3.1 Definition of the construction

The Finnish RC is a non-finite clause construction that generally consists of a participle verb form combined with a marking that expresses the semantic subject of the participle. The semantic subject is marked either by a noun or pronoun phrase as a genitive attribute of the participle (when the semantic subject differs from that of the matrix clause, DS) or by a possessive suffix added to the verb form (same subject, SS). The construction broadly corresponds to a subordinate että (‘that’) clause. It also has some clause-like properties, such as the distinction between past and non-past tenses and the use of a passive form. The morphological variants of the participle endings used in the Finnish RC are shown in Table 2.

Table 2. The morphological variants of participle used in the RC (VISK:§538)

As shown, the semantic subject of the RC is marked with a possessive suffix when it is coreferential with the matrix verb (example 4 below). When the subject is not shared, it is expressed with a genitive noun or pronoun phrase (example 3). The passive marker -tA or -ttA- (the form depends on the stem, consonant gradation and vowel harmony) may occur before the participle ending. In the past passive form, the passive and participle markers fuse into the affix -tU/dU- (VISK:§110).

Historically, the case of the different subject noun phrases was accusative, and the final n-marking in the participle endings shown in Table 2 was originally an accusative case marking. Due to historical phonological changes, the singular marking of the accusative case changed from -m to -n, making it coincide with the genitive case. The history and development of the semantic subject case marking in the RC has been studied by Forsman Svensson (Reference Forsman Svensson1983) and Ikola (Reference Ikola1960). In this paper we focus on diachronic changes in the frequency of the structure and its matrix verbs and do not systematically consider the case marking of the phrase expressing the semantic subject.

Following phonological changes, the case of the subject phrase was reanalysed as genitive, and the participle endings are now understood as fixed units (VISK:§538). The historical accusative ending -m was not used in the plural; instead, a nominative form was used. However, through analogy, in the RC, the genitive case was applied to plural occurrences as well (Itkonen Reference Itkonen1966:330; see also Penttilä Reference Penttilä1963:632, Forsman Svensson Reference Forsman Svensson1983:5). The history of the structure can be seen in Old Literary Finnish (see Section 3.2), where the plural nominative–accusative form is sometimes used instead of the genitive (Ojansuu Reference Ojansuu1909:145, Forsman Svensson Reference Forsman Svensson1983:5–6).

We gloss the participle endings as ptcp, separating the former accusative ending -n, and the case of the noun phrase expressing the RC subject as gen/acc. In the plural, we gloss the noun phrase case as gen.

The alternation between a noun phrase and a possessive suffix in marking the semantic subject of the RC, which can be understood as a type of switch-reference system (for switch-reference in general, see e.g. van Gijn & Hammond Reference van Gijn and Hammond2016), was already well established in Old Literary Finnish and has remained similar to modern times (see Section 3.2). In modern Standard Finnish, the RC does not have negation or other moods beyond indicative (VISK:§538). However, in Old Literary Finnish, negated RCs were sometimes used (see Section 3.2).

Three subtypes of RCs can be distinguished. The first type, which most indisputably corresponds to an että clause and is the most typical form of the RC in written Finnish, functions as an object for a transitive verb (see examples 3 and 4).

The second type occurs with an intransitive sensory perception verb. In this type, the matrix predicate is congruent with the construct’s subject (see example 5). Therefore, constructions of this type involve, for example, evidence based on visual or auditory perception. RCs of the second type more closely resemble the Estonian structure from which Modus Obliquus evolved (Campbell Reference Campbell, Closs Traugott and Heine1991:287).

In spoken Finnish dialects, the distribution of the first and second types of RCs is more balanced than in the written texts. In the Finnish Dialect Corpus of the Syntax Archive, 59.4% of the RCs represent the first type. In written genres, the first type is more clearly dominant, with only about 20% of the RCs representing the second type (Ikola et al. Reference Ikola, Palomäki and Koitto1989:469). In modern conversational data, the frequencies are even: half of the RC occurrences (17/34) found in an annotated corpus of Finnish everyday conversation (ArkiSyn) represent the second type, with most of these (15/17) featuring the verb näyttää (‘to seem’) as the matrix verb. This suggests that in conversational speech, the structure tends to be somewhat fixed, which aligns with the general characteristics of spoken language syntax (Hakulinen & Leino Reference Hakulinen and Leino1987:42). Since RCs are particularly common in narratives, variations in spoken datasets are likely to occur, for example, between conversations and interviews.

