1. Introduction

1Goldberg’s (1995) famous sneeze the napkin off the table example has by itself arguably come to epitomize Construction Grammar (CxG) over the years, and more particularly the constructionist approach to argument structure. It is an example that perfectly illustrates the fact that verbs may occur with arguments that are not subcategorized by the verb itself, thereby making a strong case for the existence of meaningful abstract constructions, namely Argument Structure Constructions (ASCs), that exist independently from verbs and may fuse with verbs to license creative uses. This initially led Construction Grammarians to largely focus on creative examples, even if authentic, rather than their more typical uses (Bergs 2008). As Hanks (2008: 228) rightly points out, though, “authenticity alone is not enough: evidence of conventionality is also needed”. By contrast, a more recent trend of corpus approaches has sought to provide more exhaustive accounts of constructions, thereby shifting the attention to their conventional uses (e.g. Gilquin 2010 on the causative construction; Perek 2018 on the way construction). While several constructions have since received extensive corpus-based descriptions, this is much less the case of the caused-motion construction. One reason is probably that it does not lend itself well to automatic extraction, as not a single word is lexically specified, and that even in a parsed corpus, there is no syntactic structure that allows extraction with high precision and recall (Hwang 2014). A comprehensive approach would therefore require a careful reading of data, which does not only go against the central idea in corpus linguistics of some automated process, but is also highly time-consuming and, admittedly, boring. Carefully reading a novel, while still time-consuming and not automated, is, however, arguably far from boring.

2This paper therefore aims to take a perspective that is at least partly more comprehensive in nature by examining the caused-motion construction in the first instalment of the Harry Potter series by reading through and annotating an electronic version of the book. As the analysis is limited to a single book, genre, and author, there is no claim to a representative account of the caused-motion in English, but the comprehensive account of the construction in one book makes it possible to give more due consideration to its conventional uses and to offer some appraisal of the features of the construction. In addition, Harry Potter constitutes a good case study for the caused-motion construction for two reasons. Firstly, children’s literature (Nodelman 1992: 190), and Harry Potter in particular (Nikolajeva 2003: 134), is characterized by action-oriented prose, which suggests considerable potential for the expression of caused motion, at least in a concrete sense. Secondly, “[a]s literature, indeed all verbal art, consists in an elaboration, an intensification and creative exploitation of the inherent qualities of everyday language” (Fischer & Nänny 1999: xxvi), it also gives us a chance to see the ways in which the construction is used in a creativity-prone context.

3This paper is structured as follows. Section 2 provides a literature review about the caused-motion construction, considering how it has been defined and studied in the literature and pointing out remaining blind spots. Section 3 describes the data and methods adopted in this study. Section 4 presents the results, providing first some distributional information about the construction and then zooming in on its metaphorical uses, before concluding in Section 5.

2. The caused-motion construction: definition and state-of-the-art

4The caused-motion construction can be represented as in (1). In term of form, it is made up of a complex transitive structure with a direct object and an oblique locative complement. In terms of semantics, in its central sense, it expresses the meaning of a cause X causing a theme Y to move to or along a particular goal or path Z (Goldberg 1995: 152).

5The construction is most strongly associated with the high-frequency verb put, which is considered the path-breaking verb for the construction, i.e. the verb that inherently expresses caused motion (as in the unremarkable example To put the vase on the table), and from which the construction derives its central sense. Empirical evidence also suggests that it is the most frequent verb to appear in the construction (Goldberg et al. 2005; Perek & Lemmens 2010).

6On the grounds that constructions exhibit polysemy in the same way that lexical items do, Goldberg (1995:161–162) distinguishes five different senses of the construction that are tied to different verbs:

1. Cause to move: the central sense where X causes Y to move (e.g. to push the box off the stairs)

2. Condition of satisfaction: Y moves only if the condition of motion instigated by X is satisfied (e.g. to invite someone into the room)

