A small but significant effect: lexical prediction in a selfpaced reading study
A series of studies conducted in the early 2000s reported evidence suggesting that specific lexical items and some of their features (e.g., word form and grammatical gender) could be predicted during language comprehension (DeLong et al., 2005; Van Berkum et al., 2005). These studies mainly repor...
Na minha lista:
| Autor principal: | |
|---|---|
| Outros Autores: | |
| Formato: | Dissertação |
| Idioma: | pt_BR |
| Publicado em: |
Universidade Federal do Rio Grande do Norte
|
| Assuntos: | |
| Endereço do item: | https://repositorio.ufrn.br/handle/123456789/32611 |
| Tags: |
Adicionar Tag
Sem tags, seja o primeiro a adicionar uma tag!
|
| Resumo: | A series of studies conducted in the early 2000s reported evidence suggesting that specific
lexical items and some of their features (e.g., word form and grammatical gender) could be
predicted during language comprehension (DeLong et al., 2005; Van Berkum et al., 2005).
These studies mainly reported evidence from electroencephalography experiments, but the
work conducted in Dutch by Van Berkum and colleagues (2005) included a self-paced reading
experiment, in which the authors found larger reading times for sentences containing nouns that
were unlikely in context, but still plausible. Such effect was registered before the presentation
of the critical nouns themselves in the sentences, making it impossible to attribute the difference
in reading times to explanations other than prediction. The words that triggered the reported
finding in the Dutch study were adjectives, all of which preceded the critical noun they modified
and were inflected for gender to show agreement with it. Exploring the same morphosyntactic
property, we conducted a self-paced reading experiment with 339 participants who had
Brazilian Portuguese (BP) as a first language. Our objective was to investigate whether
grammatical gender cues can be used to make predictions during language comprehension, thus
generating effects such as the ones previously reported in the literature. To do so, we created
20 experimental items that comprised two sentences: in the first, a simple situation was
introduced (e.g., “The couple looked at the restaurant menu until they could make up their
minds”); in the second, the situation unfolded either in a way that was likely (e.g., “They then
called the waitress, who wrote down the long and detailed order on her pad”) or unlikely (e.g.,
“They then called the waitress, who wrote down the long and detailed note on her pad”). The
key difference is that the critical nouns in the second sentence (‘order’ vs. ‘note’) always had
different grammatical genders in BP, which were marked in the preceding adjectives (‘long’
and ‘detailed’). If gender cues do trigger prediction effects, larger reading times were to be
expected in the unlikely condition, both for the critical noun and the words that preceded it. Our
results indicated divergent patterns in analyses conducted with log-transformed and
untransformed data: statistically significant differences in reading times prior to the critical
noun were observed in the latter analyses, but not in the former. |
|---|