joHi'H/it ol Consulting and Clinical I'sy1 « 7 f i , Vol. 44, No. 6, 910 920

Toward a Task Analysis of Assertive BehaviorRobert M. SchwartzIndiana University

John M. GottmanUniversity of Illinois

The present stud)' analyzed the components of assertive behavior. Asscrtivenessproblems were conceptualized in terms of a task analysis of the topography ofcompetent responding. One hundred one subjects who spanned the range ofassertivcncss, measured by McFall's Conflict Resolution Inventory, respondedto three sets of situations requiring refusal of an unreasonable request. Contentknowledge of an assertive response, delivery of the response under two condi-tions, heart rate, sclf-perccivcd tension, and the incidence of positive andnegative self-statements were assessed. Differences on these variables betweenlow-, moderate-, and high-assertive groups were analyzed to determine thenature of the response deficit in nonassertivc subjects. Low-assertive subjectsdiffered from moderate- and high-assertive subjects on a role-playing assess-ment requiring them to deliver an assertive response, but they did not differfrom moderate- and high-assertive subjects on their knowledge of a competentresponse or on hypothetical delivery situations. No significant differences inheart rate were observed between low-, moderate- and high-assertive subjects;however, higher self-perceived tension was found in low- compared to moderate -and high-assertive subjects. A greater number of negative and fewer positiveself-statements were reported by low- compared to moderate- and high-as-sertive subjects. The present behavior task analysis study is recommended as aclinical assessment study preliminary to investigations comparing behaviorchange interventions.

Response acquisition approaches to asser-tion training are based on a skill-deficit model.According to this view nonassertive subjectsare people with specific limited capabilities ina specific set of social situations. This ap-proach is best characterized by McFall andTwentyman (1973), who wrote thatThe therapeutic objective is to provide patients withdirect training in precisely those skills in their re-sponse repertoires. Very little attention is given toeliminating existing maladaptivc behaviors; instead,it is assumed that as skillful, adaptive responses areacquired, rehearsed, and reinforced, the previousmaladaptive responses will be displaced and willdisappear, (p. 199)

A basic question has remained unresolvedin response acquisition approaches, namely,

The authors wish to thank Pam Kcgg, RobertSetty, Jacob Pankowski, Jim Miser, and Alex Brail-man for their help in running subjects and collect-ing dala; Richard McFall for providing materialslor the present investigation; and Richard McFalland Donald Meichenbaum for providing helpfulcriticism of the article.

Requests for reprints should be sent to RobertSchwartz, Department of Psychology, Indiana Uni-versity, Bloominglon, Indiana 47401.

what is the specific nature of the deficit innonassertive subjects? Component analyses ofassertion training programs (e.g., McFall &Twentyman, 1973) provide one approach tothis question. Presumably, if modeling doesnot add significantly to the treatment effect,then the response deficit could not have in-volved a lack of exposure to skillful models.

Although the component analysis approachis useful in creating efficient interventions,there are several problems with its potentialtheoretical contribution toward specifying thenature of the response deficit. A treatmentcomponent, if effective, may be totally unre-lated to the nature of the problem it treats(Buchwald & Young, 1969). For example,although aspirin ameliorates headaches, head-aches are not a result of an aspirin deficiency.Furthermore, treatment components rarelycan claim to deal with only one deficit at atime. For example, an effective coaching com-ponent may simultaneously result in responseshaping, confidence building, and cognitiverestructuring.

It may seem in the negative case (e.g., thefailure of models to enhance the training ef-

910

TASK ANALYSIS 911

1'ccL as shown in McFall and Twentyman,1973) that some insight is gained about whatthe deficit is not. However, in the negativecase, the information gained is provocative atbest. For example, why does modeling fail toenhance the treatment program? Is it becausenonassertive subjects have seen many modelsof assertive behavior in their day-to-day ex-perience and that the information providedby the models is redundant? How does theinformation conveyed by models differ fromthat provided by coaching? Is it more induc-tive than deductive? Is it more sketchy? Or isthe modeling component poorly designed?What specifically is the response deficit thatwould make modeling ineffective and coachingeffective? Therefore, even in the negative case,a component analysis does not specify theresponse deficit with precision.

An alternative strategy for specifying theresponse deficits in nonassertive subjects issuggested by the research of Gagne (1969) inthe design of a remedial mathematics pro-gram. Suppose that some fourth-grade chil-dren in a city were incompetent in long divi-sion. Tests of addition, subtraction, multipli-cation, and the knowledge of remainderscolud be given to both children who could andcould not do long-division problems. Theintervention program would depend on thespecific performance discrepancy obtainedfrom this "task analysis" study. Such a studybegins by specifying the likely components ofa competent response and then testing theextent to which performance on the com-ponents discriminates between competent andincompetent populations.

