The Effects of Feedback Interventions on Performance: A Historical Review, a Meta-Analysis, and a Preliminary Feedback Intervention Theory (Kluger & DeNisi, 1996)

Kluger, A. N., & DeNisi, A. (1996). The effects of feedback interventions on performance: A historical review, a meta-analysis, and a preliminary feedback intervention theory. Psychological Bulletin, 119(2), 254–284.

Goals of the paper:

  1. To document both the evidence for inconsistent FI effects and the disregard for these data from the onset of FI research.
  2. To quantify the variability of FI effects and to rule out artifact-based explanations to FI effects variability.
  3. To attempt the integration of the varying theoretical and paradigmatic perspectives on FI, and propose a feedback intervention theory (FIT).
  4. To provide a preliminary test of FIT, by subjecting some putative moderators, identified by FIT, to moderator analyses of the meta-analytic effects.

[*Note: Much of this material was taken directly from the article:]

Articles in which the universal utility of feedback is called into question:

“To relate feedback directly to behavior is very confusing. Results are contradictory and seldom straight-forward.” — Ilgen, Fisher, & Taylor, 1979, p. 368

“The effects of manipulation of KR [knowledge of results] on motor learning. . .reveal. . . some violent contradictions to earlier beliefs about KR, and some glaring absences in our knowledge.” — Salmoni, Schmidt, & Walter, 1984, p. 378

“Feedback does not uniformly improve performance.” — Balcazar, Hopkins, & Suarez, 1985, p. 65

“Few concepts in psychology have been written about more uncritically and incorrectly than that of feedback.. . . Actually, feedback is only information, that is, data, and as such has no necessary consequences at all.” — Latham & Locke, 1991, p. 224

“The major culprit in sustaining the unwarranted assumptions about FIs is a lack of a FI theory. In the absence of a FI theory, Fl-related hypotheses were largely derived from the behavioristic law of effect (Thorndike, 1913, 1927). However, these hypotheses were inconsistent with data in many ways (e.g., Annett, 1969). More recent and tenable hypotheses were derived from theories that included feedback as a theoretical component (e.g., goal setting theory, Locke & Latham, 1990; control theory, Carver & Scheier, 1981, Annett, 1969, Podsakoff & Farh, 1989; multiple-cue probability learning paradigm (MCPL), Balzer, Doherty, & O’Connor, 1989; social cognition theory, Bandura, 1991, and a variant of learned helplessness theory, Mikulincer, 1994) but these are limited only to some of the known Fl-induced processes (e.g., to motivation or to learning processes, but not to both). As a result, recent FI research is carried out by isolated pockets of researchers who share either a theoretical or a paradigmatic orientation” (p. 254).

Definition of Feedback Interventions (FIs): “actions taken by (an) external agent (s) to provide information regarding some aspect(s) of one’s task performance” (p. 255). Similar to “knowledge of performance” interventions (Ammons, 1956), “augmented feedback” (Annett, 1969; Salmoni et al., 1984), or “extrinsic feedback” (Annett, 1969; Frese & Zapf, 1994).

  • FIs include knowledge of results (KR) interventions—the focus of much of
    the original research in this area—but FIs are broader in scope than KR interventions.
  • The definition includes FIs for a wide spectrum of tasks, such as test performance, memory tasks, physical tasks, attendance behavior,
    complying with regulations, etc.
  • The authors focus on task-performance FIs, including KR interventions, but exclude research dealing with feedback that is not part of an intentional intervention by an external agent.

Of the usable 131 papers, 607 effect sizes were extracted. These effects were based on 12,652 participants and 23,663 observations (reflecting multiple observations per participant). The average sample size per effect was 39 participants. The weighted mean (weighted by sample size) of this distribution is 0.41, suggesting that, on average, FI has a moderate positive effect on performance. However, over 38% of the effects were negative (p. 258).

The single most influential theory in this area is Thorndike’s (1913) law of effect, i.e., a positive FT was equated with reinforcement and a negative FI with punishment. Both a positive FI and a negative FI should improve performance because one reinforces the correct behavior and the other punishes the incorrect behavior. The law of effect was never sufficiently detailed to account for the inconsistent findings (p. 258).

FIT has five basic arguments:

  1. Behavior is regulated by comparisons of feedback to goals or standards.
  2. Goals or standards are organized hierarchically.
  3. Attention is limited and therefore only feedback-standard gaps that receive attention actively participate in behavior regulation.
  4. Attention is normally directed to a moderate level of the hierarchy.
  5. FIs change the locus of attention and therefore affect behavior.

Both the law of effect and FI theories view behavior as goal directed. To achieve goals or standards, people use feedback (whether provided by an intervention or not) to evaluate their performance relative to their goals. The result of a comparison of FI to a goal or a standard creates a feedback sign (positive or negative evaluation of one’s performance relative to the goal)—an argument accepted by researchers across a variety of theoretical orientations (e.g., Bandura, 1991, social cognition theory; Mikulincer, 1994, learned helplessness theory; Podsakoff & Farh, 1989, control theory; Locke & Latham, 1990, goal setting theory).

Control theory — when a discrepancy is noted, people are motivated to reduce it. The single system (consisting of a goal, feedback, comparison of the two, and an action to reduce sensed discrepancy) is referred to as a negative-feedback-loop, also known as test-operate-test-exit. The discrepancy can be eliminated by changing behavior to change the future feedback, by changing the standard so it matches the present feedback, by rejecting the feedback, or by escaping the situation (physically or mentally) that signals discrepancy.

Goal setting theory — people are motivated to achieve a goal, rather than eliminate the discrepancy. One can strive to attain the goal, change the goal, reject the feedback, or abandon commitment to the goal.

“…the multitude of coping mechanisms is a theoretical challenge because one needs to predict a priori which of these coping mechanisms will be activated…To predict the effects of FI on task motivation, one needs to know a priori the strength of both the goal and the feedback and the likelihood that either of them can be changed” (p. 260).

Four strategies of eliminating the feedback-standard gap:

  1. Change the behavior. When the feedback sign is negative, 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 high (e.g., high self-efficacy; Bandura & Cervone, 1983). Adding an FI to a goal-setting intervention is likely to further block any other coping mode except for changing behavior. (See research on goal-setting interventions, Locke & Latham, 1990).
  2. Abandon the standard. This occurs especially when the discrepancy is perceived to have a low likelihood of being eliminated through actions (e.g., Bandura, 1991; Mikulincer, 1988b).
  3. Change the standard. People may lower the standard when they receive negative feedback and when they cannot or do not want to abandon the standard. Alternatively, people may raise their standard when they receive positive feedback (Lewin, Dembo, Festinger, & Sears, 1944) and therefore produce improvements in future performance.
  4. Reject the feedback message. It seems that a negative feedback sign is more likely than a positive feedback sign to lead to feedback rejection (e.g., Ilgen et al., 1979).

The Limitation of the Feedback-Standard Comparisons Argument: (p. 260)

  • Needs to account for the role of multiple standards (e.g., a norm; a prior expectation (Ilgen, 1971; Ilgen & Hamstra, 1972; Kluger, Lewinsohn, & Aiello, 1994); past performance levels, (Carver & Scheier, 1990; Hsee & Abelson, 1991); performance of other groups; and an ideal goal (Lewin et al., 1944). Multiple standards influence the affective reaction to FIs (Bandura, 1991; Ilgen, 1971; Ilgen & Hamstra, 1972; Kluger et al., 1994; Locke & Latham, 1990). These findings suggest that various feedback-standard discrepancies are weighted and summed into an overall affective evaluation of the FI.
  • Cannot account for various findings regarding detrimental FI effects on learning. Adding an FI to CAI programs impairs learning relative to Fl-free CAI programs (Carroll & Kay, 1988; Lepper & Gurtner, 1989) or at best has no effects on CAI learning (Wise, Plake, Pozehl, Barnes, & Lukin, 1989). Jacoby, Mazursky, Troutman, and Kuss (1984) reported that seeking outcome feedback was negatively correlated with performance.
  • Does not incorporate findings regarding the effects of Fl-induced affect on performance. FIs strongly influence both pleasantness (e.g., Isen, 1987) and arousal (Klugeretal., 1994). Pleasantness has both inhibitory and facilitating effects on cognition and performance. Easterbrook’s (1959) cue-utilization hypothesis: arousal increases attention to focal cues and reduces attention to peripheral cues (Forgas, Bower, & Moylan, 1990; Isen, 1987; Mano, 1992; Christiansen, 1992).

