A DYNAMIC MULTILEVEL MODEL OF TASK MOTIVATION LINKING PERSONALITY, AFFECTIVE REACTIONS TO FEEDBACK AND SELFREGULATION By REMUS ILIES A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2003
Copyright 2003 by Remus Ilies
iii ACKNOWLEDGEMENTS This dissertation was supported in part by the Meredith Crawford Fellowship in Industrial and Organizational Psychology, awarded by the Human Resources Research Organization (HumRRO). I would like to express special thanks to my dissertation chair, Tim Judge. During my doctoral education, Tim has greatly influenced my development as a researcher and a teacher, and has inspired and guided the research presented in this dissertation. I would also like to thank my dissertation committee members, James Algina, Amir Erez, and Henry Tosi for their advice and support. Finally, I wish to acknowledge the support that I have received from my wife, Doina Ilies. The emotional support and encouragement received from her have helped me immensely over the course of my graduate studies.
iv TABLE OF CONTENTS page ACKNOWLEDGEMENTS...............................................................................................iii LIST OF FIGURES...........................................................................................................vi LIST OF TABLES............................................................................................................vii ABSTRACT.....................................................................................................................vi ii CHAPTER 1 INTRODUCTION...........................................................................................................1 Motivation and Emotion.................................................................................................2 A Dynamic Model of Task Motivation...........................................................................4 2 LITERATURE REVIEW..............................................................................................10 Integrative Goal Setting Theory...................................................................................10 Feedback, Performance, and Goals...............................................................................13 Goal Regulation: Mediating Mechanisms................................................................16 Moderating Influences..............................................................................................21 Dispositional moderators......................................................................................22 Situational influences............................................................................................29 3 STUDY 1..................................................................................................................... ..32 Method......................................................................................................................... .32 Participants................................................................................................................32 Experimental Design and Procedure.........................................................................33 Performance Task.....................................................................................................34 Measures...................................................................................................................34 Analyses....................................................................................................................35 Results........................................................................................................................ ...43 Discussion..................................................................................................................... 48 4 STUDY 2..................................................................................................................... ..54 Method......................................................................................................................... .55 Participants................................................................................................................55 Experimental Design and Procedure.........................................................................55
v Performance Task.....................................................................................................56 Measures...................................................................................................................56 Analyses....................................................................................................................58 Results........................................................................................................................ ...58 Discussion..................................................................................................................... 64 5 STUDY 3..................................................................................................................... ..66 Method......................................................................................................................... .67 Participants................................................................................................................67 Procedure..................................................................................................................67 Measures...................................................................................................................68 Analyses....................................................................................................................71 Results........................................................................................................................ ...73 Mediating Mechanisms.............................................................................................75 Moderating Effects....................................................................................................81 Dispositional moderators......................................................................................82 Situational moderators..........................................................................................84 Discussion..................................................................................................................... 87 6 GENERAL DISCUSSION............................................................................................89 Findings....................................................................................................................... .89 Implications..................................................................................................................9 2 Theory Development................................................................................................92 Implications for Practice...........................................................................................95 Limitations.................................................................................................................... 96 Contribution..................................................................................................................9 7 APPENDIX A INTERNET SCREENS FOR THE MULTI-TRIAL EXPERIMENT........................101 B REMOTE ASSOCIATES TEST.................................................................................105 C THE POSITIVE AND NEGATIVE AFFECT SCHEDULE (PANAS).....................106 D GOAL COMMITMENT MEASURE.........................................................................107 E ACADEMIC SELF-EFFICACY MEASURE.............................................................108 LIST OF REFERENCES.................................................................................................109 BIOGRAPHICAL SKETCH...........................................................................................119
vi LIST OF FIGURES Figure page 1-1. Conceptual representation of a dynamic model of task motivation.............................7 2-1. Goal-setting effects on performance...........................................................................10 2-2. Conceptual model of the idiographic feedback-goal mediation process....................17 2-3. Moderating effects on the feedback-goal relationship...............................................22 3-1. Illustration of a mediating effect................................................................................37 3-2. Partitioning the total variance in goals for Sample 1..................................................53 5-1. Path Model 1.............................................................................................................. .72 5-2. Path Model 2.............................................................................................................. .73 5-3. Path Model 3.............................................................................................................. .80 5-4. Path Model 4.............................................................................................................. .80
vii LIST OF TABLES Table page 3-1. Means, standard deviations, and intercorrelations for dispositional variables...........43 3-2. Parameter estimates and variance components for the null model.............................44 3-3. Within-individual standard deviations........................................................................44 3-4. Parameter estimates and variance components for Model 1 and Model 2.................46 3-5. Parameter estimates for Model 3 and Model 4...........................................................49 4-1. Means, standard deviations, and intercorrelations for dispositional variables...........59 4-2. Parameter estimates and variance components for the null model.............................60 4-3. Within-individual standard deviations........................................................................61 4-4. Parameter estimates and variance components for Model 1 and Model 2.................62 4-5. Parameter estimates for Model 3 through Model 7....................................................63 5-1. Means, standard deviations, and intercorrelations for the study variables.................74 5-2. Fit indices for Model 1...............................................................................................75 5-3. Standardized parameter estimates for Model 1..........................................................76 5-4. Standardized parameter estimates for Model 2..........................................................77 5-5. Standardized parameter estimates for Model 4..........................................................82 5-6. Dispositional moderator analyses results for neuroticism..........................................83 5-7. Dispositional moderator analyses results for BIS.......................................................84 5-8. Moderator analyses results for goal commitment.......................................................85 5-9. Moderator analyses results for academic self-efficacy...............................................86
viii Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy A DYNAMIC MULTILEVEL MODEL OF TASK MOTIVATION LINKING PERSONALITY, AFFECTIVE REACTIONS TO FEEDBACK AND SELFREGULATION By REMUS ILIES August 2003 Chair: Timothy A. Judge Department: Management Though much has been written on the effects of performance feedback on future motivation and performance, the results reported in the literature are inconsistent and often contradictory. Negative feedback (feedback indicating failure in accomplishing a goal), for example, has been found to increase, decrease, or have no effect on subsequent motivation and performance. It is argued here that the existing literature lacks a theory of dynamic self-regulation explaining how feedback and goals are interrelated across time, and what factors influence the nature of the self-regulatory processes. I develop such a theory here and present a dynamic model of task motivation that was designed to explain how individuals regulate their task behaviors and goals across time, in the presence of performance feedback. New in this model is the proposition that an affective process mediates the relationship between feedback and subsequent goals, in addition to the cognitive link through self-efficacy which has been previously identified in the literature. In addition, I identify a number of dispositional factors (such as neuroticism and
ix extraversion, affectivity, chronic self-regulatory focus, and goal orientation) that should moderate the strength of the goal regulation process. Three studies comprising seven independent samples totaling almost 1,400 participants and involving personality assessments, dynamic multi-trial experiments, and longitudinal data collection over the course of a semester were conducted to seek support for the hypothesized relationships. The results provided support for the hypothesized within-individual effect of feedback concerning previous performance on subsequent goal setting and for the mediating role of affect and emotions in explaining this relationship across time. The data collected for the third study supported the moderating role of individualsÂ’ chronic activation of their behavioral inhibition system on the relationship between negative goal-discrepancy feedback and negative emotional reactions to the feedback. The moderating role of goal commitment and academic self-efficacy was not supported by the data. Implications for feedback, goal-setting and self-regulation theories are discussed.
1 CHAPTER 1 INTRODUCTION What makes people act in certain ways, why do they choose to expend effort on certain activities over others, why are some individuals more successful in performing specific tasks than are others? Why do some individuals decide to try harder while others give up after feedback signaling failure (or success)? These are questions that have long preoccupied psychology scholars, and any attempt to answer such questions should include a discussion of the concept of human motivation. The Latin root for the word Â“motivationÂ” is Â“movereÂ” and means Â“to move.Â” Thus, in the most basic sense, the study of motivation is the study of what makes people move. Motivational theories attempt to explain why people behave the way they do, and the quest for such explanations has been at the center of basic psychology as a field of scientific inquiry. The study of motivation has also been central to applied psychology and organizational behavior. Knowing what makes individuals work harder on a task, explaining the differences among individuals in the amount of effort expended on specific activities, and understanding the motivational consequences of feedback have important implications for matching people with jobs, designing jobs and incentive plans, and deciding how to implement performance feedback systems, to name only a few of the potential applications of motivation theories in organizations. This dissertation is concerned with the study of task motivation. In an attempt to contribute to the understanding of the psychological processes that Â“moveÂ” people, I will propose a process model of task motivation that can be used to understand how people
2 regulate their actions over time. This model is most relevant to work motivation because it includes concepts and features that are consistently involved during task performance at work. Motivation and Emotion Rationalist theories of human behavior, which date back to Plato and Aristotle, assumed that the form of human existence that differentiates humans from other species Â– rationality Â– is the driving force behind the supremacy of the human race, and primordial human features such as emotions disrupt human rationality and thus have undesirable influences on human behavior. Following this rationalist assumption, basic psychology has long emphasized rationality and cognition, whereas the study of emotion has been focused on minimizing the effect of emotion on behavior. In the organizational literature, the classic rationalistic assumption and the focus on rationality and cognition from basic psychology have been reflected in the adoption of a cognitive paradigm for the study of behavior in organizations. That is, organizational scholars formulated cognitive models of traditional organizational topics such as leadership and motivation. Within this cognitive paradigm, when they were studied, emotions were either viewed as the outcome of a cognitive evaluation process (Muchinsky, 2000), or as phenomena that should be prevented by institutionalizing norms of rationality (Ashforth & Humphrey, 1995). As a result of the strong emphasis on rationality in organizational life, Â“phenomena associated with process, noninstrumentality, qualitativeness, spontaneity, or subjectivity are often viewed pejoratively Â– as Â‘irrationalÂ’ or Â‘arationalÂ’ Â– and therefore to be avoided or controlledÂ” (Ashford & Humphrey, 1995, p. 102).
3 In contrast to the cognitive paradigm, in the past two decades, there has been a growing interest in the study of emotions, affect and temperament in basic psychology (e.g., Watson, 2000). More recently, this interest in emotions and affect has permeated organizational research, and the organizational literature has seen an increased interest in the experience, expression, and management of affect and emotions at work (Ashkanasy, Haertel, & Zerbe, 2000; Fisher & Ashkanasy, 2000; Fox & Spector, 2002; Lord, Klimoski, & Kanfer, 2002; Weiss, 2001). Weiss (2001), in the introduction to a special issue on affect in the workplace, notes that Â“there has been an explosion of research on the topic over the past decadeÂ” (p. 1). However, work motivation has not been part of this Â“explosion,Â” most likely because of the cognitive focus of work theories of motivation. Developed within the rationalist perspective, traditional theories of motivation consist of cognitive explanations of action. In the organizational literature, most prominent theories of motivation Â– goal setting theory (Locke & Latham, 1990), social cognitive theory (Bandura, 1986), resource allocation theory (Kanfer & Ackerman, 1989), and expectancy theory (Vroom, 1964) Â– all have a strong cognitive focus. Such cognitive focus made the integration of affect into motivation theory difficult, which explains the relative paucity of theory and research examining the influence of affective constructs on motivational processes. Even though sporadic attempts to study the influence of emotions or affect on work or task motivation have been made (e.g., Saavedra & Earley, 1991; Venkatesh & Speier, 1999), there is no conceptual model of work motivation that includes affective or emotional constructs other than job satisfaction (e.g., Locke, 1997). Furthermore, most empirical examinations of antecedents or consequences of task or work motivation, even
4 when they are based on longitudinal designs, have investigated motivational differences between individuals and have either inferred that similar processes operate within individuals, or have assumed that there is no systematic within-individual variance in motivation and thus focused on explaining individual differences in motivational constructs or outcomes. That is, the organizational literature is mute with respect to empirical investigations of within-individual motivational processes that would explain how individuals regulate their behaviors across time. Though there has been empirical research on motivation that examined motivational constructs across time, such as Kanfer and AckermanÂ’s (1989) skill acquisition studies, or Phillips, Hollenbeck, and IlgenÂ’s (1996) and Vance and ColellaÂ’s (1990) studies of goal-regulation, the typical investigation of motivational effects across time involved modeling cross-sectional relationships and did not attempt to model within-individual relationships between motivational constructs. A Dynamic Model of Task Motivation In this dissertation, I present a process model of task motivation that integrates a conceptual model of Â‘cognitiveÂ’ work motivation with emotion and affect theory, attempting to explain how individuals regulate their behaviors during task performance, and the role of emotions and affect in interpreting performance feedback and influencing subsequent behavior. I use LockeÂ’s (1997, p. 375) model of work motivation that Â“integrates several decades of empirical research on work motivation and selected elements of eight work motivation theoriesÂ” as a starting point and argue that in order to explain motivation across time (understand how individuals regulate their behavior across
5 time), affective processes should be a included in a dynamic model of task motivation. To build this argument, I use basic motivation and affect theory. At the most basic psychological level, organisms are motivated to seek rewards (appetitive motivation) and avoid threats (aversive motivation), in order to survive. Gray (1978, 1979, 1981, 1990) has developed a behavioral theory of motivation from the fundamental premises that (a) the appetitive and aversive motivational systems can be distinguished from one another and (b) these two systems have opposing influences on the probability of a specific behavioral response (Fowles, 1987). Behavioral motivation theory postulates that two distinct brain mechanisms control appetitive and aversive motivation (Gray, 1981), and individual differences in the sensitivity, or chronic activation, of these systems reflect major personality dimensions. Gray (1981) proposed that individual differences in the functioning of these motivational systems reflect neurological differences and represent the two major personality dimensions of anxiety and impulsivity. Besides linking personality to motivation, GrayÂ’s theory is important because it provides an explanation of individual differences in brain functioning based on personality traits (Matthews & Gilliland, 1999). The neurobehavioral system regulating approach behaviors is called the Behavioral Activation System (BAS; Gray, 1981), or the Behavioral Approach System (BAS; Fowles, 1987), or the Behavioral Facilitation System (Depue & Iacono, 1989; Watson, 2000), and is activated by stimuli signaling reward (or relief from punishment) (Gray, 1981, 1990). The system regulating avoidance behaviors is called the Behavioral Inhibition System (BIS) and is activated by stimuli signaling punishment (or frustrative nonreward) (Gray, 1981, 1990).
6 In sum, the BAS regulates approach motivation and the BIS regulates avoidance motivation, and, according to Gray (1981), individual differences in the sensitivities of these systems are reflected in the personality dimensions of impulsivity and anxiety, respectively. In addition to the behavioral tendencies which they regulate and the personality dimensions which they reflect, these two broad motivational systems contain both emotional and cognitive components (e.g., Fowles, 1987; Watson, 2000). Emotions play a central role in explaining how the behavioral motivation systems work. The BAS is believed to regulate the experience of positive emotions and moods, while the BIS regulates negative emotions and moods (Gray, 1990). Stimuli from the environment influence peopleÂ’s affective states, and the resulting affective states will reinforce behavioral motivation. For example, appetitive stimuli activate approach behaviors leading to rewards, which induce positive affect. The experience of positive affect will reinforce the approach response to such appetitive stimuli. Thus, favorable cues lead to positive affect which is associated with BAS activation, and individuals tend to engage in approach behaviors when they experience positive emotions or moods. Conversely, when individuals experience negative emotions that signal an unfavorable situation, these negative emotions will reinforce avoidance behaviors because negative emotions activate the BIS. Applying the previous argument to explain task motivation, it follows that affect should have a central role in mediating the relationship between performance feedback and subsequent motivation to work on the task. The role of cognitive mediators notwithstanding (e.g., performance feedback influences subsequent motivation through self-efficacy judgments), in my view, emotions and affect are crucial in understanding
7 motivated behavior over time. Following Locke (1997), I conceptualize motivation as the regulation of behavior through personal goals, self-efficacy judgments, and task strategies. In addition to these classic components of self-regulatory motivation, I include affective constructs Â– emotions and mood Â– in the model. In its simplest conceptual form, this dynamic model is presented in Figure 1-1. Figure 1-1. Conceptual representation of a dynamic model of task motivation Explaining how the self-regulatory processes depicted in Figure 1-1 work across time (or across multiple trials) is at the center of the dynamic task motivation theory presented in this dissertation. The model is designed to explain how individuals regulate their task behaviors across time: task behavior leads to performance feedback which influences subsequent behavior through what I call idiographic (within-individual) selfregulatory processes that include personal goals, self-efficacy, task strategies, and emotions and mood.1 Before explaining the relationships among the constructs from the model, I must acknowledge that the model can be further developed. For example, in its current form, the model does not include variables such as the source or credibility of the performance feedback, which may moderate the influence of feedback on other variables (e.g., Podsakoff & Farh, 1989). 1 This is a within-individual model; the self-regulatory approach attempting to explain motivation and behavior across time gives this model its dynamic nature. Performance Feedback Self-Regulatory Processes Personal Goals Task Self-Efficacy Task Strategies Emotions and Mood Subsequent Task Behavior Task Behavior
8 Brief and Hollenbeck (1985), in their article investigating self-regulatory processes and performance, contend that Â“an approach to employee motivation which places person characteristics in a more dynamic context can be formulated; and, such an alternative approach may enrich current levels of understandingÂ” (Brief & Hollenbeck, 1985, p. 198). Specifically with respect to performance feedback, Ilgen and Davis (2000, pp. 553-554) note: Â“Encoding performance does not occur in a vacuum; perceptions are framed in contexts and result from a dynamic interaction between characteristics of the performance context and individual dispositions.Â” Accordingly, I will investigate the influence of person characteristics (dispositions) on the within-individual process of selfregulation. That is, I will investigate whether the self-regulation process works differently for different individuals, according to their dispositional characteristics. In addition, I investigate moderating situational effects. Thus, there are three distinct, yet interrelated classes of mechanisms that I formally investigate in this dissertation: within-individual (ideographical) self-regulatory processes, dispositional moderating effects, and situational influences.2 In the following chapter, I review the goal-setting theory of motivation and the literature relevant to each of the three classes of mechanisms investigated in this project (within-individual processes, dispositional moderating effects, and situational influences), I summarize previous research suggesting support for various links proposed in the dynamic task motivation model developed here, and then I formally state my expectations with respect to how motivational processes work over time (how individuals regulated their goals across time). In Chapters 3, 4, and 5, I present a series of studies 2 Dispositions can also have direct effects on motivation (see Judge & Ilies, 2002), and so can situations. Because this project is mostly concerned with the way in which motivational constructs are interrelated within individuals and across time, I will not investigate such direct effects in this dissertation.
9 designed to seek support for the hypotheses proposed in this paper. Finally, in Chapter 6, I offer a general discussion that integrates the findings of the three studies conducted, and outlines the implications, contribution, and limitations of this research.
