Introduction

§ 1 The York Cycle of Mystery Plays is a set of 50 extant Medieval plays (or pageants) that retell the Biblical history of the world, from Creation to the Last Judgement. These plays were performed annually during the Corpus Christi holiday in the city of York starting sometime after the year 1376 and until the late 1560s (Beadle and King 1984). Each pageant was assigned to one of the craft-guilds of the city, which was in charge of its performance, including selecting and training actors and building a wagon-stage and costumes. A manuscript containing the text of 47 of the pageants and written between 1463 and 1477 for the official York city records survives to this day. Additionally, an alternate version of the text of the Creation of Adam and Eve pageant and a fragment of an alternate Coronation of the Virgin pageant both survive in this manuscript, and an alternate Incredulity of Thomas pageant survives in the Sykes manuscript (York City Archive number ACC.104/G.1; Beadle 2009).

§ 2 Today, almost all aspects of the York Plays are subjects of academic study and debate (Beadle and King 1984). For example, knowledge of the costumes worn by the actors and the props used on-stage is limited to a few surviving enumerations of pieces, and their exact appearances are unknown. The process and requirements for selecting the actors is described only in cryptic terms. Stage directions are rare in the manuscript, leaving a lot of questions unanswered about the physical performance of the Plays. The manuscript also includes cues for songs, but the extent to which the actors had to sing and the use or lack of musical instruments is unknown. The nature and appearance of the pageant wagons themselves are also an area of debate, as only scarce and ambiguous accounts of their construction survives (Beadle and King 1984).

§ 3 However, it is on the language used in the York Cycle that this paper will focus, and specifically on its use of alliteration. Alliteration is an audible repetition of a sound at the beginning of multiple words of a line. For example, it is clearly illustrated in these lines, from the beginning of York play 26:

Vndir the ryallest roye of rente and renowne
Now am I regent of rewle this region in reste,
Obeye vnto bidding bud busshoppis me bowne
And bolde men that in batayll makis brestis to breste.
To me betaught is the tent this towre-begon towne,
For traytoures tyte will I taynte the trewthe for to triste.
The dubbyng of my dingnité may noyot be done downe,
Nowdir with duke nor dugeperes, my dedis are so dreste.

§ 4 The first two lines alliterate on the sound R, then the next two on the sound B, followed by two lines of T and two lines of D. Unsurprisingly, like most other aspects of the York Plays, the meaning and purpose behind the use of alliteration by some characters is a point of debate, and this debate will be the topic of Section 2. Section 3 will then present our algorithm for measuring alliteration and Section 4 will present the results that we obtain with this algorithm on the York Plays. Section 5 will discuss the meaning of these results and the possible significance of alliteration that can be inferred from them, before ending with some concluding remarks in Section 6.

Related research

§ 5 The significance of the use of alliteration in the York Cycle has long been a subject of debate. For some readers, alliteration, along with meter and verse, can be used to differentiate between two or three distinct authors or composition stages of the Plays. This idea has been studied and debated among Medievalists since it was first proposed over a century ago, when Hohlfeld pointed out meter irregularities in five of the plays that he suggested represented an earlier stage of composition (Holfled 1889). Davidson added three more plays to this irregular list (Davidson 1892), and Gayley distinguished an entire third category to Hohlfeld’s two (Gayley 1907). Greg then took the idea in a new direction (Greg 1914); he pointed out a fundamental mistake in the metrical assumptions of Hohlfeld, Davidson, and Gaylay, and created instead three new divisions of the plays of the Cycle based on his theory of meter. Chambers (1949) continued the work of Greg and reclassified three plays that Greg had misclassified. This branch of research finally led to the work of Jesse Byers Reese (1951), who introduces observations about alliteration in the classification scheme. To be sure, Hohlfeld’s original work had noted that some of the plays used alliteration, but he made no case of it when studying the plays; Reese points out that, in fact, the lines with irregular meter on which Hohlfeld based his theory are simply lines written as alliterative verses. Likewise, Reese points out that plays that Greg considered completely different and separated into distinct categories also used very similar alliterative verse forms. Alliterative verses, Reese shows, transcend the divisions of authors and composition periods neatly separated in the five landmark studies mentioned previously. The result is a very strong argument that a study of alliterative verse is needed to fully understand the composition of the York Cycle. Reese conducts such a study, and concludes that there are thirteen plays written in purely alliterative verse (Reese 1951, 649), namely 1, 16, 26, 28, 29, 30, 31, 32, 33, 36, 40, 44, and 45, as well as alliterating portions in plays 34, 17, 25, and 41 (for play titles see Table 1). Reese does not go as far as assigning these 13 plays to a separate author or stage of development in the Cycle, but he does set them apart from the other 34 plays of the Cycle as belonging to a different school of dramatic writing.

§ 6 An alternate interpretation of the use of alliteration in the Plays is that it is used to mark the roles of different characters in the Plays. Smith first suggested this when she noted that good characters such as God and Jesus spoke in verse while evil characters like Herod, Pilate and Caiaphas used alliteration heavily (Smith 1963). This noted role differentiation lines up with the Augustinian influences that are believed to have shaped the Plays (Johnston 1993). Indeed, Augustine associates different attributes to good and evil, with good represented by stasis and evil represented by chaos (De doctrina christiana I.8 [Martin 1962, 11]); similarly, it has been argued that alliteration gives more energy and movement, and therefore chaos and evil, to the speeches of the characters (Hayes 2013). However, not everyone agrees; even before the Augustinian connection was made, Epp (1989, 153) noted observations contradictory to those of Smith, pointing out for instance that Jesus and God do use alliteration in some of their speeches. Epp advocates instead that alliteration is used for a variety of reasons, including to represent a passionate speech, to signal important moments in the plays, or to elevate certain characters. He proceeds to give examples of multiple characters, good and evil, who used alliteration at various moments in the Cycle.

§ 7 In this paper, we propose a study of alliteration that will differ sharply from the others done so far over the past century: a purely statistical study. To this end, we developed an algorithm to read and count the number of alliterating lines of each character in the Cycle, which we present in the next section.

Alliteration statistical algorithm

The York Cycle

§ 8 For this research[1], we used the text of the York Cycle based on the aforementioned surviving manuscript from the York city records, edited in 1982 by Richard Beadle and publicly available from the University of Michigan Library. This version of the Cycle was chosen over the updated 2009 and 2013 editions simply because it is the one available in electronic format, which is a requirement for our work. The corpus obtained from that site has 47 plays, with the statistics listed in Table 1. In total, the entire Cycle has 381 characters and 13,307 spoken lines of dialogue.

