Computation of Conditional Probability Statistics by 8-Month-Old Infants

A recent report demonstrated that 8-month-olds can segment a continuous stream of speech syllables, containing no acoustic or prosodic cues to word boundaries, into wordlike units after only 2 min of listening experience (Saffran, Aslin, & Newport, 1996). Thus, a powerful learning mechanism capable of extracting statistical information from fluent speech is available early in development. The present study extends these results by documenting the particular type of statistical computation–transitional (conditional) probability–used by infants to solve this word-segmentation task. An artificial language corpus, consisting of a continuous stream of trisyllabic nonsense words, was presented to 8-month-olds for 3 min. A postfamiliarization test compared the infants' responses to words versus part-words (trisyllabic sequences spanning word boundaries). The corpus was constructed so that test words and part-words were matched in frequency, but differed in their transitional probabilities. Infants showed reliable discrimination of words from part-words, thereby demonstrating rapid segmentation of continuous speech into words on the basis of transitional probabilities of syllable pairs.     

Transitional probabilities and positional frequency phonotactics in a hierarchical model of speech segmentation

The present study explored the influence of a new metrics of phonotactics on adults’ use of transitional probabilities to segment artificial languages. We exposed French native adults to continuous streams of trisyllabic nonsense words. High-frequency words had either high or low congruence with French phonotactics, in the sense that their syllables had either high or low positional frequency in French trisyllabic words. At test, participants heard low-frequency words and part-words, which differed in their transitional probabilities (high for words, low for part-words) but were matched for frequency and phonotactic congruency. Participants’ preference for words over part-words was found only in the high-congruence languages. These results establish that subtle phonotactic manipulations can influence adults’ use of transitional probabilities to segment speech and unambiguously demonstrate that this prior knowledge interferes directly with segmentation processes, in addition to affecting subsequent lexical decisions. Implications for a hierarchical theory of segmentation cues are discussed.     

Chunking or not chunking? How do we find words in artificial language learning?

What is the nature of the representations acquired in implicit statistical learning? Recent results in the field of language learning have shown that adults and infants are able to find the words of an artificial language when exposed to a continuous auditory sequence consisting in a random ordering of these words. Such performance can only be based on processing the transitional probabilities between sequence elements. Two different kinds of mechanisms may account for these data: Participants may either parse the sequence into smaller chunks corresponding to the words of the artificial language, or they may become progressively sensitive to the actual values of the transitional probabilities between syllables. The two accounts are difficult to differentiate because they make similar predictions in comparable experimental settings. In this study, we present two experiments that aimed at contrasting these two theories. In these experiments, participants had to learn 2 sets of pseudo-linguistic regularities: Language 1 (L1) and Language 2 (L2) presented in the context of a serial reaction time task. L1 and L2 were either unrelated (none of the syllabic transitions of L1 were present in L2), or partly related (some of the intra-words transitions of L1 were used as inter-words transitions of L2). The two accounts make opposite predictions in these two settings. Our results indicate that the nature of the representations depends on the learning condition. When cues were presented to facilitate parsing of the sequence, participants learned the words of the artificial language. However, when no cues were provided, performance was strongly influenced by the employed transitional probabilities.     

Can infants map meaning to newly segmented words? Statistical segmentation and word learning

The present experiments investigated how the process of statistically segmenting words from fluent speech is linked to the process of mapping meanings to words. Seventeen-month-old infants first participated in a statistical word segmentation task, which was immediately followed by an object-label-learning task. Infants presented with labels that were words in the fluent speech used in the segmentation task were able to learn the object labels. However, infants presented with labels consisting of novel syllable sequences (nonwords; Experiment 1) or familiar sequences with low internal probabilities (part-words; Experiment 2) did not learn the labels. Thus, prior segmentation opportunities, but not mere frequency of exposure, facilitated infants' learning of object labels. This work provides the first demonstration that exposure to word forms in a statistical word segmentation task facilitates subsequent word learning.     

