Second Language Research 16,2 (2000); pp.

103–133

Missing Surface Inflection or

Impairment in second language
acquisition? Evidence from tense
and agreement
Philippe Prévost Université Laval and Lydia White McGill
University

In this article, two accounts of the variable use of inflection in adult second
language (L2) acquisition are examined. The Missing Surface Inflection
Hypothesis (MSIH) proposes that L2 learners have unconscious knowledge
of the functional projections and features underlying tense and agreement.
However, learners sometimes have a problem with realization of surface
morphology, such that they resort to non-finite forms (e.g. Haznedar and
Schwartz, 1997; Prévost and White, 1999). The Impaired Representation
Hypothesis (IRH) claims that L2 inflection is essentially impaired, due to
lack of functional categories, features or feature strength (e.g. Eubank,
1993/94; Meisel, 1997). These views make different predictions for adult L2
acquisition. Spontaneous production data from two adult learners of French
and two adult learners of German are examined. The data show that finite
forms do not occur in non-finite contexts, that learners exhibit syntactic
reflexes of finiteness and that inflected forms largely show accurate
agreement. These results suggest that adult L2 learners represent finiteness
and agreement at an abstract level, rather than being impaired in this
domain, supporting the MSIH.

I Introduction
It is well known that second language (L2) learners exhibit
optionality in their use of inflectional morphology, with tense and
agreement markings sometimes being present and sometimes
absent in L2 production data. This is true during the course of
acquisition, as shown, for example, by the research on morpheme
acquisition orders conducted during the 1970s (e.g. Dulay and Burt,
1974; Bailey et al., 1974; for relevant discussion, see Zobl and
Liceras, 1994). In some cases, variable use of inflectional
morphology is found even in the endstate grammar (Lardiere,
Address for correspondence: Philippe Prévost, Université Laval, Département de Langues,
Linguistique et Traduction, Pavillon de Koninck, #2284, Québec, PQ, G1K 7P4, Canada;
email: [Link]@[Link]
© Arnold 2000 0267-6583(00)SR164OA
104 Missing surface inflection

1998a; 1998b). While the fact of variation is uncontroversial, there

is relatively little agreement as to what such optionality implies and,
in particular, whether it indicates major impairment to the
interlanguage grammar (ILG) in the domain of functional
categories or whether it is the consequence of some more
superficial problem.
In this article, we examine variability in use of verbal inflection
in L2 French and German, considering the incidence of finite versus
non-finite morphology and the contexts in which each is found, as
well as verbal agreement. Following earlier work (Lardiere, 1998a;
1998b; Prévost and White, 1999), we argue that variability in adult
L2 performance does not reflect a deeper lack of functional
categories or features associated with tense and agreement. Rather,
L2 learners have difficulties with the overt realization of
morphology.

II Functional categories
The properties that we are interested in relate to verbal functional
projections. Verb (V) features have to be checked in Infl (I)
(Chomsky 1995). If V features are strong, the verb raises overtly to
I for feature-checking purposes. If V features are weak, the verb
raises covertly (after spell out). A number of syntactic consequences
result: in languages like French and German, the L2s being
considered here, V features are strong; finite verbs raise, resulting
in well-known alternations between the positions of finite and non-
finite verbs.
The tree in (1) illustrates V raising in French. The finite verb
raises to I, over negation or adverbs (as in (1a)), in contrast to the
non-finite verb, which remains in the VP (1b). Thus, finite verbs in
French appear to the left of negation or adverbs, whereas non-finite
verbs appear to the right (Pollock, 1989).
 Philippe Prévost and Lydia White 105

1) CP

Spec C'

C IP

Spec I'

(a) Jean I NegP

parti pas VP

ti
(b) Jean peut pas partir

German is V final, as well as I final on many analyses (den

Besten, 1983). In main clauses, the finite verb raises through I to
C, as illustrated in (2), while some other phrase raises to Spec CP,
resulting in the verb second (V2) effect. Nonfinite verbs, on the
other hand, remain in final position. As in French, then, there are
overt syntactic consequences of the +/– finiteness distinction, such
as V2 vs. V final, as well as the same alternation with respect to
negation as is found in French.

2) CP

Spec C'
Erj
C IP
hati
Spec I'
tj
VP I
ti
den Mann gesehen ti

While there is considerable agreement on the syntactic

consequences of feature strength, namely presence or absence of
verb raising, there is less consensus regarding the morphology. On
the one hand there are those who argue for a very close connection
between morphological paradigms and feature strength, such that
absence of a paradigm implies absence of the relevant functional
projections or feature strength (e.g., Clahsen et al., 1996). On the
other hand, there are those who argue that abstract featural
106 Missing surface inflection

properties cannot be closely associated with morphological

paradigms, given that there are languages which exhibit the
syntactic reflexes of strong features without having rich morphology
(Sprouse, 1998).

III Variability in L1 acquisition

One kind of variation in the incidence of verbal inflection which
has been much investigated in first language (L1) acquisition is the
Optional Infinitive (OI) (Wexler, 1994) or Root Infinitive (Rizzi,
1993/94) phenomenon, whereby children’s main clauses vary
between having a finite or a non-finite main verb. Although
accounts of what is going on differ considerably, there is general
agreement that OIs are structurally determined: when the child uses
non-finite forms in place of finite ones, these are indeed non-finite
and are found in structural positions typical of non-finite verbs. For
example, the child’s non-finite verbs do not raise past negation in
French and are not found in the V2 position in German. When the
child uses finite forms, on the other hand, these are found in
structural positions typical of finite verbs. Variation in the use of
finiteness, then, is not random.
In earlier work, Wexler (1994) argued that the OI stage can be
accounted for on the assumption that Tense is underspecified in
early child grammars; Hoekstra et al. (1997) argue that it is Number
that is underspecified. On these accounts, functional categories in
the grammars of L1 acquirers are represented in a manner that
differs from the adult grammar. Rizzi (1994) and Haegeman (1995)
offer a somewhat different analysis, involving truncation. They
argue that early grammars possess the same set of functional
categories as adult systems, and that none of them are
underspecified. However, there is no restriction on what projection
can serve as the root of a matrix declarative. The child’s structure
may be truncated at any point below CP, such that root VPs, root
IPs, and root NegPs are possible, in addition to CPs. Variability,
then, is the consequence of different roots being projected for
different utterances. Nevertheless, incidence of finiteness is
structurally constrained: root VPs are non-finite, whereas other
roots are finite. Both the underspecification theory and the
truncation theory assume that non-finite forms are indeed non-
finite in L1 acquisition, showing the expected syntactic properties
of non-finite verbs with respect to verb placement. The issue is not
that the child does not know the difference between finite and non-
finite forms; rather, optionality is the result of the child’s not
knowing that main clauses must be finite (on the OI account) or
 Philippe Prévost and Lydia White 107

that they must be CPs (on the truncation account).

A number of researchers have suggested that optionality in
incidence of finite forms in L1 acquisition reflects a problem with
realization of morphology, rather than any problem with functional
categories or features as such (Borer and Rohrbacher, 1997; Boser
et al., 1992). On this approach, clauses containing non-finite lexical
verbs are assumed to be finite (contra Wexler and Rizzi), containing
a finite auxiliary which is null.1 Again, non-finite forms occur in the
positions expected of non-finite verbs; in other words, they do not
raise.
The above accounts concentrate on the issue of finiteness.
However, they also have something to say about the realization of
agreement morphology. Borer and Rohrbacher (1997: 26) argue
that abstract functional projections can exist in the grammar in the
absence of overt morphology and propose that ‘the absence of overt
agreement markers . . . appears to be motivated by the desire to
avoid incorrect forms whose morpho-phonology has not been fully
acquired’. Poeppel and Wexler (1993) show that, while agreement
morphology is often lacking in early L1 acquisition of German,
when agreement is present it is accurate; for example, if a verb
occurs with 3rd person singular morphology, the subject will be 3rd
person singular. (As they point out, the reverse contingency does
not hold, since a 3rd person singular subject may also be found with
a verb form not marked for agreement.)
To summarize, while disagreeing over whether or not child
grammars allow non-finite main clauses, the above accounts of
variability in L1 acquisition agree that:
1) in the case of tense morphology, there may be variation between
finite and non-finite forms but if a verb form is finite, it has
finite properties, whereas if a form is non-finite, it has non-finite
properties;
2) in the case of agreement morphology, variation is between forms
being present and appropriate, or being absent; ‘faulty’
agreement is argued not to be found (e.g. 3rd person
morphology with 2nd person subject).
We turn now to a consideration of variability in L2 acquisition.

