1. INTRODUCTION
In languages with contrastive tone, verse, in particular rhyming verse, can be a challenge. When tone is not an issue, only the segmental material and stress of the word final rhymes have to match, but in languages with lexical tone there is a further aspect to consider and poets make use of it in different ways. Middle Chinese poetry (with four tones) in the Tang dynasty (7th–10th century), for instance, was highly regulated with a rigid poetic style. In the four canonical schemes of heptasyllabic and pentasyllabic lines, pairs of lines ended with alternating tonal patterns, first with an ‘oblique’ tone and then with an ‘even’ tone (Downer & Graham Reference Downer and Graham1963; Chen Reference Chen1979). Nevertheless, alternating lines did not necessarily match in tone perfectly because only the first tone belonged to the oblique category (considered to have steady state pitch), while all the others (falling and/or rising contours) belonged to the even category (Chen Reference Chen1979). Thus, in a quatrain, the even lines all end in an oblique tone, but may not have the same tonal melody (i.e. it could be falling or rising). Another method of tonal use in poetry is discernible in Shona, where downdrift, a phonological device of tonal steps, is significantly exploited. A line in modern Shona poetry offers a stretch, typically pronounced as one breath group which is characterized by downdrift (Chimundu 1989).
Within the Scandinavian family, certain Norwegian and Swedish dialects have a phonemic contrast in tone, labelled Accent 1 and Accent 2. While examining the role of Norwegian tonal accents in poetry (trochaic tetrameters in Ibsen's Peter Gynt and three other poems by Welhaven, Bjørnson and Hamsun), Jacobson & Sommerfelt (Reference Jakobson and Sommerfelt1979) conclude that the poets make noticeable use of the contrast between Accent 1 and 2 in expressing a variety of styles and emotions. For instance, in the hero's playful dialogue with his mother in Ibsen's Peer Gynt, the variation of tones in rhyming words sometimes acts like a ‘pun-like accompaniment of whimsical motifs’ (Jakobson & Sommerfelt Reference Jakobson and Sommerfelt1979:181), e.g. ˈmorgen2 ~ ˈsorgen1 ‘morning’ ~ ‘sorrow’, ˈkjerren2 ~ ˈmerren1 ‘cart-def.sg’ ~ ‘mare-def.sg’, ˈbukken1 ~ ˈrukken2 ‘ram-def.sg’ ~ ‘line-def.sg’.Footnote 1 When she in turn expresses a ‘challenging menace’ (Jakobson & Sommerfelt Reference Jakobson and Sommerfelt1979:181) to her son, the accents of the final downbeats are subtly changed. For instance, on the fourth, most offensive line, Ibsen uses only Accent-2 words, thus ending with an Accent-2 downbeat (4th line fakter2 ‘gestures’), while the surrounding lines always end in Accent-1 downbeats (1st line på 1 ‘to’, 2nd line få 1 ‘receive’, 3rd line ˈakter 1 ‘stern’, 5th line grant 1 ‘plainly’). Again, the tonal contrast is used to increase the emotional expressiveness.
Noticeably however, although Modern Swedish does have tonal contrast, matching tonal accents in verse has never been established as a general rule. Well known Swedish poets like Carl Michael Bellman or Hjalmar Gullberg ignore accents for the purposes of matching rhymes – cf. ˈvänner2 ‘friend-pl’ ~ ˈkänner1 ‘feel-pres’; ˈTristan1 (proper name) ~ ˈlistan2 ‘list=def.sg’.Footnote 2 Even seventeenth-century poetic scholars like Arvidi mention the words ‘tone’ and ‘accent’ while discussing poetic principles, but these terms relate only to quantity or stress (cf. Kock Reference Kock1878, Frankel Reference Frankel1999 and references therein). For instance, in his rhyming dictionary, Arvidi (Reference Arvidi1651) has no qualms about grouping together words with different accents as in ˈhandel1 ‘commerce.sg’ ~ ˈmandel2 ‘almond.sg’ or ˈalen1 ‘alder=def’ ~ ˈgalen2 ‘madadj’. Kornhall (Reference Kornhall1994) examines the accent patterns in rhymed verse composed by two 20th-century Swedish poets, Bertil Malmberg and Erik Axel Karlfeldt. Kornhall suggests that although these poets appear to unconsciously use the accents to form certain accent patterns, they do not systematically match the accents of rhyming pairs.
One figure in 18th-century Swedish literature, however, did pay attention to tonal contrasts in poetry. Anders Nicander (1707–1781), although a rather inconspicuous figure in Swedish literature, is of crucial importance for the study of Swedish tone. He was one of very few people who had the instinct of an orthoepist while composing poetry himself. Even though verse is often taken as evidence for metrical stress,Footnote 3 Nicander's poetry has the added advantage that we know that his rhymes were written with an awareness of prosodic factors other than stress. Furthermore, not only was he conscious that tones ought to be matched in a perfect rhyme, his poetic manual describes in great detail what may or may not be rhymed (Nicander Reference Nicander1737). Kock (1878: 23f.) drew attention to Nicander's observation that words like [taːlən] and [gaːlən] are not acceptable rhymes because their accents differ, in spite of obvious segmental overlap. These words reflect the basic surface tonal contrast between a monomorphemic disyllabic word (ˈgalen2 ‘mad’) and a monosyllabic stem plus the definite plural ending (ˈtal=en1 ‘number=def.pl’) in modern Standard Swedish.Footnote 4
In this paper, we analyze Nicander's theoretical treatise on verse as well as his poetry, to shed light on the conflict between lexical tone and segmental material in 18th-century Swedish verse, and to ascertain to what extent today's accents have remained unchanged. In what follows, we will present an analysis of Nicander's manual and verse in three sections. Section 2 provides a brief description of contemporary Swedish tonal word accents. In section 3, we discuss Nicander's theory of rhymes and accents, and in section 4, we compare the conclusions we have drawn from Nicander's theoretical remarks to the rhymes in his own verse. We are primarily interested in the following questions: How well established was the tonal opposition in 18th-century Swedish? Did lexical accent in Nicander's Swedish differ from modern Swedish? In particular, did clitics and inflectional suffixes behave differently with respect to accent? An examination of Nicander's verse and theoretical approach to poetry and end-rhymes should provide an insight into the tonal system of 18th-century Swedish, and by comparing this to the modern Standard Swedish system, we could then draw conclusions concerning how the tonal system has changed in the last three centuries.
2. ACCENT ASSIGNMENT IN MODERN SWEDISH
Like all Scandinavian dialects with a phonemic tonal contrast, Standard Swedish differentiates between Accent-1 and Accent-2 words. We briefly sketch a few morphophonological approaches relevant to analysing Nicander's poetry. Until recently, Accent 2 was generally assumed to be the lexically specified accent (cf. Gussenhoven & Bruce Reference Gussenhoven and Bruce1999, Riad Reference Van Der Hulst2003 and references therein; also Kristoffersen Reference Kristoffersen2000 for Norwegian). According to Gussenhoven & Bruce (Reference Gussenhoven and Bruce1999), the accent contrasts arise from the differences in tonal alignment, where Accent 2 is H*L with a lexical high tone aligned to the stressed syllable, and Accent 1 is HL* with the lexical low tone aligned to the stressed syllable. In this approach, morphemes can either be Accent-2 inducing or neutral. A privative view of the accent contrast was first advocated in Elert (Reference Elert1963), claiming that Accent 2 is the lexical tone while Accent 1 is the default. This approach has been maintained in Riad (Reference Van Der Hulst2003), where stems and morphemes can bear lexical Accent 2, marked with a floating H.
