Portal de Periódicos da CAPES

A Quantitative Analysis of Toponyms in a Manuscript of Marco Polo’s Devisement du monde (London, British Library, MS Royal 19 D 1)

2017; University of Chicago Press; Volume: 92; Issue: S1 Linguagem: Inglês

10.1086/694170

2040-8072

Mark Cruse,

Historical and Archaeological Studies

Previous articleNext article FreeA Quantitative Analysis of Toponyms in a Manuscript of Marco Polo’s Devisement du monde (London, British Library, MS Royal 19 D 1)Mark CruseMark Cruse Search for more articles by this author Mark Cruse, Arizona State University ([email protected])PDFPDF PLUSFull Text Add to favoritesDownload CitationTrack CitationsPermissionsReprints Share onFacebookTwitterLinked InRedditEmailQR Code SectionsMoreMarco Polo’s Devisement du monde (Description of the world) is widely recognized as one of the most significant works of the European Middle Ages. It was the first account of travel into Asia composed by a layman in the vernacular, and the longest such account when it appeared. Polo spent more time in the East than any previously known European, his access to the Mongol elite was unprecedented, and he provided a wealth of observations on Eastern peoples, customs, plants, animals, and many other phenomena. Polo’s work is the first in Europe to mention Indonesia, Japan, Madagascar, Sri Lanka, and numerous cities in China and elsewhere and thereby represents an extraordinary contribution to Western geographical knowledge. The work’s survival in 141 manuscripts and fragments in Latin and various European vernaculars, the wide range of patrons and owners of these books, and its early success in the print era all attest to the account’s appeal. Its profound influence may be gauged too in the texts and maps produced throughout Europe that drew on or cited it as an authority in the late Middle Ages and well beyond and in the many abridgments, translations, and editions made of it in the centuries since.1However, as scholars have long noted, Polo’s account presents many authorial and textual problems.2 The text says that Polo dictated his account while in prison in Genoa in 1298 to a man generally identified as Rustichello da Pisa, but how exactly this process unfolded and how the text was transmitted thereafter remain vexed questions. It is moreover unclear to whom the narrative voice belongs in many passages, a lack of clarity that not only can be confusing but also undermines the text’s claims to present the “pure truth” of what Polo himself saw or heard from trustworthy sources. There is also the major question regarding which version of the text is most authoritative. There is no holograph or definitive urtext, and scholars are not sure which manuscript is the oldest. Pride of place is generally given to manuscript F (Paris, Bibliothèque nationale de France, MS fr. 1116), a Franco-Italian copy seemingly produced 1310–20 in Venice that is close to the progenitor, or progenitors, of many later copies.3 Yet the Latin manuscript Z (Toledo, Toledo Cathedral, MS Zelada 49.20) contains passages not found in F that are also considered authentic, which has led to speculation that Polo or others expanded the text early in its history.4 As with other influential and widely disseminated medieval works, there is no definitive Polo text.Like much recent scholarship, the present study is concerned with the problem of textual instability in Polo’s account. However, my interest is not in authorship, narration, or the establishment of an authoritative text, but rather in the high degree of variance among the toponyms in one of the earliest copies of Polo’s account. Toponyms posed a particularly great challenge to the scribes who copied the early Polo manuscripts because so many of these words were unknown. They appeared in no encyclopedia and almost certainly on no map—they were words from nowhere. As a result, a scribe who encountered variant forms of an unknown toponym in his exemplar had no way to render it consistently with any confidence, as he could not know which form was correct. Moreover, if he could not make out a toponym’s lettering, or found the spelling orthographically or phonetically incongruous, or could not distinguish between the preceding article or preposition and the toponym itself, a scribe might easily change the word’s form, consciously or not. Normally, scribes used context and their own linguistic knowledge to decipher words, but neither was of any help with unknown proper names, which functioned more like neologisms or, often, hapaxes. As a result, inconsistencies in the spelling of toponyms could be transmitted from exemplar to copy, and small changes in the spelling of a toponym could accumulate quickly over the course of the text, such that a single toponym might have several forms that are so diverse as to render their resemblance invisible, especially when more than one scribe was involved. The variance may be so great that it is impossible to infer that the same place is being discussed in different