4), as opposed to the feminine counterpart of the same utterance in (5): (4) I nestr-i doi omp-s the.m.pl our-m.pl two.m.pl man.m-pl 'Our two husbands' (5) L-i-s nestr-i-s do-s femin-i-s the-f-pl our-m-pl two-f.pl woman-f-pl 'Our two wives' That the plural marker -s is receding in Friulian further questions the solidity of this feature in the classification of Rhaeto-Romance. 8 It is thus a shared retention across Rhaeto-Romance languages and further investigation would potentially reveal that this trait exhibits a more or less conservative character across Ladin, Friulian and Romansh varieties.
The other main morpho-syntactic feature upon which classification of Rhaeto-Romance pivots consists of the retention of word-final -s in the second singular verb paradigm of virtually all moods and tenses. In Gallo-Italic, the loss of -s in the verbal morphology goes hand in hand with the loss of the sigmatic plural: the morphologically salient suffix -s is lost altogether. As for Frignanese and Lamonat, the second singular -s has been lost in the verb morphology; however, under closer scrutiny it is possible to establish that this occurred relatively recently compared to most NIDs. As shown in Table 3, the second singular ending in -e is in stark contrast to the morphological forms of the first and third person singular, which show zero marking (-Ø). 8 An anonymous reviewer suggests that the solidity of plural marker -s as a distinguishing trait of Rhaeto-Romance can be further weakened by the fact that, diachronically, Friulian and Ladin displayed both -i and -s as plural markers for masculine nouns . This can still be seen synchronically in root allomorphy. Given Table 3, in Frignanese and Lamonat, the loss of the morphologically salient -s ending must have taken place after the loss of the word-final atonic vowels -e, -o, -u and -i. Indeed, the second person ending in -e is conserved, whereas the first person ending, allegedly in -o, has been lost.
In conclusion, at the morpho-syntactic level, both Frignanese and Lamonat exhibit characteristics that lie in between Gallo-Italic and Rhaeto-Romance with respect to the absence of the morphologically salient word-final -s. In both varieties, it is possible to reconstruct the recent loss of word-final -s by investigating how these languages morphologically encode plural number and express the second person singular verb inflection. In this respect, Rhaeto-Romance's identifying morpho-phonological features can be seen as characteristic traits of a previous stage of Gallo-Italian within Northern Italo-Romance. In the next section, we provide a socio-historical account to explain how and why Lamonat and Frignanese, and, more generally, peripheral NIDs, retained such conservative features.

Peripheral NIDs: A Geo-Sociolinguistic Model of Linguistic Innovation
In his structural outline of the dialects of Italy, Maiden (233) explains how these Latin-derived dialects are part of a geographical continuum, whereby the degree to which varieties differ from one another gradually increases with their geographical distance. Defining geo-linguistic borders between two or more dialects is consequently non-trivial and requires in-depth knowledge of vernaculars and their historical background.
As for Emilia Romagna, Emilian and Romagnol form two distinctive linguistic groups. From a historical perspective, the cultural-linguistic diversification between Emilia and Romagna is thought to have arisen at the time of the arrival of the Longobards in Emilia during the sixth century ce. This occupation occurred while the whole region was under Byzantine siege and did not affect modern Romagna until two centuries later, thus influencing the linguistic and cultural practices of Emilia at a deeper level . A further dialectal division of Emilia Romagna transcends the historical development of Emilian as separate from Romagnol and unifies the region in a tripartite horizontal sub-division. According to , there are three geo-linguistic strata cutting across the region: (i) the northernmost stratum running alongside river Po; (ii) the southernmost area which follows the foothills of the Tosco-Emilian Apennines and whose south border coincides with the Rimini-La Spezia isogloss; (iii) the historical road Via AEmilia in the middle, 9 which is thought to have contributed the most to the linguistic diversification of the region.
In this respect, the conservative features of Frignanese can be attributed to the relative geographical distance between these speech communities and the Via AEmilia. The Via was a very significant trading route, which allowed ease of access among traders and, with them, their linguistic innovations. Though it is undisputed that the Via AEmilia played a key role in shaping the linguistic repertoire of the region, the exact mechanism is still an area of debate, with some scholars attributing to it the lack of a koine in the region (Hajek 271). Similarly, the conservative features of Frignanese that bring it closer to Rhaeto-Romance compared to the other neighbouring varieties of Emilian cannot be attributed solely to the geographic location of the Via AEmilia or the mere historical happenings that occurred in the region. Rather, a more detailed diffusion model of linguistic innovation must be proposed.
From a macro-geographical perspective, one instantaneous observation can be drawn by looking at the map of northern Italy (and its surroundings) at the peripheral collocation of the linguistic areas of Frignanese, Lamonat and Rhaeto-Romance. As described in Andersen (39), Bartoli was one of the first Romance linguists to posit a solid theory explaining the linguistic divergences between peripheral and central areas. 10 Bartoli's norms of aree laterali 'lateral areas' and aree isolate 'isolated areas' seem to provide an explanatory account -at least in part -for how such conservative features can still be found in Frignanese and Lamonat, while also being present in the putative Rhaeto-Romance language family. Indeed, 'lateral areas' are peripheral zones in which innovations arrive at a later stage than in intermediate areas, where, by contrast, they spread more easily; 'isolated areas', on the other hand, are geographically isolated areas whose speech is expected to exhibit conservative forms. In order to further expand on Bartoli's theory, we must turn our attention to the geographical aspects of the areas from both a macro and micro perspective, 11 together with considerations of their socio-historical nature.
Starting with Frignanese, we have already mentioned the linguistic fragmentation of the area and the absence of a koine, and how these aspects might have been induced by the role played by the Via AEmilia. The notion of aree laterali exemplifies the central geo-linguistic stratum of the region (here intended as described by Hajek 271). This stratum is where the main cities of Emilia lie along the Via AEmilia, specifically Piacenza, Parma, Reggio Emilia, Modena and Bologna, and also where more innovative varieties are spoken (see, for example, Rognoni 1-12 for a discussion on present-day urban Modenese). The innovation and erosion of some marked morpho-phonological features are in turn explained by the effect of levelling caused by contact between different varieties converging on the Via AEmilia for commercial and trading purposes. Andersen (40) refers to this as the endocentric community attitude in his Adoption Theory (previously ' centrifugal force' in Saussure's works 281), according to which a speech community open to contact with other communities will become more prone to the process of levelling and therefore innovation, with the latter further aided by the dynamics of language acquisition (Andersen 47). Comparatively, Frignanese is spoken in the southernmost geo-linguistic stratum of Emilia Romagna, and still retains features now absent from the urban speech of Modena, being de facto a 'lateral area' in the macro-region.
In addition, given Frignano's geographical position as a rural, mountainous area in the region and province of Modena, it meets the criteria to be classified as an 'isolated area' at the micro-level. Bartoli's notion of aree isolate as linguistic areas where conservative features are retained due to the lack of exposure and contact with other varieties -often linked to absence of external communication and trade -is still central to the main existing theories of language change. Across NIDs, we propose that those varieties that are more conservative, and hence show more resemblance to Rhaeto-Romance, simultaneously exhibit the characteristics of lateral areas at the macro level and isolated areas at the micro level. In the case of Frignanese, it is a variety spoken in a rural mountainous area away from the primary and secondary trade routes (i.e. isolated area at the micro level), within a wider linguistic area at the side of the central Via AEmilia zone (i.e. lateral area at the macro level).
Bartoli's classification can be integrated via a more standard sociolinguistic model used for the spread of language innovations, namely the wave and gravity model ; see also . This model is based on population density and the geographical distance between urban foci (densely populated cities), and towns and villages (loosely populated towns and rural places) as the main criteria for how innovations spread. Innovations jump between highly populated cities following the gravity model: the more populated a city, the faster the innovation will be attracted to it. Once the same innovation is established within the highly populated urban setting, it starts spreading outwards like ripples on a pond, first reaching loosely populated smaller towns and then rural areas. In this case, the wave and gravity model could explain why the province of Bologna, with Bologna acting as a gravitational field, exhibits both Central Emilian and Eastern Romagnol features, serving as a linguistic area of transition between the two macro-varieties. Given the geographical location of Frignanese, the wave model predicts that a linguistic innovation would spread from Modena, and possibly from Bologna, in a similar fashion in all directions; nonetheless, Frignanese presents linguistic features that are not commonly found in Modenese and its surrounding areas, as well as being relatively discrete from Bolognese. In this respect, the wave and gravity model is not alone sufficient to account for the peripheral nature of certain NIDs, hence the need to integrate it with Bartoli's geo-spatial linguistic classification. We propose that the ripples consisting of the innovative feature (i.e. in a wave model of linguistic innovation diffusion) do not spread in a concentric fashion, but rather linguistic areas that, in Bartoli's terms, are isolated at the micro level and lateral at the macro level tend to resist the spread of innovations. In simpler terms, the mountainous area of Frignano breaks the concentric spread of innovations that are instead diffused all around it, as if water ripples were to find a rock jutting out of the surface of a pond. As a result, innovations do not affect that particular area. In light of this metaphor, we propose to name the aforementioned mechanism the pond rock effect within the existing wave model, whereby the 'pond rock' consists of a linguistic area that is lateral at the macro level and isolated at the micro level, in line with Bartoli's geo-spatial typology. As per evidence provided in Vitali (6) and Loporcaro (108) in support of our proposal, diatopic variation in Frignano appears in discontinuous pockets, which is in turn consistent with the pond rock effect.
Related also to Bartoli's notion of isolated areas, one final aspect must be taken into consideration when describing the conservative behaviour of certain peripheral NIDs. Specifically, Andersen's (39) centripetal force must be integrated into our proposal. According to Bartoli's conception, isolated areas (in Andersen 55) whose speech communities do not trade nor communicate on a regular basis, and therefore lack external contact, tend to reinforce their local features and are thus more likely to retain conservative linguistic traits. In this respect, the pond rock effect shields dialect areas from linguistic innovation, and Andersen's (39) centripetal force contributes to the retention of existing linguistic traits. Frignano and its pockets of conservative speech areas seem to fit with this: Zocca and Pavullo, separated by the River Panaro, were historically the main foci of the region in terms of trade, and their markets used to serve their respective slope/side of the river. Due to distance and lack of transport links, the cities of Modena and Bologna were only reached sporadically and mainly for administrative duties. Geographical boundaries such as a major river and mountains, coupled with long distances and lack of a good transport led to the rural communities of Frignano relying mainly on these villages of Zocca and Pavullo. It is therefore plausible that the most selfsufficient villages had little external contact and so did not go through a process of levelling, thus retaining some of the conservative features that were once common to all NIDs and synchronically are more closely related to Rhaeto-Romance.
In the case of Lamonat, this dialect exhibits phonological, morpho-syntactic and lexical characteristics of all the main dialect areas of the region, namely Venetan, Ladin and Gallo-Italic (De Cia 43-55). We previously claimed that Lamonat's features resemble an intermediate stage between Rhaeto-Romance and Gallo-Italic. The cause of this amalgamation of features must lie within the geo-political history of the dialect area of the Lamonat speech community, which neatly fits the sociolinguistic model proposed above.
The Lamonat speech community is located in a mountainous area on the west side of the valley that links the cities of Primiero and Feltre. Its isolated mountainous location, away from the main trade routes linking the most populated towns of the area, allowed Lamonat to maintain and further develop its pre-Venetan linguistic features. In this regard, Lamon was untouched by an important historical trade route within the region, namely, the Via di Schenèr, 'Schenèr Trail', which instead ran along the east side of the valley . The dialect area has hence been historically isolated at the micro-level in Bartoli's terms. The most compelling evidence of Lamon's isolation comes from the linguistic situation on the east side of the valley, where the Schenèr Trail ran. The Sovramontino speech community here demonstrates a heavier influence of Venetan, yet such features are mostly archaic (see . From a speculative point of view, these features might have been the result of a period of isolation by the Sovramontino speech community, which possibly began in 1631 when the town of Sovramonte was plagued by the Black Death . This historical event is important, as it might have dictated a deviation towards the Lamon territories in the trade route that linked Feltre to the Primiero valley (Province of Trento), thus a deviation in the Schenèr Trail. From that point onwards, Lamonat might have been in sporadic contact with a more innovative variety of Venetan. This contact was sporadic in the sense that the route deviation very probably ran at the base of the valley that separates Lamon and Sovramonte, still leaving the core Lamonat territories relatively unexposed. Nonetheless, this hypothesis should be corroborated by solid historical evidence that could show how synchronic linguistic (micro-)comparison is an important tool in carrying out historical research. Finally, at the macro level, the Lamonat territories historically qualify as a lateral area in Bartoli's geo-spatial linguistic typology. In 1420, Lamon, along with the territories of the city of Feltre, were incorporated into the Most Serene Republic of Venice : until the end of the eighteenth century, Lamon constituted a border territory of the Republic of Venice, away from the central territories of the new socio-political ruling class where contemporary Venetan developed.
In conclusion, Frignanese and Lamonat share similar socio-historical dynamics and geopolitical locations responsible for the retention of archaic features, rendering them closer to Rhaeto-Romance compared to other NIDs. Moreover, this provides evidence against the unity and autonomy of Rhaeto-Romance, which can be simply regarded as an array of very conservative Gallo-Italic varieties. Frignanese and Lamonat qualify as peripheral NIDs, whose speech communities are located in linguistic areas that are lateral and locally isolated in Bartoli's terms.

Conclusion
The genetic sub-classification of the peripheral dialects of Frignanese and Lamonat is less straightforward than that of most NIDs. The analysis and discussion of these two NIDs seems to weaken further the unity and autonomy of Rhaeto-Romance as a language family. Data from Frignanese and Lamonat supports the view that Ladin, Friulian, Romansh and their peripheral dialect areas are conservative or ' crystallized' NIDs: ultimately, they offer precious insights into the full array of the frozen developmental stages of Gallo-Italic. It is important to note that the features discussed here are not exclusive to Frignanese and Lamonat, but are also present in several other NIDs (Benincà,, strengthening our overall conclusion that the features which distinguish Rhaeto-Romance within Northern Italo-Romance were once common to virtually all NIDs. We also predict that those NIDs demonstrating some resemblance to Rhaeto-Romance and which, crucially, are geographically outside the Rhaeto-Romance dialect area, qualify as peripheral NIDs, and conform to the extra-linguistic characteristics outlined in our geo-sociolinguistic model of linguistic innovation. For instance, in Bartoli's terms, they can be regarded as both lateral linguistic areas at the macro level, and isolated linguistic areas at the micro level. This creates a pond rock effect which operates within the frame of the gravity and wave model and shows linguistic evolution spreading by bouncing off incoming linguistic innovations. At the same time, Andersen's (39) centripetal force is responsible for reinforcing the existing linguistic norms of the isolated speech communities, forming pockets of highly conservative linguistic varieties across NIDs.