Katharine Graf Estes

The link between statistical segmentation and word learning in adults

Many studies have shown that listeners can segment words from running speech based on conditional probabilities of syllable transitions, suggesting that this statistical learning could be a foundational component of language learning. However, few studies have shown a direct link between statistical segmentation and word learning. We examined this possible link in adults by following a statistical segmentation exposure phase with an artificial lexicon learning phase. Participants were able to learn all novel object-label pairings, but pairings were learned faster when labels contained high probability (word-like) or non-occurring syllable transitions from the statistical segmentation phase than when they contained low probability (boundary-straddling) syllable transitions. This suggests that, for adults, labels inconsistent with expectations based on statistical learning are harder to learn than consistent or neutral labels. In contrast, a previous study found that infants learn consistent labels, but not inconsistent or neutral labels.

Learning about sounds contributes to learning about words: Effects of prosody and phonotactics on infant word learning

This research investigates how early learning about native language sound structure affects how infants associate sounds with meanings during word learning. Infants (19-month-olds) were presented with bisyllabic labels with high or low phonotactic probability (i.e., sequences of frequent or infrequent phonemes in English). The labels were produced with the predominant English trochaic (strong/weak) stress pattern or the less common iambic (weak/strong) pattern. Using the habituation-based Switch Task to test label learning, we found that infants readily learned high probability trochaic labels. However, they failed to learn low probability labels, regardless of stress, and failed to learn iambic labels, regardless of phonotactics. Thus, infants required support from both common phoneme sequences and a common stress pattern to map the labels to objects. These findings demonstrate that early word learning is shaped by prior knowledge of native language phonological regularities and provide support for the role of statistical learning in language acquisition.

Infants Generalize Representations of Statistically Segmented Words

The acoustic variation in language presents learners with a substantial challenge. To learn by tracking statistical regularities in speech, infants must recognize words across tokens that differ based on characteristics such as the speaker’s voice, affect, or the sentence context. Previous statistical learning studies have not investigated how these types of non-phonemic surface form variation affect learning. The present experiments used tasks tailored to two distinct developmental levels to investigate the robustness of statistical learning to variation. Experiment 1 examined statistical word segmentation in 11-month-olds and found that infants can recognize statistically segmented words across a change in the speaker’s voice from segmentation to testing. The direction of infants’ preferences suggests that recognizing words across a voice change is more difficult than recognizing them in a consistent voice. Experiment 2 tested whether 17-month-olds can generalize the output of statistical learning across variation to support word learning. The infants were successful in their generalization; they associated referents with statistically defined words despite a change in voice from segmentation to label learning. Infants’ learning patterns also indicate that they formed representations of across word syllable sequences during segmentation. Thus, low probability sequences can act as object labels in some conditions. The findings of these experiments suggest that the units that emerge during statistical learning are not perceptually constrained, but rather are robust to naturalistic acoustic variation.The acoustic variation in language presents learners with a substantial challenge. To learn by tracking statistical regularities in speech, infants must recognize words across tokens that differ based on characteristics such as the speaker’s voice, affect, or the sentence context. Previous statistical learning studies have not investigated how these types of surface form variation affect learning. The present experiments used tasks tailored to two distinct developmental levels to investigate the robustness of statistical learning to variation. Experiment 1 examined statistical word segmentation in 11-month-olds and found that infants can recognize statistically segmented words across a change in the speaker’s voice from segmentation to testing. The direction of infants’ preferences suggests that recognizing words across a voice change is more difficult than recognizing them in a consistent voice. Experiment 2 tested whether 17-month-olds can generalize the output of statistical learning across variation to support word learning. The infants were successful in their generalization; they associated referents with statistically defined words despite a change in voice from segmentation to label learning. Infants’ learning patterns also indicate that they formed representations of across-word syllable sequences during segmentation. Thus, low probability sequences can act as object labels in some conditions. The findings of these experiments suggest that the units that emerge during statistical learning are not perceptually constrained, but rather are robust to naturalistic acoustic variation.

