Comparing Child Outcomes of Physical Punishment and Alternative Disciplinary Tactics - A...
Transcript of Comparing Child Outcomes of Physical Punishment and Alternative Disciplinary Tactics - A...
-
7/29/2019 Comparing Child Outcomes of Physical Punishment and Alternative Disciplinary Tactics - A Meta-Analysis
1/38
Clinical Child and Family Psychology Review, Vol. 8, No. 1, March 2005 ( C2005)
DOI: 10.1007/s10567-005-2340-z
Comparing Child Outcomes of Physical Punishment and
Alternative Disciplinary Tactics: A Meta-Analysis
Robert E. Larzelere1,2 and Brett R. Kuhn1
This meta-analysis investigates differences between the effect sizes of physical punishmentand alternative disciplinary tactics for child outcomes in 26 qualifying studies. Analyzingdifferences in effect sizes reduces systematic biases and emphasizes direct comparisons be-tween the disciplinary tactics that parents have to select among. The results indicated thateffect sizes significantly favored conditionalspanking over 10 of 13 alternative disciplinary
tactics for reducing child noncompliance or antisocial behavior. Customary physical punish-ment yielded effect sizes equal to alternative tactics, except for one large study favoring phys-ical punishment. Only overly severe or predominant use of physical punishment comparedunfavorably with alternative disciplinary tactics. The discussion highlights the need for betterdiscriminations between effective and counterproductive use of disciplinary punishment ingeneral.
KEY WORDS: children; parenting; discipline; punishment; spanking.
Uncertainty about the effects of physical pun-ishment on children has persisted despite decadesof research. Two major perspectives have emergedrecently. The first is an unconditional anti-spanking
perspective, advanced by both social scientists(Gershoff, 2002; Straus, 2001) and advocacy groups(EPOCH-Worldwide, 2004). In response, at least 13countries have passed laws banning all physical pun-ishment by parents (EPOCH-Worldwide, 2004).
The second perspective, which has been calledthe conditional-spanking perspective (Benjet &Kazdin, 2003), has attempted to identify condi-tions under which spanking may be beneficial or atleast not detrimental to children. The conditional-spanking perspective emphasizes the parenting con-text and manner of implementation, which may
distinguish effective from counterproductive usesof punishment more than its form (e.g., physicalor nonphysical). In one sense the disciplinary ac-
1Psychology Department, Munroe-Meyer Institute, University of
Nebraska Medical Center, Nebraska.2Address all correspondence to Robert E. Larzelere, Psychology
Department, MMI, 985450 Nebraska Medical Center, Omaha,Nebraska 68198-5450; e-mail: [email protected].
tions of parents in most cultures have reflected aconditional-spanking perspective until recently. In19941995, for example, 94% of American parentsand 52% of Canadian parents of 3- and 4-year-olds
reported using physical punishment at least occasion-ally (Larzelere, 2004; Straus & Stewart, 1999). Theconditional-spanking perspective holds that spank-ing should be investigated under the conditions forwhich parents have considered it advisable beforeimposing a spanking ban on parents (Bauman &Friedman, 1998; Baumrind, Larzelere, & Cowan,2002; Eysenck, 1993; Friedman & Schonberg, 1996b;Larzelere, Baumrind, & Polite, 1998).
Two recent literature reviews from these twoperspectives did little to resolve the issue. Gershoffs(2002) meta-analysis concluded that physical pun-
ishment was linked positively to immediate com-pliance, but negatively with 10 other outcomesin children and families. In a qualitative review,Larzelere (2000) concluded that causal evidenceshowed that nonabusive spanking of 26-year-oldsproduced more beneficial than detrimental child out-comes when it was used to enforce milder disci-plinary tactics such as reasoning or time-out, es-pecially in subcultural groups that support its use.
1
1096-4037/05/0300-0001/0 C 2005 Springer Science+Business Media, Inc.
-
7/29/2019 Comparing Child Outcomes of Physical Punishment and Alternative Disciplinary Tactics - A Meta-Analysis
2/38
2 Larzelere and Kuhn
Benjet and Kazdin (2003) recently compared thetwo reviews and concluded, A top priority for re-search on spanking would seem to be a comparisonof spanking with alternative procedures that alreadyhave considerable evidence in their behalf (p. 215).The current meta-analysis attempts to address thispriority by investigating the studies included in eitherreview that examined one or more alternative dis-ciplinary tactics in addition to physical punishment.It also investigates several methodological problemsthat could explain the discrepant conclusions fromthe two reviews (Benjet & Kazdin, 2003).
To provide a context for this meta-analysis, webriefly summarize the methodological problems thathave hindered definitive conclusions about physicalpunishment. We then clarify why a meta-analysisusing differences in effect sizes between physicalpunishment and disciplinary alternatives can reduce
these methodological problems.
Methodological Issues
The spanking controversy persists largely be-cause pervasive methodological problems have per-mitted a wide range of interpretations. These prob-lems include predominantly correlational research;failing to discriminate among nonabusive, customary,and overly severe use of physical punishment; mea-suring disciplinary practices and child outcomes fromthe same information source; and failing to rule out
plausible alternative explanations.The strongest evidence against physical punish-
ment in Gershoffs (2002) thorough meta-analysisconsisted of longitudinal correlations, i.e., zero-ordercorrelations between physical punishment and sub-sequent child outcomes. Although such correlationsare consistent with a causal effect (Smith, 2002),their pattern is typical of most corrective interven-tions (Larzelere, Kuhn, & Johnson, 2004). In post-treatment comparisons, recipients of corrective inter-ventions will compare poorly to those not needingsuch interventions, whether the intervention is deliv-
ered by physicians (e.g., radiation treatment), educa-tors (Head Start), psychologists (marital counseling),or parents (punishment).
Consider radiation treatment as an example. Pa-tients who received radiation treatment last year aremore likely to experience cancer-related symptomsthis year than those who did not receive (or need)radiation treatment. Longitudinal zero-order corre-lations would indicate that radiation treatment is
associated with increased cancer-related symptoms.Of course, the initial presenting problem (cancer)is the causal factor underlying that correlation be-cause it leads to both the corrective intervention(radiation treatment) and the subsequent outcome(cancer). Consequently, zero-order longitudinal cor-relations cannot discriminate effective corrective in-terventions from those that are counterproductive.
Second, most of the research on physical pun-ishment lumps together nonabusive and custom-ary punishment with overly severe forms of physi-cal punishment. For example, 65% of the studies inGershoffs (2002) meta-analysis included overly se-vere physical punishment in their measure, accord-ing to Baumrind et al. (2002). Examples ranged fromvaguely defined punitive discipline (6% of thestudies), composite measures of the frequency andseverity of physical punishment (29%), and the in-
clusion of extreme violence (31%), such as slappingin the face (seven studies), beating up (three studies),or hitting with a fist and causing bruises and cuts (onestudy).
Third, many studies of disciplinary tactics havebased the antecedent and consequent variables onthe same source of information. Typically, moth-ers reported both their disciplinary tactics and theirchilds behavior. In retrospective studies, grown chil-dren reported both their current functioning and thedisciplinary tactics they received earlier in life. Thissame-source bias has been shown to inflate associ-ations between disciplinary tactics and adverse out-
comes (Yarrow, Campbell, & Burton, 1968).Finally, plausible alternative explanations of the
data on physical punishment have not been ruledout, resulting in widely discrepant explanations forthe varied outcomes across studies. Consider thestrongest evidence of the effectiveness of spanking.Four small-randomized clinical studies found thatspanking was effective in reducing defiance in clin-ically oppositional 26-year-olds (Bean & Roberts,1981; Roberts; Day & Roberts, 1983; Roberts, 1988;Roberts & Powers, 1990). The difference in effectsizes between those four randomized studies (mean
d = 1.21) and the 113 non-randomized studies (meand = .35) in Gershoff (2002) approached the largestdifference ever found in a meta-analysis (Lipsey &Wilson, 1993). This difference could be explained byone or more of the following confounded interpre-tations. Compared to the non-randomized studies,Roberts four randomized studies (1) had causallystronger evidence, (2) limited spanking to two open-handed swats under the supervision of a clinical
-
7/29/2019 Comparing Child Outcomes of Physical Punishment and Alternative Disciplinary Tactics - A Meta-Analysis
3/38
Physical Punishment vs. Alternative Tactics 3
psychologist, (3) used spanking only to enforce com-pliance with time-out, (4) applied only to childrenfrom 2 to 6 years of age who (5) were clinically re-ferred for oppositional behavior problems, and (6)focused on decreases in defiance in the clinic as theprimary outcome. Whereas advocates of the anti-spanking viewpoint consider the type of outcome(short-term compliance) to be the crucial distinction(Gershoff, 2002; Straus, 2001), conditional-spankingresearchers emphasize the stronger causal evidence,the specific conditions in the randomized studies(e.g., the childs age, the discipline situation), and theway in which spanking was implemented (Baumrindet al., 2002; Larzelere, 2000).
