In Chapter 9, I offer a discussion related to the main theoretical contributions of this study. In particular, I will refer to the following three insights gained from the qualitative and quantitative analyses of the 24 CPs presented in this book:

1. a) The CPs’ semantic evolution correlates with the semantic scope and evolution of the morphologically and semantically related simple verbs (see Chapters 5, 6 and 7).

2. b) Semantic changes in the CPs precede syntactic ones, but the data suggest that differences between them may, eventually, even out (see Section 8.2).

3. c) When we look at the frequencies with which the analytic CPs occur in 19th- and 20th-century British fiction, we see that they are mostly decreasing or are consistently very low (see Sections 6.1, 6.5, 7.1.1, 7.1.2 and 7.4). This contrasts with the general trend of PDE to be replete with analytic structures (see e.g. Hawkins Reference Hawkins2019).

I will now elaborate on how these findings tie in with three concepts known to be well-supported functional principles at work in various languages. These are the principles of:

1. i) competition

2. ii) iconicity

3. iii) economy of expression

Although it is clear that language change is rarely driven by one factor alone (see e.g. Hinrichs & Szmrecsanyi Reference Hinrichs and Szmrecsanyi2007; Szmrecsanyi et al. Reference Szmrecsanyi, Rosenbach, Bresnan, Wolk and Hundt2014; Szmrecsanyi Reference Szmrecsanyi, Börjars, Denison and Scott2013) and that functional principles are often interwoven, this study concentrates on the effect that the semantic scope and evolution of the morphologically (and semantically) related simple verbs have on the CPs. In this chapter, I will ask: why does the semantic scope and evolution of the simple verb correlate with the semantic scope and evolution of the CP? The answer that is provided draws on the concept of competition, which, originally a psycholinguistic effect, has been shown to affect not only synchronic language use but also language change (see e.g. Zehentner Reference Zehentner2019; Reference Zehentner2022; Schlüter Reference Schlüter2005).Footnote ¹ On the basis of this concept, I develop a model of semantic space that accommodates both the simple verb and the CP and that shows how semantic changes of the simple verb result in changes of the CP (see Section 9.1).

Next, I turn to the mismatches between form and meaning observed in the CPs and ask how they can finally be resolved. Section 9.2 explores the form–meaning correspondence captured by the notion of iconicity and shows in what ways iconic tendencies in the language can be held responsible for the fact that meaning and form do not entirely drift apart.

Finally, my findings showing that we have fewer (rather than more) CPs in the 20th century than in former stages of English run counter to the trend of the English language becoming increasingly analytic. I argue that the reverse development attested for in this study has to do with competition between the simple verb and the CP. In addition, this development is in line with speakers’ increasing preference for more rather than less economic structures (Section 9.3). While the principles of competition and iconicity explore the relation that obtains between different linguistic constructions or between their forms and their meanings, the principle of economy shifts the focus onto language users, explaining why speakers prefer the simple verbs over morphologically and semantically related CPs.

9.1 The CP and the simple verb: Competition in semantic space

In this study, I have made predictions as to the evolution of the CPs on the basis of the semantic scope and evolution of the morphologically and semantically related simple (or prepositional) verbs (e.g. notice, provide for, use).Footnote ² The findings presented in Chapters 5, 6 and 7 have supported the hypothesis that a CP’s degree of semantic specialization correlates with the semantic scope and evolution of the morphologically and semantically related simple verb, attesting to a 73% confirmation rate for the hypothesis of ‘no specialization’ (eight out of eleven cases; see Section 7.4) and a 100% confirmation for the hypothesis of ‘specialization/restrictedness’ (see Section 6.5). I will now introduce a model of competition which can account both for cases of specialization and ‘no specialization’ (out of the set of CPs which have a semantic overlap with their morphologically related simple verbs).

The model that I would like to offer is based on the idea that the simple verb and the CP compete with each other in semantic space.Footnote ³ I will illustrate this by focussing on the CP take notice of and its morphologically related simple verb notice. On the basis of this pair of examples, I will show a) how I envisage the semantic coverage of the simple verb to be related to that of the CP and b) how changes in the semantics of the simple verb may affect the evolution of the CP.

