Word links from Troilus to Othello and Macbeth

The Bard, 2/1, 1978, pp. 4-22

The predominant view of Shakespearean scholars is to fall in line with the suggestion of Sir Edmund Chambers that Troilus and Cressida was written in or about the year 1602. For instance, in her Introduction to the Cambridge University Press edition [1], Alice Walker has no difficulty in believing that the play Troilus and Cressida, appearing in a Stationers' Register entry of 7 February 1603, was the play that Shakespeare wrote, "for it is not very likely that the Chamberlain's men would have had two Troilus and Cressida plays at the same time". She hinks, moreover, that the play was all written at one time, and not over any period of five or ten years, and, indeed, that there was no authoritative revision of the text of the play after its first writing. These views are fairly typical.

     It is, then, very interesting to find Miss Hotine [2] arguing for a much later date, for preference about 1608. This is on historical and psychological grounds, and particularly because the play seems to her a peace‑time rather than a war­time play, and so most unlikely to antedate 1604. In view of this clash of opinion, it seemed worth while to look for extraneous sources of evidence, i.e. in the vocabulary of Troilus and Cressida in relationship to the vocabulary of other plays of the canon. In the Chambers chronology Troilus and Cressida is placed after Hamlet and before All's Well That Ends Well. We should, then, find the rare‑word vocabulary showing relationships to these two plays above all others on the Chambers dating, but with later plays if Miss Hotine's argument is sound.

      The basis of the analysis that follows is a card index prepared by the present writer [3] from Bartlett's Concordance, checked, corrected and amplified by comparison with the Harvard Concordance of Marvin Spevack. Textual disagree­ments between the two were checked against the Cambridge editions of the plays, which were taken as authoritative. The card index includes only and all words found in two or more plays but with not more then ten citations, i.e. the rarest but not unique words. Classification followed the general principles of the O.E.D., but particularly distinguishing as different word homonyms functioning as different parts of speech. The distribution of the citations of card‑index words throughout the canonical works is shown in Table 1. Link words were divided into two groups, (a) of the very rarest appearing altogether not more than six times, and (b) commoner though still unusual words appearing seven to ten times. Columns 3 and 7 of the Table show the numbers of links of these two groups of words with other plays, against which are set in columns 4 and 8 the numbers which would be expected on the hypothesis of purely random association. On that hypothesis every play should have approximately that number of associations with Troilus which would be proportionate to its own particular share of the total indexed treasury. It will then be found that some plays are in deficiency in the number of links and others have an excess. Whether an excess is statistically significant or not can be calculated by simple arithmetic combined with the use of standard statistical tables.


    How the sums were worked out can be shown by an example. There were 15266 citations of words appearing two to six times in the index, or 14638 omitting the appearances in Troilus. Macbeth's share was 388 citations, or 0.026506 of the whole. This is the fraction of the 1485 citations provided by Troilus which, on the basis of random chance, one would expect to find in Macbeth, i.e. 39.3619 appearances. Instead 60 were actually found, i.e. in excess to 152%. The excess of observation over expectation is statistically significant at the 0.0025 level. In the same way, the seven‑ to ten‑word citations in the card index totalled 7960 with Troilus citations subtracted, Macbeth's share being 179, or a fraction of 0.022487. This fraction of the total of 1540 links provided by Troilus to other plays is 34.6300, whereas the observed number is 40, an excess of 116%, but this time not statistically significant. All the numbers of links found to excess in the various plays are shown in columns 5 and 9 of Table 1, with the degree of statistical significance attaching to them in columns 7 and 10. It is the plays that appear both in column 5 and in column 9 which particularly engage our attention. The observed and the expected numbers relating to them, of two‑ to six‑ and seven‑ to ten‑ words, can be summed; and we then arrive at the figures of Table 2.


We can see from this table that predominant linkage of Troilus is with what are commonly thought to be later plays, and especially Othello and Macbeth. The association of Troilus with Othello is, perhaps, not very surprising. They are in any case thought to have been written about the same time, and in each the theme of a woman's frailty or infidelity is developed at length, much more so in Othello than in Troilus.

     To explore the literary and possibly chronological, and not merely the statistical significance of these observations, it is worth putting them on record. In what follows is shown first the act, scene and line reference of the link word, as shown in Spevack's Harvard Concordance, then the link word itself, as it appears in Troilus, then its equivalent appearances, first in Othello, then in Macbeth, likewise with Spevack's line references. For quotation the modernised spelling of the Cambridge University paperback editions of the plays has been used. Troilus was edited by Alice Walker, Othello by Alice Walker and John Dover Wilson, and Macbeth by John Dover Wilson.

This collection of material provokes a number of speculations. Although the words themselves are thematic in nature, and not mere 'functional' or 'filler' words, they seem to be used by Shakespeare as independent entities, i.e. in contexts widely differing in emotional overtones and dramatic impact. It is only occasionally that they lead to parallel passages. 1.2.74 BAREFOOT is used in very much the same way by Pandarus and by Emilia, and may well represent a contemporary proverbial usage. 1.3.153 STRUTTING shows both Ulysses and Macbeth expressing the same withering contempt of their creator for the stage actor. 3.2.189 DUSTY provides two comparable reflections on the triumph of time, many times more poetically powerful coming from Macbeth than from Cressida. And with 5.3.84 DOLOURS Cassandra forehears and Macduff records the devil's symphony of human agony brought on in the ruin of a state.

     Also noteworthy are three remarkable double links where two link words closely placed in one play are found again in another, still in close approximation. First of these are 1.2.238 DIRT and 1.2.241 DOLTS in a trivial context in Troilus, brought together again at 5.2.164 and 5.2.163 in Othello at a moment of high drama and tension. 4.4.88 PROMPT and 4.4.145 ALACRITY neighbour one another in Troilus, even though separated by 57 lines; they come together at 1.3.232 in Othello in the same line. In Troilus 5.2.24 JUGGLING and 5.2.48 PALTER stigmatize a mood of sourness and contempt; but coming together in Macbeth 5.8.19 and 5.8.20 they are notes in a cry of despair.

      In their contexts these link words are, in the main, laden with affect; but they are vessels into which a variety of affects can be poured, and can spill out in any direction. They tend to be connected with particular characters, especially with char­acters who have the gift of eloquence, e.g., as the following table 3 shows, with Ulysses in Troilus and with Othello in Othello. Nevertheless the poet Macbeth gets only his expected shire of these words and lago rather less than his. Conspicuously deficient is Pandarus, who has many lines but few link words. Cressida has no more, but is dowered with fewer lines. The mean‑spirited do rather poorly in this respect. The deviations from expectation are statistically significant with all three of Ulysses, Pandarus and Othello.  

     Just as these words are not evenly distributed between characters, so they are not evenly distributed over the whole of a play. In Troilus there are 38 occurrences in 1.3, i.e. 164% of the 23.2 to be expected from its number of lines, a highly significant deviation statistically. 11.2 is also rich in these words, 150% of expectation, and 11.3 as well with 148%. IV.5 is a long scene of 293 lines, and quite a dramatic one, but has only 35% of expectation (6 instead of more than 17), a statistically significant defic­iency, though the reason for it is not clear.

     In all three plays these link words show a tendency to come up in runs and clusters. If one appears, there is more than a normal chance that it will be shortly followed by another. If one classifies the intervals between one link work and the next, as is shown in Table 4, the short intervals preponderate. However, if there is a break in the flow of the writing, as when there is a change of scene, the interval extends again.


    A lot of attention has been paid to Shakespeare's imagery, which also occurs in clusters; and a number of non‑logical mterative image clusters have been discovered. They seem at times to have had a compulsive effect on Shakespeare's flow of thought, even inappropriately. The rare words we have been considering frequently have the quality of an image. Nouns tend to be concrete rather than abstract; verbs are often verbs of action and movement; adjectives frequently imply one or other of the senses. The connection between one link word and the next is often straightforward, so that one would not be surprised on coming across the pair in a word‑association test: 1.3.41,42 liquid, moist; 1.3.171, 173 alarm, cough; 2.2.33, 35 shroud, devout; 2.3.164, 166 dispose, peculiar; 3.2.79, 80 imposition, difficulty; 4.1.70, 72 bawdy, contam­inated; 5.1.61 fitchew, lizard; 5.3.31 doff, harness; 5.4.8 luxurious, drab; and some others. In an almost equal number of cases the combination of the two or more words strikes one as odd, peculiar in an imaginative way: 1.3.125 chaos, suffocate; 1.3.194, 195 dirt, weaken, exposure; 1.3.206, 207 batters, poise (n); 2.3.206, 208 batters, insolent; 5.2.191, 193, 194, croak, parrot, drab. In yet another third of all these clusters the connection between the words is occult, legitimate in the context, but showing a wide leap in thought: 1.3.339 imputation, poised; 2.2.63, 66 enkindled, traded, dis­tastes; 2.2.173, 175 decision, humanity; 2.3.118, 120 virtuously, unwholesome; 2.3.128, 129, 130 underwrite, predominance, ebbs; 4.2.12 wights, grasps; 4.4.48, 50 distasted, genius; 4.4.86, 88 sweeten, prompt; 5.2.47, 48 Guardian, foh, palter; 5.8.18 separates, frankly.


One meets the same kinds of connectedness between link words in Othello and in Macbeth. For the record: Othello 1.3.136, 137 imminent, insolent; 1.3.232 alacrity, prompt; 1.3.273, 275 adversity, privately; 1.3.327, 329 poise, preposterous; 1.3.355, 356 sanctimony, barbarian; 2.1.231, 236, 239 conveniencies, position, humane; 3.3.79 82 peculiar, poise; 3.3.90, 92 perdition, chaos; 3.3.232, 234 foh, position; 3.3.276, 280 forked, generous; 3.4.36, 39, 41 moist, devout; 3.4.72, 74 prophetic, skilful; 4.1.28, 29 convinced, blab; 4.2.29 cough, hem; 4.2.78, 81 bawdy, impudent; 5.2.163, 164 dolt, dirt. Macbeth 1.3.17, 18, 20 shipman's, drain, lid; 2.1.52, 53 off'rings, sentinel; 2.3.60, 61, 62 livelong, feverous, parallel; 4.3.28, 32, 36, 40, 42 leave‑taking, basis, grasp, gash, uplifted; 5.2.4, 5 alarm, excite; 5.2.21, 23 giant's, pestered; 5.5.23, 25, 27 dusty, struts, idiot; 5.8.19, 20, 24, 26 juggling, palter, gaze, underwrit.

     The links running from the rare‑word vocabulary of Troilus to Othello and Macbeth are numerous and striking. What is more important, they are in great excess over what might be expected on a basis of chance coincidence. Such a level of significance, or "confidence", absolutely requires some rational explanation and cannot just be dismissed. A reasonable and satisfying hypothesis would be that all three plays are written near to one another in time. What that time might be, in calendar terms, is left open. In the Chambers chronciogy, Othello was written in 1604 and Macbeth in 1605. Though the vocabulary findings certainly suggest a placing of Troilus later than Hamlet and before All's Well, they do not carry us far enough to combine happily with Coriolanus, Timon and Pericles, the plays attributed by Chambers to the vicinity of 1608.

     So far we are taken by the material reported above. However, what began as an investigation ad rem has expanded into the rudiments of a study of Shakespeare's use of the unusual and rare elements in his vocabulary. The link words we have discussed are only a sample from a much larger storehouse. Even so they warrant some provis­ional ideas about the processes of word‑association in Shakespeare's thought. These words are ones particularly pregnant of meaning. They carry a more than normal load of emotional tension and dramatic impact. Their distribution from line to line and scene to scene during the course of a dramatic argument is shown not to be random but to follow a law‑like mode.

     The great statistical thinker, Udny Yule (4), in his pioneer studies of literary vocabulary, found himself led to a remarkable conclusion. Studying a writer's works, one after another, counting the number of different words used so far, one reached no sign that his vocabulary was nearing exhaustion. Add another work, and another crop of new words could be added to the total. With a penetration that anticipated the advances of the psychologists, Yule concluded that for writers generally, and indeed for each of us, a man's treasury of words includes every word that he has ever read with understanding. The word, once read, enters the storehouse to remain there, however unused, however deeply buried and overlaid. As time passes with it still unused, a stimulus of ever greater energy or specificity is needed to drag it up from the depths. Shakespeare is credited with having used some seventeen thousand words. Some, indeed, he invented for himself, but out of elements he already had available. All the others must have come to him from reading; and if his reading provided him, not only with the 17,550 independent words counted by Alfred Hart (5), but, presumably, several times that number, his reading must have been vast indeed.


[1] The Works of Shakespeare: Troilus and Cressida, ed. Alice Walker (Cambridge 1969) xxxviii. Miss Walker, however, also notes (x) that "many [critics] have thought it was begun about the time of The Merchant of Venice and completed c. 1602 or even later (c. 1606‑8)".

[2] Margaret Hotine, 'Troylus and Cressida: historical arguments for a 1608 date', The Bard 1(1977)153‑61.

[3] Previous work by the present writer based on Shakespeare's vocabulary of rare words includes four communications to Notes and Queries: 'A statistical note on A Lover's Complaint', NQ 218 (1973) 138‑40; 'Shakespeare: word links between poems and plays', NQ 220 (1975) 157‑63; 'Word links with The Merry Wives of Windsor', NQ 220 (1975) 169‑71; 'Word links with All's Well that Ends Well', NQ 222 (1977) 109‑12.

[4] G. Udny Yule, The Statistical Study of Literary Vocabulary (Cambridge 1944).

[5] Alfred Hart, 'Vocabularies of Shakespeare's plays', Review of English Studies 19 (1943) 128‑40; 'The growth of Shakespeare's vocabulary', ibid. 19 (1943) 242‑54.