Talking about Objects

Post on 06-Jan-2016

49 views 0 download

Tags:

description

Talking about Objects. Jason Kahn & Jennifer E. Arnold UNC – Chapel Hill Amlap. Did you get the reading group organized for this semester yet?. Did you get the reading group organized for this semester yet?. You wanted a reading group this semester, too?. - PowerPoint PPT Presentation

Transcript of Talking about Objects

Talking about Objects

Jason Kahn & Jennifer E. ArnoldUNC – Chapel Hill

Amlap

Did you get the reading group organized for this semester yet?

Did you get the reading group organized for this semester yet?

You wanted a reading group this semester, too?

Did you get the reading group organized for this semester yet?

You wanted a reading group this semester, too?

Why does “reading group” get reduced?

• Givenness Explanations • Probability Explanations

Why does “reading group” get reduced?

• Givenness Explanations– Uttered words become

given in a discourse (Prince 1992)

– Given information has a different discourse status/representation (Halliday 1967)

– Given discourse information gets reduced on subsequent mention (Fowler & Housum, 1987)

• Probability Explanations

Why does “reading group” get reduced?

• Givenness Explanations– Uttered words become

given in a discourse (Prince 1992)

– Given information has a different discourse status/representation (Halliday 1967)

– Given discourse information gets reduced on subsequent mention (Fowler & Housum, 1987)

• Probability Explanations– Context increases the

probability of mentioning a word

– Highly probable words get reduced (Bell et al., 2009)

– Words that convey little information also get reduced (Levy & Jaeger 2007)

Well, what kind of reading group did you want?

[Reading group]

Bard & Aylett 2004; Brown-Schmidt 2009

Cognitive Mechanisms for Reduction

Adapted from Levelt 1989

Cognitive Mechanisms for Reduction

Givenness

Cognitive Mechanisms for Reduction

Givenness

•Facilitation•Pre-planning

Cognitive Mechanisms for Reduction

Givenness

•Facilitation•Pre-planning

Probability

Cognitive Mechanisms for Reduction

Givenness

•Facilitation•Pre-planning

Probability

•Associative facilitation•Pre-planning

Cognitive Mechanisms for Reduction

Givenness

Probability

(Arnold, 1998; Givón, 1983; Tily & Piantodosi, 2009)

Facilitation-Based Reduction

• Any stimulus that activates representations that are used for language production should lead to some measurable amount of reduction. This includes probability, givenness, and their combination.

Novel Predictions

• Referential events should “count” for reduction

Novel Predictions

• Referential events should “count” for reduction

• We should observe reduction from non-linguistic information

Novel Predictions

• Referential events should “count” for reduction

• We should observe reduction from non-linguistic information

• We should observe a difference between linguistic and non-linguistic information

Linguistically Given

Non-linguistically (conceptually)

Given Predictable

Linguistic

Non-linguistic

Control

X

Experimental Layout (Exp 1)

XXX

Linguistically Given

Non-linguistically (conceptually)

Given Predictable

Linguistic

Non-linguistic

Control

X

Non-linguistic Stimuli Should Elicit Reduction

XXX

Linguistically Given

Non-linguistically (conceptually)

Given Predictable

Linguistic

Non-linguistic

Control

X

Linguistic Stimuli Should Elicit More Reduction

XXX

Linguistically Given

Non-linguistically (conceptually)

Given Predictable

Linguistic

Non-linguistic

Control

X

XXX

We Will Return to Predictability

Instruction-giving Task

Speaker Listener

Speaker Listener

Move

Speaker Listener

“The accordion rotates right”

Move

Speaker Listener

“The accordion rotates right”

x 3

Move

Speaker Listener

“The toothbrush. The belt. The accordion.”

Move

Speaker Listener

“The accordion rotates right”

“The toothbrush. The belt. The accordion.”

Move

Speaker Listener

“The accordion rotates right”

x 3

Move

Speaker Listener

Move

Speaker Listener

“The accordion rotates right”

Move

Speaker Listener

“The accordion rotates right”

x 3

Move

1)Or… Or…

1)

2)

Or… Or…

“[Onset] The toothbrush shrinks.”

1)

2)

3)

Or… Or…

“[Onset] The toothbrush shrinks.”

