Lesson 11: Context-dependent memory and Grant et al. (1998)
Interpret data from a modified Grant et al. (1998) experiment
SkillsContext-dependent memoryMemory research and data
Worked example
A student modified Grant et al. (1998). Twenty Year 12 students read a one-page article. Ten, chosen at random, read it and took a 20-question test with the same instrumental music playing (same context); the other ten read it with music and took the test in silence (different context). Test scores:
| Same context | 14 | 12 | 16 | 11 | 15 | 17 | 12 | 14 | 13 | 15 |
|---|---|---|---|---|---|---|---|---|---|---|
| Different context | 12 | 11 | 15 | 10 | 14 | 13 | 9 | 13 | 12 | 12 |
Same context: standard deviation 1.91. Different context: standard deviation 1.79. An unpaired t-test gave , .
(a) State a null and an alternative hypothesis. (b) Calculate the mean score of each group. (c) Draw a conclusion. (d) Suggest one way the uncertainty in the conclusion could be reduced.
Step 1: (a) Hypotheses
Null: there will be no difference in mean test score between students tested in the same context as study and students tested in a different context. Alternative: students tested in the same context will have a different mean test score from students tested in a different context.
Step 2: (b) Same context mean
Step 3: (b) Different context mean
Step 4: (c) Conclusion
The same-context group scored 1.8 marks higher on average, and , so the difference is statistically significant. Reject the null hypothesis: matching the study and test contexts improved test scores, which supports context-dependent memory as in Grant et al. (1998).
Step 5: (d) Uncertainty
The p-value is close to 0.05 and each group had only 10 students, so there is a real chance this is a type I error (a false positive). Repeating the experiment with a larger sample, or with all four conditions of the original 2 × 2 design, would reduce the uncertainty.
A small sample gives large standard errors; with p near the threshold, a replication would show whether the effect is reliable.
Answer
(a) Null: no difference in mean scores between same- and different-context groups; alternative: a difference. (b) Same 13.9, different 12.1. (c) p = 0.043 < 0.05: significant; reject the null; matching contexts improved scores. (d) Larger sample or replication (p is near 0.05 with n = 10 per group, so a type I error is possible).
Try a similar one
A fresh question of the same kind, checked as you answer, with hints and a full solution.
Method
Hypotheses name the IV and the DV and the population.
Means from totals; compare p with 0.05.
Qualify the conclusion: sample size, p near the threshold, one kind of context.
Watch out for
Saying the experiment proves the hypothesis: it supports it.
Writing a directional alternative hypothesis when the test was two-tailed.