Table 1: Details of the York Plays corpus used in this research.

Play Title Lines of dialogue Number of characters
1 The Fall of the Angels 163 6
2 The Creation 172 1
3 The Creation of Adam and Eve 96 3
4 Adam and Eve in Eden 99 3
5 The Fall of Man 180 5
6 The Expulsion 168 3
7 Cain and Abel 139 4
8 The Building of the Ark 151 2
9 The Flood 325 8
10 Abraham and Isaac 383 5
11 Moses and the Pharaoh 410 10
12 The Annunciation and Visitation 253 4
13 Joseph’s Trouble about Mary 308 5
14 The Nativity 154 2
15 The Shepherds 134 3
16 Herod and the Magi 394 13
17 The Purification 460 6
18 The Flight into Egypt 231 3
19 The Slaughter of the Innocents 281 8
20 Christ with the Doctors in the Temple 288 9
21 The Baptism 175 4
22 The Temptation 210 4
23 The Transfiguration 240 8
24 The Woman Taken in Adultery / The Raising of Lazarus 209 12
25 The Entry into Jerusalem 544 16
26 The Conspiracy 313 10
27 The Last Supper 187 8
28 The Agony in the Garden and the Betrayal 308 17
29 Christ before Annas and Caiaphas 405 11
30 Christ before Pilate I: The Dream of Pilate’s Wife 560 13
31 Christ before Herod 437 9
32 The Remorse of Judas 391 8
33 Christ before Pilate II: The Judgement 487 12
34 The Road to Calvary 350 10
35 The Crucifixion 311 5
36 The Death of Christ 416 15
37 The Harrowing of Hell 408 15
38 The Resurrection 456 12
39 Christ’s Appearance to Mary Magdalen 149 2
40 The Supper at Emmaus 198 3
41 The Incredulity of Thomas 199 5
42 The Ascension 276 9
43 Pentecost 228 11
44 The Death of the Virgin 194 15
45 The Assumption of the Virgin 328 22
46 The Coronation of the Virgin 160 8
47 The Last Judgement 379 14

Phonetic transcription software

§ 9 In order to pick out alliterations in the text, it is necessary for us to develop software that can make out the phonetic reading of the text based on the spelling of the words. We started by defining 30 different sounds that occur in English words. We have listed these sounds, with illustrative examples taken from the York Cycle, in Table 2. Then, the algorithm we developed works in two steps. In the first or processing step, a word is converted into a phonetic string. To do this, the word is read letter by letter, and each letter is replaced by a sound from Table 2 based on its context (i.e. the surrounding letters in the word) using one of over 110 pronunciation rules we defined. The processing rules defining the pronunciation of the letter A are given as an example in Table 3. Note that the rules are considered in order, so that in the case where several rules could be applied only the first applicable one is used. In the second or post-processing step, the phonetic strings are simplified by removing all silent sounds and duplicate sounds.

Table 2: List of sounds recognized by our software, with examples from the York Cycle.

Sound Symbol York Cycle Example (phonetic equivalent in parenthesis)
Silent letter 0 waste (wast0), knawyn (0nawin)
Note: this symbol is eliminated in post-processing
Short A a parformed (parfOrmd), askid (askid), saande (sand)
Long A A mase (mAz), begane (bEgAn)
Regular B b beam (bEm), bee (bE)
Soft C, Soft S s sygne (sign), cytte (sit)
Hard C k clapped (klapd), cloke (klOk), waxe (wAks)
Regular SH C fische (fiC), shames (CAmz)
Regular D d dore (dOr), brother (brodr), þerfore (derfOr)
Short E e leffis (lefiz), themselfe (demself)
Long E E beete (bEt), clean (klEn)
Regular F f fitte (fit), offerrand (Oferand)
Hard G g galilee (galilE), goyng (going)
Soft G, Regular J j joly (joli), jornay (jOrnA)
Audible H h hurte (hurt), hatred (hatrd)
Short I i lymbo (limbO), pilatus (pilatuz)
Long I I wide (wId), wyde (wId)
Regular L l devil (dEvil), fully (fuli)
Regular M m comyn (komin), somme (som)
Regular N n sonne (son), slang (slang)
Short O o renowne (rEnoWn), companye (kompanI)
Long O O own (OWn), tow (tO)
Regular P p spoken (spoken), appered (apErd)
Regular R r broder (brodr), carre (kar)
Regular T t spotte (spot), wittes (witz)
Short U u creaturis (krEturiz), fun (fun)
Long U U poors (pUrz), tribute (tribUt), 3one (yUn)
Regular V v doves (dovz), verilye (vErilI)
Regular W w whallis (waliz), saw (saw), twelffe (twelf)
Regular OW W grounde (grWnd), foundynge (fWnding)
Regular YE y youreselfe (yWrEzelf), yett (yet), 3is (yoiz)
Regular Z, Hard S z dragons (dragonz), lazare (lazAr)

Table 3: Rules for pronouncing the letter A.

Rule York Cycle Example (phonetic equivalent in parenthesis)

if 'a' is followed by 'i' or 'y':

sound ← A

playe (plA)

if 'a' is preceded by 'o':

sound ← O

yoare (yOr)

if 'a' is preceded by 'e' and followed by 'u':

sound ← O

jeauntis (jOntiz)

if 'a' is preceded by 'e':

sound ← E

creaturis (krEturiz)

if 'a' is followed by a consonant followed by 'e':

sound ← A

begane (bEgAn)

if 'a' is followed by 'u':

sound ← O

braunchis (brOnCiz)

if 'a' is followed by a final 'r' or 'l', optionally followed by 's':

sound ← 0

werkar (werkr), kepars (kEprz)

otherwise:

sound ← a

waste (wast)

Alliteration detection

§ 10 Once the words are rendered as phonetic strings, the algorithm can detect alliterating lines automatically. This actually creates a new problem: while hearing alliteration is easy for us humans, a formal algorithmic definition of the feature has not, to our knowledge, been written before. Moreover, some English linguists define it not simply as a repeating sound but also take into account stressed and unstressed syllables as well as meter and rhythm of the lines (Matonis 1984). However, for our algorithm, we opted to use a simpler phonetic definition of alliteration. In our software, an alliterating line is any line that satisfies one of these two sets of conditions:

§ 11 Alliterating conditions 1:

  • The most common initial sound of words is used in at least three words in the line, excluding stopwords;
  • These words represent at least half the words of the line, excluding stopwords;
  • There is only one most common initial sound (in other words there aren’t two or more equally common initial sounds).

