Both theories derive some implicatures (baseline > 0)
Both theories trigger implicature for "some" in UE context
Both theories derive triggered (not just possible) for "some"
Neither theory triggers implicature for "all" (top of scale)
NeoGricean predicts DE blocking
RSA predicts DE blocking (via @cite{potts-etal-2016} lexical uncertainty)
Both theories now predict DE blocking
NeoGricean predicts task effect (contextualism)
RSA model incomplete: doesn't handle task effects
NeoGricean models task effect, RSA incomplete
Observed verification task rate from @cite{geurts-pouscoulous-2009}
Instances For
NeoGricean baseline (35%) is close to observed (34%)
RSA baseline (50%) is far from observed (34%)
NeoGricean closer to @cite{geurts-pouscoulous-2009} data than RSA
Both theories now agree on DE blocking
Agreement check on task effect
Rate matching using interface function
RSA doesn't match data within tolerance
NeoGricean is closer to data
NeoGricean full coverage report
Equations
- One or more equations did not get rendered due to their size.
Instances For
RSA full coverage report
Equations
- One or more equations did not get rendered due to their size.
Instances For
NeoGricean has no incomplete phenomena (but some out of scope)
RSA has incomplete phenomena that could be extended
RSA covers reference games (NeoGricean doesn't claim to)
NeoGricean explicitly marks reference games as out of scope
NeoGricean coverage summary (legacy)
Equations
Instances For
RSA coverage summary (legacy)
Equations
Instances For
@cite{geurts-pouscoulous-2009} task effect data as a test case
Equations
- Comparisons.ScalarImplicatureTheories.geurtsPouscolousTaskEffect = { inferenceRate := 62, verificationRate := 34, significant := true }
Instances For
DE blocking test case from ScalarImplicatures data
Equations
- One or more equations did not get rendered due to their size.
Instances For
Test NeoGricean against DE blocking data
RSA captures DE blocking (via @cite{potts-etal-2016} lexical uncertainty)
NeoGricean matches task effect pattern
RSA model incomplete for task effect (can't represent QUD manipulation)
Compare both theories to observed verification rate
NeoGricean captures the Geurts & Pouscoulous task effect data
Equations
- One or more equations did not get rendered due to their size.
NeoGricean captures DE blocking pattern
Equations
- One or more equations did not get rendered due to their size.
RSA captures DE blocking pattern (via @cite{potts-etal-2016})
Equations
- One or more equations did not get rendered due to their size.