10 CHAPTER 2 LITERATURE REVIEW Integrative Goal Setting Theory The motivational technique of goal setting is perhaps the most valid scientific motivational theory that has been applied to work motivation (e.g., Pinder, 1984; Wood & Bandura, 1989). In short, the most recent version of the goal setting theory specifies that assigned goals influence performance through self-set (personal) goals and selfefficacy judgments (e.g., Latham & Locke, 1991; Locke, 1997). In addition, self-efficacy also influences personal goals. Schematically, this model is presented in Figure 2-1 below.1 Figure 2-1. Goal-setting effects on performance Goals that are assigned by legitimate leaders or supervisors influence peoplesÂ’ self-set (personal) goals because individuals typically commit to the assigned goals (e.g., 1 Ability also influences performance both directly, and indirectly through self-set goals and self-efficacy (e.g., Latham & Locke, 1991). In the interest of parsimony, ability effects are not reviewed here. Assigned Goals Performance Self-Set Goals Self-Efficacy
11 Latham & Locke, 1991). Furthermore, assigned goals are an indicator of assessed capability to perform the task, thus they influence self-efficacy. If goals can influence performance, what types of goals have more beneficial influences on performance? Two main goal attributes are thought to have positive influences on task performance: goal difficulty and goal specificity (Locke, 1997; Locke & Latham, 1990). Given that they have sufficient ability and knowledge to perform the task, people who set more difficult goals will perform at a higher level than those who set less difficult goals (e.g., Locke, 1997). Â“Goal specificity refers to the clarity of the goal or the degree to which the goal refers to an explicit versus a vague performance outcomeÂ” (Locke, 1997, p. 380). Although goal specificity by itself does not necessarily influence performance (Locke, 1997), people who are striving to accomplish goals that are both specific and difficult routinely outperform people trying for other types of goals (e.g., easy goals, or Â‘do your bestÂ’ [vague] goals) (Locke & Latham, 1990). In addition to being specific and difficult, goals have to be attainable in order to have a maximal beneficial effect on performance (Locke & Latham, 1990). Social cognitive theory (e.g., Bandura, 1986; Bandura & Locke, 2003) proposes that a central self-regulatory process of motivation works though peopleÂ’s beliefs in their personal efficacy. Perceived self-efficacy concerns individualsÂ’ judgments of their capabilities to mobilize motivational and cognitive resources and to formulate task strategies such as to exercise control over actions needed to perform a specific task. Strong positive self-efficacy beliefs increase peopleÂ’s level of motivation as reflected in how much effort they will exert on the task and how long they will persevere (e.g., Bandura & Cervone, 1983; Wood & Bandura, 1989). Increased levels of effort and
12 perseverence usually lead to performance improvements; thus, high levels of self-efficacy will usually be associated with high performance levels. Self-efficacy also influences performance indirectly, through its effect on personal goals. Individuals who have greater confidence in their capability to perform a specific task will naturally set higher goals for their performance on the task than those who lack such self-confidence (Button, Mathieu, & Aikin, 1996; Earley & Lituchy, 199). The literature on goal setting is vast and generally supportive of the positive effects of difficult and specific goals on task performance (Mento, Steel, & Karren, 1987). Reviews of the goal setting literature are numerous: Locke, Shaw, Saari, and Latham (1981) present an enumerative review of the goal setting literature; the metaanalysis of Mento et al. (1987) shows support for predictions derived from goal setting theory; and LockeÂ’s most recent papers (e.g., Locke, 1997) present comprehensive reviews of the literature and further conceptual refinements of the theory. Various types of studies have showed support for the effects of goals, and, cumulatively, this support is impressive: Goal setting has been studied most often at the individual level and is supported by the results of several hundred studies, conducted in both laboratory and field settings, using 88 different tasks and more than 40,000 subjects in eight different countries, employing time spans from minutes to years, involving different measures of performance, including creativity (Shalley, 1991), and many forms of goals, for example, assigned, self-set, and participative. (Locke 1997, p. 381) Given the sheer amount of empirical evidence in support of goal setting theory, I will not present a comprehensive review of this literature here. Instead, I will focus on the interrelationships between goals and feedback, which are at the center of the dynamic model of task motivation developed in this dissertation.
13 Feedback, Performance, and Goals As noted, goal-setting theory was developed as a technique to enhance individualsÂ’ motivation in the workplace. As such, applications of the theory have been extensively used in organizations (e.g., management by objectives; Mento et al., 1987). In order to implement motivational programs based on goal setting, it is important to study what moderates the effects of goals on performance. One such moderator is performance feedback (Locke & Latham, 1990). Latham and Locke (1991, p. 226) concluded that their review of the effects of feedback and goals on performance Â“leads to the conclusion that goals and feedback together are more effective in motivating high performance or performance improvement that either one separately.Â” Furthermore, Mento et al. (1987) provided meta-analytic evidence for the beneficial effect of feedback in goal setting interventions. Performance feedback is also essential in understanding how employees regulate their goals and behaviors across time. Locke (1997 p. 384), after describing the moderating role of feedback (knowledge of results) on the effects of goals on performance, notes: Â“On the other side of the goal-feedback coin, goals mediate the effect of knowledge of results on subsequent performance.Â” From this point of view, performance feedback is a process variable that explains motivation within individuals, which is exactly the perspective that I take in this dissertation. That is, I examine the processes used by individuals to regulate their motivation across time, and I investigate individual differences in the nature and strength of these idiographic processes. Toward that end, I first review the conceptual links between performance feedback and goals across time, and then I turn to mediating processes and moderating effects.
14 The effect of feedback on goals has to be examined within the broader framework that considers the various mechanisms that explain the effect of feedback interventions on task performance. This broad area has generated a large body of research, but the findings have often been contradictory or inconsistent (Kluger & DeNisi, 1996). Latham and Locke (1991, p. 224) note that Â“few concepts in psychology have been written about more uncritically and incorrectly than that of feedback.Â” Kluger and DeNisi (p. 277) argue that lack of a unified theory makes it impossible to interpret such findings: Â“Without a comprehensive theory, there is no way to integrate the vast and inconsistent empirical findings.Â” The simplest framework explaining the effects of feedback on performance assumes that a basic regulatory mechanism of behavior is the Â“evaluation of and reaction to a feedback-standard comparisonÂ” (Kluger & DeNisi, 1996, p. 259). In a multi-trail setting, comparing task feedback with the goal set for the specific task, leads to the categorization of feedback as positive (when the goal is met or exceeded) or negative (when the goal is not met). The feedback-comparison-reaction framework assumes that people are motivated to align their performance with their standards and they attempt to do so by striving to reduce or eliminate any discrepancy between performance and standards (both positive or negative). From a static standpoint (at a certain point in time), individuals have two main choice alternatives to eliminate the gap between performance (i.e., feedback) and standards (goals): to adjust their behavior or their standards (Kluger & DeNisi, 1996). If such gap is negative, it can be eliminated by (a) putting forth more effort to improve performance (direct effort regulation) or (b) adopting a lower standard (goal regulation).
15 When performance is equal to or higher than the standard, the gap can be eliminated by (a) exerting less effort (direct effort regulation) or (b) increasing the goal (goal regulation). The classic feedback-comparison-reaction model stipulates that most often people regulate their effort (i.e., increase effort following negative performance) rather than adjusting their goals (e.g., Podsakoff & Farh, 1989). Recently though, several authors have questioned this assumption and attempted to identify situational or dispositional conditions under which the assumption holds (vs. when the effect is reversed and individual adjust their goals and not their effort level) (e.g., Cron, Slocum, & VandeWalle, 2001; Ilgen & Davis, 2000; Kluger & DeNisi, 1996). In contrast to the feedback-comparison model, according to goal setting theory (e.g., Locke & Latham, 1990), people are motivated to achieve goals, rather than eliminate discrepancies between performance and standards. Thus, according to this theory, following negative performance feedback, people will increase their effort levels while maintaining their goals. When their performance matches the standard (positive feedback), according to the goal setting theory, people create positive discrepancies (setting goals higher than their past levels of performance; Phillips et al., 1996) to motivate their behaviors, and they are generally motivated to accomplish these new goals. After they receive negative performance feedback, people will generally reduce negative discrepancies between performance and standards by increasing effort levels and maintaining their goals. Over time, the relationship between feedback and future goals will likely be positive (and not null) because as feedback becomes more negative people are likely to start adjusting their goals downwardly, and this relationship will become
16 increasingly strong over a wide range of negative performance feedback.2 Following positive feedback, people will generally increase their goals but, like in the case of negative feedback, the exact nature and magnitude of this relationship will vary according to the characteristics of the task and will also be influenced by dispositional moderators (examined in the next section). In sum, for both positive and negative feedback ranges, previous research suggests a positive relationship between performance feedback and subsequent goals within individuals (i.e., across the positive feedback range, the more positive the feedback is, the higher the subsequent goal will be; across the negative feedback range, the less negative the feedback is, the higher the subsequent goal will be), but it also suggests that the idiographic self-regulatory process may work differently for different individuals, or in different situations. In the following sections, I investigate the psychological mechanisms explaining how individuals adjust their goals following performance feedback, and then I examine how different individuals use these mechanisms. Goal Regulation: Mediating Mechanisms Figure 2-2 presents my conceptualization of the dynamic processes that mediate the feedback-subsequent goals relationship. Perhaps the most intuitive way to explain this model is to view it from a multi-trial perspective where individuals set goals before each task trial, and receive performance feedback after performing each task trial. Task behavior (performance) leads to feedback, which influences subsequent goals. A major goal of this dissertation is to study the mediating processes through which individuals interpret performance feedback. 2 Under extreme or repeated negative feedback individuals are likely to withdraw from the task.
17 Figure 2-2. Conceptual model of the idiographic feedback-goal mediation process A cognitive explanation of the effects of performance feedback on future goals stipulates that feedback impacts subsequent goals through its effect on task self-efficacy (e.g. Latham & Locke, 1991). Positive feedback (indicating that an individual has performed better than his/her initial goal) raises self-efficacy (Latham & Locke, 1991). In contrast, negative feedback leads to decreased self-efficacy. In turn, self-efficacy judgments influence goals. According to social cognitive theory, people rely on past performance to judge their task self-efficacy and set personal performance goals (Bandura, 1991; Wood & Bandura, 1989). People with high self-efficacy set goals that are higher than previous performance (positive discrepancy creation; Phillips et al., 1996), whereas those with low levels of self-efficacy will set goals that are lower than their past level of performance. The mediating role of self-efficacy is included in the model in the interest of comprehensiveness, but studying this mediating effect is not central to this dissertation. Thus, I will not offer formal hypotheses with respect to the role of self-efficacy in explaining self-regulation within individuals. Kluger and DeNisi (1996), in a comprehensive meta-analytic review on the effect of feedback interventions on performance, explain that the classic feedback-standard comparison argument is inadequate for explaining the effect of feedback on performance Performance Feedback Affect Self-Efficacy Subsequent Task Goal
18 because it ignores feedback-induced affect and its effects on future performance. With respect to goal setting, Brockner and Higgins (2001, p. 47) note that Â“emotional consequences of goal attaintment/nonattainmentÂ” are an aspect of goal-setting theory that has been neglected. Thus, in addition to the cognitive mediation effect through selfefficacy, I propose that performance feedback also influences subsequent goals through an affective mechanism: performance feedback signaling success or failure (in reaching the initial goal) influences individualsÂ’ positive and negative affect which activate the behavioral approach or avoidance system, respectively. Indeed, there is empirical evidence suggesting that goal attainment or goal progress is associated with positive affect, whereas nonattainment or lack of progress is associated with negative affect (e.g., Alliger & Williams, 1993; Carver & Scheier, 1990). Then, depending on which behavioral system is more active, individuals will increase their goal (when BAS is more active) or decrease it (when BIS is more active) before engaging in the next task trial. In sum, positive performance feedback leads to positive moods and emotions which, in turn, lead to increased goals. Negative feedback leads to decreased goals through negative moods and emotions. Even though they did not examine the role of affect or emotions in the goal regulation process, Williams, Donovan, and Dodge (2000) studied goal and performance regulation in 25 varsity track and field athletes with a longitudinal design. These authors found evidence for both downward goal revision following negative feedback and positive discrepancy production following positive feedback and noted that their results are consistent with social cognitive theory (Bandura, 1986) and would be difficult to explain by control theory (Carver & Scheier, 1981).
19 Though much has been written on the conceptual relationship between affect and motivation, very few studies present direct empirical investigations of the nature and magnitude of this relationship. Erez and Isen (2002) reviewed the literature on the link between positive affect and motivation and concluded: Â“Â… a closer look at the body of literature that does link the two constructs reveals very little actual empirical work integrating motivation theory and work on affect.Â” Erez and Isen (2002) examined the relationship between positive affect and expectancy components of motivation. In their first study, these authors found that positive affect influenced participantsÂ’ perceptions of expectancy and valence, and their performance on an anagram-solving task. In the second study presented by Erez and Isen (2002), positive affect influenced all three components of expectancy motivation (expectancy, instrumentality, and valence). Specifically with respect to the mediating role of affect in explaining the relationship between performance feedback and subsequent goals, Cron, Slocum, and VandeWalle (2001) found that negative performance feedback (failing to meet oneÂ’s goal on a course exam) was positively related to subsequent goals (the less negative the feedback, the higher the subsequent goal was), and the relationship was partly mediated by negative emotional reactions. The results of Cron et al. (2001), though suggestive, come from cross-sectional analyses. Such analyses, even though the design was longitudinal, cannot separate within-individual from cross-sectional effects. Summarizing the expectations derived from the theoretical considerations reviewed above, I expect that feedback concerning past performance will influence subsequent goal-setting and that both positive and negative affect will have a mediating effect on the feedback-subsequent goals relationship. Thus,
20 H1: Across time, performance feedback will have a direct effect on subsequent goals such that feedback indicating better performance will be associated with higher subsequent goals than feedback indicating lower performance. and H2: The effect of performance feedback on subsequent goals will be mediated by positive and negative affect. It is important to note that the predictions proposed in this dissertation concern a continuum of performance feedback, which can range from negative feedback indicating a large negative discrepancy between performance and goals, to feedback indicating that the goal has been matched, to positive feedback that indicates a large positive discrepancy between performance and goals. This Â‘continuousÂ’ approach to the study of performance feedback, unlike the categorical approach (i.e., investigating the effects of feedback sign; Kluger & DeNisi, 1996), is particularly well suited for the study of withinindividual relationships, and for investigating cross-level moderating effects.3 Though the study of cognitive mechanisms is not a principal focus of this dissertation and thus I do not offer formal hypotheses with respect to these mechanisms, I acknowledge that affect may also influence subsequent goals indirectly through a cognitive mechanism: by influencing judgments of task self-efficacy. The simplest explanation of the effects of affective states on judgments is offered by the mood congruency effect. Mood congruency theory is largely based on associative network models of memory (e.g., Bower, 1981), which suggest that emotions impose an organizational structure on concepts in memory and stimulate similarly valenced 3 In a multi-level modeling framework, the slopes of the within-individual relationships can be modeled, across individuals, with equations that include moderator variables as predictors.
21 memories and cognitions. According to mood congruency theory, during positive affective states, positively valenced information and cognitions become activated, whereas negative affective states make association with negatively valenced memories and cognitions more likely. In this way, affective states are believed to influence beliefs and judgments. Indeed, empirical evidence suggests that positive mood is associated with increased levels of self-efficacy in performance or achievement situations (Baron, 1990; Forgas, Bower, & Moylan, 1990). Recently though, more sophisticated models of the interrelationships between affect and cognitions have been developed (e.g., the mood-asinformation model [Schwarz & Clore, 1983]; the mood-as-input model [Martin, Ward, Achee, & Wyer, 1993]), and these models specify that the mood congruency effect will only be observed under certain conditions. Moderating Influences Both situational and dispositional factors can influence the within-individual processes that regulate goals and performance across time, in the presence of performance feedback. With respect to negative feedback, Brockner, Derr, and Laing (1987, p. 319) noted: This interest [in the effect of negative feedback on subsequent motivation] is partially attributable to the fact that failure feedback has the potential to elicit a wide variety of motivational responses. That is, negative feedback may decrease, increase, or have no effect on individualsÂ’ strivings, depending upon certain situational and dispositional factors. Investigating the influence of dispositional factors on the within-individual regulatory processes that link feedback with affect on goals is a central objective of this dissertation. Below, I review the literature relevant to dispositional moderating effects, and I present formal hypotheses with respect to these effects.
22 Dispositional moderators In this section, I develop hypotheses with respect to dispositional moderating effects on the feedback-subsequent goals relationship. Because this relationship is expected to be partly mediated by affect (as specified in H2), moderator effects on the feedback-affect and affect-subsequent goals relationships could be specified as well. Such potential moderating influences are portrayed in Figure 2-3 below: Figure 2-3. Moderating effects on the feedback-goal relationship The dispositional moderating effects on the main feedback-goals relationship, conceptually, represent individual differences in sensitivity to feedback, as reflected in individualsÂ’ goal setting tendencies. Studying individual differences in motivational reactions to feedback is a research area that has generated substantial interest among researchers (e.g., Kluger & DeNisi, 1996), and understanding how personal characteristics are connected to the feedback process can have important organizational implications (Weiss & Adler, 1984). For this reason, and in the interest of parsimony,I formulate my expectations with respect to the dispositional moderators of the main feedback-goals relationship, but I acknowledge that the same dispositional factors may have similar moderator effects on the feedback-affect and affect-subsequent goal relationships. MODERATING INFLUENCES Performance Feedback Subsequent Task Goal Affect
23 In the introduction, I made a case for the role of the basic behavioral systems of approach and avoidance (BAS and BIS) in regulating human motivation. In this section, I attempt to show that the various dispositional characteristics that have been linked to individualsÂ’ sensitivity to feedback are all indicators of chronic levels of BIS and BAS activation in individuals.4 Toward that end, I first review theoretical considerations and empirical evidence with respect to dispositional influences on individualsÂ’ reactions to feedback. Neuroticism and extraversion. The introductory chapter presents a description of the two motivational-behavioral system components of GrayÂ’s (1981; 1990) Reinforcement Sensitivity Theory (RST) Â– namely the BAS and the BIS. In short, RST specifies that the BAS deals with appetitive motivation and approach behaviors, whereas the BIS deals with aversive motivation and avoidance behaviors. Gray (1990) also argues that BAS activity controls the experience of positive feelings and the BIS is responsible for the production of negative feelings. Gray suggested that the traits of anxiety and impulsivity reflect individual differences in the strength of the BIS and BAS, respectively. Trait anxiety is responsible for individual differences in typical response patterns to punishment and frustration (non reward). Individuals who score high on trait anxiety measures are mainly motivated to avoid punishment and thus have a tendency to inhibit their behavior in novel environments. In contrast, trait impulsivity is related to enhanced reward sensitivity. Impulsive individuals are mainly motivated by rewards, thus they have a tendency to exhibit approach behaviors, especially in novel situations. Because anxiety is a facet of 4 A direct measure of chronic BIS/BAS activation was developed by Carver and White (1994), and includes four dimensions: BIS, BAS Reward Responsiveness, BAS Drive, and BAS Fun Seeking. I use this measure in one of the studies.
24 the broader personality dimension of neuroticism (Costa & McCrae, 1992), and impulsivity belongs to the broad domain of extraversion (e.g., McCrae & Costa, 1985), it follows that with regard to within-individual processes that link performance feedback to subsequent goal setting (through affective processes), neuroticism should moderate the strength of the feedback-goals relationship when feedback is negative, and extraversion should moderate the relationship among positive feedback and goals.5 Thus, individuals who score high on neuroticism should display stronger idiographic relationships between negative performance feedback and their subsequent goals than those scoring low on neuroticism, whereas people scoring high on extraversion should have their subsequent goals influenced by positive feedback more strongly than those scoring lower on extraversion. Though following a different path to derive the conceptual connection between neuroticism and extraversion, and the BIS and the BAS, Carver, Sutton, and Scheier (2000) contend that by focusing on approach and avoidance tendencies in defining extraversion and neuroticism, personality scholars would rely less on factor analysis and more on conceptual explanations of behavior. These authors concluded: Â“The notion that approach and avoidance tendencies are fundamental qualities underlying behavior addresses this problem [over reliance on factor analysis] by suggesting a fundamental basis for the existence of each of these trait dimensionsÂ” (Carver et al., 2000, p. 748). Larsen and Ketelaar (1989), in an experimental study designed to test the moderating effect of neuroticism and extraversion on peopleÂ’s susceptibility to mood inductions, induced pleasant and unpleasant moods using manipulated feedback 5 The correspondence between anxiety and neuroticism, and between impulsivity and extraversion is not perfect as anxiety and impulsivity likely lie at a 30-degree angle from neuroticism and anxiety, respectively (Pickering, Corr, & Gray, 1999).
