The Middle voice construction in Romanian – a corpus based analysis This paper offers an analysis of the two middle constructions se and isi in Romanian, (1) and (2) respectively. In Romanian there is more than one way to express the middle voice construction and this can be considered a very significant language internal variation that needs to be studied. The claim that the two middle voice constructions cover the same semantic spectrum, as they were taken to be by previous linguists and traditional grammarians, is not acceptable. It will be shown that the two constructions differ in the semantic fields that they cover, in the grammatical associations and most important in the way speakers conceptualize them. By using a self built corpus of Romanian from two different genres, narrative and newspaper, one can conclude that there are main differences in the verbal frames that each middle construction applies to. Moreover, even in cases where these constructions semantically overlap, major differences in the lexical collocations, grammatical paradigms as well as in their frequency of usage, exist. It will be shown that the isi-middle construction occurs more frequently with speech verb, cognitive verbs, intentional perception verbs, and emotional verbs; whereas se-middle construction occurs most often in association with spontaneous event verbs and movement verbs including translational motion verbs. Moreover, it will be argued that, unlike shown in traditional grammars, the isi-middle construction is the one that appears more often with grooming verbs and not the se-middle construction. Finally, both genre specific differences and across genre tendencies evident from the data such as the fact that the se-construction is more frequently used than the isi -construction, across genres, are dealt with. By using corpora this paper finds a variety of usages of the middle voice constructions that where not analyzed before because of the lack of natural data and the solely usage of introspective data. I will argue that in the case of isi- middle construction there is an extension to other semantic domains such as external possession (3) and causative constructions (4), while it is not so for the se-middle construction. Hence, this paper makes two major contributions. First it takes an empirical and data based approach to language and so to the middle voice construction finding main semantic differences between the two middle constructions. Secondly it discusses major usages that were overlooked because of the previous non-empirical and based on introspection approaches to the notion of middle voice in Romanian. In previous literature no clear predictions are made as to when each of the functions occurs. This paper discusses implications of the data for our conception of Middle constructions from the point of view of meaning and form. In conclusion, in order to give a better account of the Middle voice in Romanian a usage based analysis has to be provided. Thus, a reconsideration of the functions and semantics of the Middle voice in Romanian is called for. (1) Dar cand ninsoarea [[se]] inteti si se transforma in viscol but when snow MM got stronger and MM transform in storm But when the snow got stronger… (2) El [[isi ]]aduse aminte de stiuca, pe care o chema ... He MM brought mind of the pike….. He remembered the pike (3) El isi cioplea fluierul. (External possession with benefactive meaning) He MM carve the flute. He carved his flute (for himself) (4) El [[isi]] supara nevasta. (Causative) He MM makes angry wife. He makes his wife angry. Corpora Number of words Romania Libera Corpus 138745 Fairy Tales 96,393 Raw SE 1094 1588 Semantic mapping in the narrative genre: Se-middle 3% 2% 1% 2% Percentage 1% 2% 0% spontaneous emotion 5% 31% 2% 9% 3% 10% 126.8236 60.70088 ISI 1101.151 446.2639 Isi-middle cognitive grooming intentional perception translational speech change in body posture activity reciprocal spontaneous activity emotion modal 28% 126 216 Frequency per word SE speech action non-translational motion 8% ISI cognitive perception 8% Frequency mental translational grooming,body care body action indirect middle existencial mental middle passive non-translational Avram, Mioara 1997. Gramatica Pentru Toti. Bucuresti: Humanitas Barber, E.J.W. 1975. Voice, beyond the Passive. BLS 1, 16.24 Biber, Douglas. 1999. Investigating Language Use through Corpus-Based Analyses of Association Patterns. In Usage based Models of Language. Barlow & Kemmer (eds.), 287-313. CLS 1 Publication Chomsky, Noam. 1965. Aspects of the theory of syntax. Cambridge: MIT Press. Ch. 1; pp. 3-62. Cienki, Alan. 1993. Experiencer, Possessors and Overlap Between Russian Dative and u+Genitive. Proceedings of the 19th Annual Meeting the Berkeley Linguistic Society, 76-89 Daniliuc, Laura & Radu. 2000. Descriptive Romanian Grammar, an outline. Lincom Europa. Kemmer, Suzanne. 1993. The Middle Voice. Amsterdam: John Benjamins Publishing Company. Kemmer, Suzanne. & Barlow, Michael. 1999. A Usage based Conception of Language. In Usage based Models of Language. Barlow & Kemmer (eds.), CLS 1 Publication Langacker, Ronald. 1993. Reference-point constructions. Cognitive Linguistics 4, 1: 1-38 Payne, Doris L. & Barshi, Emanuel. 1999. External Possession, Typological Studies in Language 39, Amsterdam: John Benjamins Publishing Company.