Although these four methodological problemsare often acknowledged, the extent to which theyundermine research conclusions has received insuf-ficient attention. Suppose radiation treatment were
studied in the same way that researchers have investi-gated physical punishment. Borrowing statistics fromGershoffs (2002) thorough meta-analysis, two-thirds(65%) of studies of radiation treatment would haveincluded excessive dosages of radiation (Baumrind etal., 2002), 58% would have been cross-sectional stud-ies, and only 4% would have taken into considera-tion the presence or severity of cancer. Would it besurprising that patients who received radiation treat-ment last year had higher rates of cancer both lastyear and this year, compared to those who did notreceive (or need) radiation? A meta-analysis of ra-diation treatment using predominantly correlational
studies would come to the same conclusions as Ger-shoffs (2002) meta-analysis, specifically that radia-tion treatment is consistently linked to detrimentaloutcomes. As aptly noted by Straus (2001), validcausal conclusions require controlling for the effectsof initial child misbehavior. Otherwise, initial childmisbehavior may lead to more disciplinary tactics aswell as worse child outcomes, which would accountfor the associations found by Gershoff (2002).
Rationale for a Meta-Analysis
of Differential Effect Sizes
This meta-analysis attempts to reduce these per-vasive methodological problems by (1) distinguishingamong four types of physical punishment, (2) basingeffect sizes on each studys strongest methodologicalevidence whenever possible, and (3) analyzing differ-ential effect sizes between physical punishment andalternative disciplinary tactics.
To address the lumping problem, we distin-guish among conditional spanking, customary phys-ical punishment, overly severe physical punishment,and predominant use of physical punishment. Con-ditional spanking (as labeled by Benjet & Kazdin,2003) refers to spanking under the limited condi-tions that have been associated with better child out-comes (e.g., spanking when a 26-year-old refuses tocomply with time out). The purpose of distinguish-ing this category is to determine whether spankingis associated with better outcomes than alternativetactics even under ideal conditions. Customary phys-ical punishment represents the manner in which par-ents typically use physical punishment. The purposeof this category is to investigate whether typical useof physical punishment is associated with better orworse outcomes than alternative tactics. Overly se-vere physical punishment includes the use of exces-
sive force, hitting with an object, or slapping in theface (Baumrind et al., 2002). Finally, predominantus-age indicates that physical punishment is the parentsprimary disciplinary method, i.e., it is preferred overmilder disciplinary tactics.
This meta-analysis bases effect sizes on thefindings from each study that are methodologicallystrongest. For example, our effect sizes are basedon results that take initial child misbehavior into ac-count from distinct sources of information, when-ever possible. This choice contrasts with Gershoffs(2002) decision to base effect sizes on correlationsfor the sake of consistency, ignoring methodologi-
cally stronger findings in several studies.Finally, this meta-analysis estimates differences
in the effect sizes of physical punishment vs. alter-native disciplinary tactics, using identical methodswithin the same study. If the apparently detrimen-tal child outcomes reflect causal effects unique tophysical punishment, then the effect sizes of physi-cal punishment should compare poorly to the effectsizes of alternative disciplinary tactics. On the otherhand, if detrimental child correlates of physical pun-ishment represent methodological artifacts, then theeffect sizes of alternative disciplinary tactics should
appear equally detrimental.A methodology for analyzing differences be-tween effect sizes is already well established for ran-domized studies. It is based on the differential ef-fect size contrasting post-treatment outcomes from atreatment and a control group. For the usual effectsize measure (d), this is the same as calculating aneffect size for each group (e.g., improvement frompre-to-post) and then using the difference between
-
7/29/2019 Comparing Child Outcomes of Physical Punishment and Alternative Disciplinary Tactics - A Meta-Analysis
4/38
4 Larzelere and Kuhn
those two effect sizes. This equality is based on twoassumptions. First, the treatment and control groupmust have identical pre-treatment scores, which ran-domization guarantees in the long run.3 The secondassumption is that the effect sizes for the treatmentand control groups are based on the same stan-dard deviation. When these assumptions apply, typi-cal meta-analyses of randomized clinical trials can beconsidered equivalent to analyses of differences be-tween effect sizes. The only distinction in our meta-analysis is that it compares two treatments (disci-plinary tactics) with each other rather than treatmentand control groups.
The major advantage of analyzing differencesbetween effect sizes, however, is for non-randomizedstudies, which dominate this literature. Causal con-clusions can be supported from correlational stud-ies only to the extent that plausible alternative in-
terpretations have been ruled out (Larzelere et al.,2004; Shadish, Cook, & Campbell, 2002). This prin-ciple applies to individual studies as well as to meta-analyses. For example, an alternative explanation forthe positive correlation between physical punishmentand subsequent antisocial behavior is that the childsinitial antisocial behavior may increase both the fre-quency of physical punishment and subsequent an-tisocial behavior. Just as individual studies controlfor this possibility by using initial child misbehavioras a covariate, this meta-analysis uses differences be-tween effect sizes to control for initial child misbe-havior.
A second advantage of analyzing differences be-tween effect sizes is that they allow researchers todirectly compare realistic disciplinary choices. Bas-ing effect sizes on simple associations between a dis-ciplinary tactic and a child outcome implicitly com-pares parents who use that disciplinary tactic withthose who do not use it. Instead of choosing betweena given disciplinary tactic and doing nothing, parentstypically choose between two or more alternative dis-ciplinary responses (Ritchie, 1999). Differences in ef-fect sizes are better suited for such comparisons.
In summary, this meta-analysis uses differences
between effect sizes to control for confounds thatinfluence all disciplinary tactics, e.g., selection biasdue to initial child misbehavior. It falls short of be-ing causally definitive, however, because it rules outonly some plausible interpretations of the underly-ing empirical evidence. This strategy is a substantial
3When pre-test scores differ, relative pre-post gains provide afairer comparison than post-treatment differences.
improvement over typical meta-analytic methods forcorrelational data because it controls for importantconfounds and rules out alternative interpretationsassociated with them. At the very least, the cur-rent meta-analysis can determine whether the corre-lationally based effect sizes are uniquely detrimen-tal for physical punishment, are more detrimental forsome disciplinary tactics than others, or are equallydetrimental for all disciplinary tactics. Making thesedistinctions is a crucial step toward designing morecausally informative studies in the future. The re-sults also have important implications for how phys-ical punishment should be used, if at all, and whichalternative disciplinary tactics might be used instead.
METHOD
Literature Selection
Research studies were selected for this meta-analysis from recent reviews by Gershoff (2002) andLarzelere (2000). Both reviews attempted to be ex-haustive within their inclusion criteria for at leastthe previous 26 years. Additional selection criteriainclude the following: (1) The study must have inves-tigated one or more recommended alternative dis-ciplinary tactics as well as physical punishment, us-ing similar research methods. (2) The children hadto average less than 13 years old at the time of thediscipline. Most retrospective studies were excluded
because they pertained to physical punishment ofteenagers, based on the finding that retrospective re-ports of physical punishment correlated most highlywith mothers reported physical punishment at 1214 years old (Stattin, Janson, Klackenberg-Larsson,& Magnusson, 1995). To be included, retrospectivesurveys had to ask specifically for disciplinary tacticsat a younger age. (3) Selected studies had to inves-tigate at least one child outcome, excluding studiesthat investigated only parental outcomes.
These criteria yielded 26 studies that investi-gated physical punishment and one or more al-
ternative tactics, summarized in Table I. Onlyeight of these studies were included in both previ-ous reviews. Eleven studies from Gershoffs (2002)meta-analysis were excluded from Larzeleres (2000)review because they were cross-sectional (sevenstudies) or used overly broad measures of punish-ment (three studies). One other study was incorrectlyexcluded from Larzeleres (2000) review, because itdid specify a younger age in its retrospective survey
-
7/29/2019 Comparing Child Outcomes of Physical Punishment and Alternative Disciplinary Tactics - A Meta-Analysis
5/38
Physical Punishment vs. Alternative Tactics 5
TableI.EffectSizesofPhysicalPu
nishmentandAlternativeDisciplinaryResponsesonChildOutcomesBasisofeffects
ize&
Age,Gender
Discipline
Effect
discrepanciesfrom
Study
(Parent)a
Sample
N
tacticb
Outcome
sizec(d)
Gershoff(2002)
Short-termcompliance
RandomizedClinicalTrials
BeanandRoberts
(1981)
26B,G(M)
Clinic-referredfor
disruptivebehavior
problems
8
2-swatspankback-up
fortime-out(TO;
conditionalphysical
punishment[PP])
Observedcompliance
rateto30paren
tal
commands
2.80d
Gainfrompre-testtopost-testmeans.
Gershoffcomparedpos
t-testmeansfor
thecompleteparenttra
iningprocedure
vs.ano-treatmentcontrolgroup
8
Child-determined
releasefrom
time-out(TO)
Compliance
1.07
Gainfrompre-topost-test
8
2-swatspankback-up
fortime-out(TO;
conditionalPP)
Time-outsuccess
(Time-outsrequired
during30
commands)
1.61
Numberoftime-outsneededcompared
tothechild-determined
releasefrom
time-out.Outcomenot
inGershoff
DayandRoberts
(1983)
25B,G(M)
Clinic-referredfor
disruptivebehavior
problems
8
2-swatspankback-up
fortime-out(TO;
conditionalPP)
Compliancerateto30
parentalcommands
3.45d
Gainfrompre-testtopost-test.Gershoff
comparedthepost-testsforthespank
vs.thebarrierback-ups
8
Barrier(roomTO)
back-upfor
time-out(TO)
Compliance
3.51
Gainfrompre-topost-test
Roberts(1988)
2-6B,G(M)
Clinic-referredfor
behaviorproblems
9
2-swatspankback-up
forTO(conditional
PP)
Compliancerateto
30+
parental
commands
1.60d
Gainfrompre-testtopost-test.Studynot
inGershoff
9
Barrier(roomTO)
back-upforTO
Compliance
1.45
Gainfrompre-topost-test
9
2-swatspankback-up
forTO(conditional
PP)
Numberoftime-o
uts
(TOs)andTO
escapespriorto
complianceto1
0
commands
.13
MeannumberofTOsand
TOescapes
comparedtothebarrierback-up
RobertsandPowers
(1990)
26B,G(M)
Clinic-referredfor
behaviorproblems
9
2-swatspankback-up
forTO(conditional
PP)
Complianceratetoup
to60parental
commands
1.34c
Gainfrompre-testtopost-test.Gershoff
comparedthespankpo
st-testvs.the
other3back-upscombined
9
Barrier(roomTO)
back-upforTO
Compliance
2.17
Gainfrompre-topost-test
9
Restraintback-upfor
TO
Compliance
1.15
Same
9
Child-determinedTO
release
Compliance
1.52
Same
9
2-swatspankback-up
forTO(conditional
PP)
TOsuccess(children
withoutexcessive
TOsorexcessiv
e
TOescapes)
.44
Proportionofchildrenwithsuccessful
TOcompliance,comparedto
child-determinedTOre
leasegroup.
OutcomenotusedbyG
ershoff
9
Barrierback-upfor
TO
TOsuccess
1.02
Same
-
7/29/2019 Comparing Child Outcomes of Physical Punishment and Alternative Disciplinary Tactics - A Meta-Analysis
6/38
6 Larzelere and Kuhn
TableI.Continued
Basisofeffectsize&
Age,Gender
Discipline
Effect
discrepanciesfrom
Study
(Parent)a
Sample
N
tacticb
Outcome
sizec(d)
Gershoff(2002)
9
Restraintback-upfor
TO
TOsuccess
.22
Same
Within-subjectsequentialanalyses
Larzelereetal.(1996)
23B,G(M)
Volunteers
38
Slaphandorspank,
whetherreasoning
wasalsousedornot
(customaryPP,
contrastedwith
nexttwotactics)
Delayuntilnext
recurrenceof
disobedience,
comparedtotypical
delaysforthatchild
.05
Deviationsfromparticipa
ntsmean
delays,comparedwithother(i.e.,no
punishmentorreasonin
g).Couldnot
replicateGershoff
Reasoning,whether
usedwithphysical
ornonphysical
punishmentornot
Delayuntil
disobedience
recurrence
.01
Same
Nonphysical
punishment
(time-outor
privilegeremoval),
whetherusedwith
reasoningorPPor
not
Delayuntil
disobedience
recurrence
.02
Same
Reasoning&PP(no
nonphysical
punishment;
conditionalPP
combinedthiswith
Reasoning&PP&
nonphysical
punishment)
Delayuntil
disobedience
recurrence
.02
Same
Reasoning&
nonphysical
punishment&PP
(partofconditional
PP)
Delayuntil
disobedience
recurrence
.17
Same
Reasoningalone(no
physicalor
nonphysical
punishment)
Delayuntil
disobedience
recurrence
.06
Same
Nonphysical
punishmentalone
(time-outor
privilegeremoval;
noreasoningorPP)
Delayuntil
disobedience
recurrence
.02
Same
Reasoning&
nonphysical
punishment(noPP)
Delayuntil
disobedience
recurrence
.08
Same
-
7/29/2019 Comparing Child Outcomes of Physical Punishment and Alternative Disciplinary Tactics - A Meta-Analysis
7/38
Physical Punishment vs. Alternative Tactics 7
Ritchie(1999)
3B,G(M)
Volunteersfrom
birthrecords
90
Spank(conditional
PP)
Immediatered
uction
inprobabilityof
defiance
.97
Fromimmediatelypriorp
robabilityof
defiancecomparedtoimmediately
subsequentprobability
ofdefiance.
StudynotinGershoff
Reasonoroffer
alternatives
Dropindefiance
.02
Same
Threatenorverbal
powerassertion
Dropindefiance
.08
Same
Privilegeremoval
Dropindefiance
.33
Same
Time-out
Dropindefiance
.60
Same
Noresponse(ignore)
Dropindefiance
.02
Same
Physicalpower
assertion
Dropindefiance
.45
Same
Spank(customaryPP)
Immediatered
uction
inphysical
or
passive
noncompliance
.07
Fromimmediatelypriorp
robabilityof
twononcompliancetyp
escomparedto
theirimmediatelysubsequent
probability
Reasonoroffer
alternatives
Reductionin
noncompliance
.18
Same
Threatenorverbal
powerassertion
Reductionin
noncompliance
.02
Same
Privilegeremoval
Reductionin
noncompliance
.24
Same
Time-out
Reductionin
noncompliance
.16
Same
Noresponse(ignore)
Reductionin
noncompliance
.33
Same
Physicalpower
assertion
Reductionin
noncompliance
.20
Same
Correlationalsequentialana
lyses
Chapmanand
Zahn-Waxler(1982)
1029mos.,B,
G(M)
Volunteers
24
Physicalcoercion(PP
orrestraint)
withoutreasoning
(customaryPP)
Immediatecom
pliance
.09
Comparedtotheoverallcompliance
rate.CouldnotreplicateGershoff
Physicalcoercion&
reasoning
(conditionalPP)
Immediatecom
pliance
.02
Comparedtotheoverallcompliancerate
Reasoning(withor
withoutverbal
prohibition)
Immediatecom
pliance
.22
Same
Verbalprohibition
Immediatecom
pliance
.15
Same
Lovewithdrawal
(includingignoring
andtime-out)plus
anyofabovetactics
Immediatecom
pliance
.40
Same
Minton,Kagan,and
Levine(1971)
27mos.B,G
(M)
Volunteers
70
PPasproportionof
observed
misbehavior
(predominantPP)
Disobedience
requiringma
ternal
reprimand
.55
Averagedcorrelationsforboysandgirls.
SameasGershoff
-
7/29/2019 Comparing Child Outcomes of Physical Punishment and Alternative Disciplinary Tactics - A Meta-Analysis
8/38
8 Larzelere and Kuhn
TableI.Continued
Basisofeffectsize&
Age,Gender
Discipline
Effect
discrepanciesfrom
Study
(Parent)a
Sample
N
tacticb
Outcome
sizec(d)
Gershoff(2
002)
Explanationsof
reprimandsas
proportionof
misbehavior
Observed
disobedience
.44
Averagedcorrelationsforboysandgirls
Antisocialbehavior
Statisticallycontrolledlongitudinalstudies
LarzelereandSmith
(2000)
6
9atT1,811at
T2,B,G(M)
Nationalsampleof
youngmothers
7
85Frequencyspankedin
pastweek
(customaryPP)
Antisocialbeha
vior2
yearslater
.23
Meanantisocialfor1or
moretimesper
weekvs.nouseofthe
disciplinary
tactic,controllingforexternalizing
problemsatTime1,fiveother
variablesand6interactionsofthese
variableswiththedisc
iplinarytactic.
Unpublished,notinG
ershoff
7
85Frequencyprivileges
removedinpast
week
Antisocialbeha
vior
later
.21
Same
7
85Frequencygrounded
inpastweek
Antisocialbeha
vior
later
.20
Same
7
71Frequencyallowance
removedinpast
week
Antisocialbeha
vior
later
.10
Same
7
85Frequencysentto
roominpastweek
Antisocialbeha
vior
later
.18
Same
Larzelereetal.
(1998)
2
3atT1,4at
T2;B,G(M)
Volunteers
3
8
Slaphandorspank
(PP)without
reasoning,as
proportionof
misbehavior
incidents(averaged
withReasoning&
PPforpredominant
PP)
Disruptivebehavior
20monthslater
.41
Meanpartialcorrelation
ofproportional
usagewithsubsequentdisruptive
behavior,controllingforinitial
disruptivebehavior.N
otinGershoff
(multiplereportsfrom
samestudy)
Reasoning&PP
(proportionaluse;
partof
predominantPP)
Disruptivebehavior
20monthslater
.32
Same
ReasoningwithoutPP
ornonphysical
punishment
Disruptivebehavior
later
.80
Same
Nonphysical
punishment
(time-outor
privilegeremoval)
Disruptivebehavior
later
.20
Same
Reasoning&
nonphysical
punishment
Disruptivebehavior
later
.10
Same
-
7/29/2019 Comparing Child Outcomes of Physical Punishment and Alternative Disciplinary Tactics - A Meta-Analysis
9/38
Physical Punishment vs. Alternative Tactics 9
Within-subjectsequentialan
alyses
Larzelereetal.
(1996)
23B,G(M)
Volunteers
38
Slaphandorspank
(PP),whether
reasoningwasalso
usedornot
(customaryPP,
contrastedwith
nexttwotactics)
Delayuntilnext
recurrence
of
fighting,compared
totypicaldelaysfor
thatchild
.08
Deviationsfromparticipantsmean
delays,comparedwith
other(i.e.,no
punishmentorreasoning).Gershoff
didnotincludethisoutcome
Reasoning,whether
usedwithphysical
ornonphysical
punishmentornot
Delayuntilfighting
recurrence
.01
Same
Nonphysical
punishment
(time-outor
privilegeremoval),
whetherusedwith
reasoningorPPor
not
Delayuntilfighting
recurrence
.36
Same
Reasoning&PP(no
nonphysical
punishment;
(conditionalPP
combinedthiswith
Reasoning&PP&
nonphysical
punishment)
Delayuntilfighting
recurrence
(deviation)
.07
Same
Reasoning&
nonphysical
punishment&PP
(partofconditional
PP)
Delayuntilfighting
recurrence
.81
Same
Reasoningalone(no
physicalor
nonphysical
punishment)
Delayuntilfighting
recurrence
.09
Same
Nonphysical
punishmentalone
(time-outor
privilegeremoval;
noreasoningorPP)
Delayuntilfighting
recurrence
.26
Same
Reasoning&
nonphysical
punishment(noPP)
Delayuntilfighting
recurrence
.63
Same
Uncontrolledlongitudinalstudies
McClellandand
Pilon(1983)
5atT1,31atT2;
B,G(M)
Kindergartensample
78
ExtentofPP,
combining
frequencyand
severity(severePP)
NeedforPow
er26
yearslater
.42
Meanofcorrelationswith
Needfor
Powerformalesandfe
males.Study
notinGershoff
Reasoning
NeedforPow
er
.00
Non-significa
ntr
Privilegeremoval
NeedforPow
er
.00
Non-significa
ntr
-
7/29/2019 Comparing Child Outcomes of Physical Punishment and Alternative Disciplinary Tactics - A Meta-Analysis
10/38
10 Larzelere and Kuhn
TableI.Continued
Basisofeffectsize&
Age,Gender
Discipline
Effect
discrepanciesfrom
Study
(Parent)a
Sample
N
tacticb
Outcom
e
sizec(d)
Gershoff(
2002)
Lovewithdrawal
NeedforPower
.00
Non-signifi
cantr
Sears(1961)
5
atT1,12atT2;
B,G(M)
Kindergartensample1
60
ExtentofPP,
combiningseverity
andfrequency
(severePP)
Antisocialagg
ression
7yearslater
.14
Correlationswithantisocialaggression7
yearslater.Gershoff
averaged6
correlations(cross-se
ctionalandwith
prosocial,ambiguous,andantisocial
aggression)
Privilegeremoval
Antisocialagg
ression
.12
Correlationswithlater
antisocial
aggression
Lovewithdrawal
Antisocialagg
ression
.11
Same
Yarrowetal.(1968)
4
B,G(M)
Nurseryschoolsample5
8
Useofphysical
punishmentfor
vignettesabout
extreme
disobedience
(conditionalPP)
Teacher-rated
aggressionin
nurseryschool2
monthslater
.38
Correlation.Gershoffu
sedthe
correlationofseverit
yofall
punishmentforaggressionwith
concurrentaggressiontowardparents
(mother-report)
Useofreasoningfrom
vignettes
Laterschool
aggression
.28
Correlation
Useofscoldingfrom
vignettes
Laterschool
aggression
.24
Same
Useofprivilege
removalfrom
vignettes
Laterschool
aggression
.38
Same
Useofisolationfrom
vignettes
Laterschool
aggression
.18
Same
Useofdivertingfrom
vignettes
Laterschool
aggression
.47
Same
Useoflove
withdrawalfrom
vignettes
Laterschool
aggression
.24
Same
Retrospectivestudies
Watson(1989)
0
5atT1;17at
T2;B,G(M,
F)
NationalMerit
Scholarshipfinalists
&average
test-takers
2
500Parent-reported
spankingand
possiblytimeout
beforeage6,
(customaryPP)
Youth-reporte
d
hostility&m
ilder
(obloquial)
problems
.09
Correlationswithhostilityandobloquial
problems,using.00fornon-significant
rs.Gershoffusedonlythesignificantr
withonehostilityme
asure
Privilegeremovaland
assigningextra
dutiesbeforeage6
Hostilityandmilder
behaviorproblems
.13
Correlationswithhostility&obloquial
problems,using.00fornon-significant
rs
Uncontrolledcross-sectionalstudies
Strausand
Mouradian(1998)
2
14B,G(M)
Randomsampleof
twocounties
7
44
Howoftenspanked,
slappedorhitthe
childduringthe
past6months,
controllingfor
severe
out-of-controlPP
( c o n d i t i o n a l P P )
Antisocialand
impulsivebehavior
.14
F-valuesforPP,controllingforsevere
PP,3otherdisciplina
rytactics,4other
variables,andtheirinteractionswith
PP.Gershoffprobablyusedgraphed
meanantisocialscores,whichcouldnot
becomparedwithalternativetactics
-
7/29/2019 Comparing Child Outcomes of Physical Punishment and Alternative Disciplinary Tactics - A Meta-Analysis
11/38
Physical Punishment vs. Alternative Tactics 11
Percentageof
spankingsinwhich
motherssaidthey
lostitdueto
anger(severePP)
Antisocialan
d
impulsivebehavior
.28
F-valuesforseverePP,w
ithabove
controls
Howoftentheyused
disciplinary
reasoning,privilege
removal,and
time-outduring
past6months
Antisocialan
d
impulsivebehavior
.39
F-valuesforthesealternativedisciplinary
tactics,withabovecon
trols
Substanceabuse
Retrospectivestudies
Tennantetal.(1975)
014atT1,M=
23atT2;B
(M,F)
USArmysoldiers
5044
Spanking(customary
PP)
Frequentuse
of
hashish,alcohol,
amphetam
ines,and
opiates
.24
Percentageofmostandleastfrequent
usersreportingbeings
panked.
Averagedacross4substances.Not
amongGershoffs11o
utcomes
Non-contact
punishment
Frequentuse
of
hashish,alcohol,
amphetam
ines,and
opiates
.08
Percentageofmostvs.le
astfrequent
usersreportingreceivingthis
punishment.Averaged
across4
substances
Watson(1989)
05atTime1;17
atTime2.B,
G(B)
NationalMerit
Scholarshipfinalists
&average
test-takers
2500
Parentalreportof
spankingand
possiblytimeout
beforeage6
(customaryPP)
Youth-repor
ted
alcoholusage
.02
Correlation.Outcomeno
tinGershoff
Withdrawalof
privilegesand
assigningextra
dutiesbeforeage6
Youth-repor
ted
alcoholusage
.10
Correlatio
n
Conscience&resistancetote
mptation
Uncontrolledlongitudinalstudies
Grinder(1962)
56atT1,1112
atT2;B,G
(M)
Kindergartensample
140
ExtentofPP,
combiningseverity
andfrequency
(severePP)
Resiststemp
tationin
forbidden-toylab
test6yearslater
.24
Thepercentageofchildrenwhoresisted
temptationforhighvs.lowuseofthe
disciplinarytactic.Cou
ldnotreplicate
Gershoffsbeneficialeffectsize
Privilegeremoval
Resiststemp
tation
.10
Same
Isolation
Resiststemp
tation
.01
Same
Lovewithdrawal
Resiststemp
tation
.37
Same
Uncontrolledcross-sectionalstudies
Aronfreed(1961)
12B,G(M)
Sixth-gradesample
120
Primarilysensitization
(PP&uncontrolled
verbalassaults)vs.
primarilyinduction
(lovewithdrawal,
encouraging
responsibility,and
explanations;based
onresponsesto
vignettes;
predominantPP)
Internaland
external
motivation
sfor
moralcorr
ectionsin
projective
story
completions
.16
2
2contingencytables
with6internal
and6externalmoralm
otivations,
counting9non-signific
antassociations
as.00.Gershoffusedo
nlythe
significantassociationwithreparations
-
7/29/2019 Comparing Child Outcomes of Physical Punishment and Alternative Disciplinary Tactics - A Meta-Analysis
12/38
12 Larzelere and Kuhn
TableI.Continued
Basisofeffect
size&
Age,Gender
Discipline
Effect
discrepancies
from
Study
(Parent)a
Sample
N
tacticb
Outcome
sizec(d)
Gershoff(2002)
Burtonetal.(1961)
4B,G(M)
Privatenurseryschool
sample
77Interviewerratingof
PPasusual
disciplinetechnique
(predominantPP)
Resiststemptation
(labtest)
.35
From2
2contingencytable.Gershoff
didnotincludethisoutcome
Reasoningratedas
usualtechnique
Resiststemptation
.08
2
2contingencytable
Scoldingratedas
usualtechnique
Resiststemptation
.25
Same
Privilegeremoval
ratedasusual
technique
Resiststemptation
.63
Same
Isolationratedas
usualtechnique
Resiststemptation
.12
Same
Interviewerratingof
frequencyof
spanking,slapping,
andshaking(severe
PP)
Resiststemptation
.00
Tablenotgiven,estimate
dat.00dueto
non-significance
Ratedfrequencyof
reasoning
Resiststemptation
.61
2
2contingencytable
Ratedfrequencyof
scolding
Resiststemptation
.00
Non-significantassociation
Ratedfrequencyof
privilegeremoval
Resiststemptation
.34
2
2contingencytable.
Ratedfrequencyof
isolation
Resiststemptation
.00
Non-significantassociation
Ratedfrequencyof
lovewithdrawal
Resiststemptation
.33
2
2contingencytable
Interviewerratingof
frequencyof
spanking,slapping,
andshaking(severe
PP)
Conscience(initial
childactionsafter
wrongdoing)
.62
Gershoffusedonlythiscorrelationfrom
thestudy
Ratedfrequencyof
reasoning
Conscience
.36
Correlation
Ratedfrequencyof
scolding
Conscience
.00
Non-significantr
Ratedfrequencyof
privilegeremoval
Conscience
.32
Correlation
Ratedfrequencyof
isolation
Conscience
.30
Correlation
Ratedfrequencyof
lovewithdrawal
Conscience
.00
Non-significantr
-
7/29/2019 Comparing Child Outcomes of Physical Punishment and Alternative Disciplinary Tactics - A Meta-Analysis
13/38
Physical Punishment vs. Alternative Tactics 13
Lytton(1977)
2B(M,F)
Volunteers
90
Maternalandpaternal
frequenciesofPP
(customaryPP)
Compliance(two
measures)and
conscience
(1)
.04Averageofbetasof2pare
ntsfor3
outcomes,assuming.00
for
non-significantpredictors.Gershoff
onlyusedtheonesignifi
cant
correlation
Ratingofmothers
induction(e.g.,
explanationof
orders)
Compliancea
nd
conscience
.06Non-significantrs,butmostlyinthe
indicateddirection,acco
rdingtothe
text
Ratingofmothers
verbal
psychological
punishment
(criticism,
withdrawaloflove)
Compliancea
nd
conscience
.11Averageofbetasforthree
outcomes,
assuming.00fornon-sig
nificant
predictors
Maternalfrequencyof
lovewithdrawal
Compliancea
nd
conscience
.06Non-significantrs,butmostlyinthe
indicateddirection,acco
rdingtothe
text
Searsetal.(1957)
5B,G(M)
Kindergartensample
160ExtentofPP,
combiningseverity
andfrequency
(severePP)
Conscience
.41Correlationsofhighvs.lowuseofeach
disciplinarytacticwithc
onscience,
similartoGershoff
Reasoning
Conscience
.37
Same
Privilegeremoval
Conscience
.14
Same
Isolation
Conscience
.00Non-significantr
Lovewithdrawal
Conscience
.18Correlationsofhighvs.lowlove
withdrawal&conscienc
e
Yarrowetal.(1968)
4B,G(M)
Nurseryschoolsample
86
Reporteduseof
physical
punishmentfrom
vignettesabout
extreme
disobedience
(conditionalPP)
Conscience(m
aternal
report)
.02Correlation.Thisoutcome
notin
Gershoff
Useofreasoningfrom
vignettes
Conscience
.22
Correlation
Useofscoldingfrom
vignettes
Conscience
.18
Same
Useofprivilege
removalfrom
vignettes
Conscience
.22
Same
Useofisolationfrom
vignettes
Conscience
.30
Same
Useofdivertingfrom
vignettes
Conscience
.32
Same
Useoflove
withdrawalfrom
vignettes
Conscience
.04
Same
-
7/29/2019 Comparing Child Outcomes of Physical Punishment and Alternative Disciplinary Tactics - A Meta-Analysis
14/38
14 Larzelere and Kuhn
TableI.Continued
Basisofeffectsize&
Age,Gender
Discipline
Effect
discrepanciesfrom
Study
(Parent)a
Sample
N
tacticb
Outcom
e
sizec(d)
Gershoff(2
002)
Prosocialbehavior
Uncontrolledlongitudinalstudies
Zahn-Waxler,
Radke-Yarrow,
andKing(1979)
1524monthsB,
G(M)
Volunteers
16
Proportionaluseof
PPinemotionally
charged
motherchild
interactions
(predominantPP)
Reparationsan
d
altruism4.5months
later
.58
Averagecorrelationsfortime-lagdata
forreparationsandfo
raltruism.
Gershoffprobablyusedtheseplus
contemporaneousrs
Explanationswith
affect(proportional
use)
Reparationsan
d
altruismlate
r
1.25
Same
Positivesuggestions
(proportionaluse)
Reparationsan
d
altruismlate
r
.56
Same
Explanationswith
neutralaffect
(proportionaluse)
Reparationsan
d
altruismlate
r
.10
Same
Physicalrestraint
(proportionaluse)
Reparationsan
d
altruismlate
ron
.26
Same
Unexplainedverbal
prohibitions
(proportionaluse)
Reparationsan
d
altruismlate
ron
.70
Same
Ignoring
(proportionaluse)
Reparationsan
d
altruismlate
ron
.34
Same
Retrospectivestudies
Watson(1989)
05atT1;17at
T2;B,G(M,
F)
NationalMerit
Scholarshipfinalists
&average
test-takers
2500Parentalreportof
spankingand
possiblytimeout
beforeage6
(customaryPP)
Youth-reported
altruism
.00
Correlation,using.00fo
rnon-significant
rs.OutcomenotinGershoff
Parentalreportof
withdrawalof
privilegesand
assigningextra
dutiesbeforeage6
Youth-reported
altruism
.00
Non-significantrs
Uncontrolledcross-sectionalstudies
Hall(1994)
45B,G(M)
Preschoolersfrom
highincome
families
41
Frequencyof
spankingor
slapping,from
ConflictTactics
Scale(CTS)items
(customaryPP)
Verbalpositive
son
interpersona
l
problemsolv
ing
task
.18
Correlation.Gershoffusedonlyther
withnonverbalnegativesonasimilar
task,buttheequivalentrwasnot
reportedforreasoning
Reasoning(%of
maximumpossible
scoreonthreeCTS
items)
Verbalpositive
son
interpersona
l
problemsolv
ing
task
.44
Correlati
on
-
7/29/2019 Comparing Child Outcomes of Physical Punishment and Alternative Disciplinary Tactics - A Meta-Analysis
15/38
Physical Punishment vs. Alternative Tactics 15
Self-esteem
Retrospectivestudies
Larzelere,Klein,
Schumm,and
Alibrando(1989)
012atT1,M=
21atT2;B,G
(M,F)
HomeEconomics
collegestudents
157
Spankingfrequency
(customaryPP)
Self-esteem
.10
Correlationinfullrepo
rtmentionedin
articlesfootnote.Ou
tcomenotin
Gershoffduetonon-significantrs
Alternative
punishments
(time-out,privilege
removal,
restitution)inall
threeagegroups
Self-esteem
.08
Correlation
Watson(1989)
05atT1;17at
T2;B,G(M,
F)
NationalMerit
Scholarshipfinalists
&average
test-takers
2500
Parentalreportof
spankingand
possiblytimeout
beforeage6
(customaryPP)
Youth-repo
rted
neuroticism,
self-acceptance,and
senseofwell-being
.03
Correlationswiththree
outcomes,using
.00fornon-significan
trs.Outcomenot
inGershoff
Parentalreportof
withdrawalof
privilegesand
assigningextra
dutiesbeforeage6
Neuroticism
,
self-acceptance,&
well-bein
g
.00
Non-significantrs
Uncontrolledcross-sectionalstudies
Coopersmith(1967)
1012B(M)
Fifth-gradersselected
forconsistencyor
inconsistencyon
measuresof
self-esteem
63
Predominantuseof
physical
punishmentrather
thannexttwo
tactics,whenrules
areviolated
(predominantPP)
Self-esteem
.42
Proportionsofthehigh
andlow
self-esteemgroupsw
ithPPas
predominantdisciplinarymethod,
comparedtothenexttwo,tactics
similartoGershoff
Predominantuseof
lovewithdrawal
Self-esteem
.54
Proportionsoftheextremeself-esteem
groupsusinglovewithdrawal
predominantly
Predominantuseof
mildermanagement
tactics(restraint,
denial,isolation)
morethanabove
twotactics
Self-esteem
.86
Proportionsoftheextremeself-esteem
groupsusingmildertactics
predominantly
Stressdiscussionand
reasoningtoobtain
complianceand
cooperation,rather
thanforceor
autocraticmeans
Self-esteem
.87
Proportionsoftheextremeself-esteem
groupsthatpreferredeitherofthese
twodisciplinaryapproaches
Competency
Uncontrolledlongitudinalstudies
Crowne,Conn,
Marlowe,and
Edwards(1969)
5atT1,18atT2;
B,G(M,F)
Kindergartenchildren
83
Frequencyand
severityofspanking
byeachparent
(severePP)
Ambitious,
yet
realistica
spirations
andunus
ualgoal
shifts13yearslater
.00
Meanof15associationsforthree
measuresofphysical
punishmentand
fiveunambiguousou
tcomes.
d=
.00if
non-significant.NotinGershoff
-
7/29/2019 Comparing Child Outcomes of Physical Punishment and Alternative Disciplinary Tactics - A Meta-Analysis
16/38
16 Larzelere and Kuhn
TableI.Continued
Basisofeffectsize&
Age,Gender
Discipline
Effect
discrepanciesfrom
Study
(Parent)a
Sample
N
tacticb
Outcom
e
sizec(d)
Gershoff(2
002)
Reasoning
Ambitious,rea
listic
aspirations,etc.
.09
Meanoffiveassociation
s(onlyone
measureofreasoning)
Privilegeremoval
Ambitious,rea
listic
aspirations,etc.
.00
Meanoffiveassociation
s
Isolation
Ambitious,rea
listic
aspirations,etc.
.00
Meanoffiveassociation
s
Lovewithdrawal
Ambitious,rea
listic
aspirations,etc.
.00
Meanoffiveassociation
s
Retrospectivestudies
Watson(1989)
05atT1;17at
T2;B,G(M,
F)
NationalMerit
Scholarshipfinalists
&average
test-takers
2500Parentalreportof
spankingand
possiblytimeout
beforeage6
(customaryPP)
ScoreonNatio
nal
MeritSchola
rship
Testandrep
orted
%rankinclass
.11
Meanoftwocorrelation
s.Outcomenot
inGershoff
Parentalreportof
withdrawalof
privilegesand
assigningextra
dutiesbeforeage6
ScoreonNatio
nal
MeritTestand%
rankinclass
.19
Meanoftwocorrelation
s
Uncontrolledcross-sectionalstudies
Hall(1994)
45B,G(M)
Preschoolersfrom
highincome
families
41
Frequencyof
spankingor
slappingonConflict
TacticsScale(CTS)
item(customary
PP)
Numberofrelevant
solutionsin
interpersona
l
conflicttask
.73
Correlation.Outcomen
otinGershoff
Reasoning(%of
maximumpossible
scoreonthreeCTS
items)
Numberofrelevant
solutionsin
interpersona
l
conflicttask
.73
Correlation
aAgeinyearsunlessotherwiseindicated;M=
mean.T1=
Time1,T2=
Time2;B=
boys,G=
girls;(M)=
mothers,(F)=
fathers,(B)=
both.
bPhysicalpunishmentiscate
gorizedaseitherconditional,customary,severe,orpredominantusage.
cApositivedindicatesabeneficialassociation,i.e.,thatgreateruseofthedisciplinarytacticisassociatedwithpreferablechildoutcomes,e.g.,thetacticisassociatedwithgreater
prosocialbehaviororlessantisocialbehavior.Anegativedindicatesadetrimentalassociationbetweenthetacticandthechildoutcome.
dThemeanpre-postgainfor
achild-determinedrelease(1.295)wassubtractedfromthesedsforanalysesofeffectsizesofphysicalpunishment(e.g.,TableII).
-
7/29/2019 Comparing Child Outcomes of Physical Punishment and Alternative Disciplinary Tactics - A Meta-Analysis
17/38
Physical Punishment vs. Alternative Tactics 17
about discipline. Seven of the 26 studies in thismeta-analysis were included only in the Larzelere(2000) review. They were probably excluded fromthe Gershoff (2002) review due to having unusualchild outcome variables (three studies of substanceabuse, need for power, or realistic/ambitious aspira-tions), falling outside of her search criteria (two stud-ies), being unavailable via interlibrary loan (1), or be-ing part of a study that was already included (1).
Moderating Variables
The following variables were coded to deter-mine whether they accounted for differences in theeffect sizes of physical punishment and alternativedisciplinary tactics. A studys design was coded aseither (1) a randomized experiment, (2) a non-
randomized study that controlled for initial child mis-behavior with statistical controls or within-subjectanalyses, (3) a time-ordered study (longitudinal, ret-rospective, or sequential) in which the measure ofthe disciplinary tactic clearly preceded the child out-come measure (without controlling for initial misbe-havior), or (4) a cross-sectional design, in which thereferent periods for disciplinary tactics and the childoutcome overlapped in time.
Four types of physical punishment were distin-guished. Conditionalspanking was defined as phys-ical punishment that was used primarily to back-upmilder disciplinary tactics (e.g., reasoning or time-
out), used for defiance, or used in a controlledmanner. These definitions of conditional spankingemerged because each type demonstrated more ben-eficial outcomes (or less detrimental outcomes) thanother types of physical punishment in at least onestudy (e.g., Larzelere, Schneider, Larson, & Pike,1996; Ritchie, 1999; Straus & Mouradian, 1998). Al-though the most optimal usage might incorporateall three definitions, no study explicitly incorporatedmore than one of these definitions in its measureof physical punishment. Customary physical punish-ment was defined as typical parental usage (e.g., us-
age or frequency), without emphasizing its severityor predominance. It could have included severe phys-ical punishment, but only to the extent typical of ordi-nary usage by parents. Overly severe physical punish-ment was based on measures that gave extra pointsfor the severity of physical punishment. Examples in-cluded shaking (Burton, Maccoby, & Allinsmith,1961), severe spankings (Sears, Maccoby, & Levin,1957), or spanking when so angry that you lost it
(Straus & Mouradian, 1998). Finally, predominantuse of physical punishment included studies inves-tigating predominant disciplinary tactics (e.g., theprimary disciplinary tactic used) or proportional us-age (e.g., the proportion of disciplinary incidents forwhich the parents used physical punishment ratherthan milder disciplinary tactics).
Outcome variables were grouped into four cat-egories, consisting of compliance; antisocial behav-ior (including substance use and abuse); conscienceor resistance to temptation; and positive behaviors,competencies, or emotions. Several analyses com-bined antisocial behavior and conscience into a largercategory of misbehavior inhibition, to increase statis-tical power for testing other moderating variables.
Same-source bias was coded when an effect sizewas based solely on information provided by thesame person, as opposed to incorporating distinct
sources of information. Two age groups distinguishedchildren averaging older or younger than 7 years atthe time of the discipline.
Selected Meta-Analytic Details
Most effect sizes were calculated usingJohnsons (1989) DSTAT program, followingGershoff (2002). When a study had multiple relevantstatistics, we selected statistics that minimized themethodological problems noted by Baumrind et al.(2002), including the same-source bias and corre-
lational statistics. When a study included severalstatistics that differed on these characteristics, effectsizes were based on the stronger evidence and alsodistinguished four types of physical punishment(conditional, customary, severe, and predominantusage). In four studies, the best estimate of the effectsize controlled statistically for one or more othervariables, such as initial child misbehavior. In thosecases, the effect size was based on a standardizedregression coefficient or similar statistic, followingGlass, McGaw, and Smith (1981). In three of thosefour studies, the effect size could not be estimated
in standard deviations units of the outcome vari-able. Therefore, those studies were coded as using adistinct standard deviation unit typical of covariance-corrected coefficients, following Glass et al. (1981).Because the effect sizes from those three studies didnot differ significantly from the other 23 studies, thedistinction was dropped for the main analyses.
Effect sizes (ds) were corrected for an up-ward bias in small studies, using Hedges correction
-
7/29/2019 Comparing Child Outcomes of Physical Punishment and Alternative Disciplinary Tactics - A Meta-Analysis
18/38
18 Larzelere and Kuhn
(Lipsey & Wilson, 2001). For the analyses, each ef-fect size was weighted by a function of its sample sizeand the inverse of its extremity (Lipsey & Wilson,2001, p. 49; Shadish & Haddock, 1994, p. 268).
The calculations of effect sizes from three sets ofstudies warrant additional clarification, which is pro-vided in the appendix (the Roberts series of studies,two Larzelere studies, and Ritchie, 1999). The guid-ing principle was to base effect sizes on equivalentanalyses of physical punishment and alternative tac-tics. In addition, Larzelere et al. (1996) and Ritchie(1999) yielded different effect sizes for conditionaland customary use of physical punishment.
Two effect sizes were included from a studywhen they were relevant for distinct cells in a partic-ular analysis. For example, several studies had effectsizes for conditional spanking and for another cate-gory of physical punishment. Those studies then con-
tributed two different effect sizes for those two typesof physical punishment. This increased the samplesize from 26 studies to 32 relevant effect sizes in theinitial analyses of type of physical punishment.
Following Hedges (1994), the Q statistic wasbased on ANOVA sums of squares to test hypothe-ses about whether weighted mean effect sizes var-ied significantly by moderating factors. The Q statis-tic is distributed as 2 under the null hypothesis.Most analyses of moderating variables had missingcells because all combinations of those factors werenot represented by at least one study. Consequently,the Q statistic was based on the Type IV sums of
squares in the weighted fixed-effects ANOVA. Sta-tistical tests of whether weighted differential meansdiffered from zero used a z statistic, based on Lipseyand Wilson (2001, p. 115).
RESULTS
Effect Sizes of Physical Punishment by Research
Design and Physical Punishment Type
The effect sizes of physical punishment on childoutcomes varied significantly by type of physical
punishment, Q(3) = 9.80, p < .05, by research de-sign, Q(3) = 28.25, p < .001, and by their interac-tion, Q(5) = 17.82, p < .01. As shown in Table II,weighted mean effect sizes appeared detrimental forsevere physical punishment (d = .22) and predom-inant physical punishment (d = .21), but were nearzero for customary and conditional physical punish-ment (ds = .06 and .05, respectively). Mean effectsizes were apparently detrimental in studies using
correlational designs (d = .22), approached zero intime-ordered and controlled designs (ds = .10 and.08, respectively), and were apparently beneficialin studies employing randomized designs (d = .80).The interaction effect was due to the following excep-tions to the usual pattern of effect sizes become lessdetrimental or more beneficial as design quality im-proved: Customary physical punishment produced itsmost detrimental effect size in statistically controlledstudies rather than in cross-sectional studies, whereaspredominant usage yielded its most detrimental ef-fect size in time-ordered designs rather than in cross-sectional studies.
Differential Effect Sizes by Research Design
and Physical Punishment Type
The next analysis investigated differentialeffectsizes by research design and type of physical punish-ment. A differential effect size is the difference be-tween the mean effect size for physical punishmentand the mean effect size for alternative disciplinarytactics in the same study using the same method-ology. The results showed that neither design northe Design Physical Punishment Type interactionwas significant, Q(3) = 4.37 and Q(5) = 8.12, respec-tively. Differential effect sizes varied only by the typeof physical punishment, Q(3) = 18.26, p < .001.
Table III shows that differential effect sizes fa-
vored physical punishment over alternative tacticswhen physical punishment was defined as conditional(differential d = .29) or customary (differential d =.14). (For brevity, differential d will be shortened tod from here on, which is what it would be calledto describe differences between treatment and con-trol conditions.) In contrast, differential effect sizesfavored alternative tactics over both overly severe(d = .07) and predominant physical punishment(d = .33).
It is instructive at this point to compare the re-sults of the first two analyses. In Table II, the effect
sizes associated with physical punishment varied sig-nificantly by research design and by the Design Physical Punishment Type interaction. In Table III,however, differential effect sizes (d for physical pun-ishment minus d for alternative tactics) did not varysignificantly by research design or by its interactionwith physical punishment type. These results indi-cate partial success in reducing confounds associ-ated with correlational evidence by using differential
-
7/29/2019 Comparing Child Outcomes of Physical Punishment and Alternative Disciplinary Tactics - A Meta-Analysis
19/38
Physical Punishment vs. Alternative Tactics 19
Table II. Weighted Effect Sizes for Physical Punishment by Research Design and Type of Physical Punishment
Research designType of physical
punishment Cross-sectional Time-ordered Controlled Randomized Weighted mean d
Predominant .22 (4) .58 (1) .05 (1) .21 (6)
Overly severe .32 (3) .07 (4) .22 (7)
Customary .06 (2) .10 (4) .19 (3) .06 (9)Conditional .13 (2) .30 (2) .68 (2) .80 (4) .05 (10)
Weighted mean d .22 (11) .10 (11) .08 (6) .80 (4) .00 (32)
Note. n of studies in parentheses. A positive effect size (d) indicates that higher physical punishment scores are
associated with more beneficial or less detrimental child outcomes than are low scores on physical punishment.
All mean ds are weighted by Lipsey & Wilsons equation (2001, p. 49).
effect sizes of alternative tactics from the samestudy.
To obtain sufficient statistical power to investi-gate hypothesized moderators of the relative effec-tiveness of physical punishment and alternative tac-tics, the following analyses drop research design as afactor because it was not a significant predictor. Like-wise, predominant usage and severe physical punish-ment were combined into one category in subsequentanalyses.
Differential Effect Sizes by Outcomeand Physical Punishment Type
The next set of analyses investigated whetherthe differential effect sizes of physical punishment vs.alternatives varied by type of child outcome. The re-
sults indicated that outcome type, physical punish-ment type, and their interaction were significantlyrelated to differential effect sizes: physical pun-ishment type, Q(2) = 41.15; outcome type, Q(2) =21.15; interaction, Q(3) = 16.95, all ps < .001. As ex-pected, conditional spanking showed a more posi-tive differential effect size (d = .29) than customaryphysical punishment (d = .14), which, in turn, pro-
duced a more positive differential effect size thansevere/predominant physical punishment (d = .12;see Tables IVVI). Unexpectedly, effect sizes morestrongly favored physical punishment for misbehav-ior inhibition (antisocial behavior and conscience;d = .12) than for either compliance (d = .00) or posi-tive behavior and affect (d = .01). The interaction ef-fect reflected the fact that the differential effect sizesfor compliance varied by type of physical punishmentmuch more than for other outcomes. Both the largestnegative and the largest positive differential effectsizes occurred for compliance. Severe/predominantphysical punishment compared less favorably withalternatives for compliance than for any other out-come, whereas conditional spanking compared morefavorably with alternative tactics for compliance thanfor any other outcome. For most outcomes, differen-tial effect sizes were positive for conditional spank-
ing, approached zero for customary physical punish-ment, and were negative for severe or predominantusage.
Two of the weighted means listed above de-pended heavily upon results from the largest study,a retrospective survey of substance abuse in 5044military personnel (Tennant, Detels, & Clark, 1975).Because of its unusually large sample size, this study
Table III. Weighted Differential Effect Sizes (Physical Punishment Minus Alternative Tactics) by Research
Design and Type of Physical Punishment
Research designType of physical
punishment Cross-sectional Time-ordered Controlled Randomized Weighted mean dPredominant .37 (4) .78 (1) .22 (1) .33 (6)Overly severe .11 (3) .05 (4) .07 (7)
Customary .10 (2) .18 (4) .06 (3) .14 (9)Conditional .22 (2) .44 (2) .59 (2) .34 (4) .29 (10)
Weighted mean d .03 (11) .17 (11) .02 (6) .34 (4) .11 (32)
Note. n of studies in parentheses. A positive effect size (d) indicates that physical punishment is associated withmore beneficial or less detrimental child outcomes than are alternative tactics in the same studies. All mean ds
are weighted by Lipsey & Wilsons equation (2001, p. 49).
-
7/29/2019 Comparing Child Outcomes of Physical Punishment and Alternative Disciplinary Tactics - A Meta-Analysis
20/38
20 Larzelere and Kuhn
Table IV. Effect Sizes ofConditionala Physical Punishment Compared to Alternative Disciplinary Tactics
Mean effect size (d)
Alternative Alternative Conditionala Mean difference in
disciplinary tactic Child outcome Nstudies Nchildren disciplinary tactic physical punishment effect sizesCompliance
Reasoning Noncompliance 313 152 .03 .55 .59b
Verbal prohibition Immediate
compliance
11 24 .15 .02 .17
Threats or verbal
power assertion
Stop defiance 13 90 .08 .97 .89
Privilege removal or
time out
Subsequent
compliance
12 38 .02 .01 .01
Privilege removal Stop defiance 13 90 .33 .97 1.30
Time out Stop defiance 13 90 .60 .97 .37
Barrier (room timeout)
Compliance tocommands &
time out
346 52 1.04 .84 .20
Reasoning plus
nonphysicalpunishment
Subsequent
compliance
12 38 .08 .01 .09
Ignoring Stop defiance 13 90 .02 .97 .95
Love withdrawal Immediatecompliance
11 24 .40 .02 .38
Restraint, physical
power assertion
Stop defiance,
compliance tocommands &
TO
23,6 108 .34 .85 .51
Child release from timeout (TO)
Compliance tocommands
and to timeout
26,7 34 .02 .77 .74
Mean for compliance 7 220 .26 .68 .43b
Antisocial behaviorReasoning Aggression 22,8 96 .22 .35 .56
Scolding Schoolaggression
18 58 .24 .38 .62
Privilege removal ortime out
Aggression 12 38 .26 .26 .00
Privilege removal Schoolaggression
18 58 .38 .38 .00
Isolation School
aggression
18 58 .18 .38 .56
Reasoning plus
nonphysicalpunishment
Aggression 12 38 .63 .26 .37
Reasoning ornonphysical
punishment
Antisocial,impulsivity
19 744 .39 -.14 .25
Love withdrawal School
aggression
18 58 .24 .38 .62
Diverting School
aggression
18 58 .47 .38 .85
Mean for antisocial behavior 3 840 .35 .07 .28
-
7/29/2019 Comparing Child Outcomes of Physical Punishment and Alternative Disciplinary Tactics - A Meta-Analysis
21/38
Physical Punishment vs. Alternative Tactics 21
Table IV. Continued
Mean effect size (d)
Alternative Alternative Conditionala Mean difference indisciplinary tactic Child outcome Nstudies Nchildren disciplinary tactic physical punishment effect sizes
Conscience
Reasoning Conscience 18 86 .22 .02 .24
Scolding Conscience 18 86 .18 .02 .16
Privilege removal Conscience 18 86 .22 .02 .24
Isolation Conscience 18 86 .30 .02 .28
Love withdrawal Conscience 18 86 .04 .02 .02
Diverting Conscience 18 86 .32 .02 .34
Mean for conscience 1 86 .04 .02 .06
Grand mean 9 1050 .23 .06 .29
Note. In the last three columns, effect sizes that are positive indicate that more beneficial outcomes are associated with greater use of(a) alternative disciplinary tactics or (b) conditional physical punishment, or that (c) more beneficial outcomes are associ-
ated with conditional physical punishment than with alternative disciplinary tactics. Effect sizes are based on comparable statis-tics, minimizing the methodological problems noted by Baumrind et al. (2002) whenever possible. Means are weighted by sam-
ple size and by effect size extremity according to Lipsey and Wilsons equation (2001, p. 49), with each study contributingone mean of its relevant effect sizes. Some differential effect sizes are not an exact difference of the tabled entries due to
rounding.Studies cited: 1Chapman and Zahn-Waxler (1982), 2Larzelere et al. (1996), 3Ritchie (1999), 4Day and Roberts (1983), 5Roberts (1988),6Roberts and Powers (1990), 7Bean and Roberts (1981), 8Yarrow et al. (1968), 9Straus and Mouradian (1998).aEither (1) nonabusive backup for milder disciplinary tactics in 2- to 6-year-olds, (2) used specifically for defiance in 3- or 4-year-olds, or
(3) used in a controlled manner (not out of control due to anger) with 214-year-olds.bSignificant heterogeneity of the effect sizes contributing to this mean, Q statistic, p < .05.p < .05, significantly different from d = .00, z statistic (only performed in the right-hand column).p < .01.p < .001.
was weighted eight times more than the median-sized study in this meta-analysis. Excluding thisstudy, the weighted mean differential effect sizedropped to d = .00 for customary physical pun-
ishment and to d = .02 for misbehavior inhibition(antisocial/conscience). To avoid an overly favor-able comparison between customary physical punish-ment and nonphysical punishment based solely onthis one study, we repeated all analyses excludingthe Tennant et al. (1975) study and report the re-sults separately whenever this exclusion changed thefindings.
Tables IVVI distinguish between antisocialbehavior and conscience, even though they weretreated as misbehavior inhibition in the above anal-yses due to the small number of studies that in-
vestigated conscience. Table IV shows that, com-pared to alternative tactics, conditional spankingwas associated with greater reductions in noncompli-ance, d = .43, z= 3.08, p < .01, and antisocial behav-ior, d = .28, z= 4.11, p < .001. Conditional spank-ing did not differ significantly from alternative tac-tics in promoting the development of conscience, butthis was based on only one cross-sectional study.
Table V shows that, compared to alternatives, cus-tomary physical punishment was associated withgreater reductions in antisocial behavior, but this re-sult depended upon the largest study (Tennant et
al., 1975), d = .17 with it, z= 6.56, p < .001; d = .03without it, z= 0.81, ns. Otherwise, customary phys-ical punishment was not significantly different fromalternative tactics in its associations with other out-comes. Table VI shows that, compared to alterna-tives, severe/predominant physical punishment wasassociated with less compliance, d = .99, z= 4.43,p < .001, conscience, d = .36, z= 4.48, p < .001,positive behavior and affect, d = .36, z= 2.28,p < .05, and antisocial behavior, d = .14, z= 2.23,p < .05.
Differential Effect Sizes by Outcome, PhysicalPunishment Type, and Other Predictors
Additional analyses investigated whether thesame-source bias, the childs age, or the short- vs.long-term timing of the outcomes influenced differ-ential effect sizes and whether they modified the
-
7/29/2019 Comparing Child Outcomes of Physical Punishment and Alternative Disciplinary Tactics - A Meta-Analysis
22/38
22 Larzelere and Kuhn
Table V. Effect Sizes ofCustomarya Physical Punishment Compared to Alternative Disciplinary Tactics
Mean effect size (d)
Alternative Alternative Customarya Mean difference in
disciplinary tactic Child outcome Nstudies Nchildren disciplinary tactic physical punishment effect sizes
Compliance
Reasoning Noncompliance 313 152 .07 .04 .03
Verbal prohibition Immediatecompliance
12 24 .15 .09 .24
Threats or verbalpower assertion
Stop noncompli-ance
13 90 .02 .07 .05
Privilege removal or
time out
Subsequent
compliance
11 38 .02 .05 .03
Privilege removal Stop noncompli-
ance
13 90 .24 .07 .17
Time out Stop noncompli-
ance
13 90 .16 .07 .09
Ignoring Stop noncompli-
ance
13 90 .33 .07 .26
Love withdrawal Immediate
compliance
12 24 .40 .09 .31
Physical power
assertion
Stop noncompli-
ance
13 90 .20 .07 .13
Mean for compliance 3 152 .10 .04 .06
Antisocial behavior
Reasoning Aggression 11 38 .01 .08 .07
Non-contact
punishment
Aggression or
substanceabuse
21,4 3594 .07 .23 .31c
Privilege removal Antisocial
behavior oralcohol usage
25,6 3285 .13 .10 .03
Sent to room Antisocial
behavior
15 785 .18 .23 .05
Mean for antisocial behavior 4 6879 .10 .06 .17bc
Conscience
Reasoning Conscience &
compliance
17 90 .06 .04 .10
Verbal psychological
punishment
Conscience,
compliance
17 90 .11 .04 .07
Love withdrawal Conscience,compliance
17 90 .06 .04 .02
Mean for conscience 1 90 .04 .04 .00
Mental health
Privilege removal Neuroticism,
esteem
16 2500 .00 .03 .03
Positive behavior and affectReasoning Solutions &
verbalpositives in
problemsolving
18 41 .59 .28 .31
-
7/29/2019 Comparing Child Outcomes of Physical Punishment and Alternative Disciplinary Tactics - A Meta-Analysis
23/38
Physical Punishment vs. Alternative Tactics 23
Table V. Continued
Mean effect size (d)
Alternative Alternative Customarya Mean difference indisciplinary tactic Child outcome Nstudies Nchildren disciplinary tactic physical punishment effect sizes
Time out, privilegeremoval, or
restitution
Self-esteem 19 157 .08 .10 .02
Privilege removal Prosocialbehavior
16 2500 .00 .00 .00
Privilege removal Academic
achievement
16 2500 .19 .11 .08
Mean for positive behavior and affect 3 2698 .08 .05 .03
Grand mean 9 7281 .08 .06 14bc
Note. See Table IV.Studies cited: 1Larzelere et al. (1996), 2Chapman and Zahn-Waxler (1982), 3Ritchie (1999), 4Tennant et al. (1975), 5Larzelere and
Smith (2000), 6Watson (1989), 7Lytton (1977), 8Hall (1994), 9Larzelere et al. (1989).aCustomary physical punishment refers to typical usage by parents, e.g., via measures of usage or frequency without emphasizing
severity or predominant usage.bSignificant heterogeneity of the effect sizes contributing to this mean, Q statistic, p < .05.c
After dropping the largest study (Tennant et al., 1975), the differential effect size is not significantly different fromd = .
00 and theremaining effect sizes are homogeneous.p < .05, significantly different from d = .00, z statistic (only performed in the right-hand column).p < .01.p < .001.
main conclusions. For each new predictor, the firstquestion was whether it was associated with differ-ential effect sizes at the zero-order level (e.g., cor-relations). If so, the second question was whetherits independent contribution was significant af-ter controlling for outcome and type of physicalpunishment.
Same-Source Bias
Without controlling for other factors, differen-tial effect sizes favored physical punishment over al-ternative tactics more when the same source of infor-mation was used for parent and child variables thanotherwise, d = .15 vs. d = .03, Q(1) = 12.42, p