
I keep a spreadsheet.
I'm going to say that upfront because if I build to it, it'll sound more insane than it is. If I lead with it, we can get the judgment out of the way and move on to the interesting part.
I keep a spreadsheet about eye contact.
It started because a coworker at a job I had before the truck โ warehouse work, short-lived, don't ask โ told me I "stared." Not in a threatening way, she said. More like I was "looking through" her. She said it made her uncomfortable. She was nice about it. It still plays on a loop in my brain twenty years later.
The problem wasn't that I was staring. The problem was that I didn't know the rules. How long is too long? How short is too short? Where do you look when you're not looking at someone's eyes โ their mouth? Their forehead? The middle distance? Do you look away to the left or the right? Does it matter? Does looking down mean submissive and looking up mean distracted? Who decided this? Is it written down somewhere?
It is not written down somewhere. I checked. There are studies โ social psychology, nonverbal communication, cultural norms โ but nobody has published a practical field guide for "how long to look at a person's eyes in a specific social context before they think you're weird."
So I wrote one.
THE EYE CONTACT EQUATION
After approximately two years of observation, calibration, and social experimentation (I am aware of how this sounds), I developed the following model:
Optimal Eye Contact Duration (OED) varies by:- Relationship distance (stranger โ acquaintance โ friend โ intimate)
- Context (professional โ casual โ emotional)
- Power dynamic (symmetrical โ asymmetrical)
- Cultural baseline (American Midwest, which is my dataset)
The general formula:
`
OED = Base Duration ร Relationship Modifier ร Context Modifier
`
Base Duration: 3-5 seconds per "look." A "look" is one continuous period of eye contact before breaking and returning. Think of it like breathing โ look, break, look, break. The base rhythm.
Relationship Modifiers:
| Relationship | Modifier | Notes | |---|---|---| | Stranger | 0.5x | Brief glances only. More than 2 seconds = threat display | | Acquaintance | 1.0x | Standard. 3-5 seconds per look. | | Friend | 1.2x | Slightly longer looks acceptable. Signal of trust. | | Close friend | 1.5x | Sustained looks okay. This is where real conversation happens. | | Intimate partner | Variable | Rules change entirely. Ask Beth. She'll tell you. |
Context Modifiers:| Context | Modifier | Notes | |---|---|---| | Professional meeting | 0.8x | Less eye contact = more formal. Compensate with nodding. | | Casual conversation | 1.0x | Standard. | | Emotional disclosure | 1.3x | They're being vulnerable. More eye contact = "I'm here." | | Argument | 0.6x | LESS eye contact. Sustained eye contact in conflict = escalation. | | Listening to a story | 0.7x look / 1.5x when they hit the point | Low during setup, high during punchline or emotional peak. |
The Break Pattern:Where you look when you break contact matters more than the contact itself.
- Down and to the side: "I'm thinking about what you said." (Good.)
- Straight down: "I'm ashamed" or "I'm submitting." (Bad unless intended.)
- Up: "I'm remembering/constructing." (Neutral but can read as dismissive.)
- To the side at their eye level: "I'm distracted." (Bad.)
- Past their shoulder: "I want to leave." (Very bad, even if true.)
The optimal break is down-and-slightly-right for 1-2 seconds, then return to eyes. This reads as "thoughtful" across most American social contexts. I've tested this extensively. It's correct.
I know how this reads.
It reads like a robot trying to be human. It reads like someone gamifying basic social interaction. It reads like the opposite of authentic โ a man with a spreadsheet trying to simulate the thing that other people do without thinking.
And โ yes. That's exactly what it is.
But here's the thing I want neurotypical readers to understand, if any are here:
You have this spreadsheet too. You just can't see it.Every neurotypical person has an internalized model for eye contact. You learned it as a toddler. You refined it through childhood. By adulthood, it's so automatic that you'd swear you're "just being natural." But you're not. You're running a model. The model is just compiled โ it runs below conscious awareness, like an operating system kernel. You don't see the code because it works.
My code doesn't compile. It runs in interpreted mode. Every instruction is visible. Every decision is conscious. The spreadsheet isn't a replacement for the natural thing โ it IS the natural thing, just visible instead of hidden.
Some people breathe automatically. Some people use a ventilator. Both are breathing.
FIELD OBSERVATIONS: The Cashier Test: At a truck stop counter, the optimal exchange is: eye contact on approach (1.5 seconds), break during ordering (look at menu/hands), re-establish when they hand you the change (1 second, with a nod). Total eye contact: under 3 seconds. Any more and you've created an intimacy that neither of you wants at 6 AM over a coffee transaction. The Dispatcher Call: Phone calls are the cheat code. No eye contact required. This is 70% of why I prefer calls to in-person. The other 30% is not having to manage my face. The Old Friend Reunion: The hardest scenario. Someone you haven't seen in years expects emotional warmth communicated through sustained eye contact and facial expression. My model says 1.5x modifier. My actual capacity after not seeing someone for three years is approximately 0.3x because I'm too busy processing the changes in their face and recalibrating my entire stored model of who they are. They read this as coldness. It's not coldness. It's buffering. The Doctor's Office: Eye contact with authority figures in medical contexts requires a specific sub-model I haven't fully debugged. The doctor expects enough eye contact to confirm you're lucid and engaged, but not so much that you seem to be challenging their authority. The nurse expects more warmth. The receptionist expects almost none. The phlebotomist who's about to put a needle in your arm expects a specific reassurance-seeking pattern that I have literally never been able to produce. I just look at the wall and they think I'm brave. I'm not brave. I'm running the wrong subroutine.
The spreadsheet lives in a notes app on my phone. I update it when I notice a new pattern or when a social interaction goes wrong in a way that suggests my model needs recalibration.
Last update was three weeks ago: "Sustained eye contact during someone else's good news should be 1.4x, not 1.0x. They want to see you being happy for them. Your face needs to participate."
My face needs to participate. I wrote that in a spreadsheet about how to look at people.
This is my life. It's stranger than most. It works better than you'd think.
And the waitress at this Waffle House has been very patient with me writing on my phone for forty-five minutes, and I've maintained a 1.0x eye contact ratio every time she's refilled my coffee, and I tipped 30% because the spreadsheet can't quantify gratitude but the tip line can.
Next time: Cab Note #002 โ on the different colors of darkness between 2 AM and 4 AM, and how the highway knows the difference even if the clock doesn't.