If we extract the speeches of each of the individual speakers in the plays and combine them into a single document, we are then able to compare these composite ‘character’ texts with one another, within the same play or across a larger corpus. This allows us to view ‘characters’ in a shared, neutral space, independently of their local contexts in plays and of the larger structures of the plays from which they are extracted, such as authorial canon and genre.
In some ways, this returns us to the realities of the early modern theatre, since for actors ‘character’ was almost certainly a more important conceptual unit than ‘play’. Actors learning their parts, as Simon Palfrey and Tiffany Stern remind us, were supplied with a paper roll containing only their characters’ lines and necessary cues, and not the entire playbook. In this sense, as they note, ‘the part has a physical as well as an institutional reality’, and this physical aspect ‘facilitates not only intra-play, but inter-play references’.Footnote 1 Audiences and readers may similarly privilege ‘character’ in their appreciation of drama, as when characters like Tamburlaine or Faustus seem to loom larger than the plays in which they appear. Some characters, such as Shakespeare's Falstaff, may reappear in sequels or in entirely different plays. King Charles I famously replaced the titles of three of the plays in his copy of the Shakespeare Second Folio with the names of their main characters, Malvolio, Falstaff, and Paroles, apparently as substitute titles.Footnote 2 In 2016 Rebecca McCutcheon premiered a ‘new’ play ‘by’ Shakespeare, Margaret of Anjou, fashioned by Elizabeth Schafer and Philippa Kelly from Margaret's appearances in Richard the Third and the three Henry the Sixth plays.Footnote 3
We might expect the closest resemblances to be between characters created by the same authors, but there are reasons for similarities to appear across authorial canons as well. Impressive characters may well inspire derivatives and imitations. Recognisable stock characters are an important resource for playwrights, since they help audiences see lines of action and bring with them ready-made expectations, which can be fulfilled or frustrated. Since the repertoire of plots and settings will always be limited, shared structures will create dramatic niches to be filled by the same ‘sort’ of character, just as ecological niches in different parts of the globe may give rise to very similar evolutionary outcomes in species. Thus characters have their own identities and functions within particular plots, but they also fall into recognisable categories. When John Marston's boy actors confer on their roles in the Induction to Antonio and Mellida, with their ‘parts in their hands’, they speak in terms of such stock character types: a ‘proud’ duke who ‘strut[s]’ with his hair ‘stroke up’, a penurious lover, a fool, a parasite, a cross-dressing male lover, a braggadocio, a Stoic, and a mercurial duke's son – each describing how to ‘dispose [his] speech to the habit of [his] part’.Footnote 4
The methods of computational stylistics give us the opportunity to explore these resemblances across plays and to relate patterns in characters’ language use to broader structures in the drama. The quantitative part of our analysis is confined to the characters’ spoken dialogue, and then to what profiles of frequency among the commoner words can tell us. With these figures and profiles, we then return to the richer, more comprehensive matrix which embraces action, costume, casting, and dramatic meaning – literary history in the broadest sense – for context. The observations we make below – that Richard III is the quintessential mainstream character, Bosola and Flamineo are closer than most sequel characters, Queen Margaret and the Duke of Anjou are exact matches along the most important stylistic axes, and that Julius Caesar is a standout in its immediate theatrical context – enshrine truths of a sort, since we can trace each of them back to an empirical base. Yet in a larger sense they are best thought of as challenges to interpretation, each dependent on a particular frame of reference. It is a tribute to the complex codes in language and their robust underlying structures that we get even these occasional tantalising glimpses of deeper dramatic realities from a simple calculus of percentage counts.
This is an exercise in the empirical sociology of character. We start by looking at characters in the mass, rather than as individuals, on the basis of a literal-minded approach to language as a series of word tokens. This analysis shares the double-edged quality typical of most quantitative generalisations in the humanities and social sciences. Two people live in the same area, earn $1,000 a week (plus or minus $100), and are both vegetarian. This series of alignments is immediately suggestive, but when we interview the individuals concerned we may or may not sense some parallels in the dimensions which were our primary interest, such as world-view and personality. Even so, at least we shall have the comfort of a rigorous, explicit beginning to the search for the subtler differentiations and parallels.
Proximities between Characters
As the long history of character criticism in Shakespeare studies has shown,Footnote 5 it is tempting to treat each character in early modern English drama as one-of-a-kind, with motivations and life histories peculiar to themselves. In a sense, this is true: all characters reflect aspects of the individuality of their authors and actors, while characters based on recognisable figures, from famous rulers to infamous renegades, rely upon – and contribute towards – the shared histories of their namesakes. However, characters are also creations in a specific cultural context, products of imitation as well as invention, brought into being to figure within the particular structures of a constrained dramatic world. If we focus on the resemblances between characters rather than their differences, we can then bring out this patterned effect.
With this aim in mind, we extracted 666 characters with more than 2,000 words of dialogue from a corpus of 243 plays performed by professional companies between 1580 and 1642 (Appendix B). Confining the analysis to characters with larger speaking roles such as these narrows the overall range in size, and reduces size as a factor in comparisons. It also allows more of the ‘law of large numbers’ to come into play, so that local aberrations are evened out in a more extended to and fro of situations and motivations. We found the 100 most common function words in the plays overall and compiled a table of percentage counts for these words for each character. We then calculated a distance between each character and every other character based on a simple geometric relationship between their various word counts.Footnote 6 Our interest was to see which pairs of matching characters emerged, and to identify any wider patterns in these pairings.
