An object clause, also sporadically referred to as a comment clause in the literature (Warner 1982; Huddleston and Pullum 2002: 951), is the kind of clause functioning as the direct object of the matrix verb. In English the most common type of object clause is introduced by the complementizer that, as in I know that Peter will arrive soon, traditionally labelled as a that-clause (Quirk et al. 1985: 1049). It can be omitted, except in formal contexts, leaving an asyndetic zero that-clause, i.e. I know Peter will arrive soon. 1
Both constructions can nowadays be used interchangeably and with no apparent difference in meaning, even though various factors have been considered to account for the choice between the two alternatives. Elsness (1984: 519) lists four conditioning factors directly influencing the use of that or the zero link: (i) style, whether formal or informal, the former consistently favouring the use of that; (ii) the matrix verb, there being particular preferences towards the syndetic or the asyndetic construction depending on the nature and frequency of the matrix verb; (iii) potential ambiguity, that being preferred in cases where zero might lead to ambiguity, particularly when there are intervening elements between the matrix verb and the clause; and (iv) semantic contrast, that chosen where “the connective points to the preceding context”, thus retaining some of the anaphoric force that it used to have when it functioned as a demonstrative in previous stages of the history of English (Elsness 1984: 519). There have been other insights into the topic providing further evidence on the semantics of the construction (Storms 1966; Dor 2005), its discourse features (Thompson and Mulac 1991) or register variation (Biber 1999).
From a historical perspective, Jespersen justified the use of this complementizer in view of the existence of two originally independent sentences: I think: he is dead and I think that [demonstrative pronoun pointing to what follows, namely]: he is dead. Jespersen’s (1933: 350–351) historical explanation considered that “in the course of time that was accentually weakened […] and this weak that was eventually felt to belong to the clause instead of to what precedes, and by that very fact became what we now call a conjunction”. In the light of this, two questions are implicitly answered. The first one has to do with the nature of the original object clause link in English: it is not possible to consider the absence of that as a mere dropping of the conjunctive element (the terms omission or deletion being then inaccurate), but rather as the typical speech-based object link in the history of English (Rissanen 1991: 287–288; cf. Bolinger 1972: 14). The second has to do with the function of the zero link, as the use of that “helps to mark a clause boundary, and it tends to be deleted more as this function is less useful” (Warner 1982: 175). Therefore, zero is favoured in those contexts in which the boundaries of the clause are transparent, i.e. before other conjunctions and before pronouns. On the contrary, if a non-finite form of the verb or any other intervening element appears between the matrix verb and the clause, the use of that becomes more than expected. In this sense, some grammatical contexts of the matrix verb have been seen as constraints on the use of that or zero: the syndetic connector being often preferred if the verb occurs in a non-finite form (Rissanen 1991: 286), in the passive voice (Rohdenburg 1996, Rohdenburg 2006: 143–166) or when the matrix verb is negated (Suárez Gómez 2000: 193).
Zero that-clauses may be traced back to the early English written records (Mustanoja 1960; Fischer 1992: 313). In Old English, their frequency is low, the asyndetic forms mainly used when the subjects of the main and the subordinate clause are the same or before a complement representing the exact words of the reported proposition (Mitchell 1985: §1976ff; Traugott 1992; Palander-Collin 1997; Hosaka 2010). In Middle English, in turn, Warner estimates a rough figure of 4% of that-clauses with a zero link in John Wycliffe’s sermons (late fourteenth century), and observes that the phenomenon was governed by the following syntactic factors: the removal of the clause subject; the particular preferences of the matrix verb; the existence of intervening elements between the matrix verb and the clause; and/or the nature of the clause initial element, whether a noun phrase, a pronominal or a conjunction (Warner 1982: 170–173). The most detailed corpus-based historical survey of zero/that as complementizers, in our opinion, is Rissanen’s (1991). The use of the Helsinki Corpus allows him to show that the definite rise of zero in English takes place from the second half of the sixteenth century to the early seventeeth century, being more frequent in speech-based text types (trials, sermons) or in texts representing the oral mode of expression (comedies). After reaching its peak at the end of this century, the use of the asyndetic construction diminishes drastically in the eighteenth, plausibly as a result of the pronouncements of prescriptive grammarians (Rissanen 1991; 1999: 284–285; see also Sundby et al. 1991; Finegan and Biber 1995; López Couso 1996; Görlach 2001: 125–126).
In her study on “Finite complement clauses in Shakespeare’s English” (1990), Fanego also deals, among other aspects, with the use of that or the zero link in four plays by William Shakespeare: Romeo and Juliet (1597), The Merry Wives of Windsor (1602), King Lear (1606) and The Winter’s Tale (1623). The percentages obtained from the analysis reach 79.49% for zero and 20.51% for that, with a tendency for the former to increase in prose texts (84.17%) and diminish in verse (68.53%), in connection with formality/informality (Fanego 1990: 143). A similar pattern is observed when one or more constituents separate the subject of the complement clause from its matrix predicate, with a percentage of 92.59% when no elements intervene in contrast to 66.66% whenever they appear (Fanego 1990: 144–146). Such a high level of occurrences of the zero link lead Fanego to conclude that “Shakespeare’s predilection for the zero-clause is idiosyncratic and seems to have no parallels in either earlier or later periods” (Fanego 1990: 147). Suárez Gómez (2000) has also studied the behaviour of that and zero in the Helsinki Corpus between 1420 and 1710. She thoroughly analyses a number of internal linguistic constraints, such as frequency and specificity of meaning of the matrix verb, grammatical context –finite vs. non-finite, ditransitivity, use of auxiliary verbs – and presence or absence of intervening material, among others. She focuses on text-types within the most informal end of the register cline: drama and private letters. Rates of zero are high in both, but a higher incidence is found in the former (67.2%) when compared to the latter (53.1%), thus confirming the tendency for the zero link to be used in text-types approaching the spoken language. Interestingly, she also notices that the use of one connector or the other often depends on the addressee: “in letters addressed to a person considered inferior by the writer, object complement clauses are introduced by the complementizer zero” (Suárez Gómez 2000: 187).