3. Prevent: Y is prevented from moving (e.g. to lock someone into the room)

4. Help: X helps the motion of Y without directly causing it (e.g. to guide someone to the platform)

5. Enable: X enables the motion of Y without directly causing it (e.g. to allow someone into the room)

7While her analysis focuses largely on prototypical examples that are concrete in nature, it is a template that allows for metaphorical extensions as well, which are typically the locus for polysemy that extends to temporal or other abstract domains. One metaphorical extension that has received considerable attention is the ‘transfer caused-motion’ (e.g. give a piece of land to one’s son), which is based on a metaphor that conceives transfer of ownership as physical transfer (Goldberg 1995: 111). As metaphorical extensions can themselves involve additional extensions, this sense is further used metaphorically with verbs of communication (e.g. say something to someone, tell the truth to someone), which hinges on the ‘conduit metaphor’ that construes communication as an act of transfer (Reddy, 1979). Alongside these semantic extenstions, Goldberg (1995) also stipulates a number of semantic constraints that apply to the construction, such as (i) the impossibility for a cognitive decision to mediate between the causing event and the motion (e.g. * to persuade him into the room), or (ii) contrary to the ditransitive construction, the impossibility of drawing on the metaphor that construes “an action directed at a person as an object given to the person” (Goldberg 1995: 94) (e.g. to throw someone a glance vs. * to throw a glance to someone).

8Most subsequent studies have focused on further substantiating some of the analyses of the construction (e.g. Baicchi 2011), on challenging some of the constraints on the construction posited by Goldberg (e.g. Peña Cervel 2009), or questioning some of the broader mechanisms of Construction Grammar as proposed by Goldberg, taking the caused-motion as their object of study (e.g. Van der Leek 2000; Hampe & Schönefeld 2003; Torre 2017). For instance, Peña Cervel (2009) focuses on the verbs persuade, encourage, convince and instruct to demonstrate that the construction can in fact occur with verbs of cognitive decision, while Hampe & Schönefeld (2003) drew, among others, on Harry Potter material, to illustrate their claim that creative uses involve a great deal of low-level iconicity rather than the fusion with a highly abstract construction.

9Two corpus approaches to the construction deserve special mention. Hwang (2014), in particular, examines large amounts of corpus data and elaborates a semantic typology of the construction that is compatible with her data. However, the analysis excludes some of the senses of the construction, and the corpus query does not allow for an exhaustive extraction of the construction. In addition, the main goal is to contribute to an effort in NLP to automatically extract the construction. The second is that of Goldberg et al. (2005), which investigates the cue validity for the caused-motion of the ‘Verb Object Locative’ syntactic pattern and thereby provides a quantitative overview of the instances that are related to the caused-motion.

10All in all, no study has sought to sketch an exhaustive usage-based profile of the construction in the vein of the corpus studies in Construction Grammar reported above, though. In addition, there has been a clear propensity in the literature to focus (i) on (invented) examples that express concrete caused motion, and/or (ii) on invented or authentic creative instances of the construction. As a result, many blind spots remain, such as the range and distribution of verbs that occur in the construction, or the distribution between the different senses, or between the literal and metaphorical instances. While it is beyond the scope of this paper to comprehensively address all these questions, this paper seeks to take a step in that direction through an exhaustive examination of the caused-motion in the first Harry Potter book.

3. Data and methods

11The study relies on an electronic version of the first instalment of the series, Harry Potter and the Sorcerer’s Stone, which is made up of 78,434 words. The text was processed with the online CATMA tool (Version 6.3, Gius et al. 2021), which is a freely accessible online interface designed to let users manually annotate text based on a customized annotation scheme. This allowed me to read through the text and mark each (potential) instance of the caused-motion construction, which were then exported as a concordance in Excel for further analysis.

12Following Goldberg’s (1995: 152) definition of the construction, the caused-motion was defined as a complex transitive structure made up of a direct object and an oblique that is realized as a locative complement, either as a prepositional phrase or an adverbial. In terms of semantics, all instances corresponding to the five senses distinguished by Goldberg were retained, including their metaphorical instances1. Yet, as corpus data tends to elude neatly theoretically defined constructs, and especially since the caused-motion shares some of the form and semantics of the verb-particle and resultative constructions (cf. Goldberg 2014), it was necessary to elaborate criteria to objectively and consistently select what was to count as an instance of the caused-motion. The criteria were the following: (i) All instances that allowed the adverbial particle to come before the verb were considered instances of the verb-particle construction and thus excluded from the analysis (e.g. take one’s cloak off). As a result, almost all oblique arguments are prepositional phrases, notwithstanding a few exceptions (e.g. poke one’s head inside). (ii) Obliques expressing a resultant state rather than a (metaphorical) goal or path were considered instances of the resultative construction and thus excluded (e.g. set affairs in order). (iii) Instances with obliques that express a source argument were generally excluded, with some exceptions. It became apparent from the data that the line was sometimes thin between a source and a path argument depending on the choice of the preposition, as illustrated in examples (2) and (3):

13While the oblique in (3) could be interpreted as a source rather than a path as in (2), the few cases like those in (3) were included by analogy with the former cases.