Comparing each of the 666 characters with every other results in a total of 221,445 pairings. Of these, the closest was Tamburlaine from 1 Tamburlaine the Great and Tamburlaine from 2 Tamburlaine the Great. Figure 3.1 shows the distance between these two characters in relation to the other pairings in the set which involve a character reappearing in a sequel or sequels – the shorter the bar for a pairing, the closer are the two characters concerned.
Figure 3.1 Distances between characters in plays and their sequels, using percentage counts of the 100 most common function words combined by Squared Euclidean Distance.
Eight of the sixteen pairings have distances over 7.5 and fall into a crowded part of the overall distribution of distance scores. They are only a little closer as pairings than we would expect in a pair of characters taken at random from the full set (Figure 3.2).
Figure 3.2 Histogram of distances between 666 characters based on percentage counts of the 100 most common function words (221,445 pairings in total).
All the distances are represented in Figure 3.2, grouped in 34 cells covering the span from 0 to 85. This is a skewed normal distribution with a high peak: there is a long tail of distance scores to the high end, and the most populous cell – the range between 12.5 and 15 – contains 37,200 values, or 16.8 per cent of the whole. There are just two values under 2.5, and two over 82.5. The lowest value – that is, the shortest distance – is between the two Tamburlaine characters.
Reverting to Figure 3.1, we see that next closest pairing of a character appearing in a play and its sequel involves Byron from George Chapman's Byron's Conspiracy and the same character from Byron's Tragedy. In between (and not shown in Figure 3.1) are eight other pairings. In ascending order of distance, they are: Shakespeare's Richard III and Hoffman from Henry Chettle's eponymous play; Ithocles and Orgilus from John Ford's The Broken Heart; Herod and Antipater from the collaborative play named for them by Gervase Markham and William Sampson; two characters from different plays by John Webster – Bosola from The Duchess of Malfi and Flamineo from The White Devil – nihilistic, philosophical characters, both villainous yet not unambiguously evil; the eponymous Richard III and Volpone; Sejanus from Ben Jonson's play of that name and Cicero from Jonson's Catiline; two Thomas Middleton characters from different plays, both schemers in city comedies – Witgood, penniless nephew of a rich uncle in A Trick to Catch the Old One, and Quomodo, the rich conniving merchant from Michaelmas Term; and two characters from different plays by Christopher Marlowe – Faustus from Doctor Faustus and Barabas from The Jew of Malta.
This is a group of the very tightest pairings. The distances between them are very low indeed, belonging in the two extreme left-hand columns of Figure 3.2. Some of the characteristics of this group are as one might have predicted: all but the two involving Richard III share an author, for instance. It is a surprise, though, that two of the pairings bring together characters from the same play, since character differentiation is often regarded as an inherent characteristic of theatre: Hubert C. Heffner, for example, defined character as ‘the differentiation of one agent from another agent in the action’, and drama as ‘the art purely of character in action’.Footnote 7
Equally unexpected is the result that sequel characters, such as Falstaff in both parts of Henry the Fourth and King Edward in Thomas Heywood's two-part Edward the Fourth, have higher distances than this group. One would have thought that the conjunction of so many factors in sequel characters – the same author, genre, and dramatic niche and setting, as well as the same or similar plot dynamic – would guarantee a near perfect match, but they are exceeded in the event by quite unrelated characters. Of the face of it, this analysis suggests we should regard likeness between characters less as a matter of identity – a function of the dramatist's aim to represent a unique on-stage personality – and more as a matter of functional role, closer to the types Marston describes – proud dukes, penurious lovers, fools, parasites, and so on. Ithocles and Orgilus are cases in point. They are bitter enemies. Ithocles is determined to prevent the marriage of his sister to Orgilus, and finally murders him. Yet when viewed in the context of a vast mass of dramatic characters, they are close to indistinguishable, and resemble each other more than the Hal of the two parts of Henry the Fourth or indeed the Falstaffs from those two plays. Figure 3.3 shows how the percentages for the twenty most common function words in the set varies from the overall average for these two characters.
Figure 3.3 The z-scores for the 20 most common function words for Orgilus and Ithocles in John Ford's The Broken Heart, using averages and standard deviations from the full set of 666 characters with ≥ 2,000 words of dialogue.
The measure we use in this instance is the z-score, which is the difference between the observed count and the mean, divided by the standard deviation for that variable. The closer a word score is to the axis, the better the character's dialogue conforms to the overall average of the 243 plays. The negative scores to the left indicate that the character's count is lower than the average, and the positive scores to the right that it is higher. Me is an example of a word which is only just below the average for both characters. The general pattern, though, is that the words have high z-scores, meaning a marked variation from the average. Most vary from the average in the same direction – sixteen of the total twenty – and there are some very close matches: and, of, you, inpreposition, and have, for example. These characters are unusual, but unusual in the same way.
