1 EXCESSIVE OPTIONS AND EXCUSES By RYAN M. JOHNSON A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2008
2 2008 Ryan M. Johnson
3 ACKNOWLEDGMENTS I thank Barry Schlenker for his unending support and contagious enthusiasm for research. I thank James Shepperd, Lise Abrams, James Algi na, and Richard Johnson for their contributions to the theory, measures, computer programmi ng, statistical technique s, and physiological equipment that made this research possible. I thank my parents and step-parents for their positivity and encouragement. Finally, I tha nk Kate Sweeny for always supporting me even during the times when I might not deserve it. I am very lucky.
4 TABLE OF CONTENTS page ACKNOWLEDGMENTS ............................................................................................................... 3 LIST OF TABLES ...........................................................................................................................6 LIST OF FIGURES .........................................................................................................................7 ABSTRACT ...................................................................................................................... ...............8 CHAP TER 1 INTRODUCTION .................................................................................................................. 10 Sociological Background ........................................................................................................11 Decision Making and the Lure of Choice ...............................................................................15 Option Number and Information Overload ............................................................................ 20 Information ................................................................................................................... ...20 Option Number and Excuse Making ......................................................................................26 Regret and Responsibility ................................................................................................ 28 Persistence of Excuse Making Over Time ......................................................................31 Goals and Hypotheses .............................................................................................................33 Study 1 .............................................................................................................................35 Study 2 .............................................................................................................................40 2 STUDY 1 ................................................................................................................................44 Method ........................................................................................................................ ............44 Participants .................................................................................................................. ....44 Instruments ................................................................................................................... ...44 Procedure ..................................................................................................................... ....47 Results .....................................................................................................................................52 Manipulation Checks .......................................................................................................52 Excuse Making ................................................................................................................55 Affect ........................................................................................................................ .......61 Excuse Making and Affect .............................................................................................. 62 Discussion .................................................................................................................... ...........63 3 STUDY 2 ................................................................................................................................68 Method ........................................................................................................................ ............68 Participants .................................................................................................................. ....68 Instruments ................................................................................................................... ...68 Procedure ..................................................................................................................... ....68 Results .....................................................................................................................................71 Manipulation Check ........................................................................................................71
5 Use of Excuses: Task 1 ....................................................................................................72 Affect: Task 1 ..................................................................................................................75 Use of Excuses: Task 2 ...................................................................................................76 Affect: Task 2 ..................................................................................................................79 Path Model .......................................................................................................................79 Discussion .................................................................................................................... ...........85 4 GENERAL DISCUSSION ..................................................................................................... 93 APPENDIX A STUDY 1: QUESTIONNAIRE .............................................................................................98 B STUDY 2: FIRST QUESTIONNAIRE ............................................................................... 104 C STUDY 2: SECOND QUESTIONNAIRE ..........................................................................107 LIST OF REFERENCES .............................................................................................................111 BIOGRAPHICAL SKETCH .......................................................................................................119
6 LIST OF TABLES Table page 2-1 Main effects of option number on individual excuse item s in Study 1 ............................. 66 2-2 Factor loadings for excuse fact ors (Study 2 loadings in parentheses) ............................... 67 3-1 Main effects of option number on individual excuse item s in Study 2 ............................. 91 3-2 Option number excuse interaction on interest ................................................................ 91 3-3 Option number excuse inte raction on task-interest excuses ...........................................91 3-4 Factor loadings for excuse f actors in second ques tionnaire of Study 2 ............................. 92 3-5 Zero-order correlations between excuse s after tasks one and two and affe ct after tasks 1 and 2 ................................................................................................................. ......92
7 LIST OF FIGURES Figure page 2-1 Screen shot for car task th at participants experienced .......................................................66 3-2 The SEM models linking excuse m aking (re verse-scored: higher numbers = more excuses) after first and sec ond task and affect (higher num bers = better affect) after first and second task.. .........................................................................................................90
8 Abstract of Dissertation Pres ented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy EXCESSIVE OPTIONS AND EXCUSES By Ryan M. Johnson May 2008 Chair: Barry R. Schlenker Major: Psychology Research shows that individuals in western society have felt increasingly less responsible for the outcomes in their lives over the past four decades. This dissertation examined two potential causes of this trend by determining whether excuse making after poor decisions is facilitated by (1) information overload caused by excessive options and (2) pro-excuse social environments. In two experimental studies, participants c hose between either 6 or 18 cars with the goal of finding the car that most clos ely matched a given standard. A ll participants in both studies received failure feedback. In Study 1, participants (N = 256) were also given physiological feedback that was intended to manipulate the inte nsity of negative affect they felt after failing. After the feedback, they completed a questionnai re measuring excuses and affect. In Study 2, participants (N = 162) were eith er given or not given a salient excuse for failure before the decision task. After the decision task and ques tionnaire, they completed a memory task with failure feedback followed by another questionnaire. In both studies, participants with more options made more excuses for failure than participants with fewer options. These excuses ranged from relatively legitimate to relatively illegitimate, suggesting that participants w ith too many options disengaged their general
9 perceptions of responsibility over their decision, a nd that led them to take multiple routes of excuse making even when some were not warrante d. In Study 2, participants who were provided with an excuse for failure on the first task also made more excuses after the second task despite the excuse not being relevant to the second ta sk. This effect again suggested that making excuses in one context can lead to more excuse s in another, irrelevant contexta metaphorical "culture of excuses." Finally, structural equation modeling in Study 2 showed that making excuses may actually hurt affect i mmediately after those excuses are used but improve affect in the long run. In general, this research suggests that trends in excuse making in western culture may be in part due to pervasive information ove rload resulting from an overabundance of options in everyday decision making, and a more pro-ex cuse culture may perpetuate this effect.
10 CHAPTER 1 INTRODUCTION These studies were primarily designed to examine the effects of the number of available options on self-reported personal re sponsibility for a decision that turns out poorly. For example, imagine someone chose a product from a list of 10 options, and the product broke after only a few days. The probability that the person woul d not feel responsible for the poor outcome may differ if the number of options was higher or lo wer than 10. Would someone be more likely to defer responsibility for the poor decision if there were only three options from which to choose? What if there were 50? More over, why should option number have any effect on perceptions of responsibility at all? I hypothe sized that more options would be associated with more excuse making and perceptions of less personal responsibility. A secondary purpose of these studies was to determine if changes in reported responsibility reflect an overall change in attributional style that ca rries over to other tasks. In other words, would someone who made a lot of excuses regarding a poor decision be more likely to make excuses about some other, unrelated negative outcome at a later time? Would this chronic shift to a more external attributional style be more likely if the initial number of options was too cognitively difficult to process? What if the decision were simple (i.e., only a few options)? It was hypothesized th at people who experienced conditions that facilitated excuse making on one task would then be more likely to make excuses on later, irrelevant tasks. These questions were addressed in two studies that assessed the causal effects of option number on responsibility and some possible processe s that could be drivin g those effects. The following sections will set the stage by providi ng contextual background for the research questions and then describing the research that was conducted to an swer those questions. First, a review of some sociological trends and some examples of relevant, real-world phenomena
11 explains why the importance of these research questions goes beyond basi c laboratory research. Second, a scientific literature review combines information from social psychological and marketing sources to establish theoretical back ground for predictions invol ving the constructs of interest. Finally, two studies are pr esented that address these issues. Sociological Background Anthony Rezza was fired for e mbezzling $2000 from the federal government, which he subsequently lost gambling in Atlantic City. Ho wever, the man was reinstated in his job after suing the government because, as a compulsive ga mbler, he was not considered responsible for his actions and was thus protected by the Am ericans with Disabilities Act (Rezza v. United States Department of Justice, 1988). This laws uit is an example of someone using a self-serving excuse to defer personal respons ibility and gain an advantag e after a negative outcome. Ridiculous as it might seem, howe ver, the story reflects a trend in western society over the last 40-50 years toward a social climate that is more accepting of excuses than ever before, or, at least, more accepting of excuses at the level of the individual (as opposed to the level of the corporation, school, government, etc.). Examples of this trend are present in many aspects of contemporary society. The number of civil lawsuits in America, for example, almost tripled between 1960 and 1990, and the numbers of lawyers also increased dramatically faster than the general population (Margolick, 1991). According to some estimates, the United States has well over half the total number of lawyers in the world (e.g., Sykes, 1992) w ith over 1,000,000 total (R owley & Hall, 2006) dozens to hundreds of times more lawyers per person than any nation in the world. Between 1970 and 2000, the number of lawyers per 1,000 Ameri cans has increased 233% (Caplow, Hicks, & Wattenberg, 2001). Although it is certainly deba table whether this tremendous increase is lawyers is a cause or an effect of the shift toward less personal responsibility on the part of the
12 individual, it is clear that the tre nd toward more civil litigation implies that it is more acceptable than ever for individuals to make excuses in a variety of situations. Although one could argue that an increase in lawsuits may actually reflect an increase in the social importance of responsibility (i.e., the purpose of lawsuits is to determine who is responsible for some outcome), that argument does not account for the fact that the people bringing all these lawsuits are doing so because they do not feel (or do not want to ap pear to be) responsible for their outcomes, and that is by definition an excuse. Thus, even if legal responsibility is still deemed important at some level (e.g., corporations are more accountable than ever for potential health risks posed by their products), that ha s not affected the increase in lega l excuse-making on the part of the individual. The number of ways in which people can be categorized as minority members has also increased dramatically in recent decades. In the 1960s, minority status tended to be based on relatively concrete attributes th at people had no control over, su ch as race, sexual orientation, gender, and, to a lesser extent, religion. Now, however, virt ually every person in the United States is a member of multiple minority groups and membership in many of these groups provides a ready-made excuse for poor performan ce, health, affect, self-esteem, and many other negative traits, behaviors, and outcomes. As Charles Sykes ( 1992) argued, everyone these days feels like a victim in one way or another, and ma ny transgressions can be at least partly blamed on real or perceive d discrimination. In education, the number of students who are being classified as learning disabled has increased exponentially since the invention of the term in the 1960s. By 1996, between 7% and 15% of students were labeled as learning disabled (Sykes, 1996). Some psychiatrists emphasize that, although some learning disabl ed students have distinct physiological
13 characteristics, the vast majority have no neurological deficits, id entifiable genetic influences on their disorder, and no clear clinical descriptions (e.g., Breggin, 1991). Whether or not these students actually deserve their label is irrelevant to this point: giving excuses to students who are failing is more common now than it used to be. Similarly, clinicians have defined substantially more psychological disorders and diagnosed a s ubstantially higher percentage of the population with these disorders than ever before. Psychiatrist Irwin Savo dnik (2006) points out that, in the last 50 years, the number of iden tified mental illnesse s has risen from 107 to 365, and this is entirely due to changes in what is deemed normal by society, not by an increase in actual disorders. The result of this tremendous in crease in mental disorders is a reduction of accountability for negative outcomes (Savodnik, 2006) In other words, the more mental disorders there are, the more likely someone is to be able to use one as a realistic excuse for committing a crime, failing in school, sabotaging pe rsonal relationships, or any other type of problematic behavior. Again, whet her or not these disorders are le gitimate is beside the point the point is that there are more excuses availa ble to people than there used to be, and people seem to be very willing to use them, regardless of whether they end up better or worse off for the experience. Even measures of traditional psychological variables have shifted more towards excusemaking attributional styles. For example, scor es on Rotter's (1966) locus-of-control scale have shifted almost a full standard deviation to th e external side of the scale since 1960 (Twenge, Zhang, & Im, 2004). Relative to those with a more internal locus of control, those with a more external locus of control expect that they have less control over achieving desired goals, because they believe that random interfer ence and other external influences will have a large effect as
14 well. This pattern indicates a trend toward pe rceptions of less contro l in goal achievementan important component in excuse making (Sch lenker, Pontari, & Christopher, 2001). It is important to emphasize that the purpos e of this introduction and the studies that follow is not to condemn the use of excuses as accounts for behavior. Social psychological research has demonstrated that, in many cases, excuse-making attributional styles can alleviate negative affect (Mehlman & Snyde r, 1985), protect self-esteem (Crocker & Major, 1989; Major, Kaiser, & McCoy, 2003; McFarland & Ross, 1982), lower the risk of depression (Peterson, Luborsky, & Seligman, 1983; Peterson & Seligman, 1984; Raps, 1982), and improve health (Snyder & Higgins, 1988), among other things. Everyone uses excuses, and, in general, excuses seem to be very effective at improving the outlook, performance, and health of the user as well as improving the users impressions in others. On the other hand, there are costs associated with excuses, such as when a socially acceptab le excuse leads to worse performance on an uninteresting, unmotivating task (Johnson & Schlenker, 2008), or wh en flimsy excuses make the user appear less reliable and more narcissistic (Pontari, Schlenker, & Christopher, 2002). Therefore, it would be simple-minded to argue that a shift in a society toward more excuses is clearly a good or bad thing in ge neral; it simply is what it is. However, because more excuse making can have clear and identifiable costs and benefits, it is important to determine some of the primary cause s for a shift in personal responsibility. It is likely that a decrease in the importance of pers onal responsibility is multiply determined, and it is unlikely that any one cause is clearly the primary one. Ther efore, although th ere are surely many variables that could cause more excuse making, the studies presented will focus on the increasing number of options in an increasingly complex society. This variable has the
15 advantage of being easy to identify in society at large and easy to manipulate in a laboratory setting to determine what causal effects it may have on personal responsibility. Decision Making and the Lure of Choice Over the past 40-50 years, excuse making has covaried with an increase in the daily necessity of sifting through ever-increasing amounts of information and choosing from increasingly large choice sets. Decisions such as choosing a car or deciding which career to pursue have become significantly more complex than they were decades ago, in large part because there are so many more options than there used to be. Tasks that used to be relatively straightforward and limited (e.g., making a phone cal l) now carry with them a host of auxiliary decisions (e.g., which carrier is best?; cell phone, land line, or ISP?; block caller ID?; etc.) and necessitate even more decisions once the call is made (e.g., voicemail services offer multiple options such as leaving call-back numbers, forwarding numbers, shipping information, text messages, and/or voice messages). The prolif eration of mail-order catalogs, home-shopping networks, and the internet gives shoppers the op tion of considering an almost limitless number of products that can be examined, purchased, and shi pped without ever leavi ng home. Objectively, this pervasive presence of huge amounts of information and huge amounts of alternatives when making decisions has given people more control over many of their day-to-day decisions. When choosing virtually any well-establi shed product, today's consumer has a great deal more control over the exact specifications and attributes of that product than ever before. However, given that general perceptions of control over life's outcomes have decr eased (Twenge et al., 2004) along with the increase in objective control, it is possible that the relationship between actual control and perceived control may not be as simple as it seems. Although there are certainly countless other variables that have had some impact on gene rational changes in percep tions of control, the
16 relatively common necessity of highly complex decision making may still have an important impact on those perceptions. Is a large amount of contro l and agency in decision making a good thing? Since Seligman's (1975) research on learned helplessness, much of traditional psychological research has supported the notion that ha ving no control (or the percep tion of no control) over the outcomes in one's life can lead to lower motivat ion, well-being, confidence, and other negative outcomes (e.g., Cordova & Lepper, 1996; Deci, 1 981; Deci & Ryan, 1985; Deci, Spiegel, Ryan, Koestner, & Kaufman, 1982; Langer, 1975; La nger & Rodin, 1976; Perlmuter & Monty, 1977; Schulz, 1976; Taylor & Brow n, 1988; Zuckerman, Porac, Latin, Smith, & Deci, 1978). More specifically, decision-making resear chers have found that a lack of control over important, selfrelevant decisions leads to nega tive affect. For example, Ackerman and Gross (2006) found that, relative to students with no contro l, business students who were as ked to choose their classes for a hypothetical marketing minor (as opposed to ha ving their classes chosen for them) reported more satisfaction, desire for the option, and happi ness, and they reported less stress and regret. On the other hand, a smaller collection of res earch has shown that, although some control usually creates better outcomes than no control, it is not without its consequences. Burger (1989), for example, argued that negative consequen ces of control have b een all but ignored in the literature, and he identified several variable s that may be negatively affected by increased control, such as health, performance, and affect Since Burger's review, empirical studies have confirmed many of his predictions. For example, illusions of control have been negatively associated with trading performance in profe ssional stock traders (F enton-O'Creevy, Nicholson, Soane, & Willman, 2003), and more job control has b een associated with worse somatic health in customer-service representatives with low self-efficacy regarding their job duties (Jimmieson,
17 2000). Increasing control in general appears to be a mixed bag of consequences which vary according to moderating variables. Burger co rrectly predicted that a variety of negative outcomes can result in situations where extensiv e perceptions of contro l are not warranted (e.g., stock trading) or more control means more account ability in a situation of low self-efficacy (e.g., unconfident customer service reps). Up to this point in this lite rature review, decision-making control has been described in fairly vague terms, although there are many ways to define and operationalize control. For example, one specific way to increase control is to increase the number of decisions a person can make along the way to achieving some end. So meone who gets to c hoose the destination, airline, seat numbers, and hotel when planning a vacation should l ogically be said to have more control than someone who only c hooses the destination. Interestingly, irre spective of knowledge regarding the positive impacts of control, people ge nerally appear to asso ciate more choice with more positive outcomes, and, if gi ven a choice, they generally pref er the path that leads to as many decisions as possible. In fact, even anim als prefer the option of making more decisions. Rats (e.g., Voss and Homvie, 1970) and pigeons (e .g., Ono, 2000) prefer to make more decisions even when the outcome is exactly the same as with fewer decisions, and primates (e.g., Suzuki, 1999) also prefer to make more decisions, as long their apparent chances of success are at least as good as with fewer decisions. Humans also have inefficient preferences rega rding choice, preferring to take the path of more decisions when the odds of success are the same (Beattie, Baron, Hershey, & Spranca, 1994) or even when they are worse (Bown, R eed, & Summers, 2003). Ultimately, contrary to people's expectations when choosing how many decisions they want to make, making more decisions leads to less satisfaction before the outcome of the decisions are even known (Botti &
18 Iyengar, 2004). In summary, people prefer having more control in their ab ility to make decisions for themselves, and they prefer maximizing the numb er of decisions they get to make even when it results in no advantag e over fewer decisions. Although there are certainly many other ways to define control when referring to decision making (e.g., the total number of decisions made, the total number of attr ibutes or qualities of decisions, the social pressures associated with making certain decisions), I will only focus on onethe number of options when making a choice. The primary reason for this focus is that option number is the clearest opera tionalization of control that can be said to have increased along with excuse making at a societal level. Although it is relatively clear that people in American society currently have more options from which to choose when making typical decisions (e.g., buying products, choosing educational institutions, pursing career options, choosing leisure-time activities, etc.) than ever be fore, it is much less clear that people have more control in other ways (e.g., the number of discrete decisions that people make each day). One purpose of this research is to determine a potentially important causal variable that can explain the increase in excuse making that has occurred in recent decades, and option number is the only control-related variable that can be quantifiably determined to covary with excuse making in society in recent decades. There is less research on the topic of option number than there is on the topic of general control in decision making (i.e., the difference between having some control vs. no control), but the results seem generally consistent. Initial rese arch on the topic showed that, similar to general preferences for personal decisions and the largest number of deci sions possible, people seem to prefer more options (Kahn & Lehmann, 1991) an d more thoroughly enjoy the decision-making process when more options are offered in rela tion to fewer (Iyengar & Lepper, 2000), and more
