This paper has been carried out in the framework of several research projects on language technologies (ref. PID2020-112818GB-I00, E3/04/21, UMA-CEIATECH-04). The author would like to thank two anonymous reviewers for their comments and suggestions.
1Recent years have witnessed a dramatic change of paradigm in phraseology research. Traditional methods have given way to various quantitative, distributional, and computational approaches. Particularly relevant to this paper are three emerging theoretical threads (and their complementary methodologies): Construction Grammar, Computational Phraseology and Corpus Linguistics (Dobrovol’skij & Piirainen [2018], Fellbaum [2019], Goldberg [2019], Corpas Pastor & Colson [2020]).
2We adopt a constructionist approach to the usage-based study of property resultatives (Gries [2013]). Resultative constructions are schematic patterns that convey a change of state caused by the completion of an action or event. When expressing motion, resultatives lexicalise the manner in which the action is performed and indicate the trajectory of the movement and the result outside the verbal unit by means of an adjectival or prepositional phrase (Levin & Rappaport Hovav [2006]). Instances like drive someone up the wall, drive someone out of his/her mind or drive someone mad illustrate metaphorical extensions of the caused motion construction where states are considered locations. In this paper, the terms resultative construction and resultative will be used interchangeably to encompass all subtypes.
3Our initial claim is that this array of idiomatic expressions are just instances of the same type of resultative constructions which convey a person’s change of mental state (typically negative). Differences among instances are due to various collocational preferences that seem to operate at various levels, including grammaticalisation and coercion processes.
4The rest of the paper is organised as follows. Section 1 revolves around core notions of Construction Grammar (CxG), with a special focus on resultative constructions. Section 2 will describe the methodology used in this paper. Resultative constructional idioms of this type will be extracted from two giga-token corpora of contemporary English through protocolised tasks in the form of heuristic strategies and steps. Section 3 will present the results and main findings of our corpus-based study of idiomatic property resultatives, with a focus on pairings of form and meaning (Martí Solano [2020], Corpas Pastor [2021]). Corpora will be analysed in a number of ways, including the frequency of this construction in English (in the two varieties selected), the basic elements of its syntactic structure, and the semantic and informative constraints which operate on these constructions, the lexical filledness of slots, usage and diatopic restrictions, etc. In the concluding part (Section 3), the relevance of our study will be discussed, with special attention to notions like “idiomaticity”, “productivity” and “variability”, among others, which will be revisited from a corpus-based and computational perspective.
To the best of our knowledge, this is one of the first studies based on giga-token corpora that explores idioms as part of more general property resultative constructions.
5Construction Grammar (conventionally abbreviated to CxG) refers to a family of formal and usage-based constructionist approaches that understand language as an intricately structured, semantically motivated network of symbolic units, referred to as constructions (Hoffmann & Trousdale [2013]). Constructions have been characterised in different ways. For the purposes of this paper, the usage-based definition proposed by Goldberg [2006: 5] will be adopted:
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.
- 1 The notion of “construction” could be considered an extended version of the Saussurean concept of l (...)
6Each construction is considered to be symbolic because it comprises a pairing of a particular form with a particular meaning and/or function. In other words, constructions have a form side that is conventionally associated to a meaning pole which is by default non-compositional (it cannot be derived or computed from the assembly of individual parts).1 These conventionalised form-meaning pairs vary greatly in size, complexity and schematicity (Croft & Cruse [2004)], ranging from morphemes (e.g., dis-, -ed), to words (e.g., hand, daredevil), filled idioms or partially lexically-filled phrasal patterns (e.g., make a mountain out of a molehill, send someone to Coventry), and more abstract patterns, like the covariational conditional construction [the Xer… the Yer] or the ditransitive construction [Subj V Obj1 Obj2]. In this constructional network, traditional idioms are no longer viewed as exceptions or irregularities, but as constructions of their own that may share certain aspects with fully productive expressions (and vice versa).
7Following this line of reasoning, an idiom like drive someone up the wall (‘to make (someone) irritated, angry, or crazy’)2 would be a special case (constructional idiom, C1) of another, more abstract construction (resultative construction, C2). Instances of C2 can be found in examples (1-4):
- 3 The two giga-token corpora used in this study are described in Section 2.1.
(1) There are times when he wants to re-create an event but to the exact detail that he remembers, and I must admit that it drives me up the wall. (AEM)3
(2) He accepted to play the king’s fool as long as this did not do violence to his conscience and did not drive him insane, allowing him to push social reform. (EnTT15)
(3) Parents and passers-by were able to push the gate open just enough to prevent it from closing completely. The child had to be physically pulled free, but escaped serious injury. (EnTT15).
(4) I found myself at the Sporting Club in Tribeca, New York City, screaming myself hoarse, painted all over with green, red, and white theatrical paint (AEM).
(5) Men often drink themselves crazy to get over the day’s experiences while women often cry their eyes out. (EnTT15)
- 4 The postverbal NP is not understood to be the object of the verb when used in isolation: cf. scream (...)
8Key components of the Resultative construction are the agent (causer), the patient (experiencer of the caused action), the result/goal (the new state) and the verbal action that causes the change of state. According to Boas [2003: 9], resultatives can be (i) transitive, i.e., based on transitive verbs occurring with an additional phrase supplying the resultative interpretation (as in ex. 1, 2 and 3); (ii) intransitive, i.e., based on intransitive verbs occurring with both a non-subcategorised noun phrase4 (NP) and a resultative phrase to yield the resultative interpretation (as in ex. 4); and (iii) based on transitive verbs with both a non-subcategorised NP and a resultative phrase (as in ex. 5).
- 5 From a purely cognitive viewpoint, the underlying metaphor would be a change of state is a change o (...)
9Goldberg & Jackendoff [2004: 536-537] use the alternative term “Property Resultative Constructions” to denominate this type of metaphorical extension where states are considered locations.5 Resultative constructions can be also considered a metaphorical extension of the Caused Motion Construction: ‘A causes B to move to C by doing D’ (Goldberg [2006]). In this respect, Levin & Rappaport Hovav [2006] point out that resultatives expressing motion lexicalise the manner in which the action is performed and indicate the trajectory of the movement and the result outside the verbal unit by means of an adjectival or prepositional phrase. Thus, drive someone up the wall (C1) can be seen as a partially lexically filled instance of the Resultative construction (C2), which is described below in two alternative ways:
FORM: [SBJ1 V4 OBJs OBL3] ↔ MEANING: ‘Agent1 causes Patient2 to become State3 by V4-ing’.
FORM: [X V Y Z] ↔ MEANING: ‘X causes Y to become Z by V-ing’, where Y is animate and Z is usually an adjectival or prepositional construction.
Following Hoffmann [2017], we will use the second description (more informal and schematic) of the form and meaning parts of these types of conventional pairings. The arbitrary nature of the two poles is represented by a bidirectional arrow.
10We argue that drive someone up the wall is dominated by the Resultative construction via an instance link, i.e., the instance inherits the syntax and semantics associated with the superordinate construction. And the other way round, drive someone up the wall would dominate C2 by a subpart link (as a proper subpart of C2 that exists independently). We also claim that the same inheritance links can be observed with regards to the other instances of resultatives in the examples above (drive someone insane, pull someone free, and freeze solid). The intricate mutual-influence relationships that govern constructions can be further illustrated by C1 with regard to the Verb-specific drive-construction (C3, ex. 3) which is both an instance of C2 and a part of C1, and so forth.
- 6 These alternative terms refer to the same phenomena, although the focus is slightly different: verb (...)
11Resultatives have been extensively analysed in the literature. A state of the art, albeit brief, would be outside the scope of this study. Instead, the interested reader is referred to the seminal works by Boas [2003], Goldberg & Jackendoff [2004] and Beavers [2012]. In this paper we will focus on a special class of resultatives exemplified above: verb-specific constructions (Croft [2012]). They have also been termed mini-constructions (Boas [2003]) and collocational constructions (Corpas Pastor [2015], [2017a]).6
- 7 The event-based frame semantic representation of drive-crazy is described by Boas [2003: 234] as:
A (...)
12Verb-specific constructions are partially lexically-filled phrasal patterns (Goldberg [2006: 215]). They are characterised as form-meaning pairings that capture generalisations at the level of verb classes. Verb-specific constructions are more substantive − and, therefore, less schematic − than general constructional schemas (e.g., ditransitive, resultative, etc.). According to Boas [2003], they are represented by an event-frame7 with its own semantic/pragmatic and syntactic specification. In this type of constructions only verbs of a given class may occur in the construction at hand, but not every member of the verb class can always do so. These constructions are always partially lexicalised regarding their verbal slot, which is instantiated by a set of verbal fillers selected in some arbitrary fashion, similarly to the collocational restrictions described in the literature (cf. Corpas Pastor [2015, 2017b]).
- 8 See also, Langlotz [2006] and Goldberg [2019].
13It should be noted that verb-specific constructions are related to other lexically-filled phrasal patterns: constructional idioms. According to Booij [2002: 320], constructional idioms are “syntactic constructions with a (partially or fully) non-compositional meaning contributed by the construction, in which – unlike idioms in the traditional sense – only a subset (possibly empty) of the terminal elements is fixed”. These symbolic units exhibit varying degrees of productivity and schematicity (Taylor [2015]).8
- 9 According to Shiota et al. [2014], uncontrolled desire triggers positive emotions that promote an a (...)
