An (unpolished) essay on Fingerprinting

A 19th-century forensic breakthrough


This is an essay written for the purpose of personal learning and enjoyment. The usual scrutiny reserved for academic publications is not applied to it. Use the content in this essay at your own discretion.

I. Preface

It all starts with The Great Ace Attorney. The great game brought me back to the late Victorian Era, when cutting-edge technology such as steam locomotives, cameras, and music boxes, captured people’s imagination. Among other things, I was quite surprised to learn that fingerprints have only started to be considered in crime investigation at the turn of the century, despite us being obviously aware of their existence – after all, we only need to stare at our fingertips. But then I realized that, in fact, I had little idea how fingerprinting works! Therefore, I am writing this article, first to briefly recount the history of fingerprinting, then to explain how fingerprints are collected and then used to identify a person, and finally discuss other uses of fingerprint technology.

II. History of fingerprinting

The use of fingerprints for authentication has quite a long history. For example, in ancient China, fingerprints and handprints were used as signatures, for example in courts and when entering a contract. So, we see that people long realized that fingerprints were a reliably distinctive feature of a person. So then begs the question: why were fingerprints not used in criminal investigation and prosecution? I can think of a few possible reasons. First, in order to use fingerprints to identify suspects, we need a fingerprint database, and constructing such databases require considerable bureaucratic effort. Second, most people lived in small, close-knit communities, and description of apparent physical features is most likely enough to uniquely identify a person. Finally, and perhaps most crucially, most fingerprints that we leave on surfaces are invisible – hence the name latent prints – and people might just have no clue that fingerprints could be lifted from the crime scene, and the “how” was thus left unanswered. So, it wasn’t until the late 16th century that European academics and doctors started investigating fingerprints. They identified features such as ridges (^ shaped lines), spirals, and loops, and some posited that fingerprints were unique to each individual.

Fast forward to the 19th century. In 1880, Dr. Henry Faulds, a Scottish surgeon then working in Japan, published a paper on the usefulness of fingerprints for identification, and proposed a method to record them with printing ink. When he returned to London a few years later, he couldn’t convince the London police to adopt his idea. At that time, the Bertillon system, developed in 1883 by French police officer and biometrics researcher Alphonse Bertillon [1], was prevalent in Europe and America. Anthropometric measurements such as head length and middle finger length were combined with a mugshot to uniquely identify a person in the police records.

Figure 1 - The Bertillon system.

It was based on solid scientific evidence and proven to be quite accurate and effective, although it had its drawbacks. For one, when the subject was a child (not fully grown) or a woman (long hair), the measurements would be less accurate. Also, measurement required accurate tools and trained staff, which were not readily available in some places. Fingerprints on the other hand are permanent and unambiguous, and all it lacked was a system to classify and sort the fingerprint data, for easy entry and retrieval.

It took less than a decade for such systems to be invented. Henry Faulds, and independently Francis Galton, attempted to devise their own classification systems, but both systems had deficiencies for filing purposes. Sir Edward Henry and his team, then based in India, developed the Henry system for criminal classification. The Argentine system was devised by Juan Vucetich, an Argentine chief police officer. Yet another system, the Conlay system, was employed in certain parts of America. Soon enough, police officials realized the superiority of dactyloscopy (fingerprint identification) over the Bertillon system of arthrometry, and in 1901 following a trial in India, the UK fingerprint bureau was established at Scotland Yard, and fingerprints were gradually accepted in courtrooms as credible evidence.

The Henry system continued to be in use until the 1990s, following the development of automated systems. Nowadays in America, the FBI maintains the nation-wide Integrated Automated Fingerprint Identification System (IAFIS), which uses many more features than the Henry system to classify fingerprints.

III. Uniqueness of fingerprints

Before delving into how fingerprint identification works in practice, let us put the claim that “fingerprints are unique” under scrutiny. After all, if fingerprints were not unique, it would be possible for an innocent person to be accused as a criminal (although the person would have to be at the wrong place at the wrong time AND have the wrong fingerprints, a most unfortunate coincidence indeed).

Figure 2 - Are fingerprints really unique?

