Gary P. Latham Joseph L. Rotman School of Business, University of Toronto, Toronto, Ontario, Canada
This article is based in part on Edwin A. Locke's G. Stanley Hall lectures at the annual meetings of the American Psychological Association in 1999 and of the Southeastern Psychological Association in 2000. It is also based on Gary P. Latham's presidential address to the Canadian Psychological Association in 2001 and on his invited address to the American Psychological Society in 2001.
Correspondence may be addressed to: Edwin A. Locke, 32122 Canyon Ridge Drive, Westlake Village, CA 91361. Electronic mail may be sent to firstname.lastname@example.org.
In the 1950s and 1960s, the study of motivation in North American psychology was not considered a respectable pursuit. The field was dominated by behaviorists, and"motivation" was argued by them to lie outside the person in the form of reinforcers and punishers. When internal mechanisms were acknowledged, as in drive reduction theory, it was said that they were primarily physiological.
McClelland, a nonbehaviorist, argued for the existence of internal motives, such as need for achievement, but these were asserted to be subconscious (McClelland, Atkinson, Clark, & Lowell, 1953) and hence measurable only by projective tests. Behaviorists, drive reductionists, and advocates of subconscious motives all agreed that introspection was not a valid method of understanding human motivation. This ruled out the possibility of studying the conscious regulation of action.
An exception to the anticonsciousness zeitgeist was the work of Ryan. Anticipating the cognitive revolution in psychology, Ryan (1970) argued that"it seems a simple fact that human behavior is affected by conscious purposes, plans, intentions, tasks and the like" (p. 18). For Ryan, these, which he called first-level explanatory concepts, were the immediate motivational causes of most human action.
Lewin and his colleagues (e. g. , Lewin, Dembo, Festinger, & Sears, 1944) studied conscious goals, or levels of aspiration, years prior to Ryan's work. However, they treated levels of aspiration as a dependent rather than an independent variable. Mace (1935), a British investigator who was perhaps less influenced than others by American behaviorism, was the first to examine the effects of different types of goals on task performance. His work was largely ignored, however, except for a citation in Ryan's classic text with Smith on industrial psychology (Ryan & Smith, 1954).
Goal-setting theory was formulated inductively largely on the basis of our empirical research conducted over nearly four decades. It is based on Ryan's (1970) premise that conscious goals affect action. A goal is the object or aim of an action, for example, to attain a specific standard of proficiency, usually within a specified time limit. As industrial-organizational psychologists, our primary interest has been to predict, explain, and influence performance on organizational or work-related tasks. Thus, we focused on the relationship between conscious performance goals and level of task performance rather than on discrete intentions to take specific actions (e. g. , to apply to graduate school, to get a medical examination). The latter type of intention has been studied extensively by social psychologists, such as Fishbein and Ajzen (1975).
The first issue we addressed was the relationship of goal difficulty to performance. Atkinson (1958), a student of McClelland, had shown that task difficulty, measured as probability of task success, was related to performance in a curvilinear, inverse function. The highest level of effort occurred when the task was moderately difficult, and the lowest levels occurred when the task was either very easy or very hard. Atkinson did not measure personal performance goals or goal difficulty. Moreover, his task-difficulty findings have not been replicated when task performance goals were measured. 1 We found a positive, linear function in that the highest or most difficult goals produced the highest levels of effort and performance. Goal difficulty effect sizes (d) in meta-analyses ranged from. 52 to. 82 (Locke & Latham, 1990). Performance leveled off or decreased only when the limits of ability were reached or when commitment to a highly difficult goal lapsed (Erez & Zidon, 1984).
