Introduction: root based or derivational morpho-syntax?
In their recent paper Alexiadou and Doron (2012) propose a universal analysis1 of the derivation that involves active, middle and passive constructions in languages of the world, taking into consideration the morpho-syntactic angle of this derivation. In their model the transitive and intransitive constructions based on alike-looking verbs are not derived from each other. For instance the verbs in the sentences below are not connected by any derivational means:
John broke the vase
The vase broke
The vase was broken
In other approaches derivational links are recognized between the verbal form in (1) and these in (2) and (3), together with necessary morpho-syntactic re-arrangements.
The element connecting all the three predicates in Alexiadou and Doron’s (2012) approach is the root. When placed in appropriate structures, together with properties brought in by particular heads added to the structures, the root answers for the varied characteristics of the above verb-argument combinations. In other words, particular verbal uses are not morpho-syntactically, or otherwise, derived from each other. The superficial similarity of the verbs in terms of phonological properties and argument structures associated with them is due exclusively to the fact that the verbs contain the same minimal lexical element – the root. Such a view of morpho-syntax gains popularity nowadays, as represented by works of Arad (2005), Pylkkänen 2008, Alexiadou and Anagnostopoulou (2004), Alexiadou (2010), Alexiadou and Doron (2012), and others.
According to the views prevailing in morpho-syntax so far, the verbal forms above derive from each other. A position along these lines is argued for by Chierchia (2004), Reinhart and Siloni (2004), (2005), Everaert, Marelj and Siloni (2012), and others who see intransitive variants as produced from transitive verbs by special morpho-lexical operations reducing valency values of the basic lexemes. This approach has been criticized because not all intransitives, even these with characteristic non-active morphology2, possess corresponding transitive forms.
For instance in Polish we have intransitive verbs with unaccusative semantics ( e.g. change of state verbs) and characteristic marking, which in this language is reflexive for non-active and non-passive verbal uses: e.g. pocić się ‘sweat’, chmurzyć się ‘become cloudy’, dowiadywaćsię ‘get to know’, ulatniać się ‘evaporate’, zżywać się ‘get used to’, wściekać się ‘get angry’, wdawać się ‘grow to be like’, wynużać się ‘emerge’, wymykać się ‘elude’, wradzać się ‘from one’s birth be like sb.’, etc. These verbs have no transitive counterparts – prospective bases, as well as no other lexemes which could constitute convincing sources for their derivation: related lexemes (whenever they appear) are de-verbal resultative adjectives, which could arguably be based on these very verbs: zachmurzony ‘clouded’, spocony ‘sweaty’, zżyty ‘befriended’, wściekły ‘crazy’, wynużony ‘surfaced’, wrodzony ‘in-borne’.
The position taken by derivationists, such as Reinhart, is that in such problematic cases a potential basic verb with transitive characteristics exists, and it just happens not to be used. The disadvantages of such a position are obvious, so it might seem that the opposite view wins: both types of structures, i.e. transitive and intransitive, are derived directly from roots.
The viability of the root-base hypothesis depends on some verifiable linguistic regularities.
First of all it should follow that the intransitive verbs derived from verbal roots with a specific morphology should/could have different properties from intransitive verbal forms3 without such morphology: morphologically marked intransitives would have complex structures, dissimilar from those of unmarked, less complex verbs. The varied degree of complexity should translate into differences in linguistic behavior of the two classes of intransitives (cf. sec 3.).
Then, there should be no systematic similarities in idiosyncratic/sub-regular behavior obtaining between paired (form-wise) transitive and intransitive verbs: sporadic alternations should appear in a haphazard way with intransitive and transitive verbs, and not with transitive-intransitive pairs. If such alternations appear in pairs of verbs – derivational relationships between them seem convincing (cf. sec. 4).
In this text we ask whether morpho-syntactic data in Polish from the sphere of reflexive marking speak in favor of the derivationist or root based morphology. We will show with respect to the two points above that the derivationist approach scores better. No doubt much more research has to be conducted to arrive at a definite answer in this respect. Still for now we will show that Polish shows no expected dissimilarities between bare and marked unaccusatives and there exist striking similarities between paired transitives and intransitives. Consequently, the derivationist position gains some support from the Polish data.
Morpho-syntactic formations and the middle voice head
As this paper is concerned with the morpho-syntactic phenomena pivoted on the anticausative formation, let us first introduce the rudiments of Alexiadou and Doron’s (2012) account of the non-active verbal morphology, i.e. the area wherein anticausativization belongs. Alexiadou and Doron (2012) claim that three systems of morphological marking can be distinguished for verbal valency and voice among the languages of the world. In some languages, like English, the passive voice is marked with non-active morphology, while the remaining valency classes bear the active marking. In languages like Modern Greek, on the other hand, one type of morphology – e.g. passive – marks a variety of verbal morpho-syntactic classes (anticausatives, reflexives, reciprocals, dispositional middles, medio-passives and passives) as distinct from truly active verbs. In the third type, there is active morphology for regular transitive verbs, another morphological marking for reducted valency verbs (middles, reflexives, reciprocals, dispositional middles, medio-passives) and the third kind – for passive forms. Polish belongs to the third group. Particular distinctions among valency classes and their morphological markings are attributed to the presence of different structural heads in verbs. The head which is of special interest to us is the middle (voice) head – μ. The overt morphology which accompanies the presence of the middle head may be the same that realizes the passive voice (with a different head) in e.g. Modern Greek. This whole area of morphology in a given language is called non-active morphology. Alexiadou and Doron (2012) state that in Romance languages the morphology associated with the middle head is reflexive. Polish seems to be another language where reflexive morphemes accompany ‘middle’ verbal uses. These uses, to be presented in greater detail below, encompass various cases of Alexiadou and Doron’s (2012:3) non-active morphology:
anticausatives (i.e. spontaneous events like break, open)
reflexives which are mostly limited to verbs of body care (wash, comb) and naturally reciprocal events (meet, kiss)
dispositional middles (This book sells well)
Out of these classes, in the languages which have more than one non-active morphological pattern, passives are realized with another head and other morphemes than the rest. The remaining categories share overt morphology, though the structures deriving them will differ, while the middle head is the constant factor. Polish seems to belong to the class of these mixed type languages as it has two non-active morphological devices. Special morphology derives passives, while all the remaining cases can be realized with the use of reflexive morphology. Below we will illustrate the relevant structures in English, Greek and Polish to give an idea how meanings correlate with morphological exponents. As the first group, we introduce anticausatives:
English: The window broke.