14The Drive Y Z construction also classes as a constructional idiom, as its meaning/function is not entirely compositional. For instance, in ex. 6 the non-compositional nature of this construction shines through the adverb literally, while negative prosody is implied by the danger of remembering a traumatic experience (notice the use of the connector or). The interplay between literal and idiomatic readings is also present in ex. 13. Negative prosody is marked in ex. 7 (notice the use of expletives goodammit, damn) and also implicit in ex. 11 and 12. Like most idioms, the metaphorical basis of the construction functions as an intensifying element: compare irritate somebody versus drive somebody mad and drive somebody up the wall. An example of explicit intensification can be also found in ex. 8, 9, 13 and 15 (notice a frequent use of exclamation marks and gradation, like so much, nearly, etc.). In some cases, the emotions are intense, difficult to control, but not necessarily negative,9 as in ex. 15. Resultatives with APs wild and crazy (and occasionally mad) may also carry sexual connotations (ex. 8-10). This is particularly the case in some resultatives with PPs like over the edge, as they tend to convey intense desire and lust (ex. 16).
- 10 Araneum Anglicum Maius (AAM, cf. 2.1.).
(6) But it’s also so horrific and terrible that it must quickly be forgotten or it could literally drive you insane. (AAM)10
(7) Goddammit I know. This is driving me crazy. They want it faster and it has to support all these other damn systems. (AAM)
- 11 enTenTen15 (enTT15, cf. 2.1.).
(8) I love you so much. You drive me wild. (enTT15)11
(9) These girls are very open minded and would drive you crazy with their sexy figures and erotic talks. (enTTT15)
(10) The Changeling is a gripping and darkly comic tale of how love and sex drive us mad, and one of the most powerful tragedies ever written. (enTT15)
(11) They smell large amounts of money and they circle like sharks in bloody water. Money is like pheromones to them and it drives them into a frenzy with desire to obtain it. (enTT15)
(12) Injustice and cruelty drives him to despair and loss of faith. (enTT15)
(13) The crumbs I can deal with – it was the footprints and smudges that drove me batty! (enTT15)
(14) When my son was very young, and I was telling him off, I said, “You’re driving me up the wall!”. He instantly imagined himself actually driving me up the wall, and burst out laughing. When I realised what he was laughing at, I had to join in. (enTT15)
(15) It took a ton of self control, because the tempting peanut buttery chocolate smell alone was nearly driving me insane (AAM).
(16) As always, because we have such a close bond, we orgasmed together, well it was actually her third when I finally let myself go into her to drive her totally over the edge. (enTT15).
15In general, constructional idioms have often been considered as just idiomatic sequences, but this view ignores the fact that these are not isolated cases but productive constructions that can be used with a variety of fillers, mainly in the verbal slot (Corpas Pastor [2021]). In this respect, resultative verb-specific constructions could also be classed as a special type of constructional idioms.
- 12 Non-compositional meaning aspects are given in curly brackets.
16In this light, examples (1-2) and examples (6-15) instantiate higher- and lower-order constructions: (i) a resultative construction, (ii) a verb-specific resultative construction, and (iii) a resultative constructional idiom. They all contribute to the meaning/form pairings of a given instance, as lower-order constructions are inter-dependent on higher-order constructions, and vice versa. We will refer to this construction conglomerate as the Drive-class Y Z construction (Drive-class construction, for short). This special case of the Resultative construction can be described12 as follows:
FORM: [X DRIVE Y_animate Z] ↔ MEANING: ‘X causes Y to become Z (insane, mentally unstable, very unsettled and/or irritated) {intensification, metaphorical implications, and usually negative prosody}’
- 13 Usage-based approaches consider two types of frequency: (a) high token frequency that frequently le (...)
17Cognitive Construction Grammar (Goldberg [1995], [2006], [2013]) and Radical Construction Grammar (Croft [2001], [2013]), consider frequency13 as a distinctive feature: symbolic units emerge through repeated experience with actual instances of constructions and their generalisations, and this also fosters entrenchment. The higher the input frequency of a particular construction (construct and/or pattern) is, the stronger will be its entrenchment in the neural work (Hoffmann [2013: 315]). In many approaches to CxG, constructions are equated to mental representations that are learned through language use and are particularly sensitive to a number of factors, such as type and token frequency or prototypicality, among others. For this reason, usage-based constructionist approaches are particularly amenable to corpus-based methods (Yoon & Gries [2016]).
This section will include a description of the corpora selected for the study, as well as the heuristic data analysis approach applied.
18It is an established fact that to study resultatives a large corpus is needed. According to Boas [2003: 11], “in order to construct an adequate theory of resultative constructions, we should not restrict ourselves to a limited set of data, but should instead aim at collecting large amounts of empirical data in order to cover the subject of study in its entirety”. In the case of idiomatic resultative constructions, corpus data are even more necessary due to their low number of occurrences. Compare the frequency of drive (someone) up the wall in the corpus enTenTen15 (628 tokens/ 0.04 per million tokens) with the figures for drive (2,169,187/ 140.75) and the prepositional phrase up the wall (2,826/ 0.18).
19Boas [2003] used the British National Corpus (BNC) of 100 million words in his account of English resultatives. He studied the Resultative construction with drive consisting of an animate object with an adjective or prepositional phrase synonymous with mad in the BNC corpus. He concluded that this type of verb-specific resultative construction (drive-crazy) is very productive in English, it is partially lexically filled and largely conventionalised (like the majority of resultative constructions), and it shows a clear preference for adjectives or adjectival phrases (APs) (mad, crazy, insane, wild, nuts, batty, dotty, crackers) over prepositional phrases (PPs) (to madness, to insanity, to distraction, to suicide, to despair, to desperation, up the wall, into a frenzy, over the edge).
- 14 We carried out a corpus-based contrastive analysis of ‘insanity’ idioms with mad/loco in comparison (...)
- 15 enTenTen15 (enTT15) is a 15 billion Web corpus that was web-crawled in 2015. See a full description (...)
- 16 News on the Web (enNOW) is a 13.9-billion-word corpus of English with data from web-based newspaper (...)
- 17 The Global Web-based English (GloWbE) is a 19-billion-word Web corpus of English with texts from tw (...)
20In Corpas Pastor [2021], we studied the Drive Y Z construction14 in three very large web-crawled corpora: enTenTen1515, News on the Web16, and The Global Web-based English corpus17. We compared Boas’s [2003] results with occurrences of ‘insanity’ synonyms in the three giga-corpora analysed. We focused on APs and related PPs (to madness, to insanity). Our findings are summarised in Table 1:
Table 1. Distribution of ‘insanity’ synonyms in the Drive Y Z construction (Corpas Pastor [2021])
|
BNC
|
enTenTen15
|
enNOW
|
GloWbE
|
|
mad
|
108
|
5,094
|
821
|
429
|
|
crazy
|
70
|
13,514
|
2297
|
1168
|
|
insane
|
23
|
3,771
|
443
|
333
|
|
wild
|
22
|
2,194
|
244
|
102
|
|
nuts
|
18
|
68
|
615
|
410
|
|
batty
|
4
|
468
|
53
|
37
|
|
dotty
|
4
|
9
|
2
|
1
|
|
crackers
|
4
|
8
|
4
|
4
|
|
to madness
|
5
|
763
|
19
|
8
|
|
to insanity
|
1
|
402
|
9
|
16
|
- 18 Verb frequencies were obtained from the enTenTen15, as this corpus seems to contain the higher numb (...)
- 19 The findings in Corpas Pastor [2021] suggest that the preferred exemplar or construct in World Engl (...)
21In the aforementioned paper we argue that the present situation differs from Boas’ account [2003] in that the inventory of ‘insanity’ adjectives and causative verbs has changed substantially. New adjectives have appeared (mental, frantic, berserk, potty, loopy, bonkers, bananas), while others tend to be used much less (dotty, crackers). At the same time, the set of verbs used in the drive construction with the most typical or central adjectives (mad, crazy, insane, wild) seems to have expanded as well: make, get, send, turn18. The large data retrieved from the giga-token corpora also show that the Drive X Y construction formally occurs with the central verb drive and two central ‘insanity’ adjectives (mad and crazy), that are varietally dependent: the central adjective of the construction seems to have moved from mad to crazy in World English (and non-British varieties), with a strong preference for mad in British English.19
22We also concluded that giga-token corpora provide much richer data on this type of low-occurring constructions. However, frequency differences across three corpora could be determined by their different size, composition and/or type: enTenTen15 and GloWbE are web-crawled reference corpora of mixed genres that include a large number of English varieties, while enNOW is a monitor corpus that contains web-based newspapers and magazines from 2021 to the present time (its size increases around two billion words each year).
- 20 For a detailed description of the Aranea family of corpora, see Benko [2014].
- 21 Both corpora have been web-crawled automatically from Internet resources, filtered by language, pre (...)
- 22 https://www.sketchengine.eu/.
23The differences in the results obtained by Boas [2003] and Corpas Pastor [2021] could be partially explained by the small-scale corpus used by the former. For this study, two giga-token corpora of English have been selected: the enTenTen15 (enTT15) and Araneum Anglicum Maius (AAM or Araneum, for short), a 1.2 billion-word Web corpus created in 2013 and released in 201520. These two corpora are roughly comparable in terms of compilation method (automatic creation), pre-processing, time frame, mixed genres and language varieties covered21. Both of them are available through Sketch Engine22. They differ in size: while both of them are huge corpora of contemporary English (over 1 billion words), enTenTen15 is twelve times bigger than the Araneum that, in its turn, is twelve times bigger than the BNC. See Table 2 for a summary description of the two corpora:
Table 2. Corpus size and components
- 23 Tokens includes word and non-word tokens (e.g., “man” versus the semi-colon “;”).