“In summary, the assumption that behavior is regulated through feedback-standard comparisons and discrepancy reduction is too simple…By adding additional assumptions presented below, the preliminary FIT sets the foundation to accommodate these challenges” (p. 261).

Additional FIT assumptions:

  • Hierarchy (p. 261-2). Negative-feedback-loops are organized hierarchically. Negative- feedback-loops at the top of the hierarchy contain goals of the self, whereas those at the bottom of the hierarchy contain physical action goals (e.g., open the door). Loops that are high in the hierarchy can supervise the performance of lower level loops, such that the output of higher level loops may be the change of goals for lower level loops.
  • Attention. At any level of the hierarchy, there may be discrepancies in
    many negative-feedback-loops, but only those loops receiving attention are acted on.
  • Normal locus of attention. Attention is normally directed to a moderate level of the hierarchy, that is, not to the ultimate goals of the self or to the
    detailed components of an ongoing activity.
  • FI effect of locus of attention. FIs command, and often receive, considerable attention. FIs are unlikely to be ignored because any FI has potentially serious implications for the self.

FIT: Integrating the assumptions

Three levels of linked processes involved in the regulation of task
performance (highest to lowest): meta-task processes involving the self -> task-motivation processes involving the focal task -> task-learning processes
involving the task details of the focal task.

Processes that occur above the focal task level = “meta-task processes” to indicate that these processes have the potential to control the focal task processes. Meta-task processes include:

  • Processes that link the focal task with higher order goals, such as the evaluation of the implication of task performance for the self
  • Processes that have considerable effects on performance, such as attention to the self, affect, and possibly framing effects.
  • Nonfocal task processes and nonfocal task-learning processes, such as a motivation to retaliate against the feedback messenger (M. S. Taylor, Fisher, & IIgen, 1984) and learning that the feedback sender is untrustworthy. Such processes may not be in themselves meta-task processes; but in the context of FI, they are likely to be activated by meta-task processes to serve higher order goals.

FI Effects on Task-Learning Processes

When people are confronted with subjective failure that they want to overcome, they first try to work harder (Wood & Locke, 1990). Working harder is the output of the motivational process, and it is accomplished by activating programs or scripts for action that are available from past experience. These programs are
lower level negative-feedback-loops supervised by the motivational processes. They are activated by default because they require only the allocation of little additional cognitive resources. If working harder fails, people may try to work smarter by generating a hypothesis regarding means for improved performance. According to a model proposed by Wood and Locke (1990) to account for the effects of goal setting on complex tasks, a motivated performer first activates an universal strategy that can work on most tasks. The universal strategy is to expend more effort, persist, and focus attention on the tasks. If the universal strategy fails, people may search for a task-specific plan; if the latter fails or is not available, people may try to develop a new strategy. The universal strategy and the task-specific strategy correspond to the task-motivational level and the task-learning level in FIT, respectively.

Learning processes may also be activated directly by FI cues. FI cues that refer to components of the task (e.g., you are not using your thumb for typing) are likely to direct attention to learning processes and generate working hypotheses, or at least cause their reevaluation.

Fl-induced attention to learning processes does not guarantee an improvement in performance.