We also compared the effect of specific, difficult goals to a commonly used exhortation in organizational settings, namely, to do one's best. We found that specific, difficult goals consistently led to higher performance than urging people to do their best. The effect sizes in meta-analyses ranged from. 42 to. 80 (Locke & Latham, 1990). In short, when people are asked to do their best, they do not do so. This is because do-your-best goals have no external referent and thus are defined idiosyncratically. This allows for a wide range of acceptable performance levels, which is not the case when a goal level is specified. Goal specificity in itself does not necessarily lead to high performance because specific goals vary in difficulty. However, insofar as performance is fully controllable, goal specificity does reduce variation in performance by reducing the ambiguity about what is to be attained (Locke, Chah, Harrison, & Lustgarten, 1989). Goal studies have also compared the effects of learning versus performance goals and proximal versus distal goals. These results are discussed below in relation to the moderating effects of task complexity.
Expectancy and Social-Cognitive Theories
Goal-setting theory appears to contradict Vroom's (1964) valence-instrumentality-expectancy theory, which states that the force to act is a multiplicative combination of valence (anticipated satisfaction), instrumentality (the belief that performance will lead to rewards), and expectancy (the belief that effort will lead to the performance needed to attain the rewards). Other factors being equal, expectancy is said to be linearly and positively related to performance. However, because difficult goals are harder to attain than easy goals, expectancy of goal success would presumably be negatively related to performance.
The apparent contradiction between the two theories is resolved by distinguishing expectancy within versus expectancy between goal conditions. Locke, Motowidlo, and Bobko (1986) found that when goal level is held constant, which is implicitly assumed by valence-instrumentality-expectancy theory, higher expectancies lead to higher levels of performance. Across goal levels, lower expectancies, associated with higher goal levels, are associated with higher performance.
This within-between distinction is not an issue in social-cognitive theory (Bandura, 1986, 1997). Self-efficacy (task-specific confidence) is measured by getting efficacy ratings across a whole range of possible performance outcomes rather than from a single outcome (Locke et al. , 1986). The concept of self-efficacy is important in goal-setting theory in several ways. When goals are self-set, people with high self-efficacy set higher goals than do people with lower self-efficacy. They also are more committed to assigned goals, find and use better task strategies to attain the goals, and respond more positively to negative feedback than do people with low self-efficacy (Locke & Latham, 1990; Seijts & B. W. Latham, 2001). These issues are addressed further below.
Goals affect performance through four mechanisms. First, goals serve a directive function; they direct attention and effort toward goal-relevant activities and away from goal-irrelevant activities. This effect occurs both cognitively and behaviorally. For example, Rothkopf and Billington (1979) found that students with specific learning goals paid attention to and learned goal-relevant prose passages better than goal-irrelevant passages. Locke and Bryan (1969) observed that people who were given feedback about multiple aspects of their performance on an automobile-driving task improved their performance on the dimensions for which they had goals but not on other dimensions.
Second, goals have an energizing function. High goals lead to greater effort than low goals. This has been shown with tasks that (a) directly entail physical effort, such as the ergometer (Bandura & Cervone, 1983); (b) entail repeated performance of simple cognitive tasks, such as addition; (c) include measurements of subjective effort (Bryan & Locke, 1967a); and (d) include physiological indicators of effort (Sales, 1970).
Third, goals affect persistence. When participants are allowed to control the time they spend on a task, hard goals prolong effort (LaPorte & Nath, 1976). There is often, however, a trade-off in work between time and intensity of effort. Faced with a difficult goal, it is possible to work faster and more intensely for a short period or to work more slowly and less intensely for a long period. Tight deadlines lead to a more rapid work pace than loose deadlines in the laboratory (Bryan & Locke, 1967b) as well as in the field (Latham & Locke, 1975).
Fourth, goals affect action indirectly by leading to the arousal, discovery, and/or use of task-relevant knowledge and strategies (Wood & Locke, 1990). It is a virtual axiom that all action is the result of cognition and motivation, but these elements can interact in complex ways. Below is a summary of what has been found in goal-setting research:
1. When confronted with task goals, people automatically use the knowledge and skills they have already acquired that are relevant to goal attainment. For example, if the goal involves cutting logs, loggers use their knowledge of logging without the need for additional conscious planning in their choice to exert effort and persist until the goal is attained (Latham & Kinne, 1974).
