Transitivity Alternations in English and Norwegian: Experimental Investigations

NORWEGIAN SENTENCE PROCESSING IN THE SPOTLIGHT It has been reported in the literature that people reading sentences correctly anticipate a `gap´ in the next set of words coming up when they hear a question word like who/hvem or what/hva at the beginning. This effect has been shown clearly for English as a slowdown of processing using a variety of different scientific measures. Not only do readers know there is going to be a `gap´, they also have a bias in English to interpret that gap in direct object position. We decided to test this in Norwegian using an eye-tracking experiment . We were particularly interested in whether it made a difference whether the first verb after the hva or hvem was transitive or intransitive. In other words, do readers use the information about whether the verb usually occurs with a direct object or not to help them decide where to ` expect´ the gap. In English, speakers are strongly biased to expect the gap after a transitive verb, according to much replicated experimental evidence. However, nobody had yet tested this effect in Norwegian. We were uncertain as to what to expect since Norwegian word order is freer than in English, and the position of the direct object (if any) in a sentence is less predictable. Norwegian also has slightly more prepositional structures than direct objects when compared to English, according to a corpus check. The initial finding of our experiment was that there was a general slowdown for sentences with a gap, but there was no particular position in the sentence where readers experienced that slowdown in Norwegian, in contrast to what has been reported for English. So far, the particular effects reported for English have been assumed to be part of a universal and automatic system of language parsing. We are intrigued by the possibility that the grammar of Norwegian is constraining the processor in ways that have not yet been noticed and reported in the psycholinguistic literature. Our tentative conclusion is that the particular language one speaks has a strong effect on the cognitive process of reading in ways that the scientific community has not so far anticipated. Our two main experiments to date have shown that English differs from Norwegian in two striking respects. The first concerns the interpretation of sentences like `The door opened' , which in English is a simple intransitive while in Norwegian is expressed as a transitive verb with a reflexive marker. Our psychological test matching sentences with viewed events showed that this surface difference in form did indeed correlate with a difference in the way those sentences conceptualised the described events. In a nutshell, the Norwegian sentence type seemed to require the door to somehow be responsible for its own fate, while in English the opening door was conceived of as purely externally caused. One of the main outcomes of the project as a whole is that we discovered that it is unwise to extrapolate from English to expectations of how processing is going to take place in Norwegian. Assumptions about commensurability in processing (although standard in the literature) turned out to be unwarranted, and this made our project more complex in implementation but also potentially more exciting. We think that the project shows that, although it is difficult to construct experiments minimally comparing two distinct languages, such work is urgently important if we are to understand linguistic processing more generally and not just rely on the results emerging from studies of English.

In this project, we use modern techniques of reaction time based psycholinguistic experimentation to compare the Lexica of Norwegian and English with respect to transitivity alternations. In English, many verbs freely alternate between transitive (causat ive) and intransitive (inchoative) uses with no mediating morphology. In Norwegian, these alternations are mostly found with morphologically related pairs, or no alternation is possible at all. Previous research has argued that lexical decision tasks are sensitive to the internal semantic complexity of the event structure representation of a verbal item (McKoon and Macfarland 2002,Gennari and Poeppel 2003). However, contributions to the latencies from ambiguity, and morphological cohorts has never be en fully controlled for. Many studies concentrate on frequency as the primary factor to be controlled for, but effects of ambiguity and contextual diversity (Baayen 2010) have been shown to have strong and non-trivial direct effects on response time late ncies. The two languages chosen here for comparison offer the possibility of comparing similar lexical semantic patterning within databases that have minimally different internal morphological patterning. In this way, we hope to isolate the effects of m orphological complexity and contextual diversity and establish whether indeed there is any evidence for some verbs having a more complex event structure than others. The findings are central to theories of the lexicon and of the causative-inchoative alte rnation in particular, where elaborate event structure templates have been claimed to reside in the lexicon, and where the prediction is that lexical complexity of representation should have an impact on psycholinguistic performance. So far, the experime ntal evidence for these theories has been patchy. We aim to provide the first exhaustive comparative study of this classic alternation that directly controls for morphological relatedness and contextual diversit