Two or III Feet Apart: Oral Recitation, Roman Numerals, and Metrical Regularity in Capystranus

Abstract

The Middle English verse romance, Capystranus, contains 32 numerical expressions in fewer than 580 lines, some of which are spelled out, others are in Roman numerals, while the rest are in mixed orthographical/numerical forms. Values range between one and 100,000 and contain up to four individual lexemes requiring as many as six syllables to be expressed orally. Although it was initially suspected that such complex expression might be silently “skipped over”, the verse shows that the poem’s metrical regularity is dependent upon full vocal/acoustic realization of all numbers expressed in Roman notation. While this does not preclude silent reading, it does at least suggest the text was suitable for oral recitation. This paper shows how even complex Roman numerals were either suitable or suitably adapted for quick and easy interpretation. Propositions are provided concerning the choice of orthographical representation or Roman notation and its treatment. In this romance it seems that line length was less important than a number’s position in a line, as well as the ease with which a number could be “subitized” or mentally visualized and approximated, resulting in high frequency, low-value numbers usually being realized as lexical items to be written out, while high-value numbers conceived of more abstractly would be represented symbolically. The exceptions seem to be certain high frequency numbers of account (e.g., 20, 100, 1000), which could be provided orthographically or numerically despite difficulty subitizing them, while common low-value numbers in time expressions could be expressed numerically as they would appear on time pieces.

Introduction to the Text

The relatively obscure late fifteenth-century verse romance Capystranus recounts the raising of the Turkish siege of Belgrade by John Hunyadi and John of Capistrano in 1456; although it could have been written at any time between that date and 1515 when it was first printed, Stephen Shepherd notes that “a number of obvious corrupt rhymes” suggests that it had probably been circulating in some form for a considerable time beforehand (1995, p. 391). While his conclusion seems to be accepted by Rhiannon Purdie (2008, p. 171), Bonnie Millar-Heggie holds that the romance was produced closer to the terminus ante quem of 1515 (2002, p. 131), the date of the earliest attested printing; for her part Éva Róna (somewhat confusingly) posits that it was both based on eyewitness accounts as well as being composed specifically for the press (1966, pp. 351–352). Whatever the case, it today exists in three fragments, each apparently printed by Wynkyn de Worde, the first of which (London, British Library, incunable C. 71. c. 26 (Short Title Catalogue no. 14649)) is the most complete. The second printing in 1527 (London, British Library, C. 40. m. 9(18) (STC no. 14649.5)) only preserves the first 120 lines, while the third of 1530 (Oxford, Bodleian Library, Douce frag. f. 3 (STC no. 14650)) preserves lines 161-397.

In none of these versions is the romance complete, a point which might account for it receiving so little modern scholarly attention, despite Anastasija Ropa’s assertion that such frequent printing—as well as the inclusion of expensive woodcut illustrations—“could, perhaps, testify to its popularity with the [contemporary] audience” (2015, pp. 255, 258). Its longest version ends imperfectly with the besieged Christians despairing before presumably turning the tables and achieving the unexpected victory reported in the historical record. Ropa further notes that “current anxieties about the Ottoman threat” would have lent the romance particular appeal; she suggests that the lack of more complete copies after three apparently successful printings indicates that—much as Anthony Edwards observed in respect to de Worde’s printings of Morte Darthur—the romance had been “literally read to destruction” (2015, p. 258).¹ Structurally the poem does adhere to what one would expect of a late Middle English verse romance, employing the twelve-line tail-rhyme stanza so typical of the genre, though—unsurprisingly in a popular work—a few six- and nine-line variants do appear. While the meter is also occasionally irregular, it nevertheless tends to adhere to the convention of having a four-beat meter in the rhyming couplets forming the A-, C-, D- and E- lines, while the intervening B-lines—the tail-rhyme occurring in every third line of a stanza—each contain the requisite three beats and rhyme together amongst themselves. Thus, despite twenty-first-century underappreciation, in terms of both content and style, the poem does seem to have all the elements needed to lend it popular appeal in the late fifteenth or early sixteenth century.

Presence of Roman Numerals in the Text and Their Suitability for Recitation

A striking point about the text is that it is particularly rich in numerical expressions, having 31 cardinal numbers and one ordinal number in its 579 extant lines (Table 1).² In other words, on average a numerical expression can be expected every stanza and a half. In a relative majority of cases, the numbers are spelled out; in a minority of cases, they are provided entirely in Roman numerals; in between, a fair number of cases are provided in mixed notation. Given such frequency, this usage might lead one to question the extent to which numbers expressed logographically or ideographically are taken into consideration in the meter’s syllable count, especially considering that some values represented wholly or partially by Roman numerals require up to four words and six syllables to express orally in the manner they are represented here (e.g., l. 335).³ Rapidly decoding such values and rendering them into speech could conceivably prove disruptive to an oral delivery; neuropsychological studies have shown that recognition of the phonological realization of a numerical value (i.e., how to pronounce it) is more rapidly realized when it is spelled out orthographically rather than represented symbolically, either in Arabic or Roman numerals, suggesting that such representation is not conducive to oral recitation. In their study of the processing skills involved in reading Arabic numbers, Besner and Coltheart (1979, p. 467) note:

When a word is written in a syllabic or alphabetic script, its orthographic representation specifies its pronunciation, since the individual symbols making up such a representation map onto individual syllables or phonemes. In contrast, when a word is written in a pictographic or ideographic script, there is no direct representation of pronunciation. The point is that theories of reading which propose that a printed form is initially converted into a phonological representation which is then used throughout subsequent stages of the reading process are much more easily formulated in relation to the reading of alphabetic or syllabic scripts than pictographic or ideographic scripts.

In other words, sounding out numerical values is easier when they are written orthographically rather than represented symbolically. Furthermore, Roman notation has been empirically demonstrated to be more difficult to decode than Arabic numbers. In another study on the neural correlates involved in processing Roman numerals, Nobuo Matsuka and his team observed that “Roman code generally takes more time to process than numbers written in digits or in alphabetical code” (Masataka et al. 2007, p. 282).

Table 1 List of all numerical expressions in order of occurrence

In view of such evidence, it was initially speculated that numbers expressed in Roman notation might be silently “skipped over” at this late date, because Arabic numbers had been increasingly employed since the thirteenth century (Hill 1915, p. 11), and by the mid-fourteenth century “cumbersome Roman numerals had given way to Arabic figures in most contexts” (Picard 2017, p. 71).⁴ Although traditional Roman numerals continued to be used for display purposes in the manufacture of artisanal products, such as clocks, by the turn of the sixteenth century Arabic numerals had supplanted them for most practical purposes and, “owing chiefly to the influence of printing”—such as our copies of Capystranus—“the forms [of Arabic numerals] became practically fixed” (Smith 1916, p. 193). In his seminal study on the use of Arabic numerals in Europe, Hill notes that their functional utility and preference in printing made them common by 1500, and that “after about 1510…the use of Arabic numerals had become universal” (Hill 1915, pp. 9–10). As such, it would seem logical to conclude that the retention of Roman numerals in the first printing of Capystranus, occurring half a decade later, was more for display purposes than practicality; much like the costly illustrations which suggest that the text was meant to be admired visually, it would not be a stretch to conclude that, like the stylized script, the Roman numerals were meant to be seen as much as read.

Furthermore, skipping over such numbers need not necessarily be considered careless reading when the recognized tendency to inflate statistics is taken into account. Even in serious chronicles, Verbruggen notes that “Fantastically exaggerated numbers were accepted and repeated” (1997, p. 6). Stephen Morillo concurs that “tremendously exaggerated numbers appearing in medieval chronicles were either repeated or increased in subsequent works” (1999, p. 115), while Toni Mount likewise observes that “medieval chronicles exaggerated both the numbers engaged and the casualties” (2016).⁵ Capystranus abounds with such overestimations, reporting that 100,000 Ottomans on dromedaries attacked and killed 20,000 defenders on “Maudeleyne nyght” (ll. 444–455), though twentieth-century scholarship estimates that only about 30,000 Ottomans participated in the siege, while professional troops under Hunyadi and Capistrano numbered barely 4000 (Setton 1984, pp. 174–177). Such over-reporting was taken as granted—and indeed, even today there is a tendency to knowingly repeat disproven figures which have become traditional (e.g., Norwich 1982, p. 269)—so it therefore seems reasonable to assume that contemporary readers might be less concerned with establishing the exact value given, and more inclined to interpret any substantial figure represented by Roman numerals as the equivalent of “heaps” or “masses”.

Taken altogether, the preponderance of numbers written out, the mounting obsolescence of Roman numerals, their inclusion in a printing meant to be admired visually as much as read, and the frequent triviality of precise values might indicate a tendency towards silent reading, or at least a disregard for the numerical expressions, but ultimately it seems that in all cases the numbers’ pronunciation lends regularity to the meter, including quite large and complex numbers expressed wholly or partially in Roman notation. This hints at an important oral component in reading the text “out loud” even at this late date, something that would also require a fairly immediate decoding of the numerical expression.

Numbers Expressed Orthographically in Their Entirety

When listed selectively (Table 2), it can immediately be seen that a majority of the numbers—17 of the 32 examples, just over half (53.125%)—are expressed orthographically in their entirety (two instances each in lines 2 and 113, and one each in the remaining lines in which numbers are only expressed orthographically). That such a high percentage of numerical expressions is spelled out rather suggests that the acoustic phonemic component of these words was at least as important as their semantic value. This is itself perhaps a strong indication that the text would have been at least suitable for oral recitation.

Table 2 Numbers expressed orthographically in their entirety

Numbers Expressed Entirely in Roman Notation

Six examples—thus 18.75% of the lines containing numerical values—are entirely provided in Roman numerals (Table 3); four of these examples consist of a single character, which represent the numbers “hundred” (i.e., “.C.” l. 161), “thousand” (i.e., “.M.” l. 433), “five” (i.e., “.v.” l. 443) and “ten” (i.e., “.x.” l. 444) respectively. Each of these single-character glyphs is both preceded and followed by a punctus, the former two being capitalized, the latter two being lower case. It is perhaps noteworthy that throughout the text all ideographs representing a value lower than 100 are consistently in lower case (including “fifty,” i.e., “l”), while all such logographs representing values of 100 or higher are in upper case.⁶ This distinction remains in place even when the former and latter groups are combined to represent numbers of intermediate or mixed values. (This point is obscured in Shepherd’s edition, in which all numerical values are regularized to lower case.)

Table 3 Numbers expressed entirely in Roman numerals

The third example in Table 3 (l. 394)—representing the number “six”—is comprised of two symbols, both lower case letters, though curiously there is only one punctus preceding the numerical expression, and no punctus following it. Finally, the first example in the text of a number in the tens of thousands—representing the value 80,000—consists of four lower-case symbols (lxxx) for “eighty” separated by a punctus from a following capital “M” to represent “thousand,” the whole being both preceded and followed by puncti.

The supposition that the text was intended to be suitable for recitation is here supported by the paucity of values expressed entirely in Roman notation—only six out of 32—four of which consist of a single character, and therefore would require simple recognition of a single glyph rather than any mental arithmetic. Likewise, the alphabetic digraph to represent “six” would probably be a familiar “clock number” characteristically appearing on time pieces, and as a high frequency number of accounts and measurements common in day-to-day life (e.g., a half dozen, half a foot, a half shilling, etc.); indeed, it is provided here in a time expression, which, as shall be seen, seems to prompt symbolical rather than orthographical notation, and the traditional ideograph would undoubtedly be familiar enough require no computation. Even if it were unfamiliar, however, it is the sort of number like nine (represented by “IX” in their paper) that Nobuo Masataka et al. (2007, p. 282) concede is not very difficult to decode in their study of post-pubescent acquisition of Roman numerals amongst individuals never before exposed to the system. None of these expressions could therefore be expected to impede fluid recitation when reading out loud.

