2 Introducing ProTag constructions

Prior to Mycock (Reference Mycock2019), the ProTag constructionFootnote ³ in English had been the subject of sociolinguistic research (for example, Snell Reference Snell2008; Moore & Snell Reference Moore, Snell, Gregersen, Parrott and Quist2011) and received brief mention in some grammars of varieties of British English (for example, Biber et al. Reference Biber, Johansson, Leech, Conrad and Finegan1999; Shorrocks Reference Shorrocks1999), but systematic investigation of its form and functions was lacking. The history of this construction has received even less attention, with the result being that what is known about the existence of ProTags and their features in earlier varieties is limited and partial. In this article, we set out for the first time the defining characteristics of the ProTag construction in EModE and then compare it to its counterpart in PDBE in order to determine the extent to which they differ, and hence how the construction has changed over this period. We begin with a review of Mycock's (Reference Mycock2019) findings in relation to PDBE in order to introduce what is likely to be an unfamiliar construction and to highlight those features of it which may be subject to variation and therefore must be considered in any research on it, including this study of ProTags in EModE.

3 Data and methodology

Mycock's (Reference Mycock2019) findings indicate that the ProTag construction is relatively rare in PDBE. The scant reference to this construction in the literature on EModE leads one to suspect that the same is likely to be true of this earlier variety of English. There is also the issue that tags are overwhelmingly a feature of spoken English, making the study of tags in historical varieties even more of a challenge. The closest available approximation to speech is text from contemporary drama, with comedy in particular representing an important source (Culpeper & Kytö Reference Culpeper and Kytö2000). It was this type of data that was the basis for Tottie & Hoffmann's (Reference Tottie and Hoffmann2009) research on tag questions in English in the sixteenth century.Footnote ⁷

We decided to focus our searches for ProTags on dramas by playwrights in which tag question use was relatively common on the basis that these were most likely to yield examples of other tag constructions. In this way, we sought to optimise the possibility of identifying a set of examples that would give a fuller picture of the form and functions of ProTag constructions as they occurred in EModE despite the likelihood that this construction would occur relatively infrequently. Works by Jonson, Marlowe and Shakespeare provided 55 per cent (n = 75, out of a total of 136) of all of the examples of tag questions identified in Tottie & Hoffmann's (Reference Tottie and Hoffmann2009) study,Footnote ⁸ so preliminary work was undertaken to determine whether ProTag constructions were also a feature of these playwrights’ works. An initial search revealed examples of ProTags in the works of all three. For this first systematic investigation of ProTags in EModE, we have therefore chosen to focus on the dramas of Jonson, Marlowe and Shakespeare. Our corpus is smaller than the one investigated by Tottie & Hoffmann (Reference Tottie and Hoffmann2009) because of this narrower scope, concentrating as it does on fewer playwrights and a smaller total number of plays (69 versus 197). This reflects the more labour-intensive nature of our task, which necessarily involved searching for and analysing individual words, viz. all instances of nominative personal pronouns and demonstrative pronouns, in the dramatic works of these three playwrights, rather than strings of words that together form multiword tags. Our results provide an interesting point of comparison with previous work on the history of question tags, and augment and extend our knowledge of tag constructions more generally in the history of English.

Focusing on these three playwrights gives a corpus of 69 plays that represents the period from around 1587 (the earliest possible date of performance for Marlowe's Dido, Queen of Carthage) to 1637 (Jonson's death). Whilst the plays attributed to Marlowe and Jonson are generally accepted without contention, recent scholarship on collaborative authorship has nuanced the Shakespearean canon. To reflect these most recent studies, and to take a maximalist approach to inclusion in our corpus, we include the complete plays attributed to Shakespeare in the New Oxford Shakespeare (Taylor et al. Reference Taylor, Jowett, Bourus and Egan2016). A full list of the works included in our corpus is provided in table A1 in the Appendix.