The purpose of the present investigationwas to determine what components are neces-sary in order to perform a competent assertiveresponse. The assertive response was definedto include measurable responses from the cog-nitive, physiological, and overt responseclasses. Low-assertive, moderate-assertive, andhigh-assertive subjects were compared to de-termine which components of assertive be-havior differentiated between groups withinthe three response classes mentioned above.For the purpose of this study, the definitionof assertive behavior has been limited to re-

fusal behavior, that is, refusing an unreason-able request.1

The components assessed within the cogni-tive system included positive and negativeself-statements, that is, innerstatements orthoughts that would make it easier or harderto deliver a convincing refusal. When con-fronted with unreasonable requests, it is possi-ble that assertive people make self-statementsthat are adaptive in terms of their ability torefuse. The unreasonable request may alsoelicit self-statements in nonassertive subjectsthat focus on the fear of being disliked or onhaving a moral responsibility to help everyoneregardless of the situation. Meichenbaumfound that test-anxious clients (Meichenbaum,1972), speech-anxious clients (Meichenbaum,1971), and phobic clients (Meichenbaum,1971) produce negative self-statements thatare maladaptive in terms of the desired per-formance. In the present study, the cognitiveself-statements as they relate to the assertionsituations were assessed by the AssertivenessSelf-Statement Test (ASST) devised for thisstudy.

Within the physiological system, the com-ponent measured by the present investigationwas heart rate. In treating nonassertive sub-jects, McFall and Marston (1970) found thatbehavior rehearsal resulted in a reduction inheart rate measured after McFalFs BehaviorRehearsal Assertion Test (BRAT); controlgroups demonstrated an increase in heart rate.Since a reduction in heart rate appears to bean outcome of McFall's treatment program,it has been used as the physiological measurein the present study. In addition, subjectswere asked to rate their self-perception oftension on a 7-point scale after performingassertive responses.

To separate knowledge of the content of acompetent response from its delivery, threesets of problematic situations that require anassertive response were administered to the

J In a pilot study with 60 undergraduates, a gen-eral assertion scale (Galassi, DeLo, Galassi, & Bas-tein, 1974) was administered with McFall andLillesand's (1971) CRT. The correlation between thetwo scales was .72, so it is likely that the inabilityto refuse an unreasonable request is strongly relatedto general assertion problems.

912 ROBERT M. SCHWARTZ AND JOHN M. GOTTMAN

Abilily 1o deliveran assertive responsein direct confrontation

(BRAT)

Cognitiveself-statements

(ASST)

Ability to deliver an assertiveresponse in indirect situations

(HYPO)

Knowledge of the content ofa good assertive response

(AKO

F.IGTJRK 1. Hypothetical task analysis of the assertive response. (Boxes lower on the flowchart repre-sent prerequisite behaviors or mediating process for competent performance.)

subjects. To assess the content knowledge ofthe response, the Assertiveness KnowledgeInventory (AKE) was devised. This inven-tory presents unreasonable requests in writtenform and requires a written refusal responseto determine whether the subject knows whatan assertive response entails, f t was assumedthat the written nature of the task wouldminimize other possible response components(e.g., physiological arousal or negative self-statements) that might occur as the task ap-proached reality. To assess the ability to de-liver the response orally under limited cir-cumstances, the Hypothetical BehavioralRole-Playing Assertion Test (HYPO) wasdevised. These situations were presented onaudio tape and an oral response was required,but the subjects were told to imagine thatthey were only modeling a good assertiveresponse to show a friend how to do it. Thetask assessed the ability to construct an as-sertive response and deliver it orally underhypothetical and "safe" circumstances. Again,

an attempt was made to make the situationunrealistic to reduce possible responses fromother classes that may be elicited by the realsituation. Finally, a shortened form of theBehavioral Role-Playing Assertion Test (theReduced Behavior Rehearsal Assertion Test;RBRAT) was used to assess the content anddelivery of the assertive response under cir-cumstances that simulated reality as much aspossible. Here subjects were told to imaginethat they were being confronted with an un-reasonable request and to respond orally asthough they were actually talking to the per-son making the request. This task was de-signed to determine the subject's ability toconstruct and deliver an assertive responseunder circumstances that approximate reallife.

