The present paper reports on an investigation into an English un-participle pattern that is called unpassive, or is described as an adjectival passive. The main characteristic of the pattern is an (adjectival) past participle prefixed by un-, which is used as a predicative complement to a verb. Besides the different terms used for the pattern, there is also some indeterminacy with respect to its particular form. All of the descriptions focus on the verb BE, but mention is also made of GO and REMAIN. That is, the specifications of the pattern’s formal side differ to some extent. To provide information on this issue and to get hold of potential (verb-related) differences in the pattern’s function, we conducted an empirical analysis from a usage-based construction grammar perspective. Our focus is on the form-function interplay of the pattern in order to gain information about its constructional status and its exact formal and semantic make-up. The database selected for this study is the BNC, from which all occurrences of ‘verb plus un-participle’ were extracted. The data were submitted to collexeme and covarying collexeme analyses to identify the spectrum of meanings/functions associated with these forms, and distinctive collexeme analyses were carried out to see whether the un-participles found pattern differently with the individual verbs. The results indicate that, on closer examination, the un-participle construction does not represent a homogeneous category, but must be seen as a schematic template of related, though different, usage events that may have expanded analogously from a prototype construction. On the basis of our analyses and informed by findings from developmental studies, we suggest that the related constructions form a network.

La présente thèse porte sur l’acquisition du passif en français chez des enfants de 3 à 6 ans. Notre premier objectif était de tester l’effet du potentiel adjectival des passifs sur la compréhension des passives actionnelles et non actionnelles. Selon Borer et Wexler (1987), Babyonyshev et al. (2001), Hirsch et Wexler (2006) et Oliva et Wexler (2018), les enfants ne parviendraient pas à comprendre facilement les passifs verbaux, car certains principes de leur grammaire n’auraient pas atteint une certaine maturation syntaxique qui leur autoriserait la computation de ces passifs. Ils pourraient néanmoins utiliser une stratégie adjectivale pour les comprendre en analysant ces passifs comme des passifs adjectivaux, dont la représentation syntaxique serait accessible à leur grammaire. Pour tester cette hypothèse, nous avons mené une première expérimentation auprès de 98 enfants, âgés de 3 à 6 ans, et de 16 adultes, qui ont réalisé une tâche d’appariement phrase-image. Les résultats révèlent que les passives actionnelles ayant un faible potentiel adjectival sont les mieux comprises par les enfants, ce qui va à l’encontre d’une hypothèse adjectivale. Pour expliquer ces résultats, nous avons avancé que l’affectivité avait un effet facilitant sur la compréhension du passif (Nguyen et Snyder, 2018). Les passives actionnelles présentant un degré plus élevé d’affectivité seraient mieux comprises par les enfants. Les résultats de cette expérimentation appuient également l’hypothèse de Paolazzi et al. (2015, 2016) et Grillo, Alexiadou, Gehrke et al. (2019). Comme ces passives ne revêtent pas d’ambiguïté adjectivale/verbale, leur traitement serait moins complexe que les passives actionnelles présentant un potentiel adjectival élevé et qui peuvent être analysées comme des passives adjectivales ou verbales en l’absence du complément en par. Notre deuxième objectif de recherche était de mesurer l’effet du complément en par sur la compréhension des passives actionnelles et non actionnelles. Dans une deuxième expérimentation, 96 enfants, âgés de 3 à 6 ans, et 16 adultes ont effectué une tâche d’appariement phrase-image. Les résultats indiquent que la présence du complément en par n’a aucun effet sur la compréhension des passives actionnelles. Nos deux expérimentations reproduisent aussi l’effet Maratsos (Maratsos et al., 1985) : un délai acquisitionnel est observé pour la compréhension des passives non actionnelles. Les résultats de la deuxième expérimentation montrent que les passives courtes non actionnelles sont mieux comprises que les passives longues non actionnelles. Ainsi, notre recherche en conclut que l’acquisition des passives ne se fait pas d’un seul coup, mais qu’elle est largement déterminée par la présence du complément en par. Ce résultat appuie l’hypothèse de Paolazzi et al. (2015, 2016) et Grillo, Alexiadou, Gehrke et al. (2019). Les passives courtes, qui, sans le complément en par, seraient interprétées comme des passives adjectivales, seraient réanalysées en passives verbales lorsque ce complément est ajouté. Cette réanalyse serait couteuse cognitivement, ce qui expliquerait la moins bonne compréhension des passives longues non actionnelles. Notre thèse propose un ordre d’acquisition des constructions passives. Les passives courtes actionnelles seraient acquises à 3 ans, et les longues, à 4 ans. Les passives courtes non actionnelles seraient maitrisées à 5 ans alors que les longues le seraient après l’âge de 6 ans. Somme toute, nos résultats mettent en évidence l’importance du type de verbe (actionnel/non actionnel), qui est plus déterminant dans la compréhension du passif que l’effet du potentiel adjectival et du complément en par. Nos résultats remettent également en cause l’hypothèse d’une stratégie adjectivale et révèlent l’importance du type de prédicat (évènementiel/statif) et de l’ambiguïté adjectivale/verbale dans la compréhension des phrases passives. / The present dissertation focuses on the acquisition of the French passive construction by young children aged 3 to 6 years old. Our first objective was to test the effect of