Loss of the case system

The most significant changes between Classical Latin and Proto-Romance (and hence all the modern Romance languages) relate to the reduction or loss of the Latin case system, and the corresponding syntactic changes that were triggered.^[1]

The case system was drastically reduced from the six-case system of Classical Latin. Although five cases can be reconstructed for Vulgar Latin nouns (nominative, accusative, genitive, dative, and ablative), for Proto-Romance this had been reduced to three: nominative, accusative-ablative, and genitive-dative. This system is preserved best in pronouns. In the West, the genitive-dative disappeared with the genitive replaced by de + ablative and the dative by ad + accusative.^[2] This left only two cases: nominative and oblique. (However, a morphologically unmarked genitive, the so-called juxtaposition genitive, syntactically still discernible, survives in Old French and Old Occitan, also leaving traces in Old Italian and some modern Italian dialects.^[3]) Some of the older Gallo-Romance languages (in particular, Old French, Old Occitan, Old Sursilvan and Old Friulian, and in traces Old Catalan and Old Venetian) preserved this two-case system well into the literary period, and in Ibero-Romance languages, such as Spanish and Portuguese, as well as in Italian (see under Case), a couple of examples are found which preserve the old nominative. In the East, a genitive-dative made entirely of dative forms^[4] was retained but the nominative and accusative-ablative eventually merged.

Concomitant with the loss of cases, freedom of word order was greatly reduced. Classical Latin had a generally verb-final (SOV) but overall quite free word order, with a significant amount of word scrambling and mixing of left-branching and right-branching constructions. The Romance languages eliminated word scrambling and nearly all left-branching constructions, with most languages developing a rigid SVO, right-branching syntax. (Old French, however, had a freer word order due to the two-case system still present, as well as a predominantly verb-second word order developed under the influence of the Germanic languages.) Some freedom, however, is allowed in the placement of adjectives relative to their head noun. In addition, some languages (e.g. Spanish, Romanian) have an "accusative preposition" (Romanian pe, Spanish "personal a") along with clitic doubling, which allows for some freedom in ordering the arguments of a verb.

The Romance languages developed grammatical articles where Latin had none. Articles are often introduced around the time a robust case system falls apart in order to disambiguate the remaining case markers (which are usually too ambiguous by themselves) and to serve as parsing clues that signal the presence of a noun (a function that used to be served by the case endings themselves).

This was the pattern followed by the Romance languages: In the Romance languages that still preserved a functioning nominal case system (e.g., Romanian and Old French), only the combination of article and case ending serves to uniquely identify number and case (compare the similar situation in modern German). All Romance languages have a definite article (originally developed from ipse "self" but replaced in nearly all languages by ille "that (over there)") and an indefinite article (developed from ūnus "one"). Many also have a partitive article (dē "of" + definite article).

Latin had a large number of syntactic constructions expressed through infinitives, participles, and similar nominal constructs. Examples are the ablative absolute, the accusative-plus-infinitive construction used for reported speech, gerundive constructions, and the common use of reduced relative clauses expressed through participles. All of these are replaced in the Romance languages by subordinate clauses expressed with finite verbs, making the Romance languages much more "verbal" and less "nominal" than Latin. Under the influence of the Balkan sprachbund, Romanian has progressed the furthest, largely eliminating the infinitive. (It is being revived, however, due to the increasing influence of other Romance languages.)

Other changes

• Loss of phonemic vowel length, and change into a free-stressed language. Classical Latin had an automatically determined stress on the second or third syllable from the end, conditioned by vowel length; once vowel length was neutralized, stress was no longer predictable so long as it remained where it was (which it mostly did).
• Development of a series of palatal consonants as a result of palatalization.
• Loss of most traces of the neuter gender.
• Development of a series of analytic perfect tenses, comparable to English "I have done, I had done, I will have done".
• Loss of the Latin synthetic passive voice, replaced by an analytic construction comparable to English "it is/was done".
• Loss of deponent verbs, replaced by active-voice verbs.
• Replacement of the Latin future tense with a new tense formed (usually) by a periphrasis of infinitive + present tense of habēre "have", which usually contracts into a new synthetic tense. A corresponding conditional tense is formed in the same way but using one of the past-tense forms of habēre.
• Numerous lexical changes. A number of words were borrowed from the Germanic languages and Celtic languages. Many basic nouns and verbs, especially those that were short or had irregular morphology, were replaced by longer derived forms with regular morphology. Throughout the medieval period, words were borrowed from Classical Latin in their original form (learned words) or in something approaching the original form (semi-learned words), often replacing the popular forms of the same words.