These specific components of assertive be-havior all appear to be relevant to successfulperformance. The relationship of these com-ponents can be conceptualized in a hierarchi-cal task analysis, with performance of the

TASK ANALYSIS 913

JR.HRAT representing the terminal behavioralobjective (see Figure 1). The skills measuredby tasks assessing knowledge of the content ofa good assertive response are prerequisite toperformance on the RBRAT. But before theterminal behavior can be performed, the heartrate responses, self-perceived tension, andcognitive self-statements may intervene toaffect delivery. The form that these responsestake can either be adaptive or maladaptive interms of the terminal goal behavior. Sincethis study is exploratory in nature, no spe-cific hypotheses were offered as to the natureof differences on these components of the as-sertive response,

METHODSubjects

Forty-seven male and 54 female college studentsparticipated in the experiment. They ranged fromextremely nonasscrtivc to highly assertive as mea-sured by the Conflict Resolution Inventory (CRI)developed by McFall and Lillcsand (1071). A num-ber of subjects were recruited on the basis of theirown evaluation of their degree of asscrlivcncssthrough announcements in several undergraduatepsychology classes. All subjects were formally testedfor level of asscrtiveness by the CRI. Subjects wereassigned to cither low-, moderate-, or high-assertivegroups on the basis of their CRI scores. Classificationwas done with a bivariatc criterion using both asser-tion and nonassertion scores on the CRI. Low as-scrtives had to earn an assertion score of 13 or lessand a nonassertion score of 18 or more; moderateassertivcs had to earn an assertion score of between10 and 20 or a nonassertion score of between 11 and17; and high assertivcs had to earn an assertionscore of 21 or more and a nonassertion score of 10or less. There were 32 low assertives, 41 moderateassertives, and 28 high assertives, with approximatelyequal numbers of males and females in each group.

ProcedureSubjects were introduced to the experiment and

told that the purpose was to find out more abouthow people react in situations requiring assertive be-havior in order to develop a training program to helppeople who have a problem in this area.

While subjects filled out the CRI, their heart ratewas recorded to obtain a base rate for later com-parisons. They also indicated their level of tensionon a 7-poinl scale to provide a base rate of sclf-perccivcd tension. After subjects were administeredthe CRI and randomly assigned to a counterbalanc-ing order, they were presented with three sets ofstimulus situations in which they were confrontedwith unreasonable requests. The three sets of situa-tions were presented in counterbalanced order for allgroups, and heart rale was again recorded before

and during Ihc (ir.sl. and last situations on HieRBRAT. After responding to all the assertive situa-tions, the ASST was administered to assess the posi-tive and negative self-statements.

Dependent MeasuresThe subjects' written and tape-recorded responses

on the AKI, HYPO, and RBRAT were rated inde-pendently by two "blind" judges using a S-pointscale from 1 (unqualified acceptance) to S (unquali-fied refusal; McFall & Twentyman, 1973). Raterintcrcorrelation on the AKI was .92, and a t lestindicated that there was no difference between thetwo raters, t ( 88 )= .27 , p — ,79. Rater intcrcorre-lation on the HYPO was .56, and there was no dif-ference between raters, t(88) = .27, p~ .7<). Raterintercorrelation on the RBRAT was .90, and therewas no difference between the two raters £(88) =.39, p = .70. Overall reliability (as measured by in-terrater correlation) on the three tasks was .79.

Heart RaleHeart rate was measured using a spatially dis-

placed finger plcthysmograph that activated a photo-electric cell. It was placed on the ring finger ofthe nonwriting hand. A base rate was taken whilethe subject was filling out the CRI. Recordings werelater made before and during the first and sixthRBRAT situations.

Self-perceived TensionSubjects were asked to rate how nervous they felt

on a scale from 1 to 7 (1 — not all nervous and 7 =extremely nervous). This measure was taken duringthe CRI as a base rate and immediately after thefirst and sixth RBRAT situations.

Cognitive Self-statementsImmediately after responding to all of the 18

assertive situations, the subject was given the ASST.This is a 34-itcm questionnaire with 17 positiveself-statements that would make it easier to refusethe request and 17 negative self-statements thatwould make it harder to refuse. Examples of eacharc as follows:

Positive: I was thinking that it doesn't matterwhat the person thinks of me; I was thinking thatI am perfectly free to say no; I was thinking thatthis request is an unreasonable one.

Negative: I was worried about what the other per-son would think about me if I refused: I wasthinking that it is better to help others than tobe self-centered; I was thinking that the otherperson might be hurt or insulted if I refused.

Subjects were asked to indicate on a scale from1 to 5 how frequently these self-statements charac-terized their thoughts during the preceding assertivesituations (1 — hardly ever and 5 — very of ten) .

