Quaker Prophetic Language in the Seventeenth Century: A Cross-Disciplinary Case Study

Abstract

This paper explores three themes: (i) a short, empirical research account of the linguistic realization of seventeenth-century Quaker prophecy using digital corpus-based tools; (ii) a practical description of how those tools can be used in interdisciplinary research such as the prophecy study; and (iii) a reflective section that considers the advantages, potential richness but also challenges of embarking on an integrated piece of research that straddles established academic disciplines. The ‘prophecy’ analysis comments on the nature of prophecy from a linguistic perspective. It includes positive and negative connotations observed in the data contrasted with non-Quaker texts (including the Bible), and also how Quaker prophetic style changed during the second half of the seventeenth century. The secondary purpose of the paper is to demonstrate the value of departing from traditional, well-established approaches in a discipline such as religion. Quaker studies scholars are familiar with the exercise of grappling with unfamiliar approaches, concepts and specialist vocabulary in order to learn about new insights that they might not otherwise encounter. The present quantitative-based study of Quaker prophesying is a fresh attempt to bring new life to this aspect of historical Quaker writings.

1. Introduction

This paper has three different but related aims. Firstly, it explores the nature and extent of the language of prophetic discourse published by over a hundred different Quakers in the second half of the seventeenth century. The apocalyptic message contained in these pamphlets, prose broadside texts and books was the main reason Friends developed their initial energetic campaign of preaching and pamphleteering. My research findings address these primary questions: What was the nature of the Quakers’ prophetic message, how was it expressed, and how does the impending ‘Day of the Lord’ message (as they saw the imminent end times) compare to the biblical book of Revelation upon which their eschatological cry is based? I present analytical findings from several aspects of Quaker prophecy: (i) the positive or negative nature of the range of Quaker eschatology; (ii) Quaker anti-Calvinist understandings; (iii) how the style and content of prophetic language changed in the published texts of many early Quaker writers. Quaker discourse is touched on in its wider early Modern sense beyond the present-day one of apocalypse.

The second aim of the paper is to introduce in more general terms a methodology little used previously in Quaker studies scholarship but employed in the prophecy enquiry introduced above (see also Roads 2012, 2017a, 2017b) (This method relies on simple corpus-based analytical techniques that yield quantitative results; these can be combined with qualitative observations. In the final part of the paper, I offer a very brief overview of the issue of interdisciplinary research generally before addressing some questions that an approach such as mine throws up. What are its advantages, to what uses can quantitative analytical techniques be put, and what are the challenges a researcher might face? I hope readers of this paper will feel able to engage with the crossover of disciplines referred to throughout.

2. Historical Prophecy and the Corpus Data

This section introduces my Quaker corpus, the nature of historical prophecy and linguistic issues connected to such analysis.

2.1. The Quaker Corpus

My research makes use of nearly 200 Quaker-authored texts (~638,000 words) ranging from 1650–1699. Although compiled randomly, one purpose of making the collection was to spread the net as widely as possible so that a breadth of representation could be achieved, thus enabling a measure of generalisation to be put forward. More on the representativeness of the corpus, and therefore its reliability, in Section 3.2. Many of the early Quaker published texts either include or are primarily concerned to express an eschatological message. However, George Fox, as one of the founders of Quakerism, urged Friends to warn of a need to repent before the imminent end of time (as he and other Quakers saw it), and an eschatological warning was not his primary message, which was rather one of discovery of a new way of accessing the Truth (Ambler 2007, pp. 180–86) and a mission to proclaim this discovery to the world. It was other Quakers who felt more impelled to take it upon themselves to disseminate in print their urgent message of repentance. Barbour and Roberts (2004, pp. 47–148) assign a chapter on Tracts to Proclaim the Day of Visitation. Gwyn (1986) devotes an entire book to the subject as do Dandelion et al. (1998). Moore (2000, pp. 61–67) reports on the ‘eschatological standpoint’ of some of the Quakers, describing it as their imminent expectation of ‘the Kingdom of God’ on earth; Part 1 (pp. 21–103) of Smith’s (1989) study covers this topic extensively.

Not all prophetic writing at this time has the sense of ‘predictive’, however. In the seventeenth century, the term denoted more than predictive discourse and included such forms of utterance as public preaching, reading and interpreting the Scriptures in persuasive discourses that referred to the present as well as the future, and making divinely inspired prophetic judgements (Smith 1989; Aune 1991; Font Paz 2009). My enquiry into such texts takes as a starting point the linguistic expression of futurity as the central expression of the speech act of religious predictive prophesying.

2.2. The Nature of ‘Prophecy’

Much has been written concerning Christian prophecy, starting from Old Testament biblical texts to the later prophetic writings of the New Testament and then the revival of prophetic visions and outpourings during the seventeenth century. Aune (1991, p. 198) comments, as part of his study on early Christianity, that the act of prophesying ‘can be understood in many different ways’. Sweeney (1996, pp. 18–30) finds at least seventeen forms of prophecy, including prophetic judgement, announcements of a sign or event, ‘woe’ oracle, vision report and proclamation of Christ’s coming. All of these forms are found in the Quaker writings, but the most common variety published during the early years of the Quaker movement was the foretelling, apocalyptic form.

In order to disentangle those (parts of) Quaker texts that express prediction (mainly connected to non-spiritual utterances) from prophecy, it is necessary to identify and then exclude such elements before analysing the remainder. Searle and Vanderveken (2009) claim that a writer or speaker who is forecasting is making:

… a special kind of prediction that is based on relatively clear signs of how something seems to be shaping up. To foretell is to ‘tell’ in advance, often something rather vague.

