By Benjamin Moody
People have some amazing systems for storing objects and recognizing them. In general, when people are trying to recognize shapes, they rotate the image in their minds until they can see that it matches up with a basic pattern encoded in their memory. This process is called mental rotation (Shepard 1982).
There are two different possibilities for how this system might work. The storage of images in the mind could be based on a retinal reference frame or a gravitational reference frame. This means that images would either have to be rotated to be upright in a subject's field of vision or relative to the ground. There is evidence that both are used, depending on which is easier (Friedman 1996).
People "learn" to use mental rotation if they are exposed to it often. For example, people have higher scores on mental rotation tests if they have been playing a computer game which requires mental rotation (De Lisi 1996).
People also recognize shapes more easily when they appear larger to the eye, but the actual size does not have a direct effect on mental rotation speed. This means that a small picture close up and a large picture farther away will rotate at the same speed if they appear the same size on the retina (Suzuki 1988).
Applying these ideas to reading raises some new questions because of the way written words are formed from sequences of letters. People can usually recognize letter shapes without rotating them individually. When a person is reading, they recognize the letters and put them in order to compare with words in memory. However, this ordering is disrupted when the letters are lined up in a different direction. The person must go back to the letters, figure out the word orientation, and reread each letter in that direction. (Koriat 1989)
There is a question as to whether gender plays a role in how well people do in mental rotation tasks. Some references say there is a difference, others say there isn't (Uecker 1993). It is commonly believed that men do better than women with mental rotation tasks.
As Friedman and Hall note, the findings of their recent study (Friedman 1996) "do not address the issue of whether the frame of reference for representing familiar objects is viewer- or object-centered" (i.e., retinal or gravitational). Letter forms in words are unquestionably familiar objects to experienced readers.
Most of these studies use rotation of objects, not words, and those that do use words use only individual words, not a block of text. Also, although studies indicate that people rotate objects to the upright in their mind before recognizing them, very few of these studies address the idea of what the upright actually is. Is the upright determined by gravity, or by the position of the person's head? My project, which studies the effect of head tilt on reading, explores this open question.
My experiment is about mental rotation. I want to find out whether people rotate text to be upright relative to gravity or relative to the orientation of the head before reading it. To do this, I have created an experiment consisting of seven tests in which subjects will read text on a computer screen and the computer will time them to see how long it takes them to read.
I have found many papers on the subject, written by various people, but most of these use rotation of objects, not words, and those that do use words use only individual words, not a block of text. Also, although studies indicate that people rotate objects to the upright in their mind before recognizing them, very few of these studies address the idea of what the upright actually is. Is the upright determined by gravity, or by the position of the person's head?
I think that people will rotate text to be upright relative to gravity before reading it. I think that text oriented horizontally will, on the average, be easier to read than text oriented in another direction. I think this because of the observation that people recognize objects equally easily no matter what orientation their heads are at.
In my experiment, subjects were shown text on a computer screen. This was run by a computer program I wrote in Microsoft Windows Logo. MSWLogo is not sold by Microsoft. It is free software made by Softronics, Inc. and can be downloaded from their web site (www.softronix.com). The screen was set to 640 x 480 resolution and the text was always the same size. They were timed by the computer to see how long it took them to read text normally, rotated 90° clockwise, rotated 90° counterclockwise and with the subject's head horizontal to the left and right, aligned with the subject's head and aligned with the gravitational upright. The results were then stored in a file for analysis. I collected data in three locations: at home, at school, and at the hospital where my mother works.
I assigned numbers to each of the tests. Tests 0-2 were done with the subject's head in an upright position. Test 0 was reading text normally. Test 1 was done with the text rotated 90clockwise. Text in test 2 was rotated 90 counterclockwise. Tests 3 and 4 had the subject's head horizontal, tilted to the left. The text in test 3 was horizontal. The text in test 4 was rotated 90 counterclockwise. Tests 5 and 6 had the subject's head horizontal, tilted to the right. Test 5 was horizontal, test 6 was rotated 90clockwise.
Before testing, subjects were asked three questions. They were asked whether they were male or female, left or right handed, and how old they were. They were instructed to not move their heads during the reading.
I took text to display from two children's books, George and Martha Round and Round and George and Martha Rise and Shine, by James Marshall. I altered this text slightly, to make each page 30 words long. I wanted to use 30 words because it fit very nicely on the windows Logo screen, even when rotated. I did not use individual words. One reason is that many advanced readers recognize words as single objects, and I wanted to see how reading an entire page would be affected by rotation. I also wanted to get more of a difference between the times required to read each orientation.
This graph shows my basic results for all subjects. It displays percentage of the baseline reading speed (average of Tests 0, 4, and 6) achieved during mental rotation not aligned with gravity (MR-G, Tests 1 and 2) and aligned with gravity (MR+G, Tests 3 and 5). The dot in the center represents the average percentage of baseline reading speed achieved and the lines up and down represent this average plus and minus one standard deviation.