Greek 4 (from Alexiadou and Doron 2012:5): i supa kaike ‘The soup burnt’
Polish: Okno się zbiło ‘The window broke’, Zupa się spaliła ‘The soup burnt’
Reflexive and reciprocal meanings are expressed alike:
English: She washes every day ; They fight fiercely.
Greek: i Maria htenizete ‘Mary combs herself’
Polish: Maria się myje ‘Mary washes’
Dispositional middles again are formed alike, though in Polish they are infrequent formations, mostly ousted by impersonal constructions:
English: This book reads easily.
Greek: Afto to vivlio diavazete efkola ‘This book reads easily’
Polish: Ta szkoła już się buduje osiem lat ‘This school has been being built for eight years’5
Medio passives represent a variety of constructions, where traces of non-active morphology show up, yet the presence of an external instigator is implied:
English: The job got done.
Greek: To vivlio diavastike ‘The book was read’
As Alexiadou and Doron (2012) claim that medio-passives appear in lexically conditioned cases where passives cannot be formed, we assume that for Polish such structures as the following will constitute examples for medio-passives: To się wie ‘It is known, it is obvious’, To się czuje ‘It is felt’, etc. Passives are not acceptable with the roots contained in these forms, so medio-passives take over. Like in passives, in medio-passives the participations of an (unnamed) instigator (e.g. experiencer ) is implied.
In Polish all the above verbal uses require the presence of the reflexive clitic element się, which although movable, nevertheless constitutes a morpho-syntactic complex with the verbal stem.6
Polish reflexive morphology performs also other functions which are less apparently non-active. It marks reflexiva tantum, e.g.: bać się ‘be scared’, podobać się ‘appear pretty to sb.’, szwendać się ‘loiter’, ohajtnąć się ‘get married, colloq.’, przytrafić się ‘happen’, kłaniać się ‘bow’, ulatniać się ‘disappear, colloq.’, etc., which show a variety of meanings. Some of them may be seen as non-active, but it is not a rule. Apart from reflexiva tantum, there are also isolated antipassives 7 with the reflexive morpheme: pakować się ’pack up’, grzebać się ’procrastinate’, zastrzegać się ‘protest’, przedrzeźniać się ‘mimick, colloq.’. Also occasional unergatives are created in this way: stawiać się ‘oppose, colloq.’, wtryniać się ‘butt in, colloq.’. However, these uses, which are not in the class of non-actives according to Alexiadou and Doron’s (2012) terminology, are few and far between and consequently the overall picture of the reflexive morphology in Polish classifies Polish as a system with two types of non-active morphology: middle – expressed with reflexive forms and passive. In the remaining parts of this paper we will concentrate on anticausatives (sec. 3)and dispositional middles (sec. 4) to see how well they befit the picture proposed by the two authors, the picture which is to account for certain morpho-syntactic properties in universal terms. We will show that in spite of theoretically adequate reflexive morphology, Polish does not fall in with Greek, English or Hebrew. We will also try to investigate what the reasons for this different behavior might be.
Active and non-active anticausatives (Alexiadou and Doron 2012)
Alexiadou (2010), Alexiadou, Agnostopoulou and Schäffer (2006), Alexiadou and Doron (2012) propose that anticausative/unaccusative8 verbs are of two forms in languages of the world.
Unaccusatives with active morphology and adjectival bases have simple structure of the form in (4) below (adopted from Alexiadou and Doron 2012):
γ R λe[dry (e,x)]
x √[R dry ] λxλe[dry (e,x)]
where γ is a causative head, which accounts for the derivation of the relevant verbs with active, unmarked inflection. They also mention an option where no cause is specified and no voice head appears in the structure, as it would be superfluous with verbs without a trace of external arguments. Such anticausative verbs would not have the causative (or any other) voice head.
Unaccusatives with non-active morphology in the form of e.g. passive marking, as in Greek, or reflexive morphology, as is the case in Polish suszyć się ‘to dry’, have an additional layers of structure introduced by special non-active middle voice head μ:
Modified from Alexiadou and Doron (2012:15):
μ λe[suszyć9(e, x)]
x μ λx λe[suszyć(e, x)]
μ √[R suszyć] λxλe[suszyć(e,x)]
This voice head (Alexiadou and Doron 2012:14) "modifies the root by reclassifying it with respect to its requirement of an external argument". In other words, while the root requires e.g. an external causer, the middle verb does not require it any longer:
Jan/wiatr stłukł wazon ‘John/wind broke the vase’ vs. Flakon stłukł się (od wibracji) ‘The vase broke (from vibrations)’.