The ASST was conscnsually validated on an inde-pendent sample of 37 college students. Only those

914 ROBKRT M. SCHWARTZ AND JOHN M. GOTTMAN

items H i n t o b t a i n e d a, <)(1% agreement as lo whetherthey wen: positive, or negative in terms of facilitat-ing or interfering with refusal behavior were used.

Subjects were also asked to respond to an itemthat asked them to indicate which of four sequencesbest characterized their thought process in termsof the order in which they made positive and nega-tive self-statements. The sequences were as follows:Coping (—h), at first negative and later positive;unshaken doubt ( ), at first negative and laternegative; unshaken confidence (-|- - | - ) , at first posi-tive and later still positive; giving up (-\—), atfirst positive and later negative. This was intendedto assess whether subjects tended to sequence theirpositive and negative self-statements in differentways at different levels of assertivencss.

Problematic SituationsThe situations were worded similar to the follow-

ing: You have been standing in the ticket line althe movie theater for about 20 minutes. Just asyou are getting close to the box office, three peoplewho you know only slightly from your dorm comeup to you and ask if you would let them cut infront of you.

All situations were taken from the CRI. Situa-tions within each task were selected from the CRIlo representatively sample refusal items with vary-ing situational contexts, difficulty, and to whom therefusal was directed (to a close friend, friend, oracquaintance). The three tasks were sets of situa-tions that differed in the following ways:

1. For the AKI, six situations were presented inwritten form, and the subjects were required to re-spond in writing with what they thought was a"model" refusal response. The AKI was designed toassess the subject's content knowledge of a goodassertive response. CRI items were 2, 14, 20, 24,27, and 30.

2. For the HYPO, six situations were presentedon tape. In responding to the situations orally, thesubjects were told to imagine that a friend hadgiven in to an unreasonable request because thefriend did not know how to refuse. The subjectswere asked to imagine that the friend wanted toknow what to say at the time. The subject's responsewas tape recorded. CRI items were 4, 6, 10, 25,31, and 35.

3. For the RBRAT, six situations were presentedon tape in an attempt to create a simulation of areal situation. The subjects were told to imagine inas much detail as possible that they were beingconfronted by an unreasonable request. They weretold to respond naturally as though they were ac-tually talking to the person making the request. CRIitems were 12, 16, 22, 29, 32, and 33.

Since all situations for the three inventories weretaken from the CRI, it is important to ask whetherthe three inventories have similar psychometric prop-erties. To answer this question an independent sam-ple of 60 undergraduates took the CRI. Nonasser-tion scores were computed for the CRI. Also, thetotal number of items on which (he subject said he

or she would not refuse, and the to ta l number olitems on which the subject expressed discomfort-were computed separately for the AKI, HYPO, andRBRAT. A principal components analysis found onelarge nonasserlion factor that accounted for 54.30%of the variance. This factor had a high loading forCRI nonassertion (.85) and high loadings for AKI,HYPO, and RBRAT discomfort scores (.85, .65, and.79, respectively) but lower loadings for refusalscores on the AKI, HYPO, and RBRAT (.27, .50,and .50, respectively). Hence, discomfort scores wereused to test whether assertion items of differentialdiscomfort had been assigned to the three inven-tories. A repeated measures F test resulted in an Fratio of .71, p > .50, with refusal means of 2.30,2.43, and 2.52, respectively, for the AKI, HYPO,and RBRAT. Therefore, it appears that the threeinventories do, in fact, have similar psychometricproperties.

RESULTSData were analyzed for each dependent

variable using a 3 X 6 analysis of variancedesign with three groups (low, moderate, andhigh asserliveness) and six orders of admin-istration of the AKI, HYPO, and RBRATsituations. There were no significant GroupsX Order interactions. Results are presentedseparately for the order and groups maineffects.

Order EffectsSignificant order main effects were ob-

tained for the knowledge of content (AKI)situations, F ( 5 , 83) = 2.90, p - .019, andfor the RBRAT situation, F ( 5 , 83) - 3.96,p = .003. Subsequent tests using Tukey'shonestly significant difference (HSD) testshow that performance on the AKI was bestfor those subjects who received the AKI last;that is, after the HYPO and BRAT, Tukey'sHSD = 3.98, obtained difference(83) = 4.38,p < .05. Performance on the RBRAT wasbest for those subjects who received theRBRAT last (i.e., after responding on boththe AKT and the HYPO), Tukey's HSD =4.93, obtained difference(83) = 6.44, p < .OS.

