Search

本研究はコーパスおよび文処理実験の2つの側面から，主語を示すカラの使用法を検討した。まず，『現代日本語書き言葉均衡コーパス中納言』に収録された新聞，広報紙および雑誌の3つのジャンルで「名詞＋カラ＋動詞」のパターンを検索した。カラが主語で使用されるのは，新聞コーパスの696件中１件，広報紙コーパスの1,138件中１件，雑誌コーパスの3,053件中13件のみであった。カラ格の主語が通常は使用されないことを示した。次に，主語を表すカラが，主語のガと同じような特性を持つかどうかを確認するために，文処理実験を行った。ガ格主語は動詞句内から移動したといわれるが，カラ格の主語は動詞句内に留まっているという仮説に基づき，時の副詞句を挿入して，ガ格主語の正順語順のASOVとSAOVの語順，カラ格主語の正順語順のASOVとかき混ぜ語順のSAOVの反応時間と正答率を測定した。その結果，反応時間および正答率ともに，ガ格主語文およびカラ格主語文にスクランブル効果はみられなかった。したがって，ガ格の主語と同様に，カラ格の主語も動詞句外に位置し，同様な主語性を持つということを実証した。, The present study investigated the subjective property of kara using corpus search and a sentence processing experiment. Based on three different genres of corpora, newspapers, public announcements, and magazines taken from the Balanced Corpus of Contemporary Written Japanese (BCCWJ), the “noun + kara + verb” pattern was searched and classified according to the 11 different usages of kara. Kara used as subjects were counted as once out of 696 in newspapers, once out of 1,138 in public announcements, and 13 times out of 3,053 in magazines. Based on the frequency patterns of the 11 kara usages, entropy and redundancy were calculated in each corpus. These indexes showed similar values across the three corpora, suggesting that the kara use for subjects is rare. In addition to the corpus search, the present study also employed an experiment on the processing of sentences containing kara-marked subjects. More specifically, it was investigated whether kara-marked subjects share the same features as ga-marked subjects. In modern Japanese ga-marked subjects are generated inside the VP domain and then moved to the specifier position of the IP as [IP NP-ga [VP t NP-o V]] (c.f., Kitagawa, 1986; Kuroda, 1988; Hasegawa, 1999). In contrast, Ueda (2003) and Inoue (2002) claim kara-marked subjects must stay in the inside the domain of νP as [IP [νP NP-kara [VP NP-o + V]]]. By inserting a time adverb (e.g., yesterday, today, tomorrow) before and after the subject NP, sentence stimuli were created in two canonical orders of [IP Adv NP-ga [VP NP-o V]] and [IP NP-ga Adv [VP NP-o V]] for the subject NP-ga, and in canonical order [IP Adv [νP NP-kara [VP NP-o + V]]] and scrambled order [IP NP-kara Adv [νP t [VP NP-o + V ]]] for the subject NP-kara. Accordingly, the processing of each sentence was compared to determine relating features of ga- and kara-marked subjects. A 2 (NP-ga and NP-kara) × 2 (word order) ANOVA for reaction times and accuracies indicated the significant main effect for NP type, but not for word order; additionally, no interactions were found. As with ga-marked subjects, the findings reveal that kara-marked subjects are moved to the specifier position of the IP. Consequently, kara-marked subjects share the same subjective property with ga-marked subjects.