Constituent order

From Pin Eight
Jump to: navigation, search

In linguistic typology, languages are classified by the constituent order or word order of the subject (or agent), verb, and object (or patient) in a transitive statement. SVO and SOV are most common, followed by VSO, with VOS, OSV, and OVS rare.

Attested orders

In several languages, breaking the verb V into an auxiliary that carries tense and agreement (the "tensed" verb, labeled T) and the non-finite main verb (the "predicate" or "participle", labeled P) can help clarify some ordering phenomena that the SVO system conflates. In sentences with no auxiliary, the main verb occupies T.

  • English is STPO: "Sam has eaten oranges." Questions invert into [W]TSPO, and simple tenses of verbs other than be gain do-support in questions rather than eliding P: "Has Sam eaten oranges?" Intransitive sentences with an adverb may allow moving the whole verb before the subject: "There goes the neighborhood"; "All of a sudden was seen a thundering herd of angry elephants." [1]
  • Modern Romance languages are SoTPO, where pronoun objects (lowercase o) go on the opposite side of the verb compared to noun objects. The TP compound doesn't split as easily as it does in English, and thus questions invert into [W]oTPSO: "Has eaten Sam oranges?"
  • Breton is TSPO in progressive aspect but PTSO in its do-supported simple present:[2] "Is Sam eating oranges" but "Eat does Sam oranges."
  • Biblical Hebrew is traditionally analyzed as VSO. A later analysis proposes that SVO is more common outside narratives, in which a conjunction fuses with the verb and pulls it to the front.[3]
  • Welsh is TSPO: "Has Sam eaten oranges." The standard order for simple tenses in Irish and Welsh is VSO, but do-support has become common in colloquial Welsh.[2] Celtic languages form the participle P by attaching a tense preposition to a gerund (called a verbnoun by Celticists), literally "in eating" or "after eating". Glenda Newton has a theory on how this arose in Old Irish.[4]
  • Yosondúa Mixtec is TPSO: Has eaten Sam oranges. Its strict VSO structure, which keeps even serial verb constructions to the left of the subject and allows only adverbs to precede the verb, is in fact far more common worldwide than the TSPO of Celtic, which appears to derive from SVO.[5]
  • Some languages are topic-prominent in their syntax, distinguishing "given" noun phrases (those providing topic or background information) from "new" noun phrases (those making a comment). Classical Arabic has been described as VGN (verb-given-new), where an object that is given ordinarily precedes the subject. When a subject is both given and "heavy" (long), the sentence tends to flip into SVO, with inna "indeed" often preceding the subject to mark it as a topic.[2]
  • Dutch, German, and Kashmiri are V2 (verb second) in main clauses, where an adverb can push the subject after the finite verb. This results in STOP/AdvTSOP: "Sam has oranges eaten" but "Yesterday has Sam oranges eaten." Rarely the main verb is fronted into PTSO:[6] "Eaten has Sam oranges." "Scrambling" in Wikipedia gives the German example Erwähnt hat er das nicht (lit. "Mentioned has he that not") for "He didn't mention that".
  • Japanese is SOPT: "Sam-ga oranges-o eaten has." So are subordinate clauses in most V2 languages.
  • Though Chinese is generally SVO, it also uses serial verb constructions that can be seen as prepositional phrases preceding the verb. This gives Mandarin an optional SOV order, repurposing a word meaning "hold" as a preposition marking the direct object: 我把苹果吃了 Wǒ bǎ píngguǒ chīle (lit. I HOLD-apple eat-PAST) "I ate the apple."

In fiction:

  • Yoda's dialect from the Star Wars films is usually POST: "Eaten oranges Sam has." Occasionally it's POTS, especially for a short S:[7] "Eaten oranges has Sam." Without the distinction between tensed and main verbs, it could be analyzed as VOS, OVS, or OSV. Not all Yoda quotes are in this order, as for very important points, Yoda makes an effort to use the STPO order of standard Galactic Basic.
  • Among the languages of Noen, Nognese is TSPO, and Noeneg is even more V2 than German, with auxiliary support in all tenses. One Noeneg-based interlanguage is PTSO, which Noeneg speakers understand as P fronted past T and Nognese speakers understand as T forming the conjugation of P.

Be careful with the notation used here, as some combinations have other meanings:


Not all 24 orderings of STPO are equally likely in languages. Several tendencies are common, for which linguists have given cognitive functional explanations of varying quality.[8]

  • Subject before object reflects the tendency of subjects to be topics.
  • As the head of the entire sentence, T is commonly found next to one or both of its immediate dependents S and P.
  • P and O are commonly found together, forming a verb phrase (VP), a constituent with P as the head.
  • Orderings with S and O adjacent are more likely to have overt case marking on nouns. In fact, languages that lose noun case tend to put at least part of the verb between the subject and object.[9]
  • A language tends toward one head directionality (either head-dependent or dependent-head) for most constructions.

In an SOV language where the verb agrees with the subject in person and number, the agreement has to jump over the direct object to land at the verb. In addition, the mapping of each noun phrase's grammatical case to its thematic role is not certain until the verb. This can prove challenging for adult L2 learners, encouraging them to switch to SVO, as Latin speakers did when the language gradually became Italian and Old English speakers did under the Norman regime that produced Middle English. The fact that Chinese, English, Spanish, and other SVO languages have over twice as many speakers as SOV languages is evidence for a tendency of widely spoken languages to switch to VO order.[10]

The "brace" orders (TSPO and STOP) reflect a compromise between keeping the verb phrase contiguous and keeping head directionality consistent. When explaining the absence of T?SVO (V in situ whether or not an initial finite auxiliary is present, or SVO/TSPO) and the rarity of brace orders outside Celtic and Germanic, Alexandra Zepter of Rutgers has set forth six violable constraints in an optimality theory that results in a small number of underlying word orders.[5]

A language that tends toward the majority of clauses having one argument may end up reanalyzed as verb-peripheral. One researcher proposes this as part of how Celtic became VSO.[11]

There even appears to be a correlation with phonology. Languages with few phonemes show simpler syllable structure (often mostly CV), with less information per syllable. This leads to more agglutinative morphology, more overt case, and somewhat more tendency to remain head-final (SOV).[12] Languages with more phonemes tend toward more consonant clusters and closed syllables. Many phonemes per syllable mean fewer syllables per word, which means it's easier for the next syllable to be an independent word, producing a tendency toward more isolating morphology. This increases homophony and word class (part of speech) ambiguity, which in turn produces more rigid syntax and more opaque idioms and set phrases, and this usually moves the verb before the object (VO) to separate it from the subject.[13]

Another phonologically motivated phenomenon is Wackernagel's law. Languages, at least Indo-European ones, tend to put something unstressed in a main clause's second position. In Latin, certain connective sentence adverbs such as enim "for" occupy this spot. In German, Old English, and other V2 languages, this is the finite verb, particularly when auxiliary, leaving STOP as a sometimes brief stop on the way to SVO of Modern English. In early Celtic, this was a set of verb prefixes detachable from the verb stem, including those that modify the meaning and the object pronoun.[14]

Luke Maurits measured the word order of the language of thought by the relative time needed to recall the action, agent, and patient of an event. Test subjects recalled the agent most quickly and the action least quickly, leading to a conclusion of a subject–object–verb language of thought (SOVLOT).[8] This finding, however, runs up against findings that the mental pathways for prelinguistic communication and for phrase structure processing behave differently in their ordering. This is borne out by appearance of OV order in home sign, the rudimentary sign languages created by deaf children of hearing parents, even though creole languages are overwhelmingly SVO. Hearing speakers of English, Chinese, Spanish, and Italian (all SVO) ended up using the same sort of SOV structures seen in home sign when challenged to communicate with only gestures, failing to transfer their native language structures to sign. Both Italian speakers and speakers of Turkish (SOV) ended up signing a main clause before the subordinate clause and both in SOV order. Speakers of both languages also understood OV signed utterances faster than VO and VO speech synthesis faster than OV.[15]

Order within a phrase

Comparing the order of constituents across head-final, weakly head-initial, or strongly head-initial languages can reveal other insights as to universal tendencies. When order consistently reverses between OV and VO languages or between AN and NA languages, this shows that particular aspects of a thing or an event are more tightly or more loosely bound in speakers' minds to the thing or event. OV languages, for example, show adverbs in the order time-manner-place-verb, whereas VO languages have them in the opposite order: verb-place-manner-time. Yet having demonstratives and numerals out of harmony with head directionality is common across languages, as these modifiers are "light" and "mobile" compared to other modifiers of the noun (namely adjectives, genitives, and relative clauses).[16]