IV Variability in L2 acquisition
Superficially, at least, L2 acquisition presents a similar picture, in
that incidence of tense and agreement morphology is variable. At
1
Radford (1992) also argues for a null finite Aux, once functional categories have emerged
in the child grammar.
108 Missing surface inflection

issue is whether the variability takes the same form as in L1

acquisition or whether it is of a different nature, and, if so, whether
this reflects some form of permanent ‘impairment’ to the ILG.
Broadly speaking, there are two kinds of accounts of variable use
of inflectional morphology in L2. On the one hand, there are
researchers who propose that ILGs contain abstract functional
categories and features, and that they show syntactic consequences
of these properties, such as verb raising, case-assignment, presence
or absence of null subjects, etc. Variability in morphology is argued
to reflect difficulties in identifying the appropriate morphological
realization of functional categories; that is, the problem lies in
mapping from abstract features to their surface morphological
manifestation (Epstein et al., 1996; Grondin and White, 1996;
Haznedar and Schwartz, 1997; Lardiere and Schwartz, 1997;
Lardiere, 1998a, 1998b, 2000; Prévost and White, 1999). On the other
hand are those who argue that L2 grammars suffer from some form
of impairment in the functional domain, resulting in variability,
either temporarily or permanently (Meisel, 1991, Eubank, 1993/94;
Beck, 1998; Eubank et al., 1997; Meisel,1997; Beck, 1998).

1 Missing surface inflection

Following Haznedar and Schwartz (1997), we refer to the first
approach as the Missing Inflection Hypothesis, emending this term
to Missing Surface Inflection Hypothesis (MSIH) to make it clear
that it is at the surface morphological level that inflection is
assumed to be absent, rather than at the abstract featural level. This
position has been advanced independently by a number of
researchers; some have simply suggested that it must be the case
without pursuing the issue in detail (e.g., Epstein et al., 1996;
Grondin and White, 1996). Recently, researchers have begun to
investigate the properties of missing surface inflection in some
detail (Haznedar and Schwartz, 1997; Lardiere, 1998a; 1998b;
Martohardjono et al., 2000; Prévost and White, 1999).
Arguing against an OI phenomenon in L2 acquisition, Haznedar
and Schwartz (1997) looked at data from a Turkish child learning
English as L2. While they found evidence of use of both finite and
non-finite verb forms, they found little evidence of other properties
that are normally associated with the OI stage, such as incidence of
null subjects and accusative case in place of nominative. They also
found that agreement was accurate when it occurred (as was also
observed by Grondin and White (1996) for child L2 French).
Haznedar and Schwartz argue that there is no evidence for a
syntactic deficit, such as underspecification of Tense, in their
 Philippe Prévost and Lydia White 109

subject’s English; rather, the child’s non-finite morphology was

indicative of missing inflection. They speculate as to whether the
same phenomenon might be found in adult L2 acquisition.
Lardiere (1998a; 1998b) examined the L2 English of an adult
Chinese-speaker, Patty, who had resided in the USA for 18 years
and whose grammar is clearly at its end state. Lardiere shows that
incidence of tense morphology in spontaneous production is at
about 35%, while 3rd person singular agreement is less than 17%
(excluding the copula and auxiliary be). In spite of this low level of
overt morphology, Patty shows a variety of syntactic phenomena
which suggest that she has tense and agreement, with appropriate
weak values, at an abstract level. For example, Patty shows 100%
correct incidence of nominative case assignment and complete
knowledge of the fact that English verbs do not raise, suggesting
that her problems are in the domain of morphological mapping
rather than being due to any deficit in functional features as such.
Prévost (1997) and Prévost and White (1999) looked for evidence
of truncation in L2 acquisition but found this only in the case of
child L2 learners. The behaviour of adults was more consistent with
missing surface inflection. To summarize their findings, co-
occurrence of finite and non-finite main verbs was found in both
the child and the adult data, suggesting variability. In the child L2
data, non-finite forms occurred in different contexts from finite
forms; the distribution of non-finite verbs was structurally
determined and consistent with truncation. The adult data, on the
other hand, showed finite and non-finite verbs occurring in the
same contexts (in CPs, with subject clitics, preceding negative
markers, etc.). Many more (apparently) non-finite verbs were found
in the adult corpora and the adults produced these forms
throughout the period of observation.
All the accounts of missing L2 inflection described above suggest
that non-finite or bare stems are being used as substitutes for finite
forms. In such cases, then, main verbs bearing the infinitival marker
are not in fact non-finite but are being used as default forms
exhibiting properties of finite verbs. This assumption differs
crucially from most accounts of variation in L1 acquisition, where
non-finite forms are argued to be truly non-finite.2

2 Impairment
The accounts assuming missing surface inflection contrast with
views that argue for some kind of impairment to the ILG, either
2
But see Phillips (1996) for a different view: he assumes that non-finite main verbs in L1
acquisition are in fact finite but that they do not raise.
110 Missing surface inflection

globally or locally, at the level of functional categories or features.

We will refer to this position as the Impaired Representation
Hypothesis (IRH). Meisel (1991; 1997) effectively assumes global
impairment. He argues that L2 grammars lack agreement (on the
grounds that infinitives and bare stems are common in place of
forms appropriately inflected for agreement) and that finiteness
and verb-placement are unrelated (see also Clahsen, 1988). Meisel
(1997) found that non-finite forms occur in finite positions, in
particular in the V2 position in L2 German. However, if finiteness
and verb placement are truly unrelated, we would expect not only
to find non-finite forms in finite positions but also finite forms in
non-finite positions. Meisel argues that this is indeed the case for
two of the three adults that he studied; in their L2 German
negatives, the negator appears to the left of finite verbs and modals
in some cases. He concludes that adult L2 acquisition is
substantially different from child L1 acquisition. In particular,
Universal Grammar (UG) is no longer involved; rather the adult
proceeds in terms of linear sequencing. By implication, then,
functional projections like IP do not exist in the ILG and abstract
agreement and finiteness distinctions are not involved in L2
acquisition.
A more local kind of impairment is proposed by Beck (1998),
Eubank (1993/94) and Eubank et al. (1997), who argue that feature
strength (at least with respect to finiteness) is essentially impaired
or inert in the ILG. This was originally assumed by Eubank to be
a temporary phase but is now argued to be permanent (Beck, 1998).
Optional verb placement is the consequence of inert feature
strength: if a feature is neither strong nor weak, a verb can appear
in either a raised or an unraised position.
What is not so clear is why Beck and Eubank assume that a verb
will be finite when it raises and nonfinite when it does not. Absence
of functional projections, features or feature-checking mechanisms
in fact implies randomness or mismatches in the use of inflectional
morphology (Borer and Rohrbacher, 1997; Lardiere, 2000). As
Borer and Rohrbacher (1997: 28) point out:

Without AgrSP, the syntactic appropriateness of fully inflected forms cannot

be checked, nor can syntactically inappropriate forms be excluded. A grammar
without AgrSP should therefore allow well-formed morphological forms to be
distributed randomly, as the mechanism designed to check their
appropriateness in a specific syntactic context is non-existent. A similar
rationale predicts that without a TP, children should make tense errors, failing
to check base-generated tensed forms against the properties of T . . . randomly
distributed inflected forms should abound.
 Philippe Prévost and Lydia White 111

We believe that the same rationale applies if TP and AgrP are

present but features or feature values are not, that is, if features are
impaired or inert as to strength. In such a case, there would be
nothing to prevent featural mismatches. As far as we can see, this
indeed implies that there should be no contingency between
finiteness of the verb and verb placement. Both morphological and
syntactic variability should be found; in other words, incidence of
finite/non-finite morphology will not be structurally determined.