When we look at the first attestations of the simple verb notice and the CP take notice of in the OED, we see that notice is first mentioned in the year 1447 (OED online), whereas the CP take notice of first occurs more than a hundred years later (1573; see OED online). Since languages tend to avoid cases of complete synonymy,Footnote ⁴ it makes sense to find out to what extent the simple verb and the CP overlap. In order to do so, we need to elaborate on the relation between form and meaning in semantic space.

Like many other models of competition (for an overview, see De Smet et al. Reference De Smet, D’hoedt, Fonteyn and van Goethem2018), the present model relies on the assumption that languages in general are more or less isomorphic in the sense that they reserve one form for one meaning and vice versa (cf. e.g. Bolinger Reference Bolinger1977: x; see also Haiman Reference Haiman1980):

Returning to semantic space, this means: if one item has an excessive semantic coverage, the other should cover far less of the semantic ground because the semantic overlap between the simple verb and the CP should be kept to a minimum (thus optimizing form–function mapping).Footnote ⁵ Applied to the case of take notice of and notice, the broad semantic coverage of the simple verb notice should prevent the periphrastic construction from also having an extensive semantic coverage. The data that have been provided on the normalized frequencies of these items in BNC wrid.1 support this assumption: while the simple verb notice has as many as 227 occurrences pmw in BNC wrid.1 (and therefore is in FB 7; see Section 5.2), take notice of is much rarer, accounting for no more than 9.4 occurrences pmw (and thus ranging in FB 5; see Table 6.2, Section 6.1). Provided that the normalized frequencies of the verbal expressions are an adequate proxy of their semantic generality (see the argumentation provided in Section 4.2), we can conclude that the CP take notice of is semantically much more narrow than the simple verb notice.

When we look at the more specific distribution of the simple verb and the CP in semantic space, three models suggest themselves. I will refer to them in Figure 9.1 (Model 1), Figure 9.2 (Model 2) and Figure 9.3 (Model 3). While all of them share the idea that the number of meanings in semantic space is distributed across the simple verb and the CP (synchronically), they differ as to what the relation between the two items is. While the simple verb and the CP are clearly distinct in terms of their meanings in Model 1 (in Figure 9.1), there is some semantic overlap between them in Model 2 (in Figure 9.2). In Model 3 (in Figure 9.3), then, the meanings of the CP are entirely covered by the (broader) meanings of the simple verb. We will now discuss each of these models in more detail before we decide which one comes closest to mapping semantic competition between the simple verb and the CP.

Figure 9.1 Model 1 of notice and take notice of in semantic space (synchronic perspective)

Figure 9.2 Model 2 of notice and take notice of in semantic space (synchronic perspective)

Figure 9.3 Model 3 of notice and take notice of in semantic space (synchronic perspective)

The two models which are diametrically opposed to each other are Model 1 (in Figure 9.1) and Model 3 (in Figure 9.3). While Model 1 assumes that there is no context in which both the simple verb notice and periphrastic take notice of can occur, Model 3 is based on the reverse assumption – that is, that all the contexts in which take notice of is used can be replaced by the simple verb notice. In order to decide if one of these assumptions is correct, we will have to answer the following two questions:

1. 1) Do the data attest to cases of variation between the simple verb and the CP?

2. 2) Can all occurrences of the CP be replaced by the simple verb?

Let us start by focussing on the first question. If the data attest to cases in which both the CP and the simple verb can be used, we can rule out Model 1 (in Figure 9.1) for the reason that there needs to be some semantic overlap between the simple verb and the CP (however small this may be). Certainly, there are contexts in which the simple verb can be replaced by the CP. One such example is quoted in (1).

Since we can say he didn’t seem to take notice of the pouring rain without a major change in meaning, we can rule out Model 1 (in Figure 9.1) as one that does not account for variation between the simple verb and the CP. Let us now turn to the second question. Can take notice of can be replaced by the simple verb notice in all the contexts in which it occurs? The answer is clearly in the negative. One example of a context that does not allow for variation (with the surrounding syntactic structure kept unaltered) is given in (2).

