Behavior Modification

The effects of rewards on people’s test scores has been the subject of many studies. Under the headline “Learn-&-Earn Plan Pays Off: Scores Soar At Cash-For-Kids Schools” In The New York Post Kelly Magee and Yoav Gonen reported about the results of a program that rewarded students for obtaining higher scores on tests.

An overwhelming number of schools participating in a controversial program that pays kids for good grades saw huge boosts — up to nearly 40 percentage points higher — in reading and math scores this year, a Post analysis found.

About two-thirds of the 59 high-poverty schools in the Sparks program — which pays seventh-graders up to $500 and fourth-graders as much as $250 for their performance on a total of 10 assessments — improved their scores since last year’s state tests by margins above the citywide average.

The gains at some schools approached 40 percentage points.

The report about the effects of this public policy intervention for improving students’ scores is news, at least at present. However, it turns out that these sorts of data have been available for quite a while.

Clingman and Fowler (1976) examined the effects of rewards on the scores of primary aged children with different pre-test scores taking a vocabulary test. First they determined the children’s scores on one version of the test, creating groups of children whose scores were in the top, middle, or bottom third for the participating students. Four weeks after they established the children’s scores, they administered the test again under different conditions. Some children in each group took the test under the same standard conditions (“no reward”); some took received tiny candies for each correct answer (“contingent reward”); and some received a number of candies equal to the number received by a peer in the reward condtion, but the candies were given to them all at once (“non-contingent reward”).

From Clingman & Fowler, 1976

The accompanying graph shows the results of the study by Clingman and Fowler (1976). The metric is the number of points difference between the two testings; if a student’s score stayed essentially unchanged (i.e., was 100 on both tests), he would be coded as having a change of “0.” It is interesting to note that most children has lower scores on the second testing. However, the change in scores for the children with the lowest initial scores when they received rewards is striking. Technically, this is called an “interaction”: One condition (rewards) affected one of the groups (low pre-test), but other conditions made little difference for any of the groups.

There are many other studies examining similar phenomena. I’ve listed only a few here for the purposes of illustration; there are entire literatures that are not represented in this list (e.g., group-vs-individual rewards; rewards and intrinsic motivation), so please don’t take this as a thorough analysis. It’s just a hint. But, do take it as an indicator of the generality of the finding.

The main idea is that test scores are affectedd by the conditions in effect when the testing occurs. Probably some children perform well regardless of the conditions; we might infer that they have learned to work hard during testing. But some other students only show their best performance when the testing conditions are optimal. As educators, we ought to (a) ensure that there are such optimal conditions in place during testing and (b) teach students to perform their best regardless of conditions.

Link to the NY Post article. Flash of the electrons to Joanne Jacobs for an entry in her blog that tipped me to the report.

References

Miller, J., & Eller, B. (1985). An examination of the effect of tangible and social reinforcers on intelligence test performance of middle school students. Social Behavior and Personality, 13, 147-157

Traub, R., Hambleton, R., & Singh, B. (1969). Effects of promised reward and threatened penalty on performance of a multiple-choice vocabulary test. Educational and Psychological Measurement, 29, 847-861.

Taylor, C., & White, K. (1982). Effects of reinforcement and training on group standardized test performance. International Journal of Rehabilitation Research, 5, 362-367.

Terrell, F., Terrell, S., & Taylor, J. (1981). Effects of type of reinforcement on the intelligence test performance of retarded Black children. Psychology in the Schools, 18, 225-227.

Maheady, L., Sainato, D., & Maitland, G. (1983). Motivated assessment: The effects of extrinsic rewards on the individually-administered reading test performance of low, average, and high IQ students. Education & Treatment of Children, 6, 37-46.

Kieffer, D., & Goh, D. (1981). The effect of individually contracted incentives on intelligence test performance of middle- and low-SES children. Journal of Clinical Psychology, 37, 175-179.

Clingman, J., & Fowler, R. (1976). The effects of primary reward on the I.Q. performance of grade-school children as a function of initial I.Q. level. Journal of Applied Behavior Analysis, 9, 19-23.

Fowler, R., & Clingman, J. (1977). The influence of intrinsic and extrinsic reward on the intratest performance of high- and low-scoring children. Psychological Record, 27, 603-610.