V. Cvetko Orešnik and J. Orešnik, Natural Syntax: Three-value naturalness scales 235 Varja Cvetko Orešnik and Janez Orešnik Filozofska fakulteta, Ljubljana Natural Syntax: Three-value naturalness scales Slovenski jezik – Slovene Linguistic Studies 6 (2007): 235–249 Naravna skladnja je deduktivna teorija v razvoju in veja teorije naravnosti. Sodbe o naravnosti so ubesedene v lestvicah naravnosti, ki sledijo iz osnovnih meril (ali »aksiomov«), naštetih v začetku sestavka. Napovedi teorije se izračunavajo v t. i. izpeljavah, katerih glavni sestavini sta par lestvic naravnosti in pravila o ujemanju med soodnosnimi vrednostmi naravnosti. Ločimo vzporedno in križno ujemanje, ki sta v dopolnjevalni razvrstitvi. Križno ujemanje je obvezno v izpeljavah, ome- jenih na skrajno nenaravno okolje. Lestvice naravnosti vsebujejo zvečine po dve vrednosti, npr. >nat (+, –) / človeško; nekatere vsebu- jejo po tri vrednosti, npr. (za slovenščino) >nat (ednina, množina, dvojina), ali so še bolj zapletene. Ta sestavek obravnava lestvice s po tremi vrednostmi in prispeva tole: omejevanje okolja neke izpeljave na najbolj desno ali na najbolj desni vrednosti take lestvice naravnosti, ki vsebuje tri vred- nosti, zahteva križno ujemanje. To je ponazorjeno s slovenskim, ruskim in angleškim jezikovnim gradivom. Druge objave o naravni skladnji: Kavčič 2005a–b, Orešnik 1999, 2000a–b, 2001a–f, 2002, 2003a– c, 2002/03, 2004. V tem seznamu so samo dela, napisana v angleščini. Natural Syntax is a developing deductive theory, a branch of Naturalness Theory. The natural- ness judgements are couched in naturalness scales, which follow from the basic parameters (or “axioms”) listed at the beginning of the paper. The predictions of the theory are calculated in the deductions, whose chief components are a pair of naturalness scales and the rules governing the alignment of corresponding naturalness values. Parallel and chiastic alignments are distinguished, in complementary distribution. Chiastic alignment is mandatory in deductions limited to unnatural environments. Naturalness scales are prevailingly two-value; for instance, >nat (+, –) / human. Some are three- value – for instance (for Slovenian), >nat (singular, plural, dual) – or even more complicated. This paper deals with three-value scales and features the following chief finding: the restriction of the environment of any deduction to the rightmost value or to the two rightmost values of a three- value naturalness scale requires chiastic alignment. This is exemplified with Slovenian, Russian and English material. Earlier publications on Natural Syntax: Kavčič 2005a–b, Orešnik 1999, 2000a–b, 2001a–f, 2002, 2003a–c, 2002/03, 2004. This list cites only works written in English. Natural Syntax is a (developing) deductive linguistic theory that determines the pre- suppositions on the basis of which a (morpho)syntactic state of affairs can be made pre- dictable, and thus synchronically explained. The two basic kinds of presuppositions are naturalness scales and rules of alignment among corresponding values of any two scales. 236 Slovenski jezik – Slovene Linguistic Studies 6 (2007) Every (morpho)syntactic state of affairs is represented by two comparable variants. Natural Syntax contains no generative component. We begin by listing the criteria with which Natural Syntax substantiates natural- ness scales: (a) The parameter of favourable for the speaker and of favourable for the hearer. What is favourable for the speaker is more natural, the speaker being the centre of com- munication. Expressed in a scale: >nat (favourable for the speaker, favourable for the hearer). This view of naturalness is commonplace in linguistics (Havers 1931: 171), under the names of tendency to economise (utilised first of all by the speaker) and tendency to be accurate (mainly in the hearer’s interest). (b) The principle of least effort (Havers 1931: 171). What conforms better to this prin- ciple is more natural for the speaker. What is cognitively simple (for the speaker) is easy to produce, easy to retrieve from memory, etc. (c) Prototypicality. What is nearer to the prototype is more natural for the hearer. The speaker favours non-prototypicality. (d) Degree of integration into the construction. What is better integrated into its con- struction is more natural for the speaker. (e) Frequency. What is more frequent tokenwise is more natural for the speaker. What is cognitively simpler for the speaker is used more. (However, the reverse does not obtain: what is natural for the speaker is not necessarily more frequent.) (f) Small v. large class. The use of (a unit pertaining to) a small class is more natural for the speaker than the use of (a unit pertaining to) a large class. During speech small classes are easier for the speaker to choose from than are large classes. (g) Specialised v. non-specialised use. The specialised use of a category is more natu- ral for the speaker than its non-specialised use. Suppose that a language has reflexive personal pronouns. These are specialised for expressing reflexivity (whereas other personal pronouns are not specialised for expressing reflexivity, although they do express it under certain conditions) and their use for expressing reflexivity is very natural for the speaker: >nat (+, –) / reflexive personal pronoun expressing reflexivity. (h) Given a construction, the movement of a unit to the left is more natural for the speaker than the movement of a unit to the right. (Movement to the left is more natural than non-movement; movement to the right is less natural than non- movement.) (i) Acceptable v. non-acceptable use. What is acceptable is more natural for the speak- er than what is not acceptable. The very reason for the acceptability of a syntactic unit is its greater naturalness for the speaker with respect to any corresponding non-acceptable unit. (j) What is more widespread in the languages of the world is more natural for the speaker (the typological criterion). What is cognitively simpler for the speaker is realised in more languages. V. Cvetko Orešnik and J. Orešnik, Natural Syntax: Three-value naturalness scales 237 The basic format of our naturalness scales is >nat (A, B), in which A is favour- able for the speaker and B is favourable for the hearer. A and B are the “values” of the scale. Whenever two basic scales are called for, the other assumes the shape >nat (C, D). Two expanded scales are allowed, viz. >nat (A + B, B) and >nat (A, A + B); they are valid if the corresponding scale of the format >nat (A, B) is valid. Exemplification below. The naturalness scales are supported by the above criteria of naturalness (hence- forth, axioms). Normally it suffices to substantiate any scale with one criterion, which backs up either value A or value B of the scale; the non-supported value is allotted the only remaining position in the scale. Of course, a scale may be supported with more than one criterion. Any clash among the criteria applied to a scale is to be handled with constraints on the combinations of criteria. So far only a few constraints have been formulated; I have not yet encountered much useable crucial language data. The naturalness scales are an essential part of deductions, in which Natural Syn- tax expresses its predictions about the state of affairs in language data. An example of a deduction: English. The numerical indication of frequency normally consists of a cardinal number followed by the word times – e.g., four times – except that there are one-word expressions available for the lowest numbers: once, twice and archaic thrice (Collins Cobuild 1990: 270–71). The two variants: the type once and the type four times. 1. The assumptions of Natural Syntax: 1.1. >nat (type once, type four times) I.e., the type once is more natural than the type four times. – According to the criterion of least effort, item (b) in the list of axioms. 1.2. >nat (low, non-low) / number I.e., any low number is more natural than any non-low number (Mayerthaler 1981: 15). – Low numbers are more easily accessible to the speaker. According to the criterion of favourable for the speaker and of favourable for the hearer, item (a) in the list of axioms. 2. The rules of parallel alignment of corresponding values: 2.1. value A tends to associate with value C, 2.2. value B tends to associate with value D. See Note 4.1 below. 3. The consequences: If a language distinguishes between low and non-low numbers in numerical indi- cations of frequency such that one kind of number uses the pattern four times and the other kind of number uses the pattern once, then it is the low numbers that tend to use the pattern once and it is the non-low numbers that tend to use the pattern four times. Q.E.D. (The reverse situation is not expected.) 4. Notes 4.1. Value A of scale 1.1 (= the type once) tends to combine with value C of scale 1.2 (= low number). Value B of scale 1.1 (= the type four times) tends to combine with value D of scale 1.2 (= non-low number); similarly in the remaining deductions, with 238 Slovenski jezik – Slovene Linguistic Studies 6 (2007) the proviso that the alignment (unlike here) is sometimes chiastic. Chiastic alignment will be explained below. 4.2. Natural Syntax cannot predict the cut-off point between low and non-low numer- als. In every deduction, the rules of alignment play a prominent role; compare item 2 in the above deduction. The alignment rules regulate the combinations of correspond- ing values of the two naturalness scales mentioned in the deduction. The alignment can be parallel or chiastic. Suppose that the two scales are >nat (A, B) and >nat (C, D). Parallel alignment pairs value A with value C, and value B with value D. Chiastic alignment pairs A with D, and B with C. A paramount question is when the alignment is parallel and when chiastic. Par- allel alignment is the default case. Experience based on work with a number of ex- amples has shown that chiastic alignment is necessary whenever a given deduction is limited to language data obtaining within an “extremely unnatural environment”. This is defined as value B of the scale >nat (A, B), provided the scale cannot be ex- tended to the right; i.e., if there is no such value that would be even less natural than value B. The question of “extremely unnatural environment” in three-value natural- ness scales will be addressed below. At the time of this writing, the state of the art cannot explain why there are two kinds of alignment and why they are distributed as they are. This paper deals with naturalness scales that contain three values each instead of the usual only two values; for instance, >nat (singular, plural, dual) (valid for Slove- nian). We shall inquire whether chiastic alignment obtains only within the (rightmost) dual or perhaps even within the (likewise rightmost) non-singular; i.e., the plural and the dual together. The object of Natural Syntax is (morpho)syntax. Given the scarcity of suitable syntactic examples, here we shall extend Natural Syntax to encompass morphology and even morphonology, in which the axioms and mechanisms of Natural Syntax apply equally well as in syntax, to the best of our knowledge. However, we shall continue to use the label Natural Syntax. It can be shown with Slovenian material that chiastic alignment does apply with- in the dual, as in the following deduction: (1) Slovenian. The dual of the noun oko ‘eye’ is očesi. Such a dual does not primarily mean ‘organ of sight’, but refers to corns (on a toe), eyes (on a potato), etc. Only if the dual is expanded with the numeral dva/oba ‘two/both’, does the noun phrase dve/obe očesi also mean ‘organ of sight’; for instance, v vojni je izgubil obe očesi ‘he lost both his eyes during the war’, pri sinovih pustolovščinah zatisniti obe očesi ‘to close both eyes to his son’s adventures’ (examples from SSKJ s.v. oko). This situation is restricted to the dual (SS00 272). The two variants: oko meaning ‘organ of sight’ and having other meanings, within the dual. – The deduction proceeds in the extremely unnatural environment “dual” 1. The assumptions of Natural Syntax: V. Cvetko Orešnik and J. Orešnik, Natural Syntax: Three-value naturalness scales 239 1.1. >nat (+, –) / oko ‘organ of sight’ I.e., the meaning ‘organ of sight’ is more natural than other meanings of oko. – The meaning ‘organ of sight’ is first of all associated with humans and higher animals, whereas the other meanings occur prevailingly in the physical world. According to the criterion of favourable for the speaker and of favourable for the hearer, item (a) in the list of axioms. 1.2. >nat (–, +) / dva/oba ‘two/both’ I.e., the absence of the numeral dva/oba ‘two/both’ is more natural than its pres- ence. – According to the criterion of least effort, item (b) in the list of axioms. A special case of 1.2: 1.2.1. >nat (+/–, +) / dva/oba ‘two/both’ I.e., optional presence of the numeral dva/oba ‘two/both’ is more natural than its obligatory presence. – The scale assumes the permitted expanded format >nat (A + B, B) and is automatically valid because the corresponding basic scale 1.2 has been substantiated. 2. The rules of chiastic alignment: 2.1. value A tends to associate with value D, 2.2. value B tends to associate with value C. 3. The consequences: If a language distinguishes, within the dual, between oko meaning ‘organ of sight’ and oko having “other” meanings such that one meaning is obligatorily accom- panied by dva/oba and the other meaning is not obligatorily accompanied by dva/oba, then it is the meaning ‘organ of sight’ that tends to be obligatorily accompanied by dva/oba and it is the “other” meanings that tend not to be obligatorily accompanied by dva/oba. Q.E.D. (The reverse situation is not expected.) A similar situation in the plural requires parallel alignment: (2) Slovenian. The plural of the noun oko is oči if the meaning is ‘organ of sight’, and očesa if the meaning is different (SS00 272). The two variants: oko meaning ‘organ of sight’ and having other meanings, with- in the plural. 1. The assumptions of Natural Syntax: 1.1. >nat (+, –) / oko ‘organ of sight’ I.e., the meaning ‘organ of sight’ is more natural than other meanings of oko. – The meaning ‘organ of sight’ is first of all associated with humans and higher ani- mals whereas the other meanings occur prevailingly in the physical world. According to the criterion of favourable for the speaker and of favourable for the hearer, item (a) in the list of axioms. 1.2. >nat (oč-i, oč-es-a) / plural of oko I.e., the plural oči is more natural than the plural očesa. – The plural oči contains two morphs, whereas the plural očesa contains three morphs. According to the crite- rion of least effort, item (b) in the list of axioms. 