Before discussing how construction grammar relates to grammaticalization, a brief introduction of its key concepts is necessary. Construction grammar (Croft 2001, Langacker
2005, Goldberg 2006, Hilpert 2014) is a theory according to which all linguistic knowledge boils down to knowledge of constructions. Constructions are the minimal units of knowledge and they consist of a linguistic form paired with a meaning. The concept of linguistic form is to be understood broadly. A construction can consist of a single word, such as cat, but can also consist of more complex patterns such as the X-er and Y-er (e.g. the more the merrier). More abstract patterns such as the ways the comparative is expressed in English (more X, most X, X-er and X-est) also constitute constructions. The COMPARATIVE itself (as a grammatical concept) is also a construction (and is written in capitals when used as such).
The view that constructions are the basic and only unit of linguistic knowledge is complemented by the related assumption that constructions are thought to be organized in a network called the construct-i-con. This network is the mental representation of constructions that individual speakers have and therefore involves a certain organization and links between these constructions. For instance, the substantive lexical construction more X is linked to a more abstract COMPARATIVE construction. Constructions can be ranked according to their degrees of schematicity and abstractness. Thus, the COMPARATIVE construction is more schematic and abstract than the more X construction. The terms macro-, meso- and micro- constructions are sometimes used to label constructions, from more to less schematic/abstract (Traugott 2008: 236).
For the present purpose, the definition of what constructions are for the present purposes is taken from Goldberg (2006: 5):
Any linguistic pattern is recognized as a construction as long as some aspect of its form or function is not strictly predictable from its component parts or from other constructions recognized to exist. In addition, patterns are stored as constructions even if they are fully predictable as long as they occur with sufficient frequency.
The first part of this definition posits that a construction must not be fully predictable from its component parts (non-compositionality), which is a logical consequence of the assumption that constructions are the minimal units of knowledge. This is how more intelligent is not an example of a construction, because it can be fully predicted on the basis of the two constructions it is made of, namely more X and intelligent. More intelligent is called a construct, which denotes an instantiation of a more general construction (i.e. more X). The second part of the definition points to an exception to non-compositionality. Highly frequent elements, such as I don’t know, are also considered as constructions, despite having a
meaning fully predictable on the basis of the more general constructions it involves. This is because frequent elements are conventionalized and tend to replace possible alternative. If a speaker is asked a question to which they don’t know the answer, they will likely respond with I don’t know, instead of This information is unknown to me, or I lack this piece of knowledge. This is why I don’t know is likely stored as a construction in its own right in the construct-i-con. The inclusion of frequency into the definition also implies that highly frequent elements such as common plurals (e.g. cats) can also be regarded as constructions if frequent enough. This blurs the line between the concept of construct and construction because the earlier example of more intelligent could be considered as rather frequent enough to be considered as its own construction. This particular question regarding how frequent a construct needs to be in order to be considered as a construction goes however outside the scope of the present discussion (see Bybee 2003, 2013).
Language change can be studied from a construction grammar point of view and the terms “constructionalization” and “constructional change” are often used to describe this approach (e.g. Traugott and Trousdale 2013). Constructionalization is the creation of new form-meaning pairings, while constructional changes are changes that affect the features of an existing construction. Constructional changes are believed to precede and succeed most cases of constructionalization. The creation of new form-meaning pairings (i.e. constructions) is mainly made possible by prior changes in an already existing construction.7
For example, the Modern English construction a lot of developed from the Old English word hlot which used to refer to a concrete object, mainly a piece of wood to select an individual, in the context of elections and of drawing lots. The word hlot was also used to refer to a share or unit of something gained by such drawing lots. In Middle English, partitive uses of lott are attested, along with the preposition off, such as aniʒ lott Off Moysœsess lare (any part of Moses’ teaching), as well as uses with a meaning close to a “group of”, such as Summ lott off gode sawless (certain groups of good souls), as reported in Traugott and Trousdale (2013: 23-26). In those examples, lott is a head, followed by a modifier (off NP) and it denotes units that are part of a larger whole. These uses of lott are still lexical as they refer to a concrete unit, but since this unit is part of a larger whole, lott also has a relational (i.e. partially grammatical) quality. It is only in later attestations from the early 19th century that lot of can be regarded as fulfilling the grammatical function of quantifier, similar to Present-Day English, meaning “a large quantity of” and where it is no longer a head, but a
7 Traugott and Trousdale (2013: 29-30) distinguish between gradual constructionalization and instantaneous
modifier instead. This change has been described by the following schema which shows the reanalysis of the structure of the construction (i.e. its form) and its associated meaning:
[[Ni [of Nj]] ↔ [parti - wholej]] > [[N of] Nj]] ↔ [large quant - entityj]]
(Traugott and Trousdale 2013: 25)
This illustrates that constructionalization (i.e. creation of a new construction) may require constructional changes to take place first. Partitive uses of lexical hlot/lott off paved the way for structural reanalysis and for uses as a quantifier. These changes are also followed by host- class expansion, such as current uses of a lot of with much more abstract concepts that are not inherently quantifiable (e.g. a lot of courage). This shows that constructional changes can also occur after constructionalization has taken place.
