(Adjunct) islands and the finiteness effect
description
Transcript of (Adjunct) islands and the finiteness effect
(Adjunct) islandsand the finiteness effect
Dan Michel Grant Goodall
UC San Diego1
Overview of talk
What does it all mean?About islands? About grammar and processing?
Exp. 6: Complement clauses
Finiteness effect everywhere?
Exp. 2-5: Eliminating a confoundIs it finiteness or the overt argument?
Exp. 1: Adjunct islands
Is there a finiteness effect?
Islands and finiteness
Why might finiteness matter to islands?
Islands
• Domains in which gap is not possible, despite earlier filler
wh-phrase … [ … __ … ] …
*What did Mary eat pie [while John drank _ ]?
Two views of islandsAccumulationIslands result from accumulation of several
independent processing difficulties (filler-gap dependency, complex syntactic structure, etc.)
E.g. Kluender (1998, 2004), Hofmeister & Sag (2010)
DisruptionIslands result from otherwise unproblematic
element that may disrupt filler-gap dependency (bounding node, intervening feature, etc.)
E.g. Ross (1967), Chomsky (1986), Rizzi (2004)
Things to keep in mind
• These two views are not mutually exclusive.
• Accumulation often associated with processing.Disruption often associated with grammar.
But these associations aren’t logically necessary.
Role of finiteness in islands
• Finiteness has been claimed to be important for wh-islands:
a book which I can’t figure out…a. [what to do about __]b. ?? [what I should do about __]
(from Ross (1967))Finiteness effect:Finite clause is
more resistant to gap.
Role of finiteness in islands
• And for subject islands:
We investigated what [the campaign…a. ?*to preserve __ ] had harmed the forest.b. *that preserved __ ] had harmed the forest
(adapted from Phillips (2006))Finite clause is
more resistant to gap.
Role of finiteness in islands
• Adjunct islands are less often discussed. Many have assumed there is no effect.
Who did John go home…a. ?? [after kissing __] b. * [after he kissed __]
(See Szabolcsi (2006), Truswell (2011))Finite clause is
more resistant to gap?
How to view the finiteness effect?AccumulationIslands result from accumulation of several
independent processing difficulties (filler-gap dependency, complex syntactic structure, etc.)
E.g. Kluender (1998, 2004), Hofmeister & Sag (2010)
DisruptionIslands result from otherwise unproblematic
element that may disrupt filler-gap dependency (bounding node, intervening feature, etc.)
E.g. Ross (1967), Chomsky (1986), Rizzi (2004)
FinitenessIntrinsically difficult for processing.Should see finiteness effect very generally.
Suggested in Kluender (2004), Hofmeister (2007).
FinitenessNot intrinsically difficult. Should see effect with some dependencies.
Suggested in Cinque (1990), Manzini (1992), Truswell (2011).
Non-Ameliorating Ameliorating
y/n-Q Who did the carpenter restore the antique table after…
he negotiated with _? negotiating with _?
wh-Q Did the carpenter restore the antique table after…
he negotiated with the buyer? negotiating with the buyer?
5.69 (1.09) 5.72 (1.07)p = 0.68
y/n
wh
n.s.
***
What accumulation looks like
n.s.
***
y/n
wh
Main Effect
***
Series11
2
3
4
5
6
7
wh
y/n
Accumulation
Non-Ameliorating Ameliorating
y/n-Q Who did the carpenter restore the antique table after…
he negotiated with _? negotiating with _?
wh-Q Did the carpenter restore the antique table after…
he negotiated with the buyer? negotiating with the buyer?
5.69 (1.09) 5.72 (1.07)p = 0.68
y/n
wh
n.s.
***
What disruption looks like
n.s.
***
y/n
wh
Main Effect
***
Series11
2
3
4
5
6
7
wh
y/n
Disruption
Format for experiments
• 195-220 participants, all UCSD students.• Non-native or non-English-dominant speakers
excluded.• 2 x 2 design, where one factor is question-
type: wh- vs. yes/no question• Each participant sees at least 4 tokens of each
type, mixed with at least 40 fillers.• Latin square design, randomized. • Acceptability judgment task, 7-point scale
Experiment 1: Adjunct islands
• Do adjunct islands also have finiteness effect?
• Both make similar predictions. If they are both on the wrong track, we need to know!
Accumulation: Yes, definitely!If finiteness is intrinsically difficult, it should be here too.
Disruption: Yes, probably.If finiteness disrupts wh-dependencies elsewhere, it probably will here also.
