Computational Modeling of the Segmentation of Sentence Stimuli From an Infant Word-Finding Study.

Computational models of infant word-finding typically operate over transcriptions of infant-directed speech corpora. It is now possible to test models of word segmentation on speech materials, rather than transcriptions of speech. We propose that such modeling efforts be conducted over the speech of the experimental stimuli used in studies measuring infants' capacity for learning from spoken sentences. Correspondence with infant outcomes in such experiments is an appropriate benchmark for models of infants. We demonstrate such an analysis by applying the DP-Parser model of Algayres and colleagues to auditory stimuli used in infant psycholinguistic experiments by Pelucchi and colleagues. The DP-Parser model takes speech as input, and creates multiple overlapping embeddings from each utterance. Prospective words are identified as clusters of similar embedded segments. This allows segmentation of each utterance into possible words, using a dynamic programming method that maximizes the frequency of constituent segments. We show that DP-Parse mimics American English learners' performance in extracting words from Italian sentences, favoring the segmentation of words with high syllabic transitional probability. This kind of computational analysis over actual stimuli from infant experiments may be helpful in tuning future models to match human performance.

Developmental change in English-learning children's interpretations of salient pitch contours in word learning.

To efficiently recognize words, children learning an intonational language like English should avoid interpreting pitch-contour variation as signaling lexical contrast, despite the relevance of pitch at other levels of structure. Thus far, the developmental time-course with which English-learning children rule out pitch as a contrastive feature has been incompletely characterized. Prior studies have tested diverse lexical contrasts and have not tested beyond 30 months. To specify the developmental trajectory over a broader age range, we extended a prior study (Quam & Swingley, 2010), in which 30-month-olds and adults disregarded pitch changes, but attended to vowel changes, in newly learned words. Using the same phonological contrasts, we tested 3- to 5-year-olds, 24-month-olds, and 18-month-olds. The older two groups were tested using the language-guided-looking method. The oldest group attended to vowels but not pitch. Surprisingly, 24-month-olds ignored not just pitch but sometimes vowels as well-conflicting with prior findings of phonological constraint at 24 months. The youngest group was tested using the Switch habituation method, half with additional phonetic variability in training. Eighteen-month-olds learned both pitch-contrasted and vowel-contrasted words, whether or not additional variability was present. Thus, native-language phonological constraint was not evidenced prior to 30 months (Quam & Swingley, 2010). We contextualize our findings within other recent work in this area.

Relating referential clarity and phonetic clarity in infant-directed speech.

Psycholinguistic research on children's early language environments has revealed many potential challenges for language acquisition. One is that in many cases, referents of linguistic expressions are hard to identify without prior knowledge of the language. Likewise, the speech signal itself varies substantially in clarity, with some productions being very clear, and others being phonetically reduced, even to the point of uninterpretability. In this study, we sought to better characterize the language-learning environment of American English-learning toddlers by testing how well phonetic clarity and referential clarity align in infant-directed speech. Using an existing Human Simulation Paradigm (HSP) corpus with referential transparency measurements and adding new measures of phonetic clarity, we found that the phonetic clarity of words' first mentions significantly predicted referential clarity (how easy it was to guess the intended referent from visual information alone) at that moment. Thus, when parents' speech was especially clear, the referential semantics were also clearer. This suggests that young children could use the phonetics of speech to identify globally valuable instances that support better referential hypotheses, by homing in on clearer instances and filtering out less-clear ones. Such multimodal "gems" offer special opportunities for early word learning. RESEARCH HIGHLIGHTS: In parent-infant interaction, parents' referential intentions are sometimes clear and sometimes unclear; likewise, parents' pronunciation is sometimes clear and sometimes quite difficult to understand. We find that clearer referential instances go along with clearer phonetic instances, more so than expected by chance. Thus, there are globally valuable instances ("gems") from which children could learn about words' pronunciations and words' meanings at the same time. Homing in on clear phonetic instances and filtering out less-clear ones would help children identify these multimodal "gems" during word learning.

Very Young Infants' Sensitivity to Consonant Mispronunciations in Word Recognition.

Before they start to talk, infants learn the form and meaning of many common words. In the present work, we investigated the nature of this word knowledge, testing the specificity of very young infants' (6-14 months) phonological representations in an internet-based language-guided-looking task using correct pronunciations and initial-consonant mispronunciations of common words. Across the current sample (n=78 out of 96 pre-registered), infants' proportion looking to the target (named image) versus the distracter was significantly lower when the target word was mispronounced, indicating sensitivity to phonological deviation. Performance patterns varied by age group. The youngest group (6-8 months, n=30) was at chance in both conditions, the middle group (9-11 months, n=21) showed significant recognition of correct pronunciations and a marginal mispronunciation effect, and the oldest age group (12-14 months, n=27) demonstrated the mature pattern: significant recognition and a significant mispronunciation effect. Ongoing work is completing the pre-registered sample size.

