The Strength of the Relation Between Performance-Approach and Performance-Avoidance Goal Orientations: Theoretical, Methodological, and Instructional Implications

Abstract

In current research on achievement goal theory, most researchers differentiate between performance-approach and performance-avoidance goal orientations. Evidence from prior research and from several previously published data sets is used to highlight that the correlation is often rather large, with a number of studies reporting correlations above .50. The large magnitude of this correlation raises questions and warrants further investigation. The size of the correlation also varies substantially across studies; thus, several potential moderators were considered. Minimal evidence for moderation was found, with little variability in relations as a function of fear of failure, culture, and specificity of the goal assessment. There was some evidence of variability in the correlation based on age, perceived competence, and assessment instrument. The article concludes by highlighting theoretical, methodological, and instructional questions that arise as a result of the large correlation and making recommendations and guidance for research, instructional practice, and theory advancement.

ACKNOWLEDGMENTS

We thank Amanda Durik, Andrew Elliot, and Allison Ryan for their helpful feedback and suggestions on an earlier version of this manuscript.

Notes

In our review of this work, we focus on classroom and school goal structures rather than experimental laboratory studies, as the majority of the laboratory studies either assign students goals or attempt to activate certain goal orientation “schemas” and thus do not provide clear evidence regarding how the learning context shapes goal orientations (Linnenbrink, 2004).

²Hulleman et al. (2010) found that the strength of the correlation between achievement and performance-approach goals varied as a function of scale type (AGQ: = .14, Patterns of Adaptive Learning Survey [PALS]: = –.01; other published: = .01; custom: = –.02) and whether the majority of the scale items assessed normative ( = .14), evaluative ( = –.14), performance no-goal ( = .01), or no clear majority ( = .03). In addition, the correlation between performance-avoidance goals and achievement varied based on scale type ( = –.20, PALS: = –.13; other published: = –.08; custom: = –.09); there were no significantly differences based on the coding of the goal items as normative, evaluative, and so on. For interest and achievement goal type, there were no significant moderators for performance-approach goals. For performance-avoidance goal orientations, results varied based on scale type (AGQ: = –.20, PALS: = –.05; other published: = –.11; custom: = –.02). The correlation between performance-avoidance goals and interest decreased when all of the items were coded as assessing performance-avoidance ( = .03) in comparison to no items assessing performance-avoidance directly ( = –.13).

³Huang (2011) reverse-scored indicators of negative affect such that a negative correlation indicates a positive association between the achievement goal and negative affect.

These data sets are not meant to be representative of all studies, but as a group they are similar to the samples used in much of the extant literature.

TABLE 1 Descriptors of Previously Published Data Sets Reanalyzed for Moderator Analyses

We thank an anonymous reviewer for bringing this important point to our attention.

⁶Significant levels between pairs of correlations were calculated using Fisher's r to Z transformation; p value cutoffs were adjusted using the Bonferroni adjustment.

We were not able to statistically compare the strength of these correlation coefficients, which were taken from the same sample rather than independent samples, as there was not sufficient information reported about the intercorrelations across subject domains in Bong (2005). However, the overlapping confidence intervals suggest the correlations are similar.

Because both the original and revised scales appear in Midgley et al. (2000), it is very difficult to determine which scales researchers have employed in their research as researchers typically cite the 2000 manual but do not clearly state the version of the scale. Hulleman et al. (2010) did not differentiate between these two versions of PALS in their meta-analysis, likely due to this difficulty in determining which scale was used. Thus, we cannot clearly analyze whether the strength of the correlation shifts between the original and revised version.