Juho Ruohonen and Juhani Rudanko: Infinitival vs gerundial complementation with afraid, accustomed, and prone: Multivariate corpus studies

Ruohonen and Rudanko present a collection of five detailed corpus-based case studies, along with an introduction and a conclusion. The studies focus on variation and change in the domain of English non-finite complementation, and more specifically on the factors underlying the variation between to-infinitives and gerundial complements, as in examples (1) and (2).

In line with variationist methods, the authors carefully delimit “the data to only those contexts that are functionally parallel” (Tagliamonte 2012: 10) before using statistical methods to determine the relative importance of different factors explaining the variation (Tagliamonte 2012: 11). In the case studies, this restricts the data to lexical predicates that select both the infinitival and gerundial complement patterns, and further to those tokens that involve subject control, i.e. co-reference between the matrix subject and the understood subject of the complement clause. The use of multiple regression analysis to examine the relative weight of various factors constraining the variation is argued by the authors to set apart the collection as “the first book-length study of non-finite complements in English with such an orientation” (p. 3). Table 1 provides a basic summary of the data focus of the five case studies.

Chapter 3 is the only chapter that starts from a particular complement pattern. In this chapter, the authors zoom in on degree complement constructions as in (3) (complements licenced by a degree modifier like too rather than by an adjectival predicate) to discuss the sometimes thorny issue of distinguishing these from to-infinitival complements where both are possible. The other four chapters focus on particular adjectival complement-taking predicates: afraid, accustomed and prone. The three predicates are characterized by the authors as belonging to different semantic classes, following Quirk et al. (1985: 1226–1230): afraid is characterized as a dispositional or modal adjective, which involves “the attribution of potential or propensity for an entity to act or function in a certain way”; accustomed as an aspectual adjective, which ascribes to the matrix subject “tendencies and recurrent patterns of behavior”, and prone combines these modal and aspectual semantics (p. 7).

In what follows I will give a brief description of a selection of factors constraining the variation between to-infinitival and gerundial subject-controlled complements in the context of the same predicate. After that I will give a summary of each case study before ending with an overall evaluation. A first explanatory factor relates to the different semantics of the two complement patterns. Besides the semantics of specific lexical predicates, the main factor focused on in the book is the Choice Principle, which predicts that a different non-finite complement pattern will be preferred depending on whether the understood subject of the complement clause has an agentive role [+Choice] as in (1) or a non-agentive role [−Choice] as in (2) in the event described by the complement clause. The prediction, borne out in (1) and (2), is that the infinitive will be preferred in [+Choice] contexts, and the gerundial clause in [−Choice] contexts (e.g. Rudanko 2017: 20).

A second set of features has traditionally been grouped under the so-called Cognitive Complexity Principle, which states that “more explicit grammatical alternatives tend to be preferred in cognitively more complex environments” (Rohdenburg 1996: 149). The central idea is that certain linguistic contexts come with a high processing cost, and induce the speaker to choose the most explicit, or transparent, linguistic pattern. Infinitival and gerundial clauses are non-finite forms lacking obligatory subject expression and tense/mood markings and can therefore both be seen as less explicit than finite complement clauses. At the same time, however, the diachronic rise of gerundial clauses, often at the expense of earlier to-infinitival patterns, has been shown to be delayed significantly in contexts of complexity (e.g. Vosberg 2003). Gerundial clauses are also less explicitly recognizable as clausal complements due to their external (and sometimes also internal) nominal behaviour (De Smet 2010). Based on such considerations, Rohdenburg (2016: 472, cited on p. 84) argues that to-infinitival complement clauses can indeed be considered the more explicit variant in comparison to gerundial clauses. Syntactically complex environments then include cases where phrases are extracted from the complement clause, passive constructions, cases with insertions between the main and the complement clause, or with longer constituents. An important complexity factor focused on in the book is the Extraction Principle, which states that the infinitive will be preferred over the gerundial clause “in environments where a complement of the subordinate clause is extracted (by topicalization, relativization, comparativization, or interrogation etc.) from its original position and crosses clause boundaries” (Vosberg 2003: 308). The authors consider cases of adjunct extraction as in (4) besides complement extractions.

A third type of explanation is the so-called ‘Horror Aequi’ principle (e.g. Rohdenburg 1996: 175, fn. 2; Vosberg 2003), i.e. the avoidance of formally (near-)identical grammatical expression types in near proximity. A possible example would be the avoidance of two consecutive to-infinitives as in to be afraid to go home. Besides semantic, syntactic complexity-related and stylistic factors, extralinguistic factors have also been discussed, albeit to a lesser extent, in the literature. In the book, the impact of text type, regional variety of English, and inner versus outer circle varieties of English is taken into consideration.

