Theoretical background of the presented data

Full overview of the theory concerning adjective valence and the structure of the adjective valence database is to be found in a printed volume which is an outcome of this project together with the online version. The book will be published in 2017 with the title Adjective valence in the ancient languages, ed. Dita Frantikova, Baar-Verlag, Hamburg.

The database contains adjective valence examples in selected ancient languages. Besides the general information on the language and selected bibliography, it contains examples of every possible paradigmatic slot of adjective valence in the language. Where the example was not found, but is possible, it is labelled “not found”. Where it is structurally impossible, it is labelled “no data” (e.g. when the language does not use a selected case, e.g. ergative).

In languages where no valence structures are found, the inevitable reason is the size of the corpus. Every language with coherent texts, even of a limited volume, such as Old Persian, has proved to contain syntactic structures of adjective valence. Where the texts are restricted to funeral inscriptions, votive texts or inscriptions on vessels, such structures may be missing. As the adjective valence expands the basic clause structure, the fact that it is missing in short inscriptions is not striking and certainly should not be understood as a proof of such structures missing in the language altogether. Languages with minor corpora without adjective valence attestations are included only in case their corpus was carefully examined and allows for such conclusion.

Adjectives – attributive vs predicative position

Adjectives are found either in the attributive or predicative position. The attributive modification as described by Sproat and Shih (1988, for Chinese) can be direct or indirect. The direct attributive modification is achieved by a simple attributive position and the indirect by forming an additional structure layered on top of a predicative modifier, such as a relative clause. Attributive modification can also be achieved by transponing the predicative adjective into attributive position. In such cases, the valence of the attributive adjective does not differ in type or dependent from that in the predicative position[1]. For the sake of preventing an overlap of data, in the AdjectiveValenceDatabase, we confine only to predicative adjectives.

In the set of languages treated in our database, all can use adjectives in both attributive and predicative position. We are aware that in the world´s languages, this is not the rule: as per Morzicky (2014:65), some languages seem to lack either attributive or predicative adjectives entirely. Examples may be found in Baker (2003:206-207). An example of a predicative-only language is Slave (an Athabaskan language, Baker 2003:194, citing Rice; besides also langauges Ika and Edo, see Dixon 2004), and an example of an attributive-only language is Vata (Niger-Congo language; Baker 2003 apud Koopman 1984).

Valence in linguistic theory

The term valence was introduced to modern linguistics by L. Tesnière in his posthumous “Eléments de syntaxe structural”  , where he compares verbs to molecules and discusses the relationship between the subject, direct object and indirect object; his overall focus is on verbs. Being the author of the term, Tesnière was not a pioneer neither of the comparisons of language to biological structure, nor of the valence concept itself. The comparison of language structures to biology was introduced in print a year earlier, in Hockett’s discussion of surface and deep grammar (1958:249). With American structuralist tradition, following Bloomfield (1933), clause structures with the verb as the nucleus have already for some time been in focus of linguists and the concept of valence, regardless of the terminology, was being explored (e.g. Harris 1957; Erben 1972). The research of valence later stretched from verbs to non-verbal predicates – nouns and adjectives (Chomsky 1970:190 pp.; Herbst 1983, Sommerfeld 1971, 1977). Although adjectives are most frequently found in the attributive position (generally non-valence position; see p. XX), they are also used as parts of the predicate where they can fulfil the role of a head of a valence construction. Sommerfeld (1971:115) refers to adjectives that take more than one relational constant as “relative adjectives”.

The valence of a lexical item is understood as its inherent relationality that allows it to govern a number of arguments (Haspelmath 1991 [2004:1130]). The configuration of arguments that a lexical item (or a predicate) can govern is understood as its valence pattern.

