In this book, I have studied the semantic and syntactic evolution of 24 different CPs, all of which have the structural composition make/take + NP, with the head noun of the NP obligatorily followed by a preposition. The central hypothesis that has been tested against British fiction from the 16th to the 20th century (with a main focus on the 19th and 20th centuries) reads: a CP’s degree of semantic specialization correlates with the semantic scope and evolution of the morphologically and semantically related simple verb (Hypothesis 1).

In order to test Hypothesis 1, it has, in a first step, been contrasted with the more comprehensive claim that morphological relatedness between the CP and the simple verb alone is sufficient for competition to arise (for the empirical results, see Table 6.1 in Section 6.1 and Table 7.2 in Section 7.1). The empirical findings have confirmed Hypothesis 1 in the sense that the semantic scope and evolution of the CP correlates with that of the semantically related simple verb but not with a simple verb that is only morphologically related. Yet, the hypothesis makes better predictions for some CPs than for others: Hypothesis 1 has been confirmed in 73% of all cases (for eight out of eleven CPs) that have morphologically and semantically related simple verbs which do not have an extensive semantic coverage or are becoming more narrow in terms of their meanings (Type II-CPs)Footnote ¹ and categorically applies to those CPs whose morphologically and semantically related simple verbs are very broad (ranging in FB 7 or FB 8) or have been extending their range of application in the course of time (Type I-CPs).

That Hypothesis 1 makes excellent predictions for Type I-CPs has been accounted for by reference to the concept of competition. Here, I have argued, it is the wide semantic scope of the simple verbs that leads the CPs to become deferred to a niche role, in which they adopt a specialized function. In the case of take notice of, make mention of and make answer to (with the latter being extinct or almost extinct in the 20th century), this is the function of an NPI, which is confined to non-assertive contexts. In the case of take leave of, specialization shows in the CP’s inclination for contexts that contain possessive determiners and, in the case of make use of, the CP is in its incipient stage of becoming attached to positive value adjectives and upward-scaling quantifiers. Whether, in the long run, make use of and take leave of will also develop into polarity-sensitive items (which is what the 20th-century data suggest; see Section 6.4) remains to be shown by future research.

The concept of competition has also provided an explanation for why we obtain a lower confirmation rate of the hypothesis for CPs which do not undergo specialization (Type II-CPs). Since, in these cases, the simple verbs are much more narrow in terms of the meanings they cover, this leaves room for other factors to enter the stage: these may be other morphologically (and semantically) related CPs (e.g. have doubts about in the case of make doubt of and pay heed to/give heed to in the case of take heed of), semantically but morphologically non-related periphrastic constructions (e.g. say sorry for and express regret in the case of make apology for or pay attention to and take notice of in the case of take heed of) or usage-based factors associated with the nominal level, as in the case of the frequencies of the head nouns. Here, the study has revealed that the overall frequencies of the head nouns correlate with the semantic restrictedness of Type II but not of Type I-CPs in 20th-century English. In addition, a case study on genre effects and the phonological weight of the head nouns presented in Section 7.1.2 suggests that these factors, too, impinge on the semantic–syntactic status and evolution of Type II-CPs. The findings presented thus allow for the conclusion that the predictive power of the simple verbs positively correlates with their frequencies of occurrence and that the predictive power of potential other factors inversely correlates with the frequencies of the simple verbs (i.e. where they are very high, other factors hardly stand a chance of influencing the semantic scope and evolution of the CPs). Future studies will have to cast more light on the relation that obtains between the token frequencies of the ‘predictors’ involved (e.g. the simple verbs, head nouns) and the strength of competition with the CPs.

Finally, the concept of competition has helped us to explain why certain CPs do not fall under the scope of the hypothesis. This applies to 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. The reason is that these CPs are semantically so different from their simple verbs that they do not compete with them in semantic space (i.e. do not fall under the scope of Hypothesis 1). As outlined in Section 9.1, it would be intriguing to study their role further by applying psycholinguistic tests measuring whether semantically non-related verbs get activated alongside semantically related ones. Since such experiments would have to take into consideration both the role of semantic competitors and the role of morphological/phonological competitors (in the sense that make sense of is very likely to get activated alongside sense; see also Section 9.1), they will broaden the perspective adopted in this study by showing that selection processes cannot be accounted for by semantic competition alone.

