Singularity

What you see

Each Singularity trial is one search field: a square grid of 3 × 3 up to 6 × 6 tiles, with one item on each tile. Every item is alike except one. Your job is to find the odd one and mark it.

Every item is a small caret (an upward chevron) with a few dots under it. It carries four features:

• Shade: how dark the ink is. This is lightness, never colour, so colour-vision differences do not matter.
• Angle: the direction the caret points, in steps of 45 degrees (eight possible directions).
• Size: how long the caret's arms are.
• Dots: how many dots sit under the caret, from one to six.

Above the field, the Time pill counts down. Click a tile to mark it; click the same tile again to unmark it. With a keyboard, Tab into the field, move with the arrow keys and press Space to mark. When you are happy, check your answer with Enter or the Check answer button.

Three details of the timing are worth knowing exactly:

1. The Time pill stops at your first mark and never restarts, even if you unmark.
2. If the pill runs out, the field is not locked: mark your best guess. Nothing is lost except time.
3. The time that counts for your score runs from the moment the field appears until you check your answer, not until you mark.

What changes as levels rise

Levels 1 to 7 are single-feature fields: every item but one is identical, and the odd item differs in exactly one feature. The fields grow from 9 to 36 items and the differences shrink until the odd item might be one dot or one 45-degree turn away from the rest.

Levels 8 to 10 are combination fields, and they work differently. The field is made of two families of items. Suppose the two features involved are angle and dots: half the field points right with two dots, and the other half points up with four dots. The odd item points right and has four dots. Every feature it has is shared with half the field. Only the pairing is unique.

At levels 9 and 10, when shade is not one of the two features involved, the distractors also get a slight random flicker in shade (up to 4 or 5 steps on the 100-step scale). The flicker never makes an item the odd one; it is noise designed to distract you.

What it measures

Singularity measures selective attention in visual search: picking one target out of a crowd of distractors. It feeds the Selective attention facet of our Attention dimension, alongside Target Practice, and carries 20 per cent of the dimension's weight.

The task descends from the visual search paradigm, and in particular from Anne Treisman and Garry Gelade's feature integration theory. Their key observation was that a target defined by one unique feature, if it is different enough from its neighbours, seems to pop out: it is found about as fast in a big display as in a small one. A target defined only by a combination of features does not pop out; it needs attention to move from item to item, so each extra item adds time.

Researchers measure this as a search slope: extra milliseconds per item on screen. Jeremy Wolfe, whose Guided Search model is the main modern account, describes pop-out searches as having slopes near, but typically a little above, 0 ms per item, while a hard search such as finding a letter T among letter Ls runs at roughly 20 to 50 ms per item. Wolfe also notes that feature integration theory "was not quite correct": combination searches can be guided by the features you know the target has, which is exactly the trick the method below uses.

Singularity also depends on perceptual precision. At level 7 the odd item may differ by a single dot or a 45-degree turn, and at every level the size of the difference shrinks. Search speed only helps if your eye can register the difference when it lands on it.

How it is scored here

Each field is scored right or wrong. You marked the odd item, or you did not. There is no partial credit for a nearby tile.

Speed counts too. Your time on a field runs from the moment it appears until you select Check answer, and it feeds two figures.

In your game score, time adjusts the value of a correct answer. The app compares your time with the time expected at that field's difficulty: 2.5 seconds at difficulty 1, plus a quarter of a second for every step up the 10-point difficulty scale, so 4.75 seconds at the top. Accuracy is weighted three times as heavily as speed:

• Finishing a correct field in half the expected time raises its value by about 19 per cent.
• Taking twice the expected time lowers it by about 16 per cent.
• The effect is capped at three times faster or slower than expected.
• A wrong answer scores zero however fast it was. So does a skipped field.
• The first field of a session, and any time under 0.15 seconds or over 60 seconds, counts on accuracy alone.

The difficulty of each field is measured from the field itself: how many items it holds (the biggest factor), whether it is a combination field, how much shade flicker there is and how tight the timer is. Your game score is built from difficulty earned per field. On our scale, 50 means you are performing at what the difficulty you work at predicts.

The game rating out of 1000 on your session report is 50% accuracy, 30% level reached and 20% speed. Speed there is the median time of your correct fields against the expected time for each field's level (2.5 seconds at level 1 up to 4.75 seconds at level 10). Matching the expected time earns half the speed share, 100 points; a third of the expected time or less earns all 200; three times or more earns none. With fewer than three correct fields in the session, speed is left out and the rating is 62.5% accuracy and 37.5% level.

Speed never changes your level: only right and wrong do.

How the level moves

In adaptive practice, every correct field moves you up one level and every wrong or skipped field moves you down one level. So you will oscillate around the level where you are right about half the time. That is the point of the design: your working level tracks your real ceiling.

Where you will meet it

Odd-one-out visual search appears in at least one major game-based hiring assessment, and the underlying skill of spotting a small difference quickly is tested in several ordinary formats. We have no link with any of the vendors named; these are the same family of task, not our game.

• HireVue's game-based assessment includes a game also called Singularity. A 2023 validation study, written with a HireVue author, describes it as an attention task in which you "identify a unique shape amongst increasing numbers of distractor shapes, and with a decreasing number of unique features of the target shape". The same study reports that a model combining HireVue's cognitive games correlated about .5 with traditional cognitive ability tests. HireVue says each of its games takes about three minutes.
• SHL's Checking test is "designed to measure your ability to compare information quickly and accurately": the same fast, precise comparison, applied to strings of letters and numbers.
• Criteria's mobile game-based test, GAME, lists attention to detail among what it measures, in about five to six minutes of mini-games.

