Janez Orešnik Ljubljana 9LD ICELANDIC CONSONANT LENGTHENING R'ULE AND MODERN ICELANDIC INFIXATION OF /~ Survey and summar;y. § l: Old Icelandic had a Consonant Lengthening Rule in its morphological component, whose task was to lengthen postvocalic stem final l' Q, ~ in certain inflexional forms, e.g •. in the nom. sg. st611 of stol~ "chair". The rule is formulated in § 1.1, and illustrated in § 1.2, where some of its exceptions are listed as well. The formulation of the rule presented in § 1.1 is defended ·in some detail in § l. 3 •. In § l .4 the ordering relations of the rule are discussed (it operates earli er than an:y pb.onological rule), and in § 1.5 i ts gradual disappearance from the language is described. § 2: In Modern Icelandic, beginning before the Reformation, some of the functions of the Old Icelandic Consonant Lengthening Rule were gradually taken over by the (morphological) mechanism for the assignment of endings. The mechanism assigns, among other '1 things, an infix /~, which serves as an inflexional marker, either alone or in combination with suffixed desinences; e.g. Modern Icelandic nom. sg. stoll contains the infixed /~/ (not found in an:y other form of the word), which thus continues, in this case, the function of the Old Icelandic Lengthening Rule. §3 defends the position that, say, the ~ of the Old Icelandic dat. sg. f. storri is_not due to the operation of the Consonant Lengthening Rule upon stem final .!'. of st6r-, but to the substitution of an ending -E,Ei for the stem final .!'. of st6r-. Three arguments in favour of this position are stated in §3.2, where the diachronic origin of forms such as n±-tt, n,f-rri, Swedish pret. trodde, etc. is claimed to be a consequence of analyseš analogous to that of st6rri defended here. In §3.3 are listed some claims about language acquisition and the properties of grammars that are implied by the author•s positi.on concerning the derivation of storri and similar forms. § 3.4 briefly discusses the fate of some Old Icelandic ~ initial endings in the further history of the language. §3.5: § l and those parts of ff2-4 depending on §1 are to be· considered 229 Linguistica XIII prelilllinary. § 4.1 discusses the lexical ent:cy and the derivations of an Old and Modern Icelandic morphologically regular lexical item (ouinn "open") in whose derivations the Consonant Lengthening Rule applied in Old Icelandic. Also mentioned are the less known variants such as gen. pl. ~ for liainna, and .dat. sg. f. opnri for opinni. The importance of the latter type for the C?rrect formulation of the Modern Icelandic Vowel Syncope Rule is stressed. ~4.2 treats a few idiosyncratic lexical_ items: l:ltill; mikill (also dat. sg. f. mikl-ri), annarr (also gen. sg. f. annr-ar and dat. čauru(m)), ~; ~; lus, d:ls, etc. 1 1.1. Formulation of the Consonant Lengthening Rule Old Icelandic generative morphology contains a Consonant Lengthening Rule whose main objective is to lengthen stem final consonants in certain inflectional forms. The rule is stated in (1), q.v. It reads: Any postvocalic stem final /l n s/ is lengthened (1) Old Icelandic Consonant Lengthening Rule (Morphological) { (a) { 2' {JAV (b) :] ~ [+long] /V_ J#{(c) (d) in the strong cases: gen. pl., gen. sg. f., dat. sg. f. in short comparatives in the strong nom./acc. sg. m. of adjectives and in the nom. sg. of strong masculine nouns in the 2.3p. sg. pres. ind. if either Condition A or Condition B is satisfied: Condition A: The form is an adjective and either subcondition (a) or subcondition (b) is satisfied: Subcondition (a): The form is in one of the following strong cases: gen. pl., gen. sg. f., dat. sg. f. Subcondition (b): The form is a short comparative, i.e. one whose comparati ve marker does not begin wi th underlying /a/. ; Condition B: The form does not contain any inflexional ending, and it.satisfies either subcondition (c) or subcondition (d): Subcondition (c): the form is a strong masculine noun in the nom. sg., or it is an adjective whose stem ends in_!!, in the strong nom. or acc. sg. m. Janez OREŠNIK Subcondition (d): The form is in the 2nd or 3rd p. sg. pres. ind. "Stem final" means: •Word final or followed by an inflexional encling. The noun and the affixed article are separated by a word boundary on the phonological level (Orešnik 1972, which is valid, m.utatis m.utandis, tor Old Icelandic as well); e.g. in the definite acc. sg. hamar-inn "the hammer", !:. is stem final and word final.2 By adjectives I here mean all words that receive ad~ectival desinences in th.eir inflexion; thus, lllailj' pronouns, numerals and participles are adjectives. Similarly noUn8 are here words receiving substantival desinences in their inflexion. 1.2. Illustrations ot, and exceptions from, the Consonant Lengthening Rule. For examples ot rule (1), see (2). (2) Illustrations of rule (l) Case (a): ~- "blessed, happy": gen. pl. saell-a, gen. sg. t. saell-ar, dat. sg. f. saell-i ~- "iiiuniticent": gj?full-a, -~, -i ~- "fine, beautiful": vaenn-a, -~, -i opin- "open": opinn-a, -~, -i v:ls- "wise": v:!ss-a, -.!!!:, -i ~- "alternate, different": :fmiss-a, -~, -i Case (b): ~: short comparative stem sael~, e.g. saell-i s.i2ful-: gj?tull-, e.g. g.i2tull-i (and g,i<(flar-i) ~-: ~-, e.g. vaenn-i opin-: ( opn-ar-i) } v:Ls-: ('v:i'.s-ar-i) no suitable ~-: __ · examples Case (c): stol- "chair": nom. sg. m. stoll djgtul- "devil": dj<(full ~-:~ ~-: gj2tull ~- "stone": ~ ~- "heaven": himinn ~-: ~ (a:cc. sg. m. vaen-an, *~)° opin-: opinn, acc. sg. m. opinn ~- "one, alone": nom. sg. m. einn, acc •. „ sg. m. einn, ~ l.s- "ice": iss , ( ~-=~ ~-:~ Case (d): skin- "shine": 2.:;p.. sg. pres. ind. skinn ~- "pour, pump": eyss fil- "will n: fil! ?7;1 Linguistica XIII The rule has a number of exceptions, notably from cases (c-d); see (3). Notice that cases such as 2.3p. sg. pres. ind. ~ of (3) Examples of exceptions from cases (l c-d) Case (c): No lengthening of. the stem final /l n s/ in the nom. sg. m.: ,22!!- m. "son": nom. sg~ .22!! besicie ~ and very rare ~. ~ (Noreen 1923:201) Magnus:- propeJ:'. name: nom. sg. Map;niis, *Map;nuss For additional, quite numerous, examples see Noreen 1923:210, 248-49, 268, etc. Case (d): no lengthening in the 2.3p. sg. pres. ind. :. vil- "will": 2.3p. sg. pres. ind. sometimes .Yi1 beside the regular .!.llJ;. (and .!il.::!'.) sk:Ln- "shine": sk:Ln beside the regular stinn fr,i6s.., "freeze": fr:fs beside the regular fr:fss, cf. Noreen 1923:210. venja "accustom" instead of ~. or ~ of ~ "steal" beside stell, etc., are not exceptions from the Consonant Lengthening Rule (1), but involve alternations of endings: stel-r.contains the ending /r/, and thus does not satisfy the structural description of rule (1), case (d), which requires that the consonant to be lengthened is WORD final; the variant stell has no ending. Similarly, the 2p. sg. pres. ind. ~ of vilja "will", contains the ending /t/; there is also a variant with the ending /r/, .!ll.::!'