- 24 Sketch Engine does not provide information about the number of types per subcorpus (existing or cre (...)
- 25 See Benko [2014] and the information about this corpus at http://ucts.uniba.sk/aranea_about/index.h (...)
|
enTenTen15
|
Araneum Anglicum Maius
|
|
Tokens23
|
15,411,682,875
|
1,200,023,361
|
|
Words
|
13,190,556,334
|
888,466,066
|
|
Types
|
46,275,610
|
4,696,677
|
|
Documents
|
33,655,541
|
1,159,878
|
|
Subcorpora
|
|
BrE
|
Tokens24
|
667,443,008
|
65,859,463
|
|
AmE
|
Tokens
|
179,947,609
|
8,913,437
|
|
Genres
|
|
Mixed genres: Arts, Business, Computers, Games, Health, Home, News, Recreation, Reference, Regional, Science, Shopping, Society, Sports.
|
Mixed genres (non-specified).25
|
24General English (World English) and the two main varieties (British and American) will be considered. World English encompasses all English varieties included in each corpus (including American and British English), plus documents in English with non-country specific top language domains (TLD), such as .eu, .org, .gov, .com, .net, etc. The enTenTen15 corpus includes subcorpora of British and American English by default, as well as other subcorpora of Australian, Canadian, European, Indian, New Zealand and South African varieties. The Araneum corpus does not include subcorpora of language varieties. In this case we have created two subcorpora from documents whose URL ends in the country specific Internet domains .uk and .us.
25Our methodology is partially based on Boas [2003] and Corpas Pastor [2021]. It includes a recursive data collection protocol which consists of three heuristic phases:
-
Initial selection of matrix verbs and resultative phrases;
-
Quantitative analysis based on distributional data;
-
Extraction of further instances through query patterns.
- 26 Only the verb put has been added to List A as it is a component of the instance put sb. on edge.
- 27 We have added round the bend, around the bend, out of one’s mind and out of one’s wits, which are d (...)
26First, all slot fillers identified by Boas [2003] and Corpas [2021] for the Drive-class Y Z construction have been selected and sorted according to their corresponding slot (Y or Z), and per pattern within slot Z (APs, PPs). List A encompasses all matrix verbs (‘drive-mental state’: e.g., drive, send). List B includes resultative APs (e.g., mad, potty, bonkers) and List C encompasses resultative PPs (to insanity, up the wall). Lists A26 and C have been further enlarged with other resultative PPs that occur in idiomatic resultative constructions with drive27, which we have extracted randomly from various dictionaries. Table 3 displays the base fillers for the verbal slots (i.e., verbs that fuse in this construction) and the resultative phrases (APs and PPs).
Table 3. Base slots fillers
|
A
|
B
|
C
|
|
Verbs
|
APs
|
PPs
|
|
drive
|
bananas
|
around the bend
|
|
get
|
batty
|
into a frenzy
|
|
make
|
berserk
|
on edge
|
|
put
|
bonkers
|
over the edge
|
|
render
|
crackers
|
out of mind
|
|
send
|
crazy
|
out of one’s mind
|
|
set
|
dotty
|
out of one’s wits
|
|
turn
|
frantic
|
round the bend
|
|
loopy
|
to despair
|
|
insane
|
to desperation
|
|
mad
|
to distraction
|
|
mental
|
to insanity
|
|
nuts
|
to madness
|
|
potty
|
to suicide
|
|
wild
|
up the wall
|
27Lists A, B and C enable us to extract distributional data for this construction automatically.28 To this end, a PHP script has been written to extract data from the Sketch Engine JSON API29 (application programming interface)30. The script generates the necessary CQL (corpus query language) query patterns by means of the matrix verbs and resultative phrases in Table 3. Fig. 1 shows a fragment of the script.
28The script uses automated HTTP requests to communicate with the API. Araneum and enTenTen15 (sub)corpora have been queried automatically in a sequential fashion. Four CQL query patterns have been launched in order to extract frequency data for each construct (V+AP or V+PP pattern) from each corpus: (i) MatrixVerb_AP_enTT15, (ii) MatrixVerb_PP_enTT15_PP, (iii) MatrixVerb_AP_AAM, and (iv) MatrixVerb_PP_AAM. Then, the API answers have been stored by means of tables.
29By way of illustration, for the query [lemma="drive"][tag="(NP.*|PP)"][lemma="crazy"] in one of the corpora, we obtained the frequency result "concsize": 1114, which we entered in the corresponding table. In addition, to create the subcorpora we have used two general query patterns that encompass all constructs in pattern V+AP or V+PP per corpus. These pattern-filtered subcorpora are intended to be used for qualitative analysis of the data.
Figure 1. Script screenshot
30The third step of our protocolised method allows us (a) to uncover (novel) slot fillers not included in Table 3, and (b) to extract additional instances of fillers in Table 3 which could not be retrieved/quantified automatically in the previous phase of our analysis, due to sequence discontinuity, wrong POS and parsing errors, typos, substandard spellings, grammatical errors, among others. For example, sequences like “as in “Without Umbra stabilizing the sword, it drove whoever touched it insane” (enTenTen15), “dive him mad” (Araneum), “is driving me mad”, “druv me crazy”, “drovey me crazy” (enTenTen15), etc. can only be retrieved and quantified this way.
31To this end, six pattern-based search sequences have been used in the form of CQL advanced queries (see below). Results have been used to enlarge/revise the inventory of slot fillers and to refine the quantitative analysis performed in the previous phase.
1) V. + Pers. Pron./Pers. Noun + Resultative APs (List B, Table 3)
|
[tag="V.*" & lemma!="(get|drive|make|put|render|send|set|turn)"] [tag="(NP.*|PP)"][tag="J.*" & lemma="(bananas|batty|berserk|bonkers|crackers|crazy|dotty|insane|mad|
mental|nuts|frantic|loopy|potty|wild)"]
|
2) V. + Pers. Pron./Pers. Noun + Resultative PPs (List C, Table 3)
|
[tag="V.*" & lemma!="(get|drive|make|put|render|send|set|turn)"] [tag="(NP.*|PP)"]([lemma="a?round"][lemma="the"][lemma="bend"]| [lemma="into"][lemma="a"][lemma="frenzy"]|[lemma="to"] [lemma="distraction|insanity|madness|suicide|despair|desperation"]| [lemma="over"][lemma="the"][lemma="edge"]|[lemma="on"][lemma="edge"]| [lemma="out"][lemma="of"][lemma="mind"]|[lemma="out"][lemma="of"] [tag="PP.+"][lemma="mind|wits"]|[lemma="up"][lemma="the"][lemma="wall"])
|
3) Matrix V. (List A, Table 3) + Pers. Pron./Pers. Noun + Adj.
|
[tag="V.*" & lemma="(get|drive|make|put|render|send|set|turn)"] [tag="(NP.*|PP)"][tag="J.*" & lemma!="(bananas|batty|berserk|bonkers|crackers|crazy|dotty|insane|mad|mental|nuts|frantic|loopy|potty|wild)"]
|
4) Matrix V. (List A, Table 3) + Pers. Pron./Pers. Noun + Prep. + N.
|
[tag="V.*" & lemma="(get|drive|make|put|render|send|set|turn)"] [tag="(NP.*|PP)"][tag="IN"][tag="N.*" & lemma!="distraction|insanity|madness|suicide|despair|desperation"]
|
5) Matrix V. (List A, Table 3) + Pers. Pron./Pers. Noun + Prep. + Det. + N
|
[tag="V.*" & lemma="(get|drive|make|put||render|send|set|turn)"] [tag="(NP.*|PP)"][tag="IN"][tag="DT"][tag="N.*" & lemma!="bend|frenzy|mind|wall"]
|
6) Matrix V. (List A, Table 3) + Pers. Pron./Pers. Noun + Prep. + Adj. pos. + N
|
[tag="V.*" & lemma="(get|drive|make|put||render|send|set|turn)"] [tag="(NP.*|PP)"][tag="IN"][tag="PP.+"][tag="N.*" & lemma!="mind|wits"]
|
32This section summarises the main findings of our study. We have performed quantitative and qualitative analyses of the construction under scrutiny in the two corpora selected (enTenTen15 and Araneum). Quantitative data will be provided by means of tables and also as part of the discussion. Tables 4-11 have been obtained automatically through the script described in Section 2. Due to the low frequency of most slot fillers, only raw figures have been entered (i.e., total number of occurrences in both corpora). Normalised frequencies (also retrieved automatically via the script) are provided throughout the discussion when appropriate. Further instances and any other distributional data, obtained by means of the six pattern-based CQL queries (phase 3), will be provided as part of the discussion.
33Tables 4-11 are organised per matrix verbs in descending order of frequency of co-occurrence with crazy (central AP) in enTenTen15: drive (8,481), make (2,152), get (128), send (45), turn (44), put (20), set (9) and render (1). Raw frequencies are provided for both enTenTen15 and Araneum (World English at the top; British English below, right, American English, left below). For instance, the figures in Table 5 for drive someone insane are to be read as follows: 2,073 occurrences in enTenTen15 (24 in the British subcorpus, 25 in the American subcorpus), and 223 occurrences in the Araneum (7 in the British subcorpus, 2 in the American subcorpus).