14Applying these criteria to the data yielded a total of 544 instances of the caused-motion construction. The data were then further coded for (i) the verb lemma, (ii) the sense following the taxonomy outlined above, and (iii) literal vs. metaphorical use. This was complemented by qualitative observations about semantics or idiomaticity when relevant.

4. Results

4.1. A quantitative overview of the caused-motion

15Thanks to the exhaustive annotation of the construction, we can get a sense of the actual usage of the construction by considering a number of quantitative dimensions, namely the distribution of (i) the verbs that occur in the construction, (ii) the senses of the construction, and (iii) the literal vs. metaphorical uses.

16For the 544 tokens, the verb-slot of the construction is instantiated by 161 verb types. Of those, 89 types (i.e. 55.3% of all types) are hapax legomena. Although such a high proportion of hapaxes is likely due to the small size of the corpus, it is nonetheless indicative of a highly productive verb-slot. Table 1 lists the verb that occur with a minimum frequency of five, while Table 2 exemplifies some hapax legomena, most of which are verbs of motion (e.g. grab, expel, dump, lift, swing), although others belong to different verb classes with, most notably, several verbs of verbal communication (e.g. tell, yell, murmur). As can be seen from Table 1, the top 3 verbs are the versatile high-frequency verbs take, put, and get. Contrary to expectations, take is the most frequent verb in our data set, while put, which has been reported as the most frequent verb to occur in the construction in ICE-GB (Perek 2009, in Perek & Lemmens, 2010), here only arrives in second place after take. More strikingly, take and put respectively account for a mere 8% and 7% of the data, while Perek (2009) found put to account for 24% of the caused-motion instances in his study. While this discrepancy may be due to a largely automated and thus less exhaustive account of the construction in Perek (2009), it further testifies to the diversity of verbs used in the construction here, which is in all likelihood also in part an effect of the literary genre wherein lexical diversity is an intrinsic part of the art. In fact, even the high-frequency verbs are used in a range of different senses that introduce diversity in their use. Next to concrete and prototypical uses of the caused-motion (e.g. take this boy to the hospital wing, put the money into a small leather pouch, get your letters to you), they are also found in a range of metaphorical instances (e.g. take a drink from the bottle, put the cat out of his mind, get them out of trouble).

17In terms of semantics, Figure 1 shows the distribution of the data across the five senses of the construction with 95% confidence intervals. Strikingly, the central sense of ‘cause to move’ takes the lion’s share (80.1%), followed by the ‘Prevent’ (11.6%), ‘Help’ (4.2%), ‘Condition of satisfaction’ (2.6%), and finally ‘Enable’ (1.5%) senses.

Figure 1. The distribution of the five senses of the caused-motion

18Each instance was further coded for literal or metaphorical meaning. An instance was considered metaphorical if the theme and/or path or goal were abstract and could therefore not entail actual motion (e.g. drawing attention to herself), or if the entire instance was an idiomatic expression (e.g. put just one toe out of line). Figure 2 represents the distribution of the metaphorical vs. literal uses in the data, which shows that literal uses account for 71% of the data against 29% of metaphorical uses. Since metaphorical uses are generally underrepresented in the literature on the construction, the next section will zoom in on these uses in particular.

Figure 2. Literal vs. metaphorical uses of the caused-motion

4.2. A qualitative look at the metaphorical uses of the caused-motion

19A close inspection of what lies under the hood of metaphorical uses reveals a range a different phenomena. First of all, as also noted by Hwang (2014: 59–60), metaphoricity takes the form of a gradient, with some instances being arguably more metaphorical than others. Consider the following examples in which abstract entities have been underlined:

20In (4), only one of the theme or path/goal are abstract entities, while in (5), both are abstract. Somewhat more different in nature still are conventionalized idiomatic expressions as those in (6), which abound in the data. Note how even among idiomatic expressions, it is possible to retain some actual motion, as in (6a-b): example (6a) involves the motion of Harry by means of pulling, but not the pulling of Harry towards his feet, but rather the pulling up of Harry so that he stands on his feet. By contrast, the other instances in (6) are arguably a notch more metaphorical as they do not involve any concrete motion, and most instances involve at least one abstract entity.

21Secondly, some metaphorical instances can be grouped as belonging to the same semantic field, testifying to some preferential patterns of metaphorical extensions. Two such groups are particularly prominent in the data. The first set relates to the semantic field of verbal communication and is a pattern used to report speech, as can be seen in the examples in (7). In a book of fiction that is rife with dialogues, this is perhaps not unexpected. These instances draw on the ‘transfer caused motion’ extension, and more particularly the ‘conduit metaphor’, viz. the metaphorical template via which utterances are construed as objects that are moved towards the hearer. A total of 40 instances are found in the data, with as many as 20 different verb types. The two most frequent verbs are conventionally tied to the pattern of reporting (e.g. say, tell), but the pattern is also found with verbs of manner of speech that are used rather creatively as can be seen in (7b-d).

22The second group, made up of 19 instances for 8 verb types, pertains to the semantic field of visual perception and is exemplified in (8). Although not typically discussed in relation with the caused-motion construction, this group can be said to fall under the broader label of ‘visual motion events’, which can be defined as “event[s] in which the gaze of an agentive perceiving experiencer is conceptualized as something emanating from an experiencer’s eye(s)” (Cappelle 2020: 238). Although the focus of visual motion events tends to lie on intransitive uses of verbs followed by locative prepositional phrases (e.g. to look over something, to stare across the cafeteria), the instances with the caused-motion reify the gazing event as an object far more strongly by means of the deverbal nouns of visual perception (a glance, a look) that function as the themes of verbs of motion (throw, cast, take).

23Putting this in relation with the constraints that Goldberg (1995) has posited for the caused-motion, examples (8a-c) are particularly interesting. All three are examples of light verb constructions that draw on the metaphor that construes “an action directed at a person as an object given to the person” (Goldberg 1995: 94). As was detailed in Section 2, Goldberg in fact argues that this metaphorical extension is not possible with the caused-motion construction on the grounds that an example like *to throw a glance to someone is ungrammatical. While the to preposition does not occur in this configuration in the data, several instances, with different verbs, are found with the directional preposition at. In addition, several other light verb constructions drawing on this metaphor, but not in the context of visual perception, occur in the data (e.g. make a call to, give attention to), thereby further challenging this constraint.

24Although the present discussion of metaphorical uses has focused on only the most salient aspects that the data reveal, it has hopefully brought to light some of the properties of metaphoricity involved in the use of the caused-motion, thereby also showing what is to be gained from a usage-driven approach to the description of constructions.

5. Conclusion

25This small-scale study has sought to offer some novel insights in the use of the caused-motion construction by taking an exhaustive approach to the construction in one piece of text, namely Harry Potter and the Sorcerer’s Stone. This has made it possible to pay due attention to the typical uses of the construction as found in authentic data and to provide a quantitative perspective on several dimensions of the construction, like the distribution of verbs found in the construction or the distribution of its different senses. It goes without saying that focusing on a single book by a single author largely limits the generalization of the present quantitative findings as they are certainly biased to some extent by the genre of (children’s) literature and the author’s idiolectal preferences. Yet, from a quantitative perspective, it has at least shown the productive potential of the construction given the wide range of verbs found in such a small dataset, and given some order of magnitude of the distribution of the different senses of the construction and of the literal vs. metaphorical uses. As to the qualitative analysis of the metaphorical uses, it has illustrated the range of phenomena that metaphorical uses can encompass, namely (i) the gradient that metaphorical uses take, and (ii) the different kinds of metaphorical extensions that are typical of the caused-motion, that is, many highly conventional idiomatic expressions alongside two more specific prominent semantic groups. Finally, relying on authentic material has also offered evidence that challenges some intuition-based constraint that had been posited, underlining the importance of verifying such claims empirically. All in all, it is hoped that this study has shown what can be gained from a more exhaustive and usage-based approach to the caused-motion construction and that it may prompt a more wide-scale and representative study in the future.