Explicitness of the anaphor

An anaphoric expression may have different degrees of explicitness, varying between a maximally explicit pole occupied by a lexically specified NP, and a maximally implicit pole, occupied by zero. This variation obtains in relative constructions, too. On the one hand, the head nominal may be repeated in the relative clause. shows this for the preposed relative clause; shows it for the postposed relative clause.

It is no coincidence that both examples feature the correlative diptych: lexical representation of the head in both clauses is largely restricted to the adjoined relative construction.

While repetition of the head nominal is the minority strategy in the correlative diptych, the default anaphoric strategy is illustrated by and .

The distribution of the head noun and the resumptive pronoun in the two variants of the correlative construction follows the normal rules of anaphora: While the order of main clause and relative clause is opposite in the plain and the inverse correlative diptych, the order of appearance of the head noun and the pronoun remains the same in both variants: the noun in the initial clause, the pronoun in the final clause (with an empty head in the particular case of ).

The strategy which represents the antecedent by an anaphoric pronoun in the relative clause is typical of the postnominal type. It is illustrated by . Although this particular example features, at the same time, an extremely remote syntactic function of the relativized position, which would require some overt representative of the relativized item in just any language which can relative on this position, one may generalize that representation of the antecedent by a personal pronoun in the postnominal clause is quite usual even for hierarchically higher syntactic functions, including the direct object.

On the other hand, the relative clause may just display an empty place which may be interpreted as a zero anaphor of the referent of the complex nominal. shows an empty place in the direct object position of a prenominal relative clause.

The remaining positional type is the circumnominal relative clause. Here the question of anaphora or empty-place formation does not arise, since the head occupies the relativized position.

Given the series of – , it gets clear that explicitness of the anaphor marking the relativized position correlates with the degree of desententialization of the relative clause: If it is adjoined, it differs minimally from an independent sentence. In this context, the rules of normal anaphora apply, with the consequence that the anaphor may be maximally explicit and by default is a pronoun. If the relative clause is embedded, it is but a subconstituent of an NP. In this context, only a highly grammaticalized form of anaphora is possible, coded by a weak pronoun or even by zero.

Empty-place formation and indices

The resumptive pronoun illustrated by above has a subtype which is the representation of the nucleus by a bound index. The example series ff illustrates the main relativization strategy of Tojolabal.

Data from: José Gómez Cruz, Cláusulas relativas en tojol-ab'al. [Unpublished handout. San Cristóbal de las Casa: CIEAS] 21-05-2016; with emendations.

The series illustrates the relativization of the set of syntactic functions from the intransitive () and transitive () absolutive via the ergative (), indirect object (), nominal (“possessive”) attribute () down to the comitative () and causal () adjunct.¹ Down to this position, the strategy is very uniform (local and temporal adverbial clauses, however, involve an interrogative pro-adverb): Relativizing into absolutive function involves an asyndetic relative clause. Then there is a subordinator (grammaticalized from a demonstrative pronoun) which is optional for the ergative function and obligatory from there downwards. What is presently of importance, however, is the representation of the head in the relative clause: it is consistently achieved by pronominal indices. This includes the paradigm indicating absolutive function (B), except that the third person is zero.

Since these indices are bound, they do not appear in the same syntagmatic position that an NP – instantiated by a lexical NP or a free pronoun – would occupy. This is why this strategy is regarded, in certain descriptions, as a gap strategy. However, a gap in a linguistic construction is recognized in comparison with another construction which bears a paradigmatic relation to the former and has an overt expression in the same place. Now the question is which other construction is chosen for this paradigmatic comparison.

Certain descriptive models (alias “linguistic theories”) postulate a structure of simple verbal clause in which all complement positions are filled by lexical NPs. This is a clause type that practically does not occur in texts of many languages, and much less in the same text position occupied by the relative clauses above. Thus, such descriptive models postulate a gap in these cases not on account of some alternative construction of the same language where the gap is filled, but on account of some fiction.

A natural methodological decision would seem to be to compare the relative clause with an independent declarative clause which paraphrases the relative clause and would be able to take its place without affecting the natural flow of the text. In Tojolabal as in many other languages, such a clause would have the same structure as the relative clauses in – : the indices are the device provided by the language system in order to anaphorically resume an NP of the preceding sentence. If this is the construction with which the relative construction is paradigmatically related – or in a dynamic view: from which it is derived – then none of the relative clauses of the above sequence contains a gap. Instead, syntactic positions interpreted as being occupied by some term given in the immediate context are signaled by an indexing strategy motivated and established independently of relative clause formation. In other words, empty place formation in relative constructions of such languages employs no specific strategy.

Needless to say, the case of the gap strategies as they occur in English or Mandarin and are discussed in the previous section is different.

Anaphora, empty-place formation and orientation

In the prenominal relative construction, there is an additional reason why rules of normal anaphora are without force: The relative order of the relativized position and the head nominal reverses the canonical order of antecedent and anaphor. If an anaporic relation obtained between the relativized position and the head, it would have to be cataphora. However, then there would have to appear a cataphoric demonstrative in the relativized position. This actually never happens. This strongly indicates that the relationship between the relativized position and the head of a prenominal relative construction is not an anaphoric one.

If the relative clause is not finite, it may be oriented towards its open position as an adjectival is oriented towards its head. Its modifying slot is then directly occupied by the head itself; in other words, no anaphora obtains. This is the case in , where the relative participial is oriented towards the indirect object position.

Up to now, it would appear that the two prenominal relative clauses of and represent an essential difference concerning the strategy which relates the relative clause to its head: would provide an empty place which, being some kind of zero anaphor, is occupied by the nucleus through some quasi-anaphoric operation, while the relative participial of would be oriented towards the open position, which, being a modificative slot, would be directly occupied by the head. Whether or not it is appropriate to postulate such a categorical difference between the two strategies may be safely left to linguistic theory. Here it suffices to see that the anaphoric relation between the head and the relativized position shrinks to a direct syntactic relation by grammaticalization. This is the same process which applies in the marking of syntactic functions of dependent NPs by indices on the verb. is a schematic illustration of what has happened in the change from standard to colloquial French.

In .a from standard French, the initial noun phrase is left-dislocated, thus not a constituent of the subsequent clause, and the resumptive pronoun il appearing in the latter is the subject of the verb and bears a normal anaphoric relation to the left-dislocated NP. In #b from colloquial French, the construction has been grammaticalized. The clause boundary between the initial constituent and the rest disappears, Jean becomes the subject of the verb, and il is reduced to a cross-reference index on the verb. While here one may want a theory by which the subject NP occupies the subject argument position of the verb only by mediation of the index, such a contraption becomes unnecessary in the version #c (of standard and colloquial French), where the subject NP doubtless occupies the verbal argument position without any mediation.

Coming back to relative clause formation and to and in particular, we see that there is an intimate relation between anaphora, empty-place formation and orientation. Ultimately, all three of these operations serve the function of helping some referent occupy some open position.

¹ The same preposition as in is used for instruments.