2. The rules of parallel alignment: 2.1. value A tends to associate with value C, 240 Slovenski jezik – Slovene Linguistic Studies 6 (2007) 2.2. value B tends to associate with value D. 3. The consequences: If a language distinguishes, within the plural, between oko meaning ‘organ of sight’ and oko having other meanings such that one meaning assumes the form oči and other meanings assume the form očesa, then it is the meaning ‘organ of sight’ that tends to assume the form oči and it is other meanings that tend to assume the form očesa. Q.E.D. (The reverse situation is not expected.) A similar example follows in deductions (3) and (4): (3) Slovenian. If the dual is used in reference to a pair (of objects), the noun is almost obligatorily preceded by the numeral dva/oba ‘two/both’; for instance, obe roki ‘both hands/arms’. Otherwise the dual does not refer to a pair (Derganc 2003: 172). The two variants: the dual roki ‘hands/arms’ (not forming a pair) and the dual obe roki ‘both hands/arms’ (forming or not forming a pair). – The deduction proceeds in the extremely unnatural environment “dual”. 1. The assumptions of Natural Syntax: 1.1. >nat (roki, obe roki) I.e., roki is more natural than obe roki. – According to the criterion of least effort, item (b) in the list of axioms. 1.2. >nat (+, –) / a pair I.e., being a pair is more natural than not being a pair. – The pair has a special place in nature, not least of all with humans. According to the criterion of favourable for the speaker and of favourable for the hearer, item (a) in the list of axioms. A special case of 1.2: 1.2.1. >nat (+/–, –) / a pair I.e., not being a pair optionally is more natural than not being a pair obligatorily. – The scale assumes the permitted expanded format >nat (A + B, B) and is automati- cally valid because the corresponding basic scale 1.2 has been substantiated. 2. The rules of chiastic alignment: 2.1. value A tends to associate with value D, 2.2. value B tends to associate with value C. 3. The consequences: If a language distinguishes, within the dual, between the type roki and the type obe roki such that one type can mean a pair and the other type cannot mean a pair, then it is the type obe roki that tends to allow the meaning of a pair and it is the type roki that tends not to mean a pair. Q.E.D. (The reverse situation is not expected.) (4) Slovenian. The plural roke can mean a pair unless it is preceded by a numeral (Derganc 2003: 172). The two variants: the plural roke meaning a pair and not meaning a pair. 1. The assumptions of Natural Syntax: 1.1. >nat (+, –) / a pair I.e., being a pair is more natural than not being a pair. – The pair has a special place in nature, not least of all with humans. According to the criterion of favourable for the speaker and of favourable for the hearer, item (a) in the list of axioms. V. Cvetko Orešnik and J. Orešnik, Natural Syntax: Three-value naturalness scales 241 1.2. >nat (–, +) / numeral I.e., the absence of a numeral is more natural than its presence. – According to the criterion of least effort, item (b) in the list of axioms. A special case of 1.2: 1.2.1. >nat (–, +/–) / numeral I.e., obligatory absence of a numeral is more natural than optional absence of a numeral. – The scale assumes the permitted expanded format >nat (A, A + B) and is automatically valid because the corresponding basic scale 1.2 has been substanti- ated. 2. The rules of parallel alignment: 2.1. value A tends to associate with value C, 2.2. value B tends to associate with value D. 3. The consequences: If a language distinguishes, within the plural, between roke meaning a pair and roke not meaning a pair such that an accompanying numeral is allowed in one case and disallowed in the other case, then it is roke meaning a pair that tends to disallow an accompanying numeral and it is roke not meaning a pair that tends to allow an ac- companying numeral. Q.E.D. (The reverse situation is not expected.) We proceed to a few deductions whose language material encompasses both the plural and the dual. It turns out that these numbers together serve as an extremely un- natural environment, which of course demands chiastic alignment. First we consider the plural and dual of the Slovenian noun človek ‘man’. In the non-singular of the noun človek the suppletive stems človek- and ljud- are used. The plural contains only ljud-, whereas in the dual the genitive and the locative display ljud- and the remaining cases are based on človek-. This distribution exhausts the possible distribution of suppletive stems: there is regulated distribution (here in the plural) and non-regulated distribution (here in the dual). The corresponding natu- ralness scale is >nat (non-regulated, regulated) / suppletion. (Regulated suppletion is favourable for the hearer because of its regularity, hence it must be mentioned in posi- tion B of the scale.) Examples from various languages show that non-regulated sup- pletion prevails under more natural conditions than regulated suppletion. Consider the present tense of the Slovenian verb biti ‘be’: the singular displays non-regulated suppletion (sem, si, je), the non-singular displays regulated suppletion (s-va, s-ta, s-mo, s-te, s-o); the singular is more natural than the non-singular. (The personal pronouns of Slovenian – noun phrases of high naturalness – show only non-regulated suppletion in their declension; for instance, jaz ‘I’ – mene ‘me’, etc.) This makes it possible to formulate the following deduction involving suppletion: (5) Slovenian. The plural of the noun človek uses only the suppletive stem ljud-, whereas the dual uses the suppletive stems človek- and ljud- in complementary dis- tribution (SS00 271). The two variants: the suppletive stems človek- and ljud- within the non-singular. – The deduction proceeds in the extremely unnatural environment “non-singular”. 