The first of these six examples, however—in which 80,000 is represented by the rather cumbersome (and technically incorrect) “.lxxx.M.”—is somewhat unexpected. This is the sort of number which Masataka’s team would presumably put in the category of their “MMXVII”, which they state “should take longer to compute simply because more manipulation is necessary” (2007, p. 282). Nevertheless, the round number represented by “.lxxx.” can only be construed as the value represented uniquely by the English word “eighty”,⁷ and the rest of the expression consists of the relatively common logograph “.M.” for 1000. While this expression might indeed require rapid decoding, it is neither terribly complex nor ambiguous. Indeed, by being divided into two discrete units, the first representing the word “eighty”, which is separated by a punctus from the following one-character symbol representing the word “thousand”, this complex variation on Roman notation is arguably more easily deciphered than the correct Roman notation for this value, ${\bar{\hbox{L}}}{\bar{\hbox{X}}}{\bar{\hbox{X}}}{\bar{\hbox{X}}}$, in which the value of 80 is raised to the thousands by inclusion of macrons over the figure. Not only are the individual words “eighty” and “thousand” here signaled independently—aiding the reader to enunciate them separately—the representation “.lxxx.M.” avoids the possibility of missing the macrons or misinterpreting them as a flourish on the page, which would perhaps have been even more problematic if, as Shepherd posits, the romance circulated in manuscripts before being printed (1995, p. 391). As such, it seems unlikely that the manner in which this value is represented would impede the rhythm of oral recitation by a competent reader—indeed, it would probably facilitate accurate performance even better than the standard method of signaling the thousands place—again suggesting that this text was suitable for being read out loud.

Numbers Represented by Mixed Expressions

Finally, ten lines—slightly less than a third of the total (31.25%)—have numerical values that are given in a combination of numbers spelled out and Roman notation (Table 4), all of which express a multiple of hundreds or thousands All individual component values expressed orthographically in these examples are also represented by Roman numerals elsewhere in the text with the exception of the seventh example on the table (l. 450), the “two” of 2000, a value that also appear at the end of line 350, where it is likewise spelled out. Similarly, all of the values represented by alphabetical Roman numerals in the mixed examples also appear orthographically elsewhere.

Table 4 Mixed expressions to represent numbers

The second example in Table 4 (l. 335), representing 26,000, deserves particular attention. Although the value 26 never occurs anywhere else in the text, both the values 20 and six do, the former always in mixed expressions where it is either spelled out or represented in Roman notation, while the latter only occurs in one other instance where it is a whole number on its own represented by a Roman-numeral digraph. Here they have been deliberately split apart into their separate values with the ones (or unit) position preceding the tens position in the increasingly archaic Germanic fashion, the two values both semantically linked and lexically separated by the conjunction “and”. This also seems indicative of oral reading, for while “.xxvi.” could undoubtedly be printed as easily as “.lxxx.”, it perhaps risked being misread as “twenty-six”, which would disrupt the meter. It seems that rather than risk the ambiguity that might result in a misreading when reading out loud—the only situation in which a misinterpretation would matter—the words in this instance are spelled out in order to show exactly how the figure should be enunciated. Furthermore, as with 80,000 (above), the value of thousand is conveniently marked by a separate logograph rather than a relatively inconspicuous macron, facilitating recognition of the need to pronounce “thousand” separately by allowing the reader to interpret the units individually. This method is consistently preferred in all nine mixed expressions—eight of which refer to numbers over 4999, where the standard convention is to use a macron—including, curiously, 500 represented as “Fyue.C.” instead of the traditional one-character ideograph, “D”, which as an infrequently used symbol might have caused hesitation.⁸ Taken altogether, therefore, it again seems that representation of these elevated numerical values suggests that the text was indeed meant to be suitable for oral recitation.

With the exception of the above-mentioned 26,000, all of these combinations also consist of simple, two-part expressions wherein a single round number represented in Roman notation—usually a single character to represent the thousands or the hundreds place (“M” or “C”, respectively)—is complimented by a value represented orthographically in a different digit place (the ones or tens) to provide a complete integer. The exception is the first instance of the value 20,000 (l. 376), where—unlike lines 532 and 544 where the value is expressed as “Twenty.M.”—the value 20 is represented by the digraph “.xx.” with the value 1000 written out orthographically as “thousande”. Although perhaps not as frequently employed in day-to-day life as “six”, 20 was likewise a common monetary value (20 shillings in a pound, etc.), and as a score it would presumably have been as familiar as a dozen. Furthermore, the mental arithmetic required to decode the digraph is minimal—analogous to Masataka’s “‘MMMM’ (four thousand)”, which “requires one to sum up the symbols to come up with the answer” (2007, p. 282), but which ultimately is relatively simple. So again, it would seem that representing this value in Roman notation would in no meaningful way impede oral recitation of the text.

