Effectiveness of Materna1 Differential Feedback to Young Children`s

Effectiveness of Materna1 Differential Feedback to
Young Children's Utterances:
A Sequential Anaiysis
Constance M. Baillie
Mount Saint Vincent University
Submitted in partial fuifillment of the requirements for
the degree of Masters of Arts in School Psychology
at
Mount Saint Vincent University
Halifax, Nova Scotia
August, 1994
Copyright by Constance Baillie, 1994
1*1
National Library
ofCanada
Bibliothéque nationale
du Canada
Acquisitions and
Bibliographic Services
Acquisitions et
services bibliographiques
395 Wellington Street
Ottawa ON K I A ON4
395. me Wellington
OtiawaON K1AON4
Canada
Canada
your iYe
v m nittirence
Our fi& None refér(MCB
The author has granted a nonexclusive licence dowing the
National Library of Canada to
reproduce, loan, distribute or sell
copies of this thesis in microfonn,
paper or electronic formats.
L'auteur a accordé une licence non
exclusive permettant à la
Bibliothèque nationale du Canada de
reproduire, prêter, distribuer ou
vendre des copies de cette thèse sous
la forme de microfiche/film, de
reproduction sur papier ou sur format
électronique.
The author retains ownership of the
copyright in this thesis. Neither the
thesis nor substantial e e a c t s fiom it
rnay be printed or otherwise
reproduced without the author' s
permission.
L'auteur conserve la propriété du
droit d'auteur qui protège cette thèse.
Ni la thèse ni des extraits substantiels
de celle-ci ne doivent être imprimés
ou autrement reproduits sans son
autorisation.
Dedication
1 would like to dedicate this in recognition and appreciation of those who have
believed in and supporteci me throughout my university years.
Thank you to m y family whose phone calls, letters and visits have been the
bright spots in many days, and have meant more than they knew. These were the
things that helped me keep perspective, realize the importance of family in living a
happy and heaithy Iife, and l e m that while work should be enjoyed, it rernains only a
part of life.
A speciai thank you to my very dear fiiend, my fiance Robert, whose endless
support and encouragement kept me going during the times when 1 felt most like
giving up. Wetve been able to care for and be sensitive to each others needs
throughout this early part of our lives; 1 know that we can do the same in the years
that follow. it is with happy anticipation that 1 await whatever Our future may bring.
iii
Abstract
Revious research has provided evidence that mothers of young children offer
differential feedback to various aspects of their children's speech (e.g., grammar and
ambiguity). These results favour the nurture side of the nahue/nurture controversy,
which maintains that the environment influences children's acquisition of linguistic
grammar. However, to uphold the nurture perspective, it is necessary not only to
show that differential feedback exists within children's environments, but aiso that
children are aware of and are able to use the information contained in such feedback
to irnprove the quality of their speech. The curent study examines children's
responses immediately following maternai feedback for evidence of improvement in
quality of the subsequent responses over the children's initial utterances. The method
of sequential analysis is used to examine the interaction of four mother-child dyads
during videotaped playsessions which occurred in theû own homes when the children
were two years of age and again at age three. Results provide strong support for the
existence of differential maternal feedback (negative evidence) to the grammaîicality
and ambiguity of children's utterances. Although maternal feedback does seem to
influence children's utterances, no evidence was found to show that feedback facilitates
language development.
Acknowledgement
1 would like to thank Dr. David Funow for his insightful guidance as thesis
advisor, and for his much appreciated words of encouragement and expertise
throughout this study. His "pep talks" gave me confidence to believe that 1 r e d y
could accomplish what at times seemed to me almost beyond reach. Genuine thanks
are also extended to Dr. Kim Kienapple for his patience and assistance with the
analyses and their interpretation, and to Dr. Jennifer McLaren for her helpful critiques.
1 would like to thank Karen M. L. Skene, experienced research assistant and
good fiend, for her careful coding for the reliability checks. Gooci luck with d l your
future endeavours, Karen; I'll be watching for your hrst publication. A sincere thank
you is extended to Robert Johnson for the t h e he dedicated to data entry and
preparing tables.
Finally, 1 would like to thank Dr. Chris Moore at Dalhousie University for the
use of his archive of videotaped interaction, fiom which the data for the current study
were collected.
Table of Contents
Dedication
Abstract
...
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
...................................................
iv
.............................................
v
Acknowledgements
List of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii
Introduction . . . . . . . . . . . . . . . . . . . . . . i . . . . . . . . . . . . . . . . . . . . . . . . . . .
1
....................................................
18
Method
..............................................
18
.............................................
18
..............................................
24
....................................................
27
Subjects
Procedure
Analysis
Results
Discussion
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
List of Tables
Table 1.
Reliabilities for Child and Mother
Interaction Categories
Table 2.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
Frequencies of Child Utterances within
Grammatidity Category
Table 3.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
Frequencies of Child Utterances withui
Arnbiguity Category
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
. . . . . . . . . . . . . . . . . . . 31
Table 4.
Frequencies of Maternai Response Types
Table 5.
Standard Scores and Coefficients for Analyses of
Differential Matemal Responding (Six Feedback Types)
to Grammaticality Across Age Groups
Table 6.
.....................
33
Standard Scores and Coefficients for Analyses of
Differential Matemal Responding (Six Feedback Types)
to Ambiguity Individually at 2 Years and 3 Years
ofAge
Table 7.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
Standard Scores and Coefficients for Analyses of
Three Categones of Differential Matemal Responding to
Grammaticaiity Across Age Groups
Table 8.
......................
37
Standard Scores and Coefficients for Analyses of
'ïhree Categories of Differential Matemal Responding to
Ambiguity Across Age Groups
vii
..........................
38
Table 9.
Standard Scores and Coefficients for Analyses of the
Effects of Three Categories of. Matemal Feedback to
Grammaticaiity Across Age Groups
Table 10.
......................
40
Standard Scores and Coefficients for Analyses of the
Effects of Three Categories of Maternal Feedback to
Grammaticality Individually at 2 Years and 3 Years
of Age
Table 11.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
Standard Scores and Coefficients for Analyses of the
Effects of Three Categories of Maternal Feedback to
Ambiguity Across Age Groups
Table 12.
..........................
46
Standard Scores and Coefficients for Analyses of the
Effects of Three Categories of' Matemal Feedback to
Ambiguity individually at 2 Years and 3 Years of Age . . . . . . . . . 48
viii
Effectiveness of Feedback
1
Effectiveness of Materna1 Differential Feedback to
Young Children's Utterances:
A Sequential Analysis
The controversy over the relative effects of nature versus nurture has k e n an
active one within the social sciences, including psychology, for many years. The
study of language development within the field of psychology has not been exempt
fiom this debate. Currently, the nahire/nurture controversy in the language acquisition
literature is centered around arguments concerning the existence or non-existence of
feedback by caregivers to children conceming the acceptability of their speech.
Traditionally, the contention has been that the environment provides no helpful
basis for the Leaming of linguistic p m m a r by young children. This view supports the
idea that the ability to acquire grammatical language is an innate capacity which
develops independently of environmental input, a position obviously supportive of the
nature side of the nature/nurture controversy. More specifically, this perspective
maintains that young children rewive little or no feedback from their mothers as to the
gmmrnaticality or ungrammaticality of their speech. in other words, mothers do not
respond differentially to grammatical and ungrammatical utterances of children, and
thus children have no instruction towards their acquisition of linguistic grammar.
The classical proponents of this view were Brown and Hanlon (1970). Brown
and Hanlon (1970, p. 46) argued that the proposition, "Syntactically correct utterances
corne to prevail over syntactically incorrect utterances through the selective
Effectiveness of Feedback
7
administration of signs of approval and disapproval" cannot be mie unles parental
approval and disapproval are in fact appropriately contingent on syntacticai acniracy.
According to the results of their study (Brown & Hanlon, 1970, pp. 46-47), which
examined the pre-schwl years of t h e children, "parental reactions do not even meet
the minimal circumstance of appropriate contingency
...";approval and disapproval
were not primarily linked with the grammaticality of the utterance, but with the tnith
value of the proposition. Thus, they concluded that children do not receive feedback
fkom their environment as to the grammaticality of their speech. ln addition, because
some ungrammatical or immature foms were used by al1 three children in their study,
Brown and Hanlon took this to mean children are alike in their innate knowledge,
language-processing routines, preferences and assumptions which they bring to the
problem of language acquisition. Furthemore, they sumised that the only force
toward grammaticality operating on the child is the occasional conflict between hisher
theory of the structure of the Language and the data he/she receives.
More recently, studies have found that, in fact, children do receive different
feedback to their grammatical utterances compared to their ungrammatical utterances
(e.g. Bohannon & Stanowia, 1988; Demetras, Post, & Snow, 1986; Furrow, Baiiiie,
McLaren, & Moore, 1993; Funow & Moore, 1990; Hirsh-Pasek, Treiman, &
khneideman, 1984; and Penner, 1987). This feedback does not take the form of
mothers directly telling children, "No, that's wrong, dont say it that way", the explicit
feedback demanded by Brown and Hanlon (1970). However, these studies, which
Effectiveness of Feedback
3
included children ranging in age from 2 to 5 years, collectively show that the type of
response a mother makes to her child's utterance does stem to reflect whether that
utterance was grammatical. This type of materna1 responding which provides
differential feedback at a more implicit level to children's utterances has been labelled
negative evidence, as described by Marcus (1993). According to Maras, negative
evidence is a corrective response to children's ungrammatical utterances that can occur
only if parents are in some way sensitive to the grammaticality of their childrents
utterances. In contrast with negative evidence, positive evidence is the input
(sentences) which children hear in their daily interactions with others (Marcus, 1993).