passages.As I hope to show, the variance in the Polo toponyms invites analysis that draws on codicological, philological, and quantitative approaches. Examination of the oldest Old French copy of the account shows the scribal uncertainty that attended the copying of the toponyms. By distinguishing between familiar and unfamiliar toponyms, by quantifying the number of variants and the degree of orthographic and phonetic variance for each toponym, and by assigning the occurrences to specific scribes, we may form a precise picture of the nature, extent, and perhaps the reasons for this variance. Unfamiliarity, combined with the problems of letter recognition and word separation in textualis script, led to a high percentage of variance and to many kinds of variants. Yet rather than see these variants as errors, we should analyze them as forms of reader response. An analysis of these toponyms in their manuscript context, devoid of annotation, enables us to encounter Polo’s text as the strange and complicated work that it was to its earliest readers. Ultimately, the ways in which scribes responded to the unfamiliar toponyms in Polo’s account are of interest because the scribes’ choices reflect not only scribal practice but also the processes by which the new geographical information in Polo’s text was absorbed by medieval readers.London, British Library, MS Royal 19 D 1 and the Problem of Toponymic VarianceThe eighteen Old French manuscripts and fragments of Polo’s account preserve one of the earliest versions of the text.5 The oldest of these manuscripts is London, British Library, MS Royal 19 D 1, which dates to 1336, only twelve years after Polo’s death, and was produced for King Philip VI of France (reigned 1328–50).6 BL Royal 19 D 1 is a compilation of historical and travel texts seemingly prepared to aid Philip VI and his council as they planned a crusade, which in the end was never launched.7 Between 2001 and 2009, a team under the direction of Philippe Ménard edited the Polo text in BL Royal 19 D 1 and published it in six volumes. The editors chose this copy because of its age, proximity to the source, and linguistic conservatism.8 The Old French version of Polo’s text was gallicized over time; BL Royal 19 D 1 preserves an early stage of the text’s transition from Franco-Italian to Old French, although the editors seem to consider it a mildly gallicized copy of an already translated exemplar rather than an extensive translation in its own right.9 Nothing is known of this exemplar, but it is possible that BL Royal 19 D 1 descends from a translation of the manuscript given in 1307 by Marco Polo to Thibaut de Chepoy, envoy of Philip VI’s father Charles of Valois.10A monumental achievement, the Ménard edition is also an example of the approaches that have long governed the editing and study of Polo’s text. The edition draws on the work of generations of philologists and historians who have sought to trace Polo’s place names back to their source words in Arabic, Chinese, Mongolian, Persian, and other languages and thereby to identify precisely what sites were meant.11 In this way, scholars hoped to reconstruct Polo’s route, to evaluate the veracity of his account, and to make his text accessible.12 And indeed, almost all the toponyms have been traced to original sites.13It is obvious that this etymological analysis of toponyms is valuable and necessary, but it also has important implications for how we understand the medieval reception of Polo’s account. An example is the rubric for chapter 71 in BL Royal 19 D 1, which states that this chapter is about the kingdom of “Erguuil” (Fig. 1). This is clearly the intended spelling—the final minim before the final l has the superior oblique stroke indicating the letter i. Further in this chapter, the word appears again (Fig. 2). This time there is no superscript stroke to indicate the letter i, only five sequential minims that appear to be connected along the baseline. Here it is not clear how the word should be rendered, a typical problem with textualis script. However, despite the clarity of the first reading and the imprecision of the second, in the Ménard edition the word is rendered as “Erguiul” in both cases, with no note indicating a variant or a vexed reading. In the glossary, “Erguiul” is defined as the “Mongol name for Liangzhou, a kingdom and city in China, today known as Wuwei (Gansu).”14 Thus is this toponym regularized and linked to a modern site.15Fig. 1. “Erguuil” in the rubric to chapter 71 of Marco Polo’s Devisement du monde, British Library, MS Royal 19 D 1, fol. 80r.View Large ImageDownload PowerPointFig. 