The development of fingerprints happens when we are in our mother’s womb. 6-7 weeks in, the foetus starts growing thick pads on his palms and feet. In the beginning, the pads look like poorly fit gloves, but in a few weeks they begin to smooth over our fingers, and the way they fold determines the fingerprint patterns. This process is heavily affected by environmental factors such as pressure of amniotic fluid and position of the foetus, and therefore considered highly random, making it (almost) impossible for individuals to share the same patterns. Even identical twins have different fingerprints!

Despite common belief of the uniqueness of fingerprints, such claim has not been rigorously verified. Moreover, even if the claim were true, fingerprints recorded in the database or lifted from a crime scene are not perfect representations of the prints on the person’s hand, as the former could be incomplete or of limited resolution. Even the same person’s fingerprint impression on different occasions could be slightly different. This is a gaping hole in the credibility forensic science that experts should strive to address. For example, in the Madrid bombing attack on 11 March 2004, a fingerprint was lifted from one of the scenes and FBI used it as evidence to accuse Brandon Mayfield of perpetrating the attack. Mayfield was in fact innocent, and indeed later the Spanish investigators recognized that the lifted fingerprint did not belong to him. Unfortunately, he was detained for an extended period of time until his name was cleared. It seems that, in practice, fingerprint analysis is prone to error just like any human endeavor.

In terms of theoretical uniqueness, Jim Morrow, a maths professor in U Washington, wrote an interesting note to estimate the probability that no two people ever lived have the same fingerprints. He made a lot of assumptions and so the conclusion may not be entirely accurate, but it could be quite an interesting read to see how mathematics can potentially be used to tackle something “empirical” such as fingerprints.

Therefore, the answer to the question “are fingerprints unique?” is that: “We believe so. We don’t really know, and we ought to know.”

IV. Collecting Fingerprints

There are three types of fingerprints that can be found in a crime scene: patent prints, latent prints, and plastic prints. Patent prints are prints observable by the naked eye. Collecting them is easy enough: simply take a photograph. Latent prints, prints hidden from plain sight, takes a bit more work to collect. So are plastic prints, prints that record the three-dimensional aspects of the skin, for example those formed on wax, paint, and drying blood.

We can collect latent fingerprints on smooth, nonporous surfaces – for instance glass, metal, and concrete – via dusting called dusting. The thin ridges that extend beyond the dermis constitute a person’s fingerprints. When the fingers are pressed against a surface, the sweat and oil on the ridges leave a fingerprint pattern on it that, with a little physical development, can be brought back to light. In forensic investigation, fingerprint powder is used to reveal latent fingerprints. On dark surfaces, white or other light-colored powder is used, and on light surfaces – you guessed it – black or other dark-colored powder is used. Chemicals in fingerprint powder need to serve two purposes: provide color (dark/light) for effective visualization and adhere to the residues left by a fingerprint. Organic color pigments such as black ferric oxide are used for the purpose of visualization, and inorganic chemicals such as lead and copper are used for the purpose of adhesion. For the latter purpose, particles that are mechanically well attracted to residue deposited by friction ridge skin, as well as moisture and oil, are sought after. The size and charge of the particles also matter. The two types of particles are mixed to form the fingerprint powder used in dusting.

We now have fingerprint powder; how does the rest of the procedure go? In order to lift a latent fingerprint from a smooth nonporous surface (the easiest surface to work with), we first apply a small amount of fingerprint powder to the region of interest. Blow excess powder away if necessary. Next, use a brush with fine, soft bristles to gently brush away the powder that does not attach to the fingerprint. Hopefully a clear fingerprint will form. Then, we take a photograph of the developed print, and lift the print using a transparent adhesive tape. Finally, we tape the lifted print to an appropriately colored card (latent lift card) for preservation. [2]

As for prints left on porous surfaces such as paper, the surface is first processed with chemicals and then physically developed. The chemicals react with amino acids and inorganic salts on the print residue and change color, making it possible to photograph the revealed prints. There are other techniques such as alternate light source and superglue preprocessing, which prevents contamination of the crime scene by fingerprint powder, as well as special techniques for collecting prints from other difficult surfaces, but we won’t go into details here.

Lastly, we look at the three-dimensional plastic prints. They can be easily captured using photographs. A procedure called casting is used to preserve fingerprints, for example to store as evidence. Casting is quite simple in principle: liquid plaster is poured over the fingerprint and hardened to make a cast of the impression.

Figure 3 - A plastered print.

V. The Henry Classification System

After collecting the requisite fingerprints, the next step is to compare them to the suspects’ fingerprints, or even look them up in the fingerprint database. If we store the prints simply as images, looking them up will be a nightmare. We need to encode the fingerprints with serial and numerical values in an intuitive way, for easy recording and retrieval. In this section we will go through the Henry system [3], which as mentioned earlier is one of the earliest fingerprint filing systems. [4]

Most fingerprint filing systems are based on the presence or absence of three basic patterns: arches, loops, and whorls. An arch is formed when the fingerprint contour makes a hill-like shape. A loop is formed when fingerprint ridges make a U-turn (more precisely, the ridges should pass through the imaginary line between the core and the delta). A whorl is formed when the fingerprint contour goes around like a whirlpool. The diagram should make it clear how to distinguish between the three patterns.

Figure 4 - arches, loops, and whorls.

It was Francis Galton who first identified these patterns and attempted to use them to classify collected fingerprints. However, the classification system was not refined enough, and many fingerprint data got binned in the same class, making database lookup tedious. Both Vucetich and Henry believed that fingerprints would be the future of criminal identification, and each came up with improvements over the Galton system. [5]

Under the Henry classification system, each set of 10 fingerprints is assigned a two-column identifier. There are SIX (!) components: primary, secondary, sub-secondary, major, key, final. They are written in two-column form, in the order of:

key -> major -> primary -> secondary -> sub-secondary -> final

from left to right. Refer to the table below for a brief description of each component. The right-hand fingers are labeled 1 to 5 from thumb to pinky; the left-hand fingers are labeled 6 to 10 from thumb to pinky.

Loop count1 (right thumb)Whorl 2,4,6,8,102 (right index)2,3,45 loop count
6 (left thumb)Whorl 1,3,5,7,97 (left index)7,8,9(10 loop count)
Before getting into how the ensemble of 10 fingerprints is eventually assigned a Henry code, we must first discuss how each individual fingerprint is classified. We have previously mentioned arch, loop, and whorl, and now we shall learn how to further classify each of the three types of patterns.
  • Arches: arches are the rarest kind of fingerprints, totalling less than 10% of all. Arches do not have U-turn or circuit patterns, and ridge lines tend to flow from one side of the finger, rise near the middle, then fall and exit from the other side of the finger.

    There are two types of arches: plain arch (AA) and tented arch (TT). A plain arch has a less steep curve in the middle, whereas a tented arch tends to have an angled, upward thrust in the middle. The distinction is not cut-and-dry, and investigators may leave room for interpretation by marking the fingerprint with two (or perhaps even more) possible types. Therefore, sometimes one sees a fingerprint being typed as “A?T”, meaning that it could be a plain arch or a tented arch, depending on who’s looking at it.

  • Loops: elaborating on the basic definition given previously, loops should possess three defining characteristics: a sufficient unspoiled recurve, the presence of a delta, and a positive ridge count (the ridge count is the number of ridges between the core and the delta). The ridge count will be one of the parameters used in the Henry code.

    Another parameter is whether the loop is ulnar or radial. The names come from the two bones in our forearm: the radius and the ulna. If the loop enters and exits from the thumb side, it is a radial loop; otherwise, it is an ulnar loop. The ulnar loop is much more common than the radial loop.

    A fingerprint that is almost a loop, satisfying only two of the three defining characteristics, is classified as a tented arch.

    On a fingerprint card, if the fingerprint is a loop, we write the ridge count on the top right of the square (as a convention we add 50 to the ridge count of a radial loop), and at the bottom we use a slanted line towards the pinky side to indicate an ulnar loop and the letter R/r to indicate a radial loop. The capital R is used for index fingers, and the small r is used for other fingers. This will be significant later when we go through the Henry system.