Accepting the view of a privative contrast, Lahiri, Wetterlin & Jönsson-Steiner (Reference Lahiri, Wetterlin and Jönsson-Steiner2005a), nevertheless claim that Accent 1 is the lexically specified accent, and Accent 2 is assigned by default to all polysyllabic words lacking lexical accent specification and contain at least a disyllabic trochee. Accent 1 can be lexically specified on prefixes (be-, för-, ent-), suffixes (-era, -ell, -ant, etc.) or stems (ˈfänrik ‘ensign’, ˈtaxi ‘taxi’, ˈfredag ‘Friday’, januˈari ‘January’, etc.). The presence of lexically specified accent blocks the postlexical default rule, and thus Accent 1 dominates. Elsewhere, all words that do not contain a disyllabic trochee – polysyllabic or monosyllabic – surface with Accent 1.
Lahiri et al. (Reference Lahiri, Wetterlin and Jönsson-Steiner2005a) further draw attention to the fact that Scandinavian tonal dialects (Norwegian as well as Swedish) show different phonetic realizations of the two contrasting accents. Consequently, they do not represent lexical accent with a particular tone, but instead morphemes bearing lexical Accent 1 are assigned an abstract diacritic [̽]. The diacritic indicates that a morpheme (word or affix) is a carrier of lexical Accent 1 and is tonally interpreted on a dialect-specific basis. In this way, the precise phonetic correlate of the lexically specified Accent 1 is left open such that the lexical tone can be manifested by a high or low tone depending on the dialect (e.g. L* in Central Swedish but H* in Southern Swedish). Accent 2, the unmarked tone, also has its own dialect-specific phonetic manifestation.
An advantage of the approach taken by Lahiri et al. (Reference Lahiri, Wetterlin and Jönsson-Steiner2005a) is that it phonologically deals with cases like the prefixes be-, för-, ent- without having to resort to exceptional or additional rules.Footnote 5 In earlier approaches, a word like ˈfalla2 ‘to fall’ is assigned Accent 2 because of the infinitive -a which is assumed to carry the accent. The problem then is to explain why a form like beˈfalla1 that also contains the same infinitive -a has Accent 1. One solution in privative terms is to assume that the prefix triggers a process of deaccenting (Riad Reference Riad1998: fall-a2 > ˈfalla2; be-fall-a2 > deaccenting > beˈfalla1).Footnote 6 Similarly, Bruce & Hermans (Reference Bruce and Hermans1999) propose a context where, despite the Accent-2 inducing infinitive suffix, words that have ‘a pre-head before the left-headed foot’ (Bruce & Hermans Reference Bruce and Hermans1999:613) will usually be assigned Accent 1 (with the exception of a subcategory of nouns). In contrast, Lahiri et al. (Reference Lahiri, Wetterlin and Jönsson-Steiner2005a) do not assign lexical accent to the infinitive -a; rather, non-prefixed infinitive forms like ˈfalla2 get Accent 2 by default since the word contains a disyllabic trochee and no lexical specification. The prefix be-̽, however, is lexically specified for Accent 1 which blocks the default Accent 2 from applying to infinitives with this prefix: be̽-fall-a. Accent 1 from the prefix is assigned to the entire word and be̽falla1 is thus Accent 1.
The tables in (1)–(3) show how accent assignment in Swedish works within this framework. Here we list suffixes (inflectional endings, separated by a hyphen), clitics (definite endings, separated by an equals sign) and stressed and unstressed prefixes chosen specifically to relate to Nicander's work.
(1) Accent assignment with syllabic suffixes vs. cliticsFootnote 7
The examples in (1) show that accent assignment works differently with the definite endings as compared to regular plural suffixes. We follow the standard assumption that accent assignment remains unaffected by the addition of the definite endings (singular=en/ncommon =et/tneuter; plural=en/na/a) in Swedish (cf. Riad Reference Riad1998:65). Lahiri, Wetterlin & Jönsson-Steiner (2005b) argue that this difference is due to the fact that the definite endings in Swedish and Norwegian phonologically behave like clitics which are added after accent assignment. Thus, the sg.def forms in (1) are Accent 1 despite the fact that they form disyllabic trochees. The plural endings, on the other hand, form part of the accent-assignment domain and the plural forms get Accent 2 in (1a, b). The addition of the definite plural clitic has no effect on the accent of the word (cf. columns 2 and 3 in (1)). The specified {be-̽} prefix, however, induces Accent 1 on the form it attaches to even in the plural; cf. (1c).
In (2) we see that both stems and grammatical morphemes can carry lexical tone and that lexically specified accent always dominates.
(2) Accent assignment with lexically specified words and affixes
In (2a, b), the disyllabic words are lexically specified for Accent 1. Despite the fact that ˈfänrik1 contains a disyllabic trochee it is Accent 1 since the lexical specification blocks the application of default Accent 2. Lexical specification dominates no matter what is added to the stem. The noun ˈfänrik1 remains Accent 1 with the plural suffix (ˈfänrikar1) or clitic (ˈfänriken1, ˈfänrikarna1). The same is true for poˈet1 in (2b) which remains Accent 1 in all suffixed and cliticized forms, even though the plural forms provide the prosodic environment where Accent 2 could potentially apply if there were no lexical specification intervening (poˈeter1, poˈeterna1). The examples in (2c–f) demonstrate that the addition of a lexically specified prefix dominates and overrides any other accent. The infinitive forms in (2c, e) get Accent 2 as the default because they are not lexically specified and contain a disyllabic trochee. However, when lexically specified prefixes like/be×-/or/för×-/are involved as in (2d, f), the lexical accent overrides the accent of the stems and the prefixed forms receive Accent 1.
The examples in (3) show that words with stressed prefixes, as in (3a) – as opposed to words with unstressed prefixes – are always assigned Accent 2 in Standard Swedish following the same pattern as compounds, cf. (3b), which are Accent 2 irrespective of accent specification of individual words, cf. (3c).
(3) Stressed prefixes and compounds
Thus, the lexically specified word ˈt
mos1 gets Accent 1 in the singular and in the plural, but the compound ˈtrmosflaska2 in (3c) gets Accent 2 despite the lexical specification of the first member.Footnote 8
As we have seen so far, certain stems and grammatical morphemes play a dominant role in accent assignment when they are specified for Accent 1. Furthermore, the definite endings behave like clitics, i.e. they are invisible to accent assignment. Did these morphemes have the same tonal properties about two-and-a-half centuries ago? Dialectal differences have given rise to speculations regarding accent change. For instance, Riad (1998:84, fn. 23) suggests that words with unstressed prefixes like be- could have had Accent 2 at an earlier period as in the present-day Älvdalen dialect instead of Accent 1 of Standard Swedish. Kock (1878:80–86) points to reports suggesting that Accent 2 on words with unstressed prefixes also occurred in 19th-century Stockholm Swedish. However, Kock suggests that the earliest be-words had Accent 1, and that Accent 2 in these forms appeared later. He provides evidence from earlier authors that other dialects (Östergötland, Västergötland, Göteborg) had Accent 1 during the same time. Further, Kock also refers to the 1836 grammar from Svenska Akademien which has Accent 1 for be-words. Thus, Kock assumes a temporary change in Stockholm Swedish through the influence of Accent 2 in the simplex words, which have again reverted back to Accent 1 later.