1. The assumptions of Natural Syntax: 1.1. >nat (non-regulated, regulated) / suppletion 242 Slovenski jezik – Slovene Linguistic Studies 6 (2007) I.e., non-regulated suppletion is more natural than regulated suppletion. – Regu- lated suppletion is favourable for the hearer because of its regularity. According to the criterion of favourable for the speaker and of favourable for the hearer, item (a) in the list of axioms. 1.2. >nat (plural, dual) I.e., the plural is more natural than the dual. – Many more languages use the plu- ral than the dual. If a language has the dual it has the plural as well. According to the typological criterion, item (j) in the list of axioms. 2. The rules of chiastic alignment: 2.1. value A tends to associate with value D, 2.2. value B tends to associate with value C. 3. The consequences: If a language distinguishes, within the declension of the noun človek, between the plural and the dual such that one number uses regulated suppletion and the other number uses non-regulated suppletion, then it is the dual that tends to use non-regu- lated suppletion and it is the plural that tends to use regulated suppletion. Q.E.D. (The reverse situation is not expected.) 4. Note. The stem človek- is also used in the singular. This is not included above be- cause a deduction that would involve the stem človek- even in the singular cannot be formulated within Natural Syntax. We add a similar Russian example: (6) Russian. The 1st-person imperative pojdëm means ‘let the two of us go’ (thus the dual), whereas the 1st-person imperative pojdëmte means ‘let’s go’ (thus the plural) (Loporcaro 2006: 145, referring to the more basic study Xrakovskij & Birjulin 2001, which is not accessible to us). The two variants: the 1st-person imperatives pojdëm and pojdëmte. – The deduc- tion proceeds in the extremely unnatural environment “non-singular”. 1. The assumptions of Natural Syntax: 1.1. >nat (pojd-ë-m, pojd-ë-m-te) I.e., the form pojdëm is more natural than the form pojdëmte. – The form pojdëm contains three morphs, and the form pojdëmte contains four morphs. According to the criterion of least effort, item (b) in the list of axioms. 1.2. >nat (plural, dual) I.e., the plural is more natural than the dual. – Many more languages use the plu- ral than the dual. If a language has the dual it has the plural as well. According to the typological criterion, item (j) in the list of axioms. 2. The rules of chiastic alignment: 2.1. value A tends to associate with value D, 2.2. value B tends to associate with value C. 3. The consequences: If a language distinguishes between the hortatives pojdëm and pojdëmte such that one hortative expresses the dual and the other hortative expresses the plural, then it V. Cvetko Orešnik and J. Orešnik, Natural Syntax: Three-value naturalness scales 243 is pojdëm that tends to express the dual and it is pojdëmte that tends to express the plural. Q.E.D. (The reverse situation is not expected.) As can be seen, deduction (6) produces the expected predictions only if chiastic alignment is applied. If we accept the above deductions as correct we must also consent to the follow- ing generalisation: Generalisation A (valid for three-member naturalness scales). If the environment of a deduction consists either of the rightmost value or of the two rightmost values of a scale, the environment is classified as extremely unnatural and the alignment must be chiastic. We will adhere to Generalisation A until some counterexample forces us to change it. The naturalness scale covering the grammatical number is not the only three- value scale. Here we shall consider only the scale regulating the comparison of adjec- tives and adverbs; the scale deserves some attention because of Slovenian data. The comparison scale is >nat (positive, comparative, superlative). According to Generalisation A we expect chiastic alignment in deductions whose environment is restricted to the rightmost value of the scale – i.e., to the superlative – as well as in deductions whose environment is limited to both rightmost values at once; i.e., both to the comparative and to the superlative. In the remaining cases we expect parallel alignment; we shall test a deduction within the comparative and (for special reasons) a few deductions encompassing all three grades. We begin with deduction (7), restricted to the superlative and requiring chiastic alignment: (7) English. The superlative. The adverb very is allowed only before the synthetic superlative; for instance, it was the very best performance (Huddleston & Pullum 2002: 1169). The two variants: the synthetic and the analytic superlative. – The deduction proceeds in the extremely unnatural environment “superlative”. 