If these differences in structure correlate with differences in behavior of morphologically marked and unmarked anticausatives, then the distinction could be felt to be motivated for Polish. However, if they fail to diverge, then the conception becomes questionable. This distinction into two classes of anticausatives is argued for in detail in Alexiadou (2010). She makes a note of the problematic issue and admits that anticausatives with structures in (8) and (9) do not show differences in behavior: they can appear indiscriminately with the by itself phrase in Greek, Hebrew and English and cannot appear with the agentive by phrase. Still she insists that some differences can be noticed: the formations of verbs in (8) and (9) correlate with chosen classes of roots, i.e. with externally caused ones in Greek for the forms with non-active (in Greek – passive) morphology (9), while internally caused – with active morphology verbs (8). Leaving aside the problem with establishing on practical grounds which roots10 are externally and which are internally caused, the distinction into morphologically marked with non-active morphology externally caused anticausatives and bare – internally caused anticausatives does not seem to work in Polish: here the same roots appearing in the same clauses can derive both marked and unmarked anticausatives11. Below we will illustrate the phenomenon of such doublets. They are very numerous in Polish, and if we take seriously the claim that roots condition the choices of anticausative formations, we are stuck with the forms like these in (11). Below (12-14) we will show that both classes of anticausatives in Polish appear in the same syntactic environments as well.
czerwienieć – czerwienić się ‘grow red’, zielenieć – zielenić się ‘grow green’, schnąć –suszyć się ‘grow dry’, moknąć – moczyć się ‘get wet’, chłodnąć (obs.) - chłodzić się ‘get cold’, potnieć – pocić się ‘sweat’, etc. etc.
Moreover, anticausatives with and without marking appear with the same PPs. Contrary to some descriptions of the Polish language12, we have established on the basis of the corpus studies that both types of anticausatives can appear in Polish with by itself phrase. In (9) we illustrate such situations with anticausatives (from NKJP) with morphological reflexive marking, in (10) – without the marking13:
drzwi zamykają się same podczas jazdy ‘the door closes by itself while in motion’
brama otwiera się sama ‘ the gate opens by itself’
auta zwykle niepsują się same ‘usually cars do not brake down by themselves’
bądź takim, abyś nie musiał czerwienić się sam przed sobą ‘be so that you do not have to grow red by yourself faced with yourself’
lakiery muszą schnąć same ‘top coats must dry by themselves’
więdnie samo to co złe a dobre - pięknie szeleści i idzie dalej ‘whatever is evil wilters by itself, whatever is beautiful rustles and goes on’
kochane ciało chudnie samo ‘the loved body loses weight by itself’
drzewa nie rosną same, tak po prostu ‘trees do not grow by themselves, just like that’
Like in English, Greek and other languages, in Polish, according to the data taken from NKJP, by itself phrase can appear with the two types of anticausatives indiscriminately.
Similarly, when we consider this very same corpus, we do not perceive differences in other PPs that can modify the events described by the pairs of anticausatives, thus failing to testify to their distinct structure. According to Alexiadou and Doron (2012), unmarked anticausatives should be formed from intrinsically caused roots. When there is a need to refer to the external argument of the agentive type, such roots may be equipped with non-active morphology and accompanied by agentive PPs14. This could be expected of Polish doublets: we may think that both bare and reflexively marked forms serve the purpose of distinguishing the events with and without agentive participants. Unfortunately, no traces of the suggested dichotomy can be found. Typical agentive phrases, like the ones accompanying passive structures in Polish, fail to appear with anticausatives altogether: przez Jana, which has the agentive interpretation in a passive structure: Ta praca została wykonana przez Jana ‘This work was done by John’, gains a completely different meaning with anticausatives. The phrase expresses a very indirect causation, never agentivity, and it may accompany bare and marked anticausatives indiscriminately, e.g. : Ania poczerwieniała przez Jana, Ania zaczerwieniła się przez Jana ‘Anna reddened because of John (‘s presence, something he did, etc.)’. That it does not stand for an Agent-participant is made clear by such NPs complementing the preposition which are emphatically non-agentive: Ania poczerwieniała przez wiatr, Ania zaczerwieniła się przez wiatr ‘Anna grew red because of the wind’.
A typical prepositional phrase which appears with anticausatives of both kinds contains the preposition od and it introduces non-agentive causers with both types of anticausatives without the slightest discrimination, as the examples in (14) below show:
topniejących odlawy zboczy lodowcowych ‘melting from lava glacial slopes’ vs.
opony topią się od tarcia ‘tiers melt from friction’
w błocie twardniejącym odmrozu ‘in the mud hardening from frost’ vs. olej roślinny utwardzający15 się odtlenu atmosferycznego ‘plant oil hardening from atmospheric oxygen’
czerwieniał od niewyładowanego gniewu ‘he reddened from suppressed anger’ vs. czerwieniąc się odrozkoszy ‘reddening from pleasure’
potnieli odupału‘they sweated from heat’ vs.pociły się odupału ‘they sweated from heat’
Atmosfera gęstnieje oderotyzmu ‘The atmosphere thickens from eroticism’ vs. sytuacja, osadzona w surowym, pustym krajobrazie, zagęszcza się odemocji ‘the situation situated in severe, empty environment, thickens from emotions’
Może dlatego, że mieszkają w namiotach i są wściekli, bo cały czas mokną oddeszczów ‘Perhaps because they live in tents and are angry, since they get wet from rain the whole time’
moczyły się odpotu, a nie przemakały ‘they got wet from sweat, but did not soak through’
płonące oduderzeń piorunów zabudowania ‘the outhouses burning from the striking thunderbolts’ vs. trzy razy paliła się odpioruna ‘it burnt three times from the striking thundrbolt’
If any differentiation is observed at all, it is the greater tolerance of bare anticausatives for accompanying PPs (in terms of numbers of relevant examples), than of those with non-active morphology. In the light of the hypothesis that the first group represents internally conditioned roots, while the other – externally conditioned ones, it is clearly unexpected. In fact the anticausative doublets in Polish and their uniform behavior undermine the sense of an analysis along the lines of Alexiadou and Doro (2012) for Polish in particular, and also more generally - as their theory has universal ambitions.
