1 / 28

How do we know what words mean?

How do we know what words mean?. Jenni Rodd j.rodd@ucl.ac.uk jennirodd.com. This lecture. Overview of topic An example experiment (Rodd et al., in press) Suggested experimental methods Practical issues. The practical stuff…. http:// jennirodd2ndyearlab2012.wordpress.com

mervyn
Télécharger la présentation

How do we know what words mean?

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. How do we know what words mean? Jenni Rodd j.rodd@ucl.ac.uk jennirodd.com

  2. This lecture • Overview of topic • An example experiment (Rodd et al., in press) • Suggested experimental methods • Practical issues

  3. The practical stuff… http://jennirodd2ndyearlab2012.wordpress.com (can access via jennirodd.com) Place to access course materials e.g., these slides, relevant papers, informed consent etc ask questions (use the ‘comment’ facility) discuss issues Or email me: j.rodd@ucl.ac.uk Or visit my office (209; on Tuesday afternoons)

  4. Sentence Comprehension To understand the meaning of a sentence we must: • Identify which words are present • Retrieve the meanings of these words • Combine these meanings together Ambiguous words make this more difficult e.g “The shell was fired towards the tank” Semantic ambiguity is extremely common

  5. Meaning Dominance Key factor in semantic disambiguation • Relative frequencies of meanings e.g., “pen” Dominant: writing implement Subordinate: animal enclosure Strong influence on reading times

  6. Dominance Norms Generated using word association task e.g., Twilley et al., 1994 PEN: “pencil”, “ink” 91% “pig”, “sheep” 4% MARCH: “April”, “month” 51% “walk”, “soldier” 47% Norms used to categorise individual meanings as: dominant, balanced, subordinate

  7. Meaning Preferences: Stable or Flexible? Use of norms assumes dominance is relatively stable • across individuals • within individuals Alternative view: Preferences are highly flexible Strongly influenced by individual/recent experience

  8. Example sentences with ambiguous words • the match ended as a tie • the head of the local branch was replaced when the company was reorganized • the star had many fans who came to all his concerts • the waist of the jeans was very narrow • a spade was not the suit that the card player wanted • the coach started to brake too late

  9. Experiment 1: Aim How flexible are listeners’ meaning preferences? Can an encounter with an ambiguous word affect meaning preferences after 20 minute delay?

  10. seal Expt 1: General MethodRodd et al., (in press) Three Stages • Prime phase : Semantic relatedness task (~6 min) 2. Filler task: Digit span (~8 min) 3. Test phase: Word association task (~18 min) Does Prime influence responses at Test? The seal came up onto the bank of the river shore? ?

  11. Expt 1: Stimuli • Prime phase 59 Sentences taken from Rodd et al., (2005) • All contain two ambiguous words either homonyms (e.g., “pen”) or non-homographic homophone (e.g., “night”/”knight”) • Usually less frequent meaning • Filler task: Digit Span Randomly generated number strings (3 to 9 digits) 3. Test phase 118 words from previously heard sentences heard

  12. Experiment 1: Predictions Will we find ‘meaning priming’? • Would improve communicative efficiency • Listeners are sensitive to distributional information (Dominance effects)

  13. Expt 1: Results Test phase: Word association Responses coded as consistent/inconsistent e.g., “The knight/night began to charge on his horse” • consistent: soldier, army, armour • inconsistent: day, dark, moon Misheard and ambiguous responses excluded Proportion ‘consistent’ responses calculated

  14. Expt 1: Results Significant priming (p<.001) 30% increase in consistent responses Control Primed

  15. Conclusions • Our preference for a particular meaning of an ambiguous words is relatively flexible • One encounter with a meaning can bias interpretation after 20 minutes Other recent experiments with this method suggest that this priming/learning effect • only occurs if the listener is attending to the sentence • is not affected by a change in the voice between prime and test (for unfamiliar voices)

  16. Unanswered Questions(Soon to be answered!) • Can printed words influence how we process later spoken words? (and vice versa) • How long can priming effect last? • Can a sufficiently strong prime completely overturn lifetime preference? • Does our childhood experience still influence our processing? • Does the situational context affect processing? • Is there more priming within a (known) speaker? • Do we take into account the identity of the speaker? • Can a prime in another language affect processing? • Does the ‘meaningfulness’ of the prime matter

  17. Suggested Method Overview The word association task will be at the centre of these experiments. What factors influence the word association responses to any given ambiguous word?

  18. Suggested Method 1:Web-based Individual Differences Experiment Web based experiment • Similar to the word association experiment your recently completed Compare how different groups of people process the same set of words… • People with different hobbies (now or in the past) • People from different language/cultural backgrounds • People of different ages

  19. Suggested Method 1:Web-based Individual Differences Experiment e.g. “SPADE”

  20. Suggested Method 1:Web-based Individual Differences Experiment e.g. “TUBE”

  21. Suggested Method 2:Web-based Priming Experiment Web based experiment • Similar method to Rodd et al., (in press) • PRIMES • (optional) Filler task (e.g. digit span, maths problems) • TEST The prime could be anything that you think might influence later processing of the ambiguous words… e.g., pictures, printed words, words in other language

  22. Advantage of web experiments • Easier to recruit naïve subjects • SurveyMonkey (or similar) sites provide free, easy to use software (note that free version does have limitations in terms of easy access to data) But • Difficult for auditory experiments (visual expts can’t use ambiguities with different spellings like night/knight) • Less control than lab-based experiments

  23. Suggested Method 3:Non-Web Priming Experiment Web experiment may be unsuitable if: • you want to use sound recordings or videos e.g., effects of identity on priming • you want to control situational cues e.g., lecture theatre vs. canteen Present primes using laptop/MP3player/IPad Record word association responses IMPORTANT: If you test local participants you MUST keep a record of who you have tested and pass this on to me.

  24. Experimental Design Issues: Priming Experiment Need to think about experimental design: • Two groups of participants – primed vs. unprimed • Two groups primed in different ways e.g., prime bark-tree vs bark-dog • More complex ‘rotated’ design Appropriate baseline is essential Strongest effects when priming low-frequency meanings i.e. prime pen-enclosure not pen-paper

  25. Experimental Design Issues: Stimuli Fillers – word association task should include set of irrelevant filler items to distract from purpose Easy to obtain lists of ambiguous words from online papers (but don’t limit yourself to these items)

  26. The Plan Week 1: Choose your question/hypothesis Decide on method Week 2: Pilot & refine experiments. Begin data collection Week 3: Finish data collection & analyze results Week 4: Report results

  27. The practical stuff… Ethics Experiments following general approach will be covered by my ethics application. Note: only healthy adult participants no strongly emotional stimuli relatively easy tasks ok to be vague about aims, but avoid clear deception Statistics May need to conduct subjects and items analysis average consistency responses across either subjects and/or items (see website)

  28. The practical stuff… http://jennirodd2ndyearlab2012.wordpress.com (can access via jennirodd.com) Place to • access course materials e.g., these slides, relevant papers, ethics application, informed consent and demographic information • ask questions • discuss issues Or email me: j.rodd@ucl.ac.uk Or visit my office (209; on Tuesday afternoons)

More Related