Epistemic judgments in language
Behavioural data and perspectives for neurological validation
Alexandra Kratschmer - Valentina Bambini
Institut for Sprog, Litteratur og Kultur, Aarhus Universitet - Laboratorio di Linguistica, Scuola Normale Superiore, Pisa
How do we avoid this?
Behavioural study
Introduction
What is epistemicity in language?
Any aspect regarding truthfulness/probability of a linguistic content unit according to linguistic signals provided by
the speaker
Epistemic judgments can be placed on scales, e.g. going from “true” to false” (see Kronning, 2005)
Coding by the speaker: Concerning the content: epistemic modality
the force a speaker attributes to a content expressed in an utterance ("true", "probable", "necessarily true")
Decoding by the hearer/reader: Concerning the interpretation of the speaker’s linguistic signals conveying
his/her epistemic judgments – what did the speaker mean?
What is going on in our brains, when we decode the speaker’s signals concerning epistemic modality?
What is going on in our brains, when we try to decode utterances with literal and non-literal meanings?
Linguistic and Cognitive
Modelling of Epistemicity
Brain Mapping of Metaphor
Processing
(based on Kratschmer 2006)
(based on Bambini et al. 2006)
Subjects: 10 graduate students (5 F)
A possible modulation to study epistemic processing is
offered by the Italian verb sembrare (engl. “to seem”)
Sembrare contains the semantic aspect of epistemic
quantification, i.e., the speaker’s judgment of the truth
value (of the predication subordinate to sembrare)
epistemic factor:
non-literal
reading
That lawyer seems
a civil lawyer
That lawyer
seems an actor
That lawyer is a
shark
assertive
categorization
(non modalized)
categorization
modalized by
sembrare
comparison
based on
sembrare
metaphor
true
(T)
 true
 true
false
(TR: i.e. true
with a reserve)
(TC: i.e. true as
a comparison)
(F)
 true
(TC)
Cognitive components in
the neural network for
metaphor comprehension
Aims of the study
Verifying the actual distinctability of the four readings in contextualized examples
Detecting stimuli designing problems
Isolating “well functioning” examples qualifying as stimuli in neurolinguistic tests
Strengthening of theoretical assumptions
+
random between
0-1000 ms
A shark.
Task: Semantic association: choosing between
two adjectives presented AFTER the target
sentences; metaphor comprehension is an
implicit task (i.e., not metalinguistic)
Memory
Memory
Sensory-based
Sensory-based
processes
processes
3000 ms
ferocious
max
2500 ms
until
completion
Results:
Metaphors compared to
literal sentences activated
a distributed network of
cerebral regions
(disconfirming the clinical
hypothesis on the selective
involvement of the right
hemisphere), including
traditional language areas
plus regions responsible for
extra-linguistic higher order
cognitive functions (memory,
theory of mind, etc.)
Converging Perspectives
Studying the Neural Basis of Epistemicity
The literal/non literal readings of sembrare resemble the literal/figurative readings of metaphorical expressions
We hypothesize that epistemic judgment is a complex cognitive operation, both linguistic and extra-linguistic,
supported by a widely distributed neural network
Our aim is to identify such a neural network and the cognitive processes involved. Steps:
Behavioural studies
What are the epistemic modulations subjects are sensitive to?
What are the correct stimuli to modulate the epistemic ability in the brain?
fMRI study
geographical
Theory
ofof
mind
Theory
mind
Metaphor
Metaphor
comprehension
comprehension
Imagery
Imagery
Do you know what that lawyer is?
Stimuli: 43 groups of four sentences;
Did you hear the speech of that lawyer? He is an intern.
in each group the target word appears Did you hear the speech of that lawyer? He seems a civil lawyer.
Did you hear the speech of that lawyer? He seems an actor.
once with one of the four meanings;
Did you hear the speech of that lawyer? He is a shark.
words balanced for frequency and
constant sentence structure to prevent
confound effect; random order inside
each group to prevent mechanical
choices
Task: Determining the epistemic degree of the literal reading of the target sentence in its context:
true (T), true with a reserve (TR), true as a comparison (TC), false (F)
Cognitive components
expected in the neural
network for epistemic
judgment
Attention
Attention
Emotion
Emotion
Subjects: 11 students
s
Can we validate this cognitive postulate?
epistemic factor:
literal reading
That lawyer is an
intern
Our behavioural test proved that naive speakers are able
to operate subtle epistemic operations
We hypothesize that a complex cognitive system
underlies this ability (a folk epistemology system?):
language, theory of mind, attention...
We hypothesize that a widely distributed neural network
supports such a system
Our hypothesis is supported by results on metaphor
Experimental Design
m
The two readings can be classified as:
a) literal reading – categorization (with a reserve)
b) non-literal reading - comparison
On the basis of theoretical
work on epistemic modality,
we elaborated an epistemic
continuum
The epistemic continuum
Do you know what that fish is? A shark.
Do you know what that lawyer is? A shark.
Do you know what that politician is? A senator.
00
75
Consider the different values on a cognitive scale of truth
assessment:
a) Il computer sembrava rotto (“The computer seemed
broken”); formalized by an epistemic factor ” ≤ true”
b) Il computer sembrava senziente (reagiva come un
essere umano) (“The computer seemed conscious” (it
reacted like a human being)); formalized by an epistemic
factor “< true”
Stimuli: 40 sentence pairs; in each pair the
target word appears once
with a literal, once with a metaphorical
meaning; 40 fillers
We think that a good way to
modulate the epistemic
ability is to include the
literal/non literal distinction
into the true/false dichotomy
Conclusions
Language
Language
Semantic
Semantic
integration
integration
Language
Language
Syntactic
Syntactic
reorganization
reorganization
The study of mental processes underlying
epistemological skills may support the
hypothesis of a dedicated “folk
epistemology” system. Over the last
decades many cases of domain-specific
cognitive systems with dedicated
inferential resources have been identified
in infants and higher primates. These
include folk physics, folk biology, folk
arithmetics, folk psychology, among
others. Can the ability to judge truth and
epistemic reliability be accounted for in
terms of a folk epistemology system?
(European Review of Philosophy 8 (2008),
Heintz et al., 2008)
Results
High statistical correlation
between intended and
decoded readings (range:
94,71 - 98,23 %)
There is a context-related
distinctability of the four
readings
References
100,00
98,00
96,00
94,00
92,00
90,00
The four categories lie on an epistemic continuum
 V and F lie each on their own end of the scale: higher correct decoding%
 VR and VP are adjacent: lower correct decoding %
Strengthening of theoretical assumptions
10 of the proposed example groups sanctioned as “well functioning”; other
example groups adjustable
T
TR
TC
F
Bambini,V., C. Gentili, E. Ricciardi, P. Pietrini. 2006. Neural Correlates of
Metaphor Comprehension Assessed by fMRI. Human Brain Mapping.
Kratschmer, A. 2006. Catégorisation vs comparaison : une question de
quantification épistémique. Propos de modèle interprétatif semanticopragmatique modulaire des constructions italiennes avec sembrare/parere.
To appear in : Cahiers Chronos.
Kronning, H. 2005. Polyfoni, modalitet och evidentialitet. Om epistemiska
uttryck i franskan, särskilt epistemisk konditionalis. Arbejdspapirer 3,
Sprogligt Polyfoninetværk. RUC. 71-99.
Acknowledgments
This project is funded by the research priority area Cognition,
Communication & Culture (CCC), Faculty of the Humanities, Aarhus
University
Contact
Alexandra Kratschmer: romak@hum.au.dk
Valentina Bambini: v.bambini@sns.it