2. If the path to the goal is not a matter of using automatized skills, people draw from a repertoire of skills that they have used previously in related contexts, and they apply them to the present situation. For example, Latham and Baldes (1975) found that truck drivers who were assigned the goal of increasing the weight of their truck loads made modifications to their trucks so that they could better estimate truck weight before driving to the weighing station.
3. If the task for which a goal is assigned is new to people, they will engage in deliberate planning to develop strategies that will enable them to attain their goals (Smith, Locke, & Barry, 1990).
4. People with high self-efficacy are more likely than those with low self-efficacy to develop effective task strategies (Latham, Winters, & Locke, 1994; Wood & Bandura, 1989). There may be a time lag between assignment of the goal and the effects of the goal on performance, as people search for appropriate strategies (Smith et al. , 1990).
5. When people are confronted with a task that is complex for them, urging them to do their best sometimes leads to better strategies (Earley, Connolly, & Ekegren, 1989) than setting a specific difficult performance goal. This is because a performance goal can make people so anxious to succeed that they scramble to discover strategies in an unsystematic way and fail to learn what is effective. This can create evaluative pressure and performance anxiety. The antidote is to set specific challenging learning goals, such as to discover a certain number of different strategies to master the task (Seijts & G. P. Latham, 2001; Winters & Latham, 1996).
6. When people are trained in the proper strategies, those given specific high-performance goals are more likely to use those strategies than people given other types of goals; hence, their performance improves (Earley & Perry, 1987). However, if the strategy used by the person is inappropriate, then a difficult performance-outcome goal leads to worse performance than an easy goal (Audia, Locke, & Smith, 2000; Earley & Perry, 1987). For a detailed discussion of the relation of task goals and knowledge, see Locke (2000).
The goal-performance relationship is strongest when people are committed to their goals. Seijts and Latham (2000a) found goal commitment questionnaires to have high reliability and validity. Commitment is most important and relevant when goals are difficult (Klein, Wesson, Hollenbeck, & Alge, 1999). This is because goals that are difficult for people require high effort and are associated with lower chances of success than easy goals (Erez & Zidon, 1984).
Two key categories of factors facilitating goal commitment are (a) factors that make goal attainment important to people, including the importance of the outcomes that they expect as a result of working to attain a goal, and (b) their belief that they can attain the goal (self-efficacy).
Importance.There are many ways to convince people that goal attainment is important. Making a public commitment to the goal enhances commitment, presumably because it makes one's actions a matter of integrity in one's own eyes and in those of others (Hollenbeck, Williams, & Klein, 1989). Goal commitment can also be enhanced by leaders communicating an inspiring vision and behaving supportively. In field settings (e. g. , Ronan, Latham, & Kinne, 1973) and laboratory settings (e. g. , Latham & Saari, 1979b), the supervisor's legitimate authority to assign goals creates demand characteristics.
An alternative to assigning goals is to allow subordinates to participate in setting them. The theory is that this would make goals more important to the person because one would, at least in part, own the goals. A series of studies by Latham and his colleagues revealed that, when goal difficulty is held constant, performances of those with participatively set versus assigned goals do not differ significantly (e. g. , Dossett, Latham, & Mitchell, 1979; Latham & Marshall, 1982; Latham & Saari, 1979a, 1979b; Latham & Steele, 1983). Erez and her colleagues (Erez, 1986; Erez, Earley, & Hulin, 1985; Erez & Kanfer, 1983), however, reached the opposite conclusion.
Working collaboratively, with Locke as mediator, Latham and Erez explored reasons for their contradictory findings. They found that from a motivational perspective, an assigned goal is as effective as one that is set participatively provided that the purpose or rationale for the goal is given. However, if the goal is assigned tersely (e. g. ,"Do this ...") without explanation, it leads to performance that is significantly lower than for a participatively set goal (Latham, Erez, & Locke, 1988). Meta-analyses of the effects of participation in decision making on performance, for those studies that measured performance objectively, yielded an effect size of only. 11 (Wagner & Gooding, 1987a, 1987b).
Subsequently, Locke, Alavi, and Wagner (1997) found that the primary benefit of participation in decision making is cognitive rather than motivational in that it stimulates information exchange. For example, Latham et al. (1994) found that with goal difficulty level controlled, participation in goal setting had no beneficial effect on performance. However, people who participated with others in formulating task strategies performed significantly better and had higher self-efficacy than those who did not participate in formulating strategies.
Monetary incentives are one practical outcome that can be used to enhance goal commitment. However, there are important contingency factors. The first is the amount of the incentive; more money gains more commitment. Second, goals and incentive type interact. When the goal is very difficult, paying people only if they reach the goal (i. e. , a task-and-bonus system) can hurt performance. Once people see that they are not getting the reward, their personal goal and their self-efficacy drop and, consequently, so does their performance. This drop does not occur if the goal is moderately difficult or if people are given a difficult goal and are paid for performance (e. g. , piece rate) rather than goal attainment (Latham & Kinne, 1974; Latham & Yukl, 1975; T. Lee, Locke, & Phan, 1997).
Latham (2001) developed an empathy box to help managers identify nonfinancial outcomes that employees expected as a result of committing to or rejecting a specific difficult goal. In a study where the goal was to reduce theft, when self-efficacy regarding honest behavior was high, actions taken to change outcome expectancies led to a significant decrease in stolen material (Latham, 2001).
Self-efficacy.As noted, self-efficacy enhances goal commitment. Leaders can raise the self-efficacy of their subordinates (a) by ensuring adequate training to increase mastery that provides success experiences, (b) by role modeling or finding models with whom the person can identify, and (c) through persuasive communication that expresses confidence that the person can attain the goal (Bandura, 1997; White & Locke, 2000). The latter may involve giving subordinates information about strategies that facilitate goal attainment. Transformational leaders raise the efficacy of employees through inspiring messages to and cognitive stimulation of subordinates (Bass, 1985).
For goals to be effective, people need summary feedback that reveals progress in relation to their goals. If they do not know how they are doing, it is difficult or impossible for them to adjust the level or direction of their effort or to adjust their performance strategies to match what the goal requires. If the goal is to cut down 30 trees in a day, people have no way to tell if they are on target unless they know how many trees have been cut. When people find they are below target, they normally increase their effort (Matsui, Okada, & Inoshita, 1983) or try a new strategy. Summary feedback is a moderator of goal effects in that the combination of goals plus feedback is more effective than goals alone (Bandura & Cervone, 1983; Becker, 1978; Erez, 1977; Strang, Lawrence, & Fowler, 1978).
Control theory (Carver & Scheier, 1981) also emphasizes the importance of goal setting and feedback for motivation. The assumptions that underlie control theory, however, are questionable (Locke, 1991a, 1994; Locke & Latham, 1990). In essence, the theory is based on a machine model derived from cybernetic engineering (Powers, 1978). The source of motivation is asserted to be a negative feedback loop (such as that characterizing a thermostat) that eliminates goal-performance discrepancies. The natural state of the organism is, by implication, one of motionlessness or rest.
Control theory is in effect a mechanistic version of Hull's drive reduction theory, which was abandoned decades ago. However, machines do not possess internal motivational states and do not have goals of their own. Their"goals" are those of the machine's builders. Furthermore, discrepancy reduction is a consequence rather than a cause of goal-directed behavior. As Bandura (1989) stated, goal setting is first and foremost a discrepancy-creating process. Motivation requires feed-forward control in addition to feedback. After people attain the goal they have been pursuing, they generally set a higher goal for themselves. This adoption of higher goals creates rather than reduces motivation discrepancies to be mastered. "Self motivation thus involves a dual cyclic process of disequilibratory discrepancy production followed by equilibratory reduction" (Bandura, 1989, p. 38).
