SIDE EFFECTS T0 ORAL CONTRACEPTIVES: ACROSS LAGGED PANEL ANALYSIS OF AN ATTITUDE / BEHAVIOR RELATIONSHIP Dissertation for the Degree of Ph. D. MICHIGAN STATE UNIVERSITY M. CHRISTINE FALVEY 1975 LIBRARY nth imn 523:3 «1*. R ’ vi. I Umvh: m4“ ' ABSTRACT SIDE EFFECTS TO ORAL CONTRACEPTIVES: A CROSS LAGGED PANEL ANALYSIS OF AN ATTITUDE/BEHAVIOR RELATIONSHIP By M. Christine Falvey The relationship between attitudes and behavior is examined tusing a cross lagged panel technique to uncover causal vectors. An attitude questionnaire on oral contraceptives, birth control and woman's role, and a side effect questionnaire was given to 59 women who were beginning first time use of oral contraceptives. The same questionnaires were given after three months of use. The cross lagged panel analyses of the data indicates there are slight but significant causal vectors of attitude toward oral contraceptives affecting side effects, attitude toward oral contra- ceptives affecting attitude toward woman's role, and attitude toward birth control affecting attitude toward woman's role. Theoritical difficulties of the cross lagged technique con- clusions which can be drawn if only two data wave are used are discussed. Clinical implications and implications for methodology in attitude behavior research are also discussed. SIDE EFFECTS TO ORAL CONTRACEPTIVES: A CROSS LAGGED PANEL ANALYSIS OF AN ATTITUDE/BEHAVIOR RELATIONSHIP By > M. Christine Falvey A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Psychology 1975 Dedication: To my mother and father Gertude Gleason O'Connor Danial Vincent O'Connor ii ACKNOWLEDGMENTS Thanks and appreciation to all my committee members Andrew Barcley Lawrence Messe Mary Ryan and especially to William Crano for his invaluable support and guidance. iii TABLE OF CONTENTS Page LIST OF TABLES . . . . . . . . . . . . . . . . vi LIST OF FIGURES . . . . . . . . . . . . . . . . vii LIST OF APPENDICES . . . . . . . . . . . . . . . viii INTRODUCTION . . . . . . . . . . . . . . . . . 1 Cross-Lag Panel Analysis . . . . . . . . . . . 4 'LITERATURE REVIEW: MINOR SIDE EFFECTS OF ORAL CONTRACEPTIVES . 10 Introduction . . . . . . . . . . . . . . . 10 A Sample of Clinical Studies . . . . . . . . . . 12 Double-Blind Placebo Studies . . . . . . . . . . 15 .ySummary on Minor Side Effects. . . . . . . . . . 26 Studies on Psychogenic and Placebo Variables. . . . . 27 Summary . . . . . . . . . . . . . . . . . 29 Research Hypotheses . . . . . . . . . . . . . 3O METHODOLOGY. . . . . . . . . . . . . . . . . . 31 Design . . . . . . . . . . . . . . . . . 31 Instrument Construction. . . . . . . . . . . . 31 The Attitude Scale . . . . . . . . . . . . . 33 Results from Pretest. . . . . . . . . . . . . 41 Subjects. . . . . . . . . . . . . . . . . 44 Data Collection Procedure . . . . . . . . . . . 44 RESULTS 0 C O O O O O O O O O O O O O O O O C 46 Introduction . . . . . . . . . . . . . . . 46 Main Analysis and Results . . . . . . . . . . . 46 iv DISCUSSION . Clinical Implications Interrelations Between Attitudes. Methodological Implications Implications for Attitude/Behavior REFERENCES APPENDICES Research . Page 64 66 69 69 71 72 77 LIST OF TABLES Reported Side Effects for Placebo Versus True Medication (oral contraceptive) Groups, with and without warning . First Internal Consistency Test (alpha) for Each of the Scales Constructed From the Factors of the Analysis of Pretest Data 0 O O I O O O Q C I O O C O 0 Sample of Appendix 2 Showing Values of Factor Loadings and ItemrScale Correlation for Each Item. . . . . . Second Internal Consistency (alpha) Test for Each of the Reduced Scales Constructed From Pre-test Data . . . . Third Internal Consistency (alpha) Test for Each of the Reduced Scales Constructed From Pre-test Data . . . . Correlations Between Each Side Effect Item and Each AttitUde scale C O O O O O O O 0 O O 0 O 0 Significance Level of t Test for Difference Between Cross Lag Correlations. . . . . . . . . . . . Alphas For Each Scale at Time 1 (first unit) and Time 2 (three months after onset of pill use) . . . . Original Uncorrected Cross Lagged Correlations, Variable Partialed Out, and New Partial Variable Correlations. . vi Page 25 34 35 36 37 38 49 56 6O Figure LIST OF FIGURES Page Hypothetical Relationship Between Attitudes and Adjustment 0 O O O O O O O O O O O C O I O 6 Panels of A11 Possible Pairs of Variables (Correlations Corrected for Attenuation) . . . . . . . . . . 48 Model Cross Lagged Panel . . . . . . . . . . . 58 Panels of A11 Variables with Significant Cross Lagged Correlations (Values Not Corrected for Attenuation) . . 6l Panels of All Pairs of Variables with (Originally) Significant Cross Lagged Correlations with Common Variable Partialed Out. . . . . . . . . . . . 62 vii Appendix 1. Original Pretest . . . . . . . . 2. Factor Loading Item--Whole Correlation for Each of Pretest . . . . . . . . . . 3. Factor Loadings on Reduced Pretest . . 4. Frequency of Named Side Effects . . . 5. Responses to Side-Effect Items on Pretest 6. Side Effects and Attitude Questionnaires 7. Main Analysis: Correlation Matrix . . 8. Matrix of Correlations for 3 Wave Data . LIST OF APPENDICES viii Page 78. 89 93 96 99 101 103 105 INTRODUCTION It has long been the notion that our "attitudes," or how we "feel" or "think" about issues is a strong determinant of how we behave regarding these issues. Some dissenters have always been present to remind us of the perhaps more cynical possibility that our actions are not always a product of our minds but that some- times attitudes are formed after the fact to rationalize our ' actions. Whatever the truth, there is a profound assumption that belief and action are strongly tied. Even when we observe that: "It is one thing to say it, but it's another to do it," we acknow- ledge by the necessity of occasional exception our underlying acceptance of the attitude/behavior relationship. Yet consistency between attitudes and behavior does not always appear in research results. Studies fall into three categories: (1) Those which demonstrate that a certain attitude pre- dictably precedes an expected behavior; or that a change in atti- tude is followed by a corresponding change in behavior. For example, Greenwald (1965, 1966) undertook two studies which demonstrated that a communication to junior high students which stressed the importance of vocabulary learning resulted in an increase in the number of problems worked on. Kothandapani (1971) l found measured "intent to act" was a good predictor of subsequent contraceptive seeking behavior. (2) Those which demonstrate that engaging in a certain behavior is followed by a change in attitude.* Such studies have arisen in research on cognitive dissonance theory and Bem's self-perception theory. As an example here there are classic studies like Festinger and Carlsmith's (1959) in which they paid subjects $1 or $20 to engage in counter-attitudinal behavior. The subjects paid the smaller amount later indicated attitude change congruent with the behavior. Dissonance research, then, presents a circumscribed version of behavior causes attitudes hypothesis. The modified hypothesis demonstrated by this research is: behavior causes attitudes to change if such a change will reduce dissonance. This has been a much attacked and modified theory and is undoubtly over— simple. Nevertheless, it does seem to stand in some respects and contain an element of truth. Research on self-perception theory tries to confirm a less circumscribed hypothesis: Behavior occurs, is perceived and atti- tudes constructed or modified to fit the behavior. Unfortunately such confirmation is often indirect since the subject is usually tricked into believing he behaved a certain way and then attitudes *Since this is the less intuitively obvious possibility no one undertakes the compliment of a Kothandapani—like study. That is: researchers do not observe a behavior and then measure atti- tudes as a way of demonstrating the possibility that behavior causes attitudes, although the reverse, measuring attitudes and then observing behavior, has been an acceptable way to demonstrate the possibility that attitudes cause behavior. are measured. For example, in a study by Ross, Insko and Ross (1971) subjects were told they had answered an item on a questionnaire a certain way. Subsequent attitude assessment revealed that subjects' attitudes varied according to what they had been told about their (supposed) previous response. (3) Studies in which no relationship between attitudes and behavior is demonstrated; or in which results are inconsistent. Fleishmann et a1. (1955) altered foremens' attitudes toward leader- ship, but this had no effect on their behavior. In other studies Janis and Feshbach (1953) altered attitudes toward dental hygiene land Maccoby et al. (1962) altered mothers' (of infants) attitudes toward toilet training but in neither case was behavior altered. There are a number of possibilities which could explain these contradictory results: (1) Human behavior may often be inconsistent so that a person's behavior does not always reflect his attitude; (2) much of the research on attitude change has been con- ducted in a laboratory situation where attitudes were experimentally manipulated. It may be that such manipulation is too superficial or short-lived in its effect to influence behavior; (3) measures of either the attitude and/or of the behavior may be inadequate; (4) often research measures a general attitude and a specific behavior (e.g. attitude toward religion and church attendance). In these cases one would not expect good correspondence; (5) the studies have been conducted in different settings, measuring different attitudes and behaviors with different instru- ments. In other words there is no standardization of techniques. Whatever the reasons for the contradictions, they remain. In addition to this problem is the fact that many of these studies depend on correlational type relationships to show that attitudes "cause" behavior -- for example, studies which relate a range of attitudes across subjects to subsequent behavior. There are studies with a more experimental orientation which make causal inferences legitimate -- for example, studies which measure attitudes, attempt to produce change in half the subjects while using the others for controls and then compare subsequent behaviors. These studies usually depend on a rather artificial manipulation of attitudes in an experimental setting and the results are often short-lived. Cross—Lag Panel Analysis What seems to be needed, then, is a method of comparing attitudes and behavior in "real world” situations in such a way that causal inferences can be made. A recent technique, cross- lag panel analysis gives us just such a tool. Originally suggested by Campbell (1963) as a modification of Lazarfeld's 16-fold table technique, this method allows causal inferences to be made from correlational data. Detailed descriptions of this technique have appeared else- where (Campbell, 1963; Campbell and Stanley, 1963; Pelz and Andrews, 1963; Kenney, 1971; Crano et al., 1971). The article by Crano et al., gives a good non'technical explanation of the logic of the technique. The method is based on the notion of time precedence.* If one event consistently precedes another and the reverse is not true then two causal alternatives are entertained, either: (1) Event one is presumed a cause of event two; or (2) Both event one and two are caused by another common factor. In experimental designs, control of the application of the independent variable allows the experimenter to rule out the second alternative and to assume that his independent variable (event one) has caused observed changes in the dependent variable (event two). - For correlational designs there has been traditionally no way to rule out the second alternative, i.e., cause by another common factor. However, given correlational data at more than one time point it is possible to rule out the second alternative and to make the causal assumption (alternative one). Suppose, for example, that we have data on adjustment and attitudes of women on oral contraceptives just as they begin using oral contraceptives, and that we have the same measures repeated three months after onset. The correlations shown in Figure l are obtained. The synchronous correlations (r att1 adj1 and r att2 adjz) tell to what extent the variables are correlated at a given time. The autocorrelations (r att1 att2 and r adj1 adjz) give information about the test-retest reliability of each instrument. The cross *The explanation given here is a simplified one. Some of the more complex details of the technique will be discussed in the results section. .uaoaumanv< mam mouauHuu< cowsuom mHsmnoHumHoM HMUHumnuom%mll.H mustm N nv¢ [I IIAaOHumHouuoo cunmv anmHflvmu Hnu< I Nn H M]. \\ \\ AnOHumHouuoo A:0HumHouuoo mzocounoahmv .wm up u .I.1/ \\ \\ Fmaocouaushmv an mm m [I II \\ \\ h H m v 3 u \/ cm 3 u \ \ I l \ \ .\ NuumHfifim / / \ H / \ l \\ AmGOHumHouuou wowme mmouov H Nu: . 