The examples obtained for take notice of (for a series of examples, see Chapter 6) suggest that there are fewer rather than more examples that allow for variation between the two items (this being the reason why the overlap between the CP and the simple verb in Model 2 is relatively small; a more systematic analysis still needs to be carried out). What this means is that we are left with Model 2 (in Figure 9.2) as the one that most adequately describes the synchronic relationship between take notice of and the simple verb notice.

If Model 2 (in Figure 9.2) is the model that best accounts for the (synchronic) relationship between take notice of and notice, the question emerges of whether all the other CPs investigated in this study also comply with the relation outlined in Model 2 or whether they are better described with the help of Model 1 (in Figure 9.1) or Model 3 (in Figure 9.3). The answer to this question can be arrived at by reviewing those CPs which are clearly semantically distinct from their morphologically related simple verbs (see the overview provided in Table 4.3, Section 4.1). The CPs take advantage of, take charge of, take hold of, take offence at, take possession of, make difference to, make love to and make sense of all have in common that they are (for various reasons) semantically so different from their corresponding simple verbs that they are not in semantic competition. In other words: since the meanings of these verbs are entirely distinct from those of the simple verbs, the evolution of the simple verb does not seem to affect the evolution of the CP (for the empirical findings, see Table 6.1 in Section 6.1 and Table 7.2 in Section 7.1; compare also the lower predictive power of the simple verbs associated with Type II-CPs as compared to Type I-CPs).

If we were to relate these eight CPs to one of the three models, it should, obviously, be Model 1 (in Figure 9.1). In other words, Model 1 represents a suitable model for morphologically related verbal expressions that are not in semantic competition and should therefore not be subjected to the hypothesis. From here it follows that CPs for which we predict to find a positive correlation in the form stated in Hypothesis 1 should either comply with Model 2 (in Figure 9.2) or with Model 3 (in Figure 9.3). Since there is no set of CP and simple verb that allows for variation in all the contexts in which the CP occurs (the reader is invited to play this through with the set of CPs provided in Section 4.1), all the CPs which fall under the scope of the hypothesis should be described in the framework of Model 2 (in Figure 9.2).Footnote ⁶

While Model 1 (in Figure 9.1), Model 2 (in Figure 9.2) and Model 3 (in Figure 9.3) give synchronic snapshots of the semantic relation between the CP and the simple verb, we will now look at the effects of semantic competition for language change. For this matter, we turn to Model 4 (in Figure 9.4), which visualizes in what ways the semantic coverage of the simple verb and the CP may be related in diachrony (with the dotted arrows indicating language change). In line with Model 1, Model 2 and Model 3, the model is simplified in the sense that competition is restricted to the CP and the morphologically related simple verb investigated in this study (although semantic competition may extend beyond it).Footnote ⁷

Figure 9.4 Model 4 of changes in semantic space

Although Model 4 (in Figure 9.4) depicts semantic change, it also has a synchronic dimension, and it is this dimension that we will look at before we turn to semantic change: the semantic coverage of the simple verb affects the semantic scope of the CP. This is indicated by the two unbroken arrows which point from the semantically broad simple verb to the semantically narrow CP and from the semantically narrow simple verb to the semantically broad CP: depending on the semantic coverage of the simple verb, the CP is either broader or more narrow in terms of its semantic scope. Let us now turn to the dimension of semantic change.

In Model 4 (in Figure 9.4), we see dotted lines indicating that the semantic coverage of both the simple verb and the CP may change in the course of time. Now, since there is competition between the simple verb and the CP, semantic changes in the simple verb will affect the semantic coverage of the CPs. This means: if the simple verb increases in terms of its semantic density over time, the CP is likely to decrease in terms of the meanings it covers. If, on the other hand, the simple verb becomes semantically more narrow (a state in which it may still be semantically denser than the CP), the CP may in turn become semantically more general (if only to a limited degree).

