Introduction to Psycholinguistics

I presented a talk on a research paper published by Ms Jenny Saffran then at University of Wisconsin-Madison in Cornell University. This talk is for an elective subject Introduction to psycholinguistics, I chose during my masters.

Transcript

1. Words in a sea of sounds: the output of infant statistical learning Cognition 81 (2001) 149-169, Author : Jenny R. Saffran Presentation by : Mahak Gupta
2. About the author •Jenny Saffran is a Professor of Psychology at the University of Wisconsin, Madison. •Interests - Language acquisition and early cognitive development, and she also conducts research on music cognition.
3. Overview for this seminar 1.Introduction 2.Set of Experiments conducted and results a.Experiment 1 b.Experiment 2 c.Experiment 3 3.Conclusion
4. Introduction Google Images
5. Introduction •Motivation: How do infants learn about “new” words?
6. Introduction – Background and pre-requisites for Understanding of this paper -Statistical learning: Patterns in sounds exposed to infants -Statistical Learning by 8 month old infants (1996): è Experiment 1: è Research Question: Do Infants treat words and part words differently ? è Subject: 24, 8 month old infants with American english language environment. -Words : pabiku, tibudo, golatu, daropi; Nonwords: Different words with same syllables in above set; Part-words: tudaro, pigola, bikuti, budopa. Familiarization: Sound Clip -Procedure : Infants were exposed to trials with a sound sequence of 2 words and 2 nonwords from above set. -Result : The listening time was different for words and non words; more time for non words. http://www.atsweb.neu.edu/hlittlefield/CourseDocs/Acq/SaffranEtAl-1996.pdf
7. Introduction – Background and pre-requisites for Understanding of this paper (Cont'd) è Experiment 2: è Research Question: Can infants compute more difficult “statistical computations” to distinguish words from syllable strings spanning word boundaries ? For e.g in pretty#baby, If infants can distinguish between ty#ba è Subject: 24, 8 month old infants with American english language environment. -Words : pabiku, tibudo, golatu, daropi; Nonwords: Different words with same syllables in above set; Part-words: tudaro, pigola, bikuti, budopa. -Procedure : Infants were exposed to trials with a sound sequence of 2 words and 2 part words from above set. -Result : The listening time was different for words and part words; more time for part words. http://www.atsweb.neu.edu/hlittlefield/CourseDocs/Acq/SaffranEtAl-1996.pdf
8. Introduction – Learning from previous experiments •Infants have an amazing capability for word segmentation i.e discovering the boundaries between words. •Some factors influencing the direction of infant preferences are infant's age, the length of the delay between exposure and test, the closeness of the match between exposure and test. •The number of times an item is repeated or processed during familiarization can lead to either familiarity preferences after relatively little exposure, or novelty preferences after a greater number of exposure
9. But... What is the difference in Experiments conducted earlier and we're going to study now ? •In the previous experiments statistical learning was tested on exposing infants to strings of some words. •In the new experiments – Statistical learning is tested on exposing infants to some novel word-like units, rather than “probabilistically-related” strings of sounds.
10. Experiment 1 - Do Infants treat words and part- words differently ? Subject - 56 infants aged 8 month, Monolingual Half of them were assigned to English frames and half to nonsense frame conditions Target words – Condition A : pabiku, tibudo, golatu, daropi; Condition B : tudaro, pigola, bikuti, budopa Stimuli - Targets in the present test were placed at the ends of sentence frames. Familiarization – Sound clip from 1996 experiment. No. of Trials – 12 672 total trials with 10 outliers (Listening time < 1.5 sec)
11. Experiment 1 (Cont’d) Procedure: Blinking Light Blinking Light + Sound Blinking Light + Sound http://consumertraveler.com/wp-content/uploads/lap_baby.jpg
12. Experiment 1 - Results l Infants tested on targets placed in English sentential frames preferred to listen to the sentences containing words over the sentences containing part-words.
13. Experiment 2 - Can infants discriminate b/w common English phrases from NS phrases that resemble English ? Subject - 20 infants, aged 8 month, Monolingual Half of them were assigned to English frames and half to nonsense frame conditions Stimuli - Same four English Phrases and their equivalent Nonsense phrases - but without words and part-words. For e.g : I like my... ; Zy fike ny... etc Procedure - No change except there was no Familiarization. No. of Trials – 4 Jusczyk, Friederici et al. (1993) http://www.apa.org/pubs/journals/releases/xhp273644.pdf
14. Experiment 2 - Results l Preference for the nonsense frames over the English frames is a processing advantage for English materials relative to nonsense materials. l Infants find repeated English fragments as less interesting than repeated nonsense fragments because they are processed and acquired more rapidly.
15. Experiment 3 - Are familiarity with test frames influencing results ? Subject - 40 Infants aged 8 month, Monolingual Stimulus – Same English frame and same tone frame for all the four sentences in test corpus Procedure - Similar to Experiment 1
16. Experiment 3 - Results l Increased familiarity in test frames has increased the listening time of infants as compared to the experiment 1.
17. Result Comparison – Experiment 1 vs Experiment 3 0 1 2 3 4 5 6 7 8 9 10 E1 - English Frame E1 - Nonsense Frame E3 - English Frame E3 - Tone Frame words part words
18. Conclusion •We’ve seen how Infant statistical learning mechanisms are engaged in syllable tracking sequences of word segmentation tasks and this was influenced by infants listening preferences. •Infants treat words and part words differently as there's a difference in listening time with same English/NS phrases and changing words and part words. (Experiment 1) •Infants are able to discriminate b/w the English phrases with NS phrases which resemble english (Experiment 2) •The familiarity with test frames used increases the listening time of infants for the same set of words and part words. (Experiment 3)
19. References è Statistical Learning by 8 month old . J. Saffran, Richard Aslin, Elissa newport. è Words in a sea of sounds:the output of infant statistical learning. J. Saffran è Word Segmentation: The Role of Distributional Cues. J. Saffran, Richard Aslin, Elissa newport. è Ted talk by Patricia Kuhl - http://www.ted.com/talks/patricia_kuhl_the_linguistic_genius_of_babies?language=en è Google Images è Wikipedia; Statistical learning, word segmentation,