2. “Erg?l” in chapter 71 of Marco Polo’s Devisement du monde, British Library, MS Royal 19 D 1, fol. 80r.View Large ImageDownload PowerPointI choose this example not to criticize the editorial work but to point out the complexities of rendering certain toponyms in this manuscript. To their credit, the editorial team preserved the toponymic variants in almost all cases.16 Another goal of this edition is to “define” these toponyms in the index by linking them to specific modern sites. The index presents Polo’s account as a text that imparts independently verifiable geographical knowledge—a transparent, indeed modern, representation of the world. Yet this approach endows the text with a referential clarity that it did not possess for its earliest readers. King Philip VI likely had no map of China, nor is it likely that any of his advisers had been further east than Armenia.17 In BL Royal 19 D 1, as in other early manuscripts, Polo’s account forced Europeans to confront their profound lack of geographical, environmental, and ethnographic knowledge. We have grown accustomed to thinking of Polo’s text as representing a step forward in Europe’s understanding of the wider world, when in fact the high rates of variance among the account’s toponyms also reflect the geographical, cartographic, and linguistic ignorance of its early readers.It could be argued that the variance among the toponyms in BL Royal 19 D 1 is to be expected in a nonstandardized spelling system, such as that of Old French, and that study of this variance therefore contributes little to Polo scholarship. After all, medieval scribes could have many different ways to render the same word: A Linguistic Atlas of Late Mediaeval English records sixteen spellings of the first-person personal pronoun and more than five hundred of the preposition “through,” both of which were undoubtedly familiar to scribes.18 The same scribe may even render the same word or phoneme with different spellings from one iteration or line to the next. As Margaret Laing argues, such litteral substitution—the representation of the same sound with different letters—was normal, and indeed systematic, and is only perceived as strange or deficient because of the biases of modern scholars.19Litteral substitution was possible because scribes and readers knew the language they were reading and, if they could not immediately identify a word, could refer to their lexicon to decide which word or phoneme was intended.20 However, as the analysis below will show, the majority of toponyms in Polo’s account were neologisms that were not in any European reader’s lexicon, because they had never appeared in any source known to European readers. This means that readers of Polo’s text could not know whether a toponym was correctly spelled because they had no prior knowledge of the word. Indeed, in the case of neologistic toponyms, they had no way of knowing if the word before them was a word at all—if it had any currency or even existence beyond the page before them. Moreover, to change the spelling of a toponym, whether with the intention of representing the same sound or not, could in fact create a new toponym in the reader’s mind. A scribe could inadvertently add, or appear to add, a new site reference to Polo’s text.The same problem faces modern editors as faced medieval readers who encountered unknown toponyms. As the Conseils pour l’édition des textes médiévaux notes, “There exist cases where one cannot be certain that a toponym is intended as such: ‘un camp a la borde de Fay’ [a field at the Fay farm] or ‘un camp a La Borde de Fay’ [a field at La Borde de Fay]? The editor will rely on other passages in the text or on what he has observed of the usages of the region and period.”21 However, if orthographic variance means there is no observable pattern to the toponym’s form, or if the word does not belong to the known lexicon of a region or period, there is no way to reconstitute the toponym with any certainty. Toponymic variance in most European sources was presumably not a problem for medieval readers because they were usually familiar with the places in question. Geographical and lexical knowledge and textual context were sufficient to allow readers to recognize, for example, that “Traiecto,” “Treiecti,” and “Uttret” all represented the city of Utrecht.22 But such geographical and contextual knowledge did not apply to many of the toponyms in the Polo text, which is why generations of scholars have had to engage in complicated, multilingual, comparative research to clarify which places Polo intended.23While we have no evidence of whether King Philip VI and his councillors read Polo’s text in BL Royal 19 D 1, we can examine how other readers—the text’s two scribes—reacted to the neologistic toponyms. Not just mechanical copyists, medieval scribes were also textual interpreters, particularly when they confronted words about which they were unsure or that they did not know.24 In BL Royal 19 D 1, as in other copies of Polo’s account, we can see how the scribes grappled with these unknown toponyms, which clearly posed enormous interpretive challenges. Granted, it may be that the toponymic variance stems from BL Royal 19 D 1’s exemplar, in which case the scribal behavior at issue is not that of these but of earlier scribes. Nonetheless, by acknowledging and quantifying orthographic and phonetic variance instead of editing it away, we