A brief history of Downward Entailment in child language

1
A brief history of Downward Entailment in child
language
Andrea Gualmini, Luisa Meroni, Teresa Guasti,
Gennaro Chierchia, Stephen Crain (with help from
Graciela Tesan and Kermit the Frog)
2
Downward Entailment in child language

• The semantic property of DE governs
• The conjunctive interpretation of disjunction,
• and the cancellation (or reversal) of
  scalar implicatures
• II. The licensing of the NPI any
• III. Yielding valid inferences from sets to their
  subsets
• NB the structural notion of c-command is a
  prerequisite for I-III

3
The Standard View of Scalar Implicatures
The interpretation of scalar terms results from
the interaction of semantic and pragmatic
principles, e.g. Scalar Implicatures (SIs).

• E.g., semantic principles assign to disjunction
  an inclusive-or interpretation.
• And pragmatic principles constrain the
  interpretation of disjunction, to include only
  the truth conditions associated with
  exclusive-or.

4
The Standard View of Scalar Implicatures
Information Strength If a sentence S1
asymmetrically entails another sentence, S2, then
S1 makes a stronger statement than S2
S1 Fred bought a hamburger and a hot dog S2
Fred bought a hamburger or a hot dog
Because S1 entails S2, S1 makes a stronger
statement
5
Summary The Standard View

• The English words or and and are mapped onto
  the logical connectives, ? and ?, which form a
  scale based on information strength. An element
  on a scale is weaker (lt) if it is true in a
  superset of the circumstances associated with
  another element on the same scale. E.g., (or lt
  and), (some lt many lt most lt every)
• Suppose ? and ? form a scale, such that ? ? ?.
• Then if the speaker utters S(?) (i.e., uses the
  logically weaker scalar term), the statement is
  interpreted as S(?) and not S(?).
• However, scalar implicatures are cancelled in
  contexts of uncertainty

6
A Puzzle for the Standard View of SIs

• SIs fail to arise in specific linguistic
  contexts, regardless of the amount of information
  available to the speaker.
• a. Paul doesnt like John or Bill.
• b. Paul is stronger than John or Bill.
• c. Paul usually arrives at the gym before John or
  Bill.
• d. If John or Bill go to the gym, Paul goes
  swimming.
• e. Paul forbids John or Bill from going to the
  gym.

7
and it becomes even more puzzling

• The same contexts (in which SIs fail to arise)
  license the Negative Polarity Item any
• a. Paul doesnt like any other linguist.
• b. Paul is stronger than any other linguist.
• c. Paul usually arrives at the gym before any
  other linguist.
• d. If any other linguist goes to the gym, Paul
  goes swimming.
• e. Paul forbids any other linguist to go to the
  gym.

8
So far Disjunction is two-faced

• In positive contexts
• (1) Snow-white found the apple or the banana.
• Snow-white found the apple or Snow-white found
  the banana.
• The reading in (1) may receive an exclusive-or
  reading
• In negative contexts
• (2) Snow-white will never find the apple or the
  banana
• Snow-white will never find the apple and
  Snow-white will never find the banana
• The reading in (2) is limited to a conjunctive
  interpretation

9
Japanese or in positive contexts

• (a) Shirayukihime-wa ringo ka banana-o
  mitukerudesyou.
• Snow-white will find the apple or the
  banana
• ? Shirayukihime-wa ringo-o mitukeru ka
  shirayukihime-wa banana-o mitukerudesyou.
• Snow-white will find the apple or Snow-white
  will find the banana

10
Japanese or in negative contexts

• (b) Shirayukihime-wa kesshite ringo-mo banana-mo
  mitukenaidesyou.
• Snow-white will never find the apple or the
  banana
•  
• ? Shirayukihime-wa kessite ringo-mo
  mitukerarenai shi Sjirayukihime ha kesshite
  banana-mo metukerarenaidesyou.
• Snow-white will never find the apple and
  Snow-white will never find the banana

11
Kiguchisan-wa nani-mo tabeta.
Kiguchisan-NOM ANYthing eat-past. Kiguchi ate
anything.Kiguchisan-wa nani-mo
tabe-NA-katta. Kiguchi-NOM ANYthing
eat-NOT-PAST. Kiguchi didn't eat
anything.
and Japanese NPIs
12
The Semantic Core Model - Chierchia (2000)

• SIs are cancelled in contexts that license any,
  (roughly) in Downward Entailing (DE) contexts.
• Why are negative polarity items licensed in DE
  contexts? 
• Why are SIs cancelled in DE contexts?
• Presumably the answer is the same for both, so
  these operations reside in the same module of
  grammar.

