Mark my words: High frequency marker words impact early stages of language learning.

Abstract

High frequency words have been suggested to benefit both speech segmentation and grammatical categorization of the words around them. Despite utilizing similar information, these tasks are usually investigated separately in studies examining learning. We determined whether including high frequency words in continuous speech could support categorization when words are being segmented for the first time. We familiarized learners with continuous artificial speech comprising repetitions of target words, which were preceded by high-frequency marker words. Crucially, marker words distinguished targets into 2 distributionally defined categories. We measured learning with segmentation and categorization tests and compared performance against a control group that heard the artificial speech without these marker words (i.e., just the targets, with no cues for categorization). Participants segmented the target words from speech in both conditions, but critically when the marker words were present, they influenced acquisition of word-referent mappings in a subsequent transfer task, with participants demonstrating better early learning for mappings that were consistent (rather than inconsistent) with the distributional categories. We propose that high-frequency words may assist early grammatical categorization, while speech segmentation is still being learned.

Exploring the “anchor word” effect in infants: Segmentation and categorisation of speech with and without high frequency words

“…However, this meant that infants were faced with the initial challenge of finding these items in the speech stream, which may have substantially increased the difficulty of the task [106,107]. It is possible that greater benefits to segmentation would emerge once high frequency words have reached a certain threshold of familiarity (perhaps with prior exposure, or with a longer training stream), thereby facilitating the interplay between top-down and bottom-up processing suggested in prior research [33,34] (see [45] for similar arguments). In future studies, measuring infants' knowledge of the high frequency words separately (and perhaps relating this to their performance on the segmentation task) would give valuable insight into this possibility.…”

“…In a recent study with adults, Frost, Monaghan, and Christiansen [ 45 ] examined the possibility that the benefit of high frequency words may actually be twofold—with these items potentially assisting with the categorisation of new words, in addition to helping with their initial discovery [ 46 – 49 ]. This was in light of the substantial overlap observed between the high frequency words that were seen to assist segmentation in the PUDDLE model [ 43 ] and words that were found to cue grammatical categorisation in prior corpus analyses [ 2 ] which indicated that the same items could conceivably inform learning for both tasks at the same time.…”

“…Frost et al [ 45 ] examined this possibility by training adults on an artificial language comprising lower frequency bisyllabic target words and high frequency mono-syllabic marker words, which distinguished target words into two otherwise unidentifiable categories (with one marker word reliably preceding targets in each category). After exposure to a continuous speech stream, participants were tested on their ability to segment the speech into words, and form distributional categories based on marker-target word co-occurrence.…”

“…Performance was compared to that of a control condition who were trained on a language that comprised target words only. Speech segmentation scores were similar for both groups, but the marker words shaped adults’ formation of grammatical categories, suggesting that these high-frequency words may inform categorisation during the early stages of language acquisition–perhaps while learners are still discovering how to segment speech [ 45 ].…”

“…However, combining short high frequency words and longer, lower frequency words in speech means that children will be faced with the challenge of segmenting words of different lengths. This is a key feature of natural language, and while children can undoubtedly cope with this ‘in the wild’, examining how they do so has proven challenging in prior research, and whether it is possible under laboratory conditions for artificial language remains an open question (though see [ 45 ] and [ 50 ] for findings that suggest this is possible for adults). In fact, Johnson and Tyler [ 51 ] proposed that infants’ ability to segment speech was limited, such that words had to be of similar length in order for statistical segmentation to proceed.…”