This variety of relatively well-established factors forms the basis for the investigations and further elaborations of explanatory principles in the book. Because the book consists of separately published articles, these factors are (re-)introduced in different places, and it is perhaps not straightforward for the unacquainted reader to see the bigger picture. I would have personally liked to have seen a longer introduction or conclusion that zooms out from the detailed analyses of particular cases and highlights the necessity for these studies by describing in somewhat more detail what has been labelled ‘the Great Complement Shift’ (Fanego 2007; Rohdenburg 2016; Rudanko 2012; Vosberg 2006, 2009. To put it succinctly, this term refers to a series of changes in the history of English that have been argued to have effected a large-scale restructuring of its complementation system, thereby setting off English from its Germanic relatives (see De Smet forthcoming for a brief overview). Most notably, the diachronic rise of the to-infinitive, partly at the expense of subjunctive that-clauses, and the introduction and rise of verbal gerunds since Middle English have been well documented. As a result of such changes, it has been argued, English has witnessed a strong increase in non-finite – and in particular gerundial – complement patterns, which may suggest an underlying drift towards increasing functional differentiation and economy. These changes of course do not happen overnight, and do not affect all linguistic environments to the same extent, which makes persisting variation in Late Modern and Present-day English a treasure trove for studies on constraining and facilitating factors in language change.

Now that the main background and types of explanatory factors have been introduced, we can turn to the findings of the individual case studies. Chapter 2 focuses on non-finite complement clauses of the predicate afraid in Canadian and British English. The data is coded for a wide range of features (text type, Horror Aequi phenomena, the Choice Principle, Cognitive Complexity features like extraction, passivization of the complement clause, insertions between the main and complement clause, negation in either clause). The authors show that in the Canadian data, both of the examined types of non-finite complements of afraid show a downward trend in the period since the 1930s, which could be due an increase in finite complement patterns or competition with other adjectives of fear (p. 18, attributed to a reviewer’s comment). The regression model confirms the Choice Principle to be the main predictive principle in the distribution of the two non-finite patterns in both varieties of English. Especially in the Canadian data, fictional texts favour the infinitival pattern. While passive complement clauses are significantly more likely to attract gerundial forms by themselves, the regression model clarifies that this results from a correlation of passives with [−Choice], and that the factor itself is therefore secondary. Matrix negation favours infinitives in [−Choice] complements, an interaction which is significant only in the British data set. Factors like intermediate insertions, complement negation and Horror Aequi and even extractions were too infrequent across the two complement patterns to reach statistical significance.

Chapter 3 explores the important methodological question of deciding which to-infinitives are actual complements of the adjectival predicate. To be able to examine this, the authors focus on cases in which a to-infinitival complement of an adjective of fear (‘fear-complement’) is nested within a degree complement construction (‘too-complement’), which expresses that “the matrix subject has exceeded the upper limit of fear compatible with the actualization of the complement situation” (p. 53). The fear- and too-complements can be left implicit if they are sufficiently clear in the context, and can be found in the context of a third type of to-infinitive, namely that of adjuncts of result or purpose. Example (5) shows the co-occurrence of a complement expressing the cause of fear and a complement to the degree element too. Example (6) has both a degree complement and a purpose adjunct. As the authors note (p. 73), such cases of consecutive to-infinitives might be expected to be avoided on the basis of Horror Aequi considerations, even though this appears not to be the case (see also p. 80). Tokens with three consecutive to-infinitives, however, were not attested, which can be seen to be in line with Horror Aequi considerations.

The authors propose two types of diagnostics to disentangle the different patterns. Firstly, time relations play a role: in (5), the first infinitive (to be labelled lefties) is temporally subsequent to the second one (to criticise the wealthy), which suggests that the first infinitive represents the cause of fear, and the second one the degree complement which the subject is unwilling to actualize (and not an adjunct of purpose). The second type of diagnostic rests on grammatical alternations maintaining the distinct constructional semantics: fear-complements can be expressed by gerundial patterns, and affirmative too-degree complements by so-complements with polarity reversal (e.g. ‘she was so frightened that she could not walk down…’ in [6]). Still, the authors show the remaining indeterminacy of cases as in (7), where the fear-complement and too-complement analyses are equally plausible and even “virtually synonymous” (60).