Ex 1 Finite Non-finite
y/n-Q Did the carpenter restore the antique table after…
he negotiated with the buyer?
negotiating with the buyer?
wh-Q Who did the carpenter restore the antique table after…
he negotiated with _? negotiating with _?
he negotiated negotiating
Experiment 1: Adjuncts
Ex 1 Finite Non-finite
y/n-Q Who did the carpenter restore the antique table after…
he negotiated with _? negotiating with _?
wh-Q Did the carpenter restore the antique table after…
he negotiated with the buyer? negotiating with the buyer?
5.69 (1.09) 5.72 (1.07)p = 0.68
y/n
wh
n.s.
***
he negotiated negotiating
Experiment 1: Adjuncts
n.s.
***
y/n
wh
Main Effect
***
Series11
2
3
4
5
6
7
***
n.s.
Main Effect
***
wh
y/n
Yes.
• Finiteness often co-occurs with the presence of an overt subject.
… after he negotiated …… after negotiating …
• Is the finiteness effect due to: – Finiteness itself? – The extra argument (subject)?
A confound
Yes.
Exp. 2: CNPCFiniteness constant / ± extra argument
Exp. 3: Subject islandExp. 4: Complement clause
Exp. 5: CNPCExtra argument constant / ± finiteness
Ex 2 Overt argument No overt argument
y/n-Q Does the principal like the thought of …
the children learning subtraction in kindergarten?
learning subtraction in kindergarten?
wh-Q What does the principal like the thought of …
the children learning _ in kindergarten?
learning _ in kindergarten?
the children
Experiment 2: CNPC
Ex 2 Overt argument No overt argument
y/n-Q Who did the carpenter restore the antique table after…
he negotiated with _? negotiating with _?
wh-Q Did the carpenter restore the antique table after…
he negotiated with the buyer? negotiating with the buyer?
5.69 (1.09) 5.72 (1.07)p = 0.68
y/n
wh
n.s.
***
the children
Experiment 2: CNPC
Series11
2
3
4
5
6
7
y/n
wh
Main Effect: ***
Series11
2
3
4
5
6
7Main Effect: ***
y/n
wh
Interaction:n.s.
Ex 3 Overt argument No overt argument
y/n-Q Does the prosecutor know that …
the defendant presentingthe child's testimony will convince the jury?
presentingthe child's testimony will convince the jury?
wh-Q What does the prosecutor know that …
the defendant presenting_ will convince the jury?
presenting_ will convince the jury?
the defendant
Experiment 3: Subject islands
Ex 3 Overt argument No overt argument
y/n-Q Does the prosecutor know that …
the defendant presentingthe child's testimony will convince the jury?
presentingthe child's testimony will convince the jury?
wh-Q What does the prosecutor know that …
the defendant presenting_ will convince the jury?
presenting_ will convince the jury?
the defendant
Experiment 3: Subject islands
Series11
2
3
4
5
6
7
y/n
wh
Main Effect: n.s.
Interaction:n.s.
Ex 4 Overt argument No overt argument
y/n-Q Does the architect want …
the contractor to see the building plans before Monday?
to see the building plans before Monday?
wh-Q What does the architect want …
the contractor to see _ before Monday?
to see _ before Monday?
the contractor
Experiment 4: Complements
Ex 4 Overt argument No overt argument
y/n-Q Does the architect want …
the contractor to see the building plans before Monday?
to see the building plans before Monday?
wh-Q What does the architect want …
the contractor to see _ before Monday?
to see _ before Monday?
the contractor
Experiment 4: Complements
Series11
2
3
4
5
6
7Main Effect: ***
Interaction:n.s.
No Main Effect of Question: Complement clauses are not islands
y/nwh
CNPC Subject islands Complement Cl
Main effect of additional overt argument?
YES No, but same pattern YES
Effect of additional overt argument specifically in WH?
NO NO NO
Series11
2
3
4
5
6
7
Series11
2
3
4
5
6
7
Series11
2
3
4
5
6
7Eliminating a confound
Yes.
Exp. 2: CNPCFiniteness constant / ± extra argument
Exp. 3: Subject islandExp. 4: Complement clause
Exp. 5: CNPCExtra argument constant / ± finiteness
General preference for fewer arguments, but not specific to wh-dependency.
