Variables in the acceptability of transitive variants for internally caused verbs of change of state in Romanian †

This paper explores the acceptability of transitive variants for a number of seemingly internally caused verbs of change of state in Romanian, i.e. the verbs express change of state events and lack transitives. To this effect, we designed a linguistic questionnaire which tested the acceptability of transitive counterparts of these verbs, in which we varied the denotation of the subject argument, i.e. natural force vs. agent. The overall results obtained from the 33 participants in the survey revealed the influence of semantic factors on transitive acceptability. The respondents marginally accepted the transitives with natural force subjects, while they rejected the transitives with agent subjects. On the other hand, transitive acceptability varies with the age of the participants, the low scores assigned to the transitives put to the test, irrespective of the denotation of the subject argument, by the youngest group points towards the maintenance of the current valence of these verbs.


Introduction
Within a lexicalist approach to the syntax-semantics interface, Levin & Rappaport Hovav (1995) suggested the classification of verbs of change of state into externally caused verbs of change of state and internally caused verbs of change of state.Crosslinguistically, the two verb classes are defined by a series of semantic, syntactic and morphological properties.
Thus, externally caused verbs of change of state "imply the existence of an «external cause» with immediate control over bringing about the eventuality denoted by the verb" (Levin & Rappaport Hovav, 1995, p. 92), whereas internally caused verbs of change of state denote events where "some property inherent to the argument of the verb is «responsible» for bringing about the eventuality" (Levin & Rappaport Hovav, 1995, p. 91).Levin & Rappaport Hovav (1995, p. 94) put forth the following lexical semantic representation for externally caused verbs of change of state: (1) [[x DO-SOMETHING]

CAUSE [y BECOME STATE]]
As is apparent from (1), the lexical semantic representation of an externally caused verb of change of state is dyadic, comprising two subevents: a causing subevent [x DO-SOMETHING] and a change of state subevent [y BECOME STATE].Each subevent contains an argument variable whose semantic role is dictated by the position it occupies in the lexical semantic decomposition: x is part of the causing subevent and has the semantic role cause, whereas y belongs to the change of state subevent and is a patient.
An externally caused verb of change of state is derived from the dyadic lexical semantic representation in (1) through the lexical binding 1 of argument x in the mapping from the lexical semantic representation † I would like to thank the anonymous reviewers of this article for their discussions and comments.
to argument structure.The operation of lexical binding captures the idea that there is an external cause in the event expressed by the intransitive.For instance, in the case of an externally caused verb of change of state like English break "our knowledge of the world tells us that the eventuality could not have happened without an external cause" (Levin & Rappaport Hovav, 1995, p. 93).Importantly, intransitive break has a dyadic lexical semantic representation, while its argument structure is monadic.
Given their derivation from transitive bases2 , externally caused verbs of change of state show transitive and intransitive variants and take part in the causative alternation [illustrated in (2a,b)], a phenomenon exhibited by pairs of semantically related verbs with transitive and intransitive variants, in which the transitive supplies the cause for the change of state event expressed by the intransitive3 .
b.The window broke.
In languages with anticausative morphology4 (Haspelmath, 1993), externally caused verbs of change of state tend to be morphologically marked.
On the other hand, the lexical semantic representation of internally caused verbs of change of state is monadic, and contains only one event [y BECOME STATE], with the patient argument y, as represented in (3) below: Internally caused verbs of change of state are basic intransitives which usually lack transitive counterparts, irrespective of the denotation of the subject argument, agent [in (4b) and ( 5b)] or natural force [in (4c) and ( 5c)].Consequently, the verbs do not take part in the causative alternation.
b. *The rangers decayed the logs.c. *The bad weather decayed the logs.[Levin & Rappaport Hovav, 1995, p. 97, (34a,b,c)] In sum, as illustrated by the English data above, externally caused verbs of change of state express externally caused events and show transitive variants, while internally caused verbs of change of state express internally caused events and lack transitive variants.Since only externally caused verbs of change of state have transitive variants, the participation in the causative alternation is restricted to this class.
In the next section, we take a look at a number of Romanian verbs of change of state which lack transitive variants and seem to fall into the class of internally caused verbs of change of state.The investigation of the properties exhibited by these verbs, with special emphasis on the acceptability of transitive variants, might have a bearing not only on the valence of the verbs under study, but also on the relevance of postulating internally caused verbs of change of state.
For starter, the verbs seem to observe the semantic criterion ascribed to internally caused verbs of change of state as they all depict naturally occurring chemical or biological processes driven by properties of the subject entity.
While such morphological arguments challenge the basic intransitivity of a se cloci "go bad" and a se prinde (laptele) "(about milk) coagulate", in what follows, we investigate the tenability of the fixed valence (i.e. the unique intransitivity) of the Romanian verbs above.