As a rule of thumb, we will follow Riad's assumption that unless there is clear evidence to the contrary, if all Swedish dialects have the same accent on a particular morpheme, this was also the case in the 18th century. Dialectal differences warn us that some change has taken place. To determine which dialect represents the earlier stage, however, is a matter of conjecture unless we have further evidence. This is what we hope to discover from Nicander. For instance, based on evidence from Nicander, we will argue that Accent 2 on be-words in the Älvdalen dialect is an innovation. In the next section, we take a closer look at Nicander's poetic manual and examine his sometimes cryptic statements in search of information on the prosodic properties of accents and of lexically specified morphemes during his time. We then turn to Nicander's own poetry in section 4 to investigate what sorts of words he himself considered to be proper end-rhymes.
3. NICANDER'S POETIC MANUAL AND TONE
Nicander came from southern Sweden. He was born in Småland, moved to Blekinge at the age of two, attended school in Helsingborg and then went to the University of Lund. From the age of twenty-one he lived in areas where the dialects had standard Swedish accent distribution (Stockholm and Västervik). At school and later at university, Nicander is reported to have been an outstanding student with a profound knowledge of classical philology and exceptional poetic talent (Hanselli Reference Hanselli1874). He was fluent in several languages including Latin, German and French. After he left university, Nicander held different public positions, which allowed him to continue his poetic work. Besides composing his own poetry, Nicander also translated poetic works from other languages. At the age of 30, he wrote a poetic manual Oförgripelige anmerckningar öfwer swenska skaldekonsten (1737), in which he instructs readers how to write poetry in Swedish modelled on Latinate poetic rules. In the introduction, Nicander states that he aims to show that Swedish is well suited for composing verse following the rules of Latin poetry.Footnote 9 The manual consists of thirteen chapters of which only chapters I and XII are significant for our purposes. The first chapter deals with syllable quantity and outlines the basics of how Swedish words ought to be dealt with in verse. Nicander explains how quantity rules for Latin verse should be applied to Swedish verse in conjunction with regular Swedish pronunciation, with instructions on how syllables should be placed in long or short position. Chapter XII deals with end-rhymes in Swedish verse, and here we find crucial evidence for the accent system of 18th-century Swedish. All other chapters instruct the reader on verse types and do not provide any information about the spoken language.Footnote 10 In addition to the printed manual, we have also found a handwritten version of the manual (in Nicander Reference Riad1793), which is incomplete and mainly concentrates on quantity in terms of stress and strong/weak position in heroic verse. It is only in the printed manual that Nicander provides the effects of accent on verse and makes more general claims about different kinds of verse including rhymed verse. We will therefore mainly consider the printed manual here but we also refer to the hand-written version wherever necessary.
In section 3.1, we discuss the crucial terminology from the first chapter of the poetic manual, which provides the basic principles for understanding Nicander's rules for verse. We then turn to chapter XII of the manual in section 3.2, which deals with properties of end-rhymes, and proves to be the most crucial part with respect to tonal accents.
3.1 Latin verse rules and quantity applied to 18th-century Swedish
A fundamental principle of Nicander's verse practice was to assume that stressed syllables in natural language should be in strong position in verse. With this principle as background, the first chapter of Nicander's poetic manual provides us with a set of terms describing various categories of syllables. Our goal is to understand these statements and terms so that we can interpret his rules for end-rhymes.
Nicander's rules about stress, quantity and position in verse include four critical terms. Based on his manual and the examples he provides we can interpret these as follows.
(4) Interpretation of the terms lång, kort, brevis, communis
(i) LåNG
Syllables can be lång ‘long’ either (a) ‘by nature’ as Nicander puts it (i.e. intrinsically), if they contain a diphthong or a long vowel, or (b) by position, when followed by a consonant cluster. lång syllables can be placed in strong position in verse, which in turn then reflects natural word stress. The actual meaning of the term lång depends on its context. When Nicander asserts that diphthongs are lång in the very first rule, he is referring to length ‘by nature’ (intrinsic length). Thus, the aj sequence in the stressed penultimate syllable of kaja2 ‘jackdaw’ is lång by nature since Nicander regards this as a diphthong (as it still is in Norwegian today, e.g. ˈkaie). But when Nicander describes the special properties of what he calls ‘composite forms’, lång seems to imply something further. He explicitly points to certain words with unstressed prefixes – which in modern Swedish bear Accent 1 – as being different because the stressed syllable is lång in contrast to their non-prefixed forms in which they are communis, as we will see in Rule I:9 below.
(ii) KORT
Kort ‘short’ depicts syllables in weak position. Syllables can be intrinsically kort, i.e. have a short vowel (ˈklaga ‘complain.inf’, ˈhare ‘hare.sg’, beˈgripa ‘comprehend-inf’, förˈarga ‘annoy.inf’), or they can be made or treated as kort in order not to be interpreted as stressed if they are unstressed in natural speech (ˈdundrande > ˈdundra’ne), or if they are words that can be prosodically in weak position (och ‘and’, har ‘have.pres’, hon ‘she’).
(iii) BREV
The term brev basically denotes another kind of a kort syllable, which always corresponds to short unstressed syllables in natural speech, and are thus placed in weak position in verse (ˈGoliat, fiˈol). Conversely, kort syllables as described in (ii) could be heavy in natural speech.
(iv) COMMUNIS
Communis is a term that does not refer to the natural weight of syllables but to their variable treatment in verse. Syllables that are referred to as communis can be treated as lång or kort, and thus be placed in strong or weak position.
Although at first glance the term communis seems straightforward, it has a number of hidden nuances, particularly in conjunction with lång. In Rule 7 of chapter I, Nicander says:
Syllables in compounds should keep the quantity of their simplex form but that does not always happen in Swedish such as förfara has the middle vowel lång but fa in fara is communis. (Rule I:7)Footnote 11
The crucial distinction drawn here is between the infinitive form fara2 in isolation and the complex form (compound in Nicander's terminology) that has an unstressed prefix förˈfara1 ‘proceed-inf’. Whatever quantity (quantitatem) may mean here, Nicander clearly indicates that fa has different properties in ˈfara than in förˈfara. Could these properties be related to length or stress? Not really, since in both words fa is stressed and is the penultimate syllable in the word. Our conjecture is that in this rule Nicander is not simply referring to a length distinction but to a contrast in tonal accent, implying that although ˈfara2 and förˈfara1 ought to have the same tonal accent, the prefixed form has Accent 1 while the simplex form has Accent 2. When Nicander genuinely talks about quantity, he uses the term brevis. We find more support for our interpretation in Rule I:9.
(5) Rule I:9
All regular past forms have a brevis penultimate syllable such as: spelade, elskade. Exception. Irregulars, such as: hă̄de has communis penultimate syllable and so does lă̄de, but its composite form is lång, such as: belă̄de.Footnote 12
Unlike förˈfara1, both ˈspelade2 and ˈelskade2 (past tense forms of ˈspela ‘play.inf’ and ˈelska ‘love.inf’) have initial stress and Accent 2 in modern Swedish. Today both verbs have disyllabic roots (spela, elska), and consequently, the root and the infinitive forms are homophonous, unlike monosyllabic roots like far, where infinitive is ˈfara2.Footnote 13 The penultimate syllables, la and ska, in both past tense forms are the final unstressed syllables of the roots and these are brevis which can never be in strong position. However, as in ˈfara vs. förˈfara, Nicander draws attention to the communis penult of ˈlade2 ‘lay-past’ as compared to the lång penult in the prefixed form beˈlade1 ‘cover-past’. The difference between the two stressed penultimate syllables is that one belongs to an Accent 1 word and the other to an Accent 2 word. Again Nicander draws attention to differences between CV syllables, which happen to be brevem penults on the one hand (unstressed and presumably short, and cannot be in strong position), as against stressed communis versus lång penults on the other hand. Nicander points out that the composite forms beˈlade1 and förˈfara1 deviate from the simplex counterparts ˈlade2 and ˈfara2 although they ought not to. Since neither stress nor quantity nor syllable position (all are stressed and penultimate syllables) can be the cause for this concern, we hypothesize that he was conscious of the tonal accent differences and draws attention to them.Footnote 14
Thus, words with unstressed prefixes are used as examples where their stressed syllables are described as lång in conjunction with a discussion of communis and brevis. Words like förˈfara1 ‘be wasted-inf’, beˈlade1 ‘cover-past’ were always treated as lång by Nicander. These are compared to their non-composite forms (ˈfara2 ‘go-inf’, ˈlade2 ‘lay-past’), which invariably have Accent 2 in modern Swedish.