19 recent studies confirmed this pattern (Kahn & Wansink, 2004). However, although people seem to prefer as many options as possible, having more options actually causes people to be less likely to make a decision at all (Chernev 2003a; 2003b; 2006; Iyengar & Lepper, 2000; Tversky & Shafir, 1992). For example, Boatwrig ht and Nunes (2001) showed in a field study that people were more likely to buy products in a grocery store as th e number of options increased, but that trend flattened out and eventu ally reversed as the number of options became too large. Chernev (2003a; 2003b) showed that this effect of too many options reducing preferences is moderated by the salience of a premeditated ideal option. In essence, although it appears that people enjoy having variety, excessive variety also makes the decision process more difficult and often leads people to av oid making a final decision at all. Bown et al. (2003) described and compared three similar characteristics of decisionmaking control that could be described as elicitin g the lure of choice": (1) number of options, (2) number of independent decisions, and (3) number of decisions about each attribute of a given group of options. The authors made the case th at the lure of choice resulting from having a relatively larger number of separate decisions to make is different from the lure of choice that results from having a relatively larger number of separate options to choose from when making a single decision. For example, it is possible that a person could be lured by a store display with a large number of alternatives (e.g., a display of 50 different pre-made gr eeting cards) because of the opportunity to exactly match his/her exact, preferred specifi cations. However, that same person might not be lured by a stor e display that required a sequen ce of decisions that led to a previously undefined product (e.g., a make-your-own greetin g card station that requires sequential decisions about paper type, paper size, pa per color, font, ink color, graphics, etc.). This difference in preferences could exist even if the sequence of decisions in the second
20 example and the comparison of attributes in the fi rst example each required the same number of separate considerations, mental effort, and distin ct decisions. The third type of characteristic (i.e., the degree to which attribute decisions are divided into smalle r and smaller units rather than single, holistic judgments) will not be addressed in detail here, because it not as relevant to the purpose of this research. The lure of choice is demonstr ated in each of the three type s of situations, because people tend to prefer more choice in all three (Bown et al., 2003), but because they do not necessarily reflect the same processes, each deserves independent testing. This research will focus on effects related to large numbers of options in a given decision rather than ot her manifestations of decision-making control for these reasons: ther e is less research on the relationship between option number and responsibility than between general decision making and responsibility, and the option-number manifestation is most relevant to the increase in decision making in society at large. Option Number and Information Overload Information When considering how an increase in options could lead to any ps ychological outcom e, one must determine what characteristic(s) of la rger choice sets make s them different from smaller choice sets. The most noticeable characteris tic of larger choice sets is that they contain more options and, consequently, necessitate more considerations, comparisons, and cognitive load than smaller choice sets. Fo r the purposes of this research, the aspects of a choice set that result in this additional cognitive load will be lo osely referred to as information. There are many ways to define and quantify the amount of inform ation in a choice set, su ch as the actual or perceived number of options, at tributes of the options, and th e distribution (i.e., relative frequency) of attributes among the options. Res earchers have used thes e and other aspects of
21 choice sets both separately and in combina tion (e.g., Kahn & Wansink, 2004; Keller & Staelin, 1987; Lurie, 2004; Malhotra, 1982), but it is unclear which, if any, is the most important. However, before dealing with which method of m easuring information is the best, it is important to emphasize why information is important. Researchers have used the term "information overload" to describe what happens when a decision maker must make a decision in the presence of more information (in whatever form) than he/she can effectively process. Jac oby and his colleagues (Jacoby, Speller, & Berning, 1974; Jacoby, Speller, & Kohn, 1974) were the firs t to study the idea of in formation overload in consumer-choice situations. They found that people made worse decisions when there was too much information to process adequately. Their initial results were crit icized for methodological reasons, but subsequent researchers (e.g., Diehl, 2005; Lurie, 2004; Keller & Staelin, 1987; Malhotra, 1982) used a variety of different methods and operationaliz ations of decision quality to replicate the finding that more information be yond some ideal amount of information leads to worse decisions. In general, researchers have found that decision quality is positively correlated with information up to a level of information th at is not cognitively overwhelming, but, once the amount of information exceeds the amount that ca n be easily processed, decision quality declines as more information is added. Although the objective quality of d ecisions is not of interest in the studies presented, it is reasonable to assume that objective quality and s ubjective experience (e.g., regret, responsibility, satisfaction, decision confidence, etc.) would be co rrelated. It is therefor e important to examine the effects of information on choice quality, becaus e there is so little rese arch on the subjective, psychological effects of decisions made from to o much information in the form of too many
22 options. So, given that decision quality decreas es beyond some critical amount of information, what exactly is the best way to de fine and quantify information? Information quantity There are at least three ways th at information in a choice set can be increased: (1) incr easing the number of options (e.g., 5 vs 10 shirts in a choice set), (2) increasing the number of attributes (e.g., 10 shirts that differ on size and color vs. 10 shirts that differ on size, color, brand, and fabric), and (3 ) increasing the homogeneity of attributes such that each option in the set is less distinct from the others (e.g., 10 shirts that are each clearly different from all other shirts in the choice set vs. 10 shirts that ar e all relatively equivalent to all the other shirts in the choice se t). However, research indicate s that all three dimensions are relatively independent measures of information and thus do not need to be combined in studies. For example, studies that included factors for numb er of options and number of attributes did not find significant interactions between those factors (e.g., Jacoby, Spe ller, & Berning, 1974; Jacoby, Speller, & Kohn, 1974; Lussier & Ol shavsky, 1979; Olshavsky, 1979; Malhotra, 1982), and studies that included fact ors for number of options and di stribution of attributes also found no interactions (e.g., Lurie, 2004). Theref ore, because it makes sense to treat the three types of information as independent causes of de cision effectiveness, the first (i.e., number of options) was chosen as the factor of interest for this research. Future re search should include the other two factors to determine if they produce different change s in the kinds of subjective psychological variables that were m easured in the current studies. There is a distinction that is missing in almost every study on option number in choice sets. Studies that compare large and small choice sets do not typically attempt to determine cognitive differences between those conditions that go beyond the number of options, and it is easy to understand how these unidentified confounds could be problematic When manipulating
23 the number of options in an experiment with tw o choice-set conditions (i .e., one condition with a large number of options and one with a small num ber of options), researchers typically make sure to keep the number, variety, and distribution of each option's attributes the same in both conditions. Although this is a good thing because it assures that the two conditions have options that differ only in number, it confounds information in the choice set and number of options in the choice set, because a small group of options is obviously going to contain less information than a larger group of equivalently complex options. It is therefore impossible to tell if betweencondition effects are a result of having more vari ety, having more information to sift through, or some combination of the two. Lurie (2004) attempted to address this issue as it relates to decision quality. He computed a measure of uncertainty in bits, which are th e unit of measurement for uncertainty in a choice set as defined by Shannon (1949) in his theory of information structure. Generally, uncertainty is a measure of the average likelihood of an attr ibute occurring. It can al so be thought of as the inverse of information (Garner, 1962). In the case of choice se ts and decision making, uncertainty is determined by the number of opti ons, the number of attri butes that each option possesses, the number of values or levels of each attribute, and the rela tive frequency of the levels of those attributes among options across th e entire choice set (Lurie, 2004). Lurie used Garners (1962) equation in an unknown way to control for the amount of information in a group of options by measuring the degree to which attribut es of the options were evenly distributed. In short, more evenly distributed options have mo re uncertainty associated with them, because distinguishing between them is more difficult. Fo r example, if only one car in a choice set of 10 had leather upholstery (a rare o ccurrence), that attribute would provide more information (i.e., less uncertainty) than if five cars had leat her upholstery (a common occurrence) and would
24 clearly distinguish that car from others in the ch oice set, meaning that the cars would not have to be examined as closely to differentiate them. In other words, it is c ognitively easier to choose from very different options than it is to choose from very similar options so it is theoretically possible to have one choice set with 10 options th at is as easy to process as a choice set with 15 options as long as the second set has fewe r or less evenly distributed attributes. Lurie found that distribution of attributes among options and the number of options both independently affected decision quality, and, when uncertainty (measured in bits) was controlled in a regression equation that included number of options, uncerta inty remained significant and number of options did not. Thus, he argued th at uncertainty (i.e., info rmation) was the best predictor of decision quality, and in formation structure should be cl early defined in terms of bits rather than simply counting attributes. Howeve r, there are a number of reasons why a similar approach was not used in this research. First, Lurie did not clearly define how he computed his uncertainty coefficients, so it is unclear how opti on number is relevant to the computation. Also, other authors only use Garners original equation to study uncertainty as it relates to binary probability of attributes (i.e., the probability th at an attribute does or does not occur), whereas Lurie somehow used the equation (or some unknown adaptation of it) to account for attributes with three levels, one of which always occurs These ambiguities make it unclear how to replicate his method for calculating bits, which, to my knowledge, has not been used by anyone else. Second, and more importantly, a measure of unc ertainty defines the occurrence of each attribute the same way, thereby tr eating all attributes and their levels as equal, interchangeable parts. Although this manner of standardization may be strai ghtforward and parsimonious in a mathematical sense, treating the levels of each attribute the same across attributes leaves out
25 critical information contained in the units of the attributes themselves. In other words, the exact levels of each attribute and the units of those attributes are eq ually important for decision making in the real world, so standardiz ing them takes away a very impor tant piece of information for decision makers. For example, consider the ta sk of choosing a new cell phone from a group of other cell phones that are the same on every a ttribute except onetalk-time battery life. Suppose that the phones to be consid ered have three possible levels of the battery-lif e attribute: 2, 3, and 4 hours. Those options for battery life are pretty low by today s standards, and some are lower than the maximum amount of time a phone conversation coul d conceivably take, meaning that the 4-hour version would be the only one that was acceptable to many choosers. Now, suppose the phones have these levels of ba ttery life: 8, 12, and 16 hours. Mathematically, these two scenarios are equivalent, because th e battery lives differ by the same proportion. However, battery life is a less important attribute in the second situ ation, because all three phones have beyond the industrys average battery life for a phone, and it unlikely that anyone would need to talk on a phone for 8 or more hours without rechargi ng. Therefore, taking attributes out of their co ntext to compute uncertainty leads to the loss of a critical piece of the puzzle for determining the importance of attributes This problem has been cited by some as the reason why the calculation of information-struct ure coefficients has ne ver become popular in psychology in the way that it has in engineering, biology, physics, and statistics (Luce, 2003). In summary, measuring information with finite coefficients is a strate gy that is especially problematic when measuring the effects of t hose coefficients on s ubjective psychological variables, so that strategy was not used in th e studies presented. Rath er, counts of option number in each choice set will be the only measure of information, and further unique manipulations of information can be used in future research as needed.
26 Information quality The general quantity of inform ation is obviously important, but what about the quality? Quality can be define d as the information's inherent usefulness in assessing the utility of an option (Keller & Staelin, 1987). For example, suppose one is choosing a new desktop computer for work and has access to three attributes about potential computers in the choice set. The information would be highe r quality if those attr ibutes were processing speed, memory, and hard-drive size than it would be if the attrib utes were color, texture, and shape, because color, texture, and shape are less informative about the utility of a computer than processing speed, memory, and hard -drive size. Keller and Staeli n (1987) showed that decision quality is increased with higher quality inform ation (holding quantity co nstant) and decreased with higher quantity information (holding quality constant), but more importantly for this research, decision confidence was increased with higher quality information (holding quantity constant) and decreased with hi gher quantity information (holding quality constant). However, higher information quality only in creased decision quality to a point, and then the increase in decision quality leveled off, although increases in decision confidence did not level off. This pattern indicates that, as long as the attributes of a choice set are of sufficiently informative quality to make a good decision, they stop making a difference in decision quality. Thus, it is conceivable that, if a researcher provided attr ibutes that were high-quality enough, he/she could measure differences in information quantity without concern for any main effects or interactions involving information quality. The goal of the presented studies was to do just that: provide enough quality information about all of the options that quantity of options could be manipulated without interference from quality-related issues. Option Number and Excuse Making In his ap tly titled The Paradox of Choice: Why More is Less Barry Schwartz (2004) argued that people with virtually limitless options in life may f eel overwhelmed at times and
27 experience a host of negative outcomes as a result. He gave several examples of the kinds of variety people encounter when making even the si mplest decisions. On a trip to the grocery store, for example, Schwartz noticed that ther e were 285 different variet ies of cookies, 360 kinds of shampoo, and 275 different breakfast cereals. He went on to poi nt out that the huge increase in choices and personal freedom goes far beyond the consumption of material goods. For example, the American higher-education expe rience has shifted away from being mostly required courses to being mostly courses from which students can choose a huge variety of options. Even the most traditional institutions (e.g., Harvard, Princeton) now offer as many as 350 class options for fulfilling ge neral-education requirements, wh ereas a student's options for general education were very limited in the past (Schwartz, 2004). This trend is perhaps not surprising, given that people generally tend to pr efer having as many options as possible when making choices (Broniarczyk, Hoyer, & McA lister, 1998; Chernev, 2006; Kahn & Lehman, 1991). After citing an impressive list of other ways in which society must deal with more options now than ever before (e.g., hundreds of cable channels, complex health care and retirement plans, multiple utilities options, career options, etc.) and the ways that current values and advertising strategies encourage the quest for more choice (e.g., th e recent allowance of direct marketing of pharmaceutical companies to consumers puts the burden of drug choice on patients), Schwartz argues that having too ma ny choices can be problem atic for psychological outcomes such as satisfaction and regret. Recent empirical research pr ovides support for the notion that increasing a persons freedom to choose is only beneficial up to a poi nt, and more choice beyond that can lead to negative affective states. For example, having a large number of options to choose from when making a decision leads to feelings of increased regret, decreased sati sfaction, and increased
28 frustration (Haynes & Olson, 2007; Iyengar & Le pper, 2000). Moreover, larger numbers of options also cause people to be less confident in their choices (Chernev 2003a; 2003b; Iyengar & Lepper, 2000) and less likely to be willing to make a choice at a ll (Chernev 2003a; 2003b; Iyengar & Lepper, 2000; Tversky & Shafir, 1992). However, although there have been a small handful of studies regarding the effects of a la rge number of options on regret (Haynes & Olson, 2007; Iyengar & Lepper, 2000) and "self-blame" (Botti & McGill, 2006), there has been no research on the effects of excessive options on ex cuse making and responsibility, and it is that relationship that this research addresses. Regret and Responsibility At first glance, a casual reader might ask w hy it is necessary to determine the relationship between option number and excuse making. The relationship between option number and regret has already been established; doe snt that pretty much sum it up? The answer, which is no, is based on the fact that theoreti cal and empirical links between regret and responsibility are somewhat unclear, and the relationship between regret and excuse making is virtually unexplored. Research on regret and responsibility is so mewhat divided, because researchers disagree on the extent to which feelings of regret necessitate feelings of personal responsibility. On the one hand, most researchers regard regret as something that cannot exist without the perception of responsibility (Gilovich & Medvec, 1994; van Dijk, van de r Pligt, & Zeelenberg, 1999; Zeelenberg, van Dijk, & Manstead, 1998). According to this pers pective, regret is a negative affective state resulting from the belief that one should have acted differently to change a negative outcome over which one had control. Em bedded in this perspec tive is the notion that some decisions are "wrong" in retrospect, meaning that the actor should have known better and is thus responsible for the negativ e outcome (Zeelenberg, van Dijk, & Manstead, 1998). Moreover,
29 other negative affective states that ma y occur following a negative outcome (e.g., disappointment) are not the same as regret and thus do not require the presence of responsibility. On the other hand, other researchers have argued that feeling responsible for a bad outcome is not necessarily a requirement of regret (e.g., Connolly, Ordez, & Coughlan, 1997; Simonson, 1992), because it is possibl e to regret an action without f eeling very responsible for it. For example, regret and responsibility can dive rge when making a decision between a risky but potentially very valuable option and a safe but much less valuable option (Simonson, 1992). If one chooses the risky, valuable op tion and it doesn't work out, then one would regret the decision and feel responsible for it. On the other hand, if one chooses the safe option and it doesn't work out, then one would regret the decision because of the missed opportunity, and one would not feel as responsible for it, b ecause going with the safe opti on was obviously the best decision (Simonson, 1992). The issue has continued to be debated by philosophers and scientists with some even changing sides, such as when Ordez and Connolly (2000) esse ntially retracted their argument that responsibility and regret have no re lationship. The inconsistency of these results stems from three problems: (1) the definition of regret determines whether or not responsibility must play a role, (2) the focus of the questions given to participan ts (e.g., the behavior that lead to the outcome vs. the outcome itself) greatly in fluences responses, and (3 ) participants simply cannot (or do not bother to) di stinguish well between related affective states such as disappointment, regret, and sadness. The first two problems are difficult to remedy, because no one has developed a reliable, valid measurement of context-specific (i.e., state) regret (Schwartz et al.  developed a measure of trait tende ncy for regret). However, the third problem makes the first two essentially i rrelevant. Connolly and Butler (2006) demonstrated that there were high correlations between common measures of regret, disappointme nt and sadness that
30 people predicted they would feel and actual regret, disappointment and sadness that people later reported feeling. The same high correlations were found between positively-valenced predictions and actual e xperiences of rejoicing, elation, a nd happiness. Their conclusion was that people do not differentiate their emotions clearly enough to effectively distinguish between the related concepts of regret, disappointment, and sadness. The implication is that a general affective measure would be more useful. Researchers generally agree th at regret requires percepti ons of responsibility, whereas disappointment does not (Zeelenberg et al. 1998). However, if it is the case that regret, disappointment, and sadness are so closely related, then minor methodological differences (i.e., the nature of the questions being asked or the wo rding of the items) could account for differences between them, when, in fact, they are essentially all highly correlated co mponents of a general, negative experience. Therefore, although disappointment, regr et, and sadness differ notably when carefully defined as hypothetical constructs by social scientists, they may not differ in consistently meaningful ways for lay peopleat least, not until researchers develop reliable, valid scales that differentiate them. This definitional issue is relevant to th e research being presented because, although a small handful of studies involve option number and regret, those studies do not necessarily inform excuse making, and the primary purpose of the proposed studies was to determine the relationship between excuse making and option nu mber. Botti and McGill (2006) conducted the only study in the literature that contains refere nces to responsibility and information, but the manipulations of information (i.e., more or less differentiation of option a ttributes rather than more or fewer options), the ch aracteristics of the tasks (e.g., all possible outcomes were known to participants to be unpleasant before the task even started), and the measurements of
31 "responsibility" (responsibility defined as "sel f-blame"an operationali zation that is in a conceptual grey area between excuse making a nd regret) were dramatically different than anything presented in this resear ch. Therefore, the presented st udies involving different-sized choice sets and excuse making will provide useful information about a behavior that cannot be inferred from existing research. That said, it is likely that excuse making would be at least somewhat related to regret, disappointment, or some composite measurement of post-decisional affect. As such, Study 1 examined this potential relationship by measur ing how excuse making is affected by postdecisional affect, number of options, and the inter action of the two factors. All research on the effects of excessive options on regret (or simila r post-decisional affectiv e measures) have used regret as an outcome variable in combination with other outcome variables. However, these studies did not attempt to determine the direct e ffects of regret on other outcome variables, but, rather, considered the effects on each outcome in isolation. It is possible that post-decisional affect is an important causal force that leads to a variety of outcomes including excuse making, but post-decisional affect must be manipulated through random assignment to most accurately determine if that is the ca se (Spencer, Zanna, & Fong, 2005). Persistence of Excuse Making Over Time Up until this point, this literature review has implied a relationship that has not yet been explicitly stated. Namely, it is possible that getting into an excuse-making attributional style as a result of negative post-decisional affect and excessive options coul d cause one to maintain that excuse-making attributional style over time. In other words, a person who feels he/she must make excuses to account for an ostensibly bad d ecision may then be more likely to make excuses about other, unrelated bad decisions in the future. Therefore, it is possible that the simultaneous increase in both option number and excuse making over the past several decades may be more
32 than just a correlational relationshi p. It could be that the pervas ive presence of excessive options and excessive decisions causes people to make more excuses, and making more excuses causes a more permanent shift toward an ex cuse-making attributional style. Research on excuses typically has examined the construct by addressing three issues: (1) what causes people to make more excuses (e .g., Major, Kaiser, & McCoy, 2003; Schlenker, 1980), (2) what interpersonal and intrapersonal outcomes covary with making more excuses in a specific context (e.g., Brown & Josephs, 1999; Crocker & Major, 1989; Melhman & Snyder, 1985), and (3) what variables covary with chro nic excuse-making attributional styles across contexts (e.g., Follette & Jacobson, 1987). Ho wever, no one has attempted to determine experimentally if external influences can faci litate excuse making that then persists on subsequent, unrelated tasks. This effect is not difficult to imagine, because research has shown that there are many ways in whic h exhibiting a behavior can increase that behaviors frequency in the future. Research on catharsis, for exampl e, shows that acting aggr essively has a priming effect that promotes further aggressive thought s and actions in the future (Anderson, Carnagey, & Eubancks, 2003), and increasing perceptions of public acceptance of catharsis also increases later aggression (Bushman, Baumeister, & Stack 1999). More generally, people are usually expected to be consistent in their self-presentatio ns, so public commitments of any kind (e.g., behaviors, expressed attitudes, attributional styl e, etc.) tend increase the probability of similar actions and cognitions recurring in the future (Schlenker, 2003; Schlenker, Dlugolecki, & Doherty, 1994; Tice, 1992). This issue of chronic excuse making resulting from external influence was addressed in Study 2 by manipula ting the presence of a believable, socially acceptable excuse (thereby mimicking social norms of excuse availability and acceptance) and the number of options in a choice set.
