Lexical Integration Without Meaning Or Sleep
Efthymia C. Kapnoula, Stephanie Packard, Keith S. Apfelbaum, Bob McMurray, & Prahlad Gupta
Method (continued)
Procedure:
Repetition
Stem completion
sook
Training:
• What kind of information is required?
• 10 were left untrained
By examining these factors we get closer to understanding the process
of novel word learning.
• 10 (of the 20) non-words were trained
0.9
soo…
sook
• What kind of training is required?
1
sook
0.8
• Which 10 counterbalanced across participants
• Alternating repetition and stem completion blocks (x11)
Background
• Each non-word was presented once in every block.
• Learning a novel word entails the development of such properties.
• Specifically, learning a novel word entails linking its phonological
word-form with:
suit
• Its sub-lexical features
• Other lexical representations
sook
/s/
/u/
Testing:
• Visual World Paradigm:
measure of real time activation
+
/s/ /u/
/t/
• What does it take for these links to form?
• Gaskell & Dumay (2003): meaning is not necessary for a newly
learned word (e.g. cathedruke) to inhibit a similar known word (e.g.
cathedral).
/s/ /u/
/p/
• Lindsay & Gaskell (2012): sleep is not necessary.
/s/ /u/
=
/s/ /u/
/s/ /u/
/t/
suit
/k/
+
Word-splice
+
/s/ /u/
/s/ /u/
/t/
/t/
=
/s/ /u/
suit
•
800
1000
1200
22 blocks (single tokens) – 36 subs
1
1400
1600
1800
2000
/s/ /u/
22 blocks (multiple tokens) – 28 subs
11
0.9
0.9
0.8
0.7
0.6
0.5
Matching-splice
0.4
Untrained-non-word-splice
0.3
Trained-non-word-splice
0.2
Word-splice
*
ns
0.8
0.8
0.7
0.7
0.6
0.6
0.5
0.5
Matching-splice
Matching-splice
0.4
0.4
Untrained-non-word-splice
Untrained-non-word-splice
0.3
0.3
Trained-non-word-splice
Trained-non-word-splice
0.2
0.2
Word-splice
Word-splice
0.1
0.1
00
0
500
/t/
soup
/sup/
suit
/t/
Word-splice:
Slow activation
?
/suk/
suit
late
•
600
0
early
/t/
Matching-splice:
Fast activation
Subjects: 36 undergraduate students
400
0.1
/t/
late
/sut/
200
0.9
=
early
Method
soup
ns
1000
Time in ms
1500
2000
00
500
500
1000
1000
Time in ms
1500
1500
2000
2000
p = .032
• But not significant difference when training is longer (33 blocks)
Conclusions
late
early
• Is interleaved exposure to neighboring words necessary?
suit
0.2
*
• Trained-non-word-splice was not significantly different from wordsplice (p = .695).
/t/
Predictions:
sook
Trained-non-word-splice
• Only phoneme-monitoring in training (no production)
• The splicing creates a co-articulatory mismatch (CAM) between the
vowel and the final consonant.
• Can competition effects arise immediately after training?
• 20 triplets in total
0.3
Work in progress:
Non-word-splice condition
• effect only showed up 2 days after the first exposure.
• Each has two real words differing only in the place of articulation
of the final stop consonant.
Untrained-non-word-splice
• Looks to the target were significantly slower in the trainedcompared to the untrained-non-word-splice condition (p = .014).
Word-splice
Materials: 20 monosyllable non-words (e.g. sook)
0.4
Time in ms
Matching-splice
Question:
Remaining questions:
Matching-splice
0
• Following the Dahan, Magnuson & Tanenhaus (2001) paradigm, the
target word was presented in three different splicing conditions:
• Once learned, these “linkages” allow the newly learned word to
interact with its features and with other words (e.g. inter-lexical
inhibition).
• But interleaved exposure to neighboring known words is.
0.5
0
suit
soup
/k/
0.6
0.1
Proportion of looks to the target
• Words have the property of interacting with other words and with
phonological representations.
0.7
Proportion
target
the target
to the
looks to
of looks
Proportion of
What does it take for a novel word-form to become lexically integrated?
Results
Proportion of looks to the target
Theoretical motivation
/t/
Non-word-splice:
Medium Activation
When the target is spliced with another word, CAM partially
activates the competitor, inhibiting the target.
When the target word is spliced with a non-word we expect
minimal: CAM does not cue another word
Question:
• Will trained non-words create inhibition (like known words), or not
(like non-words)?
• CAM cuing a trained novel word slowed down activation of the
target word, whereas the exact same physical stimulus did not have
the same effect when the CAM cued an untrained non-word.
• This indicates that lexical engagement does not require:
• Meaning.
• Time-consuming lexical consolidation (with or without sleep).
• Interleaved exposure to old and new words.
• Lexical competition is a form of lexical engagement that plays a
critical role in resolving acoustic ambiguity (McMurray, Tanenhaus &
Aslin, 2009; McMurray et al, 2009). Our results indicate that this
competition can stem from a minimally experienced phonological
sequence that is not semantically integrated.
References
•
•
•
•
•
Dahan, D., Magnuson, J.S., Tanenhaus, M.K., & Hogan, E. (2001). Subcategorical mismatches and the time course of lexical access: Evidence for lexical competition. Language and Cognitive Processes, 16(5/6), 507-534.
Gaskell, M. G., & Dumay, N. (2003). Lexical competition and the acquisition of novel words. Cognition, 89, 105-132.
Lindsay, S. & Gaskell, M. G. (2012). Lexical integration of novel words without sleep. Journal of Experimental Psychology: Learning, Memory & Cognition, X(X), XXX-XXX .
McMurray, B., Clayards, M. A., Tanenhaus, M. K., & Aslin, R. N. (2008). Tracking the time course of phonetic cue integration during spoken word recognition. Psychonomic Bulletin and Review, 15, 1064-1071.
McMurray, B., Tanenhaus, M. K., & Aslin, R. N. (2009). Within-category VOT affects recovery from "lexical" garden paths: Evidence against phoneme-level inhibition. Journal of Memory and Language, 60, 65-91.
53rd Annual Meeting of the Psychonomic Society
[email protected]