Hirsh-Pasek et al. (1984) replicated, but also challenged, Brown and Hanlon
(1970). Whereas Brown and Hanlon studied only explicit verbal responses to
children's utterances, Hirsh-Pasek et al. considered that parents may show sensitivity
to the grammaticality of children's utterances. less explicitly (e.g., by repetitions of
child utterances). In addition, Hirsh-Pasek and her colleagues asked whether explicit
approval or disapproval would appear to be contingent on "well-fomedness" when a
wider age-range was considered (ages 2 to 5 years). The results of their study were
consistent with the h d i n g of Brown and Hanlon (1970) that explicit approvai and
disapproval are unrelateci to "well formedness". Again, approval seemed to be
govemed by tnith rather than grammaticality of children's speech. However, in
contrast with Brown and Hanlon, Wh-Pasek et al. found that for the 2-year-olds,
mothers more often repeated ill-fomed uttemnces than well-formed ones. Finally,
Effectiveness of Feedback
4
vimially al1 repetitions of the ill-formed sentences included a correction of the child's
e m r . This pattern was evident only for the 2-year-olds and not for the 3-, 4-, and
5-ycar-olds.
However, this finding suggests that mothers are sensitive to the form of
their children's utterances. Thus, awiording to Hirsh-Pasek et ai. (1984), the
environment might provide subtle cues to help the child narrow down his/her
hypotheses about candidate grammars as he/she tries to induce the grammar of hisher
native language.
Penner (1987) dso found evidencc of differential responding to children's
utterances on the basis of their grammaticality. Pemer investigated the possibility t hat
several specific types of parental responding are contingent on grammaticality of child
utterances. Six types of parental responses were observed in her study: topic
extensions, verbal approvals, expansions, repetitions, no responses, and confirming
questions. Aithough the differences for four of these response types were very small,
expansions and topic extensions showed large differences in occurrence depending on
the grammaticality of child utterances. Expansions more often followed
ungrammatical than grammatical utterances, while topic extensions occurred more
ftequently after grammatically correct child utterances. Although there is no evidence
for complete feedback that consistently distinguishes between grammatical and
ungrammatical utterances, Penner's (1987) fhdings contradia the belief that there are
no differences in parental responses to grammatical and ungrammatical child
utterances (e.g. Brown & Hanlon, 1970; and Morgan & Travis, 1989).
Effectiveness of Feedback
Bohamon and Stanowia (1988) took the study of differential responding to
children's gramrnaticality a step further. They compared several types of repetitious
responses and clarification questions of both parents and non-parents to children's
semantic, syntactic, and phonological errors to determine which responses
differentially followed children's ili-formed speech. Results indicated that both male
and fernale adults tended to request clarififation of well-formed utterances l e s than
they did utterances containhg a syntactic or phonological e m r . As well as king
differentiall y responsive to the grammaticality of children's utterances, adults provided
correct language exemplars when a child made an error. Parents provided such
"specific evidence" more often than non-parents, but al1 adults tended to respond more
often when the preceding child utterance contained a single error versus multiple
errors of the specific types. Bohannon and Stanowicz sumised that this may be a
result of difficulty in understanding a child's intended meaning in an utterance
wntaining multiple errors. The results of this study illustrate that children receive
information about their grammaticality not only from their mothers, but also from their
fathers and other adults (Bohannon & Stanowia, 1988).
A comprehensive review of the literature by Moerk (1991) noted that whereas
evidence supporthg the existence of differential responding by mothers has corne fiom
several sources, much of the research refuting the existence of feedback to children
has stemmed from the data used by Brown and his associates. Moerk re-examined 20
hours of the interaction of "Eve" and her mother with the goal of presenting clear
Effectiveness of Feedback
examples of the dynamics of materna1 negative or corrective evidence in the data.
Corrections or negative feedback were defined with respect to an antecedent utterance
of the child, in most cases the immediately preceding utterance. Moerk (1991, pp.
224-243) found feedback pertaining to (a) contentives (e.g., nouns and labels, verb
phrases, and determiners); @) functors (e.g., prepositions and bound morphemes); and
(c) syntax (e.g., syntactic rnistakes or other syntactic production problems, and
attempted prep s itional phrases). Thus,in agrezment with Hirsh- Pasek et al. (1984),
Moerkts findings illustrate that Brown's data contained evidence of irnplicit corrective
responding, rather than the explicit responding he demanded, to several aspects of the
children's speech.
The findings of these four studies point to the conclusion that children are
provided with feedback concerning the grammaticality of their utterances. in addition,
several research reports (e.g. Furrow & Moore, 1990; Hirsh-Pasek et al., 1984;
,
have reported on variables which have
Morgan & Travis, 1989; and P e ~ e r 1987)
been found to affect differential responding. in particular, parental responses have
been found to be more contingent on grammaticality when children are younger and
points out that a v e n the longer, more
their language more immature. P e ~ e (1987)
r
grammatical utterances of older children, it is reasonable that they would elicit fewer
parental expansions, and that those that did occur would be l e s linked to grammatical
errors. The utterances of older children would be easier to understand, and so more
topic extensions would ocnir in this group and would again be less linked to the
Effectiveness of Feedback
7
grammaticality of utterances (Pemer, 1987).
In the investigation of variables infiuencing differential respondhg, several
intnguing hdings point to the possibility that it may be the ambiguity of child
utterances, rather than the gxammaticality, which elicits certain types of responses.
Parents do not consciously try to teach children language by controllhg their
responses to child utterances. Pemer (1987) found that, in fact, none of the parents in
her study reported awareness of differential ~spondingbased on grammatidity.
hstead, most reported that they responded in the same way to grammatical and
ungrammatical utterances because their focus was on the content of the communication
rather than on the correctness of their children's speech (Penner, 1987).
Consistent with Penner's hdings, one explanation for why children's utterances
are more often responded to and correcteci when only a single error has occurred than
when multiple errors occur is that it is the communicative ambiguity of child
utterances to which mothers are responding (Bohannon & Stanowia, 1988). Multiple
emrs may make it difficult or impossible to decipher a child's intended meaning, and
thus to paraphrase the child's flawed sentence. This fits with certain features of
current procpssing theories of Language use (e.g. Bock, 1982;cited in Bohamon and
Stanowia, 1988) and language learning (Bates & MacWhinney, 1982;Nelson, 1981;
cited in Bohannon & Stanowia, 1988). Provision of the correct form of children's
speech errors would be ineffective if the same meaning as was the child's intention
were not reflected in the mother's corrective version (Bohannon & Stanowia, 1988).
Effectiveness of Feedback
8
Thus, one might expect that less ambiguous statements (one error) wouid be more
likely to receive a parental response than more ambiguous (multiple errors) statements.
Furrow et al. (1993) empirically tested this alternative hypothesis that it may be
the child's ambiguity as opposed to grammaticality to which mothers respond. The
purpose of their study was to discover whether mothers' response type depends
primarily on the grammaticality or the ambiguity of their children's utterances, and
whether the children's age affects whether the mother responds more to one than the
other. Using three mother-child dyads, mothers' responses to 2- and 3-year-old
children's spontaneous utterances were evaluated. Mothers' response type was found
to be contingent on the ambiguity of the children's utterances, as well as the
grammaticality. Interestingly, ambiguity held opposite effects for the two age groups.
At age 2, childrenrs ambiguous utterances were often followed by no responses on the
mothers' part, while at age 3, mothers more often responded with clarification
questions. Unambiguous utterances that occurred when the children were 2 years of
age were most often responded to with clarification questions by the mother, while at
age 3, they were followed by move-ons.
Furrow et al. hypothesized that varying
levels of arnbiguity at age 2 and age 3 might provide an explanation for the influence
of age on materna1 responding to the ambiguity of children's utterances. Accordhg to
Bohannon and Stanowicz (1988), feedback would be helpful to the child only if the
response made by the mother was appropriate for what the child had intended.
Mothers' response type was also found to be affected by the grammaticality of the
Effectiveness of Feedback
9
children's utterances. Children's utterances that were grammatid were m a t often
followed by exact repetitions, move-ons, and no responses. Those that were
ungrammatical were followed more often by clarification questions and expansions and
recasts. These findings support previous reports of mothers' differential responses to
children's language (Bohannon & Stanowia, 1988; Demetras et al., 1986; Fmow &
Moore, 1990; Hirsh-Pasek et al., 1984;and Penner, 1987) and extend the literature.
In sumrnary, research has provided evidence suppoding the contention that
differential feedback to children's utterances does exist, and that it is given in response
to various aspects of the grammaticality, as well as the ambiguity of children's speech.
Whether or not children are able to utilize the clues provided in matemal feedback to
improve their speech may be less certain. As Morgan and Travis (1989)noted,
support for the contention that language input incorporates negative information
requires not only the existence of differential feedback, but also children's king able
to perceive differences in distributions of adult responses to their grammatical and
ungrammatical speech.
