Statistical learning, implicit learning and first language acquisition: a critical evaluation of ... more Statistical learning, implicit learning and first language acquisition: a critical evaluation of age-invariance and the link to language learning outcomes Abstract. The role of distributional information in language learning, and learning more generally, has been studied extensively in both the statistical learning and the implicit learning literatures. Despite the similarity in research questions, the two literatures have remained largely separate. Here, we provide a critical evaluation of two developmental issues pertaining to these literatures: the question of age-invariance-does learning improve during development or is it fully developed in infancy-and the ability of statistical learning measures to predict variation in language learning outcomes. In the first part, we draw on combined findings from the statistical learning and implicit learning literatures to suggest that both improve during childhood, contra the age-invariance prediction. In the second part, we review existing findings on the link between statistical learning and language outcomes in development and evaluate the suitability of current SL measures for capturing individual differences across the lifespan. We present data indicating that while SL measures show moderate reliability in adults, they do not capture stable individual differences in children, undermining their utility for assessing the link to language outcomes in developmental samples. We end with a broader discussion of the way methods impact our theoretical understanding of statistical and implicit learning.
Recent work suggests that cultural transmission can lead to the emergence of linguistic structure... more Recent work suggests that cultural transmission can lead to the emergence of linguistic structure as speakers' weak individual biases become amplified through iterated learning. However, to date no published study has demonstrated a similar emergence of linguistic structure in children. The lack of evidence from child learners constitutes a problematic gap in the literature: if such learning biases impact the emergence of linguistic structure, they should also be found in children, who are the primary learners in real-life language transmission. However, children may differ from adults in their biases given age-related differences in general cognitive skills. Moreover, adults' performance on iterated learning tasks may reflect existing (and explicit) linguistic biases, partially undermining the generality of the results. Examining children's performance can also help evaluate contrasting predictions about their role in emerging languages: do children play a larger or smaller role than adults in the creation of structure? Here, we report a series of four iterated artificial language learning studies (based on Kirby, Cornish & Smith, 2008) with both children and adults, using a novel child-friendly paradigm. Our results show that linguistic structure does not emerge more readily in children compared to adults, and that adults are overall better in both language learning and in creating linguistic structure. When languages could become underspecified (by allowing homonyms), children and adults were similar in developing consistent mappings between meanings and signals in the form of structured ambiguities. However, when homonimity was not allowed, only adults created compositional structure. This study is a first step in using iterated language learning paradigms to explore child-adult differences. It provides the first demonstration that cultural transmission has a different effect on the languages produced by children and adults: While children were able to develop systematicity, their languages did not show compositionality. We focus on the relation between learning and structure creation as a possible explanation for our findings and discuss implications for children's role in the emergence of linguistic structure.
Across languages, word frequency and rank follow a power law relation, forming a distribution kno... more Across languages, word frequency and rank follow a power law relation, forming a distribution known as the Zipfian distribution. There is growing experimental evidence that this wellstudied phenomenon may be beneficial for language learning. However, most investigations of word distributions in natural language have focused on adult-to-adult speech: Zipf's law has not been thoroughly evaluated in child-directed speech (CDS) across languages. If Zipfian distributions facilitate learning, they should also be found in CDS. At the same time, several unique properties of CDS may result in a less skewed distribution. Here, we examine the frequency distribution of words in CDS in three studies. We first show that CDS is Zipfian across 15 languages from seven language families. We then show that CDS is Zipfian from early on (six-months) and across development for five languages with sufficient longitudinal data. Finally, we show that the distribution holds across different parts of speech: Nouns, verbs, adjectives and prepositions follow a Zipfian distribution. Together, the results show that the input children hear is skewed in a particular way from early on, providing necessary (but not sufficient) support for the postulated learning advantage of such skew. They highlight the need to study skewed learning environments experimentally.
The prevalence of redundancy in the world languages has long puzzled language researchers. It is ... more The prevalence of redundancy in the world languages has long puzzled language researchers. It is especially surprising in light of the growing evidence on speakers' tendency to avoid redundant elements in production (omitting or reducing more predictable elements). Here, we propose that redundancy can be functional for learning. In particular, we argue that redundant cues can facilitate learning, even when they make the language system more complicated. This prediction is further motivated by the Linguistic Niche Hypothesis (Lupyan & Dale, 2010), which suggests that morphological complexity can arise due to the advantage redundancy might confer for child learners. We test these hypotheses in an artificial language learning study with children and adults, where either word order alone or both word order and case marking serve as cues for thematic assignment in a novel construction. We predict, and find, that children learning the redundant language learn to produce it, and show better comprehension of the novel thematic assignment than children learning the non-redundant language, despite having to learn an additional morpheme. Children in both conditions were similarly accurate in producing the novel word order, suggesting redundancy might have a differential effect on comprehension and production. Adults did not show better learning in the redundant condition, most likely because they were at ceiling in both conditions. We discuss implications for theories of language learning and language change.
Proceedings of the Annual Meeting of the Cognitive Science Society, 2021
If speakers communicate efficiently, they should produce more linguistic material when comprehens... more If speakers communicate efficiently, they should produce more linguistic material when comprehension difficulty increases. Comprehension difficulty can be impacted by the message itself (studied extensively) or by properties of the listener (studied less). Here we investigate the impact of listeners' estimated proficiency on speakers' productions, using referential choice as a case study. Compared to full noun phrases (fNPs, 'the woman'), pronouns ('she') can convey similar content with less linguistic material: accordingly, speakers use more fNPs when the message is unpredictable (Tily & Piantadosi, 2009) and when listeners lack relevant information (Bard & Aylett, 2004). If referential choice is also impacted by listeners' estimated proficiency, then speakers should use more fNPs when conversing with language learners. To test this, we compared participants' descriptions of the same picture book to learners and proficient speakers. Indeed, participants used more fNPs when their interlocutors were child-or adult-learners, illustrating efficient adaptation to listener type.
Speakers tend to assign more linguistic material to less predictable elements. This tendency is t... more Speakers tend to assign more linguistic material to less predictable elements. This tendency is typically explained by a bias for an efficient trade-off between production effort and understandability and is claimed to shape linguistic structures across languages. Recent work suggests this trade-off enters the linguistic system through learning processes with learners deviating from their input by increasing marking for less predictable elements. However, no study to date has tested whether child learners also show such predictability-based marking, an important gap seeing that children are the primary learners in real-life language acquisition. A recent study showed that adults increase predictability-based marking of an optional-plural marker, in line with communicative efficiency. Here, we ask if children show a similar pattern. Results show that children, unlike adults, do not show an efficient trade-off in their productions. We discuss implications for the role of different language learners on language change.
Do language learners benefit from exposure to input that is more predictable and has lower entrop... more Do language learners benefit from exposure to input that is more predictable and has lower entropy? Frequency is known to facilitate learning (more frequent words acquired earlier). However, frequency is only one measure of the distributional structure of the linguistic input. Here, we show that entropy also impacts language learning: adults show better word segmentation in an artificial language when the sequence has lower entropy (created by making one word more frequent). Segmentation improved both for the language as a whole, and for the less frequent words, despite appearing half the number of times. These results illustrate the facilitative effect of entropy reduction on language learning. Theoretically, they show that the effect of frequency is relative, not absolute, and that language learners are sensitive to more complex measures of the environment. Methodologically, they suggest that the prevalent use of uniform distributions in word segmentation studies may underestimate learners' abilities.
Trends in language acquisition research, Aug 15, 2016
The chapter explores the distribution and content of frequent frames – recurring multiword sequen... more The chapter explores the distribution and content of frequent frames – recurring multiword sequences – appearing at the start of utterances in speech directed to young Hebrew-speaking children. Previous work has documented the existence and prevalence of such frames in several languages (English, German, and Russian). Here, analysis of a dense corpus of Hebrew child-directed speech was undertaken with two goals in mind. The first aims at examining the distribution of multiword elements in Hebrew child-directed speech, to ascertain whether frequent frames are found in a morphologically rich language like Hebrew and, if so, to see how pervasive they are compared to other languages, and how consistent across different caretakers. The second goal is to explore the content of frequent frames in Hebrew to address such questions as: Do they provide children with relevant morphological and syntactic information? Are they frequent enough to be employed in learning? Results show that frequent frames do occur in Hebrew, that they are relatively consistent across caretakers, and that they illustrate a range of grammatical relations. These findings expand our understanding of frequent frames in general, while also adding to the relatively sparse information on the nature of child-directed speech in Hebrew.
Is linguistic knowledge probabilistic? A growing body of evidence suggests that: • speakers have ... more Is linguistic knowledge probabilistic? A growing body of evidence suggests that: • speakers have knowledge about the probabilities of linguistic events • those probabilities influence speakers' choices between constructions, and their phonetic realization
During their first year, infants learn to name objects. To do so, they need to segment speech, ex... more During their first year, infants learn to name objects. To do so, they need to segment speech, extract the label and map it to the correct referent. While children successfully do so in the wild, previous results suggest they struggle to simultaneously learn segmentation and object-label pairings in the lab. Here, we ask if some of children's difficulty is related to the uniform distribution they were exposed to, since it differs from that of natural language, and has high entropy (making it less predictable). Will a low entropy distribution facilitate children's performance in these two tasks? We looked at children's (mean age=10;4 years) simultaneous segmentation and object-label mapping of words in an artificial language task. Low entropy (created by making one word more frequent) facilitated children's performance in both tasks. We discuss the importance of using more ecologic stimuli in the lab, specifically-distributions with lower entropy.
The presence of redundant marking in languages raises interesting questions about the balance of ... more The presence of redundant marking in languages raises interesting questions about the balance of different pressures in language learning and use. Speakers tend to avoid redundant elements in production: omitting (or reducing) more predictable elements. At the same time, languages maintain different types of redundant marking, such as encoding thematic assignment by both word order and case marking. Why is redundancy found in languages even though speakers seem to avoid it? Here, we propose that redundant cues can facilitate learning. We test this hypothesis in an artificial language learning study with children, where either word order alone or both word order and case marking serve as cues for thematic assignment in a novel construction. Results show that children learned the redundant language better despite having to learn an additional morpheme. We discuss implications for the effect of different cognitive pressures on language change.
While the languages of the world differ in many respects, they share certain commonalties, which ... more While the languages of the world differ in many respects, they share certain commonalties, which can provide insight on our shared cognition. Here, we explore the learnability consequences of one of the striking commonalities between languages. Across languages, word frequencies follow a Zipfian distribution, showing a power law relation between a word's frequency and its rank. While their source in language has been studied extensively, less work has explored the learnability consequences of such distributions for language learners. We propose that the greater predictability of words in this distribution (relative to less skewed distributions) can facilitate word segmentation, a crucial aspect of early language acquisition. To explore this, we quantify word predictability using unigram entropy, assess it across languages using naturalistic corpora of child-directed speech and then ask whether similar unigram predictability facilitates word segmentation in the lab. We find similar unigram entropy in child-directed speech across 15 languages. We then use an auditory word segmentation task to show that the unigram predictability levels found in natural language are uniquely facilitative for word segmentation for both children and adults. These findings illustrate the facilitative impact of skewed input distributions on learning and raise questions about the possible role of cognitive pressures in the prevalence of Zipfian distributions in language.
One of the striking commonalities between languages is the way word frequencies are distributed. ... more One of the striking commonalities between languages is the way word frequencies are distributed. Across languages, word frequencies follow a Zipfian distribution, showing a power law relation between a word's frequency and its rank (Zipf, 1949). Intuitively, this means that languages have relatively few high-frequency words and many low-frequency ones. While studied extensively, little work has explored the learnability consequences of the greater predictability of words in such distributions. Here, we propose such distributions confer a learnability advantage for word segmentation, a foundational aspect of language acquisition. We capture the greater predictability of words using the information-theoretic notion of efficiency, which tells us how predictable a distribution is relative to a uniform one. We first use corpus analyses to show that child-directed speech is similarly predictable across fifteen different languages. We then experimentally investigate the impact of distribution predictability on children and adults. We show that word segmentation is uniquely facilitated at the predictability levels found in language, compared both with uniform distributions and with skewed distributions that are less predictable than those of natural language. We further show that distribution predictability impacts learning more than distribution shape, and that learning is not improved further in distributions more predictable than natural language. These novel findings illustrate learners' sensitivity to the overall predictability of the linguistic environment; suggest that the predictability levels found in language provide an optimal environment for learning; and point to the possible role of cognitive pressures in the emergence and propensity of such distributions in language.
Statistical learning, implicit learning and first language acquisition: a critical evaluation of ... more Statistical learning, implicit learning and first language acquisition: a critical evaluation of age-invariance and the link to language learning outcomes Abstract. The role of distributional information in language learning, and learning more generally, has been studied extensively in both the statistical learning and the implicit learning literatures. Despite the similarity in research questions, the two literatures have remained largely separate. Here, we provide a critical evaluation of two developmental issues pertaining to these literatures: the question of age-invariance-does learning improve during development or is it fully developed in infancy-and the ability of statistical learning measures to predict variation in language learning outcomes. In the first part, we draw on combined findings from the statistical learning and implicit learning literatures to suggest that both improve during childhood, contra the age-invariance prediction. In the second part, we review existing findings on the link between statistical learning and language outcomes in development and evaluate the suitability of current SL measures for capturing individual differences across the lifespan. We present data indicating that while SL measures show moderate reliability in adults, they do not capture stable individual differences in children, undermining their utility for assessing the link to language outcomes in developmental samples. We end with a broader discussion of the way methods impact our theoretical understanding of statistical and implicit learning.
Recent work suggests that cultural transmission can lead to the emergence of linguistic structure... more Recent work suggests that cultural transmission can lead to the emergence of linguistic structure as speakers' weak individual biases become amplified through iterated learning. However, to date no published study has demonstrated a similar emergence of linguistic structure in children. The lack of evidence from child learners constitutes a problematic gap in the literature: if such learning biases impact the emergence of linguistic structure, they should also be found in children, who are the primary learners in real-life language transmission. However, children may differ from adults in their biases given age-related differences in general cognitive skills. Moreover, adults' performance on iterated learning tasks may reflect existing (and explicit) linguistic biases, partially undermining the generality of the results. Examining children's performance can also help evaluate contrasting predictions about their role in emerging languages: do children play a larger or smaller role than adults in the creation of structure? Here, we report a series of four iterated artificial language learning studies (based on Kirby, Cornish & Smith, 2008) with both children and adults, using a novel child-friendly paradigm. Our results show that linguistic structure does not emerge more readily in children compared to adults, and that adults are overall better in both language learning and in creating linguistic structure. When languages could become underspecified (by allowing homonyms), children and adults were similar in developing consistent mappings between meanings and signals in the form of structured ambiguities. However, when homonimity was not allowed, only adults created compositional structure. This study is a first step in using iterated language learning paradigms to explore child-adult differences. It provides the first demonstration that cultural transmission has a different effect on the languages produced by children and adults: While children were able to develop systematicity, their languages did not show compositionality. We focus on the relation between learning and structure creation as a possible explanation for our findings and discuss implications for children's role in the emergence of linguistic structure.
Across languages, word frequency and rank follow a power law relation, forming a distribution kno... more Across languages, word frequency and rank follow a power law relation, forming a distribution known as the Zipfian distribution. There is growing experimental evidence that this wellstudied phenomenon may be beneficial for language learning. However, most investigations of word distributions in natural language have focused on adult-to-adult speech: Zipf's law has not been thoroughly evaluated in child-directed speech (CDS) across languages. If Zipfian distributions facilitate learning, they should also be found in CDS. At the same time, several unique properties of CDS may result in a less skewed distribution. Here, we examine the frequency distribution of words in CDS in three studies. We first show that CDS is Zipfian across 15 languages from seven language families. We then show that CDS is Zipfian from early on (six-months) and across development for five languages with sufficient longitudinal data. Finally, we show that the distribution holds across different parts of speech: Nouns, verbs, adjectives and prepositions follow a Zipfian distribution. Together, the results show that the input children hear is skewed in a particular way from early on, providing necessary (but not sufficient) support for the postulated learning advantage of such skew. They highlight the need to study skewed learning environments experimentally.
The prevalence of redundancy in the world languages has long puzzled language researchers. It is ... more The prevalence of redundancy in the world languages has long puzzled language researchers. It is especially surprising in light of the growing evidence on speakers' tendency to avoid redundant elements in production (omitting or reducing more predictable elements). Here, we propose that redundancy can be functional for learning. In particular, we argue that redundant cues can facilitate learning, even when they make the language system more complicated. This prediction is further motivated by the Linguistic Niche Hypothesis (Lupyan & Dale, 2010), which suggests that morphological complexity can arise due to the advantage redundancy might confer for child learners. We test these hypotheses in an artificial language learning study with children and adults, where either word order alone or both word order and case marking serve as cues for thematic assignment in a novel construction. We predict, and find, that children learning the redundant language learn to produce it, and show better comprehension of the novel thematic assignment than children learning the non-redundant language, despite having to learn an additional morpheme. Children in both conditions were similarly accurate in producing the novel word order, suggesting redundancy might have a differential effect on comprehension and production. Adults did not show better learning in the redundant condition, most likely because they were at ceiling in both conditions. We discuss implications for theories of language learning and language change.
Proceedings of the Annual Meeting of the Cognitive Science Society, 2021
If speakers communicate efficiently, they should produce more linguistic material when comprehens... more If speakers communicate efficiently, they should produce more linguistic material when comprehension difficulty increases. Comprehension difficulty can be impacted by the message itself (studied extensively) or by properties of the listener (studied less). Here we investigate the impact of listeners' estimated proficiency on speakers' productions, using referential choice as a case study. Compared to full noun phrases (fNPs, 'the woman'), pronouns ('she') can convey similar content with less linguistic material: accordingly, speakers use more fNPs when the message is unpredictable (Tily & Piantadosi, 2009) and when listeners lack relevant information (Bard & Aylett, 2004). If referential choice is also impacted by listeners' estimated proficiency, then speakers should use more fNPs when conversing with language learners. To test this, we compared participants' descriptions of the same picture book to learners and proficient speakers. Indeed, participants used more fNPs when their interlocutors were child-or adult-learners, illustrating efficient adaptation to listener type.
Speakers tend to assign more linguistic material to less predictable elements. This tendency is t... more Speakers tend to assign more linguistic material to less predictable elements. This tendency is typically explained by a bias for an efficient trade-off between production effort and understandability and is claimed to shape linguistic structures across languages. Recent work suggests this trade-off enters the linguistic system through learning processes with learners deviating from their input by increasing marking for less predictable elements. However, no study to date has tested whether child learners also show such predictability-based marking, an important gap seeing that children are the primary learners in real-life language acquisition. A recent study showed that adults increase predictability-based marking of an optional-plural marker, in line with communicative efficiency. Here, we ask if children show a similar pattern. Results show that children, unlike adults, do not show an efficient trade-off in their productions. We discuss implications for the role of different language learners on language change.
Do language learners benefit from exposure to input that is more predictable and has lower entrop... more Do language learners benefit from exposure to input that is more predictable and has lower entropy? Frequency is known to facilitate learning (more frequent words acquired earlier). However, frequency is only one measure of the distributional structure of the linguistic input. Here, we show that entropy also impacts language learning: adults show better word segmentation in an artificial language when the sequence has lower entropy (created by making one word more frequent). Segmentation improved both for the language as a whole, and for the less frequent words, despite appearing half the number of times. These results illustrate the facilitative effect of entropy reduction on language learning. Theoretically, they show that the effect of frequency is relative, not absolute, and that language learners are sensitive to more complex measures of the environment. Methodologically, they suggest that the prevalent use of uniform distributions in word segmentation studies may underestimate learners' abilities.
Trends in language acquisition research, Aug 15, 2016
The chapter explores the distribution and content of frequent frames – recurring multiword sequen... more The chapter explores the distribution and content of frequent frames – recurring multiword sequences – appearing at the start of utterances in speech directed to young Hebrew-speaking children. Previous work has documented the existence and prevalence of such frames in several languages (English, German, and Russian). Here, analysis of a dense corpus of Hebrew child-directed speech was undertaken with two goals in mind. The first aims at examining the distribution of multiword elements in Hebrew child-directed speech, to ascertain whether frequent frames are found in a morphologically rich language like Hebrew and, if so, to see how pervasive they are compared to other languages, and how consistent across different caretakers. The second goal is to explore the content of frequent frames in Hebrew to address such questions as: Do they provide children with relevant morphological and syntactic information? Are they frequent enough to be employed in learning? Results show that frequent frames do occur in Hebrew, that they are relatively consistent across caretakers, and that they illustrate a range of grammatical relations. These findings expand our understanding of frequent frames in general, while also adding to the relatively sparse information on the nature of child-directed speech in Hebrew.
Is linguistic knowledge probabilistic? A growing body of evidence suggests that: • speakers have ... more Is linguistic knowledge probabilistic? A growing body of evidence suggests that: • speakers have knowledge about the probabilities of linguistic events • those probabilities influence speakers' choices between constructions, and their phonetic realization
During their first year, infants learn to name objects. To do so, they need to segment speech, ex... more During their first year, infants learn to name objects. To do so, they need to segment speech, extract the label and map it to the correct referent. While children successfully do so in the wild, previous results suggest they struggle to simultaneously learn segmentation and object-label pairings in the lab. Here, we ask if some of children's difficulty is related to the uniform distribution they were exposed to, since it differs from that of natural language, and has high entropy (making it less predictable). Will a low entropy distribution facilitate children's performance in these two tasks? We looked at children's (mean age=10;4 years) simultaneous segmentation and object-label mapping of words in an artificial language task. Low entropy (created by making one word more frequent) facilitated children's performance in both tasks. We discuss the importance of using more ecologic stimuli in the lab, specifically-distributions with lower entropy.
The presence of redundant marking in languages raises interesting questions about the balance of ... more The presence of redundant marking in languages raises interesting questions about the balance of different pressures in language learning and use. Speakers tend to avoid redundant elements in production: omitting (or reducing) more predictable elements. At the same time, languages maintain different types of redundant marking, such as encoding thematic assignment by both word order and case marking. Why is redundancy found in languages even though speakers seem to avoid it? Here, we propose that redundant cues can facilitate learning. We test this hypothesis in an artificial language learning study with children, where either word order alone or both word order and case marking serve as cues for thematic assignment in a novel construction. Results show that children learned the redundant language better despite having to learn an additional morpheme. We discuss implications for the effect of different cognitive pressures on language change.
While the languages of the world differ in many respects, they share certain commonalties, which ... more While the languages of the world differ in many respects, they share certain commonalties, which can provide insight on our shared cognition. Here, we explore the learnability consequences of one of the striking commonalities between languages. Across languages, word frequencies follow a Zipfian distribution, showing a power law relation between a word's frequency and its rank. While their source in language has been studied extensively, less work has explored the learnability consequences of such distributions for language learners. We propose that the greater predictability of words in this distribution (relative to less skewed distributions) can facilitate word segmentation, a crucial aspect of early language acquisition. To explore this, we quantify word predictability using unigram entropy, assess it across languages using naturalistic corpora of child-directed speech and then ask whether similar unigram predictability facilitates word segmentation in the lab. We find similar unigram entropy in child-directed speech across 15 languages. We then use an auditory word segmentation task to show that the unigram predictability levels found in natural language are uniquely facilitative for word segmentation for both children and adults. These findings illustrate the facilitative impact of skewed input distributions on learning and raise questions about the possible role of cognitive pressures in the prevalence of Zipfian distributions in language.
One of the striking commonalities between languages is the way word frequencies are distributed. ... more One of the striking commonalities between languages is the way word frequencies are distributed. Across languages, word frequencies follow a Zipfian distribution, showing a power law relation between a word's frequency and its rank (Zipf, 1949). Intuitively, this means that languages have relatively few high-frequency words and many low-frequency ones. While studied extensively, little work has explored the learnability consequences of the greater predictability of words in such distributions. Here, we propose such distributions confer a learnability advantage for word segmentation, a foundational aspect of language acquisition. We capture the greater predictability of words using the information-theoretic notion of efficiency, which tells us how predictable a distribution is relative to a uniform one. We first use corpus analyses to show that child-directed speech is similarly predictable across fifteen different languages. We then experimentally investigate the impact of distribution predictability on children and adults. We show that word segmentation is uniquely facilitated at the predictability levels found in language, compared both with uniform distributions and with skewed distributions that are less predictable than those of natural language. We further show that distribution predictability impacts learning more than distribution shape, and that learning is not improved further in distributions more predictable than natural language. These novel findings illustrate learners' sensitivity to the overall predictability of the linguistic environment; suggest that the predictability levels found in language provide an optimal environment for learning; and point to the possible role of cognitive pressures in the emergence and propensity of such distributions in language.
Even in relatively configurational languages, such as English, speakers frequently have a choice ... more Even in relatively configurational languages, such as English, speakers frequently have a choice between different constituent orders. Many of these word order variations have been linked to complexity (Hawkins 2005; inter alia). For example, heavy-NP shift is more likely if the shifted NP is more complex than the NP it shifts over . Other cases of word order variations, however, have not been considered in these terms. The choice between different wh-phrase orders, as in , has been said to be determined by (categorical) grammatical constraints, such as Superiority Robinson 1972, Chomsky 1973; inter alia).
In this chapter, we address the debate between single-system and dual-system models of language b... more In this chapter, we address the debate between single-system and dual-system models of language by looking at the processing of multi-word phrases. We present findings that challenge the distinction between ‘stored’ and ‘computed’ linguistic forms via two experiments. The first demonstrates parallels in the processing of words and phrases: frequent four-word phrases are processed more quickly than less frequent ones, without any evidence of a frequency threshold. The second experiment shows that idiomatic phrases prime their construction just as well as non-idiomatic phrases, suggesting that they are not stored as analyzed wholes, but instead have internal structure. Taken together, the findings undermine the empirical criteria traditionally used to distinguish between ‘stored’ and ‘computed’ forms: compositional phrases showed frequency effects, even though such effects are often thought to be a marker of lexical storage, while non-compositional forms (idioms) showed evidence of internal structure, unexpected if they are stored as unanalyzed wholes. The findings show that linguistic structures are processed in qualitatively the same way regardless of where they fall on the frequency and compositionality continua, and highlight the utility of models that deal with all linguistic experience in a qualitatively similar fashion, and allow for experience to influence the learning, representation and processing of all linguistic patterns.
The Acquisition of Relative Clauses: Processing, …, Jan 1, 2011
In this chapter, I outline the developmental path of relative clauses in Hebrew while asking more... more In this chapter, I outline the developmental path of relative clauses in Hebrew while asking more general questions about how constructions are learned. I argue that Hebrew-speaking children show a gradual expansion of uses that is sensitive to the distributional patterns in their input. This pattern, found both in comprehension and production, is consistent with usage-based predictions about how constructions are learned. Taking Hebrew relative clauses as a case-study, I show how children’s own uses become more semantically and structurally complex, and how their understanding develops to rely less on morphological cues. By looking closely at production and comprehension patterns we can see that children’s use of relative clauses, like that of other constructions, develops gradually over time in ways that are sensitive to language-general and language specific cues. Finally, I suggest that the frequency of multi-word sequences (larger than one lexical word) plays a role in children’s expansion of uses: Other things being equal, children prefer to produce construction variants with a higher chunk frequency.
Experience, Variation and Generalization: Learning a …, Jan 1, 2011
Text-book descriptions of how production develops in first language acquisition often move from b... more Text-book descriptions of how production develops in first language acquisition often move from babbling (producing syllables), through single-word utterances, to multi-word combinations. This progression emphasizes the small-to-big aspect of language learning where each stage involves larger, more structured linguistic units. In this chapter, I discuss an equally important, but often neglected, process: the move from large unanalyzed units to the identification and analysis of smaller more structured ones. I present evidence for such Gestalt processes in a range of linguistic domains (phonetics, morphology, and syntax), and suggest that they play an important role in first language learning by offering children another route into linguistic structure. I discuss implications for identifying early building blocks for language learning.
Relative clauses (RCs) have been studied extensively in language acquisition and adult processing... more Relative clauses (RCs) have been studied extensively in language acquisition and adult processing. Studies show that both children and adults find object relatives harder than subject relatives [1, 2]. Despite the similar pattern, adult difficulty is taken to reflect the increased processing demands of object relatives [1] while child difficulty is often interpreted as evidence for children's lack of adult-like knowledge of the structure, attributed to under-developed syntax [2] or to the use of non-adult processing heuristics which are ...
Are speakers sensitive to four-gram frequency? Production is affected by the frequency (and predi... more Are speakers sensitive to four-gram frequency? Production is affected by the frequency (and predictability) of linguistic material of varying sizes. Words are phonetically reduced in more predictable semantic and syntactic environments (eg, Jurafsky et al. 2001; Gahl & Garnsey, 2004; Tily et al., 2009), and in highly collocated phrases (eg I don't know, Bybee & Schiebman, 1999). Such investigations uncover the range of distributional information speakers are sensitive to.
ABSTRACT Children find object relative clauses difficult. They show poor comprehension that lags ... more ABSTRACT Children find object relative clauses difficult. They show poor comprehension that lags behind production into their fifth year. This finding has shaped models of relative clause acquisition, with appeals to processing heuristics or syntactic preferences to explain why object relatives are more difficult than subject relatives. Two studies here suggest that children (age 4; 6) do not find all object relatives difficult: a corpus study shows that children most often hear and produce object relatives with pronominal subjects.
Page 1. Phonetic Production Reflects Syntactic Probability: Evidence from Duration and Disfluency... more Page 1. Phonetic Production Reflects Syntactic Probability: Evidence from Duration and Disfluency Harry Tily, Neal Snider, Anubha Kothari, Inbal Arnon and Joan Bresnan Department of Linguistics Stanford University AMLaP, Turku August 2007 Page 2. Is linguistic knowledge probabilistic?
Abstract Speakers frequently have a choice among multiple ways of expressing one and the same tho... more Abstract Speakers frequently have a choice among multiple ways of expressing one and the same thought. When choosing between syntactic constructions for expressing a given meaning, speakers are sensitive to probabilistic tendencies for syntactic, semantic or contextual properties of an utterance to favor one construction or another. Taken together, such tendencies may align to make one construction overwhelmingly more probable, marginally more probable, or no more probable than another.
What does it mean to 'learn'a syntactic construction? Looking at relative clause acquisition in H... more What does it mean to 'learn'a syntactic construction? Looking at relative clause acquisition in Hebrew, I argue for a usage-based approach where learning constructions involves a gradual expansion of uses (from formulaic to more productive,[1]) that is sensitive to the distributional patterns in the input, and a gradual reduction of lexical specificity [2].
Focusing on children's production of the dative alternation in English, we examine whether childr... more Focusing on children's production of the dative alternation in English, we examine whether children's choices are influenced by the same factors that influence adults' choices, and whether, like adults, they are sensitive to multiple factors simultaneously. We do so by using mixed-effect regression models to analyze child and child-directed datives extracted from the CHILDES corpus. Such models allow us to investigate the collective and independent effects of multiple factors simultaneously. The results show that children's choices are influenced by multiple factors (length of theme and recipient, nominal expression type of both, syntactic persistence) and pattern similarly to child-directed speech. Our findings demonstrate parallels between child and adult speech, consistent with recent acquisition research suggesting there is continuity between child and adult grammars. Furthermore, they highlight the utility of analyzing children's speech from a multi-variable perspective, and portray a learner who is sensitive to the multiple cues present in her input.
The dative alternation the pirate will send the necklace to the princess (PP) the pirate will sen... more The dative alternation the pirate will send the necklace to the princess (PP) the pirate will send the princess the necklace(DO)• dative verbs also display bias effects:• p (PP| send)> p (DO| send)• p (PP| show)< p (DO| show)• this cannot be reduced to a difference in meaning• the two constructions mean basically the same thing• controlling for semantic properties of the arguments, Bresnan et al (2007) still find e ects of verb bias
Are all children exposed to the same linguistic input, and do they follow the same route in acqui... more Are all children exposed to the same linguistic input, and do they follow the same route in acquisition? The answer is no: The language that children hear differs even within a social class or cultural setting, as do the paths individual children take. The linguistic signal itself is also variable, both within and across speakers - the same sound is different across words; the same speech act can be realized with different constructions. The challenge here is to explain, given their diversity of experience, how children arrive at similar generalizations about their first language. This volume brings together studies of phonology, morphology, and syntax in development, to present a new perspective on how experience and variation shape children’s linguistic generalizations. The papers deal with variation in forms, learning processes, and speaker features, and assess the impact of variation on the mechanisms and outcomes of language learning.
ABSTRACT Why are children better language learners than adults? Intuitively, there is a differenc... more ABSTRACT Why are children better language learners than adults? Intuitively, there is a difference between the unstructured (yet successful) way a child learns language and the effortful and often frustrating experience of trying to master a new language as an adult. On an empirical level, non-native speakers rarely reach native proficiency in pronunciation, morphological and syntactic processing, or in the use of formulaic language and idioms.
CUNY Conference on Human Sentence Processing. March, Mar 1, 2007
Speakers frequently have a choice among multiple ways of expressing one and the same thought. Whe... more Speakers frequently have a choice among multiple ways of expressing one and the same thought. When choosing between syntactic constructions for expressing a given meaning, speakers are sensitive to probabilistic tendencies for syntactic, semantic or contextual properties of an utterance to favor one construction or another. Taken together, such tendencies may align to make one construction overwhelmingly more probable, marginally more probable, or no more probable than another. Here, we present evidence that acoustic features of spontaneous speech reflect these probabilities: when speakers choose a less probable construction, they are more likely to be disfluent, and their fluent words are likely to have a relatively longer duration. Conversely, words in more probable constructions are shorter and spoken more fluently. Our findings suggest that the di¤ering probabilities of a syntactic construction in context are not epiphenomenal, but reflect a part of a speakers' knowledge of their language.
Relative clauses have been extensively studied in language acquisition due to their complex struc... more Relative clauses have been extensively studied in language acquisition due to their complex structure and the apparent difficulty children have with them. Their study has been made more intriguing by the well-documented asymmetry between subject and object relatives. It has long been noted that children comprehend subject relatives better than they do object relatives (Correa, 1995; Friedmann & Novogrodsky, 2004; de Villiers, 1979).
Design – 2 x 2 wh-FILLER (first wh-phrase: which NP vs. bare what) wh-INTERVENER (second wh-phras... more Design – 2 x 2 wh-FILLER (first wh-phrase: which NP vs. bare what) wh-INTERVENER (second wh-phrase: which NP vs. bare who) Manipulating the accessibility by choosing either the more accessible which-phrase or less accessible bare wh-phrases.
Poster presented at the Tenth International Congress …, 2005
In has long been noted, in a variety of languages (French, English, Hebrew) and tasks (act-out, p... more In has long been noted, in a variety of languages (French, English, Hebrew) and tasks (act-out, picture-selection) that children have difficulty in comprehending object relative clauses. Two studies examining the production and comprehension of both resumptive and non-...
Human infants and adults are able to segment coherent sequences from unsegmented strings of audit... more Human infants and adults are able to segment coherent sequences from unsegmented strings of auditory stimuli after only a short exposure, an ability thought to be linked to early language acquisition. Although some research has hypothesized that learners succeed in these tasks by computing transitional probabilities between syllables, current experimental results do not differentiate between a range of models of different computations that learners could perform. We created a set of stimuli that was consistent with two different lexicons—one consisting of two-syllable words and one of three-syllable words—but where transition probabilities would not lead learners to segment sentences consistently according to either lexicon. Participants’ responses formed a distribution over possible segmentations that included consistent segmentations into both two- and three-syllable words, suggesting that learners do not use pure transitional probabilities to segment but instead impose a bias towards parsimony on the lexicons they learn.
Uploads
Papers by Inbal Arnon
language acquisition. Although some research has hypothesized that learners succeed in these tasks by computing transitional probabilities between syllables, current experimental results do not differentiate between a range of models of different computations that learners could perform. We created a set of stimuli that was consistent with two different lexicons—one consisting of two-syllable words and one of three-syllable words—but where transition probabilities would not lead learners to segment sentences consistently according to either lexicon. Participants’ responses formed a distribution over possible segmentations that included consistent segmentations into both two- and three-syllable words, suggesting that learners do not use pure transitional probabilities to segment but instead impose a bias towards parsimony on the lexicons they learn.