Some of the remarks by Rissanen (1991), Fanego (1990) and Suárez Gómez (2000) point to the possibility that the diffusion of zero that-clauses followed the pattern of a typical change from below. In sociolinguistic terms, it is meant to be a “spontaneous process” (Guy 1990: 51) which is not consciously borrowed from external prestigious norms, as in the case of so-called changes from above. It appears in the vernacular in connection with internal linguistic factors, operating below the level of social awareness. Unlike changes from above, which in contemporary western societies tend to be initiated by members of the middle ranks, the leaders of changes from below usually belong to the upper working-classes (Labov 1994: 78).
There is not, to our knowledge, an empirical sociolinguistic corpus-based study of the phenomenon in the history of English. With this aim, the present paper studies the distribution of the syndetic and the asyndetic constructions between 1424 and 1681, covering the periods of the history of English known as late Middle English and early Modern English. Our aim is to shed some light on the sociolinguistic dimension of the phenomenon. As such, our paper is organised into five different sections. After this introduction, section two describes the scope and the corpus data upon which this study is based. Section three presents the distribution of this type of complementation in terms of the type of verb, while section four presents the models of social stratification for sociolinguistc research adopted in the analysis of early English and the results obtained when one of them is plotted on the relevant data. Finally, section five contains the conclusions.
The present study deals with the use and distribution of that-clauses in the light of the following verbs, i.e. to know, to think, to say, to tell and to hope, which have been chosen in view of their high frequency in the corpus. According to some previous approaches to the topic (Elsness 1984; Rissanen 1991; Suárez Gómez 2000), that-deletion is found to be more widespread in combination with high-frequency verbs rather than low-frequency matrices. Since the present study considers the sociolinguistic dimension of the phenomenon, it becomes a must for us to select high-frequency items as the ideal input to witness the difussion of the phenomenon in late Middle English and early Modern English. Following Rissanen’s classification (1991: 274), these items come to represent three verb types, namely verbal expression (say and tell), non-verbal mental activity or state (know and think), and modal mental activity (hope).
The analysis is exclusively concerned with those object clauses in which the subordinate clause immediately follows the matrix verb since it is the context in which the alternative zero/that is bound to occur, as shown in examples 1–2 below.
I suppose his sone woll say I have done hym some plesure in thes partes (wyatt, 159.026.893). 2
Trulie I doe verily thinke that I shall not goe out of my chamber this long time (oxinde, I,182.109.1613).
Other positional combinations have been excluded insofar as “zero is impossible when the object clause precedes the main clause and that is practically impossible when the subject of the main clause begins with a push down element of the object clause” (Rissanen 1991: 274). Constructions like those shown in the examples below have also been disregarded from our analysis considering that the verb to think is not really followed by the subordinator that, which functions as a demonstrative or a relativiser instead:
I think that will bee best (osborne, 55.025.1260).
[…] and that was it I think that spoiled it (osborne, 106.046.2471).
The data used as source of evidence come from the Parsed Corpus of Early English Correspondence (text version, henceforth abbreviated as PCEEC), a corpus especially compiled for sociolinguistic research, covering nearly three hundred years of the history of English, from 1410 to 1681. 3 The use of private letters is vital for historical sociolinguistic research for the following reasons. On the one hand, as non-anonymous texts, letters allow the reconstruction of psycho-biographical information about their authors and addressees and this favours a reproduction of the sociolinguistic variables that can be correlated with linguistic production: age, gender, education, professional background, social status, social network, mobility, etc. On the other hand, correspondence is accepted as the historical genre more likely to approach the oral vernacular, showing some characteristics that belong more to the spoken than to the written register: personal involvement, interaction or personal stance, among others (Biber 1995, Biber 2001; Biber and Finegan 1997: 265–267). Letters, therefore, and especially those compiled in the PCEEC, can confidently be used in reconstructing the linguistic processes from the past, such as variation and change, that are often bred in this medium (Nevalainen et al. 1996a; Nevala and Palander-Collin 2005; Nevalainen 2007; Conde-Silvestre 2007; Palander-Collin 2010; Elspaß 2012). The present study makes use of the late Middle English and early Modern English collections of letters in the corpus. The former covers a span of eighty years (1420–1499), although we have focused on the four well-known collections of fifteenth century private correspondence, the Cely Letters, the Paston Letters, the Plumpton Letters and the Stonor Letters, which, with a total of 377,414 running words, afford specific data on a varied range of informants from the period 1424–1499. The early Modern English part of the corpus is divided into three subperiods: 1500–1569 (309,220 words), 1570–1639 (910,675 words), and 1640–1681 (555,415), amounting up to 1,775,310 words altogether (Taylor and Santorini 2006). All collections of correspondence from each of these three subperiods have been analysed.
The automatic retrieval of the instances has been carried out with the use of a freeware concordance program, AntConc 3.2.2. (Anthony 2011). The process was not straightforward, however. Even though the use of the root of the verb prompted the automatic generation of many of the instances (i.e. hop*), other searches were needed on account of morphological variation (e.g. think ~ thinks ~ thinketh ~ thought) and spelling inconsistency (think ~ thynk ~ þink ~ þynk). Once the complete set of instances was generated, further disambiguation was required so as to select the examples complying with the scope of our research.
3 Historical analysis
This section deals with the chronological distribution of that-clauses in late Middle English and early Modern English, by looking at the corpus instances in the period 1424–1681 while it also intends to shed some light on the likely influence of the type of verb on the phenomenon.
The material from the Parsed Corpus of Early English Correspondence provides us with a total of 7,048 instances of object clauses for this period, of which 4,458 are zero that-clauses and the remaining 2,590 instances appear with the complementizer that. The periods 1424–1499 and 1500–1569 amount to 1,695 and 755 examples of object clauses respectively, small figures if compared with the periods 1570–1639 and 1640–1681, with a total of 2,152 and 2,446 instances each. Table 1 below reproduces the distribution of instances in terms of period, type of verb, and whether a syndetic or an asyndetic construction is involved.