  • When the task is well practiced, attention to task details is likely to interrupt the execution of automatic scripts (well-tested hypotheses) and impair performance (Vallacher & Wegner, 1987).
  • Outcome FIs (mere KR) impede learning of complex tasks and subsequently task performance (e.g., Balzer et al., 1989). In fact, outcome FIs cause participants to experiment with successful task strategies, resulting in poorer task performance relative to no-FI controls (Hammond & Summers, 1972). This MCPL finding is consistent with FIT, that is, FIs cause a motivated recipient to test new hypotheses regarding more efficient ways to perform a task. The mere motivation to learn may backfire because the more varied and elaborate attempts at the task (i.e., decreased cognitive consistency) are often futile. Barley, Connolly, and Ekegren (1989) showed that an increase in motivation leads to an increase in dysfunctional strategy search.
  • Process FIs (i.e., information about judges’ policy rather than judges’ accuracy) are not detrimental to performance consistency, but their effects on learning are not clear (Adelman, 1981, Experiment 2; Lindell, learning are not clear (Adelman, 1981, Experiment 2; Lindell, 1976; Schmitt et al, 1977; Steinman, 1974). This suggests that FI effects on strategy may not always affect overall performance because the alternative strategy is equally effective. Indeed, in some tasks (e.g., dichotomous MCPL), process FIs induce different strategies than outcome FIs without a noticeable difference in overall performance (Castellan & Swaine, 1977).
  • Rven if attention is directed to learning processes, but the informational value of an FI is redundant with the preexisting knowledge, no FI effect on learning should be expected.

The MCPL literature suggests that for an FI to directly improve learning, rather than motivate learning, it has to help the recipient to reject erroneous hypotheses. Whereas correcting errors is a feature of some types of FI messages, most types of FI messages do not contain such information and therefore should not improve learning—a claim consistent with CAI research.

Moreover, even in learning situations where performance seems to benefit from FIs, learning through FIs may be inferior to learning through discovery (learning based on feedback from the task, rather than on feedback from an external agent). Task feedback may force the participant to learn task rules and recognize errors (e.g., Frese & Zapf, 1994), whereas FI may lead the participant to learn how to use the FI as a crutch, while shortcutting the need for task learning (cf. J. R. Anderson, 1987). Indeed, in one CAI experiment, it was found that FI was detrimental to the performance of transfer tasks (Carroll &
Kay, 1988). This finding suggests that FIs may reduce the cognitive effort involved in task performance [germane cognitive load] and therefore is detrimental in the long run. In FIT language, an FI may lead to the generation of a hypothesis designed to attain a goal of obtaining positive feedback, whereas no FI may lead to the generation of a hypothesis designed to attain a goal of performance improvement.

In summary, FIs affect the learning process by directing attention to discrepancies between the hypotheses (standards) regarding the details of task performance and the outcomes of acting on these hypotheses. If the FI is not accompanied with cues helping to reject erroneous hypotheses, it may cause the
recipient to generate a multitude of hypotheses that can reduce consistency and hence decrease performance. Even when the FI is accompanied by useful cues, they may serve as crutches, preventing learning from errors (natural feedback) which may be a superior learning mode.

FI Effects on Meta-Task Processes

FI cues and the outputs of task processes may divert attention up the hierarchy and away from the details of the task. This shift of attention may activate at least four interdependent mechanisms: mode of resolving feedback-self discrepancies, attention to the self, depletion of cognitive resources for task performance, and affective processes. Each of these processes is complex and interdependent.