Group DifferencesThe main effects for the assertiveness inde-

pendent variable were as follows:Knowledge of content ( A K I ) . No signifi-

cant differences were obtained between low-,moderate-, and high-assertive groups on theAKI, 77(2,83) - 2.n,p~ .19.

TASK ANALYSIS 915

Indirect delivery (HYPO). No significantdifferences between low-, moderate-, or high-assertive groups were obtained on the HYPOsituation, F ( 2 , 83) = .44, p = .63.

Direct delivery (RBRAT). Groups did dif-fer significantly on the RBRAT, F ( 2 , & 3 ) -8.26, p - .008. Using the Tukey HSD, allpairwise comparisons among the means onthe RBRAT were made. Only the high- andlow-assertive groups differed significantly,Tukey's HSD = 3.99, obtained difference =5.38, p < .01.

Heart rate. Pulse was recorded first as abaseline and then before and during RBRATSituations 1 and 6. No significant differenceswere obtained for base pulse, F(2, 83) = 1.34,p = .27, RBRAT 1 before pulse, F ( 2 , 83) =.28, p = .76, RBRAT 1 during pulse, F ( 2 ,83) = .55, p = .58, RBRAT 6 before pulse,F(2, 83) = .42, p = .66, or RBRAT 6 duringpulse, F(2, 83) = .65, p — .53.

To investigate the trials effect, a one-wayrepeated measures analysis of variance wasperformed. All groups showed significantchanges over trials—for low assertives, 7^(4,124) = 5.98, p < .001; for moderate assert-ives, F(4,160) = 6.13, p<.00l; and forhigh assertives, F(4,108) = 5.18, p = .001.For each group, pairwise comparisons amongtrial means were performed to determinewhich changes over trials contributed to thesignificance. No groups differed from CRTbase heart rate to before RBRAT 1, indicat-ing that the pulse reading was reliable. Eventhough all groups increased their heart ratefrom before RBRAT 1 to during RBRAT 1,only the low-assertive group demonstrated asignificant increase, HSD = 4.31, obtaineddifference(60) = 4.88, dj = 60, p < .01. ByRBRAT situation 6, no group increased theirheart rate significantly from before to duringthe situation. For low assertive, HSD = 4.31,obtained difference(60) = 3.38, p > .05; formoderate assertive, HSD = 3.87, obtained dif-ference (60) = 2.37, p > .05; for high assert-ive, HSD - 4.97, obtained difference(60) =.71, p > .05. Therefore, whatever differencesin heart rate increases may have existed be-tween low-, moderate-, and high-assertive sub-jects on Situation 1, these differences nolonger existed by Situation 6.

Self-perceived tension. Low-assertive sub-jects consistently reported themselves to bemore nervous than high assertives, with mod-erate assertives falling midway between. Thiswas true for base tension, F ( 2 , 83) = 6.57, p= .003, post-RBRAT 1 tension, f (2 ,83) =5.38, p = .007, and for post-RBRAT 6 ten-sion, F ( 2 , 83) = 5.51, p = .006. Using a one-way analysis of variance of average tension onRBRAT Situations 1 and 6 combined, groupsagain differed significantly, F(2,98) = 6.05,p < .01. Using the Tukey HSD test, it wasfound that low- and moderate-assertive sub-jects differed in self-perceived tension, HSD- .91, obtained difference(60) = 1.06, p <.05; low- and high-assertive subjects also dif-fered, HSD =1.15, obtained difference(60)= 1.28, p < .01. The moderate- and high-assertive subjects, however, did not differ sig-nificantly on self-perceived tensions, HSD =.91, obtained difference(60)= .22, p > .05.Three repeated measures analyses of variancewere performed to assess the trials effect. Allthree groups of subjects reported less tensionon RBRAT Situation 6 than on RBRAT Sit-uation 1: For low assertives, F(l, 31) = 6.00,p = .019; for moderate assertives, F(i, 40) =4.42, p = .039; and for high assertives, F(\,27) = 5.29, p = .028.

Cognitive sell-statements (ASST). Signifi-cant differences were found between low-,

Low ModerateAssertive Group

Kic . i iKK 2. Positive and negative self-statementscores as measured by Ihc ASST (85 ~ highest possi-ble score) .