.• and two without any desinence, vill, with regular lengthening of l, and vil (Noreen l923:360); only the last form is an exception from rule (1), cf·. (3). Nominative singular masculine forms such as~ "swan" and ~ "mild" are parallel to ~' ~ as far as lack of lengthening is concerned. 1.3. Comments on the formulation of the Consonant Lengthening ~ The consonant to be lengthened by the Consonant Lengthening Rule (1) is postulated to be preceded by a vowel: Post-consonantal /l n s/ .are not lengthened in the environments indicated, cf. nom. sg. m. fr,ials "free", not frjalss. The rule operates only upon stem final segments; e.g. the adjective ~- "loose" does not become llaussi, with long l• in the dat. sg. f., although the l of its 232 Janez OREŠNIK underlying representation /laus+i/ satisfies all the conditions of case (l a), excep,t the stipulation that the segment to be lengthened must be stem final. In the.cases (a-b) the consonant, to be lengthened must be followed by a vowel, .otherwise ;no lengthening takes: place; cf. gen. pl • .!!E::!:!! of ~- "accustomed" ,' *vann•ra. Case (a) of (1) must be limited to the strong cases of adjectives, for no .lengthening takes place in the weak gen. 'pl., jgen. sg. f~, and dat. sg. f.: weak gen. pl. ~ of ~~ "happy, blessed", not saell-u. - Case (a) is limited to adjectives, because nouns do not undergo lengthening in the gen. pl., ~en. sg• f., and dat. sg. f.; cf. ~- "stone", gen. pl. stein-a, not steinn-a. - Case (a) is limited to the gen. pl., gen. sg. f., and dat. sg. f., because there is no lengthening in the remaining case forms; cf. nom. sg. f. 159mul, not 6.2mull; of gamal- "old". Case (b) of (1) is limited to the SHORT comparatives, because no lengthening is observed in the LONG comparatives; the comparative of the adjectival stem v:fs- "wise" is long, v:fs-ari, and there is no lengthening of ~ (*v:fss-ari). And there is no lengthening of _E; in the long compara,tive opn-ari of opin- "open", as follows from the fact that the vowel i before _E; has unclrgone vowel syncope; vowels before lengthened consonants do not undergo sincope, cf. dat. sg. f. opinn-i. Definite weak noun case forma such as nom. sg. ~ "the rooster" is to be analyzed as hani##nn, not han#1'inn, as proved by the definite acc. sg. form hana-nn, showing that the final vowel of the noun stem wins over the initial vowel of the affixed article. There can be no lengthening of _E; in ~. because i ts _E; 'is not word final. Case (c) is limited to non-verbal in!lexional forma. Speaking of nouns first, case (c) has to be limited to the nom. sg. m., for case (c) does not apply outside nominative, or singular, or masculine, of nouns. Cf. acc. sg. aptan C* aptann) of aptan- m. "evening 11 ; nom. sg. sin ( * ~) of sin- f. •sinew"; nom. pL ~ ( * .l2!!!!) of ~- n. "kin". As to the adjectives, there is a very special provision for adjectives whose stems end in _E;: the latter is lengthened in the strong acc. sg. m. if word final; therefore acc. sg. m. ~. but ~ (* einn-an), both of ~- "one, alone". Stem final l and ~are not lengthened in the strong acc. sg. m., but this fac·t need not be mentioned extra, for there are no adjectival stems in l or ~ which satisfy the structural description of the Consonant Lengthening Rule AND stand in the strong acc. sg. m. Apart from the special case just mentioned, the stem final 233 Linguistica XIII E;l:~ are also lengthened in the strong nom. sg. m. The limitation. to this case form is necessary, because - apart from the situations enumerated in cases (a-b) and the Just discussed acc. sg. m. - there is no lengthening of word tinal l.!?;~ outs±de the nominative, or the singular, or the·masculine, or if the l ,!?;~are not word.final. Cf. nom. sg. t. and nom./acc. pl. ggmul ( *gomull) of gamal- "old"; xiom. sg. m. f~l-r ( * r21l-r) of 4ll!l- "palea (1 is not word. final). Case (d) has to be limited to non-:first persons o! verbs because no lengthening· takes place in the :first persons; cf. lp •. sg. pres. ind. sldn ( * sk:lnn), which is minililally dif!erent from the 2.3p. sg. pres. ind. sldnn •. Case (d) must be limited to the :forms Without endings because no lengthening o:f the stelli.:final consonant can be observed if it is followed by a desinence; c!. ~ ( *venn-r). The limitation to the siilSlllar is aut.omatic because there are no plural finite forms which would satisfy the structural description o:f the Consonant Lengthening .Rule (1). Case (d) must be limited to the present tense: the 3p. jsg. pret. ind. ~ of ~ "steal" does not undergo lengthening of the stem :final consonant ( * .§!fil). Case (d) is limited to the indicative mood because, say, 2p. ! sg. impera- tive, e.g. stel of stela, does not undergo lengthening ( *.~Ell)• 1.4. The ordering relations of the ConsonantLengthening Rule (1) The rule must follow the morphological rules which assign endings. This follows from (1), cases (c-d), where the Consonant Lengthening Rule (1) needs the intormation that the word on which it operates contains no ending. Unless the scanning power o:f the rules is greatly increased, such knowledge is available only a:fter all the rules which assign endings have applied. On the other hand, rule (l) precedes the (phonological) Vowel Syn- cope Rule. The lengthenilig of the stem :final consonant in, say, dat. sg. gamall~i must take place before the .Vowel Syncope Rule applies, other- wise the latt.er rule would delete the· unacce~ted .!!. o:f the form, and yield * gaml-i. 3 As far as it is known at present, the Vowel Syncope Rule is the earliest rule in Icelandic generative phonology. I have so far not found anything contradicting the assertion that the Consonant Lengthen- ing Rule (1) precedes all the phonological rules in the ordering. I have characterized the Consonant Lengthening Rule (1) as belonging to the morphological component. In this I have been led by the 234 Janez OREŠNIK circumstance that it precedes all the phonological 'rules in the ordering, and b7 the tact that the rule is tull ot morphological ini'ormation. In the absence of a principled criterion delimiting morphology .from phonology, 'J113' characterization can only be tentative. 