1. The assumptions of Natural Syntax: 1.1. >nat (synthetic, analytic) / superlative I.e., the synthetic superlative is more natural than the analytic superlative. – The synthetic superlative consists of one word, and the analytic superlative of two words. According to the criterion of least effort, item (b) in the list of axioms. 1.2. >nat (–, +) / very I.e., the absence of very is more natural than its presence. – According to the criterion of least effort, item (b) in the list of axioms. A special case of 1.2: 1.2.1. >nat (–, +/–) / very I.e., obligatory absence of very is more natural than its optional absence. – The scale assumes the permitted expanded format >nat (A, A + B) and is automatically valid because the corresponding basic scale 1.2 has been substantiated. 244 Slovenski jezik – Slovene Linguistic Studies 6 (2007) 2. The rules of chiastic alignment: 2.1. value A tends to associate with value D, 2.2. value B tends to associate with value C. 3. The consequences: If a language distinguishes between the synthetic and the analytic superlative such that one superlative can combine with very and the other not, then it is the synthetic superlative that tends to allow combination with very and it is the analytic superlative that tends to disallow combination with very. Q.E.D. (The reverse situa- tion is not expected.) It can be seen in deduction (7) that the correct result is obtained only if chiastic alignment is applied. The next deduction comprises the comparative and the superlative at the same time: (8) English. In the comparative and superlative of the type strong, a [g] is inserted between the stem and the suffix; for instance, stron[g]er, stron[g]est. There is noth- ing comparable in the type cunning; for instance, the (rare) superlative cunning-est is pronounced without any [g] (Huddleston & Pullum 2002: 1581). The two variants: the type strong and the type cunning in the comparative and the superlative. – The deduction proceeds in the extremely unnatural environment “comparative and superlative”. 1. The assumptions of Natural Syntax: 1.1. >nat (type strong, type cunning) I.e., the type strong is more natural than the type cunning. – The type strong consists of one syllable, the type cunning of two syllables. According to the criterion of least effort, item (b) in the list of axioms. 1.2. >nat (–, +) / inserted [g] I.e., the absence of the inserted [g] is more natural than its presence. – According to the criterion of least effort, item (b) in the list of axioms. 2. The rules of chiastic alignment: 2.1. value A tends to associate with value D, 2.2. value B tends to associate with value C. 3. The consequences: If a language distinguishes between the type strong and the type cunning such that the comparative and the superlative of one type inserts a [g] whereas the com- parative and the superlative of the other type does not insert a [g], then it is the type strong that tends to insert a [g] and it is the type cunning that tends not to insert a [g]. Q.E.D. (The reverse situation is not expected.) It can be seen in deduction (8) that the correct result is obtained only if chiastic alignment is applied. We proceed to the deductions involving all three grades simultaneously. A look at such deductions is necessary because of the somewhat special Slovenian conditions. However, we begin with an English example that is simpler than the corresponding Slovenian data. As mentioned above, the expected alignment is parallel. V. Cvetko Orešnik and J. Orešnik, Natural Syntax: Three-value naturalness scales 245 (9) English. In the comparison of the type strong, a [g] is inserted in the comparative and the superlative; for instance, in the comparative stron[g]er (Huddleston & Pullum 2002: 1581). The two variants: the positive strong (lacking a [g]) as against the non-positive stronger, strongest (both containing a [g]). 1. The assumptions of Natural Syntax: 1.1. >nat (positive, non-positive) I.e., the positive is more natural than the non-positive. – The positive is zero coded in many languages. According to the criterion of least effort, item (b) in the list of axioms. 1.2. >nat (–, +) / inserted [g] I.e., the absence of the inserted [g] is more natural than its presence. – According to the criterion of least effort, item (b) in the list of axioms. 2. The rules of parallel alignment: 2.1. value A tends to associate with value C, 2.2. value B tends to associate with value D. 3. The consequences: If a language distinguishes, within the type strong, between the positive and the non-positive such that one grade contains [g] and the other grade lacks it, then it is the positive that tends to lack [g] and it is the non-positive that tends to contain [g]. Q.E.D. (The reverse situation is not expected.) The reader will appreciate the tiny difference between deduction (8) involving the comparative and the superlative (and implementing chiastic alignment) and de- duction (9) involving all three grades (and applying parallel alignment). Against this background let us inspect the Slovenian comparison lep, lepši, najlepši ‘beautiful, more beautiful, most beautiful’. Even here one would expect first of all a division into the positive and the