916 ROBERT M. SCHWARTZ AND JOHN M. GOTTMAN

moderate-, and high-assertive subjects on posi-tive self-statements, P(2, 83) — 6.53, p =.003. P^ven stronger differences were found onnegative self-statements, 7/(2,83) = 36.25, p— .00001 (see Figure 2). High-assertive sub-jects had more positive and fewer negativeself-statements than low-assertive subjects;moderate-assertive subjects fell midway be-tween.2 The Tukey HSD test indicated thatonly the low- and high-assertive groups dif-fered significantly on positive self-statements,HSD - 6.41, obtained difference(60) - 7.99,P < .01. On negative self-statements, however,all groups differed significantly as shown bythe following pairwise comparisons amongthe groups—for low- and moderate-assertivegroups, HSD = 5.63, obtained difference(60)-- 8.08, /; < .01; for moderate- and high-as-sertive groups, HSD — 5.63, obtained differ-ence (60) = 7.81, p < .01; for low- and high-assertive groups, HSD = 5.63, obtained dif-ference(60) = 15.89, p < .01.

To test for an interaction between groupsand self-statements, a repeated measuresanalysis of variance was performed with twolevels of self-statements (positive and nega-tive). A significant interaction was obtained,7''(2,98) = 29.82, p < .0001.

To investigate differences between positiveand negative self-statements within groups, at test for matched samples was performed.The low-assertive group had more negativethan positive self-statements, but this differ-ence was not significant, F( 1,31) = 1.77,/> —.1,90. On the other hand, the moderate grouphad significantly more positive than negativeself-statements, ^(1,40) = 24.65, p — .001.The high-assertive group also had significantly

TAHUi 1Cm- H Q U A K K CONTINGENCY TAIN.E SHOWING TireI'lCTCCKNTAGKS OF SUBJECTS ClIOOSING KACII ()!' THE

FOUR SEW- STATEMENT SEQCKNCKS

Self--statement sequence

Unshaken Unshaken GivingAssertive Coping doubt confidence up

group ( — +) ( ) (++) (+ —)

2220

7

2274

M 2261 1282 7

more positive than negative self-statements,F(l, 27) = 66.51, p < .0001.

To investigate whether the assertive groupsdiffered in the way they sequenced positiveand negative self-statements, a chi-square con-tingency table test was performed and foundto be significant, X

2(6) = 16.01, p — .025. Agreater percentage of the high-assertive sub-jects checked the item characterized by "un-shaken confidence" (+ +) than the low-as-sertive subjects, with the moderate subjectsfalling midway in between (see Table 1).Within the low-assertive group, there wereindividual differences in the sequence of posi-tive and negative self-statements, with nopreference shown for any of the sequences(excluding unshaken confidence). Tn fact, thealternative sequences were chosen by equal(22%) percentages of low-assertive subjects.Those in the moderate group not character-ized by "unshaken confidence" did show apreference for the coping sequence (— +),with 20% choosing this sequence.

In addition to the assertive and nonassertivescores on the CRT, McFall and Lillesand(1971) calculated the difference between theassertive and nonassertive scores. Differencescores in the present investigation rangedfrom a low of —24 to a high of 34. In anattempt to gain greater descriptive and pre-dictive precision, a polynomial regression wasperformed for positive and negative self-state-ments on assertiveness. The relationship be-tween assertiveness and both positive andnegative self-statements was best described bya linear function; for positive self-statements,7 ^ ( 1 , 9 7 ) = 63.1, p < .01. Neither the quad-ratic nor the cubic terms were significant (seeFigures 3 and 4).

DISCUSSION

Items selected from the CRI to form theAKT, HYPO, and RBRAT did not differ sig-ificantly on perceived discomfort. Discomfort

2 The self-statements that distinguished low andhigh asscrtives the most tended to fall into thefol lowing categories: (a) concern about negativeself-image and fear of being disliked and (b) other-directed versus self-directed—concern for the otherperson's position, feelings, and needs.

TASK ANALYSIS 917

80

75

co 70

| 65

Is 60

<? 55u_GJ 50CO

40CO

2353025

-20 -15 -10 -5 0 5ASSERTION

10 15 20 25 30

FIGURE 3. Polynomial regression of negative self-statements on CRT assertion difference score.

scores loaded highly on the CRT nonassertion have important implications in describing thefactor. The three tests thus have similar psy- nature of the response deficit in nonassertivechometric properties, and the absence of be- subjects.tween-group differences on the AKI and Nonassertive subjects did not differ fromHYPO and their presence on the RBRAT highly assertive subjects in their ability to

-20 -15 -10 -5 0 5 10 15 20 25 30ASSERTION

FIGURE 4, Polynomial regression of positive self-statements on CR1 assertion difference score.