34A word of caution is needed as regards diatopic varieties, as not all documents can be traced back to a particular variety (e.g., TLDs .com, .org, .net, etc.), nor all .uk and .us TLDs necessarily correspond to British and American varieties. For instance, a document identified as .uk is located in Britain, but it could have been produced by a non-British speaker or in any other English-speaking country. In addition, the low number of instances may compromise results on language varieties, since corpora much larger than enTT15 are needed. Distributional data on American and British English in the tables below should be taken as indicative only.
35The verb-specific construction Drive Y Z licenses resultative constructions with the homonymous verb in its ‘drive-mental state’ sense (see Table 4). This specific sense is inherited by the verb from the general resultative construction (caused motion), and, through a process of semantic coercion, it is interpreted not as a caused change of place, but as a caused change of mental/emotional state: in this case, resultative APs and PPs in the sense of emotionally unsettled, irritated, unbalanced, etc. No wonder, then, that the verb drive is the central verbal slot filler of this particular construction (drive someone mad/mental/to insanity, etc.).
Table 4. Matrix verb drive
|
V=drive
|
Z=AP
|
Z=PP
|
|
enTT15
|
AAM
|
|
enTT15
|
AAM
|
|
bananas
|
24
|
12
|
around the bend
|
20
|
7
|
|
5 / 0
|
0 / 0
|
|
8 / 2
|
1 / 0
|
|
batty
|
286
|
44
|
into a frenzy
|
59
|
3
|
|
3 / 11
|
0 / 0
|
|
1 / 0
|
0 / 0
|
|
berserk
|
20
|
2
|
on edge
|
1
|
0
|
|
0 / 0
|
0 / 0
|
|
0 / 0
|
0 / 0
|
|
bonkers
|
208
|
45
|
over the edge
|
281
|
28
|
|
7 / 2
|
2 / 0
|
|
8 / 7
|
0
|
|
crackers
|
7
|
1
|
out of mind
|
2
|
0
|
|
0 / 0
|
0 / 0
|
|
0 / 0
|
0 / 0
|
|
crazy
|
8,481
|
1,114
|
out of one’s mind
|
121
|
19
|
|
105 / 154
|
20 / 8
|
|
1 / 2
|
0 / 0
|
|
dotty
|
5
|
1
|
out of one’s wits
|
1
|
1
|
|
1 / 0
|
1 / 0
|
|
0 / 0
|
1 / 0
|
|
insane
|
2,073
|
223
|
round the bend
|
21
|
2
|
|
24 / 25
|
7 / 2
|
|
2 / 2
|
0 / 0
|
|
mad
|
25
|
32
|
to despair
|
21
|
26
|
|
22 / 21
|
26 / 0
|
|
10 / 3
|
4 / 0
|
|
mental
|
60
|
8
|
to desperation
|
64
|
8
|
|
4 / 0
|
1 / 0
|
|
2 / 0
|
0 / 0
|
|
nuts
|
23
|
9
|
to distraction
|
315
|
38
|
|
3 / 4
|
0 / 0
|
|
11 / 4
|
4 / 0
|
|
frantic
|
20
|
8
|
to insanity
|
67
|
12
|
|
0 / 0
|
0 / 0
|
|
0 / 1
|
0 / 0
|
|
loopy
|
12
|
0
|
to madness
|
141
|
5
|
|
0 / 0
|
0 / 0
|
|
3 / 6
|
0 / 0
|
|
potty
|
21
|
3
|
to suicide
|
22
|
23
|
|
7 / 0
|
1 / 0
|
|
12 / 6
|
2 / 0
|
|
wild
|
1,030
|
57
|
up the wall
|
26
|
30
|
|
9 / 22
|
2 / 0
|
|
19 / 8
|
4 / 0
|
- 31 Standardised frequencies over 0.01 are indicated within brackets: enTT15 first and then AAM, separa (...)
- 32 Bonkers has one more occurrence than batty in AAM (45 versus 44), although they both have the same (...)
- 33 Mad has a standardised frequency of 0.01 in enTT15 and 0.03 in AAM. This could be due to the differ (...)
- 34 Berserk and frantic show the same raw frequency (20) in enTT15, but frantic appears to be slightly (...)
36The Drive Y Z construction can take all 15 ‘insanity’ adjectives in Z. However, not all APs are equally central. For instance, the adjectives with a higher number of occurrences in both corpora (see Table 4), which also exhibit the highest standardised frequencies, are: (1) crazy31 (0.55; 0.93), (2) insane (0.13; 0.19) and (3) wild (0.07; 0.05). The rest of resultative APs with drive are (4) batty (0.02; 0.04), (5) bonkers32, (6) mental, (7) bananas, (8) mad33, (9) nuts, (10) potty, (11) frantic, (12) berserk34, (13) loopy; (14) crackers, (15) dotty.
37Drive also seems to have a central position in the construction with PP resultatives. This matrix verb fuses in constructions with all 15 PP resultatives in the ‘drive-mental state’ sense. The list of preferred PP resultatives (central exemplars) can be also established by their token frequency (raw/standardised) and their coverage, i.e., whether they are instantiated in both corpora or only in one of them. We will assume that all PPs are present in enTT15 by default. Cases of phrasal resultatives that only appear in AAM will be indicated. Central are (1) to distraction (0.02; 0.03), (2) over the edge (0.02; 0.02), and (3) to madness (0.01, <0.01), followed by the rest of the PPs, all of them with normalised frequencies below 0.01: (4) out of mind, (5) to insanity, (6) to desperation, (7) into a frenzy, (8) up the wall, (9) to suicide, (10) to despair, (11) round the bend, (12) around the bend, (13) out of mind, (14) on edge and (15) out of one’s mind.
38With regard to alternations, a preference for APs over PPs is also confirmed by our data. Compare drive + insane (enTT15: 2,073 / 0.13; AAM: 223 / 0.13) and drive + to insanity (enTT15: 67 / <0.01; AAM: 12 / 0.01); drive + mad (enTT15: 25 / <0.01; AAM: 32 / 0.03) and drive + to madness (enTT15: 141 / 0.01; AAM: 6 / <0,01). Interestingly enough, drive + to madness (3rd position in the rank) is more frequent as a resultative PP than drive + mad (7th position), when compared to the rest of adjectival slot fillers. The preference for APs over PPs can be also easily attested by the number of total instances in both corpora. There are 12,263 occurrences of drive with resultative APs in enTT15, but only 1,161 with PPs (1,560 and 202 instances in the AAM, respectively).
- 35 Our analysis is based on data from the largest corpus, enTT15. Frequency data from AAM will be also (...)
- 36 Standardised frequencies in AAM: 0.3 (BrE), 0.9 (AmE).
- 37 Standardised frequencies in AAM: 0.02 (BrE), 0 (AmE).
- 38 Standardised frequencies in AAM: 0.03 (BrE), 0 (AmE).
- 39 This is in line with the findings in Corpas Pastor [2021]: results for raw frequencies indicate tha (...)
39Diatopic preferences cannot be easily established because of data sparsity (in addition to other issues mentioned above). In any case, considering coverage in both corpora and normalised frequencies ≥ 0.01, some constructs with APs appear to be more typical in American English (AmE) than in British English (BrE), and vice versa, although they do not mark diatopy in a clear way35. For instance, drive + crazy (BrE: 0.16; AmE: 0.86)36, drive + wild (BrE: 0.01; AmE: 0.12) and drive + batty (BrE: <0.01; AmE: 0.06) are more frequently found in American English, while drive + potty (BrE: 0.01; AmE: 0)37, drive + mental (BrE: 0.01; AmE: 0) and drive + bonkers (BrE: 0.01; AmE: 0.01)38 are perhaps more typical of the British variety. Results for drive + mad are, to a certain extent, contradictory: according to the standardised frequencies in enTT15, it would be more frequent in the American subcorpus (0.12) as compared with the figures for the British variety (0.03). However, in the AAM the situation depicted seems to be the opposite (BrE: 0.39; AmE: 0).39 Results for constructs with PPs are even more inconclusive as both raw and standardised frequencies are comparatively much lower.
- 40 Approximate raw frequency figures were provided from enTenTen15, as this corpus appeared to contain (...)
40While drive seems to be the central matrix verb, other verbs can also fuse in the Drive Y Z construction. According to Boas [2003], this construction occurs occasionally with make and send as verbs. Findings in Corpas Pastor [2021] confirmed a frequent use of those verbs plus two more (make: >7000, get: >1000, send: >300, turn: >100)40, as well as occasional uses of the verbs set and render in this construction. The author notes that these alternative matrix verbs seem to have undergone grammaticalisation (in the case of the four frequent verbs) or semantic specialisation due to strong restrictions in the APs inventory of slot fillers (e.g., set + wild, render + insane).
41The remaining part of the discussion will be organised around those subsets of matrix verbs that can fuse in the Drive-class construction: (a) frequent, grammaticalised verbs that represent alternative ways to convey meaning akin to the drive-‘mental state’ sense (Tables 5-8): make, get, send and turn; (b) less frequent verbs that convey meaning akin to the drive-‘mental state’ sense with strong restrictions on the resultative phrases that may follow the postverbal NP (Tables 9-11): both render and set, as well as put, which has been added to the list of matrix verbs that can fuse in the Drive-class resultative construction (cf. put some on edge, see Section 3).