V Predictions
A number of questions remain unanswered by the research
conducted so far. One issue is whether or not variation in L2
morphology is unconstrained, as might be expected on the views
that assume impairment, with breakdown in functional categories
or features, hence in feature-checking mechanisms. In other words,
is it the case that finite and non-finite forms alternate freely, while
agreement morphology is often inaccurate? Here, we consider
predictions relevant to finiteness and to agreement.
As far as finiteness is concerned, on the MSIH, finite forms are
assumed to be genuinely finite, hence they should occur in finite
positions but not in non-finite ones. Nonfinite forms, on the other
hand, are sometimes truly non-finite and at other times occur as
substitutes for finite inflection, hence their position is more
variable. Thus, we predict that finite verbs will be found only in
finite contexts. In particular, if a verb is finite, it should not be used
after a preposition or another verb such as an auxiliary or modal
in the same clause. Only non-finite forms should be used in such
contexts. Moreover, finite forms should appear before negation in
L2 French or German and should not be found following a negator.
The IRH, on the other hand, predicts no clear contingencies
between finiteness and verb position: if features or feature strength
are lacking, finite verbs should be found in both raised and unraised
positions. Thus, the two hypotheses differ as to their predictions for
finite verb forms. As far as non-finite forms are concerned, both
hypotheses make the same prediction, though for different reasons,
namely that they will occur in both finite and non-finite contexts.
As for agreement, where finite forms are used, the MSIH predicts
agreement will be appropriate, since the relevant features and
feature-checking mechanisms are assumed to be present (cf.
Poeppel and Wexler, 1993; Borer and Rohrbacher, 1997). At least
the strong form of the IRH (Meisel, 1991; 1997) predicts agreement
mismatches, since the mechanisms necessary for checking
agreement features are impaired.
112 Missing surface inflection

VI Current study
In the present article we further examine spontaneous production
data obtained longitudinally from four adults learning L2 French
and L2 German in naturalistic environments (for background
information see Table 1). The two learners of L2 French –
Abdelmalek and Zahra – were native speakers of Moroccan Arabic.
They were interviewed roughly once a month for a little over three
years by the European Science Foundation Project on L2
acquisition by Adult Immigrants (Perdue, 1984; 1993). Both learners
had immigrated to France from Morocco, where they had had no
previous exposure to French. Their first interviews took place after
one year in France, during which time they had little contact with
French. At that point, their L2 proficiency was judged to be very
limited. As for the two learners of German – Ana and Zita – they
were native speakers of Spanish and Portuguese respectively. They
were immigrants to Germany and were interviewed by the ZISA
project (Clahsen et al., 1983). They were first recorded three months
after their arrival in Germany. While neither of them had been
exposed to German in their native countries, once in Germany Ana
had much more exposure to the language than Zita. Ana also had
started taking classes in German one month prior to the first
recording session. Both learners were interviewed on average once
a month for a little less than two years. (Note that there is a break
in Ana’s data between months 14 and 23 as she returned to Spain
for 9 months. Recording resumed when she came back to
Germany.)
Utterances containing a verb and consisting of at least two
constituents were retained for analysis. Items clearly identified as
repetitions and formulaic expressions were discarded. Potentially
problematic is the fact that both French and German inflectional
paradigms contain homophonous morphemes. As can be seen in
Table 2, singular forms in French are homophonous in each of the
three verb-classes in the present tense.3 Distinctive forms are only
found in the plural (but 3rd person plural is homophonous with the

Table 1 Learner and interview details

Abdelmalek Zahra Ana Zita
L1 Arabic Arabic Spanish Portuguese
L2 French French German German
Age at onset adult 34 22 17
Number of files 25 25 17 25

3
We gloss forms which are ambiguous because of this homophony as 1/2/3S.
 Philippe Prévost and Lydia White 113

Table 2 French inflectional paradigm (present tense)

1st group 2nd group 3rd group
chanter (‘sing’) finir (‘finish’) vendre
(‘sell’)
Infinitival form chant-er ([e]) fin-ir vend-re
Past participle chant-é ([e]) fin-i ([i]) vend-u
Present tense 1/2/3 singular chant-e/-es/-e fin-is/-is/-it ([i]) vend-s/-s/-ø
1/2/3 plural chant-ons/-ez/- fin-issons/-issez/- vend-ons/-
ent issent ez/-ent

Table 3 German inflectional paradigm (present tense)

lernen (‘learn’)
Infinitival form lern-en
Past participle ge-lern-t
Present tense 1/2/3 singular lern-e/-st/-t
1/2/3 plural lern-en/-t/-en

singular in the 1st group, -ent being silent). In the case of the 1st
group, the 2nd person plural marker -ez is homophonous with the
infinitival suffix -er and past participle marker -é. In the case of the
2nd group, singular forms ending in [i] are homophonous with the
corresponding past participle. In our investigation, a verb form was
considered non-finite in the absence of clear evidence to the
contrary.
German, on the other hand, has more distinctive forms in the
singular than in the plural (see Table 3). The 1st and 3rd person
plural markers are homophonous with each other and with the
infinitival marker -en.

VII Results
The results reported in this article concern both root and non-root
contexts, including matrix declaratives, embedded clauses and
questions. We report total number of occurrences of the forms we
are interested in, collapsed across all interviews, our rationale being
as follows. According to the IRH, functional categories or features
are permanently impaired. Therefore, we should find evidence for
variable placement of finite verbs, as well as agreement problems,
throughout the data, including in the later interviews. The opposite
should obtain for the MSIH: since features are assumed to be
available in the initial state of L2 acquisition (either from the L1
114 Missing surface inflection

or from UG),4 finite verbs should not be randomly positioned in

any of the files being examined, and agreement should be accurate
when it occurs. Finite and non-finite verb forms are indeed found
in almost all the files examined (Prévost, 1997; Prévost and White,
1999). It is not the case, for example, that non-finite forms occur
solely in the earliest interviews and that finite forms emerge later
on. (We return to the issue of development in the discussion.)

1 Finite forms in non-finite contexts

a Verbs occurring after auxiliaries, prepositions and other verbs:
Table 4 summarizes the incidence of finite and non-finite verb
forms in clearly non-finite contexts, namely following a preposition,
an auxiliary or another verb, such as a modal.5 Considering the data
as a whole, the incidence of finite forms in these environments is
very low, the vast majority of verbs appropriately being non-finite.
In addition, there is no period in the data, not even in the earlier
files, during which finite forms are found in non-finite contexts or
are randomly distributed in finite and non-finite environments.
Examples of appropriate use of non-finite forms in non-finite
contexts are given in (3) through (5). In some cases, the past
participle is used instead of the infinitival form (6) and vice versa

Table 4 Finiteness of verbs following auxiliaries, prepositions and other verbs

L2 French L2 German

Abdelmalek Zahra Ana Zita

–fin +fin –fin +fin –fin +fin –fin +fin

Aux + V 180 3 98 0 12 2 32 1
Prep + V 28 3 36 1 22 1 3 0
V+V 26 8 17 1 30 2 35 1
Total 234 14 151 2 64 5 70 2
(5.6%) (1.3%) (7.2%) (2.8%)

4
Proponents of the MSIH include those advocating Full Transfer/Full Access (e.g. Haznedar
and Schwartz, 1997) as well as those advocating Full Access to UG without transfer (Epstein
et al., 1996). These views agree that the early IL grammar is not impaired or underspecified
with respect to functional categories or features, even though they disagree on the source of
the initial representation of these categories. For the present purposes, this issue can be left
open.
5
In German root clauses, non-finite verbs used with auxiliaries and modals appear in final
position. Here, we will not be concerned with the acquisition of target verb-placement: the
focus of the study is whether or not non-finite verbs bear the infinitival marker, not whether
they are correctly positioned in the clause. The predictions made by the MSIH and the IRH
with respect to morphology are not affected by the issue of verb-placement in German. See
also note 7.
 Philippe Prévost and Lydia White 115

(7). Nonfinite verb forms are also found as complements of

adjectives, with the preposition omitted (8).6
3) a. il est parti l’Espagne, lui (Abdelmalek, month 14)
he is gone-PP the Spain him
b. Malika est sortie (Zahra, month 23.7)
Malika is gone+out-PP
c. ich habe gekommen drei mal in Wohnung (Ana, month 11.7)
I have come-PP three times in house
d. meine Schwester hat nich gesehen (Zita, month 25.6)
my sister has not seen-PP

4) a. pour voir (Abdelmalek, month 15)

for see-INF
b. sans déclarer (Zahra, month 20)
without declare-INF
c. für nehmen (Ana, month 4)
for take-INF
b. ich hab zu tun (Zita, month 25.8)
I have to do-INF

