Florian Schwarz & Sonja Tiemann
The Path of Presupposition Projection in Processing: The Case of Conditionals Introduction
Eye Tracking Study: PSP Projection from Conditionals
Questions about Presupposition (PSP) Projection: • What mechanism can model it most successfully? • Does it incur online processing costs? • Schwarz & Tiemann (2012): eye tracking evidence that PSP projection out of the scope of negation takes time • Present Study: The time course of resolving PSP of wieder (‘again’) in the consequent of conditionals to global (discourse) vs. local context (antecedent of conditional)
Theoretical Background
• Context S (Sc) + conditional • PSP trigger wieder (‚again‘) in consequent • PSP support in (a) antecedent or (b) global context
• wieder scoping below or above negation • 24 items, 32 subjects, 52 unrelated filler items • Eye tracking using EyeLink 1000
CONTEXT: Tina war letzte Woche {(I) ∅ / (II) nicht} Schlittschuh laufen.
Structure of Factors: Condition Firstword Context Location LOCAL a WN I GLOBAL b NW I GLOBAL c WN II LOCAL d NW II
Tina was last If
(B)
DRT Hypothesis: Effort ≈ Projection Path length
II
~
ice-skating
{(I) nicht / (II) ∅} Schlittschuhhaufen war, not
ice-skating
was,
TARGET: dann geht sie heute bestimmt ... then goes she today certainly ...
(WN) ... wieder nicht ... (NW) ... nicht wieder ... Schlittschuhlaufen not
not
again
ice-skating
Predictions
I
p x went ice-skating before ⇒ x went ice-skating before (A2)
not
she yesterday
again
LOCAL
x Tina(x) c x went ice-skating before (C)
week
Wenn sie gestern
PSP projection in DRT (1) Sc. If p, she didn‘t go ice- skating again today.
• Predictions of DRT Hypothesis supported • Context * Location Interaction • no sig. difference between d and c
Methods & Material
Focus on two approaches: dynamic semantics vs. DRT Projection Path defined on discourse representations:
Summary of Results
a = 1 (A2-B) ∧
d = 2 (A1-B)
GLOBAL
< =
b = 3 (A1-C) ∨ c = 2 (A2-C)
Context * Location interaction (+ main effect of Location)
Processing Hypothesis: Effort ≈ Antecedent distance (clauses) LOCAL
a=1 = II d = 1 I
Dynamic Complexity Hypothesis: Effort ≈ Update Complexity (Negation) LOCAL GLOBAL I a = 4 (NegPSP) > b = 2 (PosPSP) ∨ ∧ II d = 1 (PosPSP) < c = 3 (NegPSP)
GLOBAL
<
b, d < c)
x went ice-skating today x went ice-skating before
Van der Sandt (1992), Geurts (1999)
(2) c + If p, q = c – ((c+p) – ((c+p)+q)) (defined iff (c+p) + PSPq = (c+p) • PSPq evaluated relative to c+p No way to determine location of support for PSP! Two Possible Processing Factors: • Negated PSP more complex harder to process: o PSPq = ~r: c + ~r = c? ≈
c – (c+r) = c?
o PSPq = r:
c + r = c?
• Distance of anaphoric trigger from antecedent, parallel to pronouns, plausibly measured in clauses (independent from but compatible with dynamic account)
• Further support for Schwarz & Tiemann’s (2012) proposal that PSP projection takes time and effort in online processing
General Discussion • Evidence for processing costs of PSP projection • Supports theories that derive presupposition projection by a sequence of manipulations on representations, such as DRT • Dynamic theories alone can‘t determine any distance measure as the location of the antecedent can‘t be determined (local contexts are always combination of global + local information)
• DRT Projection Path distance predicts processing effects extremely well
Local resolution of the PSP precedes global resolution • Dynamic context update of conditionals (Heim 1983):
• … and NOT … o … with clause distance, as on a plausible Processing Hypothesis, or o … with dynamic update complexity ( negation)
• A pure processing account in terms of distance from antecedent can‘t account for the interaction either (unless negation has a special and hitherto unknown role in this regard)
(A1)
PSP projection in Dynamic Semantics
• Processing effort seems to correlate with length of DRT Projection Path…
Main Results
• How do these results relate to other theories?
Significant Context * Location Interaction for Total Time, Regression Path Duration, and Rereading Duration on both the Verb and on {wieder nicht}. Also: • Main effect of Location
• Simple effect for
• Main effect of Context (marginal or non-sig for some measures; dominated by interaction)
o a < b, but not d < c o a < d (for Reg Path) o c < b (numerical only)
Furthermore: • Main effect of Firstword (NW > WN) in Firstword * Location analysis (predicted only by DRT) • Analysis with DRT Projection Path Length as sole predictor o Equally good as Context * Location analysis o Better than analysis in terms of Location only
Reading Measure Definitions • Total Time: Sum of all fixations on target region • Regression Path Duration: Sum of all fixations from first entering the target region until exiting it to the right • Rereading Duration: Sum of fixations that return to preceding text after first entering target region up unto first fixation to the right of target region
References I. HEIM (1983), “On the Projection Problem for Presuppositions”, Proceedings of the Second West Coast Conference on Formal Linguistics: 114-125.. ::: R. VAN DER SANDT (1992), “Presupposition Projection as Anaphora Resolution”, Journal of Semantics 9: 333-377. ::: F. SCHWARZ & S. TIEMANN (2012.), “Presupposition Processing - The Case of German wieder”, Proceedings of Amsterdam Colloquium 2011. ::: B. GEURTS (1999), Presuppositions and Pronouns. Elsevier. Amsterdam, New York.