(Non-)Interpetation of Subordinate Tense



(Non-)Interpetation of Subordinate Tense
Göttingen, 9.11.2009
The talk deals with the interpretation of subordinate tense, in particular with SOT context.
a. John thought Mary was sick.
b. Russ. Ivan dumal, chto Masha byla bol'na.
In an SOT-language like English, Past under Past(Past\Past) may have a simultaneous
interpretation, i.e., we have temporal agreement. In a non-SOT language we only have the
shifted interpretation.
Semantic source of temporal agreement: Temporal agreement only occurs if the
agreement tense is not interpreted.
a. Past John thought ∅-tense Mary was sick
(non-local agreement)
b. Past Ivan dumal ∅-tense Past Masha byla bol'na
(local agreement)
In English, the temporal morphology of the embedded verb is determined by the matrix
tense via a binding chain through the verbal quantifier ‘think’. In Russian, ‘dumat'/think’
breaks the binding chain. The morphology of the embedded verb is determined by an
embedded relative Past. The difference between English and Russian is derived from:
The SOT-parameter
A language L is an SOT-language if and only if the verbal quantifiers of L transmit
temporal features.
Verbal quantifiers quantify over times (e.g. fut. will) or world-times (e.g. verba dicendi et
sentiendi like believe). The SOT-parameter explains the different distribution of subordinate
tenses in SOT and non-SOT languages (e.g. the expression of simultaneity as Past under
Past in SOT, as Present under Past in non-SOT languages).
The talk will take up a recent challenge by Olga Khomitsevich against existing
accounts: verbs of perception (e.g. notice) and, occasionally, factive verbs (e.g. know)
express simultaneity by Past under Past. We will show that at least perception are not verbal
quantifiers and hence not subject to the SOT-parameter.
Göttingen, 9.11.2009
The talk will give precise formulation of the syntax and semantics of the
constructions involved. The essential idea is that temporal features are transmitted via
semantic binding; verbal quantifiers may or may not act as barriers for feature transmission.
The main tense of an attitude complement will be a semantically empty PRO, that of a
relative clause will be an obligatorily bound temporal pronoun Tpro.
Three different types of constructions: (i) Tense in relative clauses; (ii) tense in complement
clauses, i.e., tense under attitudes; (iii) tense in adjunct clauses, notably before/after-clauses.
2.1. Tense in Relative Clauses
With a Past tense matrix verb, English and Russian relative clauses mostly behave in a similar
way (different from Japanese).
Present under Past: Simultaneous reading available? (cf. (Kusumoto, 1999: chap. 2)
a. Mary talked to a boy who is crying. (morphology: Past + Pres)
deictic√, simultaneous*
b. Маша встретила мальчика, который плачет. (morphology: Past + Pres)
deictic√, simultaneous*
c. Marie redete mit einem Jungen, der weint.
Past under Past: simultaneous/independent available?
( 5)
a. Mary talked to a boy who was crying. (morphology: Past + Past)
Independent, simultaneous
b. Маша встретила мальчика, который плакал. (morphology: Past + Past)
Independent, simultaneous
Göttingen, 9.11.2009
c. Marie redete mit einem Junge der weinte.
Independent, simultaneous
Present under Future: Simultaneous/independent reading available?
a. Mary will buy a fish that is still alive. (Ogihara, 1989)
b. German
Marie wird einen Fisch kaufen, der noch lebt.
c. Masha kupit rybu, kotoraja zhivet v Bergenskom akvariume
Past\Present Perfect: simultaneous reading available?
English and German: yes
a. John will say that he has bought a fish that was alive.
(at the time of the buying!)
b. Hans wird sagen, dass er einen Fisch gekauft hat, der noch lebte/gelebt hat.
Russian: no (Imperfective Past\Perfective Past)
c. Ivan skazhet, chto on kupil rybu, kotoraya eshcho zhila v Bergenskom akvariume.
deictic tense in the relative
Tense under Attitudes
Present under Past
Göttingen, 9.11.2009
a. #John believed Mary is sick.
only double access (not treated)
b. Hans glaubte dass Marie krank ist.
Simultaneous (or DA)
c. Vanja skazal, chto Mash bol'na.
Vanja said that Masha (is) sick.
German can express simultaneity by any subjunctive form: sei (Sub1), wäre (Subj2),
sein würde (periphrastic Sub2).
Past under Past
a. John believed that Mary was sick.
Simultaneous or shifted
b. Hans glaubte, dass Marie krank war.
c. Vanja skazal, chto Masha byla bol'na.
Here, too, Germ. can express simultaneity by any subjunctive form.
Present\Past: Simultaneous/independent reading available?
Past\Past: Simultaneous/independent reading available?
English is a clear SOT-language: simultaneity under attitudes is expressed by temporal
Russian is a non-SOT-language. Simultaneity under attitudes is expressed by Present.
German can treat complements as in English, i.e., simultaneity can be expressed by
temporal agreement. In addition, Present or any subjunctive form can express
simultaneity. Conclusion: German subjunctive forms are semantically tenseless;
present forms may be tenseless.
Göttingen, 9.11.2009
Our LFs are expressions of an intensional λ-language, which is based on the types e
(individuals), i (times), v (events), t (truth-values), s (world histories). “Intensional” means
that expressions of type a express meanings of type (sa), i.e., a-intensions. For details see
the appendix.
a. John called.
b. Mary is happy.
a. Present, type i: F(N) = λw.s*.
b. Past, type i(it,t): F(P) = λw.λt.λPit.(∃t’)[t’ < t & P(t’)] (Heim, 1997)
P is a relative tense. The argument of P is always N in matrix clauses. In subordinate
clauses, P can have a time variable t as argument that is bound by a higher tense or
locally bound by a λ-operator. Past is indefinite. For the Partee Problem, see the
a. F(Johne) = λw.John
b. F(calledi(et)) = λw.λt.λx.x calls in w at t.
feature: uP
c. F(happyi(et)) = λw.λt.λx.x is happy in w at t.
no tense feature
Like happy, called has a tenseless semantics! The morphology of the latter is checked
by a semantic tense.
The time argument is the first by convention.
At DS the time argument is filled by the semantically empty pronoun PRO, which is
moved for type reasons at LF (PRO-theory of (Heim and Kratzer, 1998), 226-28).1
Derivation of (10a)
DS: [TP [T P N] [VP John [called PRO]]]
PRO-movement (with subsequent PRO deletion)
LF: N λ0 [TP [T P t0 ] PRO λ1 [VP John [called t1]]]
= λw.(∃t < s*) John calls in w at t
The temporal auxiliary be: type i(it,t)
Questions from the audience show that this theory is not yet as popular among semanticists
as it deserves to be. It shows that the λ-operator is deeply rooted in the logical syntax of
natural language.
Göttingen, 9.11.2009
F(is) = λwλt.λPit.P(t)
feature: uN
Analysis of (10b)
DS: [TP N [VP [V is PRO] [AP Mary happy PRO ]]]
PRO movement (with subsequent PRO deletion)
LF: [TP N λ1 [VP [V is t1] λ2 [AP Mary happy t2]]]
= λw. Mary is happy in w at s*
John had called.
has/had: type i(it,t)
features: has uN, iP had
uP, iP
λw.λt.λPit.(∃t’)[t’ < t & P(t’)]
(Extended Now meaning for have-pres ignored.) The perfect temporal auxiliary is a verbal
quantifier, but it has the feature iP and can therefore never transmit the feature uN unlike
[it,t P N] [λ1 [[had t1][ λ2 [John [called t2]]]]]
= λw.(∃t1)[t1 < s* & (∃t2 < t1)[John calls in w at t2]]
The future auxiliary will is the mirror image of have:
will: type i(it,t)
feature uN
λw.λt.λPit.(∃t’)[t’ > t & P(t’)]
The temporal auxiliary would has the same meaning but the feature uP. This doesn’t
capture the fact that temporal would never occurs in matrix sentence. A finer analysis
should analyse the form as a subjunctive (see the treatment of the German subjunctive in
John will call.
N [λ 1 [[will t1][ λ 2 [John [call t2]]]]]
= λw.(∃t’)[t’ > s* & John calls at t’]
(Zeijlstra, 2004) and others: There are two sorts of features, interpretable ones [iF] and
uninterpretable ones [uF]. Interpretable features check uninterpretable features.
In English, finite verb forms have uninterpretable temporal features. Present forms of
a verb have the feature [uN] “uninterpretable Present/Now”. The semantic Present N has the
feature [iN] “interpretable Present/Now”. Past forms of a verb have the feature [uP]
“uninterpretable Past”. The semantic Past tense P has the feature [iP].
Göttingen, 9.11.2009
( 21)
Some verb forms with spell out:
Present: call/calls [uN]
Past: called [uP]
Past Participle: called (no inherent temporal feature)
Infinitive: call (no inherent temporal feature)
Meaning of all these: λw.λt.λx.x calls in w at time t
Feature transmission under semantic binding.
A semantic tense P or N transmits a feature [uP]/[uN] to the time variable it binds. If
the variable is an argument of a tensed verb form, the feature has to agree with the
tense feature of the verb.
(Heim, 1994b), (Heim, 2005).