1.5. The tate of the CollSonant Len.gthening Rule (l) During the development into Modem Icelandic the Consonant Lengthening Rule (l) ·gradually lost its domain, and it is no longer in the language. In some cases it was replaced by the infixation of /~, see section 2 below. Each case of rule (1) will now be treated separately. Case (a). In word forma such as gen. pl. saell-a, gjgfull-a, vaenn-a, etc. the lorig l developed into /~l/, and long ~in many illStances into /~n/, so that ~-infixation has been in the grammar to help produce these and such forma at least since the fourteenth century, from which period are found the earliest preserved reliable examples showing the change as accomplished (Benediktsson 1963 a:- 158). In words with. stem final ~· e.g. gen. pl. v1ss-a, Ymiss-a, the endings -_!, -.!!:E_, -i were replaced by the regular adjectival endings -~, -~, -.!i_, respectively, around 1300 (Noreen 1923:200), where- upon suchwordsno longer satisfied the structural description of the Consonant Lengthening Rule (1). Not enough is known ·as yet about the time when the long non- final ~ of case forms such as gen. pl. opinna,·of opin-, was shortened. The unaffected modern pronunciation admits only short ~ after an unstressed vowel (see, e.g., Benediktsson 1963 b:l48, on Ulfana vs. ulfanna). On the other hand, Bandle (1956:104) considers it quite possible that etymological long ~ was still preserved in the said position at the time of the writing of the Guabrandsbibl{a, published in 1584. .A:D. investigation of this chronological problem is a desideratum. In conclusion, case (a) of the Consonant Lengthening Rule (1) fell into disuse gradually. It disappeared from the language when the long ~ of unaccented syllables had been simplified in pronunciation. Case (b). All short comparatives which in Old Icelandic underwent case (b) of r\lle (1) contained stem final l or ~· riever ~·. (The adje,c- tives in~ formed long comparatives in-~.) In the fourteenth century even the 11 and ~ of these comparatives were changed to /~l/ and /~, whereupon case (b) of rule (1) was superseded by ,9:-infixation. Case (c). Nominative singular case forms such as stoll, djgtull, 235 Linguistica XIII ~. gj!Cfull, ~. ~· of (2c), iinderwen:t the :ehange of Q to - fil, and !!l1 to .2E,, in_the fourteenth_ century. In th:i,s function .case (c) has been substituted for by ,9;-infixation. Nomi-native singular .C!l-Se forms such as ~. opinn lost the length of their word final -~- by the sixteenth century at the latest, if not even earlier (Bandle 1956:105-8), through a sound-law simplifying word final !!J1 preceded by an unaccented vowel into short .!!• Case· (c) thus becB.lD.e unnecessary :for such instances. A similar, presumably even earlier, change, wi th similar consequencea, ai':fected i, the nom. sg. :forms iss, viss, ~ (Noreen 1923:210). In conclusion,·case (c) disapp~ared :from the language in the sixteenth century at the latest. Case (d). Forms such as 2.3p. sg. pres. ind. aldnn lost the length o:f thei~ :final consonant before the :fourteenth century, so that they could not participate in the change o:f !!l1 to dn. The process which brought this change about is not well understood (Noreen 1923:210). Forms such as eyss simplified their ~ at the same time as :Lss, v:Lss, Ymiss, at the latest. Forms such as vill were almost exceptional in Old Icelandic; many were early replaced by .forms containing the normal 2.3p. , sg. pres. ind. ending -_!'., e.g. stell o:r ~ "steal" became ~. and fell outside the scope of case (d), since their l was not word final an:y longer. The 2p. sg. pres. ina'.. vin was early replaced by ~ and likewise fell outside the scope of the rule. Only the 3p. ~ remained and underwent the regular phonetic change of ]d to dl in the fourteenth century. The :form~ is so exceptional that it is even uncertain whether its /d/ is now produced by d-infixation. It is ll!Uch more likely that vill i; ·a suppletiv~ formati;;-n. Thus case (d) disappeared from the language in the :fourteenth century at the-latest. 2. On the infixation of d in Modern Icelandic. In Modem Icelandic generative morphology the role of the Consonant Lengthening Rule (1) has been partially assumed by new in:flexional endings involving the infixation of /~/. For instance, gen. pl. saell-a [saI~lal , of ~-, is now analyzed as the stem ~- with ,9; in:fixed before the stem :final l• plus the genitive plural case desinence /a/: /saI~l+a/; the :form contrasts with the acc. sg. f. ~. without the in:fixed /~/. In the nom. sg. m. ~ [saI~p , the ending consists o:f the in:fixed ,9; only; contrast nom. sg. f. ~. without in:fixed .!!· Generally speaking, the sound-laws producing fil < Q _and .2E, < ~ have brought about two new declensions in Icelandic, one involving masculine noun, and another adjective, stems in l and .!!:· For examples see (4). 236 Janez OREŠNIK (4) Illustratioli.s oi the infixation of .,S: ~- "happy, blesse'1", strong nom. sg. m. afil, gen. pl. saell-a,. gen. sg. f. saell-ar, dat. sg. f •· saell-i (colloquial vari- ants saell-ra, -.!:.!!!:, -.!'.!) ·. ~- "big, great": ~. m:ikill-a, mikill-ar, mikill-i (colloquial variants mikill-ra, -.!:.!!!:, -.E) stol- "chair": nom. sg. st611 gaffal- "fork": gaffall ~- "fine·, beautiful": ~' vaenn-a, vaenn-ar, vaenn-i (coll. vaenn-ra, -.!:.!!!:, -.!'.!.) ~- "stone": steinn The membership of non-verbal stems in the two declensions involving infixed .2: is predictable. Anon~verbalstem is inflected in the new wa:y if it ends in an accented vowel plus (short) l or _g, whereby the accented vowel has to pertain to the following set: the diphthongs, ~,i (z), and i· (For ~ see belowl.) Cf. ~-, stol-, ~-, ~- of ( 4). The colloquial genitive plural case forma such as saell-ra, vaenn-ra, gen. sg. f. saell-rar, vaenn-rar, dat. sg. f. saell-ri, vaenn-ri, all with infixed .2:• contrast with gen. pl. gul-ra, of ~- "yelow", and ~· of ~­ "accustomed", etc., without _2:. - In addition, a non-verbal stem is in- flected in the new wa:y if it ends in an unaccented vowel plus l• cf. gaffal-, ~- of (4). Pairs such as saell-ra vs. gul-ra of the previous paragraph are evidence against a Modern Icelandic morphological rule of .2,-infixation. Such a rule would have to mention the set of vowels consisting of the diphthongs, ~' i C;i), ,i!. This set, however, does not seem to form a nat- ural. class in the sense of Halle 1958 (1964:328 l , Chomsky & Halle 1968:335ff._, and no other rule of Icelandic grammar seems to have to ex- plicitly mention the same set of vowels. For this reason a rule listing this set of vowels must be-suspect if no case in its favour can be made. The set of vowels just discussed includes ~· Cf. the old loanword klen- "snug, little", with _9: infixed as in certain forms of ~aen- in (4). Cf. also the more recently naturalized adjectival stem ~- "nice, neat, handsome", with nom. sg. in. ~Cl}l_g_ or m (Boavarsson 1963 s.vv. m, penn; my attention has been dra~ to m- by Tryggve Skomedal, viva ~' 1966); I lack data on the strong gen. pl., gen. sg. f., and dat. sg. f. of this adjective. I know of no other Icelandic noun or adjective stems ending in accented ~· Incidentally, m- is not the only stem of the appropriate kind displa:ying optional absence of _9:-infixation. For instance, 237 Linguistica XIII pjon- in. "s.ervant, waiter'', which now has .a regular nom. sg. m. pjonn, with d, often did not undergo this infixation in the sixteenth to nine- teenth centuries (P6r.6lfssori 192.5:77). . There is also a single case of /1/ ~ /~l/ in the conjugation: the :;p. sg. pres. ind • .!fil, as against, sa;y, the lp •. sg. pres. ind. fil, of vilja "wish, want" •. The