THE RELATIONSHIP BETWEEN ( IN ) ALIENABLE POSSESSION AND THE ( THREE POTENTIAL ) FORMS OF POSSESSED NOUNS IN HUNGARIAN

The paper gives a thorough insight into the system of possible forms of (in)alienably possessed nouns in Hungarian. Its point of departure is the group of [Nominative + -j- +A] possessive forms the stem of which has an alternative (morphologically “shorter”) possessive form; such longer possessive forms are claimed to express alienable possession (see den Dikken 2015). We point out that Hungarian deverbal nominals― and especially the groups of T-nouns―play an interesting role in this system via the thematic character of their possessors (given the obvious connection between alienable possession and external argumenthood, on the one hand, and inalienable possession and internal argumenthood, on the other).

in contrast to the intrinsic relationship between verbs and their Themes (Kratzer 1996), the association of the longer possessed form with the less intrinsic semantic relationship is in harmony with den Dikken's (2015: 138) thesis based on a linguistic universal proposed by Haspelmath (2008), according to which longer possessive forms express alienable possession.
(1) Forms of possessees in some deverbal nominal constructions in Hungarian

Four Groups of Nouns in Respect of Possessed Forms
The 3 rd person singular possessedness suffix -(j)A has the following four allomorphs (in the case of singular possessees): -ja, -je, -a, and -e, distributed partly on the basis of vowel harmony, and partly on the basis of the following mysterious phenomenon, which attracts much attention in the literature.There are nouns which can appear more or less readily both with -jA and -A essentially depending on the alienable or inalienable semantic character of the possessive structure (e.g., Kiefer 1985Kiefer , 2000: 201): 201), as is illustrated by the minimal pair in (2a-a') below.As the stem of the noun that bears the possessedness suffix may also appear in two different forms, illustrated by the oftenquoted minimal pair in (5b-b'), nouns can be divided into four groups with respect to their potential alternative (3Sg) possessed forms.The relevant data is presented below in the series of examples in (2)(3)(4)(5).
What is at stake is the verification in Hungarian of a straightforward generalization by Haspelmath (2008) according to which languages tend to express alienable possession by means of morphologically richer forms than inalienable possession.This can be done either by verifying that the component -j-itself has a morphemic status responsible for the expression of alienability inside the possessedness suffix -(j)A (den Dikken 2015: 131) or by interpreting morphological richness in some less trivial way (in (5b-b'), the expression of alienability is claimed to be transferred from -j-to -á-by den Dikken (2015: 141-142)).
Let us start the overview by considering the distribution of grammaticality judgments in the group of nouns with a single stem but with a phonotactically permitted alternation between the forms -jA and -A of the possessedness suffix -(j)A (2).As in all example pairs in (2-5), the possessive structures in the primed examples are evaluated as expressions of unquestionably alienable relationships while those in the primeless ones as those of inalienable relationships, or at least of types of relationship which can be regarded as encoded in Hungarian as inalienable on the basis of analogous examples.
Part-whole relations form the trivial basis of alienable possessive structures (2a,b,d,e) with body parts as a distinguished subset (2b,e). 3Of these examples, in the inalienable constructions (2a,b,d), the -jA variants are fully unacceptable, while in the corresponding alienable possessive structures (2a',b',d'), the -jA variants are more or less marked but not unacceptable.In the latter case, the -A variants are also more or less marked (but still acceptable).As a similar distribution of grammaticality judgments can be observed in the minimal pair (2c-c'), rulers of nations can be considered to be encoded in language as inalienable parts of their nations.
