Jump to: Language | Evolution | Acquisition | Plasticity
Project : Linguistic typology and universal cognitive biases
A “universal” bias among world languages for certain statistical patterns among words would suggest that languages that exhibit these tendencies are more learnable, since their grammars more closely match human statistical learning preferences.
We seek cross-linguistic evidence on how statistical learning shapes grammars, across a large set of contemporary human languages.
- Computer simulation
- Human simulations of Iterated Language Learning (ILL)
Project : Emergence of grammar in infancy
This project aims to address how humans utilize their general SL abilities in the acquisition of natural languages.
In particular, we seek support that SL in preverbal children is a precursor and predictor of later grammatical development.
Speakers’ linguistic backgrounds influence their sequential statistical learning in an artificial language characterized by conflicting forward-going and backward-going
transitional probabilities. English-speaking adults favor backward-going transitional probabilities,
consistent with the head-initial structure of English, while Korean-speaking adults favor forward-going transitional probabilities,
consistent with the head-final structure of Korean.
We further found that English-learning infants develop this directional bias by 13
months, indicating that statistical learning rapidly adapts to the predominant syntactic
structure of the native language. Such adaptation possibly facilitates subsequent
learning by highlighting statistical structures that are likely to be informative.
For example, we found that even before they learn to speak, infant adaptation to syntactic structure is surprisingly fast
(Thiessen & Onnis, submitted). We are investigating whether such adaptation facilitates subsequent learning by highlighting
linguistic structures that are likely to be informative.
- Statistical analyses of child-directed speech (corpora)
- Behavioral infant studies
- Computer simulations
- Erik Thiessen, Carnegie Mellon University
- Sook Whan Cho, Sogang University
Project: Cognitive training for Second Language Learning: Syntax
As children, our brains become exquisitely “attuned” to our first language. This creates a paradox of success! First language experience changes subsequent learning, and we may find it harder to learn a second language as adults. We found that adults’ linguistic backgrounds influence their sequential statistical learning abilities in an artificial language characterized by conflicting forward-going and backward-going transitional probabilities (Onnis & Thiessen, 2013).
English-speaking adults favored backward-going transitional probabilities, consistent with the head-initial structure of English, while Korean-speaking adults favored forward-going transitional probabilities, consistent with the head-final structure of Korean (Onnis & Thiessen, 2013).
Subsequent testing on adults revealed the possibility to retrain monolinguals towards parsing preferences that are not consistent with their native language, suggesting training interventions to improve second language learning. There is surprising adult plasticity in the brain.
This project focuses on:
- cognitive training of sequential learning abilities
- Pre-test, post-test designs
- Event-Related Potentials (ERP)
Project 2: Cognitive training for Second Language Learning: Speech
Learning a second language takes years of effort, practice, and instruction.
Perceiving and producing novel speech sounds that do not exist in the first language is particularly hard.
This project proposes an innovative training scheme based on distribution-based learning.
The phonemes that form minimal pairs (e.g., ‘light’ versus ‘right’) can be disambiguated by paying attention to their phonetic contexts, or phonotactics.
For example stloop is not viable in English, because stl is not a viable phoneme combination in English.
A speech perception pre-test/post-test intervention design is used, and Phonetic Training is achieved via a Vocabulary learning task over multiple ‘homework’ sessions embedded in a second language course.
We manipulate the presentation of target foreign sounds such as /l/ and /r/ with two types of variables: 1) Informativeness of phonotactics, by randomly assigning one group of learners to words that contain informative distributional cues to R and L contexts, and another group to words containing uninformative phonotactics;
and 2) Modality of presentation, with one group randomly assigned to the speech modality only (a standard listen & repeat task), and one group assigned to speech and written modalities (listen & repeat + read & write).
We predict significant perceptual learning in participants are trained with informative cues, lending support to distribution-based learning.
Furthermore, speech augmented with orthographic cues should result in significant perception improvements at the end of the intervention.
Finally, the speaking task will allow researchers to record and gauge any improvement in pronunciation in the various conditions, thus capturing relevant data not only on the perception of speech (listening) but also on its production (speaking).
The findings would point to more effective means of learning and teaching pronunciation in a second language, and may be extended to the treatment of speech disorders in native-language learners.
This project focuses on:
- Perceptual training of statistical learning abilities
- Pre-test, post-test designs
- Event-Related Potentials (ERPs)
Have A Project In Mind?
We will be pleased to hear from you! Please contact us here.