Table 5. Matrix verb make
|
V=make
|
Z=AP
|
Z=PP
|
|
enTT15
|
AAM
|
|
enTT15
|
AAM
|
|
bananas
|
6
|
3
|
around the bend
|
10
|
1
|
|
0 / 0
|
0 / 0
|
|
0 / 2
|
0 / 0
|
|
batty
|
8
|
2
|
into a frenzy
|
0
|
0
|
|
0 / 0
|
0 / 0
|
|
0
|
0
|
|
berserk
|
6
|
0
|
on edge
|
9
|
0
|
|
0 / 0
|
0 / 0
|
|
1 / 0
|
0 / 0
|
|
bonkers
|
8
|
2
|
over the edge
|
7
|
1
|
|
0 / 0
|
0 / 0
|
|
0 / 4
|
0 / 0
|
|
crackers
|
15
|
0
|
out of mind
|
4
|
0
|
|
0 / 0
|
0 / 0
|
|
0 / 0
|
0 / 0
|
|
crazy
|
2,152
|
345
|
out of one’s mind
|
1
|
0
|
|
16 / 56
|
3 / 2
|
|
0 / 0
|
0 / 0
|
|
dotty
|
0
|
0
|
out of one’s wits
|
0
|
0
|
|
0 / 0
|
0 / 0
|
|
0 / 0
|
0 / 0
|
|
insane
|
246
|
36
|
round the bend
|
0
|
0
|
|
3 / 4
|
1 / 0
|
|
0 / 0
|
0 / 0
|
|
mad
|
32
|
25
|
to despair
|
0
|
0
|
|
22 / 20
|
10 / 4
|
|
0
|
0
|
|
mental
|
149
|
18
|
to desperation
|
1
|
0
|
|
12 / 1
|
0 / 0
|
|
0 / 0
|
0 / 0
|
|
nuts
|
1
|
0
|
to distraction
|
0
|
0
|
|
0 / 0
|
0 / 0
|
|
0 / 0
|
0 / 0
|
|
frantic
|
45
|
3
|
to insanity
|
0
|
0
|
|
0 / 0
|
0 / 0
|
|
0 / 0
|
0 / 0
|
|
loopy
|
26
|
2
|
to madness
|
0
|
0
|
|
0 / 1
|
0 / 0
|
|
0 / 0
|
0 / 0
|
|
potty
|
5
|
0
|
to suicide
|
7
|
1
|
|
0 / 1
|
0 / 0
|
|
1 / 0
|
0 / 0
|
|
wild
|
338
|
11
|
up the wall
|
9
|
0
|
|
8 / 6
|
1 / 0
|
|
2 / 0
|
0 / 0
|
Table 6. Matrix verb get
|
V=get
|
Z=AP
|
Z=PP
|
|
enTT15
|
AAM
|
|
enTT15
|
AAM
|
|
bananas
|
4
|
0
|
around the bend
|
3
|
0
|
|
0 / 0
|
0 / 0
|
|
0 / 0
|
0 / 0
|
|
batty
|
1
|
0
|
into a frenzy
|
5
|
0
|
|
0 / 0
|
0 / 0
|
|
1 / 0
|
0 / 0
|
|
berserk
|
1
|
0
|
on edge
|
16
|
3
|
|
0 / 0
|
0 / 0
|
|
1 / 2
|
0
|
|
bonkers
|
1
|
0
|
over the edge
|
23
|
2
|
|
0 / 0
|
0 / 0
|
|
0 / 0
|
0 / 0
|
|
crackers
|
4
|
0
|
out of mind
|
6
|
0
|
|
0 / 0
|
0 / 0
|
|
0 / 0
|
0 / 0
|
|
crazy
|
128
|
12
|
out of one’s mind
|
428
|
66
|
|
3 / 3
|
0 / 0
|
|
8 / 5
|
2 / 1
|
|
dotty
|
0
|
0
|
out of one’s wits
|
0
|
0
|
|
0 / 0
|
0 / 0
|
|
0 / 0
|
0 / 0
|
|
insane
|
19
|
2
|
round the bend
|
1
|
1
|
|
0 / 0
|
0 / 0
|
|
0 / 0
|
0 / 0
|
|
mad
|
25
|
21
|
to despair
|
3
|
0
|
|
5 / 8
|
1 / 0
|
|
1 / 0
|
0 / 0
|
|
mental
|
98
|
14
|
to desperation
|
1
|
0
|
|
1 / 4
|
0 / 0
|
|
0 / 0
|
0 / 0
|
|
nuts
|
4
|
0
|
to distraction
|
1
|
0
|
|
0 / 0
|
0 / 0
|
|
0 / 0
|
0 / 0
|
|
frantic
|
6
|
0
|
to insanity
|
2
|
0
|
|
0 / 0
|
0 / 0
|
|
0 / 1
|
0 / 0
|
|
loopy
|
0
|
0
|
to madness
|
0
|
0
|
|
0 / 0
|
0 / 0
|
|
0 / 0
|
0 / 0
|
|
potty
|
23
|
5
|
to suicide
|
0
|
0
|
|
0 / 1
|
1 / 0
|
|
0 / 0
|
0 / 0
|
|
wild
|
63
|
4
|
up the wall
|
1
|
0
|
|
2 / 2
|
0 / 0
|
|
0 / 0
|
0 / 0
|
Table 7. Matrix verb send
|
V=send
|
Z=AP
|
|
Z=PP
|
|
enTT15
|
AAM
|
|
enTT15
|
AAM
|
|
bananas
|
1
|
0
|
around the bend
|
8
|
1
|
|
0 / 0
|
0 / 0
|
|
0 / 0
|
0 / 0
|
|
batty
|
1
|
0
|
into a frenzy
|
160
|
8
|
|
0 / 0
|
0 / 0
|
|
2 / 4
|
0 / 0
|
|
berserk
|
4
|
0
|
on edge
|
3
|
0
|
|
0 / 0
|
0 / 0
|
|
0 / 0
|
0 / 0
|
|
bonkers
|
7
|
0
|
over the edge
|
742
|
37
|
|
0 / 0
|
0 / 0
|
|
13 / 15
|
1 / 0
|
|
crackers
|
1
|
0
|
out of mind
|
1
|
0
|
|
0 / 0
|
0 / 0
|
|
0 / 0
|
0 / 0
|
|
crazy
|
49
|
5
|
out of one’s mind
|
3
|
1
|
|
2 / 0
|
0 / 0
|
|
0 / 0
|
0 / 0
|
|
dotty
|
0
|
0
|
out of one’s wits
|
0
|
0
|
|
0 / 0
|
0 / 0
|
|
0 / 0
|
0 / 0
|
|
insane
|
33
|
6
|
round the bend
|
6
|
0
|
|
2 / 0
|
2 / 0
|
|
0 / 0
|
0 / 0
|
|
mad
|
20
|
10
|
to despair
|
2
|
0
|
|
1 / 0
|
0 / 0
|
|
0 / 0
|
0 / 0
|
|
mental
|
20
|
1
|
to desperation
|
0
|
0
|
|
0 / 0
|
0 / 0
|
|
0 / 0
|
0 / 0
|
|
nuts
|
1
|
0
|
to distraction
|
0
|
0
|
|
0 / 0
|
0 / 0
|
|
0 / 0
|
0 / 0
|
|
frantic
|
3
|
0
|
to insanity
|
0
|
0
|
|
0 / 0
|
0 / 0
|
|
0 / 0
|
0 / 0
|
|
loopy
|
9
|
0
|
to madness
|
0
|
0
|
|
2 / 0
|
0 / 0
|
|
0 / 0
|
0 / 0
|
|
potty
|
1
|
0
|
to suicide
|
3
|
0
|
|
0 / 0
|
0 / 0
|
|
0 / 0
|
0 / 0
|
|
wild
|
64
|
3
|
up the wall
|
4
|
1
|
|
2 / 0
|
0 / 0
|
|
0 / 0
|
0 / 0
|
Table 8. Matrix verb turn
|
V=turn
|
Z=AP
|
Z=PP
|
|
enTT15
|
AAM
|
|
enTT15
|
AAM
|
|
bananas
|
0
|
0
|
around the bend
|
0
|
0
|
|
0 / 0
|
0 / 0
|
|
0 / 0
|
0 / 0
|
|
batty
|
0
|
0
|
into a frenzy
|
1
|
0
|
|
0 / 0
|
0 / 0
|
|
0 / 0
|
0 / 0
|
|
berserk
|
2
|
0
|
on edge
|
4
|
0
|
|
0 / 0
|
0 / 0
|
|
0 / 0
|
0 / 0
|
|
bonkers
|
0
|
0
|
over the edge
|
1
|
0
|
|
0 / 0
|
0 / 0
|
|
0 / 0
|
0 / 0
|
|
crackers
|
0
|
0
|
out of mind
|
1
|
0
|
|
0 / 0
|
0 / 0
|
|
0 / 0
|
0 / 0
|
|
crazy
|
44
|
4
|
out of one’s mind
|
0
|
0
|
|
1 / 3
|
0 / 0
|
|
0 / 0
|
0 / 0
|
|
dotty
|
0
|
0
|
out of one’s wits
|
0
|
0
|
|
0 / 0
|
0 / 0
|
|
0 / 0
|
0 / 0
|
|
insane
|
17
|
0
|
round the bend
|
0
|
0
|
|
1 / 1
|
0 / 0
|
|
0 / 0
|
0 / 0
|
|
mad
|
21
|
4
|
to despair
|
3
|
1
|
|
1 / 0
|
0 / 0
|
|
0 / 0
|
0 / 0
|
|
mental
|
27
|
1
|
to desperation
|
0
|
0
|
|
1 / 1
|
0 / 0
|
|
0 / 0
|
0 / 0
|
|
nuts
|
0
|
0
|
to distraction
|
0
|
0
|
|
0 / 0
|
0 / 0
|
|
0 / 0
|
0 / 0
|
|
frantic
|
0
|
1
|
to insanity
|
0
|
0
|
|
0 / 0
|
0 / 0
|
|
0 / 0
|
0 / 0
|
|
loopy
|
0
|
0
|
to madness
|
2
|
0
|
|
0 / 0
|
0 / 0
|
|
1 / 0
|
0 / 0
|
|
potty
|
1
|
0
|
to suicide
|
0
|
0
|
|
0 / 0
|
0 / 0
|
|
0 / 0
|
0 / 0
|
|
wild
|
64
|
2
|
up the wall
|
0
|
0
|
|
1 / 1
|
1 / 0
|
|
0 / 0
|
0 / 0
|
42In a similar way to the Drive-verb specific construction, other matrix verbs of the Drive-class construction also show very clear semantic selection restrictions: their resultative predicates typically denote a negative change in the mental state of Y. This type of lexical subcategorisation seems to be inherited from the higher order construction (drive-verb specific). Regarding distribution, corpus data provide less relevant information since the number of instances retrieved is much lower. This is particularly the case with PPs. In general, the Araneum proves to be far less useful than the enTenTen15, while diatopy remains rather blurred in both corpora.
