Lesson 10 of 11 · 20 min

Investigating divided attention and memory: means, SD and the t-test

Replicate an aspect of Craik et al. (1996) with an independent groups experiment, then describe the results with the mean, standard deviation and error bars, and test them with an unpaired t-test.

What you will be able to do

  • Describe the aim, method and main finding of Craik et al. (1996).
  • Identify and operationalise the IV and DV, controlled variables and hypotheses of a class replication with independent groups.
  • Calculate the mean and standard deviation, and explain what standard error and a 95% confidence interval show.
  • Draw and read a column graph with error bars, and interpret an unpaired t-test: t, df, p and a conclusion.

Can you learn while doing something else? Students often study with a phone buzzing beside them. Fergus Craik and his colleagues tested what dividing attention does to memory, and their experiment is one you can run in class.

History

Craik, Govoni, Naveh-Benjamin and Anderson (1996)

Aim: to find out whether dividing attention affects memory more when information is being learnt (encoding) or when it is being remembered (retrieval).

Method: university students heard lists of words. In some conditions they learnt the words, in others they recalled them, while doing a second task at the same time: watching a screen and pressing the key that matched where a mark appeared, as fast as possible. Their memory and their speed on the second task were measured with full and with divided attention.

Findings: dividing attention while learning greatly reduced how many words were later remembered. Dividing attention during recall had a much smaller effect on memory, although it slowed the second task.

Conclusion: encoding needs attention; retrieval is more automatic and is protected at the expense of the other task.

Interactive

Try it yourself

Learn one list with full attention and one while responding to a dot. Then compare your recall with Craik et al. (1996). Your results stay in your browser and are not a diagnosis.

A class replication. Your class replicates the encoding part. Sixteen Year 11 students are randomly allocated to two groups of eight. Everyone hears the same 20 common nouns read at one word every 2 seconds, then has 2 minutes to write down as many as they can, in any order (free recall).

  • Full attention group: listens to the list and does nothing else.
  • Divided attention group: sorts a shuffled pack of playing cards into suits while listening.
The design of the class experiment
FeatureIn this experiment
Independent variable (IV)Attention while learning: full (listening only) or divided (sorting cards by suit while listening)
Dependent variable (DV)Number of words correctly recalled out of 20 in 2 minutes of written free recall
Controlled variablesThe same 20-word list, the same rate (one word every 2 s), the same 2-minute recall, the same room and time of day, standardised recorded instructions
DesignIndependent groups: each student is in one condition only
AllocationRandom allocation (names drawn from a hat), so differences in memory ability are spread across both groups by chance
Null hypothesis (H₀)There will be no difference in the mean number of words recalled by Year 11 students who learn the list with full attention and those who learn it while sorting cards.
Alternative hypothesis (H₁)Year 11 students who learn the list while sorting cards will recall a different mean number of words from those who learn it with full attention.

Note

Why independent groups here?

With repeated measures, each student would need two different lists, and the second time they would know what is coming (an order effect: practice or fatigue). Independent groups avoid order effects, but the groups may differ in memory ability, so random allocation matters. A repeated measures version is possible with two equally hard lists and counterbalancing (half do full attention first).

Match each part of the class experiment to its role.

The results. Each number is one student's score: words recalled out of 20.

Words recalled out of 20
GroupScoresTotal
Full attention14, 12, 15, 11, 13, 16, 12, 15108
Divided attention9, 11, 8, 10, 12, 7, 9, 1076

Definition

The mean and the standard deviation

The mean is the average score: xˉ=∑xn\bar{x} = \dfrac{\sum x}{n}. For the full attention group, xˉ=1088=13.5\bar{x} = \dfrac{108}{8} = 13.5 words; for the divided group, 768=9.5\dfrac{76}{8} = 9.5 words.

The standard deviation (ss, SD) measures how spread out the scores are around the mean: roughly the typical distance of a score from the mean.

s=∑(x−xˉ)2n−1s = \sqrt{\frac{\sum (x - \bar{x})^2}{n - 1}}

Psychology uses this sample SD, which divides by n−1n - 1. On a calculator it is sxs_x (not σx\sigma_x); in a spreadsheet STDEV.S.

Standard deviation by hand: the full attention group

  1. Find the mean

    xˉ=13.5\bar{x} = 13.5.

  2. Subtract the mean from each score

    0.5,−1.5,1.5,−2.5,−0.5,2.5,−1.5,1.50.5, -1.5, 1.5, -2.5, -0.5, 2.5, -1.5, 1.5 (they add to 0).

  3. Square each deviation and add

    0.25+2.25+2.25+6.25+0.25+6.25+2.25+2.25=220.25 + 2.25 + 2.25 + 6.25 + 0.25 + 6.25 + 2.25 + 2.25 = 22.

  4. Divide by n − 1

    228−1=227=3.142…\dfrac{22}{8 - 1} = \dfrac{22}{7} = 3.142\ldots (the variance).

  5. Take the square root

    s=3.142…=1.77s = \sqrt{3.142\ldots} = 1.77 words.