One important point has to be mentioned here to give a more honest picture of the situation in Polish. In fact the members of doublets differ with respect to whether they have more or less similar transitive forms related to them. Of course in the root based morphology of Alexiadou’s type this relationship is not derivational in nature16. Nevertheless, it may support the ‘more agentive’, or at least ‘more externally conditioned’ character of the forms with non-active anticausative morphology. What we have in mind is that się verbs are more like transitive verbs in form than bare anticausatives. In a derivational approach we would be entitled to say that they are based on the same derivational stem:
Mróz czerwienił policzki ‘Frost made the cheeks grow red’ vs. Policzki czerwieniły się od mrozu ‘ The cheeks grew red from frost’ vs. Policzki czerwieniały17 od mrozu ‘The cheeks grew red from frost’
Mężczyzna suszył buty ‘A man dried his shoes’ vs. Buty suszyły się ‘The shoes were drying up’ vs. Buty schły ‘The shoes were drying up’18 Consequently, it might be argued that the verbs with reflexive morphology have roots differently caused (externally, agentively) than bare anticausatives (internally). However, quite apart from the fact that the difference between the two anticausatives according to the Slavic tradition would not concern roots, but more morphologically complex entities which are verbal stems19, there are many pairs based on the same root where the non-active verb has no transitive counterpart (see 16 below), as well a still greater number of anticausatives with reflexive morphology and no transitive counterpart (see 17 below) to which this line of reasoning cannot be applied:
pękać20 – rozpękać się ‘bust’, promienieć - rozpromienić się ‘radiate light’, kwaśnieć - skwasić się ‘go sour’, potnieć - pocić się21 ‘sweat’, rdzewieć - rdzawić się ‘rust’, wyschnąć - rozsychać22 się ‘dry up’, zaśmierdnąć - zaśmierdzieć się ‘stink’, postarzeć23 - postarzeć się ‘grow old’.
spietrać się ‘get scared’, rozlatywać się ‘break into pieces’, akomodować się ‘accomodate’, ustatkować się ‘settle down’, narodzić się ‘get born’, zagnieżdżać się ‘get nested’, rozśnieżaćsię ‘get snowy’, przewężać się ‘get narrower’, skrupić się ‘be focused on sb. (about sth. negative)’, mgławić się ‘get foggy’, załzawić się ‘produce tears’, przekrwić się ‘get bloodshot’, rozbuchać się ‘intensify (negative)’, zalęgnąć się ‘get fertile’, wykluć się ‘hatch’, roztyć się ‘grow fat’, zakatarzyć się ‘get the running nose’, etc., etc.
Consequently, Polish neither supports the hypothesis that unmarked anticausatives should be created from internally caused roots, nor does it uphold the claim that verbs with non-active morphology have externally caused roots. It also fails to supply the data with would show that the exceptionally formed marked verbs with inherently caused roots used in the circumstances that require specifying causation show agentive PPs accompanying them. Thus the question arises – is Polish a freak of nature?
We feel that there are two possible explanations for this state of affairs. The first one undermines the suggestion that reflexive morphology used in Polish to mark non-active voice in the understanding of Alexiadou and Doron (2012) has the same properties as the passive marking functioning in the prevailing body of data from various languages discussed by them (and by Alexiadou 2010). Differences in morphological marking mechanisms may after all be reflexes of some structural details and perhaps assuming that reflexive and passive marking should be identified with each other in their functions in respective languages does not entail that all particulars in both types of languages are the same.
Another explanation may be connected with a more thorough diachronic investigation of the Polish data. It is just possible that the Polish language in the sphere of anticausatives is going through the intermediate period of change. It seems that the anticausative forms with się oust bare anticausatives in many cases and we are able to quote many bare verbs that sound obsolete or obsolescent:
Thus it stands to reason that the unruly Polish system is at the moment out of the expected equilibrium and, in consequence, it disobeys the regularities proposed for the world of anticausatives by Alexiadou and Doron (2012).
A case for the derivationist approach - anticausatives from ‘clear’ verbs
Another conclusion to be drawn from the Polish anticausative data might be that the whole system based on roots as basic units of verbal morpho-syntax is not the right track to follow and that the derivational model is more adequate. One argument for the derivational nature of anticausatives in Polish has already been given. Namely, if we consider the anticausatives with się as derived from their transitive counterparts, we can explain the otherwise inexplicable similarities between transitive and intransitive verbal forms25 in ( 15) above.
Of course, if we want to argue in favor of the derivationist approach, we are left with a number of anticausatives which look as if they were derived, but lack the relevant transitive bases and consequently their unaccusative semantics cannot be attributed to valency reduction (see 16, 17), but is accidental26.
In the root based approach such verbs do not constitute a problem since the ultimate basic unit is the root and roots are never lacking. Their unaccusative semantics is safeguarded by proposed middle voice heads, at least for most subclasses27.
Consequently, although we do not have direct evidence for the kind of root based structures proposed by Alexiadou and Doron (2012), still the root-based approach, like derivational approach, scores some points in the field of Polish anticausative morpho-syntax.
However, the derivationist position makes much sense for certain areas of Polish morpho-syntax and we would like to present one such case – also from the area of non-active verbal morphology.
The relevant body of data the so called ‘clear’ verbs, as described by Rappaport Hovav and Levin (1998), Levin (2006) for English, Segal and Landau (2012) for Hebrew, Alexiadou and Anagnostopoulou (2012) for Greek. These verbs are very interesting in themselves but we will concentrate in this text on these properties which are significant from the point of view of the derivationist vs. root based morphology debate.
First, however, let us introduce the data in some detail. ‘Clear’ verbs have a curious property in various languages which allows the users to exchange arguments in the frames in which these predicates appear without introducing any serious change in meaning. So the verbs can be followed by a direct object location and a PP expressing substance (Stuff), or the other way round: by the direct object Stuff and a location PP. In English (after Rappaport Hovav and Levin (1998) and Levin (2006)) we have the following alternation possibilities:
John cleared the bar top of glasses (External argument+verb+Location+Stuff) vs.
John cleared glasses from the bar top (External argument+verb+Stuff+Location)
The pattern of arguments in (19) is called the COS frame (Change of State), the ones in (20) – COL (Change of Location).
However, only a sub-group of verbs which appear in either of the frames may participate in the alternation. There are COS- only verbs and COL- only ones:
łamać ‘break’ (COL-only verb)
COL - Łamię gałązkę z rosnącego przy drodze bzu. ‘I am breaking a twig from a lilac bush growing near the road’
COS – *Łamię rosnący przy drodze bez z gałązki. (A.M.) *’I am breaking a lilac bush growing near the road of a twig’
leczyć ‘cure’(COS-only verb)
COS - Ja nie ośmieliłbym się leczyć kogoś zdepresji. ‘I would not dare to treat anybody for depression’
COL – *Ja nie ośmieliłbym się leczyć depresję z kogoś. *’I would not dare to treat depression from anybodyody’28 ‘Clear’ or ‘deprivation’ verbs – called thus because of their semantics29 - figure prominently in the alternating group, still the possibility of the alternation has to be established on the verb by verb basis. There have been suggestions that the alternation is possible with manner verbs, but not with result verbs, however for Polish, as we have shown in Malicka-Kleparska (to appear b), it is not predictable on the basis of the meaning of verbs which of them undergo the alternation and which do not30.
Still one striking regularity can be noticed: if a verb alternates, its dispositional middle or anticausative correspondent alternates too in a predictable way (cf. however Alexiadou and Agnostopoulou 2012 for Greek31). If, as we claim, the appearance or non appearance of the alternation is an unpredictable property of particular verbs, to find the same behavior with their dispositional middle or anticausative counterparts suggests the derivation of the latter from the former: Thus, in the simplest possible way, the properties of the base would be carried onto the derivative. Consequently the derivationist position would be naturally supported by this body of data.
The root based morphology would be equally expedient if we could show that it is the property of the root that preconditions the alternation. We shall demonstrate that it cannot be the case – in Polish verbs with the same root may or may not enter the alternation, depending on the presence of other morphological elements. On the other hand verbs which share larger portions of structure (e.g. the same prefixes and the root) behave in an identical fashion with respect to possibilities of participating in the ‘clear’ alternation. Consequently, the root based model fails to capture this regularity.
Consider for instance the verb tłuc ‘break’, which is at the same time a root. It does not tolerate either the COL nor the COS frame, it appears with a direct object, but not with two internal participants:
*Tłucze ozdobę ze szklanki ‘He breaks an ornament from a glass’, *Tłucze szklankę z ozdoby ‘He breaks a glass from an ornament’
Likewise, dispositional middles32 are not possible with this verb:
*Ozdoba tłucze się ze szklanki ’An ornament breaks off a glass’ *Szklanka tłucze się z ozdoby ‘A glass breaks off an ornament’
If the root based model was adequate, we could expect the same behavior of all the verbs based on one root, in this case the non-alternating behavior. The data do not uphold this expectation. If the verb tłuc is prefixed with ob-, the morphologically complex verb and its dispositional middle/anticausative counterparts show the same preferences for structural surroundings – they take the COL frame. The frame must be modified for anticausatives as the direct internal arguments of the transitive verb has to serve as the external argument of the anticausative/dispositional middle derivative.
Obtłukł ozdobę ze szklanki (COL structure) ‘He broke and ornament from the glass’ vs. Ozdoba obtłukła się ze szklanki (anticausative) ‘An ornament broke from the glass’ vs. Ozdoby obtłukują się łatwo ze szklanek (dispositional middle) ‘Ornaments break off easily from glasses’
We can observe similar behavior with a number of roots and verbs. Here we supply just a handful of examples:
Dziabać ‘cut, colloq.’– COL-only verb: Dziabał gałęzie z sosny ‘He cut branches from a pine tree’ vs. *Dziabał sosnę z gałęzi ‘He cut the pine tree from branches’
At the same time, the prefixed verb and its dispositional middle correlate and occur in both COL and COS frames:
Obdziabał sosnę (Location) z gałęzi (Stuff), Obdziabał gałęzie (Stuff) z sosny (Location) ‘He cut the pine tree off branches/branches from the pine tree’ vs. Sosna (Location) łatwo obdziabała się zgałęzi (Stuff) (dispositional middle) ‘The pine tree got its branches cut off easily’, Gałęzie (Stuff) łatwo obdziabały się z sosny (Location) (dispositional middle)‘Branches got cut off easily from the pine tree’
Rąbać ‘hack’ – COL-only verb
obrąbać ‘hack off’- alternating verb, dispositional middles show the same alternation:
Obrąbał sosnę z gałęzi (COS), Obrąbał gałęzie z sosny (COL)‘He hacked off branches from a pine tree’ vs.
