Combining attention training with internet-based cognitive-behavioural self-help for social anxiety: a randomised controlled trial.

The researchers compared two groups of adults diagnosed with social anxiety disorder (SAD):

**Group 1 (ICBT + ABM):** Received 2 weeks of daily attention training designed to reduce *attentional avoidance* of threatening faces, followed by 9 weeks of guided internet-based cognitive-behavioural self-help.

**Group 2 (ICBT + Control training):** Received 2 weeks of daily *control* training (a sham version of the same task that did not train attention away from threat), followed by the same 9 weeks of ICBT.

The attention training used a modified **dot-probe task**. In this task, two faces (one threatening, one neutral) appear briefly on screen, then a dot appears behind one of them. In the ABM condition, the dot always appeared behind the *neutral* face, training participants to shift attention away from threat. In the control condition, the dot appeared equally behind threatening and neutral faces (50/50), providing no systematic training.

The key question: Would adding ABM boost the effects of ICBT beyond ICBT alone?

**Total sample:** 133 participants (66 in ABM group, 67 in control group)

**Age:** Mean age 33.2 years (SD = 10.2, range 18–67)

**Gender:** 68% female

**Diagnosis:** All met DSM-IV criteria for social anxiety disorder (SAD), confirmed via structured clinical interview

**Recruitment:** Via newspaper advertisements, online forums, and mental health clinics in Sweden

**Exclusion criteria:** Current or past psychosis, bipolar disorder, severe depression (Montgomery-Åsberg Depression Rating Scale score > 35), ongoing psychological treatment, recent change in medication (past 3 months), substance dependence, suicidal ideation

**Setting:** Entirely internet-based; participants never met researchers in person

**Primary outcome measures (social anxiety):**

**Liebowitz Social Anxiety Scale – Self-Report (LSAS-SR):** 24 items rating fear and avoidance of social situations (0–3 scale each), total range 0–144, higher = worse

**Social Phobia Scale (SPS):** 20 items measuring fear of being observed by others (0–4 scale), range 0–80

**Social Interaction Anxiety Scale (SIAS):** 20 items measuring fear of social interactions (0–4 scale), range 0–80

**Secondary outcome measures:**

**Montgomery-Åsberg Depression Rating Scale – Self-Report (MADRS-S):** 9 items measuring depression severity (0–6 scale), range 0–54

**Quality of Life Inventory (QOLI):** 32 items measuring life satisfaction across 16 domains, range -6 to +6

**Attention bias measure:**

**Dot-probe task (same format as training):** Measured reaction time differences between trials where the dot appeared behind threatening vs. neutral faces. A positive score = bias *toward* threat; a negative score = bias *away* from threat (attentional avoidance).

**Assessment timepoints:** Pre-treatment, after the 2-week training phase (before ICBT started), and post-treatment (after 9 weeks of ICBT).

**Design:** Randomised controlled trial (RCT) with two parallel groups, no crossover.

**Randomisation:** Participants were randomly assigned 1:1 to ABM or control training using a computer-generated random number sequence. Allocation was concealed from researchers until after baseline assessment.

**Blinding:** Participants were blind to which condition they received (they were told both were "attention training" but not which was active vs. control). The researchers who conducted diagnostic interviews and outcome assessments were also blind to group assignment. However, the therapists providing ICBT support were *not* blinded — they knew which participants were in which group because the training phase preceded ICBT.

**Duration:**

2 weeks of daily attention/control training (one session per day, ~10–15 minutes each)

9 weeks of ICBT (8 modules, one per week, with weekly email or phone support from a therapist)

Total study duration: approximately 11 weeks

**ICBT protocol:** The internet-based cognitive-behavioural self-help programme was based on Clark and Wells' (1995) cognitive model of SAD. It included:

Psychoeducation about social anxiety

Cognitive restructuring (identifying and challenging negative thoughts)

Exposure exercises (gradual confrontation with feared social situations)

Reduction of safety behaviours

Attention training exercises (separate from the dot-probe task — these were standard CBT techniques like shifting focus from internal to external attention)

Relapse prevention

Participants received weekly written feedback from a therapist via a secure online platform, plus optional brief telephone calls.

**Statistical approach:** Intention-to-treat analysis using linear mixed models, which handle missing data under the assumption that data are missing at random. Effect sizes reported as Cohen's d (within-group and between-group). Primary analyses tested group × time interactions.

**What this design can prove:**

Whether adding ABM to ICBT produces *greater* improvement than ICBT alone (between-group comparison)

Whether ABM changes attention bias as measured by the dot-probe task

Whether changes in attention bias correlate with symptom improvement

**What this design cannot prove:**

Whether ABM alone is effective (no ABM-only group was included)

Whether ICBT alone is effective (no waitlist or placebo control group — both groups received ICBT)

Long-term effects beyond 11 weeks (no follow-up beyond post-treatment)

Mechanisms of change (mediation analyses are exploratory, not causal)

**Major methodological weaknesses:**

**No ICBT-only control:** Because both groups received ICBT, any effect of ABM could only be detected as an *additive* benefit. If ICBT is highly effective, there may be a ceiling effect that masks ABM's contribution.

**No sham ICBT:** Participants knew they were receiving an active treatment, which could inflate expectations and placebo effects.

**Therapist unblinding:** Therapists knew group assignments, potentially introducing bias in their support (though they followed a standardised protocol).

**Attention bias measurement:** The same dot-probe task was used for both training and assessment, making it unclear whether any changes reflected genuine attention bias modification or simple task practice effects.

**No manipulation check during training:** The paper does not report whether participants actually shifted their attention as intended during the training sessions (e.g., by analysing reaction time patterns trial-by-trial).

**Primary outcomes (social anxiety):**

Both groups showed large, significant improvements from pre- to post-treatment on all social anxiety measures.

**LSAS-SR:** ABM group: mean 72.2 → 43.3 (within-group d = 1.41); Control group: 72.8 → 44.9 (within-group d = 1.39). Between-group difference at post-treatment: d = 0.08 (not significant, p > 0.05).

**SPS:** ABM group: mean 34.1 → 18.9 (d = 1.10); Control group: 34.5 → 19.8 (d = 1.06). Between-group d = 0.06 (not significant).

**SIAS:** ABM group: mean 49.7 → 33.6 (d = 1.16); Control group: 50.1 → 34.2 (d = 1.12). Between-group d = 0.05 (not significant).

**Secondary outcomes:**

**Depression (MADRS-S):** Both groups improved moderately. ABM: mean 15.4 → 10.1 (d = 0.67); Control: 15.8 → 10.5 (d = 0.63). Between-group d = 0.04 (not significant).

**Quality of life (QOLI):** Both groups improved. ABM: mean 0.54 → 1.32 (d = 0.66); Control: 0.48 → 1.28 (d = 0.64). Between-group d = 0.03 (not significant).

**Attention bias:**

Neither group showed significant change in attention bias from pre- to post-treatment.

**ABM group:** Bias score changed from -5.3 ms (slight avoidance of threat) to -3.1 ms (within-group d = 0.10, not significant).

**Control group:** Bias score changed from -4.8 ms to -2.9 ms (within-group d = 0.17, not significant).

The between-group difference in bias change was d = 0.07 (not significant).

**Clinical significance:**

**Reliable improvement** (based on LSAS-SR change exceeding the reliable change index): 68% in ABM group vs. 64% in control group (not significant).

**Clinically significant change** (reliable improvement *plus* post-treatment score below clinical cutoff): 44% in ABM group vs. 42% in control group (not significant).

**Key negative finding:** The ABM training did *not* successfully modify attention bias. This means the study cannot test whether attention bias modification *would* enhance ICBT if it worked — because the training itself failed.

In plain English:

Both groups improved substantially — roughly equivalent to moving from "moderate" to "mild" social anxiety on the LSAS-SR (a drop of about 28 points out of 144).

The improvement was large: within-group effect sizes of d ≈ 1.4 mean the average participant at post-treatment scored about 1.4 standard deviations *lower* on social anxiety than at pre-treatment. For context, this is considered a "large" effect in clinical psychology (convention: 0.2 = small, 0.5 = medium, 0.8 = large).

However, adding ABM made *no detectable difference* — the between-group effect sizes were all d < 0.1, which is essentially zero. To put this in perspective: if you randomly assigned 100 people to ICBT alone and 100 to ICBT + ABM, you would expect about 51 of the ABM group to have better outcomes than the average ICBT-alone participant (a trivial difference).

The attention training itself failed to shift attention bias by any meaningful amount (d = 0.10–0.17, which is "negligible" to "small").

**Acknowledged by authors:**

The ABM training did not successfully modify attention bias, so the study cannot test the hypothesis that ABM + ICBT > ICBT alone.

The dot-probe task may not be a reliable measure of attention bias (test-retest reliability is known to be low).

No follow-up data beyond post-treatment, so durability of gains is unknown.

The sample was self-selected (internet-recruited), which may limit generalisability to clinical populations.

**Additional critical observations:**

**No active control for ICBT:** Without a waitlist or placebo control, we cannot be sure that the improvements were due to ICBT specifically, rather than time, spontaneous remission, or non-specific factors (e.g., attention from a therapist).

**Therapist unblinding:** Therapists knew which participants received ABM vs. control training. Although they followed a manual, this could subtly influence their enthusiasm or support.

**Low statistical power for moderation:** The study was powered to detect a medium-to-large between-group effect (d ≈ 0.5), but the actual between-group difference was tiny (d < 0.1). The study cannot rule out a *small* additive benefit of ABM (e.g., d = 0.2), but such an effect would be clinically trivial.

**Training duration:** Only 2 weeks of daily training (14 sessions) may be insufficient to modify attention bias. Lab-based studies that succeeded often used more sessions (e.g., 8 sessions over 4 weeks in Amir et al., 2009).

**No assessment of compliance:** The paper does not report how many training sessions participants actually completed, or whether they performed the task correctly (e.g., with sufficient attention).

**Sample characteristics:** 68% female, all Swedish, all comfortable with internet-based interventions. Results may not generalise to other populations.

**No blinding of participants to ICBT:** All participants knew they were receiving an active psychological treatment, which could inflate expectations and placebo responses equally across groups.

For someone running their own n=1 experiment:

### What to test

**Primary intervention:** Internet-based cognitive-behavioural self-help for social anxiety (e.g., using a structured programme with psychoeducation, cognitive restructuring, exposure exercises, and safety behaviour reduction). The ICBT in this study lasted 9 weeks with weekly therapist support.

**Add-on to test:** Attention bias modification using a dot-probe task designed to train attention *away* from threatening faces (or, alternatively, *toward* neutral faces). The protocol here used 14 daily sessions (10–15 minutes each) before starting ICBT.

### Minimum meaningful duration

**For ICBT alone:** At least 8–9 weeks to complete a full programme. The within-group effects in this study were large by week 9.

**For ABM:** If you want to test ABM as an add-on, run at least 2 weeks of daily training (14 sessions) before starting ICBT. However, based on this study, do *not* expect ABM to add benefit — the evidence suggests it is ineffective in this format.

### What to measure (specific metrics)

**Social anxiety symptoms:** Use the Liebowitz Social Anxiety Scale – Self-Report (LSAS-SR) or the Social Phobia Scale (SPS) and Social Interaction Anxiety Scale (SIAS). Measure at baseline, after ABM phase (week 2), and after ICBT (week 11).

**Attention bias:** Use a dot-probe task with threatening vs. neutral faces. Calculate bias score = (mean RT for incongruent trials) – (mean RT for congruent trials). A negative score = avoidance of threat; a positive score = vigilance toward threat. Measure at baseline and after ABM phase.

**Depression:** Montgomery-Åsberg Depression Rating Scale – Self-Report (MADRS-S) to track mood changes.

**Quality of life:** Quality of Life Inventory (QOLI) or a simple 0–10 life satisfaction rating.

**Compliance:** Track number of training sessions completed, time per session, and accuracy on the dot-probe task (reaction time and error rate).

### Key confounds to control for

**Expectation effects:** If you know you're receiving ABM, you might expect more benefit. Consider using a sham training (e.g., 50/50 dot placement) during a "control" phase, or use a single-case experimental design with multiple baselines.

**Practice effects on the dot-probe:** Repeated testing on the same task can improve reaction times regardless of training. Use a different set of face stimuli at each assessment, or use a different attention bias measure (e.g., eye-tracking if available).

**Therapist contact:** In this study, participants received weekly therapist support. If you're doing self-guided ICBT without support, results may differ (likely smaller effects).

**Medication changes:** If you're on psychiatric medication, keep dose stable throughout the experiment. Any change could confound results.

**Life events:** Major social events (e.g., starting a new job, moving, relationship changes) could independently affect social anxiety. Log these in a daily diary.

**Time of day:** Reaction times on the dot-probe task vary with alertness. Test at the