Previous empirical investigations into selective attention for threat in anxiety have not offered sufficiently rigorous methodological control to rule out alternative explanations for the data. In turn, these procedural problems have not permitted clear and consistent patterns in the data to emerge, which in turn, has hindered the development of a coherent theory. Six empirically tight experiments investigated attentional biases for threatening information in non-clinical anxiety. Four studies employed verbal material as the stimuli and two studies used schematic representations of human faces. Five studies (four using words and one using faces) employed a computerised emotional Stroop colour naming task. For the final experiment a modified version of the dot-probe classification task was developed to assess interference in stimulus processing when the distracting information and central task were presented in spatially separate locations. All studies used experimental designs that examined the influence of state anxiety and trait anxiety by employing student samples assigned to high trait anxious (HTA) or low trait anxious (LTA) groups on the basis of questionnaire scores. In each experiment state anxiety was manipulated through the threat of electric shock, and the stimulus material was presented to participants both within and outside of awareness. In the latter condition a backward masking procedure was used, and stimulus onset asynchronies between the target and mask were individually set for each participant. In the studies employing verbal material as stimuli awareness was assessed using a forced-choice lexical decision task, whereas for the experiments that used pictorial stimuli an analogous face-status decision task was developed. In all studies participants' data were excluded on an individual basis if their performance on the awareness checking tasks exceeded criterion levels.
In Study 1.1 threatening words that were either unrelated (e.g., cancer, danger) or related (e.g., electrocute, shock) to the source of the stressor were presented to participants. In the masked condition HTA participants were facilitated at colour naming both types of threatening words relative to the neutral control items (e.g., spoon, carpet) when performing under threat of shock, but this effect was not evident in the shock safe condition. For unmasked trials the HTA group showed significant interference in colour naming both types of threat words relative to control words when performing under the threat of shock, but not in the shock safe condition. Neither valence of the items nor the threat of shock influenced colour naming latencies in either exposure mode for the LTA group.
Study 1.2 and Study 1.3 investigated the nature of stimulus characteristics that affect differential colour naming times to masked and unmasked threat and control verbal material. In Study 1.2, HTA individuals alone were presented with the general threatening words and their length and frequency matched controls employed in Study 1.1. A second set of orthographically similar threat (e.g., cancer vs. camcer) and control (e.g., bottle vs. bittle) distracter items was also developed to investigate whether information-processing biases occur as a result of semantic processing or whether they occur on the basis of the stimulus features. For the masked trials the results indicated that participants' performance was significantly facilitated in colour naming the genuine threat words when performing under the threat of shock relative to the shock safe condition, and this effect was most evident for those who were exposed to the threat of shock in the initial part of the experiment. For the unmasked exposure trials there was a non-significant trend for genuine threat words to produce colour naming interference when performing under the threat of shock, relative to the shock safe condition. There was no difference in colour naming latencies for the orthographically similar threat and control distracters between the shock threat and shock safe conditions in either exposure mode.
In Study 1.3 HTA and LTA individuals performed a colour naming task for threatening and positively valenced words during masked and unmasked trials performed with and without the threat of shock. For the unmasked trials HTA individuals showed significantly greater interference for threat words relative to LTA individuals who showed facilitation for threat, but only when performing under the threat of shock. There were no between group differences in colour naming latencies for the unmasked threat and control items in the shock safe condition. For the masked trials neither item valence nor the threat of shock influenced colour naming latencies in either exposure mode for either group.
In Study 1.4 HTA and LTA participants performed a colour naming task on threat and neutrally valenced items. However, unlike the previous three experiments that presented masked and unmasked trials in an intermixed sequence, the mode of exposure was blocked. Half the participants within the HTA and LTA groups were shown masked exposures followed by unmasked exposures, whereas the other half received the opposite exposure mode order. The results revealed that in both exposure modes HTA participants showed significant interference for threat relative to the LTA group who showed facilitation and these effects were evident irrespective of the threat of shock. For the masked exposures however, this pattern of responding was restricted to participants who received unmasked trials prior to the masked trials. For those who received the masked exposure trials first there was no evidence of selective attention for information of which they were not aware. For the unmasked exposures, the order of exposure mode did not influence this pattern of results. The results suggest that awareness of threat might be a necessary precondition to establish selective attention effects for verbal material that is presented using backward masking procedures.
Study 2.1 and Study 2.2 assessed selective attention effects for pictorial stimuli. In Study 2.1 HTA and LTA participants performed a colour naming task in which they were presented with an intermixed sequence of masked and unmasked schematic pictures of happy, neutral, and threatening human faces. The trials were performed with and without the threat of shock. Verbal ratings were also obtained for the stimuli on the dimensions of valence, interest, arousal, and facial expression. For the masked trials the results revealed that participants were significantly slower at colour naming the threatening faces relative to the happy faces, with colour naming latencies for the neutral faces falling between the other two and not-significantly different from either. For the unmasked exposures colour naming latencies were unaffected by item valence. The pattern of responding for the masked and unmasked trials operated independently of state and trait anxiety status. The rating data confirmed that the selective attention effects observed during the masked exposures were accounted for on the basis of item valence rather than interest and/or arousal factors because the threatening and happy faces were matched on these dimensions.
In Study 2.2 HTA and LTA individuals classified a centrally located digit as odd or even as quickly and accurately as possible. At the offset of the digit they were required to verbally identify the shape of a small probe (either a circle or a square) that was presented in the periphery of the computer screen while at the same time a happy, neutral or angry face pair was presented immediately above and below the location formerly occupied by the digit. Masked and unmasked trials were presented in an intermixed sequence and participants performed these tasks both with and without the threat of shock. The results revealed that for the unmasked trials HTA individuals were significantly slower to identify the status of the probe on threat face trials relative to happy face trials when performing under the threat of shock, but no differences were observed in the shock safe condition. During the unmasked trials neither item valence nor state anxiety status affected response latencies to the probes for the LTA group. For the masked trials, the LTA group were slower to identify the status of the probe on happy face trials relative to threat face trials without the threat of shock, and slower on threat face trials relative to happy face trials when state anxiety was high. Performance on the probe status task during the masked exposures was unaffected by state anxiety status or item valence for the HTA group.
The overall pattern of results was interpreted with respect to contemporary models of attention and anxiety, and the implications for treatment of anxiety disorders are discussed. The limitations of the current thesis are noted and directions for future research are suggested.