They contrast this with the proposition that prophecy has the underlying meaning (known in speech act theory as ‘illocutionary force’) of a prediction:

[Prophesy] has the illocutionary force of a prediction with an additional, particularly with an additional, particularly authoritative model of achievement. The latter has to do with the authority of an oracle … of God or of divine revelation. The speaker presupposes that he has good reasons for the belief to the point of certitude (my italics).

(Searle and Vanderveken 2009, p. 173)

In theory, therefore, it should be possible to retrieve all the instances of this special type of prediction. The next section explains how can these be identified.

2.3. The Linguistic Realisation of Apocalyptic Prophecy

What is exhortatory prophecy in the Quaker context? To find out, I developed a set of criteria as preparation for categorising Quaker apocalyptic prophetic statements, building on a body of theoretical work (Wierzbicka 1987, pp. 269–70; Gotti et al. 2002, pp. 224–28; Gotti 2003, pp. 267–96; Clenenden 2003). The process for carrying out the computer-based searches is described in Section 3 of the present paper. I used the application Wordsmith Tools (Scott 2008), but other software packages and online resources are also available.

The list presented below contains some language-related technical terms. This cannot entirely be avoided but is typical of cases where my related discipline of linguistics rubs up against a theologically-based discussion, in this case a discussion on the nature of prophesying. Modal auxiliary verbs are verbs such as will, shall, can, might, must that are not used on their own but rather in conjunction with a main verb. They can express many meanings in English; the two forms and functions relevant to this enquiry are shall and will, which together with a main verb are one way of expressing future actions in English, for instance in the sentence: she will come back tomorrow.

Below is the final set showing which sets of occurrences I included and which I excluded from my corpus results: (The forms SHALL/WILL here denote all instances of shall, shalt, will, wilt. It is a convention in linguistics and adopted in this paper to indicate a headword or ‘lexeme’ using small capitals. This differentiates from individual word or phrase examples that are formatted simply in italics. The original spelling remains unchanged.).

Included

Excluded

SHALL/WILL instances conveying a dynamic intention. The writer’s subjective or ‘God-given’ assertion that an event was to take place but no time or natural place is stated. Utterances presumed to be spoken directly by God and using the first person singular pronoun. Biblical quotations, whether referenced as such by the writer or woven into the general text.

Adjacent conditional subordinate clauses as they could be categorized as warnings. SHALL/WILL instances according to Gotti et al.’s’s (2002)classification, signifying: volition, refusal, promise, order, prohibition, inference, possibility. Interrogatives (deemed not to be carrying sufficient force for prophecy). Instances of neutral (non-religious) prediction, habit and assurance. Occurrences of the verb prophesy as denoting a different prophecy type from apocalyptic. Any item that carried a weak sense of prophecy, based on my qualitative judgement, or was ambiguous in the context. The lexical verb will (signifying volition).

Two studies that have informed my approach are based on the grammatical realisation of prediction through certain modal auxiliaries. Gotti et al. (2002, p. 41) single out ‘prophecy’ as a sub-set of ‘prediction’ by asserting that it is the strength of a prediction in context that makes it prophetic. They arrive at this definition by analysis of modal auxiliary usage, specifically SHALL and WILL, in a historical context. The following examples from my corpus illustrate how I classify such instances (linguists refer to this as ‘coding’). Examples (1) and (2) show the type of eschatological foretelling that is a central component. References for the source texts are listed in the Appendix A:

(1)

Therefore hear this you hard hearted ones and stiffnecked, who rebel against the Lord your maker, there is a day hastening, and an hour approaching, which will make you gnash your teeth for very anguish, and weepe and howl for very sorrow, to consider how you have slighted the day of your visitation (M. 1660),

(2)

For the candle of the wicked is to be put out, and then a woful day will be known unto all those who are left in darkness; they will not know whither to go (Gillman 1663).

Examples (3) and (4), on the other hand, are coded as simple prediction as they lack the special force or other qualifying criteria necessary for classifying as prophecy:

(3)

But as Darkness increaseth in thee, so the Power of it will bind thee down as a Chain, and smother every good desire in thee (Crisp 1666),

(4)

The Lord God of Power will exalt his Princely Scepter, over all Scepters, Thrones and Dominions, Kingdoms and Countries, and none shall be able to prevent him (Evans 1663).

These decisions are necessarily subjective and there is a fuzzy boundary separating one rater’s coding as ‘prediction’ with another’s coding as ‘prophecy’. The most practical descriptions for identifying the pragmatic meanings of predictive prophesying comes from Wierzbicka (1987):

Prophesying, like predicting, involves saying what will happen in the future. There is, however, something mysterious and almost mystical about it. The prophet feels that he knows things about the future that other people don’t know, and can’t know, because they are ‘hidden’ from them. For some reason, that the prophet himself doesn’t quite understand, he can ‘see’ in his mind what other people can’t see. … He can ‘see’ … things about the future because he has somehow been chosen by God to convey publicly a message which is important ‘to everyone’. … It is inconceivable that a prophecy should include a date. Prophesying differs in this respect not only from predicting, which can be extremely precise, but even from fortune-telling.

(Wierzbicka 1987, pp. 269–70)

In addition to the criteria listed in this section, a list of ‘I-statements’, suggested by Wierzbicka for testing the nature of a potentially prophetic utterance, was helpful as a second approach for classifying some of the more ambiguous occurrences. Meanings that help to identify linguistic realizations of prophetic utterances:

• I know things about the future that other people don’t know
• I imagine I can see these things happen
• I don’t know how and when they will happen
• I know that I haven’t come to know these things in the way people are expected to come to know things
• I feel as if somebody else was speaking through me
• I think God is speaking through me
• I say: sometime in the future X will happen
• I assume it will be important for everyone
• I assume that people should do something because of that
• I say this because I feel I have to say it

Wierzbicka (1987, p. 270).