Any linguistic corpus compiled from early modern literary texts is fraught with the challenges negotiated by the practice of textual criticism. Rather than a single, authoritative source for these texts, we are faced with an array of versions, editions and states, many of which have legitimate but mutually exclusive claims to authority.Footnote ⁹ To give a notoriously problematic example, the single work Hamlet has three main texts: a quarto edition published in 1603 (known as Q1), a second quarto published in 1604 (Q2) and the text contained in the ‘First Folio’ of 1623 (F1). Thought to be based on three different sources (either manuscripts or a performance of the play), each text contains substantial differences, adding, omitting or altering material now thought to be key parts of the play. The Folio text, for instance, contains some 1,400 lines not found in Q1.

Textual criticism has offered various theories and methods as to how we might derive a single Hamlet, or a single edition of any text, from its many available iterations. Traditionally, this has prioritised the reconstruction of a text that most accurately reflects the author's intent. To reach this point, a range of evidence can be applied, from internal matters of literary style and perceived quality, to historical, documentary evidence. The final product is often a conflation of several sources, giving the modern reader an ‘ideal’ text: a new text for which there may be no exact historical precedent. Whilst these modern editions are useful tools for readers, dramatists and literary scholars, as later creations they are not appropriate for historical linguistic studies.

This study therefore follows the rationale behind the school of textual criticism known as ‘unediting’ (see Marcus Reference Marcus1996; McLeod Reference McLeod1981). Unediting acknowledges that centuries of editorial convention and tradition impose on a text certain ideological or linguistic features not found in their time. In returning to the text's earliest incarnations, practitioners of unediting read texts in their native historical contexts. This is especially pertinent due to the nature of our object of study. As is the case for PDBE, descriptive studies of EModE have neglected the ProTag construction, or omitted it entirely. It is absent, for instance, from Abbott's otherwise thorough A Shakespearean Grammar (Reference Abbott1884). In this work, we seek to raise awareness of the ProTag as a linguistic feature not only in the works of these three playwrights but also in EModE more generally.

This study's corpus therefore comprises the earliest available texts of the plays, and does not rely on modern editions except for standardising line numbers for reference purposes. In doing so, we do not intend to present these first editions as purely authorial: they are likewise subject to the influence of those involved in producing the editions such as typesetters or transcribers. However, the fact of these non-authorial influences is not problematic for linguistic study: by selecting the earliest available texts, we guarantee that the corpus is reflective of contemporaneous usage, and not influenced by a more recent editor. Whether these texts faithfully reflect the authorial intent of Jonson, Marlowe and Shakespeare or not, they are certainly examples of EModE. For this reason, we also included co-authored works such as Eastward Ho! in our corpus (written by Jonson, George Chapman and John Marston).

The theoretical rationale behind using the earliest available texts is supported by a practical exigency: the database used for the source texts, EEBO-TCP, prioritises first editions, often ignoring later editions of the text.Footnote ¹⁰ The Text Creation Partnership (TCP) is an on-going project that creates fully searchable electronic texts transcribed from early printed books; EEBO-TCP is the part of this project that uses books digitised on the Early English Books Online database (EEBO) as its source – so far, roughly 50 per cent of EEBO books have been encoded. Transcribed by humans, EEBO-TCP is more accurate than texts created with Optical Character Recognition software. As literal transcriptions of their source texts, EEBO-TCP makes no editorial intervention, representing, for instance, typographical errors as they appear in the digitised books.

To search for ProTags, the selected editions were identified in EEBO, and the link to their EEBO-TCP text followed, giving the entire text in a single browser window. A browser extension called Multi-highlight was used to highlight all demonstrative and nominative-form personal pronouns in the text.Footnote ¹¹ This extension allows the user to input multiple search strings, and highlights them all at once in the browser. One pronoun was searched at a time, but with multiple variants of spelling. This accounted for spelling and orthography variants common to EModE, for example yow for ‘you’. Given EEBO-TCPs literal transcription of the original text, we also checked for common typographical variations such as a double v for w, and errors such as yon for ‘you’.Footnote ¹² A full list of the variant forms searched for is given in table A2 in the Appendix.