Valence of an item can be defined semantically, syntactically and grammatically (The well-known description of semantic and syntactic distinction is Payne 1997:169pp.). From the semantic point of view, we understand that valence must be semantically motivated, as it is the semantic intention of the adjective, which projects on a surface level as a syntactic valence (in other terms, as argument structure) in a clause structure (Pauliny 1943:16, for verbs; Kupść 2009b:241 for adjectives). The definition of semantic roles is known for arguments of a verbal predicate (e.g. Aarts 1997:87). These can be partially applied to other predicates – adjectives and nouns[2]. In the following sentence, the adjective concerning has a compulsory argument with the role of a patient:

“They discussed problems concerning finance.”

Other arguments can take roles of locative (“published in London”), benefactive (“prepared for children”) etc. The agent position is expressed by the governing noun (the adjective subject) and thus not found in the valence frame of adjectives.

As valence is coded syntactically, it can be described in morphological (or grammatical) terms, which is the approach taken in this work. One can e.g. say that the dependent in the phrase “concerning finance” takes the accusative case. The argument of the predicate can be defined as inherent (obligatory) or facultative (optional). The distinction is a scale rather than a clear-cut division (e.g. Kupść 2009b:241). Even a syntactically obligatory argument can be omitted for semantic or pragmatic reasons, as well. In such case, we are dealing with so called general argument (discussed in PDT 2.0, part 4.1), which is deleted in the surface structure. General arguments are obvious when the adjective is used in both the attributive and the predicative position, e.g.:

“a baking woman” (attributive, no argument on the surface level)

“a woman baking a cake” (predicative, obligatory accusative argument – patient)

On the other hand, general argument needs to be distinguished from the optional argument which may not necessarily be present.

The presence or omission of an argument may also influence the contextual meaning of the adjective, e.g.

“Two soldiers were in prone position.”

“The two soldiers were prone to grumbling.”

In this example, the argument is obligatory for the semantics “inclined to” while it is optional for the semantics “straight, horizontal”.

Use of adjuncts helps to specify the adjectiveʼs meaning, e.g. “enough” in the following example:

“He was stupid enough to go sailing in January.”

Adjuncts must be distinguished from the arguments proper (see chapter Adjective valence arguments).

On the surface, the semantic valence of adjectives may have more realizations than that of verbs; it can often be expressed more than one way:

“It is right to bow before the older folks.”

“It is right that we bow before the older folks.”

Valence is found mainly (though not exclusively) with adjectives functioning as copula complements. In the sentence

“The suitable candidates were offered promising positions.”

the adjective suitable has no valence relations other than that to its head noun candidates. However, when used in predicate position, it has an accusative argument valence pattern, as shown in the following example:

“This programme is suitable for the elderly.”

Adjective valence is therefore in absolute terms less frequent than the valence of verbs, as most adjectives are used as NPM (e.g. in the modern Czech corpus, they number 75% of occurrences, see Kopřivová 2006:2. The ratio must differ across languages, as does the ratio of transitive versus intransitive use of verbs).

On contrary to verbs, adjectives do not change form or class with addition of a valence exponent.

Distinguishing adjective valence from other structures

Valence argument of an adjective must be distinguished from other parts of a clause which may resemble prototypical arguments. The need for a distinction, together with rise of disputable issues, arises in case of collocations, of use of intensifying adverbs, of adjective grading and others[3].

Collocations

In corpus linguistics, the terminology distinction between the term collocation and valence proves as important. Collocation is a statistically-oriented notion used for description of the frequency of co-occurrence of lexical units which exceeds the statistical chance.[4]

To define valence, multiple criteria must be implied: not only the criterion of frequency of occurrence but also of their dependence relation.

Intensifying adverbs

Intensifying adverbs as specificators of the adjective in an adjective phrase may head the (following or preceding) part of the phrase. In such cases, the phrase is not a valence argument of the adjective. A plausible criterion is the one based on grammaticality – whether the omission of the adverb changes the leaves the clause ungrammatical.

This makes the infinitive phrase “to be true” a dependent on the adverb rather than an argument of the adjective.