form .!fil can be accounted for on the lexical level, .see (5). Other more or less idiosyncratic trai ts· of vilja, such as-!. of the 2p. sg. pres. ind., ,,i-initial endings in the present out- side the singular of the indicative and imperative moods, and -~ in the supine, are to be accounted for by the mechanism for endings, and are undoubtedly to a great degree also regis.tered in the lexical entry of which (5) is a part. (5) Partial lexical entry o~ Modern Icelandic vilja: /vidl/ in the :;p. sg. pres. ind. /vil/ elsewhere 4 The adjective/numeral/pro~oun einn "one" has two strong ac.c. sg. m. forms, einn and ein-an, the latter being used predicatively in the meaning "alone", Eina;/'.'sson 1945:6.5. The (only) accusative singular mase. of .the indefinite pronoun ~ is ~· Since the two acc. sg. m. case forms ~ and ~ are exceptional in another respect as well (no ending instead of, or beside, the expected -~), they are most likely to be accounted for on the lexical level; this is especially clear with einn, where the difference between the acc. ~ and ~ is associated with a difference in meaning and syntactic behaviour. The cases of infixed .2: not covered by the new declensions either stem from lexical representations, or are due to the phonological sandhi rule which inserts a .2: in the context /r~n, 1/ (the rule is not treated in the present paper; notice that it must follow the vowel syncope rule in the ordering, Orešnik 1972:24). The "lexical" case can be illustrated by villa [vidlaJ "error" (as against :ti:lli c:..r1:aJ "villa"), and innumer- able other examples. 5 The declension of E. and l final masculine noun stems which undergo .2,-infixation can be ad hoc called NOUN DECLENSION (A), and the nouns which are inflected according to declension /A/ must be marked with [+Declension (A)]. Thus, stol-, gaffal-, and ~- of (4) must be marked with [+Declension (A)J , whereas, say, ~- "swan" and ~- "whale" pertain to another declensional type (inom. sg. svan-ur, hval-ur}. Biskup- "bishop", with its lack of ending in the nem. sg., belongs to a third 238 Janez OREŠNIK ,declensional type. The noun stem pjon- must o'nce have vacillated, as far a.S its nominative singular is concerned, between the declensional type of biskllp- (nem. sg. pjon) and declension (A) (nem, sg. pj6nn). - The relationships among the sundry declensional types of nouns and adjectives will not be disctissed here. The declension of the B; and 1 final adjective stems which undergo ~-infixation can be here ad hoc called ADJECTIVE DECLENSION (B), and the adjectives inflected in this way m.ust be marked with the feature [+Declension (B)] • Thus, ~-, E&lli-, · and ~- of (4) are marked with [+Declension (B)J , whereas, say, .sBd:.- "yellow" and _!!!!!- "ac- customed" are marked as pertaining to another declensional type; fil- "expensive", ·megn- "pungent", etc.·, whose strong nom. sg. m. equals their respective stems, belong to a third type. The adjective ~­ vacillates, at least as far as its strong nem. sg. m. is concerned, between the declension of fil-, ~-, and declension.(B). Just as adjectives are lexically subcategorized into several declensional types, they must be lexically subcategorized into comparison types. For instance, ~- and ~- of (4) belong to a comparison type which includes Q.-infixation in the comparative: comparati ve .saell-i, vaenn-i, etc. Their colloquial varianta saell-ri, vaenn-ri can be contrasted with comparatives such as falleg-ri of falleg- "handsome", with -& and without Q.-infixation. - There is no straightforward relationship be.tween the comparati ve type of ~-, ~-, and what has above been called declension (B). Cf. vesal- "wretched", comparative :vesl-ari/vesall(r)-.i;, and ~- "sick", comparative vesaell-i., \ 3.1. Lengthening of stem final· ·r. Historically and structurally parallel to the examples of (2) are quite as many with stem final .!:• Some are enumerated in (6), q.v., where the letters (a-d) refer to the respective cases (a-d) of rule (1) whereas (e) has no correspondence in ( 1 ). Dat. sg. f. st6rri, to take .one of the items enumerated in (6)., came into being in pre-literary Icelandic, its long .!: stemming from the amalgamation of the .!: of the ending -.ti:_·with the stem final .!:• The same is true of all the other case forma listed in. (6). Thus one way of treating these forma would be to postulate underlying, say, dat. sg. /st6r+ri/, parallel to, sa;y, J'.spak+ri/ of spak- "quiet, wise". The same is true of all the other cases of (6). However, I believe that the Old Icelandic strong dat. sg. f ~ storri must be analyzed, on the morphological level, so th:at the ending begins with a LONG _!, and is not 229 Linguistica XIII (9) Examples of lengthened stem final r 1 (a) star- "big, great": gen. pl. storr-a, gen. sg. r. st6rr-ar, dat. sg. r. storr-i okkar- "our (p.ual)i•: okkarr-a, ,...E, -i' (b) st6r.-: short comparative .stem ~-, e.g. stoerr-i. (c) st6r-: strong nem. sg. m. storr . okkar-: ~ (d) far- "go, travel": 2.3p sg. pres. ind. f.š!E!: (e) ll!l_- f. "gravelly bank": nem. sg. eyrr ~- "mare": ~ ADDED to the stem st6r-, but REPLACES its final .!'.· After the replacement, which takes place in the morphological component, we get /stor:i/, the phonological representation of the case form in question. Endings such as -.!'.& of n;f-rri, of E±- "new", which are added to stems, can be called ADDED endings. Endings such as -~ of storri, which replace one or more segments of stems, can be called REPLACING endings. In the morphological representations the segments to be replaced by re- ri1acing endings are here preceded by the REPLACEMENT SIGN (, 1 i. e. opening round bracket. The replacements ta1'e place within the morphological com- ponent, so that no replacement signs ~ppear in phonological, or later, representations. It is an_open questi.on what happens to the morpheme boundary between the segment(s) 1to be replaced and a replacing ending, during the replacement. - The same affix can be an.added ending in one form, s.nd a replacing one in another. As .far as I can see, an added ending and its homophonous replacing allomorph are related as follows. The replacing desinence is used if the stem final ·segment is phonetically similar to the initial segment of the termination, say, i.f the two segments are both oral dentals/alveolars; otherwise the. ending is added to the stem. Cf. sto(_!:+rri vs. n;f-rri. There ms:y also be cases where the distribution o.f the two allomorphs cannot be predicted in such a simple way; moreover, PHONETICALLY SIMILAR remains to be defined. 3.2. Arguments in favour o.f the analysis. The following can be said in favour of the replacing endings and o.f 'I!IJ' analysis o.f st6rri: (I) Old Icelandic nom./acc. sg. glatt, o.f glae- "glad", is here analyzed, on the morphological level, as /gla(e+t:/; i.e. in the morpho- logical component, the stem .fi.nal /e/ is replaced by the case desinence /t:/, to yield the phonological representation /glat:/, which minimally deviates .from· the systematic phonetic representation o.f the case .form 240 Janez OREŠNIK :glatt, and is ·therefore·, ceteri·s·· paribua, maximally motivated. The· alternati:Ve morphological representation /glaa.