Langston, Cognitive Psychology, Notes 5 -- Imagery
Note:  For a lot of this I relied on a very nice essay by Bruce (1996).  Her edited volume Unsolved Mysteries of the Mind doesn't really play up the mystery part as much as you'd expect, but it presents a nice discussion of some of the debates in Cognitive Psychology.
I.  Goals.
A.  Where we are/themes.
B.  Measuring imagery.
C.  Images and memory.
D.  Using images.
E.  Structure of images.
F.  Other kinds of images.
G.  Reality monitoring.
II.  Where we are/themes.  It's hard to fit imagery into the course as a separate unit since it goes with so many other areas.  I'm leaving it separate because going over how imagery interacts with the other systems will serve as a good review.  Basically, our questions are: Here's a little exercise (from Bruce, 1996).  Imagine a dinner plate.  There's some spaghetti around the top rim.  Just below that are two fried eggs side by side.  In the middle is a carrot, pointing down.  Below the carrot is a banana.  How many people have a clear mental image?  What do you see there?
Do people use images?  This is a perfect cognitive psychology question.  It's the ultimate kind of mental event.  Note that demonstrating that people do use images puts all of our methodology to the test.
On the surface, it's obvious that imagery exists.  Here's a simple example from Descartes.  Imagine a hexagon and a pentagon.  Can you mentally see the difference between the two?  Now, imagine a 999 sided polygon and a 1000 sided polygon.  Can you see that?  You can conceive of such things, but you can't image them, suggesting that imagery is something different from language and other thought processes.
Some big debates:
A.  Can you have imageless thought?  Is it possible to have a thought not accompanied by a visual, verbal, or tactile image?  Does it matter?
B.  What is a visual image?  Is it a picture in the head?  Activation of neural circuits used for vision?  Something else?
C.  How do you know what's real?  We'll finally tackle some issues raised in the first lecture by asking how you know you're remembering an image and not something that actually happened.
III.  Measuring imagery.  Most models of intelligence acknowledge that there are multiple intelligences.  You can be verbally strong, but spatially weak, and vice versa.  In other words, knowing a person's verbal ability doesn't necessarily tell you anything about their spatial ability.
How can we assess your imagery?  There are lots of tasks.  Here are three sampled from a book on intelligence by Guilford (1967).
A.  Cube folding.  I have some expanded cubes with arrows.  Mentally fold the cubes and tell me if the arrows will touch.
B.  Cube task.  I show you the initial picture of an ‘X’ on a cube.  Then, I take away the cube and tell you how the X moves.  You tell me where it ended up.
Demonstration:  There are two:
1.  Cube folding.  Will the arrows touch?  The first is “yes,” the second is “no.”  I have some you can fold if you want to check.
 Cube folding pictures
2.  Cube:  Here's a cube.  Here are the moves:  One down, one back, one down, one left, one back, one right, one up, one forward, one forward, one up.  Where is it?
 Cube picture
What can we learn from these types of tasks?  Just like digit span can tell us about short term memory capacity, this can tell us about your spatial ability.  There seems to be a lot more variability in spatial ability than verbal ability (in college populations).  Why?  Think about the skills emphasized and the tests used for admissions.
Note the parallel to working memory.  All of these tasks require storage plus processing.
IV.  Images and memory.
A.  Learning pictures.
1.  Shepard (1967):  People saw a long list of pictures or words (612).  After two hours, they took a recognition test (two alternative forced choice).  People were nearly 100% accurate for pictures, about 88% for words.  After a week, people were about 88% for both.  So, picture memory is better than word memory, but not after a long delay.
2.  Standing (1977):  Learn 1000 words, 1000 simple pictures, and 1000 bizarre pictures.  Recognition was tested two days later.  People learned 615 words, 770 pictures, and 880 bizarre pictures.
So, people are really good with pictures, and better than with words.  This is part of the evidence that will lead Paivio to propose a dual code (one verbal one visual).  We'll see that later.