Adjective grading

The four basic types of degree phrases are equative (such as “as tall as”) and comparative (“bigger than”) standards of comparison, high degree adverbs (such as “very”), and measure phrases (“six feet”). As stated by Baker (2003:218), adjectives as degree heads have not been studied much cross linguistically, as they may not be very frequent in non-Indo-European languages (Baker lists American-Indian languages which employ adverbial expressions in instances of nominal grading).

Bogal-Allbritten (2003:23) in their description of grading in Navajo claim that “a second direction of research is the treatment of a degree phrase as an argument vs. an adjunct to the adjective. The present proposal for the mechanism by which degree phrases modify AA/PA-marked builds on Schwarzschild’s recent (2010, 2011) work on this issue. Interest in this issue can also be traced back to Schwarzschild (2005), where measure phrases are analysed as adjuncts rather than arguments of the adjective. Navajo is, so far, unique in clearly marking degree phrases as adverbial when they modify syntactically intransitive adjectives.”

Adjective grading used in comparative degree phrases is one of the syntactic construction types which interfere with the inherent valence properties of an adjective (Kupść 2009b:242). Comparative constructions are generally considered non-arguments (Kupść 2009a:204). For a phrase to be understood as a case of adjective valence, one of the criteria is that it is unique to the adjective. In comparatives, the only semantic condition is that the adjective is qualifying and that it contains certain “measure for comparison”, which is in turn a condition for grading. Also, it should be possible to use the valence argument in all degrees (in the positive as well as in the comparative).

The status of particles used for comparison (English “then”, French “que”) is not that of a preposition nor of a conjuction – they are understood as homophonous words, used for a specific purpose. As shown by a following table, the grade of the adjective is another feature that characterizes an adjective valence phrase, not its inherent characteristic:

Table 1. Role of adjective grading

In case of true adjective valence, the argument connects to the adjective regardless of its grade:

“he is similar to his sister”

“he is more similar to his sister than me”

“he is the most similar to his sister of the whole family”

Here, the phrase “similar to sb.” is a case of adjective valence of the adjective “similar”.  Not so the phrase “more similar than me” which is considered an adjunct rather that an argument of “similar” as it is can be used with a generic adjective, such as “higher than me”, “nicer than me”.

Adjective valence arguments

The classification of adjective dependents is portrayed in the following table. They can be divided by their syntactic type to noun (substantive, adjective) phrases, adpositional phrases, other word forms and subordinate clauses.

Table 2. Adjective valence arguments

[2] The terminology of transitivity, used primarily for verbs, is also used for adjective predicates. One may say that an adjective is intransitive or transitive, denoting whether it is capable of binding with an argument. The distinction may be understood on semantic or syntactic ground and each approach yields different result. Bogal-Allbritten (2013:1) presents data from Navajo which show that although all adjectives in the language are semantically transitive, morphology restricts the syntactic transitivity possibilities of some of them. While adjective transitivity is a feature worthy of research, it is not discussed here. Argument structure of an adjective is far more loose than that of verbs and deciding about the complete valence possibilities of an adjective brings about many “shadow zones”. Besides that, the issue is not important for overall adjective valence description – adjectives need not be classed the same way as verbs as they pertain their unique semantic properties (see the chapter Semantic types and valence).

[3] Another situation when it is disputable whether on is dealing with an adjective valence is in case of restructuring operations (Kupść 2009b:253). By restructuring operations here are not meant restructuring predicates (Wurmbrand 2003:7) but shift of a subject to an argument-like position, e.g.

„The freshness of the tea was astonishing“ versus

*“The tea was astonishing because of its freshness.“

Kupść claims that the adjective dependent in cases such as the latter example is not a valence exponent as it has been reached by mere transposition. As finding such minimal pairs in corpora of dead languges is close to impossible, we do not take restructuring alone into account. The argument to follow is rather that of the inherent valence of the given adjective, i. e. whether the adposition or conjunction of the argument is not a general feature for (most) adjectives in the langauge.

[4] Besides that, the term may also be understood as a subtype of phraseme, where the syntactic expression has developed a fixed compound meaning, e.g. “safe and sound”, which cannot be replaced by “safe and healthy”.