Sunteți pe pagina 1din 74

Aspects of the grammar and

the processing of agreement


in English and Spanish

I. Agreement and beyond


II. Knowledge of language via access to the mind:
comprehension of language
III. Formal and conceptual sources of agreement
IV. Cross-linguistic differences in agreement: attrition vs
redundancy
V. Aspects of the repercussions of morphological attrition
VI. Summary and conclusions

1
Juan Carlos Acuña Fariña

The psycholinguistics of agreement in


English and Spanish: a tutorial overview

Lingua 119: 389-424


2009.
Elsevier: Amsterdam
2
I. Agreement and beyond
Aim: 1. to discuss aspects of the grammar and
the processing of agreement with special
emphasis on how this works in English and in
Spanish.
2. To examine how precisely agreement
interacts with the rest of the grammar.
Systematic covariance between a semantic or formal
property of one element and a formal property of
another. (Steele 1978: 610).

Terminology: controller-target (also probe and goal, or


trigger and source). They all suggest asymmetry.
3
No limits: According to Ferguson (1964), Bengali´s
pronouns have forms for:

– 12 different categories of person, number and respect.


– verbs have syncretized tense/mode/aspect and person
suffixes.
– for each tense/mode/aspect set of forms, there is
independent marking of
– 5 person/number/respect classes.
– For instance, -i is the first person ending in the present but
the second person inferior in the future, while –o is the
second person ordinary in the present and first person in the
future.

4
German: three gender classes (masculine, feminine and
neuter) and a complex set of interacting morphological,
phonological, and semantic principles (Köpcke and Zubin
1983, 1984).

-nouns commencing in kn- are masculine.


-those ending in a fricative plus ‘t’ tend to be feminine.

As Corbett (2006: 206) points out, when it


comes to agreement options, “there is such
variety that we might wonder if we are
faced with anarchy”.

5
-“a clear case of the victory of the
indexical aspect of language over its
iconic aspect”. Agreement “is not
only non-iconic but meaningless”.
(Haiman 1985: 162).
- Jespersen (1922: 352 ff.):
“superfluous” and “cumbersome”,
-Taylor (2002: 332 ff.):
“dysfunctional”.

6
Mallison & Blake (1981):

over 70% of the world´s


languages (including English).

7
(1) [We will keep] the little white
candlesticks and the comfortable red
chairs.

(2) [Nos quedamos con] lo-s pequeño-s


candelabro-s blanco-s y la-s cómoda-s
silla-s roja-s.

8
WHY BOTHER?

Why make computational room for the


Spanish redundancy in (2) when the
minimal coding of English in (1) will
suffice?

9
1.2. Agreement in the grammar
and in the mind

Psychological adequacy (Dik 1978,


1991; Kaplan & Bresnan 1982, Van Valin
& Lapolla, 1997; Franck et al. 2005;
Escribano 2007).

10
Chomsky´s (1998, 1999, 2001) notions of
feature interpretability, legibility of the core,
Phase, Phase Impenetrability and Locality.
Chomsky (2001: 13) insists that probe-goal
relations must be local “in order to minimise
search”. In order to reduce the “computational
burden” (1999: 9), he proposes that the
derivation of expressions proceeds by phase,
and that “phases must be as small as
possible, to minimise memory” (2001: 14).
The Phase Impenetrability Condition.

11
This serial thinking, and the obvious
derivational cyclicity on which it rests, is
precisely the kind of theoretical claim that can be
addressed in a laboratory. All one needs to do is
to seek to falsify it by showing evidence of
semantic penetration inside putative phase
domains.

Thornton and MacDonald (2003):

(3) The album by the classical composers … BE praised


(4) The album by the classical composers … BE played.

12
Feature mismatch (Corbett 2006: 143 ff.): the
committee have postposed any decision //
*these committee are satisfied. This calls for a
view of semantic agreement, which is no more
than a way of sanctioning mismatch (some
believe that agreement is always semantic:
Barlow 1999, Pollard and Sag 1988).
In general, mismatches echo the penetrability
or impenetrability of hypothesized
components in the grammar and the
processing of agreement. This will be an
important issue here.
13
1.3. Domains and kinds: distance
and penetrability
The Agreement Hierarchy (Corbett 1979)
Attributive > predicate > relative pronoun > personal pronoun

*These committee vs this committee are satisfied with their work

BUT:
Frequent intervention of formal processes across
large distances and of notional factors in the short
distances:
(5) A lot of people were there (*is there)
(6) The police haven´t yet made up their
14
mind/minds
III. Formal and conceptual sources of
agreement features

Agreement lies at the heart of sentence structure in that it


usually codifies the formal link between the subject and the
predicate: what signals that a clause has been created. A
core process.