25 indicating success or failure. Extraversion predicted respondentsÂ’ positive affect increases, following positive feedback, and Neuroticism predicted increases in negative affect scores after negative feedback. The effects of neuroticism and extraversion on positive and negative mood inductions have been consistently replicated by Larsen and his colleagues in other studies (e.g., Larsen & Ketelaar, 1991; Rusting & Larsen, 1997). Affectivity. Modern affect theory specifies that affective tendencies and experiences are regulated by the two broad bio-behavioral systems described earlier; the BAS controls the experience positive affect, and the BIS controls the experience of negative affect (e.g., Carver et al., 2000; Watson, 2000). Watson and his colleagues (Watson, 2000; Watson et al., 1999) contend that the principal parameters of the BAS and the BIS are the dispositional factors of Positive Affectivity (PA) and Negative Affectivity (NA), respectively. It follows that trait-PA reflects individual differences in the sensitivity of the BAS and thus should moderate individualsÂ’ reactions to positive feedback (reward), whereas trait-NA reflects individual differences in the sensitivity of the BIS so it should moderate individualsÂ’ sensitivity to negative feedback (punishment). Not surprisingly, trait measures of NA correlate substantially with neuroticism, and traitPA scores correlate highly with extraversion (see Watson, 2000). Chronic self-regulatory focus. Higgins (1997, 1998) proposed that there are important differences in the processes that explain peopleÂ’s motivation to approach pleasure and avoid pain. Higgins proposed two distinct self-regulatory mechanisms, one in which people have a promotion focus, and one in which they have a prevention focus. Regulatory Focus Theory specifies (Brockner & Higgins, 2001, p. 37):
26 PeopleÂ’s regulatory foci are composed of three factors which serve to illustrate the differences between a promotion focus and a prevention focus: (a) the needs that people are seeking to satisfy, (b) the nature of the goal or standard that people are trying to achieve or match, and (c) the psychological situations that matter to people. Promotion oriented people are seeking to satisfy growth and development needs, seek to attain goals associated with their ideal self (goals that reflect hopes, wishes and aspirations), and are motivated by positive outcomes (the presence and absence of positive outcomes are salient for promotion-focused people) (Brockner & Higgins, 2001). In contrast, prevention oriented people are predominantly driven by security needs, they seek to attain goals or standards associated with their ought self (referring to duties, obligations, and responsibilities), and for such people negative outcomes (presence or absence of such outcomes) are more salient (Brockner & Higgins, 2001). A dispositional conceptualization of regulatory focus assesses individual differences in the strength or importance of individualsÂ’ promotion and prevention foci. Self-regulatory focus strength is typically measured with response latencies exhibited by participants who complete a survey that requires them to list attributes of their ideal and ought selves, and rate themselves on these attributes (e.g., Brockner & Higgins, 2001). Carver et al. (2000) note that the self-regulatory focus framework resembles other motivational-behavioral systems that reflect differences in behavioral approach and avoidance. The relationship between the strength of the BAS and BIS and dispositional self-regulatory focus is obvious: individuals with high chronic activation levels of the BAS are generally approach-orientated and so they will have a strong promotion selfregulatory focus; whereas those characterized by high BIS activation levels are generally avoidance-oriented, thus they will have a strong prevention self-regulatory focus. It
27 follows that individuals with a promotion self-regulatory focus should be more sensitive to approach incentives (i.e., rewards; positive feedback), whereas those with a prevention self-regulatory focus should be more sensitive to avoidance incentives (i.e., punishments; negative feedback). Goal orientation. The concept of goal orientation, defined as the broad goal held by an individual facing a [learning] task, has generated substantial interest in the motivation, learning, and training literatures, as it is thought to have an effect on how individuals learn (Dweck, 1986; Dweck & Leggett, 1988). Mastery, or learning, orientation is characterized by the desire to increase oneÂ’s competence on the task. Mastery oriented individuals have been characterized as being self-directed and selfregulated learners (Seifert, 1997), they believe that intelligence is malleable and effort leads to success (Dweck & Leggett, 1988), prefer challenges (Seifert, 1995), and engage in more complex learning strategies (Fisher & Ford, 1998). Performance oriented people focus on task performance and comparisons with others, seek to prove their ability to others, and believe that intellectual abilities are immutable (e.g., Dweck & Leggett, 1988). Kanfer and Ackerman (1989) argue that motivational variables have two important functions in skill acquisition and self-regulation: (a) they drive the allocation of attentional effort from a limited resource pool, and (b) they direct the allocation of effort within the learning task. Following Kanfer and AckermanÂ’s contention about the functions of motivation described above, Fisher and Ford (1998, p. 403) suggest that Â“goal orientation may serve these functions.Â” Individuals with a high mastery orientation will direct attention to the task, and allocate effort more efficient to learning. In contrast,
28 those characterized by a high performance orientation will devote part of their effort to ego management and part to perceived performance indicators, thus exhibiting a less efficient allocation of resources to the task of learning itself. It follows that goal orientation should influence how individuals respond to task failure (e.g., Farr, Hofmann, & Ringenbach, 1993). Whereas mastery, or learning, oriented people will likely intensify their efforts and persist longer to achieve personal development, those with a performance orientation will likely withdraw from the task because failure could reveal low ability. More recently, VandeWalle (1997) proposed that goal orientation is better conceptualized as a three-factor construct. The three factors are obtained by partitioning the traditional performance orientation dimension into a Â‘proveÂ’ dimension and an Â’avoidÂ’ dimension. In this project, I adopt such a measurement approach and propose that the two factors of mastery/learning and avoidance (performance) orientation are especially relevant to self-regulation and motivation across time. Indeed, in a study that focused specifically on the influence of goal orientation on personal goals, VandeWalle, Cron, and Slocum (2001) found that learning goal orientation positively influenced personal goal levels, and avoidance goal orientation negatively influenced goal levels (proving orientation had no effect on goal level). Furthermore, Cron et al. (2001) found that the intensity of negative emotional reactions to negative performance feedback was predicted by respondentsÂ’ standing on an avoidance goal orientation measure. The theory and evidence reviewed above suggest that learning and avoidance goal orientation are related to the BAS (learning orientation) and the BIS (avoidance
29 orientation), thus I expect these two factors to influence individualsÂ’ reactions to feedback in the same way as do the other indicators of BAS/BIS activation. Summarizing the arguments presented above, I propose the following hypotheses regarding dispositional moderating effects: H3a: Following negative feedback, the feedback-subsequent goals relationship will be stronger for individuals who have increased chronic levels of BIS activation. Increased chronic BIS activation levels are indicated by: (a) high scores on the BIS scale, (b) high neuroticism and trait-negative affect scores; (c) avoidance goal-orientation; and (d) chronic prevention selfregulatory focus. H3b: Following positive feedback, the feedback-subsequent goals relationship will be stronger for individuals who have increased chronic levels of BAS activation. Increased chronic BAS activation levels are indicated by: (a) high scores on the BAS Reward Responsiveness scale, (b) high extraversion and trait-positive affect scores; (c) mastery goal-orientation; and (d) chronic promotion self-regulatory focus. Situational influences This section is concerned with examining situational influences on the withinindividual relationship between feedback and goals. When performance is negative, for example, Â“people choose to increase their effort, rather than lower the standard, when the goal is clear, when high commitment is secured for it, and when belief in eventual success is highÂ” (Kluger & DeNisi, 1996, p. 260). I contend here that constructs that moderate the effects of goal setting on performance may similarly impact the relationship between performance feedback and goals.
30 Goal commitment . Goal commitment is one of the most studied moderators of the effects of goal setting on performance. Locke (1997, p. 385) notes: Â“It is virtually axiomatic that if people are not really trying for a goal, then the goal will not lead to improved performance.Â” Indeed, Donovan and Radosevich (1998) present meta-analytic evidence for the moderating role of goal commitment, though the meta-analytically estimated magnitude of the moderating effect was relatively low. With respect to the feedback-goals-performance relationship, goal commitment is most relevant under conditions of goal nonattaintment or unsatisfactory goal progress which lead to negative performance feedback. Under these conditions, when individuals are committed to their overall goals, they will adjust their behavior (increase effort) rather than decrease their goals. It follows that for negative feedback, the feedback-subsequent goals relationship will be moderated by goal commitment such as the less committed individuals are to their overall goal, the stronger the feedback-subsequent goals relationship will be (people change their goals more easily when they are less committed to these goals). General task self-efficacy. Brockner et al. (1987, p. 319) note that theories designed to explain the effect of negative feedback on subsequent motivation Â“all suggest that individualsÂ’ postfailure expectations concerning the efficacy of their subsequent effort is an important mediator of their degree of motivation.Â” A situational variable that directly influences these efficacy expectations is general task self-efficacy. If an individual believes that he or she can successfully perform the behaviors required for satisfactory task performance, feedback that indicates unsatisfactory goal progress is likely to result in increased levels of effort. General task self-efficacy is a domainspecific general belief about oneÂ’s capability to perform a certain type of task, and it is
31 different from state self-efficacy in that it is a general belief that is consistent across performance situations (unlike state, or task-specific, self-efficacy [Chen, Gully, Whiteman, & Kilcullen, 2000]), though it will eventually be influenced by repeated performance feedback. General task self-efficacy is also distinct from, though not unrelated to, generalized self-efficacy (Sherer et al., 1982) because general task selfefficacy is domain specific, as opposed to generalized self-efficacy which is a general set of beliefs and expectations about oneÂ’s ability across task domains. In sum, I believe that (a) goal commitment, and (b) general task self-efficacy should moderate the magnitude of the within-individual relationship between feedback and future goals. For these effects, I offer the following two hypothesizes: H4: Following increasingly negative feedback, individuals who are more committed to their overall performance goal will be less likely to downwardly adjust their goals than less committed individuals. H5: Across individuals, general task self-efficacy will moderate the magnitude of the negative feedback-subsequent goals relationship such as more efficacious individuals will be less likely to adjust their goals downwardly than less efficacious individuals following negative feedback.
32 CHAPTER 3 STUDY 1 The first study was designed to test the main effects and mediation hypotheses (H1 and H2), and to investigate support for the dispositional moderation hypotheses (H3a and H3b) by examining whether differences among individuals on the hypothesized dispositional factors influence the magnitude of the within-individual relationships (i.e., examining cross-level moderator effects). Method The study was conducted in two phases. In the first phase, undergraduate students were asked to complete a personality survey measuring neuroticism, extraversion, and trait positive and negative affectivity. In the second phase of the study, participants completed an 8-trial experiment that asked them to successively set a performance goal, and perform a certain task (there were two versions of the task), for each trial. After performing the trial task, participants received performance feedback that was either real or manipulated, and then they were asked to report their current affective state. The second phase was conducted using an Internet interface that was specifically developed for this study. Participants Participants were recruited from a large introductory management class at the University of Florida. Students signed up for the study on a web page that sent the data to a secure database (the database was stored in a non-public folder on the web server). They were invited to participate by an advertisement that was placed on the course Web
33 page. Participation in the study was completely voluntary and individuals who participated received extra credit points in return for their participation. A total of 1,024 undergraduate students completed phase 1 (the personality surveys); 745 participants provided complete data for the study. Experimental Design and Procedure Participants logged on to the Internet site, read a detailed description of the task and procedure, were asked to report their momentary affective state and then to set a goal for the first trial task. The Web page for goal-setting gave participants the option to choose between nine different goal levels, ranging from 10% to 90% (i.e., I want to perform better than 10/90% of the participants in this experiment). After setting a goal for the first trial, participants were presented with the performance task and were given five minutes to work on the task. After submitting their task solutions, participants were presented with performance feedback. Half of the participants were provided with real feedback regarding their performance (how they actually performed) by programming the electronic interface to evaluate respondentsÂ’ solutions against the correct solutions and using a distribution of responses constructed a priori. The other half received manipulated feedback that ranged between 35% and 80% (i.e., for this trial, you have performed better than 35/80% of the participants), which was randomized across trials for each participant. After receiving the feedback, participants were asked to report their affect, and then they started the subsequent trial. Appendix A shows the Internet screens for the phase 2 multi-trial experiment.
34 Performance Task Two different tasks were used. The first was a brainstorming task that asked participants to list as many uses of a common object or material as they could. This task has been successfully used in prior laboratory research on goal setting motivation (e.g., Harkins & Lowe, 2000; Locke, 1982). Students were asked to list uses for the following objects/materials: (1) absorbent towel, (2) rubber tire, (3) wood, (4) ice, (5) sunlight, (6) a sheet of paper, (7) coat hanger, and (8) sand. The second task contained five items from Mednick and MednickÂ’s (1967) Remote Associates Test (RAT; see Appendix B for a sample trial), a test widely used with undergraduate students (e.g., Brown & Marshall, 2001). Measures Affect . I used the Positive and Negative Affect Schedule (PANAS; Watson, Clark, & Tellegen, 1988) with momentary instructions for measuring positive affect (PA) and negative affect (NA). The internal consistencies reliability of the PA scores ranged between .92 and .95 across the eight trials; the reliability of the NA sores was between .90 and .92 across the trials. The measure is shown in Appendix C. Neuroticism and extraversion . I used the NEO-FFI survey (Costa & McCrae, 1992) to measure the personality factors of neuroticism and extraversion. Internal consistency was .88 for neuroticism and .82 for extraversion. Positive and negative affectivity . I used the Positive and Negative Affect Schedule (PANAS; Watson, Clark, & Tellegen, 1988) with general instructions for measuring positive affectivity (trait-PA) and negative affectivity (trait-NA). The internal consistencies were .85 and .88 for trait-PA and trait-NA, respectively.
35 Analyses To test the hypothesized within-individual effects and the cross-level moderator effects, I used hierarchical linear modeling (HLM; Byrk & Raudenbush, 1992). Before proceeding with the tests of the hypotheses, I first investigated whether systematic withinand between-individual variance exists in the performance goals set by individuals. To do so, I estimated a null model which calculated the withinand betweenindividual variance in goals. The equations for the null model are presented below: Level 1: Goalij = 0j + rij (1) Level 2: 0j = 00 + U0j (2) where Goalij= individual jÂ’s goal for trial i; 0j=average goal level for individual j; 00= grand mean of goal scores; 2 = variance (rij) = within-individual variance in goals; 00 = variance (U0j) = between-individual variance in goals. Provided that the test of the null model reveals that there is substantial withinand between-individual variance in the criterion, tests of the other HLM models can be conducted. Because there were four distinct conditions in this study (two versions of the task [Â‘usesÂ’ vs. RAT] and two types of feedback [real vs. manipulated]), I estimated the null model on each of the four data sets. Below, I offer descriptions of analyses used to test the hypotheses. Hypothesis 1 . Model 1 tests the relationship between feedback and goals within individuals. At level 1, the model estimates the individualsÂ’ intercepts and slopes for predicting goals with feedback, and at level 2, because no predictors are included in the equations, the models estimates the pooled values for the level 1 parameters. In order to
36 estimate parameters using only within-individual variance, the feedback variable was centered at the individualsÂ’ means, which removed all the between-individual variance in the predictor scores. The equations for Model 1 are presented below: Level 1: Goalij = 0j + 1j (Fdij) + rij (3) Level 2: 0j = 00 + U0j (4) 1j = 10 + U1j (5) where Goalij= individual jÂ’s goal for trial i; Fdij= individual jÂ’s feedback for performance on trial i-1; 0j= level 1 intercept for individual j. 1j= individualsÂ’ slopes for predicting their goal with the feedback concerning their previous performance, across time; 00=pooled intercept; 10=pooled slope for predicting goals with feedback. As it was the case with the null model, Model 1 was estimated on each of the four data sets determined by the type of performance task (Â‘usesÂ’ vs. RAT) and the type of performance feedback (real vs. manipulated). Hypothesis 2 . To test whether affect mediates part of relationship between feedback and subsequent goals, I followed the procedures for testing mediation with regression analysis outlined by Baron and Kenny (1986) and MacKinnon and Dwyer (1993), as well as the recommendations for applying these procedures for testing mediation in multilevel models given by Krull and MacKinnon (1999). The simplest analysis for testing mediation with ordinary least squares (OLS) analysis, is to investigate whether a hypothesized mediator (M) mediates the relationship between X and Y. The conceptual model for testing this mediating effect is presented in Figure 3-1.
37 Figure 3-1. Illustration of a mediating effect Using the notations from Figure 3-1, c is the magnitude of the direct effect of X on Y, a is the magnitude of the direct effect of X on the mediator M, and the product a*b represents the magnitude of the meditating effect. Alternatively, the quantity c-cÂ’ represents the magnitude of the mediation effect. (In OLS regression on single-level data c-cÂ’=a*b [MacKinnon, Warsi, & Dwyer, 1995].) To examine whether such a mediating effects is statistically significant by testing the significance of the product a*b, three procedures based on (a) the first-order Taylor series expansion of the multivariate delta method (Sobel, 1982), (b) the second order Taylor series expansion or exact variance under condition of independence (Goodman, 1960), and (c) the estimate of unbiased variance (Goodman, 1960) have been developed (Krull & MacKinnon, 1999; MacKinnon et al., 1995). Alternatively, methods to test the significance of c-cÂ’ have been developed (e.g., Freedman & Schatzkin, 1992), but these methods are not recommended because they can have high Type I error rates under certain conditions (e.g., when a 0 and b = 0; MacKinnon Lockwood, Hoffman, West, & Sheets, 2002). X Y M X Y c a b c'
38 In multilevel modeling, when the number of groups (i.e., individuals, in this study) exceeds 100, Krull and MacKinnon (1999) recommend testing the significance of the mediation effect with the unbiased estimator test. This test involves computing the zvalue for the mediation effect as z = a * b /SQRT( b2* sa 2 + a2* sb 2 sa 2* sb 2). MacKinnon and colleagues (see MacKinnon et al., 2002) have shown that the test comparing the z-value computed with the previous formula with the unit normal distribution has low power because the product a*b is not normally distributed (its distribution is often asymmetrical with high kurtosis). To address the low power problem, MacKinnon et al. (2002) estimated empirical sampling distributions of z for various magnitudes of the mediation effect and various sample sizes on the basis of extensive simulations, and they termed the test using the empirical sampling distribution as zÂ’). In this study, I will use the zÂ’ test by comparing the z-value computed with the formula presented above with the empirical critical values for a small mediating effect and N=100 provided by MacKinnon et al. (2002). When the role of more than one mediating variable is examined in the same model, the significance of the mediating effect through each variable can be tested using formula presented above, and the proportion of the total effect mediated jointly is computed as (c-cÂ’)/c. Alternatively, the proportion of the total meditating effect can be computed as ( ! ai*bi)/( ! ai*bi + cÂ’), though different researchers diverge on whether one should use actual or absolute values of for the paths a and b, and on whether one should Â“trimÂ” the model based on significance testing before estimating this proportion. I used the (c-cÂ’)/c formula here because it is more parsimonious and seems to be less controversial.
39 Following the procedure outlined above, to test the mediation I used (a) the parameter estimates for predicting PA and NA with feedback at level 1 (to estimate a1 and a2), (b) the results for Model 1, which predicted goals with feedback at level 1 (to estimate c), and (3) the results for Model 2, which included PA and NA as withinindividual predictors of goals, in addition to the feedback predictor (to estimate b1, b2, and cÂ’). Again, in order estimate parameters using only within-individual variance, the predictors were centered at the individualsÂ’ means, which removed all the betweenindividual variance in the predictor scores. Model 2 equations are presented below: Level 1: Goalij= 0j + 1j (Fdij) + 2j (PAij) + 3j (NAij) + rij (6) Level 2: 0j = 00 + U0j (7) 1j = 10 + U1j (8) 2j = 20 + U2j (9) 3j = 30 + U3j (10) where Goalij= individual jÂ’s goal for trial i; Fdij= individual jÂ’s feedback for performance on trial i-1; PAij =momentary positive affect score reported by individual j before setting the performance goal for trial i; NAij =momentary negative affect score reported by individual j before setting the performance goal for trial i; 1j= individualsÂ’ slopes for predicting their goal with feedback concerning their previous performance, across time; 2j= individualsÂ’ slopes for predicting their goal with their momentary PA score; 3j= individualsÂ’ slopes for predicting their goal with their momentary NA score; 00=pooled intercept; 10=pooled slope for
40 predicting goals with feedback; 20=pooled slope for predicting goals with PA; 30=pooled slope for predicting goals with NA. Hypothesis 3a and hypothesis 3b . In order to estimate the within-individual parameters for predicting goals with feedback independently for negative and positive feedback (at level 1), I first constructed a Â“feedback signÂ” dummy variable reflecting whether the performance was higher than the goal for the respective trial (positive feedback) or whether is was lower than the goal (negative feedback). Second, following Raudenbush, Brennan, and Barnett (1995), I constructed a pair of dummy variables that allow estimating the within-individual intercepts for the regression of goals on negative and positive feedback independently. These two variables, x_n and x_p, were constructed as x_n = 1, when feedback was n egative, and x_n = 0 otherwise, and x_p = 1, when feedback was p ositive, and x_p = 0 otherwise. Third, in order to independently estimate the slopes for predicting goals with negative and positive feedback independently, I constructed two other variables, Fd_n and Fd_p, that were equal to the value of the feedback when the feedback sign corresponded to the variable name and were set to zero when the sign was opposite (e.g., Fd_n = actual feedback value when feedback was negative, Fd_n = 0 when feedback was positive; Fd_p = actual feedback value when feedback was positive, Fd_p = 0 when feedback was negative). To investigate whether the dispositional variables moderate the within-individual relationship between negative feedback and goals, I estimated Models 3 and 4. Both models regressed, at level 1, individualsÂ’ momentary goals on the two intercept estimating variables (x_n and x_p) and the two slope estimators (Fd_n and Fd_p). At level 2, Model 3 regressed individualsÂ’ slopes and intercepts for predicting goals with
41 negative and positive feedback on neuroticism and extraversion, respectively, and Model 4 regressed these slopes and intercepts on negative and positive affectivity. The equations for Model 3 are shown below; the equations for Model 4 are identical to the Model 3 equations with the exception of the level 2 predictors (negative and positive affectivity instead of neuroticism and extraversion). Level 1: Goalij= 1j(x_nij) + 2j(x_pij) + 3j(Fd_nij) + 4j(Fd_pij) + rij (11) Level 2: 1j = 10 + 11(N) + 12(E) + U1j (12) 2j = 20 + 21(N) + 22(E) + U2j (13) 3j = 30 + 31(N) + 32(E) + U3j (14) 4j = 40 + 41(N) + 42(E) + U4j (15) where Goalij= individual jÂ’s goal for trial i; x_nij = dummy variables equal to 1 when feedback sign was negative and zero otherwise; x_pij = dummy variables equal to 1 when feedback sign was positive and zero otherwise; Fd_nij= individual jÂ’s value of feedback for performance on trial i-1 if such feedback was negative, or zero otherwise; Fd_pij= individual jÂ’s value of feedback for performance on trial i1 if such feedback was positive, or zero otherwise; N= neuroticism; E= extraversion; 1j= individualsÂ’ intercepts for predicting their goal with feedback concerning their previous performance when such feedback was negative, across time; 2j= individualsÂ’ intercepts for predicting their goal with feedback concerning their previous performance when such feedback was positive, across time; 3j= individualsÂ’ slopes for predicting their goal with feedback concerning their previous performance when such feedback was negative, across time; 4j=
42 individualsÂ’ slopes for predicting their goal with feedback concerning their previous performance when such feedback was positive, across time; 10=pooled intercept for predicting goals with negative feedback, controlling for N and E; 20=pooled intercept for predicting goals with positive feedback, controlling for N and E; 30=pooled slope for predicting goals with negative feedback, controlling for N and E; 40=pooled slope for predicting goals with positive feedback, controlling for N and E; 11=the level 2 regression coefficient for predicting individualsÂ’ intercepts from regressing their goals on negative feedback at level 1 with their neuroticism scores; 12=the level 2 regression coefficient for predicting individualsÂ’ intercepts from regressing their goals on negative feedback at level 1 with their extraversion scores; 21=the level 2 regression coefficient for predicting individualsÂ’ intercepts from regressing their goals on positive feedback at level 1 with their neuroticism scores; 22=the level 2 regression coefficient for predicting individualsÂ’ intercepts from regressing their goals on positive feedback at level 1 with their extraversion scores; 31=the level 2 regression coefficient for predicting individualsÂ’ slopes from regressing their goals on negative feedback at level 1 with their neuroticism scores; 32=the level 2 regression coefficient for predicting individualsÂ’ slopes from regressing their goals on negative feedback at level 1 with their extraversion scores; 41=the level 2 regression coefficient for predicting individualsÂ’ slopes from regressing their goals on positive feedback at level 1 with their neuroticism scores; 42=the level 2 regression coefficient for predicting individualsÂ’ slopes from regressing their goals on positive feedback at level 1 with their extraversion scores.