We assume the conventions for semantic binding out-lined in (Heim and Kratzer, 1998). In
particular, a phrase or operator α may bind a variable via a λ-operator. λ-abstracts are
created by PRO- or WH-movement.
Mary called.
(DS) (not interpretable)
Percolation of tense features
a. Features percolate along the head line.
b. The feature of a temporal variable either agrees with the inherent feature of the
head or it is transmitted to the head (and percolates to the phrase).
Since the semantic Past is the head of the semantic tense [P N], the feature [iP] percolates to
the phrase [P N].
The LF
Göttingen, 9.11.2009
Basic assumption: The highest semantic tense of a relative clause is the pronoun
Tpro, which is obligatorily bound by a higher tense.
This idea is implicit in Kusumoto’s dissertation.2 The present formulation is due to Irene
Heim (p.c). Tpro can be read as “time pronoun”. This is reminiscent of (Partee, 1973) who
proposed that tenses should be pronouns in general (we differentiate between matrix clauses
(tense is an operator), complements (tense is PRO or TPro) and adjuncts (tense is Tpro).
Recall that von Fintel & Heim also show the existence of a Wpro (“world pronoun”); cf.
(von Fintel and Heim, 2000).
English/German relatives
Mary will buy a fish that is (still) alive. (Ogihara, 1989)
N λ 1 will(t1) λ 2 M. buy(t2) a fish WH3 Tpro2 λ 4 is(t4) λ 5 x3 alive(t5)
Cf. (Kusumoto, 1999: chap.2).
1. K. doesn’t use N. She has a distinguished variable t*, which denotes s* in the
matrix, and which is anaphoric in relatives and bound by a predicate of attitude in
2. The temporal arguments of finite verbs are variables that carry a presupposition:
pasti denotes a past time. In addition we have a relative Past.
Göttingen, 9.11.2009
= λw.(∃t > s*)(∃x)[fishw(x) & alivew(x,t) & buyw(Mary,x,t)]
will is a verbal quantifier. It transmits its temporal feature to the variables it binds, i.e.
t2 and Tpro2. More accurately will has two features uN, the inherent and the
transmitted one. Feature transmission is only possible for uN.
The deictic reading of (27) requires binding of Tpro to the matrix N:
N λ 1 will(t1) λ 2 M. buy(t2) a fish WH3 Tpro1 λ 4 is(t4) λ 5 x3 alive(t5)
= λw.(∃t > s*)(∃x)[fishw(x) & alivew(x,s*) & buyw(Mary,x,t)]
The semantics is different, and Tpro1 gets its feature directly from N.
For Past under Past constructions, the literature assumes three readings:
Mary talked to a boy who was crying.
Simultaneous, anterior, independent
The simultaneous reading is obtained by having a Tpro (and no P) in the relative clause,
where Tpro is bound to the matrix Past. The backward shifted reading is obtained by a
semantic Past in the relative, which is bound to the matrix Past, and the independent reading
has an embedded Past bound to N. Past under Past doesn’t give us convincing data that
require bound tense in the relative clause. We could have a deictic Past in all these cases.
But attitudes will give us data that can only be dealt with within the binding approach:
John said that he would buy a fish that was still alive. (Ogihara)
German relatives are treated in almost every case like the English ones. The last sentence
has a different treatment:
Hans sagte er werde (Sub1) einen Fisch kaufen, der noch lebendig sei/wäre/ist/*war.
The subjunctive (or ind. pres.) is obligatory for the shifted reading.
5.2. Russian Relatives
Russian relatives are analysed like the English ones with the exception that Present under
Future has to be deictic.
Analysis of Russian
0. The verbal quantifiers of Russian don’t transmit temporal features. For instance,
budet ‘will’ has the feature uN, but it doesn’t transmit it. (SOT-parameter)
1. Perfective Present expresses semantic Future F:
[[ F ]] = λt.λP.(∃t’)[t’ > t & P(t’)], feature iF
Göttingen, 9.11.2009
2. The analysis of tense under attitudes will require a relative PRESENT, which is
locally bound by PRO and hence never occurs in relative clauses. See below.
3. Tensed participles are interpreted like relatives.
Relative Present in Russian
[[ PRESR ]] = λw.λt.λPit.P(t)
feature: iN
The syntactic argument of PRESR has to be PRO.
Semantically PRESR is identity, i.e. void. The trivial operator merely serves the purpose to
assign the feature uN to the temporal variable it binds. This is the Present analysis for nonSOT languages of (Ogihara, 1996).
Маша купит рыбу, которая живёт в Бергенском аквариуме.
Why is the bound reading (with the state in the relative clause located in the future) not a
possible interpretation of the Russian sentence?
N λ 1 F(t1) λ 2 …buy-fut(t2)…WH Tpro1/*2 λ 3…lives(t3)…
If Tpro is bound by N, it correctly inherits uN and transmits uN to zhivjot/lives.
If Tpro is bound by F, it has the feature uF, which is in conflict with the present
morphology of zhivjot/lives.
Ol’ga budet zamuzhem za vrachom, kotoryj zhivet v Murmanske.
Olga will (be) married with a doctor who lives in Murmansk
N λ 1 budet(t1) λ 2 Ol’ga zamuzhem(t2) za vrachom WH3 Tpro1/*2 λ 4 t3 zhivet(t4)
no T-feature!
If Tpro is bound by N, it correctly inherits the feature uN and we have the deictic
If Tpro is bound by budet, it has no temporal feature (SOT-parameter) and uN of
zhivet/lives cannot be checked. budet has the inherent feature uN, and its variable
has uN, too. But uN is not transmitted. We cannot wrap Tpro into PRESR, which
would check zhivet/lives, because PRESR must have PRO as argument. Another
Göttingen, 9.11.2009
option would be that Tpro comes with a temporal feature from the lexicon.
Though a forward shifted reading for Fut\Fut is preferred, there is the deictic reading as
Через два года она выйдет замуж за человека, который завтра будет у нас в
‘Within two years she will marry a man that will be our guest tomorrow’
5.2.1. Insertion of covert Past
A complication for Russian arises from the fact that we can always insert a relative past
under Past or Future to obtain a semantic Pluperfect or Future Perfect; cf. (Paslawska and
jvon Stechow, 2003), (Grønn, 2003). This insertion does not block or affect feature
transmission from above (“counter cyclic insertion”).
(Иннокентий наперед знал-ipf-past, как утром за завтраком они с восторгом
Göttingen, 9.11.2009
Innokentij in-advance knew, how in-morning during breakfast they with pleasure
сообщат-pf-pres, что им обоим снился-ipf-past один и тот же сон ...
will-tell, that they both dreamt one and the same dream ...)
- Одинаковых снов не бывает-ipf-pres! - скажет-pf-pres им отец,
identical dreams not happen - will-tell them father,
который уже вернется-pf-pres с ночного дежурства
who already will-(have)-returned from night work
и будет завтракать-ipf-infinitive вместе с ними.
and BUDET eat-breakfast together with them
‘Innokentij knew in advance how tomorrow morning during breakfast they would
tell with great pleasure that they both had the same dream … Identical dreams do
not happen, their father, who will eat breakfast together with them after having
returned from night work, will tell them.’
N λ 1 F(t1) λ 5 THE father [WH2 already Tpro5 λ 3 P(t3) λ 4 t2 return-pf-pres(t4)…]
Here Tpro is bound by the matrix Future F and transmits the feature uF from the matrix.
The counter-cyclically inserted P in the relative does not block the transmission.
Why not choose the simpler analysis with a deictic future in the relative?
The father, who will already have returned, will say:”……
N λ1 F(t1) λ3 the father [WHx F(Tpro1) λ2 uzhe x will-return(t2)] will-say(t3):”….
In terms of truth-conditions, this analysis is hard to distinguish from the previous one. The
problem is the uzhe ‘already’ in the relative. It modifies an achievement/accomplishment. If
this were possible, the following sentence should be good, contrary to the facts:
The father will already return.
already modifies states and the covert relative Past converts the accomplishment into a
state and thus makes it a good input for the aspectual particle.
Göttingen, 9.11.2009
5.2.2. A note on tensed participles in Russian
Tensed participles behave like relative clauses and are therefore analysed similarly. Past
participles should behave like relative clauses with a verb in past tense. Hence past
participles have the inherent feature uP. Present participles can be used deictically or
shifted. In the following sentence, a deictic interpretation of the participle is preferred:
Алла познакомится со студентом, живущим на втором этаже.
lit. Alla will meet with students living-pres on the second floor
But there are cases with a shifted interpretation:
Svjashchennik blagoslovil umirajushchego krest’janina.
(U. Junghanns, p.c.)
The priest blessed the dying-pres peasant’
We describe the facts correctly if we assume that present participles have no inherent tense
feature, i.e., they behave like adjectives. This makes it possible that the Tpro in (42) is
bound by N and that the Tpro in (43) is bound by P. Apart from this difference, tensed
participles are analysed like relative clause.
In order to avoid temporal paradoxes analysed in (von Stechow, 1981), (von Stechow,
1995), (Heim, 1994a) among others, attitudes cannot be simply quantifiers over worlds.
They quantify over worlds and times (and individuals, neglected here). Hence complements
must be properties of time, type i(st). Therefore attitudes are verbal quantifiers.
