Children use multiword frequency in real

Children use multiword frequency in real-time sentence comprehension
Arielle Borovsky1 Inbal Arnon2
State University, corresponding author: [email protected] , 2Hebrew University, Jerusalem
1Florida
Experiment 1: Familiar frame processing
Introduction
Children are sensitive to multiword information and can use this to
The task
Matthews, 2008; Arnon & Clark, 2011) and comprehension
Time course
Accuracy
support learning (Romberg & Saffran, 2010), production (Bannard &
Brush
(Borovsky, Elman, Fernald, 2012). However, it is not clear if this is
your
teeth
Brush
hair
your
driven by simple lexical associations summed between individual
words or by the combinatorial frequency of the entire sequence. We
focus on children’s sensitivity to multiword frequency to ask:
Brush
1.  Do children utilize multiword knowledge in online word
recognition?
2. 
her
teeth
Brush
her
hair
Object-VerbFrequency
High
Frequency
Are they are capable of rapidly forming novel multiword
Low
Frequency
Figure 2. Looks to target relative to the
distractor from verb onset to Target onset
FrameType
associations?
We test these questions in two visual-world eye-tracking studies
with 36 5-to-8-year-olds. The studies examine the effect of
phrase frequency vs. object-verb frequency on children’s online
processing.
Frequent(FF)
Brushyourteeth
Brushyourhair
Modified(MF)
Brushherteeth
Brushherhair
Figure 1. Visual-world-paradigm task.
Participants viewed pairs of images
while hearing sentences in one of four
experimental conditions.
Children were sensitive to object-verb
frequency (β = .12, p < .01), not frame
frequency (β = -.06, p = .8). Faster to
high frequency object-verb
associations in both frequent and
modified frames (brush your teeth and
brush her teeth)
Figure 3. Time course of fixations proportions
to Target and Distactor images across
experimental conditions (in 50 ms bins). Error
cloud represents +/- 1 SEM .
Anticipatory gaze to target object following
frequent object-verb associations.
Experiment 1 asks if children’s object recognition is facilitated
Experiment 2: Novel frame processing
by multiword knowledge. To distinguish between object-verb
frequency and phrase frequency we cross object-verb
frequency (high-frequency: brush X teeth vs. low-frequency
The task
brush X hair) and frame frequency (Frequent-frame: brush
Time course
Training phase
your teeth vs. Modified frame: brush her teeth)
Test phase
Anticipatory
Window
The monkey
•  Prediction: if children utilize multiword frequency,
recognition should be facilitated following frequent vs.
rides the
bus.
1350 ms
modified frames
Experiment 2 asks if children can form novel multiword
associations that can facilitate processing after controlling for
The monkey eats the candy
The dog rides in the car.
verb-object frequency. We expose children to multiword
combinations using a training story where two sentential objects
(apple & candy) are equally predicted by summed lexical
Summed association frequencies:
Two (action-related) objects equal
associations (monkey+eat), but only a single outcome (candy) is
predicted by multi-word frequency
•  Prediction: if children utilize multiword frequency,
Object-Verb
Frequency
Combinatory
Frame
Object-Verb
Frame
recognition should be facilitated based on phrase frequency
High
Monkey-ride-bus
Ride-bus
Low
Monkey-eat-candy
…
and NOT object-verb frequency.
High
Dog-eat-apple
Eat-apple
Low
Dog-ride-car
….
Agent+
Ac:on=
Object
Monkey
Eats
candy
Busx1
Candyx1
Applex2,
Candyx1
Candyx2
Applex2
Busx1
Dog
Ridesin
Car
Applex1
Carx1
Busx2
Carx1
Busx2
Carx2
Applex1
Children anticipate combinatorial outcome:
form predictions about multiword sequence.
Discussion
References
Arnon, I., & Clark, E. V. (2011). Why brush your teeth is better than
teeth–Children's word production is facilitated in familiar sentenceframes. Language Learning and Development, 7(2), 107-129.
Bannard, C., & Matthews, D. (2008). Stored word sequences in
language learning the effect of familiarity on children's repetition of
four-word combinations. Psychological science, 19(3), 241-248.
Borovsky, A., Elman, J. L., & Fernald, A. (2012). Knowing a lot for
one’s age: Vocabulary skill and not age is associated with
anticipatory incremental sentence interpretation in children and
adults. Journal of experimental child psychology, 112(4), 417-436.
Romberg, A. R., & Saffran, J. R. (2010). Statistical learning and
language acquisition. Wiley Interdisciplinary Reviews: Cognitive
Science, 1(6), 906-914.
Figure 4. Time course of fixations towards the
Target and competitor items from sentence
onset to sentence offset. The error bar clouds
represent +/- 1 SEM. . Arrow indicates first
time bin where Target fixations significantly
diverges from equally-frequent Action-related
object (using cluster-based permutation
procedure; cluster t = 106.1, p < .0001).
Experiment 1
•  Children used strength of association between the verb and sentential object to anticipate upcoming nouns.
•  No sensitivity to the frame (phrase) frequency.
•  Experimental limitations may drive lack of frame effect:
•  Naturalistic differences in frame frequency may have been too subtle to drive processing
•  Duration of pronoun/article may not have left enough time to influence processing in advance of the spoken noun.
Experiment 2
•  Clear evidence for anticipation of combinatory outcome even when summed lexical outcome was matched
Acknowledgements
We are grateful to the families and children who participated in this
research. Angele Yazbec and the undergraduate research assistants at
the FSU Language and Cognitive Development Lab (http://lcdlab.com)
also contributed to the testing and data entry on this project.
This research was supported by a grant from the NIH to AB:
DC013638.
•  Future work needed to address whether and how summed lexical frequency may over-ride combinatorial frequency
•  Together, these findings suggest that children use more than just pair-wise associations to support real-time sentence
comprehension (though the effects may be subtle in natural language processing) and highlight the need to account
for the influence of multi-word chunks in models of language processing