fMRI explained

You want to understand why people respond differently to two offers, a brand or a price. A brain scan may seem an attractive way to get closer to that response. But what does the scan actually show, and what additional information do you need before basing a business decision on it?

fMRI can help answer a focused research question about brain processes. Its value comes from combining an appropriate task, careful analysis and a bounded interpretation. A coloured brain map does not, by itself, explain the behaviour of your target group.

fMRI stands for functional magnetic resonance imaging. The commonly used BOLD method uses changes in a blood-related MRI signal to investigate brain activity indirectly. It does not directly record thoughts or individual nerve impulses.

This article concerns BOLD-fMRI in behavioural research. The method can support questions about the spatial organisation of responses during a task. Whether that provides useful information for your decision depends on the comparison and the evidence supporting its interpretation. Ogawa et al., 1990.

In 1990, Ogawa and colleagues described MRI contrast dependent on blood oxygenation. Their animal research was an important foundation for BOLD: blood oxygenation level dependent. That historical contribution does not validate a commercial application. Ogawa et al., 1990.

When reading research, distinguish three levels: the physical signal, its relationship with neural processes and its relationship with behaviour. Strong evidence at one level does not automatically fill a gap at another. For a pricing decision, for example, you need to know how the selected measure relates both to the task and to the decision you eventually want to support outside the scanner.

Logothetis and colleagues compared electrical recordings with BOLD responses in monkeys’ visual cortex. The relationship with local field potentials highlights that BOLD does not simply represent the output of firing neurons. Logothetis et al., 2001.

The haemodynamic response unfolds over seconds; it does not provide a timeline of individual neural events. Logothetis discusses the interpretive constraints of this blood-related signal in his review of fMRI. Logothetis, 2008.

When reviewing a proposal, ask how the presentation of information, the response and the analysis fit together. If a participant sees a price shortly after an image, the design needs to account for the relationship between those events. A report should explain the comparison that was analysed, rather than only stating when something appeared on screen.

The table below is an editorial decision aid. It is not a universal ranking of research methods.

A research plan can combine several sources of information. Give each method a distinct question to answer. Adding equipment does not automatically strengthen the study. The combination is valuable when the findings support a better-founded decision together.

These steps are practical briefing considerations, not a complete scanner protocol.

1. Describe the decision. Specify the uncertainty you want to reduce. “Insights into the brain” is too broad; a specific comparison provides direction.
2. Define the task. Establish what people see, hear or choose and which differences between conditions matter. Check that participants understand the task.
3. Select complementary outcomes. Decide which choice data, response measures or ratings are needed to interpret the brain measurement.
4. Specify the analysis. Separate planned tests from exploratory analyses. Agree how unexpected findings will be reported.
5. Plan quality and feasibility. Discuss movement control, exclusions, eligible participants, sample requirements and scanning and analysis time with the research team.
6. Define the next step. State in advance which findings would lead to further research, an adjustment or retaining the current approach.

A brief becomes more concrete when it also states what will remain unknown afterwards. A task with two fixed price options does not automatically establish the optimal price for every customer and circumstance. Defining that boundary prevents a narrow experiment from having to support a much broader promise after the event.

Knutson and colleagues investigated product and price evaluation in a purchasing task. Brain responses before the decision contributed to predicting purchases within that experiment. This does not establish that the same approach predicts market sales. Knutson et al., 2007.

For your brief, the central question is what you intend to predict. A difference between conditions within one task is distinct from a prediction for new people, new products or a later occasion. Ask which data were used to develop the model and which remained independent for testing.

Also ask whether the additional measurement adds value beyond information that can be collected more simply or at lower cost. Relevant alternatives might include actual choices, purchase history or focused ratings. Added value needs to emerge from the relevant comparison; using a brain scanner is not evidence of it in itself.

Poldrack describes reverse inference: inferring a particular mental process from activity in a brain region. How informative that inference is depends partly on how selectively the region is associated with that process. Poldrack, 2006.

Examine the reasoning behind labels such as desire, trust or resistance. Which task, comparison and complementary data support the label? Could the observed response also fit a different explanation? An impressive image does not remove those questions.

Ask for a clear distinction between observation, interpretation and recommendation in the report. “We found a difference in the predefined comparison” is a different statement from “this target group trusts the brand”. The second requires evidence that specifically concerns trust.

Fictional example; not a Neurofactor client case or a research result. A supplier wants to investigate how two ways of presenting the same price relate to the evaluation of an offer. In one condition, participants see only the total amount; in the other, they also see an explanation of what is included.

The research team first asks whether people understand the explanation and which option they choose. It then considers whether an additional fMRI question is relevant and feasible. The design needs to account for the extra text in the second condition. Otherwise, a difference might reflect reading or task demands instead.

This example has no winning condition and no conversion percentage. It illustrates decisions needed before measurement. If the team cannot make a convincing case for the additional brain measurement, a behavioural study may remain the appropriate first step.

Poldrack and colleagues discuss low statistical power, analytical flexibility and insufficient replication among the risks in neuroimaging research. Their methodological review advocates more transparent and reproducible practices. Poldrack et al., 2017.

Use that background to ask concrete questions. Why is the sample appropriate for the purpose? Which exclusion rules apply? Which analyses were planned in advance? How will uncertainty and unsupported expectations be reported? Also ask how data and analytical decisions will be documented so that the route to the conclusion remains clear.

A sound proposal describes more than the expected result. It explains what small, uncertain or absent differences would mean. Distinguish a lack of convincing evidence for a difference from evidence that no relevant difference exists. Discuss with the analyst which design and level of precision would be needed for the latter conclusion.

MRI does not use ionising radiation, but screening for implants and metal objects is still necessary. Hearing protection and limiting movement also matter. Suitability must be assessed by the qualified scanner team. FDA, Benefits and Risks.

Specify responsibility for participant information, consent, data storage and any incidental findings in the project agreements. Taking part in research is not a promised medical screening. Information given to participants must match what the team actually provides and is qualified to assess.

Ask the imaging centre or research partner to assess whether the task, equipment, schedule and supervision fit together. Set out the expertise and responsibilities required. A partner’s name or availability should only enter a concrete offer once confirmed; this article does not establish a particular partnership.

An associative target group, in our working definition, is a bounded group whose members decide on the basis of similar associations within a specific choice context. A similar fMRI pattern does not establish such a group by itself.

First specify the associations and choice context and how the boundaries will be investigated. Then decide whether fMRI has a relevant complementary question to answer. Keep the investigation of similarities between people separate from attaching an appealing label to a pattern afterwards. Describe a grouping that has not yet been tested as a hypothesis.

fMRI becomes useful when a specific research question connects to a defensible design and interpretation. Ask what is measured, which comparison supports the answer and what additional testing is needed before applying the conclusion. This gives the brain measurement a clear place within your research and produces an outcome you can assess and use.