This work is currently under review for publication, so stay tuned!
Our experiment was a first attempt to test a direct, causal relation between prediction and learning. We find that prediction itself doesn’t explain how children learn novel words, but 3-year-olds and 4-year-olds who predict and redirect attention toward the novel referent were more successful in learning. That being said, learning 12 novel words in just a few minutes was clearly difficult for children, as indicated by their low accuracy at test. In sum, further experiments are needed to evaluate the role of prediction in learning, and to clarify what other factors (e.g., cognitive control) are involved during learning.
It was a great time, as always, despite the snow!
Looking forward to CUNY 2018!
Here is my research statement for my 2016 NSF graduate fellowship proposal.
Hopefully 2nd time’s the charm!
This year, I gave a talk on nonverbal prediction at BUCLD.
Below are an abstract and slides for reference. I’m working on this paper with Lauren Emberson, Alexa Romberg, and Casey Lew-Williams, so stay tuned for the publication.
Prediction may be a language learning mechanism. This idea is supported by research showing that children with larger vocabularies (MCDI, PPVT) generate verbal predictions while processing language, and flexibly update predictions in light of new information. But do predictions support language learning, or vice-versa? In the present study, we assessed nonverbal prediction and vocabulary (MCDI) in 12-24-month-old infants (n=50). Infants with larger vocabularies efficiently updated nonverbal predictions in light of new information. This work reveals that links between prediction and language learning extend beyond the verbal domain, and are apparent even in infancy. Given the vast potential for making inaccurate predictions, the ability to continuously update predictions may be crucial for learning. In ongoing experiments, I am exploring the complex relations between language input, nonverbal prediction, verbal prediction, and language learning.
Well, today I submitted my NSF proposal!
Here is my research statement.
Hopefully my work demonstrates both “intellectual merit” and “broader impacts.”
This year, I gave a talk on children’s comprehension of truth-functional negation at BUCLD.
The main point of my talk is that pragmatic support is important for processing the negative form, but other factors (e.g., processing demands) play a role as well. We found that 3-year-olds interpret negation incrementally with supportive discourse context. However, we found that 2-year-olds had difficulty switching flexibly between affirmative and negative statements (Study 2) and only successfully comprehended both forms in a blocked design with affirmative trials first (Study 3).
Here’s my poster from CUNY 2014 (Columbus, OH): CUNY 2014 Processing Negation
And the submitted abstract:
Young children’s comprehension of negation
Tracy Brookhyser, Roman Feiman, & Jesse Snedeker (Harvard University)
Negation is a crucial test case for understanding incremental semantic interpretation, because its compositional representation is inconsistent with expectations generated by lexical items (see e.g. Panizza, 2012). Early processing studies found negative sentences were initially interpreted like their affirmative counterparts, raising the possibility that early predictive processing is associative (Kaup, Lüdtke & Zwaan, 2006; Fischler, et al., 1983). However, subsequent studies with richer discourse context found adults can rapidly integrate negation into their sentence interpretation (Nieuwland & Kuperberg, 2008; Tian, Breheny & Ferguson, 2010). No such result exists with children, and recent evidence suggests children 3 and younger do not process negation incrementally (Nordmeyer & Frank, 2013). We used the visual world paradigm to examine online sentence comprehension. Children heard a story, with corresponding pictures, which established a discourse context for both the negative and affirmative sentences (e.g. DW plans to color two stars but is interrupted), then heard the critical sentence and response prompt (e.g. “DW didn’t color/colored one of the stars. Which one was it?”). We previously found adults show incremental interpretation of negation in these contexts. Adapting this method to children, we found older children (n = 16, M = 42 mos) and younger children (n = 28, M = 31 mos) performed differently (Fig 1). In our online measures, we analyzed the proportion of fixations to the affirmative target (e.g. the colored star) for the time between VP offset and NP onset. In older children, a mixed effects model found solely an effect of polarity (p = .01). In younger children, there was an interaction between vocabulary, measured by MCDI, and polarity (p = 0.01). There was a marginal effect of polarity among younger children with higher MCDI scores (p = .07). In offline measures, older children’s picture selection was above chance in affirmative (77%) and negative (86%) trials, but younger children performed at chance in affirmative (60%) and negative (52%) trials, despite good performance in unambiguous practice trials (81%). Overall, results suggest 2-year-olds have difficulty resolving competition from two similar referents, regardless of polarity, though those with higher vocabularies may do somewhat better. 3-year-olds do not appear to have particular difficulty interpreting negation. Critically, we find no evidence that initial interpretation is associative (e.g. early affirmative fixations in both conditions). In this context, negative utterances are interpreted as quickly and accurately as affirmatives, suggesting incremental compositional processes in both cases.