Mayr/Schmitt – Precedence in semantics

Topic and aims of this course in a nutshell This course deals with the treatment
of semantic e ects that are correlated with asymmetries in linear order (precedence)
rather than asymmetries in hierarchical structure (c-command). We discuss a number of
phenomena that seem to exihibit such e ects and investigate potential semantic accounts thereof. More speci cally, we will consider semantic accounts such as dynamic semantics and other discourse-representation-approaches (DRT) that are speci cally designed to deal with linear asymmetries, investigating (a) their empirical scope (how much data do they actually cover?) and (b) their conceptual foundations (how explanatory are such theories?). We will then raise the question whether the phenomena under investigation actually compel us to assume such theories or whether there are static systems that can explain the empirical facts just as well or even better. In the following, we give a more detailed outline of the empirical and theoretical issues at stake.
Precedence in dynamic semantics Dynamic semantics and other DRT approaches
o er a uni ed account of donkey anaphora and presupposition projection. In such anal-
yses, e ects of linear precedence typically found in these phenomena are recast as order
of context update e ects. The order of context update is determined by the lexical en-
tries for connectives. For instance, in the donkey anaphora example in (1) the pronoun
appears to covary in its interpretation with the inde nite even though the former is not
c-commanded by the latter. Dynamic semantics assumes the antecedent in (1) is added
to the context before the consequent. Thereby a discourse referent is established for the
context in which the consequent is later evaluated, allowing the pronoun to refer to this
discourse referent.
(1) If a farmer owns a donkey, he beats it.

Independent evidence for this type of approach comes from presupposition ltering. (2)
does not inherit the presupposition of its consequent that John used to smoke. The
reason is that the consequent is evaluated only after the antecedent has been added to
the context. This results in the tautologous conditional presupposition that if John used
to smoke, he used to smoke, i.e., the presupposition that John used to smoke gets ltered.

(2) If John used to smoke, he stopped.

Empirical questions about the dynamic treatment of precedence From what
was said above, some immediate questions arise: First, (3) where antecedent and consequent are reversed when compared to (1) cannot be a case of donkey anaphora. Recall that the lexical entry for the conditional requires that its antecedent be added to the context rst. But then no discourse referent has yet been established when the consequent is later evaluated. Why is (3) similar to (1) then? Is the inde nite generically interpreted?
(3) A farmer beats a donkey, if he owns it.

Second, an asymmetry for (4-a) and (4-b) is predicted: the former should presuppose that if John has cancer, he used to smoke. The latter should presuppose something stronger that John used to smoke|as the antecedent containing the presupposition trigger is added to the context rst. Why do both give rise to the strong inferene that John used to smoke then the so-called proviso problem?
(4) a. If John has cancer, he stopped smoking.
b. John has cancer, if he stopped smoking.

Finally, why is a covarying interpretation possible in (5-a) but not in (5-b)? Dynamic
semantics has nothing to say about bound variable interpretations and thus it cannot
be straightforwardly extended to precedence e ects found there|i.e., so-called crossover

(5) a. Every student likes his mother.
b. *His mother likes every student.

Theoretical issues for dynamic semantics In addition to such empirical questions,
dynamic semantics has come under attack for being non-explanatory. In short, the or-
der of update is claimed to be completely stipulated and thus non-explanatory. Is it
possible to make dynamic semantics explanatory? Moreover, other approaches to the
phenomena cited exist. For instance, E-type accounts of donkey anaphora work with a
static situation-based semantics, and moreover they have something to say about bound variable phenomena, too. The same is true of so-called continuation-based approaches, whose semantic composition itself has a left-right bias built in. Presupposition projection facts have been covered in trivalent semantics and static reformulations of the notion of (local) context thereby relegating order e ects to the realm of processing. None of these approaches, however, extends to all of the data discussed above. Do we need a uni ed account?
Concrete questions addressed in this course In this seminar we ask how precedence
e ects should be captured in semantics and in particular whether a uni ed approach is
desirable given the empirical situation. In the course of this we will discuss a wide variety
of anaphora phenomena and facts regarding presupposition projection. We will begin by
laying out the dynamic approach to these topics. Then we will ask whether dynamic
semantics is still needed by discussing some of the competing theories mentioned above. We will address possible and existing replies to these competing approaches from the dynamic camp. Finally, we raise the question as to which types of phenomena each of the approaches views as belonging to the same semantic class and whether this classi cation is plausible in light of the empirical facts.

Required background Introduction to formal semantics and some syntax.

Preparatory readings
 Donkey anaphora: Chierchia 1995, chapter 1. Elbourne 2005, chapter 2.
 Projection problem: Heim 1983. Schlenker 2008.
Schedule The seminar will take place in week 1 of the EGG summer school and will
consist of 10 sessions, i.e., two per day.
July 25: Introduction
 Session 1: donkey anaphora
 Session 2: the projection problem
July 26: Dynamic semantics
 Session 1: the dynamic system
 Session 2: descriptive and explanatory adequacy
July 27: Alternative approaches to anaphora
 Session 1: E-type anaphora
 Session 2: Continations
July 28: Alternative approaches to the projection problem
 Session 1: Local contexts as processing
 Session 2: Three-valued logic
July 29: Taking stock
 Session 1: Making dynamic semantics explanatory
 Session 2: Independent evidence for dynamic semantics?