A Protracted Developmental Trajectory for English-Learning Children's Detection of Consonant Mispronunciations in Newly Learned Words.

Children are adept at learning their language's speech-sound categories, but just how these categories function in their developing lexicon has not been mapped out in detail. Here, we addressed whether, in a language-guided looking procedure, two-year-olds would respond to a mispronunciation of the voicing of the initial consonant of a newly learned word. First, to provide a baseline of mature native-speaker performance, adults were taught a new word under training conditions of low prosodic variability. In a second experiment, 24- and 30-month-olds were taught a new word under training conditions of high or low prosodic variability. Children and adults showed evidence of learning the taught word. Adults' target looking was reduced when the novel word was realized at test with a change in the voicing of the initial consonant, but children did not show any such decrement in target fixation. For both children and adults, most learners did not treat the phonologically distinct variant as a different word. Acoustic-phonetic variability during teaching did not have consistent effects. Thus, under conditions of intensive short-term training, 24- and 30-month-olds did not differentiate a newly learned word from a variant differing only in consonant voicing. High task complexity during training could explain why mispronunciation detection was weaker here than in some prior studies.

Consequences of phonological variation for algorithmic word segmentation.

Over the first year, infants begin to learn the words of their language. Previous work suggests that certain statistical regularities in speech could help infants segment the speech stream into words, thereby forming a proto-lexicon that could support learning of the eventual vocabulary. However, computational models of word segmentation have typically been tested using language input that is much less variable than actual speech is. We show that using actual, transcribed pronunciations rather than dictionary pronunciations of the same speech leads to worse segmentation performance across models. We also find that phonologically variable input poses serious problems for lexicon building, because even correctly segmented word forms exhibit a complex, many-to-many relationship with speakers' intended words. Many phonologically distinct word forms were actually the same intended word, and many identical transcriptions came from different intended words. The fact that previous models appear to have substantially overestimated the utility of simple statistical heuristics suggests a need to consider the formation of the lexicon in infancy differently.

A cross-linguistic examination of toddlers' interpretation of vowel duration.

Languages differ in their phonological use of vowel duration. For the child, learning how duration contributes to lexical contrast is complicated because segmental duration is implicated in many different linguistic distinctions. Using a language-guided looking task, we measured English and Dutch 21-month-olds' recognition of familiar words with normal or manipulated vowel durations. Dutch but not English learners were affected by duration changes, even though distributions of short and long vowels in both languages are similar, and English uses vowel duration as a cue to (for example) consonant coda voicing. Additionally, we found that word recognition in Dutch toddlers was affected by shortening but not lengthening of vowels, matching an asymmetry also found in Dutch adults. Considering the subtlety of the crosslinguistic difference in the input, and the complexity of duration as a phonetic feature, our results suggest a strong capacity for phonetic analysis in children before their second birthday.

Learning phonology from surface distributions, considering Dutch and English vowel duration.

In learning language, children must discover how to interpret the linguistic significance of phonetic variation. On some accounts, receptive phonology is grounded in perceptual learning of phonetic categories from phonetic distributions drawn over the infant's sample of speech. On other accounts, receptive phonology is instead based on phonetic generalizations over the words in the lexicon. Tests of these hypotheses have been rare and indirect, usually making use of idealized estimates of phonetic variation. Here we evaluated these hypotheses, using as our test case English and Dutch toddlers' different interpretation of the lexical significance of vowel duration. Analysis of thousands of vowels of one Dutch and three English mothers' speech suggests that children's language-specific differences in interpretation of vowel duration are likely due to detection of lexically specific patterns, rather than bimodality in raw phonetic distributions.

Lexical Learning May Contribute to Phonetic Learning in Infants: A Corpus Analysis of Maternal Spanish.

In their first year, infants begin to learn the speech sounds of their language. This process is typically modeled as an unsupervised clustering problem in which phonetically similar speech-sound tokens are grouped into phonetic categories by infants using their domain-general inference abilities. We argue here that maternal speech is too phonetically variable for this account to be plausible, and we provide phonetic evidence from Spanish showing that infant-directed Spanish vowels are more readily clustered over word types than over vowel tokens. The results suggest that infants' early adaptation to native-language phonetics depends on their word-form lexicon, implicating a much wider range of potential sources of influence on infants' developmental trajectories in language learning.