+t/ or /glaa+t:/ would' obliterate the need for positing replacin.S~ endings (beside added ones) in the theory, at least as far as glatt is concerned. At the same time, however, that representation would exact a phonological representation /glaa+t/ or /glaa+t:/, which deviates from the corresponding systematic phonetic represeiitation Of glatt more, and is consequently, ceteris paribus, less motivated, than the phono'iogical representation /glat:/. : , • 1 Furthermore; the phonological representation /glaa+t(!:)/ exacts the existence of a phonological rule producing /t:/ out of /a+t(:)/ in simplex words; there is a good chance that such a rule need not be posited at all if replacing endings are introduced into the theory. The example glatt shows that, while disallowing replacing endings leads, ceteris paribus, to a simpler theory of language, that degree of simplic- ity is achieired at the·cost of the motivation for certain phonological representations. In other words, we must choose between a simpler theory and more abstract phonological representations (e.g. /glaa+t(:)/) on the one hand, and a less · simpl·e theory and more natural phonological representations. (e.g. /glat :/) on the other. I opt for the latter, and thus for rep.lacing endings. - Once replacing endings are allowed in morphology, they can be employed whenever, ceteris paribus, their use leads to better motivated phonological representations than the utilization of some other tool does. (II) If the Old Icelandic dat. sg. st6rri is postulated to originate in the morphological representation /sto(r+r::V, and not from /stor+ri/, there is no need for a rule that coalesces /rr/ into /r:/ in simplex words, in Icelandic synchronic grammar (although such amalgamation can undoubtedly take place as a historical process; st6rri HAS arisen from st6r+ri, diachronically speaking). More generally, there is presumably no need for phonological rules coalescing whichever two equal consonants into one long consonant, in simplex words. Cf. however, ~-in §;;.:;. (III) It is generally assumed that the length of .!'. in n;f-rri did not come into being through a sound-law.·It is considered an analogical creation under the influence of storri and similar forms, where ,!:!: is lautgesetzlich·(Bandle 1956:302, with further references). If so, there arises the questi'on of the mechanism which helped create -m_ on the basis of st6rri. In mY -opinion it must be postulated that the learners of the language were unable to.recognize in storri the same ending as in spak-ri and n;f-ri, although they could perceive a similarity between the auslauts of stčrri, spak-ri, and n;f-ri. They identified - and this was 241 Lingusitica XIII the first crucial step - the rri o! storri as the ending, not to be added to the s tem st6r-,. but t;" repl~e ste~ final .! of ~-. Schematically: st6r- + -.!ti.~ sto-+ -.!ti.-7' st6rri. In this pro- cess a vowel final stem came into being temporarily, and the ending was added to it. The original distribution of the ending -.!ti_ must have been construed as, "added to the vowel finai adjective stems which had arisen from E.~!inal stems through the deletion o! the stem final _!:, in the strong dative singular feminine." Soon this dist:r:i- bution must have been simplified through the omission of the information on the source of the vowel final stems to which the ending -.!ti. could be affixed. This was the sec.ond crucial step. The desine_nce -rri could now be "added to vowel final adjective stems, in the ~trong d;ti°ve singular feminine." This led to the introduction of -.Eti,. into the strong dative singular feminine case !orms of v~wel final stems such as B±- "new": the dat. sg. f. n.f-ri was replaced by D:f-rri at the beginning of the literary period. The same analysis was made in the gen. sg. f. storrar (and led to n:f.:.rar > n:f-rrar), in the gen. pl. st6rra (and led to n:f-ra > i;f-rra), in the short comparative stem ~- (and led to Itf-r-> i;f-rr-, which existed besid.e n:fj-ar-, Noreen 1923:300), in the nom. sg. m. storr (and led to n;f-r > n.f-rr). The process also took.place in nouns, so that nom. sg. m6-r "moor; sea gull" occasionally became m6-rr, and similarly in some other vowel final substantival stems (Noreen 1923:253, 256). The type represented by the nom. sg. mo-rr, JJ,f-rr was Obliterated in the further history of the language, and it is doubtful that it had ever been as fully spread as the remaining innovations mentioned above. The obliteration of the type is presumably due to the .change which also shortened the '.! of nom. storr, rendering stor. · In the same vein, I submit that the ending of the gen. sg. huss of hus- "house" is not /s/, on the ana:iogy of the -~in, say, gen. sg. barn~s of barn "child", buta long /s:/, which is substituted for the fina1 ~ of the stem hus-: hus-+ -~' hu- + -~~ hilss. The genitive ending spread outside its original domain to vowel final stems: gen. sg. mo-s of mo-r "moor; sea gull" became occasionally mo-ss, and similarly with a number of other vowel final stems (Noreen 1923:253, 255 ( where the desinence -~ is said to be-even somewhat more frequent than short -~in the neutral vo:wel final ~-stems such as bu "house, estate"J, 256, 257 (gen. hirais(s), kvaeais(s)], 269). The situation was similar in adjectives: gen. sg. m./n. ~ was analyzed as containing the replacing ending -~, and the latter spread to vowel final stems, so tha'.t 242 Janez -OREŠNIK .there arose, at least occasionally, gen. sg. m./n. n;f-ss and other similar forms (Noreen 192~:291, 2CJG). In the further development, -~ was obiiterated in vowel final stems, both substantival and adjectival, so tllat now there are gen. m6-s and n:f-s, etc. The simplification of -~ to -§_ is probably due to the change which .also affected forms such as nom. sg. 1.ss "ice" > 1.s. In the stems in .2. the simplification did not take place in the genitive singular, so that ther.e are still gen. huss, 1.ss and lauss; the_ending -~ has been preserved for unknown reaso;;;- here (Bandle 1956:99, 100). I postulate a similar development in the strong nom./acc. sg. n. nf-tt. Cases. such as nom./acc. sg. n. hv!tt o:f hv!t- "white", :foett o:f :foedd- "born", ~ of breia- "broad", were analyzed as containing the termination long ,!, /t:/, which replaced the :final dental o:f the stem in the strong nom./acc. sg. n.: hv!t-+-,!!~ hvi-+-.:!:!