Attention Bias Modification for Anxiety: A Promising Idea with Disappointing Results

Few ideas in anxiety research have generated as much excitement and subsequent disappointment as attention bias modification. The logic was elegant: anxious people preferentially attend to threat; this attentional bias maintains their anxiety; train them to attend elsewhere, and the anxiety should reduce. Two decades of research later, the idea has not delivered what it promised.

The background

People with anxiety disorders are faster to detect threatening stimuli. This was demonstrated by MacLeod, Mathews and Tata in 1986 using the dot-probe task: pairs of words appeared briefly on screen, one threatening and one neutral, followed by a probe in one location. Anxious participants were faster to respond when the probe replaced the threatening word, indicating that their attention had been drawn to it [1].

This attentional bias has been replicated hundreds of times across anxiety disorders, though it is more reliably found with some paradigms and populations than others. The finding is well established enough to be considered a feature of anxiety, even if its exact role, whether it is a cause, consequence or correlate, remains debated.

The treatment idea

If the bias maintains anxiety, then modifying it should help. Attention bias modification (ABM) typically uses a modified dot-probe task: on every trial, the probe always appears in the location of the neutral stimulus, training the person to shift attention away from threat. Over hundreds of trials, the hope is that this retrained attentional pattern becomes automatic and generalizes beyond the task.

The approach was attractive for several reasons: it was computerized and standardized, did not require a therapist, could be delivered online or at home, and targeted an automatic process that might be missed by talk-based therapies.

The rise

An early meta-analysis in 2010 found a moderate and significant effect of ABM on anxiety symptoms, reporting it as a promising intervention [2]. This generated a wave of funded research, clinical translations and even commercial applications. Studies were published at a rapid pace, many reporting positive results.

The fall

Larger, better-controlled studies began to report null or small results. A comprehensive meta-analysis by Cristea, Kok and Cuijpers in 2015 examined all available evidence for cognitive bias modification (attention and interpretation combined) across anxiety and depression. The pooled effect on anxiety symptoms was very small: Hedges’ g = 0.13. When the authors applied corrections for small-study effects and publication bias, the effect became non-significant [4].

A meta-analysis specific to social anxiety found a significant but small effect on symptom measures, but also noted substantial heterogeneity and possible publication bias [3].

The pattern across the literature is consistent with a field where early small studies found inflated effects, positive results were preferentially published, and larger pre-registered studies found little.

Why it may not work

Several explanations have been proposed for the failure:

The task may not change attention. Many ABM studies do not include a post-training attention assessment. When they do, the evidence that the training actually shifts attentional patterns is inconsistent. If the training does not modify the bias, the theoretical basis for symptom improvement collapses.

Attention bias may not be causal. The correlational evidence that anxious people attend to threat does not prove the bias maintains their anxiety. It could be a downstream consequence of anxiety, or both could stem from a shared process. A prospective study found that attention bias predicted later depressive symptoms in youth, which is consistent with a causal role [6], but single studies cannot settle the question.

The dot-probe may measure attention poorly. The dot-probe task has low reliability, meaning it may not consistently measure the same thing across sessions. If the measurement is unreliable, training based on it may be incoherent. The person is being trained on noise.

The training is boring. Hundreds of identical trials are tedious. Participants may disengage, reducing any potential effect.

The debate

Richard McNally, a prominent researcher in the field, published a commentary titled “Attentional bias for threat: crisis or opportunity?” [5]. He argued that the field faces a genuine crisis but that the core idea, that cognitive biases play a role in anxiety, is not invalidated by the failure of one particular training paradigm. The opportunity lies in developing better tasks, better measures and more theoretically grounded interventions.

Proposed improvements include:

• Gaze-contingent training using eye tracking to provide real-time feedback about where attention goes
• Personalized stimuli using words or images relevant to the individual’s specific fears
• Combination approaches adding ABM to CBT rather than using it standalone
• Better measurement using attention paradigms with higher reliability

Whether any of these refinements will rescue the approach is an open question. The history of the field suggests caution about early claims of efficacy.

What to take from this

The ABM story is worth understanding not because of what it says about attention, but because of what it says about how evidence develops:

1. A plausible theory does not guarantee an effective treatment.
2. Small early studies in psychology systematically overestimate effects.
3. Publication bias can create an evidence base that looks stronger than it is.
4. The field’s initial enthusiasm outpaced the evidence, and commercial applications were premature.

For how publication bias operates more broadly, see publication bias in anxiety research. For a primer on interpreting the effect sizes reported in these meta-analyses, see understanding effect sizes.

When to look for more support. Attention bias modification is not currently recommended as a standalone treatment for anxiety. If anxiety is affecting your life, evidence-based treatments like CBT and medication have much stronger support. Talk to your GP or a psychologist. If you are in crisis, call or text 988 (US) or Samaritans on 116 123 (UK and Ireland).