A HunGram account of four Hungarian PVCs

In Laczkó & Rákosi (2011)¹⁰⁶ we explore the tenability and implementational applicability of the approach proposed by Forst et al. (2010). In this vein, we give a detailed analysis of both the compositional and the noncompositional uses of two Hungarian noninflecting spatial PVC types and report its successful implementation. Consider the following examples of the first type.

(73) A rák ki mász-ott a folyó-ból.

the crab.NOM out crawl-PAST.3SG the river-out.of

‘The crab crawled out of the river.’

(74) Az elnök ki fej-ez-te együttérzés-é-t.

the president. NOM out head-Vsuf-PAST.3SG sympathy-his-ACC

‘The president expressed his sympathy.’

The sentence in (73) is an example of the compositional use of the preverb ki ‘out’, while (74) illustrates an utterly noncompositional use (because the simplex verb form fejezte does not exist on its own). We assume that preverbs are nonprojecting words in the sense of Toivonen (2001), and their syntactic category is PRT (short for particle).¹⁰⁷ For the analysis of (73) we need the following lexical forms for the preverb and the verb (only the relevant details are indicated in these XLE style implementational representations).

(75) a. mászik V (↑ PRED)= ‘crawl < (↑ SUBJ) (↑ OBL) >’.

b. ki PRT (↑ PRED) = ‘out < %ARG1 (↑ OBL) >’.

The verb mászik ‘crawl’ has its regular lexical entry. It is a two-place predicate with a subject and a (goal) oblique argument. The preverb ki ‘out’ in its compositional use is also a two-place predicate: it takes a verb as its first argument and a (source) oblique second argument.

In c-structure, the preverb, analyzed as the main predicate, has the customary functional head annotation, while the verb has a set of annotations containing the restriction operator encoded by the / symbol.¹⁰⁸ The interplay of these annotations results in syntactic complex predicate formation, represented in f-structure. The PRED feature in the f-structure of (73) has the following value:

(76) ‘ki < ‘mászik < [rák], NULL >, [folyó] >.

The preverb (ki ‘out’) is the main predicate, and it has a “nested” argument structure. Its first argument is the verb (mászik ‘crawl’) with its own embedded two-place argument structure.¹⁰⁹ The verb’s first argument is the subject (rák ‘crab’), and its second (oblique) argument¹¹⁰

consideration against this black-and-white view is that there is a huge grey and gradient area between the black-and-white extremes of the compositionality/productivity and noncompositionality/nonproductivity scale.

106 This discussion of Laczkó & Rákosi (2011) is an augmented version of Section 2.2 in Laczkó (2013: 380-383).

107 In using this PRT category, we also follow the practice of the English and German implementational grammars.

108 For further details, see Laczkó & Rákosi (2011).

109 For more on XLE’s restriction operator, see Section 5.4 in Forst et al. (2010) and Section 3.1.4.1 in this dissertation.

110 Which is a goal argument.

receives the zero grammatical function (NULL).¹¹¹ The preverb’s second argument is a source oblique (folyó ‘river’).¹¹² Our XLE grammar produced the following analysis.

Figure 16. Laczkó & Rákosi’s (2011: 312) XLE c-structure and f-structure of (73)

In analyzing noncompositional spatial PVCs, in Laczkó & Rákosi (2011) we also adopt Forst et al.’s (2010) XLE approach. For instance, in the analysis of (74) we employ the following lexical forms for the (independently nonexisting) verb and the preverb.

(77) fejez V (↑PRED) = ‘%FN < (↑ SUBJ) (↑ OBJ) >’

(↑ CHECK _PRT-VERB) = +

(↑ PRT-FORM) =c ki

@(CONCAT (↑ PRT-FORM) # stem %FN).

(78) ki PRT (↑ PRT-FORM) = ki

(↑ CHECK _PRT-VERB) =c +.

In the XLE notation, the %FN symbol expresses the value of the PRED feature without its argument structure, see the first line. Within angle brackets in the same line, the argument structure of this noncompositional PVC is given: it is a two-place predicate taking a subject and an object argument. The second line contains one of the two members of a CHECK feature pair. This member is defining and the other is constraining. This is an extremely useful XLE device. Its function is to regulate the obligatory co-occurrence of two elements in a particular configuration. The essence of this _PRT-VERB type CHECK feature is that it requires that the two elements involved must co-occur in a PVC configuration. The third line

111 It has been “restricted out”.

112 The important point here is that the verb mászik ‘crawl’ is strictly incompatible with a source argument.

constrains that the form of the particle in this particular instance has to be ki (out). The fourth line calls XLE’s concatenation (CONCAT) template. The function of this template is to formally combines (concatenates) the two elements, the preverb form and the verbal stem, in a string connected by the hash mark. This string serves as %FN, the value of the PRED feature without the argument structure.¹¹³ So in our analysis of (74), the PRED feature has the following value representation in f-structure (where elnök = president, együttérzés = sympathy).

(79) ‘ki#fejez < [elnök], [együttérzés] >’

As regards the lexical form of the preverb in (78), notice that in this use it has no PRED feature, it only has a FORM feature (whose value is ki), see the first line in its lexical form.