33 Goals and Hypotheses Until th is point, my review of the literature has implied an inverted-U relationship between option number and psychological outcomes su ch as regret, satisfaction, and stress. The implied relationship suggests that psychological outcomes improve as one's control increases from nothing to something, but after some varian t level of control, a dding additional control actually starts to harm those outcomes. This relationship between information overload and detrimental psychological outcomes follows from the finding that more information leads to worse decisions and worse psychological outcom es past the point of information overload (Diehl, 2005; Lurie, 2004; Ke ller & Staelin, 1987; Malhotra, 1982). For example, Ackerman and Gross (2006) showed that students who were asked about foundation courses in marketing reported the lowest regret and st ress regarding their courses when they got to choose one third of their courses relative to those st udents who got to choose none or al l of their courses. However, the ascending part of the curvilinear relationshi p was not tested in the current studies, because there has already been a great deal of rese arch on the beneficial effects of control on psychological outcomes when conditions of no cont rol have been compared to conditions of some control (e.g., Cordova & Lepper, 1996; De ci, 1981; Deci & Ryan, 1985; Langer & Rodin, 1976; Perlmuter & Monty, 1977; Schulz, 1976; Taylor & Brown, 1988; Zuckerman, Porac, Latin, Smith, & Deci, 1978). More research on th is same issue is not really necessary. At this point, one might argue that I contradi ct myself with this statement, because the previously cited research does not define control as option number and I made a point of arguing that research on other operati onalizations of control does not necessarily generalize to option number. However, this criticis m does not present a problem, because there are clear, conceptual differences between studies that compare choice sets of various sizes an d studies that compare conditions of some choice vs. conditions of no c hoice. More specificall y, a study that compares
34 a small-option condition with a large-option condition is comparing degrees of control, whereas a study that compares a small-option condition w ith a no-option condition is comparing some control and no control. Resear ch is well established on the co mparison between some control and no control, but research is not well established on the comparison be tween some control and too much control (i.e., informa tion overload). Therefore, al though adding a no-option condition to the proposed studies would serv e to replicate findings on the c ontrol vs. no control issue, the findings would not be likely to add anything meaningful to existing literature. If a no-option control condition is omitted from the design, and a curvilinear relationship is hypothesized, then it critical to establish the sm all-option condition at or past the peak of the information-overload curve. If one chooses a number of options for the small-option condition that is on the ascending part of the curve and a number of options for the large-option condition that is on the descending part of the curve, then one might find no difference between the two conditions, thereby missing an important relationshi p. Ideally, one must therefore choose option numbers for all conditions that will fall on the same side of the curve to provide an opportunity to detect any linear pattern that may exist. In the information-overload se ction of this proposal, I postulated that, because people make worse decisi ons when overloaded with information, certain negative psychological outcomes such as regret a nd disappointment would also be more likely. Therefore, to adequately test relative differences in negative psychological outcomes (and related outcomes such as excuses), all conditions needed to include at least enough options to cause information overload. In the cas e of the current studies, information overload was defined as more options and/or attributes than participants could adequately process in the allotted time, even if they did have time to actually see al l the information. Failing to achieve information overload in the condition with the smallest num ber of options would mean failing to choose a
35 condition at or past the peak of the curve, because the top of the curve should occur at the point of information overload. Research on decisi on making as it relates to option and attribute number has suggested that people cannot adequa tely process more than approximately six (Bettman, 1979; Wright, 1975) to 10 (Malhotra, 1 982; Streufert, Driver, & Haun, 1967) pieces of information (attributes or separate options) at once when choosing between options. All conditions in both studies therefore had more than this minimum requirement to ensure information overload, because the condition of th e smallest information contained six options with three attributes each, there by equaling 18 pieces of information. In summary, replicating the ascending part of the curvilinear rela tionship between option number and excuses (i.e., the difference betw een having no options and having a small number of options) would not add much to extant li terature, whereas the descending part of the curvilinear relationship is much less established and does necessitate further research. There is scant literature on the effects of small-option-number vs. largeoption-number conditions on any variables, so any additional research would be informative. The next section includes hypotheses that expand on the notion that large numbers of options may actually lead to feelings of less responsibility. Study 1 Study 1 was a 3 (moderate-arousal feedback, low-arousal feedback, or no feedback) 2 (option number: 6 or 18 options) completely between-subjects design. Arousal feedback was manipulated to facilitate a mani pulation of post-decisional affect. All participants were given a limited amount of time to choose one car that was as close as possible to a given standard, and all participants were given feedback indicating that they did a relativel y poor job. If all participants receive negative feedback, then participants on av erage should be encouraged by that failure to make some sort of excuse to protect their public image and self-image, and differences in their
36 excuse making were thus examined. All partic ipants were assumed to be feeling somewhat negative affect after the failure feedback (college stude nts typically do not en joy failure feedback on intellectual tasks), and the affect manipulation was intended to manipulate the intensity of this negative affect. Negative-affect intensity was manipulated through false physiological feedback regarding participants' arousal during and after the task. Specifically, Study 1 was intended to answer two questions: (1) do excessive options cause more excuse making, and (2) does the inte nsity of negative post-decisional affect cause differences in excuse making independently of and in combination with different-sized choice sets? Hypotheses follow for each factor in isolation as well as their interaction. Post-decisional affect Research indicates that pa rticipants who receive false physiological feedback often internal ize and interpret that feedback su ch that their actual feelings end up in the same direction as the bogus feedback. For exampl e, Batson, Engell, and Fridell (1999) showed that participants who were given false physiological feedback indicating that they were very physiologically arous ed ended up feeling more aroused than participants who were given feedback indicating that they were less aroused. Moreove r, similar false physiological feedback has been shown to affect feelings of empathy (Batson, Tu rk, Shaw, & Klein, 1995), judgments of attractiveness (Val ins, 1966), ratings of emotionali ty (Wilkins, 1971), and attitudes regarding objectionable pictures (Parkinson & Manstead, 1981). These studies collectively indicate that manipulatin g arousal feedback not only affects participants' perceptions of their arousal but also leads participants to look for some cause to which they can attribute that arousal. In Study 1, participants were given failure feedback, so it was assumed that the most salient attribution they could make for a particul ar level of arousal woul d be that the failure feedback was responsible. Thus, it was expected that participants w ho received physiological
37 feedback indicating higher levels of arousal im mediately after their performance feedback would be more likely to interpret that physiological feedback as an indication that they became relatively upset upon learning of their failure. Alternatively, participants who received physiological feedback that indicated relativel y low, unchanging arousal after receiving their performance feedback would interpret that physiologi cal feedback as an indi cation that they were not particularly bothered by the failure. A no-feedback (i.e., contro l) condition was also included to get a baseline measure of how participants would respond in the absence of any physiological feedback. When a failure occurs, people generally accoun t for that failure using four strategies (Schlenker, 1980), including excuses (a denial of responsibility), justif ications (acceptance of responsibility along with an expl anation that the transgression was not actually a bad thing), apologies (acceptance of responsibility along with signs of remorse), and denials (a denial of responsibility and a denial that anything wrong to ok place at all). Reasonable excuses have the unique advantage of facilit ating an image of effectiveness afte r a failure (Schlenker et al., 2001). Also, in addition to a negative im pression, negative affect is also likely to be present after a failure, and an excuse would be an effective way to manage that as well. Research on excuse making and affect shows that excuse-making (i.e., external) attributional styles can often reduce negative affect following a failure (e.g., McFa rland & Ross, 1982; Mehlman & Snyder, 1985; Ouwerkerk & Ellemers, 2002). Given that excuses can reduce negative affect, one can make the following hypothesis: Hypothesis 1: More negative post-decisional affect will lead to more excuses. Number of options. There are a number of reasons to expect more excuse making under conditions of excessive options. As previously stated, excuses pr ovide a way to reduce negative
38 affect resulting from a mistake, so assuming the decision doesnt work out, an excuse would help people feel less negative affect and less ineffective in their decision-making abilities, and negative affect should be highest in the condi tion with the most options (Iyengar & Lepper, 2000). In addition, believable excuses can reduc e the negative opinions of others following a mistake (Schlenker et al., 2001), and excuses become more believable as the difficulty of the task (e.g., the more options one must consider) increase s. In other words, excuses reduce several types of negative outcomes following a poor de cision, and excuses may be more believable under conditions of an excessive number of opti ons than under conditions of a more reasonable number of options, thereby increasing their e ffectiveness for intrapersonal and interpersonal goals. Research on retroactive pessimism (i.e., the tendency to assess ones chances for success more pessimistically after a poor decision than be fore the decision) provides more insight into why more options may lead to more excuses. Although it is described in a relatively different context, retroactive pessimism basically achieves the same end as an excuse when accounting for a poor decisionto make a disappointing outcome more palatable (Tyko cinski, 2001). For example, Tykocinski (2001) showed that stude nts who incorrectly chose the winner of an upcoming election later increased their estimate of the winners likelihood of victory, whereas students who correctly chose the winner showed no such shift. Similar research (Tykocinski & Steinberg, 2005) shows patterns of retroactive pessimism that correspond in many ways to lackof-control excuses (Schlenker et al., 2001), because saying that ones probability of success was relatively low in hindsight is very similar to sayi ng that one did not have a great deal of control over the outcome. Some have even classified certain types of excuses as "counterfactual
39 excuses," thereby linking the not ion of counterfactual thinking and excuse making (Markman & Tetlock, 2000). On the other hand, one could at first glance po tentially genera te reasons to expect fewer excuses with more options, but this reverse patt ern is unlikely to actually occur. For example, having more options could be argued to be s ynonymous with having more responsibility in an objective sense, because having more options mean s a greater ability to choose exactly what one wants. However, although this may be true in an objective sense (and even that is arguable), people in these situations could just as easily argue that having more options actually lowers control over the ultimate outcome, as long as there are enough options to lower one's chances of adequately processing all the information, ther eby lowering the probabili ty of a good decision. This type of lack-of-control excuse would work nicely in a sufficiently la rge choice set, because excuses that cast the actor as fail ing because of circumstances out of his/her control are generally effective at reducing negative impr essions (Schlenker et al., 2001). Similarly, excuses related to low prescription clarity (i.e., the cl arity of the goals, rules, or scri pts for success in a situation; Schlenker et al., 2001) would also be effective, because one co uld easily argue that the large amount of information made it unclear how to find a successful strategy for assimilating it. One might instead argue that regret has been shown to increase in the presence of a large number of options, and regret has been linked to more responsibility by some researchers (Gilovich & Medvec, 1994; van Dijk, van der Pl igt, & Zeelenberg, 1999; Zeelenberg, van Dijk, & Manstead, 1998). If people are feeling more responsibility as the number of options goes up, then one might predict fewer excuses. However, as stated in the pr evious section, the link between regret and responsibility is still unclear, and it is possible that people can feel regret for a poor decision while still making an excuse to improve their mood and impression on others,
40 because lay people do not think of regret as necessitating responsibil ity in the same way as many social scientists (Connolly & Butler, 2006). More over, people who can objectively be identified as having the most responsibility are sometimes motivated to make the most excuses (Schlenker & Miller, 1977). Ultimately, because excessive op tions provide several convenient excuses (e.g., lack of control, lack of pres cription clarity, retroac tive pessimism) that can serve to reduce the negative affect and interpers onal impressions created by failu re, the following hypothesis is warranted: Hypothesis 2: A larger number of options will lead to more excuses. Combination The factors of option number and post-decisional affe ct were included together in the study to test for an interaction between them. The presence of an interaction will be informative because of the hypothesized pr ocesses behind the effect of option number on excuses and the function that excuse s serve. If it is true that excuses reduce the negative affect associated with a poor decision from a large number of options, then th at effect should be strongest when the negative affect is the hi ghest. Therefore, the following hypothesis is warranted: Hypothesis 3: There will be an option number arous al interaction such that excuse making will be highest in the moderate-arousal, 18-option condition. Study 2 Study 2 was a 2 (excuse or no excuse) 2 ( option num ber: 6 or 18 options) betweensubjects design with task (task 1 and task 2) as a within-subjects factor. As in Study 1, the first task involved choosing a car in a limited amount of time that best represented a given standard, and all participants were given the same failure feedback as in Study 1. The availability of a reasonable excuse was manipulated in the description of the time limit for the task. Some participants were told that they were in a condition with very little time relative to others (excuse
41 condition), whereas other were told that they were in a condition with a lot of time relative to others (no-excuse condition). In actuality, all participants rece ived the same amount of time for the task. Measurements of excuse making and affect were taken following the first task. Finally, a working-memory task was completed that was the same for all participants. Participants were given failure feedback after that task as well, and more measurements of excuse making and affect were taken. Study 2 had two goals: (1) determine the e ffects on excuse making and post-decisional affect of the availability of an acceptable excuse and the number of options and (2) measure the conditions under which an experi mentally-induced, excuse-making attributional style persists over time. The first goal reflected an attempt to better understand the process behind any effects of option number on excuse making. The second goa l represented an attempt to determine if excuses resulting from excessive options facilitated later excuse making regarding an unrelated task. The overall design was an attempt to reflect relevant real-w orld conditions in the laboratory (i.e., conditions under which excuses were permissible or not, option number was overwhelming or not, and there were opportunities to make excuses for poor outcomes on more than one task in a given context). The study necessarily required two tasks to establish two times at which to measure excuses for the within-subjects measurement. H ypotheses follow for the excuse factor alone and in combination with the option-number factor. Excuses Recall that in Study 1, post-decisional affect was manipulated through arousal feedback to determine if more intense nega tive affect caused more excuse making. The hypothesized process in Study 1 was that people who experience more negative affect would shun responsibility for a poor decision as a way of reducing that negative affect. In Study 2, the
42 presence of a good excuse was manipulated in part to determine its eff ects on affect. If the hypothesized process and motivation in Study 1 we re correct (i.e., that people feel negative affect following a failure, and making a excuses for their failure would reduce that negative affect), then conditions under which excuses ar e more acceptable should lead to more positive affect and more excuse making in Study 2. In othe r words, because external attributions increase positive affect following failure (McFarland & Ro ss, 1982), a believable excuse already present in the decision environment should have reduc ed negative affect preceding and following the decision, because the excuse should take off some of the pressure before the decision is made and provide a reasonable self-esteem defense after the decision in the event of a failure, which all participants experienced. In addition, because the participants were accountable to others as well as themselves for their decision, the presence of the excuse in the excuse condition should have promoted more excuse making (Johnson & Schle nker, 2008). The following hypotheses are thus warranted: Hypothesis 4: The presence of an acceptable external excuse will cause less negative affect. Hypothesis 5: The presence of an acceptable external excuse will cause more excuse making. Hypothesis 6: More excuse making will be associated with less negative affect. Excuses (over time) and option number The presence of a reasonable, socially validated excuse was expected to establish a norm of acceptability regarding excuses, thereby leading to a general clim ate of more excuses in the future on all tasks in the same experimental context. This was expected to be true even on a second task that was totally unrelated to the first, because a general social context in which excuses are permissible should apply to all events in that context. Making more excuses in the pa st facilitates making more excuses in the future
43 for a variety of self-presentational reasons (S chlenker, 2003; Schlenke r, et al., 1994; Tice, 1992). Moreover, the creation of a climate of excuses was expected to be amplified by the presence of an excessive number of options, b ecause more options should have led to more excuses in general, and more excuses on one task should have led to more on the next. Following that logic, the hypotheses below are warranted: Hypothesis 7: Excuse making will be higher after th e second task in the excuse condition when compared to the no-excuse condition Hypothesis 8: Excuse making will be higher after th e second task in the 18-option condition when compared to the 6-option condition Hypothesis 9: There will be an excuse option number interacti on such that excuse making will be highest after both tasks in the excuse/ 18-option condition
44 CHAPTER 2 STUDY 1 Method Participants Two-hundred fifty six undergraduate students (182 fe males; 74 males) received partial credit toward fulfillment of a research require ment for their introductory psychology course. Students did not know the nature of the study when they signed up for it. One participant was deleted for responding with the maximum valu e for every excuse-related question, including those that were reverse scored, which indica ted that he was not reading the questions. Instruments Excuse mak ing No one has clearly es tablished the best technique for operationalizing or measuring excuse making behavior, so the scales used in the presented studies contained several different types of items that operationalized excuse maki ng in different ways. All items can be found in Appendix A. First, a list of relatively independent items gave participants a chance to make several different types of excuses using a variety of scale points (e.g., "how easy was it for you to complete the task given the amount of time you ha d?" = "Not at all easy" to 9 = "Extremely easy"and "how much personal responsibility do you feel for your performance on the test?" 1 = "No responsibility" to 9 = "A great deal of responsibility"). The advantage of these items was that each could be answered without as much of a risk that particip ants would "anchor" their responses according to a scale that was common to all items, and that would make it feasible to analyze each item independently to look at differe nt types of excuse making. For example, the item regarding the clarity of the instructions was designed specifically to ta p "prescription clarity excuses." Moreover, because the scale points were so different for each type of item, and each
45 item assessed relatively independent characteristics, the items did not group together well in factors (exploratory factor analysis of the ite ms across both studies re vealed inconsistent, relatively unreliable factors that did not always make clear con ceptual sense), so combining the items did not seem to be the best option in analysis of the data. Second, participants complete d a list of items that ga ve them a chance to reduce responsibility on a variety of dimens ions, but all items in this section used the same scale (i.e., 1 = "Hurt my performance a lot" to 9 = "Helped my performance a lot"). In the first grouping of items mentioned above, the benefit of being able to analyze each item independently is offset by the disadvantage of measurement error associated with each item. In the second grouping of items, the common scale allowed items to be anal yzed independently or grouped together into meaningful clusters that represented different latent variables of excuse making. These groupings were determined through exploratory factor analysis that reve aled three reliable categories across both studies (see Re sults section). It did not make sense to include the initial set of individual items in these factors, because there was a clear conceptual difference between using a scale that ranged from he lping to hurting performance and a scale representing degrees of responsibility from none to a grea t deal (exploratory factor analys is supported this contention). Affect Affect was measured by using a series of 10 semantic differentials (see Appendix A for items). Participants were asked to pick a point along the 7-point continuum that represented how they felt at that particular moment regarding th e decision task (as opposed to how they remembered feeling during the task). Maximization scale The Maximization Scale (13 items; Schwartz et al., 2002) was included for exploratory purposes, although strong effects were not expected. Some research indicates that maximizers have a tendency to experience a variety of more negative outcomes
46 after making decisions than satisficers (Hayne s & Olson, 2007; Iyengar, Wells, & Schwartz, 2006; Schwartz, 2004). The Maximization Scale measures the extent to which people pursue the "perfect" option when making decisions (Schwartz et al., 2002). People range on a continuum from weak maximizing tendencies (s atisficers) to strong maximizing tendencies (maximizers). When shopping for a sweater, fo r example, maximizers (relative to satisficers) will check more stores, compare prices and other attributes of the sweaters more carefully, spend more cognitive energy trying to get the "best deal ," and, ultimately, regret the final purchase more (Iyengar, Wells, & Schwartz, 2006; Schwartz et al., 2002). It is ironi c that that maximizers feel less satisfied about their purchases than sati sficers, because maximizers are actually better at getting the best deals (Iyengar, Wells, & Schw artz, 2006). Satisficers on the other hand, shop with a level of quality in mind that is sufficient, and they will purchase the first sweater they see that seems adequate rather than continuing to shop for a sweater th at might be slightly better. Given the research on the negative affect experienced by chronic maximizers in the presence of a large number of options (e.g., Haynes & Olson, 2007), it is hypothesized that maximizers will experience more negative affect after a poor decision than satisfice rs, and that negative affect may lead to more excuses. The pattern of negative affect is especially likely to be evident in the 18-option condition, because the maximizers will have more options from which to choose, and, thus, more opportunities for indeci sion and counterfactual thinking after the decision. However, the relationship between negative affect and excuse making may not be the same for maximizers as it is for satisficers when the task involves choosing the best op tion, because that skill is more important to the identity of maximizers. The im portance of decision-making skills to the identity of the participants is one of several factors that covary with maximizing tendencies, and those make predictions for excuse making difficult. Still, given that the scale is useful in predicting a
47 variety of reactions to decisions any results involving maximizati on scores should be useful in dissecting the processes behind other effects in the study. Se e Appendix A for scale items. Procedure The design was a 3 (moderate-arousal feedb ack, low-arousal feedback, or no feedback) 2 (option number: 6 or 18 options) completely between-subjects design. Maximization was controlled as an individual difference measure, because research indicates that it may moderate the effects of the manipulated independent variables (see Instru ments section for hypotheses). Participants completed the study in groups of betw een one and four. Participants had their own cubicles from which they could not see or interact with anyone else in the room. When participants arrived, they were asked to sit at a desk with a computer. After they signed the informed consent form, the experiment er handed out instructions explaining the purpose of the study. What follows is a brief summary of how the study was described to participants: In marketing, researchers often ga ther information about a certain demographic of people to whom they want to sell products. Through surveys, interviews, and experimental studies, market researchers can ev en determine exactly what charac teristics of products are most important to people. However, even with all th is information, market re searchers still sometimes make mistakes and recommend the development of products that are not the ideal choice for consumers. We are interested in testing why ma rket researchers sometimes make these mistakes, and that is the purpose of todays study. You will be asked to play the role of a market researcher in which you must choose the best option from a list of hypothetic al cars. You will be given a limited amount of time, because market researchers typically work under strict deadlines. The car design you choose will be the one that is bu ilt and sold to the public. You will be given information about the relative importance of diffe rent attributes of cars to consumers from a
48 certain group, and you will try to choose the car with the attributes that most closely match the ideal attributes of th e consumer population. The instructions also indicat ed that the researchers were interested in how physiological changes affect decision making. When particip ants finished reading the instructions, the experimenter hooked them up to bogus physiological equipment by taping electrodes to their index and middle fingers. They were told th at the equipment would be measuring skin temperature, galvanic skin response, and hear t rate, and those variables would be averaged together to create an overall measure of physiol ogical arousal. Also, they were told that the measurements would begin as soon as they starte d the task and would continue until after the task was over. Finally, participants were told that the task had been shown to be a "good predictor of decision-making ability, cognitive speed, general intelligence, and success in business (e.g., business savvy)." Th is was done to ensure that pa rticipants would have a reason to care about their performance. After the participants were hooked up to the equipment, they began the decision-making task. The task and all dependent measures we re administered on a co mputer using Microsoft Visual Basic 6.0 (see Figure 2-1 for screen shot of car task). The particip ants saw an instruction screen that reiterated the main poi nts of the study. Then, the comput er instructions indicated that the physiological equipment needed to calibrate a baseline level of arousal so that later levels could be compared to the baseline. A clock appeared on the screen and participants were instructed to sit still for 20 s econds while the equipment established a baseline. This was done to make subsequent physiological feedback more m eaningful by ostensibly comparing participants' later arousal to their own baseline, thus indica ting change rather than abstract, meaningless values.
