Chapter 3 Theories of language acquisition and the acquisition of aspect

8/18/2019 Chapter 3 Theories of language acquisition and the acquisition of aspect

http://slidepdf.com/reader/full/chapter-3-theories-of-language-acquisition-and-the-acquisition-of-aspect 1/26

Chapter

Theories

of language acquisition and the

acquisition of aspect

Asp ect constitutes an essen tial part of both lang uag e use reflected in the

lexicon and the gram m ar) and langua ge acqu isit ion reflected in the

early emergence

of

conceptual

and

l inguist ic expressions).

In the

last

tw o

decades,

this area

has not

only received extensive l inguist ic analyses

as

discussed in the previous chapter, but also attracted an en orm ous n um be r

of empirical studies

and

produced many theoretical explanations

in

both

first

and second language acquisit ion. The study of the acquisit ion of

aspect can provide

significant

ins ights in to the unders tandin g of general

theoretical issues in langua ge acqu isit ion. In this book we aim at discov-

er ing the psychol inguis t ic mechanisms that can capture how learners

acquire lexical and grammatical aspect. In this chapter, we first provide

an overview of the broad theo retical issues cur rently und er debate in first

and

second language acquisit ion, as

these

issues are important and

re levant

to the specific questions we ask W e then brief ly review exis t ing

empirical evidence from crosslinguistic studies of the acquisit ion of

aspect

in

diverse languages,

first in

child language, then

in adul t

second

langu age acqu isit ion. W e discuss various theoretical issues that are

particularly relevant

to the

empirical data

in

L I

and L2

acquis i t ion

of

aspect.

3 1 Nativist

and

functionalist approa ches

to

language acquisition

3 1 1

First

language

Current debates in language acquisit ion center around the issue of

whether language learners, childre n or adults, are equipped with inn ate

principles

or

mechanisms that enable

th e

acquisit ion

of

language. Nat iv-

is t

propone nts argue strongly for the existence of a

priori

d o m a i n -

specific pr inciples or mechanisms for language, a l though the detai ls of

their arguments vary greatly. Some argue that the innate components

involve

biologically pre-wired categories or constructs for w hich the

learner

simply needs

to find out how

they

are

instantiated

in

specific

Unauthenticated

Download Date | 4 2 16 6:29 PM



30 heories

of


fo rms in

each language

to be

learned (e.g., Bickerton 1981,

1984 .

Others argue that they involve specific linguistic principles

or

grammati-

cal categories that are genetically determined, for exam ple, C ho m sky 's

argument

that

th e

initial state

of the

language faculty

is a

biological

endowment

to the hu m an species (e.g., Ch om sky 1975, 1981,

1990 .

The issue of language innateness is an old philosophical problem (the

Plato 's problem),

but in its

modern form

is

largely attributable

to

C h o m -

sky's proposal that there

is a

Language Acquisit ion Device (LAD) which

specifically allows the human species to acquire language (Chomsky

1965, 1975, 1988, 1990).

The LAD was

later formulated

as

Universal

Grammar (UG),

an

invariant

set of

principles

and

parameters universal

to

al l

hum an languages

and

part

of the

biological endowment

of all hu-

mans. According to UG theories, only the specific values of these pa-

rameters have

to be set

according

to

various language-specific

propert ies

to which the learner is exposed. This principles-and-parameters approach

has been extensively adopted not only in studies of first language acqui-

sition (Hyams 1986, Wexler and Manzini 1987, Wexler 1998 , but also in

studies of second language acquisit ion (Epstein, Flynn, and Martohard-

jono 1996, White 1985).

19

Today,

few

people would disagree with

th e

idea that humans,

and not

other species, hav e a special, biological cap acity to learn lan gu ag e.

However, just what

constitutes

this capacity

and how it is employed in

langu age acqu isition has been a m atter of intense debate

(Elman

et al.

1996, Sam pson 1997, Tomasello 1995). N ativists argue that this capac ity

mus t

include principles specifically designed

for

language, such

as the

pro-drop param eter (whether a language allows subjectless

sentences,

Hyams 1986),

th e

binding parameter (whether

a

subject

NP can

intervene

a reflexive pro nou n and its antecedent, Wexler and M anzini 1987), and

specific

semantic distinctions (the process-state

and

punc tua l -

nonpunctual distinc tions , Bicker ton 1981, 1984 .

These

principles,

nativists claim , are already in place when children are initially faced with

the task of language learning: they are not learned. In contrast to nativ-

ists, other researchers, loosely labeled as fun ctio na lists , argue tha t

language itself, or its relevant conceptual constructs, is not an innate

capacity, and much of language is learned. They argue that nativists

often confuse universali ty with innateness (Givon 1979, 1995; DeL ancey

1998; see

further

discussion in 8.4.3). Language, in their view, has to be

considered as a communicative device in i ts social and pragmatic envi-

ronment, and language acquisit ion should be examined along with

other

Unauthenticated




Nativist and functionalist approa ches 31

areas of cognitive development (Bates and M acW hinney 1982, 1987;

Snow 1977). Language acquisition, in this view, consists of ex tracting the

patterns that hold between forms and mean ings in con tinuou s speech,

and linguistic input and the ability to analyze the linguistic input are

essential

to the

learner

in

this

process

(Bo we rm an 1985, 1989;

Li,

B u r -

gess,

and

Lund 2000; Maratsos

and

Chalkley 1980). Readily applied

to

th e

acquisition

of

tense

and

aspect, this approach would suggest that

English-speaking children's early

use of the

past tense w ith resultative

verbs may be due to the child's attempt to model past tense forms in the

mother 's

speech (Brown 1973; Shirai

and

Andersen

1995).

20

The

nativist-functionalist

debate on language acquisi t ion has recent ly

been cast into a new form of debate on the represen tation of ling uis tic

symbols

and

rules (see

Elman et

al. 1996; Klahr

and

M acWhinney 1998

for

a summary). Much of Chomsky's theory has been conceptualized as

a representational theory, according to which knowledge of language is a

set of

internalized symbolic rules

in the speaker's

mental representation.

The

acquisition

of

language

is a

process

by

which

th e

appropria te rule

system is selected, and the use of language is a behavior governed by

rules in the representation (Chomsky, 1990). A well-known rule-l ike

behavior is the fact that young English-speaking chi ldren overgeneral ize

th e

past tense form -e d

to

irregular verbs, producing

errors

like goed

breaked

and

corned

instead of

went broke

and

came

(Brown 1973,

Kuczaj

1977).

A rule-based, sym bol-representation analysis wo uld

account for this type of results by arguing that children have internalized

a linguistic rule ( adding -ed to a verb to make th e past tense ) and

applied th e rule to o broadly (Berko 1958). Thus, th e acquisition of the

English

past tense would involve two separate processes: learning of the

regular rule and learning of the irregular exceptions. Rumelhar t and

McClelland (1986) argued against this account, suggesting that rules

are

convenient linguistic descriptions and not representational entities. They

proposed that children learn both regular and irregular verbs with th e

same mechanism that can be operationalized in a parallel distr ibu ted

processing (PDF)

network,

in

which

a

large number

of

processing un its

are connected and w ork in parallel, cap turin g ling uistic behavior by

patterns of activation distributed throughout the network. In this per-

spective, overgeneralization reflects the child's ability to extract statistical

regularities

in the

input ,

and use a

pat tern product ively (e.g.,

f low

glow

and slow all take -e d to make the past, and so should blow}.