We conclude that Nicander was aware of accent differences. In this part of the manual he does not talk about ‘good’ or ‘bad’ rhymes, just about types of prosodic categories. Nevertheless it is obvious that when things do not fit, in effect Accent 1 is always involved, as in Rules I:7 and I:9 unstressed Accent-1 prefixed verbs are compared to non-prefixed verbs (ˈfara2 vs. förˈfara1, lade2 vs. beˈlade1). However, he draws attention to categories in a rather oblique manner. In the next section we will examine how these types of words fit into end-rhymes when Nicander refers to accent differences more directly.
3.2 Nicander's poetic manual and end-rhymes
Chapter XII of Nicander's poetic manual is dedicated entirely to different kinds of Swedish end-rhymes (1737:33–35) where he explicitly differentiates between the two accents. Although most of his definitions concerning rhymes are taken directly from well known Latin poetic notions, we need to discuss them because they directly relate to the issue of accent differences. Out of the five rules in this chapter, only the first four provide us with relevant information about the phonological system of Nicander's Swedish.
3.2.1 Feminine and masculine rhymes
The first of the rules of chapter XII of the manual is presented in (6).
(6) Rule XII:1
Following traditional metrical systems, Nicander distinguishes two types of end-rhymes, qwinlige rim (feminine rhymes) and manlige rim (masculine rhymes), where masculine end-rhymes consist of a final syllable that is lång. His examples of masculine rhymes are given in (7).
(7) Nicander's masculine rhymes – lång final syllables
The examples of masculine rhymes include monosyllabic words and one disyllabic word with an unstressed prefix. These words all have Accent 1 in modern Swedish. They also must have had Accent 1 in the 18th century, since Accent 2 cannot be realized on monosyllabic words or polysyllabic words with final stress where there is no disyllabic trochee available. Nicander describes the final syllables here as lång. One interpretation of lång would be that these syllables are in strong position. Since all the end-rhymes consist of words with final stress, this interpretation is valid. However, as we pointed out in section 3.1, Nicander describes composite words with unstressed prefixes and certain monosyllabic words as lång which all have Accent 1. Thus, the lång last syllable can also be referring to Accent 1. We return to this issue later in section 3.2.2.
Feminine end-rhymes match the two final syllables, the first of which is stressed and in a metrically strong position. Examples of feminine rhymes are given in (8).
(8) Nicander's feminine rhymes
The examples that Nicander gives for feminine rhymes under Rule XII:1 are all words that have Accent 2 today. However, the definition of feminine end-rhymes implies that they are disyllabic trochees, which theoretically could also have Accent 1. Indeed, Nicander also has Accent-1 feminine rhymes as we will see under Rule XII:4 in section 3.2.4.
3.2.2 Masculine end-rhymes and faithfulness to quantity
The second rule of chapter XII instructs the reader to pay attention to quantity in end-rhymes. The rule is presented in (9).
(9) Rule XII:2
In the context of end-rhymes, Nicander claims that a short syllable may not be ‘lengthened’ for the sake of matching an end-rhyme and provides examples of unacceptable rhymes. Below we add the stress and the accent of these words as they are in modern Swedish. The final rhyme is marked in bold.
(10) Mismatching end-rhymes due to stress and quantity
For Nicander, the short final syllables in fengelse and frelsare cannot be rhymed with the lång final syllables in galathé or betee. This statement suggests that these words were stressed as they are today: short unstressed -e in fengelse and frelsare may not occur in strong position and thus cannot rhyme with a lång syllable like the long stressed final -e in galathé or betee. Note that this rule also subsumes the fact that masculine rhymes (lång final stressed syllables, cf. section 3.2.1 below) cannot be paired with non-masculine rhymes. Not only are they bad rhymes, but the rhymes may not be rectified by changing a final syllable which is lång in spoken language to a kort final syllable in verse; stressed long vowels cannot be shortened nor can the unstressed ones be stressed and lengthened.
There is a further point to be made about stress in Swedish. Nicander wanted Swedish to follow the Latin stress rule, viz. heavy penults are stressed, otherwise stress falls on the antepenult. When certain words violated this pattern, for example, ˈdundrande (the heavy penult was not stressed), he resorted to syncope or metathesis to change the lång syllable into a short one ending up with ˈdundra'ne so that the naturally stressed antepenult could be put into strong position in verse. The fact that Nicander insists on maintaining the faithfulness of quantity and stress in these end-rhymes tells us that he was conscious of them in the contemporary language.
3.2.3 Feminine end-rhymes: coinciding final V or VC
Rule 3 of chapter XII, as presented in (11), instructs the reader to match final segments in feminine end-rhymes.
(11) Rule XII:3
Matching final segments in end-rhymes is a basic rule in rhyming systems universally. Here, Nicander's examples show that he was aware of subtle pronunciation differences.
(12) Matching and mismatching final segments in feminine rhymes
The words in (12a) are perfectly acceptable matches in feminine rhymes. However, the pairs in (12b, c) cannot rhyme since they have different final consonants. Nicander's insistence that final segments must match suggests that final <d> and <t> differed in pronunciation. Thus, like modern Swedish, Nicander's Swedish clearly did not devoice final consonants, even though Swedish was in close contact with Low German – a language with final devoicing – and borrowed from it extensively. The same argument holds for the final vowel: the word pair in (12d) is a matching rhyme, whereas (12e, f) do not match since the vowels in the final syllables differ. Thus, the vowels [e], [o] and [a] in the final syllables of these rhymes must have been distinct in quality as is still the case in modern Standard Swedish, whereas other Scandinavian languages like Standard East Norwegian have neutralized the vowel distinction.Footnote 15 Thus, we can conclude that final segments listed here maintained the same contrasts then as they do now.
3.2.4 Matching accent in (feminine) end-rhymes
Now we come to Nicander's most explicit statement concerning the difference in tonal accent between clitics and inflectional suffixes.
(13) Rule XII:4
This is the most important rule from our perspective. Nicander asserts here that there are words which fulfill the criteria of Rule XII:3 above (which says that the final syllables of feminine rhymes should be segmentally overlapping) but nevertheless they are still not proper rhymes since they do not have the same accent in his speech.Footnote 16 Examples of non-rhyming pairs due to conflicting accents are listed in (14). In the second column, we indicate the morphological decomposition, with suffixes separated by a hyphen and clitics with an equal sign, as in the table in (1) above.
(14) Non-rhyming pairs according to Nicander
What does Nicander mean by accent? We discuss each example in turn. The subscript numbers indicate the accent of the word in modern Swedish. If the accents were the same for Nicander as they are now, it is quite obvious that the non-matching property of these rhymes can only be due to differences in accent since, aside from the differing word onsets, the final segments of each pair are identical. However, can we be certain that the tonal accents for each pair also differed for Nicander, just as they do today?