The second line is the other (constraining) side of the CHECK _PRT-VERB coin.¹¹⁴ In c-structure, the preverb and the verb are functional co-heads.

The other noninflecting PVC type we discuss in Laczkó & Rákosi (2011) is illustrated in (80). This is an example of the compositional use of the PVC.

(80) János át lép-ett a kerítés-en.

John.NOM across step-PAST.3SG the fence-on

‘John stepped over the fence.’

The discussion is rather brief, because the only relevant difference between this type and the previous type, or, more precisely, the only property this type has and the other type lacks, is that here the particle, even when it is used compositionally, strictly constrains the case form of its oblique argument. Consequently, we propose the following lexical forms for the particle and the verb as used in (80).

(81) át PRT XLE (↑PRED) = ‘across < %ARG1 (↑OBL) >’

(↑OBL CASE) =c superessive.

(82) lép V XLE (↑PRED) = ‘step < (↑SUBJ) (↑OBL) >’.

The example in (80) is directly comparable to that in (73). The two lexical entries in (81) and (82), again, are directly comparable to (75b) and (75a), respectively. The difference between the two PVC types is captured by the constraining equation in (81).

It is also important to note that in this PVC type, too, we find the same instances of noncompositionality as in the former PVC type. For instance, it stands to reason that (83) is straightforwardly comparable to (74). Consequently, (83) allows and requires the same sort of analysis as we propose for (74).

(83) János át lép-ett a problémá-n.

John.NOM across step-PAST.3SG the problem-on

‘John got over the problem.’

113 Note that this XLE concatenation process is radically different from that assumed by Ackerman et al. (2011). In their system concatenation means the creation of a synthetic form, a morphologically complex word. By contrast, the XLE device only brings about a string in the value of the PRED feature of a complex predicate in f-structure, and the elements corresponding to the two pieces of the string (flanking the hash mark) are still two free morphemes, that is, two independent syntactic atoms in c-structure.

114 Given that XLE does not tolerate multiple entries for the same lemma in its lexicons, in our HunGram grammar we have a single lexical form for the preverb ki (out) and the two representations in (75b) and (78) are expressed disjunctively in a single entry, but this has no theoretical repercussions.

In Rákosi & Laczkó (2011) we develop an XLE analysis of inflected PVCs, again, fundamentally in the spirit of Forst et al. (2010). The two construction types that we concentrate on are exemplified in (84) below:

(84) a. Rá ugrott-ál az asztal-ra.

onto.3SG jumped-2SG the table-onto

‘You jumped onto the table.’

b. Mögé ugrott-ál az asztal-ra.

behind.to.3SG jumped-2SG the table-DAT

‘You jumped behind the table.’

(84a) contains what is often referred to as a reduplicating particle. Elsewhere such a particle functions as a case marker, and in the PVC, it is part of a dependency with a lexical noun phrase that bears the same case morphology as that spelled out by the particle (with some possible but irrelevant phonological differences). What we dub possessive particles function as postpositions elsewhere, and, when used as particles, they license an associate in dative case in the dependency, as in (84b). Here I confine myself to a brief overview of our analysis of the reduplicating type. This is the more interesting type from the perspective of the XLE treatment of PVCs.¹¹⁵ For our treatment of the possessive type, see Rákosi & Laczkó (2011).¹¹⁶

In our analysis of the reduplicating PVC type, we capitalize on a relatively widely-held view in the literature,¹¹⁷ and we make a distinction between the ordinary reduplicating particle use and the pronominal particle use of one and the same form. Consequently, on our account the particle rá is assumed to function as a phrasal pronominal element in (85a), and we treat the reduplicating particle in (85b) as a special agreement marker that became entirely bleached, losing all its semantic content. Our proposal is very close in spirit to that of Ackerman (1987, 1990, 2003).

(85) a. Rá ugrott-ál.

onto.3SG jumped-2SG

‘You jumped onto it/her/him.’

b. Rá ugrott-ál az asztal-ra / az asztal-ok-ra.

onto.3SG jumped-2SG the table-onto the table-PL-onto

‘You jumped onto the table / the tables.’

We assume the following lexical representation for the pronominal particle in (85a).

(86) rá: Pron (↑PRED)= ‘pro’

(↑CASE)= sublative (↑PERS)= 3

(↑NUM)= SG

115 The essence of this relevance is as follows. In Forst et al. (2010) we programmatically assume that this PVC type can be used compositionally and productively, and, consequently, in such cases it should be analyzed along the syntactic complex predicate formation via restriction lines. By contrast, on the basis of our findings about the behaviour of these reduplicating PVCs, in Rákosi & Laczkó (2011) we claim that even the productive-looking cases are fraught with idiosyncrasies, and, therefore, a lexical analysis along the concatenational lines is more appropriate.

116 We argue that this type is a rather marked (and speaker-dependent) construction, and it calls for a special lexical treatment. For the details, see Section 4 in Rákosi & Laczkó (2011).