Historically speaking, these data allow us to trace the development of that and zero in late Middle and early Modern English. Figure 1 reproduces the occurrence of both constructions in the four periods under scrutiny, where the figures have been normalized to a text of 10,000 words for comparison. In line with Rissanen’s (1991: 272–289) and Suárez-Gómez’s (2000: 182) approaches, the syndetic construction is preponderant in the earliest periods (24.60), although the use of zero is not as low as expected (20.24) and by the mid-sixteenth century it already predominates over that clauses. The definite rise of zero in English is found to take place in the latter part of the sixteenth century, the moment in which it clearly separates from the syndetic construction, to such extent that it amounts to 16.71 occurences in the period 1570–1640, if compared with just 6.91 instances of the syndetic construction. This picture is also mirrored in the period 1640–1681 where the zero link is observed to gain a wider acceptance, almost tripling the number of that-clauses, as they amount to 32.12 and 11.91 occurrences, respectively. Interestingly enough, that-clauses undertake the opposite line of development since they begin to decrease in the late sixteenth century, coinciding with the spread of the zero link, perhaps as a result of a “stylistic shift towards less formal ways of writing” (Rissanen 1991: 286). The asyndetic construction seems to have gained ground considerably, to the point that it concomitantly produced a push shift which restricted the scope of that-clauses, setting aside other linguistic aspects like the nature of the subject, which undoubtedly could have also played a role.
Figure 2, reflecting the same patterning in percentage, positively shows that the diffusion of zero that-clauses in the early Modern period (from 1500 onwards) was moulded in the S-like shape accepted by many linguists as typically followed by changes diffusing in time. Different stages have recently been considered within S curves: (a) incipient, at the beginning i.e. below 15% of progress; (b) new and vigorous, once the lower central part of the curve is reached (15–40%) and diffusion is accelerated; (c) mid-range changes, which have got as far as the central part (41–65%) and start to lose momentum; (d) nearly completed (66–85%); and (e) completed, at the very end of the curve (over 85%) (Labov 1994: 79–83). Within this patterning, the diffusion of zero that-clauses – once it actuates and surpasses the connector that as a variant in competition – seems to have been well-advanced at mid-range in the second period, and nearly completed in the last two periods analysed, almost reaching the upper part of the S curve in 1681. Our data, therefore, corroborates Fanego’s findings as regards the use of zero in Shakespeare’s plays; rather than an idiosyncratic predilection for it, the playwright seems to be accommodating to a general trend already well-advanced in the vernacular.
Table 2 and Figure 3 reproduce the chronological distribution of the syndetic and the asyndetic constructions as regards the type of main verb. The figures have also been normalized to a text of 10,000 words.
In the period 1424–1499 all verbs – except to tell – show slightly higher occurrences with the zero link than with that. Particularly outstanding is the number of object clauses with the main verb to say, reaching 12.6 and 11.89 with zero and that, respectively. This distribution contrasts with the other verb of saying, to tell, which shows a higher occurrence with that-clauses (8.24) than with zero (2.41). In his analysis Rissanen (1991: 288) pointed out that the relative frequency of a particular verb may have a direct bearing on the use of the zero link as they “adopt the variation pattern only when they are well established in the language” (see also: Suárez Gómez 2000: 187–188). Comparing the behaviour of say and tell in this respect may confirm this proposal, insofar as the former, which clearly favours the innovation, is notably more frequent than the latter. In fact, it is the very high figure for that-clauses after the main verb to tell which explains the preponderance of the syndetic construction over the asyndetic one. All in all, the analysis of private correspondence shows that zero that-clauses were already widespread in the latter part of Middle English, thus contradicting previous analyses – such as Warner’s (1982) based on texts written to be orally delivered, like the Wycliff sermons.
The same pattern appears in the second period, 1500–1569, with a more balanced distribution of occurrences: that-clauses still predominate after the main verb to tell (1.39), but the zero link is widespread with to hope (0.35) and specially with the high frequency verb to think, which amounts up to 5.33 instances compared with 3.46 of the syndetic link. The behaviour of to say is also remarkable, because the conspicuous pattern of Middle English is here reversed, and the use of that is slightly ahead of zero (4.85 and 4.39, respectively).
The period 1570–1639, in turn, marks off a widespread use of zero with object clauses, therefore shedding clearer light on the intrinsic preferences for one type of construction depending on the type of verb. In this vein, the verbs to think, to hope and to say are found to favour the use of zero, as they more than double the occurrence of that: the verb to think, for instance, reaches 5.2 and 1.3 occurrences with both types of constructions. Finally, the period 1640–1681 further confirms this trend insofar as these three verbs (to think, to hope and to say) show higher instances of zero. The verb to think shows 9.63 and 1.18 occurrences with the asyndetic and the syndetic construction respectively, and similar rates are found with the verb to hope, with 9.05 and 0.99 instances, respectively. The verbs to know and to tell, on the other hand, do not reach the level of the other items in the adoption of the innovation, showing similar figures for both the syndetic and the asyndetic construction. The latter, for instance, is found to be the most conservative main verb, appearing with 5.05 and 4.6 occurrences of that and zero links respectively. This picture confirms Rissanen’s and Suárez-Gómez’s remark that the variation pattern is adopted differently depending on the type of verb.
4 Sociolinguistic analysis
4.1 Sociolinguistic framework
The rationale behind the historical sociolinguistic enterprise lies in the application of the well-known uniformitarian principle: the idea that languages varied in the same patterned ways in the past as they have been observed to do today (Labov 1972, Labov 1994; Conde-Silvestre 2007; Bergs 2012: 81–82). In the context of this general principle –notwithstanding the difficulties inherent in the conceptualization of social structure, both in the present and, especially, in the past– it is a common assumption that social differences between speakers are among the driving forces for the diffusion of new linguistic elements among populations, and that “on the basis of what we know of the relationship between social order and language change in present-day societies, there is every reason to believe that similar phenomena existed in the past” (Nevalainen and Raumolin-Brunberg 2003: 139). Accordingly, in this section, the distribution of that-clauses and zero that-clauses in late Middle English and early Modern English is analysed in connection with the reconstructed social structure of the periods, in an attempt to confirm (i) the existence of social stratification for this variable, and, if such be the case, (ii) the social group or groups that lead the change in the different chronological stages of its diffusion, thus (iii) tracing the social direction of the change as actually diffusing from below, in sociolinguistic terms, or not.
