All results currently shown are from SIMULATED participants — no human data have been collected.
Results · detection probability · SIMULATED
Detection probability
Whether a change was detected at all — pressed, after onset, and localised to the right object — as a function of the design factors and the stimulus features. Catch-trial false-alarm rate: 3.4% (294 catch trials).
01 Magnitude and speed
Detection rises with end magnitude and falls with duration
Detection vs end magnitude, by duration.Means ±95% CI over trials.Detection vs speed.Speed = normalised end magnitude ÷ duration; the design has nine discrete speeds.
By duration
duration (ms)
n
detected
±95% CI
median RT (ms)
mean magnitude at detection
mean confidence
3000
591
40.3%
4.0%
2911.6
0.68
3.04
6000
545
36.7%
4.1%
5502.1
0.67
3.12
12000
538
36.3%
4.1%
10001.4
0.61
2.94
By magnitude level
magnitude level (0 small … 2 large)
n
detected
±95% CI
median RT (ms)
mean magnitude at detection
mean confidence
0
577
12.0%
2.6%
4140.2
0.27
2.09
1
538
33.5%
4.0%
4775.5
0.52
2.76
2
559
68.7%
3.8%
4635.9
0.79
3.33
Duration × magnitude
duration × level
n
detected
±95% CI
median RT (ms)
mean magnitude at detection
mean confidence
3000 × 0
206
13.6%
4.7%
2861.6
0.30
2.14
3000 × 1
190
35.8%
6.8%
3180.7
0.53
2.72
3000 × 2
195
72.8%
6.3%
2820.1
0.83
3.37
6000 × 0
193
8.3%
3.9%
5276
0.30
2.13
6000 × 1
173
30.6%
6.9%
5526.8
0.50
2.83
6000 × 2
179
73.2%
6.5%
5533
0.79
3.36
12000 × 0
178
14.0%
5.1%
6658.6
0.21
2.00
12000 × 1
175
33.7%
7.0%
10681.1
0.52
2.73
12000 × 2
185
60.0%
7.1%
10048.5
0.75
3.26
02 Type and place
Change type, eccentricity, clutter and category
Detection by change type.Types are compared at their own magnitude scales; the normalisation is within type.Where changes are missed.Detection rate by target location on an 8×5 grid; cross marks fixation.
Peripheral changes need more magnitude before they are noticed.Distribution of normalised magnitude at the moment of response among detected trials; the censored share (never detected) is shown above each box and is modelled, not dropped, in the AFT.
By change type
type
n
detected
±95% CI
median RT (ms)
mean magnitude at detection
mean confidence
appear
238
51.7%
6.4%
4138.9
0.64
3.05
brightness
233
40.3%
6.3%
5200
0.67
3.10
disappear
236
44.5%
6.4%
4723.9
0.66
3.07
hue
237
30.0%
5.8%
5180
0.68
3.08
orientation
241
23.2%
5.4%
4631.4
0.71
3.07
position
241
40.7%
6.2%
4194.7
0.60
2.85
size
248
34.7%
5.9%
4513.6
0.67
3.05
By eccentricity bin
bin
n
detected
±95% CI
median RT (ms)
mean magnitude at detection
mean confidence
central
511
52.6%
4.3%
4003
0.59
2.86
mid
568
27.6%
3.7%
5546.4
0.69
3.12
peripheral
595
34.8%
3.8%
4615.9
0.72
3.19
Eccentricity × magnitude
bin × level
n
detected
±95% CI
median RT (ms)
mean magnitude at detection
mean confidence
central × 0
175
20.6%
6.0%
4089.2
0.25
2.00
central × 1
143
52.4%
8.2%
3680.7
0.46
2.63
central × 2
193
81.9%
5.4%
3913.7
0.73
3.17
mid × 0
213
8.9%
3.8%
4302.9
0.28
2.16
mid × 1
198
26.8%
6.2%
6086.9
0.57
3.00
mid × 2
157
54.1%
7.8%
5457
0.86
3.41
peripheral × 0
189
7.4%
3.7%
4218.3
0.29
2.21
peripheral × 1
197
26.4%
6.2%
4675.2
0.54
2.69
peripheral × 2
209
67.5%
6.4%
4372.5
0.83
3.47
By clutter
clutter
n
detected
±95% CI
median RT (ms)
mean magnitude at detection
mean confidence
high
821
30.4%
3.2%
5425.9
0.70
3.14
low
853
44.9%
3.3%
4264.2
0.63
2.96
By target category
category
n
detected
±95% CI
median RT (ms)
mean magnitude at detection
mean confidence
bird
206
28.2%
6.2%
3890.3
0.62
3.05
box
224
25.0%
5.7%
6117.7
0.70
3.16
bush
158
33.5%
7.4%
6074.1
0.69
3.17
car
145
46.2%
8.1%
5438.1
0.56
2.75
house
140
72.9%
7.4%
3221.2
0.63
2.86
lamp
145
28.3%
7.3%
7446.9
0.72
3.32
person
282
33.0%
5.5%
4633.8
0.74
3.27
rock
147
36.7%
7.8%
4102.1
0.66
3.02
sign
130
49.2%
8.6%
5471.3
0.58
2.88
tree
97
46.4%
10.0%
2885.5
0.70
3.00
03 Models
Mixed-effects logistic regression
Random intercept per participant (SD = 0.9485 on the logit scale, statsmodels BinomialBayesMixedGLM (VB)); alongside a fixed-effects fit with cluster-robust standard errors. Reference change type: hue.
Coefficients (log-odds of detection)
term
mixed β
mixed SE
OR
cluster-robust β
95% CI
p
Intercept
-3.264
0.068
0.04
-3.201
[-4.10, -2.30]
p = 2.7e-12
log2_duration
-0.227
0.025
0.80
-0.179
[-0.36, 0.00]
p = 0.052
magnitude_norm
5.422
0.089
226.31
4.862
[4.26, 5.46]
p = 1.0e-56
ecc_100u
-0.349
0.025
0.71
-0.262
[-0.40, -0.13]
p = 1.4e-4
log_area_c
0.531
0.083
1.70
0.488
[0.32, 0.65]
p = 8.0e-9
contrast_c
0.723
0.304
2.06
0.352
[-0.27, 0.98]
p = 0.268
clutter_high
-0.760
0.099
0.47
-0.696
[-0.99, -0.40]
p = 4.8e-6
sal_z
0.421
0.067
1.52
0.404
[0.24, 0.57]
p = 9.3e-7
target_importance
0.849
0.110
2.34
0.817
[0.46, 1.17]
p = 5.5e-6
trial_frac
0.173
0.119
1.19
0.170
[-0.23, 0.57]
p = 0.405
type_appear
0.574
0.168
1.78
0.571
[0.16, 0.99]
p = 0.007
type_brightness
0.256
0.177
1.29
0.353
[-0.17, 0.88]
p = 0.186
type_disappear
0.293
0.165
1.34
0.333
[-0.10, 0.77]
p = 0.133
type_orientation
-0.765
0.202
0.47
-0.647
[-1.19, -0.10]
p = 0.020
type_position
0.661
0.181
1.94
0.609
[0.09, 1.13]
p = 0.023
type_size
0.218
0.178
1.24
0.252
[-0.19, 0.70]
p = 0.267
Participants differ a lot.Per-participant detection rate ±95% CI; dashed line = grand mean.
