The internal classification of the Romance languages is a complex and sometimes controversial topic which may not have one single answer. Several classifications have been proposed, based on different criteria.
Attempts at classifying Romance languages
Difficulties of classification The comparative method used by linguists to build family language trees is based on the assumption that the member languages evolved from a single proto-language by a sequence of binary splits, separated by many centuries. With that hypothesis, and the glottochronological assumption that the degree of linguistic change is roughly proportional to elapsed time, the sequence of splits can be deduced by measuring the differences between the members.
However, the history of Romance languages, as we know it, makes the first assumption rather problematic. While the Roman Empire lasted, its educational policies and the natural mobility of its soldiers and administrative officials probably ensured some degree of linguistic homogeneity throughout its territory. Even if there were differences between the Vulgar Latin spoken in different regions, it is doubtful whether there were any sharp boundaries between the various dialects. On the other hand, after the Empire's collapse, the population of Latin speakers was separated—almost instantaneously, by the standards of historical linguistics—into a large number of politically independent states and feudal domains whose populations were largely bound to the land. These units then interacted, merged and split in various ways over the next fifteen centuries, possibly influenced by languages external to the family (as in the so-called Balkan language area).
In summary, the history of Latin and Romance-speaking peoples can hardly be described by a binary branching pattern; therefore, one may argue that any attempt to fit the Romance languages into a tree structure is inherently flawed."Not only is the tree model inadequate to express the relationships between diatopically related varieties, but it may seriously distort the diachronic and synchronic study of language. Some would argue that this model works well within Indo-European linguistics, where the varieties under consideration (all written and therefore partially or fully standardized) are usually well separated in space and time and where the intervening varieties have all vanished without trace, removing any possibility of viewing the Indo-European family as a continuum. However, where the object of study is a series of now-existing varieties or a range of closely related varieties from the past, the tree model is open to a number of grave objections." (Penny 2000: 22). In this regard, the genealogical structure of languages forms a typical linkage."A linkage consists of separate modern languages which are all related and linked together by intersecting layers of innovations; it is a language family whose internal genealogy cannot be represented by any tree." (François 2014:171).
On the other hand, the tree structure may be meaningfully applied to any subfamilies of Romance whose members did diverge from a common ancestor by binary splits. That may be the case, for example, of the dialects of Spanish and Portuguese spoken in different countries, or the regional variants of spoken standard Italian (but not the so-called "Italian dialects", which are distinct languages that evolved directly from Vulgar Latin).
Criteria
The two main avenues to attempt classifications are historical and typological criteria:
Historical criteria look at the Romance languages' former development. For example, a widely employed model divided the Romance-speaking world between West and East based on whether plural nouns end in -s or in a vowel. Researchers have highlighted this is mainly valid from a historical point of view as the change appeared in Antiquity in the East (Italo-Romance, Dalmatian and Eastern Romance), while in the West plural nouns ending in -s were preserved past this stage but could be lost by more recent changes (such as aspiration of word-final -s in some varieties of Spanish and its phonetic loss in French). Another criterion taken into account is the distinction between conservative and innovative/progressive Romance languages. Generally, the Gallo-Romance languages (discussed further below) form the core "innovative" languages, with standard French often considered the most innovative of all. The phenomenon is attributed to language development in the Carolingian Empire with Northern Italy and Catalan region representing marginal areas of distribution. For example, Catalan, up to the thirteenth century, used the Romance writing modes common in the Occitan area. This created a contrast with the languages near the periphery (which include Spanish, Portuguese, Italian and Romanian) which are deemed as "conservative". Sardinian is generally acknowledged as the most conservative Romance languages, at least from a phonetic point of view. Dante famously denigrated the Sardinians for the conservativeness of their speech, remarking that they imitate Latin "like monkeys imitate men".Sardos etiam, qui non Latii sunt sed Latiis associandi videntur, eiciamus, quoniam soli sine proprio vulgari esse videntur, gramaticam tanquam simie homines imitantes: nam domus nova et dominus meus locuntur. ["As for the Sardinians, who are not Italian but may be associated with Italians for our purposes, out they must go, because they alone seem to lack a vernacular of their own, instead imitating gramatica as apes do humans: for they say domus nova [my house] and dominus meus [my master]." (English translation provided by Dante Online, De Vulgari Eloquentia, I-xi )] It is unclear whether this indicates that Sardinian still had a two-case system at the time; modern Sardinian lacks grammatical case. According to Gerhard Rohlfs France replaced Italy as a centre of diffusion of innovations around the sixth century. Typological criteria measure the structural features of Romance languages, mainly in synchrony. For example, the identification of the La Spezia–Rimini Line line, which is generally accepted as the main isogloss for consonantal lenition in Romance languages and which runs across north-central Italy just to the north of the city of Florence (whose speech forms the basis of standard Italian). In this scheme, "East" includes the languages of central and southern Italy, and the Eastern Romance languages in Romania, Greece, and elsewhere in the Balkans; "West" includes the languages of Portugal, Spain, France, northern Italy and Switzerland. Sardinian does not fit at all into this sort of division.Ruhlen M. (1987). A guide to the world's languages, Stanford University Press, Stanford. Further expansions on this are discussed below.