Gałezie obrąbują się łatwo z sosny (COL) ‘Branches hack off easily from a pine tree’
Sosna obrąbuje się łatwo z gałęzi ‘A pine tree hacks easily of branches’
gryźć ‘bite’– neither COS nor COL frame is possible, only the direct object expressed with a bare NP
obgryzać ‘bite off’ – alternating verb, dispositional middles show the same frames:
Obgryzał chleb ze skórki (COS), Obgryzał skórkę z chleba (COL) ‘He chewed the crust off a bread loaf’ vs. Chleb łatwo obgryzał się ze skórki (COS) ‘The loaf of bread got the crust chewed off ’ Skórka łatwo obgryzała się z chleba (COL) ‘The crust chewed off easily from the loaf of bread’
Examples can be easily multiplied. All of them show that an idiosyncratic property of a verb is not at the same time necessarily a property of its root. This property is nevertheless shared by complex verbs and their anticausative/dispositional middle correspondents. This regularity cannot be easily represented in a root based approach to verbal morpho-syntax. Of course there is always a way round a difficulty. One may for instance claim that the frame properties are contributed to the structure by various prefixes. However, as the difference between (24) and e.g. (26) shows that it is the prefix plus the root that decides the alternating properties and not the root or the prefix separately: The same prefix with different roots results in distinct frames; ob- with tłuc gives the COL-only frame (see 24 above), while in the remaining examples with ob- alternating frames are available (see 26-28 above).
Verbal prefixes different than ob- show parallel behavior: it is a combination of a particular root and a particular prefix which brings about a specific frame, which is then used by a dispositional middle or by an anticausative. Neither the root nor the prefix alone can be held responsible. For instance odskrobać ‘scratch away’ appears with the alternating frame33, while oddłubać ‘dig out’,odrzucać ‘throw away’ with COL-only. On the other hand ogołacać ‘lay waste’, okradać ‘steal’, ograbić ‘rob’ are COS-only verbs, while oskubać, with the same prefix, is an alternating verb.
Moreover, unprefixed verbs also have their idiosyncratic behavior in this respect: strzyc is an alternating root verb, while e.g. podstrzyc takes only a bare object.
All the corresponding dispositional middles and anticausatives have the expected frames corresponding to causative relatives, and not to roots.
The data discussed in this text show that as far as verbal morpho-syntax with non-active morphology in Polish is concerned, the root based approach does not fulfill the promise of accounting for its properties in terms of more universal principles and structures. It may be the case that the Polish anticausative morpho-syntax is in a state of flux. However, there are also arguments of a different nature against adopting this model: some data show the phonological similarity of verbal stems between transitives and one subgroup of anticausatives and we notice the argument frame corespondence within the group of ‘clear’ verbs. These facts favor the derivationist approach. It accounts in a natural way for these phenomena as it links similar forms by means of derivational rules, while the root based approach has no such mechanism. We believe that discoveries of this kind may cast doubt on the root based model as a whole. Alternately, it is not beyond belief that some areas of morpho-syntax are derivational, while others result from root structures.
Running head: Polish non-active morpho-syntax
Abstract: Alexiadou and Doron’s (2012) universal proposal concerning middle voice morpho-syntax has been applied to Polish data, and to anticausatives and dispositional middle formations in particular. It turns out that the universally valid structures do not have the expected explanatory power for Polish. One reason might be the diachronically changing anticausative system of Polish. However, certain regularities observable in Polish suggest that a root based model – like the one proposed by Alexiadou and Doron – fails to capture significant linguistic interdependencies and a derivation model seems more appropriate. These regularities include phonological likeness of transitive verbs and a subclass of anticausatives (with reflexive marking). Still more intriguing is the idiosyncratic character of argument frames (change of state vs. change of location) with ‘clear’ verbs, mirrored by the distribution of the frames with dispositional middles – the relationship which cannot be encoded in the root based model.
Key words: anticausative, , middle voice, reflexive, morpho-syntax, root
Alexiadou, A. and E. Anagnostopoulou. 2012. "Manner vs. result complementarity in verbal alternations: A view from the clear alternation". To appear in: Proceedings of North EastLinguistic Society 42.
Alexiadou, A. 2010. "On the morpho-syntax of (anti)causative verbs". In: Rappaport Hovav, M., E. Doron, and I. Sichel (eds.), Lexical semantics, syntax, and event structure. Oxford: Oxford University Press. 177-203.
Alexiadou, A. and E. Doron. 2012. "The syntactic construction of two non-active voices: passive and middle" Journal of Linguistics 48 (01). 1-34.
Alexiadou, A., E. Anagnostopoulou and F. Schäffer . 2006. "The properties of anti-causatives cross-linguistically". In: Frascarelli, M. (ed.), Phases of interpretation. Berlin: Mouton de Gruyter. 187-212.
Alexiadou, A. and E. Anagnostopoulou .2004. "Voice morphology in the Causative-inchoative alternation; evidence for a non-unifiedstructural analysis of unaccusatives". In: Artemis A., E. Anagnostopoulou and M. Everaert (eds.), The unaccusativity puzzle. Oxford: Oxford University Press. 114-136.
Arad, M. 2005. Roots and patterns: Hebrew morpho-syntax. Dordrecht; Springer.
Chierchia, G. 2004. "A semantics for unaccusatives and its syntactic consequences". In: Alexiadou, A., Anagnostopoulou, E.and M. Everaert (eds.), The unaccusativity puzzle: explorations of the syntax-lexicon interface. Oxford: Oxford University Press. 22-59.
Dermidache. H. 2005. " What do pieces of words name/ Handout". Talk presented at the University of Vienna.
Doroszewski, W. 1963. Słownik języka polskiego [Dictionary of the Polish language]. Warszawa: Państwowe Wydawnictwo Naukowe.
Everaert , M., Marelj, M. and T. Siloni. 2012. "The Theta system: an introduction". In: Everaert, M., Marelj, M. and T. Siloni (eds.), Argument structure at the interface. Oxford: Oxford University Press. 1-19.
Gussmann, E. 2007. The phonology of Polish. Oxford: Oxford University Press.
Jabłońska, P. 2007. Radical decomposition and argument structure. Ph.D., University of Tromsø.