non-positive. However, there is a compli- cation because the difference between the positive and the comparative (i.e., in the suffix) is not the same as the difference between the comparative and the superlative (i.e., in the prefix). Natural Syntax is unable to encompass such a situation in a single scale. Rather, it needs two scales: one covering the positive and the comparative, the other covering the comparative and the superlative. However, even in the latter scale the alignment is parallel, against the spirit of deduction (8) above. The situation is an artefact of the theory, which is not yet able to cope with three-value scales in a straightforward manner. If the formalism of Natural Syntax were suitably changed, it is to be hoped that Slovenian comparison would be encompassed in one scale, and consequently there would be no difficulty with alignment. (10a) Slovenian. Unlike the positive, the comparative contains a suffix; for instance, lep – lep-ši (SS00 325). The two variants: the positive lep and the comparative lepši. 1. The assumptions of Natural Syntax: 1.1. >nat (positive, comparative) 246 Slovenski jezik – Slovene Linguistic Studies 6 (2007) I.e., the positive is more natural than the comparative. – In many languages the positive is zero-coded. According to the criterion of least effort, item (b) in the list of axioms. 1.2. >nat (–, +) / suffix I.e., the absence of a suffix is more natural than its presence. – According to the criterion of least effort, item (b) in the list of axioms. 2. The rules of parallel alignment: 2.1. value A tends to associate with value C, 2.2. value B tends to associate with value D. 3. The consequences: If a language distinguishes between the positive and the comparative such that one grade contains a suffix and the other grade lacks it, then it is the positive that tends to lack the suffix and it is the comparative that tends to contain the suffix. Q.E.D. (The reverse situation is not expected.) (10b) Slovenian. Unlike the comparative, the superlative contains a prefix; for in- stance, lepši – naj-lepši (SS00 325). The two variants: the comparative lepši and the superlative najlepši. 1. The assumptions of Natural Syntax: 1.1. >nat (comparative, superlative) I.e., the comparative is more natural than the superlative. – The superlative is often coded as comparative + something. According to the criterion of least effort, item (b) in the list of axioms. 1.2. >nat (–, +) / prefix I.e., the absence of a prefix is more natural than its presence. – According to the criterion of least effort, item (b) in the list of axioms. 2. The rules of parallel alignment: 2.1. value A tends to associate with value C, 2.2. value B tends to associate with value D. 3. The consequences: If a language distinguishes between the comparative and the superlative such that one grade contains a prefix and the other grade lacks it, then it is the comparative that tends to lack the prefix and it is the superlative that tends to contain the prefix. Q.E.D. (The reverse situation is not expected.) The clash between deduction (8) and deduction (10b) constitutes a counterexam- ple to Natural Syntax. The counterexample can be neutralised by changing the theory. We suggest the introduction of the following constraint: Constraint (provisional). Whenever the language material of a deduction is lim- ited so as to require chiastic alignment, this is replaced by parallel alignment just in case the narrowness of the language material is due to the narrowness of the formal- ism of Natural Syntax. We shall adhere to this constraint until some counterexample forces us to change it. V. Cvetko Orešnik and J. Orešnik, Natural Syntax: Three-value naturalness scales 247 References Collins Cobuild English Grammar, 1990: London: HarperCollins. Derganc, Aleksandra, 2003: The Dual in Slovenian. Sprachtypologie und Universa- lienforschung 56: 165–81. Havers, Wilhelm, 1931: Handbuch der erklärenden Syntax. Heidelberg: Winter. Huddleston, Rodney, & Geoffrey K. Pullum, 2002: The Cambridge Grammar of the English Language. Cambridge, UK: Cambridge University Press. Kavčič, Jerneja, 2005a: Pleonastic Use of verbs of Speaking in Greek: An Interpreta- tion in Terms of Naturalness theory. Linguistica 45, 203–12. Kavčič, Jerneja, 2005b: The Syntax of the Infinitive and the Participle in Early Byz- antine Greek. Ljubljana: Filozofska fakulteta. Loporcaro, Michele, 2006: Crossing Form and Function: First and Second Person Plural Imperatives in the Dialect of Mesocco. Folia Linguistica 40, 135–54. Mayerthaler, Willi, 1981: Morphologische Natürlichkeit. Wiesbaden: Athenaion. Orešnik, Janez, 1999: Naturalness: The English s-Genitive and of-phrase. Studia An- glica Posnaniensia 34, 191–200. Orešnik, Janez, 2000a: Naturalness: The Preterit and Present Perfect Tenses in Ger- man. In: Kovačič, Irena, Milena Milojević-Sheppard, Silvana Orel-Kos and Jan- ez Orešnik (eds.), Linguistics and Language Studies: Exploring Language from Different Perspectives. Ljubljana: Filozofska fakulteta. Pp. 21–43. Orešnik, Janez, 2000b: Naturalness: Negators in Standard