The standard proposal By applying the comparative method, some linguists have concluded that Sardinian became linguistically developed separately from the remainder of the Romance languages at an extremely early date. Among the many distinguishing features of Sardinian are its articles (derived from Latin IPSE instead of ILLE) and lack of palatalization of and before and other unique conservations such as ‘house’ (< ). Sardinian has plurals in /s/ but post-vocalic lenition of voiceless consonants is normally limited to the status of an allophonic rule, which ignores word boundaries (e.g. [k]ane 'dog' but su [ɡ]ane or su [ɣ]ane 'the dog'), and there are a few innovations unseen elsewhere, such as a change of /au/ to /a/. This view is challenged in part by the existence of definite articles continuing forms (e.g. sa mar 'the sea') in some varieties of Catalan, best known as typical of Balearic dialects. Sardinian also shares develarisation of earlier /kw/ and /ɡw/ with Romanian: Sard. abba, Rom. apă 'water'; Sard. limba, Rom. limbă 'language' (cf. Italian acqua, lingua).
According to this view, the next split was between Common Romanian in the east, and the other languages (the Italo-Western languages) in the west. One of the characteristic features of Romanian is its retention of three of Latin's seven noun cases. The third major split was more evenly divided, between the Italian branch, which comprises many languages spoken in the Italian Peninsula, and the Gallo-Iberian branch.
Another proposal However, this is not the only view. Another common classification begins by splitting the Romance languages into two main branches, East and West. The East group includes Romanian, the languages of Corsica and Sardinia, and all languages of Italy south of a line through the cities of Rimini and La Spezia (see La Spezia–Rimini Line). Languages in this group are said to be more conservative, i.e. they retained more features of the original Latin.
The West group split into a Gallo-Romance group, which became the Oïl languages (including French), Gallo-Italian, Occitan, Franco-Provençal and Romansh, and an Iberian Romance group which became Spanish and Portuguese.
Italo-Western vs. Eastern vs. Southern
A three-way division is made primarily based on the outcome of Vulgar Latin (Proto-Romance) vowels: + Outcome of Classical Latin vowels Classical Latin Proto-Romance Southern Italo-Western Eastern short A * long A short E * long E * short I * long I * short O * long O * short U * long U *
Italo-Western is in turn split along the so-called La Spezia–Rimini Line in northern Italy, which is a bundle of isoglosses separating the central and southern Italian languages from the so-called Western Romance languages to the north and west. Some noteworthy differences between the two are: Phonemic lenition of intervocalic stops, which happens to the northwest but not to the southeast. Degemination of geminate stops (producing new intervocalic single voiceless stops, after the old ones were lenited), which again happens to the northwest but not to the southeast. Deletion of intertonic vowels (between the stressed syllable and either the first or last syllable), again in the northwest but not the southeast. Use of plurals in /s/ in the northwest vs. plurals using vowel change in the southeast. Development of palatalized /k/ before /e, i/ to in the northwest vs. in the southeast. Development of , which develops to > (sometimes progressing further to ) in the northwest but in the southeast.
Recent scholarship argues for a more nuanced view. All of the "southeast" characteristics apply to all languages southeast of the line, and all of the "northwest" characteristics apply to all languages in France and (most of) Spain yet the Gallo-Italic languages are somewhere in between. These languages do have the "northwest" characteristics of lenition and loss of gemination however other seemingly clear boundaries are often obscured by local variations: The Gallo‒Italic languages have vowel-changing plurals rather than /s/ plurals. The Lombard language in north-central Italy and the Rhaeto-Romance languages have the "southeast" characteristic of instead of for palatalized /k/. The Venetian language in northeast Italy and some of the Rhaeto-Romance languages have the "southeast" characteristic of developing to . Lenition of post-vocalic /p t k/ is widespread as an allophonic phonetic realization in Italy below the La Spezia-Rimini line, including Corsica and most of Sardinia.