13
Lots of Puzzles for the Modular View

• The inclusive-or interpretation is available for
  a number of linguistic contexts, even in
  situations of certainty.

(1) Geoff does not talk to linguists or
philosophers. (2) Geoff never talks to linguists
or philosophers. (3) Geoff is stronger than
linguists or philosophers. (4) Geoff usually
arrives at the gym before linguists or
philosophers. (5) Geoff went to work without the
car or the bike. (6) If a linguist or a
philosopher goes to the gym, Geoff goes
swimming. (7) Geoff forbids linguists or
philosophers from going to the gym. (8) Geoff
doubts that linguists or philosophers can find a
counter example to Chomsky. (9) Geoff prefers
linguists or philosophers to do the laundry. (10)
Every linguist or philosopher admires
Chomsky. (11) No linguist or philosopher admires
Chomsky. (12) No psychologist admires a linguist
or a philosopher.
14
Asymmetries Minimal Pairs

• (13) Every linguist or philosopher admires
  Chomsky.
• (14) Every psychologist admires a linguist or a
  philosopher.
• (15) If a linguist or a philosopher goes to the
  gym, Geoff goes swimming.
• (16) If a linguist goes to the gym, Geoff goes
  swimming or golfing.
• (17) Geoff usually arrives at the gym before
  linguists or philosophers.
• (18) Geoff usually arrives at the gym after
  linguists or philosophers.
• (19) Geoff went to work without the car or the
  bike.
• (20) Geoff went to work with the car or the bike.

15
Conjunctive interpretations Italian

• A conjunctive interpretation must be assigned in
  a number of linguistic contexts.

(1) Gennaro does not talk to linguists or
philosophers. Gennaro non parla con linguisti o
filosofi. (2) Gennaro never talks to linguists
or philosophers. Gennaro non parla mai con
linguisti o filosofi. (3) Gennaro is stronger
than linguists or philosophers. Gennaro e
piu forte di tutti i linguisti o filosofi. (4)
If a linguist or a philosopher goes to the gym,
Gennaro goes swimming. Se un linguista o un
filosofo vanno in palestra, Gennaro va a
nuotare. (5) Gennaro will forbid linguists or
philosophers from going to the gym. Gennaro ha
vietato a linguisti or filosofi di andare in
palestra (6) Gennaro doubts that a linguist or
philosopher can find a counterexample to
Chomsky. Gennaro dubita che I linguisti o
filosofi possano trovare un problema per Chomsky
16
The (same) Distribution of any

• (21) Geoff does not talk to any other linguists.
• (22) Geoff never talks to any other linguists.
• (23) Geoff is stronger than any other linguist.
• (24) Geoff usually arrives at the gym before any
  other linguist.
• (25) Geoff went to work without any means of
  transportation
• (26) If any other linguist goes to the gym, Geoff
  goes swimming.
• (27) Geoff forbids any other linguists from
  going to the gym.
• (28) Geoff doubts that any linguist can find a
  counter example to
• Chomsky.
• (29) Geoff prefers any other linguist to do the
  laundry.
• (30) Every linguist with any brains admires
  Chomsky.
• (31) No linguist with any brains admires Chomsky.
• (32) No psychologist admires any philosopher.

