Word Segmentation: The Role of Distributional Cues

“However, the relationship lem might first be approached by an extraction mechanism: perceptually salient units, such as between the semantics of the world and the child’s linguistic input is notoriously complistressed and certain acoustically distinctive final syllables, might be extracted and identi- cated (Gleitman,…”

“On this view, it is possible that perfor-Klatt, 1975; Oller, 1973; Umeda, 1975).”

“Words may be defined in distributional terms,boundaries, including word-initial strong syllables in English and word-final accent in along the lines of other syntactic constituents (e.g., Lyons, 1968).”

“As in the linguistic materials used by Saffran et al. (1996b), there were no acoustic markers of word boundaries.”

“Other potential sources include context- or position-sensitive allophones (Church, 1987b; Gow & Gordon, 1995; Lehiste, 1960), phonotactic cues (Brent & Cartwright, 1996; Cairns et al., 1997), and distributional (or statistical) cues (Brent & Cartwright, 1996; Saffran et al., 1996b ).”

“Sequential learning has thus been demonstrated for a variety of different aspects of language, including: speech segmentation (Curtin et al. 2005; Saffran et al. 1996a; 1996b); discovering complex word-internal structure between nonadjacent elements (Newport & Aslin 2004; Onnis et al. 2005; Peña…”

“Saffran, Newport, and Aslin (1996) did examine possible interactions between statistical and stress cues in a study with adults.”

“• they are created objectively and automatically from text; • they allow the representation of gradations of meaning; • they relate well to experimental evidence indicating that the human cognitive system is sensitive to distributional information [34, 40].”

“We (Saffran et al., 1996) demonstrated, in an explicit-learning task, that adult subjects were able to use transitional probabilities to learn the multisyllabic words of an artificial language presented as a synthesized speech stream containing no other cues to word boundaries.”

“The Human Simulation paradigm provides support for an incremental constraintsatisfaction view of how children acquire their native tongue (for related discussion, see Kelly & Martin, 1994; Saffran et al., 1996; Seidenberg & MacDonald, 1999).”

“Of interest, this same asymmetry was obtained by Saffran, Newport, and Aslin (1996) in the auditory domain, although the structure of their part triplets was slightly different (X-1-2 and 2-3-”

“These facts have caused some researchers to propose that strategies based on statistical sequencing information are a crucial first step in bootstrapping word segmentation (Thiessen and Saffran, 2003), and have provoked a great deal of interest in these strategies (Saffran et al., 1996b; Saffran et…”

“Further, both adults and 8-month-old infants use these sequential probabilities in segmenting words from connected speech sequences in artificial language studies (Saffran, Newport, & Aslin, 1996a ; Saffran, Aslin, & Newport, 1996b ).”

“Thus, irrespective of awareness, the hippocampus is well-suited to forms of statistical learning that can occur after minimal exposure [2, 3].”

“In the Saffran et al. (1996a) experiments, the infants were presented with four items during the test.”

“Finally, previous work on statistical learning [63–65] would correspond, in Figure 1, to probing (typically offline) whether subjects have learned the distribution of linguistic items in the experiment (e.”

“can provide a strong foundation for binding input into perceptually coherent and meaningful units like words (Figure 4C; Brent and Cartwright, 1996; McQueen, 1998; Saffran et al., 1996),”

“The computation of co-occurrence relations, and transitional probabilities more generally, has been shown to be a key mechanism in infants’ early language learning (Saffran et al., 1996).”

“Statistics that have been proposed for measuring these differences include “successor frequency” (Harris, 1954), “transitional probabilities” (Saffran et al., 1996), mutual information (Sun et al.”

“Saffran, Newport, and Aslin (1996) suggest that children may segment utterances at low points of the transitional probability between adjacent syllables—that is, when a syllable occurs that is surprising given its predecessor.”

“There is growing evidence that babies do possess powerful statistical learning mechanisms [23].”

“…with transitional probabilities of 0, even though there were no " dips " in these adjacent transitional probabilities as in previous statistical learning experiments (Saffran et al., 1996), and the part-triplet test items crossed the boundaries of the coherent, interleaved nonadjacent triplets.”

“The past 20 years of perceptual, linguistic, and cognitive investigation in the information-processing framework have heavily documented the notion of contingent frequency as fundamental for describing both animal and human behavior (Gallistel & Gibbon, 2000; Rescorla & Wagner, 1972; Saffran, Newport, & Aslin, 1996 ).”

“Past attempts to study the learning of statistics and rules have been conducted in domains such as artificial grammar learning (12, 13), serial reaction times (14, 15), word segmentation from fluent speech (16, 17), and pattern abstraction from strings of words (18, 19).”

“…languages containing natural language-like properties serve as a tool to investigate basic processes in language acquisition (e.g., Braine, 1971; Braine et al., 1990; Moeser & Bregman, 1972; Morgan et al., 1987, 1989; Morgan & Newport, 1981; Saffran et al., 1996, 1997; Valian & Coulson, 1988).”

“Such learning has been demonstrated across a variety of situations, including speech segmentation (Saffran et al., 1996a; Saffran et al., 1996b; Christiansen et al., 1998), learning the orthographic regularities of written words (Pacton et al., 2001), visuomotor learning (Hunt & Aslin, 2001), and tone sequences (Saffr…”

“Saffran, Newport, and Aslin (1996) suggested that the learning mechanism involved computing the transitional probability of adjacent syllables, using the peaks in transitional probability to group syllables, into words, while using the dips in transitional probability to form breaks between words.”

“…over the input to learning is now well established (Aslin, Saffran, & Newport, 1998; Braine, 1963; Gerken, 2006; Gomez, 2002; Hudson Kam & Newport, 2005; Mintz, 2002; Moeser & Bregman, 1972; Morgan, Meier, & Newport, 1987; Morgan & Newport, 1981; Saffran et al., 1996; Wonnacott & Newport, 2005).”

“These findings provide another piece of evidence to the growing literature that suggests listeners are sensitive to the statistical regularities of the language (e.g., McQueen & Pitt, 1996; Saffran et al., 1996).”

“What is the phenomenal experience of the listener of a new language such as the one used in the Saffran et al. experiments, at the beginning and end of training, respectively? When people are confronted with material consisting of a succession of elements, each of them matching some of their processing primitives, the…”

“Artificial linguistic materials have been successfully used to provide precise control over distributional information in several domains of language acquisition and processing (e.g., Braine, 1963; Gomez & Gerken, 2000; Morgan, Meier, & Newport, 1987; Saffran, Newport, & Aslin, 1996 ).”

“Given that infants are sensitive to the syllable as a unit of speech (10), computing statistical relations between syllables has been proposed to be a general mechanism that can parse sequences, generating as potential candidate words those syllable sequences that have high statistical coherence (4, 11).”

“Behavioral studies provide evidence that a wide range of segmentation cues contribute to adults' ability to segment continuous speec”

“When a set of pure tones was substituted for the speech syllables used in the original Saffran, Newport, and Aslin (1996) study, adults showed the same ability to learn tone sequences that they had shown for syllable sequences.”

“…nature (either transitional probabilities, which were higher within words than across word boundaries, or frequency of cooccurrence, which were higher for the three syllable sequences within words than across words; cf. Saffran et al., 1996b; Saffran et al., 1997; Aslin, Saffran & Newport, 1998).”

“This option was provided to keep participants from becoming bored during the task, and has been used successfully (for slightly different purposes) by Saffran, Newport, and Aslin (1996).”

“…and psychological disorders (e.g., Reber, 1993; Cleeremans et al., 1998), with many studies showing remarkable similarities in infant, child, and adult learning (e.g., Cherry and Stadler, 1995; Saffran et al., 1996a,b; Thomas and Nelson, 2001; Fiser and Aslin, 2002a,b; Gómez, 2002; Saffran, 2002).”

“…specify word boundaries, research has shown that infants as young as 7.5 months of age can use statistical regularities such as the greater co-occurrence of syllables within words than between words to successfully parse a continuous stream of speech (Saffran et al., 1996a; Aslin et al., 1998).”

“In previous experiments (e.g., Saffran et al., 1996; Peña et al., 2002), where up to 6 ‘words’ were concatenated at random, the TP from the last syllable of one ‘word’ to the first syllable of another was 0.24 (since immediate repetitions are not allowed, each ‘word’ can be followed by one of the…”

“Previous language-learning studies have shown similar findings for adults exposed to an artificial language and children or even infants exposed to the same type of language ( Saffran, Newport, & Aslin, 1996 ).”

“Indeed, humans can compute statistical relations (in particular, transitional probabilities, or TPs) that can be useful to identify lexical elements in continuous speech (e.g., Saffran, Newport, & Aslin, 1996 ).”

“The 1 Saffran, Newport, and Aslin (1996) used lengthening effect of vowel harmony and all other interacof the vowel as a cue for lexical stress.”

“…(1996) computational Incremental Distributional Regularity Optimization (INCDROP) model suggests and data from Saffran, Aslin, et al. (1996) and Saffran, Newport, and Aslin (1996) illustrate, high transitional probability between speech segments tends to promote cohesion in the percept ,…”

“Learners can discover word boundaries by computing the transitional probabilities between neighboring syllables, which are higher within words compared to across word boundaries (Saffran et al., 1996a, 1996b).”