Quantitative Linguistic Predictors of Infants' Learning of Specific English Words.

To evaluate which features of spoken language aid infant word learning, a corpus of infant-directed speech (M. R. Brent & J. M. Siskind, 2001) was characterized on several linguistic dimensions and statistically related to the infants' vocabulary outcomes word by word. Comprehension (at 12 and 15 months) and production (15 months) were predicted by frequency, frequency of occurrence in one-word utterances, concreteness, utterance length, and typical duration. These features have been proposed to influence learning before, but here their relative contributions were measured. Mothers' data predicted learning in their own children better than in other children; thus, vocabulary is measurably aligned within families. These analyses provide a quantitative basis for claims concerning the relevance of several properties of maternal English speech in facilitating early word learning.

Young Infants' Word Comprehension Given An Unfamiliar Talker or Altered Pronunciations.

To understand spoken words, listeners must appropriately interpret co-occurring talker characteristics and speech sound content. This ability was tested in 6- to 14-months-olds by measuring their looking to named food and body part images. In the new talker condition (n = 90), pictures were named by an unfamiliar voice; in the mispronunciation condition (n = 98), infants' mothers "mispronounced" the words (e.g., nazz for nose). Six- to 7-month-olds fixated target images above chance across conditions, understanding novel talkers, and mothers' phonologically deviant speech equally. Eleven- to 14-months-olds also understood new talkers, but performed poorly with mispronounced speech, indicating sensitivity to phonological deviation. Between these ages, performance was mixed. These findings highlight the changing roles of acoustic and phonetic variability in early word comprehension, as infants learn which variations alter meaning.

Prosodic exaggeration within infant-directed speech: Consequences for vowel learnability.

Perceptual experiments with infants show that they adapt their perception of speech sounds toward the categories of the native language. How do infants learn these categories? For the most part, acoustic analyses of natural infant-directed speech have suggested that phonetic categories are not presented to learners as separable clusters of sounds in acoustic space. As a step toward explaining how infants begin to solve this problem, the current study proposes that the exaggerated prosody characteristic of infant-directed speech may highlight for infants certain speech-sound tokens that collectively form more readily identifiable categories. A database is presented, containing vowel measurements in a large sample of natural American English infant-directed speech. Analyses of the vowel space show that prosodic exaggeration in infant-directed speech has the potential to support distributional vowel learning by providing the learner with a subset of "high-quality" tokens that infants might attend to preferentially. Categorization models trained on prosodically exaggerated tokens outperformed models that were trained on tokens that were not exaggerated. Though focusing on more prominent, exaggerated tokens does not provide a solution to the categorization problem, it would make it easier to solve.

The infant's developmental path in phonological acquisition.

Research on how language acquisition begins has been fragmented both in terms of scientific communities and in terms of the phenomena that are taken to characterize developmental progress. In her article, Marilyn Vihman argues for an integrative approach that takes the child's efforts at speech production as primary, and notes that infants' knowledge of how words sound may accrue over a protracted period developmentally. Here, I briefly discuss how reconceptualization of the process can help integrate perspectives previously at odds.

Two-year-olds interpret novel phonological neighbors as familiar words.

When children hear a novel word in a context presenting a novel object and a familiar one, they usually assume that the novel word refers to the novel object. In a series of experiments, we tested whether this behavior would be found when 2-year-olds interpreted novel words that differed phonologically from familiar words in only 1 sound, either a vowel or consonant. Under these conditions children almost always chose the familiar object, though examination of eye movements showed that children did detect the tested phonological distinctions. Thus, children discounted perceptible phonological variations when doing so permitted a resolution of the speaker's meaning without postulating a new word. Children with larger vocabularies made novel-word interpretations more often than children with smaller vocabularies did. The results suggest that although young children do interpret speech in terms of a learned phonological system, this does not mean that children assume that phonological distinctions imply lexical distinctions. (PsycINFO Database Record

Early Word Comprehension in Infants: Replication and Extension.

A handful of recent experimental reports have shown that infants of 6 to 9 months know the meanings of some common words. Here, we replicate and extend these findings. With a new set of items, we show that when young infants (age 6-16 months, n=49) are presented with side-by-side video clips depicting various common early words, and one clip is named in a sentence, they look at the named video at above-chance rates. We demonstrate anew that infants understand common words by 6-9 months, and that performance increases substantially around 14 months. The results imply that 6-9 month olds' failure to understand words not referring to objects (verbs, adjectives, performatives) in a similar prior study is not attributable to the use of dynamic video depictions. Thus, 6-9 month olds' experience of spoken language includes some understanding of common words for concrete objects, but relatively impoverished comprehension of other words.