17
Distribution of NPIs in Italian

• (12) Gennaro does not talk to any other
  linguists.
• Gennaro non parla con nessun linguista.
• (13) Gennaro never talks to any other linguists.
• Gennaro non parla mai con nessun linguista.
• (14) Gennaro is stronger than any other
  linguist.
• Gennaro e piu forte di qualsiasi altro
  linguista.
• (15) If any other linguist goes to the gym,
  Gennaro goes swimming.
• Se qualsiasi altro linguista va in palestra,
  Gennaro va a nuotare.
• (16) Gennaro will forbid any other linguists
  from going to the gym.
• Gennaro ha vietato a qualsiasi altro linguista
  di andare in palestra.
• (17) Gennaro doubts that any linguist can find a
  counter example to Chomsky.
• Gennaro dubita che qualsiasi linguista possa
  trovare un problema per Chomsky

18
Some Minimal Pairs NPIs

• (33) Every psychologist who read any linguistics
  book admires Chomsky.
• (34) Every psychologist admires any philosopher.
• (35) If any linguist goes to the gym, Geoff goes
  swimming.
• (36) If a linguist goes swimming, Geoff goes to
  any gym.
• (37) Geoff usually arrives at the gym before any
  linguist.
• (38) Geoff usually arrives at the gym after any
  linguist.
• (39) Geoff went to work without any means of
  transportation.
• (40) Geoff went to work with any means of
  transportation.

19
Minimal Pairs in Italian Disjunction

• (7) Every linguist or psychologist admires
  Chomsky.
• Ogni linguista o psicologi ammira Chomsky.
• Every psychologist admires a linguist or a
  philosopher.
• Ogni psicologo ammira un linguista o un filosofo.
• (8) If a linguist or a philosopher goes to the
  gym, Gennaro goes swimming.
• Se un linguista o un filosofo vanno in palestra,
  Gennaro va a nuotare
• If a linguist goes to the gym, Gennaro goes
  swimming or running.
• Se un linguista vanno in palestra, Gennaro va a
  nuotare o a correre.
• (9) Gennaro usually arrives at the gym before
  linguists or philosophers.
• G. di solito arriva in palestra prima che i
  linguisti o gli psicologi.
• Gennaro usually arrives at the gym after
  linguists or philosophers arrive.
• G. di solito arriva in palestra dopo che sono
  arrivati i linguisti o gli psicologi.

20
Minimal Pairs in Italian NPIs

• (18) Every psychologist who read any linguistics
  book admires Chomsky.
• Ogni psicologo che abbia letto qualsiasi libro
  di linguistica ammira Chomsky.
• Every psychologist admires any philosopher.
• Ogni psicologo ammira qualsiasi filosofo.
• (19) If any other linguist goes to the gym,
  Gennaro goes swimming.
• Se qualsiasi altro linguista va in palestra,
  Gennaro va a nuotare
• Gennaro talks to any linguist.
• Gennaro parla con qualsiasi linguista.
• (20) Gennaro usually arrives at the gym before
  any linguist.
• G. di solito arriva in palestra prima
  qualsiasi linguista.
• Gennaro usually arrives at the gym after any
  linguist.
• G. di solito arriva in palestra dopo qualsiasi
  linguista.

21
Downward Entailment

• Downward Entailing environments license
  inferences from a set to its subsets.
• if then
• (41) Geoff does not talk to any linguist Geoff
  does not talk to any good linguist
• (42) Geoff never talks to any linguists
  Geoff never talks to any good linguists
• (43) Geoff is stronger than any linguist Geoff
  is stronger than any good linguist
• .
• (45) No psychologist admires any linguist No
  psychologist admires any good linguist

22
Descriptive Generalization

• The conjunctive interpretation is assigned to
  disjunctive statements if and only if an NPI can
  appear in that linguistic environment

23
A Deep Generalization

• Downward entailing linguistic environment
  license NPIs and constrain the interpretation of
  disjunctive statements (to conjunctive readings)
• (cf. Horn 1989234)

24
Downward Entailment

• Downward Entailing contexts license inferences
  from a set to its subsets.
• a. John did not buy a car
• ? John did not buy a red car.
• b. None of the linguists found an example
• ? None of the linguists found a good example.
• c. Every linguist who found an example wrote a
  squib
• ? Every linguist who found a good example wrote
  a squib.

25
Downward Entailment

• Downward Entailing contexts license inferences
  from a set to its subsets.
• The determiner every is downward entailing on its
  internal argument, but
• not on its external argument
• Every
• NP If every cow ate a green vegetable, then
  every brown cow ate a
• green vegetable.
• VP If every cow ate a green vegetable, then
  every cow ate
• broccoli.