The effects of activating self-related feedback loops are moderated by a host of variables:

  • Self-efficacy. Individuals high in self-efficacy are less likely to quit a task even in the face of failure relative to those low in self-efficacy. According to FIT, low self-efficacy is a meta-task mechanism that “releases” unresolved lower level feedback loops (Lord & Levy, 1994). Specifically, this mechanism is responsible for preventing endless attempts to reduce a feedback standard discrepancy and may be activated whenever an interruption occurs in the lower level feedback loop (Carver & Scheier, 1981).
  • Anxiety (Mikulincer, 1989a). Anxious participants whose self-related goals were activated are more likely to experience cognitive interference, that is, shifts of attention away from the task and toward the unmet goals of the self.
  • Velocity of the FIs. When more than one episode of FI is available, people can assess the rate of change in their performance. When the initial FI is very negative (i.e., large feedback-standard discrepancy), only a rapid rate of improvement leads to a decision to continue with the task, where a constant rate of improvement as well as a delayed improvement lead to a decision to withdraw (Duval, Duval, & Mulilis, 1992).

Common to all variables that affect the decision to withdraw from the task is a low expectation to achieve the standard along with a shift of attention to meta-task processes. This shift interferes with performance. There are additional variables that interfere with task performance (see Carver & Scheier, 1981):

  • Attention to the Self
    • Many FI cues may direct attention to the self (e.g., normative FI is likely to divert attention away from the task to meta-task processes such as evaluating the utility of task performance for higher order goals (e.g., making a good impression; Vallacher & Wegner, 1987).
    • Attention to the self is known to improve performance of dominant tasks and debilitate performance of nondominant tasks—which concurs with objective self-awareness theory (Wicklund, 1975) and control theory (Carver & Scheier, 1981).
    • Consistent with FIT, cues of both salient negative and salient positive FI have been implicated in shifting attention to the self.
    • Praise (a type of FI) impaired the performance of a cognitively demanding task but improved the performance on a simple task—findings interpreted in light of a self-attention model (Baumeister, Hutton, & Cairns, 1990).
  • Depletion of Cognitive Resources for Task Performance
    • The attention diverted from a resource demanding activity to the nontask aspects of the intervention (meta-task processes) may cause performance loss due to competition for cognitive resources (e.g., Kanfer & Ackerman, 1989). Only if the task is automated, and therefore fewer resources are needed for its completion, then the motivation induced by the intervention may cause people to successfully work harder.
  • Affective Processes
    • Attention to the self is likely to activate affective reactions. Affective reactions may influence the way in which the available resources are used.
    • According to FIT, most affective reactions are induced by evaluation of the feedback with respect to salient self goals—which may create several feedback signs. The feedback signs are then weighted and summed into a general feedback sign. This general feedback sign is then cognitively evaluated both for its harm-benefit potential for the self and for the need to take a new action (Kluger et al., 1994). The harm-benefit appraisal is reflected in the primary dimension of mood (i.e., pleasantness), whereas the appraisal of the need for action is reflected in a secondary dimension of mood (i.e., arousal).
    • Fl-induced affect influences performance of tasks other than the one used to induce the affect.
    • The output of an activating affective process sets different standards for cognitive operation. For example, Fl-induced pleasantness may induce a framing effect (Tversky & Kahneman, 1986), where negatively framed events promote risk seeking and positively framed events promote risk aversion. The framing effect may explain diverse findings showing that a negative FI sign is often followed by greater variance in both performance and standard setting than a positive FI sign. For example, Thorndike (as cited in Adams, 1978) found that the word wrong yielded lower performance consistency than the word right (leading both him and Skinner to concentrate on rewards rather than punishment). Similarly, Lewin et al. (1944) noted that participants receiving negative FIs set goals for the next performance episode with greater variability than those receiving positive FIs, possibly reflecting a risk-seeking strategy.

FIT: Predicting the Effects of FI on Performance

Three classes of variables determine the effect of FI on performance:

  1. The cues of the FI message, which determine which standards of the recipient will receive most attention and hence affect action (Cropanzano et al., 1993; Vallacher & Wegner, 1987.).
  2. The nature of the task performed, which determines how susceptible it is to attentional shifts.
  3. Situational (and personality) variables, which determine how the recipient chose to eliminate standard-FI gaps to which the FI brought attention.