Sceptics may argue that there is also a reverse effect in the sense that the semantic evolution of the CP influences the evolution of the simple verb. Since the simple verbs are generally older and far more frequent than the CPs, however, this scenario is much less likely to occur, and, indeed, 21 out of the 24 CPs analysed have frequencies lower than ten occurrences pmw in BNC wrid.1, ranging in FBs 1–5 (see Section 9.3; see also Sections 7.4 and 6.5). This contrasts with the simple verbs, where 29.2% of all verbs range in the highest FBs 7 or 8. If we take these frequencies to be indicators of the semantic scope of the verbs (for the argumentation, see Section 4.2), then the simple verbs in general have a more extensive semantic coverage than the CPs. This finding also makes sense against the background that the CPs contain the light verbs make or take, which, typically, add an aspectual reading to the construction, making them more specific when compared to the simple verb (for further discussion, see Section 1.2). In addition, the direction of the arrows reflects the methodological procedure in the sense that the simple verbs are of interest only to the extent that they provide us with relevant information on the semantic scope and evolution of the CPs.

Let us now, in a next step, ask why the model makes better predictions for some cases than for others: the empirical analyses have attested to a 73% confirmation rate for those (morphologically and semantically related) CPs which have simple verbs that do not have an extensive semantic coverage or do not extend their semantic scope (i.e. Type II-CPs) but have categorically confirmed the hypothesis of ‘specialization/restrictedness’ for those CPs which have semantically very broad simple verbs or whose simple verbs have an extensive semantic coverage (i.e. Type I-CPs).Footnote ⁸ This raises the question of where the differences in the predictive power of the simple verbs come from. In other words, why is the predictive power lower for CPs whose simple verbs do not have an extensive semantic coverage or do not extend their semantic scope?

The answer that I would like to suggest is that in those cases where a simple verb does not have an extensive semantic coverage or is narrowing down in terms of its semantic scope (obvious from medium to low frequencies of occurrence, i.e. frequencies lower than those summarized under FB 7 and FB 8), there may be other verbal competitors in the paradigm that impinge on the evolution of the CP. The present study has shown that this applies, for example, to the case of take heed of: here, we find the morphologically competing expressions pay heed to and give heed to and the non-morphologically related verb phrases pay attention to and take notice of. When we now look at the two other CPs that contradict the hypothesis of ‘no specialization’, these, too, have verbal alternatives in the paradigm. In the case of make doubt of, this is have doubts about, which is considerably more frequent than make doubt of (see Section 7.1.2). In the case of make apology for, these are the periphrastic constructions say sorry for and express regret (for more details, see again Section 7.1.2). Section 7.3 has revealed that the semantic status of those CPs that have simple verbs with medium to low frequencies of occurrence (i.e. Type II-CPs) is additionally influenced by the frequencies of the head nouns. This means that a model predicting the semantic evolution of Type II-CPs should include usage-based factors found in domains outside the VP.

Model 5 (in Figure 9.5) and Model 6 (in Figure 9.6) visualize the difference in predictive power between CPs that have semantically broad (prepositional or simple) verbs like notice (Model 5, Figure 9.5Footnote ⁹) and those that have semantically more narrow (prepositional or simple) verbs like apologise/apologize for (Model 6, Figure 9.6). In comparison to Model 5 (in Figure 9.5), which shows the distribution for a semantically broad verb like notice, the amount of semantic space that is not filled by the simple verb is much larger in Model 6 (in Figure 9.6). In line with my argumentation, it is precisely this ‘void’ which is filled by other verbal elements in the paradigmFootnote ¹⁰ or through which usage-based factors, such as the overall frequencies of the head nouns, have a chance to affect the semantic evolution of the CPs.

Figure 9.5 Model 5 on CPs that have simple verbs with an extensive semantic coverage

Figure 9.6 Model 6 on CPs that have simple verbs with a more narrow semantic coverage

If Figures 9.5 and 9.6 correctly reflect the difference in predictive power between semantically broad and semantically narrow simple verbs and if the correlation between the semantic scope of an item and its frequencies of occurrence as stated in Zipf’s Third Law holds (see the arguments and correlation coefficients provided in Sections 4.2 and 5.1), we can conclude: higher frequencies of the simple verb are indicative of a higher degree of semantic competition between the simple verb and the CP and therefore of a higher degree of semantic specialization in the CP. While it is certainly true that non-verb-related factors may also affect the semantic status and evolution of the CP, the present study has so far only focussed on the overall frequencies of the head nouns (for a synchronic study, see Section 7.3). This effect seems to be inversely correlated with the frequencies of the simple verbs: the lower the frequencies of the simple verbs are (as seen in the case of Type II-CPs), the more relevant will be the influence of the overall frequencies of the head nouns (i.e. a significant influence in the case of Type II-CPs). It will be left for future research to enter all potential factors affecting the evolution of the CPs into a statistical model predicting their development.

