Attentional Networks Task

ANT for measuring three attentional networks — alerting, orienting, and executive control — implemented in PsychoPy. Spanish-language instructions. Originally described in Fan et al. (2002) ¹, adapted following van Vugt & van den Hurk (2017) ².

Parameters

The numbers below are defined in prerandomize.py and mirrored in ANT.psyexp. Change them in one place and the other.

Parameter	Value
Fixation duration (D1)	0.4 + uniform(0, 1.2) s → mean 1.0 s
Cue routine duration	0.4 s total, cue stim on-screen for the first 0.1 s
Cue stim size	letterHeight 0.15 (asterisk, white) — clearly larger than the 0.05 fixation cross
Cue draw order	the asterisk is drawn on top of the cross (fixation_cross enters the window's autoDraw list at block start; text_1_cue / text_2_cue are appended later when the Cue routine begins, so they render after the cross)
Fixation cross	persistent for the entire block (single setAutoDraw stim, not redrawn per trial)
Flanker font	DejaVu Sans Mono so <, >, and - have identical advance widths (essential for the neutral --<-- / -->-- flankers to align with <<<<< / >>>>>)
Target max duration	1.7 s (or earlier on f / j keypress)
Assumed mean RT	0.6 s (used only to estimate task duration)
Mean trial duration	2.0 s
Trials per block	120 (4 cues × 6 targets × 5 reps)
Block lists	10 (block_a.csv … block_j.csv)
Max blocks per run	10 (each list used at most once per participant)
Inter-block break	60 s, self-paced
Response keys	f (left), j (right)
Target offset from fixation	±0.25 height units (above / below)
Window	1920 × 1080, fullscreen (windowed in pilot mode)

A trial is Fix → Cue → Target; a block is one CSV from lists/.

Cue conditions (cue column)

Each trial uses one of four cues, flashed for 100 ms inside the 400 ms cue routine. The fixation cross stays on screen for the full 400 ms.

Code	Cue type	Behaviour
NC	None	No cue text shown (cue text positioned far off-screen)
CC	Central	Single * at fixation
DC	Double	Two * simultaneously above and below fixation
SC	Spatial	Single * above OR below fixation (chosen at random), predicting where the target will appear

Target conditions (target column)

Five-arrow flanker; the participant responds to the central arrow only.

Stimulus	Direction	Flanker	Correct key
<<<<<	Left	Congruent	f
>>>>>	Right	Congruent	j
>><>>	Left	Incongruent	f
<<><<	Right	Incongruent	j
--<--	Left	Neutral	f
-->--	Right	Neutral	j

On SC trials the target appears at the cued position; on NC / CC / DC trials it appears above or below fixation with equal probability.

Pre-randomization

Trial sequences are generated ahead of time and stored as CSVs in lists/, one row per trial:

cue,target,correct
DC,<<><<,j
NC,--<--,f
...

Ten such block lists are generated (block_a.csv … block_j.csv), each containing 120 trials = 5 copies of every (cue, target) condition in a random order. Within each block every cue appears 30 times and f/j each appear 60 times by construction.

Participant schedules

prerandomize.py also emits 1000 per-participant schedule files in schedules/, named 000.csv … 999.csv — one for every 3-digit participant code. Each schedule has 10 rows (the 10 block-list letters permuted once) and looks like:

block,list_letter,condsFile
1,a,lists/block_a.csv
2,f,lists/block_f.csv
...
10,h,lists/block_h.csv

At runtime the experiment loads schedules/{participant}.csv, walks it sequentially, and stops as soon as block > nBlocks. With the same DEFAULT_SEED, all 1000 schedule files are byte-identical.

Generating the lists

With Nix:

nix run .#prerandomize

Or directly (no third-party dependencies — stdlib only):

python3 prerandomize.py [--seed N] [--output lists] [--schedules schedules] [--log LOG.md] [--report docs/report.html]

The script:

1. Builds 10 block lists of 120 trials each, asserting that every (cue, target) condition appears exactly 5 times and that f/j correct counts are balanced.
2. Builds 1000 participant schedules, asserting that every list-letter appears exactly once per schedule and rows are sorted by block ascending.
3. Renders docs/report.html (see below).
4. Compares the seed used against the most recent entry in LOG.md and appends a new line (timestamp + seed) only if the seed has changed.

HTML report

Every run also writes docs/report.html (single page, no external dependencies). It shows:

• Timing parameters — every constant from prerandomize.py so you can confirm what the experiment will actually do.
• Experiment schedule — wall-clock estimate as a function of nBlocks (1…10): instructions duration + task time (blocks + breaks)
  • total. Useful for planning sessions.
• Sample participant schedule — the contents of schedules/000.csv, rendered as a table, so you can sanity-check the format.
• Aggregate counts across all 10 lists — bar plots of cue type, flanker congruency, and target arrow. All should be flat by construction.
• Per-list timelines — one row per block list, one cell per trial. Cell fill encodes cue type (NC / CC / DC / SC), the top stripe encodes flanker congruency, and the arrow inside encodes target direction.