After the two Tamburlaine characters in the array of characters with a very close pairing come Richard III and Hoffman. This is another reciprocal pair, since Richard III is Hoffman's closest match also. Unlike the others mentioned so far, Richard III and Hoffman are paired as a result of their shared closeness to the common patterns in the set overall. Figure 3.4 shows the ten closest matches for Richard III and for the nine other characters with very low-scoring closest matches, the ten lowest overall in the set of 666.
Figure 3.4 Distances for the ten closest matches for ten characters with very close pairings with another character.
Richard III's closest match is second to the Tamburlaine pair, but his second closest match, third closest, and so on, are lower than the corresponding matches for any of the others. The Tamburlaine of 2 Tamburlaine the Great is an example of the opposite pattern: a very low score for the closest match, with the Tamburlaine of 1 Tamburlaine the Great, is followed by a jump in distance to the next closest – Friar Bacon in Robert Greene's Friar Bacon and Friar Bungay. This Tamburlaine character's closeness to his counterpart from the other Tamburlaine play is an exception – in general, he shows no great affinity with other characters. Richard III, on the other hand, is close to a range of other characters in a way that makes him the ‘odd man out’ in this group of ten. Richard's other matches, in order of closeness, are with Volpone, Faustus, Barabas, the eponymous Sir Thomas More, Pericles, Cardinal Wolsey (from Henry the Eighth), Antony (from Antony and Cleopatra), Arden (from Arden of Faversham), Young Geraldine (from Heywood's The English Traveller), and Elinor of Castile (from George Peele's Edward the First).
Although the plays from which they are drawn differ in authorship and date, these characters have some broad features in common. All are main characters. All but Elinor and Wolsey are signalled as the protagonist by their plays’ titles, and the exceptions both have large speaking parts and important roles in the main action. Volpone, feigning illness to gull his clients, Barabas, secretly planning evil on all around him, and Elinor, contriving to increase foreign influence at court and guilty of a jealous murder, could be described as scheming deceivers, and More and Wolsey as politicians, but in the group there are also character types less obviously related to Richard's role: Hoffman – revenger; Antony, Faustus, and Arden – tragic heroes in different styles; Geraldine – a wronged lover; and Pericles – hero of romance and adventure. We have to go beyond the idea of the Machiavellian or Vice-like villain to summarise the pattern of resemblance. The negatives are instructive: none of these characters are outlandish or mannered. If we examine the characters with the greatest distance from Richard, we find a tendency to extravagance and the exaggeration of some characteristic. Seven of the ten most distant are ‘humours’ characters, for instance, in the loose sense of comic characters with a marked obsession: Fluellen from Henry the Fifth, Belleur from John Fletcher's The Wild-Goose Chase, Wasp from Jonson's Bartholomew Fair, the Humorous Lieutenant from Fletcher's play of that name, Mistress Barnes from Henry Porter's 1 The Two Angry Women of Abingdon, Zuccone from Marston's Parasitaster, and Lantern Leatherhead from Bartholomew Fair.Footnote 8
Richard III has a large, mixed part which brings him close to the overall average score on the 100 most common function words. If we add up all the differences from the mean for the 100 most common function words for the 666 characters, Richard III has the lowest total – and Fluellen from Henry the Fifth the highest. Richard III has a profile unusually close to the overall pattern of the set. This comes about partly because he has a large speaking part – 8,368 words – one of only 11 characters out of 666 with more than 8,000 words of dialogue. Larger parts come closer to the mean by dint of their greater opportunities to balance departures in one direction with departures in others.
Yet this is only part of the explanation. Figure 3.5 shows that Richard III is remarkable for consistent low scores for his pairings with other characters, even among the characters with the largest parts.
Figure 3.5 The 10 closest pairings for the 11 characters with dialogues ≥ 8,000 words, ordered by the distance to their closest match.
Richard III is the model serious main character: his profile fits snugly with those of a number of others. While other characters may have a closer best match, he accumulates more close matches than they do moving along the succession of next closest pairings (Figure 3.4), and this is true for other characters with large speaking parts, as well as for the generality of other characters (Figure 3.5). The analysis helps us see Richard in the wider perspective as a standard all-round active protagonist, with a role balancing commentary and direct interaction with others.
Richard's full, rounded part brings him to the centre of the professional drama of the period under investigation, closer than any other character to the typical practice in speechmaking of the larger characters. His is the lowest average distance from other characters in the full table. Hunchbacked he may be, and extreme in the difference between his private thoughts and public declarations, but, taken all together, he is balanced and orthodox in terms of the style of his speeches. He has much to say that is orotund and periphrastic, but he is also sometimes savagely direct: ‘I wish the bastards dead’ (4.2.19). He gives orders, ponders situations, confers with associates, wheedles, and sometimes abuses enemies. He has what is surely an unusual amount of his part in soliloquies, but they are framed in observations, addresses to absent others and to himself in such a way that they do not tip his part overall away from the centre. He woos twice at length, as well as directing political and military strategy. He traverses the full range from intimate thoughts and dreams through private exchanges to public utterances. He is colloquial and also poetical – ‘Into the blind cave of eternal night’ (5.5.15). He is supremely confident and also self-hating and doubtful. The framework of style summarises his speeches as fitting the balance that is struck by the longer characters as a set, and overlooks the idiosyncrasies of his character which are so plain to audiences focusing on plot, the content of what he says, and his actions. It helps us see the roundedness of this part in more schematic, dramaturgical terms. His virtuosity in different situations gives him a style which conforms very closely to a standard main character in a serious play, so that in these terms he appears just as much as the brother, lover, king, and warrior of his preferred public image as the murderous Machiavel he privately confesses himself to be.
