Face Blind!

Chapter 8B

Recognition System - The Essence Model

...by Pertti

An Introductory Note by Bill

As a child, each person builds a system for telling people apart. For people who are not face blind, the face is a frontrunner so far ahead of the pack of other contenders, that using it is an obvious choice. Those who choose this frontrunner never give this whole process much thought.

With the face yanked out of the race, the face blind face a bunched pack of contenders, and depending on their environment and their personal strengths, they are apt to settle on any number of schemes. In the other chapters of this book, I describe in detail how I tell people apart, and how I've found other face blind people who developed a similar scheme. But there are schemes in use quite a bit different than mine, too.

Pertti, a 40 year old man from Finland, describes the method he uses below. Since describing his method, he has, just like me, had other face blind people say their method is similar to his.

Recognition System - The Essence Model, fully revised version, September 1999

...by Pertti

Pertti The picture on this page shows me dressed for a Medieval theme Live Action Role Play. I don't usually look like that, but this is one of the best photos of me lately, and shows my hair quite nicely.

Recently I took a test where I was supposed to recognize photos of 16 people after having seen only briefly.  The first part of the test consisted of checking them immediately and the other part was after an hour had elapsed.  Based on the results, I fared worse than about 99.9% of twelve-year old children, for whom the test was meant.  But when the results were broken down, it turned out that I recognized seven girls out of eight, but only one boy out of eight.  Now, if you realize that I had three choices for each photo, even random choosing should produce 2.33 as a score.

So we can safely say that I recognize girls at an almost average level and boys not at all.  One thing that was common with all the girls was that they had long hair, whereas all the boys had short hair.  But still, these results seem to support the view that Bill has already discussed elsewhere.  There is a part of humanity whom I can recognize as effortlessly as normal people see everyone around.  This part of humanity for me consists of longhaired females with certain kinds of hairdos (I'll call them "type females" hereafter.)

Having made some experiments with photos of female celebrities who are type females, I had already earlier found out the significance of hair.  A nice-looking female could be made into a dull-looking one by simply changing her hairdo.  I also cut the face out of a dull-looking picture and inserted it into a nice-looking one.  The result was that it was still somewhat nice-looking.

So hairstyle determines for me whether a face looks nice or not.  I also found out that eyes and the area around them are important for me, but they are of no use in recognition.  They simply convey the facial expression, which for me becomes part of the face.  So when I next see a person, I might think they are a different person if their expression has changed.  But changing their expressions is something people do continually, so no wonder I have trouble.

I could never survive my daily life with those skills only.  I mean, if I only ever could recognize nice-looking females with long hair, how could I e.g. make sure that my boss was always one of them?

I do have other skills, which are based on my normal visual skills.  I have learned to recognize different brands of computers, by looking at the front panels.   I can also learn to recognize differently named individuals by looking at their body.  I do use all available visual clues, including motion.

My recognition of a person seems to be based on the following visual clues:

  • length and shape of hair
  • overall body shape and size
  • general head shape
  • posture and way of moving around
  • clothing or fashion style
  • all other visual details
  • But just like ordinary people, I do not make a list of these features, but they are stored visually instead.  So I might not be able to bring to mind verbally the details regarding any person - e.g. I have a hard time remembering whether a certain person has light or dark hair, but when I see them having their hair in a new color, I immediately spot the change.

    Yet, as you may have well guessed, this all is still not enough.  I have realized that my accuracy is somewhere around 1 in 30.  I.e. if there are about 30 persons around, I tend to find one pair whom I mix up.  This confusion seems to apply to men only -  for women I seem to have almost average skills, but those are based on long hair exclusively, so short-haired women do cause me more problems.

    When I see certain people often enough and in familiar surroundings, I learn to recognize them quite reliably.  No wonder I tend to like arranging my life so that I can easily predict whom to expect and where.  When out-of-context people appear, they usually need to greet me first, before I have even an idea I know them.

    Then again, I also use voice to confirm recognition, but based on many experiences, voice itself is not a reliable recognition means.  But when combined with visual clues, I can increase my 1 in 30 accuracy to maybe 1 in 300, which is quite near the number of people I can recognize.

    It is very hard to keep all the other information from "contaminating" the stored image of a person.  It seems that I store an impression of a person around the face - so I think I see the face, whereas what I actually see is an amalgamation of all the available information of a person.  I have noticed that in a new job people's faces change during the first month or two, based on what other information I learn about them.

    Among this information that gets stored, if known, are:

    The exact list depends on the person, and some individuals get recognized with a very small set of information.  I tend to like such people, and then have more capacity to store facts about their interests and life, to be used like normal people use similar information.  But when I need a lot of information for recognizing the person, it takes more time, and leaves me less for use in social interaction.