~ hv!tt, etc. Then the requirement that the vowel :final stem to which -tt is added originate from a dental :final stem through the deletion o:f the stem :final dental was dropped, and the -.:!:! spread to other vowel :final adjectival stems: e.g. nf-t was superseded by nf-tt. A similar development may have taken place in the 2p. sg. pret. ind. acti ve o:f 01.d Icelandic strong preteri tes whose stems ended in a vowel; e.g. b,jo-t o:f biia "dwell" became bjo-tt. One :factor contributing to the introduction o:f -tt into vowel :final preterite stems such as ~- were 2p. sg. pret. ind. :forma such a.s slo-tt of sla "strike" and hlo-tt o:f hlae,ja "laugh", where the -tt is lautgesetzlich (:from stem :final *.as: and desinential -,!). (See Noreen 1923:205, 362). But a development parallel to that postulated :for the nom./acc. sg. nf-tt may have helped as well. Even in oldest Icelandic, strong preterite stems ending in vowel + !_, occasionally also in vowel + ,!, had the termination -tt in their 2p. sg. pret. ind.: preterite stem reia- o:f r!aa "ride", 2p. reitt; s tem stoa- ·of ~ "stand", 2p. stott, etc. From such cases -tt may have spread to vowel :final preterite stems such as ~-, yielding 2p. bjo-tt. This termination would undoubtedly have survived into Moderzi Icelandic, had it not been superseded by·another ending, -st, which came into use in the :first hal:f of the sixteenth century (P6rol:fsson 1925:111). Similar developments can be observed outside Icelandic. A well known case in point is the disputed origin o:f the so-call.ed third conjugation preterite stems in -~- in Norwegian and Swedish, e.g. Swedish tro "believe", pret. trodde. It has long ago been observed that verbs with regular ~ in their preterites, such as medieval Swedish :ffdha "give birth"' (pret. :f:0dde), must have played a role in the creation o:f the 243 Linguistica XIII ,preterite stems such as~· For the history of the problem, see Jansson 194.7. C:f. also Bandle 1973, with further refere:iJ.ces. Preterite st~m fpdd- must have been analyzed as present stem f~dh- minus the ste~ final dh plus the. preteri te suffix -§&-: fPdh- + -§E;.-7 f~- + -dd-~ fpdd-. The preterite suffix -dd- was initially affixed only to vowel final weak preteritival stems that had arisen through the deletion of the stem final :!!; in preteritival stems ending in vowel plus ~· This distribution was modified so that the requirement con- cerning the origin of the vowel final stems to which -§&- could be added was dropped. Subsequently -§&- began to be affixed, in the appropriate morj>hological .context, to any vowel. final stem, also to .E'.2-, etc. What was above several times re.f'erred to as "the simpli.f'ication of the distribution" of a suffix .do.es not necessarily take place. For instance, in Icelandic the same. conditions as in Swedish and Norwegian obtained in the verb. types exemplified here by f@dha and .E'.2· yet the distribution of -dd- was not broadened to include any type of vowel final stems. On the other hand, the process can remain limited to a dialect or to a f.ew lexical items in a dialect. Such seems to be the situation with respect to preteritival -dd- in Danish, where preterites of the type ~are preponderantly found in s6stdansk dialekt. See Brs6ndum-Nielsen 1971:177""'.78, also 175 (~), 413 (~), 414 (trodde), 424 and 428 (hadde), 444 (titte). 3.3. Claims about language acguisition and grammars. The process .,as I have outlined it above makes certain claims about. language acquisition and grammars: Language. learners cannot analyze a long cohsonant as arisen from two subsequent instances of the same consonant separated from each other by a morpheme boundary (e.g. st&rr ,P.. stor+r), although they assumedly can perceive the phonetic similarity between two allomorphs of ihich one contains a long consonant C where th__e other contains a short C (e.g. th.e similarity between -,ti. and -.!:,E)• One possible way of adding endings to stems is by substituting a consonant initial· ending for the last co.nsonant (consonants?) .of the stems ending in vowel plus consonant. The. vowel final stem which is produced as a transitory phenomenon .in this process has sufficient psychological reality to.play a role in the description of the distribution of the ending.addedin the l!\anner just outlined. 6 The identification of stems can be Iliade by B.n appeal to what has happened.to those stems at earlier 244 Janez OREŠNIK stages of . del'i vations; however, the gr.ammar favours the omission of such clues from grammatical statements (rules),. as shown, e.g., by the. development trodhe > ~ d.escribed .· above. There is no overriding tendency for languages to hold the number of allomorphs of e:ny ending at a minimum. (Otherwise the ending -~ would not come into being.) Language tolerates much allomor:phic suppletion in endings. The creation of vowel final stems out of consonant final stems through the deletion .of the final consonant and the addition of .an ending to the vowel final stem thus created are two relatively independent processes. On the other hand, the historical (diachronic) amalgamation of tw0 equal consonants into one long consonant seems to proceed:unimpede(\, as forms such as dat. sg. f. storri (from pre-litera.iy stor-ri after the syncope of the vowel that once intervened between the two ;!:'s) witness. It is only when confronted with a.finishedform of the type st6[r:J1_ that language learners cannot reverse the historical process: Two equal consonants amalgamated into one long consonant are irrecoverable. It is possible that the claims made in the present section are only valid for languages in which short and long consonants contrast in simplex w.ords, such as Icelandic, Norwegian, Swedish, Old English, Finnish, Italian, Latin, etc. - There is no indication so far that the unanalyzability-of-long-consonants principle is also valid for such "late" long consonants as i'ound in, say, the contracted stem ~- (bYn :- J oi' bundinn "bound". The mechanism which has here been assumed to have brought about the change of, say, n.f-ri to nf-rri is not inherently limited to cases involving endings beginning with long consonanta. Rather, the mechanism can create "analogical" forms whenever vowel i'inal stems obtain during the affixation of replacing endings. However, cases not involving iong consonants are doubtful in Icelandic. Thus Haegstad (1942:35) reports having heard seconP. person singular preterite indicative forms dr6Cxlil of draga "draw", and slo(x]il of sla "beat", in Southern Icelandic dialect at the beginning of this century. Both forms rhyme with the 2p. sg. pret. ind. tok-st of ~"take". I interp:Z:.et the two dialect forms as having arisen through the addition of the desinence -~ to the re~pective preteritivai stems dr6- and slo-. I see the origin of -~in verbal forms such as tok-st, where final Gest] arose through a sound-law spirantizing ·the stem final stop ~ to [x] before st. At least some speakers of Icelandic must have been unable to perceive /k+st/ in the phonetically realized [xst] of t6k-st (although ~ appears on the phonetic level in the remaining forms of the 245 Linguistica XIII .pret. ind. paradigm: tok, t6k-u.m, -.:!!, -J;!), and have therefore analyzed the form as /t6(k+xst/, with the replacing ending./xst/. (Notice that the Ji to be replaced by xst is phonetically similar to the initial segment of the new termination -~.) During tlie production of to(x]~ a vowel final stem comes into existenqe: t6+~. The original distribution of the ending -~ is thatit is affixed to vowel .final stems which had arisen from consonant finli.l ste!llS through the truncation of the stem .final consonalit. The modified distribution ofthe ending is that it is affixed to vowel final stems. As a resU:lt -~ begins to compete with the ending traditionally affixed to vowel final stellis in the second person singular preterite indicative of strong verbs, -st, and in a: few cases ~[x]st, dro[x],!il) wins the battle in dialect-:- If this process took place where underlying /k/ and /s/ were involved, it is conceivable. that it occurred even when· /p/ and /s/ came into contact under similar conditions. The 2p. sg. pret. ind. of the strong verb supa "sip" is rn[f]st, a historically correct form. Some speakers must have analyzed it as /sčii(p+fst/, and from such cases a new ending -f.!il could spread to vowel final stems. Haegstad (1942:46, fn. 3) reports, from Fljoi;sdalur, East Iceland, dialectal 2p. sg. pret. ind. sa(f]~ (actually he states that the form is pronounced "almost" as just shown) of ~ "see", literary form sa-st, without any f. s16CxJst, dro(x]~ against sa[fJ~ contain com;peting replacing endings. There could just as well have arisen, say, slo(f]st and sa[x]st. A similar phenomenon can be observed in the Norwegian-Swedish third conjugation. There -dd- was generalized after vowel final stems. However, a replacing allomorph -.:!:!.- was also available: the preterite of Swedish beta is ~. morphologicallyi/bo(t+t:e/, and from such forms -tt- can spread to vowel final stems. Forms such as n~tte instead of the more usual preterite n~dde of n~ "reach" are to be found in the Bahuslen and Southern Ostfold dialects, see Seip 1949:97. Consider now the gen. sg. ~Cf]sins "the .year1' (wi th affixed article), which Haegstad (1942:46) reports from East Iceland (most probably from Fljotsdalur). The literary form is arsins, with voiceless .!:.• Haegstad implies ibidem in fn. 3 that the form is pronounced assins in Fljotsdalur. The unusual alternative ~[f]~ must be due to the substitution of -.!§. for -~· The origin of -.!§. must be sought in those genitive singular forms of l2. final nominal stems in wliich /p+s/ resulted in (fsJ ; cf. gen. sg. «kin«, pronounced [sk.IfsJ beside Csk.I:ps], of skip "ship"; gen. sg. = . J J kaups [-fs] and [-psJ of kaup "bargain", skaps [-.fs] beside [-ps} of 246 Janez OREŠNIK .skapur "wardrobe", etc. KauCtJs must have been analyzed as /koi(p+fs/, and from here /fs/ spread'to /i;. 7 3.4. The fate of the rr initial replacing endings. The Old ~celandic replacing ending -.!:.!: was abolished in the further history of the langliage, in the fourteenth century after an originally long vowei, a little later. after an originally short vowel: storr> stor, ~> l.2E.· In feminine io-stems a new ending superseded -.!:.!:: the eyrr, ~ of (6 a) are now eyr-i, ~- The Old Icelandic replacing endings -~, -~, -~ etc. preceded by an unstressed vowel have had their .!:.!: shortened by modem times, although it has not yet been ascertained when exactly this happened: annarrar > annarar, etc. The remaining .!:.!: ini tial replacing endings have been preserved: nfrri, ~. staerri~ etc. 3.5. On an alternative to the Consonant Lengthening Rule. Con- sonant Lengthening Rule (1) states a list of endings, their distribution, and the generalization (whose significance is not proved) that the endings in question somehow belong together. The option remains open that the functions of the rule should be entrusted to the mechanism for the assignment of endings. In that case we need replacing: endings such as /1:/, /n:/, /s:/, similar to the /r:/ initial endings discussed above. The morphological representation of, say, the Old Icelandic strong nom. sg. m. saell would then be /saI(l+l:/, and its phonological representation /sail:/. This solution guarantees just as natural phonological representations as the alternative one utilizing the Consonant Lengthening Rule. It is only fair to mention that certain factors actually speak in favour of the solution just outlined: (a) Notice that, although Lposit the ending -ss in the gen. sg. :!ss, lauss, I d~ not analyze the corresponding nominativ-;-singular forms fss, lauss as containing the desinence -~, but a lengthened stem final §.· This difference in treatment is based on the circumstance that the gen. sg. :-~ has had a near parallel and support in the -§_ of the genitives such as barn-s, spak-s, whereas the -~ of the nominatives was not supported in this way; the nominative·~ viewed as lengthened stem final .§. had the structural support of 11, !!!l; in the types saell, ~· whereas the genitiva! -~ enjoyed no such support. - I evaluate.this type of argument as not decisive, seeing that the para1lel cases involving .!:.!: do not lead to a:ny difficulties although I do not relegate them to the Consonant Lengthezii.ng Rule when arguments such as those just stated would warrant that. 247 Linguistica XIII (b) The .gradual erosion o:f the Consona.nt Lengthening Rule (1) described in §1.5 raises the question o:f the successive intermediate :formulations o:f the rule during its disappeara.nce by degrees. The abolishment does not seem .to :follow any imma.nent logic, but is mostly governed by sound-laws; where it is not, as in the above mentioned case (in §3.2) o:f the nom •. sg. {s ( <: :l~s) vs. gen. sg. :f.ss ( < :lss), the natural impression is that the change o:f nem• :lss to :ls is due to the substitution o:f the ending -& :for the replacing ending long ~. a.nd that the structure o:f the Consonant Lengthening Rule has no say in the process. For these reasons the :formulation o:f the C.onsona.nt Lengthening Rule and whatever :follows :from that :formulation has a preliminary character. 4. A regular paradigm and special cases 4.1. The lexical entr;y of OPINN We can now proceed to the sample derivations o:f the members of a regular paradigm, and then discuss some exceptional cases. I begin with the derivations of the sundry case forms of any Old Icelandic adjective in -inn (opinn "open", kristinn "Christian", etc.). The lexical representation of such adjectives contains just one stem, say /opin/, associated with the rule feature [+Vowel Syncope Rule]. In the morphological component first the endings are added to the stem when applicable, e.g. in the gen. sg. m. /s/, in the acc. sg •. f. /a/, in the gen. pl. likewise /a/, etc. In the strong nom./acc. sg. n. the ending is not added to the stem, but the stem final /n/ is replaced by /t/: opin- + -!