In historical sociolinguistic research the risks of anachronism should also be carefully observed: the reconstruction of the external variable social structure being specially sentitive to them (Bergs 2012: 85–88). The concept of class itself, for instance, is a post-industrial revolution construct which can hardly be applied to the sociolinguistic analysis of pre-1800 texts. Kiełkiewicz-Janowick (2012: 307) sensibly remarks in this respect:
In reconstructing the past, sociolinguists have to rely in concepts that will adequately describe historical realities and, most importantly, capture the complex relationship between language and society, without falsely assuming that for any historical period the relationships are comparable to those of the present-day. In other words, present-day description and understandings of social variables and relations should not be too readily taken as valid for historical periods. Instead, the meaning of a variable has to be recovered from the historical text that is the subject of linguistic analysis, as well as from the background writings of the historical period under study.
In their seminal Historical Sociolinguistics. Language Change in Tudor and Stuart England (2003), these two authors discuss several proposals to reconstruct social stratification in late Middle and early Modern English. Out of the different models that they discuss, we have chosen a realistic one as the basis for our approach (Nevalainen and Raumolin-Brunberg 2003: 32–37, 136–137; see also Nevalainen 1996). It is a hierarchical model (reproduced in Table 3), which aims at reconstructing ranks. Anachronism is avoided by considering several factors in combination: property rights, social function, social evaluation, lifestyle, legal position, among others. In this way, the possibility that “people developed multiple identities, defining their own positions in terms of different models at the same time” is contemplated (Nevalainen and Raumolin-Brunberg 2003: 37). 4
It draws a fundamental distinction between nobility, gentry and non-gentry, together with a side division for the clergy. In addition to this, further subdivisions are found within some of these levels, particularly based on the evidence of titles and forms of address. For instance, in the group of the gentry and the clergy, upper and lower ranks may be established for knights, baronets (entitled sir or dame) and bishops, who formed the upper gentry or upper clergy, and for squires, gentlemen (entitled master or mistress) and ordinary clergymen, who may be considered part of the lower gentry or lower clergy. Similarly, a rank for the royalty is added at the very top of the chart, whilst at the same time further subdivisions are provided for the non-gentry, both urban (including merchants, craftsmen and artificers) and rural, although no definite evidence is available in this respect. Nevalainen and Raumolin-Brunberg also propose a rank for the professional order, which includes doctors, lawyers, government officials, army officers and teachers, among others. The model is sensible, in sociolinguistic terms, insofar as it reconstructs the four intermediate orders or interior social groups which, following the curvilinear hypothesis, are relevant for the study of linguistic changes in progress (Nevalainen et al. 1996b; Labov 2001: 31–33). In view of this universal sociolinguistic principle, we have concentrated on the evidence afforded by informants from the four intermediate levels: upper gentry, lower gentry, professionals and urban non-gentry, and have disregarded data from the very upper and lower ranks –nobility and rural non-gentry. In this way, the possibility of distortions derived from the urban vs. rural habitat of speakers is also avoided. Correspondents have been allocated to one rank or another both on the basis of the information extracted from the parameter codings given in the PCEEC as well as on external biographical data. 5
The uniformitarian principle also allows historical sociolinguists to consider the role of women from the past in the diffusion of linguistic changes in progress. A universal sociolinguistic principle well known in present-day western societies is that, in changes from above women go ahead of males in the adoption of the prestige forms, and in changes from below they use higher frequencies of innovative forms than men do. This principle was formulated by Labov as the “gender paradox” and contemplates that the interaction between gender and social status tends to be noticeable once the diffusion of changes in progress is well advanced and “the stigmatized or prestige form is recognized and discussed in the speech community” (Labov 2001: 293; see also Labov 1990; Nevalainen and Raumolin-Brunberg 2003; Conde-Silvestre 2007; Kiełkiewicz-Janowick 2012: 321). Confirmation of this principle in historical sociolinguistic research is difficult. Female leadership of changes from above was conditioned by their access to the “learned and literary domains of language use” (Nevalainen 2006: 208–209) and it is well known that access to education was limited to women from the high echelons of society. This is reflected in the structure of the Corpus of Early English Correspondence itself, with a mere 20% of letters by women, overwhelmingly from the upper ranks, so that “gender differentiation below the gentry is restricted” (Nevalainen and Raumolin-Brunberg 2003: 115). The scarcity of female correspondents from the other ranks also complicates the task of detecting their behaviour in connection with changes from below. Despite this difficulty, Suárez Gómez (2000: 198–199), in her analysis of drama and private letters from the Helsinki Corpus, has noticed that women resorted more commonly to zero (54%) than to that (46%), while men preferred to use the latter (69.2%) more often than the zero-link (39.8%). In order to confirm these tendencies – checking the possibility that, in accordance with present-day practices, women behaved differently from men as regards the diffusion of zero that-clauses – a group of female informants from the upper gentry has also been considered in our analysis.
4.2 Sociolinguistic analysis: results
The results for late Middle English (1424–1499) confirm that the zero link is a typical change diffusing from below. As Figure 4 shows, the innovation was led in this period by members of the urban non-gentry (60.3%) and, at this stage, the zero link had already been accepted by the lower gentry (53%). The behaviour of the upper gentry and the professionals also complies with this characterization of zero that-clauses as a change from below, insofar as both groups present the lowest rates, 39% and 39.3% respectively. This is particularly symptomatic in the case of the professional order – especially lawyers – who would have been aware of the written norms enjoying overt prestige in the community and would, therefore, be reluctant to adopt a change coming from below. Finally, from a sociolinguistic point of view, women from the upper gentry (39.7%) do not seem to be ahead of men from the same group (39%), at least at this stage, when the use of that as a connector was still the norm (see also Calle-Martín and Conde-Silvestre 2014: 127).
Figure 5 reproduces the distribution of zero that object clauses across three oustanding social groups participating in its diffusion in this period: the urban non-gentry, the lower gentry and the upper gentry (see appendix). There is no doubt of the advance of the construction with the main verbs to say and to think, which show a regular progress in the acceptance of the innovation down the social scale: 46.9%, 65.7% and 85.7% respectively for the upper gentry, the lower gentry and the urban non-gentry, in the case of to think, and 40.4%, 61.3% and 63.5% for the same groups in the case of the verb to say. Interestingly, the upper gentry shows higher occurrences of the zero link with the main verbs to hope and to know, although the total number of instances is low to establish a definite profile here. Finally, as regards the verb to tell, the innovation is not clearly adopted by members of any of the three groups, although an increasing cline is manifested down the social scale: from upper gentry (20.7%) down to urban non-gentry (39%).
