Analysing bilingualism and biscriptality in medieval Scandinavian epigraphic sources: a sociolinguistic approach

3.1 The interplay between language, script, content, and layout

The framework proposed here revolves around the interplay between four elements – language, script, content, and layout – and focuses on the interaction of different semiotic resources used for conveying meaning. The inscriptions are categorized according to the varying relationships that these four elements can show.

This idea stems from previous research on written multilingualism, specifically the system for categorizing multilingual texts proposed by Sebba (2013). Sebba envisions three types of characteristics that can be used to describe a multilingual text: (1) language–content relationship, (2) language–spatial relationship, and (3) language-mixing type.^[16] Language–content relationships refer to the fact that parts of texts in different languages might express the same content, partly the same content, or different contents altogether, resulting in equivalent texts, overlapping texts, or disjoint texts respectively.^[17] Language–spatial relationships refer to how sections of texts in different languages are arranged visually in relation to each other, and whether the layout indicates an equal or hierarchical structure between them. Finally, the language-mixing type refers to whether different languages are mixed together in a single grammatical, textual, or visual unit.^[18] Here, Sebba’s framework is complemented with additional parameters in order to be applied to medieval epigraphic texts and, in particular, to Scandinavian bilingual and biscriptal inscriptions. These modifications are explained below, grouped under four points.

First, an important addition is that all three of the characteristics mentioned above can be analysed not only with regard to language-switching and -mixing, as Sebba proposes, but also with regard to the use of different scripts. We can thus analyse script–content relationships, script–spatial relationships, and script-mixing types in the same way. Furthermore, the addition of a language–script relationship is in order (i.e., what script(s) are used for what language(s)?).

A second modification regards the types of texts identified among the possible language–content relationships. While Sebba defines texts where different languages convey the same content as equivalent, I prefer the term duplicating; and instead of disjoint, I use the term complementary (following Reh 2004). Further, I divide the category of overlapping texts into two subgroups, identical and equivalent overlapping texts. In identical overlapping texts, content repeated in different languages or scripts is translated literally or transliterated, respectively. In equivalent overlapping texts, content repeated in different languages or scripts is conveyed through different wordings or formulae. An example of an identical overlapping text containing transliterated content is the aforementioned grave slab from Ugglum churchyard, where the commemorative formula ‘Reginmóð had the monument made over Gunnarr, son of Ásbjǫrn’, is carved in both runes and Roman letters. An example of an equivalent overlapping text is another Swedish funerary inscription, from Sjögerås (Vg 131), now lost, where the commemorative section in the vernacular and runes contains only the name of the deceased, Þórðr, while the corresponding part in Latin and in the Latin alphabet consists of the formula Hic iacet Thordo, ‘Here lies Þórðr’.^[19]

A third addition to Sebba’s framework is the type of content expressed in different languages or scripts. Whereas a distinction between overlapping texts and complementary texts, for instance, tells us whether part of the content is repeated in two different languages or scripts, a focus on the type of content which is shared or unique can provide clues to the indexical values attached to those languages and scripts. This can be a useful variable, especially when studying formulaic texts, such as funerary inscriptions, which have recurring parts with specific contents and purposes, e.g., formulae commemorating the deceased or celebrating the sponsor, prayers and invocations, datings, signatures of the inscriber, and so on (see Section 4).^[20]

Fourth, the factors “script” and “language” are understood here not only in the sense of the graphic form of the writing system used and the idiom employed, but also as subsuming features connected to scripts and languages which are susceptible to influence in a contact situation. We should thus include two other types of relationship in the analysis, i.e., those between different scripts and between different languages. Influence in the relationship between scripts – “interscriptal relationships” – can be observed in, for instance, the shape of written signs, both in their distinctive traits and in smaller details of execution. The common linguistic levels central to studies of language contact, i.e., syntax, lexicon, morphology, and phonology, can be grouped under the umbrella of “interlingual relationships”. In addition to these levels, I also include variables pertaining specifically to written language. Particularly relevant for the study of Scandinavian bilingual and biscriptal inscriptions are influences that affect the writing systems involved, their grapho-phonological correspondences and spelling practices, as well as textual features such as the composition of monumental inscriptions and the formulae chosen. Such elements, which developed over time within a certain written culture, can be transferred to another written culture when the two meet and mix together.

An overview of the relationships between language, script, content, and layout included in the proposed framework is given in Table 1. Table 2 gives an overview of the types of content arrangement in bilingual and biscriptal inscriptions.

Some of the parameters described above – e.g., palaeographic and textual influence – can in theory also be grouped together as forms of epigraphic influence. For instance, in a study on language contact in the Iberian Peninsula, Simkin (2012) includes among the indirect sources for language contact those pertaining to epigraphy, such as various influences of one script on another, changes in grapho-phonological correspondences, the use of abbreviations, and the adoption of certain formulae (Simkin 2012: 90–96). In and of themselves, different groupings of such features have no bearing on the outcome of the analysis, but one can nonetheless observe that some of these phenomena might reflect an influence between handwritten texts and epigraphic ones, rather than between epigraphic traditions. A Scandinavian example of this would be the use of standard salutation formulae, originating in charters handwritten in the Latin alphabet, that were employed in business-related texts carved in runes on small pieces of wood (see e.g., Schulte 2012 for examples). Features that are more specifically epigraphic include the material aspects of the inscriptions, e.g., the carving technique or the way the writing surface was prepared, as well as variables that belong broadly to the writing culture or even to “more general cultural influence” (Simkin 2012: 91), such as domains of literacy as well as types of inscriptions and artefacts.

3.2 From categorizing to interpreting bilingualism and biscriptality

Crucial for the relevance of the framework described above is the fact that the variation that can be captured can be interpreted both in terms of literacy and as having a sociolinguistic function. For example, the amount and type of content expressed in different languages and scripts can give us insights into their functional distribution. Their visual arrangement can provide clues about their status relative to one another. Orthographic and palaeographic choices can be seen as having a social meaning and a role in the manifestation of identity. However, the interpretation of the features analysed is not always straightforward.

Several interpretative models of language contact in modern and historical times have been proposed, seeking to connect the presence of a certain type of multilingual text to individual and societal bilingual proficiency. For example, Reh (2004: 16, Figure 10) connects duplicating multilingual texts to individual monolingualism (or insufficient multilingualism) and societal bilingualism, while she sees complementary texts as a sign of both individual and societal multilingualism.^[21] Such interpretations take into account what Reh (2004: 8) terms the technical aspect of communication, i.e., that related to language proficiency. At the same time, however, there is also an affective aspect (Reh 2004: 8–9) pertaining to the identity and status of different linguistic communities. A given duplicating text may thus not only have been produced to respond to communicative needs and to readers’ proficiency, but also to signal equality between the languages used.^[22] On a similar note, Sebba (2013) remarks that multilingual texts may reflect not only writers’ and readers’ language competence, but also the relative status, actual or aspired to, of the languages used. It could be added that multilingual texts need not attest to writers’ and readers’ language competence at all. Drawing on their own and others’ research on modern and historical multilingualism, Pavlenko and Mullen (2015: 124–128) remind us that varying motivations can lie behind language choices, and that they may reflect writers’ aspirations rather than readers’ actual proficiency, or even stem from an intention to exclude certain groups of readers. In the Scandinavian runic corpus from the high and late Middle Ages, several examples of inscriptions in barely understandable Latin bear witness to the fact that inscribers’ lack of Latin proficiency was not a hindrance to choosing that language, and that the linguistic correctness of texts was not always a goal.^[23]

It is difficult to establish a one-to-one correspondence between different types of bilingual texts on the one hand and the level of bilingual competence among writers and readers or the indexical values attributed to the languages in question on the other. Nonetheless, the diversity of the variables considered, the interdisciplinary approach, and comparisons within the corpus and with related phenomena in other corpora can still allow us to refine and advance our understanding of the relationship between the vernacular and Latin written cultures in medieval Scandinavia.

In the next section, I will use a funerary inscription from late medieval Sweden as a case study to illustrate the type of analysis described above, as well as the results it can yield. It is true that considering a wider range of material would allow us to identify patterns against which we could compare individual inscriptions, but this kind of comprehensive analysis falls outside the scope of the present article. Such an analysis is undertaken elsewhere (Palumbo forthcoming), and observations made there on the basis of all bilingual and biscriptal funerary inscriptions of medieval Scandinavia are referred to in the conclusions drawn below.

4.1 Language/script–content relationship

A first clue to the language and script proficiency of the composer, the inscriber, and the expected readers is provided by the relationship between language, script, and content. The inscription from Ukna is an example of an almost totally complementary text, i.e., the content given in the two languages and scripts is almost unique to those languages and scripts. The only element that is repeated in both languages and alphabets is the name of the commemorated person, mentioned once in Latin and Roman letters after hic iacet, and a second time in Old Swedish and runes in the exhortation to the reader.

The complementarity of this text, and the basic fact that two languages and two scripts are used, would suggest an author or inscriber who, to some degree, had to be bilingual and biscriptal, and also a model reader who was able to understand both languages and scripts. The degree of the composer’s or inscriber’s competence is hard to assess from this one inscription, as the sequence in Roman letters is very short and the Latin text is formulaic. However, we can note that the Latin parts, even when carved following runic writing conventions (see Section 4.4 below), are rendered without mistakes, which suggests a familiarity with liturgical Latin. As to the readers, their bilingual and biscriptal proficiency is even harder to evaluate, since, as previously noted, bilingual texts are not always motivated by the wish to communicate through lexical means with a bilingual audience, something that I will return to in Section 4.2.

4.2 Language/script-mixing type

Another trait of this inscription that might suggest it was composed with a bilingual and biscriptal model reader in mind is the type of language/script-mixing, which is here partly bilingual and biscriptal. The first switch from Latin to Old Swedish and from Roman letters to runes happens inside a single sentence: ‘Here lies Turgillus, son of master Gunnmundr Gás’. The switch is not marked in the linguistic structure of the sentence; instead, it is marked visually both by the presence of one of the two crosses that frame the Roman letters, and by the change of the side on which the inscription is carved. The change from Latin to vernacular is, moreover, underlined visually by the change from letters to runes. The second language switch, on the other hand, from Old Swedish to the Latin prayer, takes place between separate grammatical and visual units.

Although the language/script–content relationship covered in Section 4.1 and the type of language/script-mixing attested on the monument from Ukna allow us to envision a reader with bilingual and biscriptal competencies, there are some additional caveats to take into account. Some frequent Latin words or formulae, such as the prayer Ave Maria, might have been memorized without necessarily presupposing any knowledge of the language, and when such words were carved in Roman letters they might have been recognized visually even by illiterate readers (Kleivane 2018). Perhaps more importantly, the choice of languages and scripts in a public text is not always motivated by the linguistic skills of the majority of readers, as already mentioned in Section 3. When it comes to the readers of the inscription from Ukna, although they must have been a varied group, it is safe to assume that many of them had limited literacy in any language and script, particularly in Latin and Roman letters.

This highlights the fact that the meaning conveyed by an inscription is not limited to its lexical content. Languages and scripts in themselves convey indexical values that those who produce such texts can exploit to both present their identity and frame their environment.^[28] In order to do so, writers do not necessarily need to master the relevant languages and alphabets. Readers do not necessarily need to understand a text’s lexical meaning fully either in order to identify its indexical meaning. This is true for both modern and historical multilingual texts.^[29]

To pinpoint what indexical values may have been attributed to the vernacular, runes, Latin, and Roman script in Scandinavian epigraphy, a comprehensive study of their use in inscriptions from various contexts and times is needed. In Section 4.1.1, I highlighted some of these functions when covering the types of content expressed in different languages and scripts. The vernacular and the runes may have indexed local identity and community belonging. Where Latin and the Roman alphabet are concerned, it appears that they indexed, among other things, religion and religious authority, learned and literate culture, secular administration and power, and internationalism in many bilingual and biscriptal funerary inscriptions (Palumbo forthcoming); some of these values must have been apparent not only to the author of the Ukna inscription and its more educated readership, but also, at least to some extent, to those readers who did not have the language and script proficiency required to access the lexical content of the text.

4.3 Language/script–spatial relationship

Another important aspect of a public bilingual and biscriptal text such as the Ukna inscription is its visuality. In this case, the languages and scripts display an asymmetrical arrangement. One thing that may immediately catch the modern eye’s attention is the amount of space given to the runes, which occupy the majority of the carving surface. What must have caught the local medieval eye’s attention, however, is the presence of Roman letters, placed prominently at the top of the slab, at the beginning of the inscription and framed by crosses. In twelfth- and thirteenth-century Sweden, the use of Roman letters in monumental inscriptions was still a rare phenomenon compared to the use of runes, and it must have stood out in the linguistic landscape of the time. Both the layout and the textual structure suggest, then, that the Roman letters, in this particular case, were accorded a higher symbolic status than the runes.

4.4 Compositional elements, spelling conventions, and palaeographic traits

Another way of evaluating the relative status of and possible influence between the two written traditions is to look at the choices made regarding textual formulae, spelling conventions, and palaeographic features.

In a formulaic genre like the funerary inscriptions, a study of the textual and compositional elements can be particularly fruitful. In Section 4.1.1, I already touched on this level of analysis when discussing the commemorative formula and expression of kinship. Here, I want to highlight an additional point regarding the commemorative formula. The phrase hic iacet was completely absent from the older Viking-Age runic memorial inscriptions. Under the influence of Latin epigraphic conventions, it started appearing in the vernacular form hēr ligger ‘here lies’ in runic inscriptions from the twelfth century on.^[30] The vernacular “here lies” formula was sometimes used as an alternative to the aforementioned traditional commemorative formula widely employed on virtually all eleventh-century runestones: “[sponsor] had this stone erected in memory of [commemorated], his/her [kinship relation]”. While the use of hēr ligger, ‘here lies’, is a sign of influence from Latin epigraphy, it would in itself not have been surprising if found in a runic inscription from the late thirteenth century. But it is plausible that the choice to maintain the wording in Latin and Roman letters in the inscription from Ukna marks a wish by the inscriber (or someone else involved in the production of this text) to identify themselves with the Latin epigraphic tradition, rather than following the older runic conventions or adapting them to the new Latin-influenced customs.

Other clues on the interplay between the Latin and runic traditions can be found in the orthographic conventions used in this memorial. Apart from the mixing of languages and scripts, we can also see a mixing of orthographic practices. On the one hand, the runic text shows some conventions which have been borrowed from Latin written culture, most notably the double-spelling of long consonants. Spellings like hærræ for herra ‘master’ were not used in runic writing in earlier periods, but they appear increasingly often during the twelfth, thirteenth, and fourteenth centuries. On the other hand, the Latin text in runes presents a series of spellings which are typical of the runic conventions and which render the local pronunciation of Latin rather than following a more literate norm. For instance, the Latin conjunction et is rendered in runes as æð, where æ is normally used for an open-mid front unrounded vowel [ɛ], and ð is a rarely used rune for a voiced dental fricative (much more commonly rendered by the rune used for its unvoiced counterpart, i.e., þ, which marks the lenition of /t/ in word-final position, a development attested in medieval Latin outside of Scandinavia as well; Stotz 1996: 228–229). Another example is a rare type of n-rune which is almost exclusively used for the phoneme /n/ either when it is long or when it precedes a dental consonant.^[31] In this inscription, we find it transliterated with a small capital n, for instance in the word ventris ventris.

Such orthophonic spellings of Latin words are generally interpreted as signs of deficient schooling in Latin, and that the author or inscriber who produced them had acquired their Latin skills by ear (see e.g., Knirk 1998: 490–491; but see Steenholt Olesen 2021: 98–101 for a partly different explanation, and a bigger emphasis on sociolinguistic aspects of runic Latin). But this might not always be the case, and it is definitely not the only conceivable explanation. Orthophonic spellings of Latin words written in runes do not just appear in texts where other, clearer signs of a lack of schooling in Latin or in Latin writing conventions are evident, such as the presence of spelling mistakes or sequences that are so corrupt that they should more appropriately be labelled as pseudo-Latin. On the contrary, the orthophonic spellings also appear in runic inscriptions that show a good command of liturgical Latin, the influence of Latin writing conventions (like the double-spelling of long consonants), and a high overall level of literacy (Palumbo 2022). In these instances, such spellings might be due to the fact that the inscribers had learned the Latin they needed through the use of the runic writing system rather than through the Latin one, or that they were aware of the spelling practices and conventions used in both writing systems but chose to stick with the runic ones even when carving in Latin.

The inscription from Ukna is a case in point. Its composer and/or inscriber shows not only a very good command of the runic writing system, but also competence in Latin and in the Roman alphabet. Moreover, they distinguish clearly between the orthographic conventions employed in the two written traditions. The Latin in runes is faultless, apart from the orthophonic spellings mentioned above. On the other hand, such spellings are absent from the Latin written with Roman letters. For example, the fricative allophone of /g/, [ɣ], is in medieval runic inscriptions normally rendered with an h-rune, as in the genitive form of the name Þórgíls, i.e., Þórgílsar, rendered here as þyrhilsær. In Roman letters, the same name is written turgillus, i.e., with a g. Also, the differentiation between the velar stop [k] and the affricate [ts] (or perhaps the fricative [s]) is represented orthographically in the runic inscription by the use of different runes, as in tekum tecum and grazzia gratia, whereas no such distinction is made between hic and iacet. Given these differences, it is even more striking that this inscriber not only does not follow a more literate norm in the Latin in runes, which they could have done, but also employs rare runic innovations – the aforementioned runes ð and n  – that do not have any counterpart in the contemporary Latin-alphabet inscriptions, which makes the choice of them appear very deliberate. A possibility that was hinted at before (see Footnote 27) is that these orthophonic spellings of Latin passages may have had a performance-related purpose, allowing those who could not read Roman letters or Latin to say the prayer nevertheless, simply by reading the runes.

The Ukna inscription thus shows a consciousness in the choice of writing conventions that speaks against deficient literacy skills. On the contrary, choices such as those made here can be interpreted as having a pragmatic and communicative function, and a social function as well, playing a role in manifesting the identity of the composer, the inscriber, or the commemorated person. Rather than blending the conventions of two written traditions, and perhaps adapting the runic ones to the Latin ones, the inscriber uses them both in a parallel fashion when deemed appropriate to what they wanted to convey.

Lastly, a few brief observations can be made about the palaeographic traits in this inscription. One regards the shape of the Roman letters, which appears to be slightly less uniform than other late medieval biscriptal inscriptions – and maybe even more so when compared to Latin monoscriptal epigraphs – for instance with regard to height, spacing, and placing in the frame.^[32] Conversely, the runes are quite uniformly inscribed. This difference may indicate that the inscriber was more accustomed to carving runes than Roman letters. Moreover, the runes do not show any sign of palaeographic influence from the Latin alphabet, for instance with respect to their shape or ornamental details such as serifs.^[33] This lack of influence on the palaeographic level indicates a high degree of independence in the runic epigraphic tradition, more than two hundred years after the introduction of Latin epigraphy in Scandinavia.