~ .QE_- + -! ~opit. 8 Next the Conso.nant Lengthening Rule (1) applies in the strong nom./acc. sg.m.,gen.pl.,gen.sg.f.,dat.sg.f.; e.g. dat. sg. f. :/opin+i/ becomes /opin:+i/ by (la). In the phonological component, first the Vowel Syncope Rule applies, say, in the strong acc. sg. f., changing /opin+a/ into /opn+a/ opna. Further phonological rules will not be discussed. Sample derivations are to be found in (7), q.v. During the development into Hodern Icelandic, the long .!.! in unaccented syllables was shortened, so that gen. pl. opinna, gen. sg. f. opinnar, dat. sg. f. opinni, now contain no long consonant before their respective case desinences -~, -~. -i· Nom. sg. m. opinn rhymes with nom. sg. f. opin. In the nom./acc. sg. n. the final ! became _! through a sound-law, so that the :form is now opia. The modem lexical representation of opin- is /opin/ marked with [+Vowel Syncope Rule], except that the Vowel Syncope Rule must not operate in the strong gen.· 248 Janez OREŠNIK .pl., gen. sg. f., and dat. sg. f. although the respective· terminations begin with vowels. The bloC:king of the Vowel Syncope Rule in the three cases is probably to be described as emanating from·· the respective endings, which cantherefore.be ad hoc marked as -s_-:-VS, etc. c:r. footnote 3. (7) Partial derivations of some case fo:r:ms· of Old Icelandic OPINN . nom./acc .• · dq.t. gen. pl. npm. nom./acc • gen. sg. sg. m. sg. n. sg. f. sg. n. m./n. opin oi:iiii+u opin+a opin opit opin+s Rule(la) '- opini.+.a Rule (le) opin: Vowel Syncope opn+u orthogr. repr. opinn. opnu opinna opin opit opins Occasionally the vowel syncope.has taken place in the three cases as well. Thus there is gen. pl. lYEina of ~ "past, gone (of time period)" in a manuscript containing folk songs, probably written in VestfirElir, ca. 1670 (Helgason 1962a:XVI,3); var .• l. ~ (in a syntactically. spoiled context) and.~. Further, there is dat. sg. f. ~ of ~ "very big, over-.great", in a folk song translated from Danish and .p.reserved in a ms. written 1819 (Helgason 1962a:XXIV; 1962b:l22); in. the next stanza dat. sg. f. aernri occurs in a lin,e that is a repe ti ti on of the line containing aerni. Similar forms are even mentioned in the grammatical literature; thus Guamundsson 1922:89 adduces gen. sg. f. ~ of ~ "old" (the chapter and ve.rse o:r the form are not quoted, but the author states that he has :found it in a text); Porkelsson 1902:56 mentions gen. sg. :r. fallnar beside :fallinnar o:f fallinn "fallen". In the nineteenth century, Icelandic philologists must have had some knowledge of such forms, as witnessed by the fact that some scholars, e.g. Sveinbjorn Egilsson, even assumed their existence in Scaldic poetry here and there, to the consternation of at least one :fellow philologist; c:f. G{slason 1897:230 (gen. sg. f. ~), 231 (gen. sg. f. ~), 269-70 (gen. pl. vegna, ~). - See also th~ remarks on the exceptional gen. sg •. :r. ~ below, section 4.2.3. Even more often, the endings o:f the gen. pl. -~, gen. sg. :r. -.!!!!'., dat. sg. f. -.2:_, have been replaced by the endings -~, -~, -ri, respectively, be:fore which vowel· syncope takes place. Two such cases hav.e been mentioned in passing in the previous paragraph (~, ~). In the same manuscript as ~ ther.e is also gen. pl. kristnra o:f kristinn "Christian" (~ heidinna 'oc midt i bland kristnra), see Helgason 1962b:242. Cf. further.my 249' Linguistica XIII .remarks on the dat. sg. f. ~ below, in sec.tion 4.2.2. More examples·are to be found iil Gutimundsson 1922;89, and here repeated sub (8), q.v. Guamundsson states that he has culled.these forma from texts, although he does not quote chapter and verse. Jonsson (8) gen. pl. falln-ra of fallinn "fallen" heian-ra horfn-ra heieinn "heathen" hor!inn "having vanished" gen. sg. t. ·ge!n-rar o! ge!inn "given" dat. sg. t. born-ri of ~ "born(e)• farn-ri gulln-ri horfn-ri kristn-ri opn-ri ~"gone" gullinn "golden" horfinn "having vanished" kristinn "Christian" opinn "open" 1908:89, who mentions dat. sg. f. fl6kn-ri and gen. sg. f. flokn-rar of flokinn "complicated", states that such forms are often heard in the spoken language. The literary language avoids them (!>orolfsson 1925:88). Examples such as those of (8) had come into.being by the seventeenth centucy; P6r6lfsson ib. quotes dat. sg~ f. d.ruknre, gen. sg. f. · druknrar, and gen. pl.. druknra (o! the past part. drukkinn "drunk"), from the work of the seventeenth centucy grammarian Runolfur Jonsson. I interpret these !orms as arisen from the phonological representation with the non-contracted stem, e.g. gen. sg. f. /drukkin+ +rar/, by the Vowel Syncope Rule. The /rar/ of the representation just adduced has replaced /ar/, just as this happened, say, in the gen. sg. f. einn-ar, of eiim "one, alone", which became einn-rar. These examples, which could undoubtedly be multiplied, have so far been ignored. in the treatment of the Icelandic vowel syncope. They show that.vowel syncope does not only ta:lte place before vowel initial, but alscf before _!: initial, endings; more generally, syncope takes place before any sonorailt initial ending, the vowels and _!: being the only sonorants that occur initially in endings. 4.2. Exceptional cases. There are also some exceptional paradigms 'in whose treatment suppletion has to be resorted to - not surprisingly so, for some of the most used adjectival words are involved: lftill "little", ~ "big, great", ~ "other, second", ~ "my, mine" 250 Janez OREŠNIK .(analogo.imly pinn "thy", .!E:!!E: "his,.her, their"), ~ "that", etc• 4.2.l. Old Icelandic li!ill.~ Its lexical entry contains two stems accompanied by statements on their distribution. See (9), where sample derivations are also presented. (9) (a) Lexical entry_of Old Icelandic litill: /li tl/ 9 b~fore vowel, -: except i;ti;. strong gen. pl. , gen. sg. f~, ,dll,t. sg. f. /lftil/ elsewl;lere ·(b) Some partial derivations: nom. sg. m. gen. sg. m./n. dat. pl. gen. pl. nom. sg. f. lftil ··"lftil+s 'litl+um 15.til+a l:i'.til (la) 1 Htil:+a (le) l:i'.til: VSyncope l:i'.till litlum lftilla lf til The corresponding Modern·rcelandii:: leXical entry and sample derivations !J.re stated in (10), q.v.· (10) (a) LeXical entry of Modem: Icelandic l:l:till:_ /lihtl/ bef6re vowel, except in the strong gen·. pl., . gen. sg. f., dat. sg. f. /litil/ elsewhere (b) Some partial derivations nom. '-sg. m. gen. sg. m./n. litil litll+s ~-infix litI~l VSyncope dat. pl. gen. pl. nom. sg. f. lihtl+Ym litll+a litll litI~l+a ~ lhilla lftil A comparison of the Old and Modem Icelalidic lexical entries of lftill shows that the f orms have only undergone changes dictated by the sound- -laws. The complexity of the paradigm has essentially remained unchanged since Old Icelandic times. ~· Old Icelandic ~ "big, great". Its lexical entry is described in (11), q.v., where some sample derivations are given as well. The k was palatal /k./ even in Old Icelandic, as indicated in (11); see - J 251 Linguistica XIII .Benediktsson 1959 i'n. 23. The Modem Icel!Ul