§ 12 Alliterating conditions 2:

  • The most common initial sound of words is used in exactly two words in the line, excluding stopwords;
  • There are exactly two words in the line, excluding stopwords.

§ 13 The first set of conditions can detect alliteration in a line such as to Jesus the gentillest of Jewes generacioun, which alliterates the J and soft G sound, while the second set of conditions can catch shorter lines such as that comely to kenne, which alliterates the hard C sound once we eliminate the stopwords that and to. However, one limitation of our system is that ignoring stressed syllables will miss some sounds, such as the B sound of obeye in the line Obeye vnto bidding bud busshoppis me bowne (although we can note that the alliterating B sound of the line is still correctly detected despite missing that one word).

§ 14 The reason we ignore stopwords in all alliteration conditions is that a disproportionate number of them start with a D sound (the, this, that, though, them, themselves, they, did, done, do, does, etc.) which creates a lot of noise in the results, causing the software to detect false alliterations on that sound and masking real alliterations. Moreover, in readings of the Plays, the emphasis is normally put on content words rather than stopwords, justifying our decision to ignore them in the software. From a technical standpoint, the stopwords were filtered out by using a manually-constructed list of 1439 stopwords (including variations of spelling) that appear in the York Plays. This is the standard approach in language processing algorithms in all languages: since stopwords form a small and closed set of words, it is a simple matter to build an exhaustive list of them.

Character statistics

§ 15 With this algorithm in place, we can measure the alliteration use of the different characters in the different plays of the York Cycle. We define a character as any individual speaker in one play; so for example, Jesus in Play 23 is taken as a different character from Jesus in Play 24, and Angel 1 in Play 1 is a different character from Angel 1 in Play 21. To keep ambiguity at a minimum, we will refer to characters by name and play number when we are not clearly discussing one specific play: Jesus-23, Jesus-24, Angel 1-1, Angel 1-21.

§ 16 We count the number of lines and of alliterating lines for each character, and we compute two character statistics: the total number of alliterating lines the character utters and the proportion of all lines uttered by the character that do alliterate. Both of these values are needed to get a complete picture of the character’s use of alliteration. Indeed, two characters that alliterate 47 and 49 lines of dialogue could seem very similar, until we note that in the first case these represent 67% of the character’s spoken lines (Caiphas-33) while the second total stands at only 21% of spoken lines (Abraham-10), making the first character alliterate a lot more noticeably than the second. Likewise, the proportion is not uniquely important: a character that alliterates 100% of their lines can seem extremely important until we realize that they only speaks two lines in the entire play (Miles-36), making them no more than a background character. A character that alliterates both a large number of lines and a large proportion of their lines would be one showing a truly strong use of alliteration.

Experimental results

Phonetic transcription

§ 17 Since the quality of all our subsequent results are based on our automatic phonetic reading of the text, it is primordial to verify the quality of our phonetic transcription software. Remember that our software is meant to recognize the pronunciation of words in Middle English, which had no standardized spelling, and must therefore be indifferent to (sometimes important) spelling differences, while of course not failing to distinguish between words that should indeed be pronounced differently. We thus designed a simple experiment to test this ability. We designed a list of 381 pairs of English homophones with different spellings (such as aye and eye, core and corps, theme and team) and 390 pairs of words that are spelled similarly but pronounced slightly differently (such as bother and brother, doze and daze, mass and maze; for conciseness, we will refer to these words as nearphones). We should note in passing that words that are spelled identically but pronounced differently based on sentence context, such as the first-person present and past tense of the verb to read, cannot be distinguished by our software as it operates on spelling alone with no notion of sentence context.

§ 18 The goal of the test is to verify that the software can match together as many of the actual homophone pairs (positive cases) while distinguishing the nearphone pairs (negative cases). We thus define four possible cases:

  • True Positives (TP): when the software correctly identifies a pair of homophones as pronounced identically;
  • True Negatives (TN): when the software correctly identifies a pair of nearphones as pronounced differently;
  • False Negatives (FN): when the software incorrectly marks a pair of homophones as being pronounced differently;
  • False Positives (FP): when the software incorrectly marks a pair of nearphones as being pronounced identically.

§ 19 These are standard metrics in the area of information retrieval and classification. We can further use them to define more comprehensive metrics:

  • Homophone Precision (HP): the proportion of pairs marked as homophones that actually are correct, as defined in equation (1). This metric should be as close to 1.0 as possible, to represent that only homophones are detected as pronounced identically by our software.

    (1)

  • Homophone Recall (HR): the proportion of homophones correctly recognized as such, as defined in equation (2). This metric should be as close to 1.0 as possible, to represent that our software is not misidentifying two homophones as pronounced differently.

    (2)

  • Nearphone Error (NE): the proportion of nearphones incorrectly recognized as homophones, as defined in equation (3). This is the opposite of nearphone recall (compare with equation (2)). It should be as close as possible to zero, to represent that our software can reliably distinguish the pronunciation differences of nearphones.

    (3)

  • Accuracy (A): the proportion of homophones and nearphones that were correctly recognized overall, as defined in equation (4). This metric gives a global performance value for the algorithm over both tests. It should be as close to 1.0 as possible, to indicate that the algorithm gets mostly correct answers.

    (4)

§ 20 As a benchmark, we followed the idea presented in (Christen 2006) and compared our system to six word-to-phoneme systems. The systems implemented in the Febrl software package used by Christen (2006) are:

  • Soundex, an algorithm that uses a letter-to-number encoding table;
  • Phonex, an improved version of Soundex that uses pre-processing rules before encoding;
  • Phonix, a further improved version of Soundex that applies transformation rules on letter groups prior to encoding;
  • Fuzzy Soundex, a variation of Soundex that applies transformation q-gram substitutions on groups of letters;
  • NISIIS, another transformation-rule-based system developed independently of the Soundex family;
  • Double-Metaphone, a rule-based system that is much like ours in that its rules take into account the immediate context of the letter in the word.