34 ACOHuMHoHHoo ousmv NuumHuumu N oEHu H maHu lagged correlations (r att1 adj2 and r att adjl) tell how one vari— 2 able measured at an earlier time is correlated with the other measured at a later time. The question of causality is answered by comparing the two cross lagged correlations. There are three possibilities: (l) R attl adj2 > r attz adjl, i.e., poor attitudes con- sistently precede poor adjustment but poor adjustment does not precede poor attitudes. This means causal factors are in the direction of attitudes causing adjustment. (2) R attz adjl > r attl adjz, i.e., poor adjustment is followed by poor attitudes but the reverse is not true. This means the causal factors are such that adjustment is causing attitudes. (3) R attl adj2 = r attz adjl, i.e., the magnitude of relationship between prior attitudes and later adjustment is the same as that between prior adjust- ment and later attitudes. Such a result indicates either that there is no causal relationship between adjustment and attitudes, or that the correlation occurs because of some third variable affecting the two under investigation. Thus, depending on the results of the comparison between the two cross lagged correlations it is possible to determine either that: (1) Attitudinal variables cause adjustment status. (2) Adjustment status causes the attitude changes. (3) 0r there is no causal relationship between attitude and adjustment status. In more rigorous terms we should actually say that we can determine "the direction of the preponderance of causation" and not that "X causes Y" (however the latter terminology is imployed as less awkward). By the former terminology we specify that reciprocal causation is possible -- for example, positive attitudes causing good adjustment which results in further increases in positive attitudes. However the technique does determine which is the most preponderant variable as well as the direction of causation. It should be noted that while some authors (e.g., Campbell, 1963; Campbell and Stanley, 1963) believe the cross lagged panel technique can rule out common-factoredness, and thus allow the causal inference, there is possibly a more conservative explanation. The models constructed by Kenny (1971) are used by him only to rule out synchronous common factors. A synchronous common factor is one which causes two variables simultaneously. However, the possibility of a cross lagged common factor still exists, i.e., a common factor 'causes one variable and at a later time causes the other. The first variable then is characterized as a prior symptom of the common cause. Thus, in dealing with two variables X and Y one might attain a cross lagged correlation such that: r > r . Logically the x1Y2 x2Y2 interpretation is the X and Y do not have a synchronous common cause and thus, either: (1) X causes Y; (2) or a common third factor causes X and then at a later time causes Y. As an example of the second interpretation, suppose there is a hormone which acts quickly at one site producing result X and slowly at another site producing effect Y. X and Y would be highly correlated; the cross lagged common factor is the hormone; the change X is a prior symptom (i.e., prior to Y) of the common cause (the hormone). Thus, for the example proposed here, given results in the direction r att adj2 > r att l adjl, there are two interpretations. 2 Sandell (1971) has suggested that by using data at three or more (rather than 2) time points it is possible to rule out even the lagged common factor hypothesis and thus be left only with the causal hypothesis. Although not usually cited by psychologists some studies on attitude and behavior, using cross lagged panel analysis,* have been done. Using old data from the Survey Research Centers' Economic Behavior Program, Pelz and Andrews (1963) analyzed the extent to which variables (including attitudes) affected buying behavior. Generally the direction of causation which emerged from the analysis ' matched that of expert opinion. So there is at least some empirical evidence of attitudes causing behavior in a natrual or "real world" situation. For the study undertaken here a popular and important topic was used to look at the attitude behavior relationship with the cross lag technique: the extent to which side effects to birth control pills are related to attitudes. This topic has the added advantage of allowing an exploration into the relationship between attitudes and non-voluntary, psychosomatic behavior. *This method has been used on other psychological problems, for example, Crano et a1. (1971) used it to study I.Q. and achieve- ment in school children of different S.E.S. LITERATURE REVIEW: MINOR SIDE EFFECTS OF ORAL CONTRACEPTIVES Introduction In the use of drugs, the differentiation of pharmacogenic and psychogenic effects has always proved a problem in evaluating both the effectiveness of the drug and the validity of reported side effects. This second apsect, the validity of side effects, has perhaps nowhere been no more controversial and hotly debated than in the study of oral contraceptives. Almost every conceivable condition, ranging from cancer to hair loss, has at one time or another been attributed to oral contraceptives.{nA.recent informational pamphlet \M" published by the AMA (1970) noted more than 15 possible symptoms. Scores of articles have appeared in which the authors contend either that a certain symptom is or is not attibutable to oral conctracep— tives.'?The debate has culminated in a study by Goldzieher et a1. '- ,- (1971). As explained in greater detail below, this study investi- gated the most commonly reported minor side effects using a double blind crossover technique and a large subject population (N = 356). The authors found that, except for one high estrogen preparation, the incidence of minor side effects was no higher for women on oral contraceptives than for those on placebos. Such a result strongly suggests that the minor side effects are at least in part due to psychogenic factors. Because oral 10 ll contraception is a widely used, highly effective, and easily admin- istered form of contraception it can be considered the best form yet devised, and will probably remain so for a number of years. Minor side effects seem to be a barrier in the use of oral contracep- tives. For example, Westof and Ryder (1968), found that 21 percent of women who were using the pill discontinued because they experienced unpleasant side effects. And Herzberg (1970) found that one in three women discontinued use because of minor side effects (weight increase 35%, headache 37%, decreased libido 27% and depression 21%). The importance of discovering just what the pshychological determinants 'of these side effects are is obvious. Specifying these determinatns may give us the potential to reduce greatly the experience of minor side effects, or to identify women who will experience them and pre- scribe alternative birth control forms. If this potential is realized, it would increase the acceptability, use and continuation rates of oral contraceptives, make use easier and more pleasant, and suggest constructive measures which should be taken in introducing oral contraceptives to a new population of users (for example in developing countries). The problem of psychogenic side effects is by no means limited to oral contraceptives. Goldzieher (1968) reports that interuterine devises (IUD) also show a high discontinuation rate, often attributed to side effects identical to those reported for oral contraceptives. In a general way, side effects have been an elusive problem in the use of many other drugs (Hoingfeld, 1968). Possibly, if we can isolate the determinants of side effects for oral contraceptives, 12 this will also give us a better understanding of side effects experienced with other forms of contraception and suggest parallel determinants of psychogenic side effects for other types of medica- tion. Among the most frequently reported minor side effects of oral contraceptives are nausea, breakthrough bleeding, weight gain, reduction in libido, and depression. Bakker and Dightman (1964) classified the first three as pharmacogenic and the latter three psychogenic. The classification is based on the logic that pharma- cogenic side effects are regulated through dosage change or occur with consistency and with a specific pattern of regularity. Perhaps because this classification has been commonly accepted, much of the controversy has centered around the hypothetically psychogenic side effects, depression, weight gain, and reduction in libido. A Sample of Clinical Studies Depression. Lewis, et al. (1969) investigated the incidence of depression for oral contraceptive users. His design used two groups of women, one group was using oral contraceptives, the other was not. {HE found a significantly higher incidence of depression in the oral contraceptive groupie Lewis concluded that oral contra- ceptives were responsible for the higher incidence of depression.r However his conclusion is unwarranted given the design used.\_There is evidence that the population of women who use oral contraceptives are different from those who do not. For example, their motivation for their choice of contraceptives is different, (Nilsson and Almgren, 13 1968); they are more frequently smokers than are nonrusers (Kay, Smith and Richards, 1969); and they show a higher rate of cervical dysplasia prig£_to the use of oral contraceptives than non-users. Even if the populations were identical in all respects, Lewis's design could not differentiate between depression due to the pills and depression due to psychogenic factors involved in using the pills. Case reports of depression following use of oral contracep— tives have appeared (e.g., Daly, et al., 1967). The depression is usually attributed to the medication; however, there is not adequate ' control which would warrant such a conclusion. Reports of these two methodological types, i.e., post facto comparison between a group of women using the pill and a group not, and case reports, are fairly common, but no firm conclusion can be drawn from them. Changes in Libido. Another commonly claimed side effect of oral contraceptives is that of decreased libido; research indicates however that this is probably not a pharmacogenic effect. Because libido is difficult to measure, frequency of intercourse has often been used as an operational definition of libido. Pincus, et a1. (1959) found an increase in the frequency of coitus over the first year of oral contraceptive use, and a tendency for the frequency to decline thereafter, but not below premedication levels. Rogers and Ziegler (1969) similarly found an early increase in frequency of coitus in two groups of married subjects, one in which the wife took oral contraceptives and the other in which the husband had a vasectomy; l4 Thereafter, there was a slight decrease but not below average levels for the age group and not below, or even to, pre non-coital contra— ception. Zell and Crisp (1964) after observing 250 private patients, felt that oral contraceptives exerted a "definite improving influence" on sexual adjustment. Weight Gain. The complaint of weight gain has also been investigated by a number of researchers. Bakker and Dightman (1966) kept weight records on subjects who were using oral contraceptives. For the group there was no significant trend in gain or loss over _time. (However, there were fluctuations in some subjects: 30 percent showed an increase in weight due to fluid retention as distinct from fat deposition. Outside of this group who showed the fluid retention problem, weight gain and loss did not appear to be pill related. Field trails by Mears (1961) showed no significant change in weight for patients on oral contraceptives. On the other hand, in the field trials by Pincus (1959) the majority of subjects showed a weight gain. The difference in results may be attributable to a difference in dosages. Pincus used a preparation containing 10 mg. of norethynodrel and 0.5 mg. ethinyl estradiol 3-methyl ether, while the Mears trials used tablets containing 2.5 mg. norethynodrel and varying amounts of estrogen. In addition to the types of studies reported above, a number of articles based on the clinical experiences of psychiatrists have appeared. Litz (1969) believes that reduction in libido and depres- sion in women on oral contraceptives are psychogenic reactions, and 15 suggests that important factors in these reactions are: cultural background, attitude toward infertility, pregnancy, childbirth, and parenthood, and interaction between husband and wife. A similar short report by Gluckman (1969) stated that about one fourth of all women seen in psychiatric practice who take oral contraceptives attribute at least some of ther symptomatology to the medication. ' centering He discussed psychological reasons for their "scapegoating,' on the woman's motivation for taking oral contraceptives, and possible ambivalence about it. The author points out that there is an impor- tant distinction to be made between a woman's symptoms and her _reactions to them.