1)

2)

3)

4)

Or… Or…

“[Onset] The toothbrush shrinks.”

“[Onset] The belt expands.”

1)

2)

3)

4)

5)

Or… Or…

“[Onset] The toothbrush shrinks.”

“[Onset] The belt expands.”

1)

2)

3)

4)

5)

6)

Or… Or…

“[Onset] The toothbrush shrinks.”

“[Onset] The belt expands.”

“[Onset] The accordion rotates right.”

1)

2)

3)

4)

5)

6)

7)

Or… Or…

Analysis

• Multi-level Modeled– Information condition (ling, non-ling, control)– Control variables (syllables, imageability, etc.)– Random intercept for subject and item (cross-classified)

Reduced Onset DurationLinguistic = Non-linguistic < Control

Third Instruction: “[Onset] the accordion rotates right”

*

Reduced Object DurationLinguistic < Non-linguistic < Control

Third instruction: “[Onset] the accordion shrinks”

*

*

Linguistically Given

Non-linguistically (conceptually)

Given Predictable

Linguistic

Non-linguistic

Control

X

Non-linguistic Stimuli Elicited Reduction

XXX

Linguistically Given

Non-linguistically (conceptually)

Given Predictable

Linguistic

Non-linguistic

Control

X

Linguistic Stimuli Elicited More Reduction

XXX

Cognitive Mechanisms for Reduction

Givenness Predictability

Cognitive Mechanisms for Reduction

Givenness Predictability

Linguistically Given

Non-linguistically (conceptually)

Given Predictable

Linguistic

Non-linguistic

Control

X

Predictability May (partly) Produce This Effect

XXX

See also Lam & Watson 2009

Linguistically Given

Non-linguistically (conceptually)

Given Predictable

Linguistic

Non-linguistic

Control

X

Experimental Layout (Exp 2)

X

X

XXX

Experiment 2: Is Predictability Necessary?

• Reduce the number of instructions per trial to one

• Vastly reduce the validity of the prime – the object that gets primed is the object that moves only 12.5% (1/8th) of the time

• Changed the relationship between the experimental and control condition

7/8:1/8:

Linguistic

Non-linguistic

Experimental Layout (Exp 2)

Primed Unprimed

Linguistic

Non-linguistic

Experimental Layout (Exp 2)

Primed Unprimed

Reduced Object Duration (Ling. Only)Given < Not Given

“[Onset] the accordion shrinks”

*

“The accordion”

“The heart”

Given < Not Given

“[Onset] the accordion shrinks”

(Numerically) Reduced Object Duration

Linguistic

Non-linguistic

Experimental Layout (Exp 2)

?

*

Primed Unprimed

Cognitive Mechanisms for Reduction

Givenness Predictability

Non-linguisticpredictabilitymay be necessary

Cognitive Mechanisms for Reduction

Givenness Predictability

LinguisticpredictabilityIs not necessary

Non-linguisticpredictabilitymay be necessary

Cognitive Mechanisms for Reduction

Givenness Predictability

LinguisticpredictabilityIs not necessary

Non-linguisticpredictabilitymay be necessary

…but it may help planning

Facilitation-Based Reduction• Referential event information can

affect reduction (linguistic or not)

• Non-linguistic information can elicit reduction

• Linguistic information elicits more reduction (suggesting a facilitation effect)

• Predictability does not appear to be necessary for reduction in all cases

Facilitation-Based Reduction• Referential event information can

affect reduction (linguistic or not)

• Non-linguistic information can elicit reduction

• Linguistic information elicits more reduction (suggesting a facilitation effect)

• Predictability does not appear to be necessary for reduction in all cases

Facilitation-Based Reduction• Referential event information can

affect reduction (linguistic or not)

• Non-linguistic information can elicit reduction

• Linguistic information elicits more reduction (suggesting a facilitation effect)

• Predictability does not appear to be necessary for reduction in all cases

Well, what kind of reading group did you want?

[Reading group]

Acknowledgements

• Kellen Carpenter and Giulia Pancani (our lab managers), Jennifer Tait, Alyssa Ventimiglia, Leighanne McGill (our RA’s)

• The Cognitive Department at UNC, whose comments have sharpened our focus here

• NSF BCS-0745627 to Jennifer E. Arnold