The properties are generated by assuming PRO (or TPRO) as the highest semantic tense
(“zero-tense”). PRO is moved and creates a temporal abstract; cf. (Kratzer, 1998).
a. At 5 o’ clock Mary thought it was 6 o’clock. (sequence of tense, "simultaneity")
b. В пять часов Маша думала, что (уже) шесть часов. (non-sequence of tense,
relative present)3
6.1. English attitudes
Tense in clausal complements
Past + Past morphology as in English is possible also in Russian (see 6.2. below).
В пять часов Маша думала, что было (уже) шесть часов.
Göttingen, 9.11.2009
1. Attitude verbs are (verbal) temporal quantifiers that transmit their temporal
features. SOT-parameter.
2. The highest semantic tense of a complement is the temporal PRO. PRO has to be
moved for type reasons and thus creates a temporal abstract. This follows from
the logical type of the attitude verb; cf. (Kratzer, 1998).
Quantifier over worlds and times
(style of (Lewis, 1979))
believe, type (s(it))(i,et)
F(believe) = λw.λPs(it.)λt.λy.(∀w’)(∀t’)[(w’,t’) is compatible with everything y
believes of (w,t) in w at time t → p(w’)(t’)]
P N λ 1 t1 at 5 o’ clock Mary thought(t1) PRO λ 4 t4 λ 2 was(t2) λ 3 t3 at 6 o’clock
= λw.(∃t1 < s*)[t1 = 5 o’clock & ((∀w’,t’) ∈ DoxMary(w,t1)) t’ = 6 o’clock]
The temporal variable of subordinate was inherits its feature from the matrix P via
feature transmission through thought.
Mary thought Bill left.
P N λ 1 Mary thought(t1) PRO λ 2 P(t2) λ 3 Bill left(t3)
= λw.(∃t1 < s*) Mary thinks in w at t1 [λw’.λt2.(∃t3 < t2) Bill leaves in w’ at t3]
The complement contains its own relative Past that checks the uP of left. Hence the
shifted interpretation.
Ogihara’s evidence that Past in relatives must sometimes be bound:
John thought that he would buy a fish that was still alive.
P N λ 1 John thought(t1) PRO λ 2 would(t2) λ 3 he buy(t3) a fish WH4 Tpro3
λ 5 was(t5) λ 6 t4 alive(t6)
The variable of was in the relative inherits its feature from the matrix P via binding.
Note that the binding chain goes through the infinitive buy, which has no inherent
temporal feature.
Göttingen, 9.11.2009
6.2. Russian Attitudes
Analysis of Russian
0. The Russian verbal quantifiers budet and verbs of attitudes do not transmit
temporal features. (SOT-parameter)
1. Russian has a relative Present, which argument is by stipulation PRO.
[[ PRESR ]] = λt.λPit.P(t), feature iN; Cf. (Ogihara, 1995) with modification
Present\Past (simultaneous)
Ваня сказал, что Маша в опере.
lit. John said that Mary is in the opera.
N λ 1 P(t1) λ 2 Ivan said(t2) PRO λ 3 PRESR(t3) λ 4 is(t4) λ 5 Mary in(t5) opera
λw.(∃t1 < s*) Ivan says in w at t1 [λw’.λt2.Mary be in the opera in w’ at t2]
Ваня сказал, что Маша была в опере.
lit. Ivan said that Mary was in the opera.
N λ 1 P(t1) λ 2 Ivan said(t2) PRO λ 3 P(t3) λ 4 was(t4) λ 5 Mary in(t5) opera
λw.(∃t1 < s*) Ivan says in w at t1 [λw’.λt2.(∃t3 < t2) Mary be in the opera in w’ at t3]
Impact of the SOT-parameter: If verbs of attitude did transmit their features to the
complement, we could have the simultaneous reading as in English.
lit. Ivan said that Mary was in the opera.
N λ 1 P(t1) λ 2 Ivan said(t2) PRO λ 3 Mary was(t3) in opera
Violation of SOT-parameter
λw.(∃t1 < s*) Ivan says in w at t1 [λw’.λt2.Mary be in the opera in w’ at t2]
6.3. German attitudes: the role of subjunctive
German can analyse attitude complements like English. In addition, simultaneity under
Present or Past can be expressed either by Present or by any subjunctive form.
Hans glaubt, dass Marie krank ist/sei/wäre/?sein würde. (sim)
Göttingen, 9.11.2009
Hans glaubte, dass Marie krank war/ist/sei/wäre/?sein würde (sim)
This means that all these forms are semantically tenseless. (The Present ist can express
semantic Present and is then analysed similar to English is.)