Finally, I would like to assess how useful the concept of competition is when it comes to accounting for the evolution of the 24 CPs discussed in this study. Certainly, the relation between the CPs and the simple verbs represents no ‘classic’ example of competition between two functionally similar expressions with different forms because the simple verb is semantically much more comprehensive than the CP (compare e.g. answer – make answer to, notice – take notice of). This contrasts with such a well-known case of competition as the genitive alternation (see e.g. Rosenbach Reference Rosenbach2002; Reference Rosenbach and BealIn press), or with phonological variation as in the case of the standard and non-standard adjectives worse and worser or the mono- and disyllabic variants of past participles such as drunk/drunken in the history of English (see Schlüter Reference Schlüter2005: 74, 111).Footnote ¹¹ Still, when applied to the case of CPs and their morphologically related simple verbs, the concept of competition can account for all of the following three scenarios, labelled as Scenario A, Scenario B and Scenario C.

9.2 On the relation between form and function: The principle of iconicity

The second major theoretical contribution of this study is that it promotes our understanding of mismatches between form (syntactic changes) and function (semantic changes) inside single constructions: for three out of five Type I-CPs that undergo semantic specialization in the 19th and 20th centuries (make use of, take leave of, make mention of), we have seen that semantic changes proceed much more speedily than syntactic ones (see Section 8.2). This finding is not only fully in line with the observations made by Givón (Reference Givón, Traugott and Heine1991), Haspelmath (Reference Haspelmath1999), Heine and Kuteva (Reference Heine and Kuteva2002: 3) and Francis and Yuasa (Reference Francis and Yuasa2008) for cases of grammaticalization but also adds to the initial evidence provided by Berlage (Reference Berlage2010) on three cases of lexicalization: form–function mismatches are not genuine to cases of grammaticalization but extend to cases of lexicalization.

Since make answer to is extinct in 20th-century fiction, there is only one remaining Type I-CP, take notice of, which shows a different development. In contrast to the other three CPs, take notice of does not fully confirm the hypothesis of semantic changes preceding syntactic ones. While the relation between form and meaning is still fully in line with the hypothesis in the 16th to the 18th century, this relation changes in the 19th century in the sense that the gap between semantic and syntactic changes becomes smaller to the point that, in the 20th century, it is closed: semantic and syntactic changes now proceed in parallel (see Section 8.2). This raises the question of why the gap between semantic and syntactic changes closes in the case of take notice of and whether the development that we observe here might be indicative of a larger trend in the language to let semantic and syntactic changes not diverge too far.

The answer provided in Section 8.2 is based on the frequencies of the CPs take notice of, take leave of, make answer to, make mention of and make use of. As has been shown in Table 6.2, take notice of has a frequency of 9.4 occurrences pmw in BNC wrid.1 and is thus much more frequent than all of the remaining four CPs, whose frequencies range between 0 (make answer to) and 5.3 (make use of) occurrences pmw. If frequencies of occurrence are indicative of a CP’s degree of semantic specialization/restrictedness (see Section 4.2), then we can conclude that take notice of, although clearly an NPI (see Section 6.4), is less strongly specialized than the other four CPs (with extinction following semantic specialization in the case of make answer to). We can further infer that a yet-to-be-defined range of frequencies is necessary for a CP to arrive at a situation in which the form–meaning mismatch is resolved in favour of a parallel development of form and meaning.