117 For an overview, see Section 3 in Rákosi & Laczkó (2011).

The entire lexical form is treated as a pronoun that projects a DP, rather than a PP.¹¹⁸ Moreover, in this representation the case suffix itself does not have a PRED feature, but only a CASE feature, which can possibly be interpreted compositionally in semantic structure. This is the current state of affairs in our XLE-implementation, but nothing crucial hinges on this particular assumption. The essence of our argumentation and the analysis would not change if we handled these inflecting case markers as P-elements with a PRED feature, taking lexical or pronominal P-objects.

The lexical forms for the reduplicating particle and the lexical verb are given in (87a) and (87b), respectively.

(87) a. rá: PRT (↑PRT-FORM) = rá (↑OBL PERS) =c 3

(↑OBL CASE) =c sublative (↑ CHECK _PRT-VERB) =c +

b. ugrik: V (↑PRED)= ‘rá#ugrik <(↑SUBJ) (↑OBL)>’

(↑PRT-FORM)=c rá

(↑CHECK _PRT-VERB) = +

@(CONCAT (↑ PRT-FORM) '# %stem %FN)

The particle in this use is a nonprojecting category (PRT). Given that it is compatible with either singular or plural associates, see (85b), we take it to be underspecified for the NUMBER feature (which is formally expressed here as the absence of this feature). The particle constrains two properties of the oblique associate: its PERSON and CASE features. It is in this respect that these reduplicating particles can be considered special agreement markers. As (87a) and (87b) show, the particle is specified to form a PVC with the verb (and vice versa) via the “CONCAT template, CHECK feature and PRT-FORM specification”

machinery employed by Forst et al. (2010) and Laczkó & Rákosi (2011). Our implemented grammar analyzes (85a) and (85b) in the following way.

Figure 17. The c-structure and f-structure of (85a)

118 We follow Bartos (1999), among others, when we make a categorial distinction between inflected case suffixes and inflected postpositions (which are assumed to project a PP).

Figure 18. The c-structure and f-structure of (85b)

We give the following justification for treating even the productive-looking uses of reduplicating PVCs lexically. “The primary reason why we decided to store every attested reduplicating particle plus verb combination in the lexicon is that the majority of these combinations (both in terms of types and tokens) are in fact noncompositional. It is actually not easy to find compositional reduplicating PVCs in corpora. It should also be added that there is quite a lot of idiosyncrasy involved in whether this kind of reduplication is obligatory, possible or unavailable for any potential verbal host. As a rule of thumb, it is the inherent aspectual feature of the particle that drives the combinations. The particle rá ‘onto’, for example, has a telic nature. Thus, this particle is usually obligatory if the resulting complex is telic (88a), and it is unavailable if the intended verbal meaning is atelic (88b). However, the particle can be optional in telic complexes (89a), and it can even be obligatory in atelic ones (89b).

(88) a. Nem jövök *(rá) a megoldás-ra.

not come.1SG onto.3 the solution-onto

‘I cannot figure out the solution.’

b. Nem tartozik *(rá) Kati-ra.

not belongs onto.3 Kate-onto

‘It does not concern Kate.’

(89) a. Nem rivallt-am (rá) Kati-ra.

not yelled-1SG onto.3 Kate-onto

‘I did not yell at Kate.’

b. Nem szorul-ok (rá) Kati-ra.

not press-1SG onto.3 Kate-onto

‘I stand in no need of Kate(’s help).’

We, therefore, believe that it seems justified to subject reduplicating constructions to a lexical treatment in compositional and noncompositional cases alike. In this, we follow previous accounts that treat these particles as derivational elements (see especially É. Kiss 1998a and

Ackerman 1987, 1990, 2003)” (Rákosi & Laczkó 2011: 456-457).¹¹⁹ Thus, one of the key points here is that reduplicating PVCs are often fully compositional; however, they are far from being productive, which calls for a lexical approach.

At this point a brief comparison of our approach and the RBL approach, discussed in a detailed fashion in Section 3.1.2.3, is in order.

 In the case of noncompositional PVCs both approaches propose a fully lexical treatment, thereby respecting classical LFG’s subscription to the Strong Lexicalist Hypothesis.

o RBL employs the notion of analytic morphological object, as a marked option for lexical form representation.

o Our approach, by contrast, employs an apparatus which is capable of maintaining the

“one lexical item – one morphological word – one syntactic atom” correspondence in such a way that it can still capture the marked behaviour of (noncompositional) PVCs.

For this purpose, it applies a system of devices: efficient cross-referencing between distinct lexical items via appropriate constraining equations and CHECK-features. The analysis has been successfully tested implementationally, which can be taken to be a rather strong indication of its feasibility.

 In the case of compositional and productive PVCs the two approaches are radically different.

o RBL strictly maintains its lexicalist view (in actual fact, fundamentally it applies a uniform treatment of both major PVC types).

o Our approach, by contrast, employs a syntactic complex predication formation device, thereby violating the Strong Lexicalist Hypothesis.¹²⁰

In document Laczkó Tibor Aspects of Hungarian Syntax from a Lexicalist Perspective (Pldal 186-192)