Ex 5 Finite Non-finite
y/n-Q Do many people believe the idea …
that the squirrels buried extra food under bushes?
of the squirrels burying extra food under bushes?
wh-Q What do many people believe the idea …
that the squirrels buried _ under bushes?
of the squirrels burying _ under bushes?
buried
Experiment 5: CNPCburying
Ex 5 Finite Non-finite
y/n-Q Do many people believe the idea …
that the squirrels buried extra food under bushes?
of the squirrels burying extra food under bushes?
wh-Q What do many people believe the idea …
that the squirrels buried _ under bushes?
of the squirrels burying _ under bushes?
buried
Experiment 5: CNPCburying
Series11
2
3
4
5
6
7
Interaction: *
y/n
wh
***
**
Ex 5 Finite Non-finite
y/n-Q Do many people believe the idea …
that the squirrels buried extra food under bushes?
of the squirrels burying extra food under bushes?
wh-Q What do many people believe the idea …
that the squirrels buried _ under bushes?
of the squirrels burying _ under bushes?
buried
Experiment 5: CNPCburying
Series11
2
3
4
5
6
7
Non-finitepreferred onlyin wh condition
y/n
wh
***
**
Yes.
Exp. 2: CNPCFiniteness constant / ± extra argument
Exp. 3: Subject islandExp. 4: Complement clause
Exp. 5: CNPCExtra argument constant / ± finiteness
General preference for fewer arguments, but not specific to wh-dependency.
Preference for non-finite only in wh-dependency.
It’s finiteness.
Where do we stand at this point?AccumulationIslands result from accumulation of several
independent processing difficulties (filler-gap dependency, complex syntactic structure, etc.)
E.g. Kluender (1998, 2004), Hofmeister & Sag (2010)
DisruptionIslands result from otherwise unproblematic
element that may disrupt filler-gap dependency (bounding node, intervening feature, etc.)
E.g. Ross (1967), Chomsky (1986), Rizzi (2004)
FinitenessIntrinsically difficult for processing.Should see finiteness effect very generally.
Suggested in Kluender (2004), Hofmeister (2007).
FinitenessNot intrinsically difficult. Should see effect with some dependencies.
Suggested in Cinque (1990), Manzini (1992), Truswell (2011).
One version of disruption view
• Truswell (2011): Event Locality Condition (roughly)
Filler and gap must be within single event.Adjuncts:
Finite → independent eventNon-finite → possibly part of main clause event
Complements (of bridge verbs):Finite and non-finite: part of main clause event
Prediction: Finiteness disrupts wh-dependencies in adjunct clauses. Confirmed in Experiment 1.
Prediction: But not in complement clauses. To be tested in Experiment 6!
Ex 6 Finite Non-finite
y/n-Q Did the children believe…
the guest was bringinga cake?
the guest to be bringinga cake?
wh-Q What did the children believe…
the guest was bringing _? the guest to be bringing _?
was to be
Experiment 6: Complements
Ex 6 Finite Non-finite
y/n-Q Did the teacher report …
the students knew algebra?
the students to know algebra?
wh-Q What did the teacher report …
the students knew _? the students to know _?
was to be
Experiment 6: Complements
Series11
2
3
4
5
6
7 Main Effect: **
y/nwh
Interaction:n.s.
Finite > Nonfinite
Yes.
It’s finiteness.
No. Only in islands.
Back to the beginningAccumulationIslands result from accumulation of several
independent processing difficulties (filler-gap dependency, complex syntactic structure, etc.)
E.g. Kluender (1998, 2004), Hofmeister & Sag (2010)
DisruptionIslands result from otherwise unproblematic
element that may disrupt filler-gap dependency (bounding node, intervening feature, etc.)
E.g. Ross (1967), Chomsky (1986), Rizzi (2004)
Finiteness effect is most consistent with disruption view. It occurs with: -wh-dependencies (and not generally) -islands (and not complements)
Extra argument effect is most consistent with accumulation view. It occurs everywhere.
Grammar or processing?
Given the usual associations:Accumulation often associated with processing.Disruption often associated with grammar.
It is tempting to conclude that:Extra argument effect is a processing effect.Finiteness effect is a grammatical effect.
If so, islands are (partly) a grammatical effect.
However…
• This conclusion could change if disruption is shown to be due to processing.
• The extra argument effect does seem to be due to processing, and this degrades some already bad island violations. – So processing effects do play a role in the
unacceptability of some island sentences.
Summary of findings
Yes.
It’s finiteness.
No. Only in islands.
Finiteness → grammarExtra argument→ processing
Special thanks to:Chris Barkley Boyoung KimIvano Caponigro Robert KluenderGabe Doyle Emily MorganSimone Gieselman
Thank you!
grammar.ucsd.edu/syntaxlabResearch assistants:Adrienne LeFevreMichelle McCadden