levels of anxiety or depression. The currently used cut-off score of eight gives a specificity of 0. 78 and a sensitivity of 0. 90 for anxiety sub-scale, and a specificity of 0. 79 and a sensitivity 0. 83 for depression sub-scale (Bjelland et al. , 2002). In the present study, Cronbach’s alpha was 0. 86 for depression and 0. 84 for anxiety. The Health Anxiety Inventory (HAI) is 18-item questionnaire measuring clinical and non-clinical health anxiety. Each item is scored from 0 to 3.
The first fourteen items forms the man measurement and the last four items forms the negative onsequences measurement that assesses respondents’ perceived negative consequences of being ill. A Cronbach’s alpha of 0. 89 indicates a high level of internal consistency (Salkovskis et al. , 2002). In the present study, Cronbach’s alpha was 0. 87. The Brief Pain Inventory – Short Form (BPI) measures pain severity and the impact of pain on daily functioning (i. e. , pain inference).
Pain severity is measured by three items that ask respondents to rate their worst pain, least pain, and average pain over the past week plus one item asking about current pain. Pain inference is measured by seven items asking about how pain interferes with daily functions. One addition item is used to measure pain relief from medications. In the present study, Cronbach’s alpha was 0. 94 for pain severity and 0. 92 for pain inference. Experimental Stimuli A total of 30 real-world images were used in the study. Images were categorized into three groups: pain-related, neutral, and ambiguous. Examples were included in Appendix (see Figure 1 3).
Pain-related images clearly depict people with injuries or experiencing pain (e. g. , a man holding his leg in pain, with a acial expression of pain; a woman holding his shoulder, with a facial expression of pain). Neutral images feature people without visible injuries or pain in everyday situations (e. g. , a man sitting at a desk working; a man playing cricket). Ambiguous images picture people in uncommon situations or positions (e. g. , a man lying on the ground, with no facial expression of pain; two men supporting a woman, with no facial expression of pain), which may or may not have injurious or painful causes.
Ten ambiguous images were first collected using the google search engine. Then each ambiguous image was matched to one neutral image and one pain-related image on number of people present in the scene, both foreground and background. No significant difference on number of people present was found among three image groups (F(2, 27) = . 11, p = . 901, p2 =. 01) Low level features interfere with participants gaze and attention (Henderson, 2003; Torralba et al. , 2006). Approaches used in previous research (Schoth et al. 2015) were adopted to control for features including images’ complexity, luminance, mean contrast level, red colour saturation, green colour aturation, blue colour saturation, and overall colour saturation.
As Buodo and colleagues (Buodo, Sarlo, & Palomba, 2002) suggested that more complex images give rise to larger files, complexity was measured as the size of the compressed image in kilobytes. Luminance refers to the intensity of light reflecting from the surface. Contrast refers to the difference in luminance. Colour saturation refers to the colourfulness of a colour relative to its brightness.
All parameters, except for complexity, were measured using Photoshop CS6. One-way ANOVAS were onducted to investigate the difference in low level features between image groups. Results showed that all three types of images were comparable in complexity (F(2, 27) = . 45, p = . 643, p2 = . 03), luminance (F(2, 27) = . 77, p = 473, p2 = . 05), mean contrast level (F(2, 27) = 2. 40, p = . 110, p2 = . 15), red colour saturation (F(2, 27) = 1. 09, p = . 350, p2 = . 07), green colour saturation (F(2, 27) = . 74, p = . 487, p2 = . 50), blue colour saturation (F(2, 27) = . 36, p = . 689, p2 = . 03), and overall colour saturation (F(2, 27) = . 6, p = . 523, p2 = . 05).
Details of low level eatures were included in Appendix (see Table 1). In addition to low level features, an independent sample of 10 participants (50% female, Mean age = 25. 4, SD = 3. 03, ranges from 23 to 37) rated each image on pleasantness, arousal, and pain representation. All images were presented to participants in a random order for 3 seconds on a computer each image representation, participants were asked to indicate how pleased and aroused they felt while viewing the image by clicking on a 9-point Self-Assessment Manikin scale, with 1 indicating ‘not at all’ and 9 indicating ‘extremely’.
Participants then viewed each image again and provide a rating of the extent to which the image represents pain, on a 10-point scale with 0 indicating ‘not at all’ and 9 indicating ‘very well’, in a pen and paper based test. One-way ANOVAS suggested that images significantly differ on pleasantness, arousal and pain representation. Post hoc comparisons using Bonferroni correction indicated that, for pleasantness, the rating of neutral images (M = 5. 70, SD = . 48) was significantly higher than that of ambiguous images (M = 4. 00, SD = . 7), and the mean rating of mbiguous images was significantly higher than that pain- related images (M = 2. 70, SD = . 67).
For arousal, the rating of pain-related images (M = 5. 70, SD = 1. 25) was significantly higher than that of ambiguous images (M = 4. 10, SD = . 74), and the rating of ambiguous images was significantly higher than that that of neutral images (M = 2. 70, SD = . 48). For pain representation, participants’ ratings suggested that pain-related images best depicted pain (M = 7. 40, SD = 1. 29), followed by ambiguous images (M = 5. 26, SD = 1. 09) and neutral images (M = . 30, SD = . 5).
Therefore, the results of this analysis showed ambiguous images to fall between pain-related and neutral images in pleasantness, arousal, and pain-relatedness. Image Interpretation Task The experiment was programmed using iSurvey, an online questionnaire distributing tool from the University of Southampton. iSurvey was first launched in 2009 and has been widely used in research conducted at the University of Southampton (e. g. , Carnelley & Rowe, 2010; Foster et al. , 2014). After familiarizing with the task during practice trials, participants then completed a total of 30 experimental trials.
Each trial began with the presentation of one image for four seconds. Then participants were asked to write a brief account of what they thought was happening in the image and how people present in the image were feeling. This was followed by the presentation of three pre-set statements that described three possible interpretations provided by the researchers. Each of the statement fell into one of the following three categories: pain-related, negative, and benign. Pain-related statements explicitly mentioned injuries or the person in the image was in pain.
Negative statements described a negative but pain-free situation such as getting drunk or being upset. Benign statements contained no painful or negative descriptions. The order of statements was randomized for each participant. Participants rated the likelihood of happening for each statement on a 11-point scale with 0 indicating ‘not likely at all’ and 10 indicating ‘extremely likely’. Then the next trial began until the participant had seen all 30 images. Images were presented in full random order. A diagram of the task was included in Appendix (see Figure 4).