26
Downward Entailment

• Downward Entailing contexts license inferences
  from a set to its subsets.
• The determiner no is downward entailing on both
  arguments
• No
• NP If no cow ate a green vegetable, then no
  brown cow ate a green
• vegetable.
• VP If no cow ate a green vegetable, then no
  cow ate broccoli.

27
Downward Entailment

• Downward Entailing contexts license inferences
  from a set to its subsets.
• The determiner some is not downward entailing on
  either of its arguments
• Some
• NP If some cow ate a green vegetable, then
  some brown cow ate a
• green vegetable.
• VP If some cow ate a green vegetable, then
  some cow ate broccoli.

28
The Same Contexts License NPIs

• Every
• NP Every cow that ate any vegetable became ill.
• VP Every cow ate any vegetable.
•  
• No
• NP No cow that ate any vegetable became ill.
• VP No cow ate any vegetable.
•  
• Some
• NP Some cow that ate any vegetable became
  ill.
• VP Some cow ate any vegetable.

29
Downward Entailment

• Across natural languages, Downward Entailing
  operators give rise to inferences that closely
  resemble the equivalences expressed by the two
  De Morgans Laws (Boster and Crain, 1993 cf.
  Partee, ter Meulen and Wall, 1993).
• (4) ?(P?Q) ? (?P??Q)
• (5) ?(P?Q) ? (?P ? ?Q)

30
A Generalization Downward Entailment

• OPDE(P or Q) ? (OPDEP and OPDEQ)

31
Summary
The same linguistic principle governs I. The
interpretation of or II. The appearance of
any III. The valid logical inferences from
sets to their subsets NB c-command is a
pre-requisite for I-III
32
Scalar Implicatures in Child Language

• 1. Do children ignore SIs in DE contexts?
• 2. Do children compute SIs in non-DE contexts?

33
A case study the universal quantifier

• The two arguments of Every
• DE environment - 1st argument of Every
• Every linguist who found any example wrote a
  squib.
• non-DE environment - 2nd argument of Every
• Every linguist found any example.

34
A Minimal Pair of Linguistic Contexts

• A Truth Value Judgment task to assess children's
  interpretation of or in two different contexts.
• DE environment - 1st argument of Every
  Inclusive-or
• Every dwarf who ate a strawberry or a banana got
  a jewel
• Non-DE environment - 2nd argument of Every
  Exclusive-or
• Every dwarf ate a strawberry or a banana

35
Experiment 1 OR in the 1st argument of Every
Every dwarf who ate a strawberry or a
banana got a jewel
36
(No Transcript)
37
(No Transcript)
38
(No Transcript)
39
(No Transcript)
40
(No Transcript)
41
(No Transcript)
42
(No Transcript)
43

• Every dwarf who ate a strawberry or a banana got
  a jewel
• Fifteen English-speaking children (37 to 62
  mean 410).
• They accepted the target sentences 92 of the
  time.
• A control group (11 adults) accepted them 96 of
  the time.

44
Scalar Implicatures in Child Language

• 1. Do children ignore SIs in DE contexts?
• Answer YES, they allowed the inclusive-or
  reading
• 92 of the time
• 2. Do children compute SIs in non-DE contexts?

45
Experiment 2 OR in the 2nd argument of Every
Every space-guy took a strawberry or an onion ring
46
(No Transcript)
47
(No Transcript)
48
(No Transcript)
49
(No Transcript)
50

• Every space-guy took a strawberry or an onion
  ring
• Fifteen English-speaking children (34 to 62
  mean 52)
• They accepted the target sentence 50 of the
  time. A control group (8 adults) always rejected
  the test sentences.

51
a clarification on childrens responses

• Group I - Seven children rejected the target
  sentence 26 times out 28 trials (92.8).
• Group II - Seven children rejected the target
  sentence 2 times out of 28 trials (7.2).

52
Scalar Implicatures in Child Language

• 1. Do children ignore SIs in DE contexts?
• Answer YES, they did so 92 of the time
• 2. Do children compute SIs in non-DE contexts?
• Answer YES, some of them did so 93 of the time

53
Summary

• In keeping with the Semantic Core Model, the
  (overall) acceptance of the inclusive-or reading
  of disjunction drops from 92 in the DE context
  to 50 in the non-DE context.
• The non-adult responses of some children require
  further investigation.