The method

1. Take in the whole field for a moment

   Before you look at any single tile, let your eyes rest near the centre and take in the whole grid. On single-feature fields with a large difference, the odd item often simply appears. This costs under a second and saves many.

2. Name the standard item

   If nothing jumps out, look at two or three tiles and name what normal looks like: for example, "dark, pointing right, small, three dots". You are now searching for anything that breaks that description, which is much faster than comparing tiles with each other.

3. Sweep in bands, checking the subtle features

   Sweep the field in bands of two rows. Angle and size differences usually catch the eye on a sweep; a single extra dot or a small shade gap usually does not. On your sweep, check the dots and the darkness of each band on purpose.

4. On combination fields, split the field in two

   If you can see two kinds of item, you are in a combination field. Pick one feature of one family, say "the right-pointing ones", and search only those items for the one whose other feature belongs to the other family. You have just halved the field.

5. Confirm once, then submit

   When you find a candidate, compare it with one neighbour to confirm the difference is real, especially at levels 9 and 10 where shade flickers. Then submit. Remember that scoring time runs until you submit, not until you mark.

6. If the timer runs out, keep searching calmly

   Nothing locks when the Time pill reaches zero. A correct answer a second late is worth far more than a wrong answer on time, because a wrong answer scores zero.

Worked example 1: a single-feature field

This is a level 6 style field: 25 items, one of them different in one feature. Time yourself.

Show the solution

The odd item is in row 4, column 2: it has four dots where every other item has three. Angle, size and shade are identical everywhere.

This is the hardest kind of single-feature difference in the game. Angle or size differences often pop out; a one-dot difference usually has to be checked. The method's third step is why you catch it: you sweep with the dots in mind rather than hoping the item jumps out. If you missed it, your report's "how close your misses were" figure will show that your wrong choice was a near-identical item, which is the signature of an attention-to-detail miss rather than a random guess.

Worked example 2: a combination field

This is a level 8 style field. No single feature is unique. Find the item whose combination is.

Show the solution

First, name the two families: carets pointing right with two dots, and carets pointing up with four dots. Then split the field. Search only the right-pointing carets for one with four dots.

The odd item is in row 2, column 3: it points right (like one family) and has four dots (like the other). Every right-pointing caret except this one has two dots, and every four-dot item except this one points up.

Scanning for "the odd one" without splitting is slow here, because every item looks like half of the field. Splitting turns a search of 25 items into a search of about 13.

Tips and traps

Levelling up

Levels 1 to 4 train pop-out search. Large differences, small fields: let the odd item find you.

Levels 5 to 7 train precise search. Fields grow to 36 items and differences shrink to one dot, one size step or one 45-degree turn. The sweep and the "name the standard" step start to matter here.

Levels 8 to 10 train guided combination search. The key skill is splitting the field into its two families and searching one family for the wrong partner feature, while ignoring shade flicker. The timer drops to 3.5 seconds at level 10, but remember that the timer is advisory and, in your game score, accuracy is worth three times as much as speed.

1. Week 1
   • Three sessions of 12 to 15 fields, adaptive on
   • On every field, take in the whole grid for a moment before searching
   • Note which feature your misses involved: dots, shade, angle or size
2. Week 2
   • Four sessions; practise naming the standard item on every field
   • Use the two-row sweep on every 5 × 5 and 6 × 6 field
   • Pair with a short Target Practice session for moving selective attention
3. Week 3
   • Four sessions aiming to hold level 8 or above
   • On combination fields, say the two families to yourself before searching
   • Compare your search slope and combination cost with week 1

Check yourself

5 practice questions

1. In a single-feature field, every item is identical except one. Which difference is usually hardest to spot at a glance?
2. At level 8 a field has carets pointing right with two dots and carets pointing up with four dots. What does the odd item look like?
3. The Time pill runs out before you have found the odd item. What happens?

and 2 more

Create a free account to answerSign in

Sources

1. Treisman, A. M., & Gelade, G. (1980). A feature-integration theory of attention. Cognitive Psychology, 12(1), 97–136. doi.org/10.1016/0010-0285(80)90005-5
2. Wolfe, J. M. (2021). Guided Search 6.0: An updated model of visual search. Psychonomic Bulletin & Review, 28, 1060–1092. doi.org/10.3758/s13423-020-01859-9
3. Wolfe, J. M., & Horowitz, T. S. (2017). Five factors that guide attention in visual search. Nature Human Behaviour, 1, 0058. doi.org/10.1038/s41562-017-0058
4. Leutner, F., Codreanu, S.-C., Brink, S., & Bitsakis, T. (2023). Game based assessments of cognitive ability in recruitment: Validity, fairness and test-taking experience. Frontiers in Psychology, 13, 942662. pmc.ncbi.nlm.nih.gov/articles/PMC9891208
5. HireVue. How to prepare for your HireVue assessment. hirevue.com
6. SHL Direct. Checking test example questions. shl.com
7. Criteria Corp. General Aptitude Mobile Evaluation Game (GAME). criteriacorp.com

Put it into practice: SingularitySign in to try it