For the divided attention group (9, 11, 8, 10, 12, 7, 9, 10; mean 9.5), the squared deviations add to 18. Calculate the standard deviation, to 2 decimal places.

in words

Round to 2 decimal places.

From the sample to the population. The class is a sample. If another eight students had been in each group, the means would have come out a little differently. Two statistics say how far to trust a sample mean.

  • The standard error SE=snSE = \dfrac{s}{\sqrt{n}} estimates how much the mean would vary from sample to sample. Full attention: 1.778=0.63\dfrac{1.77}{\sqrt{8}} = 0.63; divided: 1.608=0.57\dfrac{1.60}{\sqrt 8} = 0.57. It shrinks as the sample grows.
  • The 95% confidence interval is xˉ±1.96×SE\bar{x} \pm 1.96 \times SE: a range that would contain the true population mean in 95% of samples. Full attention: 13.5±1.2313.5 \pm 1.23, from 12.27 to 14.73 words; divided: 9.5±1.119.5 \pm 1.11, from 8.39 to 10.61 words.
Summary statistics (sample SD; 95% CI = mean ± 1.96 SE)
GroupnMeanSDSE95% CI
Full attention813.501.770.6312.27 to 14.73
Divided attention89.501.600.578.39 to 10.61

Mean words recalled out of 20. Error bars show ±1 standard deviation.

Key idea

Reading error bars

Always say what the error bars show: SD (the spread of the scores), SE or a 95% CI (how precise the mean is). Bars that do not overlap suggest a real difference; bars that overlap a lot suggest it may be due to chance. Only a statistical test can say how likely chance is.

The t-test. The difference between the means is 4 words. Could two random groups of eight differ this much by chance, if divided attention made no difference at all? A two-sample t-test answers that. It compares the difference between the means with how much the scores vary within the groups.

Because each student was in one group, the scores are unpaired: use the unpaired (independent samples) t-test. In a spreadsheet: =T.TEST(full, divided, 2, 2) (two-tailed, equal variances).

Result: t=4.73t = 4.73, df=8+8−2=14df = 8 + 8 - 2 = 14, p=0.0003p = 0.0003.

Key idea

The decision and the conclusion

Decision: p=0.0003p = 0.0003, which is less than 0.05, so the difference is statistically significant. Reject the null hypothesis.

Conclusion: students who sorted cards while learning recalled significantly fewer words (mean 9.5, SD 1.60) than those who learnt with full attention (mean 13.5, SD 1.77). Dividing attention while learning reduced recall, as Craik et al. (1996) found for divided attention at encoding.

Worked example

Interpret an unpaired t-test on divided attention at recall

  • Complex familiar
  • Technology active

Another class tested divided attention at retrieval. Twenty students learnt the same 20-word list with full attention, and were then randomly allocated to recall it either with full attention or while sorting cards.

Recall conditionnMean words recalledSD
Full attention1012.12.0
Divided attention1011.42.2

An unpaired t-test gave t=0.74t = 0.74, df=18df = 18, p=0.47p = 0.47.

(a) Explain why an unpaired t-test was used. (b) Interpret the result. (c) Explain how the result relates to Craik et al. (1996).

4 steps

Try it yourself first, then check the working one step at a time.

Method

  1. Independent groups → unpaired; repeated measures or matched participants → paired.

  2. Compare pp with 0.05 and state the decision about H₀.

  3. Conclude in words with the means, and relate to the original study.

Watch out for

  • Saying the result "proves" divided attention has no effect at retrieval.

Interactive

Run the test yourself

The class's data. Switch the error bars between SD, SE and the 95% CI, then edit scores and watch tt and pp.

Common mistake

Significant is not the same as proved

"p<0.05p < 0.05" means a difference this large would be unlikely if the IV had no effect. It does not prove the hypothesis; with p=0.0003p = 0.0003 there is still a tiny chance the result is a fluke. Write that the results support the alternative hypothesis.

Note

Ethics in the class experiment

Students give informed consent in writing (with written permission from a parent or carer for students under 16, as QCAA's guidelines require), know they may withdraw at any time, and their scores are kept confidential (codes, not names). Explain the aim afterwards in a debriefing. Nothing in this experiment should cause harm or distress.

Summary

  • Craik et al. (1996) found that divided attention while learning greatly reduces memory, while divided attention during recall affects memory much less.

  • In the class replication the IV is full or divided attention while learning, the DV the number of words recalled; the design is independent groups with random allocation.

  • Mean =∑x/n= \sum x / n; the sample SD divides by n−1n - 1; SE=s/nSE = s/\sqrt{n}; 95% CI =xˉ±1.96 SE= \bar{x} \pm 1.96\,SE.

  • A column graph shows the means, with error bars labelled as SD, SE or 95% CI.

  • For independent groups, use the unpaired t-test; if p<0.05p < 0.05, reject H₀ and conclude the IV had a significant effect.

Finished the lesson?

Try the 2 quick checks above and the lesson completes itself, or mark it complete now.

Lesson 10 of 110% read

Save your progress with a free account