Frequency-dependent modulation of foveal contrast sensitivity by fine-scale exogenously triggered attention
- Department of Brain and Cognitive Sciences, University of Rochester, United States
- Goergen Institute for Data Science and Artificial Intelligence, University of Rochester, United States
- Center for Visual Science, University of Rochester, United States
- Department of Neurosciences, University of Rochester, United States
Figures
Figure 1
Fine-tuning exogenous attention within the foveola.
(A) Fine-tuning of exogenous attention within the foveola occurs. For example, when we are looking at a distant traffic light, occupying less than 1° of our visual field, it suddenly turns green, capturing our attention and prompting us to move forward. As a result of the fine-tuning of exogenous attention, contrast sensitivity could be enhanced for a narrow range of spatial frequencies, centered around lower spatial frequencies (blue) or higher spatial frequencies (green) (B). On the other hand, contrast sensitivity may be enhanced uniformly across a wide range of spatial frequencies (C).
Figure 2 with 2 supplements
Experimental protocol.
(A) Trials started with a fixation marker at the center of the monitor. Observers were instructed to maintain fixation at the center throughout the trial. After a brief flash of the exogenous cue to capture observers’ attention, two Gabor patches independently tilted (±45°) were briefly displayed, one on each side of the fixation marker. At the end of the trial, a response cue appeared, and observers had to report the orientation of the stimulus that was previously presented at the cued location. In valid trials, the exogenous cue and response cue indicated the same spatial location. In neutral trials, no exogenous cue was presented. Valid and neutral trials had the same probability of occurrence. (B) Size of the stimuli. The Gabor patches had a Gaussian window of 5.4′ standard deviation, creating a 30′ × 30′ visible region. (C) Stimuli used in the experiment. Gabor patch of all spatial frequencies tested from 4 to 20 cycles per degree (CPD). (D) 68% contour of the gaze probability distribution in valid and neutral conditions during Gabor presentation. Color represents individual observers.
Figure 2—figure supplement 1
Number of trials included for analysis by condition.
(A) Number of trials and (B) proportion of trials included for analysis in each condition (n = 7). Each line is an observer. Error bars show the bootstrapped 95% confidence intervals.
Figure 2—figure supplement 2
Average saccades onset relative to the response cue onset in valid and neutral conditions.
Only saccades that occurred after the fixation period were included in the analysis. The trials presented here were not included in main analyses examining the effects of attention. Lines represent individual observers. Error bars represent ±1 standard error of the mean (SEM). The asterisk indicates a significant difference between conditions (t-test, n = 7, p < 0.01).
Figure 3 with 4 supplements
Effect of fine-grained exogenous attention on contrast sensitivity.
(A) Psychometric functions illustrating example observers’ discrimination accuracy for Gabor patches with spatial frequencies of 4, 8, 12, and 20 cycles per degree (CPD). The size of each dot corresponds to the number of trials included at a specific contrast value. Vertical lines indicate contrast thresholds, while horizontal lines represent the accuracy level midway between chance performance and maximum performance. Error bars around contrast thresholds represent 95% credible intervals. (B) Average contrast sensitivity, calculated as the inverse of the contrast threshold, across spatial frequencies in valid and neutral conditions (n = 7). Each dot represents an individual observer. Error bars denote ±1 standard error of the mean (SEM). Asterisks indicate significant post hoc pairwise comparisons between spatial frequencies (p < 0.05). (C) Average contrast sensitivity in neutral condition against that in valid condition at each spatial frequency (n = 7). Each dot represents an individual observer. Error bars indicate the bootstrapped 95% confidence intervals. (D) Average difference in log-scaled contrast sensitivity between valid and neutral conditions across different spatial frequencies (n = 7). Each line corresponds to the log-scaled contrast sensitivities from each observer. Error bars represent the bootstrapped 95% confidence intervals. Asterisks mark post hoc pairwise comparison results between valid and neutral conditions within each spatial frequency (ps < 0.05).
Figure 3—figure supplement 1
Psychometric Weibull function fits for all observers and conditions (n = 7).
Vertical lines indicate contrast thresholds, while horizontal lines represent the accuracy level midway between chance performance and maximum performance. Error bars around contrast thresholds represent 95% credible intervals.
Figure 3—figure supplement 2
Effect of covert attention on contrast sensitivity at 2 CPD compared to the other spatial frequencies tested.
(A) Average contrast sensitivity, calculated as the inverse of the contrast threshold, across spatial frequencies, including 2 CPD in valid and neutral conditions (n = 7). Each dot represents an individual observer. Error bars denote ±1 SEM. Asterisks mark significant post hoc pairwise comparisons between 2 CPD and other spatial frequencies (ps < 0.05). (B) Average difference in log-scaled contrast sensitivity between valid and neutral conditions across different spatial frequencies (n = 7). Each line corresponds to the log-scaled contrast sensitivities from each observer. Error bars represent the bootstrapped 95% confidence intervals. Asterisks mark post hoc pairwise comparison results between valid and neutral conditions within each spatial frequency (p = 0.04 for valid vs. neutral at 2 CPD; ps > 0.17 for mean gain at 2 CPD vs. other spatial frequencies [SFs]).