15
A few relevant questions
Is agreement an essentially Do we compute agreement
formal (Bock & Eberhard 1993; differently for gender and
Eberhard 1997; Levelt, Roeloffs, number? For semantic
& Meyer 1999; Chomsky 1995, gender and morphosyntactic
1999, 2001; Carminatti 2005; gender?
Franck et al. 2005) or an Is the processing of agreement
essentially conceptual the same across domains?
phenomenon (Viglioco et al.
1996; Thornton & MacDonald
2003; Haskell & MacDonald
2003; Barlow 1999; Pollard &
Sag 1988)?
Are the grammar of agreement
and the processing of
agreement ‘in good agreement’
with each other?
Are there cross-linguistic
differences in the way
agreement is processed?
16
3.2. Sources of information for
agreement

3.2.1. Semantic overrides

(7) The police is/are enquiring into that.


(8) A number of ideas were/*was proposed.
(9) Heaps of time is/*are being wasted.

17
Huddleston & Pullum (2002: 504 ff.):
(10) Twenty dollars seems a ridiculous amount to pay
to go to the movies.
(11) Three eggs is plenty.
(12) One percent of students of students take/*takes
drugs.
(13) He withdrew his motion for what were obviously
very sound reasons.
(14) What is needed are managers with new ideas and
the will to apply them.
(15) What are going to be the deciding factors?
(16) Eggs and bacon is/*are my favourite breakfast
(17) The hammer and sickle was/*were flying over the
Kremlin

18
In processing:
So-called conceptual anaphors
Gernsbacher (1991) used three types of
antecedents:
-A basketball team.
-A book.
-A plate

(18) After college, my sister went to work


for IBM. They/it made her a good offer.
19
3.2.2. Formal overrides
Cacciari et al. (1997) on epicenes:

(19) La vittima del incidente stradale sbatté


violentamente la testa contro il finestrino. Lei (lui), perciò,
perse molto sangue e svenne.
The victim of the car accident violently slammed the
head against the window. She (he), therefore, lost a lot
of blood and fainted’.
(20) L´erede decise di andare in vacanza con I soldi
ricevuti dalla zia. Lei (lui), perciò, progettò unlongo
viaggio negli USA.
‘The heir decided to go on vacation with the money
received from the aunt. She (he), therefore, planned a
long trip to the States’.
20
L´erede …………. Lei (Lui) baseline

La vittima …......…Lui non-match

La vittima ……. …Lei match

(Cacciari et al. 1997)


21
RESULT :

Formal cues ruled the parse for epicenes


(irrespective of context facilitation).
Semantics dominated the parse of bi-
gender words.

A very opportunistic parser.

When they exist, formal morphosyntactic


cues take precedence.
22
Two-stage models : Garrod & Sanford 1990, Garrod &
Terras 2000, Rigalleau et al. 2004). The Bonding and
Resolution model of Garrod & Terras (2000). Serial
automatism.

Interactive models (Bates & MacWhinney, 1989;


Taraban & McCleland, 1988; Macdonald et al., 1994;
Tannenhaus, Spivey-Knowlton & Hanna, 2000) claim
that agreement is resolved by simultaneously weighing
the force of a number of possible sources of information
(constraints), notably the information provided by the
lexicon (including lexical meaning) and the context, but
also that provided by grammatical cues. All forces are
supposed to intervene in parallel (hence in one stage),
and the ultimate parse is the result of the competition
among the different attractors. For instance, Marslen-
Wilson, Tyler & Coster (1993).

23
3.3. The Feature Hierarchy
Hypothesis

24
3.3.1. The grammar of features

Feature Hierarchy: Person > Number > Gender


(Greenberg 1966)

Raw cognitive salience: women, fire and


dangerous things.