Unauthenticated




32 heories of language acquisition

The acquisition of the English past tense ha s served as a classic

example in the debate on rules versus connections as linguistic represen-

tations, wh ich has generated a large num ber of studies from both theo-

retical perspectives. On the one hand, symbolic theorists argue that

l anguage acqu isition shou ld be characterized as a rule-based sym bolic

process (Pinker and Prince 1988; Pinker 1991, 1999; Ling and Marinov

1993; Marcus et

al.

1992).

21

On the other hand, connectionist researchers

argu e tha t lan gu age a cqu isition is a statistical learning process in whic h

the learner forms linguistic patterns on the

basis

of the regularities

inherent in the

input-output mapping process (Rumelhart

and

McClel-

land 1986; MacWhinney and Leinbach 1991; Plunkett and Marchman

1991, 1993; Seidenberg 1997). This rule-vs.-connection debate is rooted

deeply in cognitive science in the opposition between viewing th e h u m a n

mind as a m odular system whose architecture is largely predetermined or

innate

(Ch om sky 1988; Fodor 1983; Pinker 1994) and view ing it as a

highly interactive system

in

which multiple components

can

interact

simultaneously

at all

levels (Rumelhart, McClelland

and the PDP Re-

search Group 1986; Elman

et al.

1996).

Both nat ivist and functionalist approaches have been used in

research

on the acqu isition of aspect. One impo rtant issue in this regard con cern s

the emergence

or

origin

of

semantic categories

of

lexical aspect: where

do the inherent semantic notions like states, activities, accomplishments,

and

achievements come from?

Or,

where

do the

semantic features that

define these categories (e.g., telicity, pu nc tua lity,

and

dynamici ty)

come

from? The contrasting views o f nat ivis ts an d functionalis ts are relevant to

the issue. Nativists argue that these categories or featu res and their

ontological substrates are innately specified in the human genome

e.g.,

Bick erton 1981; Pink er 1984, 1989; Sm ith 1997). For example, Bicker-

ton (19 81) proposed that semantic distinctions between states and proc-

esses and between pun ctu al and no npu nctu al verbs are biologically pre-

programmed, and that children can map them onto linguistic categories

a

priori

In contrast, func tionalists argue that the so-called innate catego-

ries may be purely cognitive categories that are salient to the learner, and

that they emerge from the ch ild s interaction with the world and their

sensitivity

to event characteristics such as mov ements, chang e of state,

and change

of

location (DeLancey 1998). Moreover,

the

learner

m ay

acquire these categories on the basis of analyzing statistical regularities in

th e

l inguistic input,

fo r

example, d istribution al properties

of

lexical

and

morphological

co-occurrences

(Li, Burgess, and Lund

2000).

In this

Unauthenticated




ativist and

functionalist appro aches

33

book, we advocate a funct ional , input-based, probabi l is t ic learning

approach to the problem of the acqu isition of aspect.

3 1 2 Second language

Research on second language acquisition has also been conducted from

both nativist and functional perspectives. Unlike research on first l an -

guage acquis i t ion,

in

which

th e

issue

of

innateness

has

been

of

para-

mount importance, research on second language acquisition has been

more concerned with the characterization of learner langu age often

referred to as inter language) , us ing var ious l inguis t ic frameworks.

Particularly important approaches in this domain inclu de those based on

sociolinguistics, pidgin-creole l ingu istics Sch um an n 1978; An dersen

1983), fu n ction al lin gu istics A ndersen 1984; Sato 1990 , and l inguis t ic

typo logy Gass 1984; Eck ma n 1977; Eck ma n, Bell,

an d

Nelson 1988 .

In these fram ew orks, the issue of in naten ess has not been addressed

head-on. Instead, researchers hav e tried to accou nt fo r un iversal aspects

of

second language acquisition

by

relying

on

multiple factors, such

as

saliency, frequency, processing cost,

an d

fo rm- func t ion mapp ing .

Inn ateness has been regarded as only on e of the possible co n trib u tin g

factors.

However,

a

group

of

researchers have attempted

to

address

more

directly the issue of innateness in L2 acquis i t ion from th e generative in

particular UG) approaches. Some of these researchers argue that UG is

accessible

in its

fu l l form

to

adul t

L2

learners,

in the

same

way as it is

available to

LI

learners Epstein,

F lynn ,

an d Martohardjono

1996 .

Others make

a

less radical assumption, arguing that

UG is

only part ial ly

available to L2 learners Schach ter 1988). Still other researchers assu me

that UG is not at all

accessible

to L2

learners Bley-V roma n 1990;

Clahsen and M uy sken 1986). Given the enorm ous disagreement am on g

researchers

on

this issue see com me ntaries

on

Epstein, Flynn,

and

Martohardjono

1996 ,

it

will

still

be a

long

way

before

a

clear picture

emerges regarding the accessibil i ty or availabili ty of UG to adu lt L 2

learning.

We need

to

keep

in

mind that

the

issue

of

access

to UG may be a

moot question

if, as

suggested above,

it is

possible

for the

learner

to

acquire the target langu age w ithou t recourse to UG . On e of the cen tral

arguments for access to UG in adult L2 acquisition relies on the empiri-

cal

observation that

L2

learners, wi thout expl ic i t ins t ruc t ion ,

can

some-

Unauthenticated




34

heories

of


times acquire target grammar not present in LI

e.g.,

constraints on NP

movement such

as

subjacency). However, this observation does

not

necessarily constitute evidence for the role of UG. It is possible that the

linguistic input alone

is

sufficient

for the

learner

to

derive such con-

straints. Elman

1990, 1991,

1992) showed

how a

co nnectionist

network

can

learn such constraints e.g., agreement relationships between items

across clause bou ndaries ) by abstracting co-occurrence patterns of

linguistic

elements

in the

input.

If

children

are

equipped with such

a

powerful

learning mechanism,

and if

that mechanism still works

fo r

adults,

there may be no need to posit UG see also MacW hinney 1993).

O ur discussion of the nativist-vs.-functionalist debate and the rule-vs.-

connection debate has so far remained general in scope. However, we

shall see in the

next sections what implications these debates have

for the

acquisition of

aspect.

3 2 First language acquisition

of aspect

The last thirty years have witnessed an enormous amount of research,

both

theoretical

and

empirical, into

the

question

of how

young children

an d

adult second language learners acquire grammatical

and

lexical

aspect in conne ction with their acquisition of tense morphology. Re-

searchers have taken

a

variety

of

perspectives

in

these studies, across

a

variety

of languages including Chinese, English, Dutch, French,

German,

Italian,

Japanese, Modern Greek, Polish, Spanish,

and Turkish.

22

In

this

section we review major empirical studies in the acquisition of aspect in

connec tion with tense. As was discussed in Chapter 1, tense an d

aspect

are closely related

categories

in adult language, and they are also intri-

cately

related

in

child language. Thus,

our

review covers

th e

acquisition

of both tense an d aspect, with primary focus on aspect.

Obviously, it would not be possible to give a detailed discussion fo r

each stud y. Our goal here is to fam iliarize the reader with the basic

emp irical patterns that occur repeatedly in languag e acquisition, and to

prepare o ur discussion for importan t theoretical issues in the next sec-

tion. W e will

give

further

detailed analyses

of

some

of

these results

in

Chapters 4 through 6, where we discuss the acquisition of aspect specifi-

cally in

English, Chinese,

an d

Japanese.

The

following review

will

g ro u p

together studies that

are

closely related, both

by

language

and by

topic.

