No question: lexicalization and grammaticalization processes in the development of modal qualifier meanings

Uses	n
%
lexical noun + complement clause	2	1
lexical ‘unchallengeable’	28	12.5
lexical ‘not at stake’	40	17.5
positive modal: epistemic	76	34
positive modal: dynamic/deontic	10	4.5
negative modal: dynamic/deontic	44	19.5
negative modal: epistemic	25	11
TOTAL	226	100

Table 2: Uses of there-clauses in WB sample

The smaller proportion of existential examples, 74, or 33%, involve lexical uses. In two, we find a pattern not attested in the historical data: the lexical noun question as such (not as part of a composite predicate) sanctions a complement clause. Using noun complement clauses, Example (21) from our sample describes an educational policy document about maths teaching: it does not contain any question addressing whether the child understands the concept, but merely statements about children learning to read and recognize digits.

(21) There are statements that a child will be able to read and recognise a four-digit number for example but no question on whether the child understands the concept." (WB)

All the other examples have the lexicalized semi-fixed idioms that emerged in Early and Late Modern English. The ‘not be challengeable’-idioms, which appeared around 1700, form 12.5% of the sample. The notion or proposition potentially under challenge is still expressed either by a prepositional phrase, as in (4) above, or a clausal complement, e.g. (22). The primary status of the idiom in the utterance is typically brought out by contrastive or equivalent assertions in the discourse context. In (4) the army’s usual wariness of soldiers crying off sick is contrasted with their not questioning Tolkien’s condition, which is hence addressable: was there really no question of his condition? (Boye & Harder 2006: 581). In (22) the exclusion of doubt expressed by there was no question is preceded by equivalent make certain.

(22) To make certain, we continued until there was no question that what we were looking at was the runway. (WB)

The ‘not be an issue’-idioms, which emerged around 1850, account for 17.5% of the sample. They take nominal complements, as in (5), in a majority of cases (90%). Clausal complements, as in (23), are very restricted (10%). In these examples, the meaning of an idea not being (raised as) an issue is also supported by contextual elements such as and there never was any question in (23).

(23) He said: `Craig is under contract until the end of the next World Cup campaign, and that won't change. There's absolutely no question of him leaving, and there never was any question no matter what happened at Wembley. (WB)

The grammaticalized modal uses of the existential pattern form, with 152 out of 226, the majority of our sample. They have come to realize in Present-day English a full-fledged system of modality-cum-polarity. Their polarity value has crystallized in a systematic association with complement types: there be no question + (that) + finite clause functions as positive emphatic marker and there be no question of + gerund functions as negative marker. These are the systematic correlations between polarity and complement types identified by Kjellmer (1998) for Present-day English. The fledgling extension of finite complements to the negative marker observed in Late Modern English (example 20) seems not to have persisted.

Importantly, both polarity types can now modify either epistemic or deontic-dynamic statements. The positive marker uses, which started off modifying truth claims, had already extended to desired actions in Late Modern English, as in (16). This extension is confirmed in our Present-day English data, with examples such as (2) above, There is no question he must stay, even though epistemic claims remain the more common option, viz. 76 tokens versus 10 dynamic-deontic statements. The negative modal marker there be no question of + gerund, which had started off conveying deontic-dynamic meanings, have come to also convey epistemic meanings, as in (3) above, There’s no question of Her Majesty turning up on your doorstep unannounced. Epistemic uses were not attested in our Late Modern English data, but in the present-day sample total 25 tokens against 44 deontic-dynamic uses.
The two possessive clauses with prepositional phrase, e.g. (24), in our sample are lexical uses in which have no question means ‘not challenge’, much like the there be no question-idiom illustrated in (12). However, a quick search on the Internet reveals that I have no question is also used in a secondary, grammaticalized way, in which it approaches the status of a positive epistemic modifier, e.g. (25). The same holds for copular matrices: the Internet contains examples like (26) in which they function as a parenthetical expressing strong certainty.

(24) I have met many homosexuals with equally unshakable masculine gender identity. They have no question about their maleness. (WB)

(25) Had he played a full year, I have no question he'd have been the consensus #1. (http://forums.rotoworld.com/index.php?/topic/267200-kyrie-irving-2011-2012/page__st__40)

(26) Christ suffered for us, it is no question, .. .(http://biblelight.net/gospel-2.htm)

Referring back to the historical data, we can see I have no question as the expression that took over the grammaticalized uses from obsolete I make no question in (14). And, whereas the copular clauses were still used mainly propositionally in Late Modern English, e.g. (15), they have now established clearly grammaticalized uses. A casual Internet search did not reveal any examples of possessives or copulars with no question qualifying deontic statements and they are definitely not used with negative polarity value. The possessive and copular clauses thus have not developed anything approaching the extended system of polar and modal values of there’s no question.

3.5. Differences between lexicalization and grammaticalization

In the previous sectiosn we have seen that the distinction between lexicalization and grammaticalization is crucial to the developmental paths of clausal expressions with (no) question. There is a general consensus that grammaticalization and lexicalization share many features, which makes it difficult to differentiate the one from the other (Lehmann 2002, Brinton & Traugott 2005). Both affect larger syntagms, not individual items, and involve semantic erosion, fusion and fixing of the component elements. However, according to Lehmann (2002: 13), a grammaticalized unit is accessed “analytically”, whereas lexicalized units are accessed “holistically”, because its internal relations have “become irregular and get lost” (2002: 13). On this view, ‘lexicalization’ applies only to complex forms becoming unanalyzable wholes as in fossilization and univerbation (Himmelmann 2004: 28).

