Health Psychology, 2nd Edition

social influence tactics, including being a credible source, being empathetic, using
flattery, agenda setting, using metaphor, using story telling, fear appeals and making
people feel guilty. Here we will consider some key principles that need to be followed
if you want to motivate someone else to change their behaviour.

Message framing

Prospect theory (Tversky and Kahneman, 1981) predicts that people process
information differently depending on whether it relates to losses (or costs) or gains

170 MOTIVATION AND BEHAVIOUR

Meta-analysis

Meta-analysis allows researchers to examine the average size of associations (or

differences) between two measures or conditions across a series of studies. When

combining effect sizes (e.g. the size of an association between two measures)

across studies it is important to weight studies for the number of participants

because a finding based on 100 participants is likely to be more reliable than one

based on 10. For example, if a correlation study found that knowledge and

behaviour were correlated at r =.05 among a sample of 100 people and another

study found the correlation to be r=.07 among a similar sample of 100 people then

the average weighted correlation (r+) would be 0.06, across these two studies.

However, if the second study had only recruited a sample of 50 then r+would be

0.0425. In this way, meta-analyses allow researchers to combine results across

studies to calculate average ‘effect sizes’. Cohen (1992) has provided guidelines for

interpreting the size of sample-weighted average correlations (r) suggesting that

r+=.10 is ‘small’, r+=.30 is ‘medium’, while r+=.50 is ‘large’.

Meta-analyses can also be used to estimate the average impact that behaviour

change interventions (see Chapter 9) have on a particular outcome (such as physical

activity levels). In this case, the most usual measure is d (rather than r). dis

calculated by subtracting the behavioural outcome score of one condition from

another (for each study) and dividing this by the standard pooled deviation, that is, the

standard deviation for both conditions combined (Hedges and Olkin, 1985). For

example, if, at follow-up, those who had received an intervention were found to

exercise 3 times a week on average while those in a no-intervention control group

exercised only 1.5 times a week on average and there was an overall standard pooled

deviation of 1.5 this would generate a dvalue of 1.0. Cohen (1992) suggests that ds

of 0.2 are ‘small’ while ds of 0.5 are of ‘medium’ size and ds of 0.8 are ‘large’ effect

sizes. Thus this would be a large and impressive effect size. This is not surprising

because in this example the intervention doubled the rate of the target behaviour

compared to the control group, indicating an unusually successful intervention.

RESEARCH METHODS 8.1