£rFinally, Orchard (1969) concluded that while some side effects are pharmacogenic, others (weight gain, loss of sexual interest and depression) probably have a large psychogenic component. He considered the following variables important: 1) increased self observation and suggestion, 2) psychiatric health of the patient, and 3) interaction between the prescribing physician and the patient. ' Conclusions drawn from such clinical observations--while IT“ interesting and suggestive--off little in the way of supporting empirical evidence. Studies discussed thus far are a small sampling of many which have appeared. But because all these studies suffer design inadequacies compared to double-blind type studies, emphasis in this literature review will be on the latter. Double-Blind Placebo Studies Goldzieher, et al., 1971. Goldzieher's subjects were women attending the Research Clinic of the Southwest Foundation for l6 contraceptive assistance. None of the women had ever used oral contraceptives before. Five preparations were used in the study, one was a placebo, the other four were all contraceptives, a high estrogen sequential (Oracon), two low estrogen combination tablets (Ovulen and Norinyl), and a no estrogen chlormadinone acetate tablet. All five preparations were identical in appearance. Assignment of the subjects to their preparations was random— ized and the identity of the preparation made by code which was not broken during the study. All the subjects on placebos and 10 percent of the true medication group were instructed to use a vagainal foam, thus use of a foam did not identify the placebo takers. There were three cross over points during the study. After the fourth cycle one half of the placebo group, one half of the low estrogen group and all of the high estrogen subjects were changed to the no estrogen preparation. Subjects initially on the no estrogen preparation were switched to the high estrogen preparation. Another important feature of the study was the use of a "probed" investigation technique to obtain the report of side effects from subjects. All subjects were specifically' questioned concerning the presence or absence of each symptom. This technique tends to give a higher incidence of symptomatology than unprobed investigation. Data analysis involved comparing percentage of reported complaints of each drug group in each cycle. Percent incidence of each complaint before onset of medication was also noted. Invari- ably, the incidence of pre—treatment complaint was higher than that during the treatment cycles, for all cycles and all drug groups. l7 Goldzieher noted that this demonstrated the inadvisability of using subjects as their own controls, and suggested that the decline in complaints was due to "an interaction with the monthly questioning process and the possible appearance of a bias." He did not attri- bute the decline to a therapeutic effect of the drug. In comparing the incidence of complaint in the placebo group to the oral contraceptive groups, the results were as follows: Nausea: In the first cycle the rate of nausea was higher for Oracon and Ovulen than for the placebo group (about 22% and 15% versus 9%). For Norinyl and chlormadinone the percentage incidence was about the same or lower than the percent for the placebo. The higher incidence for Oracon was statistically significant (p < .05). This drug con- tains the most estrogen. Vomiting: A very similar picture to that of nausea was presented here, but with a lesser degree of incidence overall: a general decline over time in percent, percent for all drugs but Oracon close to those for the placebo. Oracon users had, in the first cycle, a significantly higher percentage of vomiting (15% versus 3% for the placebo). Breast tenderness: The pretreatment range for percent incidence of breast tenderness was 8.6% to 19%. There was a decline over time for all drugs and for the placebo. In the third cycle, the incidence for the various groups was approximately as follows: placebo 7%, Oracon 2%, Ovulen 5%, Norinyl 3%, chlormadinone 3%. The results do not bear out the notion that breast tenderness is related to estro- genic content. l8 Headache: Again, the pre—treatment incidence (range 20% to 30%) declined over cycles in every case. The average incidence of head- aches for the four cycles on Ovulen were significantly higher than that of the placebo group, but no change (increase) was observed in the subjects switched from chlormadinone to Ovulen in the fifth cycle (incidence went from 11% to 7%). The average level of incidence was from 6% to 12% for all groups, except for Ovulen. The percentages were similar to those for the placebo group (e.g., in the third cycle: placebo 7%, Oracon 8%, norinyl 15%, chlormadinone 9.5%). Nervousness: Except for the first cycle with Oracon and Ovulen the incidence of nervousness for the four oral contraceptive groups was similar to that of the placebo group. The Ovulen and Oracon groups had a significantly higher incidence than the placebo group but only in the first cycle. Switching from chlormadinone to Ovulen caused no increase in incidence, and switching from Ovulen to chlormadinone caused no drop, so the significance of the higher incidence of headache for the Ovulen and Oracon group is questionable. Depression: The incidence of complaint for the placebo group ranged between 2% and 9%. A very similar picture was presented by the other four drug groups. Weight gain: Using the patient's pretreatment weight as a base line, data were grouped by percentage of patients who gained 5 lbs. or more, percentage who lost 5 lbs. or more, and percentage who remained the same (within the :_S lbs. range). By the end of the fourth cycle the percentages of women who had gained 5 lbs. or more was 30% for the placebo group and 31, 20, 19 and 28% for the drug groups. Using l9 analysis of variance the data were analyzed for significant differ- ences in trends between the preparations over the four cycles of therapy. Also a x2 test was used to see if any of the differences between preparations were significant. A significance level of a = 0.10 was used to increase the chance of finding a significant result, but none were found. The 0.10 level of .50 on a factor, (2) less than .25 on the second factor, (3) had an item-subscale correlation of > .5. It was expected that this might produce a list too short for good internal consistency but less desirable items could be added once lists of best-items were drawn up. The list yielded 11 items for scale 1, 11 items for scale 2 and 6 items for scale 3. Two items were added to lengthen scale 3. A second test of the internal consistency was made, the results of which are shown in Table 4. Table 4.—-Second Internal Consistency (alpha) Test for Each of the Reduced Scales Constructed From Pre-test Data. Attitude Scale Toward No. of Items Alpha 1 Oral Contraceptive ll .74 2 Birth Control 11 .54 3 Role 8 .80 Scale 1 was lengthened by adding 4 (less desirable) items and scale 2 was lengthened by adding 3 items. This procedure was a step by step process in an attempt to create a scale which balanced ' 37 the demands of unidimensionality, single factoredness, and internal reliability. A third analysis for internal relaibility yielded the results shown in Table 5. Table 5.-—Third Internal Consistency (alpha) Test for Each of the Reduced Scales Constructed From Pre-test Data. Scale No. of Items Alpha (rounded) 1 15 .90 2 14 .82 3 8 .80 These three scales were sound, then, in terms of reliability. Because deletion of items from a scale often alters the factor struc- ture a factor analysis was done on these new scales to check on their structure. This analysis showed that the factor structure was not appreciably changed. Of the 37 items that composed the three scales all but three loaded most heavily on the appropriate factors. Loadings obtained for this analysis are listed in Appendix 3. The three attitude scales were sound in terms of each having high internal reliability and being composed of discrete factors. The final criteria for a good questionnaire is, of course, that it be valid. But as is the case with most attitude questionnaires testing for con- struct validity is almost impossible since there is no objective way to tell if subjectively reported feelings of a particular person are in fact his real feelings. 38 A brief test of predictive validity was made by comparing results of the attitude scales with the side effects portion of the pretest questionnaire (part 3 as described on p. 27). The side effects section included seven Likert scale type questions and one open—ended question. The Likert scale questions covered seven commonly reported minor symptoms. The scores from each of the three new attitude scales were correlated with the scores of each side effects item. Table 6 shows the results of that analysis. Some of the side effect questions were worded so that a high score meant a high expectation of that side effect and some were worded oppositely. In every case (but one) the sign of the correlations was in the direction expected. The best predictive relationship between scales and side effect items was obtained with scale 1 (attitudes toward oral contraceptives). Although for all three scales the correlations were modest, they indicate that there is a consistent predictive relationship between the attitude scales and the side effect questions. Table 6.—-Correlations Between Each Side Effect Item and Each Attitude Scale. Expected Direction r of Item X Scales l - 3 Item No. Topic Area of Corr. Item X 1 Item X 2 Item X 3 62 Sex + +.19 +.l4 +.11 63 Faintness - -.25 -.21 -.09 64 Nausea - -.21 -.13 -.03 65 Skin +p +.05 -.04 +.06 66 Depression - -.28 -.14 —.08 67 Weight - -.22 -.13 -.16 68 Bloating - -.30 -.14 -.10 39 In summary, then, the attitude questionnaire created by the process described above was composed of three orthogonal factors or scales: scale 1 concerned attitudes toward oral contraceptives and consisted of 15 items; scale 2 was concerned with birth control in general and was made up of 14 items; scale 3, on woman's role, was 8 items long. Each scale was internally consistent and individual items from a scale correlated well with total scores on that scale. Each scale's face validity appeared reasonable from a subjective viewpoint and finally, each had low* but consistent predictive validity. These three subscales then appeared to compose a good questionnaire for use in the actual study. Expected Side Effects Scale (Likert). This scale, as men- tioned earlier, was composed of seven Likert type items on possible side effects. Questions were worded in such a way that subjects could express positive changes which might occur as well as negative ones. Also questions involved expected side effects. A sample questionnaire is included in Appendix 1. The items covered seven commonly reported side effect areas: interest or enjoyment of sex, faintness, nausea or vomiting, skin (acne or the improvement of acne), depression or anxiety, weight, bloating or puffiness. *The low predictive validity is not unexpected since the second variable in the correlation was single item scores (on side effects) and the subject sample was very diverse in that it was comv posed of women who had never taken the pill, women who had previously taken it and women who were then taking it. '40 This questionnaire allowed the comparison of expected side effects with attitudes as reported above. It also aided in construt— ing a less awkward and more appropriate side effect checklist for use in the study proper. Personal Knowledge of Side Effects (open-ended question). An open-ended question about what side effects the subject had heard and how frequently the subject thought that side effect occurred was included in the pretest queStionnaire. This question was included simply to get a "feel" for common or lay knowledge about side effects. Since the question was open-ended it had to be coded for analysis. Coding was very simple and clear. It consisted of assigning a number to each type of side effect rather than classifying or making fine judgments. The experimenter did the coding. In all, 32 types of side effects were recorded by subjects. These were aggregated into the following categories: (1) major side effects (such as cardiovascular disorders or other disorders poten- tially threatening to the life of the subject); (2) minor side effects (such as nausea or weight gain); (3) "positive" side effects (such as improvement of acne); and (4) unclassifiable. Appendix 4 lists the side effects in each of these categories and the frequency with which they were reported by these subjects. It should be noted that these side effects are not reported side effects, nor are they expected side effects, but simply side effects which subjects may have "heard of," regardless of their personal assessment. 41 Demographic Information. This section was composed of five questions about age, marital status, previous or present use of the pill, and level in college. Results from Pretest The results for the attitude scale portion of the questionw naire were reported in the section entitled "The Attitude Scale." Basically they were that three factors emerged on first analysis. These were easily identified by reading the questions as attitude toward: (1) oral contraceptive use, (2) birth control in general, and (3) woman's role. A correlational matrix allowed choice of good items (good in terms of high item-whole scale correlation and independence of other scales). Tests of internal reliability gave a further check as items were chosen or discarded for the new reduced scales. Finally another factor analysis showed that the original three—factor structure held up. These new reduced scales were used in the study proper. A frequency distribution showed how often various symptoms were named ("heard of") by subjects on the open-ended question. In the category of major negative symptoms those related to cardio— vascular problems were most known to subjects with 48% of subjects naming a symptom of this type. Cancer was named by 17% of the sub- jects. Less than 5% of the time the following symptoms were named: hormone imbalance, sterility, vision changes (including blindness) and birth defects or harm to future offspring. 42 In the minor negative side effects category a majority of §s (50.1%) named weight gain as a symptom they had heard of. Other frequently named side effects were nausea 34.9% changes in menstruation 11.4 changes in breast 13.1 emotional changes 17.1 Very few positive side effects were mentioned. This is probably a reflection of the type of publicity oral contraceptives have received. Oral contraceptives in fact are sometimes used in a non-contraceptive medical capacity because they can improve certain conditions. They are prescribed for acne, to reduce dysmenorrhea and excessive menstrual flow. Despite these rather common positive side effects only a few subjects mentioned any. Frequency of those mentioned is as follows: improved skin 5.1% relieves cramps 1.7 regulates period 0.6 relieves worries 0.6 reduces headaches 0.6 In all 394 symptoms were mentioned by subjects. These fall into the 31 types listed in the Appendix. Only eight of the 394 were unclassifiable. The close-ended questions on side effects that subjects thought they might have if they were taking the pill were included 43 for two reasons: (1) to be able to compare these expectations with attitudes, (2) to get an indication of the type of response and the usefulness of this format. The comparison of expected side effects and attitudes was explained in "The Attitude Scales" section when discussing predictive validity. The results on close-ended side effect questions in and of themselves are only of limited interest. Generally, answers were normally distributed over the five possible choices with the most common answer being "remain the same." The two questions which were skewed concerned sex and weight gain. Thirty-nine percent of the women felt their enjoyment of sex would increase if they were to use birth control pills. Twenty—five percent said it would increase slightly. Sixty-seven percent of the subjects expected they would experience an increase or slight increase in weight if they took the pill. (More detailed results are reported in Appendix 5). Development of the Side Effects Checklist. The answers to open—ended questions on side effects aided in indicating what type of symptoms should be used in the study proper. The format of the Likert side effect items used in the pretest was modified to make response simpler for the subjects. The format used in the study listed 22 (minor) symptoms. Subjects could indicate that they did not have the condition, that they had it but were not bothered by it, slightly bothered, bothered, or very bothered by it. This method of checking symptoms was an 44 important feature of the study since it allowed the subject to indi- cate positive as well as negative reactions to the pill. Instructions were worded to emphasize that women should not necessarily expect to experience side effects and that some women experience improvement of certain conditons. The final testing instrument, including both the attitude scales and the side effects check list, is in Appendix 6. Subjects Participants in the study were 59 young women who presented themselves at Olin Health Center or Lansing Family Planning Clinic (LFPC) for the purpose of obtaining contraceptives. The women from Olin were college students at Michigan State; the women at the Family Planning Clinic were generally younger (mean age = 16.6 years) than college age. Most in both groups were unmarried. Only women who had never before used pills and who were just beginning prescribed oral contraceptives at this visit were used in the study. Data Collection Procedure At the Health Center the woman's physician would give her a booklet of materials at the end of the appointment period. Physi- cians told their patients: "A study is being done on how women feel about birth control pills. If you think you might like to partici- pate you can go to room ____and read this information on the study. If, after reading the information, you decide not to participate, Sinmly leave the materials there. If you decide to participate you Can fill out the questionnaire." 45 Subjects from the Lansing Family Planning Clinic were approached by the experimenter at the end of their visit to the clinic. They were told a study was being done and asked if they would like to take part by filling out a questionnaire. Both sets of subjects were told they would be asked to fill out a similar questionnaire later: "So we can see how you're doing." Data consisted of subjects' responses to a questionnaire which tapped attitudes toward oral contraceptives, birth control and woman's role and a symptom checklist. Subjects filled out these forms at the initial collection period (to) as explained in the design section. For the subjects from Olin the identical questionnaires were mailed out one month (t1) and again three months (t3) after their initial clinic visit. Data collection for subjects from the Family Planning Clinic could not be handled this way. Many subjects were minors living at home and obtaining contraceptives without their parents' consent. Mailing to these Ss could have jeopardized their privacy. The clinic, however, dispensed first contraceptive prescriptions on a three months basis. No one month (t1) collection was done for these subjects, but at three months (t3) followup was done by having staff give patients the questionnaire at the end of her clinic visit. For all subjects, then, there was data at t and t for some 0 3’ subjects (those from Olin) there was also data at t1. RESULTS Introduction The main source of data for analysis was from subjects with questionnaires completed at time 0 and time 3. This was considered the main data source for analysis because it contained a sufficient number of subjects (N I 59), and the time period between collections, three months, was long enough for subjects to make an evaluation of their adjustment. The optimum data would have been from time periods 0, l and 3, but problems in getting complete followups at Olin and difficulty in obtaining data at time 1 at L.F.P.C. caused the number of subjects with data at 0, l and 3 to be too small for meaningful analysis. Secondary analyses (i.e. analyses with a less than sufficient N) was also carried out for subjects with data at times 0, l, 3 to obtain an indication of trend. Main Analysis and Results A correlational matrix comprised of the three attitude Scales and the symptom checklist at time 0 and the same scales and check- list at time 3 was constructed. From this matrix, panel diagrams were set up,£nch containing two variables at time one and the same two variables at time two. Since there were four variables at each time, six panels were generated [ (g) = 6 ] covering all permutations of the 46 47 variables. Correlations were corrected for attenuation. The corre— lational matrix is in Appendix 7, and the panel diagrams are shown in Figure 2. It was expected that the strongest influence would be between scale 1 (attitudes toward birth control) and side effects because (1) scale 1 was the best scale psychometrically and (2) scale 1 was the most direct measure since it dealt directly with g£§1_contraceptives. Also it was hypothesized that the causal vectors would operate in the direction of attitudes affecting side effects rather than vice- versa. Analysis for difference between correlations was made using Pearson and Filan's revision of the standard t-test. This adaptation was used because correlated variables are dependent on each other. The conventional t-test assumes that the variable distributions are independent. Special consideration for dependence is made by sub- tracting out the covariance. The formula used was: r '1' ab cd crab + crab - 2Y dr x Or r ab cd rabrcd In the six panels the following three cross lagged relation- ships were significant (variables and causal vectors as follows): attitudes toward oral contraceptives affecting symptoms; attitudes toward oral contraceptives affecting attitudes toward woman's role in general; and attitudes toward birth control affecting attitudes towards woman's role. 48 Aoc Abc = attitude toward birth control Awr = attitude toward woman's role SE = side effects * .3819 Aoc1 Aoc2 .5840 SE2 * .3819 Aoc- Aoc f 2 .0106 .5488 .0848 .4419 Awr Awr 1 1.104 2 * Abc1 Abc2 ~5776 .4638 .5123 Awr1 Awr2 1.104 attitude toward oral contraceptives Abe .9561 . ITO .0469 Abcz -.1173 .3122 .0636 SE1 .7104 SE2 1.1040 Awrl .1046 Awrz .0028 .1364 .1403 SE1 .7104 SE2 .3818 Aoc1 .2252 Aoc2 .6605 .5203 .3142 Abc1 .9561 Abc2 *Designates Panels with Significantly Different Cross-lags. (All correlations corrected.) Figure 2.—-Panels of All Possible Pairs of Variables (Correlations Corrected for Attenuation). 49 Since a specific prediction was made concerning the relation- ship between attitudes toward oral contraceptives and symptoms a conventional p level of .05 was used to test for significance. Less specific or no predictions were made concerning the other five rela~ tionships so a more extreme p level was used to insure against accepting a spurious correlation as significant.* Table 7 summarizes the results. Table 7.--Significance Level of t Test for Difference Between Cross Lag Correlations. Cross Log Significance Variables Correlations t Level p = .05 *AO c x S.E. .3117 -.0155 .99 one tail A x S.E. .1403 .1046 .36 n.s w.r. Ab c x S.E. -.0636 .0477 .66 n.s Ao.c. x Ab.c. .3142 .2252 .14 p < .001 A x A .4419 .0106 .99 n.s o.c. w.r. Ab x A .5123 .3831 .39 p < .001 .c. w.r. *AO c = Attitude toward oral contraceptives 3" 3> II Side effects U) [‘11 II Attitude toward birth control Attitude toward woman's role *Such a possibility, of course, increases as a function of the number of tests run. 50 As mentioned in the introductory section on cross—lagged panel analysis, some finer considerations need to be made concerning this technique. Rozelle and Campbell (1969) pointed out that there are more than two rival hypotheses to choose from as a result of a cross lagged difference. Previously, the two hypotheses entertained were A causes B vs. B causes A. But two additional ones need to be considered, specifically the negative ones: A causes an inverse change in B vs. B causes an inverse change in A.* Given these two alternative hypotheses the inference which can be made from unsymmet- rical cross lags (for example, r > . r ) is that the joint effect A1'32 BlAZ of: A directly causes B and B inversely causes A, is greater than the joint effect of: B directly causes A and A inversely causes B. In many cases it may be difficult to imagine such a relation- ship (which is probably why the additional hypothesis were not under consideration earlier). Often the inverse alternatives are ruled out simply onthe basis of implausibility. For example, in analy- sizing the relationship between intelligence and achievement Crano *That is; increases in A cause decrease in B (and decreases in A cause increases in B) vs. increases in B cause decreases in A (and decreases in B cause increases in A). I use the term "inverse change" as a short-hand way to express the negative relationship. Since there is no convenient terminology for discussing cross lags, I would suggest the following: The hypothesis that the preponderance of causation is such that A causes B, or that B causes A, be called the "direct causal hypotheses" and be notated A+B and B+A. The hypothesis that the preponderance of causation is such that changes in A cause changes in the opposite direction in B or that changes in B cause changes in the opposite direction in A be called the "inverse causal hypotheses" and be notated A+§ and B+§-where x is some unknown constant in the negative relationship between A and B. 51 et a1. (1971) ruled out the possibilities that achievement inversely affected intelligence and that intelligence inversely affected achievement as extremely implausible. In some cases, however, the competing hypotheses may have validity. Rozelle and Campbell studied grades and attendance in a foreign language class. Their cross lags (r > r ) indicated A162 GIAZ that attendances had a direct influence on grades. The authors then realized that the possibility that grades might inversely influence attendance* was equally reasonable. Rozelle and Campbell bring up the issue of, but do not directly discuss, the effect which a zero synchronous correlation between variables produces on these rival hypotheses. If the initial syn- chronous correlation is zero and the relationship changes such that B inversely causes A (the hypotheses we would like to rule out) this must necessarily be reflected in a negative cross lag correlation between B1 and A2. Conversely, then, if the initial relationship between A and B is zero, there is a significant difference between cross lag and neither cross lag is negative then the possibility that A+§-does not exist. In other words, when r = 0, the B AlBl inverse causal hypotheses need be considered only when r In the present study, we can rule out the inverse rival hypotheses for the Ao.c. x S.E. relationship, since rA151 = _ 015 *Good grades could cause a drop in attendance and poor grades could cause an increase. 52 (essentially 0) and r = —.016 (essentially 0). Even if we should S1A2 consider the slightly negative correlation of r meaningful (and 31A2 not just error variance around a zero) it is still obvious that the preponderance of causation must be in the direction of attitudes affecting side effects since the relationship between attitudes at time 1 and side effects at time 2 moves from .015 to .312. In more general terms, then, the rival negative hypotheses are most reason- able when the difference of r > r is accompanied by correla- Ale BIAZ tions such that r < r and thus r < r is due to a AlB2 A131 BlAZ AlB2 decrease of rBlAl rather than an increase of rAlBZ. The special case of a zero correlation at time 1 does not apply in the analysis of the other two significant panels (Ao c x Aw.r. and Ab.c. x Aw.r.)’ but we can deal with the problem of rival inverse hypotheses by considering the more genral arguement just presented. In both panels there is a substantial initial synchronous correlation between the two variables (r = .5488 and r = .57758). In one case, this drops to near zero at time 3; in the other case there is only a slight drop. (Refer back to Figure 2). Since (in both panels) the larger cross lag correlation is slightly less than the initial synchronous correlation, and the smaller one is consider- ably less than the initial synchronous correlation, it appears that the difference in cross lags could be partially due to the fact that there is an inverse causal relationship causing one of the cross lag correlations to drop significantly over time. For these two panels then the results suggest the joint incongrous hypotheses. That is, for panel 2:. 53 the joint effect of A + A o.c. w.r. and is greater than X + _ Aw.r. A 00c and for panel 3; the joint effect of and is greater than the joint effect of A + A w.r. o.c. and X +— Ao c A w.r. the joint effect of In other words, the results show that positive attitudes toward oral contraceptives result in rejection of the traditional woman's role, but acceptance of traditional woman's role results in positive attitudes toward oral contraceptives once these women have used the pill for a few months. Similarly, positive attitudes toward birth control results in rejection of the traditional woman's role, but acceptance of traditional woman's role results in a positive attitude toward birth control once pills have been used. It will be obvious now that the usefulness in the cross- 1ag technique is determined to the extent to which we can rule out all alternative ways to explain r l 2 A B > rB A except the hypothesis that the preponderance of causation is such that A causes B. The correlation r > r A132 B1A2 obviously rules out the possibility that 54 B causes A; and we have just shown conditions under which we can rule out the inverse hypotheses. Are there any other possible explanations for an asymmetrical cross lag? Yes, there are three other cases which must be considered. It is possible that a change in factor structure of the tests or a change in the reliabilities (most likely both would occur to- gether) of the tests could cause unequal cross lags when there was no causal preponderance. Crano et al. (1971) gave a hypothetical example. Infant intelligence tests actually tend to measure motor skills more than mental ability. If children are measured at infancy with I.Q. and motor skills tests and then given the same tests later, the factor structure of the I.Q. test will change. The early I.Q. test and the later motor skills test are likely to be fairly highly correlated because both are measuring the same trait. But the early motor skills test and later I.Q. test can only be correlated to the extent that mental ability and motor skills are related. Thus r occur simply due to a change in one > r IQlM2 MlIQ2 of the tests. Kenny (1970) has suggested a method for correcting changes in specificity of tests by estimating communality ratios and using them to correct the correlations. For each variable pair the syn— chronous correlations at time 1 are divided by the synchronous correlations at time 2. The matrix of ratios which results will be single factored if there is no change in kind or only a change in amount in the tests. The communality estimates are then used to correct the cross lagged correlations for both reliability and specificity. 55 This correction procedure cannot be used here, however. The correction was developed for use on tests which measure variables which remain stable over time —- for example, intelligence tests. For such tests, as Crano et a1. pointed out, if the synchronous corre— lations are not about equal at both time points, then we conclude that the stability assumption has been violated. In other words, we make two assumptions: the relationship between variable tests to remain stable (synchronous correlations) and the tests themselves to remain stable (auto, or test-retest correlations). For the variables and tests used in this study, neither of these assumptions should be true. We assume just the opposite: (1) that the variables are changing over time and apparent low test retest correlations are Eg£_a reflection of unstable tests but of true changes in variables; (2) that the relationship between vari- ables changes and the different synchronous correlations are a result of an initial Eg_relationship between attitudes and side effects developing over time into a relationship. Therefore, the only reasonable way to correct correlations in this study is to correct for attenuation only and to inspect the tests' reliabilities to determine if they are stable enough to assume the difference in cross log correlations is not a reflection of spurious factor structure change. Correction for attenuation on all correlations was carried out using the formula: r v =.___§X____ X r r y Xny 56 The corrected correlations were used in all calculations. Internal reliabilities for all tests remained reasonably stable as shown in Table 8: Table 8.--Alphas For Each Scale at Time 1 (first unit) and Time 2 (three months after onset of pill use). Time 1 Time 2 *A .7874 ' .8412 o.c. Ab C .6558 .6980 A .6322 .7413 w.r. S.E. .7894 .7192 *Symbols the same as in Table 7. Since it is highly unlikely that the tests could be reliable at both time points and have a change in factor structure such that one vari— able was significantly more correlated with the second over time (than the second was with the first), we assume the cross lagged differences are a reflection of true differences due to causal vectors. The second case which we must consider as a possible alter— native explanation of the cross-lagged difference is that of a cross— lagged common factor. This alternative hypothesis cannot be entirely ruled out unless there are data at three time points instead of two. As pointed out earlier, attempts to get data at three time periods were unsuccessful to the extent that the N was small (28 of the 59 subjects). The correlation matrix of the four variables at the three time points is included in Appendix 8. 57 Unfortunately, these data, while they do nothing to contra— dict the earlier results, are not consistant enough to yield any new information. The last possible alternative explanation which needs to be ruled out is the following, which I will call the "partial hypothesis."* In any case where we are considering correlations between three variables the possibility that one of the three correlations is unduly inflated by the effect of the third variable should be con- sidered. Suppose, for example, we have found a correlation between grip strength and forearm size of .95. However, the data were collected on both men and women and there was a moderately high correlation between forearm size and sex, and between sex and grip strength. Obviously, the effect of sex of the subject is enhancing the grip strength/forearm size correlation. When the effect of sex is partialed out we would find that the r for grip strength/forearm size would be more modest than the original correlation of .95. So, in any cross lagged correlation the effect of any factor common to both of the variables must be partialed out. In the case 'where we are considering a cross lagged correlation between vari- ables A and D the effect of B and C must be partialed out. In the case B's (or C's) relationship to A or D is near zero then obviously there is no necessity to partial out the effect of B (or C). In obtaining the partials for the significant cross lagged correlations of this study the usual formula was used. *This previously unconsidered rival to the causal hypothesis was suggested by Dr. Lawrence Messe, Michigan State University. 58 Figure 3.——Mode1 Cross Lagged Panel. The Variables Common to r Must be Partialed Out. B and C are Both Common to the rad' ad 59 ny - xz . yz J (1 - szz)(1 - YYZZ) As stated on page 47 , three of the six panels yielded significant differences between cross lagged correlations. The process suggested here is to reexamine these significant cross lagged correlations, and, in any case of a cross lagged correlation greater than zero, to partial out the effect of any other variable in the panel which might be unduly inflating that correlation. Figure 4 shows the three panels with (originally) signifi- cant differences between the cross lagged correlations.* Working with the data of each panel cross lagged correlations (greater than .01) were "partialed." The variable boxed off in each panel is the one to be partialed out. The results of this process are shown in Table 9. Table 9 gives each unpartialed significant cross lagged correlation,* the common element which was partialled out, and the resulting partial correlation. Figure 5 shows the results presented in Table 9 drawn out in panel diagram form with the partial cross lagged correlations inserted. *Correlations not corrected for attenuation were used, since they are more conservative, and since in two cases the correction resulted in a correlation slightly larger than 1. If a number larger than 1 is used in the above partial correlation formula an imaginary number would result. 60 .wustw Am>Hum>uomcoo whoa ..m.Hv H030H m on HHH3 HmHuuma mau muomou u mam newcouum mH GOHummDU CH mOHanum> 03 mnu ou chm:0HumHmu muH moch mom: mm3 u3< .NH3 aoaaoo m omHm mH on< umnu 303m HHH3 q oustm mo Hound qunu mnu um moame <« mmoo. u Hoa< . ~6a<fina< H6n< msmm. u Nona . Haze «mud. n Hua< A ~83 muHse aoaaoo cOHumHouuoo mowwMH mmouo .mCOHumHmuuou chmHum> HmHuumm Boz mam .uso memHuumm mHanum> .mGOHumHmuuou wowme mmouu wouomuuouca HmanHHOII.m mHan 61 Box variables are those to be partialed out of the cross lagged correlation, as indicated by dotted lines. Awr1==' ————— , Awr2 Figure 4.--Panels of All Variables with Significant Cross Lagged Correlations (Values Not Corrected for Attenuation). 62 .3108 A Aoc 0c 1 -.0126 2 .0115 .4542 .1097 SE1 SE2 .5353 .3188 Aoc Aoc 1 +.0077 2 .3872 .0670 .0744 Awr Awr 1 .7558 2 .6469 Abc1 Abc2 .0038 .3719 .3336 _ .1252 Awr Awr 1 .7558 2 Figure 5.--Panels of All Pairs of Variables with (Originally) Significant Cross Lagged Correlations with Common Variable Partialed Out. 63 As can be seen in Figure 5 the comparisons to be made now are the cross lagged correlations: r = .1097; with r = -.0126 Aoc18E2 SElec2 r = .0744; with r = .0077 AoclAwr2 Awrlec2 r = .0038; with r = .1252 AbclAwr2 AwrlAbc2 None of the differences between pairs of cross lagged correlations is significant once the correlations have been reduced by this partialing process. Thus, this final analysis, while it does not yield results contradictory to the causal hypothesis, does not allow us to rule out the rival "partial hypothesis." The possibility remains that the results cannot be explained by the causal hypothesis. However, although the differences between the partialed cross lagged correla- tions are not significant they are consistently in the direction expected. That is to say, for example, that the correlation between A and SE of .1097 is larger than the correlation between SE and ocl 2 1 A of -.0126. A larger sample size might well show significantly oc2 different cross lagged correlations, even after partialing out the effect of other variables in the panel. In summary then the results suggest but do not conclusively demonstrate that there are causal factors such thatmattitude toward oral contraceptives influence later side effects, andmattitudes toward birth control and oral contraception influence attitude toward women's role. DISCUSSION After considering all rival explanations of the results concerning attitude toward oral contraceptives and side effects, the conclusion is that either (a) the preponderance of causation is such that attitude toward oral contraception effects the exper— ience of side effects or (b) there is a cross lagged common factor which causes attitude at time 1 and then at a later date causes side effects. Or, in other words, the alternative hypothesis may be that attitude is a prior symptom of the common cause. A third possibility is that the results are an artifact of a third common variable in the panel, artifically inflating the cross lagged correlation. The second alternative is a somewhat weaker notion of causality than that commonly used. However, it still provides more meaningful information than that obtainable from conventional correlations. Previously a correlation told gnly_the degree to which two variables were related. Now, even with the weaker prior symptom hypothesis, we know both the degree to which they are related, and can make a statement about the direction of causality. In this case for example, we know that the preponderance of the relationship between attitude and side effects is not due to side effects causing negative attitudes, but that attitudes or some 64 65 prior symptom of attitudes cause side effects. And further, it seems most reasonable to assume that the prior cause, if there is one, is of a psychological nature. It is difficult to imagine a physiological variable which could cause an attitude. The results suggest that, assuming the weaker prior symptom hypothesis, there is some psycho— logical variable which causes attitude toward the pill to be, for instance, negative, and at a later time causes the subject to report many side effects. For example, suppose there is some personality variable which we could characterize as tendency to complain or to find fault or to be dissatisfied. It causes subjects to feel they will do poorly on the pill (reflected in the attitude scale) and later, once they have begun taking pills, causes them to report many symptoms. Although this seems reasonable, it does not fit the data as well as the direct causal hypothesis. The prior symptom hypothesis cannot explain why this cross lagged common personality factor does not also cause the discontented subject to report more side effects at time 1 as well at at time 2. Mathematically then, although the prior symptom hypothesis is a possibility, I believe that in this case the direct causal hypothesis makes more sense, if simply because it is more par- simonious. However the value of even the prior symptom knowledge over the degree of relationship knowledge provided by conventional corre- lation is clear. Applications of this additional information will be discussed in the "Clinical Implications" section. 66 Two additional comments on this result are in order. The first comment is a reminder that we are always speaking of the pre— ponderance of causation and do not mean to rule out the possibility that the "caused" variable can later influence the causal variable in a feed back loop situation. Over time it may well be that side effects influence attitudes which again influence later side effects and so on. The second point is that even if we accept the causal hypo— thesis concerning the three relationships between attitudes and side effects and attitudes and attitudes, the relationships are not strong. So, although we might conclude that there may be an atti- tudinal component causal in side effects there are obviously other contributory factors. In the case of side effects other psychological and/or pharmacological variables are undoubtly operating. Similarly attitude toward woman's role is undoubtly influenced by factors in addition to the immediate attitudes toward birth control and oral contraceptives. Demonstrating a significant causal (or prior symptom rela- tionship) between attitudes and behavior, however, is exciting even if the relationship is not strong. Clinical Implications Since a patient's attitude affects her subsequent experience of side effects and since 21% of partients report that they stop using pills because of unpleasant minor side effects, the results of this study suggest that efforts should be made by the patient and 67 her physician to maintain a positive attitude toward oral contracep— tive use. This would, of course, be most important just as use is beginning and in the first few months. Positive attitudes will most likely be formed by educating the patient and by emphasizing the positive aspects of the pill. For example, patients might be told that "spotting" is fairly common, but that it does not usually persist after the first few months, or be told that menstrual flow usually becomes lighter in a way that emphasizes the advantage of this side effect rather than in a way that implies that it is an abnormal condition. Positive aspects of the pill are sometimes not mentioned at all and very few subjects even know of any positive effects, as the open ended pre-test ques- tion indicated. A patient should know that her attitude can affect her adjustment and that most women adjust to the pill without serious difficulty. Further, it should be emphasized that many of the dramatic side effects reported in the news media occured during the time when the pill was first introduced and dosages were much higher ‘19:?! F199;??? .9813 Besides the clinical approach of fostering positive attitudes, the complimentary approach of identifying women with negative attitudes should be considered.‘ Woman with very poor attitudes toward oral contraception are probably a bad risk in terms of future adjustment to the pill. If such women could be identified either by informal questioning or by a testing procedure it might be possible to suggest an alternative form of birth control to them. This could avoid 68 several unhappy months for the patient in which she tries to adjust, and tries new prescriptions but continues to experience discomfort. Ultimately, she switches to another form of contraception anyway, or, even worse, simply stops taking her pills and exposes herself to the risk of an unwanted pregnancy. Too, avoiding such experience would conserve the physician's time and energy for it is he, or she, who must deal with the unhappy patient.“ Before undertaking any procedure to identify woman for recom- mendations of alternative contraception forms more research is needed. The identification procedure must be accurate and valid in order to be justifiable. There are also medical implications which go beyond adjust- ment to oral contraceptives. It seems likely that similar psycho- logical factors are involved in adjustment/maladjustment to other forms of contraception. Goldzieher (1968) reported that many of the reported "side effects" of IUD's are similar to those of the pill. There is also the more general problem of the "placebo reactor." Many medications have true pharmacological side effect. Information about these side effects is sometimes reported in the news or sometimes passed by word of mouth from patient to patient. A certain percent of patients seem to be placebo reactors. That is, given that they know of an adverse side effect to a medication and that they believe they are taking that medication, they will experience the adverse side effect. Given the results of this study and the failure of past work to identify a placebo reactor personality type, it seems likely 69 that attitudinal variables would be a fruitful avenue of investiga- tion. If attitudes or expectations of some patients cause placebo reactions, adjustment to medication might be greatly improved by the proper handling of the patient prior to prescribing a new drug. Interrelations Between Attitudes The results on the relationship between attitude scales are more difficult to interpert since the study was not specifically designed to investigate these relationships. The joint hypotheses do not seem unreasonable. Attitudes toward oral contraceptives and toward birth control seem to have a direct influence on attitude toward woman's role. However, it is also true that attitude on woman's role has an inverse effect on attitudes toward oral contra- ceptives and birth control. Perhaps women who have not used birth control and who are "traditional" are not especially favorably dis- posed toward it at first. If they do begin using birth control, however, their "traditional" attitude results in acceptance of their lot -- i.e., being responsible for birth control, and being favorably disposed toward it. Non-traditional woman on the other hand may at first favor oral contraceptives and birth control in general, but once they begin use may rebel and become a little more negative in their attitudes. Methodological Implications This study served to point out more explicitly some special conditions under which the competing inverse hypotheses could be ruled out on a mathematical basis. Specifically if rAlBl = O and 70 neither cross lagged correlation is negative then the inverse hypo— theses are not reasonable. If > 0 and one or both of the r AlBl cross lagged correlations are less than r , than the inverse AlBl hypotheses should be further considered. The problems with the cross lagged technique also became apparent. The assumption of stationarity is important in this type of analysis. Stationarity can be determined by factor analysis or a correction procedure can be employed to adjust for slight devia- tions. When studying stable factors, these procedures are adequate. In cases where we are studying variables in which we expect a significant change over time which is not constant over subjects (i.e., we expect the auto correlations, and rBle, to be rA1A2 small), wo do not expect the variables to appear stationary. Intelligence usually does not change or, if an experimental manipu— lation is introduced, it changes all scores in roughly the same way. In the real world, attitudes often do not remain the same; and forces which cause their change are not consistant and do not effect all persons the same way. The cross lagged technique should help us isolate some of the forces. But a method of controlling or accounting for changes in stationarity needs to be developed. The use of three data waves would seem to be helpful here since consistancy over more than one panel of a specific vector would make the lack-of-stationarity—hypothesis very weak. In the case of the "partial" hypothesis the case for using three data waves becomes compelling. If, on further examination this hypothesis remains as a real rival to the causal hypothesis 71 then two data wave collection must be abandoned completely in favor of three (or more) waves. Previously three waves was beginning to be the preferred method—-now it may be that it should be the required method. Implications for Attitude/ Behavior Research The attitude/behavior controversy is obviously not going to be solved by any one study. The contributions here are in demonstrating an attitudinal effect on a somatic behavior and in demonstrating a method to obtain quasi-experimental control in circumstances in which there is usually no control. The drawbacks of studying attitudes in laboratory settings have often been lamented. Until recently, there has been no satisfacotry was to study attitudes as they naturally are formed and changed. A more thorough understanding of these phenomena as they occur "in the field" will help improve the quality and contribution of subsequent experimental work. REFERENCES 72 REFERENCES American Medical Association, prepared in cooperation with the American College of Obstetricians and Gynecologists, The Food and Drug Administration and the Pharmaceutical Manu- facturers Association. What you should know about the pill, revised 9/5/70, pamphlet no. 9801-728G, 970 OP-291 500M. Bakker, C.B., and C.R. Dightman. Psychological factors in fertility control. Fertility and Sterility, 15, 1964, 559. Bakker, C.B., and C.R. Dightman. Side effects of oral contraceptive, Obstetrics and Gynecology, 28, 1966, 373. Campbell, D.T. From description to experimentation: Interpretating trends as quasi-experiments. In C.W. Harris (Ed.), Problems in measuring change. Madison: University of Wisconsin Press, 1963. Campbell, D.T., and J.C. Stanley. Experimental and quasi-experimental designs for research on teaching. In N.L. Gage (Ed.), Hand- book of research on teaching, Chicago 5, Illinois: Rand McNally, 1963. Coch, L. and J.R.P. French. Overcoming resistance to change. In E.E. Maccoby, T.M. Newcomb and E.L. Hartley (Ed.), Readings in social psychology (3rd ed.) New York: Holt, Rinehart and Winston, 1958, pp. 233-250. Crano, W.D., D.A. Kenny, and D.T. Campbell. Does intelligence cause achievement? A cross-lagged panel analysis. Journal of Educational Psychology, 1972, 23, 258-275. Cullberg, J., M. Cello and C.O. Jonsson. Mental and sexual adjust- ment before and after six month's use of an oral contraceptive, Acta Psychiatrica Scandinavica, 45, 1969, 259. Cullberg, J. Mood changes and menstrual symptoms with different gestogen/estrogen combinations: A double blind comparison with a placebo. .Acta Psychiatrica Scandinavica, Suppl. 