Once searched, Multi-highlight highlights each form in a different colour, allowing the user to scroll through the entire text and easily locate and check each pronoun. If a pronoun was identified as a potential ProTag, it was collected into a spreadsheet. These were later read more closely, referring to the context as necessary, and confirmed or rejected as ProTags. Each ProTag was located in a standard edition of the playwrights’ works to allocate it line and scene numbers for reference purposes. These data are available via the Oxford University Research Archive (www.ora.ox.ac.uk). The editions used are The New Oxford Shakespeare (Taylor et al. Reference Taylor, Jowett, Bourus and Egan2016), The Cambridge Edition of the Works of Ben Jonson Online (Bevington et al. Reference Bevington, Butler and Donaldson2014) and Christopher Marlowe: The Complete Plays (Burnett Reference Burnett1999).

A further type of search result involved an ambiguity that required evaluation on a case-by-case basis. In EModE texts, I is an acceptable spelling variant for aye, which in modern spelling editions is routinely changed to ay or aye, as suggested by Foakes in his edition of King Lear (Reference Foakes1997). In the prefatory explanation of his editorial methodology, Foakes explains that ‘Some unusual spellings are noted, especially where they may suggest the way a word was pronounced, but common variants such as “I” for “Ay” … are not recorded unless some ambiguity seems to be involved.’ Despite this editorial tendency towards aye, many examples are entirely consistent with I being a 1sg nominative pronoun used as a ProTag. For instance:

1. (10)

   1. (a) I will not trust you, I, Nor longer stay in your curst company. (Shakespeare, MND 3.2.342)

   2. (b) I doe defie him, I, (Jonson, Devil 5.8.139)

Whilst many modern editions follow Foakes’ policy and change such instances of I to ay or aye, some retain such examples as pronouns. Frequently, an editorial note explicitly stating the ambiguity follows either decision. To give one example:

1. (11) And I commend them I, that can vse art, With such iudiciall practise. (Jonson, Case 4.6.18)

Example (4) is followed in the Cambridge University Press online edition of Jonson's works with the note ‘“I”… can be understood as “ay” or “I”’ (Commentary note 19). Such editorial uncertainty, possibly related to the editors’ lack of familiarity with ProTag constructions, shows that the ambiguity remains unresolved, and they ultimately defer the responsibility of interpretation from the editor to the reader. We have therefore considered each of the ambiguous I/aye examples in their dramatic contexts, and retained those that exhibit the features of ProTags consistent with the fundamental insights in Mycock (Reference Mycock2019) but to be defined specifically for EModE in section 4. We hope that the new context offered by this research may help inform future editors’ decisions regarding this particular ambiguity.

5 ProTag constructions and tag questions in EModE: a comparison

We can compare our findings on the ProTag construction in the works of Jonson, Marlowe and Shakespeare with those of Tottie & Hoffmann (Reference Tottie and Hoffmann2009) on tag questions in order to obtain a more complete picture of tag constructions in EModE drama. The focus of Tottie & Hoffmann (Reference Tottie and Hoffmann2009) is tag questions in the sixteenth century, and as such they only examine a subset of works by Jonson and Shakespeare included in our corpus. In the discussion of tag question and ProTag construction use in the work of these authors which follows, we consider only those works analysed in Tottie & Hoffmann (Reference Tottie and Hoffmann2009) in order to make a direct comparison.

First, we observe that ProTag constructions are less frequent than tag questions in the plays examined in both our study and Tottie & Hoffmann's; see table 4.Footnote ¹⁸ (A full play-by-play breakdown is provided in table A3 in the Appendix.)

