Habituation Phase

• Nonce words were combined pseudo-randomly to create 3 sets of 3 trigrams
• Each set implemented a unique NAD across three exemplars, each with a different intervening item

​

​

​

• A cartoon drawing of a baby stayed on screen, accompanying the audio stimuli, to sustain infants’ attention

For additional information

helens.lu@ubc.ca

How Language Experience Shapes Sensitivity to Non-adjacent Dependencies in 12-month-old Infants

Helen Shiyang Lu^a, Olesia Bokhanovich^a, Toben H. Mintz^a,b

Department of Psychology, University of Southern California^a; Department of Linguistics, University of Southern California^b

Introduction

Methods

Results

Online Setup

• Structurally related elements in language often occur in linearly non-adjacent relationships

“the cat is (desperately) trying to catch the laser dot”

• Learning non-adjacent dependencies (NADs) may benefit the acquisition of word categories^1-3, and it is linked to later vocabulary development⁴
• Twelve-month-old English-hearing infants were sensitive to distributional-grammatical patterns in English involving non-adjacent relationships^5-6
• However, behavioral evidence of NAD learning has only been found with older infants⁷
• This study aims to investigate 12-month-old infants’ ability to learn NADs from an artificial language
• Language-specific features may impact learning:
• Spanish has rich concord morphology (e.g., “los perros lindos”) with rhyming inflections across phrases
• Concord morphology consistent with phrase structure can aid artificial syntax learning⁸
• Phonological similarity in segments may support NAD learning^9-10
• Bilingual children show greater success in learning statistical regularities than monolinguals, especially in more challenging contexts^11-13
• We compare Spanish-monolingual, English-monolingual, and Spanish-English bilingual infants to tease apart effects of bilingualism and of language-specific input

a {X₁, X₂, X₃} b

c {X₄, X₅, X₆} d

e {X₇, X₈, X₉} f

Test Phase

• 3 grammatical (G) trigrams

e.g.,

• 3 ungrammatical (U) trigrams

e.g.,

• Infants had not heard the exact combinations of words before
• Grammatical and ungrammatical trigrams only differed in their adherence to NAD relations

​

​

​

a X₄ b c X₇ d e X₁ f

a X₄ f c X₇ b e X₁ d

• We used an infant-controlled habituation procedure¹⁴, adapted for online testing environments

Stimuli

• 15 monosyllabic nonce words were recorded by a female speaker

e.g., “bep”, “bok”, “des”, “dob”, “feep”, “foom”, “ghan”...

• Each word was then modified to last 250 ms and had a 83-ms silence added to its end to increase intelligibility

A 250-ms silence was then added in between trigrams

Note: Each letter above represents one of the 15 nonce words.

Habituation Phase

• Infants spent, on average, 133.9 seconds (SD = 57.7) in the habituation phase

English-monolinguals (M = 133, SD = 52.8)

Spanish-monolinguals (M = 145, SD = 73.1)

Bilingual English-Spanish (M = 110 , SD = 44.8)

Test Phase

​

​

​

​

​

​

​

​

​

​

​

• English monolingual infants looked longer to U trials compared to G trials (β = 0.04, 89%CI [0.004, 0.08], PD 96.4% > 0)
• Some evidence for a weaker grammaticality effect in bilinguals than in English monolinguals (β = -0.05, 89%CI [-0.12, 0.02], PD 89.0% < 0)
• No credible evidence that the grammaticality effect differed between Spanish and English monolinguals (β = 0, 89%CI [-0.10, 0.10], PD 49.6% < 0)
• Bilingual infants spent less time on test trials compared to English monolinguals (β = -0.16, 89%CI [-0.30, -0.02], PD 96.3% < 0)

Findings suggest a role for language experience in NAD learning, though the specific contributions of bilingualism and language-specific input remain to be clarified in future work with larger, balanced samples.

Participants

• 112 twelve-month-old infants were included in the current analysis
• Their language exposure was estimated using the LEAT¹⁵

​

​

​

​

​

• Additional 118 infants participated in the study but were excluded due to the following reasons

fussiness (50), technical issues (16), distraction (9), falling asleep (1), sickness (1), premature birth (14), language exposure (24), and insufficient valid test trials (3)

References: 1) Mintz, T. H. 2006. Frequent frames: Simple co-occurrence constructions and their links to linguistic structure. 2) Mintz, T. H. 2003. Frequent frames as a cue for grammatical categories in child directed speech. 3) Chemla, E., Mintz, T. H., Bernal, S., & Christophe, A. 2009. Categorizing words using ‘frequent frames’: what cross‐linguistic analyses reveal about distributional acquisition strategies. 4) Frost, R. L., Jessop, A., Durrant, S., Peter, M. S., Bidgood, A., Pine, J. M., ... & Monaghan, P. 2020. Non-adjacent dependency learning in infancy, and its link to language development. 5) Geffen, S., & Mintz, T. H. 2015. Can you believe it? 12-month-olds use word order to distinguish between declaratives and polar interrogatives. 6) Mintz, T. H. 2006. Finding the verbs: distributional cues to categories available to young learners. 7) Gómez, R., & Maye, J. 2005. The developmental trajectory of nonadjacent dependency learning. 8) Morgan, J. L., Meier, R. P., & Newport, E. L. 1987. Structural packaging in the input to language learning: Contributions of prosodic and morphological marking of phrases to the acquisition of language. 9) Newport, E. L., & Aslin, R. N. 2004. Learning at a distance I. Statistical learning of non-adjacent dependencies. 10) Grama, I. C., Kerkhoff, A., & Wijnen, F. 2016. Gleaning structure from sound: The role of prosodic contrast in learning non-adjacent dependencies. 11) Kovács, Á. M., & Mehler, J. 2009. Flexible learning of multiple speech structures in bilingual infants. 12) de Bree, E., Verhagen, J., Kerkhoff, A., Doedens, W., & Unsworth, S. 2016. Language learning from inconsistent input: Bilingual and monolingual toddlers compared. 13) Verhagen, J., & de Bree, E. 2023. Non-adjacent dependency learning from variable input: investigating the effects of bilingualism, phonological memory, and cognitive control.14) Kominsky, J. F. 2019. PyHab: Open-source real time infant gaze coding and stimulus presentation software. 15) DeAnda, S., Bosch, L., Poulin-Dubois, D., Zesiger, P., & Friend, M. 2016. The language exposure assessment tool: Quantifying language exposure in infants and children.

Acknowledgements: This research project is funded by the Doctoral Dissertation Research Improvement Grant (BCS-2234422) from NSF Linguistics Program and by the Doctoral Research Grant from USC’s Department of Psychology.

​

Bayesian Mixed-effects Model: LookingTime ~ grammaticality * language group + (grammaticality | child)

*Error bars represent 95% credible intervals around the prediction means.