Is this reflected in theories?:

25
Cognitive Grammar: all agreement is
symbolic and predicational. Period.

Minimalism: erase uninterpretable features


(Chomsky 1999: 4). Although gender may or
may not be semantically transparent, it is
supposed to enter the derivation as a valued
feature any way, since, even if it is not
transparent to the semantic component, its value
is determined –arbitrarily- by lexical
specification. So gender and number are the
same.

26
HOWEVER:

Ritter (1988, 1991, 1993); Harley & Ritter (2002);


Carminati (2005): NUMBER should head its own
functional projection:
Radford (2004: 179):

27
A contentious claim.
Psycholinguistic findings:

Nicol (1988), Di Domenico and De Vincenzi (1995) and


by De Vincenzi (1999):

(21) a. The ballerina told the skier that the doctor would blame him
for the injury.
b. The ballerina told the skier that the doctor would blame her for the
injury.

(22) a. The landlord told the janitors that the fireman with the gas-
mask would protect him from getting hurt.
b. The landlord told the janitors that the fireman with the gas-mask
would protect them from getting hurt.

28
RESULTS: number processed
faster.

BUT: -priming.
-1000 ms. Too slow.
-Lukatella et al. (1987) found the
very opposite.

CONCLUSION:
Very weak evidence for the FHH (contra
Carminati 2005).
29
ERP (Event Related Potentials):

LAN (left anterior negativity)


N400
P600

30
31
3µ V

Agreement – Number ; 340 ms Agreement – Number ; 600 ms Agreement – Number ; 740ms 2.25

1.5

0.75

-
0.75

-1.5

2.25
Agreement – Gender ; 340 ms Agreement – Gender ; 600 ms Agreement – Gender ; 740ms
-
3µ V

32
Münte & Heinze (1994), German, English and Finnish: article + noun
and S-V agreement: LANs that seem to reflect automatic (syntactic)
processes (Neville et al. 1991; Osterhout & Holcomb 1992; Osterhout
et al. 1994; Hagoort et al. 1993; Friederici et al. 1996).

Osterhout & Mobley (1995):

(23) *The elected officials hopes to succeed.


(24) *The hungry guests helped himself to the food.
(25) *The successful woman congratulated himself on
the promotion.

RESULTS:

P600 for the three types of violation. Additional LAN in


S-V.
Gender and number: no differences.
33
Objection: gender is semantically transparent in English
pronouns.

However, Hagoort & Brown (1999); Gunter, Friederici,


& Schriefers (2000); Barber, Salillas, & Carreiras
(2004); and Deutsh & Bentin (2001): the same with
grammatical gender.

SO:
Early syntactic feature checking is the same for
both gender and number, and,
At least for English S-V agreement, agreement is a
fast, shallow, automatic process.

34
Barber & Carreiras (2005) using Spanish
materials:
(26) (27)
Agreement Agreement
ARROYO HELADO EL LAGO

Gender violation Gender violation


ARROYO HELADA LA LAGO

Number violation Number violation


ARROYO HELADOS LOS LAGO

Double violation
ARROYO HELADAS

35
RESULTS:

Rapid use of morphological cues (P600 + LAN).

No differences between gender and number in the


central part of both the LAN and the P600 effects.

Significant differences in the later part of the 700-900 ms


time window, with larger amplitudes for gender than for
number.

INTERPRETATION:

Bradley & Foster´s (1987) three-stage model of


lexical access.

36
CONCLUSION:

Computing vs storing.

No psychological evidence for the


FHH: both features are processed in
the same way, at least during the first
cycles of processing.

37
IV. Cross-linguistic differences in
agreement: attrition vs redundancy

4.1. Agreement and English attrition

Osterhout & Mobley (1995): only a LAN-type effect


was obtained for S-V agreement.

English agreement is more susceptible to semantic


interference, and that manifests itself both at the
level of grammatical sanction and at certain levels of
processing at least.

Grammar and processing out of sync.

38
Berg (1998)

Number conflicts in English and German

1 . e.g. the couple (formally singular, but semantically plural).


2. e.g. pancakes (formally plural, but semantically singular).
3. ART (+ADJ) + N (sg) + of (+ADJ) + N (pl): e.g. a gang of thugs.
4. e.g. both he and she (plural not overtly expressed, must be
inferred).
5. e.g. more than one X and fewer than two X (comparatives).
6. e.g. many an X.
7. e.g. the cause (of the accident) + copula + (to be) + bad brakes.
8. e.g. it + (to be) + the politicians who …; what interests me + (to
be) + N (pl).