49 After this calibration period, participants we re given two minutes to compare all the options of all the cars available on the screen. Particip ants chose from either six or 18 cars, and their condition was randomly assi gned by the computer. The time interval (two minutes) was pilot tested to ensure that virtua lly all participants had time to l ook at all the opti ons at least once in the 18-option condition. Participants were only ab le to look at the information for one car at a time, and the program recorded if and how many times they looked at each car as well as how long they looked at each car.1 At the top of the screen, there was a list of the three attributes on which the cars could differ and the relative importance of each to the av erage consumer. The attributes were (range in parentheses) high gas mileage (20 mpg 30 mpg), high horsepower (150hp 200hp), and high safety rating (3 stars 5 stars [ .1 star intervals]). The values of each attribute were randomly determined by the computer for each participant. The relative importance of each attribute was ostensibly rated on a 1 (not at a ll important) to 100 scale (extremely important), but no value was actually be lower than 41 or higher than 65, because having extreme numbers might make choosing too easy. The importance of the attributes was defined as 49, 55, and 61. The __________________________________________ 1Time is a crucial issue in any design involving choice sets of va rious sizes, because bigger choice sets necessarily take longer to process than smaller ones. It is thus im possible to control for time in a way that eliminates all potential confounds. There are three way to deal with the issue of time in studies involving different-sized choice sets: (1) allow the participants as much time as they wa nt, (2) control time proportionally such that larger choice sets have proportionally more time than smaller choice se ts (e.g., 2 minutes for 6-option condition and 6 minutes for the 18-option condition), and (3) keep time constant in all conditions. In the current studies, the third option was the most appropriate. The third option most closely approxim ates the real world, becaus e decision makers cannot usually spend time considering options that is proportional to the number of options. Fo r example, if one needs to purchase a new television because the previous unit broke, one likely will buy a new television within a couple of days regardless of the size of the choice set. Although time spent analyzing options is likely correlated with option number, only in rare circumstances is choice-set size exactly proportionally related to time spent considering options. The first option is inadequate, because it would not eliminate the confound of the 18-item condition taking more time, it would vastly increase the variability between participants in time and effort, and it would make the failure feedback less believable for participants who spend a great deal of time analyzing the problem. The second option is also inadequate, because although it holds time-per-option constant, it confounds total time and option number, and it is less reflective of reality.
50 computer randomly determined which value of im portance went with which attribute for each participant. Participants were able to see the importance values throughout the task, but they were only able to see the actual values of the attributes for one car at a time. The screen contained either six or 18 car-shaped butt ons labeled only by number, and the participant had to mouse over each car to see its attributes. This made quick comparisons and good decisions very difficult. After two minutes, th e program stopped allowing participan ts to see the attributes, and it indicated that they must make a choice to continue with the study. Once the choice was made, the program app eared to be processi ng the results for a moment, and then it reported that the participant finished in the 44th percentile (i.e., 56% of other participants had chosen a car that was closer to the ideal car). The actua l quality of the choice was possible to compute, but participants could no t have computed this themselves, because it is necessary to know the range of values for each attr ibute, and participants were not privy to that information. The actual quality of the choice is irrelevant to th e hypotheses of this research, and it was thus not computed. After being given their relative performance, some participants were told that they would get a chance to see a graph of their arousal that covered the time period from the calibration at the beginning of the study all the way until they clicked the button to continue. Once the button was clicked, the computer appeared to process th e data for a few moments, and it then showed a graph that indicated one of two things. In the moderate-arousal condition, the graph indicated a relatively stable amount of arousal throughout th e task followed by a moderate spike once the feedback was received. The x-axis of the gr aph was labeled with these events to help
51 participants interpre t the cause of the spike. In the low-arousal condi tion, the graph indicated relatively stable arousal during the task that was identical to that in the moderate-arousal condition, and it also indicated a very small spike in arousal after the feedback was received. In the control condition, no feedback was given. In both feedback conditions, a bogus scale was provided to help participants understand what the values of arousal on the graph actually meant. The scale ranged from 0 ("your baseline level of arousal [e.g., arousal resulting from sitting calmly"]) to 200 ("overwhelming arousal [e.g., arousal resulting from overwhelmingly exciti ng or disturbing information or activity"]). Participants in the moderate-arous al condition had a level of arousal that did not vary much from baseline (it was slightly above baseline throughout the task, risi ng slowly near the end), but it spiked to approximately 75 after the feedback was received (i.e., halfway between the levels of slight arousal and moderate arous al). Their condition was labele d "moderate arousal" instead of "high arousal," because the graph they received i ndicated moderate rather than high arousal. The feedback given to the moderate-arousal participan ts was high compared to the feedback given to the low-arousal participants, but the feedback wa s not high in an absolute sense when compared to the participants' own baselines. Arousal feedback that was too high may not have been believable, so a more moderate level was chosen for the moderate-arousal condition. Participants in the low-arousal condition ha d a level of arousal af ter the feedback of approximately 25 (i.e., halfway between baseline and moderate arousal). T hus, if they read the graph correctly, all participants within each condi tion should have interpreted the arousal on their respective graphs as representing similar levels of absolute arousal. Participants in the nofeedback condition saw no graph.
52 After the feedback, participants completed a questionnaire that in cluded all the primary dependent measures, personality measures, an d manipulation checks (see Appendix A for all items). First, participants completed closeended items that measur e their perceptions of responsibility for their decision. These questions represented some of the different kinds of excuses that are identified in th e literature, such as the thr ee types of excu ses identified by Schlenker et al. (2001) (lack-ofcontrol, lack-of-prescription-clarity, and lack-of-personalcommitment excuses) and the counterfactual types identified by Markman and Tetlock (2000). In addition, participants complete d items related to regret, disappointment, sadness, and a variety of other negative emotions that could result from the task. Personality measures and manipulation checks were completed last. Fina lly, participants were probed for suspicion, debriefed, and dismissed. Results Manipulation Checks Arousal The m ost straightforward manipulation-check item for the arousal manipulation asked "On average, how aroused did you feel AFTER the d ecision task?" (1 = "Not at all aroused" to 5 = "Extremely aroused") There was a significant main effect of arousal, F (2, 248) = 23.10, p < .0001, MSE = .75. Participants in the moderate arousal condition ( M = 2.69, SD = .95) reported feeling significantly more arousal following the task than did participants in the low-arousal condition (M = 1.99, SD = .75), F (1, 248) = 28.48, p < .0001, and participants in the no-feedback condition ( M = 1.85, SD = .90), F (1, 248) = 39.84, p < .0001. The low-arousal and no-feedback conditions did not differ, F < 1.14. There were no other significant effects, F s < 1.81. Two other items assessed participants' unders tanding of the arousal manipulation. The first item asked "According to the graph you sa w of your own arousal, how aroused were you on
53 average?" (1 = "Not at all arous ed" to 5 = "Extremely aroused"; 6 = "I did not receive arousal feedback"). A 2 (6 cars or 18 cars) 2 (mode rate arousal or low ar ousal) ANOVA revealed a significant main effect of arousal, F (1, 167) = 32.10, p < .0001, MSE = .73. The no-feedback condition was not included in this ANOVA, becau se the variance in re sponding was necessarily different for them (i.e., they should all have put "6" as a response, wher eas the low-arousal and moderate-arousal participants should have put between "1" and "5"). Participants in the moderate-arousal condition ( M = 2.41, SD = .64) reported significan tly more arousal than participants in the low-arousal condition (M = 1.86, SD = .62). There was no effect of the car factor or the interact ion in the 2 2 ANOVA, F s < 1.13. An analysis of descriptive statistics for the no-feedback participants reveal ed that the vast majority correctly reported not receiving any feedback ( M = 5.56, SD = 1.20). The next item asked "On average, how aroused did you feel DURIN G the decision task?" (1 = "Not at all aroused" to 5 = "Extremely aroused"). The part of the feedback graph during the task was not different for people in the two fee dback conditions, and thei r perceptions of their actual arousal during the task were similarly not different, F (1, 248) = 1.75, p = .27, MSE = .69. However, there was a marginal main effect (omnibus F -test) of the arousal factor on arousal felt during the task, F (2, 248) = 2.82, p = .06, such that participants in the no-feedback condition (M = 2.03, SD = .96) reported feeling more arousal duri ng the task than thos e in the low-arousal condition ( M = 1.73, SD = .77), F (1, 248) = 5.61, p < .02, but were not different from those in the moderate-arousal condition ( M = 1.89, SD = .75), F (1, 248) = 1.13, p = .29. There were no other significant effects, F s < 1.19. In summary, the arousal manipulation di d have the intended effect on arousal. Participan ts in the moderate arousal condition reported feeling more aroused than participants in
54 the low-arousal and no-feedback conditions. However, manipulating arousal was only a means to an end, because the true purpose of the ma nipulation was to manipul ate the intensity of negative affect.2 Affect Although the manipulations of arousal di d generally cause the intended effects, arousal was only manipulated for the purpose of manipulating aff ect. Thus, the affect-related items were the most telling manipulation checks regarding the effectiveness of the arousal manipulation. Participants completed a series of 11 semantic differential items, 10 of which related to cognitive and emotional states that were designed to assess participants' affective reactions to the result of the de cision task. When the 10 items re garding reactions to the test were combined (the "alerttired" dimension was not included), they showed high internal consistency, Cronbach's alpha = .91, suggesting th at the mean score on those items represented an overall measure of how negative participants felt after the task. When a 2 (6 cars or 18 cars) 3 (moderate arousal, low arousal, or no feedback) ANOVA was performed, a significant main effect of arousal emerged, F (2, 250) = 4.31, p = .01, MSE = .59. Participants in the no-feedback condition ( M = 3.76, SD = .82) were significantly less happy th an participants in the low-arousal condition ( M = 4.11, SD = .74), F (2, 250) = 8.49, p < .01, and marginally less happy than participants in the mode rate-arousal condition (M = 3.98, SD = .75), F (2, 250) = 3.27, p = .07. The low-arousal and moderate-ar ousal conditions did not differ, F (2, 250) = 1.22, p = .28. These effects were not consistent with wh at was expected from the manipulation of arousal. If the manipulation ha d the intended effect, the modera te-arousal condition should have _____________________ 2Two pilot studies were conducted in which affect was manipulated through the use of directional questions. The wording of the questions was designed to encourage participants to focus on either positive or negative affect experienced during and after the task, thereby manipulating thei r affect. The results from these studies indicated that this approach was not likely to be succes sful, so the approach used in this study was used instead, because it had a greater theoretical and empirical background.
55 led to the worst affect when compared with the low-arousal and no-feed back conditions. As it turned out, affect in the mode rate-arousal condition was in the middle of the scale, and the nofeedback and low-arousal conditions were slightly on the negative side. Three affect-related items were measured imme diately after the arousal feedback as an additional manipulation check for the arousal fa ctor. After receiving two relatively affectirrelevant items ("How much patience did you feel ?" and "How much excitement did you feel?"), participants were asked (a) "How many positive fee lings did you experience?" (1 = "None at all" to 5 = "A great deal"), (b) "How many negative feelings did you expe rience?" (1 = "None at all" to 5 = "A great deal"), and (c) "How did you feel overall about your performance on the test" (1 = "Extremely bad" to 9 = "extremely good"). The main effect of arousal was not significant for (a), F (2, 247) = 1.58, p = .21, (b), F < 1, or (c), F < 1. In summary, the manipulation checks revealed inconsistent relationships between the arousal manipulation and subsequent affect. The participants in the moderate-arousal and lowarousal conditions did not differ in meaningful ways in terms of affect. This was unexpected, because research shows that people need only to be lieve that they were more aroused to attribute the differences to salient sour ces (e.g., Batson, Engell, & Fride ll, 1999; Valins, 1966; Wilkins, 1971). However, most studies on th is topic use real-time indicators of arousal rather than a onetime indicator as used in this study, so the imp act of the one-time indicator may not have been powerful enough to elicit the expected effects on affect. Excuse Making Although including gender in the analyses for Study 1 did result in a few main effects in which males made more excuses than females (a fr equently replicated find ing in the literature, e.g., Brown & Josephs, 1999; Warner & Moore, 2004), there were no interactions and no substantive contributions from including gender as a factor. Therefore, gender was excluded
56 from all analyses reported in this section to make for more straightforward analyses. Maximization scores also failed to predict anything meaningful in the data, so they were omitted as well. As stated in the method section, the survey contained two different categories of excuse items that were analyzed in two different ways: as individual items and as clusters of items. The individual items were too distinct conceptually and statistically from each other and the other excuse items later in the scale (referred to as "com bined items" below) to be included in factors. Those items were thus analyzed individually. However, items that shared the same scale were factor analyzed (see "combined items" section) into meaningful categories and analyzed as such. Individual items A 2 (6 cars or 18 cars) 3 (mode rate arousal, low arousal, or no feedback) ANOVA was conducted on the 10 individua l excuse items near the beginning of the survey. Any differences in degrees of freedom re sulted from the fact that participants were allowed to skip questions, so not all participants answered every question. The analyses revealed several main effect s of option number (see Table 2-1 for means and standard deviations). Relativ e to participants in the 6-opti on condition, participants in the 18-option condition reported feeling the following: the task was more difficult to complete in time, F (1, 249) = 10.08, p < .01, MSE = 4.35; the instructions were less clear, F (1, 249) = 3.92, p < .05, MSE = 3.12; the task was more difficult, F (1, 250) = 3.84, p = .05, MSE = 3.43. Also, one item from the second set of excuse questions did not fit well into a ny of the three factors (factors defined in next section), so that item was analyzed in is olation. When asked to rate the degree to which "the number of attributes for each car" helped or hurt performance, participants in the 18-option condition (M = 4.70, SD = 1.42) reported significantly less help than those in the 6-option condition ( M = 5.14, SD = 1.40), F (1, 247) = 6.18, p = .02, MSE = 1.98. Some excuse
57 items in this list may seem more "accurate" (i.e ., reflecting actual differenc es in conditions) than others. However, whether accurate or not, excuse s are still excuses as l ong as they serve their intended functionreducing responsi bility and facilitating self-presentational and self-protective goals. Two items that were expected to be affected by the cars manipulation were not affected: (1) "how much personal respons ibility do you feel for your perf ormance on the test" and (2) "how much do you feel that factors beyond your personal control (e.g., background noise, testing conditions) were responsible for your performance on the test?". These items were designed to tap feelings of responsibility, but they did not differ by condition, ps = .13 and .72, respectively. When correlated with the individual items and ex cuse factors (see "combined items" section of Results) that were significantly affected by the cars manipulation, the tw o responsibility items showed very low to no correlation, r s = .00 to .19. This pattern of correlations suggests that asking specifically about feelings of responsibilit y is not the same as asking about more specific characteristics. Two main effects of arousal were significant. There was a significant main effect for the item that asked "how hard did you try on the task," F (2, 250) = 3.92, p = .02, MSE = 3.78. Participants in the moderate-arousal condition ( M = 5.21, SD = 2.11) reported that they exerted significantly less effort than those in the low-arousal condition ( M = 6.01, SD = 1.81), F (1, 250) = 7.37, p < .01, and marginally less effort than those in the no-feedback condition, ( M = 5.79, SD = 1.88), F (1, 250), p = .05. The low-arousal and no-feedback conditions did not differ, F < 1. There was also a significant main effect for the item that asked "how interesting was the test," F ( 2, 250) = 3.13, p < .05, MSE = 3.65. Participants in the moderate-arousal condition ( M = 5.89, SD = 1.74) reported that the task was significantly more interesti ng than participants in the no-
58 feedback condition ( M = 5.16, SD = 2.14), F (1, 250) = 6.26, p < .01, but not more interesting than reported by participants in the low-arousal condition ( M = 5.55, SD = 1.85), F (1, 250) = 1.79, p = .18. The lowand moderate-arousal conditions did not differ, F (1, 250) = 1.41, p = .23. There were no significant interactions between the arous al and option-number factors on any item, ps > .09. Taken together, the findings from the an alyses of individual items suggested that participants in the 18-option condition made more excuses than participants in the 6-option condition. There were also two main effects of the arousal factor, but these effects were inconsistent and difficult to interpret given the ambiguity surrounding the effects of the manipulation on affect. There were no interactions between the factors. Combined items Although the items in the second part of the excuse questionnaire could be analyzed individually, it also made sense to combine them into meaningful, statistically reliable categories to get a measure of excuse ma king that was not as likely to be biased by the wording of a single item. Moreover, because ther e were significant main effects of the optionnumber factor on almost all items in each factor combining the items made for a more efficient results section. To that end, three categorie s were compiled from the 14 items in the second excuse questionnaire. These categories were created through both a priori hypotheses and principal components analysis. Principal components analysis using oblique varimax rotation revealed three factors with Eige nvalues greater th an one (see Table 2-2 for groupings and factor loadings) that were generally consistent with a priori categories (the same three factors came out in a similar analysis on the same items from Study 2). One item ("number of attributes for each car") did not consistently load well on any factor across Study 1 and Study 2, so it was left out of the factors and analyzed i ndividually (see previous s ection for that analysis).
59 The first factor generally repr esented attributions relevant to characteristics of the testing environment (e.g., the testing environment; the in fluence of the research assistant). Lower numbers on this factor (i.e., repor ting that the characteristics hurt more or helped less) reduced responsibility by implying that aspects of the testi ng situation interfered w ith performance. This factor will henceforth be refe rred to as the environment factor One item in the group did not perfectly fit this description of the factor, however: the item la beled "your skill with computer software." That item loaded only moderately we ll on the environment factor but its inclusion in the environment factor seemed to facilitate internal consistency, a nd it loaded much more strongly on that same factor when the same an alysis was performed in Study 2. This pattern suggested that participan ts interpreted the item more in line with environmental influences than the relevance of the task to them personally. The presence of the word "software" may have been enough to enhance the salience of the situation rather than a more personal evaluation. The environment factor had an acceptable level of internal consistenc y, Cronbach's alpha = .71. The second factor represented a ttributions that were relevant to participants' interest in the task and its personal relevan ce to them (e.g., "your abstract r easoning ability"; "your interest in the task"). This factor will henceforth be re ferred to as the interest factor. Lower number on this factor reduced responsibility for poor perf ormance by implying that th e participants lacked the ability, motivation, or interest necessary to su cceed. In other words, if a task does not feel personally relevant to a person because he or sh e lacks a key component necessary for success, then he or she is not required to feel responsible for any failure s that may occur. The interest factor had a moderate level of inte rnal consistency, Cronbach's alpha = .67. The third factor was perhaps the most stra ightforward, because it was composed of items that related specifically to th e nature of the software and task itself (e.g., "number of car
60 options"; "time limit imposed by the software"). This factor will henceforth be referred to as the task factor. Items in this f actor provided a very specific ty pe of environmental excuse to participants. Lower numbers on this factor redu ced responsibility by suggesting that aspects of the task specifically (as opposed to the testing environm ent in general) interfered with performance, thereby reducing personal responsibili ty. One could not be e xpected to succeed if aspects of the task itself were fl awed or otherwise interfered with performance. The task factor had a moderate level of internal consistency, Cronbach's alpha = .57. The three factors were positively correlated with each other, suggesting that all three factors shared something in common regarding excuse-making tendencies. The environment factor correlated significantly with the interest factor, r (254) = .34, p < .0001, and the task factor, r (254) = .22, p < .001. The interest and task factors were also positively correlated, r (254) = .28, p < .0001. There were significant main effects of opti on number on all three factors (there were significant main effects on almost all the items with in each factor as well ). For the environment factor, participants in the 18-option condition ( M = 5.20, SD = .80) reported being less helped by the listed influences than t hose in the 6-option condition ( M = 5.45, SD = .88), F (1, 250) = 5.39, p = .02, MSE = .70. For the interest factor, par ticipants in the 18-option condition ( M = 4.81, SD = 1.07) reported being less helped by the listed influences than those in the 6-option condition ( M = 5.17, SD = 1.01), F (1, 250) = 7.81, p < .01, MSE = 1.06. For the task factor, participants in the 18-option condition (M = 3.74, SD = 1.03) reported being less helped by the listed influences than those in the 6-option condition ( M = 4.29, SD = 1.03), F (1, 250) = 17.86, p < .0001, MSE = 1.07. As a whole, these effects indicat ed that participants in the 18-option condition made more excuses of several varietie s than participants in the 6-option co ndition.
