What “Curious” Looks Like

How the study worked

Curious is one of those words that shows up in a brand brief and a creative review and means something different to everyone in the room. The dictionary is no help. What people need is the picture. So we asked for it.

We put 24 designs on two boards. Logos, posters, packaging, patterns, photographs and illustration, mixed together with no grouping and no order. Thirty-three people completed the study out of 113 who opened it. Each person got three green dots and three red dots per board. Green meant this feels curious. Red meant this does not. Together they placed 361 dots: 180 on the first board and 181 on the second.

Each image is scored by the share of its own dots that were green. We sorted at a 54% threshold, slightly under the platform's 60% default, because the study has a wide middle and the tightest cut-off that still separates the piles is the useful one. An image is resonant when at least 54% of its dots are green, resistant when at least 54% are red, and contested when neither side gets there.

At that bar the 24 images fall into three groups. Eight cleared it as resonant. Thirteen crossed it as resistant. Three sat within a dot or two of even. For a word nobody can define, the room agreed on most of the pictures.

Figure 3 · Every image, ranked

What won: eight images that are not done with you

Eight images cleared the 54% bar. Three of them contain a person. Five do not. They split cleanly into the two routes.

Resonance array 8 images at 54% green or better

Route one: someone is looking at something you cannot see

The raised eyebrow took 83% green, 15 of 18. It is two eyes on lime, nothing else, and the pupils sit off to one side with a brow lifted over them. That is the facial expression of the word, and the room treated it as such: of the eight people who put their very first dot on it, all eight placed a green one. The man looking up took 76%, 13 of 17. His eyes are raised toward something above the frame and a loop and a sparkle orbit his head. The silhouettes at the counter took 61%, 11 of 18, the softest of the winners. Three people with their backs to us, looking at a glowing wall whose type runs backward. We see them seeing.

In each of these the person's attention is pointed away from the viewer. That is the cue. We are not being looked at; we are watching someone else look, and we want to know at what.

Route two: the image is not finished for you

The spotlight is the top score in the study at 89%, 16 of 18. A beam of light cuts a wedge across LIGHT THE WORLD, and the words are cropped at both edges so you read them by inference. The sign collage matched it at 89% on a smaller crowd, 8 of 9: five or six lettering styles clipped and stacked, with OVER sliding across the middle and something else always peeking out from behind. The symbol grid took 79%, 15 of 19, the largest green crowd on Board 1. Dots, a toggle, stripes, chevrons and three lowercase a's in a tiled layout that reads like a code you almost know. The festival strips took 78%, 14 of 18: a poster made of tilted columns of small labels, more to read than you can take in at a glance. The overprinted A cleared at 58%, 7 of 12, one letter printed in four misregistered inks so the layers separate as you look.

None of these contain a person. What they share is that the eye has work left to do. Something is cut off, hidden behind something else, or arranged as a system with a rule you have not found yet.

Loud, but not because it is loud

All eight winners are saturated and high contrast. So are seven of the thirteen losers, including the yellow B, the purple pills, the stacked B and the big smile. Brightness gets an image noticed. It does not get it called curious. Within the bright pile, the open images won and the closed images lost, every time.

What lost: thirteen finished statements

Thirteen images crossed the 54% bar on the red side. Most of them are good work. Several are the kind of polished identity piece a studio puts on the front page. They lost for one reason: they are finished.

The first way to miss is to be a mark. The yellow B at 82% red and the stacked B at 80% red are logos. Bold, complete, closed. A logo's whole job is to resolve, and this room filed resolution under not curious.

The second way to miss is to repeat. The purple pills, the orange tile wall and the striped flame are patterns. Once you have seen one module you have seen the image, and the eye leaves.

The third way to miss is to be pretty. The twice forty numerals, the x-ray tulip and the gradient letters are delicate, translucent and calm. Beautiful, and settled. The numerals took 10 of 11 dots red, the heaviest rejection in the study.

The fourth way to miss is to have already decided how you feel. The big smile at 80% red and the sticker faces at 56% red are faces aimed at the viewer, expressing delight. An emotion that has arrived is the opposite of one that is still forming.

Resistance array 13 images at 54% red or better

The big smile is the most instructive rejection. It drew 20 dots, the second-largest crowd on its board, and 16 were red. People stopped, looked, and said no. It is a face, it has an expression, it has energy, and it fails the one test that matters: it is looking at us, and it has nothing left to find out.

The sliced letters are the surprising loser. Fragmented, cropped, aggressive, and 80% red. Cutting an image up is not the same as leaving it open. The slashes destroy the letters rather than hide them, and what comes through is hostility, not inquiry. The paper bird poster lost in the opposite register. A full cast list, a title, a finished origami object, every fact present and calm. Dense information that is all there is just reading, and reading is not curiosity.

The big discovery: the open loop

Lay the eight winners next to the thirteen losers and the split is not about color. The yellow B is as loud as the spotlight. The purple pills are as bright as the symbol grid. Loud lost as often as it won.

The split is about whether the image is finished. Every winner is open in one of two ways. Either a person in the frame is paying attention to something you cannot see, or the image itself has been cut off, layered or coded so that you have to finish it. Every loser arrives complete: a logo, a repeat pattern, a pretty object, or a face that has already decided how it feels and is showing you.

Figure 4 · Open or closed, with a face or without one

No personSubjectA person

• Open
• A person in it

83% green The raised eyebrow A face that is mid-question. Pupils off to the side, one brow up. 15 of 18 dots green, and every one of its first-position dots was green.

