Processing Ambiguous Words: Multi-sense Activation Swinney 1979 Tanenhaus 1979 presented by Shauna Eggers 1980 and Steven Moran 1979.

Slides:



Advertisements
Similar presentations
Critical Reading Strategies: Overview of Research Process
Advertisements

YOU CANT RECYCLE WASTED TIME Victoria Hinkson. EXPERIMENT #1 :
Accessing spoken words: the importance of word onsets
Timing of the brain events underlying access to consciousness during the attentional blink Claire Sergent, Sylvain Baillet, & Stanislas Dehaene.
All slides © S. J. Luck, except as indicated in the notes sections of individual slides Slides may be used for nonprofit educational purposes if this copyright.
Marslen-Wilson Big Question: “What processes take place during the period that the sensory information is accumulating for the listener” during spoken.
Human Speech Recognition Julia Hirschberg CS4706 (thanks to John-Paul Hosum for some slides)
PSY 369: Psycholinguistics Language Production: Models cont.
Evaluating the Effect of Neighborhood Size on Chinese Word Naming and Lexical Decision Meng-Feng Li 1, Jei-Tun WU 1*, Wei-Chun Lin 1 and Fu-Ling Yang 1.
The Interaction of Lexical and Syntactic Ambiguity by Maryellen C. MacDonald presented by Joshua Johanson.
Critical Thinking Course Introduction and Lesson 1
Infant sensitivity to distributional information can affect phonetic discrimination Jessica Maye, Janet F. Werker, LouAnn Gerken A brief article from Cognition.
Lexical Ambiguity in Sentence Comprehension By R. A. Mason & M. A. Just Brain Research 1146 (2007) Presented by Tatiana Luchkina.
Prosodic facilitation and interference in the resolution of temporary syntactic closure ambiguity Kjelgaard & Speer 1999 Kent Lee Ψ 526b 16 March 2006.
Spoken Word Recognition 1 Language Use and Understanding.
Using prosody to avoid ambiguity: Effects of speaker awareness and referential context Snedeker and Trueswell (2003) Psych 526 Eun-Kyung Lee.
9/22/10Psyc / Ling / Comm 525 Fall 10 Semantic Priming (Phenomenon & Tool)...armkitchentree Related prime >doctoractor < Unrelated prime nurse floor...
I Like Myself but I Don’t Know Why: Enhancing Implicit Self Esteem by Subliminal Evaluative Conditioning Author: A.P Dijkserhuis.
The mental lexicon LG 103 Introduction to psycholinguistics Celia (Vasiliki) Antoniou.
Auditory Word Recognition
Language Use and Understanding BCS 261 LIN 241 PSY 261 CLASS 12: SNEDEKER ET AL.: PROSODY.
PS: Introduction to Psycholinguistics Winter Term 2005/06 Instructor: Daniel Wiechmann Office hours: Mon 2-3 pm Phone:
Psycholinguistic methodology Psycholinguistics: Questions and methods.
Hemispheric asymmetries in the resolution of lexical ambiguity Jeffrey Coney, Kimberly David Evans Presented by Chris Evans May 17, 2006.
Validity, Sampling & Experimental Control Psych 231: Research Methods in Psychology.
Validity, Sampling & Experimental Control Psych 231: Research Methods in Psychology.
Writing with APA style (cont.) & Experiment Basics: Variables Psych 231: Research Methods in Psychology.
Subliminal Perception Zoltán Dienes Conscious and unconscious mental processes.
Writing with APA style (cont.) & Experiment Basics: Variables
PSY 369: Psycholinguistics
Abstract Cognitive control processes reduce the effects of irrelevant or misleading information on performance. We report a study suggesting that effective.
Sound and Speech. The vocal tract Figures from Graddol et al.
Wilson, “The case for sensorimotor coding in working memory” Wilson’s thesis: Items held in short-term verbal memory are encoded in an “articulatory” format.
Psycholinguistics 05 Internal Lexicon.
PSY 369: Psycholinguistics Language Acquisition: Bilinugalism.
Multiple Linear Regression A method for analyzing the effects of several predictor variables concurrently. - Simultaneously - Stepwise Minimizing the squared.
Human Memory What we usually think of as “memory” in day-to-day usage is actually long-term memory, but there are also important short-term and sensory.
Syntax Lecture 8: Verb Types 1. Introduction We have seen: – The subject starts off close to the verb, but moves to specifier of IP – The verb starts.
Chapter 9 Comparing More than Two Means. Review of Simulation-Based Tests  One proportion:  We created a null distribution by flipping a coin, rolling.
WORD SEMANTICS 4 DAY 29 – NOV 4, 2011 Brain & Language LING NSCI Harry Howard Tulane University.
Chapter 7 Experimental Design: Independent Groups Design.
IN THE NAME OF ALLAH,THE MOST MERCIFUL AND BENEFICIENT TO MANKIND CHAPTER NO 4 WORKING MEMORY WAQAS SULTANI.
Lexicon Organization: How are words stored? Atomist view  Words are stored in their full inflected form  talk –> talk  talked –> talked  toothbrush.
MEMORY. Sensory Memory Sensory Memory: The sensory memory retains an exact copy of what is seen or heard (visual and auditory). It only lasts for a few.
Systematic reviews to support public policy: An overview Jeff Valentine University of Louisville AfrEA – NONIE – 3ie Cairo.
1 Compiler Construction (CS-636) Muhammad Bilal Bashir UIIT, Rawalpindi.
Avoiding the Garden Path: Eye Movements in Context
Questions to Ask Yourself Regarding ANOVA. History ANOVA is extremely popular in psychological research When experimental approaches to data analysis.
AMSc Research Methods Research approach IV: Experimental [1] Jane Reid
N400-like semantic incongruity effect in 19-month-olds: Processing known words in picture contexts Manuela Friedrich and Angela D. Friederici J. of cognitive.
1 Cross-language evidence for three factors in speech perception Sandra Anacleto uOttawa.
Issues concerning the interpretation of statistical significance tests.
The effects of working memory load on negative priming in an N-back task Ewald Neumann Brain-Inspired Cognitive Systems (BICS) July, 2010.
The Cross-Script Length Effect: Evidence for Serial Processing in Reading Aloud Kathleen Rastle (Royal Holloway University of London), Linda Bayliss (Royal.
Research Topics in Memory
REFERENCES Bargh, J. A., Gollwitzer, P. M., Lee-Chai, A., Barndollar, K., & Troetschel, R. (2001). The automated will: Nonconscious activation and pursuit.
LEXICAL INTERFACE 5 NOV 2, 2015 – DAY 28 Brain & Language LING NSCI Fall 2015.
Neural correlates of morphological decomposition in a morphologically rich language : An fMRI study Lehtonen, M., Vorobyev, V.A., Hugdahl, K., Tuokkola.
1-1 Copyright © 2015, 2010, 2007 Pearson Education, Inc. Chapter 23, Slide 1 Chapter 24 Comparing Means.
1 Vocabulary acquisition from extensive reading: A case study Maria Pigada and Norbert Schmitt ( 2006)
Models of Production and Comprehension [1] Ling4-437.
Test Loads Andy Wang CIS Computer Systems Performance Analysis.
VISUAL WORD RECOGNITION. What is Word Recognition? Features, letters & word interactions Interactive Activation Model Lexical and Sublexical Approach.
IINDIVIDUAL LEARNING STYLE IN LANGUAGE LEARNING. Most children and adults can master some content - how they master, it is determined by individual learning.
Phonological Priming and Lexical Access in Spoken Word Recognition
Levels of Processing Memory Model (LoP)
Lexical interface 5 Nov 1, 2017 – DAY 27
Office of Education Improvement and Innovation
Phonological Priming and Lexical Access in Spoken Word Recognition
Presentation transcript:

Processing Ambiguous Words: Multi-sense Activation Swinney 1979 Tanenhaus 1979 presented by Shauna Eggers 1980 and Steven Moran 1979

Outline Background and Motivation Tanenhaus 1979 –Experiment –Results Swinney 1979 –Experiment 1 –Experiment 2 –Results Discussion

Motivation Sentence ambiguity is resolved by context –Obviously! but... How exactly does it work? –How and when do contexts come into the comprehension process? Focus: lexical access –how are contexts used for lexical disambiguation? –especially: when context occurs right before ambiguity

The Question Does contextual info interact with lexical info during the “access phase”? Two possible scenarios: –Comprehension processes are highly interactive, so contextual information is used to select appropriate word senses during lexical access –Comprehension processes are isolable and autonomous, and context is only used to select word senses after all lexical info is accessed

Two Hypotheses Prior Decision (“unitary perception model”) –Context directs lexical access such that only a single, relevant reading is ever accessed for an ambiguous word –Only one word sense is ever considered Post Decision (“multiple-meaning model”) –Context comes into play only after complete access of all info about a word –All word senses are considered

Some Assumptions Assumption 1: ambiguity increases “processing complexity” Assumption 2: processing complexity increases processing time Assumption 3: these effects hold even though most people eventually become aware of only a single meaning of an ambiguous word

Previous Studies Measure processing time!  this should give us a peek into when and how ambiguity resolution processes kick in Basic idea: stimulate with some ambiguity, measure time to select appropriate word sense Common method: Phoneme Monitoring Task

Phoneme Monitoring Task Measure length of time to process lexical ambiguity using in-line detection measures Decisions about phoneme detection following ambiguities (at various distances) reflect processing complexity How exactly it works... ?? Anyone know?

Problems with PMT Is the ol' PMT really appropriate? Problem 1: What exactly is it measuring? –Lexical access? Or some process after lex access? Hard to say... Problem 2: Confounding variables –Length and phonological properties of initial phoneme of ambiguity, its control, and the word preceding the ambiguity

Alternatives Increase sensitivity of experimental task Properties of ideal task: –Reflect access of each meaning of ambig word –Minimize temporal gap between ambig word and measure of lex access –Applicable during sentence comprehension, not just after it's already been processed Which brings us to...

Cross-Modality Semantic Priming Combine auditory and visual stimuli: –Couple auditory presentation of an ambiguous sentence with a visual lex decision task –Auditory stimuli can facilitate visual lexical decisions In English: hearing a word should prime for visual recognition of related words Assumption: lex decision reaction time reflects processing of ambiguous words

Characteristics of CMSP Holds when the primed (facilitated) word is presented visually during auditory sentence comprehension Subjects in task not necessarily aware of relationship between visually presented material and auditory material Can occur as an automated process, not under control of concious direction

Advantages of CMSP Ambiguous and target words can be presented simultaneously –Avoids “distance problems” seen in PMT Can be used during rather than after comprehension Minimizes attention drawn to experimental variables Can measure activation of each meaning of an ambiguous word

Experimental Hypothesis Prior Decision: –If strong biasing context causes only a single reading of an ambig word to ever be accessed, only lexical decisions related to that reading will be primed for. Post Decision: –If all readings of an ambiguous word are accessed, even in a strong biasing context, visual words related to each reading will be facilitated (primed for).

Evidence for Multiple Stages in the Processing of Ambiguous Words in Syntactic Contexts Michael Tanenhaus, James Leiman, and Mark Seidenberg Wayne State U. & Columbia U. Journal of Verbal Learning and Verbal Behavior

Outline Overview of study Experiment setup Methods Procedure Results Conclusions Questions

Tanenhaus et al. study variable delay naming paradigm used to study the processing of noun-verb ambiguities in sentences words used were ambiguous between two semantically distinct meanings: one noun, one verb these contexts provide syntactic information which is (typically) compatible with only one reading not like ‘organ’ and ‘organ’

Example Verb reading: John began to tire. Noun reading: John brought a rose.

Experiment setup Subjects heard the sentences Sentences were followed by the presentation of a single word on a screen Task: to name the target word aloud

Target words were: either related to meaning of the ambiguous word and biased by the context –‘They began to sink’ -- swim or related to the nonbiased meaning –‘They needed a new sink’ -- swim or unrelated to either meaning –‘They needed a new joke’ -- swim

Target words also: appeared following the sentence final ambiguous word appeared at three ‘stimulus onset asynchronies’ (intervals after sentence) at: 0ms, 200ms, 600ms latencies (the intervals)

Hypotheses 1 Tanenhaus ‘79: ‘If listeners only access a particular meaning of an ambiguous word, then latencies to name a word related to that meaning should show facilitation relative to unrelated controls.’ Translation in ’06: If listeners access only one meaning of an ambiguous word then only words related to that meaning should be primed for.

Hypothesis 2 Tanenhaus ’79: If, however, listeners access both meanings of an ambiguity, there should be equivalent facilitation to both related targets. Translation ‘06: If both senses are activated for an ambiguous word, all words related to either sense should be primed for.

Method involved 60 Wayne State U. students list of 24 ambiguous words with independent noun and verb readings (e.g. watch) two sentences constructed for each word, one assigned a noun reading, the other a verb reading ambiguous word always last word in sent.

Example sentences Noun reading: I bought a watch. Verb reading: I will watch.

Control sentences two control sentences also constructed for each ambiguous word neutral word was substituted for the ambiguous word, e.g.: I bought the cake. I will park.

Sentences four sentences associated with each ambiguous word sentence with ambiguous noun reading sentence with ambiguous verb reading sentence with control noun reading sentence with control verb reading 96 stimulus sentences total

Target target word was assigned to each ambiguous word target was either a synonym (or ‘associate’) of the ambiguous word’s noun or verb reading half of the ambiguous word received noun targets other half received verb targets

Subjects in action 96 stimulus sentences divvied into 4 blocks subjects were randomly assigned to one stimulus onset asynchrony (0, 200, 600ms) throughout the experiment were given 10 practice sentences that weren’t ambiguous to start with were then let loose to the sentence hounds experiment lasted 40 minutes

Procedure subjects listened to each sentence over head phones target word was then projected onto a screen subjects instructed to read the target word as quickly as possible into microphone

Timing timing tone at end of each sentence initiated an interval timer (0, 200, or 600ms) after the target word appeared a millisecond clock began timing the subject subjects’ responses triggered the clock to stop

Results (first the errors) there were 5760 naming latencies 415 (7.20%) were missing 308 due to subject not speaking loud enough 79 due to late 70’s mechanical failure 17 due to experimenter incompetence 11 were the subject saying the wrong word (duh)

Sentence Target Conditions target conditions were divided into three factors: target type, ambiguity, congruency target type – whether or not the target was related to the noun or verb reading of the ambiguous word –verb: ‘They all rose’ –noun: ‘She held the rose’ ambiguity type – whether or not the sent. ended with an ambiguous word –yes: ‘The table was difficult to sand’ –no: ‘The table was difficult to nick’ congruency – relationship between the sent. and target –congruent: ‘She held the rose’ – flower –incongruent: ‘They all rose’ -- flower

Analysis separate ANOVAs (Analysis of Variance) were performed treating subject and target words as random factors ANOVA – test that measures the difference between the means of two or more groups (also called ‘F test’... as in ‘f it’?)

Results 1 at 0ms stimulus onset asynchrony, priming occurred for all possible readings of an ambiguous word no regard to target word being related to contextually appropriate reading of the ambiguous word

Conclusions at 0ms priming was related to both contextually appropriate and inappropriate readings ‘It was a good trip’... travel was just as fast as: ‘They began to trip’... travel

Results 2 at 200ms congruency started to kick in there was priming only when the context and target were congruent, i.e. the target was related to the intended reading of the word

Conclusions at 200ms priming occurred only when the target word was related to the contextually appropriate meaning of the ambiguous word ‘They didn’t believe what they saw’... hammer was slower than: ‘He bought a new saw’... hammer

Results 3 at 600ms results are puzzling... priming for verbs depends only on congruency priming for nouns doesn’t seem to matter in regard to congruency because when it’s a noun all senses seem to be primed for –any ideas?

Conclusions target naming latencies depended on two factors –congruency of target word with meaning of the ambiguous word biased by the target –and latency of when the target appeared

Conclusions: overall patterns both noun and verb readings of the ambiguous word were initially accessed within 200ms the appropriate reading was selected on the basis of syntactic context contextually inappropriate readings btwn 0 and 200ms is unlikely to be decay of semantic memory because of studies by Warren (1970) more plausible possibility is active suppression of the inappropriate reading (MacKay 1970)

Swinney 1979 Lexical Access During Sentence Comprehension: (Re)Consideration of Context Effects

Two Experiments Experiment 1 –Use Cross-Modal Semantic Priming techniques to measure multi-sense activation Experiment 2 –Extension of experiment 1, that...

Experiment 1: The Task Subjects read a set of sentences through ear phones While sentences being read, word-like strings are flashed on the screen in front of them Subjects indicate whether or not string is word IDEA: Measure the decision time! –Window into lexical access –Indicates whether auditory simuli priming for visual recognition

2X2 conditions: Ambiguity and Context 3 Target words –Related to intended reading, related to other reading(s), unrelated to any reading

Experiment 1: Method 2 Variables --> 4 variations –Ambiguity: Sentences contain ambiguous word? –Context: Context biases for a specific reading? Normalization: ambig word, unambig control –Frequency and word length (num syllables) as similar as possible –Meaning – control synonymous with intended reading of ambig word 4 variations X 3 Target words = 12 conditions

Experiment 1: Materials Sentences from earlier Swinney experiments; target words manually compiled Subjects hapless undergrads performing course requirement Equipment: earphones, CRT screen, and hopefully a comfy chair

Experiment 1: Procedure Subjects placed in chair, listened to sentences, flashed target words –Instructed to indicate word-decision as quickly as possible –Kept on track (focusing on auditory stimuli) with a feint: told that experiment would focus on comprehension and recall of sentences No indication given that target words related to sentences Followed up with written questionnaire about sentence comprehension, and whether had been aware of relationship btw auditory and visual stimuli

Experiment 1: Results

Lexical decisions for words related to all readings of an ambiguity are facilitated Effect holds even in conditions where there is a strongly biasing semantic context  Strong support for the Post Decision hypothesis Follow-up question: How long does this activation last? Bring on...

Experiment 2 Extension of Exeriment 1, with modification to measure decision-making post process Method: easy peasy! Delay visual target word to three syllables after auditory ambiguous word stimulus Other than that delay, method and procedure remained the same

Experiment 2: Results

Aha! Facilitation of irrelevant word senses decays after three syllables Conclusion: All senses are initially activated, but only very briefly This correlates with Tanenhaus et al's findings

General Conclusion The winner: Post Decision Hypothesis! –All meanings of an ambig word are activated at time of stimulus Also, this effect decays over a very short time span, about 1/10 second –What does this mean for the old Phoneme Monitoring Test, and others?

General Conclusion The winner: Post Decision Hypothesis! –All meanings of an ambig word are activated at time of stimulus Also, this effect decays over a very short time span, about 1/10 second –What does this mean for the old Phoneme Monitoring Test, and others? –--> This short decay time throws off results for methods that aren't sensitive to it

Method Comparisons Tanenhaus et alSwinney stimulus onset0ms 200ms 600ms Immediately After 3 syllables variablesAmbiguity, Target Type, Congruency Ambiguity, Context “zoning out” factorNot mentionedAttempted to control for Lex decision taskRead word from screenID if is word or not Lex access measureTime until begin to read word Time until indicate decision

Questions -- Tanenbaum instructed the subjects had to read a word as quickly as possible - could they block out the sentences they hear and skew the priming? was the language dense, or the articles older than us? – ‘However, a number of these utilized tasks which appear likely to have led subjects to employ specialized processing strategies.’

Questions -- Tanenbaum How strict are their example sentences in for ambiguity, when they are aurally tested? Grammatical (12d): –‘The table was difficult to nick’ vs. –‘The table was difficult to Nick’ Lexical (17b): –‘I put it in the wine’ vs. –‘I put it in the whine’

Questions -- Swinney How did subjects indicate their decisions? Say something, like in Tanenhaus et al? Push a button?...Method maybe affects reaction timeability

The End! Further questions, comments, etc? Exeunt Shauna and Steve...