61 There were no significant main e ffects of arousal feedback or option number arousal feedback interactions, ps > .10. As with the individual items, analyses of the factors revealed that pa rticipants in the 18option condition made more excuses than partic ipants in the 6-option condition. Using factors provided the advantage of creating more reliabl e categories of excuses than could be found by using individual items, and the analysis of those factors suggested that th e increase in excuses found among 18-option participants included several categories of ex cuses that ranged from very task specific to generally situation specific to personal. Affect As mentioned in the manipula tion-checks section, affect was measured as the mean score for 10 semantic differential items, and there was a main effect of arousal feedback. However, there were no significant effects of option num ber or the option number arousal feedback interaction, ps > .24. At first glance, this lack of a signi ficant main effect would seem to suggest that there was not a direct effect of option number on affect. This lack of an effect could have been problematic, because research shows that too many options should lead to worse affect (e.g., Iyengar & Lepper, 2000). Thus, for explorat ory purposes, individual items were examined to ensure consistency with st udies showing that too many opti ons leads to worse affect. Analyses revealed four main effects. Participants in the 18-option condition ( M = 3.94, SD = 1.28) felt less content than those in the 6-option condition ( M = 4.28, SD = 1.28), F (1, 250) = 4.90, p < .03, MSE = 1.58. Similarly, participants in the 18-option condition (M = 4.33, SD = 1.16) felt more dissatisfied than part icipants in the 6-option condition (M = 4.04, SD = 1.12), F (1, 250) = 4.07, p = .04, MSE = 1.30. In addition, two items that were presented very early in the survey as manipulation checks for the arousal-fee dback factor (i.e., "how many negative feelings did you experience" and "how did you feel overall about your performan ce on the test") also
62 showed a main effect of option number. Participants in the 18-option condition ( M = 2.29, SD = 1.09) experienced more negative fee lings after the task than partic ipants in the 6-option condition ( M = 2.03, SD = .96), F (1, 247) = 3.87, p = .05, MSE = 1.06. Participants in the 18-option condition ( M = 4.39, SD = 1.63) also felt less good about their performance than participants in the 6-option condition ( M = 4.82, SD = 1.72), F (1, 250) = 4.10, p = .04, MSE = 2.85. Although there were fewer effects than expected of opti on number on negative affect, the effects that did surface were all consistent with previous research on the relationship between option number and affect (e.g., Iyengar & Lepper, 2000). Excuse Making and Affect One of the primary purposes of this research was to determine the relationship between excuses and affect as it relates to decisions from variously sized choice sets. The most straightforward way to determine this relationship was to examine the main effects and interactions involving a factor th at manipulated affect. However, given that the arousal factor made very little difference in affect or excuse making (and the differences were difficult to interpret where they did exist), other strategies were explored to expand on the relationship between affect and excuse making. Affect and excuse making were clearly correl ated such that worse affect was associated with more excuses. This correlation was quite consistent throughout the data, but perhaps the most straightforward tests involve d correlating the primary affect measure (i.e., the mean of the 10 semantic differentials) and measures of excu se making (reverse scored here for purposes of interpretation). For example, there were significant positive correlations between affect and the environment, r (254) = -.32, p < .0001, interest, r (254) = -.30, p < .0001, and task, r (254) = .22, p < .001, excuse-making factors. Worse affect was associated with more excuses in all cases. Conceptually identical correlations between affect and excuses were found almost
63 regardless of the variables involved. Thus, alth ough it was clear that affect and excuses are correlated, determining why was not so easy wi th the data from Study 1. However, this relationship was examined in more detail in Study 2. Discussion In Study 1, it was hypothesized that (a) more options would lead to more excuses, (b) worse affect would lead to more excuses, and (c) option number and affect would interact to produce the most excuses in the moderate-arousal, 18-option condi tion. The first hypothesis was supported: Participants who had the most options also made the most excuses on a variety of dimensions. In addition, participants made excuses in a way that relates back to the real-world context in which this research was embedded. Some excuses made perfect sense in the context of too many options. For example, it was perfect ly logical for participants in the 18-option condition to report that the task was more difficult to complete gi ven the time restraints, because it actually was. Similarly, the restrictions on seeing information from more than one car at a time really did hurt the 18-option participants mo re, because they had less time to waste clicking from one car to the next. However, the 18-option participants didn't just stop there; they also made excuses in a variety of domains that should not have been affected by the difficulty of the task. For example, relative to 6-option participants, 18-option participants reported that the task instructions were less clear and the research assist ant was less helpful. In fact, the environment factor that emerged from the principal components analysis might be described as representing a class of excuses that was irrelevant to the di fference in conditions. Participants who were already making more excuses for legitimate reasons also seemed to make more excuses for illegitimate reasons. In other words, it seemed that participants in the 18option condition disengaged responsibility more than partic ipants in the 6-option condition, but that disengagement spread beyond legitimate "reasons" to include categories of excuses that were not
64 legitimate. It seemed that 18-option partic ipants generalized th eir disengagement of responsibility to multiple categories of excuses, some of which were not particularly justifiable in the experimental context. Only two excuse-related items that shoul d have been affected by the option-number manipulation were not: the items regarding how much personal responsibility the participants felt for their performance and how mu ch they felt other, situationa l influences were responsible. Given the low or nonexistent corr elations between these items a nd other measures of excuses used in the study, however, it seems that asking people about how responsible they feel is not the same as providing them with the opportunity to make more specifi c excuses. Research suggests that social scientists sometimes parse psychologi cal states into more distinct dimensions than average people are capable of identifying on a questionnaire (e.g., C onnelly & Butler's  work showing that participants do not always distinguish well between affective states that researchers define very carefully as distinct ). Thus, it is possibl e that although people who engage in more excuse-making behaviors probabl y feel less responsible than people who make fewer excuses, they can't necessarily label their st ate of responsibility as effectively when asked directly. Moreover, people sometimes try to benefit from taking responsibility by claiming publicly that they are responsible for some outcome while simultaneously citing specific excuses designed to reduce that responsibil ity in a more discrete way (Schlenker, 1980). This strategy can maintain some of the social benefits of excuse making while avoiding some of the costs associated with overtly claiming a lack of responsibility. The second hypothesis (i.e., worse affect w ould lead to more excuses) received inconsistent support. Excuse ma king was associated with worse a ffect correlationally, but this effect needed to be interpreted with caution due to its correlati onal nature. The ideal support for
65 the second hypothesis would have come in the fo rm of a main effect of the arousal factor, assuming the factor produced the kind of changes in affect for which it was intended. However, although the arousal factor manipul ated arousal in the intended wa y, the manipulation of arousal did not cause a change in affect in the hypothesi zed way. Moreover, any effects of the arousal factor on excuse making were weak, inconsistent, and difficult to interpret in any meaningful way. Effects regarding that factor thus will not be interpreted in detail here. Also, possibly because of the ineffectiveness of the arous al manipulation, the third hypothesis (i.e., an interaction between affect and opt ion number) was not supported. The positive correlation between excuses and negative affect was consistent with the hypotheses, but it didn't actually say much about th e relationship between excuses and affect. It made sense that people who felt worse would make more excuses, but the design of the study did not provide a way to interpret whether or not ma king more excuses would eventually help people feel better. In other words, there was not e nough time between the opportunity to make excuses and the measurement of affect to see if part icipants' excuses actually worked. Study 2 did provide this opportunity, however, because it included two questionnaires separated by a distracting, difficult task.
66 Figure 2-1.Screen shot for car task that participants experienced
67 Table 2-1. Main effects of option number on individual excuse items in Study 1 Item 18 Cars 6 Cars M SD M SD Task difficulty in terms of time 4.70 2.15 3.77 2.01 Clarity of instructions 6.28 1.93 6.72 1.59 Overall task difficulty 4.41 1.89 3.96 1.80 Note First and third rows reverse scored such that higher numbers for all three items equal more ag reement with statement. Table 2-2. Factor loadings for excuse fact ors (Study 2 loadings in parentheses) Item Number Item Wording Factor 1 Factor 2 Factor 3 Testing Personal Task Environment Interest Characteristics 1. User-friendliness of the software 74 (.69) -.16 (-.07) .01 (.22) 2. Restrictions on seeing attributes for more than one car at a time .00 (.08) -.03 (-.21) 64 (.80) 3. Number of car options -.05 (-.00) .03 (.28) 74 (.63) 4. Your general knowledge of cars -.26 (-.09) 70 (.61) -.02 (.12) 5. Your skill with computer software 52 (.67) .37 (.04) -.06 (-.09) 6. Number of attributes for each car .17 (.24) .30 (.47) .44 (.26) 7. Your memory for details .20 (.19) .47 (.53) .33 (.24) 8. The testing environment (e.g., lighting, background noise, room temperature, etc.) 68 (.62) .25 (-.16) -.24 (.22) 9. Your abstract reasoni ng ability .08 (-.02) .66 (.74) .15 (.00) 10. Your ability to solve ma th problems .02 (.05) 70 (.73) .00 (-.04) 11. Influence (e.g., clarity, ambiguity) of the computer instructions 65 (.56) -.06 (.36) .19 (-.16) 12. Influence (e.g., instructions, appearance) of the research assistant 64 (.55) -.07 (.14) .21 (-.18) 13. Time limit imposed by the software .01 (-.11) -.01 (.12) 72 (.75) 14. Your interest in the task .12 (.24) 50 (.46) .02 (.07) Note. Factor loadings are standardiz ed regression coefficients. Item 6 did not consistently load well on any factor, and it reduced the internal co nsistency for each factor if included, so it was excluded and analyzed separately.
68 CHAPTER 3 STUDY 2 Method Participants One-hundred sixty two undergraduate students (124 females; 38 males) received partial credit toward fulfillment of a research require ment for their introductory psychology course. Students did not know the nature of the study when they signed up for it. Instruments Participants complete the Ma ximization Scale as in Study 1. Procedure The study was a 2 (excuse or no excuse) 2 (o ption number: 6 or 18 options) 2 (task) mixed-model design with task (task 1 and task 2) as a within-subjects factor. The beginning of Study 2 was the same as Study 1 through the end of the first task with two exceptions. First, there was no mention of arousal, no electrodes, and no arousal feedback given. Second, the written instructions given to participants explaine d that the researchers were most interested in the effects of limited amounts of time on complex decision making, and participants in different conditions ostensibly had different amounts of time to make their decision s. The instructions also indicated that the participants had been assigned to the "two-minute" condition, but this condition was described differently depending on the level of the ex cuse factor. In the excuse condition, the instructions described the participant's condition as being one of the shortest, most difficult conditions. Moreover, th e instructions explained that research has shown that making the types of complex decisions that participants would be maki ng is very difficult in only two minutes, but participants should st ill do their best to make the be st decision possible despite the difficult conditions. In the no-ex cuse condition, participants were told that their time condition
69 was among one of the longest in the study, and res earch had shown that most people didn't have much trouble making a good decision in that amount of time. The actual time for the first task was two minutes for all participants. The excuse manipulation could have been introd uced in one of two places in the method: before the first task or after the first task. Each possibility had advantages and disadvantages, but given my goals, the most useful position of the manipulation was prior to the first task. Manipulating the excuse factor before the first task meant that decision-making strategies used during the task may have differed according to the excuse, but that apparent methodological problem could also be a benefit. As in St udy 1, decision-making strate gies and effort were operationalized as the number of times participan ts look at the attributes for each option and the time they spend looking at each attri bute. If strategies and effort had differed according to levels of the excuse factor, then th at pattern may have revealed important information about the mediating role played by decision-making stra tegies and effort on later excuse making. However, because strategies and effo rt did not differ by the excuse factor1, any subsequent differences in excuse making could not be said to be related to actual differences in performance. In both cases, important information would be revealed about the effects that excuse presence has on decision-making strategies and effort. Conversely, if the excuse manipulation had occurred after task, then it could only have aff ected post-decisional attributions, and although the results would have been relatively straightforwar d to interpret, they would have provided less insight into what the presence of the excuse was actually doing. ________________________ 1A 2 (excuse or no excuse ) 2 (6 or 18 options) ANOVA showed that participants in the excuse condition did not differ from participants in the no-excuse condition on time elapsed before looking at all options or number of clicks per option, F s < 1. This suggests that having the excuse in mind before starting the task did not actually change the way participants approached the task or the effort that they put forth toward finding the correct response.
70 The first task proceeded exactly like the task in Study 1, except there was no ostensible measurement of arousal or manipul ation of arousal feedback. Once participants finished reading over their feedback, they were told that they would be given anothe r task to complete before the study was over. However, before that second task was completed, participants were given a questionnaire containing the same items as Study 1, except they did not complete any personality measures until after the second task. The purpose of the second task was explaine d by the experimenter as follows: "As you know, the purpose of this study is to examine th e ways in which people make decisions and remember complex information that relates to mark et research. However, we are also interested in how these processes are affected by the kinds of work conditions that real market researchers experience. Therefore, you will complete the next task in the presence of the kind of distracting background noise that people often h ear in office environments. Ot her participants in this study will be completing the task with more noise, less noi se, or no noise at all." All participants heard the same description of the task. The descrip tion and the presence of the background noise were designed to offer a good potential excuse wit hout clearly iden tifying it as such, thereby encouraging people to disengage responsibility. The second task was designed to be quite different from the first, because the hypotheses related to carry-over effects of excuse making on one task to a completely unrelated task. The second task was similar to a 1-back tas k, but it required me ntal computation of randomly chosen multiplicative functions between 0 0 and 11 11. Participants saw a new function appear every two seconds, and they had to determine whether the result was more than, less than, or the same as the resu lt from the previous function. Part icipants were told that they must answer as quickly and accurately as possible. Thus, participants were required to pay close
71 attention, make quick mental decisions, and remember the last thing they saw despite proactive interference from new information. The second ta sk lasted three minutes and was preceded by a short practice session. The practice session was n ecessary to ensure that participants fully understood the nature of the task. After participan ts finished the task, the program appeared to be calculating their performance compared to ot her participants, and they eventually saw that they finished in the 37th percentile relative to other participants in the study. After the completion of the second task, part icipants completed the same questionnaire that they completed after the first task with a few minor differences related to the differences between the tasks (e.g., questions re lated to the number of cars were dropped). Participants also completed the same personality measures as Study 1 and relevant manipulation-check items. Finally, they were probed for suspicion, debriefe d, and released with cr edit for participating. Results Manipulation Check The manipulation checks demonstrated that th e excuse manipulation was effective. Two surveys (see Appendices B and C) were comple ted during the study. In the second survey, the last question assessed the degree to which participants recalled th e excuse manipulation given at the beginning of the study. Th e question was "Given your time limit, how difficult should the first task have been according to the instructions you were given be fore the task?" (1 = "Not at all difficult" to 5 = "Extremely difficult"). A 2 (6 cars or 18 cars) 2 (excuse or no excuse) ANOVA was performed on this item. Th ere was a main effect of excuse, F (1, 156) = 61.44, p < .0001, MSE = 1.19. Participants in the excuse condition (M = 3.07, SD = 1.28) reported that the task should have been significan tly harder than participants in the no-excuse condition ( M = 1.71, SD = .87). Neither the main effect of cars or the interaction were significant, ps > .18.
72 Use of Excuses: Task 1 Like in Study 1, gender did not m ake significant contributions to any of the analyses (other than occasional main effects in which ma les made more excuses than females), so it was excluded as a factor in the an alyses for Study 2. Also, Maximi zation scores again failed to predict anything meaningful in the data, so they were omitted from the analyses reported here. Individual items A 2 (6 cars or 18 cars) 2 (excuse or no excuse) ANOVA was performed on the individual items presented after the first task (items were the same as those from Study 1). Participants in the 18-option co ndition reported that the task was more difficult to complete in time, F (1, 158) = 8.82, p < .01, MSE = 4.79; the instructions were less clear, F (1, 158) = 4.94, p < .03, MSE = 3.36; and the task was more difficult in general, F (1, 158) = 3.61, p < .06, MSE = 3.88 (see Table 3-1 for means and standard deviations). These were exactly the same items that were significantly affected by the option-number manipulation in Study 1. The excuse factor also caused a main eff ect in two of the variables listed above. Participants in the excuse condition (reverse-scored M s = 4.91 and 4.86, SD s = 2.20 and 1.99, respectively) reported that the task wa s more difficult to complete in time, F (1, 158) = 5.08, p < .03, and more difficult in general, F (1, 158) = 5.26, p < .03, than participants in the no-excuse condition (reverse-scored M s = 4.13 and 4.15, SD s = 2.27 and 2.21, respectively). These were relatively accurate descriptions of the differences in the instructions acr oss conditions. As in Study 1, there were no significant effects on the tw o individual items that mention the word "responsibility," ps > .16. There was one significant option number ex cuse interaction on the item that asked "how interesting was the test?" F (1, 158) = 4.20, p < .05, MSE = 3.92 (see Table 3-2 for means). The only significant simple effect was between no-excuse participants in the 6-option and 18-
73 option conditions, F (1, 158) = 6.93, p < .01. Participants in th e no-excuse/18-option condition expressed the least interest in the tasksignifican tly lower interest than participants in the noexcuse/6-option condition and ma rginally lower than particip ants in the excuse/18-option condition, p = .08. No other simple effects were significant, p s > .24. The overall pattern of this interaction implied that it took the combination of the denial of the most obvious excuse (i.e., the no-excuse condition) and the presence of a lot of options (i.e., the 18-option condition) to make a difference in reported interest. This effect could be the result of th e fact that participants in the no-excuse condition had the most obvious excuse (i.e., difficulty of task completion given the time allotment) taken away from them by the inst ructions, and only partic ipants in the 18-option condition felt a strong enough reduction of responsib ility for precisely this reason that they pursued a more creative way to make an excuse. De nigrating interest in the task was an effective way to reduce responsibility in the no-excuse cond ition, because when participants said they had limited interest, they excused themselves somewhat from responsibility by implying a lack of obligation regarding good performance (e.g., how could I be expected to put forth my best effort and do well in a task in which I had no interest?). In summary, the main effects on the indi vidual items from the option-number factor paralleled the effects found in Study 1: 18-option participants ma de more excuses than 6-option participants. Also, the excuse factor appeared to be effective, because par ticipants in the excuse condition responded in ways that were consistent with the provided excu se (e.g., reporting that the task was more difficult), but they also deni grated the clarity of the instructions (i.e., a prescription-clarity excuse). The interaction on re ported personal interest in the task suggested that, when task difficulty was specifically excluded as an excuse in the no-excuse condition, participants in the 18option condition had to be a little mo re creative in th eir excuses by going
74 beyond blaming their failure on the difficulty of th e task by instead blaming it on their own lack of interest. Combined items A principal components analysis re vealed the same factors as Study 1 (see Table 2-2 for loadings). The three factor s had generally acceptable internal consistency, Cronbach's alpha = .72, .65, and .65, respectively. Unfortunately, although the means for the factors differed by option number in the proper direction, only one of the effects was significant. On the task factor, participan ts in the 18-option condition ( M = 3.73, SD = 1.18) reported less help from task characteristics than participants in th e 6-option condition ( M = 4.29, SD = 1.03), F (1, 158) = 11.06, p < .01, MSE = 1.15. This effect was consis tent with the reality of the differences between the 18and 6-option conditions. The main effects of option number on the other two factors did not approach significance, ps > .30. There was also a main effect of the excuse manipulation on the task factor, F (1, 158) = 7.69, p < .01. Participants in the excuse condition (M = 3.78, SD = 1.19) reported less help from the task characteristics than participants in the no-excuse condition, ( M = 4.25, SD = 1.05). The main effects of the excuse manipulation on the ot her two factors did not approach significance, ps > .29. Finally, there was a signifi cant option number excuse in teraction on the task factor, F (1, 158) = 4.81, p < .03 (see Table 3-3 for means). Partic ipants in the excuse/18-option condition made the most excuses, differing significantl y from those in the ex cuse/6-option condition, F (1, 158) = 15.01, p < .001, and the no-excuse/18-option condition, F (1, 158) = 12.67, p < .001. The other two simple effects did not approach significance, ps > .42. The results from the combined items on the first questionnaire s uggested that it was necessary to have both a large num ber of options and a salient excuse in order to facilitate the
75 most excuse making. Moreover, the task factor re presented a dimension of the task that really did differ somewhat according to option number (i.e ., those characteristics of the task probably really did seem more harmful to performance in the 18-option c ondition), but it did not represent a dimension that really differed according to the excuse factor (i.e., part icipants in both the excuse and no-excuse conditions ex perienced equivalent task diffi culty). Thus, the interaction on the task factor that qualified the two main ef fects suggested that neither more options nor salient excuses were enough to cau se more excuse making; rath er, it took both to create the effect. Affect: Task 1 There were no significant e ffects involving either of the manipulated factors on the combined affect measure (mean on same 10 items as in Study 1, Cronbach's alpha = .94), ps > .14. A similar exploratory analysis of individual items showed a similar pattern to that found in Study 1.2 In addition, there was a significant main e ffect of the excuse factor on initial affect (first item on questionnaire: "how do you feel about your performance?"), F (1, 157) = 5.73, p < .02, MSE = 3.29. Unexpectedly, participan ts in the excuse condition ( M = 4.40, SD = 1.72) were initially less happy about their performance than those in the no-excuse condition ( M = 5.09, SD = 1.89). This effect was not significant on the combined affect measure later in the questionnaire, F (1, 158) = 2.14, p < .15, MSE = .92. This pattern suggested that having a viable ________________________ 2A similar exploratory analysis to the one performed in Study 1 was performed on the individual affect items for the purposes of replicating existing findings regarding the effect s of option number on affect. The items in Study 1 that differed significantly (i.e., contentment and satisfaction) by option number did not differ in Study 2, p s > .26, but the adjusted means were very similar to those found in Study 1. In addition, there was a significant main effect of option number on the disappointment-fulfillment item, F (1, 155) = 4.34, p < .04, MSE = 1.28. Participants in the 18-option condition ( M = 3.61, SD = 1.15) felt more fulfilled (i.e., less di sappointed) than participants in the 6option condition ( M = 3.96, SD = 1.13). This effect is also consistent with research on option number and affect (e.g., Iyengar and Lepper, 2000). It is unclear why the manipulation did not affect the same items significantly in both studies, but all the effects that did occur were consistent with previous research.