Figure 3—figure supplement 3
Relation between mean and variability of contrast sensitivity (CS), depending on whether CS is log-transformed (right panel) or not (left panel).
Without a log-transform, the mean and standard deviation (SD) of CS are almost perfectly correlated, constituting a strong violation of the homoskedasticity assumption of linear models. For the present data, log-transforming CS mostly removes this correlation (except for the 20 CPD condition).
Figure 3—figure supplement 4
Average attentional benefit calculated as the ratio in contrast sensitivity between valid and neutral conditions (n = 7).
Error bars are ±1 SEM.
Figure 4 with 1 supplement
Effect of fine-grained exogenous attention on asymptotic performance.
(A) Average asymptotic performance, defined as discrimination accuracy at maximum contrast, pooled across spatial frequencies in valid and neutral conditions (n = 7). Each dot represents an individual observer. Error bars denote ±1 SEM. Asterisk indicates a significant main effect of attention (valid vs. neutral) estimated from the GLMM (p < 0.01). (B) Average asymptotic performance across spatial frequencies in valid and neutral conditions (n = 7). Each dot represents an individual observer. Error bars denote ±1 SEM. Asterisks mark significant differences in asymptotic performance between pairs of spatial frequencies (ps < 0.05). (C) Average asymptotic performance in the neutral condition against that in the valid condition at each spatial frequency (n = 7). Each dot represents an individual observer. Error bars indicate the bootstrapped 95% confidence intervals. (D) Average difference in asymptotic performance between valid and neutral conditions across different spatial frequencies (n = 7). Each line corresponds to an individual observer. Error bars indicate the bootstrapped confidence intervals. Asterisks mark post hoc pairwise comparison results between valid and neutral conditions within each spatial frequency (ps < 0.05).
Figure 4—figure supplement 1
Maximum a posteriori (MAP) estimates (points) and 95% confidence intervals for contrast sensitivity (CS) and asymptotic performance (AP) for each experimental condition across seven observers.
Both CS and AP were transformed in the same way as used in our mixed-effects analyses.
Tables
Table 1
Average number and percentage of trials retained for analysis after filtering (mean ± SE across observers) across different spatial frequencies and cueing conditions.
Brackets indicate bootstrapped 95% confidence intervals. Observers who were unable to perform the task at 20 CPD were excluded from the 20 CPD trial counts.
| Valid | Neutral | |||
|---|---|---|---|---|
| SF (CPD) | Count | % | Count | % |
| 4 | 514 [474, 549] | 79.1 [66.1, 90.4] | 518 [486, 548] | 79.6 [67.3, 90.1] |
| 8 | 491 [417, 551] | 82.0 [73.4, 89.3] | 489 [418, 553] | 81.8 [73.3, 89.6] |
| 12 | 487 [416, 552] | 80.7 [72.9, 88.0] | 507 [444, 567] | 84.1 [77.5, 90.6] |
| 20 | 531 [504, 556] | 86.8 [82.9, 90.9] | 528 [497, 554] | 86.4 [81.1, 91.6] |
Appendix 1—table 1
Linear mixed-effects model predicting log-transformed contrast sensitivity from attention, spatial frequency, and their interaction.
| Fixed effect | Estimate | SE | 95% CI | df | t | p |
|---|---|---|---|---|---|---|
| Intercept | 1.879 | 0.148 | [1.583, 2.175] | 5.96 | 12.73 | <0.001 |
| Attention (valid vs. neutral) | 0.084 | 0.028 | [0.029, 0.139] | 5.54 | 2.99 | 0.027 |
| Frequency (8 vs. 4) | –0.566 | 0.083 | [–0.729, –0.403] | 15.95 | –6.85 | <0.001 |
| Frequency (12 vs. 8) | –0.485 | 0.083 | [–0.648, –0.322] | 15.95 | –5.87 | <0.001 |
| Frequency (20 vs. 12) | –0.997 | 0.093 | [–1.179, –0.815] | 16.04 | –10.73 | <0.001 |
| Attention × frequency (8 vs. 4) | –0.008 | 0.054 | [–0.114, 0.098] | 15.53 | –0.14 | 0.889 |
| Attention × frequency (12 vs. 8) | –0.134 | 0.054 | [–0.240, –0.028] | 15.53 | –2.48 | 0.025 |
| Attention × frequency (20 vs. 12) | 0.022 | 0.060 | [–0.096, 0.140] | 16.46 | 0.37 | 0.714 |
Appendix 1—table 2
Simple effects of attention on contrast sensitivity at each spatial frequency (SF) and their effect sizes in Cohen’s d.
