How behavioural science can fail the marshmallow test
Focusing too heavily on cognition means we can miss the bigger picture
One of the most important issues that we see in behaviour change work is the way the challenge can be defined in a narrow way – focusing in on proximal influences rather than broader, and potentially more important distal influences. So for example, with vaccine hesitancy we may see it as a function of the way we evaluate risk rather than also considering a wider range of beliefs, attitudes and values that may result in that risk evaluation. While this can lead us to focus on things which can be easier to influence, the downside is that it may also result in any intervention having less impact on the desired outcomes.
To illustrate this, let’s look at the famous marshmallow test. Invented by Walter Mischel in the 1960s, it is designed to assess self-control in children. The simple premise is that if the child is able sit alone for several minutes without eating a marshmallow, they can eat two marshmallows when the experimenter returns. So there is a rewards for self-control. After many years Mischel followed up with the children in the original experiments and found that children who were better at holding out for two marshmallows were higher achieving at school. The clear takeaway from this research was that self-control plays an important role in life outcomes.
However, what has become apparent through work by Celeste Kidd and colleagues was that wait-times are modulated by a decision-making process that considers the reliability of the environmental. Children that did the marshmallow test in a condition where the researchers has demonstrated their reliability, waited significantly longer than those in the unreliable condition. This suggests that children’s wait-times reflected perfectly rational beliefs about whether waiting would ultimately pay off.
Other work by Tyler Watts and colleagues supported this finding, with children from poorer homes having more difficulty resisting the treats than affluent children. Because of course lower affluence typically means less reliability. So it would not make sense to wait, as if so there may be no marshmallow for you, let alone two. This is not to say that self-control is unimportant, but it does not appear to be the primary causal mechanism implicated in children’s wait times.
There are a number of useful lessons here for the behaviour science practitioner:
Beware only focusing on proximal influences: We often see that both proximal (in this case self-control) and distal influences (such as socio-economic status) have roles to play and interact with each other
Lookout for multiple influences on behaviour: There can be a narrowing of perspective, often with cognition as the main focus – arguably the typical experimental paradigm tends to encourage a narrowing of focus as we seek to understand the impact on one variable at a time
Short versus long term: Distal influences typically do not have quick fixes, so it is understandable why there may be a focus on proximal influences that can look easier to achieve short term albeit marginal gains; but more needs to be done to understand how to build gradual behaviour change through tackling these more substantial distal influences
The lessons of the marshmallow experiment and its subsequent reappraisal make the case for understanding the socially embedded nature of behaviour. This is a point that is easily missed if your points of reference are a relatively narrow reading of the psychology literature, typically focused on cognition and decision making. While this is necessary, it is far from sufficient.
This is why we use market and social research methods in conjunction with a behaviour change system (we use the Ipsos MAPPS system). This allows us to properly understand the range of influences on behaviour within a clear theoretically sound framework of understanding. From this, we can then carefully design a behaviour change programme which has considered the range of proximal and distal influences and which are most effective to then tackle using intervention strategies, with the resources available. In our experience this rounded approach offers a more holistic and, therefore, effective approach to behaviour change challenges.
Keep an eye out for the next article which will be on considerations for the design of a behaviour change programme which tackles both distal and proximal influences on behaviour. This will be followed by an article on the implications of all this discussion for how we undertake intervention testing (and why we need to go beyond the Randomised Control Trial).