236, 1972, 1. 73 74 Daly, R.J., F.J. Kane and J.A. Ewing. Psychosis associated with the use of a sequential oral contraceptive. The Lancet, August 26, 1974, 444. Festinger, L. Behavioral support for opinion change. Public Opinion Quarterly, 1964, 25, 404-417. Festinger, L. and J. Carlsmith. Cognitive consequences of forced compliance. Journal of Abnormal and Social Psychology, 55, 1959, 203. Fleishmann, E., E. Harris, and H. Burtt. Leadership and supervision in industry: An evaluation of a supervisory training program. Columbus: Ohio State University, Bureau of Educational Research, 1955. Gluckman, L.K. Psychiatric aspects of failures with oral contracep- tives. New England Medical Journal, 70, 1969, 10. Goldzieher, J.W. The incidence of side-effects with oral or intra- uterine contraceptives. American Journal of Obstetrics and Gynecolggy, 102, 1968, 91. Goldzieher, J.W., et al. A placebo-controlled double-blind cross- over investigation of the side effects attributed to oral contraceptives. Fertility and Sterility, 22, 1971, 609. Grounds, I., B. Davies, and R. Mowbray. The contraceptive pill, side effects and personality: Report of a controlled double blind trial. British Journal of Psychiatry, 116, 1970, 169. Greenwald, A.G. Behavior change following a persuasive communication. Journal of Personaligy, 1965, 55, 370-391. Greenwald, A.G. Effects of prior commitment on behavior change after a persuasive communication. Public Opinion Quarterly, 22, 1966, 595. Herzberg, B. and A. Coppen. Changes in psychological symptoms in woman taking oral contraceptives. British Journal of Psychia- try, 116, 1970, 161. Honigfeld, G. Non-specific factor in treatment: 1. Review of placebo reactors; II. Review of social-psychological factors. Diseases of the Nervous System, 25, 1964, 145, 255. Janis, I.L. and S. Feshback. Effects of fear—arousing communications. Journal of abnormal and social psychology, 35, 1953, 78. Kay, C.R., A. Smith, and B. Richards. Lancet, 1969, 2, 1228. 75 Kenny, D.A. Cross—lagged and synchronous common factors in panel data. Unpublished, Northwestern University, May, 1971. Kothandapani, R. Validation of feeling, belief, and intention to act as three components of attitude and their contribution to prediction of contraceptive behavior. Journal of Personality and Social Psychology, 1971, 19, 321—333. Lewis, A., and Hoghugi, M. An evaluation of depression as a side— effect or oral contraceptives. British Journal of Psychiatry, 115, 1969, 697. Litz, R.W. Emotional factors in the success of contraception. Fertility and Sterility, 2Q, 1969, 761. Maccoby, N., A.K. Rommey, J.S. Adams, and E.E. Maccoby. Critical periods in seeking and acceptingyinformation. Paris- Standford Studies in Communication, Institute for Communi- cation Research, 1962. Mears, E. Clinical trials of oral contraceptives. British Medical Journal, 1961, November 4, 1179-1183. Nilsson, A., and Almgren, P.E. British Medical Journal, 1968, 2, 453. Orchard, W.H. Psychiatric aspects of oral contraceptives. Medical Journal of Australia, August 26, 1969, 872. Ostrom, T.M. The relationship between attitude components. Journal of Experimental Social Psychology, 1969, 5, 12-30. Pelz, D.C., and F.M. Andrews. Detecting causal priorities in panel data. American Sociological Review, 22, 1964, 836. Pincus, G., J. Rock, and C.R. Garcia. Field trials with Norethynodrel as an oral contraceptive. Proceedings of the Sixth Inter- national Conference on Planned Parenthood, New Delhi, 1959, International Planned Parenthood Association, London, 216. Richter, R.H.H. In Proceedings 5th Conference of Europe and Near East and Africa Region,ylnternational Planned Parenthood Federation, Copenhagen, July 5-8, 1969, Stephen Austin and Sons, Ltd., Hertford, England, 1966, p. 121. Rogers, D.A., and F.J. Ziegler. Changes in sexual behavior consequent to use of non-coital procedures of contraception, Psychomatic Medicine, 1968, 59, 495-505. 76 Ross, M., C.A. Insko, and H.S. Ross. Self—attribution of attitude. Journal of Personality and Social Psychology, 22, 1971, 292. Sandell, R.G. Note of choosing between competing interpretations of cross—lagged panel correlations. pggychological Bulletin, _Z5, 1971, 367. Silbergeld, S., N. Brast, and E.P. Noble. The menstrual cycle: a double-blind study of symptoms, mood and behavior, and bio— chemical variables using Enovid and placebo. Psychosomatic Medicine, 55, 1971, 411. Simon, W., and J.H. Gagon. College youth study. Unpublished research, Institute for Sex Research, Indiana University, 1968. Westoff, C.F., and N.B. Ryder. Experience with oral contraception in The United States; 1960-1965. In A.F. Goldfard (Ed.), Clinical Obstetrics and Gynecology: Oral Contraception, . 22, 1968, 734. Zell, J.R., and W.E. Crisp. A psychiatric evaluation of the use of oral contraceptives. Obstetrics and Gynecology, 25, 1964, 657. Ziegler, F.J., D.A. Rodgers, S.A. Kriegsman, and P.L. Martin. Ovula- tion suppressors, psychological functioning and marital adjustment. Journal of American Medical Society, 204, 1968, 97. APPENDICES 77 APPENDIX 1 ORIGINAL PRETEST The following is an attitude questionnaire about birth control. Your participation in filling it out is voluntary. Your responses will be anonymous. It is 292 necessary for you to identify yourself in anyway (name, student number, etc.). 78 79 INSTRUCTIONS: You should have both a questionnaire and an answer sheet. Both should be marked with the same number (if they are not ask for a new set.) On your answer sheet in the box where it says "student number" write instead the number which is on your answer sheet then fill in the corresponding numbers in the boxes below. The first questionnaire requires you to write down answers. Do this onthe questionnaire and ppp_on the answer sheet. All the other items 1 through 73 can be answered on the answer sheet. Most of the items are statements followed by a scale from 1 to 5 to indicate how strongly you agree or disagree with that statement. For each statement find the corresponding item on the answer sheet and mark the number which indicates your personal reaction to the statement. Please check carefully as you go along to make sure you are marking the appropriate corresponding item on the answer sheet. 80 ATTITUDE QUESTIONNAIRE January 4, 1972 You have probably heard about some side effects of oral contracep— tives (birth control pills). If you know of any please list them. Write down any side effect you have heard of - even if you think it may not be a real one. If you do not know of any side effect write "none" and go on to the next page. SIDE EFFECT FREQUENCY OF OCCURRENCE For those side effects which you have just listed, please write beside each the number which corresponds to how frequent you think that side effect is among women who take the pill. l = Very common 4 = Uncommon 2 = Common 5 = Very uncommon or never 3 = Uncertain 81 1. Having a career can be just as fulfilling for a woman as being a mother. 1. Strongly agree 2. Agree 3. Not sure 4. Disagree 5. Strongly disagree 2. Over-population is one of the serious problems in the world today. 1. Strongly agree 2. Agree 3. Not sure 4. Disagree 5. Strongly disagree 3. I would be very upset if I found out I could not have children. 1. Strongly agree 2. Agree 3. Not sure 4. Disagree 5. Strongly disagree 4. The government should make birth control information and methods available to everyone. 1. Strongly agree 2. Agree 3. Not sure 4. Disagree 5. Strongly disagree 5. Knowing other girls who use the pill and talking to them has undoubtly influenced my decision to use the pill. l. Strongly agree 2. Agree 3. Not sure 4. Disagree 5. Strongly disagree 6. If my doctor was opposed to birth control pills, I would go along with him and use something else. 1. Strongly agree 2. Agree 3. Not sure 4. Disagree 5. Strongly disagree 7. I would not bother to use a contraceptive unless I was having intercourse on a regular basis. 1. Strongly agree 2. Agree 3. Not sure 4. Disagree 5. Strongly disagree 8. With all the bother involved in using birth control it hardly seems worth while. 1. Strongly agree 2. Agree 3. Not sure 4. Disagree 5. Strongly disagree 9. Having me take birth control pills was more my boyfriend's idea than mine. 1. Strongly agree 2. Agree 3. Not sure 4. Disagree 5. Strongly disagree 10. I don't think anyone should have more than two children. 1. Strongly agree 2. Agree 3. Not sure 4. Disagree 5. Strongly disagree 11. I would use birth control pills regardless of who disapproved of my doing so. 12. 1. 13. 82 Strongly agree 2. Agree 3. Not sure 4. Disagree 5. Strongly disagree No other person has influenced me much in my decision to use the pill - it was very much my own idea. Strongly agree 2. Agree 3. Not sure 4. Disagree 5. Strongly disagree Girls under 18 should not be allowed to take birth control pills because the hormones in them might have a bad effect on someonewhfs not yet physically mature. Strongly agree 2. Agree 3. Not sure 4. Disagree 5. Strongly disagree Most contraceptives are actually much less "safe" than we are led to believe. Strongly agree 2. Agree 3. Not sure 4. Disagree 5. Strongly disagree If my doctor would not prescribe the pill for me then I would find one who would prescribe it. Strongly agree 2. Agree 3. Not sure 4. Disagree 5. Strongly disagree In a way, using a contraceptive which is 100% effective takes some of the excite- ment out of life. Strongly agree 2. Agree 3. Not sure 4. Disagree 5. Strongly disagree I would not encourage a friend to use birth control. Stongly agree 2. Agree 3. Not sure 4. Disagree. 5. Strongly disagree If I had a minor reaction to the pill I would keep taking them until my system got a chance to adjust. Strongly agree 2. Agree 3. Not sure 4. Disagree 5. Strongly disagree Whether a girl gets pregnant or not is as much a matter of chance as anything. Strongly agree 2. Agree 3. Not sure 4. Disagree 5. Strongly disagree Many doctors try to push birth control on to their patients. Strongly agree 2. Agree 3. Not sure 4. Disagree 5. Strongly disagree If anyone gets a reaction to the pill, it will probably be me. Strongly agree 2. Agree 3. Not sure 4. Disagree 5. Strongly disagree 25. 1. 26. 29. 1. 30. 83 The thought of side effects to the pill doesn't bother me too much. Strongly agree 2. Agree 3. Not sure 4. Disagree 5. Strongly disagree I feel that my doctor pushed the pill on me without giving me a chance to ask about other methods. Strongly agree 2. Agree 3. Not sure 4. Disagree 5. Strongly disagree I am glad I live at a time when advanced birth control devices are available. Strongly agree 2. Agree 3. Not sure 4. Disagree 5. Strongly disagree People who have large families, even though birth control is available to them, are irresponsible. Strongly agree 2. Agree 3. Not sure 4. Disagree 5. Strongly disagree I would not use birth control if the state made it illegal. Strongly agree 2. Agree 3. Not sure 4. Disagree 5. Strongly disagree I really like the pill because unlike some other contraceptives, it does not in- terfer with intercourse in anyway. Strongly agree 2. Agree 3. Not sure 4. Disagree 5. Strongly disagree I don't see much point in getting married if you don't have children. Strongly agree 2. Agree 3. Not sure 4. Disagree 5. Strongly disagree With all the new types of oral contraceptives available almost everyone can find one that is "right" for them. Strongly agree 2. Agree 3. Not sure 4. Disagree 5. Strongly disagree If I were taking the pill I would be willing to put up with some minor side effect. Strongly agree 2. Agree 3. Not sure 4. Disagree 5. Strongly disagree I would be extremely upset if I got pregnant out of wedlock. Strongly agree 2. Agree 3. Not sure 4. Disagree 5. Strongly disagree I think the whole "side effect" thing with the pill has been overdone. Strongly agree 2. Agree 3. Not sure 4. Disagree 5. Strongly disagree 43. 84 Occasionally I might have intercourse without using any form of birth control. Strongly agree 2. Agree 3. Not sure 4. Disagree 5. Strongly disagree Too much importance has been attached to the problem of overpopulation. Strongly agree 2. Agree 3. Not sure 4. Disagree 5. Strongly disagree I think I would probably adjust to the pill alright. Strongly agree 2. Agree 3. Not sure 4. Disagree 5. Strongly disagree If you have a lot of children you are not able to give each child the attention he needs. Strongly agree 2. Agree 3. Not sure 4. Disagree 5. Strongly disagree I'm afraid that taking the pill might some how effect my chances of having a baby later on. Strongly agree 2. Agree 3. Not sure 4. Disagree 5. Strongly disagree I have some reservations about taking the pill. Strongly agree 2. Agree 3. Not sure 4. Disagree 5. Strongly disagree If my future husband wanted a large family, I would have doubts about whether or not to go through with the marriage. Strongly agree 2. Agree 3. Not sure 4. Disagree 5. Strongly disagree I think that waiting until I am in mylate twenties or early thirties to have children is a good idea. Strongly agree 2. Agree 3. Not sure 4. Disagree 5. Strongly disagree Women should realize that there can be more to life than getting married and having children. Strongly agree 2. Agree 3. Not sure 4. Disagree 5. Strongly disagree I feel that the pill is one of the most natural types of contraceptives a woman can use. Strongly agree 2. Agree 3. Not sure 4. Disagree 5. Strongly disagree Personally, I'm really enthusiatic about the pill. 51. 