Euro-XPrag, Barcelona, 04 June 2011
A simple(r) account of the acquisition
of quantity implicature
Napoleon Katsos1, Gerlind Grosse2,
Cornelia Schulze2 and Michael Tomasello2
1 Research
Centre for English and Applied Linguistics, University of Cambridge
2 Max Planck Institute for Evolutionary Anthropology, Leipzig
The plan
• Part A: Latest empirical findings on the acquisition of quantity
implicature and informativeness
• Part B: Three developmental accounts:
– Memory-based
– Linguistic-scales
– Linguistic-contrast-mates
• Novel empirical findings and implications
• Informativeness: ‘give no more and no less information than
you must, taking into account the purpose of the
conversation’
Grice, 1975; Hirschberg, 1991; Horn, 1972, 1984;
Levinson, 1983, 2000; Sperber & Wilson, 1995 i.a.
John: Did you meet Linda’s parents?
Bill: I met her dad.
>> Bill didn’t meet Linda’s mom.
John: All my students passed the exams. Did yours pass too?
Bill: Some of them did.
>> Some but not all of them did.
Quantity Implicature
An inference in several steps:
1. The speaker has said p
2. There is another expression, q, which would have been more
informative and relevant
3. If he could have said q, he would have done so (because he
obeys the maxim of informativeness)
4. The fact that he didn't suggests that he can't
•
•
•
•
Do children generate Quantity Implicatures?
In what conditions?
How do they acquire this skill?
Which aspects of the process are challenging?
→ Under-informative utterances in binary judgment tasks.
Barner et al, 2010; Guasti, et al 2005; Noveck, 2001;
Papafragou & Musolino, 2003; Pouscoulous et al, 2007 i.a.
A typical critical trial (Katsos & Bishop, 2011, Cognition)...
5
Experimenter:
“Mr Caveman, what did the mouse pick up?”
“The mouse picked up some of the carrots”
Experimenter:
“Mr Caveman, what did the dog paint?”
“The dog painted the triangle.”
Mr Caveman - Description mode
% correct responses (rejecting UU)
100
90
80
70
60
SEM
50
UU
40
30
20
10
0
5y
7y
9y
age group
11y
Adults
The puzzle
• In binary judgement tasks, children younger than 6
– know the logical meaning of contrast terms
– do not reject under-informative utterances at adult-like rates
– do not reject under-informative utterances at the rate that they
reject logically false ones
• Children lack some aspect of competence:
– WM account: memory resources to generate implicatures
– Linguistic-scales account: do not spontaneously conjure the scale
of contrast
• Pragmatic tolerance account: children differ in their attitude towards
pragmatic infelicity
Katsos & Bishop (2011); Katsos & Smith (2010)
• 5 ½ and 7- year-old children (n=20)
• Stimuli are identical to experiment 1; likert scale judgement
instead of binary judgement
• If children are not able to derive Quantity Implicatures: rate
under-informative utterances as good as optimal
• If they are sensitive but also tolerant: rate them worse than
optimal but better than false
Katsos & Bishop (2011) 5 ½ year-olds and adults
‘small, big, huge strawberry’
Experimenter:
“Mr Caveman, what did the dog paint?”
“The dog painted the triangle.”
TO PARTICIPANT: “Which strawberry
would you give him?”
Experimenter:
“Mr Caveman, what did the mouse pick up?”
“The mouse picked up some of the carrots”
TO PARTICIPANT: “Which strawberry
would you give him?”
Pragmatic Tolerance
• 5 ½ children are not unable to derive implicatures. They are
tolerant towards pragmatic violations
• Adults are tolerant too...
• Kids can't do implicatures?!
• Systemantic task-dependent confound inherent in the binary
judgement task
Cummins & Katsos, 2010, JSem; Davies & Katsos, 2010, Lingua;
Katsos & Bishop, 2011, Cognition; Katsos et al, 2011, Cognition.
The puzzle is still out there...
• Developmentally, what do children need in order to generate
implicatures?
Quantity Implicature
An inference in several steps:
1. The speaker has said p
2. There is another expression, q, which would have been more
informative and relevant
3. If he could have said q, he would have done so (because he
obeys the maxim of informativeness)
4. The fact that he didn't suggests that he can't
Working Memory account
(Pouscoulous et al 2007; Reinhart, 2000)
• Children lack the working memory (WM) resources required for
steps 2 or 4.