Metrical Necessity of Pronouncing Numerical Values Expressed in Roman Notation, and Possible Criteria for Selecting Orthographical or Numerical Representation

As stated above, given the complexity of some of the numbers expressed, a few of which contain four or five syllables and several words when written out in their entirety (e.g., 20,000; 26,000; 80,000; 100,000, etc.), it was initially supposed that in some instances the Roman numeral notation was to some extent “skipped over”. The use of increasingly obsolete Roman notation, combined with the inaccurate statistical reporting therein contained, might lead a reader to consider the data thus transmitted essentially irrelevant and liable to being ignored. This course of action might seem perfectly reasonable given the inherent difficulties of converting ideographically or logographically represented numbers into speech as demonstrated in the aforementioned scientific studies. Skipping over values provided numerically would dovetail with the observations of Masataka’s team, in which their subjects had to stop briefly for analysis when confronted with certain Roman numerals even after having mastered the principles of the system (2007, p. 277). It would also conform to the observations of Besner and Coltheart that “the skilled reader employs different mechanisms for processing numbers in the forms of numerals and for processing numbers in alphabetic form” (1979, p. 470), suggesting the need to “switch tack” in the middle of a task when reading out loud. Indeed, both scientific studies indicate that the brain processes such stimuli differently—Masataka’s study even demonstrating that different parts of the brain are involved—suggesting that an abrupt change in such processing would disrupt a smooth delivery. It therefore seemed that the use of Roman notation after the advent of Arabic numerals might be more appropriate for—and thus indicative of—silent eye-reading, where information need not be enunciated for its gist to be gleaned.

An analysis of the relevant lines, however, shows that such is not the case; whenever numbers or portions of numbers occur in Roman notation, the meter requires their pronunciation as if they were written out as they are elsewhere. This point was controlled by comparing the B-lines containing Roman notation with its companion B-lines in the same stanza, and also by comparing such A-, C-, D- or E-lines (etc.) with their couplet companion in order to establish the predominant beat. While there are certain irregularities in actual syllable counts, the imperfect meter in these examples is no more striking than elsewhere in the poem, and in any case it seems the total number of unstressed syllables is less important that the maintenance of three or four beats per line; indeed, in each case the meter would be even more disrupted by skipping over the values concerned rather than enunciating them. This point holds true for both the three-beat B-lines as well as for four-beat A-, C-, D- and E-lines.

In all cases, therefore, it seems that numbers were meant to be readable, regardless of the manner in which they were represented. However, since the same value could be expressed in Roman notation in one place, but written out elsewhere, it seems that it is not the case that certain types of numbers were conventionally written orthographically while others were customarily represented symbolically, as is usually the case in Present Day English (e.g., with “one” to “nine” proscriptively being spelled out and all higher numbers being provided in Arabic numerals, notwithstanding the possibility of one-word numbers such as “twelve” or “twenty” being represented either orthographically or in Arabic numerals, but consistently so in any give text). The modern convention seems to be informed by readability, where lengthy “wordy” numbers felt to be too cumbersome to write out are usually provided numerically. But although there might not have been any specific proscriptions or prescriptions due to numerical expressions substantially affecting line length, there does seem to be a certain preference analogous to today’s practice of spelling out low-value and/or high frequency numbers; such usage seems to be informed less by concern over wordiness, however, and more for consideration of how substantially “real” or “concrete” a given value might seem. With a few exceptions (which shall be addressed), it appears that numbers liable to being subitized,⁹ or accurately estimated rapidly, were considered concrete enough in meaning to be worthy of treatment as substantive lexical words in their own right, and thus spelled out orthographically, while higher values difficult to subitize came to be conceived of more abstractly, and therefore merited symbolic representation.

Irrelevance of Line Length

Regardless of the underlining criteria used to determine which system to employ, it does not seem that the physical length of the printed line has any direct bearing on the use of numerical or orthographical expression.

Long Lines

While the longest line containing a numerical expression (l. 443) does indeed make use of a Roman numeral, the shortest example does as well (l. 444)—though it ought to be added that these contiguous lines are both time expressions, which (as shall be demonstrated) appear to make regular use of Roman numerals (Table 5). Nevertheless, four of the next six longest lines with numerical expressions all contain exclusively orthographical references to numbers (Table 6: ll. 239, 308, 382, 521), while the remaining two are mixed expressions partially provided in Roman notation (ll. 256, 385). Any of these lines could have been shortened with the use of Roman numerals, though admittedly the first and third examples in Table 6 have the number at the end of the line (ll. 239, 308), the fifth and sixth at the beginning (ll. 382, 385), positions in which (as shall be demonstrated) it seems that numerical notation was forbidden.