Proposition 1: FI effects on performance are attenuated by cues that direct attention to meta-task processes. Such cues include normative FIs, person-mediated versus computer-mediated FIs, FIs designed either to discourage or praise the person, and any cue that may be perceived as a threat to the self.

Proposition 2: Fl effects on performance are augmented by (a) cues that direct attention to task-motivation processes and (b) cues that direct attention to task-learning processes coupled with information regarding erroneous hypotheses.

Proposition 3: In the absence of learning cues, the fewer cognitive resources needed for task performance, the more positive is the effect of FIs on performance.

Proposition 4: Goal-setting interventions should augment the effect
of FI on performance.

Proposition 5: FI cues that match salient self goals of a given personality
type direct attention to meta-task processes and therefore
debilitate performance.

Fl that is too specific can direct attention below the level necessary for performance, thus causing an interference (Vallacher & Wegner, 1987). In addition, the specific information may not match the natural way people represent the task cognitively and, therefore, attenuate some benefits of Fl for learning (Ganzach, 1994). Yet, the less specific Fl also led to lower estimates of variability, suggesting that specificity has complex effects on overall performance (see Lee & Yates, 1992).

The importance of task characteristics

With very rare exceptions (e.g., Mikulincer et al., 1991; Baumeister et al., 1990), Fl researchers have ignored the theoretical importance of task characteristics. This is surprising given Annett’s (1969) comprehensive review of KR studies showing that the pattern of KR effects on motor, perceptual, and verbal tasks may be different. This may be partly due to a lack of sufficient task taxonomy, which was lamented by many authors who recognized that a taxonomy is needed to establish the boundaries of the predictive power of theories in various domains (Hammond, 1992; Wood, 1986). Indeed, Annett (1969) classified KR studies into “some kind of division,” where “divisions are somewhat arbitrary” (p. 37).

However, when Fl increases motivation, subjective task complexity would moderate the effect of Fl because motivation improves performance mostly when the task requires little cognitive resources (Ackerman, 1987; Wood, Mento, & Locke, 1987).

The importance of personality

There is no doubt that personality variables moderate the reaction to FI (Ilgen et al., 1979). Among the personality variables that are known to be involved in the reaction to Fl are self esteem (e.g., Ilgen et al., 1979), locus of control (e.g., Ilgen et al., 1979), tendency for cognitive interference (Kuhl, 1992;Mikulincer, 1989a), and altruism (Korsgaard, Meglino, & Lester, 1994). For example, negative FI is more likely to direct attention to the self among participants low in self-esteem than among those high in self-esteem, but positive FI may have the opposite effect.

Meta-Analytic Moderator Analyses: Moderators

  • FI Cues: sign, content, frequency
  • Task Characteristics: novelty, complexity, time constraint, task duration, creativity, quality-quantity, ratings-objective performance, transfer, latency, task type (physical, reaction time, memory, knowledge, following rules, vigilance
  • Situational and Methodological Variables: goal setting, threat to self esteem, External rewards-punishments, Experimental control, lab-field


1) Four moderators showed systematic relationships with d regardless of the exclusions and suggest that discouraging FIs attenuate FI effects (consistent with P1), that velocity FIs and correct solution FIs augment FI effects (consistent with P2a and P2b), and that FI effects on performance of physical tasks are lower than FI effects on other tasks—a finding that we did not anticipate.

2) Six moderators became significant after all the exclusions. These moderators suggest that praise, FIs threatening self-esteem, and verbal FIs attenuate FI effects (consistent with PI) that FIs with frequent messages augment FI effects (not predicted), and that FI effects are stronger for memory tasks and weaker for following rules tasks (not predicted). [Committing to memory and following/remembering rules occur via different mechanisms?]

3) Three moderators almost reached significance criteria of .01 (i.e., ps < .05): Computerized FI yielded stronger FI effects (consistent with P2); FIs on complex tasks yielded weaker effects (P3); and FIs were more effective with a goal-setting
intervention (P4). Yet, these effects should be treated with extra caution.

