Torque measures friction rather than tension
Where the effort at the wrench actually goes.
Torque measures friction rather than tension because most of the effort applied at a wrench is spent overcoming friction under the bolt head and in the thread, and only a modest remaining share actually goes into stretching the bolt, which means the one number a torque wrench actually reports is dominated by a quantity that has almost nothing to do with the clamping force a joint actually needs.
The quantity that actually matters in a bolted joint, the bolt's own tension, cannot be seen or felt from outside the joint at all, so everybody reaches instead for the one quantity that can be measured directly at the wrench, the twisting effort needed to keep turning the fastener, and treats that visible, controllable number as a stand-in for the invisible one it was never a precise measure of in the first place.
Introduction and overview
Turning a bolt to a specified torque feels like a direct, reliable way of controlling how tight the joint ends up, since the wrench reads a clear number and stopping at that number feels like the entire job has been done correctly. What that number actually represents is the total resistance the wrench had to overcome to keep turning the fastener, and only a modest fraction of that resistance comes from the bolt actually stretching, storing the elastic tension a joint depends on, while the majority is consumed simply sliding the bolt head or nut across its bearing surface and sliding the thread flanks past each other, both of which are governed by friction rather than by the bolt's own tension at all.
This is the practical consequence of the spring model this era has been building toward, since the bolt's stretch is the one thing a torque wrench is actually trying to control, and friction is simply standing in the way as an unwanted, variable tax on every single turn.
The shoelace-through-eyelets comparison
Pulling a shoelace tight through several eyelets requires effort that is spent partly tensioning the lace across the shoe and partly overcoming the friction of the lace dragging through each individual eyelet along the way, and a lace running through more eyelets, or through eyelets that grip more, needs noticeably more pulling effort to achieve the identical actual snugness across the shoe as a lace running through fewer or smoother eyelets. Judging how snug a shoe actually is purely by how hard the lace was pulled would be a poor method, since the same pulling effort on a stickier lace produces a looser fit than the identical effort on a smoother one, and a torque wrench is making exactly this same mistake when it reports the total twisting effort at the fastener as though that number alone described the tension actually achieved inside the bolt.
Why friction varies so much between one bolt and the next
Friction at a bolt's bearing surface and in its threads depends on a long list of factors that have nothing to do with the bolt's own tension, whether the surfaces are clean or lightly contaminated, dry or lubricated, freshly plated or slightly corroded, and even a small change in any one of these can shift how much of the applied torque gets consumed by friction rather than converted into stretch. Two bolts from the identical batch, torqued to the identical reading on the identical wrench, can end up with meaningfully different actual tensions purely because one had a trace more oil on its threads than the other, a variation the torque reading itself has no way of detecting or correcting for.
The person doing the tightening rarely has any direct way of knowing which side of that variation a particular bolt has landed on, since the wrench reports the same clean, confident number regardless of whether the friction underneath happened to be typical, unusually high, or unusually low that day, which is exactly what makes this failure mode so easy to miss in practice, nothing about the tightening process itself signals that anything has gone wrong.
The one number worth remembering
In an ordinary dry, uncoated bolted joint, roughly ninety percent of the torque applied at the wrench is typically consumed overcoming friction at the bolt head or nut face and within the thread itself, leaving only around a tenth of the applied effort actually converting into the bolt's own tension. That proportion is not a fixed law of nature, it shifts with lubrication, surface finish and thread condition, which is exactly the problem, since a torque reading calibrated for one friction condition can under- or over-tension a bolt significantly the moment that friction condition changes even slightly from what was assumed.
What follows from this
Treating a torque specification as an approximate target rather than a precise one changes how tightening actually gets done in practice, controlling and standardising the friction condition as carefully as the torque value itself, using a specified lubricant, cleaning threads before assembly, replacing rather than reusing fasteners whose surface condition is uncertain, since none of that careful torque calibration means anything if the friction it assumes turns out to be wrong on the day. It also explains why critical joints so often specify not just a torque value but the exact lubricant and surface condition that torque value was calculated against, since the torque number alone was never a complete instruction on its own.
What this does not explain
Torque control remains genuinely useful for the great majority of ordinary joints, where a rough, repeatable tension is good enough and the cost of a more precise method would outweigh the benefit, and this article's point is not that torque is a bad method in general, it is that torque's real limitation, its sensitivity to friction rather than any flaw in the wrench itself, is worth understanding before trusting it on a joint where getting tension genuinely right actually matters. The next article in this set picks up exactly this gap, covering the methods that exist specifically for the joints where torque's friction sensitivity is not an acceptable risk to carry.