(2) Inalienable/alienable forms of possessed nouns: I. Basic data, in which the difference between the variants can be regarded as -j-insertion a) a ház ablak-(*j)a a') a világ legjobb ablak-?? ( ?j)a4 the house window-Poss.3Sgthe world best window-Poss.3Sg'the window of the house' 'the world's best window' b) Ili talp-(*j)a b') a világ legbüdösebb talp-?( ?? j)a Ili sole-Poss.3Sgthe world most_smelly sole-Poss.3Sg'Ili's sole' 'the world's most smelly sole' possessed noun] in order to guarantee that the alienable interpretation is achieved in a highly uniform manner.We are aware of the fact that there are also other constructions guaranteeing alienability as the [possessor + body part] construction in medical contexts and the [classifier as a possessed noun] construction (e.g., ?pohárja egye euró a sörnek 'glass.Poss.3Sgone euro the beer.Dat' ('a glass of beer costs one euro').According to our first observations, there are slight (but fairly speaker-dependent) differences in grammaticality judgments between the different types of alienable construction.This suggests that (in)alienability is not a dichotomy but a scalar category.It goes beyond the scope of this paper to investigate this global aspect of the problem of forms of possessed nouns as well as to extend the investigation to forms of possessed nouns in plural and in non-third person.As for this latter problem, a noun like ablak 'window' (2a), for instance, has no alternative possessed forms in first person singular (ablakom 'my window' is the only form), in contrast to such nouns as gyapjú 'wool' (5c), which has a separate alienable form gyapjúm 'my wool' besides the inalienable form gyapjam.The last two examples, in which a body part (2e-e') and a sort of ruler (2f-f') are referred to, do not satisfy the above-sketched distribution of grammaticality judgments, since both kinds of interpretation can be expressed exclusively by the -jA variants.The homophonous forms (obviously belonging to the two different lexical items 'faculty' and 'arm') presented in (2d) and in (2e), thus, show different patterns of grammaticality judgments, in spite of the fact that both express part-whole relations.
In the series of examples in (3), a few phonotactic rules of Hungarian are illustrated which exclude the simultaneous occurrence of the -jA and -A variants in possessed nouns.
A noun ending in a vowel (3a-b'), for instance, has no -A variant while a noun ending in -s (pronounced as the consonant in the English word ash) (3c) has no -jA variant (see Rebrus 2014: 387-390; this exclusion also holds for all other sibilants as a very strong but somewhat speaker-dependent tendency).The consonant combination shown in (3d), however, disprefers -A.As for deciding the precise set of such consonants and consonant combinations, this morphophonological task (together with accompanying methodological questions) requires future research.
(3) Inalienable/alienable forms of possessed nouns: II.One (potential) form for phono- 'the world's biggest belly' the world most_beautiful thigh-Poss.3Sg'Ili's thigh' 'the world's most beautiful thigh' e) a csavar any*( (?) á)ja e') a szerelő legnagyobb any*( (?) á)ja the screw mother.Poss.3Sg the mechanic biggest mother.Poss.3Sg'the nut of the bolt' 'the mechanic's biggest nut' The minimal pair presented in (3e-e') above is of special interest to us, given that the noun anya 'nut (of a bolt)', which belongs to the vowel-ending subgroup shown in (3a-b) is a polysemic counterpart of anya 'mother' presented in (5b-b') below, just mentioned above as an example of nouns having two stems.This pair thus patterns with the pair of homophonous nouns presented in (2d-e') above in behaving differently in respect of accepting -jA/-A variants.
The third group consists of nouns with an alternative (idiosyncratic) possessed form from that which can be derived on-line from the nominative form via adding -jA or -A (4).Such nouns, thus, potentially have three 3Sg possessed forms.However, both the inalienable and alienable meanings are very much preferably expressed by the idiosyncratic variant and the -A variant is fully unacceptable.In the case of the noun gyomor 'stomach', for instance, the inalienable meaning can be expressed only by means of the idiosyncratic variant gyomra, providing a fully acceptable possessive structure (4a).The alienable meaning can be expressed by means of either the idiosyncratic form or the -jA variant, though both resulting structures are highly marked (4a').The minimal pair in (4e-e') illustrates such an extreme preference for the idiosyncratic variant that this can readily express both kinds of meaning, with the two other potential forms providing fully unacceptable possessive structures.