76 excuse led participants to initially be less satisfied with their performance, but that effect was not as strong later in the same ques tionnaire. A possible explanations for this unexpected finding is examined in the discussion. A similar analysis to Study 1 revealed sim ilar findings regarding th e relationship between excuse making and affect in the short term. Affect was significantly correlated with the environment factor, r (159) = -.42, p < .0001, the interest factor, r (159) = -.38, p < .0001, and the task factor, r (159) = -.35, p < .0001. Study 2 also contained a second questionnaire, and that allowed for an analysis of th e effects of excuse making over a longer period of time. This analysis is discussed later in this section. Use of Excuses: Task 2 The questionnaire that participants receive d after the second task (see Appendix C for items) was necessarily slightly different from the first questionnaire, because some items only made sense in light of the car ta sk, and other items only made sense in light of the memory task. However, care was taken to make the questionn aires as similar as possible for comparison purposes. Individual items A 2 (6 cars or 18 cars) 2 (excuse or no excuse) ANOVA was performed on the first nine excu se-related items after the second task. There were significant main effects of the option-numbe r factor on two of the items. Participants in the 18-option condition ( M s = 4.92 and 6.00, SD s = 2.41 and 1.82, respectively) report ed that the task was less interesting to them, F (1, 157) = 6.09, p < .02, MSE = 5.12, and that they tried less hard, F (1, 158) = 4.86, p < .03, MSE = 3.04, than participants in the 6-option condition ( M s = 5.80 and 6.60, SD s = 2.07 and 1.65, respectively). This effect is consistent with the hypothesis that conditions facilitating excuse maki ng in one situation in the experimental context will facilitate them in another situation in that same context.
77 There was only one significant main effect of the excuse factor on the individual items. Participants in the excuse condition ( M = 4.27, SD = 1.96) reported that they felt it was less important to them personally to do well than participants in the no-excuse condition ( M = 4.98, SD = 2.07), F (1, 158) = 4.97, p < .03, MSE = 4.09. There was also a marginal effect that was supportive of hypotheses, so it is interpreted conservatively here. Participants in the excuse condition ( M = 6.52, SD = 1.83) reported that the second task was more difficult than participants in the no-excuse condition ( M = 6.04, SD = 1.89), F (1, 158) = 2.80, p < .10, MSE = 3.40 (means reverse scored). As in Study 1 and the first questionnaire of Study 2, the items in the second questionnaire that contained the word "responsibility" were no t significantly affected by the manipulated factors, ps > .24. In summary, main effects of the option-numbe r and excuse factors showed that even on a second, irrelevant task, participants made more ex cuses if they were in the excuse condition or the 18-option condition in the first task. These excuses ranged from relatively legitimate (e.g., 18-option participants reported that the second task was more difficu lt, and that could be because they were more cognitively fatigued and really did perceive it to be more difficult) to relatively illegitimate (e.g., reporting that th ey did not feel the task was in teresting or worthwhile). The lack of any option number excuse interactions suggested that the two factors has relatively independent effects on individual excuse variables. Combined items In the questionnaire after the s econd task, it was inappropriate to designate three factors that c onceptually paralleled the ones found in questionnaire from Study 1 and the first questionnaire from Study 2, because the items were not the same. Those previous questionnaires included four items that referred specifically to the car-s election task, and it did not make sense to include those irrelevant items after the second task. Th ree of the four items
78 that were not included were the items that made up th e task factor in the prev ious factor analyses. Due to the absence of these items, in was inapprop riate to have a third factor that described the car task. The remaining items were much more consistent with the environment and interest factors found in previous questionnaires. A principal components analysis with oblique varimax rotation confirmed this reasoning by revealing two factors that para lleled the interest and enviro nment factors present in Study 1 and in the first questionnaire in Study 2 (see Tabl e 3-4 for loadings). The interest factor had acceptable internal consistency, Cronbach's al pha = .80, and the environmental factor had moderate internal consistency, Cronbach's alpha = .65. The item regarding the influence of the research assistant was excluded for being poorly correlated with both factors. If people made more excuses on the interest factor, they were esse ntially reporting that they lacked the ability or motivation to succeed. People who made more excuses on the environment factor were reporting that aspects of the e xperimental environment kept them from achieving positive outcomes. Excuses on either factor allowed people to appear less responsible for the negative performance. A 2 (6 cars or 18 cars) 2 (excuse or no excuse) ANOVA performed on the two factors revealed no significant main eff ects of either experimental factor on either excuse factor, ps > .23. The option-number excuse interaction wa s not significant for the interest factor, F < 1, but it approached significance for the environment factor, F (1,158) = 3.11, p < .08, MSE = .88. None of the means differed significantly from each other, ps > .15, so the interaction will be interpreted cautiously here. However, although th e differences were not si gnificant, it is worth mentioning that the pattern of means was similar to those seen in the significant interaction on the third factor in the first questionnaire. In other words, participants in the excuse/18-item
79 condition made more excuses than any other cond ition, and the other conditions did not appear to differ much from each other. This marginal eff ect again supported the notion that having a lot of options and a salient excuse were necessary to generate the most excuse making. Affect: Task 2 A 2 (6 cars or 18 cars) 2 (excuse or no excuse) ANOVA was performed on the 10-item affect measure (Cronbach's alpha = .94). As in th e first questionnaire, none of the effects were significant, ps > .16. A similar analysis on the first it em in the questionnaire ("How do you feel about your performance on the test?" ) failed to reveal a significant main effect of the excuse factor, F (1, 158) = 2.77, p = .10, MSE = 2.58. Although there were no obvious effects of the manipulated factors on affect, the relationship between excuse making and affect de served more attention, because many different levels of excuse making occurred within all conditions, and the amount of excuses people made was related to their affect. The following section includes a closer examina tion of that pattern. Path Model When the relationship between excuse making and affect on the second questionnaire was examined, a pattern arose that was similar to the pattern found in both studies thus far. More excuses (means reverse scored to facilitate interpretation) were significantly correlated with worse affect when the interest factor, r (160) = -.27, p < .001, and environment factor, r (160) = .37, p < .0001, of excuses were correlated with the 10-item affect measure. In other words, people who made more excuses on the second ques tionnaire also reported worse affect on the second questionnaire. In additi on, more excuses were associated with worse affect when the environmental, r (160) = -.20, p < .01, interest, r (160) = -.25, p < .01, and task, r (160) = -.22, p < .01, excuse factors (means reverse scored) from th e first questionnaire were correlated with the affect measure on the second questionnaire.
80 This persistent pattern of more excuses be ing associated with worse affect within the same questionnaire seemed reasonable, because it was probable that making excuses did not instantly improve affect. In other words, the fail ure probably caused most people to feel at least somewhat bad and make some excuses for it si multaneously, so a posit ive correlation between excuses and negative affect seemed reasonable. Conversely, given th at extant research generally indicates that excuses should he lp affect, not hurt it (e.g., Snyder & Higgins, 1988), one would expect that more excuses should eventually lead to better affect after some time has passed, but there was still a significant positive correlati on between excuses made on the first task and negative affect after the second task. However, although more excuses seemed to be associated with more negative affect throughout the study, it was possible that, controlling for individual differences in the tendency to make excuses a nd the tendency to report negative affect, one might see a different pattern between excuses made on the first questionnaire and affect reported on the second questionnaire. In other words, by contro lling for given levels of affect and excuse making at all time points through the use of simu ltaneous equations, one can see how excuses after the first task affected a ffect after the second task, cont rolling for the other relevant relationships. To address this possibility, structural equation modeling with latent variables was used. Two models were created (see Figu re 3-2) that contained four la tent variables: excuses made after the first task (tim e-1 excuses), excuses made after the second task (time-2 excuses), affect after the first task (time1 affect), and affect after the second ta sk (time-2 affect). Given that the different excuse factors that were found on th e first and second questionnaires were highly correlated within each questionnaire and there were slightly differe nt items in the factors on the first and second questionnaires, it made sense to combine them into single excuse making
81 variables to avoid the problems associated with picking some types of excuses over others without theoretical just ification. The newly generated ex cuse variables represented time-1 excuses for the first questionna ire (14 items; Cronbach's alpha = .79) and time-2 excuses (10 items; Cronbach's alpha = .76) for the second questionnaire. The time-1 and time-2 affect measures were the same 10-item measures used throughout the presented studies. The most obvious, logical model given the or der of presentation of the items in the questionnaires (i.e., time-1 excuses first, time-1 affect second, time-2 excuses third, and time-2 affect fourth) is the first model in Figure 3-2. This model tested whether time-1 excuses had a different pattern of effects on time-1 and time-2 affect. Conceptually, this model tested the hypothesis that excuses made earlier in the study caused worse affect in the short run and caused better affect in the long run. Setting up the model this way had the advantage of accounting for effects of earlier excuses on later excuses and ea rlier affect on later aff ect, thereby controlling somewhat for individual differences in responding tendencies. Moreover, the model also had the advantage of simultaneously comparing the dire ction of the relationship between excuses and affect for the short term of both questionnaires while also comparing the relationship between time-1 excuses and time-2 affect (i.e., the long-t erm relationship). Overal l, using a structural equation model in this way provided a more comp lete picture of the data than other possible measurements of the relationship between tim e-1 excuses and time-2 affect (e.g., correlating time-1 excuses with a change-score of [time-2 affect time-1 affect]; using regression to test effect of excuses at time 1 on affect at time 2 with affect at time 1 controlled). A second, less intuitive model (see second model in Figure 3-2) was created to account for the possibility that it was act ually affect that was driving th e differences in excuse making. This model was consistent with the notion that th e experience of negative affect might be what
82 drives excuse making in the first place, but it wa s more problematic than the first model in terms of the method of data collection, because the a ffect variables were colle cted after the excuse variables within each questionnaire. However, if one assumes th at the initial reaction to the failure was negative affect, then it would make sense that negativ e affect would have been the true starting point of the process, and excuse making would follow as an attempt to alleviate that negative affect (Snyder & Higgins, 1988). Time-1 affect and time-1 excuses were functionally measured at the same time, so it is probably as sensible to think of them as occurring simultaneously as it is to think of them as being measured at two sepa rate time points. Some structural equation modeling researchers (e.g., Klin e, 2005) suggest that di fferent variables that are measured within a very short amount of time can be said affect each ot her regardless of order in some circumstances. The second model included affect as the firs t in the causal chain. The model tested paths from time-1 affect to time-1 excuses and from time-2 affect to time-2 excuses. Most importantly, like the first model, the second mode l tested the path between time-1 excuses and time-2 affect, thereby testing if more time-1 excuses actually caused better time-2 affect. Conceptually, the model tested the hypothesis that wo rse initial affect caused more excuse making in the short run, and more excuse making caused better affect in the long run. Like the first model, the second model also controlled for individual diffe rences in responding by estimating the path from time-1 affect to time -2 affect and from time-1 excuses to time-2 excuses. In order to compare the fit of the two models to the data, models that were less complex than the saturated models needed to be used. As such, the path of the least theoretical significance was left out of both models. That path went from time-1 affect to time-2 excuses.
83 Leaving that path out of both models allowed the models to be compared using the Akaike Information Criterion (AIC), which is considered to be one of the most useful goodness-of-fit statistics for comparing non-ne sted models (Kline, 2005). Latent-variable path models were created in LISREL (v. 8.72) from the correlation matrix of these four variables (see Table 3-5 for zer o-order correlations and Cronbachs' alphas) by setting the error variance for each latent variable to be equal to the observed variance (i.e., 1 Cronbach's alpha) for that variable. Doing this created latent variables out of the observed variables and helped deal with the problem of measurement e rror associated with observedvariable models (J. Algina, personal communication, March 22, 2005). The first model presented in Figure 3-2 fit the data very well. Several goodness-of-fit indices (computer through the "ftb" option in LISREL v. 8.72 to avoid the overly liberal estimates that can be found in vers ions of LISREL after 8.52) were chos en to illustrate this point: RMSEA = .00, NNFI = 1.01, CFI = 1.00, GFI = 1.00, SRMR = 0.015. All presented paths were significant at p < .05. The structural coefficients in the model i lluminated a pattern that was supportive of the hypothesis that excuses should im prove affect over time. The significant paths from time-1 excuses to time-1 affect and from time-2 excuse s to time-2 affect show that, in the short run, excuse making caused worse affect. The critic al path, however, was the path from time-1 excuses to time-2 affect. That path was also si gnificant, but in the opposit e direction. Therefore, although more excuses appeared to cause worse affect in the short run, more excuses at the beginning of the study caused better affect later in the st udy. This effect is especially impressive when one considers that the si gnificant zero-order correlation between time-1 excuses and time-2 affect indicated that more time-1 excuses were associated with worse time-2 affect when other
84 relevant influences (e.g., response biases) were no t controlled. The advantage of the structural equation modeling approach was that it controll ed for the often powerfu l effects of withinsubject response biases on those items and a llowed for a more accurate estimation of the relationship between excuses a nd affect in the long run. The second model presented in Figure 3-2 di d not fit the data particularly well according to some goodness-of-fit indices. Several goodness-of-fit indices were chosen to illustrate this point: RMSEA = .23, NNFI = .77, CFI = .96, GFI = .97, SRMR = .033. Although the CFI, GFI, and SRMR were acceptable, the RMSEA and NNFI were not acceptable. Given that the model was equally as complex as the first model (i .e., one degree of freed om), these goodness-of-fit indices indicated that th e model was less appropriate. Most importantly, the AIC for this model was 27.72, whereas the AIC for the first model wa s 18.69. The difference in AIC indicated that the first model was a better fit to the data. However, although the path from time-1 excuses to time-2 affect was not significant for the second mode l, it did change signs just as in the first model. The first model was the best fitting, and that was not particularly surprising given that the paths corresponded perfectly with th e order of presentation of the va riables. More interestingly, the first model implied that, although making more excuses may actually cause people to feel worse in the short run, those excuses cause them to feel better in the long run. The second model did not fit the data well, so the path of greatest interest will be interpreted cautiously here. The path from time-1 excuses to time-2 affect was not significant, but the sign did change when compared with the zero-order correlation, provi ding more support for the notion that more excuses helped affect in the long run.
85 Discussion The prim ary goal of Study 2 was to determine if the presence of a good excuse, too many options when making a decision, or some combin ation of the two facilitated excuses following both a relevant and an irrelevant task. The secondary goal was to further examine the relationship between excuse making and affect. It was hypothesized that more options would lead to more excuses (hypothesis 2), and it was h ypothesized that both the presence of a salient excuse (hypothesis 5) and the use of excuses (h ypothesis 7) would both be associated with less negative affect. The hypothesis regarding the primary goal of the study (hypothesis 2) was generally supported. Participants who had an excuse for poor performance on the first task and participants who had the highest number of options made more legitimate and illegitimate excuses after the first task. Th e number of items affected by option number was smaller than in Study 1, but this was not surprising given that Study 2 had consider ably fewer participants, and it would have been necessary to have more participants than Study 1 to have the best chance at replicating all the same e ffects given comparable eff ect sizes (Abelson, 1995). An interaction also qualified the main eff ects regarding one measure of excuse making. That interaction was supportive of the hypothesis that the presence of a good excuse and a large number of options would have an interactive effect such that the combination of conditions would produce more excuses than the individual c onditions (hypothesis 10). Participants in the 18-option/excuse condition, which could be consid ered analogous to a real-world environment that frequently encourages excuses and contai ns excessive options, made more excuses than participants other conditions. Th is interaction was significant after the first task, and it also appeared to a lesser degree after the second task. The carry -over effect on the second questionnaire was particularly interesting, be cause there was no obvious reason for conditions
86 that applied to the first task (i.e., option number and the presence of a task-specific excuse) to have affected attributions for the second task. This same persistence outcome was found, albeit marginally in some cases, when measuring the main effects of the excuse and option-number factors on items in the second questionnaire. Pe ople were more likely to make excuses after their poor performance on the second task if th ey had been in the excuse condition or the 18option condition in the first ta sk. These main effects supported the hypotheses that excuse making would be higher when a salient excuse wa s available (hypothesis 8) and when the most options were present (hypothesis 9). These patterns lend support to the idea that a pro-excuse environment developed in the experimental context in which the study took pl ace. Conditions that made excuses more probable and socially acceptable in one situation c ontinued to have that in fluence on an unrelated situation. For example, on the second questionn aire, participants in the 18-option condition reported that the task was less in teresting to them and that they tried less hardboth items fall into the category of lack-of-ob ligation excuses (Schlenker et al ., 2001). Participants in the excuse condition reported that th e task was less personally important to them (another lack-ofobligation excuse), but they also (marginally) report ed that the task was more difficult. Some of these excuses may have been somewhat legitima te (e.g., the 18-option participants may have been more cognitively fatigued than 6-option partic ipants, so they really did perceive the second task as more difficult), but others were not. Fo r example, the 18-option a nd 6-option participants differed in their description of their effort, but they did not di ffer on their actual performance. Analyses of the number of correct responses on the second task showed no differences in performance between 18-opti on and 6-option participants, F < 1. This pattern suggests that the
87 lack-of-effort excuse that was more heavily us ed by 18-option participants was not legitimate, because if they had really put forth less effort, then they should have performed more poorly. People generally try to tailor ex cuses to fit the k nown facts about the situation to avoid seeming dishonest or unreliable, but they al so are willing to go beyond the known facts when that departure will not be obvious (Schlenker, 1980). Participants seemed to generalize experiences and attributions they had during and immediat ely after the first ta sk to attributions during and immediately after the se cond task. In summary, there was a carryover effect of too many options and a carryover effect of salient excu ses present in the experimental context that led to participants making more excuses on unrelated tasks. Social psychological research provides severa l explanations for the carryover effect of excuse making. For example, self-presentati onal research suggests that people who make excuses in a given situation establish a social identity (Schlenker, 1980; 1985) that they are motivated to maintain, in this case by being wil ling to make excuses in the future. Alternatively, research on catharsis suggests that acting on particular thoughts pr imes those thoughts in similar future behavioral contexts (Ande rson, Carnagey, & Eubancks, 2003). A secondary goal of Study 2 was to determine the relationship between excuse-making and affect regarding decision making, and althou gh some data suggested that the relationship was consistent with hypothesis 7, the most straightforward re lationship between excuses and affect (i.e., the relationship proposed in hypothesis 6) was not significant. If hypothesis 6 was correct, then people in the excuse condition should have felt more positive affect, because they knew going into the task that th ey could fail without having to take full responsibility for the failure, and they could look especially good if th ey ended up succeeding. However, this pattern did not present itself, as people in the excuse condition generally did not differ in affect from
88 people in the no-excuse condition, but when they did differ, people in the excuse condition were actually less happy. This finding was contrary to other studies that have shown that experimentally manipulating the salience of an excuse generally causes people to be happier (e.g., McFarland & Ross, 1982; Mehlman & Snyder, 1985). For example, McFarland and Ross (1982) showed that by randomly assigning people to conditions in whic h excuses were either encouraged or not encouraged, participants in the pro-excuse condition showed more positive affect than those in the anti-excuse condition. However, my study diffe red in an important way from this previous study, because the excuse in my study was offered before the performance started, whereas the excuse in many excuse studies (e.g., McFarla nd & Ross, 1982; Mehlman & Snyder, 1985) was given to participants after their performance. Therefore, particip ants in the excuse condition in the current study were aware that they were disa dvantaged relative to other groups in the study from the start, and receiving failure feedback after being stuck in a difficult condition may have been discouraging, thereby leading to negative affect. If one is told that one was randomly assigned to a very difficult situation, and then one does poorly, then what is there to be happy about? Another possibility is that having the excuse in mind thr oughout the task increased the salience of excuse making in general. Being pressured to make excuses may feel unpleasant if one has any doubts that one's excuses are not legitim ate, and it was possible that participants in Study 2 did feel that way. Mehlman and Snyder's (1985) study is relevant to this idea that making excuses only feels good if one is not high ly aware of the strategy. Those authors demonstrated that the presence of an excuse only made participants feel better when they were not highly self-conscious (manipulated through the use of a bogus-pipelinea technique that
89 convinces participants that thei r true attitudes are known by the experimenter). The presence of self-consciousness presumably increased the salien ce of any self-presenta tional and introspective concerns they may have had about excuse making. When highly self-conscious, participants with an excuse actually felt wors e (Mehlman & Snyder, 1985). In my study, being told about the excuse and thinking about it throughout the task may have been similar to a self-consciousness manipulation in that participants could not ignor e the possibility that they failed to overcome a disadvantage and then made excuses about it. The notion that being aware of one's excuses might lead to worse affect is relevant to some correlational data that helped clarify the un expected experimental e ffect that people in the excuse condition did not feel significantly better overall. The data in Study 2 indicated that making more excuses was associated with worse affect in the short run. However, because Study 2 had two questionnaires that were spaced out over a 15-minute period, it also allowed the luxury of a more long term examination. That examination showed that the people who made more excuses after the first task ended up feeling better after their failure in the second task. This pattern suggested that, when initially making excuses in th is situation, people may have been aware at some level that they were trying to disengage responsibilit y for their failure, and that made them feel worse in the short r un. However, once a little time and a few more distractions had passed, that dise ngagement of responsibility either began to feel less forced or perhaps a little more justifiable, and people ended up feeling better in the long run.