Semantic variables in the acceptability of transitive variants for internally caused verbs of change of state
The methodology used involved gathering data on acceptability judgments (traditionally called grammaticality judgments, see Schütze & Sprouse, 2013) for transitive versions of the verbs listed above, in which we varied the denotation of the entity found in subject position, i.e. natural force vs. agent.Hence, we focused on the grammatical well-formedness of sentences that depends on semantic factors related to the denotation of the entities in subject position.
33 Romanian speakers, aged 20-50, participated in the survey.They agreed to fill in a linguistic questionnaire in which they had to rate 18 sentences on a scale from 1 to 5 (a five-point Likert scale), where 1 stands for "Completely unacceptable", 2 stands for "Almost unusable", 3 means "Debatable", 4 means "Almost perfect", 5 signifies "Perfect"6 .These gradient acceptability judgments are compatible with a gradient model of grammar (Schütze & Sprouse, 2013), which advocates several degrees of grammatical acceptability, as opposed to the coarse-grained traditional model which assumed only two options: "ungrammatical' and "grammatical".Thus, the participants were asked to judge sentences according to their own grammar on a scale that ranges from the "Completely unacceptable" endpoint to the "Perfect" endpoint.
We calculated the mean acceptability rating for each sentence, which is the weighted mean, and we also chose to include the minimum and the maximum score received by each sentence.Since the sentences may receive a varied range of scores, we considered that it would be useful to include the mode, i.e. the most frequent value assigned to each sentence.
The transitives with natural force subjects received the whole range of scores (sentences [17] were assigned scores ranging from the minimum value 1 up to the maximum value 5).Only some of the transitives with agent subjects were assigned the whole range of ratings (sentences [10], [16], [18] received scores from 1 to 5), while the maximum rating of most of the others was lower than 5 (sentences [2] and [4] received maximum 4; sentences [6], [8], [12] received maximum 3; sentence [14] received maximum 2).
As these figures show, the transitives with natural force subjects are higher on the acceptability scale than the transitives with agent subjects.Our findings are in line with the assumptions made by Levin (2009), who argued that English internally caused verbs of change of state sometimes show transitives with natural force subjects, but not with agent subjects8 .Usually, agents cannot manipulate natural forces like heat, rain or humidity, therefore, agents cannot be direct causes of seemingly internally caused events, and cannot show up as subjects of these transitives.On the other hand, heat, rain or humidity can be interpreted as direct causes of such events as evinced by the judgments of the speakers consulted who marginally accepted the transitives with natural force subjects.
Note that Levin's (2009) argumentation accords with the generally accepted view in the literature according to which lexical causatives9 express direct causation (although see Neeleman & van de Koot, 2012, for a distinct opinion)10 .Thus, the subjects of lexical causatives are considered to be direct causes of the change event named by the verb.
The linguistic questionnaire that we employed gave the respondents the possibility to correct the sentences they found less acceptable and suggest alternatives.The corrections include the use of the intransitive versions of these verbs with causes introduced in adjunct phrases headed by din cauza, din pricina "because of ", datorită "due to", de la "from", de "of " or la "at".Some participants avoided using agents even in adjunct phrases headed by din cauza "because of " or datorită "due to", probably because they rule out altogether the status of agents as causes of such events.Other participants resorted to analytical/periphrastic causatives headed by a face "make" as alternatives to the lexical causatives put to the test.

Age-related variables in the acceptability of transitive variants for internally caused verbs of change of
state A variable that one needs to take into account when discussing verb valence is the age of the speakers consulted.Thus, we assume that high scores assigned to transitive variants by younger speakers would point towards a change in the valence of these verbs.
The 33 participants in our survey fall into three groups, according to their age, i.e. the 20-25 age group, the 26-35 age group, and the 36-50 age group.Each age group comprises 11 members.
The participants in the 26-35 and 36-50 age groups have university education, whereas those in the 20-25 age group either have completed their university studies or they are university students.Since from the point of view of education the group is rather homogenous, the level of education was not considered   [18] were assigned the mean score 2 (almost unusable), while the transitive in sentence [14] was assigned the mean score 1 (completely unacceptable).
Although a larger number of speakers need to be questioned before making generalizations, the rejection of the transitive variants of internally caused verbs of change of state, irrespective of the denotation of the subject argument (natural force or agent), by the youngest group may indicate a tendency to maintain the valence of the verbs put to the test.

Conclusions
Our survey has revealed that the acceptability of the transitive variants of internally caused verbs of change of state in Romanian depends on the denotation of the subject argument.Thus, the transitives with natural force subjects are placed in the middle on the acceptability scale (a five-point Likert scale), whereas the transitives with agent subjects are near the unacceptability end on the scale.
Nonetheless, the more categorical rejection of the transitive variants of these verbs by the youngest group of participants seems to point to the maintenance of the current valence of the verbs.
While there might be a semantic motivation for the basic intransitivity of the verbs that are not marked with reflexive morphology as the verbs express internally caused events in the world, the morphological marking on a se cloci "go bad" and a se prinde (laptele) "(about milk) coagulate" remains unexplained in the context in which these verbs do not show fully acceptable transitives.However, the verbs present transitive variants with different meanings (see dex), from which such specialized meanings probably developed.Although the intransitives obligatorily marked with reflexive morphology a se cloci "go bad" and a se prinde (laptele) "(about milk) coagulate" may be derived from the transitives a cloci "hatch" and a prinde"catch", the former developed specialized meanings for which the transitive variants are not fully acceptable.The verbs that are obligatorily marked with reflexive morphology and perhaps those with optional morpho-

Table 1 :
With a non-agent Descriptive statistics of the results of the test for the acceptability of transitive variants for internally caused verbs of change of state