    For type females I seem to be able to make reliable recognition based on visual characteristics only.  But for all other people, I combine the visual and non-visual attributes into what I call  the essence of a person.  This essence method of mine results in a rather accurate means of recognition, but the learning is slow and forgetting quick.  So I need to have it refreshed at least once a month.  It is rather easy to note that this means that the circle of people whom I actively engage in social interaction with is rather limited.

    There is a magical processor in me which automatically takes all the observations and maps them into a resolved essence class, i.e. a class where all the essence properties have a specific value. The process is also time-dependent, i.e. it can correct itself in real-time, provided that new data appears.  Unfortunately the retention is not very good - maybe because storing these essences takes a lot of memory capacity, and that limited space is continuously used for storing essences of new people.

    I think this system works similarly to what normal people have, except that in their case it is the face that gets most of the weight and other properties are used only for additional checking. In my case there is a big hole in the face data, and hence the 'additional checking' data gets to be the primary data for recognition.

    Note that there is an automatic weighing system, which applies more weight to essence properties that are very distinctive. I.e., a guy with long and blond hair would get much weight on his hair essence, if his ordinary context only contained males with short or dark hair. But if I saw him among long-haired males, it might turn out that my recognition would initially fail in that context for him. Then new properties would need to be added.

    An essence class is not a fixed entity which there are only a limited number of, but rather it is a hierarchical structure. At first sight all people that look like, e.g. Bill, land in the same category (long haired and bearded, nice looking, middle aged, white men).

    If there is no other input than their visual appearance, that is where the essence classification ends, and hence I recognize that person only as a representative of a specific class. There are dozens of these visual classes, but not nearly enough for everyone to have a separate class. If I have about 30 coworkers, I can usually open a separate essence class for everyone. After that it usually happens that two different people whom I don't meet often land in the same essence class and I have trouble telling them apart.

    However, if I hear the person's voice, or see him/her more, and always in a specific context, the essence class gets "deeper", i.e. contains more descriptive information.

    If I have all the essence properties stored and fresh, and if I get all that information in a real-life recognition situation (or even much of it), I can quite reliably recognize a person.

    The caveat is, of course, that in the first place there are not so many people for whom I have stored the full set of essence properties, and in the second place, in a real-life situation it is not possible to get enough information - at least not immediately.   Hence, it takes me precious seconds to ascertain whom I am actually interacting with, and lose a lot of the initial social interaction, which makes the other person think I am not interested in them.  Whereas in reality I do not yet even know whether I am interested in them, because I have not found their file in my mind.  As normal people recognize familiar faces in a split-second, they usually give no allowance for someone using seconds, minutes, or sometimes even hours for that 'simple task'.

    Another caveat is that the storage quality of the essence properties is time-dependent, and hence the more time has passed, the less specific a person's classification becomes (data deteriorates). Part of this is caused by my brain needing the space for other similar data and part by originally storing data that is bound to change (fashion style, hair length, etc.)

    Sometimes there remains a problem that two distinct people end up in the same essence class. Then I just cannot keep a separate file open for each of them. In such cases I (have to) assume that they have the same character and know the same things about me and are interested in the same things. Shortly put: I treat them as the same person, although I know they are not.

    Summary: the essence system uses all the available information for recognition. It is rather reliable in limited contexts. It fails in many situations that normal people encounter all the time, but which I avoid - specifically for this reason.

    So far, we've mainly concentrated on how we recognize people. Now we move on to look at how the deficit we face blind people have in that area affects several facets of our social lives. It is fitting that our first inquiry into that realm is into an area where Pertti made significant discoveries. We next look at how face blindness affects one's emotions.

    "Face Blind!" - Table of Contents

    Chapter 1 Introduction
    Chapter 2 Discovering Face Blindness
    Chapter 3 Physical Causes of Face Blindness
    Chapter 4 The Importance of Recognizing Others
    Chapter 5 How Most People Recognize Others
    Chapter 6 Ways To Recognize Others Without Using the Face
    Chapter 7 How Non-Face Recognition Methods Work in Practice
    Chapter 8A ...Bill: How I Tell People Apart - BACK 
    Chapter 8B ...Pertti: Recognition System - The Essence Model - YOU ARE HERE 
    Chapter 9 Effect of Face Blindness on Emotions - NEXT 
    Chapter 10 Effect of Face Blindness on Sexuality
    Chapter 11 Effect of Face Blindness on Your Social Groups
    Chapter 12 Understanding Why People Choose To Look Alike
    Chapter 13 Ways To Improve Our Lives

    Appendices

    Appendix A How To Find Medical Articles on Face Blindness
    Appendix B Getting Diagnosed (Tested) for Face Blindness
    Appendix C Links to Other Face Blind People
    Appendix D Author's Information Page

    This document is copyrighted. For information, or to contact the author, go to Appendix D, the Author's Information Page. Text of this chapter last revised September 16, 1999.