Moderator Analyses: Major Conclusions

1) Several FI cues that seem to direct attention to meta-task processes attenuate FI effects on performance, whereas several FI cues that seem to direct attention to task-motivation or task learning processes augment FI effects on performance. Specifically, both praise and FI designed to discourage were postulated to increase attention to meta-task processes and were found to attenuate FI effects.

“…ceteris paribus (everything else being equal), FI cues affect performance by changes in locus of attention: The lower in the hierarchy the Fl-induced locus of attention is, the stronger the benefit of an FI for performance” (p. 275).

2) FI effects are moderated by the nature of the task. However, the exact task properties that moderate FI effects are still poorly understood. Simple-task performance benefitted from Fls (marginally) more than complex-task performance. Task complexity had relatively low interjudge reliability (.70), reflecting perhaps the difficulty in conceptualizing task complexity (and other task dimensions) and therefore suggesting that the effect that we observed is an underestimate. Also, the performance of novel tasks seemed to be debilitated when performance was measured for a short time (i.e., performance in the initial stages of task acquisition). This effect implicates meta-task processes that render the performance of subjectively complex tasks susceptible to interference from interventions such as FI. Our findings provide only weak evidence for P3 that task type and its level of mastery play an important role in determining the effect of FI on performance.

Research gap: Although the moderating effects of task features identified by
FIT received weak support, several task dimensions moderated FI effects unexpectedly: Physical tasks and following rules tasks yielded weaker FI effects, and memory tasks yielded stronger FI effects. Our results strongly suggest that task type places a serious boundary condition on the knowledge of effectiveness of various interventions designed to improve performance (cf. Hammond, 1992). Therefore, the lack of a valid task taxonomy that can be used across vastly different tasks (e.g., vigilance, memory, and adherence to regulations) poses a serious obstacle for FI research. Moreover, even within similar types of tasks
(MCPL), the “effects of feedback seem to be very sensitive to the task environment [difficulty]” (Castellan & Swaine, 1977, p. 118).

Limitation of many studies: “…many researchers of FIs still implicitly assume that FIs increase performance, and therefore they limit their studies to comparisons of several types of FIs (e.g., a positive vs. negative FI). Without control groups, we may know more about the relative merits of several types of FI messages, but we have no idea if they are better, equal, or inferior to no intervention” (p. 276).

Implications: The identification of a number of moderators suggests that in certain situations, FI can yield a large and positive effect on performance. Specifically, an FI provided for a familiar task, containing cues that support learning, attracting attention to feedback-standard discrepancies at the task level
(velocity FI and goal setting), and is void of cues to the metatask level (e.g., cues that direct attention to the self) is likely to yield impressive gains in performance, possibly exceeding 1 SD.

When an FI increases performance through an increase in task motivation, the effect may depend on a continuous FI. Removal of such an FI may result in a reversal (Komaki et al., 1980). Therefore, the cost of maintaining a continuous intervention should be considered in evaluating such an intervention. If, however, FI affects performance through task-learning processes, the effect may create only shallow learning and interfere with more elaborate learning. Lack of elaborate learning affects the ability to use the learned material in transfer tasks  (e.g., Carroll & Kay, 1988).

In the MCPL literature, several reviewers doubt whether FIs have any learning value (Balzer et al., 1989; Brehmer, 1980) and suggest alternatives to FI for increasing learning, such as providing the learner with more task information (Balzer et al., 1989). Another alternative to an FI is designing work or learning
environments that encourage trial and error, thus maximizing learning from task feedback without a direct intervention (Frese & Zapf, 1994).

“…additional development of FIT is needed to establish the circumstance under which positive FI effects on performance are also lasting and efficient and when
these effects are transient and have questionable utility. This research must focus on the processes induced by FIs and not on the general question of whether FIs improve performance— look at how little progress 90 years of attempts to answer the latter question have yielded” (p. 278).


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: