Construction grammar is a collective term for different theoretical approaches (compare, e.g. Croft 2001; Goldberg 1995) which share some fundamental starting points. Construction grammar assumes that constructions are ‘stored pairings of form and function, including morphemes, words, idioms, partially lexically filled and fully linguistic patterns’ (Goldberg 2003:219). Constructions constitute the basic units of language and are stored in a structured network, the constructicon. More than informal approaches, there is more attention to the semantic and pragmatic characteristics of grammatical patterns and to the relations between different constructions with comparable functional qualities in construction grammar. In diachronic linguistics, this translates to research of changes within these qualities or interconstructional relations, or constructional changes. This study focuses on the evolution of the constructional semantics of the aan-construction in the 16th-century. In the next section, we will go into the features of the Dutch aan-construction, and we will clarify the reason why the 16th-century aan-construction is so interesting for further research.

Dutch is one of the languages which show the intriguing phenomenon of the dative alternation. Many verbs that occur in the double object construction can also occur with an aan-object instead of a nominal indirect object. The sets of verbs which can appear in both constructions are not identical though: there are verbs that can only occur in the double object construction, for example, benijden ‘envy’, and there are verbs with which prepositions are obligated, for example, spenderen ‘spend’.¹ The dative alternation, however, was not always present in the history of Dutch.

Dutch has been characterised for centuries by the ongoing decline of the case system or deflexion, likewise in the 16th-century (Van Der Horst 2008:795). Because of deflexion, it became harder to distinguish the accusative from the dative and therefore it also became harder to distinguish the direct object from the indirect object. As the Middle Dutch double object construction had a relatively free word order (Rens 2014), syntactic ambiguity arose. To avoid this syntactic ambiguity, the indirect object could be marked with a preposition more often. This derives from the horror aequi principle, ‘the widespread (and possibly universal) tendency to avoid the unmotivated recurrence of identical and adjacent grammatical elements or structures’ (Rohdenburg 2007:220). Van Der Horst (2008:801) demonstrates with a 16th-century example that the preposition is indeed penetrating the indirect object:

In a small-scaled corpus study, Weijnen and Gordijn (1970:127) find only a few debatable examples of the aan-construction in the 16th-century. According to them the examples still demonstrate a strong locative meaning. Weijnen and Gordijn (1970), and later Duinhoven (2003) too, assume that the productive use of the preposition aan in combination with verbs of giving is a post-medieval development. Colleman (2010:294) rightly observes that both studies do not fulfil the requirements of modern corpus research and deduces that, if their assumptions are correct, the use of aan as a prepositional marker of the ‘source’-object is older than the use of aan as a prepositional marker of the recipient. Colleman (2010:295) bases that statement on examples of the 13th-century aan-construction from the VMNW (Pijnenburg 2001) (‘Early Middle Dutch Dictionary’). Research on the Middle Dutch dative alternation (Rens 2014) has proven that the verb nemen ‘take’ was indeed compatible with the Middle Dutch aan-construction. Those observations are a good motive for further research on the semantic evolution of the aan-construction.

There is not much more known about the rise of the ‘transfer’ meaning of the aan-construction. Geleyn and Colleman (2015) examine the extent to which the 17th-century aan-construction covers the semantic field of the present-day construction, and how the construction has developed through time utilising a diachronic collexeme analysis. They observe that the semantic range of the aan-construction in the 17th-century Dutch covers well-nigh the complete semantic range of nowadays, and they call for a study on the 16th-century aan-construction. Everything seems to indicate that the ‘transfer’ meaning is a 16th-century innovation, which has developed very quickly and by the first half of the 17th-century has taken up its spot in the semantic network of the aan-construction. According to Colleman (2010), the aan-construction has expanded to the peripheral ‘transfer’ meaning from the central ‘contact’ meaning of the construction:

To illustrate the difference between the more concrete ‘contact’ and the more abstract ‘transfer’ meaning, compare (2), (3) and (4):

Case (2) is a literal contact situation in which the aan-construction is used. Case (3) is a concrete transfer in which a moment of contact also takes place between the direct object and the indirect object. In (4), however, there is no physical contact anymore between the direct object and the indirect object, though a transfer is still taking place. Colleman (2010) suggests the ‘transfer’ meaning of the aan-construction is a metaphorical extension of the original ‘contact’ meaning. The ‘transfer’ meaning of the aan-construction has further evolved and almost lost its ‘contact’ meaning. We will test this hypothesis in this study.