Unauthenticated






36 Theories of language acquisition

These researchers categorized th e enactment situations according to

aspectual fea tures, two of w hic h are of direct interest

here:

perfective

events that gave

a

clear result, covered some distanc e

in

space,

and

term inated at a predeterm ined poin t (e.g., a truck push es a car to a

garage),

and imperfective

events that

did not

lead

to any

result

and

consisted of circular movements of animals in their natural habitat (e.g.,

a fish swims in a basin). Before age 6, the children used passe compose

(perfective past) significantly more often than

present

(present)

fo r

perfective

events, and, conversely, present significantly more often than

passe compose for imp erfective events. Within the category of perfective

events, the passe compose was used almost exclusively for non -dura tive

events. On the basis of these findings, Bronckart and Sinclair concluded

that

the distinc tion between result and process is the predom inant and

perhaps

th e

only aspectual feature

in the

language

of

French children

below age 6.

Ant inucci

and

Miller

(1976)

studied

th e

spontaneous speech

of

Italian

children from 1;6 to 2;6, and showed that their child participants first

restricted their

use of the

perfective past tense form

passato prossimo to

change of state verbs that specified actions with a clear result, such as

fall close find and break (Italian equivalents). The children did not

combine activity

and

stative verbs with

th e passato prossimo but

rather

with

th e

imperfetto (imp erfective past). Intriguin gly, children m ade

th e

inflectional

endings

of

past participles, which

is

part

of passato

prossimo

agree with

the

number

and

gender

of the

object

of the

verb;

in the

adu l t

langua ge, the participle agrees on ly w ith the subject (and only when the

verb takes be as opposed to have as its auxiliary). The chi ldren ' s

invention

of a

syntactic agreement rule suggests that they interpreted

th e

past participle as specifying an attribute of the object rather than an

action performed by the subject, and so according to Antinucci and

M iller, they mu st be focusin g on the resultant state of the object affec ted.

The authors proposed that children at this early stage are not cognitively

prepared for past reference unless there is a clear result inhe rent in the

action that serves as a link between the present and the past.

Aksu (1978, Aksu-Êïò 1988) examined Turkish children's acquisi-

tion

of

tense, aspect,

and

m odali ty.

She

reported findings similar

to

those

from

Italian. H er child participants first used th e past form fo r directly

experienced events -d l with change

of

state verbs

to

mark punctual ,

resultative events. In contrast, they used the past form for indirec tly

experienced events ra/f w ith verbs indicating accom plished states and the

Unauthenticated




First language acquisition 37

progressive form -lyor wi th activity verbs. Acco rding to Aks u (1978),

these patterns of restricted co-occurrences between temp oral m orp h ol-

ogy and the inherent semantics of verbs show up in the speech of Turk-

ish children before app roximately 4;6. T he results th us led Aksu to

argue that aspectual distinctions take precedence over temporal order

distinctions in language acquisition. In a more recent study, Aksu-Êïò

1998) directly examined

th e

relationship between emergent tense-aspect

mo rph ology and lexical aspect classes (state, activity, ach ievem ent, and

accomplishment) in longitudinal data from Turkish children. T his s tudy,

confirming her earlier find ings , show ed strong associations between dl

and punctual verbs , and between -lyor and activity verbs. She also

examined

th e

child-directed speech

of a

mother , f inding s t r iking paral-

lels between th e mother s and c hild ren s speech in terms of associations

betw een m orphology and semantic verb types. However, she also fou nd

that

th e

development

of -d l is

faster than that

of

-lyor, despite

th e

h i g he r

frequency of -lyor in the inpu t . She attributed this latter

f inding

to the

cognitive saliency and the clearer fo rm-fu nction ma pping for

dl.

Lexical aspect also seems

to

inf luence young children s tense

and

aspect marking in Modern Greek, as discussed by Stephany 1981,

1997). Stephany classified verbs in Modern Greek into three categories:

statives

(e.g., know, be cold , resultatives (e.g., fall, take , and non-

resultatives e.g., cry, read . In ch ildren s non-m odal expressions,

23

resultative verbs occurred more frequently than

non-resul ta t ive

verbs

with perfective aspect, whereas the reverse was true for

imperfective

aspect. Stative verbs were exclusively combined w ith imperfective aspect.

Furthermore, there were rigid co-occurrences between tense markers and

aspect markers: past tense occurred only w ith perfective aspect and

present tense only wi th imperfect ive aspect . Stephany hypothesized that

th e

co-occurrence

patterns displayed in child Modern Greek might

derive from patterns

in the

input language.

H er

analysis indicated that

the distribution o f verb categories w ith tense and aspect m arkers in

ch ildren s language conforms surpris ingly well

to the

pattern

found

in

the

moth ers speech directed to children .

To summarize, most of the studies reviewed above suggest that young

children initially restrict their use of particular tense and aspect markers

to verbs with particular lexical aspect values, and that this restriction

might

s tem from ch ildren s modeling

of the

distributional properties

in

th e

input ,

as

B rown 1973)

and

Stepha ny (1981) hav e suggested. Th is

acquisition pattern stimulated many subsequent discussions

and

debates,

Unauthenticated




38

Theories

of


and

motivated several theoretical proposals co ncernin g

ch i ld ren ' s

semantic and morphological development. Two of the mos t prominent

theoretical proposals that have draw n on the data reviewed above are

Bickerton's Language Bioprogram Hypothesis and Slobin s Basic Child

Grammar hypothesis. In connection with these proposals, researchers

have also debated

on the

so-called De fective Tense Hypothesis ,

according to which children initially use tense markers to

encode

aspec-

tual values. We now examine these theories and issues.

3.2.2. Language

Bioprogram

Hypothesis

Basic Child Grammar

and

Defective

Tense Hypothesis

On the

basis

of

data

from

studies

of

Creole languages,

Bickerton (1981,

1984) put forth the Language Bioprogram Hypothesis. This hypothesis

has attracted attention

from

language acquisition researchers in recent

years because many

of its

arguments rest

on

interpretations

of

data

in

child ren's acquisition of tense and aspect th at we reviewed

above.

The

hypothesis

advocates that certain semantic distinctions are biologically

pre-programm ed and emerge early in hum an language acquisition.

Because the distinctions are hardwired ahead of time, the child learner

simply needs to find out how they are instantiated in specific forms in

th e language to be learned. For Bickerton, tw o important innate d istinc-

tions in the domain of aspect are between state and

process

and between

punctual

and

nonpunctual categories. Because

th e

distinctions

are by

hyp othesis in nate , early on in lang uage developm ent states will be

marked

differently from

processes,

and

punctual situations will

be

marked differently from no npu nctu al situations, probably

by the use of

different

tense-aspect markers.

Bickerton supported his claims

with

evidence

from

Creole grammars,

arguing

that in the absence of relevant input (pidgins, the predecessors of

Creoles, do not have tense-aspect markers), first-generation Creole speak-

ers inv en t tense-aspect systems to mark th e

bioprogrammed

distinctions.

Drawing in addit ion on child language data from Ant inucci and Miller

(1976), Bronckart and Sinclair (1973), Brown

(1973),

Kuczaj (1978),

and Slobin and Aksu (1980), he argued that children

first

use the tense-

aspect markers of their language to mark the distinctions between state

and process and between punctual and nonpunctual. For example, Brown

(1973) observed that children learning English do not overgeneralize th e

progressive marker -ing

to

stative verbs. Bickerton interpreted this

Unauthenticated






40

Theories of language acquisition

m ar

hypothesis. Thus, while we can clearly assign th e label nativist to

Bickerton 's hypothesis ,

w e

cannot

do so

with certainty

to

Slob in ' s

approach.