54
Possible sources of children's non-adult
responses

• Pragmatic delay hypothesis (Noveck 2000).
• This hypothesis is associated with the standard
  view of scalar implicatures. It maintains that
  children lack certain pragmatic knowledge, in
  that "adults are more likely to detect a
  violation of the quantity maxim than children."
• Processing limitation hypothesis (Reinhart 1999).
  
• This hypothesis assumes that "all linguistic
  knowledge is innate" but that children fail to
  demonstrate knowledge when a task exceeds their
  processing capacity. A case in point is the
  computation of scalar implicatures "children
  know that they have to () keep two
  representations in memory (). So they start the
  execution. But, their working memory is not big
  enough to hold the materials needed to complete
  the execution of this task."

55
Child Language

• Noveck (2000) Children lack scalar implicature
  for some
•  
• Some giraffes have long necks. children YES
  adults NO
• Some birds live in cages. children and adults
  YES
•  
• All birds live in cages. children and adults
  NO
•  
•  

56
The Felicity Judgment task - (Chierchia et al.,
2001)

• Children are presented with two alternative
  descriptions.
• Both descriptions are TRUE, but one is more
  informative.

Heres an animal
Heres a turtle
57
AND vs OR in the two arguments of EVERY

• Do children know that AND is more informative
  than OR in the external argument of every?
• Do children know that OR is more informative than
  AND in the internal argument of every?

58
Experiment 1 OR vs AND in the external argument
of Every
59
(No Transcript)
60
Id like to take this banana, but it wont fit
through the door!
61
(No Transcript)
62
Every space-guy took a strawberry or an onion ring
63

• Every space-guy took a strawberry or an onion
  ring.
• Every space-guy took a strawberry and an onion
  ring.
• 15 children (age from 32 to 60 mean 47)
  only rewarded the puppet who had used the
  conjunction and 93 of the time (56/60).

64
AND vs OR in the two arguments of EVERY

• Do children know that AND is more informative
  than OR in the external argument of every?
• Yes, they choose AND 93 of the time!
• Next Do children know that OR is more
  informative than AND in the internal argument of
  every?

65

Experiment 2 OR vs AND in the internal argument
of every
66
(No Transcript)
67
I want to help! Ill take this bunch of
flowers.
68
I also want to help! Ill look after this
turtle.
69
I want to help as much as I can!
Here is what I will take.
70
I am feeling lazy today. I will just
play with these teddy-bears
71
Thank you very much for your help! Let me
give you some water for those flowers
72
Thank you! Here is some water for the
turtle
73
Wow thats a lot of help! Let me give
you a big bottle of water
74
I dont think you need any waterIll
keep this bottle.
75
(No Transcript)
76
Every girl who picked a turtle and some flowers
received some water
Every girl who picked a turtle or some flowers
received some water
77
Every girl who picked a turtle or some flowers
received some water
78
Every girl who picked a turtle and some flowers
received some water
79
Every girl who picked a turtle or some flowers
received some water. Every girl who picked a
turtle and some flowers received some water. 15
children (ages from 46 to 61 mean 53)
expressed a preference 40 times out of 60 trials.
They favored disjunction 90 of the time on these
trials
80
AND vs OR in the two arguments of EVERY

• Do children know that AND is more informative
  than OR in the nuclear scope of every?
• Yes, they preferred AND 93 of the time!
• Do children know that OR is more informative than
  AND in the restrictor of every?
• Yes, when children had a preference, they chose
  OR 90 of the time!

81
Interpretation of the findings

• Children master all aspects of pragmatic
  knowledge that are presupposed by the computation
  of Scalar Implicatures.
• Children are able to compare alternative
  representations. Childrens non-adult responses
  in earlier studies might be due to an inability
  to construct alternative representations on-line.

82
Conclusions

• Childrens interpretation of scalar terms in DE
  and non-DE contexts is consistent with the
  Semantic Core Model
• Childrens failure to apply SIs in a Truth Value
  Judgment Task is consistent with a variant of the
  Processing Limitation Hypothesis

83
Entailment Relations Continuity

• The Continuity Assumption
• The entailment relations licensed by any DE
  operator are closely related to its meaning, and
  do not show cross-linguistic variation.
• The Continuity Hypothesis maintains that child
  language can differ from the local (adult)
  language only to the same extent that adult
  languages differ from each other (Pinker, 1984
  Crain 1991 Crain and Thornton, 1998).
• On this view, we expect Downward Entailing
  operators in child language to show the same
  semantic properties as DE operators in adult
  languages. Moreover, children should know the
  correspondence between licensing conditions of
  NPIs and entailment relations from the earliest
  stages of language development.