Morgan and Travis (1989)examined the data used by Brown and his associates
for parental responses to infieaional over-regularizations and wh-question auxiliaryverb omission errors. According to Morgan and Travis, these are exemplars of
overgeneralization for which negative evidence is most critically needed. Morgan and
Travis found that expansions and clarification questions occurred more often following
ill-formed utterances. However, these corrective responses formed only a small
Effectiveness of Feedback
10
proportion of al1 adult responses foliowing grammatical errors. In addition, corrective
responding appeared to drop out of children's input while they continued to make
overgeneralization emrs. Morgan and Travis (1989, p. 549) concluded that they had
"failed to find consistent evidence that children distinguished between [maternal]
response types", or wuld recognize such cues. Thus, accordhg to Morgan and Travis,
negative feedback was ineffectual in language leaming.
in his review of the literature, Moerk (1991) found the main argument used in
support of the "no negative evidence" view and in denial of the existence of corrective
feedback to be a denial of the effectiveness of corrections; i.e., even if corrections
exist, they are ineffective for language learning (Platt & MacWhinney, 1983; Arbib,
Conklin, & Hill, 1987; Morgan & Travis, 1989). Despite these claims, Moerk
discovered evidence supporting the effectiveness of corrective feedback dating back to
research done by Brown and Bellugi (1964; cited in Moerk, 1991). in their study,
expansions were reported to aid the acquisition of grarnmar, as "shownby the relation
between the rate of parental expansions and the rate of child language development"
(Moerk, 1991, p. 246). Evidence in support of the effectiveness of corrective feedback
has k e n provided by several researchers throughout recent years.
Reviewing earlier reports on the effectiveness of maternal feedback, Schumaker
and Sherman (1978) found that parental expansions that are contingent on child
utterances had been found to serve as reinforcers as well as models for future child
speech. In
2
study similar to one done previously by Brown and Bellugi (1964; cited
Effectiveness of Feedback
11
in Moerk, 1991), Brown, Cazden, and Bellugi-Kiima (1969) examined the frequency
of occurrence of particular features of matemal speech and correlated these with
children's acquisition of those features. Children whose mothers used possessives
more often than others were found to use possessives more fiequently than other
children. They were also found to use the prepositions most often used by their
mothers correctly 90 percent of the tirne. Therefore children's order of acquisition of
some language forms was correlated with their mother's use of those forms.
Schumaker (1976; cited in Schumaker & Sherman, 1978)) found that expansions
provided by mothers appeared to serve as rnodels for children in that about 24 percent
of the mothers' expansions were imitated by children. Ten percent of these imitations
were exact replications of the mothers' expansions. These results support the finding
that maternal feedback facilitates children's languagc growth.
Scherer and Olswang (1984) also examined the iduence of maternal
expansions on children's imitations. Results of an initial naturalistic study showed that
a positive contingency existed between matemal expansions and children's immediate
spontaneous imitations. Scherer and Olswang surmised that this relationship suggests
that children view expansions of their utterancc as an invitation to continue taiking and
to imitate their mothers' utterance. In an experimental study, Scherer and Olswang
found a systematic relationship between maternal expansions, children's imitations, and
children's spontaneous productions of newly leamed semantic relations. Chi1d.cn
progressed fiom no productions of specific utterances prior to training, to spontaneous
Effectiveness of Feedback
12
imitation following matemal expansions, to spontaneous production of these
utterances. Scherer and Olswang concludecl that results £rom both studies suggested
that the children's initial use of imitation was facilitated by the immediately preceding
expansion.
Fmar (1992) compared 2-year-old chil&ents imitations of new grammatical
morphemes contained in corrective recasts (negative evidence) wi th imitations of
grammatical morphemes contained in positive evidence provided by such discourse
types as noncorrective recasts, topic continuations, and topic changes. Results
indicated that children were more likely to imitate the grammatical morpheme
information contained in corrective recasts thaa imitate the identical information
contained in the other forms of parental discourse. This suggests that children are
differentially sensitive to the grammatical morphemes contained in sources of negative
evidence. Negative evidence appeared to hold more salience than positive evidence
for children.
In a longitudinal study, Furrow and Moore (1990) examined the relationship of
matemal feedback to the grammaticality of children's speech at age 2 with children's
langage development at age 3. ui line with previous results, differential feedback to
the grammaticaiity of the children's utterances was found, both at age 2 and at age 3,
though the amount of feedback given the older children was significantly less than that
given the younger children. Contrary with others' results coaceming the effectiveness
of feedback (Brown & Bellugi, 1964; cited in Moerk, 1991; Brown, Cazden, &
Effectiveness of Feedback
13
Bellugi-Klima, 1969; Schumaker, 1976; cited in Schumaker & Sherman, 1978;
Schumaker & Sherman, 1978; Scherer & Olswang, 1984; Farrar, 1992), this study,
which examineci the effectiveness of matemal feedback for iacilitating young children's
language growth within the p e n d of one year, found no evidence that feedback was
related to language acquisition. Furrow and Moore assessed the effectiveness of
feedback by examining children's speech ai age 3 for evidence of their language
development since age 2 having been affected by maternal differential feedback to the
grammatical aspects of their speech. To accomplish this, a mesure of maternai
sensitivity to grammaticality was computed at each age level, and conelated with
grammaticality and mean length of utterance (MW)at each age and across ages. No
signifiant correlation between feedback and language growth was obtained. However,
as suggested by Furrow and Moore, rather than assessing effectiveness of feedback
through examination of language p w t h over an extended pcriod (such as a year),
examination of more immediate effects of feedback on children's speech may be more
telling.
Boyd (1990) examineci maternai sensitivity to the grammaticality of children's
utterances with the purpose of detemiinhg whether differential feedback had any
immediate effects on the child's use of syntax. Boyd examined 2- and 3-year-old
children's utterances, the following maternal response, and the children's utterance
immediately followiag the matemal response. Her results provided evidence that
children respond differentially to matemal differential feedback, and thus that children
Effectiveness of Feedback
14
are sensitive to differential responding by mothers. Children acknowledged certain
types of differential feedback (clarification questions and extended repetitions) more
often than others.
However, mothers' differential feedback to the grammaticality of
children's initial utterances did not afZect the grammaticality of the children's
subsequent utterances. niough grammatical child utterances more typically followed
maternal clarification questions and move-ons, they did so regardless of the
grammaticality of the initial utterance. Thus, although differential feedback did appear
to be part of children's nahirai language input, Boyd's shidy provided no conclusive
evidence that the environment infiuences children's syntactic development.
In sum, it has been well documented that differential negative feedback is
given by mothers to young children's speech. In addition, researchers have fwnd
some evidence that children are aware of and are able to, in tum, respond to matemal
corrective feedback, though the evidence is mixed on this point. The rationale for the
current study stems fiom the argument in the language development literature which
maintains that even if corrective feedback exisi, it is ineffective in facilitating
children's language leamhg (Plan & MacWhinney, 1983; Arbib et al., 1987). If it
were demonstrated that children were able to use feedback to improve subsequent
speech, further evidence would be gained which might be used to refute the argument
against its effectiveness given in support of the "no negative evidence" view. Thus,
further research examining the effectiveness, or children's use, of differential maternal
feedback is warranted. Several critena for determinhg effectiveness of materna1
Effectiveness of Feedback
15
feedback have been used in the Iiterature. For example, general facilitative effects of
feedback on childrents language measurable over t h e have been examined (e.g.,
Furrow & Moore, 1990). as have b a n more immediate effects (e-g., Boyd, 1990;
Farrar, 1992). The primary purpose of the present study was to investigate the
immediate effectiveness of feedback in improving children's subsequent speech.
To examine the effectiveness of matemal feedback to children's utterances, the
child utterance following the mother's response were examined for evidence of use of
the information contained in the maternai response. If materna1 responding does
provide the child with useful feedback as to the well-formedness of hisher speech
which the child is able to utilize, child subsequent responses should show evidence of
attempts at correcting e m . One might expect, for example, that ungrammatical
child utterances followed by corrective feedback or feedback signalling that an error
has occurred, would in tum be followed by a child utterance whose error in
grammaticality has been conected or in which corrective attempts have been made.
To examine the relationships among the child's initial utterance, the matemal response,
and the child's subsequent response, sequential methods of analysis were used.
Most previous research in the field of language development, and in particular
that examining the availability of feedback to children's speech, has confonned to the
"standard" methods of analysis (i.e., analysis of variance). These methods permit the
examination of the significance of the probability of occurrence of one type of
response over another. However, these methods do not permit the examination of
Effectiveness of Feedback
16
sequences of contingencks. To examine the .effectiveness of m a t e d feedback for the
development of more mature language in young children, a method of analysis which
enables the exploration of contingent matemal responses and child utterances would be
more beneficial. This point is supporteci by Moerk (1991) who suggested that further
research into the effectiveness of differential responding would be aided most by
methods which permit examination of continuous responding.
The method of sequential analysis applied to human interaction asks "if and
how the likelihood of an action by one individual is related to the action of a social
partner" (Moran, Dumas, & Symons, 1992, p. 66). Amrding to Moran et al. (1992),
sequential analysis enables the examination of contingent relations and the question of
process. Sequential analysis was developed as a twl in the study of animal behavior
and ethology, and was first applied to human behavior in the study of early motherinfant interaction (Moran et al., 1992). Sequential analysis has been used extensively
in research studying interactive behavior, such as the behavioral dynamics of early
interactions, marital interactions, and interactions in dysfunctional families.
Sequential methods have also been used in the study of discourse pattern in
adult-child interaction. Scherer and Olswang (1984), for example, used lag sequentiai
methods in their examination of the relationship between children's spontaneous
imitations and mothers' expansions during topic-related episodes. Children's
immediate and delayed imitation of preceding expansions were compared. Their
results showed that children's imitations immediately followed matemal expansions,
Effectiveness of Feedback
17
rather than king delayed. Jose (1988) used the lag sequential technique a s a tool to
abstract a "graxnmar"of speech a d contingency fkom spoken discourse. Studying
interaction between female adults and preschool children, Jose confirmed that most
discourse is based upon t h e fundamental speech act pairings: question-answer,
statement-reply, and directive-acknowledgement. Yoder and Davies (1990) also used
sequential analysis in their study of responding by developmentally delayed children to
parental questions and topic continuations. These authors state that research on
mother-child interaction using sequential analysis indicates that replies of normally
developing children follow prompting or requcsting of verbal information and
continuance of the chilci's topic of conversation a high percentage of the tirne that
these adult utterance types are used (Yoder & Davies, 1990). Thus, scquential
analysis has k e n used successhilly in language development research, particularly in
the sîudy of communicative interaction. However, sequential methods have not been
used within the more narrow field of the study of feedback to children's speech.