Fluellen from Henry the Fifth is the character sharing least with his closest fellow-characters. His nearest match is Compass from Jonson's The Magnetic Lady, but their proximity score is 22.9, which is on the high side for matches in general, let alone for closest matches (see Figure 3.2). Fluellen's is a very unusual part – an extended study in a limited repertoire, varying little from the narrow role of a comic aficionado of military tactics and ethos imbued with exaggerated national characteristics, with a restricted range of interactions which includes no casual or intimate relationships. Figure 3.6 shows the z-scores for Fluellen for the twenty most common function words, as in Figure 3.3.
Figure 3.6 The z-scores for 20 most common function words for Fluellen (Henry the Fifth), using averages and standard deviations from the full set of 666 characters with ≥ 2,000 words of dialogue.
Fluellen is highly unusual in his recourse to is, and also an outlier in using the, and, you, inpreposition, and your very freely. These reflect his insistent, repetitive speaking style among other things as a pioneer version of ‘stage Welsh’:
I think it is e'en Macedon where Alexander is porn. I tell you, captain, if you look inpreposition the maps of the world I warrant you sall find, inpreposition the comparisons between Macedon and Monmouth, that the situations, look you, is both alike. There is a river inpreposition Macedon, and there is also moreover a river at Monmouth. It is called Wye at Monmouth, but it is out of my prains what is the name of the other river – but ’tis [= it is] all one, ’tis [= it is] alike as my fingers is to my fingers, and there is salmons inpreposition both.
Figure 3.7 shows the scores for Fluellen's ten closest matches in the set, in the context of the ten closest matches for the next nine characters with the most distant closest matches. Fluellen is exceptional even in this company of idiosyncratic one-offs. His tenth closest match is as unusually remote as his closest. This group of ten (including Fluellen) is a collection of sui generis characters, not necessarily like each other, but rather brought together because they have narrowly focused, atypical speaking parts. We find in them not individuality so much as hypertrophied aspects of one kind or another. The Humorous Lieutenant, Humphrey Wasp, Mistress Barnes, Belleur, and Zuccone all have monomanias of various sorts – respectively, hypochondria, impatience, anger, bashfulness, and jealousy. Cardinal Como and King David, like Fluellen, have lop-sided, specialist roles, as intriguer and rhapsodising monarch respectively, to match Fluellen's narrow focus on the arts of war. Ricardo is a confessional, expostulating, scheming rogue, and Syphax is a portrait of unbridled lechery and ruthless cruelty, described by A. H. Bullen as ‘so prodigiously brutal as to appear perfectly grotesque’.Footnote 9
Figure 3.7 The z-scores for the ten closest matches for ten characters with very distant closest matches, ordered by distance to the closest match.
When considered separately from their home plays and matched with each other in this way, some characters emerge as worthy of scrutiny in new ways. Richard III appears as a balanced all-rounder rather than an egregious villain. The two Tamburlaines are truly one, in a way no other sequel characters are. Bosola and Flamineo, although from different plays, share a profile, as do Ithocles and Orgilus, who share a stage. Fluellen is an experiment in writing a major but flat character, more bounded in range than any of his rivals in fixedness in our sample, whether they be humours characters, termagants, extravagant rogues, or dyed-in-the-wool villains.
Newly Created Characters of 1599
Including as full a sweep as possible, as with the 666 larger characters, has its advantages. The numbers involved help show what is genuinely unusual, and the range gives more opportunity for the unexpected to emerge. However, we can also narrow the range to get a more local perspective, restricting ourselves for instance to a set of plays which one playgoer could have readily have seen, and close enough in time for memory to allow close comparison. One year's worth of new plays provides a basis for a set like this.
1599, the last year of the old century, is an attractive starting point. This is the year singled out by James Shapiro in his historical micro-study, and he makes a good argument that the two 1599 seasons show a new generation of playwrights – Ben Jonson, Thomas Dekker, and Thomas Heywood – starting to hit their straps and providing stimulating competition for Shakespeare, to replace the challenge offered by John Lyly, Thomas Kyd, Christopher Marlowe, George Peele, and Robert Greene, who had left the scene by 1597.Footnote 10 In 1599 the Globe was built on Bankside, and Thomas Kempe left the Chamberlain's Men. Numerous studies have commented that the most obvious difference across the sweep of early modern English drama occurs around 1600. D. J. Lake observes that ‘there was a fairly sudden revolution in the linguistic practices’ of several dramatists around 1599–1600, as they began to use contracted and colloquial forms at a much higher rate.Footnote 11
What new plays might a dedicated London playgoer have seen in 1599? To be conservative, we compile a set containing the eleven extant plays which three standard bibliographical sources (i.e., DEEP: Database of Early English Playbooks, the Annals of English Drama, and the Wiggins Catalogue) agree belong to 1599Footnote 12 – Dekker's 1 Old Fortunatus and The Shoemaker's Holiday; both parts of Heywood's Edward the Fourth; Jonson's Every Man Out of His Humour; Marston's Antonio and Mellida; Shakespeare's Henry the Fifth and Julius Caesar; the collaboratively authored 1 Sir John Oldcastle; and two anonymous plays, A Larum for London and Look About You.Footnote 13 Beyond this, there are thirty professional plays for which we have titles but no more, and there may have been others of which all trace has been lost.Footnote 14
This one-year set presents us with a manageable sample of closely related larger characters to put in a pool for comparison. As before, we chose the larger characters to focus on – namely, those with 2,000 words or more of dialogue – so that we avoid the idiosyncrasies of smaller parts. Figure 3.8 shows how long each play is in terms of dialogue, as well as the number of characters with larger speaking parts, and the combined total words of these parts.
Figure 3.8 Total words spoken, number of and combined total words spoken by characters with ≥2,000 words of dialogue in 11 plays performed in 1599 by professional companies.
Every Man Out of His Humour is by far the longest play, with 37,000 words and seven larger characters who together constitute 70 per cent of the dialogue of the play. A Larum for London is the shortest, with 12,000 words. Antonio and Mellida has just one character with 2,000 words or more of dialogue – Antonio – and his part constitutes about 20 per cent of the play's dialogue. Thus the dialogue of these plays can be contributed mainly by large characters, or mainly by multiple small characters, and so on. For the purposes of this chapter, we leave this potentially fruitful topic of ‘concentrated’ or ‘dispersed’ dialogue and focus instead on the similarities or otherwise of the spoken styles of these thirty-five characters from 1599.
As before, we take the 100 most common function words as the basis for comparisons. These words are the skeleton of language, and a remarkably good guide to styles of discourse, narrative, persuasion, banter, direct or indirect address – intimate, distant, eloquent, blunt, and so on. Used together as a profile, they provide a flexible and revealing index of style. Instances of these 100 words make up almost half the words spoken in the plays, and they appear regularly in any play regardless of topic or genre. They form a framework for comparison of all or any segments of dialogue.Footnote 15
We have as our starting point the dialogue for 35 characters and their counts for the 100 most common function words. We can go back to the complete table of distances already created – that is, the data behind Figure 3.2 – and extract those scores for these characters. The lowest distance, and thus the closest pair, is Brutus and Cassius from Julius Caesar. Their proximity score is 3.18, very low in a general sense. A closer look at the z-scores for the twenty most common function words in their dialogue (Figure 3.9) reveals just how closely Brutus and Cassius match in profile.
Figure 3.9 The z-scores for the 20 most common function words for Brutus and Cassius in Shakespeare's Julius Caesar, using averages and standard deviations from the set of 666 characters with ≥2,000 words of dialogue.
None of the z-scores for these words vary greatly from the mean for the whole set of 666 characters, but I, a, my, and toinfinitive are low for both characters. For example, consider Brutus's and Cassius's speeches immediately following the assassination of Caesar:
brutus. Fates, we will know your pleasures.
cassius. Stoop, then, and wash. […] How many ages hence
brutus. How many times shall Caesar bleed in sport,
cassius. So oft as that shall be,
decius. What, shall we forth?
cassius. Ay, every man away.
Brutus and Cassius sound very similar, and for good reason. At this point, they are fellow conspirators striving to create a common interpretation of this historical moment, perfectly in tune. They speak on behalf of the group – preferring we and us to the singular I and my – and develop each other's points. It would be difficult to ascribe a speech to one or other of them if presented without a speaker prefix. Earlier in the play, a convergence of the two characters is crystallised in an image when Cassius offers himself as a mirror for Brutus, so the latter can discover his own merits, and perhaps his destiny: ‘I, your glass, | Will modestly discover to yourself | That of yourself which you yet know not of’ (1.2.70–72).Footnote 16
However, this is not always the case: overall, Brutus is distinctly more philosophical, and Cassius more pragmatic, and they become estranged as the action continues, leading to the quarrel scene, the ‘half-sword parley’ which was singled out already in the seventeenth century as a touchstone of Shakespeare's power over audiences.Footnote 17 Nevertheless, the function-word data suggests that they share a style to a remarkable degree and might have seemed more distinctive as part of a play-wide, Julius Caesar style than contrasting with each other.
Cluster analysis is a simple way of combining the various distances between characters in a single chart. It works by joining the two closest items, then treating the combination as a single entity, and joining the next two closest individual items or combinations, and so on until finally the two most remote items or combinations are added. Figure 3.10 is a cluster analysis of the 35 characters speaking more than 2,000 words of dialogue in the 11 plays from 1599. This is a local ecosystem, on a scale that allows closer examination of each character and the assurance that a single spectator might have heard all these parts performed in the latter part of the 1598–9 season and the first part of the 1599–1600 season.
Figure 3.10 Hierarchical cluster analysis of 35 character parts, using Ward's Linkage and Squared Euclidean Distance, based on counts of the 100 most common function words.
The first sub-groups formed have vertical bars closest to the left-hand border. Brutus, Cassius, and Mark Antony – the three characters from Julius Caesar large enough to be included – join first. The three Julius Caesar characters are judged to be more like each other than like any other character in the set. There are seven other plays with more than one character large enough to qualify, but in each case characters fall into two different groups in the cluster or into one group and then a wider collection of ungrouped characters. Gloucester and Prince John from Look About You, a knockabout romance of multifarious disguises set in the reign of Henry II, are at a similar level of closeness to the Julius Caesar characters, but Sir Richard Falconbridge from the same play is more remote, and the Old King (that is, Henry II) is in another part of the cluster altogether.
The Julius Caesar characters are remarkable for their consistency. If we can extrapolate from this to an audience reaction, this play might have seemed different in texture from the others, with main characters similar rather than markedly different in style. In his discussion of the dramatic output of 1599, James Shapiro argues that in crafting Julius Caesar Shakespeare was working his way to a new kind of dramaturgic texture, a departure from the (by then) well-established patterns of the history play and the romantic comedy, towards a more ‘symphonic’ form.Footnote 18 The cluster analysis supports this idea, at least to the extent that the play's characters are revealed as unusually close to each other in linguistic profile when compared with the rest of the 1599 sample. Recalling the 1599 plays the next year, our thoughtful playgoer might think in terms of an innovative play with a consistent texture, a new kind of Roman play, historical and philosophical at the same time, with some characters with a common romanitas, a collective and public-minded outlook.
Every Man Out of His Humour contributes seven characters to the collection. Macilente joins the Julius Caesar group and two kings, Edward from 2 Edward the Fourth and King Harry from Sir John Oldcastle; Puntarvolo and Fastidius Briske form their own group; Cordatus and Carlo Buffone form another, closest to each other, but still not very close; and Sogliardo and Fungoso still another. In this perspective, Every Man Out characters look diverse, in contrast to the unity of their counterparts in Julius Caesar. The character which is last to join any other is Fluellen, whom we have already discussed as an extreme odd-man-out in the larger collection of 666 characters. It is hard to imagine that the two Shakespeare plays would have seemed as exceptional as they now appear to us – and it is worth noting that, according to the analysis, King Henry from Henry the Fifth is in a mixed grouping of characters in the middle of the cluster – but it is plausible that elements of the two plays might stand out.
This is a view of new plays from the professional drama of 1599 in terms of character type, from serious all-round main protagonists who form the large group at the top of Figure 3.8, to highly specialised humours or allegorical characters. It is not really a view by ‘play’, since the consistency of Julius Caesar is a remarkable exception rather than the rule. It is not really a view by ‘author’ or by ‘genre’ either: sub-clusters of characters come from plays by four different authors (Heywood, Jonson, Marston, and Shakespeare) and four different genres (Roman tragedy, humours comedy, romantic comedy, and English history). Divisions within genres, and within sequences, are apparent. The Edward of 1 Edward the Fourth joins with a more domestic-drama group, while his character from the sequel joins with a group of other rulers and statesmen.
The characters of professional plays from 1599 might be roughly divided into the centripetal, which find matches in their own or other plays reasonably soon in the process, and the centrifugal – oddities and one-offs to various degrees, which take a while to be paired up, ‘characters’ in the full sense, like Simon Eyre from The Shoemaker's Holiday and Hobbs and the Mayor from 1 Edward the Fourth, unclubbable types like the quartet of Every Man Out characters and the goddess Fortune from 1 Old Fortunatus – and the greatest oddity of all, Fluellen.
Neighbours on Principal Components
The proximities we have been examining take the chosen variables (in our case, the 100 most common function words) and treat them all equally, offering a neutral, open framework in which to place the individual items – that is, the characters. For a third view of the relations between characters in early modern drama, we return to Principal Components Analysis (PCA) and isolate the most important underlying vectors of the characters’ language. This is a particular, weighted, selective view of the data, emphasising two factors in particular rather than taking the word counts as they come. We continue to use the 100 most common function words as our variables, but modify our sample to contain only those characters with speaking parts of 2,000 words or more from plays performed between 1580 and 1619. This results in a reduced sample of 531 characters from 197 plays. Although the larger 1580–1642 set (with 666 characters parts across 243 plays) provides a more comprehensive overview and the 1599 set (with 35 character parts across 11 plays) a tighter focus, this reduced four-decade sample presents something of a ‘middle ground’ while also keeping the analysis within the range of a notional single writing career.Footnote 19
The most important of the components which emerges from the analysis is a contrast between stately kings and choric figures on the one hand and fussy busy-bodies on the other (PC1). The first group makes considered, authoritative, finished pronouncements, while the second acts on others, reacts to others, and obsesses about themselves. In parts of speech, this is a contrast between prepositions, which are used in locutions which render the detail of a depicted world, and auxiliary verbs, which are the vehicle for close interaction and personal reflection. The extreme at one end is King David from Peele's David and Bethsabe. His speeches are well-turned, replete with stately, elaborate poetic illustration. Even when addressing a single interlocutor, the style is formal and spells out well-rounded connections:
The ‘Humorous Lieutenant’, from Fletcher's play of the same name, is at the opposite extreme. He is a valiant soldier, but also a hypochondriac. His ‘humour’ is an exaggerated preoccupation with his health and his immediate sensations. He is entirely caught up in the moment:
The second axis of differentiation uncovered by PCA (PC2) is a contrast between lovers – or, more generally, those caught up in an action and focused on personal relations (domestic or intimate) – and characters whose role is to dissect some aspect of the play-world. One style is dramatic, in the sense of here-and-now engagement between characters; the other is more descriptive, with characters creating conceptual worlds through discourse. It is a rival version of the contrast of the first principal component – a second refraction of an underlying division in the dialogue of plays – between speeches which assume the world as given and present, and those which create a world by description and narrative. This time, the important word-variables are thou, thee, and thy, and I, me, and my – more common in the dialogue of the lovers and intriguers – and the articles a, and, and the, some prepositions, and most and own – all used freely by the commentators and analysts.
Mistress Barnes, from Porter's 1 The Two Angry Women of Abingdon, has one of the lowest scores on the second principal component. She has a long quarrel with Mistress Goursey, and her spoken part is mainly an assertion of her antagonism, with few ventures into anything approaching analysis or commentary. Aside from Mistress Goursey, she also engages fiercely with her husband, her daughter, and her son, all of whom she addresses as thou at times. At the other extreme of this axis are Cardinal Como from Dekker's The Whore of Babylon, Ariosto from Webster's The Devil's Law-Case, Fluellen from Henry the Fifth, and Savoy from Chapman's Byron's Conspiracy. Each of these comment on situations to their peers, and so do not use thou, thee, or thy. They are pundits offering analysis of military arrangements, or of legal, political, and religious affairs, and so tend not to call attention to their own subjectivity through the use of I, me, or my. Their concrete and well-illustrated commentary leads to a high incidence of the articles and of the prepositions of and in. In the following example, this language profile has the effect of making it unclear whether Savoy is summarising the ambassador's aims or expressing his own:
The principal components are summaries of a great deal of information. They are principled abstractions from the use of 100 words in speeches of 531 characters from 197 plays. Characters from different plays, authors, genres, and eras are mapped together. Some of the groupings are instructive, offering a categorisation of characters which gives a fresh perspective on their place in their own plays and within the discourse of late sixteenth- and early seventeenth-century English drama sampled here. Figure 3.11 highlights some of the pairings that emerge.
Figure 3.11 PCA scatterplot of 531 character parts from 197 plays performed between 1580 and 1619, using the 100 most common function words.
Claudius from Hamlet and Tullius Aufidius from Coriolanus are placed close together and in the top left of Figure 3.11, best described as a combination of disquisitory and detached in the orientation of its characters. Claudius and Aufidius are each an antagonist to the formidable protagonist after whom their plays are named. Their fates are entangled with this powerful, dramatically charismatic other, a fact on which they both ruefully reflect. They both occupy worlds of great affairs – one political, the other military – with some formal set-pieces, kingly addresses for Claudius and the impassioned welcome to Coriolanus for Aufidius. Both focus their speeches on a single male other, whether the latter is present or absent.
We may think of Claudius as a villainous intriguer, but viewed in structural terms his dialogue places him more as someone who amplifies or explains, a commentator as opposed to someone intimately involved with others. If we read through his speeches we realise that he does not enter into any sort of quipping immediate interchanges, nor into direct participation in the action. He almost always maintains a rounded, deliberate, detached tone. Aufidius is a warrior, but his dialogue does not reflect direct action but a measured consideration of Coriolanus's impact on his city's destiny and his own.
Florimell in John Day's comedy Humour Out of Breath appears close to Beatrice from Much Ado About Nothing in the detached and interlocutory (and modern-sounding) area of Figure 3.11. Florimell is an outspoken young woman, not exactly in the way Beatrice is – Florimell goes beyond wittiness to something more akin to lewdness – but, like Beatrice, Florimell is an uncomfortable fit with the conventional gender roles of the time. Beatrice risks her marriageability should her challenge to Benedick tip her from ‘wit’ to ‘shrewishness’, and Florimell likewise if her bawdiness should define her as ‘unchaste’, though both end their respective plays with advantageous marriages.Footnote 23
Lorenzo is a garrulous, amorous old man in Chapman's comedy May Day, and in the PCA he is placed close to the Falstaff of 1 Henry the Fourth, in the lower-right, interlocutory-involved area of the graph. In pursuit of his misdirected amour, Lorenzo disguises himself as a chimney sweep. He might well be modelled on this Falstaff, or the Falstaff of The Merry Wives of Windsor. J. M. Robertson noted parallels between the language of Lorenzo and Falstaff in Merry Wives, which suggested to him that Chapman might have collaborated in the play.Footnote 24
Dido and Romeo, two lovers from early plays, Marlowe's Dido, Queen of Carthage and Shakespeare's Romeo and Juliet, are placed close together in the lower-left, or disquisitory-involved area of Figure 3.11. Dido's dialogue rarely strays from a focus on Aeneas whom she addresses as thou from the beginning, and on her own feelings and situation. Romeo gives Mercutio, the Nurse, and the Friar thou, as well as Juliet. His focus is on his situation, his feelings, and – of course – on Juliet. Romeo's ‘concern with self-perception’ and ‘autonomous, self-doubting subjectivity’ have led Lynette Hunter and Peter Lichtenfels to treat his character-part as ‘a first take on Hamlet’.Footnote 25 Unlike Hamlet, however, neither Dido nor Romeo is a notable speechmaker, or notable for staccato interchanges, and both are involved in the action rather than detached.
Queen Margaret from 3 Henry the Sixth, Queen Isabella from Marlowe's Edward the Second, and the Duke of Anjou (later Henry III of France) from the same playwright's Massacre at Paris are brought together in the same disquisitory-involved section of Figure 3.11 as Dido and Romeo, though further in the disquisitory direction and not as far in the involved direction. These characters are all from history plays performed in the early 1590s: two dealing with English history of the fourteenth and fifteenth centuries respectively, and the other treating events in France twenty years or so prior. Margaret, Isabella, and Anjou all rule countries and direct armies at various times, but they also have moments of impotent rage or lament at betrayals by spouses and allies or defeats by enemies. Although she is Queen, Margaret is repeatedly forced to plead with, berate, and exhort her hearers. Anjou (or rather Henry) later becomes King of France, but threats from the Guise faction and from his brother-in-law, the King of Navarre, render his rule insecure. Like Margaret, he is often reduced to mocking and then cursing his enemies and plotting revenge on them. He is in a contestatory role, even after his coronation.
All three of these characters are shrewd politicians, and yet they remain prey to emotion. For example, Mortimer tells Isabella she is ‘too passionate in speeches’ for a ‘warrior’ (4.4.14–15).Footnote 26 At times, these emotions beget physical violence: Margaret stabs York (1.4.177 sd.), Isabella facilitates the capture of Gaveston and wishes him slain (2.4.37–9, 68–9), and Anjou has the Guise murdered (Scene 21).Footnote 27 Charles R. Forker argues that Margaret's murderous and warlike actions may have ‘prompted’ aspects of Isabella.Footnote 28 The dialogue of this trio is entirely public, comprising well-fashioned, purposeful pronouncements, with no asides or domestic interludes. They have little time for witty analysis, courtly prevarication, or subtle distinctions.
Margaret, Isabella, and Anjou are traditional speech-makers, not fussy busy-bodies, with low scores on PC1. They are involved in intense personal relations, rather than reflective and analytical, and so score low on PC2. The long chain of arithmetic analysis, from spoken part to counts of particular words to principal components, seems to have uncovered a genuine likeness. In small fragments, Margaret and Anjou in particular are difficult to tell apart:
There are two instances of and in each of these short quotations. This is a heavily weighted word-variable to the low end of PC1. Both Margaret and Anjou use this word very freely: Margaret's z-score for this word-variable is 1.74; Anjou's is 1.70.
Late in 3 Henry the Sixth, Margaret is left to rally her son and their supporters, who are dismayed by their defeat at Barnet and need to gather their strength before returning to the fray at Tewkesbury. She begins: ‘Great lords, wise men ne'er sit and wail their loss, | But cheerly seek how to redress their harms’ (5.4.1–2). The Duke of Anjou urges unrelenting pursuit of the Huguenots with the same recourse to popular wisdom: ‘Yet will the wisest note their proper griefs, | And rather seek to scourge their enemies | Than be themselves base subjects to the whip’ (4.14–16). Margaret and Anjou are ‘sisters under the skin’ – created by different playwrights, placed in different geographic settings and in different fictional centuries, one a queen and one a king, but with an underlying common orientation to the dramatic worlds in which they are situated.
Conclusion
As this chapter has shown, statistical analysis of language provides a new way to model characters’ interactions in early modern drama. It winnows the myriad connections and contrasts between characters’ language styles and highlights particular examples, which are then a challenge to interpretation. The conjunctions and disparities can be traced back to individual instances of words in phrases, sentences, and speeches, though accumulated into totals. It is possible to define exactly how the numbers which go into the analysis came about. Given this text, and this way of counting, the number has to be exactly so. Familiar canonical texts and forgotten plays by unknown or obscure authors are treated alike, with an effect which from one perspective is a distortion – Hamlet is just a character like any other and may be lost from sight altogether in the middle of a cloud of data-points – and from another point of view is a way of glimpsing a different terrain, wider and flatter than the one we are used to. In this analysis, the motivation of characters and the twists and turns of plot are obscured, while discourse types and patterns of dramatic interaction are brought to the fore.
When given the chance to escape from their plays, characters do often group by author, in the obvious way by sequel, but also by category (e.g. protagonist, antagonist) or character type. Pairings across authorial canons also appear, and take us back to traditional stock character types (e.g. garrulous old men, wily schemers). Pairings within plays are unexpected, but their occurrence reminds us that a generic play style may prevail over local differentiation when seen in wider frames.
The representative, balanced, internally various character is an unfamiliar concept which the analysis offers, along with its complement – the specialised, hypertrophied character, an eccentric bit part magnified. This is what the characters of early modern English drama would look like to Martians who had no earthly language and could only interpret the plays through patterns of word use. Their response, however primitive and schematic, would have its own logic and might prompt an interesting conversation with those who can read the plays, explore the historical record, and see them performed, and thus have a different understanding of these precious traces of a crowded, noisy, competitive, corporate activity now more than four centuries in the past.