§ 21 We performed the same homophone and nearphone experiment described above on each of the benchmarks, and the results are presented in Table 4. It can be seen that our system far outperforms the others on the HP and NE results, but at the cost of a much lower HR. This indicates that word pairs identified as homophones by our system are much more reliable than with other systems, but a lot of homophones are missed; other systems are uniformly more liberal and catch a lot more homophones while also misclassifying a considerably higher number of nearphones as homophones. The accuracy results confirm this observation. Indeed, our system is third best, below Double-Metaphone and almost on par with Phonix, but both of these benchmarks generate their high accuracy scores by being much more liberal in marking homophones to massively increase their TP score and lower their FN score (correctly identifying most homophones) at the cost of a very high FP score (mislabelling a lot of nearphones as homophones).

§ 22 This is a good result; we designed our system to “play it safe” and avoid misreading two different words as the same. Given the purpose of our system – to recognize phonetic patterns such as alliteration regardless of spelling – we feel it is better to have a system with this behaviour. Such a system may underestimate the occurrence frequency of certain patterns, but the patterns detected will be truly present in the text. By contrast, the other systems will over-estimate the occurrence frequency of patterns by detecting a lot of patterns that are not truly present in the text.

Table 4: Experimental Results of the homophone and nearphone experiment.

System TP TN FP FN HP HR NE A
Our Algorithm 203 384 6 178 0.971 0.533 0.015 0.761
Soundex 296 272 118 85 0.715 0.777 0.303 0.737
Phonex 281 245 145 100 0.660 0.738 0.372 0.682
Phonix 354 252 138 27 0.720 0.929 0.354 0.786
Fuzzy Soundex 267 282 108 114 0.712 0.701 0.277 0.712
NYSIIS 237 291 99 144 0.705 0.622 0.254 0.685
Double-Metaphone 357 270 120 24 0.748 0.937 0.308 0.813

§ 23 Finally, we should note that our phonetic transcriptions will be used to pick out alliterations, which depend only on the initial sound of each word. Consequently, phonetic transcription errors by our system on sounds other than the initial one will have no impact on our results. This will have the practical impact of greatly increasing the accuracy of our system. In fact, in the homophone test, 364 out of 381 word pairs transcribed by our algorithm had the same initial sound, on par with four of the benchmarks and surpassing Soundex and Fuzzy Soundex.

Alliteration detection

§ 24 Having demonstrated the validity of the phonetic transcription algorithm that underlies our work, we set out to phonetically transcribe the York Plays line by line, detect the presence or absence of alliteration for each line, and then compute the total number and proportion of alliterating lines for each character. The results are presented in Figure 1. The first notable result is the omnipresent use of alliteration: of 381 different characters, only 28 use no alliterating lines at all. On average, characters use 11 alliterating lines, or 32% of their lines. The use of alliteration is also extremely varied, going from 1 line to well over 100 lines in a play, and can represent a mere 1% or a massive 90% of the lines spoken by a character. In order to further illustrate the wide diversity in the use of alliteration of the characters, we have marked in Figure 1 the data points corresponding to the 22 Jesus characters in the Cycle (in plays 20, 21, 22, 23, 24, 25, 27, 28, 29, 30, 33, 34, 35, 36, 37, 39, 40, 41, 42, 44, 46, and 47). As can be seen, even that character ranges from 1 to 60 alliterating lines, or from 2% to 75% of the character’s lines in a play. Even within the multiple appearances of a single individual, there is a very wide diversity in the use of alliteration.

Alliteration statistics of all characters in the York Plays, with Jesus characters marked.
Figure 1: Alliteration statistics of all characters in the York Plays, with Jesus characters marked.

§ 25 As we explained previously, the characters that make the strongest use of alliteration are the ones that show both a large number of alliterating lines and an above-average proportion of alliterating lines. They are quite easy to pick out from Figure 2, as the points nearest to the upper-right side of the graphic represent them. We have marked and labeled these points in Figure 2, along with the other instances of the same characters in the remaining plays for completeness. It can be immediately seen that the five characters with the most significant use of alliteration are Pilate-26, Pilate-30, Herod-31, Pilate-33, and Thomas-45.

Alliteration statistics of all characters in the York Plays, with Pilate, Thomas and Herod characters marked.
Figure 2: Alliteration statistics of all characters in the York Plays, with Pilate, Thomas and Herod characters marked.

§ 26 Another way to consider these results is to plot the number of alliterating lines spoken by each character in function of the total number of lines spoken by that character, which we present in Figure 3. This new perspective will have certain advantages: in Figure 1 minor characters that speak only a few lines could be found anywhere in the lower portion of the graph depending on what proportion of these few lines alliterated, while in this new figure they are all located in the bottom-left origin, keeping the rest of the graphic noise-free. Moreover, the graph will be clearly bounded at the diagonal (the dashed grey line in Figure 3), which is the 100% limit where all lines spoken by a character alliterate. We also included the 32% line (in solid grey in Figure 3), which is the average amount of alliteration used by characters in the York Plays. The advantage of this new graph is thus to clearly note the characters with the most salient uses of alliteration. They will be in the region between the average line and the boundary, and points more isolated from others show more striking alliteration use, while points isolated but far below the average line rather show a striking lack of alliteration. Some of the more alliterative characters are marked in Figure 3; some of them are in common with those marked in Figures 1 and 2, for comparison.

Alliteration statistics of all characters in the York Plays.
Figure 3: Alliteration statistics of all characters in the York Plays.

Significance of the alliteration

§ 27 Now let us revisit the theories on the significance of the use of alliteration in the York Cycle. Having computed the number of alliterating lines for each character in each play, it is easy to sum them to get the number of alliterating lines per play and to compare that to the values in Table 1 to get the proportion of alliterating lines in each play. That information is given in Table 5, with the plays sorted in order of their proportion of alliteration. We find that this global statistical view of the Cycle plays fits perfectly with the study of Reese (1951). Indeed, his 13 plays written in “purely alliterative verse” are exactly the top 13 plays obtained by our algorithm. This lends a lot of credence to our statistical method to study the plays.

§ 28 However, the results also call into question the line Reese drew at 13 plays. After all, play 37 shows alliteration statistics almost identical to those of play 32 while play 35 is not far below them, but neither of them are considered by Reese to be “purely alliterative” plays while play 32 is. Moreover, if one were to try to break the plays into two sets based on the use of alliteration as Reese did, logic would dictate that we should find a set with a lot of alliteration and a set with sparse alliteration and a large gap between them. In fact, the largest gap in alliteration use, representing a drop of 9% in alliterating lines between two plays, exists between plays 44 and 31. That gap also happens to neatly divide the Cycle into the set of plays with more than half their lines alliterating and those with less than half their lines alliterating. Unfortunately, that gap is exactly in the middle of Reese’s set of 13 alliterating plays. To sum up, our numbers support Reese’s observations on which plays use more alliteration than others, but they simply do not seem to support his conclusion that variations in alliteration use mark the change between two different theatrical traditions.

Table 5: Alliteration use in each play

Play Number of alliterating lines Proportion of alliterating lines
26 202 65%
30 355 63%
33 307 63%
1 102 63%
45 193 59%
40 107 54%
44 103 53%
31 193 44%
16 168 43%
36 174 42%
29 172 42%
28 126 41%
32 146 37%
37 145 36%
35 108 35%
23 80 33%
43 76 33%
2 57 33%
11 132 32%
15 43 32%
20 89 31%
47 112 30%
8 45 30%
34 90 28%
9 86 26%
19 70 25%
24 53 25%
27 47 25%
10 85 22%
12 55 22%
6 37 22%
39 32 21%
7 28 20%
18 43 19%
22 39 19%
41 37 19%
46 31 19%
5 30 17%
38 68 15%
13 43 14%
14 22 14%
4 14 14%
3 12 13%
17 55 12%
42 31 11%
25 44 8%
21 7 4%

§ 29 On the other hand, Smith (1963) suggested that alliteration is a trait of the evil characters in the Cycle, such as Herod, Pilate, and Caiaphas, while good characters such as God and Jesus avoid this verbal feature. Epp (1989) has already indicated that this is not the case, and our results fully support him: alliteration appears to be used by good and evil characters alike. For example, Figures 1 to 3 show that Jesus-22 has a level of alliteration that matches Pilate-32 and Herod-16. Moreover, one of the most powerfully alliterating characters in the entire Cycle turns out to be Thomas-45. Far from being an evil character, this is the Thomas who alone is chosen to witness the Assumption of Mary, who receives her garter, and who convinces the other Apostles that Mary is alive when they believe she has died in a poetic reversal of the Doubting Thomas story (Brawer 1972). Finally, we should note that, while Pilate is indeed a character who uses a lot of alliteration, it has been argued that he is not presented as an evil character in the York Plays (Christen 2006).

§ 30 The last theory we presented in Section 2 is Epp’s, which suggests that alliterations are used to represent passionate speech, to signal important moments, to mark certain characters, and for a variety of reasons (Epp 1989). We would disagree with his assertion that these represent a variety of reasons; they are rather a variety of examples of one reason, namely to mark the most important character of a play. Recall that the York Cycle was what Beadle calls an audiate literary experience meant to be listened to, at a time when one went out to hear a play, rather than an experience meant to be read or even seen (Beadle 2000, 169-170). Moreover, this was not one play but a set of plays, performed on several different wagon-stages by different casts of actors. As Johnston (1993) pointed out , spectators would have needed a way to identify characters easily from play to play, since recognizing them visually was not possible. She suggested that characters were recognized instead by what they said and by the use of the same words from play to play. This idea agrees completely with Beadle’s insistence that the Plays were a primarily auditory art medium (Beadle 2000). Joining her idea and Epp’s, we posit that alliteration served the purpose of helping with character identification: the repeated sounds would have immediately caught the ear of the audience and focused their attention on the character speaking them in addition to making the character’s speech easier to remember. We posit, then, that one or two characters in each play used alliteration more heavily and that character or pair of characters would be the main characters of that play, specifically the characters that spectators should pay more attention to. To verify this, we present in Figures 4 and 5 the alliteration statistics for each individual play (i.e. the same statistics as in Figures 1 and 3, but in an individual graph for each play). As before, the most significant character would be the outlier of each play who uses alliteration in a way that is strikingly different than and superior to any other character in the same play. Visually, in Figures 4 and 5, that would be the character whose data point is isolated from the other characters and closer to the upper-right corner of the graph. Table 6 lists each play and the character or characters that are most significant in it. For each one, we also offer our observations on the role of the alliterating character in the play, or in the case where there are two or three alliterating characters on their relationship. Note that while Figures 3 and 5 plotted the Cycle average line of 32% alliteration, the individual play figures that are presented in Table 6 instead plot each play’s average line, which will be different for each individual play. This will illustrate the behaviour of characters compared only to the others on stage at the same time. We feel this makes more sense for a study of individual plays.

Alliteration statistics of characters in each individual play, using data from Figure 1.
Figure 4: Alliteration statistics of characters in each individual play, using data from Figure 1.

Alliteration statistics of characters in each individual play, using data from Figure 3.
Figure 5: Alliteration statistics of characters in each individual play, using data from Figure 3.

Table 6: Most alliterating character(s) in each play.

Play Character(s) Notes
1 God-1, Lucifer-1

God and Lucifer are the two outliers in this play compared to the other four characters, as can be seen in Figure 6 below. God speaks the most lines and the most alliterating lines, but he is below the alliteration proportion average for the play while Lucifer is above. In fact, God in this play is by far the character with the lowest proportion of alliteration in his speech. The other four characters (two unnamed angels, a devil and a cherub) play only supporting roles in the play, and statistically they are clustered together. This play is about the fall of Lucifer, so it makes sense for God to have the spotlight by amount of spoken lines and for Lucifer to encroach on it subtly through alliteration. It is also interesting to note that God and Lucifer do not speak to each other in this play: God gives the opening and closing monologues, while Lucifer speaks with the other characters in the body of the play.

Alliteration statistics of characters in Play 1.
Figure 6: Alliteration statistics of characters in Play 1.

2 This play has only one character, so a comparison cannot be done.
3 God-3

As God creates Adam and Eve, he is the weak focus of alliterative attention of the play. By “weak” we mean that he has a bit more alliteration than the other characters in the play, but the proportion of alliteration used in this play is noticeably less than in other plays in the Cycle.

Alliteration statistics of characters in Play 3.
Figure 7: Alliteration statistics of characters in Play 3.

4 God-4 God shows Adam and Eve a blissful life in the perfect Garden of Eden. He remains the weak focus of alliterative attention, as in Play 3.
5 Satan-5,
Adam-5, Eve-5

This is the only play in the Cycle with three alliterating characters out of five speaking characters on stage. Satan, Adam, and Eve all have very similar alliteration statistics; Satan’s only claim to superiority is an edge on the number of spoken lines, as can be seen in Figure 8. Interestingly enough, the alliteration statistics for Satan in this play are practically identical to those of God-3 and God-4. Meanwhile, God’s statistics drop considerably compared to the previous four plays, marking him as a secondary character, along with an Angel. This seems appropriate for a play on the Fall of Man: focus and blame rests on Satan, Adam, and Eve.

