under 21 years of age but to reduce effectiveness among older recipients (perhaps due
to reactance). Thus age moderates the relationship between inclusion of normative
arguments and intervention effectiveness. Similarly, condom-use skills were effective
for men but not women. Thus gender moderated the relationship between condom
use skills training and intervention effectiveness. This emphasizes the importance of
the technique–recipient fit in intervention design.
Albarracín et al.(2005) highlight the potential utility of meta-analyses when
undertaking such ‘retrospective process analyses’. The results help intervention designers
identify candidate change processes and techniques in stage 3 of the planning process
when targeting increased condom use. However, such reviews also need to be
considered cautiously. Here we will highlight five questions that can be asked to assess
the utility of such studies. First, is the sample of studies large enough? Albarracín et al.
(2005) used a very large and representative sample of intervention descriptions. Small
samples may generate results that could be reversed as more tests became available.
Second, are the intervention descriptions accurate? Content analyses of interventions
may be based on categorization of the intervention material themselves (e.g. Abraham
et al., 2007) or detailed manuals. Descriptions from scientific papers may be less accurate
and less comprehensive. For example, Abraham and Michie (2008) found that while
approximately 9 change techniques per intervention were identified in descriptions
included in manuals, only 6, on average, were identified when categorizing content
of the same interventions from descriptions included in scientific papers.
Third, are the categorizations used derived from theorized change process and specific
enough to provide guidance to designers? Albarracín et al.(2005) used a simple set of
techniques derived from theorized change processes that were relevant to the behaviour
change target of interest, namely, condom use. This is helpful to designers. However,
when broad technique categories are employed this creates interpretation problems
because researchers may have categorized distinct techniques into one cate gory. This
is sometimes referred to as the ‘apples and pears’ problem, that is, the specificity of
the categorization operates at the level of ‘fruit’ rather than particular types of fruit. For
example, Albarracín et al.’s (2005) results do not tell us whether normative arguments
targeting descriptive and subjective norms (see Chapter 7) operate differently. Do
both types of normative techniques promote condom use equally well and are both
moderated in their effectiveness by age? Thus ideally technique categories used for
intervention content analyses should be linked to particular change processes and
be very specific. Even when this is achieved we need to be aware (as Albarracín et al.,
2005) note) that one change technique may operate through multiple change processes.
For example, changing descriptive norms (that is promoting the belief that most people
like me are doing X) may also change attitudes (doing X is rewarding).
Fourth, were the defined categories reliably applied by independent coders? Clearly
if two people reading the same descriptions cannot easily agree on what it refers to
then no further conclusions can be drawn about the categorizations undertaken
(Abraham et al., 2015).
Fifth, were all relevant characteristics of the intervention categorized? The answer
to this question is unlikely to be ‘yes’ but the question highlights the importance of
third variable effects, that is, where another feature of a set of interventions, other than
those categorized, is responsible for an observed association between change technique
inclusion and effectiveness. Albarracín et al.(2005) categorized aspects of the delivery
222 MOTIVATION AND BEHAVIOUR