In the first section of our study, we will approach the aan-construction independently to get an overview of the semantic range of the construction. On the basis of the semantic network of Geleyn and Colleman (2015:220) we will examine the meaning extensions of the 16th-century aan-construction. The different meaning extensions we can distinguish are provided in Table 2.

The semantic area of the 16th-century aan-construction is rather limited in contrast to the 17th-century aan-construction (Geleyn & Colleman 2015). A first large set of verbs that we find in our corpus data belongs to the first semantic extension: the subject acts to create a contact relationship between the direct object and the aan-object. Within the first semantic extension Geleyn and Colleman distinguish four subdivisions, namely concrete transfers (5), abstract transfers (6), actions (7) and communicative transfers (8). Note that Geleyn and Colleman regard actions and communicative transfers as subtypes of abstract transfers. All four subtypes of the first semantic extension can be found in our corpus:

In addition to the first semantic extension, the eighth semantic extension, wherein the aan-object acts as a starting point for the contact relationship between the subject and the direct object, is also very frequent in our data set. The aan-object does not represent a recipient in that case but a kind of source object. Semantic extension 8 is strongly represented by, inter alia, the verb nemen ‘take’ (9) and a number of verba sentiendi (10–11):

Een exempel nemen aan ‘take an example from’ is an expression that we still know nowadays, albeit with eenvoorbeeld ‘an example’ as direct object, and was already frequent in Middle Dutch (Rens 2014). The verba sentiendi or verbs of perception are perhaps the first verbs that could occur in the aan-construction. When we search the ONW (Schoonheim, Louwen, Mooijaart, Pijnenburg & Quak 2009) (‘Old Dutch Dictionary’) for the lemma ana, we find this instance of the aan-construction:

Although the analysis of the aan-constituents in (10)–(12) is up for discussion, it is likely that such uses have played an important role in the emergence of constructions in which the aan-constituent unambiguously acts as an object.

The other semantic extensions are not represented in our corpus. It is, therefore, likely that the 16th-century aan-construction was limited to semantic extensions 1 and 8. The remaining semantic extensions are in that case only subsequent semantic developments of the aan-construction. The numbers of Geleyn and Colleman (2015:214) seem to confirm that claim. Although each of the categories in Table 2 was detected at least several times, the majority of the 1 530 aan-constructions they have found in their 17th-century corpus belongs to the semantic extensions 1 (79.3%) and 8 (10.7%). The other semantic extensions are less common with a maximum of 41 (2.7%) attestations per semantic extension.

The diachronic collexeme analysis is a variant of the distinctive collexeme analysis designed by Gries and Stefanowitsch (2004). The distinctive collexeme analysis is a statistical test which is used for competitive constructions, such as the double object construction and the aan-construction, in order to detect distinctive collexemes, that is, lexemes which have a (significant) preference for one of the tested constructions. Gries and Stefanowitsch use the distinctive collexeme analysis for the English dative alternation in order to determine which verbs prefer the double object construction and which the to-dative.