39
40
SO:
semanticity in English

vs
syntacticity in German

41
BUT:

what happens with categories 7 and 8?:

Another example … BE the conflicts …

42
English goes syntactic now

????????

43
Production and spill over
effects: on domain
penetrability
Strength of home domains
Frequency and muscles
Two interesting consequences
of English attrition:

44
1. It makes formal linkage more susceptible
to semantic interference; and
2. It freezes an SVO order that guarantees
that the who-did-what-to-whom message-
level mapping to phrasal packages is not
compromised despite the loss of
morphological marks.

Left orientation
45
“When there is a conflict between the
preverbal and the postverbal material, the
syntactic principle usually wins out. When
the conflict arises within the preverbal
domain, the semantic principle is
preponderant. This is because the syntax
directs the speaker´s attention only to the
preverbal NP as such, not to its internal
structure”. (Berg 1998: 63).

46
3 consequences should follow:
 In comprehension, as opposed to production, syntactic
penetration of form into meaning should occur in
languages, like German, Spanish or Italian, with a
rich inflectional system.

 One should observe the same pattern of syntacticity


displayed in German in a language like Spanish.

 If English is effectively left-oriented via a marked


dependence on SVO, one should be able to see other
repercussions of those orientational reflexes over and
above the behaviour with agreement.
47
First prediction confirmed:
Epicenes (Cacciari et al. 1997, submitted) + the
ERP study of Barber and Carreiras (2005) with
all kinds of agreement violations in Spanish
(LAN + P600).
In all cases morphological marks were
processed as automatic reflexes, suggesting
that in Romance languages morphosyntactic
cues are priviledged cues for phrase
construction in the way suggested by Hawkins
(1994, 2004).
48
Further evidence:
(28) Ana fue la única de la empresa que consiguió ser
nominada para los premios.

A. Se lo oímos+ PRO decir+ al mediodía+ a su


director+ absolutamente+ pletórico+ de contento.

B. Se lo oímos+ PRO decir+ al mediodía+ a su


director+ absolutamente+ pletórica+ de contenta.

C. pro Hizo+ PRO regresar+ al mediodía+ a su


director+ absolutamente+ pletórico+ de contento.

D Se lo pro quiso+ PRO decir+ al mediodía+ a su


director+ absolutamente+ pletórica+ de contenta.

49
Shallow processing indicating automaticity
of initial bonding stage.

Even though meaning can sometimes


easily salvage misanalyses, the
autonomy of form clues prevails.

50
Second prediction:

Also confirmed by another completion


study in English and Spanish
Data: 21 structures in 18 paragraphs, 66
gaps, 21/66 target-to-filler ratio. 26 British
English speakers and 34 Spanish
speakers, with ages raging from 18 to 63.
(see pp 96-7 in text).

51
ENGLISH SPANISH
N % N %
sg.: pl. sg.: pl. sg.: pl. sg.: pl.

Category 1
a. Committee 10 11 38.4 – 42.3 32 2 94.1 – 5.8
b. Enemy 19 6 73 - 23 34 0 100 - 0
c. Crew 11 5 42.3 – 19.2 34 0 100 – 0

b. A small number of PL 2 22 7.6 – 84.6 22 12 64.7 – 35.2


c. A number of PL 1 22 3.8 – 84.6 16 18 47 - 52.9

a. Cause of accident BE brakes 23 2 88.4 – 7.6 8 25 23.5 – 73.5


b. A case in point BE the revolts 19 2 73 - 7.6 1 33 3.8 - 97

52
So: English is more
susceptible to semantic
penetration than
Spanish, and more left-
oriented.

53
4.3. The processing of English pronouns
(i.e. the right edge of the AH)

Now: on-line comprehension (not production) of


distant pronouns (not of close verbs)

????????

Osterhout & Mobley (1995) found a LAN-type effect for


S-V violations in English in comprehension, but only
P600 for pronouns.

HYPOTHESES
1. Absence of the bonding phase in English.
2. More geometry used.
54
Van Gompell & Liversedge (2003):
backwards anaphora
(29)
a. gender match
When he was at the party, the boy cruelly
teased the girl during the party games.
b. gender mismatch
When he was at the party, the girl cruelly
teased the boy during the party games.
c. control
When I was at the party, the boy cruelly teased
the girl during the party games.
55
The Gender Mismatch Effect

Similar in spirit to the active behaviour of


the processor in gap-filling routines in
long-distance filler-gap dependencies
(Stowe 1986):
Forces of a structural kind (like c-
command or Chomsky´s 1981 Principle C)
might be at play in constraining co-
indexing operations.

56
Unanchored
pronoun XXXXXXXXXXXXXX Predicted co-indexation site

57
Buttress the view defended here
that:

1. Morphology does not rule the initial


stages of the processing of pronoun-
antecedent ties in English (the right
edge of the AG).

2. Geometrically-based co-indexing is
strong at least in English.

58
SUM-1:

There is no solid evidence for the


psychological validity of feature differentiation
in either English or Spanish (Greenberg 1966:
Corbett 2006).

No support for the FH (Ritter 1991, 1993; Nicol


1988; Di Domenico and De Vincenzi 1995; De
Vincenzi 1999; Harley & Ritter 2002; Carminati
2005).

59
SUM-1:

Storing vs computing.
Bradley & Foster´s (1987) three-stage
model of lexical retrieval.

60
SUM-1:
At the grammatical level, even the S-V
domain (not as far to the right as
pronouns) may exhibit semantic
interference. However, at the processing
level, S-V ties are processed
automatically, that is, as a syntactic
process, even in English (phase
impenetrability of a core clause
construction process?).
61
SUM-1:

Therefore, the grammar of agreement and


the processing of agreement are not in
perfect agreement with each other.

62
SUM-1:

At the processing level, we need to go to


the most rightward edge of the Agreement
Hierarchy to see pronouns not running on
formal rails in English.

No bonding phase in English (Garrod &


Terras 2000´s Bonding and Resolution
model; see also Rigalleau et al. 2004).
63
SUM-1:

Semanticity varies not just with domains but


also with the level of enquiry: grammar or
processing.

In fact, it varies even more because, as Berg


has suggested, domain penetrability may be
sensitive to the direction of encoding, even
in processing: semantic interference is more to
be expected in production than in
comprehension.

64
SUM-1:

Spanish patterned with German, not


with English, on predictable grounds: its
strong morphosyntax keeps semantic
penetration at bay.

65
SUM-1:

The opposite pattern of interference


where form happens first:
The processing of epicenes and the LAN +
P600 brain waves to all agreement
violations in Romance.
These are very likely the first cues to
clause construction (Hawkins 1994, 2004).

66
SUM-2:
Left-oriented SVO clause-building rails in the
processing of cataphora (Cowart & Cairns 1987;
Van Gompel & Liversedge 2003; Kazzanina et
al. 2006).

67
SUM-3:
The function of agreement

Typically, as in English, gender


disappears where a fixed word order
takes over, and it stays where there
usually is no fixed word order.

Can we still maintain the old view that


gender is a lexical –not a computational-
thing? Maybe not.
68
Sum-3:

Number may be accessed semantically,


directly as it were, whereas gender can
only be captured by the co-variance of
form.

69
SUM-3:

The reason why gender is a better clause


constructor than number is because number
does not tell enough entities of the world apart.

Conversely, in a world of continuous variability,


arbitrary gender classes allow us to co-classify
pieces of the world ad infinitum. They provide
infinite construal, creativity.

Gender is generative

70
SUM-3:
So gender may be instantiated lexically,
but for syntactic reasons.
Formal autonomy.
Formal co-variance is a crude and finite
means of expressing infinite predications.
Lacking gender, such a crucial clause
construction role must be attributed to a
fixed word order and to the orientational
reflexes that this affords.

71
SUM:

Agreement is not “superfluous”,


“dysfunctional”, and “useless”, a
“form focused activity”, any more
than word order is.

It is a core process of grammar and of


processing, in the narrowest sense.

72
The to-do list:

Repeat completion study

Geometry in Spanish

Attraction

73
THANK YOU VERY MUCH FOR
YOUR PATIENCE

74

S-ar putea să vă placă și