(4) Inalienable/alienable forms of possessed nouns: III.An idiosyncratic form coexists with an "on-line created" form derived by means of -jA from the nominative version (while a form derived by means of -A is phonotactically possible but not acceptable) In the fourth group, the nouns have an idiosyncratic possessed form but the ending of the nominative form excludes either the -A variant (5a-c',e-e') or the -jA variant (5d-d') for the same phonotactic reasons as was discussed in connection with the examples presented in (3) above.It can be observed that the inalienable meanings can be associated only with the idiosyncratic variants (see the primeless examples in (5a-d)).The alienable meanings, on the other hand, are only associated with the variants based on the nominative form (see the corresponding primed examples).The minimal pair presented in (5e-e') with the noun falu 'village' is somewhat exceptional with respect to the inalienable meaning.Presumably this is due to the quite archaic character of the idiosyncratic variant falva: in present-day Hungarian, the nominative-form-based variant faluja is almost as acceptable as the idiosyncratic variant (NB: it is even questionable whether the possessive structure presented in (5e) is encoded as an inalienable relationship in language).As for the alienable meaning tested in (5e'), it is unequivocally the nominative-form-based variant that expresses the alienable meaning (even more preferably than in the case of the acceptability pattern typical of the corresponding variants in ( 5a',b',c',d')).
(5) Inalienable/alienable forms of possessed nouns: IV.An idiosyncratic form coexists with an on-line created form whilst phonotactics prohibit

Generalizations
In what follows, the observations about the four groups of nouns are summarized (Table 1) and generalized using the terminology defined in (6) below, which makes it possible to formulate the generalizations in a simple and elegant form ( 7).It must be noted that certain speakers refuse variant 1 forms in most cases, saying that they sound very artificial (e.g., ablakja 'its window'; cf. the generally accepted v2 variant ablaka).This phenomenon may be regarded as a kind of hypercorrection: the speakers in question are convinced that the given variants violate certain rules they learned, in spite of the fact that they have never been taught such rules.Certain variant 3 forms are also problematic for some speakers because they consider them unacceptably archaic (e.g., disznaja 'his/her pig'; cf. the "modern" v1 variant disznója).It also generally holds for all examples in (2-5) except for those marked as fully acceptable or fully unacceptable that they show quite high speaker-dependent variation.
In Table 1, the four quarters correspond to the four series of examples in (2-5).The (simplest) bottom right quarter presents the grammaticality judgments given in (3).In this group, v1 has no potential alternative, since there is no idiosyncratic variant (v3) and phonotactic rules exclude another nominative-form-based variant (v2).Furthermore, what is worth noting is that, in all cases in this group, v1 can always readily express the inalienable meaning as well as the alienable one.
The top right quarter presents the grammaticality judgments given in (5).In the corresponding group of nouns, only v3 and v1 are "in competition", indicated in the corresponding heading as '{3, 1}', since v2 is excluded for phonotactic reasons.As was observed in (5), in this group, v3 can readily express the inalienable meaning (indicated by the formula '{3, 1} → 3' in the table) and v1 can be almost as readily associated with the alienable meaning ('{3, 1} → 1'), but not vice versa.Thus the available potential variants differentiate the two kinds of meanings in a plausible way.The primary inalienable meaning belongs to the idiosyncratic variant v3 whilst the alienable meaning, calculated in the given context on the basis of some kind of mental conceptual network (Alberti and Farkas to appear, 2.1.1.2.2), 6 is expressed by the variant which can be calculated automatically, that is, by v1, referred to as the primary on-line created variant in (6a).Note that the simplified formula in (7a) below ([v1~a]) refers to this latter relationship between the alienable meaning and the on-line created form, which can be regarded as a generalization over den Dikken's (2015: 131, 141-142) -j-insertion in certain cases (2a,c,d,), and -á-insertion (5b) in other cases, plus some further morphological differences for which den Dikken (2015) does not account (5a,c,d,e). 7 6 In such a mental network, it must be calculated that, say, the possessive structure my house can refer not only to default relationships such as my owning the given house, and/or my living there, but also arbitrarily expanded relationships such as my being the homeless person who inspects the garbage cans of the house or my being the agent whose task is to make the residents fill in some questionnaire.
7 It is clear that this group can be characterized not by a difference between the competing possessee variants manifesting itself in a certain sound-size morpheme (or sequence of sounds) but by the phonetically highly varied and unpredictable difference between an automatically producible potential variant v1 and an idiosyncratic variant v3 existing for historical reasons.The variant v1 tető-je 'its roof', for instance, is different form the v3 variant tete-je in the quality and the length of the stem-final vowel (see also the minimal pairs ajtó-ja/ajta-ja 'its door', disznó-ja/diszna-ja 'its pig', tüdő-je/tüde-je 'its lung').Relative to the v1 variant gyapjú-ja 'its wool', however, the v3 variant gyapj-a does not contain a stem-final vowel of another quality but it lacks the stemfinal vowel and (hence) this form gets the -j-less version of the suffix -(j)A (somewhat similar examples with other stem-final vowels: anyá-ja/any-ja 'its mother', apá-ja/ap-ja 'its father').
The minimal pair parázs-a/parazs-a 'its ember' exemplifies the case when there is a difference (chiefly) in the length of a stem-internal vowel (also see darázs-a/darazs-a 'its wasp'), while the pair falu-ja/falv-a 'its village' illustrates the type of difference based on the phenomenon often referred to as v-insertion in synchronic descriptions (also see tetű-je/tetv-e 'its louse').falva The top left quarter of Table 1 presents the grammaticality judgments given in (4).In the corresponding group of nouns, it could be theoretically possible that all the three variants be in competition, but, as can be observed in (4), v2 cannot express either the inalienable meaning or the alienable one.The systematic unacceptability of v2 is indicated in the corresponding heading by crossing out this variant (see the notation '{3, 2, 1}' in the top left quarter of the table).The table does not present the uniformly fully unacceptable data.A generalization can be formulated which holds for all types of data that v2 and v3 mutually exclude each other; see (7e) below (* [v3 & v2]).Hence there is no noun with three more or less acceptable possessed forms (7e') (* [v1 & v2 & v3]).Another straightforward consequence of the mutual exclusion between v2 and v3 is that if a noun has two possessed forms, one of them is v1 (7e").
Variant v2 excluded, hence, both in the top right quarter and in the top left quarter: the same two variants "remain in competition".However, the outcome in the two cases is different: while in the top right quarter, the two variants differentiate the two kinds of meanings, as is formulated in (7a-a') ([v1~a, v2~i, v3~i]), in the top left quarter, both the inalienable meaning and the alienable meaning are significantly more readily expressed by v3 (cf.(7c"): [v3~a → *v1, *v2]).What is formulated in (7a-a') is a (plausible) strategy that functions only in certain domains of nouns (see the two specially framed domains with dark rims in Table 1) and not a universal generalization valid for all Hungarian nouns.
The bottom left quarter of the table, in which (in the absence of idiosyncratic (v3) alternatives) the two nominative-form-based variants v1 and v2 are in competition, shows the most eclectic picture.This picture is a reflection of the great variety according to which certain data pattern with those in the top right quarter in associating different forms with the two kinds of meanings (7a-a') whilst other data pattern with those in the top left quarter in associating the same forms with the two kinds of meanings, and moreover, both v1 and v2 can serve as this dominant form (in the case of different nouns, of course; cf.(7c-c'): [v1~i → *v2, *v3], [v2~a → *v1, *v3]).Thus in this quarter, both v1 and v2 are associated with either the inalienable or the alienable meaning (in the case of different nouns), as is registered in (7b) below ([v1~i / v2~a / v3~a]); nevertheless, it never occurs that, in the case of one and the same noun, the alienable meaning is expressed by v2 while the inalienable one by v1.This restriction, which holds for all data in all the four quarters, is formulated in (7d) as follows: if different variants are associated with the two kinds of meaning (see the two specially framed domains with dark rims in Table 1), the "sequence number" (1, 2, and 3 given in (6) above) of the variant belonging to the inalienable meaning must be greater than that of the variant belonging to the alienable meaning; the opposite association is excluded.Note that this generalization can completely cover all the data with no exception due to its formulation in which cases of equation are also accepted (included in the relation 'k≥n' in (7d)): such cases cover the nouns with a single acceptable possessed form (see the other four domains framed with light rims in Table 1).
(7) Generalizations on the (somewhat hidden) relationship between v1 versus v2,v3 and alienability (a) versus inalienability (i) a) v1~a on-line created: morphophonologically ~ semantically a') v2~i, v3~i: (a-a'): partial tendency as a good point of departure b) v1~i / v2~a / v3~a there are such counterexamples inalienable/alienable for each domain e) * [v3 & v2] variants excluding each other e') * [v1 & v2 & v3] all the three variants cannot appear simultaneously e") [v k & v n (k>n) → n=1] of two variants, one is v1 All in all, although the data in (2-5) suggest a close relationship between the automatically calculable (on-line created) variant v1 and a context-dependent alienable meaning (7a), and, parallel to this, between the other two variants and inalienable meaning (7a'), either kind of meaning can be expressed by any variant (7b).
