Attentional blink

Play a rapid serial stream with two embedded targets, then ask about both, with the second target’s report expected to suffer at short lags.

When to use it

Use this for the attentional blink paradigm: a rapid serial visual presentation (RSVP) stream contains two targets, T1 and T2, and reporting T2 suffers when it appears shortly after T1 (a lag of two or three items) but recovers at longer lags. The stream must play hands-free at a fixed rate, and both reports are collected only after the stream ends, so a trial unit is one self-advancing stream row plus two question rows, kept together through randomisation. T1’s position varies from run to run while the T1-to-T2 lag stays exact, which is what the stimOrder pin spec provides in a single row.

Trial file

type	content	stimFormat	stim1	stim2	stim3	stim4	stim5	stim6	stimScreens	stimOrder	presTime	ISI	randomBlock	lag	keyboard	key	button1
instructions	You will see a fast stream of digits with one letter hidden in it. Watch for the letter, and for the digit 0. After each stream, answer two questions.																NEXT
learn		word	4	7	K	3	0	8	1	random(pin 3@3_4 5@+2)	100*6	0*5	1	2
test		word	Which letter did you see? Press K or M.										1	2	k m	k
test		word	Was there a 0 in the stream? Press Y or N.										1	2	y n	y
learn		word	8	7	M	2	0	4	1	random(pin 3@2 5@+4)	100*6	0*5	2	4
test		word	Which letter did you see? Press K or M.										2	4	k m	m
test		word	Was there a 0 in the stream? Press Y or N.										2	4	y n	y
learn		word	3	6	K	8	4	2	1	random(pin 3@3_4)	100*6	0*5	3	0
test		word	Which letter did you see? Press K or M.										3	0	k m	k
test		word	Was there a 0 in the stream? Press Y or N.										3	0	y n	n

How it works

Each stream is one learn row with six word stimuli, stimScreens = 1 (one item per screen, consuming the whole list) at 100*6 (100 ms per item) and 0*5 (no gap, each item replaces the last), the standard RSVP rate of ten items per second. On a learn trial the response window never opens and the trial auto-advances as each presTime entry elapses, so the stream plays untouched from start to finish.

The stimOrder pin spec is the design’s engine. T1 is authored as the third stimulus and T2 (the digit 0) as the fifth, and the spec places them by those authored positions: random(pin 3@3_4 5@+2) lets T1 land at stream position 3 or 4 and puts T2 exactly 2 items later, while the distractor digits shuffle freely around them. The lag 4 stream pins T1 at position 2 (3@2) so a lag of 4 still fits a six-item stream, and the T2-absent stream pins only T1 (random(pin 3@3_4)). Each participant gets one arrangement drawn at the start of their session, and the arrangement actually shown is recorded in the stimOrder_actual results column, so the analysis can read T1’s true stream position per run. The lag column is a custom annotation recording the T1-to-T2 distance (2 where the blink is deepest, 4 for the longer lag, 0 for T2-absent streams); it is written on the whole unit, questions included, so every results row carries it.

The two test rows collect the reports: T1 identity as a forced choice between the letters used in the design, and T2 presence as yes/no, each scored by key. randomBlock ties each stream to its own questions, so the units shuffle as wholes and a question can never drift to another stream (randomBlock rather than a trialOrder spec, because spec runs split where type changes and these units mix learn and test).

Variations

• Vary the lag properly: author several streams per lag (2, 3, 4), changing only the relative pin (5@+2, 5@+3, 5@+4), and compare T2 accuracy on T1-correct trials across lags; the blink shows as a dip at lags 2 and 3.
• Faster or slower streams: 70*6 with 0*5 approaches the hard end; 150*6 softens it. Longer streams use more stimN columns with matching presTime/ISI repeats and give T1 more room to float (3@3_6 in an eight-item stream).
• Distinct targets: make T1 a colour singleton by using image stimuli (stimFormat = .png) with one differently-coloured item, keeping the same structure.
• Fixed stream order instead of floating targets: drop the stimOrder cell and author each stream’s item order literally; unit order still shuffles through randomBlock.