The effect of facial expression on contrast sensitivity: a behavioural investigation and extension of Hedger, Garner, & Adams (2015)

It has been argued that rapid visual processing for fearful face expressions is driven by the fact that effective contrast is higher in these faces compared to other expressions, when the contrast sensitivity function is taken into account (Hedger, Garner, & Adams, 2015). This proposal has been upheld by data from image analyses, but is yet to be tested at the behavioural level. The present study conducts a traditional contrast sensitivity task for face images of various facial expressions. Findings show that visual contrast thresholds do not differ for different facial expressions We re-conduct analysis of faces’ effective contrast, using the procedure developed by Hedger, Garner, & Adams (2015), and show that higher effective contrast in fearful face expressions relies on face images first being normalised for RMS contrast. When not normalised for RMS contrast, effective contrast in fear expressions is no different, or sometimes even lower, compared to other expressions. These findings are discussed in relation to the implications of contrast normalisation on the salience of face expressions in behavioural and neurophysiological experiments, and also the extent that natural physical differences between facial stimuli are masked during stimulus standardisation and normalisation.


113
effects of normalisation on the physical and perceived salience of facial stimuli, it is 114 reasonable to question the degree to which normalisation influences results from both 115 image analyses and behavioural paradigms. In particular, any consistent differences in 116 RMS contrast across facial expressions would be expected either to increase, or 117 cancel out, differences in sensitivity that can be attributed to differences in effective 118 contrast.

119
To address these questions, we conducted a replication of the image analyses 120 performed by Hedger, Garner and Adams [15] . We included face stimuli that are 121 physically matched for RMS contrast, but also faces that were physically unmatched, 122 such that they contain natural differences in both physical and apparent contrast.
123 Furthermore, we conducted a traditional contrast sensitivity task in order to 124 psychophysically test predictions from Hedger's image analysis. Here, we employed 125 facial expressions as opposed to sinusoidal grating stimuli to measure expression-126 related differences in contrast sensitivity. An important feature of this latter study is 127 that it directly addresses the association between face expression and contrast 128 sensitivity at the behavioural level. the Karolinska Directed Emotional Faces set [18] .     is displayed in Figure 4.

284
For 140 raw KDEF faces, RMS contrast was calculated across the 5 expressions.

285
Fearful KDEF faces naturally contained significantly less RMS contrast compared to 286 neutral, angry and disgust expressions, and did not differ compared to happy 287 expressions. These data are illustrated in Figure 5, and summarised in Table 2

298
For 57 raw Radbound faces, RMS contrast was calculated across the 5 expressions.

299
Fearful Radbound faces naturally contained significantly less RMS contrast compared 16 300 to all other face expressions. These data are illustrated in Figure 5, and summarised in 301

312
For 7 raw Montreal faces, RMS contrast was calculated across the 5 expressions.

313
RMS contrast for fearful Montreal faces did not differ significantly compared to any 314 other face expression. These data are illustrated in Figure 5, and summarised in Table   315 2 (d).  Table 1 (e). Here, effective contrast is measured for experimental face 320 images that were not normalised for RMS contrast, and also for versions of the same 321 faces that had been normalised for RMS contrast. Paired comparisons measured 322 differences in effective contrast between fearful faces and their expression 323 counterparts. When faces were normalised for RMS contrast, effective contrast was 324 significantly higher in fear expressions compared to happy expressions, and did not significantly differ compared to neutral, anger or disgust faces. Alternatively, for the 326 same faces that were not normalised for RMS contrast, such that they were analysed 327 in the same format as they were presented to observers, fear expressions were 328 significantly lower in effective contrast compared to both neutral and disgust 329 expressions, and did not differ significantly from angry or happy expressions.

330
For the 16 raw KDEF faces used in the present contrast sensitivity study, RMS   Together, data from the present contrast sensitivity study showed that visual contrast 341 thresholds are not influenced by differences between images of facial expressions.

342
Namely, fearful expressions portrayed by face images did not enhance observers' 343 contrast sensitivity; as was predicted by findings from Hedger, Garner and Adams [15] .

344
Fearful expressions, according to image analyses by Hedger, Garner and Adams [15] 345 are higher in effective contrast, and thus well tuned to contrast sensitivity processing.

346
This proposal was driven by data from image analyses measuring differences in 347 effective contrast between fear and neutral face images that had been normalised for 348 RMS contrast. The stimuli used in the present study were raw face images that were 349 not normalised for physical contrast in any way. We replicate measures of effective 350 contrast used by Hedger, Garner and Adams [15] to establish the extent that CSF