Interestingly, in our searches on the 1548 New Testament, only two occurrences of the second type of RCs were found, both accompanying the verb näyttää. It seems that in the oldest translation, another structure – one not reached by our corpus search – was used to convey similar meanings to those of the RC with intransitive perception verbs found in dialects and later texts. Thus, our analysis focuses on the first type of RC.

The construction can also act as a subject for intransitive verbs that can attract an että clause as a subject (e.g. ilmetä ‘to appear, to turn out’ or paljastua ‘to be revealed’). It has been suggested that the subject-acting construction is the newest addition among RC types, and it is rarely used even in modern Finnish (Ikola et al. Reference Ikola, Palomäki and Koitto1989:467). Since this type does not occur within our data, it will not be discussed in further detail here.

There are differences between spoken and written language and between different text genres in terms of the proportion of RCs sharing a subject with the matrix clause. According to Forsman Svensson (Reference Forsman Svensson1983:175), in biblical language the majority of RCs have a different subject, while in legal language the distribution is more even. The distribution of SS and DS occurrences has not been examined in the Finnish Dialect Corpus. However, our search in the ArkiSyn corpus suggests that modern conversational speech may prefer RCs with coreferential subjects. Although only 17 occurrences of the first type of RCs were found, 12 of them share a semantic subject with the matrix clause. Given the infrequency of the structure in the corpus, more data are needed for firm conclusions, which will be left for future studies. The matrix verbs of the first type of RCs are more varied in conversations than those of the second type.

Our data, consisting of different versions of the same text, allow us to examine possible changes in RC usage, particularly in terms of shared and different subjects, regardless of the text topics. Comparison of our findings with other genres will be reserved for future studies.

3.2 Referative constructions in Old Literary Finnish compared to the modern standard

In this section we explore the changes that RCs have undergone during the development of the Finnish written language, drawing on previous research, and how the guidelines concerning them in standard language manuals and grammar books have been established. The status of participle forms in grammar guides has been discussed in more detail by Forsman Svensson (Reference Forsman Svensson1983:12–23).

In Old Literary Finnish, the RCs were still developing, and previous research noted that forms deviate from the modern standard. The reanalysis of the construction – where the NP began to be understood as the subject of the non-finite verb form rather than the object of the matrix clause – was still ongoing. Plural forms, in particular, included NPs in the nominative–accusative case. When the NP was still interpreted as the object of the matrix verb, the structure more closely resembled the Latin ACI structure (Forsman Svensson Reference Forsman Svensson1983:5).

The structure also exhibited other variations. Negative RCs were sometimes used, whereas today the construction lacks a true negative form. Negation can now only be expressed in the matrix clause, which can cause ambiguity in certain cases (Ikola Reference Ikola1960:65, Hakulinen Reference Hakulinen1968:465–466). Negation is thus presented as one of the factors that favour the use of a subordinate clause instead of an RC (Forsman Svensson Reference Forsman Svensson1983:94). Of the 663 että clauses collected from the 1548 New Testament, 97 are negated (14.6%). Negation can affect the frequency of matrix verbs: e.g. verbs tietää ‘to know’ (15 occurrences) and sanoa ‘to say’ (11 occurrences) are also used with negated että clauses (see Section 5.3). In later editions, the number of negative että clauses decreases to the 50 occurrences (7.5%) of the 2020 edition, suggesting a preference for simpler expressions. Negative RCs are found in the three earliest editions (1548, 1642, 1776; eight occurrences altogether). In other Old Literary Finnish texts, negative RCs were found up until the nineteenth century (Häkkinen Reference Häkkinen1994:402, 477).

In Old Literary Finnish, another participle construction was used in functions similar to RCs. In this construction, a participle was marked with a translative case instead of an accusative (example 6).

Forsman Svensson (Reference Forsman Svensson1983) thoroughly examined the seventeenth-century translative construction, comparing it to RCs. Although we did not include the translative construction in our data search, a considerable number of RCs in the 1642 edition were changed to this construction (see Section 5.1). According to Ikola (Reference Ikola1960:96), the use of the translative case in referential contexts was typical of old spoken Finnish and was found in old dialects and other Finnic languages. It is said to be a feature of early written Finnish (e.g. Siro Reference Siro1964:145). In contrast to the modern RC, in structures expressing reported speech using the translative case, the NP corresponding to the subject of the RC is still interpreted as the object of the matrix verb (Forsman Svensson Reference Forsman Svensson1983:6). In Old Literary Finnish, there was no clear distinction between the functions of RCs and translative constructions (Forsman Svensson Reference Forsman Svensson1983:8–9), and grammar manuals presented the forms as alternatives (e.g. Wikström Reference Wikström1832:56).