918 ROBERT M. SCHWARTZ AND JOHN M. GOTTMAN

construct a written assertive response (AKl)or to verbally deliver the assertive responsein hypothetical, and therefore safe, situations(HYPO). Although results on the HYPOmust be interpreted with caution because ofreduced interrater reliabilities, the finding ofno group differences on both the HYPO andAKT indicates that perhaps low-assertive sub-jects do not suffer from a content knowledgedeficiency or from an inability to deliver anassertive response, in the sense of merelyknowing (formulating and articulating) theresponse. Nonassertive subjects do, however,lack the ability to perform these componentsof the task when confronted with a situationmore similar to the real stimulus situation.

Since the population of the present investi-gation was similar to that of McFall and hisassociates, perhaps the effective components ofcoaching and rehearsal do not involve teach-ing the content of a good assertive responseor the method of delivering it as narrowlydefined above. Instead, the coaching and re-hearsal component may focus on those dimen-sions of delivery skill that are involved inreal-life situations. This raises the question ofexactly which dimensions of delivery skill arenecessary for competent assertive behavior inreal l i fe and what specifically was operating inthe treatment program used by McFall andhis colleagues. Order effects obtained on theRBRAT and AKT suggest that the assessmentprocedure was itself an intervention thatenhanced the performance of all subjects onthe RBRAT when the RBRAT followed theAKI and HYPO, and on the AKl when theAKI followed the RBRAT and HYPO. Thissuggests that mere practice in responding maybe beneficial without coaching or rehearsal.

The present study suggests that physio-logical differences do not exist between low-,moderate-, and high-assertive groups. Thismay partially account for the results of anexperiment comparing desensitization withassertion training. Bouffard (1973), in a 2 X2 factorial design, studied the relative effec-tiveness of group desensitization, group re-sponse acquisition training, and the combinedprocedures compared to an attention placebocontrol group. Although Bouffard's range ofassertivencss, as measured by McFall and

Lillesand's (1971) CRT, was limited to mod-erate ranges, he found that desensitization didnot improve performance on McFall's BRAT.Furthermore, a combined treatment packageof desensitization and response acquisition,which would seem to be an ideal treatmentpackage, failed to improve performance onthe BRAT relative to the attention placebocontrol group. Differences in the present in-vestigation were obtained in self-perceivedtension on all situations despite the fact thatall groups were reporting less tension overconsecutive situations. These findings maysuggest that differences in competent deliverymay not be a function of physiology but whatsubjects are telling themselves about theirphysiology.

McFall and Marston (1970) did obtainheart rate changes as a result of their asser-tion training program. These differences wereon the order of 5-7 beats per minute. The re-sult of the present investigation that low-assertive subjects' change in heart rate frombefore to during was significant on RBRATSituation 1 but not on RBRAT Situation 6suggests that heart rate change could be aby-product of increased confidence producedby the graded escalation of situations oversessions. Session 2 in McFall and Marston'sstudy escalated the situations over those inSession 1 so that refusal became more diffi-cult. The placebo therapy control group wasnot exposed to additional problem situationsor situations in escalating fashion. Pulse ratechanges obtained from the treatment groupafter responding dropped an average of 6.47beats/min. This change is similar to the 5beats/min change obtained in the present in-vestigation.

The size of the effects in the present taskanalysis suggests that the most likely sourceof nonassertiveness in low-assertive subjectscould be related to the nature of their cogni-tive positive and negative self-statements.Low-assertive subjects had significantly fewerpositive and more negative self-statementsthan high-assertive subjects. Even though bothpositive and negative self-statements showedhighly significant differences, the effect wasmost dramatic for negative self-statements.This agrees with Meichenbaum and Camer-

TASK ANALYSIS 919

on's (1973) findings that a variety of patientshad thought patterns characterized by nega-tive and maladaptive self-statements. It isworth noting that not one low-assertive sub-ject in the present investigation had cognitiveself-statement scores that were similar tothose of the high-assertive group. This dem-onstrates an extremely strong and consistentgroup difference that was not spuriously pro-duced by averaging the data.

Comparing positive and negative self-state-ments within the groups, the moderate- andhigh-assertive subjects had significantly morepositive than negative self-statements; thelow-assertive subjects did not differ in theirpositive and negative self-statements. Thisindicates that highly competent assertive peo-ple have a greater discrepancy between theirpositive and negative self-statements, in favorof the positive ones. There is little doubt intheir minds about the appropriateness of theiraction. The low-assertive subjects, on theother hand, can be characterized by an "in-ternal dialogue of conflict" in which positiveand negative self-statements compete againstone another. Such a state would hardly facili-tate appropriate and effective assertive be-havior. These findings suggest that some typeof cognitive restructuring (Ellis & Harper,1961) or manipulation of cognitive self-state-ments (Meichenbaum, 1972) may be an ap-propriate form of treatment for nonassertive-ness.

