The Romance languages derive, via Latin, from the Italic branch of Indo-European. Their modern distribution is the product of two major phases of conquest and colonisation. The first, between c. 240 BC and c. AD 100 brought the whole Mediterranean basin under Roman control; the second, beginning in the sixteenth century, annexed the greater part of the Americas and sub-Saharan Africa to Romance-speaking European powers. Today, some 665 million people speak, as their first or only language, one that is genetically related to Latin. Although for historical and cultural reasons preeminence is usually accorded to European Romance, it must not be forgotten that European speakers are now outnumbered by non-Europeans by a factor of nearly three to one. The principal modern varieties of European Romance are indicated on Map 8.1. Nouniformly acceptable nomenclature has been devised for Romance and the choice of term to designate a particular variety can often be politically charged. The Romance area is not exceptional in according or withholding the status of ‘language’ (in contradistinction to ‘dialect’ or ‘patois’) on sociopolitical rather than linguistic criteria, but additional relevant factors in Romance may be cultural allegiance and length of literary tradition. Five national standard languages are recognised: Portuguese, Spanish, French, Italian and Rumanian (each treated in an individual chapter below). ‘Language’ status is usually also accorded on cultural/literary grounds to Catalan and Occitan, though most of their speakers are bilingual in Spanish and French respectively, and the ‘literary tradition’ of Occitan refers primarily to medieval Provençal, whose modern manifestation is properly considered a constituent dialect of Occitan. On linguistic grounds, Sardinian too is often described as a language, despite its internal heterogeneity. Purely linguistic criteria are difficult to apply systematically: Sicilian, which shares many features with southern Italian dialects, is not usually classed as an independent language, though its linguistic distance from standard Italian is no less than that separating Spanish from Portuguese. ‘Rhaeto-Romance’ is nowadays used as a cover term for a number of varieties spoken in southern Switzerland (principally Engadinish, Romansh and Surselvan) and in the Dolomites, but it is no longer taken to subsume Friulian. Romansh (local form romontsch) enjoys an official status for cantonal administration and so perhaps fulfils the requirements of a language. Another special case is Galician, located in Spainbut genetically and typologically very close to Portuguese; in the wake of political autonomy, galego/gallego now enjoys protected ‘language’ status in Spain, although elsewhere it continues to be thought of (erroneously) as a regional dialect of Spanish. Corsican, which clearly belongs to the Italo-Romance group, would be in a similar position if the separatist movement gained autonomy or independence from France. Outside Europe, Spanish, Portuguese and French, in descending order of native speakers,have achieved widest currency, though many other varieties are represented in localised immigrant communities, such as Sicilian in New York, Rumanian in Melbourne, Sephardic Spanish in Seattle and Buenos Aires. In addition, the colonial era gave rise to a number of creoles, of which those with lexical affinities to French are now the most vigorous, claiming some ten million speakers. In general, European variants are designated by their geographical location; ‘Latin’,as a term for the vernacular, has survived only for some subvarieties of Rhaeto-Romance (ladin) and for Biblical translations into Judaeo-Spanish (ladino). ‘Romance’ derives, through Spanish and French, from ROMĀNICĒ ‘in the Roman fashion’ but also ‘candidly, straightforwardly’, a sense well attested in early Spanish. The terminological distinction may reflect early awareness of register differentiation within the language, with ‘Latin’ reserved at first for formal styles and later for written language and Christian liturgy. The idea, once widely accepted, that Latin and Romance coexisted for centuries as natural, spoken languages, is now considered implausible. Among the chief concerns of Romance linguists have always been: the unity orotherwise of the proto-language, the causes and date of dialect differentiation and the classification of the modern variants. Plainly, Romance does not derive from the polished literary models of Classical Latin. Alternative attestations are quite plentiful, but difficultin may be stylistic artifice; inscriptional evidence is formulaic; the abundant Pompeian graffiti may be dialectal, and so on. Little is known of Roman linguistic policy or of the rate of assimilation of new conquests. We may however surmise that a vast territory, populated by widely different ethnic groups, annexed over a period exceeding three centuries, conquered by legionaries and first colonised by settlers who were probably not native speakers of Latin, and never enjoying easy or mass communications, could scarcely have possessed a single homogeneous language. The social conditions which must have accompanied latinisation – including slaveryand enforced population movements – have led some linguists to postulate a stage of creolisation, from which Latin slowly decreolised towards a spoken norm in the regions most exposed to metropolitan influences. Subsequent differentiation would then be due to the loss of administrative cohesion at the break-up of the Empire and the slow emergence of local centres of prestige whose innovations, whether internal or induced by adstrate languages, were largely resisted by neighbouring territories. Awareness of the extent of differentiation seems to have come very slowly, probably stimulated in the west by Carolingian reforms of the liturgical language, which sought to achieve a uniform pronunciation of Church Latin at the cost of rendering it incomprehensible to uneducated churchgoers. Sporadic attestations of Romance, mainly glosses and interlinear translations in religious and legal documents, begin in the eighth century. The earliest continuous texts which are indisputably Romance are dated: for French, ninth century; for Spanish and Italian, tenth; for Sardinian, eleventh; for Occitan (Provençal), Portuguese and Rhaeto-Romance, twelfth; for Catalan, thirteenth; for Dalmatian (now extinct), fourteenth; and for Rumanian, well into the sixteenth century. Most classifications of Romance give precedence, explicitly or implicitly, to histor-ical and areal factors. The traditional ‘first split’ is between East and West, located on a line running across northern Italy between La Spezia and Rimini. Varieties to the northwest are often portrayed as innovating, versus the conservative south-east. For instance, West Romance voices and weakens intervocalic plosives: SAPŌNE ‘soap’ > Ptg. sabão, Sp. jabón, Fr. savon, but It./Sard. sapone, Rum. sa˘pun; RŌTA ‘wheel’ > Ptg. roda, Sp. rueda, Cat. roda, Fr. roue, but It./Sard. rota, Rum. roata˘; URTĪCA ‘nettle’> Ptg./Sp./ Cat. ortiga, Fr. ortie, but Sard. urtica, It. ortica, Rum. urzica˘. The West also generalises /-s/ as a plural marker, while the East uses vocalic alternations: Ptg. as cabras ‘the goats’, Cat. les cabres, Romansh las chavras, contrast with It. le capre and Rum. caprele. In vocabulary, we could cite the verb ‘to weep’, where the older Latin word PLANGE˘RE survives in the East (Sard. pranghere, It. piangere, Rum. a plînge) but is completely replaced in the West by reflexes of PLORĀRE (Ptg. chorar, Sp. llorar, Cat. plorar, Oc. plourà, Fr. pleurer). In this classification, each major group splits into two subgroups: ‘East’ into Balkan-Romance and Italo-Romance, ‘West’ into Gallo-Romance and IberoRomance. The result is not entirely satisfactory. While, for example, Arumanian dialects and Istro-Rumanian group quite well with Balkan-Romance, our scant evidence of Dalmatian suggests it shared as many features with Italo-Romance as with the Balkan group. Catalan is a notorious difficulty, having been subject for centuries to alternating Occitan and Spanish influences. The unity of ‘Rhaeto-Romance’ also fails to survive closer scrutiny: Ladin groups fairly well with Friulian as part of Italo-Romance, but southern Swiss varieties share many features with eastern French dialects. ‘Family-tree’ classifications, in which variants are each assigned to a single node,give only a crude indication of relationships in Romance and tend to obscure theLatin and patterns of contact. This is readily illustrated from the lexicon. The PLANGE˘RE/ PLORĀRE example, though supportive of the East-West split, is in fact rather atypical. More common are innovations spreading from central areas but failing to reach the periphery. ‘To boil’ is Ptg. ferver, Sp. hervir, Rum. a fierbe (< FERVĒRE/FERVE˘RE), but Cat. bullir, Oc. boulí, Fr. boullir, It. bollire (< BULLĪRE, originally ‘to bubble’); ‘to request’ is Ptg./Sp. rogar, Rum. a ruga (< ROGĀRE), but Cat. pregar, Oc. pregá, Fr. prier, It. pregare (< PRECĀRE, originally ‘to pray’); ‘to find’ is Ptg. achar, Sp. hallar, Rum. a afla, but Cat. trobar, Oc. trobà, Fr. trouver, It. trovare (both forms are metaphorical – classical INVENĪRE and REPERĪRE do not survive). Among nouns, we may cite ‘bird’: Ptg. pássaro, Sp. pájaro, Rum. pasa˘re (< *PASSARE), versus Oc. aucèu, Fr. oiseau, Romansh utschè, It. uccello (< AUCELLU); and ‘cheese’: Ptg. queijo, Sp. queso, Rum. cas¸ (< CĀSEU), versus Cat. formatge, Oc. froumage, Fr. fromage, It. formaggio (< [CASEU] FORMATICU ‘moulded [cheese]’). Almost the same distribution is found in a morphosyntactic innovation: the Latin synthetic comparative in -IŌRE nowhere survives as a productive form, but peripheral areas have MAGIS as the analytic replacement (‘higher’ is Ptg. mais alto, Rum. mai înalt) whereas the centre prefers PLŪS (Fr. plus haut, It. più alto). Despite this differential diffusion and the divergences created by localised borrowingfrom adstrate languages (notably from Arabic into Portuguese and Spanish, from Germanic into northern French, from Slavonic into Rumanian), the modern Romance languages have a high degree of lexical overlap. Cognacy is about 40 per cent for all major variants using the standard lexicostatistical 100-word list. For some language pairs it is much higher: 65 per cent for French-Spanish (slightly higher if suffixal derivation is disregarded), 90 per cent for Spanish-Portuguese. This is not, of course, a guarantee of mutual comprehensibility (untrained observers are unlikely to recognise the historical relationship of Sp. /oxa/ ‘leaf’ to Fr. /fœj/), but a high rate of cognacy does increase the chances of correct identification of phonological correspondences. Intercomprehensibility is also good in technical and formal registers, owing to extensive borrowing from Latin, whether of ready-made lexemes (abstract nouns are a favoured category) or of roots recombined in the naming of a new concept, like Fr. constitutionnel, émetteur, exportation, ventilateur, etc. Indirectly, coinings like these have fed the existing propensity of all Romance languages for enriching their word stock by suffixal derivation. Turning to morphosyntax, we find that all modern Romance is VO in its basic wordorder, though southern varieties generally admit some flexibility of subject position. A much reduced suffixal case system survives in Rumanian, but has been eliminated everywhere else, with internominal relations now expressed exclusively by prepositions. All variants have developed articles, the definite ones deriving overwhelmingly from the demonstrative ILLE/ILLA (though Sardinian and Balearic Catalan use IPSE/ IPSA), the indefinite from the numeral ŪNU/ŪNA. Articles, which precede their head noun everywhere except in Rumanian where they are enclitic, are often obligatory in subject position. Concord continues to operate throughout noun phrases and between subject and verb, though its range of exponents has diminished with the loss of nominal case. French is eccentric in virtually confining plural marking to the determiner, though substantives still show number in the written language. Parallel to the definite articles, most varieties have developed deictic object pronouns from demonstratives. These, like the personal pronouns, often occur in two sets, one free and capable of taking stress, the other cliticised to the verb. There is some evidence of the grammaticalisation of anclitic and in the marking of specific animate objects. This latter is widespread (using a in West Romance and pe in Rumanian) but not found in standard French or Italian. Suffixal inflection remains vigorous in the common verb paradigms everywhere butin French. Compound tense forms everywhere supplement the basic set, though the auxiliaries vary: for perfectives, HABĒRE is most common: ‘I have sung’ is Fr. j’ai chanté, It. ho cantato, Sp. he cantado but Ptg. tenho cantado (< TENĒRE originally ‘to hold’) and Cat. vaig cantar (< VĀDO CANTĀRE) – an eccentric outcome for a combination that would be interpreted elsewhere as a periphrastic future (‘I am going to sing’). Most Romance varieties have a basic imperfective/perfective aspectual opposition, supplemented by one or more of punctual, progressive and stative. The synthetic passive has given way to a historically reflexive medio-passive which coexists uneasily with a reconstituted analytic passive based on the copula and past participle. The replacement of the Latin future indicative by a periphrasis expressing volition or mild obligation (HABĒRE is again the most widespread auxiliary, but deppo ‘I ought’ is found in Sardinian and voi ‘I wish’ in Rumanian) provided the model for two new synthetic paradigms, the future itself and the conditional, which has taken over a number of functions from the subjunctive. The subjunctive has also been affected by changes in complementation patterns, but a few new uses have evolved during the documented period of Romance, and its morphological structure, though drastically reduced in spoken French, remains largely intact. In phonology, it is more difficult to make generalisations (see the individual languagesections below and, for the development from Latin to Proto-Romance, pages 146150). We can, however, detect some shared tendencies. The rhythmic structure is predominantly syllable-timed. Stress is dynamic rather than tonal and, on the whole, rather weak – certainly more so than in Germanic; some variants, notably Italian, do use higher tones as a concomitant of intensity, but none rely on melody alone. The loss of many intertonic and post-tonic syllables suggests that stress may previously have been stronger, witness IŪDI˘CE[’i˘u-di-ke] ‘judge’ > Ptg. juiz, Sp. juez, Cat. jutge, Fr. juge; CUBI˘TU [’ku-bi-tu] ‘elbow’ > Sp. codo, Fr. coude, Rum. cot. The elimination of phonemic length from the Latin vowel system has been maintained with only minor exceptions. A strong tendency in early Romance towards diphthongisation of stressed mid vowels has given very varied results, depending on whether both higher and lower mid vowels were affected, in both open and closed syllables, and on whether the diphthong was later levelled. Romance now exhibits a wide range of vowel systems, but those of the south-central group are noticeably simpler than those of the periphery: phonemic nasals are found only in French and Portuguese, high central vowels only in Rumanian, and phonemic front rounded vowels only in French, some Rhaeto-Romance and north Italian varieties and São Miguel Portuguese. Among consonantal developments, we have already mentioned lenition, which led to wholesale reduction and syllable loss in northern French dialects. Latin geminates generally survive only in Italo-Romance, and many other medial clusters are simplified (though new ones are created by various vocalic changes). Although Latin is in Indo-European terms a centum language (with k for PIE k), one of the earliest and most far-reaching Romance changes is the palatalisation, and later affrication, of velar and dental consonants before front vowels. Only the most conservative dialect of Sardinian fails to palatalise (witness kenapura ‘Holy supper = Friday’), and the process itself has elsewhere often proved cyclic.