Alliteration statistics of characters in Play 5.
Figure 8: Alliteration statistics of characters in Play 5.

6 Adam-6 or
Eve-6
With only three characters with different statistics, it is difficult to determine the clear focus of the play. Adam has the most lines and alliterating lines in this play, while Eve has a distinctly greater proportion of alliterating lines in her dialogue. It is however possible to see a clear secondary character: the Angel, who, despite having more spoken lines than Eve, has very little alliterating focus in the text.
7 Probably Abel-7 Two characters attract attention in this play: the Angel, who delivers the opening monologue of the play and who passes judgement and sentence on Cain after he kills his brother, and Abel, the good brother who piously praises God. The Angel has the most alliterating lines, but in terms of proportion of alliterating lines he is about equal to Cain. On the other hand, Abel is far above both of them in proportion of alliterating lines, but far below in actual number of lines and alliterating lines spoken. However, we still picked Abel as the likely focus for one reason: a large section of this play, believed to contain a lot of dialogue by Cain and Abel, is now lost. If these lines were consistent with the surviving portion of the play, then Abel would have been a clear outlier in this play.
8 God-8, Noah-8 This play has only two characters with roughly equal alliteration statistics. This makes the Building of the Ark into something of a dialogue between two equals.
9 Noah-9 Noah becomes the clear main focus during the Flood.
10 Abraham-10 Abraham and Isaac use alliteration equally, but with Abraham speaking more than twice as many lines as his son, he becomes the alliteration spotlight in this play and the outlier in the statistical graphs.
11 Moyses-11, Rex Pharao-11

With nearly identical alliteration statistics and much more dialogue than the other characters in this play, Moses and Pharaoh are the two points of focus. This verbally illustrates the duel they were engaged in and further shows them as evenly-matched opponents, possibly adding to the drama of the fight and the power of Moses’ victory in the minds of the audience.

Alliteration statistics of characters in Play 11.
Figure 9: Alliteration statistics of characters in Play 11.

12 Doctor-12 Surprisingly, it is not Mary, her mother Elizabeth, or the announcing angel, but the unnamed Doctor who is by far the most alliterating character in this play. But this Doctor is an important character in the play: he delivers the opening narration, in which he becomes the first character to speak of Mary’s virgin pregnancy ordained by God for a son named Emmanuel to save mankind from sin as promised to Abraham and David – in essence, he delivers an entire statement of Christian faith and history in a single highly-alliterating speech.
13 Joseph-13 Joseph is a weak focus in this play, appropriately enough since he is also a hesitant main subject.
14 Joseph-14, Mary-14 Like play 8, this play only has two characters and presents their dialogue as a duet between two equals.
15 Pastor 2-15 While the first shepherd speaks the most lines, it is the second one who alliterates the most, and the third one is the weakest of the trio. It should be pointed out though that the difference between the three shepherds is not as strong as that between characters in other plays. In the text, all three characters say praises at the birth of Jesus.
16 Herod-16 While the three wise men show strong alliteration rates in the mid-30% to mid-40%, Herod dominates this play verbally and alliterates half his lines, perhaps as a way to show his power over the other characters.
17 Symeon-17 Simeon is statistically the weak main character of this play. In his monologues he offer lamentations, praise to God, and finally prophecy of the life of Jesus.
18 Joseph-18, Mary-18 While Joseph speaks more than twice as many lines as Mary as he leads his family into Egypt at the advice of the Angel, Mary edges him out in the proportion of alliteration used. The Angel, meanwhile, is far below the other two characters in all statistics.
19 Herod-19 This play is about the Slaughter of the Innocents, and appropriately enough Herod, the perpetrator of this act, is also the statistical outlier of the play.
20

This play, about a young Jesus teaching to the Magisters and Doctors at the Temple, does not feature a single character, duo, or trio who overshadows the other characters in the play. Instead, there is a large cluster of characters with roughly equal alliteration statistics, counting six of the nine characters in the play. These six characters all alliterate around 28% to 38% of their lines, comparable to the average alliteration rate of other characters in the Cycle.

Alliteration statistics of characters in Play 20.
Figure 10: Alliteration statistics of characters in Play 20.

21 The baptism play is also unusual, this time because it features practically no alliteration. One of the Angels is the character who alliterates the most out of the four in this play, and even so this represents only 10% of his spoken lines. In fact, Table 5 has shown that this play is the one with the least amount of alliteration in the entire Cycle.
22 Diabolus-22, Jesus-22 The Temptation play is again a duet, this time between a devil and Jesus, while two angels serve only as secondary characters. Of the two, the devil has more lines, but Jesus alliterates a greater proportion of his lines, showing an interesting struggle for the audience’s attention. In the end, the devil would have taken more of the audience’s time, but Jesus’ replies would have been more memorable.
23 Like play 20, this play is unusual for having most of its characters alliterate equivalently.
24 The third unusual play to have a large set of alliterating characters.
25 Jesus-25, Burger 2-25 This play as also features a lot of alliterating characters. Nonetheless, two characters stand out in this play’s statistics. The first is Jesus, with the greatest number of spoken lines and of alliterating lines, and well befitting a play about his triumphant entry into Jerusalem. To compare, the second-most verbose character in the play is an unnamed porter, and he speaks almost as many lines as Jesus but less than half as many alliterating lines. The second noteworthy character is simply called Citizen 2 in the play, and he has by far the greatest proportion of alliterating lines in this play. However, he seems otherwise unremarkable compared to the other seven citizens, and his importance is unknown to us (if there is any; this may be only a statistical anomaly).
26 Pilate-26, Judas-26

The Conspiracy play is the most strongly alliterating play of the Cycle, as shown in Table 5; characters here alliterate on average 65% of their lines, and no character alliterates less than half his lines. But even so, two characters stand out, while the other 8 are secondary. Pilate, the Roman master of Judea, is central to the Jewish conspiracy to stop Jesus, as he alone can order Jesus’ death and he alone must be convinced to do so. He is a positive outlier in the alliteration statistics. Judas, who here sells out Jesus and allows the conspiracy to take shape is the second statistical outlier, though he commands much less attention than Pilate. In fact, his statistics make him a negative outlier in that he uses the least amount of alliteration of any character in the play. This play is not a duet of equals like plays 8, 11, or 14, but a play with a single main character (Pilate) and an anti-hero character (Judas). Whether this was done on purpose, perhaps as a way to illustrate a relationship between Pilate and Judas or to contrast them, or whether this was only an accident is unknown.

Alliteration statistics of characters in Play 26.
Figure 11: Alliteration statistics of characters in Play 26.

27 Jesus-27 Plays 27 to 33 all seem to have similar statistical characteristics. They all feature large casts of characters (8 to 12 according to Table 1) but with only one clear outlier character (see Figures 4 and 5). In each case, this outlier character dominates the dialogue of his play, while the other characters in the same play speak on average less than 16% as many lines and almost always less than half as many lines, and thus play little more than a support role to the outlier character’s monologue. The outlier character also always uses more than the play’s average amount of alliteration. For the Last Supper play, the uncontested central character is Jesus.
28 Jesus-28

Jesus continues to be the main focus as he agonizes in the Garden of Gethsemane. All other characters have either less than a dozen lines, or use little alliteration. Looking at Figures 1 and 3, we can see that this play also has the most notable alliteration use by Jesus in the Cycle. This may indicate that a special importance was given to the Garden of Gethsemane scene.

Alliteration statistics of characters in Play 28.
Figure 12: Alliteration statistics of characters in Play 28.

29 Caiphas-29 The pattern described in Play 27 continues as we enter the Passion plays, but now Jesus becomes a supporting character as focus moves to his opponents. In this play, Caiaphas is the center of attention.
30 Pilate-30 Pilate takes the focus when Jesus is brought to him.
31 Herod-31 Herod takes the focus when Jesus is brought to him.
32 Pilate-32, perhaps Judas-32 This play is about the Remorse of Judas, yet most of the dialogue is given to Pilate questioning the Jewish authorities about Jesus. Pilate is the clear statistical outlier and focus of the play. However, Judas still gets a large number of lines, marking an exception to the pattern described in Play 27. On the other hand, Judas gets by far the lowest alliteration rate of any character in this play. In that sense, this play could even be said to follow the hero-anti-hero pattern of Play 26.
33 Pilate-33 Pilate keeps the focus as Jesus is brought to him again.
34 This play marks the clear end of the set of successive plays that followed the pattern of Play 27. Far from having one clear focus of attention, this play has four of 10 characters with a large number of lines and similar usage of alliteration. They are the three soldiers that escort Jesus and John his apostle.
35 Miles 1-35 Four soldiers argue as Jesus is on the cross. But the first soldier seems to take more space in the play, with more lines and a greater proportion of alliteration. In the action, he is the one who receives Jesus’ garments. Interestingly, Jesus makes a roughly equal use of alliteration as that soldier, but with much fewer lines. This may allow Jesus to attract attention above the more verbose soldiers, or it may have been meant to mark some kind of relationship between Jesus and that first solider.
36 Another play that, like 20, has characters distributed far too uniformly in the statistics to make it possible to pick out a single outlier.
37 Jesus-37, Satan-37 The Harrowing of Hell is an alliterative duet between Jesus and Satan, with Jesus using longer speeches with more alliterating lines but Satan using a greater proportion of alliterating lines in his speeches.
38 Centurio-38 While the topic of this play is Mary and Mary Magdalene finding the empty tomb and meeting the Angel. While Pilate is the most talkative character in the play, the greatest use of alliteration is found in speech of the Centurion who holds that Jesus was a righteous man.
39 Mary Magdalene-39, Jesus-39 Another duet between two roughly equal characters takes place when Jesus appears to Mary Magdalene. However, we should note that Mary Magdalene has a slight edge on Jesus, with a little bit more alliterating lines and a bit greater proportion of alliterating lines.
40 Jesus-40 Jesus appears to pilgrims on the way to Emmaus. The two pilgrims talk much more than he does, but Jesus still gets attention by alliterating almost every line he speaks.
41 Thomas-41

Jesus appears to four of his Apostles and dismisses Thomas’ doubts about the resurrection. While Jesus speaks the most lines, Thomas is by far the alliterative center of this play, with three times the alliteration rate of Jesus.

Alliteration statistics of characters in Play 41.
Figure 13: Alliteration statistics of characters in Play 41.

42 Jesus-42 Jesus is the main focus of this play depicting his ascension into heaven.
43 Yet another play lacking a large distinctive gap between alliterating and secondary characters. This one does seem to give a bit of an edge to three characters, who are, in order of importance in our theory, the unnamed fourth Apostle, the unnamed first doctor who mocks the apostles, and then Peter. Two of these do not seem to be particularly important to the play, so it does seem like our method fails on this play.
44 Jesus-44 15 characters are present at Mary’s death, but 11 of them have five lines of speech or less. Of the four that have a meaningful amount of dialogue, Mary speaks by far the most but also has the lowest alliteration rate. The other three, Jesus, John, and Gabriel, speak approximately the same number of lines, but Jesus still overshadows them with a noticeably greater rate of alliteration.
45 Thomas-45 The play on the Assumption of the Virgin was directly based on popular apocryphal stories about her appearance to Thomas (Wall 1970). Accordingly, it is Thomas who is the center of attention in this play. In fact, from Figures 2 and 3, it can be seen that Thomas in this play is the most striking character of the entire Cycle.
46 Jesus-46 The third and last play on Mary’s death depicts her Coronation in heaven. In this play, Mary takes a background role, while the focus is on her son Jesus as he blesses her and welcomes her to heaven.
47 Jesus-47, God-47 Like the initial play, the final one is a duet between two major characters in front of a cast of secondary characters, and like the initial play it has God giving the opening and closing monologues while the other major character speaks with the cast of minor characters in the body of the play. But unlike the initial play, the second major character is Jesus, not Lucifer, and he of course agrees with God rather than rebels against him. From this point of view, the final play thus seems to be a reprise of the first play, but with heavenly harmony now replacing the original confrontation.

§ 31 Overall, the results seem to confirm our hypothesis. It is possible to identify a single most strongly alliterating character in 27 of the 47 plays, and in all these cases that character is the single most important one in the play. Moreover, in another 12 plays the alliterations reveal a duet between two major characters, either as allies as was the case for Joseph-14 and Mary-14 or God-8 and Noah-8, or in direct confrontation to each other, as for Jesus-22 and the tempting Diabolus-22, or Moses-11 and Pharaoh-11. One play even shows a triplet of major characters, as Satan-5 tempts Adam-5 and Eve-5. This leaves only 6 plays that violate our hypothesis and one that has only one character and cannot be analyzed.