| SF | Estimate | SE | 95% CI | df | t | p | Cohen’s d | Estimate (ratio) |
|---|---|---|---|---|---|---|---|---|
| 4 | 0.151 | 0.043 | [0.061, 0.240] | 19.76 | 3.52 | 0.002 | 0.79 | 1.16 |
| 8 | 0.143 | 0.043 | [0.054, 0.233] | 19.76 | 3.34 | 0.003 | 0.75 | 1.15 |
| 12 | 0.009 | 0.043 | [–0.080, 0.099] | 19.76 | 0.22 | 0.830 | 0.05 | 1.01 |
| 20 | 0.032 | 0.051 | [–0.074, 0.137] | 21.37 | 0.62 | 0.540 | 0.13 | 1.03 |
Appendix 1—table 3
Interaction contrasts comparing attention effects between spatial frequency (SF) pairs and their effect sizes in Cohen’s d.
| SF pair | Estimate | SE | 95% CI | df | t | p | Cohen’s d |
|---|---|---|---|---|---|---|---|
| 4–8 | 0.008 | 0.054 | [–0.107, 0.122] | 16.0 | 0.14 | 0.889 | 0.04 |
| 4–12 | 0.141 | 0.054 | [0.027, 0.256] | 16.0 | 2.62 | 0.019 | 0.65 |
| 4–20 | 0.119 | 0.061 | [–0.009, 0.247] | 16.9 | 1.97 | 0.066 | 0.48 |
| 8–12 | 0.134 | 0.054 | [0.019, 0.248] | 16.0 | 2.48 | 0.025 | 0.62 |
| 8–20 | 0.111 | 0.061 | [–0.016, 0.239] | 16.9 | 1.84 | 0.084 | 0.45 |
| 12–20 | –0.022 | 0.061 | [–0.150, 0.106] | 16.9 | –0.37 | 0.717 | –0.09 |
Appendix 2—table 1
Generalized linear mixed-effects model predicting normalized asymptotic performance from attention, spatial frequency, and their interaction.
| Fixed effect | Estimate | SE | 95% CI | z | p |
|---|---|---|---|---|---|
| Intercept | 2.676 | 0.349 | [1.99, 3.36] | 7.67 | <0.001 |
| Attention (valid vs. neutral) | 0.960 | 0.178 | [0.61, 1.31] | 5.41 | <0.001 |
| Frequency (8 vs. 4) | 0.289 | 0.353 | [–0.40, 0.98] | 0.82 | 0.414 |
| Frequency (12 vs. 8) | –0.181 | 0.351 | [–0.87, 0.51] | –0.52 | 0.606 |
| Frequency (20 vs. 12) | –1.020 | 0.394 | [–1.79, –0.25] | –2.59 | 0.010 |
| Attention × frequency (8 vs. 4) | 1.127 | 0.499 | [0.15, 2.11] | 2.26 | 0.024 |
| Attention × frequency (12 vs. 8) | –0.820 | 0.506 | [–1.81, 0.17] | –1.62 | 0.105 |
| Attention × frequency (20 vs. 12) | –0.167 | 0.493 | [–1.13, 0.80] | –0.34 | 0.734 |
Appendix 2—table 2
Simple effects of attention on asymptotic performance at each spatial frequency (SF) and their effect sizes in Cohen’s d.
| SF | Estimate | SE | 95% CI | z | p | Cohen’s d | Estimate (prob. diff.) |
|---|---|---|---|---|---|---|---|
| 4 | 0.566 | 0.325 | [–0.072, 1.204] | 1.74 | 0.082 | 0.39 | 0.030 |
| 8 | 1.693 | 0.392 | [0.925, 2.461] | 4.32 | <0.001 | 0.97 | 0.077 |
| 12 | 0.874 | 0.329 | [0.230, 1.518] | 2.66 | 0.008 | 0.60 | 0.044 |
| 20 | 0.707 | 0.361 | [0.000, 1.414] | 1.96 | 0.050 | 0.42 | 0.081 |
Appendix 2—table 3
Interaction contrasts comparing attention effects between spatial frequency (SF) pairs on asymptotic performance and their effect sizes in Cohen’s d.
| SF pair | Estimate | SE | 95% CI | z | p | Cohen’s d |
|---|---|---|---|---|---|---|
| 4–8 | –1.127 | 0.499 | [–2.106, –0.149] | –2.26 | 0.024 | –0.56 |
| 4–12 | –0.308 | 0.460 | [–1.209, 0.593] | –0.67 | 0.503 | –0.17 |
| 4–20 | –0.140 | 0.488 | [–1.098, 0.817] | –0.29 | 0.774 | –0.07 |
| 8–12 | 0.820 | 0.506 | [–0.173, 1.812] | 1.62 | 0.105 | 0.40 |
| 8–20 | 0.987 | 0.537 | [–0.066, 2.040] | 1.84 | 0.066 | 0.45 |
| 12–20 | 0.167 | 0.493 | [–0.798, 1.133] | 0.34 | 0.734 | 0.08 |
Additional files
Download links
A two-part list of links to download the article, or parts of the article, in various formats.
Downloads (link to download the article as PDF)
Open citations (links to open the citations from this article in various online reference manager services)
Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)
Frequency-dependent modulation of foveal contrast sensitivity by fine-scale exogenously triggered attention
eLife 14:RP108788.
https://doi.org/10.7554/eLife.108788.3