43The inventory of resultative phrases seems to be more restricted and conventionalised, depending on the individual matrix verb. The Z slot appears to also play a role in this scale of schematicity and productivity: APs appear to be less restricted than PPs when it comes to licensing Drive-class constructions. Eight APs can be found in constructions with make, get, send, and turn (namely, berserk, crazy, insane, mad, mental, frantic, potty, and wild), whereas only three PPs are subcategorised by the above verbs (on edge, over the edge and out of mind).
- 41 It should be noted that turn is both a (caused) motion verb and a verb-specific construction in the (...)
44Matrix verbs can be ordered in a cline of productivity with regard to the total number of resultative phrases licensed: (1) get (25 = 13 APs/12 PPs), (2) send (14 = 4 APs/10 PPs), (3) make (25 = 13 APs + 8 PPs), and (4) turn (15 = 9 APs + 6 APs) within the Drive-class construction.41 Overall, APs tend to exhibit a greater degree of combinatory potential than PPs. As seen before, eight adjectives can appear with all four alternative matrix verbs. In addition, five adjectives (bananas, batty, bonkers, crackers and nuts) appear in constructions with three matrix verbs (make, get and send), while loopy can be found only with two of them (make and send). An analogous, albeit more restricted, situation can be found in PP resultative phrases. As stated before, only three PPs appear in constructions with the set of four alternative matrix verbs. Four PPs can only be found with three verbs: two with make, get, and send (around the bend, up the wall), and other two with get, send, and turn (into a frenzy, to despair); and two PPs with two particular matrix verbs: out of one’s wits with get and send, and to suicide, only with make and send. Finally, to madness can only be found with turn.
- 42 In the AAM the standard frequency of send + insane is 0.03 in BrE (AmE: 0).
45Interestingly enough, these four matrix verbs index diatopy to a certain extent. A cursory look at their AP slot fillers shows that the constructs with make seem to be more frequent in American English: make + crazy (BrE: BrE: 0.02, AmE: 0.31), make + insane (BrE: <0.01, AmE: 0.02), make + wild (BrE: 0.01, AmE: 0.3); make + crazy appears to be slightly more frequent in the British variety according to the enTT15, but data from AAM depicts a completely different scenario (BE: 0.05, AmE: 0.22). A similar situation is found with regard to get and turn. Compare get + crazy (BrE: <0.01, BrE: 0.02), get + mad (BrE: 0.01, AmE: 0.04), get + mental (BrE: 0.01, AmE: 0.02), get + potty (BrE: 0, AmE: 0.01), and get + wild (BrE: <0.01, AmE: 0.01); as well as turn + crazy (BrE: <0.01, AmE: 0.02), turn + insane (BrE: <0.01, AmE: 0.01), turn + mental (BrE: 0.01, AmE: 0.01), turn + wild (BrE: <0.01, AmE: 0.01). There seems to be a slight preference for constructs with send in British English: send + crazy, send + insane42, send + mad and send + wild have a normalised frequency of 0,01 in BrE (AmE: 0).
Table 9. Matrix verb put
|
V=put
|
Z=AP
|
Z=PP
|
|
enTT15
|
AAM
|
|
enTT15
|
AAM
|
|
bananas
|
0
|
0
|
around the bend
|
0
|
0
|
|
0 / 0
|
0 / 0
|
|
0 / 0
|
0 / 0
|
|
batty
|
0
|
0
|
into a frenzy
|
2
|
0
|
|
0 / 0
|
0 / 0
|
|
1 / 0
|
0 / 0
|
|
berserk
|
1
|
0
|
on edge
|
27
|
21
|
|
0 / 0
|
0 / 0
|
|
5 / 10
|
2 / 1
|
|
bonkers
|
0
|
0
|
over the edge
|
334
|
46
|
|
0 / 0
|
0 / 0
|
|
0 / 3
|
0 / 2
|
|
crackers
|
0
|
0
|
out of mind
|
22
|
5
|
|
0 / 0
|
0 / 0
|
|
2 / 1
|
0 / 0
|
|
crazy
|
20
|
2
|
out of one’s mind
|
466
|
70
|
|
0 / 0
|
0 / 0
|
|
17 / 6
|
3 / 1
|
|
dotty
|
0
|
0
|
out of one’s wits
|
1
|
0
|
|
0 / 0
|
0 / 0
|
|
0 / 0
|
0 / 0
|
|
insane
|
1
|
1
|
round the bend
|
0
|
0
|
|
0 / 0
|
0 / 0
|
|
0 / 0
|
0 / 0
|
|
mad
|
14
|
1
|
to despair
|
0
|
0
|
|
1 / 1
|
0 / 0
|
|
0 / 0
|
0 / 0
|
|
mental
|
20
|
5
|
to desperation
|
0
|
0
|
|
5 / 2
|
0 / 0
|
|
0 / 0
|
0 / 0
|
|
nuts
|
0
|
0
|
to distraction
|
0
|
0
|
|
0 / 0
|
0 / 0
|
|
0 / 0
|
0 / 0
|
|
frantic
|
1
|
0
|
to insanity
|
0
|
0
|
|
0 / 0
|
0 / 0
|
|
0 / 0
|
0 / 0
|
|
loopy
|
0
|
0
|
to madness
|
0
|
0
|
|
0 / 0
|
0 / 0
|
|
0 / 0
|
0 / 0
|
|
potty
|
3
|
0
|
to suicide
|
0
|
0
|
|
0 / 0
|
0 / 0
|
|
0 / 0
|
0 / 0
|
|
wild
|
20
|
1
|
up the wall
|
1
|
0
|
|
2 / 0
|
0 / 0
|
|
0 / 0
|
0 / 0
|
Table 10. Matrix verb set
|
V=set
|
Z=AP
|
Z=PP
|
|
enTT15
|
AAM
|
|
enTT15
|
AAM
|
|
bananas
|
0
|
0
|
around the bend
|
0
|
0
|
|
0 / 0
|
0 / 0
|
|
0 / 0
|
0 / 0
|
|
batty
|
0
|
0
|
into a frenzy
|
0
|
0
|
|
0 / 0
|
0 / 0
|
|
0 / 0
|
0 / 0
|
|
berserk
|
1
|
0
|
on edge
|
20
|
10
|
|
0 / 0
|
0 / 0
|
|
3 / 1
|
1 /
|
|
bonkers
|
0
|
0
|
over the edge
|
21
|
3
|
|
0 / 0
|
0 / 0
|
|
0 / 1
|
0 / 0
|
|
crackers
|
0
|
0
|
out of mind
|
0
|
0
|
|
0 / 0
|
0 / 0
|
|
0 / 0
|
0 / 0
|
|
crazy
|
9
|
2
|
out of one’s mind
|
0
|
0
|
|
0 / 0
|
0 / 0
|
|
0 / 0
|
0 / 0
|
|
dotty
|
0
|
0
|
out of one’s wits
|
0
|
0
|
|
0 / 0
|
0 / 0
|
|
0 / 0
|
0 / 0
|
|
insane
|
0
|
0
|
round the bend
|
0
|
0
|
|
0 / 0
|
0 / 0
|
|
0 / 0
|
0 / 0
|
|
mad
|
8
|
0
|
to despair
|
0
|
0
|
|
0 / 0
|
0 / 0
|
|
0 / 0
|
0 / 0
|
|
mental
|
13
|
3
|
to desperation
|
0
|
0
|
|
0 / 1
|
0 / 0
|
|
0 / 0
|
0 / 0
|
|
nuts
|
0
|
0
|
to distraction
|
0
|
0
|
|
0 / 0
|
0 / 0
|
|
0 / 0
|
0 / 0
|
|
frantic
|
0
|
0
|
to insanity
|
0
|
0
|
|
0 / 0
|
0 / 0
|
|
0 / 0
|
0 / 0
|
|
loopy
|
0
|
0
|
to madness
|
1
|
0 (0)
|
|
0 / 0
|
0 / 0
|
|
0 / 0
|
0 / 0
|
|
potty
|
1
|
0
|
to suicide
|
0
|
0
|
|
0 / 0
|
0 / 0
|
|
0 / 0
|
0 / 0
|
|
wild
|
12
|
2
|
up the wall
|
0
|
0
|
|
0 / 1
|
0 / 0
|
|
0 / 0
|
0 / 0
|
Table 11. Matrix verb render
|
V=render
|
Z=AP
|
Z=PP
|
|
enTT15
|
AAM
|
|
enTT15
|
AAM
|
|
bananas
|
0
|
0
|
around the bend
|
0
|
0
|
|
0 / 0
|
0 / 0
|
|
0 / 0
|
0 / 0
|
|
batty
|
0
|
0
|
into a frenzy
|
0
|
0
|
|
0 / 0
|
0 / 0
|
|
0 / 0
|
0 / 0
|
|
berserk
|
3
|
0
|
on edge
|
0
|
0
|
|
0 / 0
|
0 / 0
|
|
0 / 0
|
0 / 0
|
|
bonkers
|
0
|
0
|
over the edge
|
0
|
0
|
|
0 / 0
|
0 / 0
|
|
0 / 0
|
0 / 0
|
|
crackers
|
0
|
0
|
out of mind
|
0
|
0
|
|
0 / 0
|
0 / 0
|
|
0 / 0
|
0 / 0
|
|
crazy
|
1
|
0
|
out of one’s mind
|
1
|
0
|
|
0 / 0
|
0 / 0
|
|
0 / 0
|
0 / 0
|
|
dotty
|
0
|
0
|
out of one’s wits
|
0
|
0
|
|
0 / 0
|
0 / 0
|
|
0 / 0
|
0 / 0
|
|
insane
|
18
|
0
|
round the bend
|
0
|
0
|
|
0 / 0
|
0 / 0
|
|
0 / 0
|
0 / 0
|
|
mad
|
4
|
0
|
to despair
|
0
|
0
|
|
0 / 0
|
0 / 0
|
|
0 / 0
|
0 / 0
|
|
mental
|
9
|
1
|
to desperation
|
0
|
0
|
|
0 / 1
|
0 / 0
|
|
0 / 0
|
0 / 0
|
|
nuts
|
0
|
0
|
to distraction
|
0
|
0
|
|
0 / 0
|
0 / 0
|
|
0 / 0
|
0 / 0
|
|
frantic
|
1
|
0
|
to insanity
|
0
|
0
|
|
0 / 0
|
0 / 0
|
|
0 / 0
|
0 / 0
|
|
loopy
|
0
|
0
|
to madness
|
0
|
0
|
|
0 / 0
|
0 / 0
|
|
0 / 0
|
0 / 0
|
|
potty
|
0
|
0
|
to suicide
|
0
|
0
|
|
0 / 0
|
0 / 0
|
|
0 / 0
|
0 / 0
|
|
wild
|
2
|
2
|
up the wall
|
