Word order typology classifies languages by their dominant arrangement of subject (S), object (O), and verb (V) in basic declarative sentences. SOV (Japanese, Turkish, Hindi) and SVO (English, Mandarin, Swahili) together account for roughly 85% of the world's languages, with VSO (Irish, Arabic, Tagalog) as a significant minority pattern. Greenberg's implicational universals revealed that basic word order correlates with other structural properties: verb-final languages tend to have postpositions, genitive-noun order, and clause-final subordinators, while verb-initial and verb-medial languages show the mirror pattern. The head-directionality parameter captures this cluster of correlations — head-initial languages consistently place heads before complements across phrase types, while head-final languages do the reverse.
Examine parallel translations of the same sentences across SOV, SVO, and VSO languages, marking the position of the verb, its arguments, and adpositions. Test Greenberg's universals against a typological sample — check whether an SOV language also has postpositions, OV order in noun phrases, and clause-final complementizers. Analyze a language with flexible word order (like Russian or Warlpiri) to understand how morphological case can liberate syntax from rigid positional constraints.
You already know from syntactic structure that sentences have hierarchical constituent organization, and from linguistic typology that languages vary systematically in their structural properties. Word order typology asks: across the world's languages, how do they arrange the core arguments of a sentence, and what other structural properties cluster with that choice? The answer turns out to be far more constrained than chance would predict.
The six logically possible orderings of subject (S), object (O), and verb (V) are not equally distributed. SOV languages — Japanese, Turkish, Hindi, Korean, Amharic — are the single most common type, accounting for roughly 45% of all languages. SVO — English, Mandarin, French, Swahili — comes second at around 40%. VSO — Classical Arabic, Irish, Welsh, Tagalog — is a significant minority. The remaining three orders (VOS, OVS, OVS) are rare, occurring in a handful of Amazonian languages. This distribution reflects something real about how languages are structured, not random drift.
The deeper insight comes from Greenberg's implicational universals: word order in the S-O-V domain predicts word order in other phrase types. SOV languages — verb at the end — tend to place the head of any phrase at the end. They have postpositions (*Tokyo ni*, "in Tokyo") rather than prepositions, genitives before nouns (*John's book* → *John no hon* in Japanese), relative clauses before the noun they modify, and subordinate clauses before the main clause. SVO languages tend toward the mirror pattern: prepositions, noun-genitive order, postnominal relatives, and main clause before subordinate. This clustering is captured by the head-directionality parameter: a head-initial language places the head of a phrase before its complement across all phrase types; a head-final language reverses this systematically.
What about languages like Russian or Warlpiri with apparently "free" word order? These are not counterexamples — they are a different solution to the same problem. English uses positional encoding to mark who is doing what to whom: "The cat chased the dog" versus "The dog chased the cat." Swap the nouns, change the meaning. Russian uses morphological case — suffixes on the nouns themselves mark subject and object, so word order is freed up for pragmatic work: marking topic, focus, information structure, and emphasis. Warlpiri uses a similar strategy but even more radically. The information about grammatical relations is encoded in the morphology; the syntax is freed from positional constraints entirely. Recognizing that "free word order" is really "pragmatically flexible order with morphological encoding" completes the picture: all languages must mark who does what to whom — they differ only in the mechanism they use to do it.