2.4. Quaker Prophetic Discourse: Positive and Negative Connotations

The results of my analysis show that approximately a third (34%) of all SHALL/WILL occurrences are prophetic statements. This reduces to 24%, or approximately a quarter if we limit the results to unconditional statements (namely, propositions that do not begin with if …). I classify those separately as ‘warnings’ realised using conditional clauses, and as a special case syntactically I have excluded those findings from the present paper. These are the raw figures obtained via the software:

• 4449 occurrences of SHALL/WILL modal auxiliaries (all senses, not just futurity),
• 1510 occurrences of SHALL/WILL classified as eschatological prophecy statements.

Applying the criteria for identifying apocalyptic prophecy as listed above shows that some instances carry a positive or optimistic predictive prophecy, others a negative or pessimistic and a few carry a neutral sense. The results show an interesting mix of optimistic as well as pessimistic apocalyptic statements. The first set of examples (5) to (7) is comprised of negatively-loaded prophetic statements:

(5)

Behold the day is coming, ye shall wish the Mountaines to fall upon you, and the rockes to cover you, and shall seek death, but shall not finde it (Howgill and Burrough 1655),

(6)

The Lord will hew you down and consume you, and your remembrance shall stink for ever, the Lord will cast you out of his sight (Taylor et al. 1655),

(7)

In the power of his spirit in his poor contemptible people, by that which the world calls foolishnesse, will the Lord confound not only your Languages; but the very Places, Shops, and Nurseries of abomination, and all Schools, Colledges and Steeple-houses shall all be laid waste, having bin the Nurseries of the Whore and false Prophets (Zachary 1660).

The Quaker positively-loaded prophecies often convey the Christian message of hope at the End Times, or (from the point of view of the writers) the final triumph of ‘God’s people’ over the wicked and their persecutors. Many instances are either biblical or paraphrased from the Bible. Examples (8) and (9) follow:

(8)

And in the judgment, all the powers of darknesse which held you captive will vanish, the lofty spirit will be brought down, and the humble and meeke spirit raised (Zachary 1660),

(9)

And for ever blessed are all they that hold out unto the end, and keep the Word of his Patience, they shall be hid in the hollow of his hand, kept safe in his Pavilion, and a defence will he make about his people, and Zion shall be the praise of the whole Earth, a beautiful City, compact, compassed about with salvation for Walls and Bulwarks, and the Mountain of his Holiness shall be on the top of all Mountains, and no Destroiers shall be found in it; the pure in heart shall see it (Nayler 1661).

The neutrally-loaded connotations mostly concern the eschatological prophecy of the Last Judgement but without specifying whether that is to be welcome or unwelcome information for the addressees. For example: ‘everyone shall have reward according to their works’ (in Stoddard et al. 1655). The illocutionary intention was probably to give warning to the readers, but the context requires that, for the present study, we should assign the benefit of the doubt. Occurrences that were not classified as prophecy mostly concern a foretelling of some action on God’s part in the temporal world, for example: ‘Come to Gods Witness, it will abide with you; and as you abide with it, you will know its power and its leadings’ (Blackborow 1658). This group of items is coded as ‘spiritual counsel’ for the purposes of this study and not as apocalyptic prophecy.

2.5. ‘Foretelling’ Prophecy Comparisons

What else can the results tell us? One line of enquiry is to compare the findings of the prophetic set of SHALL/WILL instances with those of the Book of Revelation, the source for many of the Quakers’ apocalyptic utterances. I performed this exercise using a machine-readable text file of the King James Version.

Table 1. Comparison of all ‘foretelling’ instances of SHALL/WILL forms, normed per 10,000 words, actual figures in brackets.

	Quaker Corpus	Book of Revelation (KJV)
all occurrences of SHALL/WILL modals	69.7 (4449)	116.6 (145)
all occurrences of prophecy speech acts	23.6 (1510)	101.3 (126)

presents figures comparing these modal frequencies in both the Quaker corpus and the Book of Revelation.

Revelation shows a much higher frequency count compared to the Quaker results (101.3 per 10,000 words), but its function is primarily prophetic and so serves as a benchmark text for all Christian-era prophesying. I now turn to comparisons for predictive prophetic utterances carrying various connotations. The figures make interesting comparisons. Revelation is all about the future so one would expect relatively more instances than in the range of Quaker texts, and indeed the negative connotations found in Revelation are more frequent than the Quakers provide., But Revelation is less often connected to a positive, glorious future in the minds of doomsayers of any era and the Quakers at this time seem to be no exception.

Table 2. Comparison by frequency of context-based prophecy-connotations: Quaker corpus and the Book of Revelation, KJV.

	Quaker Corpus	Book of Revelation (KJV)
negative prophecies	15.5 (992)	31.3 (39)
positive prophecies	6.8 (439)	40.2 (50)
neutral prophecies	1.3 (85)	25.7 (32)

below presents comparisons for instances of positive, negative and neutral connotation. Observed scores for the Quaker set show just under seven words per 10,000 words compared to the biblical set of just over 40 per 10,000 words. (An explanation of norming the word count frequencies per x number of words is explained in Section 3.5). The scores for positive prophetic utterances, for example, reveal a much higher value for those found in the Book of Revelation compared to the Quaker corpus data. The difference in the negative connotations shows a relatively smaller gap.

Next, Figure 1 below presents a pie chart showing the ratios of positive versus negative prophetic utterances for each of the two datasets. The left-hand chart refers to the ratios for Revelation and the right-hand chart refers to the ratios for the Quaker corpus.

The proportion of prophetic statements in the Book of Revelation classified in the present study as positive is greater than the negatively classified set. This contrasts strongly with the early Quaker interpretation of the apocalypse according to the corpus evidence. In other words, the Quakers’ vision of the ‘Day of the Lord’ is a good deal more pessimistic than the biblical one.