Vowels

Every language has a different set of vowels from every other. Common characteristics are as follows:

• Most languages have at least five monophthongs /a e i o u/. The parent language of most of the Italo-Western Romance languages (which includes the vast majority) actually had a seven-vowel system /a ɛ e i ɔ o u/, which is kept in most Italo-Western languages. In some languages, like Spanish and Romanian, the phonemic status and difference between open-mid and close-mid vowels was lost. French has probably the largest inventory of monophthongs, with conservative varieties having 12 oral vowels /a ɑ ɛ e i ɔ o u œ ø y ə/^[5] and 4 nasal vowels /ɑ̃ ɛ̃ ɔ̃ œ̃/. European Portuguese also has a large inventory, with 9 oral monophthongs /a ɐ ɛ e i ɔ o u ɨ/, 5 nasal monophthongs /ɐ̃ ẽ ĩ õ ũ/, and a large number of oral and nasal diphthongs (see below).
• Some languages have a large inventory of falling diphthongs. These may or may not be considered as phonemic units (rather than sequences of vowel+glide or vowel+vowel), depending on their behavior. As an example, French, Spanish and Italian have occasional instances of putative falling diphthongs formed from a vowel plus a non-syllabic /i/ or /u/ (e.g. Spanish veinte [ˈbejnte] "twenty", deuda [ˈdewða] "debt"; French paille [pɑj] "straw", caoutchouc [kautʃu] "rubber"; Italian sei [ˈsɛi] "you are/six", neutro [ˈnɛutro] "neuter"), but these are normally analyzed as sequences of vowel and glide. The diphthongs in Romanian, Portuguese, Catalan and Occitan, however, have various properties suggesting that they are better analyzed as unit phonemes. Portuguese, for example, has the diphthongs /aj ɐj ɛj ej ɔj oj uj aw ɛw ew iw (ow)/, where /ow/ (and to a lesser extent /ej/) appear only in some dialects. All except /aw ɛw/ appear frequently in verb or noun inflections. (Portuguese also has nasal diphthongs; see below.)
• Among the major Romance languages, Portuguese and French have nasal vowel phonemes, stemming from nasalization before a nasal consonant followed by loss of the consonant (this occurred especially when the nasal consonant was not directly followed by a vowel). Originally, vowels in both languages were nasalized before all nasal consonants, but have subsequently become denasalized before nasal consonants that still remain (except in Brazilian Portuguese, where the pre-nasal vowels in words such as cama "bed", menos "less" remain highly nasalized). In Portuguese, nasal vowels are sometimes analyzed as phonemic sequences of oral vowels plus an underlying nasal consonant, but such an analysis is difficult in French because of the existence of minimal pairs such as bon /bɔ̃/ "good (masc.)", bonne /bɔn/ "good (fem.)". In both languages, there are fewer nasal than oral vowels. Nasalization triggered vowel lowering in French, producing the 4 nasal vowels /ɑ̃ ɛ̃ ɔ̃ œ̃/ (although most speakers in France nowadays pronounce /œ̃/ as /ɛ̃/). Vowel raising was triggered in Portuguese, however, producing the 5 nasal vowels /ɐ̃ ẽ ĩ õ ũ/. Vowel contraction and other changes also resulted in the Portuguese nasal diphthongs /ɐ̃w̃ ɐ̃j̃ ẽj̃ õj̃ ũj̃/ (of which /ũj̃/ occurs in only two words, muito /mũj̃tu/ "much, many, very", and mui /mũj̃/ "very"; and /ẽj̃/ is actually a final-syllable allophone of /ẽ/).
• Most languages have fewer vowels in unstressed syllables than stressed syllables. This again reflects the Italo-Western Romance parent language, which had a seven-vowel system in stressed syllables (as described above) but only /a e i o u/ (with no low-mid vowels) in unstressed syllables. Some languages have seen further reductions: e.g. Standard Catalan has only [ə i u] in unstressed syllables. In French, on the other hand, any vowel may take prosodic stress.
• Most languages have even fewer vowels in word-final unstressed syllables than elsewhere. For example, Old Italian allowed only /a e i o/, while the early stages of most Western Romance languages allowed only /a e o/. The Gallo-Romance languages went even farther, deleting all final vowels except /a/. Of these languages, French has carried things to the extreme by deleting all vowels after the accented syllable and uniformly accenting the final syllable (except for a more-or-less non-phonemic final unstressed [ə] that occasionally appears). Modern Spanish now allows final unstressed /i u/, and modern Italian allows final unstressed /u/, but they tend to occur largely in borrowed or onomatopoeic words, e.g. guru "guru", taxi "taxi", Spanish tribu "tribe" and espíritu "spirit" (loanwords from Classical Latin), Italian babau ~ baubau "bogeyman" (onomatopoeic, cf. English "boo!"). The apparent Spanish exception casi "almost" originates from Latin quasi "as if" < quam sī, and was probably influenced by si "if".
• Phonemic vowel length is uncommon. Vulgar Latin lost the phonemic vowel length of Classical Latin and replaced it with a non-phonemic length system where stressed vowels in open syllables were long, and all other vowels were short. Standard Italian still maintains this system, and it was rephonemicized in the Gallo-Romance languages (including the Rhaeto-Romance languages) as a result of the deletion of many final vowels. Some northern Italian languages (e.g. Friulan) still maintain this secondary phonemic length, but in most languages the new long vowels were either diphthongized or shortened again, in the process eliminating phonemic length. French is again the odd man out: Although it followed a normal Gallo-Romance path by diphthongizing five of the seven long vowels and shortening the remaining two, it phonemicized a third vowel length system around 1300 AD in syllables that had been closed with an /s/ (still marked with a circumflex accent), and now is phonemicizing a fourth system as a result of lengthening before final voiced fricatives.
• In modern spoken and literary Romanian, Slavic influences are evident in phonetics and morphology. Phonetic Slavicisms include the iotation of the initial e in words such as el, ea, este pronounced [jel], [ja], [ˈjeste] (compare Spanish: el, ella, estamos, without the Slavic iotation effect).

