Based on assessment, best-fitting value for the exponent of the psychometric functions was 1.55 for our data (Supplementary Fig. 1). We also used the reported exponent values in previous researches measuring contrast-response functions psychophysically and electro physiologically (Sclar et al., 1990;Herrmann et al., 2010a) and got the same results (data not shown).
Cue-induced Temporal Attention Modulates Contrast Perception By Response Gain
Considering the main purpose of assessing the modulation of temporal attention induced by the pre-cue, data were collapsed across the reported targets. The mean contrast response functions under discrepant pre-cueing conditions (valid, invalid, neutral) were shown in Fig. 3A. It could be seen that no matter what the pre-cue was, d' increased with the increase of target contrast and saturated at high contrast, forming a typical s-shape curve. Importantly, the two psychometric functions representing perceptual sensitivity under valid and invalid pre-cueing conditions were not overlapped with each other but had an order with the one obtained in trials with valid pre-cues being always above the line indicating invalid pre-cueing condition. The curve depicting the neutral condition constantly laid between the other two psychometric functions. The clear separation of these three functions demonstrates the influence of temporal attention on the perception of contrast.
A decrease in threshold c50 with no change in asymptote dmax induced by attention is typical in contrast gain model while response gain model is characterized by an attentional boost in asymptote dmax in concomitant with no alternation in c50. The parameters of the fitted psychometric functions representing trials with discrepant (valid/neutral/invalid) pre-cues were compared to determine how cue-induced temporal attention affects perceptual sensitivity. c50 values for valid, neutral and invalid pre-cues were 0.111 (90% confidence interval = [0.101, 0.124]), 0.118 (90% confidence interval = [0.094, 0.151]), 0.119 (90% confidence interval = [0.085, 0.163]) respectively (Fig. 3A, left column) and were not significant different from each other (valid vs. neutral, p = 0.627; valid vs. invalid, p = 0.578; neutral vs. invalid, p = 0.961). The scenario was different for dmax (Fig. 3A, left column). The dmax value of CRF representing valid pre-cue was 2.55 (90% confidence interval = [2.45, 2.64]), which was not only significantly larger (p < 0.001) than its counterpart (1.26, 90% confidence interval = [1.10, 1.42]) obtained under invalid pre-cueing condition but also differed significantly (p < 0.001) from the dmax value of the CRF under neutral pre-cueing condition (1.79, 90% confidence interval = [1.64, 1.94]). Additionally, there was significant difference (p < 0.001) between the dmax values of the two CRFs under neutral and invalid pre-cueing conditions. For the psychometric functions from individual observers, a one-way ANOVA for repeated measures was used to analyze the influence of temporal attention induced by cue by including pre-cue type as the factor (valid, invalid, neutral). Pre-cue type did not show significant influence on c50 values (F(1,6) = 0.420, p = 0.541; Fig. 3A middle column) while dmax values were affected significantly by pre-cue type (F(1,6) = 12.059, p = 0.013 ; Fig. 3A right column). These observed effects of attention modulation on dmax in combination with the fact that no influence of attention was found on c50 from both the averaged and the individual data indicate that temporal attention induced by cue modulates perceptual sensitivity d' via response gain. Furthermore, the revealed differences in perceptual sensitivities between the valid and neutral pre-cueing conditions illustrate the enhancement of temporal attention on the perception of the attended target (attentional benefit) while the discovered decrease in perceptual sensitivity when the temporal cue was invalid compared with the neutral pre-cueing condition demonstrates the impairment of temporal attention on the perception of ignored distractors (attentional cost).
