Environmental Influences on Language Acquisition and Processing
Colour knowledge permeates our lives and plays a central role in pervasive tasks like object recognition. How does this knowledge develop in children? To address this question it is important to consider which skills are at play: separation of the continuous colour space, attaching labels, and extracting colour information from visual input.
In this study, we focus on functional colour knowledge, i.e. how children and adults use their knowledge about entities to categorize them. In other words, we tap into higher level skills related to colour term acquisition, which involve not only correctly partitioning the colour space and assigning the appropriate label, but also sampling over many examples and accessing some stored common property. To test the developing ability to use colour knowledge in such a way, we asked children from the age of 3 and adults to categorize 16 animals by colour in a tablet study. A second task was to assign a colour to each group. Data collection took place during an outreach event, and a secondary online study aimed to increase our sample size to obtain data from over 100 participants evenly distributed between the age of 3 and 16 and a matched adult sample.
Colour knowledge permeates our lives and plays a central role in pervasive tasks like object recognition. How does this knowledge develop in children? To address this question it is important to consider which skills are at play: separation of the continuous colour space, attaching labels, and extracting colour information from visual input. In this study, we focus on functional colour knowledge, i.e. how children and adults use their knowledge about entities to categorize them. In other words, we tap into higher level skills related to colour term acquisition, which involve not only correctly partitioning the colour space and assigning the appropriate label, but also sampling over many...
Adult listeners can ‘listen through’ speech errors, interpreting the meaning of an utterance non-literally to infer the speaker’s presumed intent (e.g. Gibson et al, 2013; Levy, 2008). Novel words also continuously enter the lexicon and have to be distinguished from errors; in children, learning can happen with as little as one exposure (‘fast mapping’; Carey & Bartlett, 1978). The aim of this project is to show whether children and adults use similar cues to learn versus non-literally interpret novel words. We will build a computational model of the mechanisms behind non-literal inference and word learning, illuminating cue weighting changes through the lifespan.
In the appropriate contexts, infants as young as one year of age engage in non-literal inference and overcome mispronunciations (Von Holzen & Bergmann, 2018). This is modulated by the presence of a suitable target for a novel word (an unknown object) and the size of the mispronunciation (few features altered). Adults perform similarly, flexibly interpreting utterances if there is a likely known candidate especially when the mispronunciation is small (Gibson et al, 2013) and when the speaker has indicated uncertainty with pauses or disfluencies (Lowder & Ferreira, 2018). This suggests that the mechanisms used to perform non-literal inference could be similar across the lifespan. A computational framework to test this is surprisal (e.g. Gibson et al, 2013): the listener compares the utterance to known words and repairs it if there is a known candidate. Using Bayesian modeling of surprisal, child and adult performance can be equated, controlling for factors like processing speed and vocabulary size.
Children are rapid learners, acquiring new lexical items quickly in early childhood (Bloom, 1973). Learning continues through adulthood, with vocabulary size increasing until old age (e.g Verhaeghen, 2003). Yet, the learning goal changes: children need to learn an item’s first label (e.g. bug), while adults often need to learn a second label (e.g. bug and also hemipteran); as such, cue weights might differ through the lifespan. In infants, non-linguistic contextual cues modulate whether non-literal inference or learning is more likely (e.g. Pruden et al., 2006). In adults, pragmatic cues influence non-literal inferences in grammatical sentences (He likes to go to the park with his cat elicits many looks to a dog; Lowder & Ferreira, 2018). These may or may not be supported by similar mechanisms.
In our new paradigm, we test whether individuals perform non-literal inference or acquire novel items by tracking processing during listening (as in Lowder & Ferreira, 2018) and assessing how the learning rate changes with age.
Adult listeners can ‘listen through’ speech errors, interpreting the meaning of an utterance non-literally to infer the speaker’s presumed intent (e.g. Gibson et al, 2013; Levy, 2008). Novel words also continuously enter the lexicon and have to be distinguished from errors; in children, learning can happen with as little as one exposure (‘fast mapping’; Carey & Bartlett, 1978). The aim of this project is to show whether children and adults use similar cues to learn versus non-literally interpret novel words. We will build a computational model of the mechanisms behind non-literal inference and word learning, illuminating cue weighting changes through the lifespan. In the appropriate...
This archive contains the materials used and the data collected for a project to assess the relation between caregivers' beliefs on four domains (growth mindset; independence and individuality; rules and regularity; language specific), caregivers' dialogic reading behaviors, and their 19- to 21-month-old child's vocabulary size as measured with the N-CDI.
This archive contains the materials used and the data collected for a project to assess the relation between caregivers' beliefs on four domains (growth mindset; independence and individuality; rules and regularity; language specific), caregivers' dialogic reading behaviors, and their 19- to 21-month-old child's vocabulary size as measured with the N-CDI.