A broader view of lexicalization can be based on the approach to idiomatic patterns advocated by Sinclair (1991: 110), Nunberg, Sag & Wasow (1999) and Langacker (1999: 344) amongst others. They point out that many idioms are only semi-fixed and tend to keep a degree of analyzability and internal variability. These insights bear, in our view, also on the diachronic process of lexicalization. To define and recognize lexicalization, we take the result of the process of change as criterial. If change produces a linguistic element with a specific contentful meaning, functioning as a lexical item, we consider it lexicalization¹. We have seen that in the formation of idiomatic patterns with light verb and deverbal noun such as make a/Ø question, there be a/Ø question, the NP keeps some of its modification possibilities (e.g. choice of determiner), adverbs may be added, e.g. never any question, and the verb can vary too, e.g. make/have no question.
In our discussion of the data, we have made the distinction between grammaticalization and lexicalization mainly on the basis of differences on the syntagmatic axis. We have followed Boye & Harder’s (2007) criteria for distinguishing between lexical and grammaticalized uses of complement-taking predicates on the basis of their having primary or secondary discourse usage status. Giving full weight to the discourse context, we applied tests relating to ‘addressability’ such as really-queries, tags, NEG-raising, and modification by a subordinate clause, e.g. (15), (25).
Useful as these syntagmatic tests are for distinguishing lexicalized from grammaticalized uses, they do not constitute the whole picture. Linguistic patterning is not only present on the syntagmatic, but also on the paradigmatic axis. We want to argue that the fundamental differences between lexicalization and grammaticalization lie, besides primariness and secondariness in discourse usage, in the different paradigmatic relations contracted by elements that have become lexical or grammatical. We therefore first have to elucidate the different types of paradigmatic configurations that lexical items and grammatical elements are defined by.
We adhere to the (neo-)Firthian view that it is an essential feature of lexical items that they impose collocational constraints on the lexical items they co-occur with, and, as members of lexical sets, colligational constraints on specific syntactic structures they pattern with. As rightly stressed by Sinclair (1991), lexical meaning does not reside solely in the lexical ‘node’. Rather, the semantic structure of a lexical item is determined by its coselection of specific (sets of) collocates. This is a distributional view of lexical meaning, according to which the collocates are diagnostic of a lexical item’s meaning. For instance, (partial) synonymy can be established in terms of similarity between collocate clouds (e.g. De Deyne, Peirsman & Storms 2009).
Petré, Davidse & Van Rompaey (forthc.) have applied this thinking to the strings be on (the/one’s) way/road to, which have lexicalized composite predicate uses (meaning ‘go to/head for + spatial goal’) and grammaticalized aspectual uses (which mean ‘be going to + action/state/event’). Extensive corpus-based study showed that the composite predicate uses clearly behave as lexical items in imposing the coselection of spatial goals or coercing this meaning onto action-state-event goals (such as be on one’s way to work/a party, etc). As is typical of (largely) synonymous lexical items, they share many collocational restrictions while still displaying distinct preferences. Be on the/one’s road to co-occurs mainly with names of towns and with the fixed phrase road to nowhere. This can be explained by the persistence of road’s very concrete lexical meaning of “An ordinary line of communication used by persons passing between different places” (OED2). Be on the/one’s way to, by contrast, collocates with more diversified spatial nouns, which may be towns, countries, continents, or landmarks such as the airport, the coast and a quarter of their collocates are action-state-event nouns such as incident, picnic, etc. This more diversified distribution of collocates is probably motivated by the more abstract meaning of way, viz. “course of travel or movement” (OED2).
In contrast with composite predicates such as be on the/one’s road/way to, whose lexical semantics predispose them to spatial collocates, the semi-fixed idioms with (no) question impose, like the corresponding simple predicates ask, question, clear colligational restrictions, that is, they predict co-occurrence with specific syntactic structures³. As is to be expected of lexical uses of complement-taking predicates describing locutions, they take either prepositional phrases specifying the ‘matter’ (Halliday 1994: 157) being asked about or questioned, or clausal complements, typically of the reported question type. These colligational patterns are motivated by a general semantic feature in the predicate, e.g. ‘ask’, which needs to be completed by the specific content of the complement (Langacker 1999: 28), e.g. ‘about x’, ‘whether or not x’. In this respect, the idioms with (no) question behave wholly like the lexical uses of simple complement-taking predicates, which argues for viewing their formation as lexicalization⁴.
We now turn to the question what form of paradigmatic organization characterizes grammatical elements. To answer this, we turn again to the Firthian tradition, more specifically, Halliday’s (1992) thought about the paradigmatic organization of the grammar. Halliday views the oppositions within a grammatical paradigm not so much as obtaining between its members, as in the tradition of Jakobson (1971 [1939]), but, at a more abstract level, as obtaining between semantic features associated with the members. By conceiving of systemic oppositions in terms of features, it is possible to capture the interdependencies between features from different systems. Which terms of one system combine with the terms of another system is constitutive of the larger grammatical system.
We noted in Section 3.4 that in Present-day English the positive and negative modal markers have both come to combine with either epistemic or deontic-dynamic modality. We are now in a position to explicate the systemic changes involved in this. In Late Modern English, the grammaticalized clausal there-structures presented language users with the possibility of expressing the following combinations of terms from polarity and modality:

positive + epistemic