43 Results Means, standard deviations, and inter-correlations for all the dispositional variables measured in the study are presented in Table 3-1. Table 3-1. Means, standard deviations, and intercorrelations for dispositional variables M SD 1 2 3 4 1. Neuroticism 19.51 8.37 1.0 2. Extraversion 32.42 6.54 -.38** 1.0 3. Positive Affectivity 52.47 8.10 -.50** .59** 1.0 4. Negative Affectivity 26.95 9.53 .60** -.21** -.18** 1.0 Notes : M =mean, SD =standard deviation. N = 1,024 ** p < .01 (two-tailed). Table 3-2 presents estimated parameter and variance components for the null model estimated in each sub-sample. The null model analyses indicated that there was significant between-individual variance in goals for each data set (p < .01 for all samples) and that a substantial proportion of the total variance in goal levels was within individuals (34.5, 34.9, 31.8, and 31.2% for the Â‘usesÂ’-real feedback [Sample 1], Â‘usesÂ’-manipulated feedback [Sample 2], RAT-real feedback [Sample 3], and RAT-manipulated feedback [Sample 4] conditions, respectively).1 These results suggest that hierarchical modeling of these data is appropriate. The parameter estimates for Model 1 and 2, computed on each of the four subsamples, are presented in Table 3-4. To standardize the within-individual regression coefficients, I used the within-individual standard deviations of the criterion and the predictor variables presented in Table 3-3. 1 These proportions were computed as 2/( 2+ 00).
44 Table 3-2. Parameter estimates and variance components for the null model Sample/Parameters 00 2 00 Sample 1 61.34** 197.28 373.84** Sample 2 66.55** 138.96 259.20** Sample 3 54.52** 229.98 493.98** Sample 4 64.91** 121.30 267.63** Notes : N =163, 193, 178, and 211, for Samples 1, 2, 3, and 4, respectively. ** p < .01. Table 3-3. Within-individual standard deviations Sample/Variable Goal Feedback PA NA Sample 1 14.05 17.40 6.92 5.13 Sample 2 11.79 13.22 5.91 4.14 Sample 3 15.17 23.00 7.21 5.49 Sample 4 11.01 13.31 5.53 4.31 Notes : PA=positive affect; NA=negative affect. The results for Model 1 show support for the first hypothesis (H1; see Table 3-4). The pooled slope for predicting goal level with feedback was positive and highly significant in each of the four data sets ( 10 = .16, p < .001; 10 = .15, p < .001; 10 = .18, p < .001; 10 = .18, p < .001, for Samples 1-4, respectively Â– these are estimates for the path c, using the notation for testing mediation), which shows that individuals used feedback regarding their previous performance to adjust their goals, as hypothesized. The mediation hypothesis (H2) specified that positive and negative affect mediate part of the relationship between feedback and subsequent goals. Model 1 estimated the magnitude of the within-individual effect of feedback on goals, and, as noted, it showed that feedback did influence goals within individuals. In addition, within individuals, feedback significantly predicted both positive and negative affect (the standardized
45 regression coefficients were * = a1 = .23, p < .01 and * = a2 = -.17, p < .01; * = a1 = .11, p < .01 and * = a2 = -.10, p < .02; * = a1 = .29, p < .01 and * = a2 = -.29, p < .01; * = a1 = .17, p < .01 and * = a2 = -.15, p < .01 for samples 1-4, respectively; these results are not shown in the tables). As noted, to test whether positive and negative affect mediate part of the relationship between feedback and subsequent goals, I estimated Model 2 which included the two affect variables as within-individual predictors (i.e., the variables were centered at the individualsÂ’ means) of goals, in addition to the feedback predictor. Sample 1 . The parameter estimates for Model 2 showed that positive affect did predict goals ( 20 = b1 = .51, p < .001); whereas negative affect did not significantly predict goals (see Table 3-4). The mediating effect of positive affect was highly significant (zÂ’ = 3.80, p < .001), whereas the mediating effect of negative affect was not significant. Comparing the parameter estimates for Model 1 and Model 2, it can be seen that the regression coefficient for predicting goals with feedback decreased from 10 = .16 to 10 = .10, which shows that introducing the two affect variables in the level 1 regression resulted in a 38% reduction in the magnitude of the pooled regression coefficient for predicting goals with feedback. Sample 2 . Positive affect positively predicted subsequent goals ( 30 = b1 =.40, p < .01), and the mediating effect of positive affect was also significant (zÂ’ = 2.89 , p < .001). Comparing the parameter estimates for Model 1 and Model 2 shows that introducing positive and negative affect in the level 1 regression resulted in a 33% decrease in the magnitude of the pooled regression coefficient for predicting goals with feedback.
46 Table 3-4. Parameter estimates and variance components for Model 1 and Model 2 Model and Sample 00 10 10* 20 20* 30 30* 2 00 Model 1 Sample 1 61.34** .16** .20** ----176.74 376.78** Sample 2 66.59** .15** .17** ----126.06 261.09** Sample 3 54.52** .18** .27** ----204.45 497.78** Sample 4 64.92** .18** .22** ----104.68 270.50** Model 2 Sample 1 61.34** .10** .12** .51** .25** .18 .07 136.18 383.10** Sample 2 66.60** .10** .11** .40** .20** -.14 -.05 80.04 268.54** Sample 3 54.53** .11** .17** .51** .24** -.18 -.07 154.98 505.21** Sample 3 64.93** .15** .18** .31** .16** .05 .02 94.94 272.13** Notes : N =163, 193, 178, and 211, for Samples 1, 2, 3, and 4, respectively. ** p < .01, * p < .05. 00 = pooled intercept; 10, 20, 30, 40 = pooled slopes; 20*, 30*, 40* = pooled standardized slopes.
47 Sample 3 . Positive affect positively predicted subsequent goals ( 30 = b1 = .51, p < .01). Comparing the parameter estimates for Model 1 and Model 2 shows that introducing positive and negative affect in the level 1 regression resulted in a 39% reduction in the magnitude of the pooled regression coefficient for predicting goals with feedback. Computing the z-value for the product of the two path coefficients on the mediating path showed that the mediating effect through positive affect was significant (zÂ’ = 4.38, p < .001), whereas the mediating effect through negative affect was not significant. Sample 4 . Positive affect positively predicted subsequent goals ( 30 = b1 = .31, p < .01), and the mediating effect of positive affect was significant as well (zÂ’ = 3.28, p < .001). Comparing the parameter estimates for Model 1 and Model 2 shows that introducing positive and negative affect in the level 1 regression resulted in a 17% reduction in the magnitude of the pooled regression coefficient for predicting goals with feedback. Across the four samples, controlling for feedback, positive affect was a consistent predictor of goals; the effect of negative affect on goals was indeed negative but it did not reach significance in any of the samples. Similarly, the mediating effect through positive affect was statistically significant in each sample, whereas the mediating effect through negative affect was not significant. Introducing the two affect variables as level 1 predictors reduced the magnitude of the regression coefficient for predicting goals with feedback in each of the four samples Â– this reduction ranged from 17% to 39%, with a mean of 32%. I interpret these results as generally supportive of H2, though the mediation effect through was realized mainly through positive affect.
48 Finally, the data did not support the moderation hypotheses (H3a and H3b) which predicted that dispositional characteristics indicating the strength of the inhibition and activation systems will have a cross-level moderating effect on the within-individual effects of negative feedback and on the within-individual effects of positive feedback on future goals, respectively. For Models 3 and 4 (see Table 3-5), 3j and 4j represent the magnitudes of individualsÂ’ reactions to negative and positive feedback, respectively, as reflected in their subsequent goals. According to H3a, neuroticism and negative affect should positively predict the magnitude of 3j across individuals (i.e., at level 2; e.g., those who score higher on neuroticism should decrease their goals to a larger extent than those who score lower on neuroticism, following negative feedback). Similarly, according to H3b, extraversion and positive affect should positively predict the magnitude of 4j across individuals. The parameter estimates for Models 3 and 4 presented in Table 3-5 show that the cross-level effects were weak and inconsistent, and thus no consistent support for H3a and H3b was received from these data. Discussion The results of this study supported the first two hypotheses. First, the present results show that performance feedback does predict goal-regulation within individuals, as predicted by H1. I found evidence for both downward goal revision following negative feedback, and upward goal revision (discrepancy creation) following positive feedback. With respect to the upward goal revision process, these results are consistent with the positive-discrepancy creation arguments of Phillips et al. (1996). The results for both the downward and the upward goal revision processes replicate the results reported by Williams et al. (2000) in their longitudinal study of varsity track and field athletes.
49 Table 3-5. Parameter estimates for Model 3 and Model 4 Model and Sample 10 11 12 20 21 22 30 31 32 40 41 42 Model 3 Sample 1 77.52** -.48** -.15 77.75** -.55* -.28 -.19 .01 .01 .11 .00 .00 Sample 2 76.98** -.29* -.03 69.58** -.50* .04 .35 .00 .01 -.56 .02 .01 Sample 3 59.16** -.21 .09 56.93** -.28 -.11 .36 .00 .00 .29 .00 .00 Sample 4 70.77** -.29 .07 59.02** -.12 .01 .73** .00 -.01* .65* -.01 -.01 Model 4 Sample 1 57.94** -.22 .22 62.78** -.58* .16 .00 .01 .00 .42 .01 -.01 Sample 2 66.72** -.15 .15 59.01** -.46** .28 -.20 .01 .01 .37 .01 -.01 Sample 3 37.03** .05 .38 49.06** -.49 .24 .29 .00 .00 -.19 .01 .00 Sample 4 44.26** .0 .44** 39.64** .10 .37* 1.01** .00 -.01 -.08 .00 .01 Notes : N =163, 193, 178, and 211, for Samples 1, 2, 3, and 4, respectively. ** p < .01, * p < .05
50 (Though its authors have used least squares dummy variable regression analysis [Alliger & Williams, 1993] which is less precise than hierarchical linear modeling, The study reported by Williams et al. was the only study that studied goal-regulation using multilevel methods that I could locate.) Second, and perhaps most importantly, the data collected for this study strongly support the contention that the experience of basic affect is an important mechanism that explains the relationship between feedback and future goals, as specified in H2. That is, in each of four samples, I found that affect (mostly positive affect) mediated a substantial proportion the within-individual relationship between feedback and goals, and the mediating path through positive affect was highly significant in each of the four samples. Here I should point out that all the between-individual variance in the affect variables has been removed from the analyses. Thus, the results cannot be explained by differences between individualsÂ’ propensity to experience positive or negative affect. Though affect and emotions are central concepts in the behavioral regulation theory (e.g., Gray, 1990; Watson, 2000), I am not aware of previous research examining the mediating role of affect in the within-individual goal regulation process. In this respect, the results presented here are unique. With respect to the dispositional moderating effects predicted by the third hypothesis (H3), the data did not support the predictions. To measure personality, I used survey instruments that have been extensively validated in previous research, and participantsÂ’ score on these instruments were reliable in this study. Thus, low reliability of the personality score does not seem to have played a role in the weak and inconsistent cross-level results. Another methodological issue that may be related to the unsupportive
51 cross-level results concerns the power of detecting such cross-level interactions. Whereas the samples of participants were large, the number of trials was relatively small, which may have influenced the precision of estimating the level 1 slopes and thus the power to predict these slopes with level 2 variables. However, whereas a large body of literature on the power of detecting interactions in cross-sectional regression analyses exists (e.g., Aguinis, Pierce, & Stone-Romero, 1994; Aguinis, 2002), methods for estimating power to detect cross-level in multilevel studies have just started to be developed (Davidson, Kwak, Seo, & Choi, 2002), and thus it is difficult to assess whether low power was an issue in the multilevel studies included here. In sum, the results of the first study are important because they replicate previous findings with respect to the within-individual relationship between performance feedback and subsequent goals, and, more importantly, because they offer evidence for the mediating effect of affect in this within-individual relationship. These results show that goal variation within individuals is responsible for a third of the total variance (both within and between individuals) in goals, and Â– more importantly Â– that this withinindividual variance can be explained by with feedback concerning previous performance and with affect. Figure 3-2 shows the partitioning of the variance in goals for Study 1 in withinand between-individual variance, and the proportions of within-individual explained by feedback and affect. The graphical illustration highlights the important role of affect in goal regulation in that not only does affect mediate the effect of previous feedback on subsequent goals, but adding positive and negative affect as withinindividual predictors of goals (in addition to feedback) increases the percentage of within-individual variance explained by 20%! To my knowledge, this is the first
52 empirical attempt to explain the link between feedback, affect and goals across time in a multilevel study that enabled precise partitioning of withinand between-individual variance.
53 Figure 3-2. Partitioning the total variance in goals for Sample 1. Between individuals variance (65.5%) Within individuals variance (34.5%) Feedback (10.4%) Unexplained within individuals variance (69.0%) Affect (20.6%)
54 CHAPTER 4 STUDY 2 The second study is similar to the first study with the following exceptions: (a) the personality survey included a questionnaire designed to measure peopleÂ’s reward and punishment sensitivities (the BIS/BAS scales; Carver & White, 1994), VandeWalleÂ’s (1997) goal orientation measure and the Selves Questionnaire of Higgins et al. (1997), (b) after each of the eight trials of the second phase, participants were given real performance feedback that was either nominal or relative; and (c) the task involved listing words that contain a certain letter (Â“Please write as many words that contain the letter ' a ' as you can, using the text boxes below. Do not include words that start with ' A ' Â”). In this study I employed both relative and nominal feedback to examine whether the two types of feedback have different effects on subsequent goals. Given the role that social comparison processes have in the regulation of motivation (e.g., Bandura & Jourden, 1991), giving feedback that compares respondentsÂ’ performance with the average performance on a specific task (i.e., relative feedback) may have a different impact on motivation than nominal feedback that simply reports performance without providing any means of comparison with othersÂ’ performance. The second study enables testing the mediating role of affect in the withinindividual relationship between feedback and subsequent goals, and also seeks support for the dispositional moderator hypotheses.
55 Method Like the first study, this second study was conducted in two phases. In the first phase, participants were asked to complete a personality survey that included measures of neuroticism, extraversion, positive and negative affectivity, Carver and WhiteÂ’s (1994) BIS/BAS scale, goal orientation, and chronic self-regulatory focus. In the second phase of the study, conducted over the Internet, participants completed an 8-trial experiment that asked them to successively report their current affective state, set a performance goal, and perform a task, for each trial. Participants Participants were 208 undergraduate students from a large introductory course in management at the University of Florida. Participation in the study was completely voluntary and individuals who participated received extra credit points in return for their participation. Experimental Design and Procedure The data for the experimental trials were collected through an electronic interface. Subjects logged on to an Internet site, read a detailed description of the task and procedure, were asked to report their momentary affective state and then to set a goal for the first trial task. There were two different version of the goal setting Web page. For the first version (relative goals), like in the first study, the participants had the option to choose between nine different goal levels, ranging from 10% to 90% (i.e., I want to perform better than 10/90% of the participants in this experiment). The second version asked participants to set nominal goals by estimating the number of words they will be able to generate in the task trial (between zero and 40). After setting a goal for the first
56 trial, participants were presented with the performance task; after submitting their task solutions, they were presented with performance feedback. Participants received real feedback regarding their performance (how they actually performed). Those who set relative goals received relative feedback; those who set nominal goals received nominal feedback. After receiving the feedback, they were asked to report their affect, and then they went on to the next trial. Performance Task As noted, the task involved listing words that contain a certain letter (Â“Please write as many words that contain the letter ' a ' as you can, using the text boxes below. Do not include words that start with ' A ' Â”). Measures Affect . The Positive and Negative Affect Schedule (PANAS; Watson, Clark, & Tellegen, 1988) was used to measure positive affect and negative affect. The internal consistencies reliability of the positive affect scores ranged between .93 and .96 across the eight trials; the reliability of the negative affect sores was between .90 and .94 across the trials. Neuroticism, extraversion, positive and negative affectivity . I used the same measures described in study 1 to measure these dispositional variables. The internal consistency reliabilities for the four construct scores were .85, .75, .77, and .75, for neuroticism, extraversion, positive affectivity, and negative affectivity, respectively. Behavioral inhibition and behavioral approach . I used Carver and WhiteÂ’s (1994) BIS/BAS scales to measure these constructs. The reliabilities of the scores were . 76 and .71 for BIS and BAS, respectively.
57 Self-regulatory focus . IndividualsÂ’ self-regulatory focus (i.e., promotion vs. prevention; Higgins, 1998) was measured with an attitude accessibility measure by tracking response latencies to the self-guide strength measure included in the Selves Questionnaire (e.g., Higgins et al., 1997). Following Higgins et al. (1997), I asked participants to list three attributes of their ideal self and three attributes of their ought self, and rate the extent to which they would ideally be characterized by each attribute and the extent to which they actually posses each attribute. The item (attribute) scores were computed as the sum of the time that it took participants to list the attribute and the time that it took them to rate the extent to which they would ideally posses and actually possess the attribute. For each attribute, the scores were subjected to a logarithmic transformation, and then the Promotion Orientation score was constructed as the sum of the total time recorded for each of the ideal attributes; the Prevention Orientation score was computed as the sum of the total time for each of the ought attributes (both were reverse-scored so that shorter latencies are associated with higher scores). From the participants who completed all the task trials required for this study, 114 provided selfregulatory focus data (data from 20 participants who experienced slow Internet connections, 21 who had to reload at least one of the attitude accessibility measurement pages, and 15 others who had to re-log in the Internet interface for various reasons were excluded). The reliabilities of the Prevention Orientation and Prevention Orientation scales were .65 and .68. Goal orientation . I used VandeWalleÂ’s (1997) scales to measure learning and avoidance (performance) goal orientation. Alphas were .86 and .83 for learning and avoidance goal orientation, respectively.
58 Analyses To test the hypotheses with the data collected in the second study, I modeled the data at two levels: withinand between-individuals. To ascertain whether multi-level modeling is appropriate for these data, I first estimated null models in each sub-sample to investigate whether sufficient within-individual variance in trial goals exists. The analyses testing the first two hypotheses were identical to the analyses conducted to test these hypotheses in the first study. That is, I estimated Model 1 and Model 2 on each of the two sub-samples. The description and equations for Model 1 and Model 2 are included in the method section of the first study, and thus I will not repeat them here. To test the significance of the mediation paths, I used the empirical distribution-based zÂ’ test developed by MacKinnon et al. (2002). To test the cross-level moderation effects, I used five pairs of dispositional variables that indicate the strength of the BIS and the BAS: neuroticism and extraversion, negative and positive affectivity, BIS and BAS scales, prevention and promotion selfregulatory focus, and avoidance and learning goal orientation. To do so, I estimated five models (Model 3 through Model 7) that each included one pair of dispositional variables as level 2 predictors, in each of the two sub-samples. Model 3 and Model 4 were identical with the corresponding models estimated in the first study (see equations 11 through 15, in Chapter 3), Models 5, 6 and 7 had the same structure as Model 3 and Model 4, but included the three new pairs of potential dispositional moderators. Results Means, standard deviations, and inter-correlations for all the dispositional variables measured in the study are presented in Table 4-1.
59 Table 4-1. Means, standard deviations, and intercorrelations for dispositional variables M SD 1 2 3 4 5 6 7 8 9 10 1. Neuroticism 19.77 7.51 1.0 2. Extraversion 31.36 5.62 -.27** 1.0 3. Positive Affectivity 51.85 8.56 -.24** .35** 1.0 4. Negative Affectivity 27.93 10.29 .52** -.18** .20** 1.0 5. BIS Scale 14.64 3.98 .64** -.02 -.12 .37** 1.0 6. BAS Scale 16.34 2.094 .09 .30** .32** .13 .29** 1.0 7. Prevention Orientation 4.43 .54 .13 -.06 -.02 -.02 .11 -.06 1.0 8. Promotion Orientation 3.49 2.17 .15 -.09 .09 -.02 -.10 -.01 .68** 1.0 9. Learning Goal Orientation 27.80 4.68 -.14 .17* -.01 .30 -.05 .25** -.03 -.07 1.0 10. Avoidance Goal Orientation 15.90 5.11 .28** -.05 -.14 .25** .36** .07 -.04 -.02 -.17* 1.0 Notes : M =mean, SD =standard deviation. N = 114-208. ** p < .01, * p < .05 (two-tailed).
60 Table 4-2 presents estimated parameter and variance components for the null model estimated in the two sub-samples (relative performance feedback Â– Sample 5, and nominal performance feedback Â– Sample 6). Results showed there was significant between-individual variance in goals for each data set (p < .01 for both samples) and that a substantial proportion of the total variance in goal levels was within individuals (33.4% and 38.2% for Sample 5 and Sample 6, respectively).1 Thus, hierarchical modeling of these data is appropriate. Table 4-2. Parameter estimates and variance components for the null model Sample/Parameters 00 2 00 Sample 5 74.42** 196.85 392.98** Sample 6 29.81** 65.03 105.16** Notes : N = 77 and 85, for Samples 5 and 6, respectively. ** p < .01. The parameter estimates for Model 1 and 2, computed on each of the two subsamples, are presented in Table 4-4. As I did in the first study, to standardize the withinindividual regression coefficients, I used the within-individual standard deviations of the criterion and the predictor variables. The standard deviations are presented in Table 4-3. The results presented in Table 4-4 show that feedback indeed predicted goals within individuals in both samples (i.e., relative feedback [Sample 5] and nominal feedback [Sample 6]), which supports H1. Feedback did predict both positive and negative affect within individuals, in both samples (the regression coefficients were = .10, p < .01 and = -06, p < .01, and = .28, p < .01 and = -.21, p < .01, for Sample 5 and Sample 6, respectively). The within-individual mediation hypothesis (H2) was also 1 These proportions were computed as 2/( 2+ 00).
61 supported in both samples; the magnitude of the regression coefficient for predicting goals with feedback concerning previous performance decreased by 44% and 29% upon introducing the affect variables in the level-1 regression, for Sample 5 and Sample 6 respectively. The mediating effect through positive affect was significant in both samples (z = 1.98, p < .05 and z = 1.67, p < .05, in Sample 5 and Sample 6, respectively), whereas the mediating effect through negative affect was significant only in Sample 6 (z = 2.02, p < .05). (As in the first study, I computed the z-value for the product of path coefficients with the unbiased estimator formula [Goodman, 1960] and then used the MacKinnon et al.  empirical distribution-based zÂ’ test to examine the significance of the product of path coefficients.) Table 4-3. Within-individual standard deviations Sample/Variable Goal Feedback PA NA Sample 5 14.03 21.32 7.11 4.02 Sample 6 8.06 7.60 7.28 5.02 Notes : PA=positive affect; NA=negative affect. Table 4-5, which displays the parameter estimates for Models 3 through 7, shows several significant cross-level moderator effects. First, in Sample 6, neuroticism was negatively associated with the strength of the within-individual relationship between positive feedback and goals (Model 3: 41 = -.03, p < .01). Whereas this relationship makes sense, the approach-avoidance framework that I used in generating the moderating hypotheses does not predict this effect, and thus it was unexpected. Second, in Sample 5, Positive Affectivity positively predicted the strength of the negative feedback-goals relationship (Model 4: 32=.01, p < .05). Â– an effect that was not predicted.
62 Table 4-4. Parameter estimates and variance components for Model 1 and Model 2 Model and Sample 00 10 10* 20 20* 30 30* 2 00 Model 1 Sample 5 74.42** .18** .27** ----97.41 407.12** Sample 6 29.79** .34** .32** ----39.79 108.81** Model 2 Sample 5 74.42** .10* .15* .28* .14* -.21 -.06 84.90 413.04** Sample 6 29.78** .24** .23** .15* .14* -.28* -.17* 20.23 111.66** Notes : N = 77 and 85, for Samples 5 and 6, respectively. ** p < .01, * p < .05. 00 = pooled intercept; 10, 20, 30, 40 = pooled slopes; 20*, 30*, 40* = pooled standardized slopes
63 Table 4-5. Parameter estimates for Model 3 through Model 7 Model and Sample 10 11 12 20 21 22 30 31 32 40 41 42 Model 3 (N and E) Sample 5 86.23** -.21 -.22 70.88* .17 -.17 .91 -.02 -.01 .25 .03 -.03 Sample 6 28.90** .06 -.08 20.23** .16 .25 1.14* -.02 .00 1.18* -.03** -.01 Model 4 (NA and PA) Sample 5 77.29** -.01 -.04 45.42* -.42 .71 -.04 -.01 .01* -.62 .00 .01 Sample 6 28.13** .10 -.06 31.85** -.02 .00 .33 .00 .01 -.95 .00 .02* Model 5 (BIS and BAS) Sample 5 52.90** -.58 1.87 151.96* -.37 -4.79 .04 -.04* .06 3.15 .01 -.20 Sample 6 29.36** .04 -.14 23.11* .13 .40 1.18 -.01 -.01 .23 .04 -.04 Model 6 (Prev. and Prom.) Sample 5 62.70** 1.31 -4.02 124.58 37.70 -22.70 .53 -.08 .12 2.70 .81* .14 Sample 6 6.10 -2.19 -2.92 21.98 2.54 -4.61 1.44 .14 .00 1.00 .20 -.02 Model 7 (LGO and AGO) Sample 5 55.53** .86 -.30 50.84 .93 -.32 .48 .00 -.01 1.20 -.04* .00 Sample 6 26.16** .17 -.19 26.49* .34 -.27 .28 .03 -.01 -.32 .00 .03 Notes : N = 77 and 85, for Samples 5 and 6, respectively. ** p < .01, * p < .05. For each model, the cross-sectional moderator variables are listed in parentheses. N=neuroticism, E=extraversion, PA=positive affectivity, NA=negative affectivity, BIS=Carver and Whit eÂ’s (1994) BIS scale, BAS=Carver and WhiteÂ’s (1994) BAS scale, Prev.=HigginsÂ’ (1997) prevention self-regulatory focus, Prom.= HigginsÂ’ (1997) promotion self-regulatory focus, LGO=learning goal orientation, AGO=avoidance goal orientation.
64 Third, as expected, Positive Affectivity did predict the magnitude of the positive feedback-goals relationship in Sample 6 (Model 4: 42=.02, p < .05). Fourth, BIS predicted the magnitude of the negative feedback-goals relationship in Sample 5 (Model 4: 31= -.04, p < .05), but this effect was in the opposite direct from the effect predicted by H3a. Fifth, HigginsÂ’ (1997) prevention self-regulatory focus, measured as individualsÂ’ accessibility of Â‘oughtÂ’ attributes, predicted the slope of the positive feedback-goals relationship in Sample 5 and (Model 6: 41=.81, p < .05). This effect was not predicted. Finally, contrary to H3b, learning goal orientation negatively predicted the magnitude of the positive feedback-goals relationship in Sample 5 (Model 7: 41= -.04, p < .05). Discussion Like the first study, the second study offered support for the hypothesized relationship between feedback and subsequent goals (H1), and for the mediating role of affect in this relationship (H2). That is, in within-individual analyses conducted in both samples, affect mediated a substantial proportion of the effect of feedback on subsequent goals. The mediating effect through positive affect was significant in both samples, whereas the mediating effect through negative affect was significant only in Sample 6. The third hypothesis, which specified the dispositional moderator effects on the within-individual relationship between feedback and subsequent goals was not supported by these data. Though the analyses detected some cross-level moderator effects, the support for the hypothesized effects was weak and inconsistent across samples. Kluger and DeNisi (1996) proposed that Â“in the absence of learning cues, the fewer cognitive resources needed for task performance, the more positive is the effect of FI [feedback interventions] on performanceÂ” (p. 269). Taken together, the first two
65 studies employ three different tasks: the Â‘usesÂ’ task, the RAT, and listing words containing a certain letter. It can be argued that these tasks have different cognitive requirements and thus have different levels of difficulty (the RAT being the most difficult, and listing of words containing a certain letter being the least difficult). Thus, by comparing the magnitude of the within-individual relationships among feedback, affect, and goals across the three tasks, I can examine, on an exploratory basis, whether Kluger and DeNisiÂ’s proposition can be verified within the framework of the present studies. Because the Â“wordsÂ” task used in the second study is the simplest of the three (compared with the Â“RATÂ” and the Â“usesÂ”), I expected the magnitude of the feedback-subsequent goals relationship to be the largest for the Â“wordsÂ” task. The data did support this expectation; the average magnitude of the feedback-goals relationship across the two samples of participants who performed the words task was .30, which is larger than both the average magnitude for the uses task (.18) and the average magnitude across individuals who performed the RAT task (.25). The mediation effect was also stronger for the words task (average of 36.5% vs. 35.5% and 28.0% for the uses and the RAT, respectively), but the difference was rather small.
66 CHAPTER 5 STUDY 3 This study was conducted over the course of a semester, and it involved undergraduate students as respondents. The study was designed to simultaneously examine the role of emotional reactions and self-efficacy in mediating the effect of performance feedback on subsequent goals, and to seek support for the moderator hypotheses. The second hypothesis (H2) specifies that individualsÂ’ affective, or emotional reactions to feedback will mediate the relationship between feedback and subsequent goals. In the introduction, I acknowledged that self-efficacy judgments can mediate the effect of emotions on goals and can also mediate part of the feedback-goals relationship independently of affect (see Figure 2-2). In this study, I will test both whether emotional reactions mediate part of the feedback-goals relationship (H2), and whether emotions and self-efficacy have independent mediating roles in this process. Hypothesis 3 specifies that dispositional constructs reflecting chronic activation levels of the BIS and the BAS (neuroticism and extraversion, and BIS and BAS scales, in this study), will moderate the relationship between feedback (goal discrepancy feedback in this study) and future goals. This effect reflects individual differences in sensitivity to negative and positive feedback. Because behavioral motivation theory actually specifies that individualsÂ’ sensitivities to rewards and punishments (BIS and BAS) affect their positive and negative emotions differentially (i.e., the BIS primarily influences negative emotions and BAS primarily influences positive emotions; Gray, 1990; Watson, 2000),
67 the moderating effect on goal setting should actually be due to the moderating effect of BIS/BAS on the feedback-emotional reactions relationship. It follows that neuroticism and BIS should moderate the impact of feedback on individualsÂ’ negative emotional reactions to feedback, and extraversion and BAS should moderate the effect of feedback on positive emotional reactions to feedback. In this study, I will examine the moderating role of dispositions both on the feedback-subsequent goals and on the feedbackemotional reactions relationships. The fourth and fifth hypotheses predict that (a) goal commitment and (b) general task self-efficacy will moderate the impact of negative feedback on subsequent goals. I will use the data collected for this study to test these hypotheses, and also to examine the moderating role of these constructs on the negative feedback-negative emotional reactions relationship. Method Participants Four hundred and ninety three undergraduate business students from a large introductory class participated in this study. Their participation was rewarded with course extra credit. Procedure The study had two phases. In the first phase, which took place at the beginning of the semester, respondents completed a personality surveys and a measure of academic self-efficacy, and they were asked to set an overall grade goal for their performance in the course and to complete a measure of goal commitment for their overall grade goal.
68 The second phase was conducted over the course of the semester. The class involved three mid-term examinations and a final exam; before each exam, participants were asked to set a goal for their performance in the exam. For this course, exam grades were posted on the course Web page. After checking their grade, respondents were shown their initial goal for the specific exam, and then were asked to report their emotional reactions to the exam performance. Following Bagozzi, Blaumgartner, and Pieters (1998), respondents were asked to rate the extent to which they experience a list of emotions in response to the question: Â“As a result of your performance on this exam relative to your goal, to what extent do you experience each of the following emotions?Â” The list of emotions comprised the emotion terms used by Bagozzi et al. (excited, delighted, happy, glad, satisfied, proud, self-assured, angry, frustrated, guilty, ashamed, sad, disappointed, depressed, worried, uncomfortable, fearful) to measure positive and negative emotional reactions. After reporting their emotional reactions to the performance feedback, respondents were asked to set their goal and report their selfefficacy for the next examination. Measures Neuroticism and extraversion . I used the same measures described in study 1 to measure these two personality constructs. The internal consistency reliabilities for the construct scores were .85 and .81, respectively. Behavioral inhibition and behavioral approach . I used Carver and WhiteÂ’s (1994) BIS/BAS scales to measure these constructs. The reliabilities of the scores were .77 and .68 for BIS and BAS, respectively.
69 Overall course goal . I asked students to set an overall goal for their grade in the course, ranging from C= 1 to A = 8 . Goal commitment . I used Hollenbeck, Willliams and KleinÂ’s (1989) 9-item measure of grade goal commitment, using a 1-7 scale ( 1 = Not at all, 7 = Completely ). The measure is shown in Appendix D. The internal consistency of the scores was .73. Academic self-efficacy . I followed Wood and Locke (1987), and constructed a measure assessing studentsÂ’ confidence in their ability to perform in the seven task areas identified by Wood and Locke (i.e., class concentration, memorization, exam concentration, understanding, explaining concepts, discriminating concepts, and notetaking). The measure is shown in Appendix E. Participants were asked to indicate their agreement with the item statements, on a 1-7 scale ( 1 = Strongly disagree, 7 = Strongly agree ). The internal consistency of the academic self-efficacy score was .79. Exam goals . Before each exam, students were asked to set a goal for the specific exam, in term of the number of questions answered correctly (all exams had 50 questions). Exam self-efficacy . I measured exam self-efficacy with a magnitude-strength survey (Lee & Bobko, 1994; Wood & Locke, 1997). Before each exam, I asked students to indicate whether they expected to achieve each of 7 levels of performance in the exam (i.e., answer at least 20 questions correctly, at least 25, 30, 35, 40, 45 questions, and answer all 50 questions correctly) and how confident they were of attaining each level. To construct the scores, I summed the confidence rating for the levels of performance that students indicated they could achieve.
70 Performance feedback . As noted, after each mid-term exam, students were presented with the goal that they have previously set for the respective exam and with their actual exam score, and then they were asked to indicate how they feel about their performance relative to their goal. That is, the performance feedback that the participants were asked to react to was the discrepancy between actual performance and goal (goal discrepancy feedback, or GDF; Vance & Colella, 1990). In consequence, for each exam, I computed a GDF score by subtracting the goal score from the actual performance score. Thus, positive GDF scores show that the exam goal has been exceeded, null GDF scores show that the goal has been met, and negative GDF scores showed that performance felt short of the goal for the specific exam. Emotional reactions to feedback . As noted, after they were presented with their exam goals and actual performance and asked to compare the two, participants were asked to rate the extent to which they experience the emotions included in the list used by Bagozzi et al. (1998) in response to the GDF information. I combined the positive terms to form the Positive Emotional Reaction (PER) scale and the negative terms to form the Negative Emotional Reaction (NER) scale. These scales are very similar to the positive and negative affect scales from the PANAS (Watson et al., 1998), but they are specifically focused on the comparison of performance with goals. Across the three measurements of emotional reactions, the internal consistency reliabilities of the PER scores were .95, .96 and .97; the reliabilities of the NER sores were .93, .95, and .95 across the three measurements.
71 Analyses I used path modeling and regression analysis on observed variables to analyze the data collected for this study. Even though these types of analysis do not permit such rigorous partitioning of withinand between-individual variance as multilevel methods, the temporal ordering and the nature of the constructs included in this study do allow causal inferences. To test whether GDF influences goal setting (H1) and whether emotional reactions to feedback (H2) and exam self-efficacy mediate this relationship, I estimated two path models on each of two data sets. The two data sets consisted of (a) the discrepancy feedback for Exam 1, the emotional reactions to this feedback, and the selfefficacy and goals for Exam 2, and (b) the discrepancy feedback for Exam 2, the emotional reactions to this feedback, and the self-efficacy and goals for Exam 3. I did not use the Exam 4 self-efficacy and goal data because about 43% of the students who participated in this study (and received discrepancy feedback regarding their performance in Exam 3) did not take Exam 4 (they had this option in the course) and thus their goal and self-efficacy ratings for Exam 4 may have been fictitious.1 Model 1 specifies that GDF influences exam self-efficacy and exam goals exclusively through emotional reactions to feedback (PER and NER). The emotional reactions, in turn, influence exam goals both directly, and indirectly through self-efficacy. Figure 5-1 presents this model below. 1 An alternative would have been to use only the data provided by students who did take Exam 4; I decided not to use this reduced sample because it has been subject to a self-selection bias: those who did not take Exam 4 likely elected to do so because they had met their goals.
72 Figure 5-1. Path Model 1 Model 1 can give support to the mediation hypothesis (H2) if the fit of the path model is acceptable and if the path coefficients on the mediating paths (e.g., GDF to NER to goals) are significantly different from zero. Model 1 actually represents a fully mediated process, in that the impact of GDF on self-efficacy and goals is fully mediated by emotional reactions to feedback. To examine whether GDF also has a direct effect on self-efficacy and goals, I estimated Model 2. Model 2 adds direct effects from GDF to exam self-efficacy and from GDF to exam subsequent goals to the paths specified in Model 1. Model 2 is shown in Figure 5-2. Because Model 2 is a saturated model, it will fit the data perfectly and thus I will not be able to assess whether emotional reactions fully, or partly, mediated the effect of goal discrepancy feedback on self-efficacy and goals by examining fit indices. The magnitudes of the direct paths, however, will indicate whether the effect of feedback on self-efficacy and goal is partly or fully mediated by emotional reactions. Finally, to test the moderating effect of the dispositional constructs (neuroticism and extraversion and BIS and BAS) and of goal commitment and academic self-efficacy on the relationship between GDF and goals and on the relationship between GDF and Goal Discrepancy Feedback Positive Emotional Reactions Negative Emotional Reactions Self-Efficacy Subsequent Goal
73 emotional reactions to feedback, I conducted multiple moderated regression analyses for each of the two exams for which I had the required data for these analyses. Figure 5-2. Path Model 2 Results Table 5-1 presents the means, standard deviations, and intercorrelations among variables measured in this study. The overall grade goal predicted exam goals moderately strongly (r =.48 and r =.41, for Exam 2 and Exam 3, respectively, p < .01 for both), as did academic self-efficacy (r =.27 and r =.23, for Exam 2 and Exam 3, respectively, p < .01 for both) and the magnitude of the academic self-efficacy-exam goals relationship was smaller. The goal discrepancy feedback (GDF) strongly influenced both positive emotional reactions (PER; r =.52 and r =.59, for Exam 2 and Exam 3, respectively, p < .01 for both) and negative emotional reactions (NER; r =-.46 and r =-.50, for Exam 2 and Exam 3, respectively, p < .01 for both). Positive Emotional Reactions Negative Emotional Reactions Self-Efficacy Subsequent Goal Goal Discrepancy Feedback
74 Table 5-1. Means, standard deviations, and intercorrelations for the study variables M SD 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 1. Neuroticism 18.67 7.34 1.0 2. Extraversion 32.40 6.26 -.41 1.0 3. BIS Scale 17.40 4.19 .64 -.17 1.0 4. BAS Scale 16.68 2.12 -.08 .40 .14 1.0 5. Overall Grade Goal 7.52 0.87 -.04 .11 .00 .09 1.0 6. Goal Commitment 55.59 5.86 -.24 .20 -.13 .15 .14 1.0 7. Academic Self-Efficacy 37.59 5.89 -.27 .21 -.15 .19 .27 .39 1.0 8. GDF after Exam 1 -7.24 5.63 .17 -.11 .16 -.08 .00 .01 .17 1.0 9. NER after Exam 1 21.32 14.80 .13 .04 .11 .09 -.09 -.04 -.21 -.46 1.0 10. PER after Exam 1 10.20 10.15 .06 .01 .07 .02 .10 .06 .21 .52 -.56 1.0 11. Exam 2 Self-Efficacy 527.93 108.85 -.16 .05 -.12 -.02 .34 .23 .32 .20 -.33 .24 1.0 12. Exam 2 Goal 44.49 3.24 -.09 .05 -.08 .07 .48 .27 .27 .08 -.22 .14 .53 1.0 13. GDF after Exam 2 -6.47 5.67 .10 .00 .10 -.04 .04 -.04 .10 .38 -.11 .20 -.04 -.19 1.0 14. NER after Exam 2 22.43 16.58 .17 -.02 .17 .06 -.08 -.04 -.17 -.19 .49 -.22 -.19 -.13 -.50 1.0 15. PER after Exam 2 11.47 11.47 -.03 .03 .01 .02 .04 .02 .10 .18 -.14 .32 .10 .08 .59 -.59 1.0 16. Exam 3 Self-Efficacy 517.62 112.79 -.12 .14 -.04 .08 .31 .20 .30 .20 -.27 .21 .65 .42 .18 -.29 .27 1.0 17. Exam 3 Goal 44.15 3.82 -.05 .09 -.04 .08 .41 .23 .25 .14 -.20 .19 .44 .64 .10 -.19 .22 .55 1.0 Notes : M =mean, SD =standard deviation. N = 493. For r " .09 " , p < .05; for r " .13 " , p < .01; tests are two-tailed. GDF=goal discrepancy feedback; NER=negative emotional reactions, PER=positive emotional reactions.
75 As expected, GDF was positively associated with subsequent goals for both exams (r =.08 and r =.10, for Exam 2 and Exam 3, respectively) though this association was only marginally significant for Exam 2 (p < .07; it was significant at p < .05 for Exam 3). Mediating Mechanisms Table 5-2 presents the fit indices for fitting Model 1 to each of the two data sets (Exam 2 and Exam 3). As can be seen in Table 5-2, the fit was excellent, which indicates that Model 1, which specifies that emotional reactions to feedback fully mediate the effect of goal discrepancy feedback on subsequent self-efficacy and goals, represents the structure of the data rather well. Table 5-2. Fit indices for Model 1 Exam dof #2 RMSEA RMR CFI GFI RFI Exam 2 2 1.71 .00 .01 1.0 1.0 .99 Exam 3 2 2.07 .01 .01 1.0 1.0 .98 Notes . dof= degrees of freedom, RMSEA=root mean square of approximation, RMR=standardized mean residual, CFI=comparative fit index, GFI=goodness of fit index, RFI=relative fit index. Table 5-3 shows the standardized path estimates for Model 1 in each of the two data sets. As expected, GDF strongly influenced participantsÂ’ emotional reactions ( $ = .52 and $ =-.46 for positive and negative reactions in the Exam 2 data, and $ = .59 and $ =-.50 for positive and negative reactions in the Exam 3 data; all coefficients were significant at p < .01). In turn, emotional reactions did influence self-efficacy ( $ = .07, p < .05 and $ =.15, p < .01, for positive reactions in the Exam 2 and Exam 3 samples, respectively, and $ = -.29, p < .01 and $ =-.20, p < .01, for negative reactions in the Exam 2 and Exam 3 samples, respectively), and self-efficacy further influenced goals ( $ = .51, p < .01 and $ = .53, p < .01 in the Exam 2 and Exam 3 samples, respectively). The
76 effect of positive emotional reactions on subsequent goals was fully mediated by selfefficacy in the Exam 2 data (i.e., the direct path from Positive Emotional Reactions to goals was small and not significant) but positive emotional reactions did show a direct effect in the Exam 3 data ( $ = .09, p < .01). For negative emotional reactions this pattern was reversed, in that they showed a direct effect in the Exam 2 ( $ = -.07, p < .05) data but not in the Exam 3 data. Table 5-3. Standardized parameter estimates for Model 1 Path Exam 2 Exam 3 GDF to PER .52** .59** GDF to NER -.46** -.50** PER to Self-Efficacy .07* * .15** PER to Exam Goal -.02 ** .09** NER to Self-Efficacy -.29** -.20** NER to Exam Goal -.07* * .02 ** Self-Efficacy to Exam Goal .51** .53** Notes . N=493, ** p < .01, * p < .05. GDF=goal discrepancy feedback; NER=negative emotional reactions, PER=positive emotional reactions. The parameter estimates for Model 2, presented in Table 5-4, show that the direct effects from goal discrepancy feedback to subsequent exam self-efficacy and exam goals were small and did not approach significance. This pattern of results suggests that the effect of goal discrepancy feedback on subsequent self-efficacy and goals is fully mediated by emotional reactions to the feedback.
77 Table 5-4. Standardized parameter estimates for Model 2 Path Exam 2 Exam 3 GDF to PER .52** .59** GDF to NER -.46** -.50** GDF to Self-Efficacy .04 ** -.01 ** GDF to Exam Goal -.06 ** -.06 ** PER to Self-Efficacy .06 ** .16** PER to Exam Goal .00 ** .12** NER to Self-Efficacy -.28** -.20** NER to Exam Goal -.08* * .00 ** Self-Efficacy to Exam Goal .52** .53** Notes . N=493, ** p < .01, * p < .05. GDF=goal discrepancy feedback; NER=negative emotional reactions, PER=positive emotional reactions. Though the goal discrepancy feedback variable is based on theory (Vance & Colella, 1990) and such discrepancy variables have been often used in longitudinal goal setting research (e.g., Phillips et al., 1996; Vance & Colella, 1990; Williams, et al., 2000), the use of difference score variables has been heavily criticized in recent years (e.g., Edwards, 2001; Johns, 1981). Such criticisms have typically focused on the use of difference scores to express congruence or incongruence in areas such as the personenvironment fit research (e.g., Edwards, 1996; Edwards & Parry, 1993; Edwards & Van Harrison, 1993). Using difference scores to measure fit reflects what Johns (1981) called the within-person discrepancy theme when Â“an individualÂ’s reaction to the organizational environment is in part a function of the discrepancy between how this environment is perceived to be and how it should beÂ” (Johns, 1981, p. 444). The alternative to using differences scores as predictors in regression analyses proposed by Edwards and his associates (e.g., Edwards & Parry, 1993) is the use of
78 polynomial regression equations that use the variables that compose the difference score as independent predictors in the regression. However, difference scores also have their defenders (e.g., Tisak & Smith, 1994a; 1994b; also see Edwards, 1994). Tisak and Smith argue that variables measured by difference scores may reflect theoretically rich constructs, and the appropriateness of using such difference score measures should be assessed in each study. I believe the goal discrepancy measure used in this study does reflect such a theoretically rich construct, and I attempt to assess the appropriateness of using such a measure below. Conceptually, I chose to use the goal discrepancy measure primarily because evaluating performance against goals is central to both self-efficacy and control theories of motivation (Phillips et al., 1996). That is, individuals evaluate their performance against their goals and then use the evaluative information to regulate their behavior. Following this self-regulation principle, in this study, I specifically asked respondents to report their emotional reactions to their performance relative to their goal . From a methodological standpoint, three main criticisms have been leveled at the use of difference scores as predictors: (a) difference scores have potentially low reliability, (b) difference score measures potentially lack validity, and (c) predictions based on difference scores cannot be accurately interpreted.2 As noted, difference score proponents such as Tisak and Smith (1994a, 1994b) suggest that researchers should evaluate these potential shortcomings of differences score measures empirically. With respect to the potential low reliability, such effect occurs when the component variables are highly correlated (e.g., Johns, 1981). For this study, the component scores used to 2 Other problems, such as inadequate consideration of the direction of differences (Johns, 1981) or the use of quadratic terms, are not relevant to the research presented here.
79 compute the GDF variable were not strongly correlated (r =.22 and r =.36, for Exam 1 and Exam 2, respectively), which suggest that the reliability of the difference score is not much lower than the reliabilities of the components. With regard to the potential lack of validity of difference scores, Johns (1981) notes that one or both components of a difference score likely predict the criterion much better than the difference score variable, and because of this effect using component scores as independent predictors should be preferred, rather than using their difference as a predictor. In the data collected for this study, previous performance indeed predicted emotional reactions somewhat better than goal discrepancy, but the difference was very small (average magnitude of .55 vs. .52 across positive and negative emotional reactions and the two exams). In contrast, previous goals predicted emotional reactions much less strongly than goal discrepancy (average magnitude of .06 vs. .52). Finally, the criticism concerning the difficulty in interpreting results based on difference score predictors refers to the confounding of the effects of component measures. That is, if a measure reflecting the difference score between scores on measures A and B (A-B) is positively associated with the score on a measure C, it is unclear whether the relationship actually reflects a positive association between A and C, a negative association between B and C, or both. For the specific investigation presented here, as noted above, it appears that, statistically, the strong relationship between exam performance and emotional reactions to feedback is primarily responsible for the association between goal discrepancy feedback and emotional reactions to the feedback (i.e., exam performance was strongly related to emotional reactions; whereas exam goal was not).
80 To examine the mediating effect of emotional reactions for the relationship between feedback and subsequent self-efficacy and goals when feedback is constructed simply as the performance on the previous exam (and not as the goal-discrepancy), I ran Model 3 and Model 4, which are identical to Model 1 and Model 2, with the exception that previous exam performance was used as feedback instead of GDF. These two models are depicted in graphical form below: Figure 5-3. Path Model 3 Figure 5-4. Path Model 4 Previous Performance Positive Emotional Reactions Negative Emotional Reactions Self-Efficacy Subsequent Goal Positive Emotional Reactions Negative Emotional Reactions Self-Efficacy Subsequent Goal Previous Performance
81 Model 3 did not fit any of the two data sets well (e.g., RMSEA=.31 and .28; RMR=.11 and .09; CFI = .86 and .89, for Exam 2 and Exam 3, respectively), and thus it was clear that emotional reactions do not fully mediate the impact of previous performance on self-efficacy and goals. The parameter estimates for Model 4 are presented in Table 5-5 below (like Model 2, Model 4 was saturated, the fit was perfect so I do not present fit indices). As expected (given the poor fit of Model 3) the direct paths from previous performance to self-efficacy and goals were positive and significant, which shows that previous performance influences self-efficacy and goals both directly and indirectly, through emotional reactions to feedback. In addition, comparing the magnitude of the paths from previous performance and from emotional reactions to selfefficacy reveals that performance was a much stronger predictor of self-efficacy than were emotional reactions, which is consistent with Mitchell, Hopper, Daniels, GeorgeFalvy, and JamesÂ’ (1994) findings (in that study, participants reported they consider past performance more strongly than current mood when forming self-efficacy judgments). Moderating Effects As noted, in this study, I examine the moderating role of dispositional factors reflecting individualsÂ’ chronic activation of their BIS and BAS, and the moderating role of goal commitment and academic self-efficacy on (a) the relationship between feedback and subsequent goals, and (b) the relationships between feedback and emotional reactions.
82 Table 5-5. Standardized parameter estimates for Model 4 Path Exam 2 Exam 3 Previous Performance to PER .56** .61** Previous Performance to NER -.52** -.54** Previous Performance to Self-Efficacy .40** .34** ** Previous Performance to Exam Goal .35** .35** ** PER to Self-Efficacy -.08* * .00 ** PER to Exam Goal -.15** -.05 ** NER to Self-Efficacy -.17** -.10** NER to Exam Goal .01 ** .10** ** Self-Efficacy to Exam Goal .42** .45** Notes . N=493, ** p < .01, * p < .05. NER=negative emotional reactions, PER=positive emotional reactions. Dispositional moderators For the first exam, 92% of the participants (454 out of 493) did not achieve their exam goals (GDF was negative for 92% of the participants; see the negative mean GDF score in Table 5-1). For Exam 2, 85% of the participants (420 out of 493) performed at a level lower than their goal. These results are consistent with previous findings. Vance and Colella (1990), for example, note that Â“closer examination of our data reveals that personal goals remained consistently higher, on average, than performance.Â” Similarly, Williams et al. (2000) noted that Â“a significant majority of athletes set initial goals that were above their previous best performance, and in virtually every (98%) instance during the season they went into competition with goals that were above their best performance at that point in the season.Â” Because the large majority of participants received negative goal discrepancy feedback, the dispositional moderator analyses focused on examining the moderating
83 effect of neuroticism and BIS on the negative feedback-goals and negative feedbacknegative emotional reactions relationships. To do so, I conducted the multiple moderated regression analyses only on the cases with negative discrepancy feedback.3 Table 5-6 shows the results of the moderated regression analyses with exam goals and negative emotional reactions as criteria, and GDF, neuroticism, and their interaction as the predictors. Table 5-7 presents the results of the moderated regressions with BIS as the moderating variable. Table 5-6. Dispositional moderator analyses results for neuroticism Dependent variable Exam 2 Exam 3 Goal NER Goal NER $ $ $ $ Step 1 GDF .14** -.46** .10* * -.39** Neuroticism -.11* * .23** -.10* * .25** R2 .026** .228** .019* * .198** Step 2 GDF .14** -.46** .10* * -.39** Neuroticism -.11* * .23** -.10* * .24** GDF X Neuroticism .03 ** -.12** -.03 ** -.07 ** % R2 .001 ** .014** .001 ** .005 ** Total R2 .027** .242** .020* * .203** Notes . N =454 and N =420 for the Exam 2 and Exam 3 analyses, respectively, ** p < .01, * p < .05. GDF=goal discrepancy feedback; NER=negative emotional reactions before the respective exam. 3 A more rigorous but less parsimonious approach is to model the effects of negative and positive feedback and the moderator effects independently in the same regression; such analyses revealed the same results for negative feedback and basically no moderator effects for the positive feedback-goals and positive feedbackpositive emotional reactions relationships.
84 The data offered limited support for the moderating role of neuroticism for the feedback-goals and feedback-negative emotional reactions relationships. That is, in only one of the four regressions did the interaction between neuroticism and feedback significantly predicted the criterion (i.e., it predicted negative emotional reactions before Exam 2). In contrast, the moderating effect of the BIS score received stronger support, in that the interaction between BIS and feedback significantly predicted negative emotional reactions in both data sets, and it also predicted exam goals for Exam 3. Table 5-7. Dispositional moderator analyses results for BIS Dependent variable Exam 2 Exam 3 Goal NER Goal NER $ $ $ $ Step 1 GDF .13** -.44** .10* * -.39** BIS -.07 ** .20** -.06 ** .24** R2 .019* * .216** .013 ** .197** Step 2 GDF .13** -.45** .11* * -.40** BIS -.07 ** .19** -.05 ** .24** GDF X BIS .05 -.18** .12* * -.08Â†* % R2 .002 ** .033** .013* * .007Â†* Total R2 .021** .249** .026* * .204** Notes . N =454 and N =420 for the Exam 2 and Exam 3 analyses, respectively, ** p < .01, * p < .05, Â† p < .10. GDF=goal discrepancy feedback; BIS=Carver and WhiteÂ’s (1994) BIS scale; NER=negative emotional reactions before the respective exam. Situational moderators As shown in the results table below, the data failed to support the moderating role of goal commitment on the impact of goal discrepancy feedback on either subsequent
85 goals or negative emotional reactions, in both data sets (see Table 5-8). Thus, H4 was not supported. The moderating effect of academic self-efficacy received also was not supported; with the exception of one significant interaction term for predicting Exam 2 goals with the product of goal discrepancy feedback and academic self-efficacy ( $ =.12, p < .01; see Table 5-9), no interaction term had a significant effect on the dependent variables. Table 5-8. Moderator analyses results for goal commitment Dependent variable Exam 2 Exam 3 Goal NER GOAL NER $ $ $ $ Step 1 GDF .12* * -.42** .10* * -.37** GC .25** -.03 ** .24** -.07 ** R2 .076** .176** .065** .143** Step 2 GDF .12* * -.42** .10* * -.37** GC .25** -.03 ** .24** -.08 ** GDF X GC -.01 ** .04 ** -.04 ** .03 ** % R2 .000 ** .002 ** .001 ** .001 ** Total R2 .076** .178** .066** .144** Notes . N =454 and N =420 for the Exam 2 and Exam 3 analyses, respectively, ** p < .01, * p < .05, Â† p < .10. GDF=goal discrepancy feedback; GC=goal commitment; NER=negative emotional reactions before the respective exam, PER=positive emotional reactions before the exam. The significant effect detected by the moderator analyses involving academic selfefficacy was actually in the opposite direction from the direction predicted by H5 (see Table 5-9). That is, the significant interaction shows that the relationship between
86 feedback and subsequent goals is stronger for those with higher academic self-efficacy. Because all the discrepancy feedback was negative, a stronger relationships means a larger degree of downward goal adjustment following more negative feedback, which is opposite from the H5 prediction (which proposed that those with higher academic selfefficacy will be less likely to makes substantial downward goal adjustments following negative feedback). Table 5-9. Moderator analyses results for academic self-efficacy Dependent variable Exam 2 Exam 3 Goal NER GOAL NER $ $ $ $ Step 1 GDF .08Â† * -.39** .06 ** -.35** ASE .24** -.15** .23** -.14** R2 .071* * .196** .062** .157** Step 2 GDF .09* * -.40** .06 ** -.35** ASE .26** -.15** .24** -.14** GDF X ASE .12** -.04 ** .05 ** .00 ** % R2 .014** .002 ** .002 ** .000 ** Total R2 .085** .198** .064** .157** Notes . N =454 and N =420 for the Exam 2 and Exam 3 analyses, respectively, ** p < .01, * p < .05, Â† p < .10. GDF=goal discrepancy feedback; ASE=academic self-efficacy; NER=negative emotional reactions before the respective exam, PER=positive emotional reactions before the exam. In sum, I interpret the results regarding the moderating effect of goal commitment and academic self-efficacy as not supportive of the last two hypotheses (H4 and H5).
87 Discussion The third study has several important findings that clarify the processes studied in this dissertation. First, the results offered support for the mediation hypothesis (H2) with a different design, different performance task, and different time frame for the longitudinal data collections. Second, the results of the mediation analyses showed that self-efficacy plays an important role in the goal-regulation process, in that the effect of emotional reactions to goal discrepancy feedback was partly mediated by exam self-efficacy. It is important to note, however, that self-efficacy beliefs did not fully mediate the emotional reactionsfuture goals relationships, which shows that affective processes (emotional reactions) and cognitive processes (self-efficacy judgments) have independent effects on goal regulation (i.e., in the Exam 2 data set negative emotional reactions had an direct effect on goals independent of self-efficacy [r=-.07, p < .05] and in the Exam 3 data positive emotional reactions had such independent effect on goals [r=.09, p < .01]). Third, the results of this study, in general, offered moderate support for the dispositional moderating hypothesis with respect to negative feedback (H3a). This hypothesis specified that neuroticism and the BIS scale should moderate the effect of negative discrepancy feedback on (a) goals and (b) negative emotional reactions. The moderation effect seems to be stronger for the negative feedback-negative emotional reactions relationship, which is in line with behavioral motivation theory (e.g., Gray, 1990). In addition, support was stronger for the moderating effect of the BIS scale than it was for the moderating effect of neuroticism. As noted in the method section, the nature of the data (i.e., the large majority of the participants did not meet their exam goals which
88 led to negative discrepancy feedback) did not permit me to conduct analyses with respect to the moderating effects of extraversion and the BAS scale on the positive feedbackgoals and positive feedback-positive emotional reactions relationships. The moderating roles of goal commitment and academic self-efficacy were not supported by the data. In retrospect, the nature of the goal commitment measure may have contributed to the lack of a moderation effect. That is, goal commitment was measured relative to the studentsÂ’ overall goal for the course, and not relative to the specific exam goals. This measurement strategy was dictated by the design of the study (i.e., it would have been infeasible to measure both goals and goal commitment before each exam) but nevertheless may have influenced the results. Though unsupportive, these results replicate those of Williams et al. (2000), who found that goal commitment did not moderate the effect of goal discrepancy on athleteÂ’s goal regulation process. To conclude, I believe the findings of this third study add to the findings of the previous two studies presented in this dissertation by (a) replicating the mediation results with a different methodology, (b) clarifying the role of self-efficacy in the goal regulation process, and (c) lending some support to the dispositional moderator hypothesis.
89 CHAPTER 6 GENERAL DISCUSSION The results of the studies included in this dissertation have implications for both theory and practice that merit discussion, and they make an important contribution to the literature on motivation. In this section, I outline the findings, implications and contribution, as well as the limitations of this research. Findings The findings presented here provide several important insights into the psychological mechanisms involved in the dynamic self-regulation of goals across time, and offer an initial explanation of individual differences in the nature and the strength of such self-regulatory processes. Through these findings, in the dissertation, I have accomplished three major objectives. First, the present results show that performance feedback does predict goalregulation. In the first two studies, I have found strong evidence for both downward goal revision following negative feedback, and upward goal revision (discrepancy creation) following positive feedback, which is consistent with goal setting and social cognitive theory (e.g., Bandura & Locke, 2003) and with previous findings about goal change (e.g., Phillips et al., 1996; Williams et al., 2000). Because the feedback-goal revision results of the first two studies were obtained by separating withinand between-individual variance, this finding reflects strictly within-individual self-regulation. The relationship between feedback concerning previous performance and subsequent goals was also
90 supported in Study 3 which used a more typical longitudinal design based on betweenindividual variance. Second, the results presented here support the contention that basic affect and emotional reactions to feedback are important mechanisms that explain the relationship between feedback and future goals. This is perhaps the most important finding of this dissertation, and it has been supported in within-individual analyses on six independent samples totaling more than 900 participants. In addition, the mediating effect was also supported in the third study in a sample of almost 500 participants. Though this is not the first research to study the effects of performance feedback on affective constructs (e.g., Kanfer & Ackerman, 1989), it is the first investigation into affective processes that explain goal and behavioral regulation (i.e., Kanfer & Ackerman , studied selfreactions as a dependent variable). The results of the third study also revealed that task self-efficacy mediates part, but not all, of the relationship between emotional reactions to feedback and subsequent goals. To my knowledge, this is the first project that investigates the contribution of affective and cognitive processes to self-regulation simultaneously. Third, I found evidence supporting the moderating role of the BIS on the relationship between feedback and future goals, when feedback is negative. That is, the data collected for the third study was consistent with the moderating role of the Carver and WhiteÂ’s (1994) BIS scale on the relationship between negative goal discrepancy feedback and emotional reactions to feedback. However, no consistent cross-level moderator effect was detected in any of the multilevel studies.
91 The data collected for the third study did not support the moderating role of goal commitment for the relationship between negative feedback and subsequent goals. As noted, one explanation for this lack of support may reside in the way in which commitment was measured (i.e., commitment to the overall course goal and not the specific task goal). An alternative explanation concerns the nature of the goals themselves. More specifically, Brunstein and his associated (e.g., Brunstein, 2000; Brunstein & Gollwitzer, 1996) maintain that commitment to self-defining goals moderates the impact of failure on subsequent motivation. Though addressing only the inhibitive nature of failure (and not the approach-inducing nature of reward) BrunsteinÂ’s (2000, p. 349) arguments are consistent with the behavioral motivation theory and thus with the model of motivational self-regulation proposed in this dissertation: Like a punishment, failure commonly elicits as strong inhibitory force that fuels the tendency not to persist in striving for the respective goal. Abandoning an unsuccessful pursuit can indeed be quite rewarding as it diminishes feelings of distress and thereby leads to negative reinforcement. Here, however, commitment enters the stage and moderates how individuals react to goal obstacles. Although a failure experience is undoubtedly unpleasant, committed individuals have a clear idea of what they want to achieve in the future. This discrepancy between the immediate trouble and the prospect of a desired future creates the motivating force (or increase in volitional strength) that propels committed individuals to further efforts at reaching their desired goals. It follows that to the extent that the goals for the course used in the third study were not self-defining for the student participants (or were not self-defining for a large proportion of participants), the fact that goal commitment did not moderate participantsÂ’ reactions to failure is perhaps not that surprising. Also unsupported was the moderating role of academic self-efficacy (Study 3). This unsupportive result cannot be easily explained. The scores on the academic selfefficacy measure were reliable (alpha=.79); they exhibited the expected relationship with
92 other constructs measured in the thirds study (i.e., they correlated positively with exam self-efficacy and with exam goals), and the focus of the self-efficacy measure (academic performance) was highly relevant to the performance context studied. Only further empirical investigation can elucidate the role of general task self-efficacy in explain the mechanisms used by individuals to respond to failure feedback. The data did suggest support for the moderating role of task difficulty (see the discussion to Study 2), but given that fact that only three types of tasks were available for comparison, and that it was unclear whether differences among the task were mostly in terms of difficulty, complexity or creativity, this suggestive evidence should be treated with caution. Future research should investigate this moderating effect with more rigorous designs. Implications Theory Development From a theoretical standpoint, this research advances the understanding of the psychological mechanisms that individualsÂ’ use in interpreting and responding to performance feedback. I have shown that feedback influences affect, which, in turn, influences subsequent goals. Furthermore, the investigation into the process through which affect influences goals revealed that affect influences goal-regulation, in part, through feelings of self-efficacy (Study 3). The present results also suggest that, at least for negative feedback, dispositional constructs reflecting the chronic activation or strength, of the behavioral inhibition system moderate the impact of feedback on affect and emotions (Study 3). Conceptually, this impact should also be moderated by causal attributions for the level of performance (Ilgen & Davis, 2000; Weiner, 1985) and by the
93 credibility and acceptance of the feedback (e.g., Ilgen et al., 1979). Thus, a more complete model of goal regulation would include both dispositional constructs and situational components, such as performance attributions, as moderators of the impact of feedback on goals. In addition, dispositional constructs such as locus of control are likely to influence the way in which people make performance, or feedback, attributions. With respect to the hypothesized role of goal commitment, as noted, goal attributes such as the degree to which goals are self-defining for individuals may influence whether goal commitment is another moderator for the feedback-future goals relationship. This research also advances individual differences theory by examining the role of individual differences in broad bio behavioral systems (i.e., BIS and BAS) in motivational self-regulation. The results presented here suggest that the chronic activation levels of individualsÂ’ behavioral inhibition system, or the extent to which they are generally prone to respond to inhibiting stimuli, regulates their responses to failure feedback. This proposal may seem to be at odds with more traditional feedback theory. That is, traditionally, feedback researchers have examined the role of individual differences in self-esteem in regulating responses to failure (Brockner et al., 1987; Kernis, Brockner, & Frankel, 1989). However, self-esteem and the BIS may not be independent from each other. Judge, Erez, Bono, (1998) present meta-analytical evidence that self-esteem is highly correlated with neuroticism, and Judge, Erez, Bono, and Thoresen (2002) suggest that self-esteem may be an indicator of the broad domain of neuroticism. Given the clear conceptual link between neuroticism and the BIS (e.g., Carver et al., 2000), hopefully the research presented here will stimulate another
94 conceptual path for integrating, or distinguishing among, the individual differences domains of neuroticism and self-esteem: through the behavioral inhibition system. Though, unlike other studies (e.g., Phillips et al., 1996; Williams et al., 2000), I did not design the studies to test predictions from social cognitive theory vs. those from control theory, like the data from these other studies, the present data suggest that control theory, in its basic form, does not adequately explain motivational self-regulation across time. That is, the results from the first two studies clearly suggest that after meeting or exceeding their goals, individuals do not maintain their goal level and decrease effort in order to minimize the positive discrepancy between performance and goals, but rather set higher goals that motivate them to increase performance, as predicted by social cognitive theory. In other terms, the data presented here are consistent with Wood and BanduraÂ’s (1989, p. 367) assertion on the topic: A regulatory process in which matching a standard occasions inactivity does not characterize human self-motivation. Such a feedback control system would produce circular action that leads nowhere. In fact, people transcend feedback loops by setting new challenges for themselves. Finally, the conceptual model of self-regulatory motivation presented here is a dynamic model that makes predictions specifically focused at the withinand betweenindividual level. Traditional goal setting theory, by stating that Â“the simplest form and most direct motivational explanation of why some people perform better than others is because they have different performance goalsÂ” (Latham & Locke, 1991, p. 213), has a primary between-individual focus. The present research builds on the emerging stream of research in goal regulation across time (e.g., Williams et al., 2000), and advances goal setting theory by adding predictions at the within-individual level.
95 Implications for Practice From a practical perspective, understanding how individuals interpret feedback should help in designing feedback delivery systems at work. The results presented here show that feedback influences future goals and this effect was observed even when feedback was manipulated. These results have three main implications for the management of feedback systems. The first practical implication concerns goal regulation following performance that has met or exceeded the goal. Theories based on simple discrepancy-reduction mechanisms suggest that positive feedback should be withheld because employees will decrease their effort after receiving such positive feedback. The present results suggest that this may not be the case Â– in the multi-level studies included in this dissertation participantsÂ’ goals following positive feedback were directly proportional with the magnitude of the positive feedback. Second, when delivering negative feedback, managers should take individualsÂ’ personal characteristics into account. The present results show that those with increased inhibitive tendencies were more sensitive to negative feedback than others, and they were more likely to exhibit decreased motivation levels after failure. Finally, the present results suggest that negative feedback is beneficial only when the magnitude of the discrepancy between performance and standard is relatively small. Though examining when negative feedback becomes de-motivating was not the purpose of this dissertation, it is clear that at some point, after repeated or extreme negative feedback, most individuals give up their goals. In addition, the results suggest that
96 individualsÂ’ dispositional characteristics are associated with their resilience and persistence. Limitations Like all studies, this project has limitations that merit discussion. As noted, not all constructs that may add to the explanation of goal regulation were measured in these studies. I did not measure attributions, feedback acceptance, goal identity relevance or whether the feedback threatened participantsÂ’ self-esteem in the studies included in this project. A process important in self-regulation that was not studied here concerns the task strategies used by individuals to accomplish their goals. That Â“people form intentions that include plans and strategies for realizing themÂ” is specified by social cognitive theory (Bandura & Locke, 2003, p.97), and task strategies have been included in feedback models (Ilgen & Davis, 2000). A more complete model of self-regulation should include concepts and processes as those explained above. An important limitation of this research concerns the potential lack of generalizability of the findings associated with laboratory experiments that use student participants. However, I believe that the novelty of the research questions and of the methods through which I attempted to answer them justify an initial examination in controlled settings. Future research should examine whether these findings generalize to different participant populations. Another possible limitation concerns the performance task used in the experiments conducted for the first two studies. The brainstorming task used in the first study, for example, though it was extensively used in previous laboratory research on goal setting (e.g., Harkins & Lowe, 2000; Lee & Bobko, 1992; Locke, 1982), it is a very
97 simple task, and thus the results may not generalize to other performance situations. Using exam performance in Study 3 ameliorates this concern to some extent, but because Study 3 used a different methodology, whether the multi-level results generalize to more complex task is an issue that requires further investigation. Contribution The limitations of the studies conducted for this dissertation should be evaluated in light of the contributions of this research. I believe the studies presented here contribute to the general literatures on affect and behavior, goal setting and selfregulation, and their implications for task performance, and to the emerging literature on the role of emotional experiences in motivational self-regulation. This research contributes in at least five main areas to these literatures. First, this research contributes to the literature on self-regulation by taking a fresh perspective on the study of self-regulatory processes and investigating such processes within individuals and across time. In the multilevel studies, I have shown that goals vary substantially within individuals, and the within-individual variation is not stochastic error but it can be predicted by constructs included in the model developed here. In these studies, jointly, performance feedback and affect explained about one-third of the withinindividual variance in goals, for the average individual (see Figure 3-2). In addition, affect (mostly positive affect) mediated a substantial proportion of the feedbacksubsequent goals relationship within individuals (average of 33.33% across the six samples from Study 1 and Study 2), and the mediating effect was statistically significant in each of the six independent samples.
98 The multilevel results have also shown that not only does affect mediate a substantial proportion of the feedback-subsequent goals relationship, but it also has substantial independent predictive power in explaining goal regulation across time. Both affect and goals are constructs that exhibit important variations across time. The present results show that affect and goals have a dynamic relationship within-individuals, in that they vary in synchrony across time. To be sure, this is not the first study examining affect or goals across time. But, to my knowledge, it is the first study attempting to explain goal variations within individuals with momentary affect. Previous research suggested that fluctuations in state affect may translate into fluctuations in behavior at work (e.g, George & Brief, 1994; Ilies & Judge, 2002). The present results show that goal variations can be predicted by affect variation in the laboratory. It follows that goal-regulation, as a process, is likely to explain part of the connection between momentary affect, behavior, and, ultimately, performance. On a conceptual level, recently developed models of behavior in organizations, such as Weiss and CropanzanoÂ’s (1996) Affective Events Theory (AET) or Spector and FoxÂ’s (2001) model of voluntary work behavior, place affect and emotions at their center. If mood and emotions have an effect on behavior, and Â“this does not usually happen in a reflexive or nonpurposive manner with humansÂ” (Spector & Fox, 2001, p. 273), the process of goal regulation is likely to explain, at least in part, the link between emotion, action tendencies, and intentional behavior. In addition, the mediating effect of affect on the relationship between feedback and future goals suggests that feedback theory could be integrated with AET. That is, performance feedback can certainly be viewed as an affective event influencing employeesÂ’ mood, emotion and, perhaps through goals, their
99 behavior. As noted, the present results support goal-regulation predictions derived from social cognitive theory. Intentionality, forethought, self-reactiveness, and selfreflectiveness are core features of human agency (Bandura & Locke, 2003). By studying how individuals react to events such as feedback in terms of affect, and the effect of these self-reactions on goals and intentions we can perhaps understand the implications of affective events on behavior from a human agency perspective. Second, I developed a model of self-regulation based on goal setting, feedback, and behavioral motivation theory. By integrating these separate theories, I gave affect and emotions, which are constructs central to behavioral regulation theory, their rightful place in the dynamic model of goal regulation. In the light of the recent explosion of interest in examining affective process in organizational behavior, such integration is timely and should stimulate future research on the complex links between feedback, affect, goals and behavior. Third, this research contributes to goal-setting theory, by adding to the previous efforts examining the goal revision process following positive feedback concerning previous performance. That is, this research adds to the accumulating evidence of the positive discrepancy creation mechanism that individuals use to regulate their motivation and behavior, and should stimulate further investigations into positive discrepancy creation, as well as attempts to develop motivational techniques based on this process that can be used in organizations. Fourth, the studies included in this dissertation contribute to the literature on the dispositional source of task and work motivation by going beyond the investigation of direct dispositional effects on motivation and examining the role of the dispositional
100 parameters of the behavioral inhibition and activation systems in the process of goal regulation across time. This approach should stimulate further integration of the BIS/BAS framework into motivation theory. Finally, the use of multi-level methods enabled the examination of goal regulation strictly within individuals. As noted, self-regulation is a within-individual process; because multi-level methods are now available, it is important to study this process at the level at which it operates. This was one of the main purposes of this dissertation, and the studies presented have shown that using multi-level methods can prove fruitful towards accomplishing this purpose.
101 APPENDIX A INTERNET SCREENS FOR THE MULTI-TRIAL EXPERIMENT
105 APPENDIX B REMOTE ASSOCIATES TEST For this task you will be presented with three words and asked to write a fourth word which is related to all three . For example, what word do you think is related to these three? A. cookies sixteen heart The answer in this case is Â“sweet.Â” Cookies are sweet; sweet is part of the phrase Â“sweet sixteen,Â” and part of the word Â“sweetheartÂ”. Here is another example: B. poke go molasses You should have written Â“slowÂ” in the space provided. Â“Slow poke,Â” Â“go slowÂ”,Â” Â“slow as molasses.Â” As you can see, the fourth word may be related to the other three words for various reasons. Try these next two: C. surprise line birthday D. base snow dance The answers for the last two examples are Â“partyÂ” and Â“ballÂ”. As you can see, the fourth word may be related to the other three for various reasons. You will be presented with 8 successive trials, each containing five groupings of words. Some of these are not easy and you will have to think about them for a while. You have 7 minutes for each trial. Sample Trial A. falling actor dust B. lick sprinkle mines C. off trumpet atomic D. gold stool tender E. blank white lines
106 APPENDIX C THE POSITIVE AND NEGATIVE AFFECT SCHEDULE (PANAS) Instructions: This scale consists of a number of words that describe different feelings and emotions. Indicate to what extent you experience the following states right now, using this scale: 1=Very slightly or not at all 2=A little 3=Moderately 4=Quite a bit 5=Very much 1. _____ Interested 11. _____ Irritable 2. _____ Distressed 12. _____ Alert 3. _____ Excited 13. _____ Ashamed 4. _____ Upset 14. _____ Inspired 5. _____ Strong 15. _____ Nervous 6. _____ Guilty 16. _____ Determined 7. _____ Scared 17. _____ Attentive 8. _____ Hostile 18. _____ Jittery 9. _____ Enthusiastic 19. _____ Active 10. _____ Proud 20. _____ Afraid
107 APPENDIX D GOAL COMMITMENT MEASURE Instructions: Use the following response scale to indicate how true each of the following statements is about the goal you have just reported, by placing the appropriate number on the line preceding that item. Please be open and honest in responding: 1 2 3 4 5 6 7 Not at all Very little Somewhat To a moderate degree Quite a bit Very much Completely 1. __ I am strongly committed to pursuing this grade goal. 2. __ I am willing to put forth a great deal of effort beyond what IÂ’d normally do to achieve this goal. 3. __ Quite frankly, I donÂ’t care if I achieve this grade goal or not. 4. __ There is not much to be gained by trying to achieve this grade goal. 5. __ It is quite likely that this grade goal may need to be revised, depending on how things go this semester. 6. __ It wouldnÂ’t take much to make me abandon this grade goal. 7. __ ItÂ’s unrealistic for me to expect to reach this grade goal. 8. __ Since itÂ’s not always possible to tell how tough course are until youÂ’ve been in them for while, itÂ’s hard to take this grade goal seriously. 9. __ I think this grade goal is a good goal to shoot for.
108 APPENDIX E ACADEMIC SELF-EFFICACY MEASURE Instructions: Below are seven statements with which you may agree or disagree. In thinking about the typical introductory college course that you have been enrolled in, please use the 1-7 scale below to indicate your agreement with each item by placing the appropriate number on the line preceding that item: 1 2 3 4 5 6 7 Strongly disagree Disagree Slightly disagree Neither agree nor disagree Slightly agree Agree Strongly agree 1. __ During the great majority of lectures, I am able to concentrate and stay fully focused on the materials presented. 2. __ Usually I am able to memorize most of the facts and concepts covered in a collegelevel course, and then recall these facts and concepts on demand. 3. __ During exams, I am able to focus exclusively on understanding and answering questions, and I successfully avoid breaks in my concentration. 4. __ I understand a large proportion of the facts, concepts, and arguments as they are presented in lectures, tutorials or course materials (e.g., textbooks). 5. __ If others ask me, I can clearly explain in my own words most of the facts, concepts and arguments covered in the course (i.e., in lectures, tutorials or course materials). 6. __ Most of the time during a course, I can discriminate between the more important and the less important facts, concepts and arguments covered. 7. __ Most times, I am able to make understandable course notes which emphasize, clarify and relate key facts, concepts and arguments as they are presented in lectures, tutorials or course materials.
109 LIST OF REFERENCES Aguinis, H. (2002). Estimation of interaction effects in organization studies. Organizational Research Methods, 5, 207-211. Aguinis, H., Pierce, C. A., & Stone-Romero, E. F. (1994). Estimating the power to detect dichotomous moderators with moderated multiple regression. Educational and Psychological Measurement, 54, 690-692. Alliger, G. M., & Williams, K. J. (1993). Using signal-contingent experience sampling methodology to study work in the field: A discussion and illustration examining task perceptions and mood. Personnel Psychology, 46, 525-549. Ashforth, B. E., & Humphrey, R. H. (1995). Emotion in the workplace: A reappraisal. Human Relations, 48, 97-125. Ashkanasy, N. M., Haertel, C. E., & Zerbe, W. J. (2000). Emotions in the workplace: Research, theory, and practice. Westport, CT: Quorum Books/Greenwood Publishing Group. Bagozzi, R., Baumgartner, H., & Pieters, R. (1998). Goal-directed emotions. Cognition and Emotion, 12, 1-26. Bandura, A. (1986). Social foundations of thought and action: A social cognitive theory. Englewood Cliffs, NJ: Prentice-Hall. Bandura, A. (1991). Social cognitive theory of self-regulation. Organizational Behavior and Human Decision Processes, 50, 248-287. Bandura, A., & Cervone, D. (1983). Self-evaluative and self-efficacy mechanisms governing the motivational effects of goal systems. Journal of Personality and Social Psychology, 45, 1017-1028. Bandura, A., & Jourden, F. J. (1991). Self-regulatory mechanisms governing the impact of social comparison on complex decision making. Journal of Personality and Social Psychology, 60, 941-951. Bandura, A., & Locke, E.A. (2003). Negative self-efficacy and goal effects revisited. Journal of Applied Psychology, 88, 87-99. Baron, R. A. (1990). Environmentally induced positive affect: Its impact on self-efficacy, task performance, negotiation, and conflict. Journal of Applied Social Psychology, 2, 368-384.
110 Baron, R. M., & Kenny, D. A. (1986). The moderator-mediator variable distinction in social psychological research: Conceptual, strategic and statistical considerations. Journal of Personality and Social Psychology, 51 , 1173-1182. Bower, G. H. (1981). Mood and memory. American Psychologist, 36, 129-148. Brief, A. P., & Hollenbeck, J. R. (1985). An exploratory study of self-regulating activities and their effects on job performance. Journal of Occupational Behaviour, 6, 197208. Brockner, J., & Higgins, E. T. (2001). Regulatory focus theory: Implications for the study of emotions at work. Organizational Behavior and Human Decision Processes, 86, 35-66. Brockner, J., Derr, W. R., & Laing, W. N. (1987). Self-esteem and reactions to negative feedback: Toward greater generalizability. Journal of Research in Personality, 21, 319-333. Brown, J. D., & Marshall, M. A. (2001). Self-esteem and emotion: Some thoughts about feelings. Personality and Social Psychology Bulletin, 27, 575-584. Brunstein, J. C. (2000). Motivation and performance following failure: The effortful pursuit of self-defining goals. Applied Psychology: An International Review, 49, 340-356. Brunstein, J. C., & Gollwitzer, P. M. (1996). Effects of failure on subsequent performance: The importance of self-defining goals. Journal of Personality and Social Psychology, 70, 395-407. Button, S. B., Mathieu, J. E., & Aikin, K. J. (1996). An examination of the relative impact of assigned goals and self-efficacy on personal goals and performance over time. Journal of Applied Social Psychology, 26, 1084-1103. Carver, C. S., & Scheier, M. F. (1998). Attention and self regulation: A control theory to human behavior. Cambridge, UK: Cambridge University Press. Carver, C. S., Sutton, S. K., & Scheier, M. F. (2000). Action, emotion, and personality: Emerging conceptual integration. Personality and Social Psychology Bulletin, 26, 741-751. Carver, C. S., & White, T. L. (1994) Behavioral inhibition, behavioral activation, and affective responses to impending reward and punishment: The BIS/BAS scales. Journal of Personality and Social Psychology, 67, 319-333. Chen, G., Gully, S. M., Whiteman, J., & Kilcullen, R. N. (2000). Examination of relationships among trait-like individual differences, state-like individual differences, and learning performance. Journal of Applied Psychology, 85, 835847.
111 Costa, P. T., Jr. & McCrae, R. R. (1992). Revised NEO Personality Inventory (NEO-PIR) and NEO Five-Factor Inventory (FFI) professional manual. Odessa, FL: Psychological Assessment Resources, Inc. Cron, W. L., Slocum, J. W. Jr., & VandeWalle, D. (2001) Negative feedback and self-set goal level: The role of emotional reactions and goal orientation. Paper presented at the Society for Industrial and Organizational Psychology, San Diego, CA. Davidson, M. L., Kwak, N., Seo, Y. S., & Choi, J. (2002). Using hierarchical linear modeling to examine moderator effects: Person-by-organization interactions. Organizational Research Methods, 5, 231-254. Depue, R. A., & Iacono, W. G. (1989). Neurobiological aspects of affective disorders. Annual Review of Psychology, 40, 457-492. Donovan, J. J., & Radosevich, D. J. (1998). The moderating role of goal commitment on the goal difficulty-performance relationship: A meta-analytic review and critical reanalysis. Journal of Applied Psychology, 83, 308-315. Dweck, C. S. (1986). Motivational processes affecting learning. American Psychologist, 41, 1040-1048. Dweck, C. S., & Leggett, E. L. (1988). A social-cognitive approach to motivation and personality. Psychological Review, 95, 256-273. Earley, P. C., & Lituchy, T. R. (1991). Delineating goal and efficacy effects: A test of three models. Journal of Applied Psychology, 76, 717-724. Edwards, J. R. (1994). Regression analysis as an alternative to difference scores. Journal of Management Journal, 20, 683-689. Edwards, J. R. (1996). An examination of competing versions of the person-environment fit approach to stress. Academy of Management Journal, 39, 292-339. Edwards, J. R. (2001). Ten difference score myths. Organizational Research Methods, 4, 265-287. Edwards, J. R., & Van Harrison, R. (1993). Job demands and worker health: Threedimensional reexamination of the relationship between person-environment fit and strain. Journal of Applied Psychology, 78, 628-648. Edwards, J. R., & Parry, M. E. (1993). On the use of polynomial regression equations as an alternative to difference scores in organizational research. Academy of Management Journal, 36, 1577-1613. Erez, A., & Isen, A. M. (2002). The influence of positive affect on the components of expectancy motivation. Journal of Applied Psychology, 86, 1055-1067.
112 Fisher, C. D., & Ashkanasy, N. M. (2000). The emerging role of emotions in work life: An introduction. Journal of Organizational Behavior, 21, 123-129. Fisher, S. L., & Ford, J. K. (1998). Differential effects of learner effort and goal orientation on two learning outcomes. Personnel Psychology, 51, 397-420. Farr, J. L., Hofmann, D. A., & Ringenbach, K. L. (1993). Goal orientation and action control theory: Implications for industrial and organizational psychology. In C. L. Cooper & I. T. Robertson (Eds.), International review of industrial and organizational psychology (pp. 193-232). New York: Wiley. Forgas, J. P., Bower, G. H., & Moylan, S. J. 1990. Praise or blame? Affective influences on attributions for achievement. Journal of Personality and Social Psychology, 89, 56-66. Fox, S., & Spector, P. E. (2002). Emotions in the workplace: The neglected side of organizational life introduction. Human Resource Management Review, 12, 167171. Fowles, D. C. (1987). Application of a behavioral theory of motivation to the concepts of anxiety and impulsivity. Journal of Research in Personality, 21, 417-435. Freedman, L. S., & Schatzkin, A. (1992). Sample size for studying intermediate endpoints with intervention trials of observational studies. American Journal of Epidemiology, 136, 1148-1159. George, J. M., & Brief, A. P. (1994). Feeling good Â– doing good: A conceptual analysis of mood at work--organizational spontaneity relationship. Psychological Bulletin, 112, 310-329. Goodman, L. A. (1960). On the exact variance of products. Journal of the American Statistical Association, 55, 708-713. Gray, J. A. (1978). The neuropsychology of anxiety. British Journal of Psychology, 69, 417-434. Gray, J. A. (1979). A neuropsychological theory of anxiety. In C. E. Izard (Ed.), Emotions in personality and psychopathology (pp. 303-335). New York: Plenum. Gray, J. A. (1981). A critique of Eysenck's theory of personality. In H. J. Eysenck (Ed.), A model for personality (pp. 246-276). New York: Springer. Gray, J. A. (1990). Brain systems that mediate both emotion and cognition. Cognition and Emotion, 4, 269-288. Harkins, S. G., & Lowe, M. D. (2000). The effects of self-set goals on task performance. Journal of Applied Social Psychology, 30, 1-40.
113 Higgins, E. T. (1997). Beyond pleasure and pain. American Psychologist, 52, 1280-1300. Higgins, E. T. (1998). Promotion and prevention: Regulatory focus as a motivational principle. Advances in Experimental Social Psychology, 30, 1-46. Higgins, E. T., Shah, J., & Friedman, R. (1997). Emotional responses to goal attainment: Strength of regulatory focus as moderator. Journal of Personality and Social Psychology, 72, 515-525. Hollenbeck, J. R., Williams, C. R., & Klein, H. J. (1989). An empirical examination of the antecedents of commitment to difficult goals. Journal of Applied Psychology, 74, 18-23. Ilgen, D. R., & Davis, C. A. (2000). Bearing bad news: Reactions to negative performance feedback. Applied Psychology: An International Review, 49, 550565. Ilies, R., & Judge, T. A. (2002). Understanding the dynamic relationship between personality, mood, and job satisfaction: A field experience-sampling study. Organizational Behavior and Human Decision Processes, 89, 1119-1139. Johns, G. (1981). Difference score measures of organizational behavior variables: A critique. Organizational Behavior and Human Decision Processes, 27, 443-463. Judge, T. A., Erez, A., & Bono, J. E. (1998). The power of being positive: The relationship between positive self-concept and job performance. Human Performance, 11, 167187. Judge, T. A., Erez, A. Bono, J. E., & Thoresen, C. J. (2002). Are measures of self-esteem, neuroticism, locus of control, and generalized self-efficacy indicators of a common core construct? Journal of Personality and Social Psychology, 83, 693710. Judge, T.A., & Ilies, R. (2002). Relationship of personality to performance motivation: A meta-analytic review. Journal of Applied Psychology, 87, 797-807 . Kanfer, R. (1990). Motivation theory and industrial and organizational psychology. In M. D. Dunnette & L. M. Hough (Eds.), Handbook of industrial and organizational psychology (Vol. 1, pp. 75-170). Palo Alto, CA: Consulting Psychologists Press. Kanfer, R., & Ackerman, P. L. (1989). Motivation and cognitive abilities: An integrative aptitude-treatment interaction approach to skill acquisition [Monograph]. Journal of Applied Psychology, 74, 657-690. Kernis, M. H., Brockner, J., & Frankel, B. S. (1989). Self-esteem and reactions to failure: The mediating role of overgeneralization. Journal of Personality and Social Psychology, 57, 707-714.
114 Kluger, A. N., & DeNisi, A. (1996). The effects of feedback interventions on performance: Historical review, a meta-analysis and a preliminary feedback intervention theory. Psychological Bulletin, 119, 254-284. Krull, J.L., & MacKinnon, D.P. (1999). Multilevel mediation modeling in group-based intervention studies. Evaluation Review, 23, 418-444. Larsen, R. J., & Ketelaar, T. (1989). Extraversion, neuroticism, and susceptibility to positive and negative mood induction procedures. Personality and Individual Differences, 10, 1221-1228. Larsen, R. J., & Ketelaar, T. (1991). Personality and susceptibility to positive and negative emotional states. Journal of Personality and Social Psychology, 61, 132140. Latham, G. P., & Locke, E. A. (1991). Self-regulation through goal setting. Organizational Behavior and Human Decision Processes, 50, 212-247. Lee, C., & Bobko, P. (1994). Self-efficacy beliefs: Comparison of five measures. Journal of Applied Psychology, 79, 364-369. Locke, E. A. (1982). Relation of goal level to performance with a short work period and multiple goal levels. Journal of Applied Psychology, 67, 512-514. Locke, E. A. (1997). The motivation to work: What we know. Advances in Motivation and Achievement, 10, 375-412. Locke, E. A., & Latham, G. P. (1990). A theory of goal setting and task performance. Englewood Cliffs, NJ: Prentice-Hall. Locke, E. A., Shaw, K. N., Saari, L. M., & Latham, G. P. (1981). Goal setting and task performance: 1968-1980. Psychological Bulletin, 90, 125-152. Lord, R. G., Klimoski, R. J., & Kanfer, R. (2002). Emotions in the workplace: Understanding the structure and role of emotions in organizational behavior. San Francisco: Jossey Bass. MacKinnon, D. P., Lockwood, C. M., Hoffman, J. M., West, S. G., & Sheets, V. (2002) . A comparison of methods to test the significance of the mediated effect. Psychological Methods , 7 , 83-104. MacKinnon, D.P., & Dwyer, J.H. (1993). Estimating mediated effects in prevention studies. Evaluation Review, 17, 144-158. MacKinnon, D.P., Warsi, G., & Dwyer, G.H. (1995). A simulation study of mediated effect measures. Multivariate Behavioral Research, 30, 41-62.
115 Martin, L. L., Ward, D. W., Achee, J. W., & Wyer, R. S. Jr. (1993). Mood as input: People have to interpret the motivational implications of their moods. Journal of Personality and Social Psychology, 64, 317-326. Matthews, G., & Gilliland, K. (1999). The personality theories of H. J. Eysenck and J. A. Gray: A comparative review. Personality and Individual Differences, 26, 583626. McCrae, R. R., & Costa, P. T. (1985). Comparison of EPI and psychoticism scales with measures of the five-factor model of personality. Personality and Individual Differences, 6, 587-597. Mednick, S. A., & Mednick, M. T. (1967). ExaminerÂ’s manual: Remote Associates Test . Boston: Houghton Mifflin. Mento, A. J., Steel, R. P., & Karren, R. J. (1987). A meta-analytic study of the effects of goal setting on task performance: 1966-1984. Organizational Behavior and Human Decision Processes, 39, 52-83. Mitchell, T. R., Hopper, H., Daniels, D., George-Falvy, J., James, L. R. (1994). Predicting self-efficacy and performance during skill acquisition. Journal of Applied Psychology, 79, 506-517. Muchinsky, P. M. (2000). Emotions in the workplace: The neglect of organizational behavior. Journal of Organizational Behavior, 21, 801-805. Phillips, J. M., & Gully, S. M. (1997). Role of goal orientation, ability, need for achievement, and locus of control in the self-efficacy and goal-setting process. Journal of Applied Psychology, 82, 792-802. Phillips, J.M., Hollenbeck, J.R., & Ilgen, D.R. (1996). Prevalence and prediction of positive discrepancy creation: Examining a discrepancy between two selfregulation theories, Journal Of Applied Psychology, 81, 498-511. Pinder, C. C. (1984). Work motivation. Glenview, IL: Scott, Foresman. Podsakoff, P. M., & Farh, J. (1989). Effects of feedback sign and credibility on goal setting and task performance: A preliminary test of some control theory propositions. Organizational Behavior and Human Decision Processes, 44, 4567. Raudenbush, S. W., Brennan, R. T., & Barnett, R. C. (1995). A multivariate hierarchical model for studying psychological change within married couples. Journal of Family Psychology, 9, 161-174. Rawsthorne, L. J., & Elliot, A. J. (1999). Achievement goals and intrinsic motivation: A meta-analytic review. Personality and Social Psychology Review, 3, 326-344.
116 Renn, R. W., & Fedor, D. B. (2001). Development and field test of a feedback seeking, self-efficacy, and goal setting model of work performance. Journal of Management, 27, 563-583. Roney, C. J. R., Higgins, E. T., & Shah, J. (1995). Goals and framing: How outcome focus influences motivation and emotion. Personality and Social Psychology Bulletin, 21, 1151-1160. Russell, J. A., & Carroll, J. M. (1999). On the bipolarity of positive and negative affect. Psychological Bulletin, 125, 3-30. Rusting, C. L., & Larsen, R. J. (1997). Extraversion, neuroticism, and susceptibility to positive and negative affect: A test of two theoretical models. Personality and Individual Differences, 22, 607-612. Saavedra, R., & Earley, P. C. (1991). Choice of task and goal under conditions of general and specific affective inducement. Motivation and Emotions, 15, 45-65. Seifert, T. L. (1995). Characteristics of egoand task-oriented students: A comparison of two methodologies. British Journal of Educational Psychology, 65, 125-138. Seifert, T. L. (1997). Academic goals and emotions: Results of a structural equation model and a cluster analysis. British Journal of Educational Psychology, 67, 323338. Schwarz, N., & Clore, G. L. (1983). Mood, misattribution, and judgments of well-being: Informative and directive functions of affective states. Journal of Personality and Social Psychology, 45, 513-523. Shah, J., Higgins, E. T., & Fridman, R. (1999). Performance incentives and means: How regulatory focus influences goal attainment. Journal of Personality and Social Psychology, 74, 285-293. Shalley, C. E. (1991). Effects of productivity goals, creativity goals, and personal discretion on individual creativity. Journal of Applied Psychology, 76, 179-185. Sherer, M., Maddux, J. E., Mercandante, B., Prendice-Dunn, S., Jacobs, B., & Rogers, R. W. (1982). The Self-efficacy Scale: Construction and validation. Psychological Reports, 51, 663-671. Sobel, M. E. (1982). Asymptotic confidence intervals for indirect effects in structural equation models. In S. Leinhardt (Ed.), Sociological methodology (pp. 290-312). Washington, DC: American Sociological Association. Spector, P. E., & Fox, S. An emotion-centered model of voluntary work behavior: Some parallels between counterproductive work behavior and organizational citizenship behavior. Human Resource Management Review, 12, 269-292.
117 Tisak, J., & Smith, C. S. (1994a). Defending and extending difference score methods. Journal of Management, 20, 675-682. Tisak, J., & Smith, C. S. (1994b). Rejoinder to EdwardsÂ’s comments. Journal of Management, 20, 691-694. Vance, R. J., & Colella, A. (1990). Effects of two types of feedback on goal acceptance and personal goals . Journal of Applied Psychology, 75, 68-76. VandeWalle, D. (1997). Development and validation of a work domain goal orientation instrument. Educational and Psychological Measurement, 8, 995-1015. VandeWalle, D., Cron, W. L., & Slocum, J. W., Jr. (2001). The role of goal orientation following performance feedback. Journal of Applied Psychology, 86, 629-640. Venkatesh, V., & Speier, C. (1999). Computer technology training in the workplace: A longitudinal investigation of the effect of mood. Organizational Behavior and Human Decision Processes, 79, 1-28. Vroom, V. H. (1964). Work and motivation. New York: Wiley. Watson, D. (2000). Mood and temperament. New York: Guilford Press. Watson, D., Clark, L. A., & Tellegen, A. (1988). Development and validation of brief measures of positive and negative affect: The PANAS scales. Journal of Personality and Social Psychology, 54, 1063-1070. Watson, D., Wiese, D., Vaidya, J., & Tellegen, A. (1999). The two general activation systems of affect: Structural findings, evolutionary considerations, and psychobiological evidence. Journal of Personality and Social Psychology, 76, 805-819. Weiss, H. M. (2001). Introductory comments. Organizational Behavior and Human Decision Processes, 86, 1-2. Weiss, H. M., & Adler, S. (1984). Personality and organizational behavior. In B. Staw & L. L. Cummings (Eds.), Research in organizational behavior (Vol. 6, pp. 1-50). Greenwich, CT: JAI Press. Williams, K. J., Donovan, J. J., & Dodge, T. L. (2000). Self-regulation of performance: Goal establishment and goal revision processes in athletes. Human Performance, 13, 159-180. Wood, R. E., & Bandura, A. (1989). Social cognitive theory of organization management. Academy of Management Review, 14, 361-384.
118 Wood, R. E., & Locke, E. A. (1987). The relation of self-efficacy and grade goals to academic performance. Educational and Psychological Measurement, 47, 10131024.
119 BIOGRAPHICAL SKETCH My education includes an engineering diploma from the Polytechnic Institute of Bucharest, earned in 1993, and a Master of Business Administration degree, received from Iowa State University in 1999. After completing two years of doctoral course work at the University of Iowa, in 2001, I transferred to the University of Florida to complete my doctoral education. Most of my research to date can be described as investigating the influence of dispositions and affect, frequently in combination, on broad organizational outcomes such as leadership, motivation and job attitudes. I am also interested in applying behavioral genetics theory and methods to organizational behavior. My research has been published (or is forthcoming) in scholarly journals such as the Journal of Applied Psychology, Organizational Behavior and Human Decision Processes, Personnel Psychology, Psychological Methods, the Journal of Cross-Cultural Psychology and Academy of Management Learning and Education.