During the twentieth century, the functions of the translative construction became distinct from RCs and now it is mainly used with some non-factive cognitive verbs (Forsman Svensson Reference Forsman Svensson1983:291). Example 7 shows how the RC occurring with the verb luulla (‘to think, to assume’) in the 1776 Bible was reformulated as an NP in the translative case in the 1938 translation. In more recent translations in our data, a constituent in the translative case in such constructions is usually a noun or adjective phrase instead of a participle verb form.

The first Finnish language guidebooks were published in the seventeenth century to teach Finnish to civil servants and foreigners who knew Latin (Vihonen Reference Vihonen1978:39–40, Häkkinen Reference Häkkinen1994:28). The seventeenth- and eighteenth-century grammar books (Petraeus Reference Petraeus1649, Martinius Reference Martinius1689, Vhael Reference Vhael1733) do not contain rules or examples of RCs. In the nineteenth century, some books describe RCs, but the descriptions are scarce, incomplete, and scattered throughout the books (Forsman Svensson Reference Forsman Svensson1983:12–23). Examples of variations not found in our materials are sometimes given (e.g. Judén Reference Judén1818:49). The correspondence with the Latin ACI construction is mentioned in some books as late as the early twentieth century (Forsman Svensson Reference Forsman Svensson1983:32).

The rule for semantic subject marking alternation between a genitive noun phrase and a possessive suffix is spelled out in Koriander’s (Reference Koriander1859:7) grammar book. However, as mentioned earlier, the system of marking switch-reference was relatively consistent already in the 1548 translation: only 13 exceptions (out of 98 SS RC occurrences) were found where a shared semantic subject is marked by a third-person pronoun used as a reflexive pronoun and accompanied by a possessive suffix, resulting in forms like häne-ns (3sg-poss.3sg) or similar forms in other persons. By the eighteenth century, these occurrences were replaced by the reflexive pronoun itse (‘self’), and in the twentieth-century translations, by possessive suffixes. Our data include no exceptions in the other direction, where a DS RC would be marked by a possessive suffix. However, some grammar books present such examples (e.g. Setälä Reference Setälä1880:46).

In the twentieth century, the description of the RC becomes clearer in grammar guides (see e.g. Saarimaa Reference Saarimaa1930:87, Reference Saarimaa1967:194–195). Today, as mentioned in the introduction, grammar guides caution against the use of participle structures, which are considered prone to errors and can make the text difficult to read. Some inconsistencies in RCs persist, particularly in the case marking of the object in passive forms and the semantic subject case marking in existential occurrences (Häkkinen Reference Häkkinen1994:403; see also Hakulinen Reference Hakulinen1968:466). However, since passive and existential occurrences are marginal in our data and not the focus of our study, these variations are not presented here.

5.1 Distribution of constructions

Figure 1 illustrates the changes in the frequency of RCs and corresponding että clauses from 1548 to 1642. As detailed in Section 4, the data collection focused on RCs and että clauses identified in the 1548 New Testament. Consequently, other constructions present in the earliest version were not included in the data. Thus, it is not possible to assess whether Agricola’s other constructions were later replaced by RCs or että clauses in the 1642 or subsequent versions. In coding the later translations, direct quotes (DQs) and translative constructions are distinguished from other constructions. To maintain clarity in Figures 1–3, only changes involving more than five occurrences reformulated into another construction are marked with arrows.

Figure 1. Distribution and change of constructions of reported discourse, perception, and cognition from 1548 to 1642.

Figure 2. Distribution and change of constructions of reported discourse, perception, and cognition from 1642 to 1938.

Figure 3. Distribution and change of constructions of reported discourse, perception, and cognition from 1938 to 2020.

Previous studies (Ikola et al. Reference Ikola, Palomäki and Koitto1989, Häkkinen Reference Häkkinen1994:495–496, Mielikäinen Reference Mielikäinen2014:38) have noted that the 1642 translation – as well as the 1776 version – generally favours non-finite structures. Rapola (Reference Rapola1942:17) provided examples of verses where että clauses used by Agricola were converted to RCs in 1642, suggesting that että clauses may have resulted from literal translation. Figure 1 reveals a significantly higher number of että clauses in the 1548 translation compared to the 1642 edition.