Linking sounds to meanings: Infant statistical learning in a natural language

The processes of infant word segmentation and infant word learning have largely been studied separately. However, the ease with which potential word forms are segmented from fluent speech seems likely to influence subsequent mappings between words and their referents. To explore this process, we tested the link between the statistical coherence of sequences presented in fluent speech and infants’ subsequent use of those sequences as labels for novel objects. Notably, the materials were drawn from a natural language unfamiliar to the infants (Italian). The results of three experiments suggest that there is a close relationship between the statistics of the speech stream and subsequent mapping of labels to referents. Mapping was facilitated when the labels contained high transitional probabilities in the forward and/or backward direction (Experiment 1). When no transitional probability information was available (Experiment 2), or when the internal transitional probabilities of the labels were low in both directions (Experiment 3), infants failed to link the labels to their referents. Word learning appears to be strongly influenced by infants’ prior experience with the distribution of sounds that make up words in natural languages.     

A computational model of word segmentation from continuous speech using transitional probabilities of atomic acoustic events

Word segmentation from continuous speech is a difficult task that is faced by human infants when they start to learn their native language. Several studies indicate that infants might use several different cues to solve this problem, including intonation, linguistic stress, and transitional probabilities between subsequent speech sounds. In this work, a computational model for word segmentation and learning of primitive lexical items from continuous speech is presented. The model does not utilize any a priori linguistic or phonemic knowledge such as phones, phonemes or articulatory gestures, but computes transitional probabilities between atomic acoustic events in order to detect recurring patterns in speech. Experiments with the model show that word segmentation is possible without any knowledge of linguistically relevant structures, and that the learned ungrounded word models show a relatively high selectivity towards specific words or frequently co-occurring combinations of short words.     

Lexical and sublexical units in speech perception

Saffran, Newport, and Aslin (1996a) found that human infants are sensitive to statistical regularities corresponding to lexical units when hearing an artificial spoken language. Two sorts of segmentation strategies have been proposed to account for this early word-segmentation ability: bracketing strategies, in which infants are assumed to insert boundaries into continuous speech, and clustering strategies, in which infants are assumed to group certain speech sequences together into units ( Swingley, 2005 ). In the present study, we test the predictions of two computational models instantiating each of these strategies i.e., Serial Recurrent Networks: Elman, 1990 ; and Parser: Perruchet & Vinter, 1998 in an experiment where we compare the lexical and sublexical recognition performance of adults after hearing 2 or 10 min of an artificial spoken language. The results are consistent with Parser's predictions and the clustering approach, showing that performance on words is better than performance on part-words only after 10 min. This result suggests that word segmentation abilities are not merely due to stronger associations between sublexical units but to the emergence of stronger lexical representations during the development of speech perception processes.     

A role for backward transitional probabilities in word segmentation?

A number of studies have shown that people exploit transitional probabilities between successive syllables to segment a stream of artificial continuous speech into words. It is often assumed that what is actually exploited are the forward transitional probabilities (given XY, the probability that X will be followed by Y), even though the backward transitional probabilities (the probability that Y has been preceded by X) were equally informative about word structure in the languages involved in those studies. In two experiments, we showed that participants were able to learn the words from an artificial speech stream when the only available cues were the backward transitional probabilities. Learning is as good under those conditions as when the only available cues are the forward transitional probabilities. Implications for some current models of word segmentation, particularly the simple recurrent networks and PARSER models, are discussed.     

