How does Japanese word order obviate the need for relative pronouns?

Question

According to the Wikipedia page on Japanese grammar:

Head finality in Japanese sentence structure carries over to the building of sentences using other sentences. In sentences that have other sentences as constituents, the subordinated sentences (relative clauses, for example), always precede what they refer to, since they are modifiers and what they modify has the syntactic status of phrasal head. Translating the phrase the man who was walking down the street into Japanese, word order would be street down was walking man. (Note that Japanese has no articles, and the different word order obviates any need for the relative pronoun who.)

Why does the different word order obviate the need for a relative pronoun?

Answer 1

In my opinion, since Japanese is a head-final language and since the verb comes at the end of a clause, a sentence-final noun would be enough to indicate that the clause is a pre-modifier of the noun.

In English, the relative clause post-modifies its head noun and, even though a relative pronoun can be optional at times, there are certain clauses that require its use. For example, subordinate participial and subordinate passive clauses may be reduced. Relative pronouns are also optional in relative clauses in which the modified noun is coindexed with the object. The contiguous noun phrases could serve as an indicator of there being a relative clause, hence the optionality of the relative pronoun. However, subordinate clauses in the simple tenses require the use of a relative pronoun, if the noun being modified is coindexed with the subject of the relative clause, to explicitly indicate that what follows the noun is a post-modifying relative clause, not a simplex main clause in which the noun is the subject.

Present participle

The cow (that is) eating the grass

Passive

The grass (that is) eaten by the cow

Clause where head noun is coindexed with object of RC

[The grass]¹ (that) the cow eats ∅¹

[The grass]¹ (that) the cow ate ∅¹

Clause where head noun is coindexed with subject of RC

[The cow]¹ *(that) ∅¹ eats the grass

[The cow]¹ *(that) ∅¹ ate the grass

The reason that subordinate clauses in the simple tenses where the head noun is coindexed with the subject of the relative clause require a relative pronoun is that without it, the head noun could potentially be parsed as being part of a simplex clause, not the head noun of a relative clause if the relative pronoun is omitted. Participial and passive clauses, regardless of coindexation, can be reduced due to the fact that native speakers intuitively know that there is no way a participial or passive clause can be a simplex clause without its auxiliary 'be', so there is no other alternative but to compute the clause as being a relative clause.

The grass eaten by the cow (never a simplex clause, so it must be a complex NP)

The cow eating the grass (never a simplex clause, so it must be a complex NP)

In Japanese, on the other hand, whether the head noun is coindexed with the subject or the object does not matter and there is no need for relative pronouns as there is no structural ambiguity whatsoever, since nouns never come after the verb in simplex clauses; when a noun does come after the verb, the clause preceding the noun is most certainly a relative clause.

牛 が ∅¹ 食べ-た 草¹

ushi ga ∅¹ tabe-ta kusa¹

Cow NOM ∅¹ eat-PAST grass¹

The grass that the cow ate

∅¹ 草 を 食べ-た 牛¹

∅¹ kusa wo tabe-ta ushi¹

∅¹ Grass ACC eat-PAST cow¹

The cow that ate the grass

This is, however, just my opinion. Whether there is any psycholinguistic evidence for such parsing complexities is not known by me.

Answer 2

The word "need" here is a bit strong. Imbabura Quechua is head-final like Japanese, but verbs in relative clauses still have special nominalization marking (wambra-ca runa-ta ricu-rca-mi "The child saw the man" vs. wambra ricu-shca runa "the man who the child saw", runa = "man"). Special morphology on the verb of the relative clause is not "needed" and wambra ricu-rca runa would be perfectly interpretable, except that it's a fact of Quechua (and not Japanese) that relative clauses have special marking.

There is a split in Bantu between languages which require an initial head-marking complementizer for subject relative clauses, and those that don't. In those that don't there is the potential for ambiguity between "a man who bought a cow" and "a man bought a cow" – but ambiguity is avoided because (using Shona as an example) the tone of subject agreements in the two clause types tells you whether you have a relative clause vs. a main clause (thus munhu akáténgá n'ómbe "a man who bought a cow" vs munhu ákaténgá n'ómbe). Kerewe is a language that requires special head-marking on the verb which in fact differs between subject relatives and non-subject relatives: and furthermore subject relatives and object relatives have different tone patterns, thus a-yábwééne "he who saw" vs. owó yabwééné "whom he saw" (head markers are in bold). Distinct tone patterns are not needed when there are distinct head-marking prefixes; distinct head-marking prefixes are not needed when tone patterns distinguish subject vs. object relative clauses.

I would say that "need" is not an informative concept for understanding how relative clauses are structured. The direction that one should be looking is for languages that systematically do not mark relative clauses in any manner, and where there is no way to distinguish between "the man bought the cow" and "the man who bought the cow". If there are none, and this isn't just an attestation gap, then perhaps a case for "need" can be made.

Answer 3

It's a view that I've heard before, and never questioned; but I think you're right to question it.

As Snailboat points out in a comment, with a participial clause we don't need a relative pronoun in English: The man walking down the street - and the clause follows the head even though English is generally head-final. In German, an equivalent participial phrase can precede its head, and the result is very like the Japanese.

I think the difference in Japanese is that there isn't a distinct participial form that is required in this construction, but an ordinary finite verb.