84
The quantified expression None of the Ns

• Both arguments of the quantified expression none
  of the are Downward Entailing. Gualmini and Crain
  (2002) conducted an experiment testing childrens
  understanding of entailment relations in the
  second argument of none of the.
• None of the pirates found the jewel or the
  necklace ?
• (a) none of the pirates found the jewel
• and
• (b) none of the pirates found the necklace

Cf. one of the
85
A Truth Value Judgment task was designed to
assess children's understanding of the DE
properties of none of the

• Test Sentence None of the pirates found the
  jewel or the necklace.
• Situation One pirate finds the jewel.
• None of the pirates found the jewel and none of
  the pirates found the necklace. FALSE
• None of the pirates found the jewel or none of
  the pirates found the necklace. TRUE

86
Test Sentence None of the pirates found the
jewel or the necklace
This is a story about an Indian who is hiding his
valuables, because he knows that there are
pirates nearby, and he needs to go out -- to
Costco
87
Good, everything is hidden away. now I can go
to Costco. I do hope buffalo is on sale.
88

• Look, an Indian camp! Im sure we can find a lot
  of good stuff here

89
Too bad! Just a knife
90
Me too! All I found was another knife!
91

• This is really a bad day for us pirates! Just
  knives!

92
Im going back to look one more time! I am sure
there is a jewel around here somewhere, and
maybe there is a necklace here too! I know he
shops at Costco
93

• Aha! I found a jewel!

94

• The story concludes, but there is a reminder of
  the events that took place
• Test sentence None of the pirates found the
  jewel or the necklace

95

• Results 30 English-speaking children (age 310
  to 510, mean age 47) rejected the test
  sentences 105 times out of 120 (88).
• A control group of 28 adults rejected them 99 of
  the time.

96
A Clarification of Childrens Responses

• What accounts for the few non-adult responses?
• Psychologically, if not ontologically, negation
  seems to require
• or at least to strongly prefer an
  affirmative context against
• which to operate. (Horn, 1989 172)
• Consider the following pair, adapted from Wason
  (1972)
• (a) 5 is not an even number.
• 4 is an even number 5 is not an even number.
• The statement in (b) seems less awkward.

97
A Clarification of Childrens Responses

• Group I, Test sentence None of the pirates stole
  the jewel or the necklace.
• Group I - 15 children (310 - 510 mean 407)
  rejected the target sentences 48 times out of 60
  trials (80).
• Group II Test sentence, with positive lead-in
• Every pirate found a knife, but none of the
  pirates stole the jewel or the necklace.
• Group II - 15 children (401 - 508 mean
  407) rejected the target sentences 57 times out
  of 60 (95).
• Conclusions
• (a) Some of Group I childrens acceptances could
  be due to the failure to satisfy the felicity
  conditions associated with the target sentence.
• (b) Sometimes there must be a special
  linguistic antecedent.
• (c) Children tend to be overaccepting when they
  are confused.

98
Another case study every

• A Truth Value Judgment task to assess children's
  interpretation of or in two different contexts.
• DE environment - 1st argument of Every
  Inclusive-or
• Every dwarf who ate a strawberry or a banana got
  a jewel
• Non-DE environment - 2nd argument of Every
  Exclusive-or
• Every dwarf ate a strawberry or a banana

99

• Boster and Crain (1993) tested childrens
  comprehension interpretation of or in the second
  argument of the universal quantifier every.
• c) Every man will choose a dog or a cat.
• d) ? Every man will choose a dog and every man
  will choose a cat

100
What about the first argument of every?
Test sentence Every lady that bought a banana
or a cartoon of eggs got a basket

• Every lady that bought a banana got a basket
  and
• every lady that bought a carton of eggs got a
  basket
• So, children should reject the test sentence if
  only the ladies who
• bought bananas got baskets say, if the
  ladies who bought eggs were
• given paper bags