Laskowski, R. 1984a "Kategorie morfologiczne języka polskiego – charakterystyka funkcjonalna" [Morphological categories of the Polish language – functional characteristics]. In: Grzegorczykowa, R., Laskowski, R. and H. Wróbel (eds.), Morfologia [Morphology]. Warszawa: PWN. 121-163.
Laskowski, R. 1984b. "Predykatyw" [Predicative]. In: Grzegorczykowa, R., Laskowski, R. and H. Wróbel (eds.),Morfologia [Morphology]. Warszawa: PWN. 171-219.
Levin, B. 2006. English object alternations: A unified account. Ms.
Levin, B. and M. Rappaport Hovav. 2010. Lexicalized meaning and manner/result complementarity. Ms. Stanford University and The Hebrew University of Jerusalem.
Malicka-Kleparska, A. 2013a. "Decausatives in Polish: A non-reflexive analysis". To appear in: Studies in Polish Linguistics VII.
Malicka-Kleparska, A. 2013b. "Valency rearrangement phenomena and their limitations with Polish ‘clear’ verbs". To appear in: Roczniki humanistyczne 61.
Ozga, J. 1976. "Clitics in English and Polish, Poznań Studies in Contemporary Linguistics 4, 127-140.
Przepiórkowski, A., Bańko M., Górski. R. and B. Lewandowska-Tomaszczyk (eds.). 2012. Narodowy korpus języka polskiego [National Corpus of the Polish Language].Warszawa: PWN.
Pylkkänen, L. 2008. Introducing Arguments. Linguistic Inquiry Monograph 48. Cambridge, Mass.: The MIT Press.
Rappaport Hovav M. and B. Levin. 2010. "Reflections on manner/result complementarity". In: Rappaport Hovav, M., Doron, E. and I. Sichel (eds.), Syntax, Lexical Semantics, and Event Structure. Oxford: Oxford University Press: 21-38.
Reinhart ,T. and T. Siloni. 2005. "The Lexicon-Syntax parameter: reflexivization and other arity operations". Linguistic Inquiry 36 (3). 389-436.
Reinhart T. and T. Siloni. 2004. "Against an unaccusative analysis of reflexives". In: Alexiadou, A., Anagnostopoulou, E.and M. Everaert (eds.), The unaccusativity puzzle: explorations of the syntax-lexicon interface. Oxford: Oxford University Press. 159-180.
Rościńska-Frankowska, M. 2012. On the derivation of causative and inchoative verb forms in Polish. Generative Linguistics in Wrocław No.1. Wrocław: Center for General and Comparative Linguistics.
Segal, Z. and I. Landau. 2012."Clear from and Clear of: The Asymmetric syntax of detaching". Linguistic Review 29: 223-278.
1 Alexiadou and Doron (2012) support their analysis of languages of the world with the data taken predominantly from Modern Greek, Hebrew and English – the languages which, according to them, are representative of existing morpho-syntactic systems involving the relevant data.
2 We use the term ‘non-active’ after Alexiadou and Doron (2012) as a common term for anticausative, middle, dispositional middle, reflexive, reciprocal and medio-passive verbal constructions.
3 The situation would be different in the derivationist approach –where all intransitives would be derived and thus no differences among them would be necessarily expected.expected.
4 All the remaining Greek examples are taken from Alexiadou and Doron (2012).
5 The example has been taken from Laskowski (1984a:139). The dispositional middle use in Polish in our opinion is slowly dying out as its acceptance by younger people is not unanimous.
6 For a more detailed description of Polish reflexive morpheme see e.g. Laskowski (1984 a:47, 139-141).
7 See Ozga (1976) for an interesting presentation of antipassives in Polish.
8 The terms anticausative and unaccusative are used in the sources we refer to indiscriminately as synonyms. We follow this practice, although we feel that the term anticausative sounds less accurate if used about the verbs which do not have transitive correspondents.
9 The root here equals the morpheme susz, while the remaining elements are connected with verb creating and inflectional morphology. Details of verbal stem formation in Polish are to be found in e.g. Laskowski (1984b:178-198). As the problem is very extensive, we cannot present it here in any reasonable detail.
10 Roots as such infrequently appear as predicates in Polish. More frequently they are accompanied in verbs by stem producing suffixes – see Laskowski (1984b). Consequently, whether the very roots would be externally or internally caused and if they would be caused as roots is altogether problematic. We assume for the sake of this discussion that specifying the causation for roots is legitimate. Similarly, the very notion of external and internal causation does not seem very precise. For instance we assume that the verb like wilt is based on a root with internal causation, still such utterances as: It wilted because of the lack of care are possible.
11 Alexiadou (2010:15) mentions an analogous situation in Salish languages, but she repeats after Demirdache (2005) that marked and bare anticausatives cannot appear in the same contexts: they imply different causes. The Polish case seems to be different in this respect. We will show that the two classes appear with exactly the same type of causal modification.
12 Jabłońska (2007:154) insists that there is an essential difference in grammaticality between the acceptable appearance of morphologically unmarked unaccusatives with by itself and ungrammaticality of the marked ones. In Malicka-Kleparska (to appear) I show, with the help of Narodowy korpus języka polskiego (henceforth NKJP) [the National Corpus of the Polish Language] (included in the references as Przepiórkowski et al. 2012 ) that the distinction is not supported by the data. See also some examples in (12, 13).
13 There are descriptions of the polish unaccusative system which suggest that there exists a semantic (or argument structure00 difference between morphologically marked and bare anticausatives. In particular an extensive analysis to this effect has been offered by Rościńska-Frankowska (2012). The research of the corpus material, however, gainsays her findings.
