BUCLD 39 Proceedings
To be published in 2015 by Cascadilla Press
Rights forms signed by all authors
Does Negative Feedback Have an Effect on Language
Elena Kulinich1,2*, Phaedra Royle2,3, and Daniel Valois1,2,4
One of the central philosophical tenets of generative grammar is that
children acquire their mother tongue in spite of severe limitations in their
linguistic environment. This is known as the Poverty of Stimulus argument
(POS). Various aspects of the POS have been questioned over the last few years
(e.g., Pullum and Scholz 2002) thereby challenging nativism in favour of a more
data based approach to language learning. One of those challenges concerns
negative feedback, the absence of which is part of the POS argument. Indeed, a
few researchers have argued that children may benefit from contrastive negative
feedback to correct their early non-target productions (e.g., Child: I *go [+past]
to school. Parent: You WENT to school, e.g., Saxton 1997, 2000; Saxton et al.
1998; Clark and Chouinard 2000; Chouinard and Clark 2003).
Several terms are used in the literature to describe the information given to
children with respect to the ungrammaticality of their utterances: negative
feedback, negative evidence, negative input, and corrective input. Which are
sometimes used as synonyms (e.g., in Marcus 1993, negative evidence is
sometimes equivalent to negative feedback). In some case, however, a
distinction is made between these terms, as in Saxton (2000) where a tripartite
distinction between negative evidence, negative feedback and negative input is
proposed. The term negative evidence is used to denote corrective input that is
provided immediately after a child’s grammatical error in the form of a correct
alternative to this error. Negative feedback corresponds to a non-specific signal
about ungrammaticality of the child utterance (e.g., clarification question). The
term negative input is used as “a generic term to denote any kind of adult
*1Département de linguistique, Université de Montréal, 2CRBLM–Centre for Research on
Brain, Language and Music, 3École d’orthophonie et d’audiologie, Université de
Montréal and 4ISC-Institut des Sciences Cognitives, UQÀM.!
Corresponding author at: Département de linguistique, Université de Montréal, C.P.
6128, Succursale Centre-Ville, Montréal, Québec, Canada H3C 3J7.
Tel.:+1.514.343.6111, ext. 5041. Email addresses: firstname.lastname@example.org (E.
Kulinich), email@example.com (P. Royle), firstname.lastname@example.org
response, contingent on child grammatical error, which embodies information
conducive to the realignment of an overgeneralized grammar” (Saxton 1997:
140). In our study we define negative feedback as an adult (usually parental)
response to a child’s utterance that provides (explicit or implicit) negative
evidence about its ungrammaticality. For example, (1) illustrates a case of
negative evidence in which a Russian-speaking mother corrects a child’s
erroneous utterance (*dostavaet ‘she gets out’), by providing the appropriate
form of the 3sg/pres. verb (dostaet), thus signalling to the child that the form
(s)he produced is non-existant (or unconventional) in the adult grammar.
(1) Child: Mama pryamo iz rotika kostochku *dostavaet.
Mother directly from mouth stone-ACC. get-3SG.PRES.
‘Mother gets a stone out of her mouth’.
‘(She) gets (it) out’.
One of the important questions discussed in studies on negative feedback
concerns the actual availability of negative evidence to children. According to
some researchers the only feedback available to children is “noisy feedback”,
i.e., corrective responses that are given is response to both ungrammatical and
grammatical sentences. Noisy feedback is deemed to be “inconsistent across
parents, declines or disappears with age, and is probably not provided for all
types of errors” (Marcus 1993: 57) as such, it is not considered helpful. At the
other end of the spectrum, others have shown that children do have access to,
and make use of, the negative feedback that they receive (e.g., Chouinard and
Clark 2003; Saxton 2000).
The general question we address in this study is twofold: (i) does negative
feedback have an effect on error elimination and, (ii) if children are found to
correct their utterances following negative feedback, can we conclude without
doubt that negative feedback per se is responsible for this change in behaviour?
Indeed, a number of studies have shown that while parental feedback does not
play a role in language acquisition (see e.g., Brown and Hanlon 1970, Braine
1971), children have internal cognitive mechanisms that help them to determine
which ones among the utterances they hear are grammatical and which ones are
not (Marcus 1993; Pinker 1989). This point of view supports Chomsky’s (1980)
poverty of the stimulus argument.
One a different note, other researchers have found that negative feedback is
helpful to the child (e.g., Saxton et al. 1998, Saxton et al. 2005). In these cases,
the effect was more prominent in contexts where erroneous and correct forms
were immediately juxtaposed or contrasted in the conversation (Saxton’s 1997
(2) Child: He *shooted the fish.
Adult: He shot the fish.)
According to Chouinard and Clark (2003) the effect of reformulations (a
type of negative feedback providing correction) was observed for all categories
of errors including phonological (e.g., girl *dere rather than girl there),
morphological (e.g., I like *carrot rather than I like carrots), lexical (e.g., !suit
rather than coat) and syntactic errors (e.g., *Ø sun Ø gone rather than the sun is
However, one major problem with most studies on negative feedback is that
they are based on longitudinal corpus data, which might result in confounds
between parenting style and feedback effects in children’s responses. Another
problem is that they fail to answer the all-important question as to whether
corrections in child speech are punctual or long-lasting in any significant way
(Saxton et al. 2005).2 Consequently, in order to isolate the effects of adult
feedback from other potential factors influencing retreat from error, and to
assess the lasting effect of negative feedback in child language correction, we
decided to investigate feedback in a controlled environment. More precisely, we
conducted a series of elicited tasks with monolingual Russian speaking children
aged from 3 to 4 years at various point in time.
Specifically, we will address the following questions:
(1) Does negative feedback play a role in the acquisition of verb
morphology in Russian?
(2) Does recovery from verb overregularizations depend on the type of
feedback provided by adults?
Based on previous studies of negative feedback we predict that there should
be a gradient effect of feedback type on verb form production: that is Correction
> Clarification Question > Repetition > No Feedback (see methods below for a
description of feedback types).
We developed an elicitation task for Russian verbs with different types of
feedback with an aim to test whether they have effects on erroneous form
production in Russian children. Four types of feedback were tested (see below in
Procedure for examples of each type):
1) Correction (C);
2) Clarification Question (CQ);
3) Repetition (R);
4) No Feedback or Move-On (NF).
2!Another drawback is that a majority of studies have focused on English-speaking
children. It is a legitimate question to ask whether children learning other languages show
similar acquisition patterns as a function of the language structure (see Slobin 2014 on
the importance of cross-linguistic studies).!
The feedback was provided by the investigators (the first author or a trained
Russian research assistant) and was given in the context of contrast to errors (as
defined in Saxton, 1997).
In order to verify if there was any general effect of time and maintenance on
recovery from errors. the sessions were conducted according to a constant time
interval: 2 weeks between sessions 1, 2 and 3 (the same time interval that was
used by Morgan et al. (1995) for the time series analysis of corpus data), and a
longer, 4 week, interval between sessions 3 and 4.
Seventy-six monolingual native speakers of Russian, aged from 2;11 to 3;11
participated in the study but only 35 of them (Mean age = 3;4, SD = 0,36)
participated in all sessions. Data from these 35 children are presented here.
Participants were recruited at three Moscow daycares. Both parents of all
children were native Russian speakers, and the daycares were exclusively
Russian speaking. Children were randomly assigned to one of four feedback
groups: 1. Correction (n = 9), 2. Clarification question (n = 6), 3. Repetition (n =
10), and a control group 4. Without feedback (n = 10).
A storybook-style presentation with drawn pictures representing actions and
introducing different verbs was used for the study. The drawings were
specifically designed for the experiment by professional artists, and presented on
a computer screen one by one with contexts describing the pictures and
prompting for verb production by the participants.
12 early-acquired Russian verbs from two verb classes (6 in each, Table 1)
that undergo overregularization in the non-past tense were used. Age of
acquisition for all verb stimuli was controlled using data on early normal child
speech development “Mishka” (the MBCDI adapted for Russian3) and corpus
data from Kharchenko (2012). We used two groups of verbs: verbs with –ova-/-
u- suffixation as in ris-ova-t’ ‘to draw’ / ris-u-j-u ‘I am drawing’ (Class 2 verbs
in Zalizniak’s, 2003, classification) and verbs with an unproductive
morphonological (palatalization) alternation as in plaka-t’ ‘to cry’ / plach-u ‘I
am crying’ (Class 6 verbs according to Zalizniak 2003). For example, children
frequently produce errors such as *ris-ovaj-u instead of risuj-u ‘I am drawing’,
or *plakaj-u instead of plach-u ‘I am crying’, therefore applying the pattern of
3 The Russian version of the MacArthur-Bates Communicative Development Inventories
(MBCDI) « Mishka » was elaborated in the Herzen State Pedagogical University of
Russia, St.Petersburg (2002).
the most productive Russian verb class (e.g., chita-t’/chitaj-u ‘to read’/‘I am
reading’). This pattern is known as the “j-correlation” model (Ceytlin 2009) and
consists in inserting a yod /j/ in the intervocalic position at the stem/inflection
boundary in the non-past forms.
Table 1. List of stimuli items
Class 2 verbs
Class 6 verbs
davat’ ‘to give’
celovat’ ‘to kiss’
vstavat’ ‘to stand up’
risovat’ ‘to draw’
tancevat’ ‘to dance’
dostavat’ ‘to get’
plakat’ ‘to cry’
iskat’ ‘to seek’
prjatat’ ‘to hide’
skakat’ ‘to ride’
rezat’ ‘to cut’
mahat’ ‘to wave’
The children participated in four short elicitation tasks. The interviews were
conducted in a comfortable isolated room at the daycare, and usually lasted no
longer than 5 minutes for each participant. The child was presented with pictures
from the storybook-style presentation (see Figure 1) and then asked to produce
the non-past (present) form of the verbs (e.g., What is he (she) doing?). The
experiment was repeated with two-week time interval between the first three
sessions, and a four-week interval between sessions 3 and 4.
Figure 1. Drawings for plakat’ ‘to cry’ and risovat’ ‘to draw’
In each group of participants the investigator’s feedback was provided in
different way. For example, in the Correction group the conversation was
constructed as follow:
(3) Adult: Chto delaet etot mal’chik?
What do-3SG-PRES this boy
‘What is this boy doing’?
Child: On *risovaet.
‘He is drawing’.
Adult: Ne *risovaet, a risuet.
NOT draw-3Sg-PRES but draw-3Sg-PRES.
‘Not drawing (OR), but drawing’.
In the Clarification question group, overregularizations (OR) were followed
by the question What? or What is s/he doing? and in the Repetition group the
investigator repeated the child’s erroneous form4. Children of the fourth control
group did not receive any feedback, i.e., in case of an error the investigator just
moved on to the next image without any reaction.
Figure 2. Percent correct verb form production in groups with different
feedback types (for Times 1, 2, 3 and 4)
The children’s responses were coded as “target” vs. “non-target”
productions. For example, for the verb ris-ova-t’ ‘to draw’ the form risuj-et
(3sg./non-past) ‘she is drawing’ is the target form while the overregularized
form *ris-ovaj-et with the same meaning is the non-target form. Only the first
response was rated. Percent correct responses was averaged for each group. The
data were entered into an ANOVA analysis with FEEDBACK (C, CQ, R, NF) as a
4 The idea to repeat an erroneous form comes from the first author’s personal
observations on phonological error feedback, where children try to correct errors repeated
by the adult. By choosing this type of feedback we wanted to verify whether the child
considers overregularized forms as errors when they are repeated by someone else. If so,
children should try to correct the erroneous form, if not – the repeated form could
reinforce the overregularized form, which may influence results. However, our results
show that both correct and overregularized forms seem to be equally good for the child
(which is also claimed in Saxton et al. 2005)
Correction Question Repetition No feedback
between-group factor and TESTING TIME (T1, T2, T3 and T4) as a within-group
factor. An alpha of 0.05 was used to determine significance for all statistical
tests. The results show a significant effect of TESTING TIME, F(3, 31) = 19.41, p
< 0.001, and no effect of FEEDBACK, F(1, 3) = 0.57, p > 0.1. No interaction
between TESTING TIME and FEEDBACK was found, F(3,31) = 1.04, p > 0.1. This
is illustrated in Figure 2 where we can observe that all groups improve
throughout the two months testing period, independent of feedback type.
According to Clark and Chouinard (2000), Chouinard and Clark (2003), and
Saxton et al. (1998, 2005) we should see improvements in verb production (i.e.
fewer overregularizations) in the Correction group as compared to other groups.
We also hypothesized that improvements in the Correction group should be
more considerable compared with the Clarification question and Repetition
groups, and even more so with the No Feedback group, due to the nature of their
feedback. However, results in all four groups (including the control group, i.e.
the one where children received no feedback) show no differences: all children
performed better in verb production tasks from session to session. That is,
children performed the same whether they received some type of feedback, or
no feedback at all.
These results are striking given that the aforementioned studies seemed to
show that corrective feedback impacts on language learning in English-speaking
(and French-speaking, Chouinard and Clark 2003) children. However, this
discrepancy can be explained by a number of methodological differences
between our study and theirs.
In Chouinard and Clark’s (2003) longitudinal study, corpus data were
analyzed. More types of error were evaluated than in our study (phonological,
morphological, lexical, syntactic), however, only one type of feedback
(reformulation, equivalent to our Correction condition) was investigated. The
authors claim that adult reformulations of child errors help identify the locus of
the error as well as to contrast the error with the conventional (correct) adult
form. The effect of adult reformulations was determined on the basis of
children’s uptake (for example, repetition or acknowledgement of the correct
form) and increase in child speech grammaticality. However, Chouinard and
Clark (2003) do not contrast reformulations to other types of feedback, nor to no
feedback conditions. Thus their data is not inconsistent with ours. Moreover,
their sample is very small (3 English-speaking and 2 French-speaking children
aged from 2;0 to 4;0) and would need additional evidence before we can
interpret their results as conclusive.
In Saxton’s (1997) experimental study, the author compares the immediate
effect of negative evidence (i.e., reformulation of the child erroneous utterance)
(Adult: What happened? Child: He STREEP his tummy. Adult: Yes, he STREPT
him) vs. positive input (Adult: Oh look, he STREPT his tummy. Is that right?
Tell the dragon what happened?) on the acquisition of 6 novel irregular past
forms (e.g., streep/strept, pro/prew, jing/jang, etc.). Children produced the
correct irregular form as a response more often following negative evidence than
when following positive input. However, the task was run over one experiment
with only 6 items of varying irregularity type, and children were asked to repeat
the right response. The question remains whether the effect of negative evidence
is maintained over time and whether it generalized to other verbs with similar
irregular patterns. Furthermore, only one measure was taken for each verb, while
we tested children four times on each one, which allowed us to check for
Saxton et al. (1998) study long-term effects of negative evidence using a
design similar to Saxton (1997) to evaluate the cumulative effect of feedback
over a period of five weeks. Children participated in ten sessions with 3-4 days
intervals for a total period of approximately one month. Results are consistent
with Saxton (1997) and show that there is maintenance throughout the testing
period. Both these studies raise the question of ecological validity, since the
proportion of negative input exceeds levels observed in naturalistic corpora.
Also negative evidence was supplied on all occasions when errors did occur,
which never happens in a real life exchanges.
A more recent study by Saxton et al. (2005) increases its ecological validity
and the amount of material analysed. Contrary to the previous studies where
only morphological errors using nonce words were tested, eleven categories of
errors were evaluated over an interval of 12 weeks in a spontaneous speech
corpus with parent-child interactions. Again, no effects were found for positive
input. Corrective negative input was associated with improvements in child
speech grammaticality for only for three structures (the possessive pronoun, the
3rd person singular verb inflection and copula verbs). These findings lead the
authors to conclude that negative input is not essential for successful language
acquisition, which is consistent with our findings in that feedback type does not
seem to have an impact on sub-regular and irregular Russian verb acquisition.
However, the children studied in Saxton et al (2005) were much younger —
with a mean age of 2 years — than those in Saxton’s et al. (1998) and Saxton
(1997), who were 3;8 to 4;6 and 4;9 to 5;6, respectively. It is possible that these
children are not mature enough to make use of feedback provided by parents (at
least for some categories of errors). Our participants were aged 3 to 4 years, but
nevertheless did not show feedback differences in learning irregular verbs in
Russian. We do not therefore think that age is an important factor explaining
differences between our experiment and those of the aforementioned authors.
A distinct advantage of our study is that we compared four different types
of feedback (not only reformulation or recast) on verb learning. In addition,
ageing (i.e. normal learning over time) effects were controlled by using a group
who were given no feedback. However, we could not control child’s exposure to
negative input outside the experiment. Further, it is possible that exposure to
feedback in this study is not sufficient to promote learning. This is improbable
however, as Saxton (1997) observed effects with few items and only one testing
period. There might be an advantage to using novel verbs, as in Saxton’s studies
(1997,1998), because by teaching children these forms, one can eliminate the
possibility that previous exposure to positive and negative evidence will affect
their behaviour in the experiment. On the other hand, novel forms in Saxton’s
study were not controlled for similarity to frequent (or less frequent) irregular
patterns. These factors can promote or demote analogical learning strategies. By
using real verbs, we could not control for previous exposure to verb stimuli, and
negative feedback before the study, nor at same time where the experiment was
conducted. However, age of acquisition for all verbs was controlled for, and
children were expected to know them, in order for the task to be natural. In any
case, we observe that, even if the exposure to feedback is limited, the
improvements observed suggests that children recover from their
overregularization errors without negative input.
In conclusion, we do not observe any significant difference between effects
of negative feedback type, or even no feedback, on verb overregularization in
Russian-speaking children aged 3 to 4 years old. Evidence from children
acquiring Russian verb morphology shows inefficacy of negative feedback on
error elimination. Children aged between 3 and 4 years get rid of errors in verb
form production (related or not to “j-correlation” model) without corrective
feedback. Our results point to the fact that general exposure to verbs (i.e.,
positive evidence) might be sufficient to overcome overregularization
behaviour. These findings support the general hypothesis that feedback is not a
strong driver of language acquisition.
More experimental studies are needed to investigate the effects of negative
feedback with respect to different error categories (e.g., morphological vs.
syntactic or semantic), different time intervals (e.g., 3-4 days between sessions
or longer periods, such as 3-6 months) and different ages (e.g., are older children
more receptive to feedback). In addition, it would be interesting to compare
parental input styles and their effects on child learning.
This research was supported by Joseph-Armand Bombardier Canada Doctoral
Scholarships, Michael Smith Foreign Study Supplements from the Social
Sciences and Humanities Research Council (SSHRC) and Paula Menyuk Travel
Award from the Boston University to E. Kulinich. The authors would like to
thank all of the children and parents who participated in the study. We are
grateful to daycare stuff for their invaluable help. We would also like to thank
Tamara Astakhova, Roman Ronko, and Vladislav Romashev from the Institut
for Modern Linguistic Research at the Sholokhov Moscow State University for
the Humanities, and the IMLR’s supervisor Anton Zimmerling, for their
contribution in organizing testing and data collecting.
Braine, Martin D. S. (1971). On two types of models of the internalization of grammars.
In D. J. Slobin (Ed.) The ontogenesis of grammar (153-186). New York : Academic
Brown, Roger and Hanlon, Camille. (1970). Derivational complexity and order of
acquisition in child speech. In J.R. Hayes (Ed.), Cognition and the development of
language (11-53). New York: Wiley.
Ceytlin, Stella N. (2009). Ocherki po slovoobrazovaniju i formoobrazovaniju v detskoj
rechi. Moskva, Znak (Studia Philologica).
Chomsky, Noam (1980). Rules and representations. Oxford: Basil Blackwell Publisher.
Chouinard, Michelle M. and Clark, Eve V. (2003). Adult reformulations of child errors as
negative evidence. Journal of Child Language, 30, 637-669.
Clark, Eve. V. and Chouinard, Michelle M. (2000). Énoncés enfantins et reformulations
adultes dans l’acquisition du langage. Langages, 140, 9-23.
Kharchenko, Vera K. (2012). Korpus detskih vyskazyvanij. Moskva.
Marcus, Gary. F. (1993). Negative evidence in language acquisition. Cognition, 46, 53-
Morgan, James L., Bonamo, Katherine M. and Travis, Lisa L. (1995). Negative evidence
on negative evidence. Developmental Psychology, 31(2). 180-197.
Pinker, Steven. (1989). Learnability and cognition: the acquisition of argument structure.
Cambridge, MA: MIT Press.
Pullum, Geoffrey K. and Scholz, Barbara. C. (2002). Empirical assessment of stimulus
poverty arguments. The Linguistic Review, 19, 9–50.
Saxton, Matthew. (1997). The contrast theory of negative input. Journal of Child
Language, 24, 139-161.
Saxton, Matthew. (2000). Negative evidence and negative feedback: immediate effects
on the grammaticality of child speech. First Language, 20, 221-252.
Saxton, Matthew, Backley, Phillip and Gallaway, Clare. (2005). Negative input for
grammatical errors: effects after a lag of 12 weeks. Journal of Child Language, 32,
Saxton, Matthew, Kulcsar, Bela, Marshall, Greer and Rupra, Mandeep. (1998). Longer-
term effects of corrective input: an experimental approach. Journal of Child
Language, 25, 701-721.
Slobin, Dan I. (2014). Before the beginning: the development of tools of the trade.
Journal of Child Language, 41, 1-17.
Zalizniak, Andrej A. (2003). Grammaticheskiy slovarj russkogo yazyka. Moskva: