11.7: Statistical Learning

Last updated
Save as PDF

Page ID: 140569

Todd LaMarr

\( \newcommand{\vecs}[1]{\overset { \scriptstyle \rightharpoonup} {\mathbf{#1}} } \) \( \newcommand{\vecd}[1]{\overset{-\!-\!\rightharpoonup}{\vphantom{a}\smash {#1}}} \)\(\newcommand{\id}{\mathrm{id}}\) \( \newcommand{\Span}{\mathrm{span}}\) \( \newcommand{\kernel}{\mathrm{null}\,}\) \( \newcommand{\range}{\mathrm{range}\,}\) \( \newcommand{\RealPart}{\mathrm{Re}}\) \( \newcommand{\ImaginaryPart}{\mathrm{Im}}\) \( \newcommand{\Argument}{\mathrm{Arg}}\) \( \newcommand{\norm}[1]{\| #1 \|}\) \( \newcommand{\inner}[2]{\langle #1, #2 \rangle}\) \( \newcommand{\Span}{\mathrm{span}}\) \(\newcommand{\id}{\mathrm{id}}\) \( \newcommand{\Span}{\mathrm{span}}\) \( \newcommand{\kernel}{\mathrm{null}\,}\) \( \newcommand{\range}{\mathrm{range}\,}\) \( \newcommand{\RealPart}{\mathrm{Re}}\) \( \newcommand{\ImaginaryPart}{\mathrm{Im}}\) \( \newcommand{\Argument}{\mathrm{Arg}}\) \( \newcommand{\norm}[1]{\| #1 \|}\) \( \newcommand{\inner}[2]{\langle #1, #2 \rangle}\) \( \newcommand{\Span}{\mathrm{span}}\)\(\newcommand{\AA}{\unicode[.8,0]{x212B}}\)

What is Statistical Learning?

Before infants can successfully link appropriate sound sequences to their referents, they must determine the set of sounds within a speech stream that correspond to potential words. For example, infants must segment the auditory statement “that is such a cute doggie” into meaningful chunks to accurately link it to its referent (e.g., “doggie”). This ability to segment speech sounds into word-level units, termed “word segmentation,” is a critical part of the word learning process (Saffran & Kirkham, 2018). Successful word segmentation requires exposure to the patterns and probabilities of sound sequences, maintaining phonological working memory and the order of the sequence of phonemes within the stream of speech to track the transitional probability so that one can increasingly identify potential word boundaries. Infants, children, and adults are all skilled at statistical word segmentation, often referred to as “statistical learning skills” (Aslin & Newport, 2012; Aslin, 2014; Saffran & Kirkham, 2018). ^{^[1]}

Statistical learning is the implicit ability to track regularities in linguistic or other input (e.g., visual or motor) and learn from the distributional information (Saffran, 2001; Lany & Saffran, 2013). The foundational statistical learning experiments in 8-month-old infants demonstrated that young infants could segment speech into potential word units using transitional probabilities or co-occurring probability information between syllables (Saffran, Aslin & Newport, 1996; Aslin, Saffran & Newport, 1998). Researchers have argued that the ability to learn from the patterns of language a child is exposed to plays an important role in language learning (Saffran, 2001, 2003). These studies suggest that during their first year, before children begin to produce words, they start learning the patterns of the language they hear, tracking the sound combinations that correspond to potential words. Furthermore, statistical word segmentation is one of the important aspects of word learning and vocabulary acquisition in toddlers. Eighteen-month-olds’ ability to use statistical information derived from fluent speech to identify words within the stream of speech and then to map the words to meaning predicts vocabulary size at 24-months and vocabulary growth from 18 to 24 months (Ellis, Borovsky, Elman & Evans, 2021). ^{^[1]}

^{^[1]} Ellis et al., (2021). Toddlers’ ability to leverage statistical iInformation to support word learning. Frontiers in Psychology, 12, 641. CC by 4.0