14 Alexiadou and Doron (2012:3) explain this behavior in the following way: " In cases where the root does not require an external argument, the active voice describes an event without an external cause/agent. Accordingly, there is no reason for merging either middle or passive morphology to achieve the reduction of the external argument. For economy reasons, merging such morphology is nevertheless only appropriate if the active is less informative, i.e. if the described event has an external argument after all. This argument must be interpreted as agent, which is default theta role assigned when the root is not the element which selects the external argument".
15 Here the participial form is also prefixed, and not only accompanied by the reflexive clitic. No unprefixed appropriate form exists. The relationship of prefixation and anticausative formation is a problem that requires separate studies going far beyond the limits of this work.
16 With respect to Alexiadou and Doron’s (2012) particular analysis, some other doubtful points have to be mentioned. For instance when they consider the interrelationships between medio-passive and passive, they claim that in the cases where the two forms could be realized by the same morphological sequence, medio-passive is blocked or ousted by passive on the lexeme by lexeme basis. I do not see how this could be attained in the system that creates the two competing structures quite separately and what the status of such a limitation would be.
17 Bold characters mark the sensitive segmental material.
18 Occasionally we may observe some semantic differences between the doublets. For instance in this particular case suszyć się may stress more the process, while schnąć – the effect. Thus we may say: Buty suszyły się na deszczu ‘The shoes were drying in the rain’, if somebody forgot to take them inside, while we would not say: Buty schły na deszczu. However, as far as we can judge, these differences are random, since for instance no similar distinction can be observed in the case of the previous pair: czerwienieć and czerwienić się.
19 See Laskowski (1984b)
20 Roots are marked in bold characters.
21 The differences which are observable in root representations constitute phonologically conditioned alternations, see e.g. Gussmann (2007) for an overview of Polish phonology.
22 Prefixes that accompany the doublets differ in many cases, but the essential feature is the same ultimate root and similar meaning, only slightly modified by prefixation in the cases we have taken into consideration. In this particular case wyschnąć means ‘dry up’, while rozeschnąć się – ‘dry up so as to produce cracks (especially in wood)’.
23 The form postarzeć may be obsolescent but the corpus still gives examples with its use, e.g.: Niegdyś młoda, żywa dziewczyna, przy niedołężnym mężu przedwcześnie postarzała. ‘At one time young lively girl, at the side of her senile husband she grew old before her time’.
24 The examples are marked as obsolete according to Słownik Języka Polskiego [Dictionary of the Polish Language], Doroszewski (1963).
25 These similarities cannot be attributed to similarities in roots as it is stem forming affixes which evoke the relevant phonological alternations.
26 Remember, however, that not all reflexively marked verbs in Polish have non-active semantics anyway.
27 Although remember that the possibility of having headless unaccusatives is admitted by Alexiado and Doron (2012) – see the comment under (8). If so – their unaccusative semantics would have to rely on the roots themselves and those would be available in a derivational model as well as elements of underived verbs.
28 For more examples see Malicka-Kleparska (2013 b).
29 The verbs stand for actions connected with removing objects or substances from various locations in various manners.
30 The examples below, taken from Malicka-kleparska (to appear b) show that the meanings of alternating and non-alternating verbs are so close that it seems hardly viable to maintain that on the basis of their differences we may predict which verbs can alternate and which cannot:
Żołnierze konfiskowali (‘confiscated’) mienie ze szkoły (COL-only ) vs. Żołnierze grabili (‘rob’) szkołę z mienia/mienie ze szkoły (alternating COS/COL)
Joanna zbiera (‘gather, wipe out’) wodę ze stołu (COL-only) vs. Joanna wyciera (‘wipe out) wodę ze stołu/stół z wody (alternating COL/COS)
Joanna usunęła (‘remove’) brud z koszuli (COL-only) vs. Joanna spłukała (‘wash out’) koszulę z brudu/brud z koszuli (alternating COS/COL)
Złodziej okradł (‘steal’) dom z pieniędzy (COS-only) vs. Żołnierze grabili (‘rob’) szkołę z mienia/mienie ze szkoły (alternating COS/COL)
Policjant otrzeźwił (‘sober’) mężczyznę z upojenia (COS-only) vs. Kłusownik oswobodził (‘free’) padlinę z łyków/łyki z padliny ‘ The poacher freed the carcas from the trap /the trap from the carcas’ (alternating COL/COS)
Wyrzucać odpadki z kubła (COL-only) vs. Opróżniać kubeł z odpadków (COS-only)‘To throw out the shavings from the bucket’’To empty the bucket of the shavings’
31 They claim that transitives ant anticausatives of this kind differ in the sphere of entailments that they have. Certain transitives when used in the so-called truncated frame do not entail substance, while anticausatives corresponding to them do.
32 We present examples where arguments are feminine nouns as in the case of neuter or masculine ones their Nominative and Accusative cases are syncretic. It is important because the Nominative case of the NP distinguishes dispositional middles from impersonal constructions from transitive verbs, which will not be considered here. To avoid this complication the non-syncretic forms have been chosen.
33 The examples in this footnote illustrate the frames of the verbs mentioned in the text: Odskrobał szybę z lodu’ He scratched the pane from ice’, Odskrobał lód z szyby ‘He scratched the ice from the pane’; Oddłubał farbę z deski ‘He peeled the paint from the plank’, Odrzucił gałąź ze stosu ‘He threw away a branch from the pile’; Ogołocił mieszkanie z ozdób ‘He emptied the flat from ornaments’, Okradł mieszkanie z mebli ‘He robbed the flat of furniture’, Ograbił mieszkanie z kosztowności ‘He robbed the flat of valuables’; Oskubał kurę z pierza ‘He plucked the hen of its feather’, Oskubał pierze z kury ‘He plucked furthers from the hen’.