Constructional change overlaps with grammaticalization, although the two concepts are distinct. An obvious example are semantic changes within lexical categories, such as the narrowing of hound which currently refers to a specific type of dog in English, but used to refer to any kind of dog in the past. These changes fall under constructional changes (i.e. changes in the form-meaning pair), but do not involve grammaticalization. In contrast, grammaticalization can involve highly abstract developments that are not necessarily constructions themselves. For instance, does the development of modal auxiliaries in English entail the existence of the highly general construction MODAL AUXILIARIES as well? It is unsure that most speakers do this generalization and indeed have a MODAL AUXILIARY construction in their construct-i-con. This shows that some processes can pertain to grammaticalization, but not necessarily to constructional change (see Hilpert 2013: 9-13 for more details).
Despite these examples, there is a significant overlap between constructional change and grammaticalization. As discussed earlier, construction grammar proposes an organization of the contruct-i-con based on degrees of schematicity and abstractness, which is reminiscent of degrees of grammaticalization, especially when considering the increase in generality of meaning (and abstractness) often involved in grammaticalization (section 2.2.1). There are at least two main ways in which adding a construction grammar point of view enhances the understanding of grammaticalization processes that are relevant in the present dissertation.
First, construction grammar posits a clear minimal unit (also called atomic units) of study, the construction. As discussed earlier, constructions can consist of single words (e.g. cat), or more complex patterns (e.g. the X-er the Y-er). While the grammaticalization
continuum presented in Hopper and Traugott (2003: 7) involves the terms “lexical item” and “grammatical word”, they should not be taken to necessarily mean single items/words but also to include more complex constructions. In fact, Himmelman (2004: 31) explicitly states that grammaticalization applies to “constructions and not isolated lexical items”. He also points out that the grammaticalizing element itself has been given more importance as opposed to its syntactic context, giving the example of the focus on go in the future be going to + infinitive construction, to the detriment of the context around it which is highly relevant. Bybee (2003: 602) advocates for a similar position where these developments can only be studied “in the context of the construction in which the grammaticalizing element occurs”, which is also supported by Traugott (2003). This will be relevant in the subsequent chapters, in different ways. Chapter 5 deals with the synchronic measurement of grammaticalization using lists of lexical and grammatical elements. While many of these elements involve single words, they also involve more complex patterns, such as complex prepositions (e.g. as long as, in order to). In addition, in chapter 6, other complex patterns involving auxiliaries such as be going to + infinitive and Keep Ving are also included. Information regarding the parameters of
grammaticalization discussed in chapter 3 will be collected using the entire pattern, instead of its main grammaticalizing elements. A limitation however is that changes within the patterns themselves will not be taken into account, as they are not the focus of the study. Instead, the focus will be more on the syntactic context around these patterns, in the form of collocate diversity (section 3.3).
Second, since constructions are part of the construct-i-con, the links they have with more general constructions can be explored as well. This was shown to be useful for example in Hilpert (2013: 79-109) who notes that the development in English from mine to my is extremely similar to that from thine to thy, on the basis of diachronic corpus-based data. Given that these two developments are fairly similar, Hilpert argues that these two elements might in fact instantiate a single construction that is a generalization of these two. This might be relevant in Chapter 6 as it involves complex prepositions which might instantiate a higher schema. As illustrated in Hoffmann (2004, 2005), complex prepositions can involve similar patterns. For example, with regard to and with respect to are instances of the pattern with X to. The empirical investigations in chapter 6 might also highlight that similar patterns display similar developments.
Two aspects of construction grammar, namely that constructions are the minimal unit of study and that these units are linked to each other in a construct-i-con, have been highlighted
as highly relevant to grammaticalization and further discussion on this matter can be found in Hilpert (2013), Trousdale (2010), Gisborne and Patten (2011), and Fried (2013). The main defining traits of grammaticalization have been introduced in the previous sections and a summary is offered in the next section.