Although the distinctive collexeme analysis is normally used for synchronic data, Hilpert (2006) suggested to extend the use of the statistical test to diachronic language data. This application is called the diachronic collexeme analysis. Unlike the distinctive collexeme analysis, which compares the preference of verbs for competitive constructions, the diachronic collexeme analysis compares the preference of verbs for the same constructions over different time periods. The diachronic collexeme analysis can be useful to uncover shifts in the constructional semantics on the basis of the preference of verbs for a certain period (Hilpert 2006:243). Hilpert concretises his method by calculating the distinctive collexemes of the shall-construction in three periods: 1500–1640, 1640–1780, 1780–1920. On the basis of the results of his analysis, Hilpert shows that the shall-construction has gone through a process of subjectification.⁴

In the table below we present the results of our diachronic collexeme analysis for the four quarter centuries of the 16th-century, calculated with Gries (2007). The verbs showing a significant preference (95% confidence level) for a quarter century are listed per quarter century. Next to each verb the number of times that verb is found in each period is indicated. The second number indicates the collostructional strength.⁵ The higher the collostructional strength, the more the collexeme occurs in that period or construction. The collostructional strength is calculated by comparing the distributions of every collexeme and construction concerned. Note that it is possible for a collexeme to occur more in period x than in period y, but still show a preference for period y as it occurs more than other verbs in period y. For example, the verb slaan ‘hit’ is a distinctive collexeme for the first quarter century but was only attested three times in the first quarter century and five times in the second quarter century, 0 times in the third quarter century and finally one time in the fourth quarter century. The collostructional strength of the verb slaan is 2.15. If the collostructional strength is higher than 1.33, the collexeme is significantly distinctive for a certain period or construction. That means that the use of zien ‘see’ in the aan-construction occurs significantly more often in the fourth period than we would have expected from a random distribution of all attestations over the four quarter centuries. Eventually 22 collexemes prove to be distinctive: four for 1501–1525, four for 1526–1550, six for 1551–1575 and eight for 1576–1600 (Table 3).

Based on the results of the diachronic collexeme analysis we can draw up some hypotheses which we will test further in this study. Before we start the discussion of our results, we want to point out that we have compared relatively short periods of time. The semantic shifts that we observe are therefore not as pronounced as the shifts that Geleyn and Colleman (2015) establish in their research. In addition, the smallest differences are also statistically significant because of that. Nevertheless, we can note some interesting findings for the semantic evolution of the 16th-century aan-construction.

Before we discuss these findings, we would like to dwell on a discrepancy in our corpus. In the second quarter century, the verb verbeuren ‘forfeit’ is much more common than in the other quarter centuries, making the verb have the highest collostructional strength (11.62) of the investigated verbs. That high score, however, is due to the legal text genre that dominates the corpus of the second quarter century. Verbeuren ‘forfeit’ is, in fact, a verb found primarily in legal language, as evidenced by the definition of verbeuren ‘forfeit’ in the WNT (De Vries & Te Winkel 1882–1998) (‘Dictionary of Dutch Language’): ‘By your own fault losing anything worthwhile, respectively the right or the enjoyment of it’. That the verb verbeuren ‘forfeit’ shows a preference for the second quarter century has more to do with the content of the corpus, than with any semantic shifts. Obviously apart from the discrepancy in our corpus, there are some remarkable findings in our corpus.

Verbs with which the aan-object acts as a starting point for the contact relationship between the subject and the direct object appear to have at least one distinctive collexeme in every quarter century: for the first quarter century the distinctive collexeme is smaken ‘taste’ (1.33), for the second quarter century hebben ‘have’ (5.51), for the third quarter century gewaarworden ‘sense’ (2.39) and finally for the fourth quarter century zien ‘see’ (3.35), bevinden ‘find’ (1.71) and merken ‘perceive’ (1.52). Semantic extension 8 (see the semantic field section) is constantly present throughout the 16th-century.

In the last quarter century, the use of the ‘transfer’ meaning of the aan-construction seems to increase. Although we find transfer verbs in the remaining quarter centuries [voortkondigen ‘proclaim’ (1.33), aanleggen ‘lay on’ (1.63), verhalen ‘recount’ (1.54), inbrengen ‘bring in’ (1.59) and zweren ‘swear’ (1.50)], those transfer verbs show a relatively low collostructional strength in comparison to the distinctive transfer verbs of the fourth quarter century [bewijzen ‘prove’ (3.56), geven ‘give’ (1.71) and to a lesser extent schrijven ‘write’ (1.44)]. It is, therefore, likely that the ‘transfer’ meaning of the aan-construction was gaining ground during the 16th-century and that the aan-construction gradually became a full-fledged competitor of the double object construction in the ‘transfer’ meaning.