Direct intervention using cognitive self-statement modification may enhance transferof training effects. McFall and Marston(1970) found that transfer effects occurred onone of five measures in a telephone follow-upresistance to a magazine salesman. McFalland Lillesand (1971) failed to show a signifi-cant difference between treatment and assess-ment-placebo control groups in their telephonefollow-up. McFall and Twentyman (1973)reported the results for four studies disman-tling a standardized scmiautomaled assertiontraining program. In the first study no trans-fer of training was demonstrated in a tele-phone follow-up. In the second study experi-mental groups again did not show transfer oftraining in two in vivo resistance to pressuremeasures, although performance was improved

on behavioral and self-report measures inuntrained situations of the RBRAT. Thethird study again found no transfer effects fora pressuring telephone call. The fourth study,however, did result in transfer of trainingusing a modification of the all-or-none pro-cedure for measuring transfer of training usedin their third study to a more continuous pro-cedure. Although it may be that obtainingtransfer effects is a function of the assessmentprocedure, taken together it is clear thattransfer of training is an issue in response-acquisition methodology.

Meichenbaum and his associates have beensuccessful in obtaining transfer effects using acoping self-statement intervention with hos-pitalized schizophrenics (Meichenbaum &Cameron, 1973), speech-anxious subjects(Meichenbaum, Gilmore, & Fedoravicius,1971), and test anxiety (Meichenbaum,1972). Glass, Gottman, and Shmurak (1976)collaborated in a study of the relative effec-tiveness of coaching and rehearsal versus cog-nitive self-statement modification in a datingskills program for girl-shy college males. Theyfound that the greatest transfer effects tountrained laboratory role-playing situations,and ratings made by females the subjectscalled for a date, were obtained by the cogni-tive self-statement intervention. These findingsare consistent with the current task analysisstudy and suggest that transfer of training-effects may be enhanced with a cognitive self-statement assertion training intervention.

REFERENCES

Eouffard, D. L. A comparison of response acquisi-tion and desensitization approaches to assertiontraining. Unpublished doctoral dissertation, Indi-ana University, 1973.

Buchwald, A. M., & Young, R. D, Some commentson the foundations of behavior therapy. In C. M.Franks (Ed.), Behavior therapy: Appraisal andstatus. New York: McGraw-Hill, 1969.

Ellis, A., & Harper, R. A. A guide to rational living.Knglewood Cliffs, N.J.: Prentice-Hall, 1961.

Gagne, P. M. Curriculum research and the promotionof learning. In R. E. Stake (Ed.), AKRA Curricu-lum Monograph Series No. 1. Chicago: RandMcNally, 1967.

Galassi, J. P., Del.o, J. S., Galassi, M. T>., & Bastein,S. The college self-expression scale: A measure ofassertiveness. Behavior Therapy, 1974, 5, 16S-171.

Glass, C, R,, Gotlman, J. M., & Shmurak, S. H. Re-sponse acquisition and cognitive self-statement

920 ROBERT M. SCHWARTZ AND JOHN M, GOTTMAN

modification approaches to dating skills. Journalof Counseling Psychology, 1976, 23, 520-526.

McFall, R. M., & Lillcsand, D. B. Behavior rehearsalwith modeling and coaching in assertion training.Journal of Abnormal Psychology, 1971, 77, 313-323.

McFall, R. M., & Marston, A. R. An experimentalinvestigation of behavior rehearsal in assertivetraining. Journal of Abnormal Psychology, 1970,76, 295-303.

McFall, R. M., & Twenlyman, C. T. Four experi-ments on the relative contribution of rehearsal,modeling, and coaching to assertion training.Journal of Abnormal 1'sychology, 1973, 81, 199-218.

Mcichenbaum, D. IT. Examination of model charac-

teristics in reducing avoidance behavior. Journalof Personality and Social Psychology, 1971, 17,298-307.

Meichenbaum, T>. H. Cognitive modification of test-anxious college students. Journal of Consulting andClinical Psychology, 1972, 39, 370-380.

Meichenbaum, D. H., & Cameron, R. Training schizo-phrenics to talk to themselves: A means of de-veloping attcntional controls. Behavior Therapy,1973, 4, 515-534.

Mcichenbaum, D. H., Gilmore, J. B., & Fedoravicius,A. Group insight versus group dcscnsitization intreating speech anxiety. Journal of Consulting andClinical Psychology, 1971, 36, 410-421.

(Received January 12, 1976)