Table 5 Longest and shortest lines with numbers

Table 6 Supplementary table of the next six longest lines with numbers in order of occurrence

Short Lines

The same lack of concern for overall line length also appears in several short lines containing numerical expressions that could have been longer had the value been expressed orthographically. While there are a few relatively short lines with numbers entirely written out, the expressions concerned always seem to occur at the beginning or end of a line, where ideographical representation is apparently not allowed (Table 7):

Table 7 Short lines with numbers spelled out at beginning or end

Elsewhere, however, lines of similar length (or shorter) are presented with Roman numerals, at least in part (Table 8):

Table 8 Short lines with mixed numerical expressions

Initial, Final and Medial Positions; Rhyme, Stress and Aesthetics

Final Position

It initially appears that numerical expressions must be written out whenever they are rhyme-words, and such a supposition would make sense given the importance the acoustic value of the words representing those numbers has in maintaining the phonological regularity of rhyme. (It also might ensure correct pronunciation in case of dialectical variation, though texts demonstrate that transcribers were not above altering an exemplar’s dialect to their own, even in rhyme-words.) It is true that in all cases where a rhyme is provided by a numerical expression, it is written out. However, the supra-segmental value of tonic stress is of equal importance in maintaining the rhythm of verse, and, as mentioned above, this component of accentuation applies to numerical expressions written ideographically or logographically as well as orthographically. While it is possible that rhyme was considered more important (or, perhaps, simply more conspicuous) than stress, it is also possible that this is largely or partly an aesthetic convention. By their nature (in Middle English verse at least) rhyme-words habitually occur at the end of a line—and in this poem, all words at the end of a line are rhyme-words—and there are no occurrences of logographical representation of numbers in this position (Table 9).

Table 9 Terminal rhyme-words always spelled out

Initial Position

Support for the point of view that numbers in terminal position must be spelled out can also be found in the treatment of numerical expressions at the beginning of lines, where rhyme does not play a role (nor, at least in this text, does alliteration). Nevertheless, all line-initial words are invariably treated specially. They are not only always spelled out, but they are always capitalized, regardless of their part of speech, whether they are common or proper, whether they begin a sentence, clause, phrase or not, and whether or not they form part of a formulaic expression, as several seem to do (e.g., initial “twenty thousand” and “five thousand” [with variants] as seen below). Much as in modern English, where numerals are proscriptively prohibited at the beginning of a sentence, it is simply by being at the beginning of a line that they receive special treatment. While this hypothesized rule holds true for all lines in the text, it can probably best be demonstrated by examining examples where the same numerical value can be represented ideographically in the middle of a line, but never at the beginning or end (Table 10):

Table 10 Numbers expressed orthographically at head or end of lines, but in Roman notation in between

Here, in the five occurrences of the value 20 (Table 10.3), the number is written out whenever it occurs at the beginning of a line, and while it never ends a line it appears in Roman notation the one time it occurs in the middle. The number “five” is provided orthographically the two times it occurs at the end of a line and all three times it appears at the beginning, but in its sole occurrence in the middle of a line it is provided in Roman notation to express the time (Table 10.1). Likewise, “six” is spelled out in its single occurrence at the beginning of a line, and though it never appears in final position, it too is provided in Roman notation the one time it occurs in medial position, also in a time expression (Table 10.2). Other variables and preferences aside, it ultimately transpires that in this text all values that are expressed both orthographically and numerically are invariably written out at the beginning and end of lines.

Medial Position

The fact that values can be represented logographically or ideographically in the middle of a line does not, however, preclude them from being spelled out in that position, as demonstrated by the 5 examples of 20,000 above (Table 10.3), in which the value “thousand” only ever occurs medially; in these examples the integer is provided orthographically three times, but is given in Roman notation only twice. This might be misleading, however, since there actually seems to be a general preference for using Roman notation when numerical expressions occur internally, especially with what Fornaciai and Park identify as “very large” numbers of 100 and over, which are “so large that the individual elements start to form a cluttered ensemble,” putting their numerosity “beyond that range governed by subitizing” (2017, p. 1). As Table 1 indicates, mid-line numbers (or their parts) are only written out 12 times (ll. 11, 113, 162, 256 (x2), 259, 335, 349, 376, 447, 454, 521), but are given in Roman notation 16 times (ll. 128, 143 (x2), 161, 335, 376, 385, 394, 407, 433, 443, 444, 447, 450, 532, 544). Furthermore, the majority of these occurrences represent the word “thousand” (eight times: ll. 128, 143, 335, 407, 433, 450, 532, 544), followed by “hundred” (three times: ll. 161, 385, 447), and then one time each for 80 (l. 143), 20 (l. 376), 10 (l. 444), “six” (l. 394), and “five” (l. 433). So while representing a number logographically or ideographically is permitted in the middle of a line, it clearly occurs much more frequently for values in the hundreds or thousands than for values in the tens or ones.

The Special Case of Hundreds and Thousands

In fact, although they only ever occur in medial position where writing out numbers is allowed, “hundred” and “thousand” are both provided logographically more often than orthographically: three out of four times for “hundred”, and eight out of 13 times for “thousand” (Table 11).

Table 11 Treatment of “hundred” and “thousand”

Such handling could be explained if, as neuropsychological studies suggest (e.g., Anobile et al. 2014, 2016; Fornaciai and Park 2017; Kaufman et al. 1949, etc.), “hundred” and “thousand” were considered large enough to be abstract and therefore warrant symbolic representation rather than treatment as substantive lexical words. None of the other numbers provided in Roman notation, all of them of lower value, are represented ideographically or logographically with such consistency. Indeed, each of the other five values concerned only occur in Roman numerals a single time, and in at least three of those cases, it seems that the semantic and/or syntactic context in which they appear triggers an exception to an otherwise regular rule requiring them to be written orthographically (i.e., in time expressions or formulae used for time expressions).

However, despite an inability to be subitized easily, “hundred” and “thousand” were nevertheless nominal round numbers commonly used as units of quantity in accounts and statistics.¹⁰ Indeed, as their use with the indefinite article “a” or “an” suggests (ll. 161, 447, 259, 433), semantically they could also be treated as discrete units of fixed measurement in the same way that 12 or 20 could be considered “a dozen” or “a score”, respectively.¹¹ While the former condition makes it natural to express them logographically with the use of an unambiguous single-letter glyph, the latter case makes it reasonable sometimes to represent them orthographically.

Observation Concerning Low-Value Numbers

Although forms such as “ii”, “iij”, “iiij” or “iv”, etc. are well attested in other (viz. earlier) romances—and such minims and combinations are presumably also producible here given the form “.vi” in line 394—in this text such representation does not usually seem possible with such low-value numbers, which are invariably provided orthographically with the notable exception of time expressions (Table 12). Although “eight” and “nine” never occur in the text, all numbers between “one” and “seven” (inclusive) do, as well as “ten”, and these numbers are invariably provided orthographically in medial position as well as at the beginning and ending of lines except when they occur in time expressions.

Table 12 Low value numbers expressed orthographically at the beginning or end of a line and internally except in time expressions

Despite the evidence that it is permissible to present numbers in Roman notation in the middle of lines, “one”, “two”, “three” and “four” are always spelled out, and this point holds true regardless of whether the number stands alone, is cardinal or ordinal, or is a holder for a digits place in the hundreds or thousands. Interestingly, these are the numbers Kaufman identified as most easily subitized (1949), and as early as 1871, it was recognized that humans have an innate ability to immediately and precisely recognize a given number of items in this range without error (Jevons 1911). Such treatment is also generally accorded to the number “five” (spelled out 5 out of 6 times), and half the time with the number “six” (which admittedly occurs only twice)—numbers less easily subitized, but still approximated accurately with diminishing precision as the numbers increase via the “approximate number system” (ANS) (Dehaene et al. 1999; Feigenson et al. 2004)—but whenever these numbers are written out they are at the beginning or end of a line. Each appears only once in mid-line position, where higher numbers are generally presented in Roman notation, but where other low-value numbers are provided orthographically.¹² In both these instances, however, they are expressed in Roman numerals, although it seems significant that in each case such Roman notation for these relatively easily subitized numbers occurs in a formulaic time expression (Table 12.5, l. 443 and 12.6, l. 394).

