No difference in flanker effects for sad and happy schematic faces: A parametric study of temporal parameters

Abstract

Flanker effects with schematic faces have been reported to be larger for happy than for sad faces, allegedly because sad faces restrict the focus of spatial attention. We report a parametric study that fails to replicate this effect. Participants performed speeded identifications of happy or sad faces accompanied by compatible or incompatible flanker faces. We varied the temporal interval between presentation of central target and flanker faces because differential attentional effects of happy and sad faces should critically depend on this variable. In contradiction to the literature, we found large compatibility effects that were modulated by temporal parameters, but not by the emotional valence of the faces, and not in the way consistent with differential attentional modulation. We conclude that previously reported asymmetries in flanker tasks with schematic faces are not due to changes in attentional scope (mediated by emotion or otherwise), but rather to perceptual low-level differences.

Keywords:

• Emotion
• Flanker task
• Schematic faces
• Visual attention

Acknowledgments

We thank Peter Kohl for data collection and Shanley Allen, Anke Haberkamp, Neiloufar Family, and Kirstin Bergstroem for helpful comments on earlier versions of the manuscript. We also want to thank Gernot Horstmann, two anonymous reviewers, and the editor for valuable suggestions. This research was supported by Schm1671/1-5 of the German Research Foundation to TS.

Notes

¹This is mainly because the use of schematic faces is seen as an elegant way to avoid the variations in physical features (such as luminance or spatial frequency) that are difficult to control in photographs of real faces (Öhman et al., 2001).

²We used G*Power 3 (Faul, Erdfelder, Lang, & Buchner, 2007) to calculate exemplary post hoc power values for Fenske and Eastwood (2003) and our study. For example, the power to explain a small effect (effect size f=0.1) in our study (set values: Error probability=.05, sample size = 20, number of groups = 1, repetitions = 120, correlation among measures=.5) was 0.835. This is comparable to Fenske and Eastwood with a power of 0.842 in Experiment 1A (sample size = 40, repetitions = 36), of 0.539 in Experiment 1B (sample size = 24, repetitions = 36), and of 0.877 in Experiment 2 (sample size = 48, repetitions = 30).

³Just like response times, error rates can represent motor response conflicts in flanker and response priming tasks. Specifically, participants tend to produce more errors in incompatible trials; this compatibility effect usually increases with the SOA between flanker (prime) and central (target) stimuli (e.g., Schmidt, Haberkamp, & Schmidt, 2011; Vorberg et al., 2003).

⁴Indeed, we strongly suppose that the Eriksen flanker effect and the response priming effect are merely two variants of the same effect, differing only in whether targets and distractors are presented at the same or neighbouring locations. Both effects have been shown to involve a processing conflict at the motor stage, to have similar time courses with respect to stimulus–onset asynchrony, and to be dissociable from visual awareness of the primes or flankers. In our lab, we use them interchangeably. From this background, it is a puzzle to us that most studies of the Eriksen effect confine themselves to a flanker–target SOA of 0 ms, where the effect is smallest.

⁵Note that Fenske and Eastwood (2003) allowed their participants a very small amount of training. In their Experiment 1A, each person performed 14 practice trials followed by 12 blocks of 24 trials. This contrasts with two 1-hour sessions with 28 practice trials followed by 60 blocks of 28 trials in our experiment. Indeed, the response time in Fenske and Eastwood's study (averaging about 550 ms in Exp. 1A, and about 600 ms in Exp. 2) is 150–200 ms slower than the average performance in our experiment. To rule out the possibility that training plays a major role in the size or time course of the compatibility effect, we investigated the modulation of compatibility effects by central face valence over the course of our experiment. An ANOVA for SOA = 0 revealed no significant interaction of the factors compatibility, valence of the central face, and number of trials (T; in six blocks of 280 successive trials each, approximating the length of Exp. 1A of Fenske & Eastwood), [F _C×V×T (5, 95) = 1.92, p=.098, . This was also the case when we contrasted the findings in the first 280 trials with the average results in all the following ones, [F _C×V×T (1, 19) = 2.40, p=.138, .

⁶Note that Horstmann, Borgstedt, and Heumann's (2006) paper does not contain any exact replication of Fenske and Eastwood's (2003) effect. The stimuli in their Experiment 1 differed in the shape of the mouth, the eyebrows, and the eyes themselves, more approximating a naturalistic sketch of a human face rather than a stylized smiley face. Even so, the authors concluded that it was isolated features of the stimulus, not the holistic expression, that determined the flanker asymmetry.

⁷Power to explain a small effect (f=0.1) was 0.801 (sample size = 20, repetitions = 108).