Table 4. ProTag constructions and tag questions (QTs) in the works of Jonson, Marlowe and Shakespeare considered in Tottie & Hoffmann (Reference Tottie and Hoffmann2009)

Jonson is the most prolific user of both tag constructions out of the three playwrights, with a fpmw count of 218; this is also the case for question tags (Tottie & Hoffmann Reference Tottie and Hoffmann2009: 152). Tottie & Hoffmann observe that this may be due to the greater proportion of prose in Jonson's works. Certainly, the majority of the ProTag constructions in Jonson's dramas appear in prose (74%, n = 72). In Shakespeare's works, the majority appear in blank verse (80%, n = 16). Overall, 61 per cent of all ProTag examples in our corpus appear in prose, with most of the remainder appearing in blank verse (38%; 1% appear in rhyming verse). Another relevant factor is likely to be that the bulk of Jonson's plays are comedies. Hoffmann (Reference Hoffmann2006) shows that tag questions are more frequent in comedies compared to tragedies since at least the mid sixteenth century. There is some evidence that this may also be the case for ProTags in EModE: 84 per cent of examples identified in our corpus (n = 102) appear in comedies.

Tottie & Hoffmann (Reference Tottie and Hoffmann2009) supply a detailed comparison of tag questions in the works of Jonson and Shakespeare, the two most prolific users of this construction in their study. For the purposes of providing a similarly detailed comparison of two different types of tag construction in EModE, we do the same here for ProTag constructions. Tottie & Hoffmann (Reference Tottie and Hoffmann2009) discovered that Jonson's use of tag questions was distinct with respect to the distribution of polarity types compared to Shakespeare. In prose, Jonson uses a high proportion of negative anchors with positive question tags (for example, you are not Pageant Poet to the City of Millaine sir, are you. Jonson, Case 1.1.45; cited by Tottie & Hoffmann Reference Tottie and Hoffmann2009: 135) compared to Shakespeare, who uses only positive anchors with negative question tags in prose (for example, You serue Octauius Caesar, do you not? Shakespeare, JC 3.1.280; cited by Tottie & Hoffmann Reference Tottie and Hoffmann2009: 135). If we compare the polarity type of the anchor in tag questions and ProTag constructions, a difference in distribution emerges.

The data in figure 1 show that in Johnson's prose, QTs (tag questions) occur more frequently with a negative anchor (64%) than ProTags do (27%); ProTags therefore occur more frequently with a positive anchor (73%) than QTs (36%) do. By contrast, in Shakespeare's prose QTs only appear with a positive anchor, while ProTags occur more frequently with a negative anchor (56%). Thus, one tag type is paired more frequently with an anchor of a particular polarity type, but that pairing varies by playwright.

Figure 1. Anchor polarity types in prose for tag question (QT) and ProTag (PT) constructions in the sixteenth-century works of Shakespeare and Jonson considered in Tottie & Hoffmann (Reference Tottie and Hoffmann2009)

In blank verse (figure 2), neither Jonson nor Shakespeare has negative anchors in tag questions. Thus, Shakespeare never has a negative anchor in a tag question, whereas Jonson has a majority of negative anchors in tag questions, but only in prose. Both writers do, though, pair ProTags with negative anchors in blank verse, more so than with positive anchors. In the case of ProTag constructions in blank verse then, the same trend is observed in the works of both playwrights: negative anchors occur only with ProTags and not with QTs.

Figure 2. Anchor polarity types in blank verse for tag question (QT) and ProTag (PT) constructions in the sixteenth-century works of Shakespeare and Jonson considered in Tottie & Hoffmann (Reference Tottie and Hoffmann2009)

We make no claims here about why Jonson and Shakespeare's use of tag constructions should vary in the ways that they do with respect to polarity and the prose/blank verse distinction, but report this observation in order to highlight possible dimensions of variation. Our research indicates that interaction with polarity ought to be considered in relation to both ProTag constructions and tag questions in future research in order to understand more fully the properties and distribution of tag constructions as a whole.

We have seen that ProTag constructions are part of the grammar of EModE, just as tag questions are. Next, we consider how ProTag constructions have changed over time by comparing our EModE data with the results of Mycock's (Reference Mycock2019) study of ProTag constructions in PDBE.