Chapter 4 returns to the lexical predicate afraid, but now with more targeted larger data sets that allow further scrutiny of relatively infrequent patterns. A wide set of features, whether semantic (Choice Principle, matrix negation), syntactic complexity-related (extractions, insertions, complement negation and passivization) or stylistic in nature (Horror Aequi) were taken into account with an impressive eye for detail, as shown in for example the authors’ distinction between various types of negation (p. 75), between pre-complement and complement-internal insertions (pp. 72, 83) as well as between different types of Horror Aequi (p. 74), distinguishing those that involve the predicate-complement constellation (e.g. immediate precedence of the complement by a to-infinitival or of -ing pattern, as in e.g. to be afraid + complement) from those that anticipate complements coming up later (e.g. a degree complement following the complement expressing the cause of fear as in [5]). In this larger data set, extractions and pre-complement insertions do come out as significant factors strongly favouring the to-infinitive, thereby counteracting the limitations of the smaller data set in Chapter 2. Horror Aequi was only significant for the avoidance of two consecutive of -ing patterns; there was no clear effect with to-infinitival complements or degree complements. In line with the findings in Chapter 2, the Choice Principle is again highly significant for complement selection with afraid, which again interacts with matrix negation in that the latter induces a preference for infinitives in [−Choice] contexts as well. There also seems to be a moderately higher incidence of gerundial clauses in the non-inner-circle varieties in the NOW corpus compared to the distributions in the inner-circle varieties.

Chapter 5 turns to complement selection by accustomed. A similar set of semantic, stylistic and complexity-related features is investigated, with some additions such as the length of insertions, the tense, modalization, complexity and type of subject NP in the matrix, as well as the phonological complexity of the subordinate verb in number of syllables. The main factors found to favour infinitives are extractions, insertions and polysyllabic verb forms. The Choice Principle is again an important factor, which interacts with matrix negation even to the extent of reversing the infinitive-favouring effect of a lower agentive subject (p. 113). With respect to this, the authors speculate on the different semantic effect of negating an adjective such as afraid which carries a negative bias against the actualization of a states of affairs, in contrast to negating accustomed with its positive bias towards actualization. Finally, the authors note a new gerundial-favouring effect in contexts where the controlling subject is positioned immediately before the adjective (8), which functions as a postmodifier to a noun. The authors raise the question for future research if complement-taking predicates used as modifiers might “be universally inclined toward less explicit clausal structures than heads” (p. 115).

The final case study focuses on prone, which differs from the other adjectives in its occurrence with inanimate and abstract subjects. Besides features related to cognitive complexity, this chapter zooms in on semantic features such as the animacy and countability of the higher subject, and the stativity, future reference, and repeatibility of the complement situation. A further analysed factor involves the modal scope of the adjective, i.e. whether it is used as a dynamic adjective “ascrib[ing] enabling conditions to a participant” or in an epistemic sense, expressing the likelihood that a proposition holds (p. 130). The model yields a high number of statistically significant predictors, some of which, however, are very infrequent in occurrence. In line with the previous chapters, extractions, [+Choice] contexts and polysyllabic verbs favour the to-infinitival pattern. With respect to the newly added features, complement stativity, epistemic use, abstract inanimate subjects, future orientation and mass and plural subjects are found to favour the infinitive. Repeatable complement situations and concrete inanimate subjects favour the gerundial pattern. The authors explain these general preferences with reference to the use of prone as a characterizing dynamic predicate. The unmarked complement pattern, the gerundial one, is proposed to match well with the prototypical uses of prone as a characterizing predicate, which typically involve repeatable behaviours being ascribed to concrete subjects (pp. 141–142). Features that are not compatible with the adjective’s prototypical semantics, and that moreover tend to have a higher cognitive complexity due to their abstractness, favour the infinitival pattern (p. 142).

Taken together, this collection clearly expands our knowledge on determining features in complement selection. In line with the authors’ established expertise on the topic, the studies do not disappoint in terms of methodological rigour and transparency, nor in terms of innovative discussions of unexplored territory, e.g. on the partial ‘merger’ or indeterminacy between predicate complement and degree complement constructions as in (7) above. In line with most of the existing scholarship, semantic features (e.g. the impact of the agentivity of the lower subject in the Choice Principle, the dynamic vs. epistemic meanings of prone, and even the interaction of negation with the negative or positive actualization bias in afraid vs. accustomed) are confirmed in multivariate analyses to have an undeniably strong effect on complement selection. Complexity features like extractions and pre-complement insertions also come out as strong predictors in data sets of sufficient size. Moreover, relatively underexplored determinants of variation are highlighted for further research: besides effects of regional variation (on which see also Mair 2002; Van Driessche and Cuyckens 2019) and suggestions of idiolectal variation (see Cuyckens et al. 2021), fictional texts are shown to favour the adjectives’ selection of the infinitival pattern. The authors (p. 39) attribute this to semantic differences with reference to Taylor and Dirven (1991): “We might indeed expect fictional narratives to use complement clauses to refer to specific actions, which have been associated with the infinitive, while learnt prose may be expected to focus on generalities, which have been proposed as a salient semantic feature of the gerund.” This suggestion may merit a detailed separate investigation – perhaps also taking into account different types of gerunds as in Maekelberghe (2020) – especially since Biber (1999) similarly finds fiction to be the text type with the highest incidence of infinitival complements of verbs. To be continued (and looking forward to it)!