90 A B Figure 3-2. The SEM models linking excuse ma king (reverse-scored: higher numbers = more excuses) after first and sec ond task and affect (higher num bers = better affect) after first and second task. Asterisks mark structural coefficients that were significant at p < .05. Standardized coeffi cients are reported. Excuses (Task 1) Affect (Task 1) Affect (Task 2) Excuses (Task 2) .76* .81* -.50* .52* -.58* Excuses (Task 1) Affect (Task 1) Affect (Task 2) Excuses (Task 2) .68* .77* -.19* .08 -.56*
91 Table 3-1. Main effects of option number on individual excuse items in Study 2 Item 18 Cars 6 Cars M SD M SD Task difficulty in terms of time 5.03 2.21 4.01 2.21 Clarity of instructions 5.84 1.76 6.48 1.89 Overall task difficulty 4.80 2.10 4.21 1.90 Note First and third rows reverse scored such that higher numbers for all three items equal more ag reement with statement. Table 3-2. Option number ex cuse interaction on interest Excuse No Excuse M SD M SD 6 Cars 5.43ad 2.10 5.95cd 1.75 18 Cars 5.56 ac 2.15 4.80 ab 1.90 Note Values represent leve l of reported interest in task (higher number = more interest). Means not sharing subscripts differ at p < .01 Table 3-3. Option number excuse in teraction on task-interest excuses Excuse No Excuse M SD M SD 6 Cars 5.76a 1.09 5.66a .99 18 Cars 6.69 b 1.11 5.85 a 1.10 Note Values (reverse-scored) represent excuse score on third excuse com ponent (higher number = more excuses). Means not sharing subscripts differ at p < .001
92 Table 3-4. Factor loadings for excuse factors in second que stionnaire of Study 2 Item Number Item Wording Factor 1 Factor 2 Dispositional Situational 1. User-friendliness of the software -.14 80 2. Your skill with computer software -.02 80 3. Speed of the presentation of each mathematical function .05 62 4. Your memory for details 73 .11 5. The testing environment (e.g., lighting, background noise, room temperature, etc.) .09 48 6. Your abstract reasoning ability 79 .08 7. Your ability to solve math problems 77 .09 8. Influence (e.g., clarity, ambiguity) of the computer instructions .33 48 9. Influence (e.g., instructions, appearance) of the research assistant .35 .09 10. Your interest in the task 76 -.18 __ Note. Factor loadings are standardized regr ession coefficients. Item 9 did not consistently load well on either factor, and it reduced the inte rnal consistency for each factor if included, so it wa s excluded and analyzed separately. Table 3-5. Zero-order correlations between excuses after tasks one and two and affect after tasks 1 and 2 Variables ____ Excuses (1) Excuses (2) Affect (1) Affect (2) Excuses (1) .79 Excuses (2) .60* .76 Affect (1) -.51* -.34* .94 Affect (2) -.30* -.37* .68* .94_ Note Asterisks mark significant correlations at p < .0001. Alphas are on diagonal. Excuses reverse-scored so that higher numbers = more excuses. Higher numbers = better affect.
93 CHAPTER 4 GENERAL DISCUSSION This research was prim arily designed to ex amine whether the incr ease in the size of choice sets that people must si ft through on a daily basis aff ects excuse making. This is analogous to examining how a prolif eration of choices may be rela ted to the current climate of excuse permissiveness in western cu lture. This research was also designed to demonstrate that a social climate that is more permissive of excu se making in one context facilitates more excuse making in other contexts. I hypothesized that mo re options would lead to more excuse making and that providing a salient excuse to participan ts would produce more excuses even on a later, irrelevant task. Both hypotheses were supported. Thus, the primary goals of the presented studies were achieved, and the data painted an interesting picture that is deserving of future research. Together, the two studi es suggested that a pro-excuse environment can develop even in a contrived, laboratory context, and once that environment has de veloped, it facilitates excuses that are both legitimate and illegitimate and that occur on relevant and irre levant tasks. This implication is perhaps not surprising, becau se once norms for acceptable behavior are established, they can quickly ta ke hold through the natural tende ncy people have to want to appear consistent when in front of a given audience (Schlenker, 2003; Schlenker, et al., 1994; Tice, 1992). These studies thus suggest that although a pro-excuse environment could develop from the pervasive experience of information ove rload that is more co mmon today than ever before, a pro-excuse environment could also pos sibly perpetuate itself even without further influence from information overload. There were also several hypotheses regarding the processes behind the effects of information overload on excuse making. It was hypothesized that people who felt poorly would make more excuses, and people who had salient excuses and who made more excuses would feel
94 better. Testing the mediating function of vari ables like affect and excu ses is best done with experimental manipulation when possible rather than mediational analyses (Spencer, Zanna, & Fong, 2005), and that was the original goal of this research. However, the arousal manipulation in Study 1 did not succeed at manipulating aff ect in a significant and unambiguous manner, and the excuse manipulation in Study 2 actually resulted in worse aff ect in participants who were given an excuse. Thus, some of the hypotheses of interest were not supported experimentally, but they did receive some correlational suppor t. The first path model in Study 2 provided support for the idea that excuses do help alleviate negative affect, but some time is necessary for the effect to occur. In the short run, making excuses may actually result in worse affect. However, given the correlational nature of the relationship of processes in the excuse-affect relationship, effects relevant to that relati onship should be interpreted with caution, because successful experiments are the only way to draw confident conclusions about those effects. The limitations of the presented studies existed primarily in the nature of the manipulations, but those limitations were necessary as a first step in this line of research. First of all, some difficult choices had to be made regarding how to manipulate option number while dealing the necessary choice between confounding option number with difficulty, time, or several other variables that c ovary with difficulty and time The choice to confound option number with difficulty was made to most closely approximate the real world (i.e., situations in which people have a limited amount of time that is not constrained by option number to make a decision) and to maximize the possibility of s eeing differences in excuse making if they did indeed exist. In the future, however, it will be important to employ other manipulations that keep difficulty constant while covarying other necessary conf ounds with option number. For example, a logical next step would be to give proportionally more time to conditions of higher
95 numbers of options. Another possibility would be to keep time constant but vary the number of attributes or the restrictions of the software such that peopl e in the largeand small-optionnumber conditions perceive equivalent difficulty (and use equivalent cognitive resources) despite having the same amount of time. Regardless of the strategy, however, controlling for task difficulty and cognitive load will provide invalu able evidence regarding whether option number itself or some byproduct of option number is the critical component in determining excuse making. Another limitation could be found in a lack of hypothesized effects on the two items that included variations of the word "responsibility." The items that addressed participants' perceptions of responsibility di rectly did not show differences by condition in either study, but that could be the result of participants' lack of in sight into their attributio nal state. People are not always able to label their emotional states as accurately as researcher s define them (Connelly & Butler's, 2006), and the same may be true for attr ibutional states. In addition, participants may have claimed equal responsibility across conditions when the word itself was used in order to maintain the socially desirable image of "taking responsibility" for their failures, but they also took the opportunity to make speci fic excuses to lower responsibil ity in a more covert and thus socially acceptable manner. Future research should examine the persistent effect on excuse making of large choice sets and pro-excuse environments. The data on this issue from Study 2 were encouraging, but they were not strong enough or complex enough to fully establish how perv asive the persistence of an externally influenced patt ern of excuse making can be. Given the small effect sizes, more participants are obviously needed in future st udies. In addition, a two-part study in which participants return to the lab at a later date (or to a different la b during the same session) with the
96 same experimenter present would be particularly informative, because it is possible that a proexcuse environment is simply time-dependent and not context depende nt (i.e., the effects observed in Study 2 would disappear if more time was allowed to pass). If participants from an excuse or large-option-number condition still made more excuse s than participants in other conditions after more time had passed (or the en vironment changed), then that would say a lot about the robustness of the effect. It is not unr easonable to expect that this may occur, because research shows that norms that are established in a laboratory context can continue to influence participants after long periods outside of the lab (e.g., Sherif, 1936). This is likely to be especially true for participants who are made very aware that their responses have been seen by the experimenter and/or other pa rticipants in the study, and those audiences are supportive of the participants' responses. Finally, the relationship between affect and excuses needs further attention, and this relationship could still be addres sed satisfactorily with studies si milar to the ones presented, but the manipulations would need to be changed. For example, the excuse manipulation could be moved to after the task, but that alone might not be enough to induce a positive effect on affect. Instead, the manipulation itself may need to differ in a way that allows participants more latitude for making excuses while simultaneously seeming less "unfair" to the participants in the excuse condition. Moreover, a measurement of "fairn ess" and believability would probably add something to the analysis by determining if th e relationship between excuse availability and immediate affect could be explained in terms of these concepts. The a ffect manipulation would also need work, because a clean manipulation that re ally causes differences in affect is essential to determining if less negative affect truly cause s fewer excuses. The correlational relationships present in the data of this research do little for firm conclusions to that end.
97 Barry Schwartz (2004) posited th at having a huge variety of choices on a regular basis creates a "paradox of choice": people seem to be programmed to seek ou t options, but they are often worse off when they have too many. Th e presented studies add some relevant and important data to research on this topic. Alt hough excuses are not inhere ntly bad (they actually often seem to be effective in a variety of domains), the costs associated with chronic excuse making have not received adequate coverage in the excuse literature (S chlenker et al., 2001). One informative relationship can be found between locus of control and measures of cynicism, poor school outcomes, self-servi ng biases, depression, stress, a nd a host of other negative psychological outcomes. More external locus of c ontrol is associated with negative outcomes for all of these variables, and society's average locu s of control seems to be getting more external over time (Twenge et al., 2004). Thus, more research on the topics in th e presented studies is needed for at least two reasons. First, the basi c processes behind and the limits of the effects need to be more clearly defined. Second, rese arch is needed to suggest how people might overcome constant overexposure to too many opti ons and avoid the obvious problems associated with sliding into the perception that their lives are controlled by influences that are more and more external.
98 APPENDIX A STUDY 1: QUESTIONNAIRE Note: The follow ing questionnaires were presented in a program created in Visual Basic 6.0, so the format was slightly different, although the scale points were exactly the same. Participants could only see one item at a time, and they could not go back. 1. How much PATIENCE did you feel? 1........2....3..4.5 None At All A Little Some A Lot A Great Deal 2. How much EXCITEMENT did you feel? 1........2....3..4.5 None At All A Little Some A Lot A Great Deal 3. How many POSITIVE feelings did you experience? 1........2....3..4.5 None At All A Little Some A Lot A Great Deal 4. How many NEGATIVE feelings did you experience? 1........2....3..4.5 None At All A Little Some A Lot A Great Deal 5. How did you feel overall about your performance on the test? 1...2...3...4...5...6...7...8...9 Extremely Somewhat Neutral Somewhat Extremely Bad Bad Good Good 6. How easy was it for you to complete th e task given the amount of time you had? 1...2...3...4...5...6...7...8...9 Not at all A little Somewhat Very Extremely Easy Easy Easy Easy Easy 7. How clear were the test instructions? 1...2...3...4...5...6...7...8...9 Not at all A little Somewhat Very Extremely Clear Clear Clear Clear Clear
99 8. How well do you understand how your score was calculated? 1...2...3...4...5...6...7...8...9 Not at all Somewhat Extremely Well Well Well 9. How interesting was the test? 1...2...3...4...5...6...7...8...9 Highly Uninteresting Neith er Uninteresting Interesting Highly Uninteresting Nor Interesting Interesting 10. How much personal responsibility do you feel for your performance on the test? 1...2...3...4...5...6...7...8...9 No A Little Some A lot A Great Deal Responsibility of Responsibility 11. How much do you feel that factors beyond your personal control (e .g., background noise, testing conditions) were responsible for your pe rformance on the test? 1...2...3...4...5...6...7...8...9 No A Little Some A lot A Great Deal Responsibility of Responsibility 12. How much do you think your score might impr ove if you took the test again under ideal testing conditions? 1...2...3...4...5...6...7...8...9 None A Little Moderately A lot A Great Deal 13. How difficult was the test? 1...2...3...4...5...6...7...8...9 Extremely Difficult Neither Difficult Easy Extremely Difficult Nor Easy Easy 14. How important was it to you personally to do well on the test? 1...2...3...4...5...6...7...8...9 Not at all A little Somewhat Very Extremely Important Important Important Important Important
100 15. How hard did you try on the test? 1...2...3...4...5...6...7...8...9 Not at all A little Somewhat Very As Hard as Hard I Could ______________________________________________________________________________ ______________________________________________________________________________ Use the following scale for the next set of items: 1...2...3...4...5...6...7...8...9 Hurt my Hurt my Had No Effect Helped my Helped My Performance Performance On My Performance Performance A Lot A Little Performance A Little A Lot 16. User-friendliness of the software 17. Restrictions on seeing attributes for more than one car at a time 18. Number of car options 19. Your general knowledge of cars 20. Your skill with computer software 21. Number of attributes for each car 22. Your memory for details 23. The testing environment (e.g., lighting, background noise, room temperature, etc.) 24. Your abstract reasoning ability 25. Your ability to solve math problems 26. Influence (e.g., clarity, ambiguity ) of the computer instructions 27. Influence (e.g., instructions, app earance) of the research assistant 28. Time limit imposed by the software 29. Your interest in the task ______________________________________________________________________________ ______________________________________________________________________________ Answer the following questions by choosing a point along the continuum that represents how you feel about the decision task at this moment 30. Disappointment 1 2 3 4 5 6 7 Fulfillment 31. Discontent 1 2 3 4 5 6 7 Content 32. Alert 1 2 3 4 5 6 7 Tired 33. Satisfaction 1 2 3 4 5 6 7 Dissatisfaction 34. Happiness 1 2 3 4 5 6 7 Regret 35. Shame 1 2 3 4 5 6 7 Pride 36. Doubt 1 2 3 4 5 6 7 Confidence 37. Pleasure 1 2 3 4 5 6 7 Displeasure 38. Lucky 1 2 3 4 5 6 7 Unlucky 39. Elation 1 2 3 4 5 6 7 Depression 40. Joy 1 2 3 4 5 6 7 Sorrow
101 ______________________________________________________________________________ ______________________________________________________________________________ Please answer the following questions regarding your personal preferences 1.......2... .....7 Completely Strongly Somewhat Neutral Somewhat Strongly Completely Disagree Disagree Disagree Agree Agree Agree 41. When I watch TV, I channel surf, often scanning through the available options even while attempting to watch one program. 42. When I am in the car listening to the radio, I often check other stations to see if something better is playing, even if Im relativel y satisfied with what Im listening to. 43. I treat relationships like clot hing: I expect to try a lot on before I get the perfect fit. 44. No matter how satisfied I am with my job, its only right fo r me to be on the lookout for better opportunities. 45. I often fantasize about living in ways th at are quite different from my actual life. 46. Im a big fan of lists that attempt to rank thin gs (the best movies, the best singers, the best athletes, the best novels, etc.). 47. I often find it difficult to shop for a gift for a friend. 48. When shopping, I have a hard time finding clothing that I really love. 49. Renting videos is really difficult. Im always struggling to pick the best one. 50. I find that writing is very diffi cult, even if its just writing a letter to a friend, because its so hard to word things just right. I often do several drafts of even simple things. 51. No matter what I do, I have the highest standards for myself. 52. I never settle for second best. 53. Whenever Im faced with a choice, I try to imagine what all the other possibilities are, even ones that arent present at the moment. 54. Whenever I make a choice, I'm curious abou t what would have happened if I had chosen differently. 55. Whenever I make a choice, I try to get information about how the other alternatives turned out.
102 56. If I make a choice and it turns out well, I still feel like somethi ng of a failure if I find out that another choice would have turned out better 57. When I think about how I'm doing in life, I often assess opportunities I have passed up. 58. Once I make a decision, I don't look back. ______________________________________________________________________________ ______________________________________________________________________________ Please answer the following questions regarding your perceptions of the test. 59. During the decision task, you were given a fixed time limit. What was your time limit? 1........2....3..4.5 1-2 minutes 3-4 minutes 5-6 minutes 7-8 minutes 9-10 minutes 60. How well did you perform on the test in terms of your percen tile ranking? 1........2....3..4.5 0%-20% 21%-40% 41%-60% 61%-80% 81%-100% 61. How many different cars were there in the decision task? 1........2....3..4.5 1-4 5-8 9-12 13-16 17-20 62. Did you get feedback regarding your level of arousal after the task? 1........2....3..4.5 No, I received Yes, I received no arousal feedback arousal feedback 63. According to the graph you saw of your ow n arousal, how aroused were you on average? 1........2....3..4.5 Not At All A Little Somewhat Very Extremely Aroused Aroused Aroused Aroused Aroused .6 I Did Not Receive Arousal Feedback
103 64. On average, how aroused did yo u feel DURING the decision task? 1........2....3..4.5 Not At All A Little Somewhat Very Extremely Aroused Aroused Aroused Aroused Aroused 65. On average, how aroused did yo u feel AFTER the decision task? 1........2....3..4.5 Not At All A Little Somewhat Very Extremely Aroused Aroused Aroused Aroused Aroused
104 APPENDIX B STUDY 2: FIRST QUESTIONNAIRE 1. How did you feel overall about your perform ance on the test? 1...2...3...4...5...6...7...8...9 Extremely Somewhat Neutral Somewhat Extremely Bad Bad Good Good 2. How easy was it for you to complete th e task given the amount of time you had? 1...2...3...4...5...6...7...8...9 Not at all A little Somewhat Very Extremely Easy Easy Easy Easy Easy 3. How clear were the test instructions? 1...2...3...4...5...6...7...8...9 Not at all A little Somewhat Very Extremely Clear Clear Clear Clear Clear 4. How well do you understand how your score was calculated? 1...2...3...4...5...6...7...8...9 Not at all Somewhat Extremely Well Well Well 5. How interesting was the test? 1...2...3...4...5...6...7...8...9 Highly Uninteresting Neith er Uninteresting Interesting Highly Uninteresting Nor Interesting Interesting 6. How much personal responsibility do you feel for your performance on the test? 1...2...3...4...5...6...7...8...9 No A Little Some A lot A Great Deal Responsibility of Responsibility 7. How much do you feel that factors beyond your personal control (e .g., background noise, testing conditions) were responsible for your pe rformance on the test? 1...2...3...4...5...6...7...8...9 No A Little Some A lot A Great Deal Responsibility of Responsibility
105 8. How much do you think your score might improve if you took the test again under ideal testing conditions? 1...2...3...4...5...6...7...8...9 None A Little Moderately A lot A Great Deal 9. How difficult was the test? 1...2...3...4...5...6...7...8...9 Extremely Difficult Neither Difficult Easy Extremely Difficult Nor Easy Easy 10. How important was it to you personally to do well on the test? 1...2...3...4...5...6...7...8...9 Not at all A little Somewhat Very Extremely Important Important Important Important Important 11. How hard did you try on the test? 1...2...3...4...5...6...7...8...9 Not at all A little Somewhat Very As Hard as Hard I Could ______________________________________________________________________________ ______________________________________________________________________________ Use the following scale for the next set of items: 1...2...3...4...5...6...7...8...9 Hurt my Hurt my Had No Effect Helped my Helped My Performance Performance On My Performance Performance A Lot A Little Performance A Little A Lot 12. User-friendliness of the software 13. Restrictions on seeing attributes for more than one car at a time 14. Number of car options 15. Your general knowledge of cars 16. Your skill with computer software 17. Number of attributes for each car 18. Your memory for details 19. The testing environment (e.g., lighting, background noise, room temperature, etc.) 20. Your abstract reasoning ability 21. Your ability to solve math problems 22. Influence (e.g., clarity, ambiguity ) of the computer instructions 23. Influence (e.g., instructions, app earance) of the research assistant 24. Time limit imposed by the software
106 25. Your interest in the task ______________________________________________________________________________ ______________________________________________________________________________ Answer the following questions by choosing a point along the continuum that represents how you feel about the decision task at this moment 26. Disappointment 1 2 3 4 5 6 7 Fulfillment 27. Discontent 1 2 3 4 5 6 7 Content 28. Alert 1 2 3 4 5 6 7 Tired 29. Satisfaction 1 2 3 4 5 6 7 Dissatisfaction 30. Happiness 1 2 3 4 5 6 7 Regret 31. Shame 1 2 3 4 5 6 7 Pride 32. Doubt 1 2 3 4 5 6 7 Confidence 33. Pleasure 1 2 3 4 5 6 7 Displeasure 34. Lucky 1 2 3 4 5 6 7 Unlucky 35. Elation 1 2 3 4 5 6 7 Depression 36. Joy 1 2 3 4 5 6 7 Sorrow ______________________________________________________________________________ ______________________________________________________________________________ Please answer the following questions regarding your perceptions of the test. 60. How well did you perform on the test in terms of your percen tile ranking? 1........2....3..4.5 0%-20% 21%-40% 41%-60% 61%-80% 81%-100% 61. How many different cars were there in the decision task? 1........2....3..4.5 1-4 5-8 9-12 13-16 17-20
107 APPENDIX C STUDY 2: SECOND QUESTIONNAIRE 1. How did you feel overall about your perform ance on the test? 1...2...3...4...5...6...7...8...9 Extremely Somewhat Neutral Somewhat Extremely Bad Bad Good Good 2. How clear were the test instructions? 1...2...3...4...5...6...7...8...9 Not at all A little Somewhat Very Extremely Clear Clear Clear Clear Clear 3. How well do you understand how your score was calculated? 1...2...3...4...5...6...7...8...9 Not at all Somewhat Extremely Well Well Well 4. How interesting was the test? 1...2...3...4...5...6...7...8...9 Highly Uninteresting Neith er Uninteresting Interesting Highly Uninteresting Nor Interesting Interesting 5. How much personal responsibility do you feel for your performance on the test? 1...2...3...4...5...6...7...8...9 No A Little Some A lot A Great Deal Responsibility of Responsibility 6. How much do you feel that factors beyond your personal control (e .g., background noise, testing conditions) were responsible for your pe rformance on the test? 1...2...3...4...5...6...7...8...9 No A Little Some A lot A Great Deal Responsibility of Responsibility 7. How much do you think your score might improve if you took the test again under ideal testing conditions? 1...2...3...4...5...6...7...8...9 None A Little Moderately A lot A Great Deal
108 8. How difficult was the test? 1...2...3...4...5...6...7...8...9 Extremely Difficult Neither Difficult Easy Extremely Difficult Nor Easy Easy 9. How important was it to you personally to do well on the test? 1...2...3...4...5...6...7...8...9 Not at all A little Somewhat Very Extremely Important Important Important Important Important 10. How hard did you try on the test? 1...2...3...4...5...6...7...8...9 Not at all A little Somewhat Very As Hard as Hard I Could ______________________________________________________________________________ ______________________________________________________________________________ Use the following scale for the next set of items: 1...2...3...4...5...6...7...8...9 Hurt my Hurt my Had No Effect Helped my Helped My Performance Performance On My Performance Performance A Lot A Little Performance A Little A Lot 11. User-friendliness of the software 12. Your skill with computer software 13. Speed of the presentation of each mathematical function 14. Your memory for details 15. The testing environment (e.g., lighting, background noise, room temperature, etc.) 16. Your abstract reasoning ability 17. Your ability to solve math problems 18. Influence (e.g., clarity, ambiguity ) of the computer instructions 19. Influence (e.g., instructions, app earance) of the research assistant 20. Your interest in the task ______________________________________________________________________________ ______________________________________________________________________________ Answer the following questions by choosing a point along the continuum that represents how you feel about the decision task at this moment 21. Disappointment 1 2 3 4 5 6 7 Fulfillment 22. Discontent 1 2 3 4 5 6 7 Content
109 23. Alert 1 2 3 4 5 6 7 Tired 24. Satisfaction 1 2 3 4 5 6 7 Dissatisfaction 25. Happiness 1 2 3 4 5 6 7 Regret 26. Shame 1 2 3 4 5 6 7 Pride 27. Doubt 1 2 3 4 5 6 7 Confidence 28. Pleasure 1 2 3 4 5 6 7 Displeasure 29. Lucky 1 2 3 4 5 6 7 Unlucky 30. Elation 1 2 3 4 5 6 7 Depression 31. Joy 1 2 3 4 5 6 7 Sorrow ______________________________________________________________________________ ______________________________________________________________________________ 32. How well did you perform on the test in terms of your percen tile ranking? 1........2....3..4.5 0%-20% 21%-40% 41%-60% 61%-80% 81%-100% Please answer the following questions regarding your personal preferences 1.......2... .....7 Completely Strongly Somewhat Neutral Somewhat Strongly Completely Disagree Disagree Disagree Agree Agree Agree 33. When I watch TV, I channel surf, often scanning through the available options even while attempting to watch one program. 34. When I am in the car listening to the radio, I often check other stations to see if something better is playing, even if Im relativel y satisfied with what Im listening to. 35. I treat relationships like clot hing: I expect to try a lot on before I get the perfect fit. 36. No matter how satisfied I am with my job, its only right fo r me to be on the lookout for better opportunities. 37. I often fantasize about living in ways th at are quite different from my actual life. 38. Im a big fan of lists that attempt to rank thin gs (the best movies, the best singers, the best athletes, the best novels, etc.). 39. I often find it difficult to shop for a gift for a friend. 40. When shopping, I have a hard time finding clothing that I really love. 41. Renting videos is really difficult. Im always struggling to pick the best one.