The hypothesized asymmetry of semantic affiliation between v1 versus v2 and v3, beyond the fact that the natural strategy formulated in (7a-a') explicitly prevails in certain domains of nouns, also prevails in the other domains "vacuously" and "implicitly" in the following sense.By 'vacuously' we mean that there is no domain in which v1 expresses the inalienable meaning with v2 or v3 expressing the alienable meaning (7d).As for 'implicit' manifestations of the asymmetry in question, the (c)-constraints formulate them by claiming that if a variant can express the opposite kind of meaning relative to its basic character given in (7a-a'), then, in the case of the same noun, it will express (at least as readily) the other kind of meaning (the one that ab ovo suits it), too.In such cases, the given variant is referred to as a dominant one.
The constraints in (7e-e") formulate restrictions on the coexistence or, on the contrary, dominance of the three variants which do not follow from the foregoing.
Let us consider a few detailed instances or consequences of the (e)-restrictions (already discussed above).If, for instance, the idiosyncratic variant v3 expresses the inalienable meaning and phonotactics does not exclude the construction of a variant v2, the latter will be fully unacceptable as an expression of either the same inalienable meaning or the alienable meaning.The latter part of this claim is in harmony with the spirit of (7a-a') while the former part can be regarded as a unicity condition: it is needless to express the same kind of meaning in two or more ways.Therefore, practically if a noun has two (more or less acceptable) possessed forms, then the alienable meaning is expressed by v1 (7e") and the inalienable meaning either by v2 or by v3, exclusively; it follows that there is no noun with three different more or less acceptable possessed forms (7e').

RELATIONSHIP BETWEEN FORMS OF POSSESSED DEVERBAL
NOMINALS AND THE THEMATIC CHARACTER OF THEIR POSSESSORS Possessive structures of (complex-eventuality-related) derived nouns with thematic possessors fit well in the system functioning according to the constraints presented in (7) above -through placing the given types of derived noun in the appropriate quarters of Table 1; see Table 2 below.As is illustrated in (8) below (which demonstrates the entire system, only a part of which was shown in (1) in the Introduction), in the case of complex-eventuality-related derived nouns, the possessor always corresponds to a designated input argument, that is, it is always a thematic argument (see Alberti and Farkas to appear).Thematic arguments appearing as possessors are of distinguished relevance because the Agent is held to stand in a non-intrinsic relationship with the verb (Kratzer 1996), which can plausibly be considered to be related to alienability, in contrast to the intrinsic (hence, inalienable) relationship between verbs and their Themes.nominals with possessors carrying different thematic roles (Section 3).In this global picture, den Dikken's (2015) hypothesis appears as a (plausible) strategy that functions in a single domain of nouns (with competing variants v1 with -jA and -j-less v2 variants; see the lower specially framed domain with dark rims in Table 1), while in other domains, the ab ovo association of the -jA variants with alienability manifests itself in more hidden forms that (i) the alienable -jA variant (or rather, what is defined as variant v1 in (6a)) is opposed to a (phonetically varied) group of alternative idiosyncratic (v3) possessee variants (see footnote 8), or (ii) there is a dominant possessee variant (7c-c"), which simply suppresses the other potential variants (blocking or covering the differentiation according to (in)alienability).

( 6 )
Definition of three kinds of possessed variants a) Possessed variant 1 (v1): on-line created as [nominative form of the noun + -jA], unless the relevant phonotactic rules of Hungarian prohibit this; otherwise, [nominative form of the noun + -A]. 5 b) Possessed variant 2 (v2): on-line created as [nominative form of the noun + -A] if the relevant phonotactic rules of Hungarian permit both this variant and the [nominative form of the noun + -jA] variant (NB: v2 is defined in a way that it is inevitably different from v1). c) Possessed variant 3 (v3): acceptable (idiosyncratic) historical form of the noun, different from those referred to in v1 and v2, if extant.

Table 1 :
Acceptability of different variants of possessed forms depending on phonotactic and historical factors