Itkonen-Kaila (Reference Itkonen-Kaila1997:72–73) argued that the influence of classical languages is most evident in the Gospels, which are narrative texts, while Swedish and German influenced the structures in the New Testament Letters, where Agricola followed Luther’s explanations. Our quantitative analysis of the frequency of RCs and subordinate clauses revealed significant differences between books, although they may be more closely connected to the topics of the texts than to the source texts. In the 1548 translation of the Letters, the proportion of että clauses is higher than in other sections (88% versus the average 72%, p = 0.039). However, when considering the ratio of RCs to että clauses, attention is particularly drawn to the Book of Revelation rather than the Gospels. In Acts of the Apostles, että clauses account for 60% of the data and in the Gospels 52%, differences to the average not being statistically significant. In Revelation, however, että clauses only account for 21% (p < 0.001). The high frequency of RCs in Revelation was also noted by Forsman Svensson (Reference Forsman Svensson1983:130). As the Book of Revelation is largely a report of what was seen and heard, the frequency of RCs is at least partly explained by the number of occurrences of the matrix verbs nähdä (‘to see’) and kuulla (‘to hear’), which favour RCs (see Section 5.3).

Figure 1 demonstrates that many että clauses were changed to RCs, DQs, and other constructions in the 1642 edition. The shift from subordinate clauses to DQs is linked to the fact that many että clauses, used as the object of the verb sanoa (‘to say’), were converted into DQs by omitting the conjunction. Another factor favouring non-finite structures in general is that both subordinate clauses and RCs were changed to non-finite structures formed with the translative case (see examples 6 and 7 above).

Figure 2 shows how the distribution of constructions evolved further in the following editions from 1776 and 1938. From 1642 to 1776, changes were minor. However, the tendency to favour RCs seemed to be reversing, with RCs being replaced by subordinate clauses and other constructions. In contrast, in 1938, significant changes were made. Many verses were completely reformulated: both että clauses and RCs were transformed into other structures. Eighteen että clauses that had been changed in 1642 to translative constructions (and preserved in 1776) reverted to että clauses in the 1938 edition, while five were changed to RCs. However, of the RCs changed to translative constructions, fewer (five occurrences) reverted to RCs, with more being changed into other structures (eight occurrences), such as essive case marking (example 8). This trend aligns with Forsman Svensson’s (Reference Forsman Svensson1983:291) observation that in the twentieth century, the meaning of the translative construction diverged from that of RCs.

In the 1938 edition, että clauses, in particular, were replaced with other constructions. Similar changes occur in the newer editions as well, reflecting an effort to modernise biblical language by employing structures more natural to the Finnish language. In our analysis, we coded as other constructions instances where, for example, a supporting pronoun is added to the construction (example 9), or where the head of the että clause is changed from a verb to a noun (e.g. toivoa ‘to hope’ replaced by toivo ‘hope’). Että clauses were also replaced by indirect questions, especially ones beginning with the word kuinka (‘how’).

Figure 3 illustrates that in the three most recent translations examined, the changes are multidirectional. RCs decrease significantly between the 1938 and 1992 translations. In this period, many RCs used with the verbs nähdä (‘to see’, 36 occurrences) and kuulla (‘to hear’, 16 occurrences) are reformulated (see Section 5.3). With these verbs of sensory perception, in the later editions, a relative clause (example 10) or an indirect question beginning with kuinka ‘how’ is often used. The RC frequency between the data from different books of the Bible varies between 22% (Revelation) and 13% (Letters) even in the latest version, but the differences are no longer statistically significant.

Our analysis shows that että clauses and RCs are often genuine alternatives to each other, as they are used interchangeably in the same verses across different editions. The number of RCs decreases most markedly in the two most recent translations, while the number of että clauses remains relatively stable over the centuries. Despite the cautious attitude of contemporary prescriptive grammar guides towards these non-finite structures, RCs are still widely used in the most recent translations for mobile device users and audiobook listeners.

However, it should be noted that our data collection was based on the structures of the first translation. In our data, both että clauses and RCs decrease as they are converted into other structures. Although these other structures sometimes reverted to earlier formulations, a limitation of our data is that they do not capture whether more recent editions translate some verses – where Agricola used other structures – as RCs or että clauses. It remains for future research to determine the extent to which the second and third types of Finnish RCs are used, as these are rare or absent in the present data.