B.  Using imagery to learn other stuff.
1.  The concrete-abstract dimension.  Words vary on a lot of characteristics.  One is how concrete the word is (“computer” vs. “thought”).  What effect does concreteness have on memory?  Concrete words are easier to recall than abstract words.  I have a demonstration of this.
Demonstration:  Here's a list of pairs.  Some are pairs of concrete words, some are pairs of abstract words.  Learn them, cued recall.  Should remember more concrete words.
Why?  Paivio thinks it's a dual code thing.  You have a picture code that's separate in memory from your word codes.  When you get a concrete word, you can get an image in the picture code, that gives you two chances to recall.  If you get an abstract word, you only get one chance to recall (verbal code) and that makes it more likely that you will forget.  This dual code also explains why picture memory is better than word memory.  Two codes = twice the chances to find what you're looking for.
One source of evidence for dual codes is the symbolic distance effect.  The basic effect is that it is harder to make decisions about the relative orderings of things the closer they are along a dimension.  For example, think of a flea, a fly, a rabbit, a German Shepard, and a horse.  People's response times to decide which is larger are faster for comparisons like horse and flea than for comparisons like fly and flea.  The closer the objects are, the harder it is to decide.  This is best explained as an image thing.  People access a mental representation that is an analog of the real thing, compare it to another image, and “see” the result.
One source of evidence for Paivio is that picture symbolic distance tasks are faster than verbal symbolic distance tasks.  The picture can access the correct mental system directly, the verbal task has to be recoded before a comparison can be made, making pictures faster.
Note that on one hand a symbolic distance effect is kind of nuts.  If I describe an ordering to you that includes the sentence “the fly is larger than the flea” and then ask you to verify the statement “the fly is larger than the flea” it takes you longer than a statement about the fly and a horse even though it's the exact same sentence.  That would lead some to say that verbal and image systems are clearly separate.
C.  Putting imagery to work.  Mnemonic devices (memory tricks) usually take advantage of imagery.  A popular one is peg-words.  You first memorize a list that goes from one to ten (one is a horse, two is a picture, ...).  Then, when you have a new list to learn (like your shopping list) you imagine the items on the list interacting with your peg words.  So, if you need to buy a toothbrush, you might imagine a horse brushing his teeth.  When you get to the store, you start in “one is a horse” and that triggers the memory of the image, and you buy a toothbrush.
Demonstration:  I'll hand out some peg-word systems.  Practice until you've completely learned your system.  Here are the three systems:
One is a bun One is a door One is truth
Two is a shoe Two is heaven Two is data
Three is a tree Three is a gate Three is a message
Four is a door Four is sticks Four is love
Five is a hive Five is a hen Five is knowledge
Six is sticks Six is a bun Six is fun
Seven is heaven Seven is a tree Seven is time
Eight is a gate Eight is a hive Eight is have
Nine is a line Nine is a shoe Nine is beauty
Ten is a hen Ten is a line Ten is focus
Then, memorize the shopping list.  Then recall.  Compare.
One should be best, followed by two, followed by three.  Why?  One has rhyming plus imagery (the words are all concrete).  Two has imagery, but no rhyming.  Three has neither.  By the way, one is the real peg-word system if you want to use it.
V.  Using images.  Now we come to the real problem.  I can't see you having an image.  So, the data I'm about to present are based on a great deal of trust that participants are really doing it.  The basic question is to look at the relationship between pictures and images.
A.  Mental rotation.  If you physically rotate something on a table, the farther you go, the longer it takes.  Do images work this way?
CogLab:  Mental rotation exercise.
It should work.  Shepard has found this effect in a number of studies.  The farther you rotate, the longer it takes, and the relationship is pretty linear.
B.  Scanning.  If I ask you to visually scan from one spot to another, the farther you go, the longer it takes.  Kosslyn had people memorize an island, and he found a linear relationship between distance and time.
Another experiment manipulated instruction set (Kosslyn, 1976).  Some people were instructed to use images to help answer questions, others had no instructions.  Imagery people were faster to answer questions if the size of the item probed was larger (does a bee have wings).  Non-imagery people were faster on the basis of association strength (does a bee have a stinger).  So, focusing on images seems to access a different kind of information.
VI.  Structure of images.  What is an image?  Phenomenologically, what does it feel like?  What might it be?  Again, since we can't see it, this is entirely speculative.  But, here are two options:
A.  Images are propositions.  A proposition is an idea unit.  It's basically verbal in nature, but it isn't words or linguistic.  It's a kind of language of thought.  So, images feel like pictures, but they're really coded in a different kind of language.  The feeling isn't real.
B.  Images are pictures.  Sort of like what we've been building so far.  On one level, this is absurd.  A tomato is red.  A picture of a tomato is red.  Is an image of a tomato red?  No.  Is an image of an elephant physically larger than an image of a rabbit?  Some of the scanning stuff certainly suggests this (“think of a rabbit by a fly, think of his eyelash” is easier than “think of a rabbit by an elephant, think of his eyelash”).
C.  How can we tell if an image is like a picture?  Well, we could look for cases where pictures and images are different.  Let's try some of that.
1.  It's hard to reverse an image, it's easy to reverse a picture.  Look at the drawings below:
 Rabbit and cube
The one on the left could be a rabbit or a duck.  You should be able to see it both ways.  The cube on the right can flip which face is the front.  Again, you should be able to see it both ways.  Now, get an image, hide the paper, and try to flip it.  It should be hard, and you might not be able to do it.  So, images and pictures do differ.
2.  It's hard to decompose an image.  Try the demonstration.
Demonstration:  Are images like a picture in the head?  If so, then activities that a picture can support should also be supported by an image.  Let's see (based on Reed and Johnsen, 1975).
I'm going to show you a figure.  Memorize it and get a clear mental image.  When you're ready, raise your hand.  Now, I'll cover it up and ask you if certain shapes are part of the figure.  Write down the letters of all of the shapes that are in the figure.  Show that people are usually bad.
Let's try number two.  Bad again.
Most people make a lot of mistakes.  Again, images and pictures are different.
However, Finke, Pinker, and Farah (1989) argue that image reinterpretation is possible, it just depends on the kind of processing.  Overcoming low-level grouping information is hard, higher level organization should be possible.  An example:  Think of a lower-case k.  Now, imagine a circle around the k, just not touching it.  Now, cut off the lower half of the k.  What do you have?  A lot of people can do this sort of thing.  The rabbit and reinterpretation demonstrations are a different sort of processing.
D.  Maybe images use perceptual hardware (it's like seeing, only fainter).  So, an image is basically utilizing the same hardware that you use for perception.  This makes it qualitatively and phenomenologically different from processing verbal information, but still ties it to a realistic neurological foundation.
1.  We can start by looking at interference.  This is popular with cognitive psychologists.  If images use visual hardware and words use verbal hardware, pictures and words shouldn't interfere.  If they all use the same hardware, pictures and words should interfere with each other.  Try the demonstration.
Demonstration:  Are images really verbally coded (like propositions)?  We should get different patterns of interference if images and verbal codes are different.  Let's try (this is based on Brooks, 1968).  I have some letters.  The task is to go from the first corner around the letter clockwise.  At each corner, tell me “yes” if it's an outside corner and “no” if it's an inside corner.  You will respond based on a mental image.
There are two response modalities, visual and verbal.  If I assign you to verbal, you whisper your “yes” and “no” answers.  If you're visual, you look at the ‘Y’ or ‘N’ on the correct row depending on what corner you're on.  If images are visual, visual responders should be slower (more interference).  To let us know, I'll split the class in half.  Clap when you're done.  The verbal half of the room should clap first.
Now, let's look at verbal codes.  I have some sentences.  For each word, respond “yes” if it's a noun or pronoun, “no” otherwise.  We'll have the two modalities again.  In this case, the verbal modality should be interfered with.
We should get more interference between pictures and pictures than pictures and words and vice versa.  So, they seem to use different hardware.  More importantly, it's a case where images and perceptual mechanisms seem to share resources.
2.  Finke and Kosslyn (1980) measured acuity in visual and imagined fields and found the same elliptical visual fields in both domains.  In both cases, better acuity was below the horizontal axis than above.  The findings were not consistent with people's beliefs about their imaginal visual fields, suggesting that the results weren't due to demand characteristics.  Spatial frequency resolution also followed this pattern (Finke & Kurtzman, 1981).  These results suggest a similarity between visual perception and imagery.
3.  Imagery doesn't seem to produce after-effects like vision does.  For example, staring at red bars and then looking at a white background will cause you to perceive faint green bars.  Imagining red bars will produce no such after-effect (maybe?).
4.  Imagery also doesn't seem to lead to repetition priming effects.  For example, imagining words in lower case doesn't seem to improve perception of those words.  This is a relatively new area of research, so maybe something will emerge here.
5.  Imagery supports high level redescription, but not low level grouping (see above).  Vision supports both.
E.  The conclusion:  It will be hard for us to tell what images are using purely cognitive methods.
One approach to solving the imagery debate (the “what are images” question) comes from neuropsychology.  You can get a look at a person's brain as they form a mental image vs. do a verbal task.  Are the same parts of the brain involved in the two tasks?  Do images actually use visual cortex like perception?  It's sort of a mix between cognitive approaches and neurological approaches.
Another possible resolution may come from additional interference experiments.  Basically, if you tie up the parts of the perceptual system that perceive a particular kind of stimulus, that should hurt imagery as well.  If you get a consistent pattern of this kind of result, it would suggest a relationship between imaginal and perceptual systems.
VII.  Other kinds of images.  I've been acting like visual images are all there are.  Obviously, that's wrong.  Can you imagine the smell of leaves burning?  Can you imagine the taste of a steak?  Do you hear an inner voice?  (Uh-oh.)
Just a taste of this.  Smith, Reisberg, and Wilson (1992) had people reinterpret strings like “NE1 4 10S” as “anyone for tennis,” but silently.  People could interpret 73% of the strings in a quiet room.  With auditory distraction, they could do 40%.  With articulatory suppression, 21%.  In other words, the more you engaged the vocal hardware, the worse their inner voice imagery got.
This area is rich in WRR opportunities.
VIII.  Reality monitoring.  Have you ever had a dream that you later confused with a real experience?  (Uh-oh.)  Do you sometimes experience déjà vu?  Have you ever remembered something that didn't really happen?  These are all problems with reality monitoring.  How do you distinguish imagined events from real ones, especially if they use the same neural hardware, as we've been discussing above?
The basic answer is a process of summing up perceptual information and information about cognitive activity associated with the memory, and comparing the two.  A lot of cognitive with little perceptual is probably imagined.  A lot of perceptual and some cognitive is probably real.
Test?  Johnson, Foley, and Leach (1988) had people imagine words being spoken by someone else (whose voice they were familiar with) and then had that person read some words out loud.  This caused more confusion than imagining words in their own voice or some other person's voice.  In other words, changing the balance between perceptual information and internal information made discriminating between real and imagined harder.
Finke, Johnson, and Shyi (1988) showed that the less attention you pay when you're creating images (or the less work you do), the harder it is to separate imagination from reality.  People had to complete shapes like the following:
 Images to fill in
When the shapes were letters and numbers, it was hard to tell which had been seen whole and which were imagined whole.  When the shapes were novel (taking more effort to complete) people were pretty good at the task.
Brain-wise, two parts are implicated in reality monitoring.  People with frontal lobe damage have trouble with confabulation of “memories” that are false.  It is difficult or impossible to persuade them that those events did not occur.  People who have their temporal lobes stimulated tend to experience déjà vu.  It appears to arise as a result of reality monitoring, but the exact mechanism is unclear.  That should just about clear everything up on the déjà vu front (he says, tongue-in-cheek).
(This is a great section for a WRR.  Some WRRs I've had for this topic:  McNally, R.J., & Kohlbeck, P.A.  (19  ).  Reality monitoring in obsessive compulsive disorder.  Behavior Research and Theory, 31, 249-253; Anderson, R.  (1984).  Did I do it or did I only imagine doing it?)

Cognitive Psychology Notes 5
Will Langston

 Back to Langston's Cognitive Psychology Page