2.6. Quaker Non-Calvinist Understandings

Can corpus searches tell us anything about the Quakers’ anti-Calvinist prophetic writings and their rejection of predestination? Much has been written about the struggles early Quakers had when challenged about their apparent denial of Christ, Scripture and so on—see, for example, Moore (2000, pp. 98–111). They were often accused of blasphemy and imprisoned for it, as described in Manning’s paper Accusations of Blasphemy. (Manning 2009). Some of the contemporary writings of this period reflect this position, notably Robert Barclay’s (1678) Apology (see Spencer 2015, pp. 142–44, for a very readable summary). This topic deserves an article all to itself but perhaps a simple search on the lexical headword DENY can serve as brief illustration. The corpus retrieves 928 instances of this verb, excluding all gerunds. An inspection of the total active/passive forms produces useful patternings. (All these sources are cited in the Appendix A). The occurrences of the active mood forms of DENY account for 86% of the total number, for example:

(10)

The worlds Church we deny; for the true Church is the Body of Christ (Parnell 1655),

(11)

This deceiver has falsly said, we deny the Resurrection (Fuce 1659),

(12)

Those outward Forms and Ceremonies, which we deny and abhor, without exception (Roberts 1670).

Many of these examples come from the ‘doctrinal dispute’ group of texts in which the Quakers again and again describe their interpretation of their religious experiences of the Christian faith, refuting misunderstandings and objections by their opponents. One extract from the smaller sample of DENY in the passive form will have to suffice here, although the richness of the corpus findings certainly deserve more detailed scrutiny. The online Oxford English Dictionary (OED) gives several senses for this lexical item1:

(13)

And this I testifie that it is onely by grace that salvation is wrought, and not by any thing of mans own, for where grace is truly owned their self is as truly denied in all its willing & working; for when that is owned and joined to, which is Gods power, then all mans work which he has wrought, and all his fair buildings which he had framed in self, in his own power and wisdome, though glorious in appearance, yet all must be pulled down and an utter desolation must be known & witnessed of whatsoever hath been done out of the grace, out of the light, out of the life and power of God (Wight 1659, p. 2).

Sense 3a in the OED defines DENY as: ‘to refuse to admit the truth of a doctrine or tenet’, 4a has: ‘to repudiate, disown, renounce, disavow’. An obsolete sense (#10) is defined as ‘to refuse to accept’. These senses do match many of the occurrences in the corpus. The passive meaning is defined as ‘to be rejected’. There are examples of this too, for example:

(14)

Now seeing it is so, that the immediate Spirit of Christ, and its Immediate Revelation is denied by him; we demand of him, how he came to be a Minister, and whence he hath his Ability? (Burnyeate 1691).

A similar query on the headword ELECT—still in the conceptual field of non-Calvinism—retrieves 88 occurrences in 38 different publications in the corpus, including such authors as James Nayler, Edward Burrough, John Burnyeate and Dorothy White.

(15)

This is the Election, make him sure to your selves in life and death, so shall you be changed with him; and though you go down into the lower parts of the Earth in the carnal body, he shall raise you spiritual (Nayler 1661).

The present article can no more than provide a brief illustration of discourse corpus-analysis as a valuable starting point for interrogating text, as the above short examples demonstrate. Further work beckons. The final section in this part of the paper looks for indications of possible diachronic change of style as evidenced by the data.

2.7. Quaker Prophetic Style in the Corpus: 1650–1699

I wondered if there was any discernible change in style or rhetoric over the whole fifty-year period of the Quaker prophecy corpus. Moore (2000, p. 204), for example, describes the changes taking place in Quakerism from the 1660s onwards, especially in the context of the Second Day Morning Meeting’s (SDMM) work (influential from 1673 onwards). The SDMM began its reviewing and editing work in 1673 and this had a major impact on the nature of the texts before and after that crucial date. We know that after 1672 there is a closer scrutiny by the Meeting of the published output, resulting in some convergence of style and message (Hagglund 2015, pp. 483–86). (Because of the importance of this watershed date for my research, I decided to clearly divide my corpus into two sub-corpora, with texts published before 1673 listed as Qcorpus1 and texts published after that date listed as Qcorpus1).

Moore (2000, p. 214) suggests there is a reduction in the stridency of tone between the styles of the pamphlets she has studied that were published during the first two periods (broadly speaking) of seventeenth-century English Quakers. She observes:

In the course of a few years, the Quaker movement changed from being one of the most radical of the sects that were looking for the coming of the Kingdom of God on earth, and became an introverted body, primarily concerned with its own internal life, while Quaker theological statements increasingly used the language of traditional Christianity.

(Moore 2000, p. 214)

Hall (1992, p. 82), in his careful study of the work of the SDMM, concludes that it was ‘filtering out the irrational, fanatical, repetitive, illiterate or untruthful from the varied manuscripts submitted’. Moore (2012, p. 16) agrees that the SDMM’s policy was to aim for submitted texts to contain less provocative language (Moore 2012, p. 16) and to reject any proposed publications, particularly ‘apocalyptic prophecy which was seen as incompatible with the ethos of the times’ (Moore 2012, p. 11). She also reports on her ongoing research of the historical situation between 1670 and 1700 when ‘the urgent problem was to convince the authorities of Quaker theological respectability’ (Moore 2012, p. 17). Barbour and Roberts agree (Barbour and Roberts 2004, p. 574), commenting that production of apocalyptic-related pamphlets gradually lessened over this later period, publication of such texts being at its highest in the 1650s. Other scholars note the earlier style, which they suggest was disappearing in the last quarter of the century: O’Malley (1982, p. 86) cites the ‘aggressive prophetic material that had been produced in the 1650s’ and Keeble (1995, pp. 121–22) describes the earlier text style as ‘the assertiveness of Quaker discourse’ with its preference for ‘castigation and denunciation’. Bauman (1998, pp. 145–50) compares the earlier prophetic ministers that spoke (i.e., wrote) ‘with vehement intensity, to the extent … that they were frequently accused of “railing” … The Lamb’s War was not to be waged with restraint’. He summarises these changes as evidence of ‘the routinisation of prophetic religion’ and notes that by even the 1670s, ‘all tone and action was to be moderated.’