Consonants

Most Romance languages have similar sets of consonants. The following is a combined table of the consonants of the five major Romance languages (French, Spanish, Italian, Portuguese, Romanian).

Key:

• bold: Appears in all 5 languages.
• italic: Appears in 3–4 languages.
• (parentheses): Appears in 2 languages.
• ((double parentheses)): Appears in only 1 language.

Notable changes:

• Spanish has no phonemic voiced fricatives (however, [β ð ɣ] occur as allophones of /b d ɡ/ after a vowel and after certain consonants). The equivalent of /v/ merged with /b/, and all the rest became voiceless. It also lost /ʃ/, which became /x/ or /h/ in some other dialects.
• The western languages (French, Spanish, Portuguese) all used to have the affricates /ts/, /dz/, /tʃ/, /dʒ/. By the fourteenth century or so, these all turned into fricatives except for Spanish and dialectal Portuguese /tʃ/. (Spanish /ts/ ended up becoming /θ/, at least in Northern, Central and some parts of Southern Spain; elsewhere, it merged with /s/, as in some other Romance languages, such as French and Portuguese.) Romanian /dz/ likewise became /z/.
• French, and most varieties of Spanish, have lost /ʎ/ (which merged with /j/). Romanian merged both /ʎ/ and /ɲ/ into /j/.
• Romanian was influenced by Slavic phonology, most notably in the palatalization of consonants in the plural form (for example pomi [pomʲ] and lupi [lupʲ]).

Most instances of most of the sounds below that occur (or used to occur, as described above) in all of the languages are cognate. However:

• Although all of the languages have or used to have /tʃ/, almost none of these sounds are cognate between pairs of languages. The only real exception is many /tʃ/ between Italian and Romanian, stemming from Latin C- before E or I. Italian also has /tʃ/ from Vulgar Latin -CI-, and from -TI- following a consonant (elsewhere /ts/). Romanian also has /tʃ/ from -TI- before a stressed syllable. Former French /tʃ/ is from Latin C- before A, either word-initial or following a consonant; Spanish /tʃ/ is from Latin -CT-, or from PL, CL following a consonant; former Portuguese /tʃ/ is from Latin PL, CL, FL, either word-initial or following a consonant.
• Italian and former Romanian /dz/ (from some instances of Vulgar Latin -DI-) are not cognate with former western /dz/ (from lenition of /ts/).

Lexical stress

Word stress was rigorously predictable in classical Latin except in a very few exceptional cases, either on the penultimate syllable (second from last) or antepenultimate syllable (third from last), according to the syllable weight of the penultimate syllable. Stress in the Romance Languages mostly remains on the same syllable as in Latin, but various sound changes have made it no longer so predictable. Minimal pairs distinguished only by stress exist in some languages, e.g. Italian Papa [ˈpaːpa] "Pope" vs. papà [paˈpa] "daddy", or Spanish límite [ˈlimite] "[a] limit", present subjunctive limite [liˈmite] "[that] [I/(s)he] limit(s)" and preterite limité [limiˈte] "[I] limited".

Erosion of unstressed syllables following the stress has caused most Spanish and Portuguese words to have either penultimate or ultimate stress: e.g. Latin trēdecim "thirteen" > Spanish trece, Portuguese treze; Latin amāre "to love" > Spanish/Portuguese amar. Most words with antepenultimate stress are learned borrowings from Latin, e.g. Spanish/Portuguese fábrica "factory" (the corresponding inherited word is Spanish fragua, Portuguese frágua "forge"). This process has gone even farther in French, with deletion of all post-stressed vowels, leading to consistent, predictable stress on the last syllable: e.g. Latin Stephanum "Stephen" > Old French Estievne > French Étienne /etjɛn/; Latin juvenis "young" > Old French juevne > French jeune /ʒœn/. This applies even to borrowings: e.g. Latin fabrica > French borrowing fabrique /fabʀik/ (the inherited word in this case being monosyllabic forge < Pre-French *fauriga).

Other than French (with consistent final stress), the position of the stressed syllable generally falls on one of the last three syllables. Exceptions may be caused by clitics or (in Italian) certain verb endings, e.g. Italian telefonano [teˈlɛːfonano] "they telephone"; Spanish entregándomelo [entɾeˈɣandomelo] "delivering it to me"; Italian mettiamocene [metˈtjaːmotʃene] "let's put some of it in there"; Portuguese dávamos-vo-lo [ˈdavɐmuʒvulu] "we were giving it to you". Stress on verbs is almost completely predictable in Spanish and Portuguese, but less so in Italian.

Nominal morphology

Nouns, adjectives, and pronouns can be marked for gender, number and case. Adjectives and pronouns must agree in all features with the noun they are bound to.

Number

The Romance languages inherited from Latin two grammatical numbers, singular and plural; the only trace of a dual number comes from Latin ambō > Spanish and Portuguese ambos, Old Romanian îmbi > Romanian ambii, Old French ambe, Italian ambo, ambedue, entrambi.

Gender

Most Romance languages have two grammatical genders, masculine and feminine. The gender of animate nouns is generally natural (i.e. nouns referring to males are generally masculine, and vice versa), but for non-animate nouns it is arbitrary, a grammatical category with no correspondence to natural gender.

Although Latin had a third gender (neuter), there is little trace of this in most languages. The biggest exception is Romanian, where there is a productive class of "neuter" nouns, which include the descendants of many Latin neuter nouns and which behave like masculines in the singular and feminines in the plural, both in the endings used and in the agreement of adjectives and pronouns (e.g. un deget "one finger" vs. două degete "two fingers", cf. Latin digitus, pl. digiti). This behavior happens also in Italian with a restricted number of words (e.g. un uovo "an egg", il braccio "the arm" masculine in the singular, le uova "the eggs", le braccia "the arms" feminine in the plural). Another noteworthy exception is Asturian, which uses a neuter gender to refer to abstract, collective and uncountable entities. It appears in adjectives (la xente galbaniego "lazy people", l'agua frío "cold water", pensar escoso "fruitless thinking"), possessives (el dineru mío "my money", la sidra vueso "your cider"), one neuter article (Lo guapo ye... "The beautiful [thing] is...", Lo que cal agora ye colar "The convenient [thing] right now is to get out"), one neuter pronoun (elli/ella/ello "he/she/it") and some nouns (un pelu → el pelo "one hair → the hair", un fierru → el fierro "an iron bar → iron [material]". The neuter gender does not admit plural forms emerging from metaphorical uses, changing to masculine or feminine (l'agua frío → les agües fríes "cold water "→ "cold waters"). Some of these neuter traces also feature in Spanish, such as ello ("it").

Such nouns arose because of the identity of the Latin neuter singular -um with the masculine singular, and the identity of the Latin neuter plural -a with the feminine singular. A similar class exists in Italian, although it is no longer productive (e.g. il dito "the finger" vs. le dita "the fingers", l'uovo "the egg" vs. le uova "the eggs"). A similar phenomenon may be observed in Albanian (which is heavily Romance-influenced), and the category remains highly productive with a number of new words loaned or coined in the neuter ((një) hotel one hotel(m) vs. (tri) hotele three hotels (f)). (A few isolated nouns in Latin had different genders in the singular and plural, but this was an unrelated phenomenon; this is similarly the case with a few French nouns, such as amour, délice, orgue.)

Spanish also has vestiges of the neuter in the demonstrative adjectives: esto, eso, aquello, the pronoun ello (meaning "it") and the article lo (used to intensify adjectives). Portuguese also has neuter demonstrative adjectives: "isto", "isso", "aquilo" (meaning "this [near me]", "this/that [near you]", "that [far from the both of us]").

Remnants of the neuter, interpretable now as "a sub-class of the non-feminine gender" (Haase 2000:233), are vigorous in Italy in an area running roughly from Ancona to Matera and just north of Rome to Naples. Oppositions with masculine typically have been recategorized, so that neuter signifies the referent in general, while masculine indicates a more specific instance, with the distinction marked by the definite article. In Southeast Umbrian, for example, neuter lo pane is 'the bread', while masculine lu pane refers to an individual piece or loaf of bread. Similarly, neuter lo vinu is wine in general, while masculine lu vinu is a specific sort of wine, with the consequence that mass lo vinu has no plural counterpart, but lu vinu can take a sortal plural form li vini, referring to different types of wine. Phonological forms of articles vary by locale.^[6]

Case

Latin had an extensive case system, where all nouns were declined in six cases (nominative, vocative, accusative, dative, genitive, and ablative) and two numbers.^[1] Many adjectives were additionally declined in three genders, theoretically leading to a possible 6 × 2 × 3 = 36 endings per adjective. In practice, some category combinations had identical endings to other combinations, but a basic adjective like bonus "good" still had 14 distinct endings.

In all Romance languages, this system was drastically reduced. In most modern Romance languages, in fact, case is no longer marked at all on nouns, adjectives and determiners, and most forms are derived from the Latin accusative case. Much like English, however, case has survived somewhat better on pronouns.

Most pronouns have distinct nominative, accusative, genitive and possessive forms (cf. English "I, me, mine, my"). Many also have a separate dative form, a disjunctive form used after prepositions, and (in some languages) a special form used with the preposition con "with" (a conservative feature inherited from Latin forms such as mēcum, tēcum, nōbīscum).

The system of inflectional classes is also drastically reduced. The basic system is most clearly indicated in Spanish, where there are only three classes, corresponding to the first, second and third declensions in Latin: plural in -as (feminine), plural in -os (masculine), plural in -es (either masculine or feminine). The singular endings exactly track the plural, except the singular -e is dropped after certain consonants.