The primary goal of the present study was to examine the effectiveness of
feedback for children by lwking at their subsequent responses following materna1
feedback. A secondary goal was to examine developmental trends in ternis of the
effect of the age of the child on his/her ability to use feedback to mend hisher
subsequent speech. This study afso explored the use of sequential analysis in the
study of feedback and its effectiveness. Specific hypotheses were that corrective
matemal feedback to ungrammatical and ambiguous child utterances would result in
Effectiveness of Feedback
18
the chiid's subsequent use of a grammatical or unambiguous (respectively) utterance
immediately following the maternal response, and that older children would be better
able than younger children to successfully repair their utterances based upon maternal
feedback. Loglinear models were used to determine the extent to which matemal
feedback was dependent upon the grammaticality and ambiguity of the child's
utterances, the extent to which children used .effectively maternal feedback to correct
their errors, and whether the age of the child affected hisher ability to make use of
maternal differential feedback.
Method
Subiects
The subjects for this study were four mother-child (daughter) dyads,
videotaped when the children were 2;O years of age (MW 1.27; 2.91; 2.65; 1.54) and
again at age 3;O (MLU 3.81; 3.99; 3.82;4.49). Al1 subjects were members of twoparent, middle-class families.
Procedure
The mother-child dyads were individually videotaped in their homes when the
child was 2;O years of age and again at age 3;O during lunchtirne, while reading
together, and while playing together. Mothers were told that the purpose of the study
was to look at conversations between mothers and children during the course of
normal everyday activities. Transcripts of the mother-chld verbal interaction during
the sessions were prepared. Al1 data for the current study were collected from the 45-
Effectiveness of Feedback
19
60 minute play sessions.
A mutually exclusive and exhaustive set of codes was used to code the
mother-child interaction which occurred during the play session. This was an event-
based cociing system, the events king the actual mother and child utterances and the
the
lapses separahg them. Event-based codllig and analysis were appropnate for
the current research since it is the functional, rather than the psychological, dynamics
of social interaction which were of primary interest (Moran et al., 1992). Each event
(Le., child utterance, matemal response, or tirne lapse) was coded on papa while
viewing the videotaped play session. Lapses of time in which no talk occurred were
included as events as their inclusion enables a more accurate representation of the
interaction between mother and child (Yoder & Davies, 1990). Recording of events
(discrete or mornentary frequency behaviors) was continuous so that no gaps existed in
the data, a necessary condition of sequentiai analysis (Stevenson, Roach, Ver Hoeve,
& Leavitt, 1990). Child utterances were codeci as to their (a) grammaticality; @)
ambiguity; (c) relationship with the previous child utterance; and (d) relationship with
the previous matemal response. Matemal responses were d e d for their relationship
with the previous child utterance in the interaction.
rammaticality. Each child utterance occumng during the play session was
assesseci, according to adult standard English.syntax, as either grammatical or
ungrammatical. The critena used to d e h e grammaticality of utterances were similar
to those used by Boyd (1990) and Furrow and Moore (1990). G r a m t i c a l utterances
Effectiveness of Feedback
20
includeâ (a) syntacticaiiy correct sentences (e.g., 1 want the blue truck), @) directives
(e.g., Stop it), and (c) direct replies to the mother's preceding utterance. Depending
upon the context, the child's direct reply may have been an incomplete sentence (i-e., a
single word or a sentence kagrnent). In order for a single word to be well-fonned, it
had to have been a proper noun, in plural form, or an acknowledgement (e.g., Ys).
-tia
utterances included (a) mistakes in word order (e.g., This 1 want), @)
omissions of an obligatory grammatical morpheme (e.g., Give me truck), and (c) errors
in matching gender, tense, case, or number between subject and verb (e.g., They is
eating ice-cream).
Any other child utterances judged to be ungrammatical by the
rater(s) was also coded as a syntactic e m r . Utterances which were incomplete (cutoff) were coded as k>complete. Utterances which were totally or partially inaudible or
incomprehensible on the videotape were d e d as Jncornprehensible.
b b i g u i t y . Al1 child utterances were judged as either ambiguous or
unambiguous. Arnbiguity referred to the clarity of the child's meaning within a
particular utterance. Utterances which were fully understood were coded as
Unambguous, while those unclear in referent were coded as m-oug
.
As such,
this study examined semantic ambiguity, rather than pragmatic arnbiguity as did
Funow et al. (1993). Based upon a system used by Konefal and Fokes (1984),
ambiguous utterances included those utterances which (a) lacked detail, @) containcd
an empty word (e.g., stuff, thing), (c) contained misarticulations which obscure
semantic meaning, (d) specified a wrong referent, and (e) were spoken while the
Effectiveness of Feedback
21
listener was inattentive. Similar utterances may have been ambiguous in one situation
and unambiguous in another, depending upon the context in which they were made.
in addition, an utterance which was ungrammatical may have k e n unambiguous (Le.,
clear in intent), and a grammatical utterance may have been ambiguous (Le., unclear in
intent). As with grammaticality ooding, utterances which were incomplete (cut-off)
were coded as w l e t e . Utterances which were totally or partially inaudible or
incomprehensible on the videotape were coded as -rehensible.
utter-.
Each child utterance was coded for
its relationship with the previous child utterance. Similar to systems based upon
Bohannon and Stanowia (1988), child utterances were categorized as one of the
following: Repetition, Expansion, Recast/Revision, Topic Continuation, Topic Change,
No Response (2-second interval), No Opporhmity for Response, No Initiation,
Unknown, Response to Matemal Utterance, or No t Contingent. Itepet-
included
word for word reproduction of al1 of the child's preceding utterance, as well as
contracted repetitions (e.g., Chiid: We went to the playground and played; Child: We
went to the playground). b ~ a n s i ~ included
11~
child responses which repeated part or
al1 of his/her preceding utterance, but added new information. Recasts/Revi&
were
coded if the child repeated major elements of hisher preceding utterance with no new
information added, or for child responses that substituted some of the elements in the
- C
child's utterance yet retained the meaning. TpOic
without revision
included al1 child utterances which did not repeat hisher prior utterance and continued
Effectiveness of Feedback
22
the topic of conversation. TOD~C
included al1 child response. which served to
R
change the topic of conversation. The No -
(2-second interval) category
included al1 cases for which a maternal utterance and the following child utterance
were separated by an interval of two seconds or greater. The No Cbpo-
for
included ail cases in which no child response occurred between two
consecutive mother utterances made within two seconds of each other. For cases in
which the child did not initiate verbal interaction following a maternal No Response,
No wtiation was recordeci. Incomprehensible or incomplete child utterances were
coded as Unknown in terms of their relationship to the prior child utterance in the
interaction, as were utterances which followed ones coded as such. m n s e to
was ooded for child utterances which provided a response to the
Matemal Utter-
immediately preceding mother utterance. Finally, a child utterance was d e d as
Conti-
with the previous utterance in the mother-child interaction in cases in
which a break greater than 4 seconds omirred in the interaction, and for cases where
the child utterance followed a maternai failure to initiate conversation after a childts
No Response. M i n g of mntingency enabled indication of occasions when two
consecutive utterances were separated by a lapst in t h e so that their relationship to
one another became questionable.
D
with
responE. Finally, each child utterance was
coded for its relationship with the prior maternai response. Parri passu, codes for this
category were identical to those for coding the relationship of each child utterance
Effectiveness of Feedback
23
with the previous child utterance, discussed above, with the following exceptions. The
category
to -Ufteranrp-
was not utilized. In addition to the
remaining categories, the following were coded. C1;iriocation Ouestions were coded if
the child's response questioned the mother's preceding utterance without requesting
new information. This class included requests for confirmation (e.g., yes/no questions,
repetition with rising intonation, and tag questions), specific queries (question was
aimed at child repeating some part of the utterance), and nonspecific queries (question
was aimed at mother repeating the whole utterance). ~couragers/Acknowled~ents
included instances in which the child's response showed acceptance of the mother's
preceding utterance, but did not continue the conversation (e-g., Hm-hmm). Child
utterances which immediately followed a No Response (2-second interval) or a No
Response Opportunity by the mother were coded as
Followiqg No R e s p m or
nse, respective1y.
M a t m l r e s - m . The type of matemal response to each child utterance was
also assessed. Mother's responses were divided into one of the following classes
according to their relationship with the previous child utterance: Repetitions,
Expansions, Recasts/Revisions, Enmuragers/Acknowledgements, Topic Continuations
without Revision, Topic Changes, Clarification Questions, No Responses (2-second
interval), No Opportunity to Respond, No initiation, Tum Following No Response by
Child, Turn Following No Opporhinity for Response by Child, Unknown, and Not
Contingent. These categories were coded using criteria identical ro those discussed in
Effectiveness of Feedback
relation to the coding of child utterances.
Reliable use of sociaiiy based coding schemes requires specificity with respect
to the behaviors king measured. Coding of child and mother utterances required
observer judgement in discriminating utterance types. An experienced research
assistant was trained in the use of the coding system and interobserver agreement was
calculated using Cohen's kappa. (See Table 1) Reliability ranged fkom 0.71 to 1.00.
For children's interaction, the mean reliability for Grammaticality and Ambiguity was
0.86, for Relationship with Previous Child Utterance 0.84, and for Relationship with
the Revious Maternai Response 0.81. The mean reliability for Matemal Response
was 0.84.