Data from the period 1500–1569 (Figure 6) confirms some of the tendencies already detected in late Middle English. Despite the low input by informants from the urban non-gentry, they still appear as leaders of this change from below (66.7%), followed, as in the preceding period, by the lower gentry (58.3%). In the same vein, the information drawn from male members of the upper gentry shows that they still lagged behind the rest as regards the acceptance of this innovation (46.6%). The linguistic behaviour of members from the professional group is interesting: our data shows that their rates of acceptance of the innovation increased from 39.3% in late Middle English to 54.2% in 1500–1569, becoming the third group from the top. Absence of reliable data makes it difficult to interpret the linguistic behaviour of female informants from the upper gentry. They show the highest rate of occurrences (68.75%), but this evidence is only based on sixteen items and it is doubtful whether it points to a genuine case of female leadership or is just a mere distortion of the scarce available data. Still, if that were the case, a clear contrast with the late Middle English behaviour of the same group arises, insofar as their rates were at the time similar to those of the male members from the gentry. This could be taken as a clue that gender affiliation of changes in progress – both in the present and in the past – only manifests itself when they start to be noticeable in chronological and social terms – at mid-range in this case – although often gender advantage by women is independent of social embedding, as Nevalainen and Raumolin-Brunberg (2003: 131) have shown for a number of changes in Tudor and Stuart England. In the case of zero that-clauses, the results obtained in later periods may be useful to confirm or reject this hypothesis.
Regarding progress of the innovation in connection to type of verb (see Figure 7), to say, to think and to know seem to have behaved as the anchor verbs for this innovation to develop in the period 1500–1569, especially among members of the professional group who, in comparison with the ME data, show the highest increase in the use of the zero link: 50.6%, 65.4% and 54.9% respectively. Avoidance of that with these verbs is also outstanding among the upper gentry, particularly in the case of to say (45.8%) and to think (55.75%). In contrast, the main verbs to hope and, especially, to tell are more resistant to this syntactic innovation (see appendix). The behaviour of each type of verb among the lower gentry and the urban non-gentry does not present a clear patterning, possibly due again to the scarcity of evidence available. The same applies to the verbs to hope and to tell, which, as a result, are not represented in Figure 7.
In the period 1570–1639 the zero link spread from mid-range to the nearly complete stage, reaching the upper central part of the S curve representing diffusion of changes in time. This means that wider awareness among members of the community would have been expected. William Labov (2001: 171) has studied the behaviour of changes from below in present-day American English in connection with socio-economic status and has noticed a tendency for them to remain stable or even decline among their natural leaders – the upper working classes –, while they spread faster among the middle classes, in accordance with widespread levels of acceptance. Zero in the early seventeenth century seems to have behaved like this (Figure 8): it reaches 72.1% among the upper gentry, 66.7% among correspondents from the lower gentry and remains at 63.6% among the informants from the urban non-gentry. Even members of the professional orders increased their production of subordinate clauses without the connector that to 77.5%, thus showing that the original change from below was reaching a stage of acceptance by the speech community at large, which started to threaten the maintenance of the syndetic construction. Data by female informants from the upper gentry – reaching 83.6% in this period – could confirm the gender affiliation of this change, which, therefore, is only noticeable once the change in progress “has passed its incipient phase” (Nevalainen and Raumolin-Brunberg 2003: 130). If this is the case, gender affiliation would be expected to remain constant until the near completion of the change. As such, this is the historically expected behaviour of women that participate in changes from below: they show a conspicuous pattern of diffusion from the lower to the higher ranks, with female speakers leading the process before upper-rank male ones (Nevalainen and Raumolin-Brunberg 2003: 115–116). James Milroy and Lesley Milroy (1993) have proposed that women, rather than adopting forms which are already prestigious, by actually adopting innovations endow the affected forms with prestige and promote their subsequent spread. The extent to which the prominent pattern found in our data confirms this hypothesis for the past is an insoluble question (also Kiełkiewicz-Janowick 2012: 323). 6
In general, data from the fourth period (1640–1681) points to the early phase of decline of the zero link. Rissanen (1991, 1999: 284–285) had already noticed that the phenomenon reached its peak in the seventeenth century and then plunged down in the eighteenth as a result of a prescriptive bias. Our results show a noticeable reduction in the repertoire of most groups when the periods 1570–1639 and 1640–1681 are contrasted (Figure 8): from 72.1% to 63.4% among males of the upper gentry, from 66.7% to 63.5% among the lower gentry and from 77.5% to 61.6% among the professionals. Only female informants from the upper gentry seem to have kept a stable use of zero as complementizer in this period (84.1%). Data from the urban non-gentry is again scarce at this stage, but it points to the maintenance of the asyndetic structure among members of the group where it originated, with figures raising from 63.6% to 77.7%. This patterning is typical of a change from below that becomes thwarted once it has surpassed the mid-range stage in the S curve. A similar recessive behavior has been noticed for other changes in progress in Early Modern English –like third person singular -s (which later would catch on) or the relativizer the which – and confirms the historical validity of the observation that “for a new form to be generalized […] it had to be adopted by the upper strata” (Nevalainen and Raumolin-Brunberg 2003: 150). Reluctance by male informants from the upper gentry could, in this case, be the reason behind the decline of the zero link in the first place. Our findings also show that in private correspondence – i.e. the genre most clearly approaching the vernacular – zero that-clauses started to decline in the late seventeenth century, much earlier than Rissanen had noticed in the multi-genred Helsinki Corpus. Incidentally, this observation could dissociate the decline of the asyndetic link from the expressed prescription by eighteenth-century grammarians and be taken as another clue that their pronouncements were often the sequel to changes already in progress, so that, as Tieken-Boon van Ostade (2006: 270) has remarked, they were often sanctioning natural tendencies in the language.