110 42. I find that writing is very diffi cult, even if its just writing a letter to a friend, because its so hard to word things just right. I often do several drafts of even simple things. 43. No matter what I do, I have the highest standards for myself. 44. I never settle for second best. 45. Whenever Im faced with a choice, I try to imagine what all the other possibilities are, even ones that arent present at the moment. 46. Whenever I make a choice, I'm curious abou t what would have happened if I had chosen differently. 47. Whenever I make a choice, I try to get information about how the other alternatives turned out. 48. If I make a choice and it turns out well, I still feel like somethi ng of a failure if I find out that another choice would have turned out better 49. When I think about how I'm doing in life, I often assess opportunities I have passed up. 50. Once I make a decision, I don't look back. ______________________________________________________________________________ ______________________________________________________________________________ 51. During the first task, you were given a fixed time limit. What was your time limit? 1........2....3..4.5 1-2 minutes 3-4 minutes 5-6 minutes 7-8 minutes 9-10 minutes 52. Given your time limit, how difficult should th e first task have been according to the instructions you were given before the task? 1........2....3..4.5 Not At A Little Somewhat Very Extremely All Difficult Difficult Difficult Difficult Difficult
111 LIST OF REFERENCES Abelson, R. P. (1995). Statistics as principled argument Philadelph ia, PA: Lawrence Erlbaum. Ackerman, D., & Gross, B. (2006). How many choi ces are good? Measurement of the effects of course choice on percepti ons of a marketing option. Journal of Marketing Education 28 69-80. Anderson, C. A., Carnagey, N. L., & Eubanks, J. (2003). Exposure to violent media: The effects of songs with violent lyrics on a ggressive thoughts and feelings. Journal of Personality and Social Psychology, 84 960-971. Batson, C. D., Engell, C. L., & Fridell, S. R. (1999). Value judgments: Testing the somaticmarker hypothesis using fals e physiological feedback. Personality and Social Psychology Bulletin, 25 1021-1032. Batson, C. D., Turk, C. L., Shaw, L. L., & Klei n, T. R. (1995). Information function of empathic emotion: Learning that we value the others' welfare. Journal of Personality and Social Psychology, 68 300-313. Beattie, J., Baron, J., Hershey, J. C., & Spranca, M. D. (1994). Psychological determinants of decision attitude. Journal of Behaviora l Decision Making 7, 129. Bettman, J. R. (1979). An information processing theory of consumer choice Reading, MA: Addison-Wesley. Boatwright, P. & Nunes, J. C. (2001). Reducing assortment: An attribute-based approach. Journal of Marketing, 65 50-63. Botti, S. & Iyengar, S. (2004). The psychological pleasure and pain of choosing: When people prefer choosing at the cost of subsequent outcome satisfaction. Journal of Personality and Social Psychology 87 312-326. Botti, S. & McGill, A. L. (2006). When choosing is not deciding: The effect of perceived responsibility on satisfaction. Journal of Consumer Research, 33 211-219. Bown, N., Read, D., & Summers, B. (2003). The lure of choice. Journal of Behavioral Decision Making 16, 297-308. Breggin, P. R., (1991). Toxic Psychiatry New York: St. Martins Press. Broniarczyk, S. M., Hoyer, W. D., & McAlister, L., (1998). Consumers' perceptions of the assortment offered in a grocery categ ory: The impact of item reduction. Journal of Marketing Research, 35, 166-176.
112 Brown, R. P., & Josephs, R. A. (1999). A burden of proof: Stereotype relevance and gender differences in math performance. Journal of Personality & Social Psychology, 76 246257. Burger, J. M. (1989). Negative reactions to increases in perceived personal control. Journal of Personality and Social Psychology, 56 246-256. Bushman, B. J., Baumeister, R. F ., & Stack, A. D. (1999). Catharsi s, aggression, and persuasive influence: Self-fulfilling or self-defeating prophecies? Journal of Personality and Social Psychology, 76, 367-376. Caplow, T., Hicks, L. & Wattenberg, B. J. (2001). The first measured century. Washington, D.C.: AEI Press. Chernev, A. (2003a). Product assortment and individual decision processes. Journal of Personality and Social Psychology, 85 151-62. Chernev, A. (2003b). When more is less and less is more: The role of ideal point availability and assortment in consumer choice. Journal of Consumer Research, 30 170-83. Chernev, A. (2006). Decision focus and consumer choice among assortments. Journal of Consumer Research, 33 50-60. Connolly, T. & Butler, D. (2006). Regret in economic and psychological th eories of choice. Journal of Behavioral Decision Making, 19 139-154. Connolly, T., Ordez, L. D., & Coughlin, R. (1997) Regret and responsib ility in the evaluation of decision outcomes. Organizational Behavior and Huma n Decision Processes, 70 7385. Cordova, D. I., & Lepper, M. R. (1996). Intrinsic motivation and the process of learning: Beneficial effects of contextua lization, personalization, and choice. Journal of Educational Psychology, 88 715. Crocker, J., & Major, B. (1989). Social stigma a nd self-esteem: The self-p rotective properties of stigma. Psychological Review, 96 608-630. Deci, E. L. (1981). The psychology of self-determination Lexington, MA: Heath. Deci, E. L., & Ryan, R. M. (1985). Intrinsic motivation and self-determination in human behavior. New York: Plenum Press. Deci, E. L., Spiegel, N. H., Ryan, R. M., Ko estner, R., & Kaufman, M. (1982). The effects of performance standards on teaching styles: The behavior of controlling teachers. Journal of Educational Psychology, 74 852-859.
113 Diehl, K. (2005). When two right s make a wrong: Searching too much in ordered environments. Journal of Marketing Research, 42 313-322. Fenton-O'Creevy, M., Nicholson, N., Soane E., & Willman, P. (2003). Trading on illusions: Unrealistic perceptions of cont rol and trading performance. Journal of Occupational and Organizational Psychology, 76, 53-68. Follette, V. M. & Jacobson, N. S. (1987). Impor tance of attributions as a predictor of how people cope with failure. Journal of Personality and Social Psychology, 6 1205-1211. Garner, W. R. (1962). Uncertainty and structure as psychological concepts New York: Wiley. Gilovich, T. & Medvec, V. H. The tempor al pattern to the experience of regret. Journal of Personality and Social Psychology, 67 357-365. Haynes, G. & Olson, J. (2007). You Choose, You Lose? The Effect of Too Many Attractive Alternatives on Choice Diffic ulty, Satisfaction, and Regret Poster presented at the annual meeting of the Soci ety for Personality and Social Psychology, Memphis, Tennessee. Iyengar, S. S., and Lepper, M. R. (2000). When choice is demotivating: Can one desire too much of a good thing? Journal of Personality and Social Psychology, 79 995-1006. Jacoby, J., Speller, D. E., & Berning, C. K. (1974). Brand choice behavior as a function of information load: Replication and extension. Journal of Consumer Research, 1 33-42. Jacoby, J., Speller, D. E., & Kohn, C. A. (1974). Brand choice behavior as a function of information load, Journal of Marketing Research, 11 63-69. Jimmieson, N. L. (2000). Employee reactions to behavioural contro l under conditions of stress: The moderating role of self-efficacy. Work and Stress, 14, 262-280. Johnson, R. M. & Schlenker, B. R. (2008). Excuses and productivity: Th e consequences of a good excuse. Manuscript submitted for publication, University of Florida, Gainesville. Kahn, B. E., & Lehmann, D. R. (1991). Modeling choice among assortments. Journal of Retailing, 67 274-299. Kahn, B. E., & Wansink, B. (2004). The influence of assortment structur e on perceived variety and consumption quantities. Journal of Consumer Research, 30 519-533. Keller, K. L. & Staelin, R. ( 1987). Effects of quality and quan tity of information on decision effectiveness. Journal of Consumer Research, 14 200-213. Kline, R. B. (2005). Principles and Practice of St ructural Equation Modeling (2nd ed.). New York: The Guilford Press.
114 Langer, E. J. (1975). The illusion of control. Journal of Personality and Social Psychology 32, 311. Langer, E. J., & Rodin, J. (1976). The effects of choice and enhanced persona l responsibility for the aged: A field experiment in an institutional setting. Journal of Personality and Social Psychology, 34, 191. Luce, R. D. (2003). Whatever happened to information theory in psychology? Review of General Psychology, 7 183-188. Lurie, N. H. (2004). Decision making in info rmation-rich environmen ts: The role of information structure. Journal of Consumer Research, 30 473-486. Lussier, D. & Olshavsky, R. W. (1979). Task co mplexity and contingent processing in brand choice. Journal of Consumer Research, 6 154-165. Major, B., Kaiser, C. R., & McCoy, S. K. (2003). It 's not my fault: When and why attributions to prejudice protect self-esteem. Personality & Social Psychology Bulletin, 29 772-781. Malhotra, N. K. (1982). Information load and consumer decision making. Journal of Consumer Research, 8, 419-430. Margolick, D. (1991, August 13). A ddress by Quayle on justice proposa ls irks bar association. The New York Times p. A14. Markman, K. D. & Tetlock, P. E. (2000). "I couldn't have known": Accountability, foreseeability and counterfactua l denials of responsibility. British Journal of Social Psychology, 39, 313-325. McFarland, C., & Ross, M. (1982). Impact of causa l attributions on affec tive reactions to success and failure. Journal of Personality & Social Psychology, 43 937-946. Mehlman, R. C., & Snyder, C. R. (1985). Excuse theory: A test of the self-protective role of attributions. Journal of Personality & Social Psychology, 49 994-1001. Olshavsky, R. W. (1979). Task complexity an d contingent processing in decision making: A replication and extension. Organizational Behavior and Human Performance, 24 300316. Ono, K. (2000). Free-choice preference under uncertainty. Behavioural Processes, 49 11. Ordez, L. D. & Connolly, T. (2000). Regret an d responsibility: A reply to Zeelenberg et al. Organizational Behavior and Huma n Decision Processes, 81 132-142.
115 Ouwerkerk, J. W. & Ellemers, N. (2002). The benefits of being disadvantaged: Performancerelated circumstances and conseque nces of intergroup comparisons. European Journal of Social Psychology, 32 73-91. Parkinson, B. & Manstead, A. S. R. (1981). An examination of the roles played by meaning of feedback and attention to feedb ack in the "Valin's Effect." Journal of Personality and Social Psychology, 40 239-245. Perlmuter, L. C., & Monty, R. A. (1977). The importance of perceived contro l: Fact or fantasy? American Scientist, 65 759. Peterson, C., Luborsky, L., & Seligman, M. E. (1983) Attributions and depressive mood shifts: A case study using the symptom-context model. Journal of Abnormal Psychology, 92 96-103. Peterson, C., & Seligman, M. E. (1984). Causal expl anations as a risk f actor for depression: Theory and evidence. Psychological Review, 91 347-374. Pontari, B. A., Schlenker, B. R., & Christopher, A. N. (2002). Excuses and character: Identifying the problematic aspects of excuses. Journal of Social & Clinical Psychology, 21 497516. Raps, C. C. (1982). Attributiona l style among depressed patients. Journal of Abnormal Psychology, 91, 102-108. Rezza v. United States Department of Justice, No. 87-6732, 1988 U.S. Dist. LEXIS 11472, (E.D. Pa. October 17, 1988). Rotter, J. B. (1966). Generalized expectancies for internal versus external control of reinforcement. Psychological Monographs : General & Applied, 80 1-28. Rowley, I. & Hall, K. (April 3, 2006). Once liti gation-averse, the country is rushing to fill a shortage of attorneys, Business Week, 3978, 46. Savodnik, I. (2006, January 1). How we live: Psychi atry's sick compulsion: turning weaknesses into diseases. The Los Angeles Times p. 1. Schlenker, B. R. (1980). Impression management: The self-c oncept, social identity, and interpersonal relations Monterey, CA: Brooks-Cole. Schlenker, B. R. (1985). Identity and self-identification. In B. R. Schlenker (Ed.), The self and social life (pp. 65-99). New York: McGraw-Hill. Schlenker, B. R. (2003). Se lf-presentation. In M. R. Leary and J. P. Tangney (Eds.), Handbook of self and identity (pp. 492-518). New York: Guilford.
116 Schlenker, B. R., Dlugolecki, D. W., & Doherty, K. J. (1994). The impact of self-presentations on self-appraisals and behaviors: The power of public commitment. Personality and Social Psychology Bulletin, 20 20-33. Schlenker, B. R. & Miller, R. S. (1977). Egocentr ism in groups: Self-ser ving biases or logical information processing? Journal of Personality and Social Psychology, 35 755-764. Schlenker, B. R., Pontari, B. A., & Christopher, A. N. (2001). Excuses and character: Personal and social implications of excuses. Personality & Social Psychology Review 5, 15-32. Schulz, R. (1976). Effects of control and predic tability on the physical and psychological wellbeing of the institutionalized aged. Journal of Personality and Social Psychology, 33 563. Schwartz, B. (2004). The paradox of choice : Why less is more New York: HarperCollins. Schwartz, B., Ward, A., Monterosso, J., Lyubomirsky, S., White, K., & Lehman, D. R. (2002). Maximizing versus satisficing: Ha ppiness is a matter of choice. Journal of Personality and Social Psychology, 83 1178-1197. Seligman, M. E. P. (1975). Helplessness: On depression, development and death San Francisco: Freeman. Shannon, C. E. (1949). The mathematical theory of communication Urbana: University of Illinois Press. Sherif, M. (1936). The psychology of social norms Oxford, England: Harper. Simonson, I. (1992). The influence of anticipating regret and responsibility on purchase decisions. Journal of Consumer Research, 19, 105-118. Snyder, C. R., & Higgins, R. L. (1988). Excuses: Th eir effective role in th e negotiation of reality. Psychological Bulletin, 104 23-35. Spencer, S., Zanna, M., & Fong, G. (2005). Estab lishing a causal chain: Why experiments are often more effective than mediational anal yses in examining psychological processes. Journal of Personality and Social Psychology 89 845-851. Streufert, S., Driver, M. J., & Haun, K. H. ( 1967). Components of response rate in complex decision making. Journal of Experimental Social Psychology, 3 286-295. Suzuki, S. (1999). Selection of forcedand free-choice by monkeys ( Macaca fascicularis ). Perceptual and Motor Skills, 88 242. Sykes, C. J. (1992). A nation of victims New York: St. Martins Press.
117 Sykes, C. J. (1996). Dumbing down our kids New York: St. Martins Press. Taylor, S. E., & Brown, J. D. (1988). Illusion and well-being: A social ps ychological perspective on mental health. Psychological Bulletin, 103 193. Tice, D. M. (1992). Self-concept change and self-pre sentation: The looking glass self is also a magnifying glass. Journal of Personality and Social Psychology, 63 435-451. Tversky, A. & Shafir, E. (1992). Choice under co nflict: The dynamics of deferred decision. Psychological Science, 3 358-361. Twenge, J., Zhang, L., & Im, C. (2004). It's beyond my control: A cross-temporal meta-analysis of increasing externality in locus of control, 1960-2002. Personality and Social Psychology Review 8, 308-319. Tykocinski, O. E. (2001). I never had a chan ce: Using hindsight tactics to mitigate disappointments. Personality and Social Psychology Bulletin, 27 376-382. Tykocinski, O. E. & Steinberg, N. (2005). Copi ng with disappointing ou tcomes: Retroactive pessimism and motivated inhi bition of counterfactuals. Journal of Experimental Social Psychology, 41, 551-558. United States Census Bureau (n.d.). Statistical abstract of the United States Retrieved April 21, 2006, from http://www.census.gov/prod/2004pubs/04statab/educ.pdf Valins, S. (1 966). Cognitive effects of false heart rate feedback. Journal of Personality and Social Psychology, 4 400-408. van Dijk, W. W., van der Pligt, J., & Zeelenberg, M. (1999). Effort invested in vain: The impact of effort on the intensity of disappointment and regret. Motivation and Emotion, 23 203220. Voss, S. C., & Homzie, M. J. (1970). Choice as a value. Psychological Reports, 26 912. Wilkins, W. (1971). Perceptual di stortion to account for arousal. Journal of Abnormal Psychology, 78, 252-257. Wright, P. (1975). Consumer choice st rategies: Simplifying vs. optimizing. Journal of Marketing Research, 12, 60-67. Zeelenberg, M., van Dijk W. W., & Manstead, A. S. R. ( 1998). Reconsidering the relation between regret and responsibility. Organizational Behavior and Human Decision Processes, 74, 254-272.
118 Zuckerman, M., Porac, J., Latin, D., Smith, R., & Deci, E. L. (1978). On the importance of selfdetermination for intrinsically motivated behavior. Personality and Social Psychology Bulletin, 4, 443.
119 BIOGRAPHICAL SKETCH Ryan Johnson was born in Sacram ento, California in 1980. He moved to Gainesville, Florida at age 6 and remained through graduate school. He received a B.S. in psychology in 2003, a M.S. in social psychology in 2005, and he pl ans to receive a Ph.D. in social psychology in May 2008, all from the University of Florida.