passive's adjectival potential on the comprehension of actional and non actional passives. Following Borer & Wexler (1987), Babyonyshev et al. (2001), Hirsch & Wexler (2006), and Oliva & Wexler (2018), children are not able to comprehend verbal passives because some principles of their grammar have not reached a certain syntactic maturity, which would allow them to compute these passives. They could however use an adjectival strategy to understand these passives by analyzing them as adjectival passives. To test this hypothesis, we conducted a first experiment with 98 children, aged 3 to 6 years old, and 16 adults, who carried out a sentence-picture matching task. Results show that actional passives with a lower adjectival potential are the ones which are the best understood by children, which do not support an adjectival hypothesis. To explain these results, we have suggested that affectivity have a facility effect on the comprehension of passives (Nguyen & Snyder, 2018). Actional passives which have a higher degree of affectivity would be the ones which are the best understood by children. The results of this experiment also support the hypothesis of Paolazzi et al. (2015, 2016) and Grillo, Alexiadou, Gehrke et al. (2019). Because these passives do not take an adjectival/verbal ambiguity, their treatment would be less complex than actional passives which have a high adjectival potential, and can be analyzed as adjectival or verbal passives without the by-phrase. Our second research objective was to measure the effect of the by-phrase on the comprehension of actional and non actional passives. In a second experiment, 96 children aged 3 to 6 years old and 16 adults did a sentence-picture matching task. Results show that the presence of the by-phrase has no effect on the comprehension of actional passives. Our two experiments reproduce the Maratsos effect (Maratsos et al., 1985): an acquisitional delay is observed for the comprehension of non actional passives. The results of the second experiment showed that short non actional passives are better understood than long non actional passives. Therefore, our research concludes that the acquisition of the passive construction does not occur at once, but is largely determined by the presence of the by-phrase. This result supports the hypothesis put forward by Paolazzi et al. (2015, 2016) and Grillo, Alexiadou, Gehrke et al. (2019): Short passives which, without the by-phrase, would be interpreted like adjectival passives, should be reanalyzed as verbal passives when this complement is added. The reanalysis would be cognitively costly, which would explain the poor comprehension of long non actional passives. This dissertation suggests an order of acquisition of passive constructions. Short actional passives would be acquired at 3 years old and long actional passives, at 4 years old. Short non actional passives would be mastered at 5 years old while long non actional passives would be acquired after 6 years old. On the whole, our results highlight the importance of verb type (actional/non actional), which is more decisive in the comprehension of passives than the effect of adjectival potential and the by-phrase. In this way, our results challenge an adjectival strategy and reveal the importance of the predicate type (eventive/stative) and of the adjectival/verbal ambiguity on the comprehension of passive constructions.

This study investigates the adjectival passive, in accordance with Beedham's (2005, 1982) analysis of the passive as an aspect, with the caveat that telicity is an optimal, not sufficient, condition. The affinity of the adjectival passive with attributive participles and the existence of implicit agents in adjectival passives has divided opinion amongst linguists. The thesis deploys grammaticality judgment questionnaires surveying 1043 2nd participles and a corpus-based study investigating 1035 2nd participles. A subsective gradience (Aarts 2007, 2006, 2004) is modelled on five morpho-syntactic properties of 2nd participles: attributive function without modification, attributive function with modification, adjectival, verbal and prepositional passive, measuring formally the ability of 2nd participles to function like adjectives. The thesis consists of seven chapters. Chapter one introduces the research questions, adjectival passives and theoretical background. Chapter two reviews the aspect analysis, telicity, offers a qualification, and sets the theoretical approach. Chapter three is about the data and methodology. Chapter four discusses the affinity between adjectival passive and attributive participles. Chapter five discusses subsective gradience. Chapter six discusses the implications of the findings. Chapter seven gives a summary and conclusion. The empirical findings in our study provide further evidence in support of a subsective gradience in 2nd participles indicative of how ‘adjectival' a participle can be, on a continuum or gradient ranging from ‘verby' 2nd participles – relatively low compatibility with adjectival properties – to very adjectival 2nd participles. 2nd participles in this study are shown to have an inherent meaning of ‘action + state'. 2nd participles which form adjectival passives function attributively and form verbal passives. However, a 2nd participle functioning attributively does not entail that it will form an adjectival passive. There is evidence that attributive un- participles can host manner adverbials. It was also found that the interpretation of attributive participles goes beyond a simple passive/perfect dichotomy, and there are cases whereby a 2nd participle modifies an NP that is not an argument of the corresponding verb. This study makes a contribution to the wider analysis of the adjectival passive and provides further support for the similarity between adjectival and verbal passives.