Time Expressions

Indeed, it is perhaps due to the nature of these numerical expressions being related to a clock—on whose dial the hours would have been conventionally indicated by Roman numerals—that the ideographic representation is made here, but there are admittedly no other mid-line occurrences of either number to use as controls for comparison. In any case, a similar treatment is given to “ten” in that value’s sole occurrence in this text where, following mention of “From.v. of the clocke” in the previous line, it likewise refers to an hour: “Tyll.x. on the other daye” (l. 444); perhaps by analogy, this treatment is then extended to the sole occurrence of “80,000,” which occurs with the same arguably formulaic structure (i.e., [prep] + [number] + [event]), but in a different context: “Tyll.lxxx.M. were layde bedene” (l. 143).¹³ Admittedly such a high-value number would probably deserve symbolic representation on its own merits, because it is large enough to surpass the discrimination threshold of Weber’s Law for numerical estimates, making it more numerically abstract rather than concretely lexical (see, e.g., Anobile et al. 2014). Given this evidence, it is tempting to presume that 10 or 12 and under—numbers either instantaneously subitized (1–4) or in the ANS range where they are “greater than the subitizing range, but sparse enough to be individuated as single items” (Anobile et al. 2016)—are more readily written out because they are in some way more tangible or concrete, and therefore conceived of more “lexically”, the exception being time expressions, presumably because the familiar image of the Roman numbers on a clock would be so intrinsically associated with giving the hour that they were retained in such situations.

Conclusions

In any case, there nevertheless seems to be a curious trend towards representing low-value numbers—especially 10 or 12 and below, which could presumably be readily visualized mentally and therefore worthy of a concrete lexical form—as orthographical words. The apparent exception to this trend is when such numbers were used in time expressions where, perhaps due to the appearance of Roman numerals on timepieces, it was deemed suitable to represent such numbers ideographically. Larger and therefore comparatively more abstract numbers that were difficult to reckon mentally, on the other hand, had a tendency to be represented wholly or partially in Roman notation, and with numbers over 100, it seems that the “heavy groups” were broken up into their components of ones and tens places separated by a punctus from hundreds and thousands places for ease of reading. While the latter groups representing the hundreds and thousands places could simply be abstractly represented with an easy-to-interpret logograph, as high frequency values capable of being expressed in a single word, they could likewise be represented orthographically, not unlike modern English.

Whatever the case, it nevertheless seems that despite expressions that would presumably take a certain amount of mental arithmetic to decipher, the values of purely ideographical, logographical or mixed expressions were decrypted rapidly into words—at least quickly enough to cause no pause in the rate of reading or recitation noticeable enough to disrupt the meter. This does not mean that silent reading did not or could not occur—it might have—but it nevertheless suggests that these texts with numbers represented ideographically or logographically as well as orthographically were still suitable for recitation. This at least indicates that to some extent basic numeracy in Roman notation was an integral component of literacy. The meter required that all the syllables be enunciated whenever a value occurred, regardless of whether those numbers appear as written words or Roman numerals.

While recognizing that the conclusions of this admittedly limited study of a single text in three imperfect sixteenth-century incunabula cannot be deemed entirely conclusive for the wider practice of Roman notation throughout the Middle English verse texts, they could nevertheless be useful for indicating trends, especially after the printing press and innovations in book-keeping had normalized the use of Arabic numbers. It might also be worthwhile to consider the extent to which trends identified here were the culmination of tendencies observed in the manuscript tradition that existed before widespread printing and innovations in accounting had popularized Arabic numbers. Indeed, the principles proposed for governing the use of Roman numerals seem to be largely if not consistently followed when a perfunctory comparison is made with the four extant manuscripts of the earlier Amis and Amiloun (dating from the fourteenth to the sixteenth centuries), and in which there is sometimes a discrepancy in the way the same numerical expression is provided in the different manuscripts. It is therefore hoped that this study can be used as a basis for further investigation into those and other texts with an eye to better elucidating authorial or scribal selection of ideographic or alphabetical representation, but ultimately corroborating the observation made here that numerical expressions were generally taken into consideration in establishing a line’s meter, regardless of the manner in which they were represented.