Learning builds on learning: infants' use of native language sound patterns to learn words.

The current research investigated how infants apply prior knowledge of environmental regularities to support new learning. The experiments tested whether infants could exploit experience with native language (English) phonotactic patterns to facilitate associating sounds with meanings during word learning. Infants (14-month-olds) heard fluent speech that contained cues for detecting target words; the target words were embedded in sequences that occur across word boundaries. A separate group heard the target words embedded without word boundary cues. Infants then participated in an object label learning task. With the opportunity to use native language patterns to segment the target words, infants subsequently learned the labels. Without this experience, infants failed. Novice word learners can take advantage of early learning about sounds to scaffold lexical development.

Domain general learning: Infants use social and non-social cues when learning object statistics.

Previous research has shown that infants can learn from social cues. But is a social cue more effective at directing learning than a non-social cue? This study investigated whether 9-month-old infants (N = 55) could learn a visual statistical regularity in the presence of a distracting visual sequence when attention was directed by either a social cue (a person) or a non-social cue (a rectangle). The results show that both social and non-social cues can guide infants’ attention to a visual shape sequence (and away from a distracting sequence). The social cue more effectively directed attention than the non-social cue during the familiarization phase, but the social cue did not result in significantly stronger learning than the non-social cue. The findings suggest that domain general attention mechanisms allow for the comparable learning seen in both conditions.

From Tracking Statistics to Learning words: Statistical Learning and Lexical Acquisition

By age 1, infants display remarkable sensitivity to the sound structure of their native language. Statistical learning, the process of detecting structure in the environment by tracking patterns in the input, is hypothesized to contribute to infants’ early learning about sound. The present paper explores how infants’ ability to track distributional information in the speech signal contributes to a fundamental aspect of language development, linking sounds with meanings in word learning. Previous research has demonstrated that infants detect several cues that mark where words begin and end in the fluent stream of speech (e.g., transitional probability, phonotactic regularities). Tracking such patterns may allow infants to isolate individual words, making them available to be associated with referents. Even very early in vocabulary development, statistical learning about which sound sequences are likely or unlikely to occur within words in the native language may also shape word learning. We propose that early experience with sound sequence regularities provides infants with a foundation for lexical acquisition.

Intersecting Constraints on Label Learning: Effects of Age, Label Properties, and Referential Context

ABSTRACT This research investigates the development of constraints in word learning. Previous experiments have shown that as infants gain more knowledge of native language structure, they become more selective about the forms that they accept as labels. However, the developmental pattern exhibited depends greatly on the way that infants are introduced to the labels and tested. In a series of experiments, we examined how providing referential context in the form of familiar objects and familiar object names affects how infants learn labels that they would otherwise reject, nonspeech sounds. We found evidence of the development of intersecting constraints: Younger infants (14-month-olds) were more open to learning nonspeech tone labels than older infants (19-month-olds), and younger infants were more open to the influence of referential context. These findings suggest that infants form expectations about labels and labeling contexts as they become more sophisticated learners.

Flexibility in Statistical Word Segmentation: Finding Words in Foreign Speech

The present experiments investigated the flexibility of statistical word segmentation. There is ample evidence that infants can use statistical cues (e.g., syllable transitional probabilities) to segment fluent speech. However, it is unclear how effectively infants track these patterns in unfamiliar phonological systems. We examined whether English-learning 14-month-olds could segment continuous speech in an artificial language consisting of Mandarin Chinese syllables. Syllable transitional probabilities provided the only reliable word boundary cues. In Experiment 1, we found that males segmented the language; they reliably discriminated words from the language versus across-word sequences. Females did not display reliable discrimination. In Experiment 2, when the test items were designed to have less overlap in their statistical structure, males and females displayed successful learning. Variation in direction of preference was revealing about infants’ processing in the tasks. Taken together, the findings indicate that infants’ statistical word segmentation abilities are sufficiently robust for processing speech streams containing foreign sounds.