5) a. je te fais manger (Abdelmalek, month 17.7)

I you make eat-INF
b. je veux partir (Zahra, month 21.7)
I want leave-INF
c. er muß zuhause Übungen machen (Ana, month 13.5)
he must at+home practice do-INF
d. ich weiß nich machen (Zita, month 11.7)
I know not do-INF

6) a. il faut parti (Abdelmalek, month 27.7)

it must leave-PP
‘We must leave.’
b. je peux servi (Zahra, month 17)
I can serve-PP

7) a. il a faire (Zahra, month 21.7)

he has do-INF
b. ich habe machen ein Jahre mehr (Ana, month 11.7)
I have do-INF one year more
c. mein Schwager is schon gehen (Zita, month 24.4)
my brother-in-law is already go-INF

8) a. c’est difficile chercher un patron (Abdelmalek, month 20.5)

it is difficult look+for-INF a boss
b. c’est pas bon manger à la cuisine (Zahra, month 23.7)
it is not good eat-INF in the kitchen
c. für mich is es egal fahren nach alte Stadt (Ana, month 13.5)
for me is it equal drive+INF to old city
6
The following abbreviations will be used: INF = infinitive; PP = past participle; S = singular;
P = plural; 1 = 1st person; 2 = 2nd person; 3 = 3rd person; SUBJ = subjunctive; MASC =
masculine; FEM = feminine.
116 Missing surface inflection

Many of Abdelmalek’s errors involving finite verbs in non-finite

contexts occurred with the expression of obligation, ‘il (expletive)
faut + V’, as illustrated in (9). Here, the second verb should be an
infinitive but Abdelmalek produces a finite verb.

9) a. il faut marche (Abdelmalek, month 36.7)

it must walk-1/2/3S
‘We must walk.’
b. il faut paye(s) (Abdelmalek, month 38.7)
it must pay-1/2/3S
‘You must pay.’

The verb falloir in fact subcategorizes for either a non-finite CP

(10a) or a finite verb in the subjunctive (10b). Abdelmalek very
often omits complementizers in the finite case, producing utterances
such as (11a). He also uses a high number of subjectless CPs
throughout his interviews, as illustrated in (11b). Thus, it might be
the case that the finite verbs following the expression il faut in (9)
involve a finite CP with both the complementizer and the subject
missing: il faut [CP Ø [IP Ø marche]].

10) a. il faut partir

it must go-INF
‘We have to leave.’
b. il faut que nous partions
it must that we leave-SUBJ+1P
‘We have to leave.’

11) a. il faut ø tu partes (Abdelmalek, month 27)

it must you go-SUBJ+1/2/3S
‘You have to leave.’
b. parce que ø peut pas donner (Abdelmalek, month 17.7)
because can-1/2/3S not give-INF

As for L2 German, most finite verbs which incorrectly appear in

non-finite contexts either bear the 1st person singular -e marker
(12) or appear as bare stems in the case of Ana (13). The verbs may
in fact be non-finite in both cases (for L2 acquisition, see Meisel,
1991; 1997; for L1 acquisition, see Boser et al., 1992). In particular,
Meisel points out that -e forms can be regional variants of verbs
ending with the infinitival marker -en. (For further discussion see
the section ‘Verbal agreement’ on page 120.)

12) a. für lerne Deutsch is nicht gut (Ana, month 13.5)

to learn-1S German is not good
b. du willst nich arbeite hier (Zita, month 24.4)
you want not work-1S here
 Philippe Prévost and Lydia White 117

13) ich will diese Jahre fahr nach Spanien (Ana, month 13.5)
I want this year drive-Ø to Spain
mit mein Eltern
with my parents

b Verb placement and negation: A standard test for finiteness in

languages like French and German involves verb placement with
respect to the negative. Finite verbs have strong features and raise
over negation, whereas non-finite verbs do not; see (1) above.
According to the IRH, feature strength determining verb-
placement is impaired. In consequence, finite and non-finite forms
should be found on either side of the negative pas in French or
nicht in German. In contrast, the MSIH holds that abstract feature
strength is unimpaired; hence, finite verbs should systematically
precede the negator in L2 French and L2 German.7 The results
confirm this prediction, as shown in Table 5. The systematic
placement of finite verbs before the negator suggests that verb-
movement has occurred and, hence, that the features and feature
strength involved in raising (and checking) are known to the
learners. In addition, correct verb-placement as regards finite verbs
is observed from the very first instances of negative clauses, which
leads us to conclude that appropriate feature strength was posited
early on. Examples of correct negative placement are presented in
(14) and (15).
14) a. mais on peut pas dormir (Abdelmalek, month 17.7)
but one can-1/2/3S not sleep-INF
b. i mange pas (Zahra, month 26.7)
he eat-1/2/3S not

15) a. ich studiere nicht (Zita, month 3.7)

I study-1S not
b. ich spreche nicht Deutsch (Ana, month 4.5)
I speak-1S not German

Table 5 Verb placement with respect to negation

L2 French L2 German

Abdelmalek Zahra Ana Zita

V-Neg Neg-V V-Neg Neg-V V-Neg Neg-V V-Neg Neg-V

Finite 90 3 135 0 82 2 74 4
Non-finite 6 44 7 5 9 12 13 29

7
In German embedded clauses, the negator precedes the clause-final finite verb. However,
the L2 German data reviewed here suggest that the two learners of German (whose L1s are
SVO) have not acquired the V-final property of German (see Prévost, 1997). Negative
embedded clauses are thus treated on a par with negative root clauses in which the finite
verb raises above negation.
118 Missing surface inflection

As for non-finite forms, Table 5 reveals that there is variability

in verb placement. With the exception of Abdelmalek, whose non-
finite forms occur almost exclusively after the negator, the other
subjects show non-finite verb forms in raised and unraised
positions.8 (Similar variability is reported by Müller (1998) in a case
study of Bruno, an adult learner of L2 German.) We assume that
non-finite forms occurring after the negative are genuinely non-
finite: verb movement does not take place, hence the verb appears
after the negator, giving Neg V order. However, when non-finite
forms are used as substitutes for finite ones, verb-raising occurs,
yielding the order V Neg.

2 Overuse of finiteness and non-finiteness compared

In Prévost and White (1999), we report incidence of non-finite verb
forms in a number of finite contexts. Table 6 summarizes those
results from three contexts involving the projection of Infl:
1) in CPs, such as questions and subordinate clauses, the projection
of C implies Infl;
2) DP subjects must appear in Spec of IP for case assign-
ment/checking reasons;
3) clitic subjects typically attach to the verb in Infl.
If Infl is projected, verb-movement will occur for feature-checking
purposes, in which case the verb should be finite and should bear
Table 6 Incidence of non-finite verbs in finite contexts (as a percentage of all
verbs in finite contexts)
L2 French L2 German

Abdelmalek Zahra Ana Zita

CPs with non-finite 54/146 (37%) 65/222 (29.3%) 13/137 (9.5%) 18/82 (22%)
verbs
DP subjects with 15/68 (22%) 32/103 (31.1%) 23/206 (11.2%) 54/250 (21.6%)
root infinitives
Subject clitics with 168/700 (24%) 120/512 (23.4%)
root infinitives
Total 237/914 217/837 36/343 72/332
(25.9%) (25.9%) (10.5%) (21.7%)

8
The figures presented in Table 5 differ slightly from what is reported in Prévost and White
(1999) (their Table 7). This is due to the fact that in the present study we looked at both
root and non-root contexts while the negation data in Prévost and White (1999) were
restricted to root declaratives. Here, we also consider negative clauses involving a complex
tense (e.g. Aux-Neg-V–fin) as instances of both V+fin-Neg and Neg-V–fin. Negative clauses
whose main verb was a bare stem other than a modal or a first person singular form in L2
German were discarded.
 Philippe Prévost and Lydia White 119