Apart from certain constructions, subjunctive forms are not licensed in matrix
sentences. Therefore they have to be licensed. Accordingly, German attitude predicates
must have the feature [iSub], which leaves open whether the embedded form is uSub1 or
uSub2. Since indicative forms are possible under attitudes, German indicative forms are
neutralised with respect to mood. Here are some lexemes with features:
( 55)
a. glaubt
iSub, uN
b. glaubte
iSub, uP
c. ist
(no mood, no tense)
d. war
(no mood) uP
e. sei
uSub1 (no tense)
f. wäre
uSub2 (no tense)
g. würde
uSub2 (no tense)
Hans glaubte dass Marie krank ist/sei/
wäre/ sein würde
uSub1 uSub2
Hans sagte, er werde einen Fisch kaufen der noch lebe/ lebt/leben würde/?lebte
uP,iSub uSubj1
uSub1 √
uSub2 uSub2
If subjunctive forms occurred only under attitudes, it would not be necessary to distinguish
between subjunctive 1 and subjunctive 2. We need the distinction however to describe the
syntax of subjunctive conditionals correctly. They require the subjunctive 2 in the if-clause.
6.4. Shifting under Attitudes and Aspect
(Altshuler, 2009) (referring to work of (Khomitsevich, 2007)) objects against the traditional
view that Past\Past means backward shifting in non-SOT languages like Russian (Altschuler’s
2009 SuB talk). He says that embedded imperfectives in the past can have a simultaneous or a
backwards shifted reading. Embedded perfectives are always anterior.
To simplify the exposition, we haven’t considered aspect so far. The fact that we have
‘simultaneity’ with Past\Past if the embedded verb is imperfective is easily explained by
considering the meaning of the imperfective.
( 58)
Göttingen, 9.11.2009
a. Imperfective
[[ IP ]] = λw.λEvt.λt.(∃e)[E(e) & t ⊆ τ(e)]
(vt, it)
(vt, it)
b. Perfective
[[ PF ]] = λw.λEvt.λt.(∃e)[E(e) & τ(e) ⊆ t]
(In order to solve Dowty’s imperfect paradox, the IP should be modalized. This is ignored
here.) A more complete analysis of (51) would therefore be:
( 59) Ivan said that Masha was in the opera (Ваня сказал, что Маша была в опере.)
P N λ 1 I. said(t1) PRO2 P(t2) λ 4 IP(t4) λ 5 was(t5) λ 6 M. in_op(t6)
λw.(∃t < s*) I. says in w at t λt’.λw’.(∃t’’ < t’)(∃s)[t’’ ⊆ τ(s) & M. is in the op. in s
in w]
We have backwards shifting in the embedded clause. But since the VP expresses a state and
the event time is in the time of the state, the state might hold on at the “subjective now” t’.
This is a point stressed emphatically in (Klein, 1994).
If on the other hand the embedded VP is in the perfective, we have the anterior
reading. (According to Khomitsevich this is what her informants observe indeed.) For
Ivan skazal, chto Masha vyshla
Ivan said Masha left
P N λ 1 I. said(t1) PRO2 P(t2) λ 4 PF(t4) λ 6 M. left(t6)
λw.(∃t < s*) I. says in w at t λt’.λw’.(∃t’’ < t’)(∃e)[ τ(e) ⊆ t’’ & M. leaves in e in w]
An achievement is “quantized”, i.e. it doesn’t apply to subintervals. If its time is before the
subjective now t’, it is over at t’ and we have anteriority.
6.5. Tense under Perception Verbs
Khomitsevich observes that Past\Past frequently expresses simultaneity in perception
constructions even for achievements/accomplishments. She correctly claims that this is an
effect of the semantics: the fact/event perceived must be simultaneous with the perception.
Verbs of perception can be analysed in the style of (Higginbotham, 1983). It follows that
they are not verbal quantifiers and hence not subject to the SOT-parameter.
Dina videla, chto/kak voda lilas’/l’jotsja iz vedra. (Altshuler, 2004)
Dina saw that/how water poured/pours from basket.
‘Dina saw that/how water was pouring from the basket’
According to Khomitsevich, perception verbs express simultaneity either by embedding a
Göttingen, 9.11.2009
Past or a Present. If these verbs were analysed like attitudes, Present would be expected.
This should be a sort of evidential reading. If Past and Present were freely interchangeable
in these constructions without a change in meaning, then that would be a problem.
(Higginbotham, 1983) analyses perceptions as relation between an individual and an
event, the eventive reading. The complement describes the event and may, of course,
localise it in time. The truth-condition of (61) is something like this:
There was a water pouring e & Dina saw e.