The ‘transfer’ meaning of the aan-construction is a metaphorical meaning extension of the ‘contact’ meaning. That ‘contact’ meaning emerges in the use of the verbs koppelen ‘connect’ (2.41), verbinden ‘connect’ (3.19) and hangen ‘hang’ (1.71). The verbs are distinctive for the third or fourth quarter century. Since these verbs are often combined with an aan-constituent that can be interpreted as a spatial prepositional phrase, we have been very strict during the manual analysis: once the aan-constituent could be replaced by a prepositional phrase with a different spatial preposition, the aan-constituent was considered as a locational prepositional phrase and therefore not taken into account. Since koppelen ‘connect’, verbinden ‘connect’ and hangen ‘hang’ clearly point to ‘contact’, it is not surprising that they often occur in the aan-construction.

Based on the increase in the use of transfer verbs in the aan-construction, we conclude that the 16th-century aan-construction is more and more open to metaphorical contexts. The question now is whether we can observe the same evolution in the relationship between the aan-construction and the double object construction in 16th-century Dutch, which we will examine in the next section.

Instead of comparing historical periods of a construction (see the diachronic collexeme analysis section) the distinctive collexeme analysis compares competing constructions within the same time period in order to ascertain which lexemes have a preference for one of the tested constructions. Since the distinctive collexeme analysis works the same way as the diachronic collexeme analysis, there is no need for further explanation about the operation of the statistical test. Table 4 presents the results of the distinctive collexeme analysis for the second quarter century. Only 21 out of the 27 examined verbs occur in the second quarter century in one or both constructions. The remaining verbs are therefore not included in the statistical test. The distinctive collexeme analysis (CollStr > 1.30; p < 0.05) yielded nine distinctive collexemes: three for the double object construction and six for the aan-construction. The table also shows the non-significant verbs and their preferred construction, because the test is based only on a small data set, so that even small shifts in constructional preference can be observed.

Among the verbs that show a preference for the aan-construction, we can distinguish two major groups. The first group of verbs are concrete transfers. Thus, there is a spatial displacement of the direct object. This verb group includes the significant verbs zenden ‘send’ (4.11) and brengen ‘bring’ (2.01), and the non-significant verbs overdragen ‘transfer’ (0.92), dragen ‘carry’ (0.76) and overzenden ‘ship’ (0.54). The distinctive collexemes of the double object construction, geven ‘give’ (19.68), schenken ‘gift’ (2.22) and verkopen ‘sell’ (1.93), on the other hand can represent both concrete and abstract transfers. Although wedergeven ‘give back’ (0.54) and overbrengen ‘transfer’ (0.27) only indicate concrete transfers, they also show a preference for the double object construction. However, their preference for the double object construction is not significant. All other verbs that show an (albeit not significant) preference for the double object construction only denote abstract transfers, actions or communicative transfers. So it seems that in the second quarter century especially verbs that denote concrete transfers have a preference for the aan-construction.

With the verbs zenden ‘send’ (4.11), schrijven ‘write’ (2.24) and brengen ‘bring’ (2.01), there is also a degree of conceptual distance between the subject and the indirect object. Thompson and Koide (1987:405–406) indicate that the conceptual distance between the agent and the recipient correlates with the ‘linguistic’ distance between the agent and the recipient in the sentence. In the aan-construction, the distance between the two roles becomes bigger, so the aan-construction appears to be a suitable way to indicate the distance to be covered.

The second group of verbs that prefer the aan-construction are the verbs with which the aan-object realises a source role or a dual role. Verhalen ‘recount’ (7.80) and nemen ‘take’ (1.88) are verbs with which the aan-object realises a source object. Note that the meaning of verhalen ‘recount’ is different in the 16th-century from nowadays. One of the meanings of verhalen we can find in the VMNW (Pijnenburg 2001) reads ‘recover, demand, (re)claim. In the construction iet verhalen ane enen’.

Using the verb verzoeken ‘request’, we will explain what is meant by the dual role of the aan-object. The aan-object mainly fulfils the role of recipient with the verb verzoeken ‘request’, as it is the recipient of the request. However, the subject expects (to receive) something of the aan-object, as is illustrated by the following example:

The aan-object in (13) performs a dual role: it is the recipient of the request, but at the same time it is the supposed ‘source’ of peace. That interpretation of the object is certainly up for discussion but does not sound that illogical, as the aan-object also takes the role source with the verbs verhalen ‘recount’ and nemen ‘take’. Colleman (2010) already pointed out that the aan-object occurred as a source object before it occurred as a recipient. Both findings would seem to indicate that the ditransitive transfer in the aan-construction emerged from the use of privative verbs, that is, verbs of dispossession, in the aan-construction. Verbs with which the aan-object occupies a so-called dual role, such as verzoeken ‘request’, maybe formed a bridge between both uses of the aan-construction.

Again, we carry out a distinctive collexeme analysis, but this time for the fourth quarter century. In the fourth quarter century, 25 of the 27 examined verbs occur at least once in one or both constructions and are thus taken into account. Eight of them were found to be distinctive collexemes: two verbs for the double object construction and six verbs for the aan-construction (Table 5).

Among the distinctive collexemes of the aan-construction we mainly find the same verbs as in the second quarter century: schrijven ‘write’ (7.61), nemen ‘take’ (6.13), verzoeken ‘request’ (4.96), verhalen ‘recount’ (3.40) and zenden ‘send’ (2.05). A few shifts in the collostructional strength can be observed; however, they are not remarkable. Overbrengen ‘transfer’ (1.35) is a new distinctive collexeme and fits within the group of concrete transfers. It is also an example of the type of particle verbs with a strong preference for the aan-construction in present-day Dutch (Colleman 2009a:605). It is remarkable that overzenden ‘ship’ nowadays has a strong preference for the aan-construction, but in the fourth quarter century a (non-significant) preference for the double object construction (0.21). However, the figures of overzenden ‘ship’ are low, and it does show a slight preference for the aan-construction in the second quarter century. Also the verbs brengen ‘bring’ (0.73) and dragen ‘carry’ (0.29) still show a preference for the aan-construction, but no significant one, as the CollStr is not higher than 1.33. Also the ballistic verb werpen ‘throw’ (0.29) shows a non-significant preference for the aan-construction.

Among the other verbs showing a (non-significant) preference for the aan-construction, we find a whole new group of verbs, that is, verbs which denote abstract transfers: the action verbs doen ‘do’ (0.70), plegen ‘commit’ (0.67) and bewijzen ‘prove’ (0.38), and the communicative verbs tonen ‘show’ (0.44) and wijzen ‘point’ (0.43). While those verbs still preferred the double object construction over the aan-construction in the second quarter century, they prefer the aan-construction in the fourth quarter century. From this we can cautiously conclude that the collostructional semantics of the aan-construction has expanded to include abstract transfers during the 16th-century.

The majority of verbs indicating abstract transfers still show a preference for the double object construction: zeggen ‘say’ (1.69), toeschrijven ‘attribute’ (1.26), vertellen ‘tell’ (0.84), onderwerpen ‘subject’ (0.42) and toestaan ‘allow’ (0.31). Verbs that can realise both concrete and abstract transfers, geven ‘give’ (16.95), schenken ‘gift’ (0.63) and verkopen ‘sell’ (0.15), decreased in collostructional strength but still show a preference for the double object construction. Based on these findings we can conclude that not all abstract transfer verbs experienced a sudden reversal in constructional preference, though the meaning of the aan-construction has clearly expanded by its use of abstract transfers.