Perception – discrimination

by Neil Rickert

Perceptual discrimination is the act of distinguishing between different items that are in the perceptual field.  In this post, part of my series on perception, I will look at barcode scanning to illustrate what discrimination is, and its role.  And I will use this example to further clarify the distinction between direct perception and indirect perception, at least as I use those terms.

These days, we see bar codes on many of the items that we purchase.  And the store clerk typically uses a scanner to read that bar code and identify which item we are purchasing.

Bar codes

You can find a detailed description of bar codes at this Wikipedia entry, and at a variety of other sites on the net.  There are actually several bar codes in use, but they all work in much the same way.  In simple terms, a bar code consists of a sequence of parallel bars, each of which is either white (or a light color) or is black (or a similar dark color).  In order to read the bar code, the reader must identify the individual bars, or, in other terminology, must discriminate between the individual bars.  And, having identified a bar, the reader must distinguish between whether that bar is a light color (such as white) or a dark color (such as black).  Thus we need both discrimination between bars, and discrimination between colors for the bars.  Once the bars and their colors have been discriminated, this information about the code could be represented in the computer as a string of binary digits, with a digit for each bar where the digit is 0 for a light color and 1 for a dark color.

In practice, there is some encoding going on, so that the end result of the scan is not just the sequence of binary digits suggested above.

Discrimination

Discriminating colors and discriminating bars cannot be completely separated.  In looking at a bar code, you can see a sharp contrast where a black bar abuts a white bar.  So discrimination begins by looking for the sharp transition in reflected light between adjacent bars of different color.  However, that is not the entire story.  There can be two adjacent  bars of the same color, and in that case there may be no sharp transition.  So the first step for the bar code reader is to determine the approximate size of each bar, based on the transitions that can be seen.  The  encoding used ensures that there are enough color transitions to be able to determine the size of the bars.  And then, once the size is known, two adjacent bars of the same color can be discriminated on the basis of distance measured across the bar area.

In practice, the bar code reader does not even know where to find the bar code.  So it must begin by scanning the general area to find the kind of color transitions sought.  The encoding used requires that there be some guard bars to help recognize what is a bar code, and what are color transitions due to other markings on the product label.  So the scanner starts by scanning for transitions, looking for a pattern of transitions that matches what it generally expected to be a bar code.  Having found a potential code, it then attempts to discriminate the individual bars, and then the colors of each bar, so as to read the code.

The most common bar code reader technology uses a laser to scan across the broad region looking for transitions.  A light sensor receives the light reflected from the laser beam striking the surface of the product.  The motion of the laser beam is coordinated by the search for signal transitions coming from the light sensor.  This is the kind of active search for an expected pattern that is analogous to what proponents of direct perception expect our perceptual system to be doing.

Indirect bar code reading

I am not aware that there are any indirect bar code readers, but we can speculate about how one might work.  The scanning equipment would form a pixel map of the entire region where a bar code is expected, and then a computer program would examine the pixel map, and attempt to find the bar code there, and to discriminate between the bars in the pixelized image.

Forming a pixelized image, already requires discrimination.  Each pixel represents a small portion of the region, and the pixel map itself is the result of discriminating into multiple small regions, so that the composite pixel map approximately represents the entire region.

Once the pixel map has been produced, a second level of discrimination is required to distinguish between parts of the pixel map that represent each bar.

Direct vs. indirect perception

I can now explain, more precisely, the distinction that I am making between direct and indirect.  Whether perception is direct or indirect depends on how we discriminate between items in our perceptual field.

Perception is direct, if our perceptual system is directly discriminating between items in our perceptual field, based on whatever means are available.  Perception is indirect if the perceptual system is making a fine grained discrimination of the entire perceptual field, without concern to what is in that field, and is then doing a secondary discrimination on the range of data formed by the first level of discrimination.

In terms of human vision, we are using direct perception if our visual system is using the sharp transition in signal at a retinal receptor, as the eye moves in saccades.  We are using indirect perception, if those signal transitions are not part of how we discriminate between items in the visual field.  If we are using indirect perception, presumably the use of saccades would be a mere implementation detail of how internal pixel maps are generated.

My own view is that we are using direct perception.  The usual bar scanners directly discriminate the bars, because that is the simplest and most reliable way of doing the job.  Looking at our visual system as the result of evolution, it seems to me that evolutionary processes would have come up with similar direct ways of discriminating items, for the same reasons of simplicity and reliability.

Summary

I have described the general idea of discrimination, and how such discrimination is important for identifying detail.  And I have explained the differences between direct and indirect perception in terms of how that discrimination is done.

Advertisements

3 Comments to “Perception – discrimination”

  1. “Looking at our visual system as the result of evolution, it seems to me that evolutionary processes would have come up with similar direct ways of discriminating items, for the same reasons of simplicity and reliability.”

    We mustn’t forget that evolution does not have to produce reliability or simplicity or any particular advantageous trait. This is a common misconception by many — that evolution only produces more favorable traits over time (I’m saying that you’ve made this error, but thought it relevant to your claim). It only has to produce a net advantage to the organism — just enough to survive and increase rates of reproduction relative to competing organisms. Our means of perception can be complicated and unreliable, as long as the level of complexity and reliability in combination with the OTHER traits of the organism, are good enough to allow the organism to survive and reproduce accordingly. So while we can say that the reliability of perception will probably tend to increase over time based on the principles of natural selection (if all else is equal), evolution can just as likely produce indirect perception or less reliable perception over time as long as the advantages of the organism outweigh the shortcomings of its perceptual mechanisms (e.g. if the defensive mechanisms of the organism are enough to compensate for its perceptual system’s relatively low correlation to reality) simply because we are not evolving one trait at a time.

    Like

  2. “Perception is direct, if our perceptual system is directly discriminating between items in our perceptual field, based on whatever means are available……We are using indirect perception, if those signal transitions are not part of how we discriminate between items in the visual field.”

    As I mentioned in an earlier post of yours, it’s also very possible that we are using a combination of direct and indirect perception, that is, we could have some perceptual discrimination occurring based on direct signal transitions in combination with internal pixel mapping (representationalism of some form). This seems the most likely to me as the brain fills in gaps of information based on some unreal representation, that is, the gaps may be filled in through the use of representationalism (indirect perception), and the “non-gaps” may be the result of the direct perception component. If our brain didn’t fill in gaps creating optical illusions among other things, exclusive direct perception would seem more plausible — but since optical illusions are evident in our perception, it’s clear that perception has to involve the qualitative equivalent of an inclusion of data that isn’t really there (a representation).

    Like

    • As I mentioned in an earlier post of yours, it’s also very possible that we are using a combination of direct and indirect perception

      That’s possible, and I have not ruled that out.

      There’s a lot of misinformation about direct perception. I have been mainly trying to make the case that it is not mysterious or magical, and that there are criteria that could be used to determine empirically which we use in particular circumstances. This post laid out some criteria that could potentially be used. Ultimately, it will be up to experimenters to investigate this.

      Like

%d bloggers like this: