How to Analyze a Signature

The question with an answer: did the same hand write both?

This is forensic document examination, and it is the version of signature analysis that holds up. The examiner is not asked what kind of person signed the cheque. They are asked whether the signature on the cheque and the signatures on file were produced by the same hand.

That is a comparison, and comparisons can be checked. You can hand someone a stack of genuine and forged samples, collect their answers, and count how many they got wrong. People have done exactly that, and the numbers are further down.

The useful thing about this framing is that it survives contact with the obvious objection. Nobody signs their name identically twice. Your own signature varies with the pen, the surface, whether you are standing at a counter, and how much of a hurry you are in. Examiners expect that variation. What they are looking for is whether the variation falls inside one person's normal range.

Why a forgery looks wrong before you can say why

Here is the mechanism, and it is the single most useful idea on this page.

A genuine signature is written. A forgery is drawn.

Your own signature is one of the few pieces of handwriting you have automated. You do not assemble it letter by letter. You run a motor program, roughly the way you do not think about the individual movements involved in tying a shoe. It comes out fast, and speed leaves physical evidence on the paper.

A forger is doing something else entirely. They are copying a shape they have seen, which means watching the model, checking their progress, and correcting as they go. That is drawing. It is slow, and slowness leaves its own evidence.

Three consequences follow, and they are what the whole checklist below is built on.

First, speed tapers the ends. When you sign fast, the pen is already moving before it touches the paper and still moving when it leaves. The stroke fades in and fades out. A slow, deliberate start puts the pen down first and then moves, which deposits a small blob of ink and leaves a blunt end.

Second, speed smooths the curves. Fast curved movements come out clean because the hand is running a stored pattern rather than steering. Slow ones wobble, because a hand being consciously guided along an unfamiliar path produces small tremors. Loops and curves are where this shows first.

Third, speed does not permit second thoughts. Genuine signatures rarely have pen lifts in the middle of a stroke, because there is no reason to stop. Forgers stop, because stopping is when you look back at the model.

The eight things an examiner actually looks at

In rough order of how much weight they carry.

1. Line quality. The fluency of the stroke itself. Smooth and confident, or hesitant and broken. This is the first thing looked at and often the last word.
2. Starts and stops. Tapered, or blunt and ink-heavy. See above for why this matters more than it sounds like it should.
3. Pen lifts. Not whether they exist, but where. Everyone lifts the pen somewhere. The question is whether they lift in the same places as the reference samples.
4. Tremor. Small involuntary wobble, concentrated in curves and loops. Worth noting that genuine tremor also comes from age, illness, cold hands, and a bad writing surface, which is exactly why this is item four and not item one.
5. Patching. Places where the pen went back over a stroke to fix a shape. A person signing their own name has no reason to retouch it. A forger unhappy with a letter does.
6. Pressure pattern. Genuine signing varies pressure through the stroke, heavier on downstrokes, lighter on the way up. Careful copying tends toward uniform pressure throughout.
7. Proportion, not size. Absolute size is nearly worthless: it changes with the space available, which is why your signature is smaller on a card reader than on a contract. The ratios between letter heights are much more stable.
8. The bits nobody thinks to copy. Stroke order, the angle a loop is entered at, where an i-dot lands relative to the stem, how a terminal stroke exits. A forger concentrates on overall resemblance. These details are invisible to that effort, which is precisely what makes them useful.

Notice that none of these are about what the signature looks like. They are all about how it was produced. A forgery that looks perfect and a forgery that looks clumsy fail in the same places.

Does the training actually matter?

It does, and unusually for this subject, somebody measured it.

In proficiency testing comparing professional document examiners with non-examiners on signature comparison, the examiners declared a genuine signature to be a forgery in 0.49% of cases. The non-examiners did it in 6.47% of cases, about thirteen times as often. Going the other way, examiners accepted a forgery as genuine 7.05% of the time against 26.1% for non-examiners.

Sit with the second pair of numbers for a moment. Even trained examiners miss roughly one forgery in fourteen. Untrained people miss one in four. If your plan for detecting a forged signature is to look at it carefully and trust your judgment, that last figure is your realistic accuracy.

(Related work has found professionals outperforming control groups on writer identification at p < 0.001, so the effect is not a fluke of one study. The field is not without critics: the 2009 US National Academy of Sciences review of forensic science put handwriting comparison among the disciplines whose scientific foundations it considered underdeveloped. Both things are true. The training helps a great deal, and the discipline is less standardised than courtroom confidence implies.)

Sources: Kam et al., signature authentication by forensic document examiners and accuracy and reliability of forensic handwriting comparisons, PNAS.

The question without an answer: what kind of person is this?

The other thing called signature analysis is graphology, which reads personality from handwriting. Big signature means confidence, rightward slant means sociability, and so on.

It does not survive testing. Reviews of the literature have repeatedly found practitioners performing no better than chance at predicting personality or job performance, and the readings feel accurate mostly because they are broad enough to fit almost anybody. We went through the specific claims, trait by trait, in what your signature says about you, including the ones that are simply fun rather than true.

The short version: interpretation of character is the part that fails, and comparison of hands is the part that works. If someone offers to do both from the same sample, you are getting one service and one performance.

The one measurement that predicted something real

With that said, there is a genuine finding in this area, and it is a strange one.

Ham, Seybert and Wang measured the signature size of roughly 600 chief executives using signatures pulled from S&P 500 annual filings, and used it as a proxy for narcissism. Firms with larger-signing CEOs showed more overinvestment in research and acquisitions, and lower profitability and operating cash flow. The CEOs in question were paid more, both in absolute terms and relative to their peers.

It is worth being precise about what this does and does not show. Nobody read a personality off a page. Signature size was used as a cheap, consistent measure across hundreds of companies, and that measure correlated with later financial behaviour. That is a very different claim from a graphologist telling you that your capitals reveal ambition, and it is a much better-supported one.

The paper is Narcissism is a bad sign: CEO signature size, investment, and performance, Review of Accounting Studies, 2018.

What you cannot conclude

A few limits worth stating plainly, because this is the part usually left out.

You cannot reach a firm conclusion from one reference sample. Examiners want several, written close in time to the document being questioned, because a single sample tells you nothing about the person's normal range of variation.

You cannot do much with a photocopy or a low-resolution scan. Pressure evidence is gone, fine tremor is below the resolution, and compression can both hide real patching and invent artefacts that look like it.

And a signature that fails several of the eight checks is not proof of anything. It is a reason to get an original document in front of someone qualified. The gap between "this looks wrong to me" and "this is a forgery" is where the 7.05% lives.

Frequently asked questions