French. Papiere zur Lin- guistik 62/63, 39–48. Orešnik, Janez, 2001a: Naturalness: The Passive in the Mainland Scandinavian Lan- guages. Sprachtypologie und Universalienforschung 54, 26–34. Orešnik, Janez, 2001b: Naturalness: Some Slovenian (Morpho)syntactic Examples. Slovenski jezik – Slovene Linguistic Studies 3, 3–31. Orešnik, Janez, 2001c: Naturalness Expressed in sEm-Values. Linguistica 41, 5–19. Orešnik, Janez, 2001d: Naturalness Expressed in sYm-Values. Folia Linguistica 35, 321–35. Orešnik, Janez, 2001e: Naturalness Applied to Slovenian (Morpho)syntax. Wiener slavistisches Jahrbuch 47, 135–58. Orešnik, Janez, 2001f: A Predictable Aspect of (Morpho)syntactic Variants. Pred- vidljiv vidik (obliko)skladenjskih dvojnic. Ljubljana: Slovenska akademija zna- nosti in umetnosti. Orešnik, Janez, 2002: Naturalness in English: Some (Morpho)syntactic Examples. Linguistica 42, 143–60. Orešnik, Janez, 2003a: Naturalness: Some Norwegian (Morpho)syntactic Examples. Norsk lingvistisk tidsskrift 21, 57–70. Orešnik, Janez, 2003b: Naturalness in English: (A) The Genitive, (B) The Pronouns. Linguistica 43, 119–40. Orešnik, Janez, 2003c: Naturalness: Some English (Morpho)syntactic Examples. Stu- dia Anglica Posnaniensia 39, 77–101. Orešnik, Janez, 2002/03: Naturalness in English: The Infinitive Clause. Klagenfurter Beiträge zur Sprachwissenschaft 28/29, 147–60. 248 Slovenski jezik – Slovene Linguistic Studies 6 (2007) Orešnik, Janez, 2004: Naturalness in (Morpho)syntax: English Examples. Jezikovna naravnost v (obliko)skladnji – angleški zgledi. Ljubljana: Slovenska akademija znanosti in umetnosti. SSKJ = Slovar slovenskega knjižnega jezika, 1970-91. 5 volumes. Ljubljana: Državna založba Slovenije. [Dictionary of Standard Slovenian] Xrakovskij, Victor S. & Leonid A. Birjulin, 2001: Imperative Sentences: Theoretical Problems. In: Xrakovskij, Victor S. (ed.), Typology of imperative constructions. Munich: Lincom Europa. Pp. 3–50. Prispelo oktobra 2006, sprejeto decembra 2006 Received October 2006, accepted December 2006 Naravna skladnja – lestvice s po tremi vrednostmi naravnosti V naravni skladnji vsebujejo nekatere lestvice naravnosti po tri vrednosti narav- nosti, npr. (za slovenščino) >nat (ednina, množina, dvojina). Sestavek se ukvarja z naslednjim vprašanjem: če je neka izpeljava gradivsko omejena na eno ali več vre- dnosti naravnosti take lestvice – kakšno je tedaj ujemanje med lestvicama izpeljave, vzporedno ali križno? Odgovor se glasi takole. Sama dvojina ali dvojina + množina kot okolje zahte- vajo križno ujemanje. Vsa druga okolja, npr. samo množina, zahtevajo vzporedno ujemanje. Te možnosti so v sestavku ponazorjene z izpeljavami, katerih jezikovno gradivo je naslednje (številke se nanašajo na oštevilčenje izpeljav): 1. Dvojina slovenskega samostalnika oko, tj. očesi. 2. Množina slovenskega samostalnika oko, tj. oči in očesa. 3. Slovenska dvojina obe roki. 4. Slovenska množina roke. 5. Dvojina in množina slovenskega samostalnika človek. 6. Ruski hortativ pojdëm ‘pojdiva’ in pojdëmte ‘pojdimo’. 7. Angleški presežnik, raba prislova very ‘brezpogojno’ pred njim. 8. Angleški primernik in presežnik, stopnjevanje. 9. Angleško stopnjevanje, vse tri stopnje. 10a. Slovensko stopnjevanje lep – lepši. 10b. Slovensko stopnjevanje lepši – najlepši. Glavni namen naravne skladnje je določati pogoje, pod katerimi so razmere v nekem jezikovnem gradivu napovedljive. Natural Syntax: Three-Value Naturalness Scales In Natural Syntax certain naturalness scales contain three naturalness values each; for instance, (for Slovenian) >nat (singular, dual, plural). The paper discusses the following question. If the language material of a deduction is restricted to one or several naturalness values of a three-value scale, what kind of alignment is used: parallel or chiastic? V. Cvetko Orešnik and J. Orešnik, Natural Syntax: Three-value naturalness scales 249 The answer is that, as an environment, the dual or the dual + the plural require chiastic alignment. All other environments – for instance, the plural – require parallel alignment. These possibilities are exemplified in the paper with deductions whose language material is as follows (the figures refer to the numbering of the deduc- tions): 1) The dual of the Slovenian noun oko ‘eye’; i.e., očesi. 2) The plural of the Slovenian noun oko; i.e., oči and očesa. 3) The Slovenian dual obe roki ‘both hands/arms’. 4) The Slovenian plural roke, literally ‘hands/arms’. 5) The dual and the plural of the Slovenian noun človek ‘man’. 6) The Russian hortative pojdëm ‘let the two of us go’ and pojdëmte ‘let’s go’. 7) The English superlative, the use of very before it. 8) The English comparative and superlative, their comparison. 9) The English comparison, all three grades. 10a) The Slovenian comparison lep – lepši ‘beautiful, more beautiful’ 10b) The Slovenian comparison lepši 7– najlepši ‘more beautiful, most beautiful’ The chief aim of Natural Syntax is to determine the conditions under which the situation in some language material can be predicted. 250 Slovenski jezik – Slovene Linguistic Studies 6 (2007)