0
|
0
|
|
0 / 0
|
0 / 0
|
|
0 / 0
|
0 / 0
|
46The second subset of alternative matrix verbs that license the Drive-class construction are put, render, and set. Set and render can be considered causative verbs which imply a change of position/physical place (caused motion). Like put, set is also a highly polysemous verb. These three verbs show a progressive semantic specialisation within this construction, which is particularly noticeable through their highly conventionalised preferences for a handful of resultative phrases, and vice versa.
47Compared with the former group (matrix verbs make, get, send, turn), only a subset of APs can combine with the verbs in the second group (set, render and put). Five adjectives can combine with these three verbs (berserk, crackers, insane, mad and wild); dotty and nuts can be AP slot fillers in constructions with put and render, potty can occur with put and set; and finally, dotty only collocates with render. A similar progressive cline of restrictions is observed with regard to the resultative PPs: only two (on edge and over the edge) can combine with both put and set within the construction, and the other phrases can combine with just one of these three verbs: three with put (into a frenzy, out of mind, up the wall), one with set (to madness) and one with render (out of one’s minds). This would place put as the most productive matrix verb of this second subset (13 = 8 Aps + 5 PPs), followed by set (9 = 6 APs + PPs) and render (9 = 8 APS + 1 PPs), the latter being selected by only one PP. As in the case of the first subset, PPs exhibit a relatively lower combinatory potential than APs.
- 43 Standardised frequencies in AAM: 0 (BrE), 0.22 (AmE).
- 44 Standardised frequencies in AAM: 0.03 (BrE), 0.11 (AmE).
48Due to the low number of instances found in both corpora, it is not possible to draw conclusions about diatopic differences within constructions licensed by this second subset. However, there are some interesting findings. Although constructs like put + wild (BrE: <0.01/ AmE: 0) and and put + out of mind (BrE: 0.03/ AmE: <0.01) seem to appear slightly more frequently in the British variety, some other constructs with put and set index primarily American English: set + mental (BrE: 0/ AmE: 0.01), set + wild (BrE: 0/ AmE: 0.01), set + over the edge (BrE: 0/ AmE: 0.01), set + on edge (BrE: <0.01/ AmE: 0.01), put + mental (BrE: <0.01/ AmE: 0.01), and, more specially, put + over the edge (BrE: 0/ AmE: 0.02)43 and put + on edge (BrE: 0.01/ AmE: 0.06)44.
49Finally, Figures 2-3 summarise the combinatorial preferences of matrix verbs and resultative phrases in the Drive-class construction. White cells indicate that the combination of matrix verb and resultative phrase can be found in both corpora (enTT15 and AAM). Cells in light blue indicate combinations which have been found in only one corpus (generally, enTT15), and cells in dark blue mean that the combination is not found in any of the corpora used in this study
Figure 2. Combinatory potential: matrix verbs and Aps
Figure 3. Combinatory potential: matrix verbs and PPs
50The final stage in the methodology deployed in this study was intended to unveil the productivity of this type of resultative construction by exploring novel slot fillers, as well as their combinatory potential. To this end, six pattern-based search sequences were launched (see Section 2). CQL searches 3, 5 and 6 have not retrieved any relevant data. That means that the inventory of resulting APs in Table 3 is exhaustive, i.e., no additional adjectival slot fillers for the Drive-class construction have been found in the corpora. It also means that there are no novel PPs in the patterns [preposition + determiner + noun] and [preposition + possessive + noun] (e.g., round the bend, out of one’s wits).
- 45 The most frequent PPs found in the two corpora are to destruction (EnTT15: 95; AAM: 6), to depressi (...)
51However, new prepositional resultative phrases have been retrieved by CQL search pattern 4 [preposition + noun] (e.g., to madness). The ‘drive-insanity’ resultative appears to be very productive, according to the number of new slot fillers retrieved. PPs stick systematically to the pattern with preposition to and nouns that denote a (typically) negative mental state or emotion: to destruction, to depression, to frustration, to dementia, to delirium, to hatred, to cruelty, to anger, to fury, to repentance, to grief, to idiocy, to distress or even to frenzy (cf. into a frenzy), among others45. Interestingly enough, these PPs tend to fuse only in the construction with the matrix verb drive, with the Drive-class Yanimate notable exception of put + to grief, with 71 occurrences in the enTT15 corpus. It is worth mentioning that novel PPs seem to combine freely with those in Table 3, as can be seen below (ex. 17):
(17) “This bill bans non-scientific ‘therapies’ that have driven young people to depression and suicide,” the governor tweeted. (AAM)
52This could be interpreted as indicative of a lower-order verb-specific drive construction, a dominance of the PPs with to as a subpart link, and the interplay between the Drive-class construction and the verb-specific drive construction. This scenario of high numbers of novel slot fillers also correlates with a higher degree of schematisation and a general intensifying function of the construction. This could explain the large number of instances of drive + to success in both corpora (enTT15: 456; AAM: 30), and other nouns with similar positive connotations (to victory, to perfection, to excellence, etc.).
53The construction appears to be even more productive with regard to the verbal slot fillers. CQL query 1 looked into new verbs that fuse in the construction with resultative APs. Novel verbs found in corpora tend to appear in combination with central, frequent adjectival slot fillers: mainly with crazy, and, to a lesser extent, insane, mad, and wild. There are no instances with other APs in Table 3. Novel verbs with crazy show delexical meanings and grammaticalisation (bring, do, generate, cf. ex. 18), or, else, enriched construal meanings which also exhibit a certain degree of intensification (ex. 19): they incorporate caused-motion resultative meaning (sprint, push) and the drive-‘mental state’ meaning (mess, jangle, shit). Particularly numerous are the instances found in enTT15 with verbs related to sex (ex. 20), which also convey additional emphasis (rub, sex, fuck, excite). Verbs that fuse in constructions with insane and mad are less numerous and fall in one of the semantic groupings mentioned above (ex. 21): delexical (bring/generate + insane), enriched construal meanings (strike + mad, run + mad, scare + insane) or sexual connotations (suck + mad, wank + wild, fuck + wild).
(18) I have one with a minimal bottom shelf (like an inch and a half off the floor). And oftentimes it holds my dogs. It generates me crazy, but for reasons unknown they love to lay on it. (enTT15)
(19) Trey lets him go, then starts interjecting chords and rhythms that just push Page crazier and crazier. (enTT15)
(20) “Peter, I’ll be thinking of you throughout, even if it does happen like we have planned and Ray is on top of me on the bed in the motel room, fucking me crazy. (enTT15)
(21) When the Furies struck you mad you couldn’t deal with your pain rationally, when you got your sanity back you needed to be strong for your mother. (enTT15)
54They all convey intensification, mainly through the adjective slot filler. In some cases (ex. 22), the typical resultative AP fuses as a separate construction into other constructions, inheriting the construal intensification aspect, akin to degree adverbials (‘in a high degree’, ‘very much’).
(22) Than most of the old timers not help [sic] should have loved him so well still though fascinated him mad and that trading thoughts already investment funds despite her concerns and turned back and forth in front of the mirror. (enTT15).
55Another interesting finding with regard to constructs with verbs denoting sexual intercourse is that they license inanimate Y (patient/experiencer) whose construal interpretation involves coercion as animate. This is particularly the case of (usually male) sexual organs and the anaphoric pronoun it (ex. 23).
(23) Her silky soft feet are all wrapped around his juicy tool, wanking it wild until it’s hard. (enTT15).
56CQL 2 search pattern revealed the verbal productivity of the construction with PPs in the Z slot. The resultative phrases with higher combinatory potential are over the edge, to madness, to despair, to suicide, to distraction, to insanity, and to desperation. They tend to favour verbal slots around similar semantic groupings as the AP resultatives discussed above: some delexical or grammaticalised verbs (take, bring), a majority of caused-motion verbs (push, shove, draw, drag, nudge, move, steer, lead, stir, tip, bring, etc.), as well as causative verbs that convey a certain degree of manipulation (force, coax, compel, impel), verbs which focus on the construal meaning ‘drive-mental state’ as a deliberate act of annoyance (goad, tease, tempt, sting, annoy). See ex. 24- 29 below.
(24) The harshest challenges can lead one to despair, suicidal thoughts, and insanity. (enTT15)
(25) What tipped her over the edge though was the incestial relations between him and Elaine, whom she believed to be Natasha, only to see "Natasha" pursue someone else at the ball. What exactly was her mental state that day when she killed most of the royal family with a poisoned dagger? (enTT15).
(26) Even after hearing the music so many times, Maggie’s performance at The Current moved me to distraction. (AAM)
(27) But now, in his extremity, he turned upon him, presenting the enormity of his sin and the hopelessness of pardon, that he might goad him to desperation. (enTT15).
(28) In India, they said that the cotton that had been genetically engineered would provide 1,500kg per acre. But the company, after lying to farmers, pushing them to suicide, had to admit that it is only 500 kg per acre. (enTT15)
(29) This journey of delicate indiscretions, lost dreams and brutish actions leads Blanche to madness, aided by her handsome, masculine brother-in-law, Stanley Kowalski. (enTT15)
57While the construction seems to be extremely productive with regard to the verbal slot fillers that can fuse with certain resultative PPs, truth is that their combinatory potential seems to be even more conventionalised than the instances with resultative APs. Not all these novel slot fillers exhibit a similar combinatory potential: lead and push fuse in the Drive-class constructions with all those resultative PPs, tilt, trip, nudge, pull or carry only collocate with over the edge, while goad can combine with over the edge, to desperation and to madness, but no instances have been found with to despair, to distraction, to insanity or to suicide. Another example is the verbal slot filler bring: it combines with over the edge, to despair, to distraction, to madness and to suicide, but no instances have been found with to distraction or to insanity.
58Restrictions also operate at the level of the PPs. By way of example, to desperation tends to have only bring, lead and push, while over the edge can take far many more verbal fillers than the other resultative PPs. The construal meanings of the verbal slots that fuse in the construction with over the edge tend to be coerced by this particular PP through a process of grammaticalisation, intensification and pragmatic specialisation (semantic prosody). Connotations related to intense feelings and emotions are usually negative (emotional distress), as Y (patient slot) is so irritated or unsettled that he or she loses their self-control, gets absolutely worked out and/or behaves in an extreme way (ex. 25, 30 and 31).
(30) Homicide detective Kay Griffith never had a problem controlling her anger—until murder suspect Tommy Rayne finally pushes her over the edge. After assaulting him, she’s suspended from her job. (AMM)
(31) One of his former TV producers in the Middle East, Serene Sabbagh, resigned from Fox recently because of its “bias and racism”. What tipped Sabbagh over the edge was the bombing of Qana. (enTT15)
59However, this is not necessarily always the case. There are many instances of this construct with over the edge and slot fillers like push, take, tip, tilt, bring, where the physical symptoms of strong feelings are directly associated with intense (almost uncontrolled) sexual pleasure (ex. 32-33):
(32) I thruster my hips a couple of times to bring me over the edge as my cock started shooting hot baby juice. (enTT15).
(33) He was now on the verge of ejaculating himself, my own orgasm acting as the impetus to take him over the edge. (enTT15).
60Finally, intensification is also a shared feature of all the constructs retrieved through CQL query 2 (cf. ex. 24-34). An interesting case is to distraction. This resultative PP exhibits a wide range of potential slot fillers to form constructs that convey intensification, similarly to degree adverbials (cf. APs mad or crazy above). See ex. 34:
(34) Society or culture may pressure us to marry “at least once.” Our family and friends may pressure us to distraction to marry. (For those who are divorced this is even harder to deal with.) (AAM).
- 46 Cf. also other intensifying constructions with the pattern to Npl: to pieces, to bits (e.g. love so (...)
61Some more instances found in both corpora are “love her to distraction”, “torments me to distraction”, “work myself to distraction”, “annoy me to distraction” (AAM); and “become addicted to distraction”, “were plagued to distraction”, “be familiar, almost to distraction”, “bore us to distraction”, “be tempted to distraction”, “pushed to distraction”, “was obsessed to distraction” (enTT15), among many others.46
62This paper contributes to the radical change of paradigm brought about by computational and corpus-based approaches to the study of phraseology and idiomaticity. It also contributes significantly to the theoretical strand it adheres to: Construction Grammar. A corpus-based methodology has been developed in order to study a type of resultative construction: the Drive Y Z construction. To this end, we have retrieved instances semi-automatically using a script and six CQL search queries. Our heuristic protocol has allowed us to establish the combinatory potential of base slot fillers and to uncover novel, productive instances of this base construction.
63Our findings confirm our initial hypothesis. Exemplars like send someone up the wall, drive someone insane, love some to distraction, fascinate someone mad, lead someone to suicidal thoughts, which have traditionally been classed as idioms, collocations or free-word combinations are, in fact, instances of the same type of construction. The metaphoricity, collocability, semantic coercion, grammaticalisation, intensification and productivity observed in all exemplars analysed can be traced back to a complex network of higher-order and lower-order constructions within the Drive-class Yanimate Zstate construction. To put it short, the Verb-specific drive construction stems from and contributes to the higher-order Resultative construction Vcaused-motion Y Z), and, at the same time, represents the starting point of the Drive-class construction.
64The notion of central slot fillers is also a powerful theoretical concept that helps explaining inheritance links with associated lower-order and higher-order constructions. Central slot fillers fuse with novel slot fillers in Resultative constructions in a recursive fashion through instance and part links, which also explains combinatory potential and restrictions. This process appears to be highly conventionalised, as only particular collocational and pattern preferences are licensed. This intricate network of constructions can also explain productivity, idiomaticity, variability and language change.
65Our study also reveals the importance of using large-scale data. Giga-token corpora are needed to retrieve meaningful results. Insufficient data provide less significant findings. In particular, our results demonstrate that big data enables more detailed and precise description of the construction in question with all its variations, restrictions, and mutual interplay between its components, listing the sets of fillers found in the corpora, summarising findings on productivity and diatopy. However, to handle such extensive data effectively it has to be adequately retrieved and processed. In this paper, a protocol for semi-automatic data retrieval and processing has been introduced and illustrated. This is another substantial contribution of this study.
66Given the above findings, we conjecture that the differences observed in Boas [2003] as compared to Corpas Pastor [2021] could be due to language change, but also most probably to the volume of data used. In a similar fashion, the way data are processed is also relevant. Unlike Corpas Pastor [2021] (and Boas [2003]), results in this study have been retrieved semi-automatically (with the exception of a limited number of manual counts), which has made the study of very large volumes of text possible.
67This paper also confirms that larger corpora provide more (and better) quality findings than smaller corpora, even in the case of giga-token corpora. In this respect, the enTenTen15 corpus has provided more informative results than the Araneum Anglicum Maius. However, even larger corpora are needed to study diatopic variation. While some indicative findings have been reported, the number of tokens included in a huge corpus like enTenTen15 remains largely insufficient to establish relevant differences between English varieties. Another issue is coverage of data in giga-token corpora. For instance, the available corpus data suggest that some constructs within the Drive-class seem to convey intense sexual connotations. However, it is not clear whether these results could be biased, as a result of different document selection criteria among corpora.
RSS feed