5.2 Same-subject and different-subject constructions

In Figure 4 we show how RCs and että clauses are distributed into SS and DS occurrences in each translation. In the first three translations, no statistically significant difference between että clauses and RCs was found regarding whether they express SS or DS structures. However, in the 1938 translation, a difference becomes apparent and increases further in the 1992 translation: RCs are reduced, particularly in structures where the semantic subject of the RC is different from that of the matrix clause. As presented in Section 3.1, genitive attributes of DS RCs make the structures more complex than SS RCs, which are marked with possessive suffixes. At the same time, the number of SS että clauses decreases. We assume that from the twentieth century onwards, translators have paid increasing attention to the complexity of structures. According to Hanna Lappalainen (oral communication), a linguist involved in the 2020 New Testament translation working group, RCs were not deliberately avoided in the text, but sentence length was considered. In the most recent translation, the number of different-subject RCs does not decrease further, while the number of SS että clauses slightly increases.

Figure 4. The number of RCs and että clauses having the same or different subject as the matrix clause in the Finnish Bible editions.

In Section 3.1 we presented a small search in the contemporary conversational Finnish corpus (ArkiSyn), considering SS and DS RCs. Although only a few occurrences were found, their distribution suggests that in everyday conversations, SS RCs could account for more than half of the use of the first type of RCs. Of course, the topic of the text or conversation seems to play a role in the occurrence of this feature. We suggest that the tendency of modern biblical language to prefer SS RCs as the overall use of RCs declines may indicate that SS RCs are perceived as easier and more accessible structures. It would be worthwhile investigating changes in the frequency of SS and DS constructions further across different types of spoken and written language data available, for example, in newspapers or academic texts.

The distribution of SS and DS RCs is naturally linked to matrix verbs, as some of verbs used in the matrix clauses prefer DS occurrences. For example, the sensory perception verbs nähdä (‘to see’) and kuulla (‘to hear’), which are the most frequent matrix verbs in the data, are more commonly used with DS structures.

5.3 The set of matrix verbs used with referative constructions and että clauses

In this section we examine whether the set of matrix verbs used with RCs and että clauses changed over the time span from the sixteenth century to the present. Figure 5 depicts the number of verb lexemes used with the constructions examined in different editions of the Bible. There is no clear change in the number of verb lexemes used with RCs in the data, only a slight downward trend in the number of lexemes used with että clauses. This trend is explained, in part, by the addition of supporting pronouns in että clauses (see example 9 in the previous section).

Figure 5. Number of matrix verb lexemes for RCs and että clauses in the Finnish Bible editions.

Että clauses are more versatile than RCs in the data: in all translations, there are many matrix verbs with which että clauses are used but not RCs. For these verbs, some other non-finite structure is more typical than an RC as an alternative to an että clause. For instance, the verbs rukoilla (‘to pray’), neuvoa (‘to advise’), and pyytää (‘to ask’) are used with MA-infinitive illative forms (VISK:§121). The alternation between että clauses and MA-infinitive illative forms is already present between the 1548 and 1642 translations (see example 11).

Based on our intuition as native Finnish speakers, RCs would be possible with some relatively frequent matrix verbs that only occur with että clauses in the data. Such verbs include lukea (‘to read’, 10 occurrences in all editions combined), kirjoittaa (‘to write’, 35 occurrences), vannoa (‘to swear’, 15 occurrences), and rukoilla (‘to pray’, 173 occurrences). Whether this is due to influence from source languages could be determined by a systematic examination of the corresponding structures in the source texts. Some verbs, e.g. muistaa ‘to remember’, pelätä ‘to fear’, and tahtoa ‘to want’, were only used with että clauses in the first edition but later they also occur with RCs.

The number of verb lexemes used in the data with both että clauses and RCs remains similar throughout the period. The five most common matrix verbs in both structures – nähdä (‘to see’), kuulla (‘to hear’), luulla (‘to think, to assume’), sanoa (‘to say’), and tietää (‘to know’) – are consistently used in all six editions, although their frequency varies between editions.

Thus, these structures show no signs of crystallising to accompany only certain verbs. Most of the verbs that occur only with RCs in the data are individual occurrences, and we assume that in larger data, they could occur with että clauses as well. However, there are tendencies in the data for certain verbs to be significantly more frequent with one or another of the structures under consideration, despite their alternativeness. We show that these preferences also change in time.