In the first pair, galen Footnote 17 is a monomorphemic disyllabic word with initial stress and there is no reason to expect any tonal change in the last two centuries. The word talen could potentially have been the verb tala + present-plural, but from the context it is quite clear that it refers to the noun tal ‘number’ in the definite plural. Since the definite plural clitic does not form an accent domain of its own today (cf. examples in (1) and (2) above) and never did (cf. Elstad Reference Elstad and Jahr1983), the mismatching characteristic of this pair must have been due to the tone – Accent 1 talen1, Accent 2 galen2. In pairs (14b) and (14c), Nicander distinguishes between nouns with definite clitic endings and verbs that have the present plural suffix. Modern Swedish lacks this suffix. However, had it remained, ˈgrinen1 (grin=def.pl) ‘grin’ would also form a minimal pair with the verb ˈgrinen2 (grina-pres.pl) ‘whine’. Again, from the context, it is obvious that Nicander is referring to the verbal present plural form and not the nominal form. Thus, the non-rhyming nature of these pairs must be due to the differences in tonal accent. Similar to modern Swedish, Nicander makes a distinction between monosyllabic Accent-1 words with a definite singular (common) or plural (neuter) clitic =en (tal=en, swin=en) and monomorphemic disyllabic words (galen) or monosyllables with a present plural suffix -en (grin-en, gråt-en) that have Accent 2 (cf. the tables in (1)–(3)). Obviously, complex disyllabic words where the endings contribute to a tonal difference do not qualify as matching rhymes for Nicander.
Could accent refer to anything other than differing tonal accent in this rule? We do not think so. Rule XII:4 refers directly to the previous rule, which discusses feminine rhymes constituting polysyllabic words with non-final stress. Consequently, the rule only refers to feminine rhymes that consist of disyllabic trochees and do not have a final lång syllable that masculine rhymes have (cf. Rule XII:1). This rule and the examples in (14) confirm our interpretation of Rule XII:1 above that feminine rhymes can be either Accent 1 or Accent 2. Comparing the examples in (12), where Nicander's mismatching rhymes point to segmental deviations, with the mismatching rhymes in (14), where there are no segmental differences, it is obvious that Nicander was sensitive to the tonal distinction between these word pairs. He clearly considered that both tonal incompatibilities as well as segmental differences contribute to non-matching rhymes.
3.3 Summary
The aim of this section has been to correctly interpret what Nicander's poetic manual tells us about the distribution of tonal accents in 18th-century Swedish. In section 3.1 we deduced that while discussing different categories of syllables, Nicander distinguishes unstressed prefixed Accent-1 verbs from their underived Accent-2 counterparts. In section 3.2, we concluded that proper end-rhymes in Nicander's verse must match in segments, stress and tonal accents. We deduced that masculine rhymes, which are described as having a final lång syllable, could only have had Accent 1. Feminine rhymes that have two final syllables without a final lång syllable could be either Accent 1 or Accent 2. Nicander's claim that feminine rhymes must have matching accents provided the vital piece of evidence confirming that he was aware of tonal accent distinctions. His examples then reinforce our claim that he is referring to a tonal difference in the stressed syllable. The examples also convey the information that there must have been a tonal difference between the monosyllabic roots with a definite clitic (which are Accent 1 today) and disyllabic roots or roots with unmarked syllabic suffixes (which bear Accent 2 in modern Swedish). Thus, when it comes to end-rhymes, Nicander explicitly points out that there are words that segmentally look the same and have stress on the same syllable but differ in terms of tonal accent and thus do not rhyme (cf. (13) above).
We conclude that Nicander was aware of tonal accents, and furthermore, that – to a great extent – the distribution of the tonal accents must have been the same in 18th-century Swedish as it is today in modern Swedish. We now turn to end-rhymes of Nicander's own poetry.
4. NICANDER'S OWN VERSE
End-rhymes should match in accent, if they are to be considered ‘proper’ rhymes – this is Nicander's central claim about suffixed, cliticized and disyllabic words in chapter XII on end-rhymes. We examined all of Nicander's printed rhymed verses included in Hanselli's (1874) collection. Altogether these contained 108 lines, which gave us a total of 56 end-rhymes of varying types and length. In addition, we examined a further selection of Nicander's handwritten verses found in Nicander (Reference Riad1793). This selection consists of another 31 verses, altogether 271 lines with 134 end-rhymes. All of the rhyming words are listed in the Appendix.
To what extent do Nicander's own rhymes match in tonal accents – this is our central point of interest. Since famous Swedish poets have ignored accents when composing rhymes for centuries, it would not be astonishing if Nicander did the same even if he holds forth on ‘proper’ rhymes in his manual. A careful examination of his poetry presents a complex picture – some rhymes are matching in accent (i.e. the accents as they are in modern Swedish) and some do not. Could the variability be due to an inability to match accents or because Nicander's accentual patterns were different? As we will see, it turns out that for Nicander, end-rhymes were the rightmost foot of a word beginning with the stressed nucleus, which is the frame of the feminine or masculine rhymes. Consequently, suffixed or cliticized words could constitute end-rhymes, while only the rhyme of the second word of compounds was critical. The same holds true for prefixed words with stressed or unstressed prefixes where the stem was the crucial constituent, giving rise to apparent accent mismatch. We will conclude that there has been no change in accent pattern from Nicander's time till the present.
4.1 Rhymes with simplex and suffixed words
The majority (74%) of Nicander's end-rhymes are simplex and suffixed words. Of these, 45% are monosyllabic masculine end-rhymes, as in the examples in (15).
(15) Monosyllabic end-rhymes
Of the simplex and suffixed feminine end-rhymes in Nicander's verses, the majority (76%) are monomorphemic or suffixed forms which take Accent 2 in modern Swedish – see (16) below. Since the accent of these words does not differ in modern Swedish dialects, there is no reason to believe that they were any different in the 18th century. For the infinitives, Nicander included what nowadays would be disyllabic roots (with disyllabic imperatives, cf. note 13) as well as monosyllabic roots. Both infinitive forms end in <a>: ˈvörda2 ‘respect.inf’, modern Swedish imperative ˈvörda2; ˈbära2 (bär-a) ‘carry-inf’, modern Swedish imperative ˈbär1.
(16) Feminine rhymes – disyllabic inflected and simplex Accent-2 words
Only 24% of Nicander's feminine rhymes correspond to Accent-1 words in Standard Swedish, simplex as well as inflected and cliticized forms. Some examples are given in (17).
(17) Feminine rhymes: disyllabic inflected, cliticized and simplex Accent-1 wordsFootnote 18
In (16) and (17), we see that Nicander matched both the final segments and the accents of the words in a rhyme, but that grammatical categories were not necessarily matched as long as segments and accents match. The rhymes in (16a) and (17a, c) are segmentally overlapping but are morphologically different. The pair in (16a), börda2, vörda2, are disyllabic stems, where the first is a noun and the second a verb. The word pairs in (17a, c) are segmentally overlapping, but the suffixes do not belong to the same grammatical category. Even so, they have matching accents, at least in modern Swedish, viz. Accent 1, since the present tense form gläder is underlyingly monosyllabic and the plural form fäder takes Accent 1, which is generally the case for umlauted plurals. In (17c), the word kläder, which in modern Swedish only exists in the plural, rhymes with fläder, both of which have underlying monosyllabic stems and are Accent 1. In Rule XII:4 Nicander cautions the reader against rhyming definite singular and plural forms with present plural forms of verbs (which are obsolete in modern Swedish), and we find that he adheres to his warning in his own rhymes. These forms can often be segmentally overlapping but they have different accents. Hence, in the feminine rhymes, Nicander never rhymes forms containing the definite clitic with the present plural or even the nominal indefinite plural, which normally bears Accent 2 in modern Swedish. Accordingly, in (17b) Nicander rhymes the definite singular forms of two monosyllabic words. Recall that in Rule XII:4, he specifically draws our attention to the accent difference in such instances.
