Date post: | 14-Apr-2018 |
Category: |
Documents |
Upload: | aditya-kaushik |
View: | 217 times |
Download: | 0 times |
7/27/2019 Grounding+Language+in+Action.pdf
http://slidepdf.com/reader/full/groundinglanguageinactionpdf 1/25
Grounding Languagein Action
HUL 281
Mind, Machine and Language
7/27/2019 Grounding+Language+in+Action.pdf
http://slidepdf.com/reader/full/groundinglanguageinactionpdf 2/25
How language conveys meaning
• Two main approaches:
• The dominant approach - treat language as a symbol manipulation
• Meaning conveyed by using abstract, amodal, and arbitrary symbols words) combined by syntactic rules.
• abstract - same word to express different things
• amodal - same word is used when things are spoken about or writte
• arbitrary - word bear no relationship to the word’s referent
7/27/2019 Grounding+Language+in+Action.pdf
http://slidepdf.com/reader/full/groundinglanguageinactionpdf 3/25
Motor Resonance and Languag
• language of action evokes motor resonance
• 2 type of evidence:
• neural: cognitive neuropsychology
• behavioral: psycholinguistics (interference effects)
7/27/2019 Grounding+Language+in+Action.pdf
http://slidepdf.com/reader/full/groundinglanguageinactionpdf 4/25
Neural Evidence: Pulvermueler (20
• Pulvermueler et al. (2001): hearing/reading action verbs
• produces somatotopic activation in theprimary motor cortex
• EEG study, movement vs. lexical decisiontask
• kick (leg), pick (arm), lick (face)
7/27/2019 Grounding+Language+in+Action.pdf
http://slidepdf.com/reader/full/groundinglanguageinactionpdf 5/25
Neural Evidence: Pulvermueler (20
• Hauk et al. (2004): fMRI study, movement vs. passive reading
7/27/2019 Grounding+Language+in+Action.pdf
http://slidepdf.com/reader/full/groundinglanguageinactionpdf 6/25
• Pulvermueler et al. (2005): TMS study, movement vs. Passive read
7/27/2019 Grounding+Language+in+Action.pdf
http://slidepdf.com/reader/full/groundinglanguageinactionpdf 7/25
Behavioral Evidence: Indexical Hypoth
of the Meaning• meaning based on action (and affordances)
• transformation of language to meaning in 3 steps:
• 1 words => perceptual symbols
• 2 perceptual symbols => affordances
• 3 affordances + syntactic construction => action: (un)doable= (mis)under
• 4 grammatical form: order of simulation, how to combine affordances
• language is made meaningful by cognitively simulating the actions implied bsentences
7/27/2019 Grounding+Language+in+Action.pdf
http://slidepdf.com/reader/full/groundinglanguageinactionpdf 8/25
Behavioral Evidence: Action-Senten
Compatibility Effect• Glenberg and Kaschak (2002): comprehending a sentence that implies acti
direction interferes with real action in the opposite direction
• Movement from and towards the body: when incongruent with the sentencmovement => reaction time is longer
• Task: sensical and nonsensical sentences, button Y/N near and far from th
• John gave you a pizza. You gave a pizza to John.
• Also with abstract transfer sentences (Glenberg et al., 2008)
• Mary told you a story. You told a story to Mary.
7/27/2019 Grounding+Language+in+Action.pdf
http://slidepdf.com/reader/full/groundinglanguageinactionpdf 9/25
Chinese Room Argument: Harnad’s (1
version of Searle’s (1980) Provides a compelling intuition as to why meaning cannot be conveyed solely by t
syntactic relations among abstract symbols
• Subject - foreign lady with a Chinese dictionary
• Task – to look for meaning of abstract symbols to decode the meaning of a sente
• Result - No matter how many of these abstract symbols she relates to one anothe
never going to determine the meaning of the sentence• Lesson learnt - abstract symbols of language must be grounded, or mapped, to t
if they are to convey meaning
In other words - if one has only abstract symbols at one’s disposal, determination of correct mapping is impossible
7/27/2019 Grounding+Language+in+Action.pdf
http://slidepdf.com/reader/full/groundinglanguageinactionpdf 10/25
Affordance
potential interactions between bodies and objects (Gibson, 1979; Tucker & Ellis, 19
Example - a chair affords - sitting - adult humans
mice or elephants - wrong sorts of bodies to sit in an ordinary chair
meaning of the situation arises from meshing the affordances of different bodies and objects
For e.g. meshing the affordances of chair and light bulb toaccomplish the goal of changing the bulb.
7/27/2019 Grounding+Language+in+Action.pdf
http://slidepdf.com/reader/full/groundinglanguageinactionpdf 11/25
Indexical Hypothesis (IH)
• 3 processes transform words and syntax into an action-based meaning:
First, words and phrases are indexed or mapped to perceptual symbols
Unlike abstract symbols, perceptual symbols are modal and non arbitrary
Based on the brain states underlying the perception of the referent
Second, affordances are derived from the perceptual symbols
arbitrary symbols = arbitrary => no affordances can be derived
perceptual symbols = non arbitrarily => affordances can be derived
7/27/2019 Grounding+Language+in+Action.pdf
http://slidepdf.com/reader/full/groundinglanguageinactionpdf 12/25
Third, affordances are meshed under the guidance of syntactic constru
For e.g., the affordances of an upright vacuum cleaner and of a coat to
accomplish the goal of hanging up the coat affordances do not mesh to guide action => non-sensible sentence
Example : “Hang the coat on the upright vacuum cleaner” is sensible
“Hang the coat on the upright cup” is not sensible
Affordances - vacuum cleaner can be used as a coat rack but not cupNeither based on explicit previous learning nor on abstract symbols
Lastly, grammatical form of the sentence directs a cognitive simulationcombines various object’s affordances to convey meaning
7/27/2019 Grounding+Language+in+Action.pdf
http://slidepdf.com/reader/full/groundinglanguageinactionpdf 13/25
Experiment 1
• Participants presented with series of sensible/nonsense sentences
• asked to distinguish between both kinds
• one independent variable, implied sentence direction (toward/ away)
“Put your finger under your nose,” => action toward the body
“Put your finger under the faucet,” => action away from the body
• Participants never instructed to consider the implied direction; just se
7/27/2019 Grounding+Language+in+Action.pdf
http://slidepdf.com/reader/full/groundinglanguageinactionpdf 14/25
Experimental Set-Up
• Specially constructed button box to record actual response direction (3 but
• Buttons differed in distance from body – near, middle and far
• pressing middle button – initiates visual representation of sentence
•
yes-is-far condition – sensible sentence – middle to far button – away from• yes-is-near condition – sensible sentence – middle to near button – toward
• Major dependent variable - time between presentation of the sentence a
of the middle button (to move to the near or the far button)
7/27/2019 Grounding+Language+in+Action.pdf
http://slidepdf.com/reader/full/groundinglanguageinactionpdf 15/25
IH Predictions
• towards sentence - simulation of actions toward the body and vice v
• According to IH, interference between simulation and action
toward sentence – interfere with - yes-is-far condition
away sentence - interfere with - yes-is-near condition• prediction for a statistical interaction between implied sentence direc
actual response direction.
This interaction is referred to as the action – sentence compatibility effect (ACE
7/27/2019 Grounding+Language+in+Action.pdf
http://slidepdf.com/reader/full/groundinglanguageinactionpdf 16/25
Sentence Types
• Total 80 sensible sentence pairs were used
• Half of them i.e. 40 toward/away pairs were in the imperative
• The concrete transfer pairs (20 toward/away pairs) described transfephysical object between “you” and another person.
• 20 abstract transfer pairs described a nonphysical transfer, such as
“Liz told you the story/You told Liz the story ”
7/27/2019 Grounding+Language+in+Action.pdf
http://slidepdf.com/reader/full/groundinglanguageinactionpdf 17/25
Experiment 1 : Results
• Analyses were conducted on the proportion of correct judgments, as well as on the reading time
• Error due to practice effects and outliers were taken care of by taking mean of readings
7/27/2019 Grounding+Language+in+Action.pdf
http://slidepdf.com/reader/full/groundinglanguageinactionpdf 18/25
Results (Contd.)
• Significant interaction between response direction and implied direction i.e. AC
• Stronger effect for two types of transfer sentences, as compared with imperati
• Consistent with IH and inconsistent with abstract symbol theories of meaning
• Merely understanding a sentence can facilitate or interfere with a physical respo
• Read Time (imperative sentences) < Read Time (transfer sentences)
• Significant interaction between sentence direction and sentence type
7/27/2019 Grounding+Language+in+Action.pdf
http://slidepdf.com/reader/full/groundinglanguageinactionpdf 19/25
EXPERIMENT 2
• Designed to replicate and modestly extend the major effects fromExperiment 1
• Participants responded with their left hands
• Aim - to determine whether ACE reflects action-planning specific fodominant hand (right-handed subjects)
7/27/2019 Grounding+Language+in+Action.pdf
http://slidepdf.com/reader/full/groundinglanguageinactionpdf 20/25
Analysis of error rates:
• Fewer errors on the toward sentences (4%) than on the away sentences (7%)
• Fewer errors for imperative sentences (4%) than for concrete (6.5%) or abstratransfer sentences
Results : Most of the results were similar to Experiment 1
Other significant finding - effect of sentence type
7/27/2019 Grounding+Language+in+Action.pdf
http://slidepdf.com/reader/full/groundinglanguageinactionpdf 21/25
EXPERIMENT 3
• Designed to test a spatial location alternative to the IH
• Participant did not move the hand.
left index finger - over the yes button (either near to or far from the body)
right index finger - over the no button (either far from or near to the body).
IH Predictions: No relevant interfering action during response => No ACE
Results: little evidence for an ACE interaction => interference arises from action no
7/27/2019 Grounding+Language+in+Action.pdf
http://slidepdf.com/reader/full/groundinglanguageinactionpdf 22/25
• Trimmed mean r.t.(Towards) = 1,742 msec < Trimmed mean r.t.(Away) = 1,80
• Significant interaction of sentence direction and sentence type
•No speed – accuracy tradeoffs, errors on toward sentences (6%) < away sentenc
7/27/2019 Grounding+Language+in+Action.pdf
http://slidepdf.com/reader/full/groundinglanguageinactionpdf 23/25
Discussion
• The results from Experiments 2 and 3 accomplished three goals.
1. First, they demonstrated replicability of the ACE.
2. Second, demonstrated that ACE is unlikely to reflect detailed actionplanning at the level of particular muscles.
3. Third, contrast indicates that ACE depends on action, and not solespatial location of the responses.
7/27/2019 Grounding+Language+in+Action.pdf
http://slidepdf.com/reader/full/groundinglanguageinactionpdf 24/25
Meshing Affordances using Gramm
• Double object construction
• “Subject– verb – object1 – object2”
• Meaning - subject transfers object2 to object1
•Notion of denominal verbs
• Example - “The woman crutched the goalie the ball” is meaningful
“The woman eggshell the goalie the ball” is not
7/27/2019 Grounding+Language+in+Action.pdf
http://slidepdf.com/reader/full/groundinglanguageinactionpdf 25/25
THANK YOU