Figure 9 shows the behaviour of each type of verb for each group of informants at the mid-range to “nearly-completed” stages; in this case, to hope – which in the two previous periods was still resistant to the spread of the innovation – shows the highest rates, especially among the upper gentry (89.3%), the professionals (95.6%) and the urban non-gentry (100%). The main verb to tell, which had systematically lagged behind the rest in late Middle English and the first half of the sixteenth century, also shows a high increase in the third period analysed, with percentages of 75% for the urban non-gentry, 65.3% among the professionals and 48.4% among member of the upper gentry. A snowball effect, like the one recognized by Ogura and Wang (1996) in lexical diffusion, is clearly manifested in the last stages of progress of the asyndetic construction. Data for 1640–1681 in Figure 10 is also useful to detect the main verbs first affected by the recession of zero: a tendency for that to be omitted with the verbs to know and to say – which had led the pattern in previous periods – is noticeable among the lower gentry and the professionals, with percentages for to know declining from 66.7% in 1570–1639 to 22.2% in 1640–1681 among the former, and from 64.5% to 32% and from 66.7% to 63.1% for the same two groups for the latter.
The interpretation of data from the Parsed Corpus of Early English Correspondence allows us to reach the following conclusions.
From a purely chronological perspective, we think that our analysis complements previous ones based on different materials. Data extracted from private correspondence – a genre generally thought to reflect the vernacular in the past – affords a new perspective, which may remain hidden when more formal text-types are dealt with. Particularly, we have detected higher rates of the zero link in late Middle English, when that was still the main complementizer, in comparison with earlier estimates, like Warner’s (1982). Our data, however, confirms the early sixteenth century as the period when the use of the zero link overtook that, in line with Rissanen’s and Suárez-Gómez’s observations. The PCEEC also supports their proposal that the absence of a connector was well advanced in the seventeenth century. In this sense, we have been able to locate the zero link at the mid-range to nearly completed stages of diffusion along the S curve in the period 1570–1639 and throughout the seventeenth century when it reached a percentage of 70% in the corpus – in parallel to a decline of that. Previous studies, and present-day observation, confirm that the zero link declined in the eighteenth century, when its diffusion as a change in progress was thwarted, possibly due to the pronouncements of prescriptive grammarians against it. Our data has also shown that, in chronological terms, the decay of the asyndetic construction in private correspondence – i.e. in the vernacular – predates this calculation: it actually seems to have started to decline among all social ranks in the late seventeenth century. We believe that this could support the idea that prescriptive grammarians sometimes were just sanctioning general tendencies spreading in the language (Tieken-Boon van Ostade 2006: 270).
From a sociolinguistic perspective, we have been able to correlate the zero link with a reconstructed social structure valid for the periods analysed: late Middle English and early Modern English. With this aim, we have followed Nevalainen and Raumolin-Brunberg’s (2003) proposal of a hierarchical, non-anachronistic model – based on contemporary evidence – already tested in historical sociolinguistic research. The distribution of the zero link in social terms confirms that it is a change from below: a characterization which had already been anticipated by noticing its regular diffusion in the vernacular text type. The social patterning in the early stages of diffusion (1424–1499 and 1500–1569) corroborates this hypothesis: it was led by members of the urban non-gentry and then reached the lower gentry. The behaviour of the upper gentry and the professionals – both utterly rejecting the innovation at this stage – can also be a clue that we are dealing with a change from below. Members of the latter group, through their contact with legal texts, would have been especially receptive to changes diffusing from above in written, formal texts, rather than to those coming from the oral medium, in informal contexts. An interesting insight, regarding the social groups leading the change in later stages, is that once the zero link entered the mid-range phase, it increased among the groups that originally had rejected the innovation, the upper gentry and the professionals, but remained stable or even declined among its original leaders, the urban non-gentry. This historical pattern seems to coincide with tendencies observed by Labov (2001) in some varieties of present-day American English.
The linguistic behaviour of female informants also stands out. The information from letters by women from the upper gentry – the only ones that had access to education and could write their own letters at the time – points to a clear shift from a retracted behaviour in late Middle English, when the percentage of the asyndetic construction by women is similar to that of men from the same group, to higher rates once the zero link surpasses the other variant in competition – the complementizer that – in 1500–1569. Women from the upper gentry were also the leaders of the change in the following period, 1570–1639. This may confirm again that we are dealing with a change from below in sociolinguistic terms, insofar as it diffuses from the low to the high ranks with women leading its diffusion before upper-rank men. Additionally, this historical behaviour – in parallel to present-day situations– leads us to ponder the insoluble question as to whether female speakers endowed the innovations they adopted with the necessary prestige for them to be subsequently adopted by others (Milroy and Milroy 1993). As regards the decline of the zero link in the early eighteenth century, our analysis also corroborates the observation that changes in progress from below would not become general if they were not adopted by members from the upper layers of society. Probably due to reluctance on the part of upper gentry males, the asyndetic construction started to decline systematically in all social groups – upper gentry, lower gentry and professionals – except among those that had promoted it in the first place, i.e. the urban non-gentry.
Our analysis has also attempted to correlate the social distribution of the main verbs followed by zero that-clauses: to say, to tell, to know, to think or to hope. Notwithstanding some fluctuations, the clearest social patterning of the innovation in the early stages – 1424–1499 and 1500–1579–is shown when the main verbs to think and to say are considered, while to hope and to tell still show high rates of that. This situation changes radically once the zero link reaches the new and vigorous stage (1570–1639) with this complementizer spreading like a snowball to the verbs to hope and to tell, which show the highest percentages in comparison to the other matrix verbs. Incidentally, to say, which had led the process of diffusion in the early stages, is the first – together with to know – to leave the innovation behind once its diffusion is thwarted in 1640–1681.
To conclude, we believe that this change in progress during late Middle and early Modern English is sensitive to the sociolinguistic approach and that the application of the variationist methodology, even though it cannot be as refined as in the case of modern societies and materials, can illuminate the ups and downs of the competition between that and zero in object clauses along the history of English.
The authors are also grateful to the two anonymous reviewers for their helpful suggestions and comments.
Parsed Corpus of Early English Correspondence, text version. 2006. Compiled by Terttu Nevalainen, Helena Raumolin-Brunberg, Jukka Keränen, Minna Nevala, Arja Nurmi and Minna Palander-Collin, with additional annotation by Ann Taylor. Helsinki: University of Helsinki and York: University of York. Distributed through the Oxford Text Archive.
Anthony, Laurence. 2011. AntCONC (version 3.2.2) [Computer Software]. Tokyo, Japan: Waseda University. http://www.antlab.sci.waseda.ac.jp/.
Bergs, Alexander T. 2012. The uniformitarian principle and the risk of anachronisms in language and social history. In Juan Manuel Hernández-Campoy & Juan Camilo Conde-Silvestre (eds.), The handbook of historical sociolinguistics, 80–92. Oxford & Malden, MA: Wiley-Blackwell. Google Scholar
Biber, Douglas. 1995. Dimensions of register reconstruction. A cross-linguistic comparison. Cambridge: Cambridge University Press. Google Scholar
Biber, Douglas. 2001. Dimensions of variation among eighteenth century registers. In Hans-Jürgen Diller & Manfred Görlach (eds.), Towards a history of English as a history of genres, 89–110. Heidelberg: Carl Winter. Google Scholar
Biber, Douglas & Edward Finegan. 1997. Diachronic relations among speech-based and written registers in English. In Terttu Nevalainen & Leena Kahlas-Tarkka (eds.), To explain the present: Studies in the changing English language in honour of Matti Rissanen, 253–275. Helsinki: Société Néophilologique. Google Scholar
Bolinger, Dwight. 1972. That’s that. The Hague: Mouton. Google Scholar
Calle-Martín, Javier & Juan Camilo Conde-Silvestre. 2014. A sociolinguistic analysis of zero that-clauses in late Middle English. In Michael Bylinsky (ed.), Studies in Middle English. Words, forms, senses and texts, 119–133. Frankfurt am Main: Peter Lang. Google Scholar
Conde-Silvestre, Juan Camilo. 2007. Sociolingüística histórica. Madrid: Gredos. Google Scholar
Conde-Silvestre, Juan Camilo & Juan Manuel Hernández-Campoy. 2004. A sociolinguistic approach to the diffusion of chancery written practices in late fifteenth-century private correspondence. Neuphilologische Mitteilungen 105. 135–152. Google Scholar
Conde-Silvestre, Juan Camilo & Juan Manuel Hernández-Campoy. 2013. Tracing the generational progress of language change in fifteenth-century English: The digraph <th> in the Paston Letters. Neuphilologische Mitteilungen 114. 279–300. Google Scholar
Elspaß, Stephan. 2012. The use of private letters and diaries in sociolinguistic investigation. In Juan Manuel Hernández-Campoy & Juan Camilo Conde-Silvestre (eds.), The handbook of historical sociolinguistics, 156–169. Oxford & Malden, MA: Wiley-Blackwell. Google Scholar
Finegan, Edward & Douglas Biber. 1995. That and zero complementizers in late Modern English. Exploring archer from 1650–1990. In Bas Aarts & Charles F. Meyer (eds.), The verb in contemporary English, 241–257. Cambridge: Cambridge University Press. Google Scholar
Fischer, Olga. 1992. Syntax. In Norman Blake (ed.), The Cambridge history of the English language. Volume 2: 1066–1476, 207–408. Cambridge: Cambridge University Press. Google Scholar
Görlach, Manfred. 2001. Eighteenth-century English. Heidelberg: Carl Winter Universitätsverlag. Google Scholar
Hosaka, Michio. 2010. The rise of the complementizer that in the history of English. In Merja Kytö, John Scahill & Harumi Tanabe (eds.), Language change and variation from Old English to Late Modern English. A festschrift for Minoji Akimoto, 69–78. Frankfurt am Main: Peter Lang. Google Scholar
Huddleston, Rodney & Geoffrey K Pullum. 2002. The Cambridge grammar of the English language. Cambridge: Cambridge University Press. Google Scholar
Jespersen, Otto. 1933. Essentials of English grammar. London: George Allen and Unwin. Google Scholar
Kiełkiewicz-Janowiak, Agnieszka. 2012. Class, age and gender-based patterns. In Juan Manuel Hernández-Campoy & Juan Camilo Conde-Silvestre (eds.), The handbook of historical sociolinguistics, 307–331. Oxford & Malden, MA: Wiley-Blackwell. Google Scholar
Labov, William. 1972. Sociolinguistic patterns. Philadelphia: University of Pennsylvania Press. Google Scholar
Labov, William. 1994. Principles of linguistic change. Volume 1: Internal factors. Oxford: Blackwell. Google Scholar
Labov, William. 2001. Principles of linguistic change. Volume 2: Social factors. Oxford: Blackwell. Google Scholar
López Couso, M. José. 1996. A look at that/zero variation in Restoration English. In Derek Britton (ed.), English historical linguistics 1994. Papers from the 8th international conference on English historical linguistics, 271–286. Amsterdam and Philadelphia: John Benjamins. Google Scholar
Milroy, James & Lesley Milroy. 1993. Mechanisms of change in urban dialects. The role of class, social network and gender. International Journal of Applied Linguistics 3. 57–77. CrossrefGoogle Scholar
Mitchell, Bruce. 1985. Old English syntax. 2 volumes. Oxford: Clarendon Press. Google Scholar
Mustanoja, Tauno. 1960. A Middle English syntax. Part 1: Parts of speech. Helsinki: Société Néophilologique. Google Scholar
Nevala, Minna. 2004. Address in early English correspondence. Its forms and socio-pragmatic functions. Helsinki: Société Néophilologique. Google Scholar
Nevalainen, Terttu. 1996. Social stratification. In Terttu Nevalainen & Helena Raumolin-Brunberg (eds.), Sociolinguistics and language history. Studies based on the corpus of early English correspondence, 57–76. Amsterdam: Rodopi. Google Scholar
Nevalainen, Terttu. 2006. Mapping change in Tudor English. In Lynda Mugglestone (ed.), The Oxford history of English, 178–211. Oxford: Oxford University Press. Google Scholar
Nevalainen, Terttu. 2007. Introduction. In Terttu Nevalainen & Sanna-Kaisa Tanskanen (eds.), Letter writing, 1–11. Amsterdam & Philadelphia: John Benjamins. Google Scholar
Nevalainen, Terttu & Helena Raumolin-Brunberg. 1996a. The corpus of early English correspondence. In Terttu Nevalainen & Helena Raumolin-Brunberg (eds.), Sociolinguistics and language history. Studies based on the corpus of early English correspondence, 39–54. Amsterdam: Rodopi. Google Scholar
Nevalainen, Terttu & Helena Raumolin-Brunberg. 1996b. Social stratification in Tudor English? In Derek Britton (ed.), English historical linguistics 1994. Papers from the 8th international conference on English historical linguistics. Edinburgh 1994, 303–326. Amsterdam & Philadelphia: John Benjamins. Google Scholar