6 ProTag constructions from EModE to PDBE

When we compare the formal properties and functions of ProTag constructions in EModE with their counterparts in PDBE, two differences are particularly striking, setting aside the difference in default pronoun case mentioned in section 1. The first is the relative proportions of personal and demonstrative ProTags in the two varieties. The second is the range of functions that ProTag constructions have in each variety. In EModE, personal ProTags occur more frequently than demonstrative ProTags (77% to 23%, respectively); see table 5. Compared to our results for EModE, the pattern in PDBE according to Mycock (Reference Mycock2019) is practically the reverse. Mycock (Reference Mycock2019) reports a far higher proportion of demonstrative ProTags compared to personal ProTags in PDBE: 80.6% to 19.4%. This difference could possibly be due to most of her corpus data being written rather than spoken, and to the fact that the distribution of the ProTag construction appears to be restricted. However, Mycock (Reference Mycock2019) interrogated one corpus comprising only spoken data from a region in which both personal and demonstrative ProTags are clearly available: the Diachronic Electronic Corpus of Tyneside English (DECTE; Corrigan et al. Reference Corrigan, Buchstaller, Mearns and Moisl2012–). If we isolate the results from the DECTE corpus, as in table 5, we find the same pattern of relative frequency exists in slightly different proportions: 70.3 per cent demonstrative ProTags to 29.7 per cent personal ProTags.

Table 5. ProTags in our EModE corpus, in PDBE (Mycock Reference Mycock2019) and in DECTE (Corrigan et al. Reference Corrigan, Buchstaller, Mearns and Moisl2012–, as reported in Mycock Reference Mycock2019)

When we consider frequency within the group of demonstrative ProTags, the occurrence of that also differs between EModE and PDBE. In EModE this appears more frequently than that (this = 15.6%, that = 4.9%), whereas in the PDBE corpora and in DECTE, that is by far the most common demonstrative ProTag (56.2% and 64.6%, respectively).

Compared with our EModE data, therefore, Mycock's (Reference Mycock2019) PDBE data, whether entirely spoken (DECTE) or majority written, exhibit a clear difference in relation to the occurrence of demonstrative versus personal pronouns as tags and with respect to the relative frequencies of the two singular demonstrative ProTags. Why should this be? It is not the case that singular demonstratives are in competition with a 3sgn personal ProTag as no such form is attested in EModE or PDBE: a singular demonstrative can clearly have a 3sgn anchor point; see (5).

We hypothesise that singular demonstrative ProTags have expanded in terms of their possible referents between EModE and PDBE. Specifically, we propose that singular demonstrative ProTags were restricted to referring to entities in EModE, but in PDBE they can refer to entities and other types of semantic objects such as propositions as well. That is, they have developed a textual or discourse deictic use. The distinction can be rather difficult to tease apart, but in PDBE examples such as the one in (31) the referent of the ProTag appears to be something other than an entity.

1. (31) Brian Jackson was very pleased with himself. From his promising interview with the redoubtable Mrs Fitton, he had driven directly to Mansfield Memorial Hospital with medical records in mind. Initiative, that. (BNC-C8D/3779)

Along with what appears to be its core speaker-oriented (subjective) function (‘Stating a fact/opinion’) and other hearer-oriented (intersubjective) functions (‘Challenging’, ‘Conversational joking’; ‘Desired action’) in EModE and PDBE, it appears that in PDBE the ProTag Construction additionally has a discourse-oriented use, exemplified in (31), which creates textual coherence. Narrog (2016) identifies increased discourse-orientation as a strong tendency in meaning change and the final stage of grammaticalisation following speaker-orientation (subjectification) and hearer-orientation (intersubjectification). Extension of the possible referents of singular demonstrative ProTags between EModE and PDBE to include parts of the discourse supports the developmental pathway proposed by Narrog (2016).