A comparison of emotionally-charged language between prophetic texts before and after 1673 strengthens the impression of a less strident, forceful tone in many of the later tracts. (Barbour and Roberts 2004, p. 53, claim that the texts actually became more ‘strident’ between 1660 and 1673). Figure 2 gives some examples, presented in concordance lines2 that show the immediate context of high-frequency terms observed in the Quaker eschatological discourse: DARK, DESTROY, JUDGE, WICKED. All forms of these words are included in each headword.

Retrieval of exhortatory prophetic language used by many Quaker authors can be compared by relative percentage points across the two sub-corpora. The headwords investigated are: REPENT, WARN, REPROVE, INIQUITY, PLEAD, CURSE, PERISH, ACCUSE, TREMBLE, CONVINCE, PERSUADE, PUNISH, HOWL, and SHAKE (see Figure 3 below). All these lexical items are of relatively high frequency in the data. The histogram shows that occurrences of REPENT, already a favorite word, actually increase during the fourth quarter of the century. There is an overall decrease in occurrences except for the verbs CONVINCE, PERSUADE and SHAKE, and, in general, this colorful language is less evident in Qcorpus2, that is, post 1673. I therefore estimate a reduction of strength of the underlying force for expressions of urgent warning.

For the lexical item woe, only a quarter of the occurrences and its underlying pragmatic force of prophesying are found in the post-1673 sub-corpus. However, there is an increase in tracts entitled ‘testimony’ or ‘declaration’ as a type of preaching rhetoric. As the century came to a close, there were several extra-linguistic historical events both within the Quaker movement and in England in general. This resulted in an increase of a specific type of Quaker prophetic genre that I call ‘spiritual counsel’ as well as similar texts that relate their personal experience. There is one indication that the organisation of the emerging Quaker movement was starting to become more institutionalised (Bauman 1998, p. 137) towards the end of the seventeenth century. This is evident in the increasing frequency of formulaic phrases found in the texts. One of these contains the headword CLEAR, as in these little snippets:

• I have cleared my conscience/my self,
• for the clearing of my conscience/of truth,
• to clear that which was upon me.

These often occur as the closing phrases to published correspondence, termed ‘open letters’ by Green (2000, p. 413), texts that aim to persuade individuals or groups to repent and change their behaviour. The phrase with its variants expressed frustration at the lack of understanding on the part of the recipient and an expression of ‘washing my hands’ of the apparent sinfulness. More research into this and similar formulaic phrases both for the seventeenth century and into the eighteenth century could prove instructive. However, change is never clear-cut and there is a fuzzy boundary even after 1673. But change there is. This section concludes the presentation of research findings into early Quaker prophetic language. Section 3 describes the methodology and processes employed in this corpus-based approach to textual enquiries.

3. Corpus-Based Tools and Techniques Explained

Non-linguistics specialists may feel daunted by the unfamiliar terms and processes used in the interrogation of electronic texts through specific software. This section takes these readers through some straightforward explanations.

3.1. The Corpus-Based Approach to Textual Studies

This section introduces, describes and explains some of the available processes in a corpus-based enquiry. Certain technical terms and methods mentioned in the section on prophecy are covered in general terms. It first responds briefly to the question posed in the introduction regarding the benefits of carrying out quantitative linguistic analysis in text-based investigations.

There are arguments for and against relying on quantitative methods for research into discourse, literary criticism and other aspects of historical textual scholarship. The traditional, qualitative approach relies on close reading and detailed knowledge of the subject matter. Corpus research can complement that by providing objective grounds for intuitive and possibly biased or opinion-based arguments. It can also confirm, extend or refute declarative statements that are believed to be true but have never been tested. Often, qualitative research is perforce small-scale; corpus findings can scale up and then generalize from much larger samples of datasets. The type of research questions will be different compared to qualitative hypotheses, simply because of the way that data are interrogated. As the sections below will illustrate, search queries must be specific and language-based. Asking a dataset to find all the metaphors, for instance, will not work; questions must be formulated more precisely, often in an iterative progression. On the other hand, relatively small samples can lead to extravagant claims, an over-reliance on the raw results can mask the bigger picture, and the increasingly sophisticated software architecture can blind us to the content of the material under investigation. Increasingly, researchers are finding inspirational ways of combining the two approaches: traditional and quantitative. With these thoughts in mind, let us turn to the practicalities.

3.2. Representativeness

There are several competing approaches to the construction of a corpus that can be said to be representative of the language community or writers it aims to stand for. In order to claim, as I do, that there are distinctive features of Quaker language and discourse that are traceable throughout the canon of seventeenth-century Quaker texts regardless of the writer, the data must be as robust as possible in order to stand as representing that material. Some selection or exclusion is necessary within the general aim of randomness. The full corpus details are listed in the Appendix A.

