Home / questions about bias / questions and bias

questions and bias - questions about bias

questions and bias-questions about bias

Mica questions and bias
The Italian particle mica co-occurs with negation in polar questions to signal a `bias' on the part of the
speaker; its presence further affects the distribution of such questions. This is illustrated by the Italian
minimal pair in (1), which are differentiated by contexts A and B; these contexts systematically manipulate
Clara's prior expectations as well as recent evidence.
(1) Context A: Clara invites Luca over for dinner and cooked many things. However, Luca barely
touches any food. Clara asks him:
a. ! Non hai mica gia' mangiato? b. # Non hai gia' mangiato?
NEG have.2.sg MICA already eaten NEG have.2.sg already eaten
`Have you already eaten?' `Have you not already eaten?'
(implies: C. is disappointed that L. ate.)
c. # Did you not already eat?
(2) Context B: Clara invites Luca over for drinks late in the evening. However, Luca gets there and asks
if she has any food. Clara asks: (1a): #. (1b): !. (1c): !.
We propose that both Italian negative questions and mica questions signal a mismatch between the ques-
tioner's prior expectations and recent evidence, but achieve this by different means. Italian negative ques-
tions, like English `inner negation' polar questions such as (1c) (Ladd 1981, Bu?ring and Gunlogson 2000,
van Rooy and Safarova 2003, AnderBois 2011a) imply that there is recent evidence for the negative content
proposition, and a prior expectation for the truth of negation's prejacent; this is accomplished via pragmatic
reasoning along the lines of van Rooy and Safarova (2003). While the kind of mismatch signaled by mica
questions is the same as regular negative questions, there are two key differences, the direction of the mis-
match, and its strength. The first puzzle, then, is how it is that mica reverses both of these implications
from a regular negative polar question: Clara in (1a) implies that she had hoped Luca wouldn't eat, but that
his behavior suggests that he has. Moreover, the bias in plain negative questions is described as `weak',
but the bias involved in mica questions is intuitively stronger: while negative questions are compatible with
very weak prior expectations and weak evidence, mica questions involve a much stronger degree of prior
expections. (The strength of the bias resembles that of English outer negation questions, e.g. didn't you
already eat? (Ladd 1981, Romero and Han 2004, AnderBois 2011a), though the content is much different.)
Question type prior expectation recent evidence
(3) Data summary: [non [mica [ ]]] ?
[non [ ]] ?
Syntax Prior syntactic work (Cinque 1976, Zanuttini 1997, Garzonio and Poletto 2009) has shown that
mica is morphosyntactically a negative element: it can appear in a concord chain with non, and its licensing
is parasitic on non. In polar questions mica can also appear in sentence-initial position without non, and
without any meaning difference, as in (4). Following Cinque (1976) we assume that `autonomous' mica has
moved to Spec,NegP and serves as the only realization of negative features in the chain.
(4) Mica hai gia' mangiato? (`Autonomous' mica) (5) Mica questions, Logical Form
Cinque's proposal implies that autonomous and non-autonomous NegP
mica both involve the same features at LF, and we will adopt this
idea here. Moreover, we will assume that at LF, mica always ap- mica
pears in the specifier position that it surfaces in when autonomous, [uNEG,iMICA] Neg
leading to the LF structure in (5). (non)
Semantics & pragmatics Our proposal has two parts. First, bias in Italian plain negative polar questions
is a pragmatic inference (following several authors on English inner negation questions; van Rooy and
Safarova 2003, Farkas and Bruce 2010, AnderBois 2011b). Second, mica in a question compositionally
contributes a biasing implication, blocking the pragmatic computation of negative question bias.
The general idea of pragmatic accounts of negative questions is that a hearer reasons about why a speaker
would choose a negative prejacent over a positive one. We will adopt van Rooy and Safarova's 2003 account
in particular: a negative question signals that the `utility value' of the negative proposition is higher than
that of the positive. (We set aside the utility-based account of outer negative questions, which do not pattern
with any of the types discussed here.) In the present examples, utility corresponds to informativity/surprisal,
and so in a negative polar question, the negative proposition has a higher surprisal value than the positive
one, relative to a previous belief state. The utility value account thus derives that regular negative questions
imply a prior expectation for . While van Rooy and Safarova do not directly address the recent evidence
component in their formalization, this also follows from straightforward pragmatic reasoning: if a speaker
signals that they had a prior expectation that , but is asking a question that may be resolved with ? , then
they must have had some reason to revise that prior expectation downwards.
The particle mica, on the other hand, clearly makes some compositional contribution. While we have
set aside data outside of questions, mica can appear in other speech act types, including assertions and
commands, with a similar meaning. We propose that mica has grammaticalized a version of the prior
expectation component: it implies that its prejacent was expected on a prior belief state of the questioner.
While it is unclear what variety of implication this is (it is non-at-issue and non-defeasible), for the sake of
discussion we will treat it as a presupposition following the syntactic literature (Cinque 1976), and a version
of this is formalized in (6).
(6) mica w,t,c= p st . p presupposes: t < t : w Doxsc(w)(t ) : p(w )
The presupposition is that there is a past set of doxastic alternatives for the speaker (sc) where p holds.
Since we are assuming an LF where mica scopes over negation, p is always negative (i.e. p = ? in
the table above), and therefore the strong prior expectation component of mica questions follows. This
presupposition blocks the pragmatic reasoning sketched above: the presupposition is simply incompatible
with the inference that the surprisal value of the negative proposition p is high, and in fact predicts that
it should be low. This also leads to a reversal in the recent evidence inference, which we assume is also
derived pragmatically for mica questions (which are still a marked form): if the speaker's prior expectation,
but not necessarily present one is in support of ? , and they are asking a question about , then they must
have received some evidence to weaken that past belief.
Further predictions While mica questions often have a bouletic component aligned against , our account
predicts that this is not necessary. Two uses of mica show that this predictions is correct: polite requests, and
incredulous questions. In polite questions the prior expectation is used to provide the hearer with an `out'
for a request, by indicating that that expectation was correct. In incredulous questions, the prior expectation
simply signals a strong mismatch with the present scenario.
(7) Non avresti mica una sigaretta?
NEG have.CON.2sg MICA a cigarette? `Do you have a cigarette by any chance?'
(8) Context C: Mario likes buying lottery tickets. His sister Laura knows that his chances of winning
are slim. One day, Mario comes home with lottery ticket in one hand and champagne in the other.
Laura: Non hai mica vinto la lotteria?
NEG have MICA win the lottery `Have you won the lottery??'
In summary, we have given a semantic/pragmatic account of a novel type of biased question in Italian,
marked by the negative element mica. The proposal is that mica grammaticalizes a pragmatic inference
found with ordinary negative questions, that the questioner's prior epistemic state made the prejacent of mica
highly likely. In contributing this inference non-defeasibly in combination with negation, mica consequently
causes a reversal in reasoning about what sort of recent evidence would lead to asking a question, and thus
the difference between negative and mica questions follows directly.
Selected references: Cinque, G.: 1976, "Mica". Annali della Facolta` di Lettere e Filosofia dell'Universita`
di Padova 1. ? Ladd, R.: 1981, A first look at the semantics and pragmatics of negative questions and tag
questions. Proc. CLS 17. ? Van Rooy, R. and M. Safarova: 2003, On Polar Questions. Proc. SALT 18.

What are the categories of bias? What types of bias are there in statistics? Selection bias. Self-selection bias. Recall bias. Observer bias. Survivorship bias. Omitted variable bias. Cause-effect bias. Funding bias.