Thus, with monomorphemic or suffixed, or cliticized words, Nicander's rhymes combine and match accents of entire words. He always matches accents even if he uses varying grammatical categories in his rhyme pairs: indefinite singular/infinitive (börda2 ~ vörda2), present ~ indefinite plural (gläder1 ~ fläder1), indefinite plural ~ indefinite singular (kläder1 ~ fläder1), etc. In none of the examples does he pair words of different grammatical categories (definite singular ~ present plural, etc.) where the accents do not match.
Up to now, all suffixed words were inflected forms that essentially consisted of one foot. However, there are also rhymes where Nicander uses words with derivational suffixes, which consist of more than one foot. These words with multiple feet show us a more complex pattern of end-rhymes and accents. In (18), complex words derived with the suffix {-het} are rhymed with simplex and inflectionally suffixed words.
(18) Rhyme with derivational suffixes
The complex words themselves all have Accent 2 in modern Swedish. However, based on the metrical pattern in the rhymes, it is obvious that lycksaligheten ‘bliss-def.sg’ has main stress on {salig} which exists in varieties of modern Swedish where it has Accent 1. The crucial point is that these rhyming pairs do not match in word accent in Standard Swedish. What Nicander rhymes here is clearly the rightmost foot: (18a) {-het} ~ {ˈvet}, (18b) {- |heten} ~ {ˈveten}, (18c) {- |het} ~ {poˈet}. The second member of these three rhyming pairs are prosodic words with their own accent, but the first member -het is a derivational ending. In (18d) there are three rhyming words, {- |heten} ~ {poˈeten} ~ {proˈpheten}, and again the first member is -het. Neither {-het} nor {-heten} exists as an independent lexical word but each can bear secondary stress. We return to these words for further discussion in section 4.3, after we have contended with similar complex patterns in compounds and prefixed words.
4.2 Rhymes with compounds and prefixed words
As we showed earlier, the majority of Nicander's end-rhymes consist of matched monomorphemic words and inflected words. These consist of one prosodic foot, and the accent of the whole word is matched in rhymes except for words with -het. However, in Nicander's verse, 26% of the rhymes also involve compounds and prefixed forms. In (19), monosyllabic words are rhymed with compounds which have a monosyllabic second member. In modern Standard Swedish these compounds are always Accent 2 (with the contour realized on the whole compound, cf. Bruce Reference Bruce1977) with stress on the first syllable and secondary stress on the last syllable, which is also in strong position.
(19) Compounds in rhymes
In all but four cases in Nicander's rhymes with compounds, the second member is monosyllabic and matches in accent with its monosyllabic rhyming counterpart. Both primary and secondary stressed syllables of the compounds are in strong position in his rhymes. Note, however, that according to the modern Swedish accentual system the accent of the whole compound obviously mismatches with the accent of the monosyllabic rhyming counterparts as in band1 ~ Göthaland2.
For prefixed words, Nicander rhymed them with simplex and inflected words and again, we find some accent asymmetries, presented in (20). In the third column, we also give the morphological decomposition of the complex words with the gloss.
(20) Prefixed words rhymed with simplex and inflected suffixed words
In these verses, Nicander always rhymes the rightmost foot beginning with the stressed vowel, which happens to be a prosodic word. In (20a, b) monosyllabic words (ˈsvår1, ˈhar1) are rhymed with monosyllabic stems preceded by unstressed prefixes (beˈstår1, förˈsvar1). However, in (20c–f), the words are suffixed but in each rhyme only one of the words is prefixed. Again the accents of the prefixed words do not always coincide with the suffixed word that they are rhymed with.
In Nicander's verse, there are also a few instances where two prefixed words are rhymed, as in (21). Here the same rhyming pattern holds – he rhymes the last foot – and stressed prefixes are never in a metrically strong position.
(21) Rhymed prefixed words
Taking into consideration all the different types of end-rhymes, we find that the rightmost foot is central for complex words including compounds. The strong position of the end-rhymes is invariably a foot which heads a (main or secondary) stressed syllable. Once we consider the final foot, then the accent asymmetries disappear. Nevertheless, other possibile explanations also need to be considered and we turn to these in the next section.
4.3 Nicander's end-rhymes and accent
To match accents in end-rhymes is not a principle that is followed in modern Swedish and has never been generally accepted as necessary in Swedish verse. Nevertheless, Nicander's idea of a perfect end-rhyme was one where not only segments but also accents matched. In his manual, Nicander made explicit reference to accents and end-rhymes only for inflected, simplex and cliticized words. He made no reference to compounds or derived words with derivational prefixes or suffixes. Our inferences concerning accents on prefixed words were based on his description on quantity. Consequently, it is revealing to see what principles of accent matching he uses in his own rhymed verse, where compounds and prefixed words are included.
Following his own principles as laid out in the manual, Nicander's rhymes with simplex or inflected and cliticized words always match in tonal accents of modern Swedish: mot1 ~ bot1, kläder1 ~ fläder1, börda2 ~ vörda2, bygden1 ~ dygden1. And indeed, there are no exceptions to this rule in his own end-rhymes. However, when it comes to derivationally affixed words and compounds, he has a more complex setup, where the accents do not match when we consider the complete word forms, as can be seen in the following rhymes: ˈmåttlig |het2 ~ vet1, nåd1 ~ ˈöfver |dåd2, grifter2 ~ beˈdrifter1, ˈGöhta- |land2 ~ band1. There could be three possible explanations of why he would disregard these tonal patterns: (i) Nicander did not always take care to match accents in his own end-rhymes with compounds and prefixed words, and only matched final segments; (ii) the accents of the mismatching rhymes reflect a different accent pattern in the 18th century; or (iii) Nicander did not consider the tonal structure of the entire word in derivatives, compounds and prefixed words, but matched the tone of the rightmost prosodic foot which could itself be a nonprefixed inflected or uninflected word. We consider each in turn.
(i) Could it be that Nicander ignored accents when his rhymes consisted of derived words and compounds? Nicander's statements concerning what one ought to consider as a good rhyming pair and what one should not use allows us to conclude that he was aware of the distinctive accents and understood quite clearly the need to match accent in rhyming pairs, and it would be unexpectedly inconsistent of him to ignore his own rules when writing rhymed verse himself.Footnote 19 Moreover, the majority of Nicander's rhymes consist of simplex and suffixed words (see Appendix) and, as we have seen, they invariably match in accent. Nicander notes the differences where simplex vs. affixed words and compounds are concerned. Recall that the rules cited in section 3.1 above state that derivatives and compounds should keep the accent of the simplex forms (cf. upvakne ~ vakne), but that this is not always the case. In particular, Nicander points to words prefixed by be- and för- as being different. For example, the penultimate vowel in förˈfara is lång but the penult in fara is communis. Thus, he surely would not deliberately ignore the accents for such end-rhymes. The reason behind the lack of accent matching in rhymes with prefixed words must lie elsewhere.