Modern Israeli Hebrew and other Semitic languages are strongly head-initial, with SVO or VSO clauses, NA noun phrases, and gerund-N-N deverbal phrases. Definite NPs are the N the A, and definite construct genitives are N the A. Only the adverbs meaning "very", "more", and "less" can appear between the and A. "Most", "all", and cardinal numbers at least two behave as nouns that precede a genitive NP to make a phrase meaning a set of that many: Card N (indefinite) or Card the N (definite). By contrast, cardinal "one" is an adjective; "one" as a noun means "one from [some collection]". Numbers at least second that follow the noun are adjectives with ordinal meaning. As in English, subset size binds tighter than the subset's position in a larger set: "first two pages" translates to Hebrew words meaning "two pages first". (Arabic number words are ordered differently.) Hebrew adjectives also show more or less the mirror image of the canonical English adjective order. Deverbals can take one of two orders: V S ACC O for an action or V O of S for the result.[17]


  1. Raymond Hickey. "Internal and external forces again: changes inword order in Old English and Old Irish". Language Sciences 24 (2002) 261–283. Accessed 2020-06-01. Cites Stockwell (1984, p. 583)
  2. 2.0 2.1 2.2 Steve Hewitt. "Arabic: verb-subject-object or verb-given-new? Implications for word order typology". Conference on Communication and Information Structure in Spoken Arabic, 2006. Accessed 2013-11-11.
  3. Robert Holmstedt. Basic Word Order in the Biblical Hebrew Verbal Clause, Part 3. 2011-05-16. Accessed 2012-06-16.
  4. In French and other VO languages with rich verb agreement, the verb moves up to T (the tensed verb position), which comes before adverbs that precede the object. Old Irish took this a step further as a conjunction was eroded to a main clause marker, causing fronting the verb to become grammaticalized first as a way to connect clauses into a narrative, similarly to the Biblical Hebrew theory, and then as the default ordering. Glenda Newton. "The development of head movement: The rise of verb-initial word order in Old Irish". Proceedings of ConSOLE XIV, 2007. Accessed 2016-08-13.
  5. 5.0 5.1 Alexandra Zepter. Phrase Structure Directionality: Having a Few Choices. New Brunswick: Rutgers, 2003-10. Accessed 2019-01-22.
  6. Andrew C. Wetta. "A Construction-based Cross-linguistic Analysis of V2 Word Order". Accessed 2014-02-04. Via jlovegren. "Answer to German is SOV: should it not have been 'Ich ein Berliner bin'?". Linguistics Stack Exchange, 2011-11-21. Accessed 2014-02-04.
  7. Mark Liberman. "Unclear of Yoda's syntax the principles are, if any". 2005-05-20. Accessed 2015-11-02.
  8. 8.0 8.1 Luke Maurits. Representation, information theory and basic word order. University of Adelaide, 2011-09. Accessed 2018-08-14.
  9. Hickey, 2002. Cites Vennemann, 1974.
  10. Christian Benz and Morten H. Christiansen. "Linguistic adaptation at work? The change of word order and case system from Latin to the Romance Languages". 2010-04-04. Accessed 2019-11-22.
  11. Hickey, 2002. Cites a manuscript by Disterheft and a multidimensional definition of transitivity by Hopper and Thompson, 1980.
  12. Gertraud Fenk-Oczlon and August Fenk. "Cognition, quantitative linguistics, and systemic typology". Linguistic Typology 3(2):151-178, 1999-01. DOI: 10.1515/lity.1999.3.2.151
  13. Gertraud Fenk-Oczlon and August Fenk. "Complexity trade-offs between the subsystems of language". Language complexity: typology, contact, change. John Benjamins, 2008-01. DOI: 10.1075/slcs.94.05fen
  14. Hickey, 2002. Cites Wackernagel, 1892, p. 342); Lass, 1994, p. 226–8; Harris and Campbell, 1995, 215ff.; Watkins, 1963.
  15. Alan Langus, Marina Despor. "Cognitive systems struggling for word order". Cognitive Psychology 60 (2010) 291–318. Accessed 2020-06-01.
  16. Hickey, 2002. Cites John A. Hawkins. Word Order Universals. New York: Academic Press, 1983.
  17. Ur Shlonsky. "The form of Semitic noun phrases". Lingua, 2003-09-02. doi:10.1016/j.lingua.2003.09.019. Accessed 2020-04-13.

External links