Nevalainen, Terttu & Helena Raumolin-Brunberg. 2003. Historical sociolinguistics. Language change in Tudor and Stuart England. London: Longman/Pearson Education. Google Scholar
Ogura, M. & W. S. Y Wang. 1996. Snowball effect in lexical diffusion. The development of –s in the third person singular present indicative in English. In Derek Britton (ed.), English historical linguistics 1994. Papers from the 8th international conference on English historical linguistics. Edinburgh 1994, 119–141. Amsterdam & Philadelphia: John Benjamins. Google Scholar
Palander-Collin, Minna. 1997. A medieval case of grammaticalization, methinks. In Matti Rissanen, Merja Kytö & Kirsi Heikkonen (eds.), Grammaticalization at work. Studies in long-term developments in English, 371–403. Berlin & New York: Mouton de Gruyter. Google Scholar
Palander-Collin, Minna. 1999. Grammaticalization and social embedding. I think and methinks in Middle and Early Modern English. Helsinki: Société Néophilologique. Google Scholar
Palander-Collin, Minna. 2010. Correspondence. In Andreas H. Jucker & Irma Taavitsainen (eds.), The handbook of historical pragmatics, 677–703. Berlin & New York: Mouton de Gruyter. Google Scholar
Quirk, Randolph, Sidney Greenbaum, Geoffrey Leech & Jan Svartvik. 1985. A comprehensive grammar of the English language. London & New York: Longman. Google Scholar
Rissanen, Matti. 1991. On the history of that/zero as object clause links in English. In Karin Aijmer & Bengt Altenberg (eds.), English corpus linguistics: Studies in honour of Jan Svartvik, 272–289. London & New York: Longman. Google Scholar
Rissanen, Matti. 1999. Syntax. In Roger Lass (ed.), The Cambridge history of the English language. Volume 3: 1476–1776, 187–331. Cambridge: Cambridge University Press. Google Scholar
Rohdenburg, Günter. 2006. The role of functional constraints in the evolution of the English complementation system. In Christiane Dalton-Puffer, Dieter Kastovsky, Nikolaus Ritt & Herbert Schendl (eds.), Syntax, style and grammatical norms. English from 1500–2000, 143–166. Frankfurt am Main: Peter Lang. Google Scholar
Sundby, Bertil, Anne Bjørge & Kari E Haugland. 1991. A dictionary of English normative grammar, 1700–1800. Amsterdam & Philadelphia: John Benjamins. Google Scholar
Suárez Gómez, Cristina. 2000. That/zero variation in private letters and drama (1420–1710). A corpus-based approach. Miscelanea. A Journal of English and American Studies 21. 179–204. Google Scholar
Taylor, Ann & Beatrice Santorini. 2006. The parsed corpus of early English correspondence. http://www-users.york.ac.uk/~lang22/PCEEC-manual/.
Tieken-Boon van Ostade, Ingrid. 2006. English at the onset of the normative tradition. In Lynda Mugglestone (ed.), The Oxford history of English, 240–273. Oxford: Oxford University Press. Google Scholar
Traugott, Elizabeth Closs. 1992. Syntax. In Richard M. Hogg (ed.), The Cambridge history of the English language. Volume 1: The beginnings to 1066, 168–289. Cambridge: Cambridge University Press. Google Scholar
Warner, Anthony. 1982. Complementation in Middle English and the methodology of historical syntax. London & Canberra: Croom Helm. Google Scholar
Huddleston and Pullum also define object clauses introduced by the subordinator that as “expanded declaratives”, or simply “that-declaratives”; contrarily, those without an introductory subordinator are called “bare declaratives” (Huddleston and Pullum 2002: 951).
The text version of PCEEC has been favoured over the parsed one. Even though acknowledging that the latter could have substantially simplified the retrieval of instances, the former was processed with AntConc in order to obtain all the instances of the matrix verbs automatically, and later manual disambiguation was needed in order to classify the instances in terms of a) the type of complementation, whether syndetic or asyndetic; and b) the social background of the speaker. The parsed version has been discarded considering that manual disambiguation is an added asset in this kind of analysis for a more reliable taxonomy of the instances.
The authors discuss other models, including a ‘conflict’ one also based on contemporary accounts. It distinguishes the poorest/common sort from the better/richer sort, introducing as a third term middle/middling sort. Other historical sociolinguists have applied this conflict model in research on a number of socio-pragmatic variables in Tudor and Stuart England (Palander-Collin 1999; Nevala 2004). It may, however, be biased if we assume, with Nevalainen and Raumolin-Brunberg, that it represents “and elite view, possibly written for justifying the existing social inequality” (2003: 37).
This model has been fruitfully applied to trace the sociolinguistic diffusion of changes from above in the history of English, such as spelling standardization (Conde-Silvestre and Hernández-Campoy 2004; 2013). Nevalainen and Raumolin-Brunberg (2003) have also applied it – in a somewhat reduced format – to a variety of changes diffusing from below in Tudor England, which positively stratify in social terms: third person singular present indicative –s, you as subject form of the second person personal pronoun, the replacement of mine/thine with my/thy as possessive determiners, the decline of multiple negation, and the spread of periphrastic do, among others.
Nevalainen and Raumolin-Brunberg (2003: 110–132) have confirmed the leadership by female informants at the new and vigorous stage in a number of changes in progress in Tudor and Stuart England: you as subject form of the second person personal pronoun, my/thy replacing mine/thine as possessive determiners before both vowel and consonant sequences, and third person singular present indicative –s.
About the article
Published Online: 2015-05-06
Published in Print: 2015-05-01
Funding: The present research has been funded by the Spanish Ministry of Science and Innovation (grant number FFI2011-26492) and by the Autonomous Government of Andalusia (grant number P11-HUM7597). These grants are hereby gratefully acknowledged.