(ii) Perhaps the fact that the accents, as they are in Standard Swedish now, fail to sometimes match for compounds and affixed words because the accents were different in the 18th century? As was already mentioned at the end of section 2 above, one hypothesis is that the prefixed infinitive forms like betala1 were earlier Accent 2 and later changed to Accent 1, as in modern Standard Swedish (cf. Riad Reference Riad1998:84, fn.23). This would account for rhyme pairs like skära2 ~ förfära1, påkalla2 ~ befalla1 by assuming that förfära and befalla simply bore the accent of the nonprefixed form and thus would have been Accent 2 for Nicander. However, to be able to use these rhymes as evidence in support of such a hypothesis, one would also have to assume that Nicander matched accents of the entire complex words; but then it is difficult to understand why he allowed rhymes with compounds (band1 ~ ˈGötha- |land2, ˈundan |tag2 ~ lag1), or disyllabic prefixed words (ˈöfver |dåd2 ~ råd1), and with derivationally suffixed words like ˈmåttlig |het2 ~ vet1. These types of complex words always bear Accent 2 today, and they are rhymed with monosyllabic words, which are and were invariably Accent 1.
An alternative explanation for the mismatching accents in the compounds in (22) could be based on Kock's claim that in the Swedish spoken at the time, it was not uncommon to reverse the stress pattern of compounds (as in some present day dialects of North Swedish). These compounds would then have obligatorily Accent 1. Given this situation, tillˈbeder, påˈkalla, Göthaˈland, öfverˈdåd would all be Accent 1 and match in accent perfectly with ˈneder, beˈfalla, ˈband, ˈnåd. If this were true for Nicander, then the accent matching needs no further explanation – he matched Accent-1 words, and this would be compatible with our view that he was sensitive to accent. There are, however, other mismatching rhymes that cannot be explained in this way, for example, ˈgrifter2 ~ beˈdrifter1, ˈskära2 ~ förˈfära1. Furthermore, this argument does not favour Riad's claim that words like befalla1 may have been Accent 2 earlier. If rhyming påkalla with befalla is due to Accent 2 matching (as Riad would have it), Kock's reasoning that påkalla had prominence on kalla, and must have therefore been Accent 1, does not hold.
In Southern Sweden, there are dialects where certain compounds can bear Accent 1. One could argue that Nicander's early southern Swedish background could have influenced the accents in these mismatching rhymes. However, these particular compounds and words with the prefix över- that Nicander uses are Accent 2 in southern Swedish as well. Thus, the mismatching rhymes cannot be a dialectal problem and cannot be explained by assuming that the unstressed prefixed words were Accent 2 in Nicander's time.
(iii) Instead, could it be that where Nicander used more complex word structures which involved more than one prosodic foot he did not consider the entire lexical item for matching accents? Let us go back to the examples of rhymes with derivational suffixes in (18) above: ˈmåttlig |het2 ~ ˈvet1, ˈlycksalig |heten2 ~ ˈveten2. What is Nicander's system? As we already mentioned at the end of section 4.1 above, words with derivational suffixes consisting of two feet follow a more complex pattern of end-rhymes and accents. According to (18), Nicander takes the last foot in heterosyllabic words with derivational suffixes and rhymes them with a monosyllabic words or inflected words. In (18a) the derivational suffix {-het} is a separate foot that takes secondary stress. Nicander treats it as a masculine rhyme and matches it with vet. In (18b) the derivational suffix {-het} together with the definite clitic {-|heten=en} also form a separate foot with secondary stress on the penult. It is therefore regarded as a feminine rhyme which he matches with veten. Note that although {-|heten} is not a lexical word, and has no intrinsic accent, Nicander feels free to rhyme an Accent-2 inflected prosodic word with a derivational suffix plus the definite singular clitic: vet-en ~ -het=en.
Does the same principle hold of rhyming a monosyllabic or suffixed word with the second element of derivational prefixed words and compounds which bore secondary stress? In (22) and (24) we repeat a few examples from Nicander's verses that involve compounds rhymed with monosyllables, and prefixed words rhymed with simplex or complex words with inflectional suffixes.
(22) End-rhymes involving prefixed words
If we only consider the second stressed foot of the complex words in these rhymes, we find remarkable consistency with end-rhymes always matching in tonal accents. In (22a, b) we have two examples of compounds rhymed with simplex words. Nicander places the first syllable as well as the last syllable of the compounds in strong position, which means that they had the same stress as today. Main stress is on the first syllable, and secondary stress on the first syllable of the last member in compounds. Nicander does not match the accent of the whole compound for the end-rhyme. Instead he only matches the accent of the rightmost prosodic feet: a monosyllabic word, which is a foot on its own, with the monosyllabic second member of these compounds, dåd1 ~ nåd1. Or the rhymes consist of inflected words, matching in accent, ˈgifvit2 ~ ˈdrifvit2. Note that the nonprefixed form is not necessarily directly related to the prefixed word.
We have now deduced that Nicander only considers the accent of the last prosodic foot of compounds and words with derivational suffixes. If we assume the same principle of rhyming for prefixed forms as well, everything falls into place. Looked at it in this light, Nicander now succeeds in matching the accents of all his rhymes, as the examples in (23) illustrate.
(23) End-rhymes involving unstressed prefixed words
The word pair in (23a) happens to match in accent, but we can also see that the rightmost prosodic foot is the lexical word drifver1 which makes a perfect match with gifver1. In (23b) the accent of the prefixed form does not match the accent of the word with which it is rhymed. However, the second foot of the prefixed form again is a lexical item drifter2, which matches the accent of grifter2. The rhymes in (23c–d) are infinitive forms which have mismatching accents. But as soon as we consider just the accent of the rightmost foot we get the match of lexical items with the matching accents, as in falla2 and kalla2. Riad's explanation that bedrifter1 rhymes with grifter2 because the former (i.e. words with unstressed prefixes) bore Accent 2 in earlier times does not explain why Nicander would then permit fördrifver1 to rhyme with gifver1.
Hence, our prediction is that Nicander would never rhyme a prefixed word with a non-prefixed word if the stems (inflected or bare stems) did not match in accent. That is, a prefixed word such as beˈsviken1 ‘disappointed’, where the base form in isolation sviken2 (svika- past participle ‘to be let down’) bears Accent 2, cannot rhyme with a noun in the definite singular such as viken1 ‘bay=def.pl’. This would also be the prediction of Riad since he would assume the unstressed prefixes like beˈsviken were earlier Accent 2. There is only one example which contradicts this hypothesis, where the non-prefixed form can be one of two possible stems, both unrelated to the prefixed word: beˈdragen1 ‘to deceive-past part’ ~ dagen1 ‘day=def’. The stressed foot of the prefixed form could match the verb dragen2, which would be Accent 2, as well as a noun dragen1 ‘stroke=def’, which is Accent-1. Presumably Nicander chose the Accent 1 option. However, we make an even stronger prediction, namely that Nicander could not rhyme stressed disyllabic prefixed words like ˈundan |tagen2 ‘exception=pl.def’ with a word like slagen2 ‘slap past part’ although these words would match in accent perfectly. This is because tagen1 ‘grip=pl.def’ in isolation bears Accent 1 and could never rhyme with an Accent-2 word slagen2. Such rhymes should be possible if Nicander was considering the accent of the entire prefixed word. Rather, in both instances (unstressed as well as stressed prefixes), although the accents of the complex words matched, rhymes would not be possible if the base words, included in the final prosodic foot, did not. All problematic rhymes can be accounted for – förˈdrifver1 (drifver1) ~ gifver1, beˈdrifter1 (drifter2) ~ grifter2, ˈöfver |gifvit2 (gifvit2) ~ drifvit2, ˈundan |tag2 (tag1) ~ lag1, etc. – if we assume that Nicander considered the accent of the nonprefixed word, or in rare cases the stressed derivational suffix. As we can see from Nicander's rhyme pairs in the Appendix, this hypothesis is borne out.