As we presented above, the semantics of matrix verbs determine how natural it is to use them in SS or DS constructions. Of the typical matrix verbs of RCs, kuulla (‘to hear’) only occurs with DS constructions, as does nähdä (‘to see’), with a few exceptions. When the number of RCs used with these verbs was reduced in the 1938 edition, the number of DS constructions was also reduced.

When considering the verbs used in the data as matrix verbs of both SS and DS RCs, the tendency in the twentieth century still seems to be for the number of DS RCs to decrease. For example, of the 15 DS RC structures having the verb tietää (‘to know’) as a matrix verb in 1938, 13 were changed to tietää että clauses in the 1992 edition. Only two of the five SS RCs with tietää were changed: one into an että clause and the other by reformulating the verse completely.

Next we will look in more detail at the changes that have occurred in some of the most frequent matrix verbs that are used with both structures. We will focus on translations where the most significant changes in the number of structures were found. First, we examine the differences in the matrix verbs used between the 1548 translation and the subsequent 1642 translation. Then, we will look at how the 1938 translation differs from the 1776 translation and how the 1992 translation differs from the 1938 translation.

In the oldest translation of the New Testament (1548), both RCs and että clauses appear with a total of 21 matrix verb lexemes. Most of these verb lexemes only occur once or twice. Figure 6 shows the frequency of 7 most common matrix verbs (N > 20) used with RCs and että clauses in Agricola’s translation (1548). A separate bar marks the share of SS occurrences.

Figure 6. The most frequent matrix verbs of RCs and että clauses in the 1548 and 1642 editions.

As we have shown, between the 1548 and 1642 versions, the number of että clauses decreases and the number of RCs increases. Figure 6 shows that RCs increase with some of the most common verbs, particularly tietää (‘to know’) and kuulla (‘to hear’). The verb tietää generally prefers että clauses, likely because the phenomena to be known can be complex to verbalise. However, the share of RCs among the objects of the verb tietää first increases and then decreases (see Figure 7). The relationship between the two most common perceptual matrix verbs, nähdä (‘to see’) and kuulla (‘to hear’), is interesting: both prefer RCs but, in the early translations, nähdä more strongly than kuulla. Perhaps due to the analogy with nähdä, the number of RCs used with kuulla increases. In the twentieth-century editions, however, both decrease with RCs (see Figure 7). In the case of että clauses, a significant decrease occurs with the verb sanoa (‘to say’): many of these structures become direct quotes, so that only the apparently superfluous conjunction että is omitted, and the wording of the quoted part does not usually change much.

Figure 7. Matrix verbs of the RCs and että clauses in the 1776, 1938 and 1992 editions.

In the 1776 translation, both RCs and että clauses appear with a total of 16 matrix verb lexemes; in 1938, 21 lexemes, and in 1992, 22. The distribution of seven most frequent verbs (N > 20) is presented in Figure 7. The set of verbs differs from that shown in Figure 6 in that the verb ymmärtää (‘to understand’) is replaced by the verb tahtoa (‘to want’). In the 1548 translation, tahtoa was not used with RCs, and in 1992, ymmärtää only occurs with että clauses. The verb tahtoa ‘to want’ is a peculiar case in the sense that although its meaning would fit being used with both DS and SS RCs, only DS RCs are used. When the semantic subject of something that is wanted is the same as the person wanting, another non-finite form, the A-infinitive is used (e.g. tahdo-n lähte-ä want-1sg leave-inf ‘I want to leave’ instead of *tahdo-n lähte-vä-ni want-1sg leave-ptcp-poss.1sg).

From 1776 to 1992, the data presented in Figure 7 show a general decrease in the use of RCs, with a fairly steady decrease across most typical matrix verbs. Figure 7 also illustrates that for these verbs, DS RCs are the ones primarily declining. Regarding the complements of the verb luulla (‘to think, to assume’), almost all SS occurrences are RCs, while DS occurrences are että clauses. The verb tahtoa ‘to want’ behaves contrary to the observed tendency, as its use with RC increases in the 1992 edition. The 2020 translation, by the way, introduces to biblical language another verb with the same meaning, haluta,Footnote ⁷ which also occurs fairly evenly with both että clauses and DS RCs.

The structures associated with the verb sanoa (‘to say’) undergo another change in the twentieth-century editions, with both RCs and että clauses decreasing. The 1992 edition shows a broader range of matrix verbs than earlier translations, including väittää (‘to claim’) as well as other verbs indicating oral communication, such as puhua (‘to talk’), huutaa (‘to shout’), and kysyä (‘to ask’), which replace sanoa.