§ 32 Since the most important character in play will often also be the one that will get the most dialogue, it should not be surprising to find in Table 6 that the most vocal character is often the one that also gets the most noteworthy alliteration use. However, Table 6 notes several interesting cases where this connection does not hold. Play 44 is a perfect example where one character, Mary-44, speaks as many lines as the next three most talkative characters combined, yet she uses the least amount of alliteration of these four individuals. On the other hand, the fact that in many cases the central character both speaks the most lines and uses more than the average number of alliterating lines in a play does not contradict our thesis; quite the opposite, it indicates that those two techniques were consciously used together to create a striking auditory effect.

§ 33 Moreover, the analysis of Table 6 highlights what appear to be patterns in the statistical distribution of characters in the Plays. The most striking one was described in our discussion of play 27; it is the pattern of a play with at least 8 speaking characters on stage and with a single major character that dominates the spoken dialogue and who uses a proportion of alliteration equal or greater to the play’s average. This is the case for plays 9, 19, 27, 28, 29, 30, 31, 33, 42, and 45, and also somewhat ambiguously for play 32 (if we ignore the large number of lines given to Judas) and play 25 (if we assume that Citizen 2 is a statistical anomaly). Some plays have a similar pattern of having a single major character that dominates the dialogue and alliterates a lot, but in front of a much smaller cast of secondary characters: they are plays 3, 4, 12, 13, 17, and 35. Finally, some plays have a related pattern of having a single major character, but one that only establishes a weak dominance on the other, by alliterating either more lines but a proportion less than the play’s average, or vice versa, by alliterating a great proportion of lines in only a short speech. These are plays 6, 7, 10, 15, 16, 38, 40, 41, 44, and 46. There are also patterns for duets. The first is the duets of equals, or of two characters with very similar statistics. These occur in plays 8, 14, and 39, the three plays of the Cycle that have only two characters, and in play 11, in front of a large cast of 8 background characters. Then there is the duet of a dominating character that has the greatest number of alliterating lines and a greater-than-average proportion of alliteration, and an anti-hero character that also speaks a great number of lines but alliterates much less than average for the play. This pattern occurs in play 26 and ambiguously in play 32; interestingly, both of the plays are about Judas’ betrayal, and both of them feature Pilate as the dominating character and Judas as the anti-hero. Finally, there is the uneven duet where one character dominates in the number of alliterating lines and the other dominates the proportion of alliterating lines. These occur in plays 1, 18, 22, 37, 47, and ambiguously in play 25. However, if there is an underlying logic to the choice of which character dominates in alliterating lines or proportion, we cannot determine it at this point. Graphical examples of at least one of each of these patterns are included in Table 6 and in Figures 6 to 14.

Conclusion

§ 34 This paper makes two important contributions. The first is a novel algorithm to define and measure alliteration in Medieval English documents. Our second contribution has been to use this algorithm on the York Cycle of plays in order to gain new insights on the use and significance of alliteration in those plays. Indeed, we have proposed and demonstrated with experimental results that alliteration seems to be used as a tool to attract the attention of medieval viewers to the most significant character in each individual play. To be sure, more work remains to be done to expand on our study. Most notably, six plays clearly did not fit our model, as we pointed out in our analysis in Table 6 and as can be seen from the results in Figures 4 and 5. The reason for these exceptions is unknown at this time and clearly calls for further research. Likewise, studies could be done on the possible patterns of alliteration that seem to emerge from our results and on their possible significance or relationship to the themes of the plays they appear in. Finally, this work could be the first step in a thorough algorithmic and statistical study of the features of the York Plays, which could help shed new light on this important Cycle.

Notes

[1]. Richard Khoury is the author to contact to obtain copies of the data and software.

Works cited

Martin, Joseph, ed. 1962. S. Aurelii Augustini De doctrina christiana. In CCSL 32, 1-167. Turnhout: Brepols.

Beadle, Richard, ed. 1982. The York Plays. Accessed on March 2, 2015. http://name.umdl.umich.edu/York.

Beadle, Richard, and Pamela M. King, eds. 1984. York Mystery Plays: A selection in modern spelling. New York: Oxford University Press.

Beadle, Richard. 2000. Verbal texture and wordplay in the York Cycle. Early Theatre 3:167-184.

Beadle, Richard, ed. 2009. The York Plays. Early English Text Society S.S., vol. 23. New York: Oxford University Press.

Brawer, Robert A. 1972. The characterization of Pilate in the York Cycle Play. Studies in Philology 69.3: 289-303.

Chambers, Edmund Kerchever. 1949. English literature at the close of the Middle Ages. Oxford: Clarendon Press.

Christen, Peter. 2006. A comparison of personal name matching: Techniques and practical issues. In Sixth IEEE international conference on data mining workshops. Hong Kong, China. 18-22 December.

Davidson, Charles. 1892. Studies in the English Mystery Plays. Transactions of the Connecticut Academy of Arts and Sciences, IX.

Epp, Garrett P. J. 1989. Passion, pomp and parody: Alliteration in the York Plays. Medieval English Theatre 11.1: 150-161.

Gayley, Charles Mills. 1907. Plays of our forefathers and some of the traditions upon which they were founded. New York: Biblio-Moser.

Greg, Walter Wilson. 1914. Bibliographical and textual problems of the English Miracle Plays. The library, third series, V.

Hayes, Douglas W. 2013. The rhetoric of stasis and chaos in the York Judgement Play. Medieval and Early Modern Institute virtual symposium.

Holfled, Alex. 1889. Die Altenglischen Kollektivmisterien, unter besonderer berücksichtigung des Verhältnesses der York und Towneley-Spiele. Anglia, XI.

Johnston, Alexandra F. 1993. The word made flesh: Augustinian elements in the York Cycle. In The centre and its compass: Studies in medieval literature in honor of Professor John Leyerle, edited by Robert A. Taylor, James F. Burke, Patricia J. Eberle, Ian Lancashire, and Brian S. Merrilees, 225-246. Kalamazoo: Western Michigan University Press.

Matonis, Ann T. E. 1984. A reexamination of the Middle English alliterative long line. Modern Philology 81.4: 339-360.

Reese, Jesse Byers. 1951. Alliterative verse in the York Cycle. Studies in Philology 48.3: 639-668.

Smith, Lucy Toulmin. ed. 1885, reprint 1963. The York Plays. New York: Russell & Russell.

Wall, Carolyn. 1970. The apocryphal and historical backgrounds of ‘The appearance of our lady to Thomas’ (Play XLVI of the York Cycle). Medieval Studies 32: 172-192.