In a next step, we want to ask: why should speakers want to resolve form–meaning mismatches at all? In order to answer this question, we will turn to the principle of iconicity, which assumes a correspondence between form and meaning in the sense that differences in meaning should be reflected in differences in form (see e.g. Haspelmath Reference Haspelmath2008; Bolinger Reference Bolinger1968; Reference Bolinger1977; Haiman Reference Haiman1980; see also Rohdenburg Reference Rohdenburg, Müller and Fischer2003; Reference Rohdenburg, Hundt, Nesselhauf and Biewer2007; Reference Berg, Fischer, Akita and PernissBerg In press). As in Section 9.1, we will here concentrate on the type of iconicity called (paradigmatic) isomorphism (see Haspelmath Reference Haspelmath2008: 2; Haiman Reference Haiman1980: 516; Croft Reference Croft2003: 105–10), according to which one form is associated with one meaning and one meaning with one form (see also Bolinger Reference Bolinger1977: x).Footnote ¹² This principle argues not only against the existence of synonymy (two forms, one meaning) and homonymy (two meanings, one form; see also Haiman Reference Haiman1980: 516) but can also be considered one of the driving principles behind language change (see also Dik Reference Dik1988: 98–9; quoted from De Smet et al. Reference De Smet, D’hoedt, Fonteyn and van Goethem2018: 200).

While languages certainly tolerate a state in which form and meaning are not related in a 1:1 fashion (see e.g. De Smet et al. Reference De Smet, D’hoedt, Fonteyn and van Goethem2018: 201), I would like to argue that they eventually strive for a closer mapping of form and meaning. This will be demonstrated for English at the example of our CPs: if constructions already exhibit a high degree of semantic specialization but formal changes still lag behind (see the cases of make use of, take leave of and make mention of), this permanently contradicts a close mapping of form and function. Take the case of make use of, which becomes gradually confined to positively oriented value adjectives and upward-scaling quantifiers but for which this change is not (yet) reflected in the modifier slot in the sense that the proportion of modifiers (and modifiers + determiners) is still as low as 37% of all cases in the 20th century (see Table 8.6, Section 8.2). Or take the case of take leave of, which, in the 20th century, is restricted to possessive pronouns in the determiner slot. The proportion of possessive determiners, overall, however, is still relatively low, accounting only for 45% of all cases (see Table 8.7, Section 8.2). The same is true for the CP make mention of, which becomes confined to the determiners no, no more and any but whose determiner slot is only filled in 77% of all cases in 20th-century BrE (see Table 8.8, Section 8.2).

If isomorphism is a driving force behind language change, this will support a development along the lines of take notice of – that is, a development in which semantic changes (may) precede syntactic ones but in which semantic–syntactic mismatches are finally resolved. If this is so, the scenario in the case of the CPs would parallel such cases of synonymy and homonymy as evolve accidentally in the process of sound change (see Haiman Reference Haiman1980: 516). Just as violations of the ‘one form – one meaning’ principle are remedied in the case of sound changes,Footnote ¹³ the gap between formal and semantic changes should eventually be closing. Future research needs to explore which ranges of frequencies need to be met to overcome formal–semantic mismatches.

9.3 The decline of CPs: Analyticity and economy of expression

In a last step, we will test the claim of whether the CPs investigated defy the pervasive trend in the English language to increase in analytic structures. In a first step, let us focus on the low frequencies of the CPs in 20th-century fiction: when we zoom in on the 24 CPs analysed, we see that 21 of them range in FBs 1–5 in 20th-century British fiction, which means that their frequencies do not extend to more than 9.9 occurrences pmw. More than that, seven CPs even range in FBs 1–4, with no more than a maximum of 0.99 occurrences pmw. This leaves us with only three CPs which have frequencies of ten or more occurrences pmw. These are listed in Table 9.1.

These findings are certainly not in line with the growing trend of the English language to be replete with analytic constructions. Given that Berg (Reference Berg2014a: 502) and Hawkins (Reference Hawkins2019) stress the fact that light verb constructions involving productive verbs like make, take, give or have are particularly popular in PDE (for quantitative data on Early Modern English, see Claridge Reference Claridge2000: 111) and that Kytö (Reference Kytö, Brinton and Akimoto1999) and Traugott (Reference Traugott, Brinton and Akimoto1999: 255–6) attest to an increase in the use of CPs (as opposed to simple verbs) from the Middle English period to the Early Modern English period (see Kytö Reference Kytö, Brinton and Akimoto1999: 198; Traugott Reference Traugott, Brinton and Akimoto1999: 255–6), these findings seem unexpected. Considering, however, that all of the CPs investigated in the present study have a morphologically and, in most cases, semantically related counterpart, while many others do not (see e.g. take place, take part with, give control to, give leave to, take the liberty),Footnote ¹⁴ the findings start to make sense: the low degree of analyticity observed in the present study seems to be motivated by competition between the CP and the simple verb (no matter how extensive the semantic overlap between them is). In order to substantiate this claim, let us compare the frequencies of occurrence attested for those simple verbs and CPs that have a semantic overlap. This overview, summarizing the findings presented in Sections 5.2, 6.1, 7.1.1 and 7.1.2, is presented in Table 9.2.