  • Whorls: a whorl is a fingerprint pattern that consists of ridges that form a circuit, surrounded by two deltas. The circuit can either be a closed curve or take a spiral shape. There are four main types of whorls: plain whorl (P/p), double loop whorl (D/d), central pocket whorl (C/c), and accidental whorl (X/x). The plain whorl is by far the most common and is defined by the circuit ridge patterns crossing the line segment connecting the two deltas. A central pocket whorl is when the circuit ridge pattern does not touch the line segment connecting the two deltas (but does not have other extraneous features). A double loop whorl has, well, two loops, usually forming a zen-like shape. An accidental whorl has multiple characteristics of various fingerprint types, such as a loop and a whorl combined. Another key parameter is whorl tracing. Tracings are either inner (I), outer (O), or meeting (M). What we need to do is to trace an imaginary ridge line from the left delta to the side of the right delta. During the tracing, if there is a gap on the current path, we drop down one ridge layer. If we end up at least three ridges below the right delta, we have an outer tracing; if we end up at least three ridges above the right delta, we have an inner tracing; in other cases, we have a meeting tracing. We sum up the information in the table below:
    ArchesPlain arch (AA/a)
    Tented arch (TT/t)
    LoopsUlnar loop (U)
    Radial loop (R/r)
    WhorlsPlain whorl (P)
    Double loop whorl (D)
    Central pocket whorl (C)
    Accidental whorl (X)
    Secondary classification
    Loop ridge countNumber of ridges between core and delta
    Whorl tracingInner (I), Outer (O), Meeting (M)

    Now we are ready to describe the components of the Henry classification system.

    • Primary: for the top number, add up the presence of whorls in the 2,4,6,8,10 finger in binary representation (right index = 16, right ring = 8, left thumb = 4, left middle = 2, left pinky = 1), then add 1 to it. Similar for the bottom number. This results in two numbers between 1 and 32.
    • Secondary: write down the subclasses that the index fingerprints belong to.
    • Sub-secondary: if any of the middle, ring, and pinky fingers are one of the rarer subclasses that are encoded with small letters (a/t/r, see the table above), we use dashes and letters to indicate the locations of the arches. Otherwise, we look at the 2,3,4 and 7,8,9 fingers. For whorls, we write down I/O/M depending on whorl tracing, and for loops we write down I/O depending on whether the ridge count exceeds the threshold for the corresponding finger – I for under, O for over. The thresholds are 9.5/10.5/13.5 for index/middle/ring fingers.
    • Key: the key is the ridge count of the first loop from finger 1 to finger 10, ignoring however the pinky fingers. It is possible that there is no key, if no non-picky fingers have prints that are loops.
    • Major: the major concerns itself with the thumbs. For whorls, we write down I/O/M depending on whorl tracing, and for loops we write down S/M/L (small: 1-11, medium: 12-16, large: 17+) depending on the ridge count. Like the sub-secondary, there is no major if either of the thumbs are one of the rarer subclasses that are encoded with small letters.
    • Final: the final relates to the pinky fingers. It records the ridge count of one pinky finger that has a loop pattern (right pinky has higher priority); if neither is a loop then the final field is left empty.
    Let’s look at the example below.
    Figure 4 - Example of a set of fingerprints classified using the Henry classification system.
    For the primary, the whorls at 4, 6, 10 means the upper primary is 8+4+1+1=14. Analogously, the whorls at 1, 5, 9 means that the lower primary is 16+4+1+1=22.

    For the secondary, both index fingers have ulnar loops (pointing away from the thumb), so U/U is recorded.

    For the sub-secondary, the top is OOI because: right index finger is a loop with ridge count 19 >= 10, right middle finger is a loop with ridge count 15 >= 11, and right ring finger is a whorl with inner tracing. The bottom follows a similar reasoning.

    For the key, we look at finger 2 because it is the first finger with loop pattern. We write down 19 as it is the ridge count of finger 2. For the major, we simply write down the whorl tracings of the thumbs, as both are whorls.

    Lastly, the final field is left empty because neither of the pinky fingers have loop patterns.

    The Henry classification system was quickly deemed vastly superior to the Bertillonage, and at the turn of the century, a fingerprint bureau was set up in England and the Henry system replaced Bertillonage as the standard criminal biometric system. Until very recently, most criminal fingerprint identification systems were based on the Henry system of documentation, or modifications of it.

    VI. Non-criminal uses of fingerprints

    Nowadays, our fingerprints have become an indispensable part of our identity. Even for people without a criminal past, we readily use fingerprints for identification, for example in smartphone locks and immigration control. Apple touch ID for instance works by “using capacitive touch to take a high-resolution image from small sections of your fingerprint from the subepidermal layers of your skin”. [6] Other than basic fingerprint types, it also maps out precise details of finer patterns. The way Apple ensures privacy is to “[create] a mathematical representation of your fingerprint and [compare] this to your enrolled fingerprint data”, which I believe amounts to some form of hashing. Therefore, according to their claim, your complete fingerprint data is never stored in the phone. The false positive rate is calculated to be 1 in 50,000, so at least statistically it is more secure than a four-digit passcode.

    VII. Summary/Reflection

    We have gone through the history of fingerprinting and discussed fingerprint collection and classification, mostly for criminal identification. This article is far from comprehensive; as this article is long overdue and I have run out of energy to expand on it, we have to call for a premature end. Notable omissions include the technological aspect of modern mass collection of fingerprints and the associated security and privacy considerations. I have gone from having almost no understanding of fingerprinting to having a working understanding. I am glad to have undertaken this journey and I find this experience incredibly empowering. I would like to explore other interesting topics by writing similar expositions.


    • Simon A. Cole. “History of Fingerprint Pattern Recognition”. Automatic Fingerprint Recognition Systems, pp. 1-26, 2004.
    • R. B. Fosdick. “Passing of the Bertillon System of Identification”. Journal of Criminal Law and Criminology, vol. 7, pp. 363-9, 1915.
    • National Research Council. “Strengthening forensic science in the United States: a path forward”. National Academies Press, 2009.
    • Brian Yamashita and Mike French. “Latent print development.” The fingerprint sourcebook, pp. 155-222, 2011.
    • Robert B. Stacey. “Report on the Erroneous Fingerprint Individualization in the Madrid Train Bombing Case”. FBI Forensic Science Communications, vol. 7, no. 1, 2005.
    • Laura A. Hutchins. “Systems of Friction Ridge Classification”.
    • The fingerprint sourcebook. Washington, DC: US Department of Justice, Office of Justice Programs, National Institute of Justice, 2011.
    • Jim Morrow, “Fe-Fi-Fo-Thumb”. Retrieved from
    • “A Simplified Guide to Fingerprint Analysis”. Retrieved from
    • “How does fingerprint work?”. Retrieved from
    • “Which are the different techniques for preservation of patent print and plastic print? Both are the type of fingerprint.” Retrieved from
    • “Why do we have fingerprints?” Retrieved from
    • Bret Little, “Henry Classification Secondary and Subsecondary.” Retrieved from
    • Apple Inc. “About Touch ID advanced security technology”. Retrieved from
    • (And obviously Wikipedia)


    [1] As a side note, Alphonse Bertillon was involved in the Dreyfus affair – a controversial case in France concerning the alleged treason of Captain Alfred Dreyfus – as an expert witness.

    [2] I tried lifting fingerprints on my wooden desk using flour and a small brush from a vacuum cleaner. Apparently, the bristles of the brush were not even and soft enough, and the experiment failed. I guess being a forensic investigator is not that easy, huh... Also, I just realized that we should be applying the powder to the brush, then dust the surface, not pour the powder onto the surface then brush the excess off...

    [3] Not to be confused with Dr. Henry Faulds, who featured in the “History of fingerprinting” section.

    [4] It puzzles me how investigators would be able to determine which finger a collected fingerprint comes from. I believe that in certain cases, for example where complete palm prints or thumb prints are found, it is easy, but in other cases determining the finger seems to be based on experience and trial-and-error.

    [5] Readers interested about the Vucetich system may read the referenced chapter “Systems of Friction Ridge Classification”.

    [6] Capacitive touch works by drawing a small electrical charge from the point of contact. Compared with resistive touch which uses (physical) pressure, capacitive touch can detect lighter touches and with greater accuracy. (Information from


    Copyright 2016-present George Cushen.

    Released under the MIT license.

Kam Chuen (Alex) Tung
Kam Chuen (Alex) Tung
PhD Candidate in Computer Science