In future studies, it will be highly interesting to explore if semantic competition extends beyond the CPs investigated in this study: while the present study has focussed on those CPs that contain the light verbs make and take, future research may complement the picture by including CPs with such other light verbs as give, have or do. For the verb give, this will, for example, involve the following pairs of CP and simple verb: give instruction to – instruct, give help to – help or give notice of – notice. Crossing language boundaries, it might be worth exploring the same question for other Germanic languages like German: is there competition between the light verb construction and a morphologically (and semantically) related simpler verb, too? Although German does not make as extensive use of multi-word verbs (here: light verb constructions) as English (see Berg Reference Berg2014a: 502; Hawkins Reference Hawkins2019: 5–6), competition may drive language change in both languages. Possible candidates are eine Frage stellen – fragen (ask a question – question) or eine Anweisung geben – anweisen (give an instruction – instruct).

On a more general level, future studies might explore whether the model of competition that has been developed to account for the semantic evolution of the CPs also holds for other types of system-internal competition. Cases of classical variation (clearly distinguished from the type of competition discussed in this study) involve the dative alternation (see Zehentner Reference Zehentner2019; Reference Zehentner2022) or the genitive variation (see Rosenbach Reference Rosenbach and BealIn press; for more cases, see Section 9.1). Against the background of the present study, one could dig deeper into semantic changes affecting the dative and the genitive alternations over time, asking: how has the semantic scope of the competing variants (e.g. the DOC and the prepositional construction or the s-genitive and the of-construction) changed in the course of time? Is, for example, the s-genitive encroaching on semantic ground that has before only been represented by the of-construction, and is the semantic specialization attested for the DOC (see Colleman & De Clerck Reference Colleman and De Clerck2011) motivated by semantic competition with the prepositional construction? While some work has been presented on the semantic scope of single variants (e.g. on the DOC, see Colleman & De Clerck Reference Colleman and De Clerck2011; Zehentner Reference Zehentner2019; Reference Zehentner2022), the question of whether one variant has changed in semantic scope because the other one has changed in its semantic extension has, to the best of my knowledge, not been sufficiently explored yet.

In contrast to many ‘classical’ cases of variation, the competition described in the present study is (highly) asymmetrical in the sense that the simple verbs tend to cover much more of the semantic ground than the CPs. It will be theoretically interesting to see how other asymmetrical cases of competition develop in language history. One possible candidate for such an alternation is the evolution of phrasal verbs such as look up or put out compared to their morphologically and semantically related simple verbs like look or put. As far as I know, there is as yet no study that looks into competition between these two items and explores the question of whether the evolution of the phrasal verb can be predicted on the basis of the evolution of the simple verb. If the model presented in Section 9.1 can be extended to phrasal verbs and their morphologically and semantically related simple verbs, the semantic scope and evolution of the simple verb should affect the semantic coverage of the phrasal verb.

In addition, one may look into the question of whether there is competition between the phrasal verb and its semantically but non-morphologically related single-word correlate: such phrasal verbs as put out, break down or bring about (for a much more comprehensive list, see Claridge Reference Claridge2000: 46–56) tend to have different meanings depending on the context they occur in, with the semantically related simple lexeme generally more narrow in meaning than the phrasal verb. This is illustrated with the help of the phrasal verb break down in examples (1) and (2). While break down in example (1) can be said to correspond to the simple verb demolish, the meaning expressed in example (2) cannot be covered by the same simple verb but has to be paraphrased along the lines of divide or classify (for the different meanings of break down, see also the OED online).

It will be exciting to see to what extent competition between the simple verb and the periphrastic construction can account for the spread of the phrasal verb in language history. As in the case of the CPs, this will involve questions of semantic specialization. Here, however, the relation between the simple verbs and the periphrastic constructions is reversed: the simple verbs are semantically relatively narrow, while the phrasal verbs cover much more of the semantic ground and tend to even expand their range of application (especially as they move from a literal to a figurative meaning, as illustrated in examples (1) and (2)).