The same system underlines many other modern Romance languages, such as Portuguese, French and Catalan. In these languages, however, further sound changes have resulted in various irregularities. In Portuguese, for example, loss of /l/ and /n/ between vowels (with nasalization in the latter case) produces various irregular plurals (nação – nações "nation(s)"; hotel – hotéis "hotel(s)").

In French and Catalan, loss of /o/ and /e/ in most unstressed final syllables has caused the -os and -es classes to merge. In French, merger of remaining /e/ with final /a/ into [ə], and its subsequent loss, has completely obscured the original Romance system, and loss of final /s/ has caused most nouns to have identical pronunciation in singular and plural, although they are still marked differently in spelling (e.g. femme – femmes "woman – women", both pronounced /fam/).

Noun inflection has survived in Romanian somewhat better than elsewhere.^[7]: 399 Determiners are still marked for two cases (nominative/accusative and genitive/dative) in both singular and plural, and feminine singular nouns have separate endings for the two cases. In addition, there is a separate vocative case, enriched with native development and Slavic borrowings (see some examples here) and the combination of noun with a following clitic definite article produces a separate set of "definite" inflections for nouns.

The inflectional classes of Latin have also survived more in Romanian than elsewhere, e.g. om – oameni "man – men" (Latin homo – homines); corp – corpuri "body – bodies" (Latin corpus – corpora). (Many other exceptional forms, however, are due to later sound changes or analogy, e.g. casă – case "house(s)" vs. lună – luni "moon(s)"; frate – fraţi "brother(s)" vs. carte – cărţi "book(s)" vs. vale – văi "valley(s)".)

In Italian, the situation is somewhere in between Spanish and Romanian. There are no case endings and relatively few classes, as in Spanish, but noun endings are generally formed with vowels instead of /s/, as in Romanian: amico – amici "friend(s) (masc.)", amica – amiche "friend(s) (fem.)"; cane – cani "dog(s)". The masculine plural amici is thought to reflect the Latin nominative plural -ī rather than accusative plural -ōs (Spanish -os); however, the other plurals are thought to stem from special developments of Latin -ās and -ēs.

A different type of noun inflection survived into the medieval period in a number of western Romance languages (Old French, Old Occitan, and the older forms of a number of Rhaeto-Romance languages). This inflection distinguished nominative from oblique, grouping the accusative case with the oblique, rather than with the nominative as in Romanian.

The oblique case in these languages generally inherits from the Latin accusative; as a result, masculine nouns have distinct endings in the two cases while most feminine nouns do not.

A number of different inflectional classes are still represented at this stage. For example, the difference in the nominative case between masculine li voisins "the neighbor" and li pere "the father", and feminine la riens "the thing" vs. la fame "the woman", faithfully reflects the corresponding Latin inflectional differences (vicīnus vs. pater, fēmina vs. rēs).

A number of synchronically quite irregular differences between nominative and oblique reflect direct inheritances of Latin third-declension nouns with two different stems (one for the nominative singular, one for all other forms), most with of which had a stress shift between nominative and the other forms: li ber – le baron "baron" (barō – barōnem); la suer – la seror "sister" (soror – sorōrem); li prestre – le prevoire "priest" (presbyter – presbyterem); li sire – le seigneur "lord" (senior – seniōrem); li enfes – l'enfant "child" (infāns – infantem).^{[9]: 36–39}

A few of these multi-stem nouns derive from Latin forms without stress shift, e.g. li om – le ome "man" (homō – hominem). All of these multi-stem nouns refer to people; other nouns with stress shift in Latin (e.g. amor – amōrem "love") have not survived. Some of the same nouns with multiple stems in Old French or Old Occitan have come down in Italian in the nominative rather than the accusative (e.g. uomo "man" < homō, moglie "wife" < mulier), suggesting that a similar system existed in pre-literary Italian.

The modern situation in Sursilvan (one of the Rhaeto-Romance languages) is unique in that the original nominative/oblique distinction has been reinterpreted as a predicative/attributive distinction:^[7]: 381

• il hotel ej vɛɲiws natsionalizaws "the hotel has been nationalized"
• il hotel natsionalizaw "the nationalized hotel"

Personal pronouns

Unlike in English, a separate neuter personal pronoun ("it") generally does not exist, but the third-person singular and plural both distinguish masculine from feminine. Also, as described above, case is marked on pronouns even though it is not usually on nouns, similar to English. As in English, there are forms for nominative case (subject pronouns), oblique case (object pronouns), and genitive case (possessive pronouns); in addition, third-person pronouns distinguish accusative and dative. There is also an additional set of possessive determiners, distinct from the genitive case of the personal pronoun; this corresponds to the English difference between "my, your" and "mine, yours".

Articles

Latin had no articles as such. The closest definite article was the non-specific demonstrative is, ea, id meaning approximately "this/that/the". The closest indefinite articles were the indefinite determiners aliquī, aliqua, aliquod "some (non-specific)" and certus "a certain".