The possibility of singular demonstrative ProTags referring to elements other than specific entities may also provide a reason for their much greater frequency compared to plural demonstrative ProTags in PDBE (78.4% compared to 2.2% of all occurrences in Mycock's study). The predominance of that over this in PDBE is likely related to the fact that ProTags are turn medially or turn finally located discourse-pragmatic variables. If, as we have proposed, a ProTag can refer to a previous portion of the discourse, the use of the distal demonstrative singular pronoun to refer to what has gone before is expected without ruling out completely the occurrence of a proximal ProTag with a similar kind of referent.

Another area in which PDBE differs from EModE relates to the possible functions of ProTag constructions. In the EModE data, we identified ProTag constructions with the functions ‘Desired action: Demand’ and ‘Question’. Neither of these functions was identified in the Mycock (Reference Mycock2019) study of PDBE data. They are, of course, possible functions of tag questions in PDBE; recall that the classificatory system we have adopted was proposed by Barron et al. (Reference Barron, Pandarova and Muderack2015) to account for tag-question functions in British English and Irish English. Thus our research indicates that, in terms of their possible functions, ProTag constructions could be argued in some ways to have had more in common with tag questions in EModE than in PDBE as both could be used with the ‘Question’ function in the earlier variety.

Two functions of ProTag constructions are attested in PDBE that we did not find in our EModE data: ‘Statement: Acknowledging response’ and ‘Desired action: Give’. We note that both are rarely attested functions in PDBE: in a search of over 980 million words, ‘Acknowledging response’ accounted for 0.9 per cent of all examples (n = 7) and ‘Desired action: Give’ accounted for 0.1 per cent of all examples (n = 1) (Mycock Reference Mycock2019). Further research is required to determine if ProTag constructions could ever be used with either of these functions in EModE.

Given the differences in the types and quantity of data available, one should be cautious of drawing conclusions about the functions of ProTag constructions in EModE and PDBE based on their relative frequency. However, what is clear from the data presented in table 6 is that by far the most common primary function in both varieties is ‘Statement: Stating a fact/opinion’.

Table 6. ProTag constructions’ primary functions in our EModE corpus, in PDBE (Mycock Reference Mycock2019) and in DECTE (Corrigan et al. Reference Corrigan, Buchstaller, Mearns and Moisl2012–, as reported in Mycock Reference Mycock2019)

If we assume that ‘Statement: Stating a fact/opinion’ is equivalent in Tottie & Hoffmann's (Reference Tottie and Hoffmann2009) functional classification system to the category ‘Attitudinal’ (defined as expressing the speaker's attitude and therefore a subjective function); and that the ‘Challenging’ category is the same in both systems, that Hortatory (defined as having a directive function by Tottie & Hoffmann) is equivalent to ‘Desired action’, and that ‘Confirmatory’ (defined by Tottie & Hoffmann as clearly seeking and receiving an answer without any strong affective function) is equivalent to ‘Question’ (making these intersubjective functions), we can make a comparison of the functions of tag questions and ProTag constructions in EModE and PDBE; see table 7.

Table 7. Tag question and ProTag construction functions in EModE and PDBE

^a Because of low numbers, hortatory QTs were included in the category ‘Other’ which represented 4 per cent of the total.

The data in table 7 indicate that tag question and ProTag constructions were predominantly used with different functions in EModE, and that the same is true in PDBE. The strength of the association of ProTag constructions with one particular primary function, ‘Stating a fact/opinion’, may explain why the ‘Question’ function was lost. If ProTag constructions occurred only rarely with the ‘Question’ function, this function may have fallen out of use.

In summary, we have identified two main differences in the ProTag construction's form and functions between EModE and PDBE: (i) in EModE personal ProTags are much more frequent than demonstrative ProTags, but in PDBE this situation is reversed; and (ii) ProTag constructions have a different range of possible functions in PDBE than in EModE, notably in EModE ‘Question’ is a possible function of both tag questions and ProTag constructions.