Paradigm and Hypotheses

This document plots the results of a 30 min version of Real Objects Attentional Blink, In this task, subjects performed an attentional blink paradigm in which they reported the identity of two colorful real world object targets amongst grayscale real world object distractors. T1 showed up as either the 4th or 6th image and T2 appeared at a lag of either 1, 2, or 8 images from T1. At the end of each trial, subjects reported which of four state x exemplar objects they saw for T1 and T2 using the a, s, d, and f keys. This experiment is meant to test whether subjects lose all information about T2 when this target is “blinked” or whether the loss of information is graded. If the loss is all-or-none, then we expect that subjects will be equally likely to report all 3 of the incorrect exemplar x state combinations when they are incorrect. If they lose information about the target in a graded manner, then we expect that the subjects will be more likely to report the T2 object as being of the same exemplar but different state or same state but different exemplar than different exemplar and different state. This would suggest that they maintained some information about T2 even when they were incorrect.

This data is being collected as an experimental sample. Data collection began on 3/3/2021. The future goal is to model this data using some flavor of General Recognition Theory, though there are currently plots attempting to re-create the analysis performed by Brady, Konkle, Alvarez, & Oliva, 2013. This analysis seems to be failing due to differences in the difficulty of state and exemplar discrimination.

Note: No error bars yet! Subjects who had an accuracy of less than 30% overall on T1 reports were excluded.

Data cleaning note: The file titled “./Mon%20Mar%2008%202021%2001:30:05%20GMT-0800%20(PST).txt” is incomplete somehow, with only about 1/4 of the typical file size. This subject has been excluded due to incomplete data.

Subjects’ (N = 91) overall T1 performance, before exclusion

Subjects’ (N = 75) performance on task, after exclusion

Statistics

T1 accuracy ANOVA

Effect DFn DFd SSn SSd F p p<.05 ges partial_eta_squared
(Intercept) 1 74 83.342 3.713 1661.142 0 * 0.953 0.957
trialLag 2 148 0.062 0.402 11.358 0 * 0.015 0.133

T1 accuracy t-tests

comparison t_val p_val d_val
Lag 1 vs. Lag 2 -4.946601 0.0000046 -0.2935557
Lag 1 vs. Lag 8 -2.800064 0.0065135 -0.1895032
Lag 2 vs. Lag 8 1.733439 0.0871826 0.1042748


T2 accuracy given correct T1 response ANOVA

Effect DFn DFd SSn SSd F p p<.05 ges partial_eta_squared
(Intercept) 1 74 47.994 2.134 1664.486 0 * 0.937 0.957
trialLag 2 148 1.628 1.092 110.266 0 * 0.335 0.598

T2 accuracy given correct T1 response t-tests

comparison t_val p_val d_val
Lag 1 vs. Lag 2 -0.9274653 0.3567002 -0.1059216
Lag 1 vs. Lag 8 -11.9999480 0.0000000 -1.3803714
Lag 2 vs. Lag 8 -12.0209874 0.0000000 -1.3198716

Demographics

The average age of the sample is 20.5866667 years.

The (messy) gender breakdown is as such:

gender count
1
F 4
female 15
Female 30
FEMALE 1
M 4
male 8
Male 10
nonbinary 1
Nonbinary Sex Female 1