The Gorilla Detection Exercises at Dawn – A Theory of Radiology Nightfloat

As a rite of passage as well as part of the regular work schedules of a radiologist, a resident trainee must take on the role of independent interpretation for exams that come into the hospital at night.  I happen to work at a place where attending backup is readily available by phone, but an attending radiologist is not in-house at night.  This provides an abundance of learning opportunities.

After finishing one week of radiology night duties as one of two trainees, I’ve begun to think how the progression of the night always seem to follow some pattern, and what that means for a radiologist trainee on call.


First, it’s probably useful to introduce the concept of a pareto-efficient curve. The curve explains the relationship between two desirable but partially mutually exclusive qualities.  For example, a radiologist wants to be very fast at interpreting studies.  A radiologist also wants to provide very high quality interpretations.  Alas, we cannot do both at the maximal capacity.  One might imagine the relationship between the two to look like this:


Standard pareto-efficiency curve

The curve is only meaningful if you’re actually working at full capacity.  If you’re just chilling out, then you would be somewhere inside the quarter-circle and can improve on both dimensions by moving towards the curve.

For example, when the nightfloat shift starts, there are usually other people around, and the responsibility for the call resident is minimal for the first 30 minutes, and so we might be working at less-than-full capacity:

The night is just starting.


Then, as the evening call attending and other fellows leave, the workload begins to build up.  It’s not really a big deal because we can easily manage a certain amount of work at this level.  Usually this happens around 10PM.

Within the limitation of keeping turnaround time reasonable, usually it makes sense to optimize our minds to deliver the best detection / diagnostic quality interpretation.  This is where most of us work to pick out minor findings that may or may not be clinically important but ultimately make us feel useful:

Other radiologists start to leave, so it’s time to ramp up the effort.

Between 10 and 10:30PM, all the radiologist leave and the night team – resident and fellow – are officially responsible for everything at all the acute care units in the health system.

It usually just means working harder, and at some point we keep up by working at our maximal capacity, hitting the pareto-efficient curve:

The workload builds, and we work harder to keep up.

The workload builds, and the team works harder to keep up.

Unfortunately the radiology workload doesn’t scale with our individual work capacities, and this is where things get interesting.


As the cases keep coming, the phone is really starting to ring as the health system switches to the night teams who may inherit patients they know less about than the day team does.  In any case, answering phone call becomes one of the major responsibilities of the night resident.


Because the trainee is already working at the maximal capacity, the only directions to move are now restricted by the pareto-efficient curve.

And really, there is only one direction you can move in this situation.

Radiologists must balance between delivering quality interpretation and timely turnaround because a lapse in either is a disservice to the patient.

This is where sometimes we decide that the priority for detection must be given to the clinically important findings.  This is where grammatical errors and typos start to show up, as the trainee tries hard to ensure clinically significant diagnoses are made but stops thinking about run-on sentences, 2 mm liver cysts, and the use of “stable” vs “unchanged.”

As shown in the graph above, this phenomenon is not necessarily a good or bad thing, but simply something that happens in the name of better patient care.

The Gorilla Detection Exercise

It is worth taking a little time to explain what this means.  Many people have already seen the gorilla illusion by Dan Simons and Chris Chabris.  (If you’ve not seen the video before and want to give it a whirl, try it here – even though you already know the surprise).

In brief, the gorilla illusion highlights a phenomenon called inattentional blindness.  It shows that when an observer is focused on identifying specific findings on a video, he/she can miss something as egregious as a gorilla walking through a crowd.

In radiology, we have all had experiences where an unexpected finding that would’ve otherwise been obvious somehow becomes invisible as we focus on other findings.  In the literature it is referred to as satisfaction of search.

Radiology nightfloat is a grand exercise in gorilla detection.  Clinically insignificant findings may not necessarily be as important for the on-call radiologist to identify, but the danger comes when we risk diverting so much attention on speed that we miss major findings.

On the pareto-efficient curve, one might call this the Gorilla Detection Threshold, the arbitrary line past which major “gorilla” findings start to get missed. In the context of my institution, “the zone of Major Changes” might be also appropriate.


Between the case workload and GDT, there is only a small room of flexibility for the radiology nightfloat team to deliver optimal care.

The Choice

Everything that has happened so far in the schematic is observational – as a nightfloat resident you don’t have a choice but to follow a minor variation of this pattern.  We all end up somewhere in the vicinity slightly east of the GDT and north of (or perhaps buried inside of) the red zone around 2AM.

No one can stay at 100% capacity all evening – the belief to the contrary is itself an misconception not too different from the memory illusion.  As we fatigue through the night, we can no longer stay on the pareto-efficient curve.  However, we may still have the ability to choose which direction to fall off.

Most people would agree that we should still try hard to stay clear of the gorilla detection threshold.  But sometimes it is inevitable that we will move so dangerously close to it.  After all, you can see the worklist build up, and you feel the pressure of more phone calls due to delayed turnaround.  But we can’t see our own inattentional blindness to important findings.


The red tides are high. The gorilla is roaring. You’re exhausted and can’t work at 100%.  Which way will you go?

In The End

Ultimately the greatest learning arises from being able to triage between the most important and less important events and needs transpiring throughout the night. A radiology resident’s attention span is the most valuable resources at night and must be spent with care.  More importantly, this resource can be partially renewed with periodic short breaks, snacks, and even just having another human being in the same room with whom to discuss cases.

At the end of the day, it will most certainly turn out to be one of the most important educational – or at least most memorable – experiences in my career.  After all, even senior faculty members still tell stories of their own on-call experience as rookie residents on independent night shift.


Howard Chen on GithubHoward Chen on LinkedinHoward Chen on Wordpress
Howard Chen
Vice Chair for Artificial Intelligence at Cleveland Clinic Diagnostics Institute
Howard is passionate about making diagnostic tests more accurate, expedient, and affordable through disciplined implementation of advanced technology. He previously served as Chief Informatics Officer for Imaging, where he led teams deploying and unifying radiology applications and AI in a multi-state, multi-hospital environment. Blog opinions are his own and in no way reflect those of the employer.

3 responses to “The Gorilla Detection Exercises at Dawn – A Theory of Radiology Nightfloat

  1. Pingback: If You Must Multitask, Do It Well » Figure Stuff Out

Leave a Reply