Learning across languages: bilingual experience supports dual language statistical word segmentation

Bilingual acquisition presents learning challenges beyond those found in monolingual environments, including the need to segment speech in two languages. Infants may use statistical cues, such as syllable‐level transitional probabilities, to segment words from fluent speech. In the present study we assessed monolingual and bilingual 14‐month‐olds’ abilities to segment two artificial languages using transitional probability cues. In Experiment 1, monolingual infants successfully segmented the speech streams when the languages were presented individually. However, monolinguals did not segment the same language stimuli when they were presented together in interleaved segments, mimicking the language switches inherent to bilingual speech. To assess the effect of real‐world bilingual experience on dual language speech segmentation, Experiment 2 tested infants with regular exposure to two languages using the same interleaved language stimuli as Experiment 1. The bilingual infants in Experiment 2 successfully segmented the languages, indicating that early exposure to two languages supports infants’ abilities to segment dual language speech using transitional probability cues. These findings support the notion that early bilingual exposure prepares infants to navigate challenging aspects of dual language environments as they begin to acquire two languages.     

Testing the limits of statistical learning for word segmentation

Past research has demonstrated that infants can rapidly extract syllable distribution information from an artificial language and use this knowledge to infer likely word boundaries in speech. However, artificial languages are extremely simplified with respect to natural language. In this study, we ask whether infants’ ability to track transitional probabilities between syllables in an artificial language can scale up to the challenge of natural language. We do so by testing both 5.5‐ and 8‐month‐olds’ ability to segment an artificial language containing four words of uniform length (all CVCV) or four words of varying length (two CVCV, two CVCVCV). The transitional probability cues to word boundaries were held equal across the two languages. Both age groups segmented the language containing words of uniform length, demonstrating that even 5.5‐month‐olds are extremely sensitive to the conditional probabilities in their environment. However, neither age group succeeded in segmenting the language containing words of varying length, despite the fact that the transitional probability cues defining word boundaries were equally strong in the two languages. We conclude that infants’ statistical learning abilities may not be as robust as earlier studies have suggested.     

Learning in reverse: Eight-month-old infants track backward transitional probabilities

Numerous recent studies suggest that human learners, including both infants and adults, readily track sequential statistics computed between adjacent elements. One such statistic, transitional probability, is typically calculated as the likelihood that one element predicts another. However, little is known about whether listeners are sensitive to the directionality of this computation. To address this issue, we tested 8-month-old infants in a word segmentation task, using fluent speech drawn from an unfamiliar natural language. Critically, test items were distinguished solely by their backward transitional probabilities. The results provide the first evidence that infants track backward statistics in fluent speech.     

The metamorphosis of the statistical segmentation output: Lexicalization during artificial language learning

This study combined artificial language learning (ALL) with conventional experimental techniques to test whether statistical speech segmentation outputs are integrated into adult listeners’ mental lexicon. Lexicalization was assessed through inhibitory effects of novel neighbors (created by the parsing process) on auditory lexical decisions to real words. Both immediately after familiarization and post-one week, ALL outputs were lexicalized only when the cues available during familiarization (transitional probabilities and wordlikeness) suggested the same parsing (Experiments 1 and 3). No lexicalization effect occurred with incongruent cues (Experiments 2 and 4). Yet, ALL differed from chance, suggesting a dissociation between item knowledge and lexicalization. Similarly contrasted results were found when frequency of occurrence of the stimuli was equated during familiarization (Experiments 3 and 4). Our findings thus indicate that ALL outputs may be lexicalized as far as the segmentation cues are congruent, and that this process cannot be accounted for by raw frequency.     

Eye movements reveal the on-line computation of lexical probabilities during reading

Skilled readers are able to derive meaning from a stream of visual input with remarkable efficiency. In this article, we present the first evidence that statistical information latent in the linguistic environment can contribute to an account of reading behavior. In two eye-tracking studies, we demonstrate that the transitional probabilities between words have a measurable influence on fixation durations, and using a simple Bayesian statistical model, we show that lexical probabilities derived by combining transitional probability with the prior probability of a word's occurrence provide the most parsimonious account of the eye movement data. We suggest that the brain is able to draw upon statistical information in order to rapidly estimate the lexical probabilities of upcoming words: a computationally inexpensive mechanism that may underlie proficient reading.     