Thus, our interpretation is that in rhymes Nicander only looked at the frame of masculine and feminine end-rhymes since these are the only two possible forms of rhymes that he allows in his theory of verse. The frame of masculine and feminine rhymes always contains segments that are simplex or suffixed/cliticized words. Although he notices discrepancies in his manual, the accent pattern of prefixed words as a whole is disregarded insofar as his end-rhymes are concerned. Therefore, since Nicander always matched the accent of the rightmost prosodic foot for his rhymes, prefixed words in rhymes do not provide any evidence whatsoever that the words with unstressed prefixes that are special today were different in Nicander's language.
5. CONCLUSION
Modern Swedish verse usually disregards tonal accents, and it has never been established as a general rule in Swedish poetry that the principle of matching accents is important for rhyme. Even famous poets like Carl Michael Bellman (1740–1795) and Hjalmar Gullberg (1898–1961) ignored accent differences when composing rhyming verse. Within this tradition, Anders Nicander stands out in that his rules clearly state that words must have the same accent to make a perfect rhyme. His Rule XII:4 (cf. section 3.2.4 above) warns the aspiring poet against using rhymes that match segmentally but not tonally. After careful examination of Nicander's poetic manual and his rhymes, we have been able to show that his comments reflect tonal accent in natural speech of his time. He was acutely sensitive to accent and actively attempted to match it in his own end-rhymes. Comparing his rhymed words with their modern Swedish counterparts, we found that Nicander used simplex, suffixed and cliticized words as rhyme pairs only when they match in accent today – and there is no reason to assume that relevant accents have changed in the meantime: ˈbörda2 ‘burden.sg’ ~ ˈvörda2 ‘respect.inf’, ˈgläder1 (gläd-er) ‘please-pres’ ~ ˈfäder1 (fader.pl-̈) ‘father. pl’. Further support comes from his manual where he particularly observes that nouns hosting definite enclitics and nouns in the indefinite plural are bad rhymes (cf. section 3.2.4 above).
Not only was Nicander sensitive to such accent contrasts, his awareness of the accents led him to make use of them in verse. His examples call attention to the accent differences between words such as ˈtalen1 (tal=en ) ‘number=def.pl’ vs. ˈgalen2 ‘mad’. Such contrasts are found in Swedish today. Thus, we could conclude that Nicander's suffixes and clitics formed different accent assignment domains as they do in modern Swedish.
However, in Nicander's rhymes involving more complex structures with derivational suffixes, prefixes and compounds, we ascertained that half of his rhymes do not have matching accents in the modern language if we consider the accent of the whole complex word structures. The explanation is that in these instances Nicander did not match the accent of the entire complex words, but only of the final prosodic feet, and these feet could be inflected or uninflected words or even stressed suffixes (e.g. förˈdrifver1 (drifver1) with ˈgifver1, beˈslutit1 (ˈslutit2) with ˈflutit2, ˈöfver |flöd2 (ˈflöd1) with ˈstöd1, ˈmåttlig |het2 with ˈvet1). This interpretation does not mean that Nicander did not recognize accent differences with prefixed words and compounds, nor that the accent patterns of unstressed or stressed prefixes were different in his time. He only considers the final prosodic foot which is within the frame of the masculine and the feminine rhymes that he describes in his manual. When discussing quantity, Nicander notes that words with initial be- or för- behave differently compared to other ‘compounds’, and that they have different accents from their nonprefixed forms (cf. section 3.1 above: ˈfara2 vs. förˈfara1, lade2 vs. beˈlade1). Thus, there are clear indications in Nicander's manual that the unstressed Germanic prefixes in modern Swedish, which bear lexical Accent 1, were always Accent 1 and never Accent 2 (cf. Lahiri et al. Reference Lahiri, Wetterlin and Jönsson-Steiner2005a).
Two alternative explanations regarding Nicander's rhymes in complex words have been proposed. Consider first the pair ˈGötha |land2 and band1 . Kock argued that compounds like Göthaland could have had main stress on the second constituent rather than the first, and therefore must have been Accent 1 as are monosyllabic words like band. If it were true that the accents of the complex forms always matched, then prefixed words like ˈpåkalla would have been Accent 1 to match the Accent 1 of beˈfalla. The opposite speculation is offered by Riad (Reference Riad1998), who suggests that words with unstressed prefixes be- or för-, as in beˈfalla, were originally Accent 2 to match the Accent 2 of ˈpåkalla. However, this account does not explain why Nicander rhymed förˈdrifver1 with ˈgifver1. If förˈdrifver originally bore Accent 2, this rhyme would be inexplicable.
We feel that neither explanation accounts for the complete picture, which is that Nicander only considered the accent of the rightmost foot (from the stressed vowel on) for his rhymes and that accent structure has remained unchanged since. We suggest that beˈfalla1 rhymes with ˈpå |kalla2 and förˈdrifver1 rhymes with ˈgifver1 because the accents of the inflected stems (which include the final foot) are the same: ˈdrifver1 ~ ˈgifver1 and ˈfalla2 ~ kalla2. Our hypotheses are supported, as mentioned earlier, by Nicander's detailed descriptions of the use of accents in rhyming and the differences between the accents in simplex and complex words (cf. section 3). Further, our claim makes the strong prediction that Nicander could not rhyme stressed disyllabic prefixed words which have always been Accent 2, like ˈundan |tagen2 ‘exception=pl.def’ with a word like slagen2 ‘slap-past part’, although at first glance both forms match perfectly in accent. This is because the nonprefixed stem tagen1 ‘grip=pl.def’ in isolation bears Accent 1 and would not rhyme with slagen2. Further, Nicander's rhymes in the Appendix bear out our accent matching hypothesis.
Thus, Nicander's manual and verse constitute a valuable source of information on tonal accents in 18th-century Swedish. His own verse together with his comments on what ought and ought not make up a perfect rhyme indicate that Nicander was fully aware of the tonal accent contrast of his time. Based on his rhymes, we can conclude that the accent pattern has not changed since his time. Since matching tonal accents has never been common in verse, we are fortunate to have this unique source of information about Swedish tonal accent as it was two centuries ago.
APPENDIX
A. Simplex masculinee end-rhymes
B. Simplex feminine end-rhymes
C. Rhymes involving derivationally affixed words and compounds
ACKNOWLEDGEMENTS
This research was made possible by funding from the Deutsche Forschungsgemeinschaft (German Research Foundation: SFB 471 and the Leibniz-Prize awarded to Aditi Lahiri) and the Ministry of Science and Culture, Baden-Württemberg. Our grateful thanks to Frans Plank and Allison Wetterlin for hours of discussion on verse and tone, and especially for reading the manuscript with such care. We also wish to thank Patrik Bye, Tomas Riad, Carlos Gussenhoven, Astrid Kraehenmann, Christoph Schwarze and all the participants of the TIE workshop at Schloss Freudental, March 2005, for their critical and helpful comments. We are also very grateful to the anonymous reviewers for their many helpful views and comments. Special thanks to Ralph Hafner at the University of Konstanz library who made it possible for us to obtain a copy of Anders Nicander's Oförgripelige anmerckningar öfwer swenska skaldekonsten from Stockholms Stadsbibliotek. We also want to thank Mathias von Wachenfeld at Linköpings Stadsbibliotek who helped us accessing and copying Nicander's handwritten material referred to in this paper.