Compilers of electronic corpora have differed over the advantages of sampling text extracts of equal size or of including full-text items regardless of size (see, for example, Sinclair 1991, p. 19; Biber 1993, p. 252; Meyer 2002, p. 44). The Quaker corpus contains both full texts and text extracts. I arrived at this compromise because of the large variation of full-text sizes available to me. The shortest text in Qcorpus1 has 602 words (Robeson 1662) and the shortest in Qcorpus2 is a London Yearly Meeting minute from 1696 containing just 428 words. The largest file, an outlier with 13,686 words, comes from Isaac Graye (1657). My findings are dependent on the confidence with which I can make generalizations from probabilistic data about the distinctiveness of early Quaker style and language, as well as the identification of certain typical features. My confidence is firmly based on the comprehensive variety of text types, sizes, authorships and dates that are the basis of the corpus. My collection is restricted to pamphlets and other printed holdings available at Friends House, London, in the Library of the Religious Society of Friends. In that way, I could be certain of verifying the artifacts at first hand in terms of edition and physical state of each copy. The transcriptions were laboriously made through a combination of keyboarding, scanning plus optical character recognition, and speech recognition. Texts already online and available could for the most part not be relied on for accuracy sadly, nor always for integrity of original spelling or punctuation.

The robustness of the findings supports my generalizations from probabilistic data about the distinctiveness of early Quaker style and language, as well as the identification of certain typical features. This is so both for the present study and other research I have published (see for example, Roads 2014, 2017b]). My confidence is firmly based on the comprehensive variety of text types, sizes, authorships and dates that are the basis of my Quaker corpus.

3.3. Corpus-Based Techniques

Different software applications offer a range of tools for corpus-based analysis. The examples below are all derived from Wordsmith Tools 5 (Scott 2008, Liverpool: Lexical Analysis Software) First, the ASCI files (sometimes called text files) containing the relevant texts must be loaded into the application. Then word lists can be generated either alphabetically or by frequency. Usually percentages are given (see Figure 4, right-hand column) for how frequent a word is across all the texts combined. Other statistical information is available but is not discussed here. Figure 4 below is a sample of a wordlist from the Quaker corpus, generated by descending order of frequency. It will be observed that the most frequent words are grammar words, not content (lexical) ones. This is typical of most corpora.

Individual words or short phrases can be used as search terms in the concordancer tool. This is also true if a grammatical part of speech is of interest, such as adjectives or interjections. Figure 5 shows selected concordances lines for the phrase within and without. The lines are sorted alphabetically in this case by the first word (in red) to the left of the key phrase, although other possibilities are available for using the sort function to search for patterns of use that might otherwise be hidden. This aids quantitative analysis in a different approach to the more usual close reading of a text that can produce qualitative findings.

By searching on a ‘keyword’ or phrase, it is possible to produce a list of all occurrences of that item, with a small amount of co-text, normally a certain number of words or characters (determined by specific research needs), either side of the central item. As Figure 6 shows, variant spellings can be accommodated. Figure 6 and Figure 7 present two examples of concordancer output (another chance too, to inspect examples of the node word DENY). In Figure 6, collocating words to the 1st left are shown in red. In Figure 7, the second order sort shows collocations in green to the 1st left of the node column; the SET column on the far right is used to code for L (=literal sense) and F (=figurative sense).

Words or phrases collocating with the node item are thrown into sharp relief visually and nuances more easily detected. The visible co-text of each line can be increased to suit the research aim and it is also possible to locate and visit the full textual source of each line within the corpus.

3.4. Collocations

A search on the keyword power retrieves a number of words that appear frequently as co-text, for instance: dark power, false power. Positive senses include eternal power, pure power. The same method searching on spirit and spirits brings up instances of envious spirits but also heavenly spirit.

Table 3. Results of search queries of collocations relating to the items power and spirit(s).

	Negative Collocations	Positive Collocations
power	dark, condemning, cursed, destroying, Devil’s, earthly, false, Satan’s, wicked.	eternal, living, pure, higher, invisible, just, mighty, pure, righteous, spreading, unalterable, unlimited.
spirit(s) [several meanings of spirit]	damned, envious, false, fiery, frothy, idolatrous, malicious, prejudiced, ranting, self-exalting, unclean, unsavoury.	able, heavenly, holy, pure.

shows both these sets of findings.

3.5. Using the Corpus-Based Approach—Some Techniques

A certain amount of learning faces a researcher the first time they consider trying the corpus-based approach. Used effectively, new findings become possible through the discovery of patterns of use or (in)frequency of linguistic features. Comparisons can be made with other collections of text or even just two single texts, depending on the research questions. However, there are limitations to the approach and creativity has to be applied in order to choose the right search terms of the corpus that has been loaded into the software. The search terms have to be precise so there is no point in interrogating a corpus for, say, all sarcastic comments in general, or choosing too specific a phrase. A key Quaker phrase, much used by Friends in the present-day is that of God. However, a richer seam is tapped if a close reading is made before performing a search. Paraphrases such as something of God or within me/you/thee may suggest breaking down a concept into a range of syntax and vocabulary.

A second difficulty arises in making comparisons. Is the comparison valid? Are the texts similar enough that differences are significant? The discourse purposes of the comparison corpus or corpora need to be close or the results will be meaningless. For example, I consulted the Corpus of English Religious Prose (COERP) as a possible comparator to my full Quaker Historical Corpus (QHC) but discovered that mainstream ‘religious’ texts comprise sermons, treatises and instructional material such as catechisms. Much early Quaker published work (pamphlets, prose broadsides and tracts) is polemical; only a relatively small proportion of the Quaker corpus covers what I term ‘spiritual counsel’, sermons, for example. The COERP could serve for simple word frequencies but I found it to be of limited use for anything more introspective. Similarly, comparing seventeenth-century Quaker texts and present-day writing by Friends (for example using a version of Quaker Faith and Practice, Britain Yearly Meeting, as a corpus) will produce findings, certainly, but those results may not tell us much that could serve as reliable generalizations3. Again, the formulation of good research questions is crucial.

There are two technical aspects to be considered before plunging into the world of corpus linguistics. The first is the benefit of acquiring a reasonable knowledge of English grammar and syntax terms and concepts. This is important when pursuing search queries or in avoiding errors that might wreck the validity of the results. A very readable grammar book such as the (corpus-derived) Longman Student Grammar (Biber et al. 2002) would make a good source of support.

The second aspect is more problematical and it concerns the statistical interpretation of findings. Many corpus linguists provide a range of quantitative measurements in their studies, and the range of software such as Wordsmith Tools, Lancsbox, or AntConc (see bibliography for details) are tempting to use in one’s observations. If a Quaker studies researcher is willing to fully understand the statistical findings on offer that is good, but students for example have been known to come to unsafe conclusions based on poor understanding of these methods. Quaker research studies, especially for historical texts are likely to be working with comparatively small datasets. It is important not to over-generalize quantitative findings from small samples. In my view, statistical findings are not essential in our field unless researchers are very sure of their ground. My own experience is to make good use of two simple quantitative tools. The first, essential one if making comparisons between corpora of different sizes, is to ‘norm’ the raw data; that is to say, to perform a calculation that scales up the scores to per 1000 words or 10,000 words. This is done by dividing the raw score by the total word-count of the corpus or the individual text, and multiplying this by a certain number of words such as 100,000. The comparisons will then be meaningful even if the two or more datasets are of different sizes. The second tool that can make sense of quantitative findings is that of ratio. In Section 2.4, we compared certain positive and negative connotations. This can usefully be displayed graphically, as in Figure 8. The left-hand chart refers to findings in Qsub-corpus1 and the right-hand chart refers to findings in Qsub-corpus2.

We have now looked at issues and practicalities for any scholar considering embarking on combining the broad academic disciplines of religion and corpus analysis. In the final part of this paper, we go wider and discuss some implications of such an integrated approach.

4. Quaker Studies and Corpus Linguistics as Integrated Study

The final part of this paper is a meta-reflection on the original research approach in our first part which then looks more widely at the nature of this innovation, bringing out some of the potential challenges contained within it.

4.1. Issues in Interdisciplinary Research

Quaker studies has become a welcoming, broad discipline, open to many academic approaches. We first look at some of the issues regarding interdisciplinary research (henceforth IDR) in general before drilling down into our particular specialism. There is increasing recognition that academic disciplines benefit from a little cross-fertilization such as IDR provides. First, though, I will try to differentiate between interdisciplinary and multidisciplinary. Definitions are hard to come by but for the purpose of the present paper, I suggest that multidisciplinary implies scholars from different disciplines collaborating by drawing on their own knowledge and method base. Conversely, the key word for interdisciplinary research is ‘integration’ combining knowledge and methods from different disciplines, using a synthesis of approaches in such a way that the resulting findings could not be separated out but stand as a new set of insights. There is room in the field of Quaker studies for both approaches. The advantages and challenges are present in both. No longer need scholars work alone in their own watertight silos. Instead, there can be advantages in looking at ideas or methods that seem more at home in other seemingly unrelated disciplines. My own line of work straddles exactly this sort of boundary. My subject home is located in the natural language usage found in historical Quaker writings; however, my work sits somewhat at a tangent to theology, social science, history or literary criticism. This can make for miscommunication with colleagues who believe someone working in one aspect of Quaker studies to be more of an expert in these types of related fields than is the case. A special case of IDR is the domain of digital humanities, where scholars working in a variety of humanities research require an understanding of basic elements of computer manipulation of data, including keeping abreast of software applications, and techniques such as text annotating, processing and mining (and, indeed, what those terms even mean). Such colleagues, as Gardiner and Musto (2015) observe, will increasingly be aware of how such digital concepts may impact various branches of the humanities in terms of innovative approaches to traditional disciplinary thought. The danger to be avoided in this kind of collaboration is the tendency to ‘us’ and ‘them’: IT specialists versus ‘pure’ humanities researchers (Gardiner and Musto 2015, pp. 166–68).

Coming closer to the area of the research described in Section 1 and Section 2 of the present paper, I offer two examples of work that specifically integrates religion and linguistics. The first uses methods from corpus linguistics and critical discourse analysis to examine patterns of representation around the word Muslim in a large corpus of British newspaper articles. Both the methodology and the results of this study would be of interest to members of several academic communities, including religions, interfaith studies, media studies and linguistics (see Baker et al. (2013)).

In the field of historical religious thought, Kohnen et al. (2011) report findings from a study investigating early modern English religious prose as a ‘conservative language variety resistant to language change and linguistic innovation’. They apply quantitative analytical methods using the corpus previously referred to in Section 3.5 to compare religious language with the standard language variety used in all written genres of the period. Although they claim this to be in the evolving tradition of IDR, their work is firmly located within the discipline of corpus linguistics and, I suggest, should be evaluated as such.

These and other examples raise the question, and, indeed, challenge, of evaluating interdisciplinary studies. Strang and McLeish put forward a convincing argument for meeting these challenges (Strang, Veronica and McLeish, Tom. 2015). They insist that:

IDR is foundational rather than merely additive to academic practice and can contribute usefully to discipline-based research as well. Its value lies in this potential as much as in its capacity to address complex problems and to integrate research creatively. Evaluating IDR encounters difficulties when approaches based in single disciplines are applied to it. It requires fresh criteria, developed within an interdisciplinary context.

(Strang, Veronica and McLeish, Tom. 2015, p. 3)

They believe that this (not so new) approach deserves either fresh outlets for publication that ‘value and recognise high quality IDR’ or that existing publications be prepared to accept the challenge on behalf of their readership. Perhaps the publication in which the present paper appears is one such innovative vehicle. Strang and McLeish (2015, p. 4) accept that much IDR derives from collaboration between academics from different disciplines, but the reality also includes individual scholars who are able to straddle more than one in order to go beyond their main area of expertise and draw on methods or theories from elsewhere.