The apes’ edge: positional learning in chimpanzees and humans

A wide variety of organisms produce actions and signals in particular temporal sequences, including the motor actions recruited during tool-mediated foraging, the arrangement of notes in the songs of birds, whales and gibbons, and the patterning of words in human speech. To accurately reproduce such events, the elements that comprise such sequences must be memorized. Both memory and artificial language learning studies have revealed at least two mechanisms for memorizing sequences, one tracking co-occurrence statistics among items in sequences (i.e., transitional probabilities) and the other one tracking the positions of items in sequences, in particular those of items in sequence-edges. The latter mechanism seems to dominate the encoding of sequences after limited exposure, and to be recruited by a wide array of grammatical phenomena. To assess whether humans differ from other species in their reliance on one mechanism over the other after limited exposure, we presented chimpanzees (Pan troglodytes) and human adults with brief exposure to six items, auditory sequences. Each sequence consisted of three distinct sound types (X, A, B), arranged according to two simple temporal rules: the A item always preceded the B item, and the sequence-edges were always occupied by the X item. In line with previous results with human adults, both species primarily encoded positional information from the sequences; that is, they kept track of the items that occurred in the sequence-edges. In contrast, the sensitivity to co-occurrence statistics was much weaker. Our results suggest that a mechanism to spontaneously encode positional information from sequences is present in both chimpanzees and humans and may represent the default in the absence of training and with brief exposure. As many grammatical regularities exhibit properties of this mechanism, it may be recruited by language and constrain the form that certain grammatical regularities take.     

The secret is in the sound: from unsegmented speech to lexical categories

When learning language, young children are faced with many seemingly formidable challenges, including discovering words embedded in a continuous stream of sounds and determining what role these words play in syntactic constructions. We suggest that knowledge of phoneme distributions may play a crucial part in helping children segment words and determine their lexical category, and we propose an integrated model of how children might go from unsegmented speech to lexical categories. We corroborated this theoretical model using a two‐stage computational analysis of a large corpus of English child‐directed speech. First, we used transition probabilities between phonemes to find words in unsegmented speech. Second, we used distributional information about word edges – the beginning and ending phonemes of words – to predict whether the segmented words from the first stage were nouns, verbs, or something else. The results indicate that discovering lexical units and their associated syntactic category in child‐directed speech is possible by attending to the statistics of single phoneme transitions and word‐initial and final phonemes. Thus, we suggest that a core computational principle in language acquisition is that the same source of information is used to learn about different aspects of linguistic structure.     

INCIDENTAL LANGUAGE LEARNING:

Abstract —Two experiments investigated the performance of first-grade children and adults on an incidental language-learning task. Learning entailed word segmentation from continuous speech, an initial and crucial component of language acquisition. Subjects were briefly exposed to an unsegmented artificial language, presented auditorily, in which the only cues to word boundaries were the transitional probabilities between syllables. Subjects were not told that they were listening to a language, or even to listen at all; rather, they were engaged in a cover task of creating computer illustrations. Both adults and children learned the words of the language. Moreover, the children performed as well as the adults. These data suggest that a statistical learning mechanism (transitional probability computation) is able to operate incidentally and, surprisingly, as well in children as in adults.     

The curse of knowledge: first language knowledge impairs adult learners' use of novel statistics for word segmentation

We investigated whether adult learners' knowledge of phonotactic restrictions on word forms from their first language impacts their ability to use statistical information to segment words in a novel language. Adults were exposed to a speech stream where English phonotactics and phoneme co-occurrence information conflicted. A control where these did not conflict was also run. Participants chose between words defined by novel statistics and words that are phonotactically possible in English, but had much lower phoneme contingencies. Control participants selected words defined by statistics while experimental participants did not. This result held up with increases in exposure and when segmentation was aided by telling participants a word prior to exposure. It was not the case that participants simply preferred English-sounding words, however, when the stimuli contained very short pauses, participants were able to learn the novel words despite the fact that they violated English phonotactics. Results suggest that prior linguistic knowledge can interfere with learners' abilities to segment words from running speech using purely statistical cues at initial exposure.     