From a theoretical viewpoint, one of the most intriguing findings related to the CPs’ specialization is the evolution of polarity-sensitive items – in particular, the evolution of NPIs. On the basis of the analyses presented in this study, the CPs take notice of and make mention of can now be added to the idioms listed in the Cambridge Grammar of the English Language as one particular subclass of NPIs (Huddleston et al. Reference Huddleston, Pullum and Bauer2002: 823). In contrast to the idioms listed there, however, these CPs are lexically specified to a much lower degree than, for example, give a damn, see a thing or can be bothered. It remains for future research to explore whether the number of CPs that turn into NPIs can still be extended. From the research presented in this study, it follows that one other potential candidate for NPI status is take leave of, which has attested to a highly significant increase in non-assertive contexts from the 19th to the 20th century (though with 17.0% still attesting to a relatively low proportion of non-assertive contexts). An additional analysis based on the American COCA corpus (here, restricted to the time span 2015–19) has shown that this ratio of non-assertive contexts stays the same for the more recent time span investigated (see Section 6.5). Future research will need to explore more thoroughly whether this means that the increase in non-assertive contexts attested for between the 19th and the 20th century (in British fiction) has come to a halt or whether it is still ongoing.

Theoretically interesting is also the fact that one CP which has undergone specialization (make answer to) falls out of use in the 20th-century corpus studied (and has a frequency as low as 0.01 occurrences pmw in the entire BNC and of 0.02 occurrences pmw in COCA). A similar situation obtains for the CP make doubt of, which, too, has become specialized and whose frequencies are so extremely low in 20th-century English (0.05 occurrences pmw) that it is only fair to assume that this CP will, before long, also become extinct. The development of these two CPs thus raises the question of whether CPs that show a very high degree of specialization (with extremely low frequencies of occurrence), will, in a next step, not be used any more. With reference to make mention of, which has a lower normalized frequency in the 19th century than make answer to but still exists in the 20th century (with only a marginally lower frequency than it had in the 19th century), it has been argued that this need not be so (see Section 6.5). Future research will have to come up with more CPs that undergo specialization in order to see whether there is a statistical trend for CPs that have become strongly specialized to eventually cease to exist. By comparing CPs which pattern along the lines of make mention of to those which follow in the footsteps of make answer to, it may ideally be possible to come up with more specific predictions as to when the specialization of an item is followed by extinction and when this item continues to exist in a niche role.

Another (main) research interest of the present study has been the speed with which semantic and syntactic changes affect the CPs. The three CPs make use of, take leave of and make mention of, all of which undergo specialization, support the claim of a semantic–syntactic mismatch in the sense that semantic changes precede syntactic ones (Hypothesis 2). The remaining Type I construction, take notice of (with make answer to being extinct in the 20th century), is theoretically extremely interesting for the fact that, here, semantic changes only precede syntactic ones up to a certain point. This has been explained along the lines of isomorphic tendencies in the language – that is, speakers’ striving for a 1:1 correlation between form and meaning. I have argued that semantic and syntactic changes will not entirely drift apart if the CPs pass a certain frequency threshold.Footnote ² In other words, while semantic changes may affect constructions faster than syntactic ones, this mismatch is ideally just an interim state, which is finally overcome: as in sound change, where cases of homonymy and synonymy may temporarily arise but may subsequently be resolved (see Haiman Reference Haiman1980: 516), speakers’ preference for a match between form and function reduces the gap between semantic and syntactic changes. For take notice of, this process starts in the 18th century and comes to a close in the 20th century, when semantic and syntactic changes start to run in parallel (see the findings presented in Section 8.2).

Future studies may extend the comparison of formal and semantic changes to a range of other CPs. For a start, these should be CPs which contain light verbs different from make and take. Possible candidates are give, do and have as in give account of, do justice to or have access to. Although each of these light verbs brings with it its own history of word meanings and grammatical patterns (see also Kytö Reference Kytö, Brinton and Akimoto1999),Footnote ³ we can imagine a scenario in which these CPs, too, attest to form–meaning mismatches in their specialization process and in which, provided that the CPs range in yet-to-be-established frequency classes (see the argumentation provided in Sections 8.2 and 9.2 as well as in this section), these mismatches are finally resolved. In a next step, the analyses could be extended to CPs that involve verbs other than light verbs. In line with Mastrofini (Reference Mastrofini, Pompei, Mereu and Piunno2023), we may think of lightness not to be a property genuine to such general verbs as make, have, do, take and give but as a pattern that arises as full lexical verbs enter a CP, thus turning into aspectual devices. Following Mastrofini (Reference Mastrofini, Pompei, Mereu and Piunno2023: 207), such possible light verb extensions are to break an agreement, to conclude an argument or to adopt an idea. There is at least the chance of them also attesting to form–meaning mismatches which may or may not finally be resolved. It remains to be explored how these extended CPs behave diachronically and in what ways they are comparable to the constructions investigated in this study.

In this study, the focus has been on the semantic and syntactic changes that occur inside each single construction rather than on grammaticalization and lexicalization as the more comprehensive frameworks of language change. Towards the end of this book, I would like to ask whether the semantic–syntactic mismatches observed in the development of CPs have any implications for the more general frameworks of lexicalization and grammaticalization. As outlined in Section 2.1, the development of the CPs has been conceptualized as a process of lexicalization. This means that this study provides substantial empirical evidence for the fact that semantic–syntactic mismatches are not genuine to cases of grammaticalization (see Francis & Yuasa Reference Francis and Yuasa2008). Rather, they support the initial evidence provided in Berlage (Reference Berlage2010) to the effect that semantic–syntactic mismatches also happen in lexicalization processes.

The third theoretical contribution that this study makes relates to the more global question of whether the trend of the English language to become ever more analytic (see e.g. Hawkins Reference Hawkins2019) is without exceptions. The answer that this study provides is clearly in the negative: 88% (21/24) of the CPs analysed are very infrequent in 20th-century British English (ranging in FB 1–5), whereas 58% (14/24) of all simple verbs range in the higher or highest frequency bands 6–8, thus refuting the claim that English becomes increasingly analytic (Hypothesis 3). In addition, this study has shown that all of those CPs which have semantically related simple verbs either show a decrease in FBs as we move from the 19th to the 20th century or have consistently low frequencies of occurrence (ranging in FBs 1–5; see Table 9.3 in Section 9.3). This finding is in line with Szmrecsanyi (Reference Szmrecsanyi, Nevalainen and Traugott2012) and Neubauer (Reference Neubauer2024) on the decreasing analyticity of the English language attested for in inflectional and derivational morphology.

The low frequencies of the CPs in 20th-century English are all the more surprising given that studies on Middle and Early Modern English have shown the CPs to be flourishing and that PDE is replete with light verb constructions (see e.g. Hawkins Reference Hawkins2019; Berg Reference Berg2014a). Is the dataset studied in this book then sufficiently different from other (historical) studies on CPs? In Section 9.3, I have argued that the CPs studied here can specialize (and thus decrease in terms of their frequencies of occurrence) precisely because functions that have been fulfilled by the CPs are taken over by the corresponding simple verb. This makes the present set of CPs stand out from the much more heterogenous group of light verb constructions discussed elsewhere (see e.g. Claridge Reference Claridge2000; Berg Reference Berg2014a; Hawkins Reference Hawkins2019). This study has provided empirical evidence supporting my line of argumentation (see Table 9.2 in Section 9.3): comparing the usage frequency of sets of CPs and simple verbs with a semantic overlap, we see that the simple verbs are used much more extensively than the CPs. Also, only those CPs whose semantics do not show any overlap with those of their corresponding simple verbs (i.e. take advantage of, take hold of, make love to) have more than ten occurrences pmw in 20th-century English. I have argued that in those cases where the speaker has a choice, the speaker’s preference for the shorter simple verb over the longer analytic construction is motivated by the principle of economy of expression.

For now, it will be intriguing to explore whether the trend towards non-analytic forms in the English language, obvious from the fact that the frequencies of the CPs are lower than those of their semantically related simple verbs and have either been decreasing in the course of time or stayed at a consistently low level, can also be confirmed for other areas of grammars. These need to be (grammatical) domains in which at least some of the semantic functions represented by one item can be taken over by the competing other item. In the following, we will see whether the increase of the non-analytic form at the expense of the analytic form is further supported by the evolution of the dative, the genitive or the comparative alternation.

In a first step, we will stick with the domain of the VP. Rather than look at CPs, however, we will turn to the dative alternation, which, in semantically comparable environments, involves competition between the synthetic double object construction as in she gave the monkey the banana and the analytic prepositional construction as in she gave the banana to the monkey.Footnote ⁴ In Early Middle English, the prepositional object construction was significantly increasing to the detriment of the double object construction (see Zehentner Reference Zehentner2022: 38). Rosenbach (Reference Rosenbach and BealIn press) now shows, however, that this development is reversed from the second half of the 15th century onwards, when the synthetic double object construction is regaining ground at the expense of the analytic prepositional construction, even to the extent that the development comes full circle: as in Old English times, the double object construction re-adopts the role of the majority variant in Modern English. This finding clearly suggests that English is not generally becoming increasingly analytic over time.

While the aforementioned changes take place in the VP, there are also changes towards more syntheticity in the NP, exemplified at the example of the genitive construction. The genitive construction has, as its two variants, the s-genitive (as in the king’s daughter) and the of-construction (as in the daughter of the king), qualifying as the synthetic and analytic constructions, accordingly. Rosenbach (Reference Rosenbach2002: 184–92) shows that, despite English becoming increasingly analytic, the synthetic s-genitive is on the increase between 1400 and 1630. A more recent study (Rosenbach Reference Rosenbach and BealIn press) further supports this trend: from the second half of the 15th century onwards, the s-genitive is gaining ground at the expense of the of-construction. This parallels the situation that we find in the case of the CPs’ decreasing frequencies at the expense of the simple verbs.

Let us, finally, turn to the adjective phrase (AP) to see whether here, too, previous studies have found any counterexamples to the general trend of the English language to become more analytic. We will here concentrate on the comparative alternation – that is, the alternation between the synthetic -er variant (as in colder) and the analytic more-variant (as in more cold). Following Mondorf (Reference Mondorf, MacWhinney, Malchukov and Moravcsik2014: 226–8), we can argue that the synthetic -er variant increases in contexts that are easy to process while, after the 18th century, the analytic variant increasingly compensates for more demanding contexts, as are, for instance, constituted by the presence of complements as (for example) in he is more proud of his good results (for a similar compensation strategy in Greek, see Karkaletsou & Alexiadou Reference Karkaletsou and Alexiadou2023). This means one variant is not replacing the other, but we see a functional division of labour between the synthetic and analytic forms. While this functional division of labour seems to persist in the case of the comparative alternation, it has been resolved in the present study: here, CPs which have simple verbs with an extensive semantic coverage become specialized, thereby severely reducing their domain of application.

To conclude, the present study has contributed to our understanding of semantic specialization processes and to the ways in which language change can be predicted. By taking into consideration the semantic scope and evolution of the simple verb, I have been able to successfully predict the semantic specialization of the CPs at a minimum accuracy rate of 73% (with a 100% accuracy for CPs that have simple verbs with an extensive semantic coverage). Although future studies still need to develop comprehensive statistical models that take in all relevant factors predicting the evolution, the data strongly suggest: the role of the simple verbs must not be overlooked when it comes to predicting the semantic–syntactic evolution of the CPs.

Finally, I have shown that the well-known principles of competition, iconicity and economy of expression offer an explanation for why a) some CPs undergo specialization while others do not, b) semantic and syntactic changes do not entirely drift apart and c) non-analytic structures are preferred over analytic ones where these have (at least) a (minimal) semantic overlap. In doing so, this study has strengthened the role that functionally motivated principles play for language change.