• WM implicated in adults:
– S-A-T, and dual-load tasks (Bott & Noveck, 2004; De
Neys & Schaeken, 2007)
– Antoniou & Katsos (this conference): WM individual
differences contribute to implicature rates
• Children:
– Better performance when distractors are removed,
easier / more frequent words used
– No individual differences / dual load studies.
Linguistic-scalar account
(Barner et al 2010, Guasti et al 2005)
• Besides knowing the logical meaning of the contrast mates
(e.g. ‘some’ and ‘all’), hearer must also be able to
spontaneously activate the relevant scale (<some, all>) upon
encountering a scalar term ('some').
→ What young kids lack is the ability to spontaneously evoke
the relevant scale (step 2), even though they know the
meaning of the contrast terms
2 /3 display
3 /3 display
• Are some of the animals sleeping?
In 2/3: kids say ‘Yes’; in 3/3: kids say ‘Yes’
• Are only some of the animals sleeping?
In 2/3: kids say ‘Yes’; in 3/3: kids say ‘Yes’(!)
• Are all the animals sleeping?
In 2/3: kids say ‘No’
• Are only some of the animals sleeping?
In 3/3: kids say ‘Yes’ even though they know the meaning of 'all'
A simple(r) linguistic account (Katsos & Bishop, 2011)
'Constrast mates'
Kids will draw quantity inferences, as long as they know the meaning of
the terms that contrast. So, step 2 is no big deal, as long as they know
that there exists an expression for that step.
• The ‘find the box’ game:
– A reference disambiguation task (picture-selection)
– A real-life motivation for the child to derive as much relevant
information as possible from their interlocutor’s utterance
Linguistic-contrast account
• Prediction-1
– If kids reliably know the meaning of 'all', or 'NP&NP'
then they should pass the test-trial 'some' or 'NP'
• Prediction-2
– If kids that know the meaning of 'all', or 'NP&NP'
then they will experience a slow-down in choosing
the under-informative box
German-speaking participants:
3-1/2-year-olds (n=24; 12 female)
5-year-olds (n=24; 12 female)
Adults (n=12; 6 female)
Order of ad hoc & scalar blocs, position of boxes, order of
instructions counterbalanced.
1st trial of each bloc repeated as third, to avoid guessing pattern.
In critical trials a sticker was hidden under both the weak and
strong term box.
`
• Correlational analyses:
• 3-1/2-year-olds:
– (2/3 or 3/3) 'some' correlates 3/3 on 'all' (rho = .41 *)
– (2/3 or 3/3) 'NP' correlates 3/3 on 'NP & NP' (rho = .41 *)
• 5-year-olds
– (2/3 or 3/3) 'some' correlates with 3/3 on 'all' (rho = .40 *)
• Performance with the stronger contrast-mate correlates with
performance with the critical weaker contrast-mate
2 (Felicity: Control v Under-info) x 2 (Type: Scalar v Adhoc) ANOVA
Felicity F(1, 23) = 22,48, ***; Type F(1, 23) = 1,72, n.s. Felicity x Type F(1, 23) = 1,31, n.s.
Magnitude of effect of Felicity doesn't correlate with performance on contrast mate
Response times
2.5
2
1.5
test
contrast
control
underinfo
1
0.5
0
scalar
ad hoc
Summary: experimental findinds
• 3 ½-year-olds pass ad hocs, but fail scalar quantity implicatures
• 5-year-olds pass both
• Pass/fail on the implicature trials is related to pass/fail with the
stronger contrast mate.
• Reaction time on under-informative utterances is longer
compared to felicitous controls
• Quite possible that working memory makes a contribution.
• Also possible that kids do need to learn scales.
Quantity Implicature
The linguistic-contrast account:
1. The speaker has said p
2. There is another expression, q, which would have been more
informative and relevant
3. If he could have said q, he would have done so (because he
obeys the maxim of informativeness)
4. The fact that he didn't suggests that he can't and that he is
communicating this.
Much younger kids may pass Quantity Implicature
tasks
Quantity inference is similar in structure to novel word
learning by mutual exclusivity (Grassmann, subm.):
– The speaker has asked for p
– There is another expression that he could have used,
q, which I do know what it refers to
– If he had meant q, he would have said so
– The fact that he didn't suggests that he doesn't
We are going to look for the age when kids know NP & NP, we
predict they should do the Quantity Implicature
As long as we don't use binary judgement tasks...
Thank you!
[email protected]