Arrows

What you see

Each Arrows trial is a block of rows, not a single row. You press Start, a 3-2-1 countdown runs, and then the rows arrive one after another at a pace the app sets. You do not click through them; the clock does.

Every row goes through the same beats:

1. A fixation cross (+) sits in the middle of the strip for a randomised 300 to 600 ms. The random length stops you from timing your press to a rhythm.
2. The row appears: 5, 7 or 9 arrows side by side. The centre arrow is the target and is drawn tinted, so you always know where to look. The rest are flankers.
3. After the exposure time the row vanishes and the strip is left empty. You can still answer during this blank.
4. The row's deadline is measured from the moment the row appeared. The next row's fixation begins 120 ms after the deadline passes.

You answer with the left and right arrow keys, or by pressing the Left and Right buttons under the strip. Only your first answer on a row is kept, and answers during the fixation cross are ignored. A row you let pass without answering is recorded as no answer.

Three kinds of row

Every block mixes three conditions in exactly equal numbers, shuffled so that no condition appears more than three times in a row. Within each condition, left and right targets are split as evenly as the count allows.

• Congruent: every flanker points the same way as the centre. The flankers agree with you, so these rows feel easy.
• Incongruent: every flanker points the opposite way. This is where the cost lives.
• Neutral: the flankers are outline diamonds that carry no direction. They give a baseline: flankers are present, but they push nowhere.

What changes as levels rise

The level sets five things: how many rows are in the block, how long each row stays on screen, the deadline, how many arrows make up a row, and whether the flankers are slightly tilted.

The tilt at levels 8 to 10 matters more than it looks. Tilted flankers are not identical to the target any more, so the row looks busier, and your eye has to find the one clean horizontal arrow in the middle. On neutral rows the diamonds are never tilted.

What it measures

Arrows measures interference control: the ability to act on the one relevant signal while irrelevant signals nearby are trying to trigger a different action. In our attention map it is the only game on the interference-control facet, and within the Attention dimension it carries 22 per cent of the weight, the joint largest share with Pulse.

The task descends directly from the Eriksen flanker task. In 1974 Barbara and Charles Eriksen asked people to identify a target letter in the centre of a short row and showed that nearby "noise" letters linked to the other response slowed them down, even though everyone knew exactly where the target would be. They concluded that people cannot prevent the processing of noise letters that fall within about one degree of the target. The arrow version became standard later; it is the conflict component of the Attention Network Test that Jin Fan, Michael Posner and colleagues published in 2002. Participants judge whether a central arrow points left or right while arrows on either side point the same way or the opposite way.

The standard explanation is response competition. Your visual system cannot fully shut out things that sit a fraction of a degree from where you are looking. The flankers are processed in parallel with the centre arrow, they partially activate their own response, and the correct response has to win against that activation. The difference between your incongruent and congruent reaction times, the flanker effect or interference cost, is the classic measure of how much the irrelevant arrows got through.

How big is the cost? In the original Attention Network Test study, 40 adults were on average 84 ms slower on incongruent rows than on congruent ones, with a standard deviation of 25 ms. So a cost somewhere between about 60 and 110 ms is ordinary. Your report shows your own figure as Extra time when the outer arrows disagreed.

How it is scored here

Our scoring is written so that you can see exactly what earns credit.

• A row is correct when you pressed the direction of the centre arrow and your press landed at or before that row's deadline.
• A row is an error when you pressed the wrong direction, or pressed after the deadline in the short gap before the next row.
• A row is an omission when you did not answer at all. Omissions score the same as errors: zero.
• Very fast presses under 150 ms are flagged as anticipations in your trial review, because no one genuinely reads and answers a row that quickly. The flag does not change the row's credit: a fast press in the right direction still counts as correct.

The block's score is simply the share of rows you got right. A perfect block scores full marks; 15 right out of 18 scores 83 per cent.

Speed inside the deadline does not add points. Your reaction times are recorded and your report shows them, including your interference cost, shown as Extra time when the outer arrows disagreed (the median time on correct incongruent rows minus the median on correct congruent rows), and that cost as a share of your congruent-row time. Those are diagnostics. The score itself asks one question per row: right direction, in time?

Where speed does matter is through the level. Higher levels shrink the deadline, and every block is given a measured difficulty from its own settings: tighter deadlines weigh most, then longer rows and tilted flankers. Your game score is built from difficulty earned per trial, so clearing a level 8 block cleanly is worth far more than clearing a level 3 block cleanly. On our scale, 50 means you are performing at what the difficulty you work at predicts.

How the level moves

In adaptive practice the level changes after every block:

• Every row correct: up one level.
• Some rows correct, some not: stay where you are.
• Nothing correct: down one level.

This is strict by design. At level 10 a block has 27 rows, and one slip holds you in place. Expect to plateau for several blocks at the level where your accuracy tops out, and treat that plateau as information, not failure.

Where you will meet it

Flanker tasks are a staple of attention research and have found their way into several game-based hiring assessments. None of these is our game, and we have no link with any of the vendors, but they belong to the same family of task.

• Arctic Shores, a game-based assessment vendor, describes a level called "Direction" in its candidate help centre: you press Q or P depending on which way the arrow in the centre is pointing. Arctic Shores does not publish which trait each task feeds and says tasks vary with what each employer includes.
• pymetrics (now part of Harver) lists attention and focus among the traits its game suite measures. An independent audit published at the ACM FAccT conference describes a core set of twelve games "derived from peer-reviewed psychological studies". Candidates report that one of the games is called Arrows; prep sites describe it as responding to arrows under a rule that can change with their colour, which adds rule switching on top of direction conflict. That description is candidate-reported, not official.
• Research and clinical batteries use the arrow flanker as the conflict part of the Attention Network Test, which is why you may meet it in university studies or cognitive screening.

The method

1. Park your eyes on the centre before the row arrives

   During the fixation cross, look at the cross itself, not at the buttons or the keyboard. The target will appear exactly there. Moving your eyes after the row appears costs time you cannot afford at high levels.

2. Rest a finger on each key

   Keep your left index finger on the left arrow key and your right index finger on the right arrow key, or both thumbs over the Left and Right buttons on a phone. Pressing with one finger that has to travel adds time and invites errors.

3. Read only the tinted arrow

   Narrow your attention to one character width. The tint is there to help: the target is the only coloured arrow. Treat the flankers as texture, not information.

4. Answer as the direction registers, not after a check

   Do not look at the row, decide, and then look again to confirm. At level 5 you have 1,400 ms; most of that should be spare. One read, one press.

5. If the row vanished before you were sure, still answer

   The blank after the row is still inside the answer window. If you caught the direction but the row disappeared, press anyway. An omission is always zero; a best read is usually right.

6. Reset after a mistake

   When you catch yourself pressing the flankers' direction, do not rush the next row to make up for it. The next fixation cross is your reset. Look at it and start fresh.

Worked example 1: four rows at speed

Here are four rows of the kind you see from level 4 up: (a) and (d) have seven arrows, as at levels 4 to 6, and (b) and (c) have nine with tilted flankers, as at levels 8 to 10. Give yourself one glance each and say the answer out loud before opening the solution.

Show the answers

• (a) Right. Incongruent: six flankers point left, the centre points right. This is the row that catches people who read the whole row as a pattern.
• (b) Left. Congruent with tilted flankers. The tilt does not change any flanker's direction, only its angle. Everything agrees, so this should be your fastest answer.
• (c) Right. Incongruent with tilt. The tilted flankers make the one level arrow easier to spot if you look for "the flat one in the middle".
• (d) Left. Neutral. The diamonds point nowhere, so there is no conflict, but the row still has to be read.

If you said "left" for (a) or (c), you answered the flankers. That is the exact error the game is measuring.

Worked example 2: reading your own report

Suppose a level 5 block (18 rows, deadline 1,400 ms) ends like this:

What is the block score, what is your interference cost, and what happens to your level?

Show the solution

• Score: 16 correct out of 18 rows is 16 ÷ 18, about 89 per cent.
• Interference cost: 590 − 470 = 120 ms. Your report would also show this relative to your speed: 120 ÷ 470 is about 26 per cent.
• Level: not every row was correct, but some were, so you stay at level 5.

The pattern matters more than the numbers. Both losses were on incongruent rows, and one of them was a row you did not answer. That points to hesitation: the conflict slowed you enough that the deadline passed. The fix is not "try harder", it is the method above: eyes parked on the centre, one read, one press.

Tips and traps

Levelling up

Three things get harder as you climb, and each needs a different adjustment.

Shorter exposure (1,200 ms down to 300 ms). At high levels the row is gone before a slow reader has finished. The fix is preparation: eyes on the centre during the cross, fingers resting on the keys. You are not reading faster; you are wasting less time before reading starts.

Longer rows (5, 7, then 9 arrows). More flankers mean more competing signal, and smaller arrows. Keep your focus narrow: you only ever need the middle character.

Tighter deadlines (2,000 ms down to 900 ms). At level 10 you have 900 ms from onset. That is still generous for a correct read; most people's hesitation, not their reading speed, is what runs out the clock.

Tilted flankers (levels 8 to 10). Use the tilt: the target is the only perfectly level arrow.

1. Week 1
   • Three sessions of 4 to 6 blocks at your starting level, adaptive on
   • After each session, note your interference cost and your incongruent errors
   • Practise the ready position: eyes on the cross, a finger on each key
2. Week 2
   • Four sessions; aim for every row correct rather than fast rows
   • When you hold a level for three blocks, do one more block focusing only on incongruent rows
   • Try the flanker drill in the Attention subject lessons as a warm-up
3. Week 3
   • Four sessions; alternate Arrows with Pulse so both control skills get fresh practice
   • From level 8, practise spotting the one level arrow among the tilted ones
   • Compare your cost with week 1, but judge progress by accuracy at your level

Check yourself

5 practice questions

1. A row shows five arrows. The centre one points right and the other four point left. What should you press?
2. At level 5 the row is on screen for 600 ms and the deadline is 1,400 ms. You catch the direction just as the row disappears, at about 650 ms. What is the best move?
3. Your correct congruent rows had a median of 450 ms and your correct incongruent rows a median of 560 ms. What is your interference cost?

and 2 more

Create a free account to answerSign in

Sources

1. Eriksen, B. A., & Eriksen, C. W. (1974). Effects of noise letters upon the identification of a target letter in a nonsearch task. Perception & Psychophysics, 16, 143–149. doi.org/10.3758/BF03203267
2. 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. doi.org/10.1162/089892902317361886
3. Hedge, C., Powell, G., & Sumner, P. (2018). The reliability paradox: Why robust cognitive tasks do not produce reliable individual differences. Behavior Research Methods, 50(3), 1166–1186. doi.org/10.3758/s13428-017-0935-1
4. Diamond, A. (2013). Executive functions. Annual Review of Psychology, 64, 135–168. doi.org/10.1146/annurev-psych-113011-143750
5. Arctic Shores help centre, article on the Direction level. help.arcticshores.com
6. Harver. Gamified assessments. harver.com/gamified-assessments
7. Wilson, C., et al. (2021). Building and auditing fair algorithms: A case study in candidate screening. Proceedings of ACM FAccT 2021. doi.org/10.1145/3442188.3445928
8. Psychometric Success. Pymetrics games (candidate-reported descriptions). psychometric-success.com

Put it into practice: ArrowsSign in to try it