Found in Translation: Late Bilinguals Do Automatically Activate Their Native Language When They Are Not Using It

In their paper “Do Bilinguals Automatically Activate Their Native Language When They Are Not Using it?”, Costa, Pannunzi, Deco, and Pickering (Cognitive Science, 2017) proposed a reinterpretation of Thierry and Wu's (2004, 2007) finding of native language‐based (Chinese, L1) ERP effects when they tested Chinese–English late bilinguals exclusively in their second language (English, L2). Using simulations in a six‐node Hebbian learning model (three L1 nodes, three L2 nodes), Costa et al. suggested that form overlaps in L1 between otherwise unrelated words create a persistent relationship between their L2 translations. In this scenario, words in the nascent L2 lexicon overlapping in their L1 translations would become linked during learning because of the form overlap in L1; once the L2 words are learned, the direct link between them would be sufficient to generate robust, apparently “L1‐mediated” priming without requiring any activation of L1 translations. Costa et al. contend that links between L2 words remain beyond the learning phase, even after links to L1 representations have been severed, and thus that their model affords an alternative account to (but not a rebuttal of) Thierry and Wu's claim of language non‐selective activation—or automatic activation of translation equivalents—in late bilinguals. In this response, we build on Costa et al.'s original simulation code, showing that it can only reproduce L1‐independent priming when implementing the L1 disconnection in their particular way. By contrast, when severing inter‐language connections bidirectionally, their model fails to retain any sizeable influence of L1 form overlap on L2 activations. The model is not the theory, however, and we discuss several issues that would need to be addressed in further attempts to model language non‐selective activation in late bilinguals.     

The Role of Age of Acquisition in Bilingual Word Translation: Evidence from Spanish-English Bilinguals

The present research tested the hypothesis that the age at which one’s first language (L1) words are learned influences language processing in bilinguals. Prior research on bilingual language processing by Kroll and colleagues has suggested that memory links between L1 words and conceptual representations are stronger than memory links between one’s second language (L2) word and conceptual representations. We hypothesized that the strengths of memory links between L1 words and conceptual representations are stronger for words learned early in life than for words learned later in life. Support for the hypothesis was obtained in bilingual translation experiment with 36 Spanish–English bilinguals. Participants translated L1 words into L2 and L2 words into L1. Half of the L1 words were learned early in childhood (early AoA words), and half were learned later in life (late AoA words). The L2 words were translation equivalents of the L1 words tested; the average age at which L2 words were learned was age 7. Target words were presented either in random order or blocked by semantic category. Translation times were longer when trials were blocked by semantic category (i.e., categorical interference) occurred only when early AoA L1 words were translated into L2. Implications for current models of bilingual memory are discussed.     

Infants’ statistical word segmentation in an artificial language is linked to both parental speech input and reported production abilities

Individual variability in infant's language processing is partly explained by environmental factors, like the quantity of parental speech input, as well as by infant‐specific factors, like speech production. Here, we explore how these factors affect infant word segmentation. We used an artificial language to ensure that only statistical regularities (like transitional probabilities between syllables) could cue word boundaries, and then asked how the quantity of parental speech input and infants’ babbling repertoire predict infants’ abilities to use these statistical cues. We replicated prior reports showing that 8‐month‐old infants use statistical cues to segment words, with a preference for part‐words over words (a novelty effect). Crucially, 8‐month‐olds with larger novelty effects had received more speech input at 4 months and had greater production abilities at 8 months. These findings establish for the first time that the ability to extract statistical information from speech correlates with individual factors in infancy, like early speech experience and language production. Implications of these findings for understanding individual variability in early language acquisition are discussed.