• Closed
• A person in it

80% red The big smile A face that has already decided. Delight aimed straight at the camera. 16 of 20 dots red, the biggest rejection on Board 1.

• Open
• No person

89% green The spotlight No person, but a beam searching across words that are cut off at the edges. 16 of 18 dots green, the top score in the study.

• Closed
• No person

82% red The yellow B No person, and nothing left to resolve. A bold, complete mark. 14 of 17 dots red, and the fastest no on Board 1.

OpenStructureClosed

The word's literal symbol proves the point from the other direction. The question-mark stickers, a logo whose smiley face is drawn from a question mark, landed at 53% green, one dot short of the line. The room did not respond to a symbol of asking. It responded to the act of looking, and to things it had to look at longer.

The twin tests

The cleanest pair in the study is two faces with strong expressions and opposite scores.

Both are faces. Both are expressive. Both are high contrast and graphic. The only difference is tense. An expression that is still forming reads as curious. An expression that has arrived reads as the opposite. Gaze direction carries it: away from us wins, at us loses.

The second pair is two pieces of transparent, overlapping typography.

Same device, opposite verdict: 58% green against 91% red. The overprinted A looks like a mistake you want to inspect. The twice forty numerals look like a decision someone already made. Overlap reads as curious only when it looks unresolved.

The three that split the room

Only three images landed between 46% and 54% green, and all three sit at 52% or 53%. They are not a wide middle. They are three near-ties, and each one is a half-match for the rule.

Contested 3 images between 46% and 54% green

The profile in the dark drew the largest crowd on either board, 21 dots, and split them 11 to 10. The man is looking out of the frame, which is route one, but the image is a polished tech advertisement with copy about delivery routes sitting next to his face. Half the room saw the gaze. Half saw the ad. At the first-dot level it is just as split, four green and three red.

The glowing figure is a human silhouette with no face at all, dissolving into light. There is a person, but nothing to tell you where their attention is. Nine green, eight red. The question-mark stickers are a logo built from the literal symbol of the word, and the room treated them the way it treated the other logos, almost. Eight green, seven red, one dot short of the line. A symbol of asking is not the same as a picture of someone asking.

Where the first dot went

The platform records the order each person placed their dots, so the first dot on each board is a clean read of the gut. Two things stand out.

People look for the yes first. Twenty-three of the 31 people who placed dots on Board 1 opened with a green dot, and 26 of 31 on Board 2. The red dots came after, as a correction. And the first green dot went to the same images the final verdict favored: the symbol grid and the festival strips on Board 1, the raised eyebrow, the spotlight and the man looking up on Board 2. The first red went to the logos. The yellow B took six first-red dots, more than any other image on its board.

Figure 5 · Where the first dot went

The gut and the considered verdict agree. That matters for anyone using the result: the open-loop reading is not something people reasoned their way to over a minute. It is what they saw in the first few seconds.

Attention is not endorsement

The profile in the dark drew 21 dots and split them down the middle. The big smile drew 20 and went 16 to 4 red. The sign collage drew 9 and went 8 to 1 green.

Plot crowd size against verdict and there is no relationship. Big crowds sit at the top, the bottom and the middle of the chart. The jazz it up study found the same thing: an image that stops the room has only stopped it. What the room decides next is a separate measurement.

Figure 6 · Crowd size against verdict

What separates green from red

Across both boards, the same six differences separate the green pile from the red one.

The last row is the one to notice. Color does not sort the piles. Both saturated and soft images were rejected, and every winner is saturated only because every open image in this set happened to be. The structure rows do all the work.

What to do with this

The next time a brief says curious, the word comes with a specification instead of a mood. It is one of the more useful entries in the field guide to vague creative feedback: give the viewer something to look for, and do not finish the thought for them.

One warning before you run with this

This study measured one word. The rejected designs might win confident, refined or joyful in a heartbeat, with the same 33 people. Use the result as a translation: when the word shows up in a brief, you now know what it is asking for. Do not stretch it into a general verdict on the work.

Methodology and confidence

33 completions across 113 opens, 361 dot placements over 24 images on two boards: 180 on Board 1 and 181 on Board 2. Each person had three green and three red dots per board, so green and red totals are roughly equal on each board by design, and salience is zero-sum: a dot placed on one image is a dot not placed on another. Two people completed the study without placing dots on one of the boards; 31 people placed dots on each board. One green dot on Board 1 was dropped above the top edge and is excluded, which is why that board counts 180 of 181 placements.

Images were sorted at a 54% threshold. The platform's results screen defaults to 60%; at 60% the resonance array loses the overprinted A, the resistance array loses the purple pills and the sticker faces, and the contested group grows to six. No image changes sides. With 33 people this is a directional study with a firm center. Per-image crowds run from 9 to 21 dots, so single scores between 54% and 70% carry real uncertainty; the sign collage cleared at 89% on nine dots and should be read as a tendency. The findings that bear weight are the ones that repeat. The open-loop pattern holds across eight separate winners and thirteen separate losers, the gaze rule holds for every image with a person in it, and five rejections at 80% or above on ten or more dots are the kind of agreement a sample this size rarely produces by chance.

Three passes would sharpen it. Run the same boards against confident, where the prediction is a near inversion. Anchor on the raised eyebrow to see whether the people who chose it also chose the cropped type, which would confirm that the two routes are one instinct. And add a second face that looks away but smiles, to separate gaze direction from expression.

Frequently asked questions