Previous study using similar paradigm to investigate the effect of temporal attention on sensitivity has demonstrated that sensitivity d' was comparable for T1 and T2 (Denison et al., 2017). We also divided our data based on whether the reported target was T1 or T2 and analyzed the CRFs of the two subgroups for each pre-cueing condition to assess whether temporal attention modulated the perception of T1 and T2 with the same pattern. Regardless of the reported target, the CRF denoting perceptual sensitivity under valid pre-cueing condition was always at the top, with the psychometric function representing neutral pre-cueing condition in the middle and the curve describing invalid pre-cueing condition at the bottom of the three (Fig. 3B left column, Fig. 3C left column). When the reported target was T1, the c50 values for different pre-cueing conditions were quite similar (valid: 0.111, 90% confidence interval = [0.095, 0.129]; neutral: 0.114, 90% confidence interval = [0.079, 0.164]); invalid: 0.097, 90% confidence interval = [0.057, 0.151]) (Fig. 3B, left column) and did not differ significantly from each other (valid vs. neutral, p = 0.897; valid vs. invalid, p = 0.466; neutral vs. invalid, p = 0.667). However, CRFs representing discrepant pre-cueing conditions had significantly different dmax values (valid vs. neutral, p < 0.001; valid vs. invalid, p < 0.001; neutral vs. invalid, p < 0.001) with the largest dmax value under the valid pre-cueing condition (2.50, 90% confidence interval = [2.35, 2.67]), the smallest dmax value under invalid pre-cueing condition (1.09, 90% confidence interval = [0.93, 1.30]) and the medium dmax value when the pre-cue was neutral (1.68, 90% confidence interval = [1.47, 1.98]). The analyze of one-way ANOVA for repeated measures showed that pre-cue type did not significantly impact c50 value (F(1,6) = 0.03, p = 0.868; Fig. 3B middle column) but had significant influence on dmax value (F(1,6) = 23.40, p = 0.003; Fig. 3B right column). Same pattern was found when T2 was the reported target. There were no significant differences (valid vs. neutral, p = 0.509; valid vs. invalid, p = 0.075; neutral vs. invalid, p = 0.435) among the similar c50 values of different pre-cueing conditions (valid: 0.112, 90% confidence interval = [0.096, 0.126]; neutral: 0.122, 90% confidence interval = [0.091, 0.167]); invalid: 0.147, 90% confidence interval = [0.094, 0.228]) (Fig. 3C, left column) but the dmax values for discrepant pre-cueing conditions (valid: 2.59, 90% confidence interval = [2.45, 2.74]; neutral: 1.90, 90% confidence interval = [1.69, 2.20]); invalid: 1.46, 90% confidence interval = [1.21, 1.79]) differed significant from each other (valid vs. neutral, p < 0.001; valid vs. invalid, p < 0.001; neutral vs. invalid, p = 0.014). Meanwhile, the significant impact of pre-cue type on dmax was observed with one-way ANOVA for repeated measures (F(1,6) = 4.14, p = 0.043; Fig. 3C right column) but pre-cueing method was not a significant influential factor for c50 (F(1,6) = 1.13, p = 0.330; Fig. 3C middle column). Based on these results, it could be concluded that temporal attention induced by cue modulated psychometric function by response gain no matter the reported target was T1 or T2, indicating the independence of the modulation pattern on the order of the reported target.
Together, these results illustrate that in our experiment perceptual sensitivity for contrast was modulated by temporal expectation which was manipulated by cue. Additionally, our data was fitted by response gain model but not contrast gain model or the mixture model. Meanwhile, our data also showed perceptual tradeoffs due to temporal attention, illustrated by the enhanced perceptual sensitivity for the target occurring at the attended time point and the worsened perception for the distractor happening at the unattended time point.
Impacts Of Cue-induced Temporal Attention On Reaction Time
The effects of cue-induced temporal attention on RT were also observed with a similar pattern as for perceptual sensitivity d'(Fig. 4), illustrated by fastest RTs on trials with valid pre-cues (Mean ± SEM: 0.511 ± 0.015 s), slowest RTs on trials that were invalid pre-cued (0.645 ± 0.019 s), and intermediate RTs on neutral trials (0.538 ± 0.008 s). The mean RTs of different conditions were submitted to a three-way analysis of variance (ANOVA) for repeated measures with pre-cue type (valid, invalid, neutral), target contrast (seven levels) and reported target (T1, T2) as three factors. It was not surprising that significant main effect of pre-cue type was observed (F(2,12) = 55.440; p < 0.001). There was also a significant main effect of reported target (F(1,6) = 7.967; p = 0.028), reflecting a faster discrimination for tilt of T2 than T1(Fig. 4B, C), which might be due to subject being more prepared for orientation discrimination at the appearance of late T2 than early T1. No significant main effect of target contrast was found (F(6,30) = 3.290; p = 0.399), indicating that in our experiment RT was not influenced by the contrast of target, and other factors that could lead to changes in RT such as motor preparation (Correa, 2012) or criterion changes (Carrasco & McElree,2001) played more influential roles on RT. The two-way interaction effects of the three pairs were not significant (Pre-cue type × Target contrast: F(12,72) = 1.301, p = 0.303; Pre-cue type × Reported target: F(2,12) = 3.973, p = 0.064; Target contrast × Reported target: F(6,36) = 2.554, p = 0.093). The three-way interaction effect of all three factors was also not significant (F(12,72) = 1.297; p = 0.307).
The results on RT demonstrate that the observed changes in contrast sensitivity induced by temporal attention could not be attributed to speed-accuracy tradeoff.