Romance languages have both indefinite and definite articles, but none of the above words form the basis for either of these. Usually the definite article is derived from the Latin demonstrative ille ("that"), but some languages (e.g. Sardinian, Old Occitan, and Balearic Catalan) have forms from ipse (emphatic, as in "I myself"). The indefinite article everywhere is derived from the number ūnus ("one").

Some languages, e.g. French and Italian, have a partitive article that approximately translates as "some". This is used either with mass nouns or with plural nouns—both cases where the indefinite article cannot occur. A partitive article is used (and in French, required) whenever a bare noun refers to specific (but unspecified or unknown) quantity of the noun, but not when a bare noun refers to a class in general. For example, the partitive would be used in both of the following sentences:

• I want milk.
• Men arrived today.

But neither of these:

• Milk is good for you.
• I hate men.

The sentence "Men arrived today", however, (presumably) means "some specific men arrived today" rather than "men, as a general class, arrived today" (which would mean that there were no men before today). On the other hand, "I hate men" does mean "I hate men, as a general class" rather than "I hate some specific men".

As in many other cases, French has developed the farthest from Latin in its use of articles. In French, nearly all nouns, singular and plural, must be accompanied by an article (either indefinite, definite, or partitive) or demonstrative pronoun.

Due to pervasive sound changes in French, most nouns are pronounced identically in the singular and plural, and there is often heavy homophony between nouns and identically pronounced words of other classes. For example, all of the following are pronounced /sɛ̃/: sain "healthy"; saint "saint, holy"; sein "breast"; ceins "(you) tie around, gird"; ceint "(he) ties around, girds"; ceint "tied around, girded"; and the equivalent noun and adjective plural forms sains, saints, seins, ceints. The article helps identify the noun forms saint or sein, and distinguish singular from plural; likewise, the mandatory subject of verbs helps identify the verb ceint. In more conservative Romance languages, neither articles nor subject pronouns are necessary, since all of the above words are pronounced differently. In Italian, for example, the equivalents are sano, santo, seno, cingi, cinge, cinto, sani, santi, seni, cinti, where all vowels and consonants are pronounced as written, and ⟨s⟩ /s/ and ⟨c⟩ /tʃ/ are clearly distinct from each other.

Latin, at least originally, had a three-way distinction among demonstrative pronouns distinguished by distal value: hic 'this', iste 'that (near you)', ille 'that (over there)', similar to the distinction that used to exist in English as "this" vs. "that" vs. "yon(der)". In urban Latin of Rome, iste came to have a specifically derogatory meaning, but this innovation apparently did not reach the provinces and is not reflected in the modern Romance languages. A number of these languages still have such a three-way distinction, although hic has been lost and the other pronouns have shifted somewhat in meaning. For example, Spanish has este "this" vs. ese "that (near you)" vs. aquel (fem. aquella) "that (over yonder)". The Spanish pronouns derive, respectively, from Latin iste ipse accu-ille, where accu- is an emphatic prefix derived from eccum "behold (it!)" (still vigorous in Italy as Ecco! 'Behold!'), possibly with influence from atque "and".^[12]

Reinforced demonstratives such as accu-ille arose as ille came to be used as an article as well as a demonstrative. Such forms were often created even when not strictly needed to distinguish otherwise ambiguous forms. Italian, for example, has both questo "this" (eccu-istum) and quello "that" (eccu-illum), in addition to dialectal codesto "that (near you)" (*eccu-tē-istum). French generally prefers forms derived from bare ecce "behold", as in the pronoun ce "this one/that one" (earlier ço, from ecce-hoc; cf. Italian ciò 'that') and the determiner ce/cet "this/that" (earlier cest, from ecce-istum).

Reinforced forms are likewise common in locative adverbs (words such as English here and there), based on related Latin forms such as hic "this" vs. hīc "here", hāc "this way", and ille "that" vs. illīc "there", illāc "that way". Here again French prefers bare ecce while Spanish and Italian prefer eccum (French ici "here" vs. Spanish aquí, Italian qui). In western languages such as Spanish, Portuguese and Catalan, doublets and triplets arose such as Portuguese aqui, acá, cá "(to) here" (accu-hīc, accu-hāc, eccu-hāc). From these, a prefix a- was extracted, from which forms like aí "there (near you)" (a-(i)bi) and ali "there (over yonder)" (a-(i)llīc) were created; compare Catalan neuter pronouns açò (acce-hoc) "this", això (a-(i)psum-hoc) "that (near you)", allò (a-(i)llum-hoc) "that (yonder)".

Subsequent changes often reduced the number of demonstrative distinctions. Standard Italian, for example, has only a two-way distinction "this" vs. "that", as in English, with second-person and third-person demonstratives combined. In Catalan, however, a former three-way distinction aquest, aqueix, aquell has been reduced differently, with first-person and second-person demonstratives combined. Hence aquest means either "this" or "that (near you)"; on the phone, aquest is used to refer both to speaker and addressee.

Old French had a similar distinction to Italian (cist/cest vs. cil/cel), both of which could function as either adjectives or pronouns. Modern French, however, has no distinction between "this" and "that": ce/cet, cette < cest, ceste is only an adjective, and celui, celle < cel lui, celle is only a pronoun, and both forms indifferently mean either "this" or "that". (The distinction between "this" and "that" can be made, if necessary, by adding the suffixes -ci "here" or -là "there", e.g. cette femme-ci "this woman" vs. cette femme-là "that woman", but this is rarely done except when specifically necessary to distinguish two entities from each other.)

Loanwords

Romance languages have borrowed heavily, though mostly from other Romance languages. However, some, such as Spanish, Portuguese, Romanian, and French, have borrowed heavily from other language groups. Vulgar Latin borrowed first from indigenous languages of the Roman empire, and during the Germanic folk movements, from Germanic languages, especially Gothic; for Eastern Romance languages, during Bulgarian Empires, from Slavic languages, especially Bulgarian. Notable examples are *blancus "white", replacing native albus (but Romansh alv, Dalmatian jualb, Romanian alb); *guerra "war", replacing native bellum; and the words for the cardinal directions, where cognates of English "north", "south", "east" and "west" replaced the native words septentriō, merīdiēs (also "noon; midday nap"; cf. Romanian meriză), oriens, and occidens. (See History of French – The Franks.) Some Celtic words were incorporated into the core vocabulary, partly for words with no Latin equivalent (betulla "birch", camisia "shirt", cerevisia "beer"), but in some cases replacing Latin vocabulary (gladius "sword", replacing ensis; cambiāre "to exchange", replacing mūtāre except in Romanian and Portuguese; carrus "cart", replacing currus; pettia "piece", largely displacing pars (later resurrected) and eliminating frustum). Many Greek loans also entered the lexicon, e.g. spatha "sword" (Greek: σπάθη spáthē,^[17][18] replacing gladius which shifted to "iris", cf. French épée, Spanish espada, Italian spada and Romanian spată); cara "face" (Greek: κάρα kára,^[19] partly replacing faciēs); colpe "blow" (Greek: κόλαφος kólaphos,^[20][21] replacing ictus, cf. Spanish golpe, French coup); cata "each" (Greek: κατά katá,^[22] replacing quisque); common suffixes *-ijāre/-izāre (Greek: -ίζειν -izein,^[23] French -oyer/-iser, Spanish -ear/-izar, Italian -eggiare/-izzare, etc.), -ista (Greek: -ιστής -istes^[24]).^[25]

Lexical innovation

Many basic nouns and verbs, especially those that were short or had irregular morphology, were replaced by longer derived forms with regular morphology. Nouns, and sometimes adjectives, were often replaced by diminutives, e.g. auris "ear" > auricula (orig. "outer ear") > oricla (Sardinian origra, Italian orecchia/o, Portuguese orelha, etc.); avis "bird" > avicellus (orig. "chick, nestling") > aucellu (Occitan aucèl, Friulian ucel, Neapolitan auciello, etc.); caput "head" > capitium (Portuguese cabeça, Spanish cabeza, French chevet "headboard"; but reflexes of caput were retained also, sometimes without change of meaning, as in Italian capo "head", alongside testa); vetus "old" > vetulus > veclus (Dalmatian vieklo, Italian vecchio, Portuguese velho, etc.). Sometimes augmentative constructions were used instead: piscis "fish" > Old French peis > peisson (orig. "big fish") > French poisson. Verbs were often replaced by frequentative constructions: canere "to sing" > cantāre; iacere "to throw" > iactāre > *iectāre (Italian gettare, Portuguese jeitar, Spanish echar, etc.); iuvāre > adiūtāre (Italian aiutare, Spanish ayudar, French aider, etc., meaning "help", alongside e.g. iuvāre > Italian giovare "to be of use"); vēnārī "hunt" (Romanian vâna, Aromanian avinari) > replaced by *captiāre "to hunt", frequentative of capere "to seize" (Italian cacciare, Portuguese caçar, Romansh catschar, French chasser, etc.).

Many Classical Latin words became archaic or poetic and were replaced by more colloquial terms: equus "horse" > caballus (orig. "nag") (but equa "mare" remains, cf. Spanish yegua, Portuguese égua, Sardinian ebba, Romanian iapă); domus "house" > casa (orig. "hut"); ignis "fire" > focus (orig. "hearth"); strāta "street" > rūga (orig. "furrow") or callis (orig. "footpath") (but strāta is continued in Italian strada). In some cases, terms from common occupations became generalized: invenīre "to find" replaced by afflāre (orig. "to sniff out", in hunting, cf. Spanish hallar, Portuguese achar, Romansh dial. anflar, Southern Italian asciare, acchiare, Romanian afla 'to find out'); advenīre "to arrive" gave way to plicāre (orig. "to fold (sails; tents)", cf. Spanish llegar, Portuguese chegar; Romanian pleca), elsewhere arripāre (orig. "to harbor at a riverbank", cf. Italian arrivare, French arriver) (advenīre is continued with the meaning "to achieve, manage to do" as in Middle French aveindre, or "to happen" in Italian avvenire) . The same thing sometimes happened to religious terms, due to the pervasive influence of Christianity: loquī "to speak" succumbed to parabolāre (orig. "to tell parables", cf. Occitan parlar, French parler, Italian parlare) or fabulārī ~ fābellāre (orig. "to tell stories", cf. Spanish hablar, Dalmatian favlur, Sardinian faeddare), based on Jesus' way of speaking in parables.

Many prepositions were used as verbal particles to make new roots and verb stems, e.g. Italian estrarre, Aromanian astragu, astradziri "to extract" from Latin ex- "out of" and trahere "to pull" (Italian trarre "draw, pull", Aromanian tragu, tradziri), or to augment already existing words, e.g. French coudre, Italian cucire, Portuguese coser "to sew", from cōnsuere "to sew up", from suere "to sew", with total loss of the bare stem. Many prepositions and commonly became compounded, e.g. de ex > French dès "as of", ab ante > Italian avanti "forward". Some words derived from phrases, e.g. Portuguese agora, Spanish ahora "now" < hāc hōrā "at this hour"; French avec "with" (prep.) < Old French avuec (adv.) < apud hoc "with that"; Spanish tamaño, Portuguese tamanho "size" < tam magnum "so big"; Italian codesto "this, that" (near you) < Old Italian cotevesto < eccum tibi istum approx. "here's that thing of yours"; Portuguese você "you"^[26] < vosmecê < vossemecê < Galician-Portuguese vossa mercee "your mercy".^[27]

A number of common Latin words that have disappeared in many or most Romance languages have survived either in the periphery or in remote corners (especially Sardinia and Romania), or as secondary terms, sometimes differing in meaning. For example, Latin caseum "cheese" in the periphery (Portuguese queijo, Spanish queso, Romansh caschiel, Sardinian càsu, Romanian caş), but in the central areas has been replaced by formāticum, originally "moulded (cheese)" (French fromage, Occitan/Catalan formatge, Italian formaggio, with, however, cacio also available in much of Italy; similarly (com)edere "to eat (up)", which survives as Spanish/Portuguese comer but elsewhere is replaced by mandūcāre, originally "to chew" (French manger, Sardinian mandicare alongside pappare, Romanian mânca(re)). In some cases, one language happens to preserve a word displaced elsewhere, e.g. Italian ogni "each, every" < omnes, displaced elsewhere by tōtum, originally "whole" or by a reflex of Greek κατά (e.g. Italian ognuno, Catalan tothom "everyone"; Italian ogni giorno, Spanish cada día "every day"); Friulian vaî "to cry" < flere "to weep"; Serbo-Croatian (Dubrovnik) otijemna "sail pole" < Dalmatian < antenna "yardarm". Sardinian in particular preserves many words entirely lost elsewhere, e.g. eja "yes" < etiam "also/yes/indeed", emmo "yes" < immo "rather/yes/no", mannu "big" < magnus, nàrrere "to say" < narrāre "to tell", and domo "house" < (abl.) domō "at home". Sardinian preserves some words that were already archaic in Classical Latin, e.g. àchina "grape" < acinam, while Sicily and Calabria typically have forms with initial /r/: ràcina.

Latinisms

During the Middle Ages, scores of words were borrowed directly from Classical Latin (so-called Latinisms), either in their original form (learned loans) or in a somewhat nativized form (semi-learned loans). These resulted in many doublets—pairs of inherited and learned words—such as those in the table below:

Sometimes triplets arise: Latin articulus "joint" > Portuguese artículo "joint, knuckle" (learned), artigo "article" (semi-learned), artelho "ankle" (inherited; archaic and dialectal). In many cases, the learned word simply displaced the original popular word: e.g. Spanish crudo "crude, raw" (Old Spanish cruo); French légume "vegetable" (Old French leüm); Portuguese flor "flower" (Galician-Portuguese chor). The learned loan always sounds (and, in writing, looks) more like the original than the inherited word does, because regular sound change has been bypassed; likewise, the learned word usually has a meaning closer to that of the original. In French, the stress of the modern form of the learned loan may be on the "wrong" syllable vis-à-vis Latin, whereas the stress of the inherited word always corresponds to the Latin stress: e.g. Latin vipera (stress on /i/) vs. French vipère, learned loan, and guivre/vouivre, inherited.

Borrowing from Classical Latin has produced a large number of suffix doublets. Examples from Spanish (learned form first): -ción vs. -zon; -cia vs. -za; -ificar vs. -iguar; -izar vs. -ear; -mento vs. -miento; -tud (< nominative -tūdō) vs. -dumbre (< accusative -tūdine); -ículo vs. -ejo; etc. Similar examples can be found in all the other Romance languages.

This borrowing also introduced large numbers of classical prefixes in their original form (dis-, ex-, post-, trans-) and reinforced many others (re-, popular Spanish/Portuguese des- < dis-, popular French dé- < dis-, popular Italian s- < ex-). Many Greek prefixes and suffixes (hellenisms) also found their way into the lexicon: tele-, poli-/poly-, meta-, pseudo-, -scope/scopo, -logie/logia/logía, etc.