A third moderator of goal effects is task complexity. As the complexity of the task increases and higher level skills and strategies have yet to become automatized, goal effects are dependent on the ability to discover appropriate task strategies. Because people vary greatly in their ability to do this, the effect size for goal setting is smaller on complex than on simple tasks. Meta-analyses (Wood, Mento, & Locke, 1987) have revealed goal difficulty effect sizes (d) of. 48 for the most complex tasks versus. 67 for the least complex tasks. For specific difficult goals versus a goal to do one's best, the effect size was. 41 for the most complex tasks versus. 77 for the least complex tasks.
Because people use a greater variety of strategies on tasks that are complex than on tasks that are easy, measures of task strategy often correlate more highly with performance than do measures of goal difficulty (Chesney & Locke, 1991). In addition, there are often goal-strategy interactions, with goal effects strongest when effective strategies are used (Durham, Knight, & Locke, 1997).
R. Kanfer and Ackerman (1989) found that in an air traffic controller simulation (a highly complex task), having a performance-outcome goal actually interfered with acquiring the knowledge necessary to perform the task. People performed better when they were asked to do their best. However, Winters and Latham (1996) showed that the fault was with the type of goal that had been set rather than with the theory. They found that when a specific difficult learning goal rather than a performance goal was set, consistent with goal-setting theory, high goals led to significantly higher performance on a complex task than did the general goal of urging people do their best.
Another factor that may facilitate performance on new, complex tasks is the use of proximal goals. Latham and Seijts (1999), using a business game, found that do-your-best goals were more effective than distal goals, but when proximal outcome goals were set in addition to the distal outcome goal, self-efficacy and profits were significantly higher than in the do-your-best condition or in the condition where only a distal outcome goal had been set. In dynamic situations, it is important to actively search for feedback and react quickly to it to attain the goal (Frese & Zapf, 1994). As Dorner (1991) noted, performance errors on a dynamic task are often due to deficient decomposition of a distal goal into proximal goals. Proximal goals can increase what Frese and Zapf (1994) called error management. Proximal feedback regarding errors can yield information for people about whether their picture of reality is aligned with what is required to attain their goal.
Personal Goals as Mediators of External Incentives
What Locke (1991b) called the motivation hub, meaning where the action is, consists of personal goals, including goal commitment, and self-efficacy. These variables are often, though not invariably, the most immediate, conscious motivational determinants of action. As such, they can mediate the effects of external incentives.
For example, assigned goal effects are mediated by personal or self-set goals that people choose in response to the assignment, as well as by self-efficacy. The relationships among assigned goal difficulty, self-set goal difficulty, self-efficacy, and performance are shown in
Figure 1 - Relationships Among Assigned Goals, Self-Set Goals, Self-Efficacy, and Performance
Figure 1. Observe that assigning a challenging goal alone raises self-efficacy because this is an implicit expression of confidence by a leader that the employee can attain the goal. The correlation between self-set goals and self-efficacy is higher when no goals are assigned.
The mediating effect of self-set goals and self-efficacy on monetary incentive effects was noted earlier (T. Lee et al. , 1997). However, not all incentive studies have found a mediating effect. Wood, Atkins, and Bright (1999) showed that incentive effects were mediated by instrumentality or outcome expectancies rather than by goals and efficacy. Also, when summary feedback is provided without any goals, the feedback effects are mediated by the goals that are set in response to the feedback (Locke & Bryan, 1968). Bandura and Cervone (1986) found that both goals and self-efficacy mediated feedback effects. Self-efficacy is especially critical when negative summary feedback is given because the person's level of self-efficacy following such feedback determines whether subsequent goals are raised or lowered.
As noted earlier, the benefits of participation in decision making are primarily cognitive rather than motivational. However, Latham and Yukl (1976) and Latham, Mitchell, and Dossett (1978) found that employees who were allowed to participate in setting goals set higher goals and had higher performance than those who were assigned goals by the supervisor. The higher the goals, the higher the performance. Finally, Kirkpatrick and Locke (1996) found that goals and self-efficacy mediated the effect of visionary leadership on employee performance.