It is obvious that ‘see’ is no verbal quantifier. The analysis of the construction has an
interesting consequence for the semantics of aspect. Here is the result we want (in our
Russian example the perception verb itself is also in the imperfective aspect):
λw(∃e)(∃t < s*)[t ⊆ τ(e) & water pouringw(e)
& (∃t’ < s*)(∃e’)(t’ ⊆ τ(e’)) seew(Dina,e’e)]
The first thing to notice is that we have two independent deictic Pasts, so no temporal
agreement is needed. The other feature to notice is that the existential quantifier on top
binds the aspect variable e that occurs in the main and in the subordinate clause. This means
that the event variable in the aspect is not always locally bound by an existential quantifier.
Here is the compositional build up of the LF:
kak voda lilas’ iz vedra ‘how the water was pouring from the basket’
kak ‘how’ is a semantically empty WH-operator that has to be moved on logical grounds
and thus creates a λ-abstract. The operator originates in the event argument of the
imperfective aspect IP. So we have to modify the usual semantics:
Göttingen, 9.11.2009
( 65)
a. Imperfective
usually: [[ IP ]] = λw.λEvt.λt.(∃e)[E(e) & t ⊆ τ(e)]
(vt, it)
here: [[ IP ]] = λw.λe.λt.λE.E(e) & t ⊆ τ(e)
usually: [[ PF ]] = λw.λEvt.λt.(∃e)[E(e) & τ(e) ⊆ t]
(vt, it)
here: [[ PF ]] = λw.λe.λt.λE.E(e) & τ(e) ⊆ t
b. Perfective
We can regain the usual existential analysis for aspect by existential closure of the variable
In order to be able to have the how-clause as an object, we prefix it by a covert
existential quantifier ∃ of type (vt,(vt,t)). This makes the complement clause a generalized
quantifier that we can QR from the object position.
Dina videla, kak voda lilas’.
[CP ∃ kak3 P(N) λ 2 IP(e3)(t2) λ 1 voda lilas’(e1)]
λ 3 [S P(N) λ 5 (∃e6) IP(e6)(t5) λe7 Dina videla(e3)(e7)]
F(videla) = λw.λe1.λe2.λx.e1 is a seeing of e2 by x in w
features: uP, uIP
A note to feature transmission: temporal features are transmitted across the aspect operator
to the variable that the latter binds.
This analysis explains the fact that Past\Past may express simultaneity in perception
Here are examples by Partee (“belief based on perception”).
A complication arises from the fact that, according to Khomitsevich, Present and Past
are freely interchangeable in these constructions and occur even in conjunctions:
Göttingen, 9.11.2009
Ona gljadela, kak sizye kol’ca ot sigary Azarello uplyvali v kamin i kak kot lovit ikh
na konec shpagi.
‘She watched as Azarello blew (past!) smoke-rings at the fireplace and the cat spiked
(present!) them on the end of his sword.’ (Barentsen, 1996, p.21, (18), Master i
This conjunction presumably expresses an event plurality. But then the present in the second
conjunct cannot be explained by the analysis given so far.
1. The temporal features of finite verbs are licensed by semantic binding by a semantic
2. Tense agreement has several analyses:
a) Non-local agreement is mediated via Tpro, which is obligatorily bound by a higher
tense (relative clauses). (Related to Principle B)
b) Local agreement is licensed by a verbal quantifier, e.g. a verb of attitude. It is
mediated by (temporal) PRO.
3. Variation in tense distribution between languages has the following sources:
a) The SOT-parameter
b) Difference in interpretation of semantic tenses (e.g. relative PRES in Japanese and
4. Not all verbs with complements are verbal quantifiers, e.g. verbs of perception are
not. Hence these are not subject to the SOT-parameter and tense in those complements has an
independent interpretation.
We introduce an intensional λ-language. “Intensional” means that expressions of type a
express meanings of type (sa), i.e., a-intensions.
Recursive definition of the interpretation function [[ . ]] F,g
Göttingen, 9.11.2009
1. Let α be a lexical entry of type a . Then [[ α]] M,g = F(α).
2. Let x be a variable of type a. Then [[ x]] M,g = λw.g(x), g(x) in Da.
3. FA: Let α have type b and daughters β of type ab and γ of type a.
[[ α ]] M,g = λw.[[ β]] M,g (w)([[ γ]] M,g (w))
4. IFA: Let α have type b and daughters β of type (sa)b and γ of type a.
[[ α]] M,g = λw. [[ β]] M,g (w)([[ γ]] M,g)
5. PM: Let α have type a and daughters β and γ of the same type.
[[ α]] M,g = λw.λx.[[ β]] M,g (w)(x) & [[ γ]] M,g (w)(x)
6. Abstraction: Let x be a variable of type a and let α be an expression of type b.
[[ λx α]] M,g = λw.λu ∈ Da.[[ α]] M,g[x/u](w)
I didn’t turn off the stove. (Partee, 1973)
= I didn’t turn off the stove at a particular time in the past
Tomorrow, John will have left.
= There is a time after s* in tomorrow which is after John’s leaving time in
Our indefinite Past is an existential quantifier and comes with a domain restriction. So the
refined lexical entry is this:
Contextually restricted Past: type (it, i(it,t))
(Musan, 2002)
[[ P ]] = λC. λt.λQ.(∃t’)[C(t’) & t’ < t & Q(t’)]
The analysis of the Partee sentence is then:
The Partee Problem
not [PC N] λt. I [turn-off(t) the stove]
If g(C) = {t | t ⊆ [11 a.m., 12 a.m.]}, the LF means:
¬(∃t < s*) t ⊆ [11 a.m., 12 a.m.] & I turn off the stove at t
John will have left at six.
We obtain the correct result if we assume the following pragmatic principle:
An embedded tense or temporal auxiliary adds the “content” of the next super
ordinate tense to its restriction provided the super ordinate tense is compatible
with the tense in question.
Göttingen, 9.11.2009
N λt WILLC1(t) λt haveC2(t) λt john left(t)
g(C1) = {t | t a time}
g(C2) = {t | t > s*}
= (∃t)[t > s* & (∃t’)[t’ > s* & t’ < t & John leaves at t’]]
The narrative progression observed for Past+Perfective should also be accounted via the
temporal restriction. But we haven’t elaborated this yet.
Altshuler, Daniel. 2004. A simultaneous perception of things: SOT in Russian. Snippets 8:5-6.
Altshuler, Daniel. 2009. Flashback discourses and the meaning of the Russian (im)perfective.
In Sinn und Bedeutung 14. University of Vienna, Austria.
Grønn, Atle. 2003. The Semantics and Pragmatics of the Russian Factual Imperfective,
Faculty of Arts, University of Oslo: Doctor Artium Thesis.
Heim, Irene. 1994a. Comments on Abusch's theory of tense: Manuscript, MIT.
Heim, Irene. 1994b. Puzzling reflexive pronouns in de se reports: Handout from Bielefeld
Heim, Irene. 1997. Tense in compositional semantics: MIT lecture notes.
Heim, Irene. 2005. Features on bound pronouns. Ms. Cambridge/Mass.
Heim, Irene, and Kratzer, Angelika. 1998. Semantics in Generative Grammar: Blackwell
Textbooks in Linguistics. Oxford/Malden, MA: Blackwell.
Higginbotham, James. 1983. The Logic of Perceptual Reports: An Extensional Alternative to
Situation Semantics. Journal of Philosophy 80:100-127.
Khomitsevich, Olga. 2007. Dependencies Across Phases. From Sequence of Tense to
Restrictions on Movement: LOT. Utrecht.
Klein, Wolfgang. 1994. Time in Language. London, New York: Routledge.
Kratzer, Angelika. 1998. More Structural Analogies Between Pronouns and Tenses. In SALT
VIII, eds. D. Strolovitch and A. Lawson. Cambridge, Mass.: Ithaca: CLC-Publications.
Kusumoto, Kiyomi. 1999. Tense in embedded contexts, Department of Linguistics,
University of Massachusetts at Amherst: Ph.D. dissertation.
Lewis, David. 1979. Attitudes De Dicto and De Se. The Philosophical Review 88:513-543.
Musan, Renate. 2002. The German Perfect. Its Semantic Composition and its Interactions
with Temporal Adverbials: Studies in Linguistics and Philosophy.
Dordrecht/Boston/London: Kluwer Academic Publishers.
Ogihara, T. 1995. The Semantics of Tense in Embedded Clauses. Linguistic Inquiry 26:663679.
Ogihara, T. 1996. Tense, Attitudes, and Scope. Dordrecht: Kluwer.
Ogihara, Toshiyuki. 1989. Temporal Reference in English and Japanese: University of Texas
at Austin.
Partee, Barbara. 1973. Some Structural Analogies between Tenses and Pronouns in English.
Journal of Philosophy 70:601-609.
Paslawska, Alla, and von Stechow, Arnim. 2003. Perfect Readings in Russian. In Perfect
Explorations, ed. Rathert Alexiadou, von Stechow. Berlin: Mouton de Guyter.
von Fintel, Kai, and Heim, Irene. 2000. Intensionality. Ms.
von Stechow, Arnim. 1981. Indexicals in De Se Contexts.
von Stechow, Arnim. 1995. On the Proper Treatment of Tense. In SALT V, eds. Teresa
Göttingen, 9.11.2009
Galloway and Mandy Simons, 25: Cornell University.
Zeijlstra, Hedde. 2004. Sentential Negation and Negative Concord. Utrecht: LOT.