2 4

Related

to the

arguments

of

Basic Child Grammar

are the

empir ical

findings

that

chi ldren

initially

restrict, or underg eneralize, the use of

certain tense-aspect markers.

In

particular, children initially tend

to use

tense markers (e.g. ,

ed

in English) to encode aspectual distinctions

(e.g.,

telicity

or resultativity).

This

le d

some authors

to

argue

for an

aspect

before

tense

hypothesis in child language (see our discussion in 3.3.1

fo r

a

related hypothesis

in

second language).

For

example, Bloom

and

her colleagues (Bloom,

Lifter,

and Hafitz 1980; Rispoli and Bloom

1985) proposed that aspect precedes tense

in

child E nglish, since

th e

initial

use of tense markers by their English-speaking children was

always

r e dundan t

with

lexical aspect (e.g., ed always occurred

with

complet ive or resultative verbs). Bronckart and Sinclair

(1973)

suggested

that

French-speaking chi ldren use tense forms to encode aspectual

differences before age 6. Similarly, An tinucci and Miller (1976) arg ued

that Italian-speaking children

at the

earliest stages

are not

able

to

encode

temporal order relationships with tense morphology because of cognitive

insufficiency.

Smith (1980) wa s the first to argue against this type of proposal .

Focusing

on

Bronckart

and

Sinclair's

(1973)

study

in

particular,

she

showed that

a

similar experiment

in

English

did not

replicate

th e

French

results. Eng lish-speak ing children between

th e

ages

of

4;7

and

6;6

produced both simple past forms

and

past progressive forms

by age 5,

al though

perfective events

(i.e., telic

situations) elicited more simple past

forms

and

im perfectiv e events (i .e., atelic situations) more past prog res-

sive forms. Smith also drew on data from children's spontaneous speech

(Brown's and Kuczaj ' s data), and showed that children between

2;5

and

5;8

used past tense form s with both perfectiv e (i.e., telic)

and

imperfec-

t ive (i.e., atelic) verbs.

Weist

et

al. (1984) echoed Sm ith's objection,

and

argued

in a

more

systematic manner. They took issue

with

Bronckart and Sinclair (1973),

Antinucci and Miller (1976), and Bloom, Lifter, and Hafitz (1980),

labeling

th e

claims made

in

these studies

th e

Defective Tense Hypothe-

sis .

By

using experimental

and

naturalistic data

from th e

acquisition

of

Polish, Weist

et al.

(1984) claimed that

at

early stages

of

acquisition,

contrasts in aspect emerged simultaneously with contrasts in tense (both

of w hich

are

grammaticized

in

Polish), thus providing counterexam ples

Unauthenticated




irst

language

acquisition 41

to the aspect before tense hypo thesis. T hese a uth ors showed that the

Polish

children did not use

tense markers redundantly

or

defectively;

they used tense inflections to express deictic relationships as in adult

Polish,

and

those inflections were

not

combined solely

with

verbs that

encode a resultant end-state. Weist and his colleagues made the same

point in other studies (W eist 1983; W eist, W ysocka, and Ly ytin en

1991):

yo ung children

as

early

as 2;6 are

able

to

unders tand

and

produce

th e

basic contrast between perfective and imperfective aspect. These authors

pointed

out that olish children s ability to make deictic tense references

at an early age is consistent with Piaget s theory that by the end of the

sensorimotor stage, children are conceptually capable of thinking about

prior and subsequent events. Because the debate on the Defective Tense

Hypothesis has

emerged

as important, we

will

take up the related issues at

several points in the remainder of this chapter.

3 2 3

A

critical analysis o f empirical

and

theo retical issues

In

the above sections we presented an overview of the m ajor patte rns

found

in

children s acquis i t ion

of

tense

and

aspect

in

d iverse languages,

along with several theoretical exp lana tions. W e presented these studies

and

arguments

in a

rather non-committal fashion.

In

this section,

we

will

take a critical look at some of the empirical and methodological issues

that have led to confusion in the l i terature. In particular, we will discuss

three sets of issues: the notion of aspect, the im pact o f m etho do logy and

age

differences

on

research findings,

and the

relative versus absolute

tense-aspect associations.

Three levels o f

the

not ion o f

aspect

In

Chapter

1

(1.1.2),

we

differentiated g ramm atical aspect from lexical

aspect

by

stating that grammatical aspect marks

a

verb with

a

par t icu lar

viewpoint

toward the described situatio n, whereas lexical aspect involv es

the inherent temporal meanings of a verb. In Chapter 2 (2.2), we cau-

tioned against a con fusion between the sem antic contents of verbs and

th e temporal properties of situation s to whic h verbs refer. Th us, there are

three levels of the notion of aspect that we should pay attention to :

Unauthenticated




42

Theories

of


(1) The tempo ral properties of situations as part of reality,

categorizable on a conceptual level. This notion of aspect

belongs

to

what Lyons (1977) calls

the

denotata

of

verbs;

(2) The semantic contents of linguistic items that encode situ-

ational properties, such as whether a verb characterizes a

situation as incorporating a temporal boundary or a result.

This notion

of

aspect

is

what

we

call

in

this book lexical

aspect ;

(3) The aspectual perspectives that languages exploit to express

whether a particula r situation encoded by a particular verb

is

in

progress, ongoing,

or

completed, e.g., perfective

and

imperfective. This notion of aspect is what we call in this

book grammatical

aspect .

As discussed earlier, there is no one-to-one correspondenc e between

these levels although they often overlap. However, several studies that

we

reviewed

above have confused situational properties, lexical aspect, and

gram m atical aspect, and some of the disagreements between dif fere nt

authors

clearly stem from su ch confu sions. Often, the same term as -

pect has been used to refer to

different

levels of the notio n of aspect,

and the same notion of lexical aspect has been referred to with different

terms.

The

well-known aspect before

tense

hypothesis

is a case in

point. Bloom, Lifter, and

Hafltz

(1980)

claimed that their results are

consistent with the principle

aspect

before tense in linguistic theory.

Their finding,

as

discussed earlier,

was

that children initially

use

tense

markers to encode inherent properties of verbs. However, the linguistic

principle of aspect before tense, according to Woisetschlaeger (1976)

and

Bybee (1985), specifies that grammatical aspect markers are gener-

ally positioned closer

to the

verb stem than

are

tense markers

in the

world's languages. Since what Bloom, Lifter, and Hafitz meant by aspect

is ,

in fact, lexical aspect, their findings are different from the linguistic

theory

of

aspect before tense.

25

Similarly,

th e

findings

in

Aksu

(1978)

and Ant inucci and Miller (1976) are different from th e linguistic princi-

ple. These findings

may be

characterized

as lexical

aspect prior

to

tense

in

child language.

26

There is also a conf usion between lexical aspect and temp oral pro per -

ties of situations in studies of the acquisition of tense-aspect morphol-

ogy. Temporal properties

of

situations

are

often assumed

to be

directly

mapped onto verb meanings, and terms such as aspectual distinctions

Unauthenticated




irst


43

or aspectual perspectives are used am bigu ously with respect to

whether they apply to the real-world denotata of verbs or to the inherent

semantic properties of verbs. This confusion may be attributable to

Vendler (1957), w ho originally used th e term time schemata of verbs .

It

is not

clear

whether time schemata of verbs are semantic properties of

verbs

or

temporal properties

of

events,

or

both. Thus, questions often

arise as to whether

Vendler 's

classifications concern verb types or event

types. A n example that well illustrates this confusion is Bicker ton ' s

state-process

distinction and punctual -nonpunctual dis t inct ion .

Bickerton

considered these distinctions as innate, universa l, and lan -

guage -indepen dent distinctions on a conceptual level (Bickerton 1981),

yet

he labeled them semantic distinction s and discussed them as

though they were the same as distinctions between stative and

process

verbs

or between punctual and nonpunctual verbs. For example, Cziko

(1989a) and Bickerton (1989) debated w hich semantic components are

at

stake

in the

punctual-nonpunctual distinction. Interestingly, Bickerton

(1989)

criticized Cziko (1989a)

for

using

th e

te rm pu nc tua l -

nonpunctual

verbs rather than pun ctual-n onpu nctu al s i tuations . In

reply, Cziko pointed out that the con fusion originated w ith Bic kerto n

(1981) him self, and argued that Bick erton (1989) still took the state-

process

distinction, if not the punctual-nonpunctual distinction, as

referring

to

verb semantics rather than

to

situations.

Confusions of

this kind have also resulted

in

misinterpretations

of

findings on the

acquisition

of

tense

and

aspect.

For

example, Bickerton

(1981), among others, compared

Bronckart

and Sinclair's (1973) study

of French directly w ith An tinucci

and

Miller 's (1976) study

of

Italian.

However, these studies are not directly comparable, because Antinucci

and

Miller dealt with

th e

role

of

inherent verb meanings

in

ch i ld ren ' s

acquisition, while Bronckart and Sinclair were concerned with the role of

temporal properties of situations (events or actions, in the ir terms). For

example, in Bronckart and Sinclair th e situations described by il a

marche

'he

walked'

and // a

saute

su r

chaque

barriere et il est

alle dans

la ferme 'he jumped over each fence and he went into the fa rm ' were

both classified

a priori

as perfective durative actions , but the two

sentences

obviously differ w ith respect

to

their lexical aspect:

th e first is

atelic while the second involves two verb phrases, both of which are telic.

Our

discussion points

to a

serious problem

in the

literature,

and we

think

that

it is

important

to

distinguish these different levels

of

aspect

theoretically and empirically. Of course, a clear distinction between these

Unauthenticated




44 Theories

of


levels

is often empirically

difficult ,

and w e are not suggesting that they

be

treated completely separately.

As we

discussed

in

section 2.2,

th e

correspondence between seman tic contents

of

lexical aspect

and

s i tu-

ational

properties

is not

arbitrary,

and

they

often

overlap

in

practice.

In

th e

research

w e

report here,

w e

a t tempt

to

honor these different levels

of

aspect, while recog nizing that a strict d ifferentiation between them may

become trivial in some

cases.

esearch

methodologies and age differences

It

is

wel l known that differences

in

methods

of

data collection

m ay

lead

to

discrepancies

in

estimates

of the age at

w hich chi ldren acquire

a

g iven

syntactic

or semantic

function

(e.g.,

Bowerman

1978a).

In studies of the

acquisi t ion of

tense

and

aspect, there

are

clear discrepancies between

conc lus ions that are based on longi tu d inal , spontaneous speech data

versus those that are based on experimental data. Stud ies based on

longi tudinal

data tend to report on earlier phases of acquisition and point

to

earlier acquisition

of the

relevant forms than

d o

s tudies based

on

experimental data. For example , Ant inucc i and Miller

(1976)

and

Bloom, Lifter, and Hafitz (1980), who used

longitudinal

data, claimed

that there were initial restricted uses of tense-aspect marking around age

2, and

they

further

noted that already

by

about 2;2, these restrictions

began to relax. How ever, results

from

Bronckart and Sinclair s

exper i -

mental s tudy were very different

in

th is regard: chi ldren s

use of passe

compose

and present in French was closely correlated with perfective and

imperfective

properties

of

events until they were

6

years old.

Not only discrepancies betwee n long itud inal and experimental stud ies

exist in the

estimate

of the age at

w hich tense-aspect m arking s develop,

bu t there are also differences between s tudies that use the same method-

ology. For example, both Bronckart and Sinclair (1973) and McShane

and Whittaker (1988) used experimental methods , bu t these tw o s tudies

found

d i f ferent

developmental t rends: unl ike chi ldren

in

Bronckart

and

Sinclair s s tud y, chi ldren in M cShane and W hi t taker s s tudy d id no t

initially

(i.e., at age 3) use

ed

and

in g differently

fo r different situations

(telic vs. atelic, iterative vs. noniterative, etc); they did so only at a later

age

(i.e.,

at

ages

4 and 5), and

their awareness

of the

contrastive aspectual

relation between the past tense and the progressive w as not complete even

by age 6.

A ks u s (1978) s t udy , d r awin g

on

both longi tudinal

and

experime ntal data, is again

different

w ith regard to the de velopm ental

trend. She suggested that Tu rkish ch ildren have an initial association

Unauthenticated




First language acquisition 45

between lexical aspect distinctions and tempo ral m orph ology , but they

abandon this association and shift to the adult usage of -dl ô ç À ò and

-lyor between

ages 4 and

4;6.

Lo ngitud inal and experimental studies in the dom ain of tense and

aspect mark ing both have their virtues

and

problems.

On the one

hand,

experim ental studies, unlik e longitu dina l ones,

are

unable

to

trace pa t-

terns of tense-aspect acquisition before about 2;6, because the usual

experimental tasks, such

as

telling

a

story

or

describing

an

acted-out

situation, are too difficult for very young children

before

that age. The

difficulty

o f

running experiments

with

very young chi ldren

can

some-

t imes

be

seen

in

studies that

use

both longi tudina l

and

expe r imen ta l

methods: for example, Smith (1980) used longitudinal data from chi l-

dren ranging in

ages

f rom 2;5 to

5;8,

but her experimental data were

obtained from much older children (ages 4;7 to 6;6). On the o ther hand,

experimental studies have advantages over longitu dina l studies

in

that

factors that contribute to chi ldren s

comprehension

and use of

tense-

aspect markers can be more precisely controlled. A s pointed out by

Fayol, Abdi ,

and Gombert (1989),

studies based

on

longi tudinal analyses

are

limited

by

their inability

to

d i sentangle potent ia l ly confounded

variables, for example, to disentangle th e

effects

of d i f ferent lexical

aspect patterns on childr en s selection o f tense-aspect markers. In the

research w e report on in this book, we have used both expe rim ental and

longi tudinal

m ethods in order to assess child ren s acq uisitio n of aspect

under

di f ferent c ircumstances

and in

di f ferent modali t ies

(e.g. , compre-

hension and product ion) .

Finally,

th e

specific tasks used

in the

studies

m ay

also affect

th e

results

as well as the interpretations. Previous studies in the acq uisition of tense

and aspect have used various tasks, inclu ding com prehension , imitation,

and product ion. Each o f these tasks could lead to s l ight ly

d i f fe ren t

outcomes, and sometimes even the same type of task could lead to

different

results. For example, in an experimental setting one could

collect child ren s pro duc tive speech

from

e i ther spontaneous interact ions

or

f rom experimental ly-el ic i ted condit ions.

These tw o

ways

of

data

collection

m ay

lead

to

di f ferences.

In

spontaneous interact ions, ch i ld ren

can funct ion within th e l imit o f their cognitive and l inguistic potential:

they

can perform as they wish, avoid ing lin gu istic structures or lexical

items

for

which they

are not

ready.

In

contrast,

in

elicited situations,

th e

experimenter often forces children

to

explore their full potent ia l because

they have

to

perform

as

required

by the

task, which often go s b ey o n d

Unauthenticated




46 Theories

of


their competence. In some cases, even subtle differences in methodology

could cause discrepancies. McShane

and

Whittaker (1988) pointed

out

that

in an enactment task in which children are asked to describe what

happens, it makes a difference whether the toys and props

remained

visible to the

chi ld

wh ile the child is describing. They suggested that

children tend to produce more present tenses when the toys are in sight,

but

more past tense forms when they are removed

from

th e scene. In the

research

we

report

on in

this book,

we

have used

a

variety

of

tasks

in

order to unravel mechanisms underlying the acquisition of aspect,

bearing in mind the impact of task differences.

elative versus absolute associations

As

discussed earlier, there

has

been

an

intense debate

on the issue of

whether children initially use tense morphemes only to mark aspect, such

that

early past-tense marking occurs only

with

telic verbs (e.g., accom-

plishments

and

achievements)

and not

atelic verbs

(e.g.,

activities

and

states). Whereas Bloom

and

colleagues held

to the

idea that there

is

indeed such an association, Weist and colleagues opposed this proposal,

arguing

that

in

early child Polish, tense

and

aspect

emerge

simultane-

ously and independently. In a subsequent exchange of views, Rispoli and

Bloom (1985)

and

Bloom

and

Harner

(1989)

argued that

th e

term

defective

is inappropriate in describing child ren's development, and

that nobody really proposed

th e

Defective Tense Hypothesis. Th ey

pointed

out

that what

is

essential

is

that inherent verb me anings stron gly

influence

the course of chi ldren 's acquisition of tense morphology.

It

seems that

th e

gist

of

this debate

is

that while Bloom

and her

col-

leagues discussed the acquisition of tense and aspect in terms of fre-

que ncy, proba bility, and likelihood of co-occurrence, Weist et al . viewed

the issue in a non -probabilistic way. As Andersen (1989) pointed out,

Weist

et al.

were criticizing what

can be

called

th e absolute

Defective

Tense Hypothesis,

a

strong, all-or-none version

of the

hypothesis.

A c-

cording

to an

absolute version

of the

hypothesis, only telic verbs receive

past-tense inflection; tense distinctions will be redu nda nt and on ly

accompany aspectual distinctions; and only references to

immediate

past

situations

will

be present in ch ildr en 's speech (Weist et al. 1984: 348).

Weist

et al . 's

claim

is

correct

as

long

as

their criticisms

are

targeted

against the absolute version of the hypothesis - a popular view at that

time. How ever, less stringent versions of the Defective Tense Hypo thesis

m ay hold true; namely, past inflections occur predominantly with

Unauthenticated




First language acquisition

47

achievement and accomplishment verbs in the early stages, and

imper -

fective

past marking, which begins later, occurs predominantly with

state/activity verbs initially. Bloom and Harner (1989) and Andersen

(1989) both independently reanalyzed the data in Weist et al. (1984),

and showed that there was, as found in other studies, a strong association

between past tense, perfective aspect, and

telic

verbs in the Polish data as

well.

27

Such association patterns can also be found in Smi th ' s 1980)

experimental data that we reviewed above (3.2.2).

In this book, we will present our own studies that suggest a relative

rather than absolute picture of the associations in early child language.

W e arg ue tha t there are strong a ssociations between lex ical aspect and

particu lar tense-aspect ma rkers, jus t as Bloo m , Lifter, and Hafitz

1980)

have proposed, but the associations are probabilistic rather than absolute

(Li 1990;

Li and

Bo w erm an 1998; Sh irai 1991, 1993; Sh irai

and

A n d e r -

sen 1995).

In this book, w e further a t tempt to unders tand the computa-

tional mechanisms that lead to such probabilistic associations in language

acquisition - associations that are sometimes stronger

(e.g.,

Chinese,

English,

and

Italian),

and

sometimes weaker (e.g., Japanese, German,

and

Polish).

3 3 econd


of aspect

3 3 1 Major crosslinguistic results

Research on the acquisition of aspect in L2 started around the m id 1980s

under

th e

inf luence

of

early LI acquisition studies reviewed above.

An derse n (1986, 1989, 1991) was am on g the first to investigate the

relationship b etween lexical aspect and tense-aspect morpho logy. He

analyzed longitudinal, conversational data from

tw o

adolescent nat ive

English speakers learning Spanish in a naturalistic setting. His analyses

showed the follow ing developmental p aths: past perfective (preterite)

appeared earlier than past

imperfective

(imperfect), and the order in

which verbs emerged with past perfective was achievem ent

->

accom-

plishment

-^

activity -> state , whereas the order in which verbs emerged

with past imperfective was state

->

activity

-^

accompl ishment

->

achievement. Thu s, perfective aspect starts from one end of the cont in-

uum of Vendler ' s categories, and imperfective aspect starts from the

other end. In terms of semantic features, perfective past is first associated

with [+pu nctual, +telic, +dy nam ic], whereas imp erfective past with

Unauthenticated




48 Theories

of


[ -punctual , -telic,

-dynamic].

This pattern

is

largely consistent with results

from

LI

studies.

In

another study, Ramsay (1990) expanded

the

context

of naturalistic

learning to classroom Spanish learners. She used an oral

narrative task with a picture storybook in her cross-sectional study at an

Am erican univ ersity. Her results showed that classroom Span ish learners

clearly follow

a

similar developmental pattern

as

naturalistic learners

in

Andersen's s tudy.

Studies

of the

acquisition

of

English

as L2

also reported associations

between lexical aspect and tense-aspect morph ology similar to those

observed

in

studies

of LI

acquisition. Numerous case studies done

by

Andersen's s tudents

at

UCLA reported strong associations between past

tense marking

and

telic verbs,

and

between progressive marking

and

activity

verbs in learners' spon taneou s speech (Robison 1990;

Yoshitomi

1992; Mishina 1993; Huang 1993; Taylor 1987; Gushing 1987). Such

associations

were also found in cross-sectional studies, for exam ple, in

Bardovi-Harlig and Bergström (1996), who tested classroom ESL learn-

ers in a written film-retell task. Bardovi-Harlig and Reynolds

(1995),

in a

further

large-scale cross-sectional

study

using

a cloze-type

28

fil l-in-the-

blank test, showed that learners had more difficulty in using past tenses

fo r

activities

and

states than

for

achievements

in

obligatory past contexts

(for more detailed discussions of L2 acquisition of English, see

4.4.2).

The acqu isition of Italian as L2 shows similar patterns. G iacalone

Ramat (1995) studied untutored L2 learners of Italian from various

LI

backgrounds

in a

longitudinal s tudy.

She

reported that

th e

past partici-

ple, which is part of p ss to prossimo i.e., perfective past, emerged

earlier than th e imperfective past, and was used mostly with punc tua l and

telic verbs (i.e., achievements

and

accomplishments).

The

past participle

also appeared with some activity verbs, but less frequently than with

achievement

and

accomplishment verbs. Imperfective past forms first

appeared with stative verbs

for all

learners. Giacalone Ramat stated that

her

Italian

L2

study confirms An dersen's idea that past participles

(or

preterit in the Spanish data) and imperfect spread

from

th e

opposite

corners

of

Vendler 's classification,

th e

first

from

th e

corner

of

p u n c t u a l

and telic verbs, the second from the corner of stative

verbs

(Giacalone

Ramat

1995: 300). Th at is, as in the case of Spanish acquisition, past

perfec tive started w ith achievem ents, and past imperfec tive with states.