Learning novel phonological neighbors: Syntactic category matters.

Novel words (like tog) that sound like well-known words (dog) are hard for toddlers to learn, even though children can hear the difference between them (Swingley & Aslin, 2002, 2007). One possibility is that phonological competition alone is the problem. Another is that a broader set of probabilistic considerations is responsible: toddlers may resist considering tog as a novel object label because its neighbor dog is also an object. In three experiments, French 18-month-olds were taught novel words whose word forms were phonologically similar to familiar nouns (noun-neighbors), to familiar verbs (verb-neighbors) or to nothing (no-neighbors). Toddlers successfully learned the no-neighbors and verb-neighbors but failed to learn the noun-neighbors, although both novel neighbors had a familiar phonological neighbor in the toddlers' lexicon. We conclude that when creating a novel lexical entry, toddlers' evaluation of similarity in the lexicon is multidimensional, incorporating both phonological and semantic or syntactic features.

Processing of lexical stress cues by young children.

Although infants learn an impressive amount about their native-language phonological system by the end of the first year of life, after the first year children still have much to learn about how acoustic dimensions cue linguistic categories in fluent speech. The current study investigated what children have learned about how the acoustic dimension of pitch indicates the location of the stressed syllable in familiar words. Preschoolers (2.5- to 5-year-olds) and adults were tested on their ability to use lexical-stress cues to identify familiar words. Both age groups saw pictures of a bunny and a banana and heard versions of "bunny" and "banana" in which stress either was indicated normally with convergent cues (pitch, duration, amplitude, and vowel quality) or was manipulated such that only pitch differentiated the words' initial syllables. Adults (n=48) used both the convergent cues and the isolated pitch cue to identify the target words as they unfolded. Children (n=206) used the convergent stress cues but not pitch alone in identifying words. We discuss potential reasons for children's difficulty in exploiting isolated pitch cues to stress despite children's early sensitivity to pitch in language. These findings contribute to a view in which phonological development progresses toward the adult state well past infancy.

Young toddlers' word comprehension is flexible and efficient.

Much of what is known about word recognition in toddlers comes from eyetracking studies. Here we show that the speed and facility with which children recognize words, as revealed in such studies, cannot be attributed to a task-specific, closed-set strategy; rather, children's gaze to referents of spoken nouns reflects successful search of the lexicon. Toddlers' spoken word comprehension was examined in the context of pictures that had two possible names (such as a cup of juice which could be called "cup" or "juice") and pictures that had only one likely name for toddlers (such as "apple"), using a visual world eye-tracking task and a picture-labeling task (n = 77, mean age, 21 months). Toddlers were just as fast and accurate in fixating named pictures with two likely names as pictures with one. If toddlers do name pictures to themselves, the name provides no apparent benefit in word recognition, because there is no cost to understanding an alternative lexical construal of the picture. In toddlers, as in adults, spoken words rapidly evoke their referents.

The acquisition of abstract words by young infants.

Young infants' learning of words for abstract concepts like 'all gone' and 'eat,' in contrast to their learning of more concrete words like 'apple' and 'shoe,' may follow a relatively protracted developmental course. We examined whether infants know such abstract words. Parents named one of two events shown in side-by-side videos while their 6-16-month-old infants (n=98) watched. On average, infants successfully looked at the named video by 10 months, but not earlier, and infants' looking at the named referent increased robustly at around 14 months. Six-month-olds already understand concrete words in this task (Bergelson & Swingley, 2012). A video-corpus analysis of unscripted mother-infant interaction showed that mothers used the tested abstract words less often in the presence of their referent events than they used concrete words in the presence of their referent objects. We suggest that referential uncertainty in abstract words' teaching conditions may explain the later acquisition of abstract than concrete words, and we discuss the possible role of changes in social-cognitive abilities over the 6-14 month period.

Self-directed speech affects visual search performance.

People often talk to themselves, yet very little is known about the functions of this self-directed speech. We explore effects of self-directed speech on visual processing by using a visual search task. According to the label feedback hypothesis (Lupyan, 2007a), verbal labels can change ongoing perceptual processing-for example, actually hearing "chair" compared to simply thinking about a chair can temporarily make the visual system a better "chair detector". Participants searched for common objects, while being sometimes asked to speak the target's name aloud. Speaking facilitated search, particularly when there was a strong association between the name and the visual target. As the discrepancy between the name and the target increased, speaking began to impair performance. Together, these results speak to the power of words to modulate ongoing visual processing.