Language Learning and Processing Lab
Prof. Inbal Arnon
Department of Psychology - Hebrew University of Jerusalem
The learnability consequences and sources of Zipfian distributions in language
While the world’s languages differ in many respects, they share certain commonalities: these can provide crucial insight on our shared cognition and how it impacts language structure. In this project, we explore the learnability sources and consequences of one of the most striking commonalities across languages: the way word frequencies are distributed. Across languages, words follow a Zipfian distribution, showing a power law relation between a words' frequency and its rank. Intuitively, this reflects the fact that languages have relatively few high frequency words and many low frequency ones, and that frequency does not decrease in a linear way. The source of this distribution has been heavily - debated, with ongoing controversy about what it can tell us about language. We propose that such distributions confer a learnability advantage, leading to enhanced language acquisition in children, and to the creation of a cognitive pressure to maintain similarly skewed distributions over time. We explore this proposal using an innovative combination of computational, mathematical, experimental, and corpus-based tools.
Starting Big
Starting Big: The role of multiword units in first and second language learning (ISF 527-12, BSF 2011107)
​
Why are children better language learners than adults despite being worse at a range of other cognitive tasks? This project offers a new perspective on this long-standing question by highlighting the differential role of multiword units in child and adult language learning. In a nutshell, we propose that children are better at certain aspects of language learning because they learn from larger and less-analyzed units: while children utilize multiword units (like I-don't-know) in the learning process, adults will tend to learn from individual words – a tendency that will hinder learning of certain grammatical relations (Arnon, 2010). As part of this project, we’ve shown that children rely on multiword information in processing (Arnon & Clark, 2011); that such units serve as building blocks in learning (Arnon, McCauley & Christiansen, 2017); and that they can facilitate learning of certain grammatical relations in adult learners (Arnon & Ramscar, 2012; Siegelman & Arnon, 2015).
​
Together with Stewart M. McCauley & Morten H. Christiansen, we recently published a paper in the Journal of Memory and Language documenting Age-of-Acquisition effects for multiword phrases.
The developmental trajectory
of statistical learning
The developmental trajectory of statistical learning (ISF funding 2016-2020)
Infants, children and adults are capable of extracting recurring patterns from their environment through statistical learning (SL), an implicit learning mechanism that is considered to have an important role in language acquisition. Research over the past twenty years has shown that SL is present from very early infancy and found in a variety of tasks and across modalities. While SL is well established for infants and adults, little is known about its developmental trajectory during childhood, leaving several important questions unanswered.
The proposed research aims to examine two major issues that have significant theoretical implications for our understanding of SL and that have been under-studied to date. The first is the nature of age-related changes in SL: does SL improve with age, like many other cognitive capacities, or is it an early-maturing capacity that does not change after infancy? The second is the nature of individual differences in SL: does SL vary between children, and if so, what are the sources and consequences of such variation for language learning?
​
New paper out on the reliability of statistical learning tasks among children (Arnon, 2019).
Studying the emergence of linguistic
structure in the lab
Experimental work in the field of language evolution suggests that cultural transmission can lead to the emergence of linguistic structure as speakers’ weak individual biases become amplified through iterated learning (Kirby et al. 2008). Interestingly, there is little evidence for similar patterns in children, despite such evidence being crucial for evaluating the role of learning biases in the emergence of linguistic structure since children are the most frequent learners in the actual process of cultural transmission and have been claimed to play a unique role in the emergence of structure. In this project, we use a novel child-friendly ILM paradigm (Raviv & Arnon, 2016) to explore the emergence of structure in child learners.