Giacalone Ramat (1997) also reported

a

high correlation between p ro-

gressive aspect and activity verbs for Italian learners, which is consistent

with

th e

English data

in

both LI

and L2

acquisition.

Unauthenticated




Second language acquisition

49

Housen's (1994) study of the acquisition of Dutch as L2 also showed

associations between lexical aspect of verbs and tense-aspect m orp ho l-

ogy. He analyzed conversational data from a native Eng lish speaker

learning Dutch

at two

different points

in

time

(with a

one-year interval),

and fou nd that present tense form s correlated more w ith du rat ive and

Stative

verbs, whereas past tense forms with pu nc tua l and dy nam ic verbs.

This correlation was stronger at time point 1 and weaker at time point 2.

He concluded that his finding s are basica lly in line w ith those of other

second language studies (e.g., Andersen 1991; Rothestein 1985;

Robison

1990) (Housen 1994: 276).

29

Tw o

studies

of

French classroom learners

at

American universit ies

also showed similar results. K ap lan's (1987) study indicated that

the

learners tended to use perfective past forms

passe

compose)

with

per-

fective

events (corresponding rough ly to achievements and acc om plish-

ments),

and use

present forms

present)

with

imperfective processes

(corresponding roug hly to activities and states). Imperfective past im -

parfaif) w as

late

in

development.

Salaberry

(1998) also

found

that

his

second-semester students

of

French showed

a

strong preference

for

us ing

the perfective past

with

achievement verbs, both in his cloze-type test on

the verb form and in written narratives (the Pear S tory , Cha fe 1980).

Learners associated Stative verbs strongly w ith imp erfective past;

in

contrast, the nativ e speaker control group used both im perfective and

perfective for stative verbs, w hich shows that the learners' use of im per -

fective

morphology

is

less flexible than native speakers.

To summarize, results from the above studies provide converging

evidence on the association between lexical aspect and tense-aspect

markers in L2 acquisit ion of diverse languages. These results are hig hl y

consistent with

L I

acq uisition resu lts (see

3.2.1)

that children init ial ly

restrict uses of perfective-past forms to telic verbs and imperfective forms

to

atelic verbs.

5.3.2.

Aspec t Hypothesis nd contr sting views

The

above crosslinguistic

L2

data have

led

some researchers

to

formula te

th e Aspect Hyp othesis (Andersen and S hirai 1994; B ard ovi-H arlig

1995a; Robison 1995),

30

according to wh ich the follow ing genera liza-

tions can be made (Shirai 1991: 9-10; Bardovi-Harlig and Bergström

1996: 312; Andersen

and

Shirai 1996: 533).

Unauthenticated




50 Theories

of


(1) Learners use (perfective) past marking on achievement or

accomplishment verbs, eventually extending

use to

activity

and

state verbs.

(2) In

languages that encode

the perfective-imperfective

dis-

tinction

morphologically, imperfective past appears later

than perfective past, and imperfect past marking begins

with

stative

and

activity (i.e., atelic) verbs,

and

then extends

to

accomplishment

and

achievement (i.e., telic) verbs.

(3) In languages that have

progressive

aspect progressive

marking begins with activity verbs, and then extends to ac-

complishment/achievement

verbs.

(4) Progressive ma rking is rarely incorrectly overextended to

stative

verbs

(in

L I

acquisition).

These generalizations can be schematically represented as in Table

3.1.

The

numbers represent order

of

acquisition, from

the

earliest (i.e.,

1) to

th e

latest (i.e., 4).

Table 3.1. Predicted

order

of

development

of

tense-aspect morphology

State Ac tivity Accom plishment Achievement

(Perfective) Past 4 <===== 3 <======= 2 <======= 1

Progressive

? <===== 1 ========> 2 =======> 3

Imperfective 1 =====> 2 ========> 3

=======>

4

? combination rarely

occurs

The generalizations of the Aspect Hypothesis are intended to apply to

th e

acquisition

of

tense-aspect morphology

in

both

LI and L2.

However,

Generalization

(4 )

seems

to be

mainly applicable

to

LI

and is

problem-

atic for L2

acquisition. Rob ison's (1990) case study

of a

Spanish learner

of English shows

a

frequent

use of

progressive morp holog y w ith stative

verbs,

inclu ding incorrect uses (such

as liking,

-wanting,

having . Out of

the 43

progressive uses,

12

(28 ) were w ith stative verbs. Thu s there

appears to be a child-adul t difference in the use of progressive with

stative verbs, since children rarely,

if

ever, generalize progressive m ar kin g

to stative verbs, as discussed earlier (Brown 1973; Kuczaj 1978).

Despite

th e

generality

of the

Aspect Hypothesis

in

account ing

for L2

acquisition of tense-aspect morphology, several recent studies have

found counterexamples to the generalizations of the hypothesis. Berg-

ström (1995;

se e

also Bergström 1997;

Bardovi-Harlig and

Bergström

Unauthenticated




Second language

acquisition

1996) studied

th e

acquisition

of

French

by

classroom learners

and she

used both a cloze-type test and a written narrative task in which learners

recounted the story line of a Chaplin silent film. The learners showed a

lower marking of past perfective passe compose} on accompl ishment

verbs than on activity verbs. For ex ample, the au thor reported the fo l-

lowing resul ts on the use of perfect ive past marking by beginning

learners in a cloze-typ e test: 79.6 for achiev em ent, 60.3 for

accom-

plishm ent, 73.8 for activity, and 38.5 for states. Th us, activities

showed higher markings

with passe

compose than did accomplishments,

fo r

th e

lowest proficiency group. This

goes

against Andersen's (1989)

prediction that learners

will

tend to use perfective past marking m o r e

frequently with

accom plishment verbs than w ith activity verbs.

Also important is

Bergström's

observation that learners at the highest

proficiency level in her stud y showed the highest con gru ence with the

Aspect H ypothesis, tha t is, the highest correlation of past perfective form s

with

achievements. This result directly conflicts

with

predict ions

of the

Aspect Hypothesis, since the hypothesis would predict that less advanced

learners show more conspicuous effect of lexical aspect than advanced

learners.

31

Still,

Bergs tröm's findings

were consistent

with

one of the

predications of the Asp ect Hy pothesis, tha t is, the spread of imp erfectiv e

impar fait) from

stative to achievement.

Another set of counter-evidence to the Aspect Hypothesis comes from

Hasbun (1995) and

Salaberry

(1997, 1999). In their cross-sectional

studies

of classroom learners of Spanish, both authors observed a high

association of per fectiv e past w ith telic verbs, and im perfec tive past with

atelic verbs. However, they both

found

that this association was not

strong at the beginning level, but became stronger with increas ing

proficiency.

Hasbun used a written film-retell task, while Salaberry used

an

oral film-retell task

and a

cloze-type written test.

In conn ection with these disagreements,

researchers

have also debated

on whether learners ' early morphology codes tense or aspect . Andersen

(1991)

proposed that early m orpho logy encodes lexical aspect.

In

contrast, Klein, Dietrich,

and

No yau (1995) argued that learners' early

past tense forms code tense and not aspect, on the basis of their extensive

longi tudinal studies of n atural is t ic learners acqu iring various Eu rop ean

languages. Salaberry

(1997,

1999) also claimed that early past perfective

forms in

learners '

speech

code

tense, rather than aspect, and on ly later

on is

there

a

stronger correlation

of

perfective past

with

lexical aspectual

value. Rohde (1996) did not find a strong correlation of progressive

Unauthenticated




52 heories of language acquisition

aspect w i th activity verbs in his longi tudinal L 2 data from chi ldren

acquir ing

English,

and

suggested that tense

is

more important than aspect

fo r L 2 acquisition . Buc zow ska and Weist (1991) claimed that the com -

prehens ion

of

tense-aspect markers

by

Polish learners

of

English sup-

ports the notion that tense is acquired before aspect.

Thus,

it

appears that although there

are

general patterns

in the

acqui -

sition of tense-aspect m orph ology in L2, as the Aspect

Hypothesis

claims,

there

are

also important variat ions between f indings from differ-

ent

studies.

A n

important task

for us is to

account

for why these

varia-

t ions

arise.

As in studies of LI acquisition (see 3.2.3), disagreements in L2 acqui-

sition research also involve terminological

and

conceptual confus ion

concerning th e notion of aspect. A s noted earlier, th e term

aspect

can

refer to three

different

levels of aspect — situation, lexical aspect, an d

grammatical aspect. For exam ple, Housen (1993, 1994) attemp ted an

analysis

of L2

Dutch tense forms

and

three semantic notions: temporal

reference

(past or nonpast), aspect (perfective or

imperfective)

and

lexical aspect (punctual

or

durative;

Stative or

dynamic). Housen (1993)

claimed that

th e

learners' interlanguage

is

tense-prominent rather than

aspect prominent, on the grounds that the learners ' past tense forms were

associated with past time reference rather than wi th perfective aspect, and

that

th e

predicted association between tense forms

and

inherent aspect

w as

not very strong. Giacalone Ram at (1997) and Jordens

(1999)

also

cited Housen ' s (1993) stud y as evidence for a tense-prominent L 2

Dutch. However, as Andersen (1989) pointed out, the Aspect Hypothesis

primarily concerns

th e

level

of

lexical aspect,

not

grammatical

aspect.

Moreover, Housen (1993, 1994)

did not

make clear

h o w th e

perfect

ve-

imperfective

dis t inction

w as

coded

in his

analysis,

and

Housen (1994:

286) gave a strong disclaimer about the reliabili ty of this coding. Thus,

although Housen 's proposal that the interlanguag e w ill be tense-

prominent i f

both

L I and L2 are

tense-prominent

is an

interesting

proposal, i t seems that a clear distinction needs to be m ade betw een

grammatical aspect and lexical aspect in addressing whether an in ter lan-

guage system is tense-prominent or aspect-prominent.

Meisel (1987) also argued against the proposal that L2 past tense

mark ing i s primari ly aspectual. He suggested that there is no systematic

evidence that aspectual distinctions are prominent in L2 acquisition, even

though

in LI

acquis i t ion

a nd Creole

languages

th e

aspectual system

m ay

be grammaticized before the tense system. However, his cri ticism was

Unauthenticated




econd language

acquisition 53

also leveled against

th e

idea that

L2

tense morphology codes g rammat i -

cal

aspect (Meisel 1987: 220). Again, wha t

is

important

in L2

acquisition

of

the past tense morpho logy is lexical aspect and not gram m atical

aspect, as the Aspect Hypothesis assumes.

Evident ly,

as in the

case

of

LI acquisition, there

are

many issues that

require more systematic analyses

at

both

th e

level

of

description

and the

level of explanation. We will a t tempt to address them and present an

account that reconciles the conflicting

f indings

and arguments in chap-

ters that follow.

3 4

Summary

At

th e

beginning

of

this chapter

w e

outlined

tw o

contrasting approaches

to language acquisition: the nativist and the functional . W e discussed

their major arguments

and

implications

for

first

and

second language

acquisition, particularly, as exem plified in the acquisition of lexical

aspect

and

tense-aspect morphology.

W e

reviewed

th e

major empirical

findings

in the acquisition of

tense

and aspect, and clarified a number of

conceptual, methodological, and theoretical issues that are important to

this literature. W e took a close look at some of the nativist proposals,

especially the Langu age Bioprogram H ypothesis advocated by Bickerton

(1981, 1984),

and the

empirical data that this hypothesis

rests on. We

also examined

a

similar hypothesis,

th e

Basic Child Gramm ar prop osed

by Slobin

(1985),

in connection

with

the crosslinguistic evidence on

children s early use of tense-aspect morp hology . Finally, to the extent

that L2 acquisition mirrors or differs

from processes

in

L I

acquisition,

w e

reviewed

the

major empirical results

of a

number

of

studies, discussed

the Aspect Hypothesis (Andersen and Shirai

1994),

a nd examined

relevant theoretical issues in second language acquisition.

In

recent

years,

proposals that children organize their g ram m ar

according to semantic or syntactic contrasts that are innately specified

(e.g. , Language Bioprogram Hypothesis) or especially salient ahead of

linguistic experience (e.g., Basic Child Gram m ar) have come increas-

ingly under

fire

from researchers who stress yo un g child ren s ex trao rdi-

nary skills at detecting patterns in the linguistic input

(e.g.,

Behrens

1993a, 1993b;

Bowerman

1985, 1989, 1996; Choi and Bowerman

1991;

Snow 1977)

and at

extracting correlations

and

prototypes from these

patterns (L i 1990; Li and Bowerman 1998; Li, Burgess, and Lund 2 ;

Stephany 1981; Shirai 1994; Shirai and Andersen

1995).

In a very

Unauthenticated




54 heories

of


recent rethinking, even Slobin himself (1997, in press) has raised serious

questions about

his

earlier proposal that there

is a

universal

set of

seman-

tic notions that are

privileged

for mapping onto grammatical mor-

phemes. He now suggests that the gramm atical morphem es found in a

language

at any

particular time,

and the

meanings they express,

are

best

seen as the outcome of continuous diachronic processes of grammatici-

zation - processes shaped not by priori templates for either gram ma ti-

cal forms or for their possible meanings, but by various kinds of com-

munication pressure. Learning the meanings of grammatical morphemes

is, according to Slobin's reformulation, no different in kind

from

learn-

ing

the

meanings

of

content words: both belong

to the

general domain

of

concept formation. In line

with

these approaches to the problem of

how learners map form and m eaning , we believe that it is possible to

account for learners acquisition of lexical and gramm atical aspect

without recourse to innate or prelinguistic constructs of the kind in the

language

bio pro gram or Basic Ch ild Gram mar. Instead, it is possible to

account for the acquisition

process

by appealing to learners' analyses of

the distribution

of

tense-aspect markers

in the

speech they hear, perhaps

as operationalized through connectionist principles

(L i

1999, 2000a,

in

press-a;

Li and MacWhinney

1996;

Rumelhart and

McClelland 1986).

In

this perspective, the acquisition of the lexicon and m orpholog y is a

dynamic, emergent process in which lexical and morpho logical catego-

ries emerge out of the interaction of multiple sources of information

(Elman

et

al„ 1996;

Li,

2000b,

in

press-b; MacWhinney, 1999).

In the

next

few chapters, we present in-depth examinations of English, Chinese,

Japanese, and connectionist networks that advocate such a general

functional

proposal concerning the acquisition of aspect.

Chapter 3 Theories of language acquisition and the acquisition of aspect

Documents