4.2. Practical Aspects of IDR

A word should be said now about the practicalities of dealing with two or more sets of conventions in the world of cross-discipline research. Here, the field is still relatively in its infancy in a number of humanities disciplines (two example studies involving religion are: Wiist et al. 2010; Payne 2015). Publishers, editors, funders and examiners who may be well-trained and even distinguished in one discipline sometimes display discomfort in accepting academic or editorial conventions required from another discipline. Findings from a 2016 study by the British Academy for the Humanities and Social Sciences on the variety of academic situations that can result in IDR nevertheless cite advantages to embarking on such ventures (British Academy for Humanities and Social Sciences 2016). These include the cross-fertilisation of scholars learning and applying methods from unfamiliar disciplines, and collaboration between colleagues working in different and, on the surface, unlikely disciplines but where both sets are searching for areas of common interest. The emergence of disciplines such as digital humanities have implications for new blends of approaches in many other fields.

However, the report notes that the one serious barrier to encouraging such work is that of reliable evaluation, as observed by Strang and McLeish (2015). This can lead to a diminishing of the value of IDR in the wider academic community. Clearly, this is not a reason from walking away from attempts. As we have seen, there is validity in facing the challenges, both in how such studies can be reliably evaluated and how policy-makers can influence funders. A very practical book offering detailed advice for any early career scholars contemplating the IDR route is that by Lyall et al. (2011). Somewhat quirky chapter headings include: ‘Planning the Expedition’, ‘Charting a Course for an Interdisciplinary Career’ and ‘Navigating the Interdisciplinary Landscape’.

This brings me back full circle to my own research into discourse analysis of early Quaker prophesying using corpus-based methods. I trust there is something new here for anyone working in the broad reach of the Quaker studies field. What interest, however, might there be for my fellow linguists (corpus-based or not)? In that parallel world, the focus is frequently on theoretical issues connected to fields such as style, syntax, pragmatics, socio-linguistics, lexis, semantics or discourse. All of those rely on scientific analysis into either present-day or historical language, focusing on synchronic findings (across the same time period) or diachronic (looking at change over time). Computational linguistics goes even further into the scientific realm by developing and testing new methods of digitally-based investigations into areas such as big data, obscure data or more sophisticated methods of automatic retrieval of language-based output.

So what then are the implications of the present paper for both applied linguists (looking at the discourse of persuasion, stance and hidden attitudes, socio-linguistics, for example), and also for those with no direct interest in language per se? There are two strands to this answer. Firstly, the findings on early Quaker prophecy presented above. My findings indicate that a Quaker perspective on the apocalypse, based on the Book of Revelation as well as their own primary experience, is unduly (and erroneously?) pessimistic in its urgent warning. For many pamphleteers of the time, there was less joy in their message than literal ‘doom and gloom’. This argument has not been put forward elsewhere, to my knowledge. It should be of interest to scholars working in the early modern field of Christianity, dissent, doctrinal disputing and allied disciplines. It contextualizes Biblical studies, radicalism and the struggle of religious minorities, among other possibilities. The topic crosses over into wider academic fields of study such as print culture generally, and pamphleteering and the production and dissemination of early broadsides more particularly.

Secondly, a feature of my study that relates to other historical enquiries is that the corpus data derive largely from the social phenomenon known academically today as ‘from below’. This is comparable to the description in 1654 by the Quaker Richard Hubberthorne of his new community of Friends as ‘the under sort of people’ (in The Mittimus Answered, quoted by Moore 2000, p. 64). We might say that such writers were producing the equivalent of today’s ‘blogs’. In other words, we are now hearing voices of the mass of ordinary people previously in danger of being ignored by historians working from sources more focused on the public picture. I can illustrate this with reference to two twenty-first century publications concerned with early Quaker thought. Gwyn (2015, pp. 204–207) reaches for just three key Quaker writers (George Fox, James Nayler and Margaret Fell) when constructing his four-page summary of early Quaker eschatology as he interprets it. Moore (2000) has performed close-reading on many more texts (but from 1649–1666 only). Many of those do feature in my own corpus. However, even here she only refers to fourteen sources in her chapter on The Kingdom of the Lord, including Fox, Farnworth, Nayler, Howgill, Dewsbury, (Moore 2000, pp. 60–74). My own approach examines nearly 200 writers and covers half a century of published texts. The corpus-based approach enables non-linguists to gain much more extensive pictures and patterning as a way of empirically validating, extending or even contradicting the knowledge already put forward by close reading of a restricted range of texts.

A further example of fruitful collaboration with a scholar from another field is the conference paper given in 2016 on the early Quaker approach to knowledge and learning (Gill and Roads 2016). This integrated literary criticism through the perspective of digital corpus analysis, involving a creative sharing of our different disciplinary approaches. The paper provided a novel perspective for conference attenders coming from historical and theological backgrounds, and unused to the ways of corpus linguistics.

4.3. Concluding Remarks: Drawing Together the Threads

This article has addressed three distinctive areas of research. The first part presented a fine-grained enquiry into the language of early Quaker prophecy, leading to new insights into how this discourse was realized linguistically by a large number Quaker writers in the seventeenth century, and how their approach and sense of urgency compared to an important source for them: the Book of Revelation in the Bible. The ability to scale up the findings on which to base these generalizations about Quakerism stems from the nature of simple quantitative corpus techniques. The second part unpacked some of the specialist terms, techniques and tools used in the non-related discipline of corpus studies, and showed how these may be applied to aspects of text-based study which have so far received little attention in the Quaker Studies field. In the third part, we looked more widely at the nature of IDR and the implications for embarking on, evaluating and publishing such research. Some areas of academia have been combining and integrating approaches for many years now, paving the way for minds to be opened to new combinations of disciplines. My own subjective impression is that the field of religious studies has yet to catch up and embrace such innovations but there is much to look forward to in the coming decades. Who knows what this diversity of approach and method can reveal? Only time and confidence can break down old barriers.