Table 9.2 shows that the simple verbs are used to a much larger extent than the CPs. More specifically, the ratio of items that range in FB 6 or higher (i.e. FB 7 or 8) amounts to 50% of all simple verbs, but there is no CP that, in 20th-century fiction, ranges in FB 6. Indeed, the difference between these two sets is statistically highly significant (x² = 8.2; df = 1; p < 0.01; Yates’ correction applied).

In a second step, we will now look at how the frequencies of the 16 CPs on display in Table 9.2 have developed across the 19th and 20th centuries. If the development of the CPs defies the trend of the English language to become ever more analytic, the majority of them should show either decreasing or consistently low frequencies of occurrence (here defined as ranging in FBs 1–5). The frequency counts produced in Table 9.3 are based on the analyses conducted in Sections 6.1, 7.1.1 and 7.1.2. Since the decrease in frequencies has been considered genre specific in the cases of make allowance for and make provision for (see Section 7.1.2), these two CPs are excluded from Table 9.3, leaving us with a total of 14 CPs.

The findings summarized in Table 9.3 clearly confirm our assumption: half of the 14 CPs on display show a decrease in FBs from the 19th to the 20th centuries, with the other half having consistently low frequencies of occurrence. In fact, there is only one CP (i.e. take account of) which shows an increase in frequencies as we move from the 19th to the 20th century, but since this increase takes place at such a low frequency level (changing from 0.8 to 2.2 occurrences pmw in the respective centuries), take account of clearly still qualifies as a low-frequency item. No doubt, these findings support the studies advanced by Szmrecsanyi (Reference Szmrecsanyi, Nevalainen and Traugott2012) and Neubauer (Reference Neubauer2024) on the decreasing analyticity of English. While the contributions made by Szmrecsanyi (Reference Szmrecsanyi, Nevalainen and Traugott2012) and Neubauer (Reference Neubauer2024) relate to inflectional and derivational morphology, the present study contributes empirical evidence from the syntactic realm of the VP.

While I have so far suggested that the decline in analyticity can be accounted for by system-internal competition between the CP and the simple verb, I will, in a final step, shift the focus from the linguistic constructions onto the language speaker: the increasing preference for the non-analytic simple verb (in cases where there is a semantic overlap between the CP and the simple verb) may have to do with speakers’ increasing striving for economy of expression (for attestations of this trend, see e.g. Biber Reference Biber, 205Aitchison and Lewis2003; Mair Reference Mair2006; Smitterberg Reference Smitterberg2021; for the principle of economy, see Croft Reference Croft2003: 102). Not only are (very) long and phonologically bulky head nouns inside such CPs as make allowance for, make apology for and make provision for generally avoided (showing in extremely low frequencies of these CPs in the 20th century, ranging between 0.2 and 0.8 occurrences pmw; see Table 7.8, Section 7.1.2) but the analytic construction itself is also much longer and syntactically more complex than the non-analytic construction (compare e.g. make mention of and mention). The speaker’s preference for shorter and less time-consuming expressions may therefore (at least in part) account for the decline of the (more complex) CPs in favour of the simple verbs. This assumption is fully in line with the observation that frequencies of more than ten occurrences pmw are only found for CPs whose meanings are so clearly different from those of the simple verbs (i.e. take advantage of, take hold of and make love to) that speakers do not have a choice between the simple verb and the CP but have to turn to the analytic construction invariably.Footnote ¹⁵ In those cases where the CP and the simple verb are in semantic competition (see Table 9.2), the analytic construction is avoided in favour of the more economic, non-analytic alternative.