Open docs/report.html in any browser to review the lists before running the experiment.

Reproducibility & seed

The default seed is set at the top of prerandomize.py (constant DEFAULT_SEED). Running the script with the same seed always produces byte-identical CSV files. To regenerate with a new randomization, change DEFAULT_SEED (or pass --seed) and re-run; LOG.md will be appended with the new seed and the run timestamp. LOG.md is the audit trail: the most recent entry identifies which seed produced the current contents of lists/ and schedules/.

Implementation notes

The PsychoPy experiment (ANT.psyexp) consumes the pre-randomized files. The participant-info dialog has two relevant fields:

Field	Range	Purpose
participant	000…999 (or any string)	Numeric IDs map directly to schedules/{ID}.csv. Non-numeric IDs (e.g. pilot) are MD5-hashed deterministically to a 3-digit code
nBlocks	1…10	Number of blocks the participant will run

At experiment start, code_welcome builds schedules/{participant_id:03d}.csv, asserts the file exists, and initialises block = 0. The outer scheduleLoop then reads the 10 rows sequentially. The inner trials loop's conditionsFile is set per-row via $condsFile, and its nReps is 1 if block <= n_blocks else 0, so schedule rows past nBlocks skip the trial loop entirely. The BlockEnd break routine has the matching skipIf = block > n_blocks, so the participant only sees a break message after blocks they actually ran.

Trial structure

Fix  (0.4 – 1.6 s)
 → Cue (0.4 s, cue stim during first 0.1 s)
 → Target (≤ 1.7 s; ends on keypress)

The fixation cross is drawn once per block (turned on at the first trial of each block via setAutoDraw=True in code_fix, turned off in code_blockend) and stays visible continuously throughout the block. This avoids the 1–2-frame gaps that would appear if the cross were re-drawn in every routine, so the cross's appearance is not itself a per-trial temporal cue — only the jittered fixation duration is.

The flow is Welcome → Intro → 10× scheduleLoop[trials → BlockEnd] → End, with the schedule loop's row count effectively capped at nBlocks (rows beyond that run the empty inner loop and skip BlockEnd). The break message after each block reads "Este es el final del bloque {block} de {n_blocks}." so the participant always sees their progress.

Per-participant reports

After every experiment run, build per-participant HTML reports from each PsychoPy CSV in data/:

nix run .#report

This scans every data/*.csv, writes docs/reports/<csv-stem>.html for each finished run, and rebuilds docs/index.html so it lists every available report newest-first. Old-schema CSVs (without the new block / list_letter / condsFile columns) are skipped silently.

Each participant report has four tabs:

• Overview — run metadata, aggregate Alerting / Orienting / Conflict cards (with reference distributions from Fan & Posner), per-block scores table, and aggregate mean RT broken down by cue type and by flanker congruency. Scores are computed on correct trials only:

  • Alerting = RT(NC) − RT(DC) (no cue minus double cue)
  • Orienting = RT(CC) − RT(SC) (centre cue minus spatial cue)
  • Conflict = RT(IC) − RT(CG) (incongruent minus congruent flankers)

• Per-block — one card per block: score row, interactive Bokeh timeline (RT per trial coloured by cue, hover gives target + congruency

  • correctness, errors marked with an X), and a per-cue RT histogram.

• RT distributions — Bokeh histogram + KDE overlays of every correct RT, one figure coloured by cue type and one by flanker congruency. Click a legend entry to hide a series.

• Bayesian — closed-form Normal–Normal conjugate posterior for each score given Fan & Posner literature priors (μ_A=40, μ_O=50, μ_C=98 ms; σ=100 ms throughout). Per-block scores are treated as observations. The page shows the prior–reference–posterior density overlay per score, posterior summary table (μ, σ, 95% CrI), and individual block scores rugged at y=0.

Reports are self-contained (Bokeh JS/CSS inlined) so they open offline, and the visual style matches docs/report.html (the prerandomization report) and the N-back report site.

Footnotes

1. Fan, J., McCandliss, B. D., Sommer, T., Raz, A., & Posner, M. I. (2002). Testing the efficiency and independence of attentional networks. Journal of Cognitive Neuroscience, 14(3), 340–347. ↩

2. van Vugt, M. K., & van den Hurk, P. M. (2017). Modeling the effects of attentional cueing on meditators. Mindfulness, 8, 38–45. ↩