52. 85 Strongly agree 2; Agree 3. Not sure 4. Disagree 5. Strongly disagree If I had a reaction to birth control pills, I would ask my doctor to try another type of pill, rather than take me off them. Strongly agree 2. Agree 3. Not sure 4. Disagree 5. Strongly disagree Oral contraceptives are no more harmful than many other types of medication, such as aSpirin, which people use all the time. Strongly agree 2. Agree 3. Not sure 4. Disagree 5. Strongly disagree If oral contraceptives were not easily available around here I would switch to something else rather than go out of my way to get them. Strongly agree 2. Agree 3. Not sure 4. Disagree 5. Strongly disagree I wish there was another way to keep from getting pregnant besides having to use birth control. Strongly agree 2. Agree 3. Not sure 4. Disagree 5. Strongly disagree I would never have intercourse without using some form of birth control. Strongly agree 2. Agree 3. Not sure 4. Disagree 5. Strongly disagree It is a shame that in this day and age there are still people who are uninformed about birth control. Strongly agree 2. Agree 3. Not sure 4. Disagree 5. Strongly disagree I think certain negative information about birth control pills has been deliberate- ly kept from the public. Strongly agree 2. Agree 3. Not sure 4. Disagree 5. Strongly disagree If I had a minor reaction to the pill I would stop taking it. Strongly agree 2. Agree 3. Not sure 4. Disagree 5. Strongly disagree Even if my birth control supplies cost a lot of money, I would somehow find a way to buy them. Strongly agree 2. Agree 3. Not sure 4. Disagree 5. Strongly disagree 53. l. 56. l. 55. 1. 56. l. 57. 86 I dislike the idea of using a contraceptive like the pill which affects my hormonal system. Strongly agree 2. Agree 3. Not sure 4. Disagree 5. Strongly disagree Oral contraceptives are no more effective in preventing pregnancy than some of the other types of contraceptives which are available. Strongly agree 2. Agree 3. Not sure 4. Disagree 5. Strongly disagree One of the most important things in life is to have a family and raise children. Strongly agree 2. Agree 3. Not sure 4. Disagree 5. Strongly disagree. There are a lot of people who simply cannot adjust to the pill. Strongly agree 2. Agree 3. Not sure 4. Disagree 5. Strongly disagree It seems like a nuisance to have to take a pill everyday if you are using oral contraceptives. Strongly agree 2. Agree 3. Not sure 4. Disagree 5. Strongly disagree If I had trouble with one type of contraceptive I would keep searching until I found one that was right for me. Strongly agree 2. Agree 3. Not sure 4. Disagree 5. Strongly disagree It is too bad that the traditional role of woman as a wife and mother is losing it's meaning for some people. Strongly agree 2. Agree 3. Not sure 4. Disagree 5. Strongly disagree If my boyfriend didn't uSe a contraceptive, then I certainly would. Strongly agree 2. Agree 3. Not sure 4. Disagree 5. Strongly disagree I find discussing birth control embarrassing. Strongly agree 2. Agree 3. Not sure 4. Disagree 5. Strongly disagree 87 IF YOU WERE GOING TO TAKE BIRTH CONTROL PILLS WHAT EFFECT DO YOU THINK THEY WOULD HAVE ON YOU FOR THE FOLLOWING AREAS? 62. 63. 64. 65. 66. 67. 68. My enjoyment of, or interest in, sex would: 1. increase 2. increase slightly 3. remain the same 4. decrease slightly 5. decrease Feelings of faintness would be: 1. more frequent 2. slightly more 3. no more or less frequent 4. slightly less frequent 5. less frequent Feelings of nausea or vomiting would be: 1. more frequent 2. slightly more 3. no more or less frequent 4. slightly less frequent 5. less frequent My skin would: 1. improve 2. improve slightly 3. not change 4. get slightly worse 5. get worse Feelings of depression or anxiety would be: 1. more frequent 2. slightly more frequent 3. no more or less frequent 4. slightly less frequent 5. less frequent My weight would: 1. increase 2. increase slightly 3. not change 4. decrease slightly 5. decrease Feelings of being bloated or "puffy" would: 1. increase 2. increase slightly 3. be no more or less than presently 4. decrease slightly 5. decrease 69. 70. 71. 72. 73. 88 How old are you? 1. Eighteen or under 2. Nineteen 3. Twenty 4. Twenty-one 5. Over twenty-one what is your marital status? 1. Single 2. Married 3. Seperated 4. Divorced 5. Widowed Have you ever taken oral contraceptives (birth control pills)? I. Yes 2. No Are you presently taking oral contraceptives? 1. Yes 2. No What is your level in school? 1. Freshman 2. Sophomore 3. Junior or Senior 4. Graduate 5. Other APPENDIX 2 FACTOR LOADING AND ITEM-~WHOLE CORRELATION FOR EACH ITEM OF PRETEST 89 90 Appendix 2 Factor Loading and Item--Whole Correlation for Each Item of Pretest Next Highest Highest Factor Next Correlation Correlation with Highest Highest with with Item # High Load Loading Loading Scale Scale 1 3 -.5164 .2911 .5715 -.2639 2 2 -.4404 -.3754 .4871 -.4536 3 3 .6686 .1750 .6330 .2421 4 2 .5391 -.2533 -.5361 -.3077 5 2 .3321 -.2094 -.3801 -.2068 6 1 .5438 .1739 .5650 .1975 7 2 -.2779 -.0913 .3616 .0525 3 2 -.6127 .1686 .6403 .3328 9 4 .5177 .0840 .1382 .0580 10 3,4 -.5244 .3639 -.6429(3) .4252(2; 11 l -.4399 -.2719 —.5458 -.3391 12 1 -.2193 -.0408 -.2841 -.0598 13 2 -.3789 +.1588 +.4491 .2475 14 1,4 .2668 .2023 .3753 .2854 15 1 -.4007 —.2575 -.4832 -.2598 16 2 -.5461 +.1970 +.6163 .2660 17 2 -.4868 -.2040 -.S701 .3198 13 1 -.4646 -.3157 —.5489 -.4062(2) 19 4,2 .3644 —.3409 .4602(2) .2070 20 4 .4142 .1893 .2684 .1479 21 4 .3385 .2826 .3424 .2175 22 1 -.7S39 -.l311 -.7668 -.3045 4. ~— —-—-.1; 91 Next Highest Highest Factor Next Correlation Correlation with Highest Highest with with Item # High Load Loading Loading Scale Scale 23 4 .4748 .2134 .3141 .2374 24 2 .6548 -.1597 -.6036 -.3331 25 4,3 -.4636 -.4430 -.S474(3) —.3000 26 2 -.4691 .2004 .5616 .3193 27 l -.3903 .2571 -.4357 -.2422 28 3,2 -.2794 -.2618 .3484(2) .36Ol(3) 29 1 .3446 .2882 -.4307 -.1340 30 1 -.6892 .1928 -.7072 -.3484 31 3 .1472 .1231 .0742 .0450 32 l -.6391 -.0632 -.5666 -.1134 33 2 .3566 .2154 .4654 .1048 34 3,2 .5399 -.4097 .6221 .5243 35 l -.5445 -.2285 -.5664 -.1964 36 4 .3164 .2548 -.3141 -.2658 37 1 .4170 .3257 .4878 .4259 33 l .7654 .1335 .7939 .3231 39 3 -.6022 .0991 -.6343 -.1941 40 3 -.5748 .0700 ~.5860 -.1772 41 3 -.6013 .2586 +.6171 -.3095 42 1 -.4368 -.2292 -.5175 -.1886 43 1 -.7698 .2617 -.8128 -.4254 44 1 -.7527 -.0374 -.6847 -.1833 45 1 -.6132 -.1970 -.5374 -.1351 92 Next Highest Highest Factor Next Correlation Correlation with Highest Highest with with Item # High Load Loading Loading Scale Scale 46 1 .4820 -.1191 +.5082 .2322 47 1 .3422 -.0881 .3292 .0334 43 2 .3876 .0870 -.4300 -.1554 49 2 -.6462 -.1593 -.5964 -.3406 50 l .3729 -.2021 .4215 .3109 51 1 .7063 -.1426 .6828 .2605 52 2 .4426 -.3015 -.4975 -.3260 53 l .6447 -.2120 .7047 .3555 54 1 .4295 -.3026 .4750 .3571 55 3 .6798 .2270 .6558 .2098 56 1 .3598 .2055 .3222 .1299 57 1 .3690 -.1121 .4222 .1826 53 2 .6593 -.1813 -.6279 -.3315 59 3 .6680 -.1979 .7194 .3891 60 2 .6050 -.1425 -.6008 -.3132 61 2 4.4840 .1766 .4469 .2103 *In cases where an item loaded highest on one scale but had highest correla— tion with a different scale,this is indicated by noting the different scale in parenthesis. APPENDIX 3 FACTOR LOADINGS ON REDUCED PRETEST 93 Factor Loadings on Reduced Pretest 94 Appendix 3 Old New Highest (pretest) (reduced test) Factor Factor Factor Loading Item Number Item Number 1 2 3 Factor 1 1 -0.1600 -0.1336 0.5645 III 3 2 0.1819 0.0958 -0.4554 III 4 3 -0.1710 —0.5344 0.0715 II 6 4 0.5646 0.0929 0.0653 I 8 5 0.0715 0.6818 -O.1205 II 11 6 -0.4623 -0.3071 0.2219 I 13 7 0.1299 0.4255 -0.0551 II 16 8 0.1192 0.6563 -0.0696 II 17 9 0.1909 0.5236 -0.1188 II 19 10 0.1191 0.4068 -0.0513 II 22 11 -0.7604 -0.1151 0.0641 I 24 12 —0.0811 -0.5722 0.2882 II 25 13 -O.1042 -0.2398 0.3802 III 26 14 0.1130 0.5818 -0.1541 11 30 15 -0.6842 -0.2009 0.1107 I 32 16 -0.6014 0.0037 -0.0656 I 33 17 -0.0638 0.3462 -0.0528 II 35 18 -0.5971 —0.0230 0.0361 I 38 19 0.7524 0.1752 —0.0327 I 39 20 -0.0871 -0.0559 0.5952 III 40 21 0.0245 -0.0951 0.6462 III 41 22 0.0575 -0.2882 0.6261 III 42 23 -O.4972 -0.0629 0.1012 I 43 24 -0.8077 -0.2310 0.0172 I 44 25 -0.7549 -0.0058 0.0005 46 26 0.4858 0.0188 -0.2964 I 48 27 0.0839 -0.3045 0.1968 II 49 28 0.3836 0.1568 —0.1266 II 9S Old New Highest (pretest) (reduced test) Factor Factor Factor Loading Item Number Item Number 1 2 3 Factor 51 29 0.6890 0.1096 -0.0771 I 52 30 -0.2650 -0.4507 0.0058 II 53 31 0.6939 0.2575 0.0394 I 54 32 0.4426 0.1746 —0.0369 I 55 33 0.3675 0.0750 -0.0694 III 58 34 -0.1937 -0.5752 0.1399 II 59 35 0.1936 0.2975 —0.5956 III 60 36 -0.1159 -0.5506 0.2125 II 62 37 0.1339 0.4736 -0.0365 II APPENDIX 4 FREQUENCY OF NAMED SIDE EFFECTS 96 1 l ‘. r _.-g‘ .. ‘r' 97 Appendix 4 Frequency of Named Side Effects Percent Mentioning Symptom this Symptom as Cardiovascular Disorders 48 % E Cancer 16.6 g Hormone Imbalance 4.6 a: Sterility 3.4 2 Vision Changes 4.0 g Harmful to Future Offspring 2.3 Change in Vaginal Flora or Fauna 1.7 'Change in Hair (increase facial or thinning head) 4.0 Acne 4.0 Increased Skins Pigmentation 6.3 Menstral Change 11.4 Nausea 34.9 Bloating 4.6 E” Headache 6.9 g Dizziness 6.3 E Cramps 4.0 g Weight Gain/Loss 50.1 E Excess Vaginal Discharge 1.7 Change in Breast 13.1 Change in Skin (e.g., hives, rash) 2.3 Change in Libido 4.6 Change in Appetite 1.7 Emotional Change 17.1 Tiredness 1.7 98 Symptom Percent Mentioning this Symptom POSITIVE Improved Skin Helps Relieve Cramps Regulates Period Relieves Worries Reduce Risk of Cancer Reduce Frequency of Headache Change in Fertility (increase) Unclassifiable 5.1 1.7 0.6 0.6 0.6 0.6 1.7 4.6 Death I ,. i -_—.__ _ APPENDIX 5 RESPONSES TO SIDE-EFFECT ITEMS ON PRETEST 99 19“» 8 100 1,! Nm.o NH.H No.4 Nm.5m Nm.wm NN.5H NH N N 00 No mN Human NsHmH No waHHmmH NN.n NN.H NH.mN N5.mm No.NN OH N 55 so me uanmz No.m No.4 NN.m Nm.os NQ.NN No.8 mH N CH Hm OH NH aOHmmmuame Ho mNaHHmmm Nm.o NN.H NH.¢ NH.¢5 Nm.qH Nm.NH NH m 0H ow 0N Nm Namaw>ousaH :Hsm No.m NN.N No.o Nm.mm Nm.NN No.N 5H 5 H NOH o4 4H 666:6: Ho mwaHHmmm N3.N Na.N No.0 NN.NN NN.oH NH.H NH m H emH NH N mamaunHmH Ho mwaHHmmH N5.m NH.H NH.H NN.mN NH.mN N5.mm o N N Nm «4 so ~66 No Hawasomam > m N N. m. n m. n n m n n S 3 8 I. a 9 S B B u B B I. S S S S O 3 a 1. a a u u. S S 8 S r\ T. T. T. Co I. co 8 no u. m u. .4 a .4 T. TL A A ummumum co mEmuH uumwwmlmvflm cu noncommmm m xwwcmaa< APPENDIX 6 SIDE EFFECTS AND ATTITUDE QUESTIONNAIRES 101 APPENDIX 7 MAIN ANALYSIS: CORRELATION MATRIX 104 1115 .mwmwommn Gazaou aw cm>Hm mcoauawuomwv manmwum>e oooo.H ommo. NHNN. Nome. mmmm. memo. nmqo.| Hemm. w magmaum> oooo.H ommm. choc. memo. mans. Nmmm. 05mm. 5 wannaum> oooo.H swam. ammo. menu. moqo. mmmm. o odnmfium> oooo.H oNHo.I snoo. mnoa. moan. n oanmwum> oooo.~ Nooo.| «cwo.l mHHo. e oun~wu~> oooo.H mama. mumm. m wanmwum> oooo.H meme. N manwwum> oooo.~ H mapmwum>s Amuoaaev Amuaaahv AnnmaHsv Annmafiev AHNMEHHV AHuwawhv Aaumafiev Aaumawav w manmuum> n manmwum> o anmHum> n manmwum> a manmwum> m maanum> N wanmwum> H wanmaum> muuwwwm «mam mHom m.cw803 Houuaoo suNHm COHunoomuucoo kuo xHuumz :OHHMHouuoo n Kamcmaa< muomwwm wwwm umwmxamc< can: «Hom n.sw803 Hopscoo nuNHm cowumoomuuaou Homo APPENDIX 8 MATRIX OF CORRELATIONS FOR 3 WAVE DATA 106 .1‘--———.—.____1 107 NH oooo.H .ua> HH mHHH. oooo.H .uo> T. 7:; tho.u monn. oooo.H oH .um> ommn. nHmH. moon. oooo.H m .Na> H 2.3 Na 38:6 2.8 m 08H» us oHou n.smeo: vumzou okuHuu< m as: us Houucoo 5an v.83» 02532 m 05H» um mo>Humouauucou Hmuo canto» ovsuHuu< N 03H» us nuuowmo ova N 05H» um oHoH n.saso: vhasou ovzuHuu< N 03H» us Houucoo :uqu neuron ovSHHuu< N oEHp us mo>Huaooaupcou Homo canto» ovzuHuu< H oaHu us uoowwo ova H 06H» us oHou o.cmaoz unwrap ovnuHuu< H osHu um Hoaucou :HNHQ vuaxou ovaumuu< H oEHu an mo>HumooaHucou Hauo vuazou ovstuu< II II II II II II I II II I HNMVIfl‘OI‘QOfiS HNoo. mNmN. vnmo.n ONOH. vam. mo~o.: NNNN.I ooHH. wNmo.u owHw. HNoH.u hvnH. memo. van. NHom. owMN. mHmo.u mmoo. eoov. mmon. NNvo.n NHHN. mmon. mmHH. oONN. won. owNo. nNNm. ano. avoo.n NnHo.u MONN. oooo.H mHoo.n mnHN.n smoo.n mmon. vmvm.u mHmn.n NNNH.I oooo.H mHmo.n momH. nvnH. moms. vwvH. momH. coco.H HHom. QVON.I nmNN. ONHm. mmvn. oooo.H mvHH.n mHmm. Nmov. HNom. oooo.H mmHH.u meN.c mme.c oooo.H mmvm. Noom. oooo.H anm. oooo.H m .Hd> h .hm> 0 .Hm> m .hw> v .hw> m .Hw> N .HMH) H .hw> mama o>mz n How mcoHumHOHnou mo xHHumz m wamcwaa¢ .ha> .pm> .ua> .hm> .um> .ua> .u¢> .ua> .ua> .um> .Ha> .Hm> NH HH 0H m a n w v m N H .ua> .ua> .um> .ua> .um> .Nm> .Ha> .um> .um> .Nm> .Nm> .um>