Selective Attention and Health Anxiety: Ill-Health Stimuli are Distracting for Everyone

A Repeated Measures ANOVA revealed the same pattern of results for both the letter and the word visual search tasks, when the WI or the IAS were used individually, and when the measures were combined to categorize individuals into high and low health anxious groups.

Power calculations were conducted using the Effect Size Generator element of ClinTools Version 4.1 (Devilly 2007). As the methodology for assessing attentional biases in health anxiety within this study is relatively unique, details entered into ClinTools were based upon three research articles investigating information processing and anxiety. Specifically, Rassin et al. (2008) who reported a medium effect (r = .29) of health anxiety during a visual search task with non-health specific stimuli. When converted to Cohen’s d using ClinTools (Devilly, 2007) r = .29 is equivalent to d = .61. Owens et al. (2004) Stroop paradigm utilising health specific words among high and low health anxious individuals, revealing a large effect (d = .97); and Rinck et al. (2003) who observed a medium to large effect (d = .73) between generalised anxiety patients and controls during a visual search task involving anxiety-related words. Results suggested an adequate sample size of approximately 28 participants per group to achieve power of .8 in a two-tailed analysis with an expected medium to large effect size. Therefore, the current group sample sizes were deemed sufficient to produce suitable power for the analyses.

In total 6, 12, or 18 items were presented in the word and letter task, for target present and absent trials. This allowed the same number of items to be displayed within the present and absent trials, preventing participants from using counting to respond to target, present versus absent (e.g., 6 items = absent, 7 items = present).

Regression coefficients were not calculated for the error scores (ms) due to insufficient data (i.e., few errors within some conditions).

The authors would like to acknowledge the input from one of the reviewers suggesting this further analysis.

Originally, ANCOVAs were applied to the data sets, with scores from the STAI-S, STAI-T, and BDI as covariates. However, the assumption of linear relationships between the covariates at each level of the independent variables were violated for error data, in the absent condition, for both letter and word tasks. In order to run the same analyses for both the RT and error data to check for speed-accuracy trade-offs, the data were investigated using mixed factorial ANOVA. Initial analyses revealed, results were not influenced by speed-accuracy trade-offs and the pattern of results produced using both errors and mean RT per condition including display size, was similar to that reported using the regression based approach.

One methodological suggestion which may have been considered to ensure priming effects in the visual search task was to block trials by distractor type (we thank a reviewer for this insightful suggestion). However if this procedural adjustment had been made it is likely that the distractor words may have lost their valence and meaning value, in turn minimising any priming effects. Researchers have previously demonstrated that repeated presentation of a word type results in habituation or a loss of valence to that word type (e.g., McNally et al. 1990).