Table 7 Overuse of non-finiteness vs. overuse of finiteness

Obligatory finite Obligatory non-finite
contexts contexts

+finite –finite –finite +finite

L2 French Abdelmalek 767 243 (24.1%) 278 17 (5.8%)
Zahra 755 224 (22.9%) 156 2 (1.3%)
L2 German Ana 389 45 (10.4%) 76 7 (8.4%)
Zita 434 85 (16.4%) 98 6 (5.8%)

inflectional morphology. In the data we examined, however, the

verb form was sometimes non-finite, as can be seen in Table 6.
If non-finite forms occurring in finite positions are in fact default
substitutes for finite forms, then one expects the overuse of non-
finite forms to be higher than the overuse of finite forms. Table 7
compares the incidence of non-finite forms in finite positions (from
Tables 5 and 6) to incidence of finite forms in non-finite positions
(from Tables 4 and 5). As Table 7 demonstrates, the incidence of
non-finite verbs in finite contexts is much higher than the incidence
of finite verbs in non-finite contexts for 3 of the 4 subjects. In other
words, while non-finite forms occur in both non-finite and finite
positions, finite forms are largely restricted to finite (raised)
positions. The contingency between finiteness and context is
significant for all learners but Ana, who was the most advanced
(Abdelmalek: χ2 = 47.91, p < .0001; Zahra: χ2 = 39.909, p < .0001;
Zita: χ2 = 7.817, p < .01; Ana: χ2 = .288, p > .05). Such a contingency
is problematic for the IRH, since positioning of verbal forms should
not depend on finiteness if feature strength is impaired. In contrast,
results are consistent with the MSIH, which assumes that syntactic
consequences of finiteness are known to learners. As a result, finite
verb forms are associated with finite features and appear in raised
positions. Nonfinite forms appear correctly in non-finite contexts
and also as a default in finite positions.

3 Accuracy on agreement
The MSIH predicts that when agreement is found, it should be used
correctly since agreement features and feature-checking
mechanisms are assumed to be unimpaired. The IRH (at least,
Meisel’s version) makes the opposite prediction: erroneous
agreement should be found, since checking cannot take place and
there is no way of excluding agreement mismatches. Here we
consider two kinds of agreement: verbal agreement in French and
German and clitic agreement in French clitic-doubling
120 Missing surface inflection

constructions, on the assumption that subject clitics are agreement

markers which appear in Infl (Auger, 1994).

a Verbal agreement: We investigated knowledge of agreement by

identifying inflected verbs. In each case, we looked at person and
number inflection on the verb and then at whether the subject of
that verb was appropriate, given the inflectional morphology. In
other words, following Poeppel and Wexler (1993), we look at
whether agreement is accurate when it is present and not at
whether a subject is found with an agreeing verb form (contra
Meisel, 1991). In addition to the regular inflectional paradigms
presented in Tables 2 and 3, both languages include verbs with
irregular paradigms involving suppletion (e.g. être (‘be’), avoir
(‘have’), aller (‘go’) in French and sein (‘be’) in German).
Suppletive forms have traditionally been excluded from the
investigation of knowledge of agreement and from the study of the
relationship between morphological and syntactic knowledge. Yet,
as Lardiere (2000) reminds us, suppletive forms are endowed with
features that need to be checked. They are thus indicative of
knowledge of agreement and it is appropriate to examine accuracy
in production of suppletive forms. (Furthermore, suppletive forms
avoid the problem of homophony which arises with regular verbs,
since almost all persons are realized with a distinct form.)
As can be seen in Table 8, agreement is largely correct in the L2
French data. Accurate use of inflectional marking is around 95%.9
While results on regular affixation are difficult to interpret, since
many agreement markers are homophonous, it is noteworthy that

Table 8 Accuracy in use of verbal agreement (L2 French)

Abdelmalek Zahra

Correct Incorrect Correct Incorrect

Regular affixation 447 (94.7%) 25 552 (93.4%) 39
homophonous inflection 445 (94.7%) 25 538 (93.2%) 39
distinctive inflection 2 (100%) 0 14 (100%) 0
être, avoir and aller forms 264 (97.8%) 6 156 (98.7%) 2
Total 711 (95.8%) 31 708 (94.5%) 41

9
Some inflected forms involving the masculine 3rd person plural clitic ils (‘they’) were
excluded from analysis since the clitic is homophonous with the corresponding singular form
il (‘him’). Thus, there is no way to tell whether the verb form is appropriate or not, as shown
in (i), where the verb is 3rd person singular and the clitic is ambiguous between singular and
plural. The NP that the clitic doubles is plural.
i) les autres i est parti (Abdelmalek, month 34.5)
the others they/he(?) is-3S gone
 Philippe Prévost and Lydia White 121

accuracy on the distinct suppletive forms of être, avoir and aller is

almost 98%.
Examples of appropriate person/number agreement are shown in
(16). Errors found in the regular paradigm mostly involved verbs
of the 3rd group, pronounced incorrectly with a final [t] or [d], as
in (17a). All Abdelmalek’s errors and 90% of Zahra’s errors were
of this type. In French, a root-final [t] or [d] is pronounced in the
singular form verbs of the 1st group in the present tense (see Table
2). (For example, je/tu/il chante(s) ([Sãt] ‘I/you/he sing(s)’) and
je/tu/il garde(s) ([ga d] ‘I/you/he keep(s)’.) It is thus possible that
the learners were overgeneralizing this regular inflection pattern to
verbs of the 3rd group like partir (‘go’) and vendre (‘sell’) where
root-final [t] or [d] are normally silent. As for the few incorrect
suppletive forms, most of them are person errors – as shown in
(17b) and (17c) – involving overgeneralization of 3rd person
singular forms, such as a (‘has’) and est (‘is’).

16) a. il donne une bière (Abdelmalek, month 14)

he give-1/2/3S a beer
b. moi j’aime pas de monde (Abdelmalek, month 25)
me I like-1/2/3S not some people
c. elle (l’) a donné le prof (Zahra, month 12)
she (it) have-3S given the teacher
d. quand le matin ils sortent le travail (Zahra, month 29.2)
when the morning they leave-3P10 the work

17) a. moi parte (Zahra, month 15.5)

me go-1/2/3S
b. j’a fait la bagarre avec lui-même (Abdelmalek,
I have-3S made the fight with him month 52.7)
c. quand moi est pas là (Zahra, month 39.5)
when me is-3S not there

In L2 German, accuracy is also very high, as can be seen in Table

9 (for similar observations, see also Clahsen, 1988). This is
particularly obvious as far as suppletive forms are concerned where
accuracy is comparable to that in L2 French. Results on regular
affixation were expected to be somewhat more revealing in L2
German since there are more distinct morphological forms in
German than in French. However, what we found was relatively
mixed: accuracy is very high on 3rd person singular -t for both Zita
and Ana; Ana shows total accuracy on 2nd person singular -st, while
Zita is somewhat less accurate; both learners show lowest accuracy

10
Given the fact that these learners pronounce the final consonant in singular forms of verbs
of the 3rd group, this form could be 1/2/3S, even though this would not be the case for a
native speaker of French.
122 Missing surface inflection

Table 9 Accuracy in use of verbal agreement (L2 German)

Ana Zita

Correct Incorrect Correct Incorrect

Regular affixation 1S (-e) 155 (74.5%) 53 70 (70.7%) 29
2S (-st) 17 (100%) 0 33 (80.5%) 8
3S (-t) 46 (95.8%) 2 90 (90.9%) 9
Total 218 (79.8%) 55 193 (80.7%) 46
sein forms 1S (bin) 19 (100%) 0 12 (100%) 0
2S (bist ) 0 0 2 (100%) 0
3S (ist) 230 (93.9%) 15 111 (97.4%) 3
3S-Past (war ) 5 (100%) 0 0 0
1P (sind ) 16 (100%) 0 0 0
3P (sind ) 28 (87.5%) 4 2 (100%) 0
Total 298 (94%) 19 127 (97.7%) 3
kann/will/muß stems 1/3S 45 (95.7%) 2 48 (96%) 2
Total 561 (88.1%) 76 368 (87.8%) 51

(albeit higher than chance) on -e. Finally, accuracy is very high on

modals such as können (‘can’), müssen (‘must’) and wollen (‘want’)
which do not bear suffixes in the 1st and 3rd person singular.
(Müller (1998) reports that Bruno also shows high accuracy of
agreement with -t and -st and lower with -e.)
Examples of correct agreement are given in (18) and incorrect
agreement in (19).

18) a. er sagt der Herr (Ana, month 4)

he say-3S the man
b. ich lese wenig (Ana, month 7.4)
I read-1S little
c. wir lernen (Zita, month 6.5)
we learn-1P
d. hunderte du bekommst (Zita, month 15)
hundred you get-2S

19) a. er kaufe ein Blume (Ana, month 4)

he buy-1S a flower
b. Michael spiele hier (Zita, month 10)
Michael play-1S here
c. die Männer arbeite ganze Tag (Zita, month 23.5)
the men work-1S whole day

The relatively high number of erroneous -e forms in the data are,

on the face of it, somewhat problematic for our hypothesis. These
forms co-occur with cases of -e appropriately agreeing with a 1st
person subject (over 70% showing appropriate agreement). Most
of the errors involve a 3rd person singular subject, as in (19a) and
 Philippe Prévost and Lydia White 123

(19b). One possibility is that the -e forms are non-finite, as discussed

earlier and assumed by Meisel (1991). If these forms are non-finite,
then they simply constitute further instances of missing surface
inflection, with a non-finite form being used in place of a finite one.
However, in that case one might expect a higher incidence of verbs
ending in -e in genuinely non-finite positions. Ana produced only 3
-e forms following a preposition or another verb (compared to 64
verbs ending in -en), while Zita produced only one such form
compared to 70 -en forms (see Table 4 above). However, Müller
(1998: 136) reports that Bruno does use -e in non-finite contexts,
including after prepositions and modals. Sixteen per cent of his verb
forms occurring after modals end with -e. Another possibility is that
-e represents some kind of default agreement.
Turning now to the other errors, Zita’s incorrect usage of -st is
largely lexically-driven: half of her errors involve the verb sagen
(‘say’), as illustrated in (20a). Ana’s errors on ist (‘is’) and
will/kann/muβ (‘want/can/must’) occur in plural contexts – (20b and
20c) – often involving the subject die Leute (‘the people’) which is
singular in Spanish, Ana’s mother tongue (la gente). This might
explain her use of singular agreement in this case (see also Müller,
1998).

20) a. mein Schwager sagst auch (Zita, month 22)

my brother-in-law say-2S too
b. aber die Leute ist ein bißchen müde (Ana, month 13.5)
but the people is-3S a little tired
c. viele Leute will nich <commenzar> <estudiar> (Ana, month
many people want-ø not start study 4.5)

In general, our results differ from Meisel (1991), who reports

generally low accuracy in verbal inflection. However, his study and
ours contrast in the method used to assess knowledge of agreement.
Meisel first looked for subjects and then checked whether the verb
showed appropriate agreement with the subject. He found that
many verbs did not agree with the subject, either displaying the
non-finite marker -en or not being inflected at all, which led him
to conclude that knowledge of agreement was impaired. In contrast,
we agree with Poeppel and Wexler (1993) that the most appropriate
way to determine knowledge of agreement is to look at whether a
form, if inflected, shows accurate agreement. Our results suggest
that agreement is in fact in place: when an inflected form is used,
agreement is largely accurate (see also Clahsen, 1988). The claim of
the MSIH is that the non-finite or bare forms that occur are in fact
finite defaults, rather than evidence of incorrect agreement.
124 Missing surface inflection

b Subject clitics (L2 French): The investigation of subject

clitics is restricted to L2 French since German does not have
such elements. Assuming that subject clitics are agreement
markers appearing in Infl (Auger, 1994), the MSIH predicts that
when they occur in clitic-doubling constructions, they should
agree with the co-referring subject NP. Under the IRH, such
agreement is not expected since agreement features are held to be
impaired. Results in Table 10 show that subject clitics agree in
person, number and gender with the NP to a large extent, in support
of the MSIH.
Examples in (21) show subject clitics agreeing with the NP in
both finite and non-finite environments.11 In (21d), the gender
assigned to the NP is masculine instead of feminine, but the
agreeing clitic is also masculine.

21) a. il est parti l’ Espagne, lui (Abdelmalek, month 14)

he is-3S gone the Spain him
b. moi j’ai pas de quarante francs (Abdelmalek, month 14)
me I have-IS not some forty francs
c. monsieur il arriver (Zahra, month 18.5)
mister he arrive-INF
d. le fille il a volé le pain (Zahra, month 27.7)
the-MASC girl he has-3S stolen the bread

Most errors involved the masculine 3rd person singular clitic il

(‘he’), which appears to be used as yet another default form.
Erroneous usage of il involves gender (where it replaces the
corresponding clitic elle), as in (22), or person, as in (23). Such
errors constitute 95% of Abdelmalek’s clitic errors and 78.5% of
Zahra’s.

Table 10 Subject clitics in clitic doubling constructions

Abdelmalek Zahra

Clitic Clitic does Clitic agrees Clitic does

agrees with not agree with subject not agree
subject with subject with subject
Finite verb forms 103 17 111 8
Non-finite verb forms 17 2 30 6
Total 120 19 (13.5%) 141 14 (9%)

11
What is at stake here is the agreement between the clitic and the NP, regardless of
the form of the verb. In Prévost and White (1999), we present data showing that both
finite and non-finite verb forms are found with subject clitics in the case of adult learners
We argue that the incidence of non-finite verbs with clitics is supportive of the MSIH. See
also Table 6.
 Philippe Prévost and Lydia White 125

22) a. la sécurité sociale, il paye (Abdelmalek, month 51.5)

the-FEM security social he pay-1/2/3S
b. la dame i travaille avec son mari (Zahra, month
the-FEM woman he work-1/2/3S with her husband 26.7)

23) a. toi il a raison (Abdelmalek, month 24)

you he has-3S right
b. moi aussi il parle avec la dame (Zahra, month 41)
me too he speak-1/2/3S with the woman

To summarize the results on agreement, in the case of verbs in

L2 French and German as well as subject clitics in L2 French,
agreement is largely accurate. When an inflectional suffix bearing
agreement is found on a verb, it is appropriate for the subject in
question; when a clitic is found, it is appropriate for the NP that it
doubles. There are two main exceptions to this observation:
1) in L2 German, the 1st person singular ending -e is found with
subjects that are not 1st person singular;
2) in L2 French, the 3rd person masculine singular clitic il is found
with subjects that are not 3rd person masculine.
We suggest that these may be default agreement markers.

VIII Discussion and conclusion

Both hypotheses discussed in this article assume variability in
incidence of inflected verb forms in L2 acquisition. The MSIH and
the IRH make similar predictions about non-finite morphology,
assuming that non-finite forms will be found in finite contexts. In
the case of the MSIH, this is because non-finite forms are resorted
to as a default (in a sense to be made clearer below). Although non-
finite in form, they behave syntactically like finite verbs. In the case
of the IRH, on the other hand, incidence of non-finite forms in
finite contexts is due to a breakdown in feature-checking
mechanisms, with the consequence that non-finite forms can occur
anywhere.
Where the two hypotheses differ is over their claims about finite
morphology, as well as agreement, and this is what we have
concentrated on in the present article. Our L2 French and German
data show that, with a few exceptions, finite forms do not occur in
non-finite contexts: they are not found after a preposition or a
negator, and they are not used along with another verb in the same
clause. These results suggest that adult learners are in fact
distinguishing between finite and non-finite morphology and that
finite forms do not substitute for non-finite. Finite forms rarely
126 Missing surface inflection

occur in low positions, suggesting that finite features are present in

the ILG rather than being impaired. In addition, we find little
evidence of ‘faulty’ agreement, in the sense of mismatches (or
feature clashes) between person, number and gender inflection on
a verb/clitic and the corresponding subject. In general, then, the
results support the MSIH.
By contrast, the data fail to confirm the IRH. We have argued
that the implication of missing functional categories or features
(Meisel, 1997) and inert feature strength (Eubank et al. 1997; Beck,
1998) is that finite verb forms should be relatively variable. If
features are lacking in IL grammars, feature-checking mechanisms
cannot apply and finite forms should appear in any position, finite
or non-finite. If feature strength is impaired, there is no motivation
for finite verbs to raise systematically; hence, they may remain
within the VP. Thus, the strong contingency between verb place-
ment and finiteness uncovered in this study is problematic for the
IRH (pace Eubank and Beck, 1998). Meisel’s approach further
predicts a high incidence of faulty agreement since all features,
including phi-features, are held to be missing. Again, this is not
supported by the data. Eubank and Beck do not make the same
predictions as Meisel on accuracy of agreement, since only V-
features related to finiteness are held to be inert. However, it is not
clear under their approach why certain features should be subject
to such inertness and others not.
The IRH predicts greater variability involving finiteness than the
MSIH, in other words, finite forms occurring in non-finite positions
and non-finite forms in finite positions. In contrast, the variability
assumed by the MSIH is more constrained; variability is limited to
(apparently) non-finite forms which may be found in finite contexts.
In that respect, the MSIH makes a clear distinction between
knowledge of surface morphology and knowledge of the abstract
features underlying movement (see Clahsen (1988) who also argues
for a dissociation between knowledge of abstract agreement and
surface morphology). Though the L2 learners investigated here
produced main verbs with infinitival markers, they also showed
knowledge of finiteness and its syntactic consequences with respect
to negation, supporting the claim that they have set L2 feature
strength to [+strong].
It is of course possible for proponents of the IRH to argue that
although abstract functional categories, features or feature values
are impaired, L2 learners can be relatively successful in the domain
of morphology by resorting to ‘parasites’ which patch up deficient
aspects of the grammar (Beck, 1997) or to explicitly learned
inflectional paradigms (Bley-Vroman, personal communication;
 Philippe Prévost and Lydia White 127

Meisel, personal communication). In that case, apparently successful

L2 acquisition of morphology tells one nothing, being taken to
indicate successful ‘parasites’ rather than appropriate knowledge of
both abstract features and surface morphological realization. Lack
of success, on the other hand, is more revealing. On the IRH
(assuming parasites or explicit learning), it is not clear why, when
mistakes are made, they should be in one direction rather than
another.
We have argued that L2 learners have abstract features for
finiteness and agreement in their interlanguage representation, as
evidenced by the syntactic and morphological behaviour of finite
verbs. They do, however, exhibit problems with the surface
morphological realization of particular forms, sometimes resorting
to default forms; in other words, there are what Lardiere (2000)
calls ‘mapping problems’ between surface forms and abstract
features. This raises the question of what mechanisms underlie the
appearance of defaults. In other words, what is their featural
content? One possible solution is provided by Distributed
Morphology (DM) (Halle and Marantz, 1993), which makes a clear
distinction between grammatical features (such as tense, person,
etc.) on a given inflected form and the node that hosts it in the
syntax.12
In DM, each inflected form is assumed to be associated with a
bundle of features, such as tense, person, number and gender. For
lexical insertion to take place, the features of the vocabulary item
must be consistent with the features of the terminal node in the
syntax where it is to be inserted. While the features of a syntactic
node will be fully specified, features of a lexical item may be
partially specified – in that some features may be lacking – or
underspecified – in that some feature values may not be specified
(Lumsden and Halefom, 1998). The features of the lexical item do
not need to exactly match all the features of the hosting node: it is
sufficient that they form a proper subset of the feature bundle of
that node. In the absence of an exact match, there is a competition
between potential candidates for insertion, the winner being the
form with the most features that match those of the terminal node.
Our assumption is that L2 learners have acquired the relevant
features of the terminal nodes in the syntax (from the L1, from UG
or motivated by L2 input). Their problems lie with the feature
specification of the associated lexical items. In the grammar of an
adult native speaker, non-finite forms are specified as [–finite] and
12
See Lumsden and Halefom (1998) for a development of DM which is consistent with
minimalist checking theory, dispensing with Halle and Marantz’ distinct level of
morphological structure.
128 Missing surface inflection

finite ones are [+finite]. In addition, finite forms are associated with
[± past] and phi-features (person, number and gender). Non-finite
forms cannot be inserted in any position where finite inflection is
required because of mismatching features; the same is true of non-
agreeing forms. Now suppose that in adult L2 acquisition, non-finite
forms are underspecified with respect to finiteness (i.e. they are α
finite). If so, then non-finite forms may be inserted into a node
involving the feature [+finite]. This would explain the appearance
of non-finite forms in finite positions. Finite forms, on the other
hand, are not underspecified, hence cannot be inserted into non-
finite positions. The same argument applies to the German -e form
and the French clitic il, which we assume to be underspecified for
agreement features (i.e. they are α person, α number, α gender),
yielding agreement mismatches with the subject (for related
proposals, see Müller, 1998). In all three cases, insertion may take
place because there is no conflict between what is required by the
terminal node and what is specified on the verb form in question.
A related proposal is made for L1 acquisition by Ferdinand
(1996). She argues that in the L1 acquisition of French, the 3rd
person singular form serves as a finite ‘elsewhere’ form (in our
terms, a default) which is not specified for person or number. As a
result, it shows up with 1st and 2nd person singular subjects and
with plural subjects. Initially, only the elsewhere form is found. This
is followed by a stage where the more specified forms are acquired
but the elsewhere form is still used. Finally, only fully specified verb
forms are found. Overgeneralizations of elsewhere forms are only
in one direction: while a 3rd person singular verb may occur with
subjects that are not 3rd person singular, plural verb forms do not
occur with singular subjects, nor do 1st person verb forms occur
with 3rd person subjects, suggesting that these are fully specified
for relevant featural content. In addition to elsewhere forms on
verbs, Ferdinand proposes that L1 acquirers have an elsewhere
form of subject clitics, underspecified for agreement features. In L1
acquisition, this takes the form of a reduced vowel, whereas in our
L2 data, the 3rd person singular clitic il acts as a default form, as
we have seen.
Acquisition of L1 morphology is always successful (except in
pathological cases). Once morphosyntactic and semantic features
have been mapped to a particular form, that association is
permanent, preventing non-finite forms from surfacing in finite
positions, as well as agreement mismatches (except for the short
period of variability between elsewhere forms and agreeing forms
noted by Ferdinand). In other words, once the more fully specified
forms are acquired, they replace underspecified forms, which can
 Philippe Prévost and Lydia White 129

no longer surface. Hence, one sees a developmental progression in

L1 acquisition. The same is only partially true of L2 acquisition. In
the case of default agreement, there was in fact some evidence of
a developmental progression: in Ana’s case, for example, 3rd person
singular forms were delayed. She did not produce -t forms
consistently until month 11.7. Before that, many verbs ending in -e
were found with 3rd person singular subjects. This was followed by
a period of overlap between -t forms and -e forms. Finally, the
number of errors on -e went down while -t forms were produced
more regularly (see also Clahsen, 1988). Müller (1998) also reports
that Bruno shows a developmental progression, with forms
underspecified for agreement subsequently replaced by forms
which are fully specified. So far, then, incidence of defaults in L1
and L2 acquisition appears to be parallel, involving the use of
underspecified or default forms before the appropriate agreeing
forms have been acquired.
Despite these similarities, there are also considerable differences.
Our data show that non-finite forms alternate with finite forms
throughout the period of observation in both L2 French and L2
German and that non-finite forms never disappear (for details, see
Prévost and White, 1999). Furthermore, Lardiere (1998a; 1998b)
shows that tense and agreement morphology remain variable in an
end state L2 English grammar. What seems to be going on, then, is
that even when more fully specified forms are acquired, they do
not necessarily always ‘win’ in the competition for lexical insertion,
so that underspecified forms continue to surface. This suggests that
access to the more fully specified lexical entries is sometimes
blocked. That is, even when a form specified for the relevant
features has been acquired, it becomes temporarily irretrievable
from the lexicon. We speculate that this might be due to processing
reasons or to communication pressure, in which case one might
expect the problem to affect different kinds of language use
differentially. For example, L2 learners might be expected to
perform more accurately on an untimed grammaticality judgement
task (where they have time to access the relevant representation)
than in spontaneous production or in timed tasks.
In conclusion, we have argued that L2 learners’ treatment of
finite morphology is not arbitrary: when a verb is finite, the relevant
syntactic reflexes are found, as well as appropriate agreement. We
have suggested that problems of adult L2 learners relate to the
mapping of specific morphological forms to abstract categories. We
do not, however, wish to imply that the problem is trivial. Clearly,
there is a real difference between L1 and L2 acquirers in this
domain: normal L1 acquirers always acquire the appropriate
130 Missing surface inflection

inflectional morphology of their mother tongue and use it

consistently, whereas L2 learners often do not. However, despite
lack of consistent use of verbal morphology, when it occurs, it is
systematic, suggesting that there is no impairment at an abstract
level.

Acknowledgments
We should like to thank Jürgen Meisel for making his data available
and Wolfgang Klein for permission to make use of the ESF data,
as well as Maria Beck, Laurent Dekydspotter, Lynn Eubank,
Bonnie Schwartz and Rex Sprouse for helpful discussions. This
research was supported with a research grant from FCAR. Earlier
versions of this article were presented at GASLA 1998, SLRF 1998
and BU 1998 and we thank the audiences for their comments.

IX References
Auger, J. 1994: Pronominal clitics in Québec colloquial French: a
morphological analysis. Doctoral dissertation, University of
Pennsylvania.
Bailey, N., Madden, C. and Krashen, S. 1974: Is there a ‘natural sequence’
in adult second language earning? Language Learning 24, 235–43.
Beck, M-L. 1997: Viruses, parasites and optionality in L2 performance.
Paper presented at Second Language Research Forum, Michigan
State University, October 1997.
––––– 1998: L2 acquisition and obligatory head movement: English-
speaking learners of German and the local impairment hypothesis.
Studies in Second Language Acquisition 20, 311–48.
Borer, H. and Rohrbacher, B. 1997: Features and projections: arguments
for the full competence hypothesis. In Hughes, E. Hughes, M. and
Greenhill, A., editors, Proceedings of the 21st Annual Boston
University Conference on Language Development. Somerville, MA:
Cascadilla Press, 24–35.
Boser, K., Lust, B., Santelmann, L. and Whitman, J. 1992: The syntax of
CP and V2 in early child German (ECG): the strong continuity
hypothesis. In Broderick, K., editor, Proceedings of NELS 10.
University of Massachussetts: Graduate Linguistics Students
Association, 51–65.
Chomsky, N. 1995: The Minimalist Program. Cambridge, MA: MIT Press.
Clahsen, H. 1988: Parameterized grammatical theory and language
acquisition: a study of the acquisition of verb placement and inflection
by children and adults. In Flynn, S. and O’Neil, W., editors, Linguistic
theory in second language acquisition. Dordrecht: Kluwer, 47–75.
Clahsen, H., Eisenbeiss, S. and Penke, M. 1996: Lexical learning in early
syntactic development. In Clahsen, H., editor, Generative perspectives
 Philippe Prévost and Lydia White 131

on language acquisition: empirical findings, theoretical considerations,

crosslinguistic comparisons. Amsterdam: John Benjamins, 129–59.
Clahsen, H., Meisel, J. and Pienemann, M. 1983: Deutsch als Zweitsprache.
Tübingen: Narr.
den Besten, H. 1983: On the interaction of root transformations and lexical
detective rules. In Abraham, W., editor, On the formal syntax of the
West Germania. Amsterdam: John Benjamins, 47–131.
Dulay, H. and Burt, M. 1974: Natural sequences in child second language
acquisition. Language Learning 24, 37–53.
Epstein, S., Flynn, S. and Martohardjono, G. 1996: Second language
acquisition: theoretical and experimental issues in contemporary
research. Brain and Behavioral Sciences 19, 677–714.
Eubank, L. 1993/94: On the transfer of parametric values in L2
development. Language Acquisition 3, 183–208.
Eubank, L. and Beck, M-L. 1998: OI-like effects in adult L2 acquisition.
In Greenhill, A., Hughes, M., Littlefield, H. and Walsh, H., editors,
Proceedings of the 22nd Annual Boston University Conference on
Language Development. Somerville, MA: Cascadilla Press, 189–206.
Eubank, L., Bischof, J., Huffstutler, A., Leek, P. and West, C. 1997: ‘Tom
eats slowly cooked eggs’: thematic-verb raising in L2 knowledge.
Language Acquisition 6, 171–99.
Ferdinand, A. 1996: The development of functional categories: The
acquisition of the subject in French. The Hague: Holland Academic
Graphics.
Grondin, N. and White, L. 1996: Functional categories in child L2
acquisition of French. Language Acquisition 5, 1–34.
Haegeman, L. 1995: Root infinitives, tense and truncated structures.
Language Acquisition 4, 205–55.
Halle, M. and Marantz, A. 1993: Distributed Morphology and the pieces
of inflection. In Hale, K. and Keyser, S.J., editors, The view from
building 20: essays in linguistics in honor of Sylvain Bromberger.
Cambridge, MA: MIT Press, 111–76.
Haznedar, B. and Schwartz, B. 1997: Are there optional infinitives in child
L2 acquisition? In Hughes, E., Hughes, M. and Greenhill, A., editors,
Proceedings of the 21st Annual Boston University Conference on
Language Development. Somerville, MA: Cascadilla Press, 257–68.
Hoekstra, T., Hyams, N. and Becker, M. 1997: The underspecification of
number and the licensing of root infinitives. In Hughes, E., Hughes,
M. and Greenhill, A., editors, Proceedings of the 21st Annual Boston
University Conference on Language Development. Somerville, MA:
Cascadilla Press, 293–306.
Lardiere, D. 1998a: Case and tense in the ‘fossilized’ steady state. Second
Language Research 14, 1–26.
––––– 1998b: Dissociating syntax from morphology in a divergent end-state
grammar. Second Language Research 14, 359–75.
––––– 2000: Mapping features to forms in second language acquisition. In
Archibald, J., editor, Second language acquisition and linguistic theory.
Oxford: Blackwell, 102–29.
132 Missing surface inflection

Lardiere, D. and Schwartz, B.D. 1997: Feature-marking in the L2

development of deverbal compounds. Journal of Linguistics 33,
327–53.
Lumsden, J. and Halefom, G. 1998: A minimal theory of inflection.
Unpublished manuscript, Université de Québec à Montréal.
Martohardjono, G., Kessler, K. Finger, I. and Argodale, P. 2000: Variability
in L2 English grammatical morphemes: the question of functional
categories in a performance versus knowledge paradigm. In Mizera,
G. and Juffs, A., editors, Proceedings of GASLA IV, Pitt Working
Papers in Linguistics, 4.
Meisel, J. 1991: Principles of Universal Grammar and strategies of
language learning: some similarities and differences between first and
second language acquisition. In Eubank, L., editor, Point counterpoint:
Universal Grammar in the second language. Amsterdam: John
Benjamins, 231–76.
––––– 1997: The acquisition of the syntax of negation in French and
German: contrasting first and second language development. Second
Language Research 13, 227–63.
Müller, N. 1998: UG access without parameter setting. In Beck, M-L.
editor, Morphology and its interfaces in second language knowledge.
Amsterdam: John Benjamins, 115–63.
Perdue, C. 1984: Second language acquisition by adult immigrants: a field
manual. Rowley, MA: Newbury House.
––––– editor, 1993: Adult second language acquisition: crosslinguistic
perspectives. Cambridge: Cambridge University Press.
Phillips, C. 1996: Root infinitives are finite. In Stringfellow, A., Cahana-
Amitay, D., Hughes, E. and Zukowski, A., editors, Proceedings of the
20th Annual Boston University Conference on Language
Development. Somerville, MA: Cascadilla Press, 588–99.
Poeppel, D. and Wexler, K. 1993: The full competence hypothesis of clause
structure in early German. Language 69, 1–33.
Pollock, J.-Y. 1989: Verb movement, Universal Grammar, and the structure
of IP. Linguistic Inquiry 20, 365–424.
Prévost, P. 1997: Truncation in second language acquisition. Doctoral
dissertation, McGill University.
Prévost, P. and White, L. 1999: Accounting for morphological variation in
second language acquisition: truncation or missing inflection? In
Friedemann, M.-A. and Rizzi, L., editors, The acquisition of syntax.
London: Longman, 202–35.
Radford, A. 1992: The acquisition of the morphosyntax of finite verbs in
English. In Meisel, J., editor, The acquisition of verb placement.
Dordrecht: Kluwer, 1–22.
Rizzi, L. 1993/94: Some notes on linguistic theory and language
development: the case of root infinitives. Language Acquisition 3,
371–93.
Sprouse, R. 1998: Some notes on the relationship between inflectional
morphology and parameter setting in first and second language
acquisition. In Beck, M., editor, Morphology and the interfaces in
 Philippe Prévost and Lydia White 133

second language knowledge. Amsterdam: John Benjamins, 41–67.

Wexler, K. 1994: Optional infinitives, head movement and the economy of
derivations. In Lightfoot, D. and Hornstein, N., editors, Verb
movement. Cambridge: Cambridge University Press, 305–50.
Zobl, H. and Liceras, J. 1994: Functional categories and acquisition orders.
Language Learning 44, 159–80.