101
Downward Entailment and Every
This is a story about some ladies at the market.
The farmer wants to give the ladies something to
put the food in.
102
I am sorry, but I do not have anything for that
huge pepper.
103
Do you have anything for me?
104
Sure, you can take this basket.
105
I can have a basket too.
106
Where can I put these eggs? If I put them in a
basket, they are gonna brake.
107
You can put them in a paper bag.
108
You can have a paper bag too.
109
(No Transcript)
110

• 18 Children (age from 311 to 58 - mean 51)
  participated in the experiment. Each child was
  presented with four trials
• Test sentence Every lady that bought a banana or
  a cartoon of eggs got a basket.
• Children rejected the target sentence on 68
  trials out of 72 (94)

111
Finally, c-command

• The news that Noam had not won was a surprise to
  some/any of the linguists.
• The news that Noam had won was not a surprise to
  some/any of the linguists.
• The bear who never laughed expected to find
  some/any dogs at the party.
• The bear who laughed never expected to find any
  dogs at the party.

• DE operators must c-command an NPI to license it

112
A cross-linguistic generalization

• The news that Noam had not won was a surprise to
  some/any of the linguists.
• La notizia che G. non ha vinto ha sorpreso
  alcuni/nessun linguisti.
• The news that Noam had won was not a surprise to
  any of the linguists.
• La notizia che G. ha vinto non ha sorpreso
  nessun linguisti.
• The bear who never laughed expected to find
  some/any dog at the party.
• Lorso che non rideva mai pensava di trovare
  qualche/nessun cane alla festa.
• The bear who laughed never expected to find any
  dog at the party.
• Lorso che rideva avrebbe mai pensato di
  trovare nessun cane alla festa
• DE operators must must be in a certain
  structural relationship (c-command)
• (a) to license an NPI
• (b) to guarantee the conjunctive interpretation
  of the disjunction operator.

113
Test sentences

• 1. Merlin The girl that stayed up late did not
  get a dime or a jewel
• 2. Merlin The girl that didnt go to sleep got
  a dime or a jewel
• Inclusive-or the conjunctive interpretation
• Exclusive-or interpretation is permitted

114
E.g., Japanese 1. negation c-commands or
Merlin Osokumade okiteita onnnanokowa
koin-mo houseki-mo morawanai darou. The
girl that stayed up late will not get a dime or a
jewel 2. negation doesnt c-command or
Merlin Nemuranak-atta onna-no-ko-wa koin-ka
houseki(-ka)-wo morau deshou.
The girl that didnt go to sleep will get a dime
or a jewel
115
Downward Entailment C-command
This is a story about two girls who each had a
tooth come out.
116
Well, I know I should go to sleep, but I really
want to see what the Tooth-Fairy looks like, so
Im gonna stay up.
117
Ok, I have some dimes and some
jewels!
118
Merlin The girl that stayed up late will not
get a dime or a jewel Merlin The girl that
didnt go to sleep will get a dime or a jewel
119
This girl is asleep, Ill give her a dime --
heres a jewel too
120
Why are you awake? You know you are supposed to
be asleep!
121
I know! But I wanted to see what you looked
like. But, I lost a big tooth, so you have to
give me something!
122
Ok, ok, calm down. Ill give you a
jewel
123
Merlin I said that the girl that stayed up late
would not get a dime or a jewel Merlin I said
that the girl that didnt go to sleep would get a
dime or a jewel
124
11 children (age from 46 to 59 mean 51)
participated in the first condition of the
experiment. I predicted that the girl that
stayed up late would not get a dime or a
jewel The child subjects rejected this statement
87 of the time.
125
11 children (age from 42 to 510 mean 51)
participated in the second condition of the
experiment. Each subject was presented with two
trials. I predicted that the girl that didnt
go to sleep would get a dime or a jewel The
subjects accepted this statement 90 of the time.
126
This concludes the brief history of Downward
Entailment in child language

• In child language, the property of DE governs
• I. The conjunctive interpretation of or, and
  the cancellation
• of scalar implicatures
• II. The licensing of the NPI any
• III. Yielding valid inferences from sets to their
  subsets
• IV. The structural notion of c-command is a
  prerequisite for I-III

127

• THANK YOU!