Typology

Linguistic typology is contrasted with genealogical linguistics on the grounds that typology groups languages or their grammatical features based on formal similarities rather than historic descendence. The issue of genealogical relation is however relevant to typology because modern data sets aim to be representative and unbiased. Samples are collected evenly from different language families and emphasize the importance of lesser-known languages in gaining insight into human language.

Speculations of the existence of a (logical) general or universal grammar underlying all languages were published in the Middle Ages, especially by the Modistae school. At the time, Latin was the model language of linguistics although transcribing Irish and Icelandic into the Latin alphabet was found to be problematic. The cross-linguistic dimension of linguistics was established during the Renaissance. For example, Grammaticae quadrilinguis partitiones (1544) by Johannes Drosaeus compared French and the three "holy languages:" Hebrew, Greek, and Latin. The approach was expanded by the Port-Royal Grammar (1660) of Antoine Arnauld and Claude Lancelot, who added Spanish, Italian, German, and Arabic. Nicolas Beauzée's 1767 book includes examples of English, Swedish, Lappish, Irish, Welsh, Basque, Quechua, and Chinese. The conquest and the conversion of the world by Europe gave rise to "missionary linguistics," which produced first-hand word lists and grammatical descriptions of exotic languages. That work is accounted for in the "Catalogue of the Languages of the Populations We Know" (1800) by the Spanish Jesuit Lorenzo Hervás. Johann Christoph Adelung collected the first large language sample with the Lord's Prayer in almost 500 languages (posthumous, 1817). More developed 19th-century comparative works include Franz Bopp's 'Conjugation System' (1816) and Wilhelm von Humboldt's "On the Difference in Human Linguistic Structure and Its Influence on the Intellectual Development of Mankind" (posthumous, 1836). In 1818, August Wilhelm Schlegel made a classification of the world's languages that had three types: (i) languages lacking grammatical structure like Chinese; (ii) agglutinative languages like Turkish; and (iii) inflectional languages, which can be synthetic like Latin and Ancient Greek, or analytic like French. The idea was later developed by others including August Schleicher, Heymann Steinthal, Franz Misteli, Franz Nicolaus Finck, and Max Müller. The word "typology" was proposed by Georg von der Gabelentz in his Sprachwissenschaft (1891). Louis Hjelmslev proposed typology as a large-scale empirical-analytical endeavor of comparing grammatical features to uncover the essence of language. Such a project began from the 1961 conference on language universals at Dobbs Ferry. Speakers included Roman Jakobson, Charles F. Hockett, and Joseph Greenberg, who proposed 45 different types of linguistic universals, based on his data sets from 30 languages. Greenberg's findings were mostly known from the 19th-century grammarians, but his systematic presentation of them would serve as a model for modern typology. Winfred P. Lehmann introduced Greenbergian typological theory to Indo-European studies in the 1970s. During the 20th century, typology based on missionary linguistics became centered on SIL International, which today hosts its catalogue of living languages, Ethnologue, as an online database. The Greenbergian, or universalist, approach is accounted for by the World Atlas of Language Structures, among others. Typology is also done within the frameworks of functional grammar including functional discourse grammar, role and reference grammar, and systemic functional linguistics. During the early years of the 21st century, however, the existence of linguistic universals became questioned by linguists proposing evolutionary typology.

The below table indicates the distribution of the dominant word order patterns of over 5,000 individual languages and 366 language families. SOV is the most common type in both of them although that is much clearer in the data of language families, including isolates. "NODOM" represents languages without a single dominant order. • Type — Languages — % — Families — % • SOV — 2,275 — 43.3% — 239 — 65.3% • SVO — 2,117 — 40.3% — 55 — 15% • VSO — 503 — 9.5% — 27 — 7.4% • VOS — 174 — 3.3% — 15 — 4.1% • NODOM — 124 — 2.3% — 26 — 7.1% • OVS — 40 — 0.7% — 3 — 0.8% • OSV — 19 — 0.3% — 1 — 0.3% Though the reason of dominance is sometimes considered an unsolved or unsolvable typological problem, several explanations for the distribution pattern have been proposed. Evolutionary explanations include those by Thomas Givon (1979), who suggests that all languages stem from an SOV language but are evolving into different kinds, and by Derek Bickerton (1981), who argues that the original language was SVO, which supports simpler grammar employing word order, instead of case markers to differentiate between clausal roles. Universalist explanations include a model by Russell Tomlin (1986) based on three functional principles: (i) animate before inanimate, (ii) theme before commentm, and (iii) verb-object bonding. The three-way model roughly predicts the real hierarchy (see table above) assuming no statistical difference between SOV and SVO, and no statistical difference between VOS and OVS. By contrast, the processing efficiency theory of John A. Hawkins (1994) suggests that constituents are ordered from shortest to longest in VO languages and from longest to shortest in OV languages, which give rise to the attested distribution. His approach relies on the notion that OV languages have heavy subjects, and VO languages have heavy objects, which is disputed.

A second major way of syntactic categorization is by excluding the subject from consideration. It is a well-documented typological feature that languages with a dominant OV order (object before verb), Japanese for example, tend to have postpositions. In contrast, VO languages (verb before object) like English tend to have prepositions as their main type of adpositions. Several OV/VO correlations have been uncovered. • Correlation — VO languages — Example (VO) — OV languages — Example (OV) • Adposition type — prepositions — of..., than..., on... — postpositions —...of,...than,...on • Order of noun and genitive — noun before genitive — father + of John — genitive before noun — John-of + father • Order of adjective and standard of comparison — adjective before standard — taller + than Bob — standard before adjective — Bob-than + taller • Order of verb and adpositional phrase — verb before adpositional phrase — slept + on the floor — adpositional phrase before verb — floor-on + slept • Order of verb and manner adverb — verb before manner adverb — ran + slowly — manner adverb before verb — slowly + ran • Order of copula and predicative — copula before predicate — is + a teacher — predicate before copula — a teacher + is • Order of auxiliary verb and content verb — auxiliary before content verb — want + to see Mary — content verb before auxiliary — Mary see-to + want • Place of adverbial subordinator in clause — clause-initial subordinators — because + Bob has left — clause-final subordinators — Bob has left-because • Order of noun and relative clause — noun before relative clause — movies + that we saw — either — we-seen + movies or movies we-seen

Several processing explanations were proposed in the 1980s and the 1990s for the above correlations. They suggest that the brain finds it easier to parse syntactic patterns that are either right- or left-branching but not when it is mixed. The most widely-held such explanation is John A. Hawkins's parsing efficiency theory, which argues that language is a non-innate adaptation to innate cognitive mechanisms. Typological tendencies are considered as being based on language users' preference for grammars that are organized efficiently and on their avoidance of word orderings that cause processing difficulty. Hawkins's processing theory predicts the above table but also makes predictions for non-correlation pairs, including the order of adjective, demonstrative and numeral in respect with the noun. The theory was based on corpus research and lacks support in psycholinguistic studies. Some languages exhibit regular "inefficient" patterning, which include the VO languages Chinese, which has the adpositional phrase before the verb, and Finnish, which has postpositions. However, there are few other profoundly exceptional languages. It is suggested more recently that the left-right orientation is limited to role-marking connectives (adpositions and subordinators), stemming directly from the semantic mapping of the sentence. Since the true correlation pairs in the above table either involve such a connective or, arguably, follow from the canonical order, orientation predicts them without making problematic claims.

Another common classification distinguishes nominative–accusative and ergative–absolutive alignment patterns. In a language with grammatical cases, the classification depends on whether the subject (S) of an intransitive verb has the same case as the agent (A) or the patient (P) of a transitive verb. If a language has no cases but has as its word order AVP or PVA, a classification may reflect whether the subject of an intransitive verb appears on the same side as the agent or the patient of the transitive verb. Bickel (2011) has argued that alignment should be seen as a construction-specific property, rather than a language-specific property. Many languages show mixed accusative and ergative behaviour (such as ergative morphology marking the verb arguments on top of an accusative syntax). Other languages, called "active languages," have two types of intransitive verbs; some of them ("active verbs") join the subject in the same case as the agent of a transitive verb, and the rest ("stative verbs") join the subject in the same case as the patient. Still other languages behave ergatively only in some contexts (the "split ergativity" is often based on the grammatical person of the arguments or on the tense/aspect of the verb). For example, only some verbs in Georgian behave that way and, as a rule, do so only in the perfective aspect.

Linguistic typology also seeks to identify patterns in the structure and distribution of sound systems among the world's languages. That is accomplished by surveying and analyzing the relative frequencies of different phonological properties. Sample relative frequencies are given below for certain speech sounds formed by obstructing airflow (obstruents). The relative frequencies show that contrastive voice commonly occurs with plosives, such as the English "neat" and "need," but occurs much more rarely among fricatives, such as the English "niece" and "knees." According to a worldwide sample of 637 languages, 62% of them contrast voicing in stops but only 35% do so in fricatives. In the vast majority of those cases, the absence of voicing contrast occurs because there is a lack of voiced fricatives and because all languages have some form of plosive (occlusive), but some languages have no fricatives. This chart shows the breakdown of voicing properties among languages in the aforementioned sample: • Total • 117 — 218 — 395 (62%) • 44 — 198 — 242 (38%) • Total — 221 (35%) — 416 (65%) — 637 Languages worldwide also vary in the number of sounds that they use. Languages can have very small phonemic inventories (Rotokas, with six consonants and five vowels), or very large inventories (!Xóõ, with 128 consonants and 28 vowels). An interesting phonological observation that was found by the data is that the larger the consonant inventory of a language, the more likely it has a sound from a defined set of complex consonants (clicks, glottalized consonants, doubly articulated labial-velar stops, lateral fricatives and affricates, uvular and pharyngeal consonants, and dental or alveolar non-sibilant fricatives). Of that list, only about 38% of languages in a survey of over 600 with small inventories (less than 19 consonants) contain a member of the set, and 32% of average languages (19-25) contain at least one member and 38% of large consonant inventories (more than 25 consonants) contain a member of this set. It is then seen that complex consonants are in proportion to the size of the inventory. Vowels contain a more modest number of phonemes, with the average being 5–6, which 51% of the languages in the survey have. About a third of the languages have larger than average vowel inventories. More interesting, though, is the lack of relationship between consonant inventory size and vowel inventory size. Here is a chart showing the lack of predictability between consonant and vowel inventory sizes in relation to each other: • Small — Average — Large — Total • Small — 47 — 153 — 65 — 265 (39%) • Average — 34 — 105 — 98 — 237 (35%) • Large — 34 — 87 — 57 — 178 (26%) • Total — 115 (17%) — 345 (51%) — 220 (32%) — 680