Data for the grammaticality, the ambipity, and the relationship of each child
utterance and the type of each matemal response was entered into a data file. The
created data file was input h t o a speciaily written program which transformed the data
into a format compatible with the data reduction program in preparation for analysis.
Analvsis
The behavior record was examined to.count the presence or absence of
specified sets of adjacent or near adjacent behaviors (Stevenson et al., 1990). As
described by Stevenson et al. (1990, p. 56), for each question of interest, a "target"
was selected as a potential eliciting utterance type and a "criterion" (or "critena") was
selected as a potentiai response. The event(s) following the occurrence of the selected
target utterance type were then examined for the presence or absence of the chosen
Effectiveness of Feedback
Table 1
ReliabiIities (Cohen's Ka~na)for CMd and Mother Interaction Cateaories
Child Utterance
Moîher
Grammaticaiity
Ambiguous
Rel'nship to
Prev. Child
utt .
Rel'nship to
P m . Child
utt.
Rel'nship to
P m . Child
utt .
2 Y-
.79
.79
-83
.83
.85
3 years
-81
-75
-8 1
.7 1
-80
2 years
1.O0
3 Y=
.83
Subjcct
one
h o
three
four
Effectiveness of Feedback
26
critenon response type(s). For example, to test the hypothesis that corrective feedback
to ungrammatical child utterances would result in children's subsequent use of a
grammatical utterance, ungrammaticai child utterances could be specified as the target
behavior, and matemal recasts (for example) and grammatical subsequent child
utterances as the criterion utterance types. This analysis would examine the likelihood
of an ungrammatical utterance followed by a maternal recast (rather than any other
coded materna1 response type) king followed by a grammatical utterance, as opposcd
to any remaining possibility within the Grammaticality category . The nul1 hypothesis
in sequential analysis states that there is no sequential dependence between the
behavior of two interactants (Moran et al., 1992).
Thc tallies resultiug bom counting the occurrence or non-occurrence of targets
and critenons were organized in separate 2 X 2 (for two item sequences) or 2 X 2 X 2
(for thne item sequences) contingency tables for each set of selected target utterances
and critenon responses. The rows represented the presence or absence of the criterion,
and the columns represented the presence or absence of the target response(s). This
process was followed to obtain contingency tables for criterion response types in
relation to target utterance types at negative one, zero, and one lag steps.
Using a log-linear program in the PC version of SPSS/PC+, log-linear
procedures (logit model) were used to determine the direction and statistical
significance of the obtained contingencies between child utterances and maternal
feedback (Bakema.& Gottman, 1986). Essentially, log-linear analysis is the
Effectiveness of Feedback
27
quivalent of an anaiysis of variance (ANOVA) for categorical data. This type of
a d y s i s is based on the assumption that the logarithm of the expected values is a
linear function of variable main effects and interactions (Dixon, 1985). As described
by Stevenson et al. (1990), positive standard parameter estimates (z-values) indicate
an increased likelihood of the target behavior following the criterion, while negative
standard parameter estimates indicate a suppression, or decreased likelihood, of the
target behavior following the criterion.
The primary goal of this report was to examine children's ability to utilize the
matemal feedback afforded them to correct the grammaticality and ambiguity in their
subsequent utterances. It was expected that children would be shown to be able to
make use of this feedback to correct their utterances, and that results would indicate
that children at age 3 were better able than children at age 2 to do so. Research into
maternal responding to children's utterances has demonstrated some fairly consistent
results @ohannon & Stanowicz, 1988;Furrow et al., 1993;Kirsh-Pasek et al., 1984;
Penner, 1987). It was expected that these patterns of results would be confmed.
Results
Frequency counts of each code within each category of the coding system were
tabulated for mother-child interaction across age groups, as well as individually for
children age 2 and 3 years. A total of 3370 interactions were coded, 1624 of which
occurred when the children were 2 years of age, and 1746 at age 3. Within the
Grarnmaticality category, grammatical child utterances comprised 37.5 percent of the
Effectiveness of Feedback
28
total interaction of children, ungrammatical 28.0 percent, incomplete utterances 3.8
percent, incomprehensible or inaudible utterances 4.1 percent, and absent utterances
(no response by the child) 26.5 percent. (See Table 2) Within the Ambiguity
category, unambiguous child utterances accounted for 52.1 percent of the interaction,
ambiguous utterances 13.4 percent, incomplete utterances 3.8 percent,
incomprehensible or inaudible utterances 4.1 percent, and absent utterances 26.5
percent. (See Table 3) The Matemal Response category was subdivided into several
response types. (See Table 4) Matemal topic continuations comprised 15.9 percent of
the interadion, topic changes 14.9 percent, and acknowledgements or encourages 11.2
percent. Additional codes within the Matemal Response category of the coding
system not of interest during analyses (e. g., No initiation, Turn following no
opportunity, No opportunity for response) comprised 36.3 percent of the total materna1
interaction, 7.6 percent of which were utterances which were not contingent upon the
child's preceding utterance and therefore not considered feedback.
Prior research into differential maternai responding to children's utterances has
provided relatively consistent evidence that mothers do respond differentially to the
grammaticality of children's speech (Bohannon & Stanowia, 1988; Furrow et al.,
1993; Hirsh-Pasek et al., 1984; Penner, 1987). Analyses were done to conf~rmthese
fïndings. Materna1 response types were combined to produce six response types
commoniy investigated in the liierature. These hcluded matemal repetitions,
expansions and recasts, move-ons (topic continuations and topic changes),
Effectiveness of Feedback
29
Table 2
Frequencies of Child Utterances within Grammaticality Cateeoxy
2 years
3 YPercentage
Grammatical
Ungrammatical
lncomplcte
Incomprehcnsible
Absent
Percentage
Eff'éctiveness of Feedback
30
Table 3
Freeuencies of Child Utterances within Ambimiitv Cateaorv
-
Frequency
Unambiguous
Ambiguous
(ncomplcte
hcomprchcnsible
Absent
Percentage
Frequency
Percentage
Effectiveness of Feedback
Table 4
Freauencies of Matemal Response T m
2 yean
Response
Rcpctition
Expansion
Rccast
Topic Continuation
Topic Change
No Rcsponse
Acknowlcdgcment
Confirmation Quest
Non - Spccific Qucst
Spccific Quest
-
Non Contingent
Other
Frquency
3 years
Percentage
Frequency
Percentage
Effectiveness of Feedback
,
acknowledgements, no respomes, and clarification questions ( c o ~ a t i o n nonspecific, and specific). Across ags, m a t e d expansions and recasts (z = 7.07, p c
.OS), clarification questions (z= 1.97, p < .OS), and no respomes (Z = 3.66, p < .05)
were found with significantly greater frequency foliowing ungrammatical utterances.
(See Table 5) Altematively, move-ons (2 = 750, p < -05) and acknowledgements (Z
= 6.63, p < -05) occurred with greater probability following grammatical than
ungrammatical utterances. These results appear to be consistent with previous
hdings.
Matemal differential responding to the ambiguity of children's utterances was
also examined. Because previous research suggested that varying levels of arnbiguity
at age 2 and age 3 may trigger contrasting patterns of matemal responding (Furrow et
ai., 1993), feedback to ambiguity was examined individually for children at 2 and 3
years of age. (See Table 6) Expansions and recasts were likely to follow ambiguous
child utterances at age 2 (2 = 2.84, p c -05) and at age 3 (z= 5.43, p < .OS), as well
as unambiguous child utterances at age 2 (2 = 3.77, p < .05) and at age 3 (z= 2.04, p
< .OS). Examination of the coefficients showed that expansions and recasts were more
likely to follow 2 year olds' ambiguous (coeff. = 0.2106) and unambiguous (coeff. =
0.2223) utterances and 3 year olds' ambiguous utterances (coeff. = 0.3650) than 3 year
olds' unambiguous utterances (coeff. = 0.1230). Move-ons followed unambiguous
utterances at age 3 (2 = 6.75, p < .OS) with greater probability. Acknowledgements or
encouragers occurred with increased probability following unambiguous child
Effectiveness of Feedback
33
Table 5
Standardized Scores IZ - Values) and Coefficient for Analyses of Differential Materna1
Respondina (Six Feedback Types) to Grafnmaticality Across Ape Groups
Maternai Respow
Rcpctition
Grammatical
Ungrammatical
Expansion/Rccast
Grammatical
Ungrammatical
-
Movc On
GramrnaticaI
U ngrammatical
Acknowledgmcnt
Grammatical
Ungrammatical
Clarification Question
Grammatical
Ungrammatical
No Response
Grammaticai
Ungrammati cai
*p < .O5
Coefficient
-
Z Value
Effectiveness of Feedback
34
Table 6
St andardized Scores [Z - Values) and Coefficients for Analvses of Differential Matemal
Res~onding(Six Feedback T-ypes) to Arnbimity lndividuali~at 2 Years and 3 Years
M a t e d Rcsponse
-
--
- -
--
-- . -
-
Z Value
Coefficient
--
-
-
-
-
-
-
-
Coefficient
Z - Value
- --
Repetition
Unambiguous
0.099
0.83
0.079
0.73
Ambiguous
0.07 1
0.43
0.199
1.45
Unambiguous
0.222
3.77*
O. 123
2.04*
Ambiguous
0.21 1
2.84*
0.365
5.43*
Unambiguous
0.075
1.41
0,177
6.75*
Ambiguous
0.086
1.40
-0.019
-0.43
Expansioflccast
Move
- On
Effectiveness of Feedback
Table 6 (continued)
Maternai Rcsponse
Coefficient
Acknowledgement
Unarnbiguous
0.268
Ambiguous
O.044
Clarification Ques.
Unambiguous
-0.062
Ambiguous
0-204
No Rcsponse
Unambiguous
-0.004
Amb iguous
0.343
-
Z Value
Coefficient
Z - Value
Effectiveness of Feedback
utterances at age 2 (2 = 5.08, p < .OS) and age 3 (2 = 8.17,p < .OS). Clarification
questions were more likely to follow ambiguous child utterances at botb age 2 (2 =
2.23, p < -05) and age 3 (2 = 4.19,p < .OS). No responses followed with greater
probability ambiguous utterances at age 2 (2 = 4.72, p < .05). No significant effect
was found for matemal repetitions.
To further explore the dynamics of maternai feedback, maternai responses were
categorized into three classes. Corrective matemal feedback included expansions,
recasts, and confirmation clarification questions. Maternal feedback more generaily
signalling the occurrence of an error in the child's previous utterance included
expansions, recasts, confirmation clarification questions, non-specific clarification
questions, and specific clarification questions. Maternal feedback signalling
acceptance of the child's previous utterance included repetitions, topic continuations,
topic changes, ackmwiedgements, and no responses. Analyses were done to
investigate the differential occurrence of these three broad classes of materna1
feedback to the grammaticality and ambiguity of children's utterances across ages.
(See Tables 7 and 8) Results indicated that corrective feedback was significantly more
likely to follow ungrammatical (Z = 7.21, p < .05) and ambiguous (z = 6.61, p < .OS)
child utterances. Feedback signaihg the occurrence of an error significantly more
often followed ungrammatical (z = 6.52, p < .05) and ambiguous (2 = 6.59, p < .OS)
child utterances. Feedback signalhg the acceptance of children's utterances was
significantly more likely to follow grammatical (z= 12.14, p < .OS) and
Effectiveness of Feedback
37
Table 7
Standardized Scores ( Z - Values1 and Coefficient for Analvses of Three Catenories of
Differential Matemal Res~ondinpto Gramrnaticality Across Ane Grou~s
-
Fecdback
Corrective
Grammatical
Ungrammati d
Signal of Error
Grammatical
Ungrammaticai
Signal of Acccptance
GrammaticaI
Ungrammatid
--
Coefficient
-
p
.
-
-
Z Value
Effectiveness of Feedback
38
Table 8
Standardized Scores (2- Values) and Coefficient for Andvses of Three Cateaories of
Differential Matemal Feedback to Ambipity Across Age G~OUDS
-
Feedback
Corrective
Unambiguous
Ambiguous
Signal of Enor
Unambiguous
Ambiguous
Signal of Accçptance
Unambiguous
Ambiguous
--
--
-
Coefficient
-
Z Value
-
Effectiveness of Feedback
39
ungrammaticai (2 = 6.91, p < .Os), and unambipous (z = 15.44, p < .OS) child
utterances. Examination of the associated coefficients show that sipals of acceptance
were more likely to follow gmmmaticai (coeff. = 0.294) and unambiguous (coeff =
0.320) utterances than ungammatical utterances (coeff. = 0.150).
Having provided evidence of the differential occurrence of these three classes
of maternal feedback, their influence on the grammaticality and ambiguity of children's
subsequent utterances was investigated. Influence on the probability of occurrence of
children's no responses (absent utterances) was not investigated, since it was
improvement in grammaticality and ambiguity that was of interest. Across ages (ages
2 and 3 combined), there were few significant influences on the grammaticality of
children's utterances. No significant effect of maternal corrective feedback or
feedback signalling the occurrence of an e m r were found. Results showed a
significant suppression of the likelihood that a grammatical child subsequent response
would follow a grammatical initiai utterance (t = 6.13, p < .OS) given an intervening
maternai signal of acceptance (t = -2.80, p c .05). (See Table 9) Exarnùied
individually at age 2 and age 3, this effect was significant only for 2 year old chüdren.
Grammatical utterances were more likely to be followed by a second grammatical
utterancc (2 = 2.66, p c .OS), however this probability was significantly reduced by the
mother's signal of acceptance of the child's initiai utterance (Z = -2.23, p < .05). (See
Table 10) No significant effects were found for the influence of maternal corrective
feedback or feedback signalling the occurrence of an e m r on the grammaticality of 2
Effectiveness of Feedback
40
Table 9
Standardized Scores (Z - Values) and Coefficient for Analyses of the Effects of Three
Cateaories of Materna1 Feedback to Grarnrnaticalitv Across Aae Groups
Tenn
Squence
-
Gram - Corrective Gram
Child 1 by Child -1 by Morn O
CMd 1 by Child -1
Gram - Corrective - Ungram
ChiId 1 by Child -1 by Morn O
Chiid 1 by Child -1
-
-
Ungrarn Corrective Gram
Child 1 by Child -1 by Morn O
Child 1 by Child -1
Ungram - Corrective Ungram
ChiId 1 by Child -1 by Morn O
-
Child 1 by Child 1
Gram - Sig. Error - Gram
ChiId 1 by Child -1 by Morn O
Child 1 by Child -1
-
Gram - Sig. Error Ungram
Child 1 by Child -1 by Morn O
-
Child 1 by Child 1
-
-
Ungram Sig. Error Gram
Child 1 by Child -1 by Morn O
Child 1 by CbiId -1
Ungrarn - Sig. Error
Ungram
-
Child 1 by ChiId -1 by Morn O
Child 1 by Child -1
Coefficient
-
Z Value
Effectiveness of Feedback
41
Table 9 (continued)
-
Term
Sequence
-
Gram - Sig. Accept Gram
Child 1 by Chiid -1 by Mom O
Child 1 by CMd -1
Gram - Sig. Accept - Ungram
-
Child 1 by Chiid 1 by Mom O
Child 1 by Child - 1
-
-
Ungram Sig. Acccpt Gram
Child 1 by Child - 1 by Mom O
Child 1 by Child - 1
Ungrarn - Sig. Acccpt
Ungram
-
Child 1 by Chiid -1 by Mom O
Child 1 by Child -1
-
-
Z - Value
Effectiveness of Feedback
Table 10
-
Standardized Scores (2 Values) and Coefficients for Analyes of the Effects of Three
Catesories of Maternai Feedback to Grarnmaticalitv individuallv at 2 Years and 3 Years of
Age
Gram - Corrective
Gram
Coefficient Z - Value Coefficient Z - Value
Term
Squcnce
-
Child 1 by Child -1 by
Mom O
Child 1 by Child -1
Grarn - Corrective Ungram
Child 1 by Child -1 by
Mom O
-
Child 1 by Child 1
-
Ungram Corrective
- Grarn
Child 1 by Child -1 by
Morn O
-
ChiId 1 by Child 1
Ungram - Corrective
- Ungram
Child 1 by Child -1 by
Mom O
Child 1 by Child -1
Effectiveness of Feedback
43
Table 10 (continued)
Sequence
Gram-Sig.Enor-
Gram
-
Coefficient Z Value Coefficient Z - Value
Tenn
ChildlbyChild-Iby
Mom O
Child 1 by Child - 1
Gram - Sig. Error Ungrarn
Child 1 by Chiid -1 by
Mom O
Child 1 by Child -1
-
Ungram - Sig. Error - Chdd 1 by Child 1 by
Gram
Mom O
Child 1 by ChiId -1
Ungram - Sig. Error - Child 1 by Child -1 by
Ungram
Mom O
Child 1 by Child - 1
Effectiveness of Feedback
44
Table IO (continued)
Tenn
Gram - Sig. Accept Gram
Child 1 by Child -1 by
Mom O
Child 1 by Child - 1
Gram - Sig. Accept Ungrarn
Chiid 1 by Child -1 by
Mom O
Child 1 by Child -1
Ungram - Sig. Accept Child 1 by Child -1 by
Mom O
- Gram
Child 1 by Child -1
Ungram - Sig. Accept
- Ungram
Child 1 by Child - 1 by
Mom O
Child 1 by Child -1
Coefficient
Z - Value Coefficient Z - Value
followed by
i;fi,
however
:5
intervened, L 11:-
iaI
i:
.i
..a.
feedback s i s .
resulted in
rp !..
nî
a a
.,., ,qential
Feedback sig
!!
relationship t
-''-erit..
fi,!
Results show€
,: ..'
Is ivhile
*
.
Effectiveness of Feedback
45
or 3 year old children's utterances.
In addition to increasing the likelihood of occurrence of interaction sequences,
matemal feedback also reduced the probability of the occurrence of specific sequences
of child utterances. Grammatical initial utterances were significantly likely to be
followed by a grammatical subsequent utterance for children 2 years of age, however
when corrective feedback and feedback signalling the occurrence of an error
intervened, no sequential relationship existed. Likewise, ungrammatical initial
utterances were significantly likely to be followed by ungrammatical subsequent
utterances at age 2, though the intervention by mother with corrective feedback,
feedback signalling the occurrence of an emr, and feedback sipalling acceptance
resulted in no sequential relationship betweea ungrammatical child utterances.
Feedback signalling acceptance of children's initial utterances resulted in no sequential
relationship between consecutive grammatical and ungrammatical utterances, while
without this feedback there was a significant probability that grammatical initial
utterances would be followed by ungrammatical subsequent utterances at age 2, and
vice versa for the combined data.
Severai significant effects were sho& for the innuence of ail three classes of
matemal fecdback on the ambiguity of children's utterances. (See Tables 11 and 12)
Results showed that, without materna1 intervention, unambiguous initial child
utterances were significantly likely to be followed by an unambiguous subsequent
utterance. However, when a matemal signal of acceptance followed an unambiguous
EEectiveness of Feedback
46
Table 1 1
-
Standardized Scores IZ Valuesl and Coefficient for Analyses of the Effects of Three
Categones of Matemal Feedback to Arnbiauiîy Across Atze Grou~s
Sequence
Unambig - Corrective - Unambig
Tenn
Child 1 by Child -1 by Morn O
-
Child 1 by Child 1
Unamb ig - Corrective - Ambig
Child 1 by Child - 1 by Morn O
-
Child I by Child 1
-
Arnbig Corrective - Unarnbig
Child 1 by Child -1 by Morn O
Child 1 by Child - 1 by Mom O
Child 1 by Child 1
Ambig - Corrective - Ambig
Child 1 by Child - 1
Unambig - Sig. Error - Unambig
Child 1 by Child -1 by Morn O
Child 1 by Child -1
Unambig - Sig. Error - Ambig
Child 1 by Child -1 by Morn O
Child 1 by Child -1
Ambig - Sig. Error - Unambig
Child 1 by Child -1 by Mom O
Child 1 by Child - 1
-
-
Ambig Sig. Error Arnbig
Child 1 by Child -1 by Mom O
Child 1 by Child -1
Coefficient
Z - Value
Effectiveness of Feedback
47
Table 1l (continued)
Tem
Scquence
Unambig - Sig. Acccpt - Unambig
Child 1 by Child -1 by Mom O
Child 1 by Child -1
-
-
Unambig Sig. Accept Ambig
Child I by Chiid -1 by Morn O
Child 1 by Child - 1
Ambig-Sig-Acçept-Unambig
ChiIdlbyChild-IbyMomO
-
Child 1 by Child 1
Ambig - Sig. Accept - Ambig
Child 1 by Child -1 by Mom O
Child 1 by Child -1
Coefficient
Z - Value
Effectiveness of Feedback
48
Table 12
-
Standardized Scores [Z Values1 and Coefficient for Analyses of the Effects of Three
Cateeories of Matemal Feedback to Ambimitv Individuallv at 2 Years and 3 Years of Ape
Unambig - Corrective Child 1 by Child -1 by
Morn O
- Unambig
Child 1 by Child - 1
-
-
Unambig Corrective Chiid 1 by Child 1 by
Mom O
- Ambig
Child 1 by Child - 1
Ambig - Corrcctive
Unambig
-
Child 1 by Child -1 by
Mom O
-
Child 1 by Child 1
Ambig - Corrcctive Ambig
Chiid 1 by Child -1 by
Morn O
Child 1 by Child -1
Effecûveness of Feedback
49
Table 12 (continued)
-
Coefficient Z VaIue Coefficient Z - Value
Scqucnce
Tenn
Unambig - Sig. Error
- Unambig
Child 1 by Child -1 by
Morn O
ChiId 1 by Child -1
Unambig - Sig. Error
- Ambig
Child 1 by Child - 1 by
Mom O
Child 1 by Child -1
Arnbig - Sig. Error
Unambig
-
Child 1 by Child - 1 by
Mom O
-
Child 1 by Child 1
-
Ambig Sig. Error
Ambig
-
Child 1 by Child -1 by
Mom O
Child 1 by Child - 1
Effectiveness of Feedback
50
Table 12 (continued)
Term
Squcnce
Unarnbig - Sig.
Child 1 by Child - 1 by
Accept Unambig
Mom O
-
Child 1 by ChiId -1
-
Unambig Sig.
Acccpt - Ambig
Child 1 by Child -1 by
Mom
O
Child 1 by Child -1
Ambig - Sig. Accept
- Unambig
Child 1 by Child -1 by
Mom O
Child 1 by Child -1
Ambig - Sig. Accept
- Ambig
Child 1 by Child -1 by
Mom O
ChiId 1 by Child -1
-
Coefficient Z Value
Coefficient Z - Value
Effectiveness of Feedback
51
initial child utterance, the likelihood that the subsequent utterance would also be
unambiguous was significantly reduced (z= -3.82, p < .OS). Investigation of this
effect showed signifiace oniy at age 3; the likelihood of the occurrence of an
unambiguous subsequent utterance following an unambiguous initial utterance (Z =
3.72, p < .OS) was significantly reduced by a materna1 signal of acceptance following
the initial utterance (Z= -2.97, p < .OS).
With respect to ambiguous initial child utterances, no sequential ~lationship
existed between the occunence of an ambiguous initial utterance and an unambiguous
subsequent utterance (z = -0.70).
However, when a matemal signal of emor followed
an ambiguous initial utterance, the l i k e l i h d that the subsequent child utterance would
be unambiguous was significantly reduced
(z = -2.26, p < .OS).
Analysis of this effect
individually at age 2 and age 3 showed a signifiant effect only at age 3. At age 3,
ambiguous utterances were not likely to be followed by an unambiguous subsequent
utterance (2 = -1.99, p < .OS); a matemal signal of e m r following an ambiguous
initial utterance further reduced the likelihood of the occurrence of an unambiguous
subsequent utterance (2 = -2.48, p < .OS). Finally, no sequential relationship existed
between an ambiguous initial utterance and an ambiguous subsequent utterance (Z= -
1.05); however the interjection of a maternal signal of acceptance following an initial
ambiguous utterance reduced the likelihood of the occurrence of an ambiguous
subsequent child utterance (2 = -2.26, p < .OS).
As with gramrnaticality, matemal feedback to ambiguity also reduced the
Effectiveness of Feedback
52
probability of the OCCUTrence of specific sequences of child utterances. Unambiguous
initial utterances were signincantly Likely to be followed by a unambiguous subsequent
utterance for children, however when corrective feedback (for 2 year olds) and
feedback signalling the occurrence of an error (for 2 and 3 year olds) intervened, no
sequentiai relationship existed. Likewise, ambiguous initial utterances were
significantly iikely to be followed by ambiguous subsequent utterances at age 2,
though the intervention by mother with corrective feedback, feedback signalling the
occurrence of an error, and feedback signalling acceptance resulted in no sequential
relationship between ambiguous child utterances.
Discussion
The primary goal of this study was to investigate whether children are able to
use materna1 feedback to correct the grammaticality and ambiguity of their speech. It
was expected that children would use the information contained in materna1 feedback
to correct or improve the quality of their subsequent utterance following the provision
of feedback by the mother. Though results indicate that differential materna1 feedback
is provided to both grammaticality and ambiguity, they do not provide definitive
evidence that feedback results in children's subsequent use of a grammatical or
unambiguous utterance.
The results show that corrective feedback, feedback signalling error, and
feedback signalling acceptance are provided differentially in response to both the
grammaticality and the ambiguity of children's utterances. Matemal corrective
Effectiveness of Feedback
53
feedback and feedback more generally signaiiing the 0cCuffenc-e of an error in the
child's previous utterance are contingent upon ungammatical and ambiguous child
utterances. Matemal feedback signalling the acceptance of children's speech more
often occurred following grammatical and unambiguous utterances. This illustrates
that mother responses which follow ungrammatical or ambiguous child utterances are
likely to provide negative evidence of incorrect or ambiguous forms, while responses
which follow grammatical or unambiguous utterances are l e s likely to provide such
information. Accordhg to Marcus (1993), negative evidence can occur only if parents
are in some way sensitive to the form of their children's utterances. These results,
therefore, provide support for the contention that mothers are sensitive to the form of
their children's speech and provide feedback to their children accordingly. This
opposes Brown and Hanlm (1970) who maintained that approval and disapproval were
linked with mith value rather than syntactical accuracy.
Matemal feedback seems to have little effect on the grammaticality of
children's subsequent utterances at age 2 or age 3. No significant results were found
for corrective feedback or feedback signalling error. For 2 year old children, feedback
indicating matemal acceptance of a child's initial grammatical utterance seems to
suppress the likelihood that a grammatical subsequent child utterance will follow. It
may be that 2 year old children are l e s likely to produce grammatical utterances in
general, but that when they do, their mother responds by showing acceptane.
More significant effects were found for the ifluence of materna1 responding to
Effectiveness of Feedback
54
the ambiguity of children's utterances than to the gra~nmaticality. These results show
that when corrective feedback or feedback signalling an e m r are given iollowing an
unambiguous utterance, the result is a diminished likelihood that the subsequent
utterance will be unambiguous. Feedback of this nature (corrective or signailing error)
offered following an unambiguous utterance may actually conhise the child, possibly
resulting in the child's no response, for instance, following this episode. in addition,
results indicate that unambiguous utterances followed by a materna1 signal of
acceptability are l e s Likely to be followed by a subsequent unambiguous utterance, but
only at age 3, and that the likelihood of an ambiguous utterance following a matemal
signal of acceptability is increased. It may be that the child, given an indication that
hisher previous utterance was acceptable, moves on to a new topic of conversation
which results in a more ambiguous utterance.
Results also show that ambiguous child utterances followed by a maternai
signal of e m r are less likely to be followed by a subsequent unambiguous child
utterance. It may be that children do not fully correct the ambiguity of their previous
utterance following a signal of error. Instead, they may repeat the correction offerd
by their mother or correct only the portion of their utterana for which feedback was
given, resulting in a second ambiguous utterance.
in sum, the results of this study show that maternai responding does affect
children's utterances. Interestingly, the influence of matemal feedback on the
grammatiality and ambiguity of children's subsequent utterances appears to be
Effectiveness of Feedback
55
consistent1y in the opposite direction t O that hypothesized. However, children's no
responses (absent utterances) following feedback were not examined. A s they
comprise 30.2 percent of the children's interaction at age 2 and 23.0 percent at age 3,
it rnay be that they play a major mle in the pattem of results found. Examination of
the effect of matemal responding on the probability of the occurrence of a subsequent
no response by children would be interesting. It may be that given feedback, children
becorne contemplative which results in a subsequent no response. None the l e s , the
results of this study would seem to suggest that matemal feedback is used by children
in a consistent manner to facilitate their acquisition of language. To examine more
specifically the way in which children respond to matemal feedback, it would be
useful to investigate the relation of the child's subsequent response to their initial
utterance and to the preceding matemal response. insufficient data were available in
this study to examine these categories of response.
Results of this study support earlier research hdings with respect to
differential materna1 responding to both grammaticality and ambiguity. In fact, these
results not only replicate previous findings, but strengthen the body of evidence
supporthg differential responding by mothers. Because sequentiai analysis requins a
mutually exclusive and exhaustive cocüng system, analyses compare the occurrence of
a single code against the occurrence of the remaining codes within a specific category,
as opposed to methods used in earlier research which simply compare the occurrence
of two selected codes within a category. As such, these analyses not only compare
Effectiveness of Feedback
56
maternai responding to grammatical and ungrammatical (or unambiguous and
ambiguous) child utterances, but rather grammatical versus al1 other possible codes
within the grammaticality category, and ungrammatical vernis al1 other possible codes
within the grammaticality category. Revious studies have examined mother-child
interaction more as a dichotomy, ignoring aspects of interaction which may M u e n c e
responding by both mother and child. Sequential analysis thus provides a more
wmplete picture of the nature of the interaction between mother and child.
Furthemore, since the occurrence of a single code is compared with the
occurrence of several remaining codes, much more "noise" is introduced into the
analyses, reducing the probability that a code will be found to occur significantly more
often within the interaction. Thus, using this analysis procedure, an effect must have
greater strengh to reach statisticai significance. The results of this study show that
children are provided with differential feedback which occurs fiequently and
consistently enough within the context of the entire interaction that it still attains
statistical significance. Thus, there is not so much "noise" in mother-child interaction
that differential responding becornes lost.
In terms of the dynamics of mother-child interaction, the types of cornparisons
made using sequential analysis would appear to more closely resemble the actuai
conversational environment. interaction between mother and child consists of the
occurrence of grammatical and ungrammaticai, unarnbiguous and ambiguous utterances
within the context of the conversation as a whole. Verbal interaction also includes, for
Effectiveness of Feedback
57
example, incomplete and incompreheasible utterances as well as lapses of time in
which no rrsponses are made by the interadants. To respond differentially to the
g m a t i c a l i t y and ambigity of the child's utterances, the mother must separate these
aspects from other attributes of the child's interaction. Likewise, for children to use
information containeci in maternal feedback, they must be able to discem pertinent
feedback from other aspects of the mother's speech, then apply the information toward
the improvement of theû own utterances.
The use of sequential methods in this study permit the exploration of the
applicability of these methods to the study of matemal negative feedback and its
effectiveness for child language development. One potential drawback to the use of
sequential methods in this area may be the amount of data which may be required to
provide signifiant results. For each parameter specified for analysis, the amount of
data required to provide adequate ce11 counts inmeases. Though this factor is a crucial
one in studies using small data sets, it should not pose diffinilty for larger
investigations. The benefits of sequential methods in terms of analysis seem to
outweigh the limits it imposes.
A potential limitation of this study may exist in the use of assessrnent of
children's ability to use maternal feedback to correct their subsequent utterances to
judge the effectiveness of matemal feedback in the faditation of language
development. This method does not allow the examination of the possibility that
children's response to maternal feedback may be delayed beyond the subsequent
Effectiveness of Feedback
58
response diredly following the feedback (i. e., children may not respond, resulting in
an absent utterance subsequent to feedback), or of cumulative effects of maternal
feedback on children's language development. However, previous research (Furrow &
Moore, 1990) did not find evidence of the effectiveness of differential matemal
responding to child speech by correlating maternal responding with the children's
language level at age 3 as compand to age 2. As Fu-
and Moore suggested, other
mediating factors which codd influence the effects of feedback on language
development over such an extended penod as a year may have affected the results of
their study. Thus, examination of the immediate effects of corrective feedback on
children's grammaticality and ambiguity seemed warranted.
A sccond limitation of this study may exist in the generality of the coding
system used, particularly with respect to the categories used to code matemal response.
Using the current system, maternal responses were coded in terms of their general
relationship with the preceding child utterance. For example, if a matemal response
were coded as having been a recast of the preceding child utterance, the recorded code
indicated only that a matemal recast occurred. This system assumed that mothers
signal occurrence of emrs in a general sense, which provides the child with the
opportunity to
coma his/her following utterance. The use of a more specific coding
system which indicata the specific portion of the child utterance k i n g recast (e. g.,
verb, subject, object) may provide more detailed information as to how mothers
facilitate language learning. More detailed coding might also provide a c l w e r view
Effectiveness of Feedback
of the way in which children use matemal feedback provided them in their daily
interactions for language p w t h . However, the use of a very specific coding system
will require a larger amount of data for analysis.
Zn conclusion, results of this study provide strong evidence for differential
matemal responding, within the context of interaction as a whole, to the
gra~nmaticaiityand ambiguity of 2 and 3 year old children's utterances. interestingiy,
matemal feedback seemed to have a greater influence on ambiguity than
grammaticality. However, no evidence was found to show that children correct the
grammaticality and ambiguity of their subsequent responses immediately following the
provision of feedback. As suggested earlier, i n addition to examining the
grammaticality and ambiguity of the child's response following feedback, it may be
helpful to investigate the relation of the childfs subsequent response to his/her initial
utterance and to the motherfs response to l e m more about the effectiveness of
matemal differential feedback for language development.
Effectiveness of Feedback
60
References
Arbib, M. A., Conklin, E. J., & Hill, J. C.(1987). From m
a theory to -.
New York: Oxford University Press.
. .
Baiceman, R., & Gottman, I. M. (1986). Obse-eradn:
m u c t i m
analvsk. New York: Cambridge University Ress.
Bohannon, J. N. & Stanowia, L. (1988). The issue of negative evidence: Adult
rPsponses to children's language errors. R e v e l o ~ c h o h g y 2,40.,
684-689.
.
Boyd, G. (1990). me sinnificance of maiemal feedback on childrms
*
1
-
. .
developmea. Unpublished master's thesis, Dalhousie University, Halifax, Nova
Scotia.
Brown, R., Cazden, C., & Bellugi-Klima, U. (1969). The child's grammar from 1 to
m. In J. P. Hill (Ed.), Minnesota s m s i a on child psvchology,
Vol. II @p.
28-73). Minneapolis: University of Minnesota Ress.
Brown, R. & Hanlon, C. (1970). Denvational cornplexity and the order of acquisition
in child speech. In J. R. Hayes (Ed.), -on
and the d e v e l ~ e n of
t
@p. 11-53). New York: John Wiley & Sons, Inc.
Demetras, M. J., Post, K. N., & Snow, C. E. (1986). Feedback to f k t language
Leamers: The role of repetitions and clarification questions. J o u d of
275-292.
Dixon,'W. J. (Ed.) (1985). BMPD statistid software. Berkeley: University of
Effectiveness of Feedback
61
California Press.
Farrar, M. 1. (1992). Negative evidence and grammatical morpheme acquisition.
Developmental
280, 90-98.
Furrow, D., Baillie, C., McLaren, J., & Moore, C. (1993). Differential responding to
two and three-year-old's utterances: The roles of grammaticality and
of 363-375.
ambiguity.
Furrow, D., & Moore, C. (1990,July). Mothers' feedback to children's utterances: A
u t u d i n a l study. Paper presented at the 5th International Congress for the
Study of Child Language, Budapest, Hungary.
Hkh-Pasek, K., Treiman, R., & Schneiderman, M. (1984). Brown & Hanlon
revisited: Mothers' sensitivity to ungrammatical forms. Journal of C m
81-88.
Jose, P. E. (1988). Sequentiality of speech acts in conversational structure. Journal of
istic Research, 17(1b 65-88.
Konefal, J. A., & Fokes, J. (1984). Linguistic analysis of children's conversational
. .
repairs. J o u d of P s y c h o ~ i s t i cResearch, 13(1), 1-11.
Marcus, G. F. (1993). Negative evidence in language acquisition. Cognition, 46 5385.
Moerk, E. L. (1991). Positive evidence for negative evidence. First
a
219-251.
Moran, G., Dumas, J. E., & Symons, D. K. (1992). Approaches to sequential analysis
Effectiveness of Feedback
62
and the description of coniingency in behavioral interaction. J3ehaviod
&s-
65-92.
Morgan, J. L. & Travis, L. L. (1989). Limits on negative information in language
input.
Jounial of 531-552.
P e ~ e r S,. G.(1987). Parental responss to grammatical and ungrammatical child
utterances.
D e v e l o m 58, 376-384.
Platt, C. B., & MacWhinney, B. (1983). Error assimilation as a mechanism in
language learning. Journal of Child- 1
401-414.
Scherer, N. J., & Olswang, L. B. (1984). Role of motherst expansions in stimulating
chiidrents language production. Journal of Smech
and Hean'wResearch, 27,
387-396.
Schumaker, I. B., & Sherman, 1. A. (1978). Parent as intervention agent: From birth
onward. In R. L. Schiefelbusch (Ed.),
intervention stratehd= @P.
237-3 16). Baltimore: University Park Press.
Stevenson, M. B., Roach, M. A., Ver Hoeve, J. N., & Leavitt, L. A. (1990). Rhythm
in the dialogue of infant feeding: Retexm and term infants. mant Rehavia
and Develo~m-
a 51-70.
Yoder, P. J., & Davis, B. (1990). Do parental questions and topic continuations elicit
replies bom developmentally delayed children?: A sequential analysis. J o u d
pf Speech and
Research, 33,563-573.