may gain a clearer picture of how Polo was read and interpreted by his earliest readers. We may strip away these toponyms’ strict referentiality and discover other meanings and functions. We may develop a more detailed picture of how far Polo’s text diverged from the geographical texts and the maps of its time, and thereby see more clearly the processes by which its information was assimilated into European culture.Quantifying Scribal Uncertainty in BL Royal 19 D 1My analysis of the toponyms in this copy of Polo’s account began with the transcription into Word and Excel of every occurrence of every toponym from the high-resolution scans of BL Royal 19 D 1 on the British Library website. I determined that 249 different places are named in this copy of Polo’s account. This list does not include words for directions, which may at times represent geographical zones; or for generic places, such as “the islands of India”; or for most monuments, such as the tomb of the khans. I then divided these toponyms into two groups: one of toponyms that occurred only once in the text, the other of toponyms that occurred more than once (Table 1). The toponyms that occur only once I discarded because they cannot be subjected to variant analysis. I then took the 162 toponyms that occurred more than once and classified them according to whether they had variants (Yes/No) and would have been familiar to the scribes (Yes/No/Maybe). I based this classification primarily on Pelliot’s Notes on Marco Polo, which lists many (though not all) variants for Polo’s toponyms and many of the European and Asian sources in which they appear. Toponyms for European locations or for extra-European locations mentioned in Western sources were tagged “familiar” or “maybe familiar.” For example, “Abasie,” “Acre,” and “Alixandre” I tagged as familiar, given their prevalence in encyclopedias, chronicles, romances, and on maps, while “Badasiam” I tagged as “maybe familiar” because, although not prevalent, it was from c. 1200 the source of Latin and vernacular words for “ruby,” and because the land is mentioned in Hayton’s Fleur des histoires of 1307.25Table 1. Toponym Occurrence (Devisement du monde, British Library, Royal 19 D 1)Number of Discrete ToponymsMore Than One OccurrenceOne Occurrence249162 (65.1%)87 (34.9%)View Table ImageThe issue of familiarity is crucial because we would assume that toponyms that were known to scribes would have a lower rate of variance than unfamiliar toponyms and would therefore provide a control group. I did not control for variance across the entire text or across sample excerpts of nontoponyms because my focus is variance as it correlates to the familiarity of toponyms only. Of the 162 toponyms occurring more than once, 121 (75 percent) are not familiar, twenty-five (15 percent) are familiar, and sixteen (10 ten percent) are “maybe familiar” (Table 2). Among the “not familiar” group, seventy-three (60 percent) possess at least one variant and forty-eight (40 percent) possess no variants (Table 3). Of the twenty-five familiar words, eight (32 percent) possess at least one variant and seventeen (68 percent) possess no variants (Table 4). Among the “maybe familiar” group, eight (50 percent) possess at least one variant and eight (50 percent) possess no variants. As we would expect, the majority of unfamiliar toponyms have variant forms, while the majority of familiar toponyms do not. In statistical terms, these percentages show that unfamiliarity and variance are positively correlated: an unfamiliar toponym is more likely to have a variant than a familiar or maybe familiar toponym.Table 2. Familiarity of Toponyms with More Than One OccurrenceNumber of Discrete ToponymsNot FamiliarFamiliarMaybe Familiar162121 (75%)25 (15%)16 (10%)View Table ImageTable 3. Variance of Unfamiliar ToponymsNumber of Discrete ToponymsWith VariantsWith No Variants12173 (60%)48 (40%)View Table ImageTable 4. Variance of Familiar ToponymsNumber of Discrete ToponymsAt Least One VariantNo Variants258 (32%)17 (68%)View Table ImageWhile correlation does not prove causation, my hypothesis is that the correlation between unfamiliarity and variance is causal and not random. We can refine this analysis further by noting that the forty-eight “not familiar” toponyms with no variants almost all adhere to the same placement pattern: in thirty-three of the forty-eight cases (69 percent), the occurrences all appear on the same folio, while in fourteen cases of forty-eight (29 percent), they appear on adjacent folios (Table 5). There is only one outlier. What these numbers suggest is that scribes rendered unfamiliar toponyms consistently when the occurrences were clustered and the word was therefore visible or fresh in memory. If we exclude these proximate occurrences, the variance rate among the “not familiar” toponym group effectively increases to 100 percent; that is, every “not familiar” toponym that is copied more than once, save one, has at least one variant.26Table 5. Placement of “Not Familiar” Toponyms with No VariantsNumber of Discrete ToponymsAll Occurrences on Same FolioAll Occurrences on Adjacent FoliosOccurrences More Than 2 Folios Apart4833 (69%)14 (29%)1 (2%)View Table ImageOf the 162 toponyms occurring more than once, eighty-nine (55 percent) have one or more variants and comprise the Variant Group that is the focus of this analysis (Table 6), while seventy-three (45 percent) have no variants and are therefore discarded. Within the Variant Group, seventy-three (82 percent) toponyms are unfamiliar, eight (9 percent) are familiar, and eight (9 percent) are “maybe familiar” (Table 7), which would seem to provide further confirmation of the correlation between lexical novelty and variance. The data set indicates the following for each toponym in the Variant Group: the first occurrence of the word, the number of times it appears, the hand, and the number of variants it possesses.27 Thereafter the data set indicates the variants (first variant, second variant, etc.) and their characteristics: frequency, hand, Levenshtein distance from first occurrence, and phonetic variance ratio (the number of sounds that differ from the first occurrence). The variants are numbered according to the order in which they appear in the text: the “first variant” is the first to appear after the first occurrence of the word, the next variant form is the “second variant,” and so on.Table 6. Variance of Toponyms with More Than One OccurrenceNumber of Discrete ToponymsWith One or More Variants (“Variant Group”)With No Variants16289 (55%)73 (45%)View Table ImageTable 7. Variant Group: Degree of FamiliarityNumber of Discrete ToponymsNot FamiliarFamiliarMaybe Familiar8973 (82%)8 (9%)8 (9%)View Table ImageThe Variant Group consists of 240 discrete toponyms. Of these, as noted earlier, eighty-nine (37 percent) are first occurrences, which leaves 151 toponyms (63 percent) that are variants (Table 8). These percentages maintain when we broaden the analysis to include all occurrences of toponyms in the Variant Group. There are a total of 616 occurrences of toponyms in the Variant Group, of which 220 (36 percent) are first occurrences or their identical repetitions, and 396 (64 percent) are variants or their identical repetitions (Table 9). Fifty-three (59.6 percent) of the eighty-nine toponyms in the Variant Group have only one variant (Table 10), which may at first seem to indicate a relatively low frequency of variance. Yet perhaps the most striking ratio in the data set is that between the total number of first occurrences (220) and the total number of first variants (252) (Table 11). Keep in mind that the number of first variants could be as low as eighty-nine (the number of discrete first occurrences). What the first variant group demonstrates is the considerable instability of these toponyms. The scribes did not devise consistent renderings of these toponyms but instead were more likely to depart from an initial form than to repeat it. Another way to make this point is to look at the ratios of first occurrences to variants. As we see with the total number of first occurrences and first variants, the ratio (220:252) is nearly one-to-one, with a slight advantage to the first variants. But the ratio of first occurrences to variants across the entire variant group is nearly one-to-two in favor of the variants, and this is true whether we mean discrete toponyms (89:151) or the entire universe of toponyms (220:396) in the Variant Group. These ratios show that the scribes, after copying a first variant, frequently favored the first variant form or went on to produce other variants. Another way to say this is that, when copying a toponym for a second, third, or nth time, scribes were almost twice as likely to use a variant form as they were to repeat the first-occurrence form.Table 8. Variant Group: Number of First Occurrences and of VariantsNumber of Discrete ToponymsFirst OccurrencesVariants24089 (37%)151 (63%)View Table ImageTable 9. Variant Group: Number of First Occurrences and Their Identical Repetitions vs. Number of Variants and Their Identical RepetitionsNumber of ToponymsFirst Occurrences and Identical RepetitionsVariants and Identical Repetitions616220 (36%)396 (64%)View Table ImageTable 10. Variant Group: Variants per ToponymNumber of ToponymsWith Only 1 VariantWith Only 2 VariantsWith Only 3 VariantsWith Only 4 VariantsWith Only 5 VariantsWith Only 6 VariantsWith Only 7 VariantsWith Only 8 Variants8953 (59.6%)24 (27%)7 (7.9%)2 (2.2%)1 (1.1%)1 (1.1%)0 (0%)1 (1.1%)View Table ImageTable 11. Variant Group: Total Number of First Occurrences and of First VariantsNumber of ToponymsFirst Occurrences and Identical RepetitionsFirst Variants and Identical Repetitions472220 (46.6%)252 (53.4%)View Table ImageThis discussion of variant forms raises two important issues: scribal hand distribution and degree of variance. The Polo text in BL Royal 19 D 1 occupies folios 58r to 135r (quires 9 to 19). Of the seven hands found in the entire manuscript, Hands 2 and 3 are found in the Polo text. Hand 2 is responsible for the majority of the text: folios 58r to 85v (quires 9 to 12) and folios 118r to 135v (quires 17 to 19). Hand 3 copied folios 86r to 117v (quires 13 to 16; beginning of chapter 81 to end of chapter 156).28 Because of the structure of Polo’s account, which proceeds sequentially from place to place, only four toponyms in the Variant Group were copied by both scribes.29 There is an almost perfect separation of word distribution between the two, which simplifies analysis of their approaches. What we see is that the proportion of variant forms employed by the two scribes is almost exactly the same. Of the eighty-nine first-occurrence forms in the Variant Group, Hand 2 copied fifty-two (58.4 percent) and Hand 3 thirty-seven (41.6 percent), which correlates almost exactly to the number of folios each copied: Hand 2 copied forty-six (59 percent) of seventy-eight folios, Hand 3 twenty-two (41 percent). The similarities between these scribes extend to the entire range of discrete toponyms that each scribe copied (Table 12). Of the 139 discrete toponyms copied by Hand 2, fifty-two (37.4 percent) are discrete first occurrences, and the remaining eighty-seven (62.6 percent) are discrete variants. For Hand 3, of the 103 discrete toponyms copied, thirty-seven (35.9 percent) are discrete first occurrences and the remaining sixty-six (64.1 percent) are discrete variants. There is, in other words, no significant difference between the rates of variance in the toponyms copied by these two individuals.Table 12. Comparison of Discrete Toponym Distribution for Hand 2 and Hand 3 Hand 2 (Out of 139 Toponyms Copied)Hand 3 (Out of 103 Toponyms Copied)Number of Discrete First Occurrences52 (37.4%)37 (35.9%)Number of Discrete Variants87 (62.6%)66 (64.1%)View Table ImageThe data set also allows us to scrutinize the degree of variance within the Variant Group. As noted above, litteral substitution is typical scribal behavior and does not necessarily suggest uncertainty about a word. However, as pertains to toponyms, if the spelling differences represent the same sounds but are nonetheless significant, or if the spelling differences also entail phonetic changes, we should look more carefully at the degree and nature of the discrepancy. Spelling and phonetic changes are especially important to this discussion because toponyms with different orthographic and, especially, phonetic renderings may suggest that the scribe (or exemplar) is improvising readings and unable to decide upon a correct or consistent form. At the extreme, the scribe may be unaware that the variants are in fact variants of the same word and not distinct words. In short, the degree of variance between first occurrences and variant forms may offer a clearer picture of the degree of scribal uncertainty than a simple tally of variants.I calculated orthographic variance according to a standard Levenshtein distance metric, which counts the number of deletions, insertions, or substitutions needed to transform one word into another. The source string was always the first occurrence, to which each variant was compared as the target string. For example, “Bangala” is a first-occurrence form. Here, each letter position is indicated by a superscript number: B1a2n3g4a5l6a7. The first variant of this toponym is Balanga = B1a2l3a4n5g6a7. Positions 3, 4, 5, and 6 are different, thus the Levenshtein distance is 4. The distance between “Aucata” and “Caca” is 3 (two deletions plus one substitution), and so forth. I used the same approach to calculate phonetic variance, but this time by counting the difference in phonemes instead of individual letters: B1an2g3a4l5a6 / B1a2l3an4g5a6. Positions 2, 3, 4, and 5 are different, so the phonetic variance ratio is also 4. Of course, orthographic change does not always correspond to phonetic change—the examples of “Caciauf/Casiauf” and “Singuy/Synguy/Cynguy” show only litteral substitution. Among first variants, the majority of words (56.1 percent) are different from the first occurrence form by only one letter (Table 13). We might therefore assume that there is also little phonetic change, and thus a low degree of instability.Table 13. Levenshtein Distance between First Variants and First OccurrencesNumber of Discrete ToponymsDistance of 1Distance of 2Distance of 3Distance of 4Distance of 58950 (56.1%)20 (22.5%)11 (12.4%)7 (7.9%)1 (1.1%)View Table ImageWe might also assert that phonetic change is a more certain measure of scribal uncertainty than is orthographic variance, because it suggests complete unfamiliarity with a word. Here we come to another significant set of data points, for although the majority of first variants differ in only one letter from their first-occurrence parents, seventy-two (81 percent) of these eighty-nine first variants exhibit at least one phonetic change. In other words, although only one letter is altered in the majority of these first variants, in most cases this letter change also results in a different pronunciation. Across the e