This graph contains data from 49 subjects. The probability of the results displayed here being completely random was determined to be less than 1 in 100,000 using a paired t test. This is expressed using a p- (probability) value, with lower values representing more significant results; thus for this result, p < 0.00001. This means that these results are very significant statistically. They show that there is a very definite effect of gravity on the ability of people to rotate text.
The ratio, R, of MR+G to MR-G reflects this effect of gravity on mental rotation. The average value of R for the 49 subjects was 1.18, indicating that mental rotation speed is increased by 18% on average if the text to be read is oriented horizontally with respect to gravity.
This graph shows differences between men and women. Again, the large dots represent averages and the lines represent plus and minus one standard deviation. The male data is colored blue; the female data is red.
The people included in my study were 22 females and 27 males. R is significantly different between the females and the males in my study, with a greater effect observed among the males (p < 0.06).
Although these results were quite different, I believe that, since I had mostly male students take my test, the two groups do not represent all ages well and this data may not be accurate.
In both groups, the effect of gravity, shown by the difference between MR-G and MR+G, was significant (p < 0.0001 for the females, p < 0.01 for the males).
This graph demonstrates the various differences between right handed (blue green), left handed (purple), and ambidextrous (dark yellow) people. The basic outline is the same as for the other graphs. This appears to show us that right handed people in general perform mental rotation tasks more rapidly than left handed people, although this observation may not be accurate because only 8 left handed subjects were tested. No conclusion can be drawn from my ambidextrous data since I had very few subjects (only 3!). The probability of the difference between MR-G and MR+G in this small group being completely random is about 50% (by a paired t test).
The effect of gravity on mental rotation is highly significant in the right-handed group (p < 0.0001) and also significant in the smaller left-handed group (p ~ 0.1).
This graph shows four age groups and the differences in their mental rotation skills. The groups are close to the same size. Group 1 contained subjects aged 10 and younger. This group had the highest MR-G index, and the second-lowest MR+G index of the four groups, and showed no significant difference (p ~ .37) between MR-G and MR+G (i.e., no significant effect of gravity on mental rotation). The other three groups all showed a significant effect of gravity on mental rotation (p < 0.1 for the 11-15 g roup, p < 0.001 for the 16-40 group, and p < 0.01 for the 41 and older group).
The three older groups showed an apparent decrease in mental rotation skills in both categories with increasing age, but only the changes in MR-G (general mental rotation ability) are significant (for example, the difference in MR-G between the 11-15 group and the 41 and older group is significant, with p < 0.01).
My experiment showed many things. First, it showed that it is equally easy to read text at the same orientation as your head, no matter what that orientation is. This means that it is just as easy to read text that is rotated when your head is rotated as it is to read upright text when your head is upright.
The next important thing my experiment showed was that there was no difference between clockwise and counterclockwise rotation. This means that people rotate text equally easily left and right.
Using these two observations, I simplified the seven variables into three:
I further simplified my results by normalizing the last two variables and created the following new variables:
I found the average of these two variables for all subjects and graphed them.
I found that, in all cases, it is easier to read text that is oriented with your head than in another orientation. I also found that it was easier to read rotated text if it was upright relative to gravity. My hypothesis stated that people would rotate text to be upright relative to gravity before reading. It appears now that people can use either the gravitational or retinal upright, but they prefer the retinal upright. An outline of the mental process could be as follows:
1. Read each letter horizontally relative to the eye and store the word.
2. Does it make sense?
Yes - use this word
No - continue
3. Get head orientation and find gravitational upright.
4. Read letters horizontally relative to gravity.
5. Does it make sense?
Yes - use this word
No - continue
6. Find angle of word.
7. Look at a letter and find orientation.
8. Get head orientation.
9. Is it closer to the head orientation than to gravity?
Yes - go to 10
No - go to 12
10. Rotate to head orientation.
11. Read word horizontally relative to head and stop.
12. Rotate to gravitational upright.
13. Read word horizontally relative to gravity and stop.
There were some differences in the different groups. I found that there was more of an advantage of having text upright relative to gravity in males than females. I also found that there was a decline in general mental rotation ability as people aged. I found no significant difference between right and left handed people.
Koriat, A and Norman, J (1984). What is rotated in mental rotation? Journal of Experimental Psychology: Learning Memory Cognition, 10 (3), 421-34
Koriat, A and Norman, J (1989). Why is word recognition impaired by disorientation while the identification of single letters is not? Journal of Experimental Psychology: Human Perception and Performance, 15 (1), 153-163
Friedman, A and Hall, DL (1996). The importance of being upright: use of environmental and viewer-centered reference frames in shape discriminations of novel three-dimensional objects. Memory & Cognition, 24 (3), 285-295
DeLisi, R and Cammarano, DM (1996). Computer experiences and gender differences in undergraduate mental rotation performance. Computers in Human Behavior, 12 (3), 351-361.
Suzuki, K and Nakata, Y (1988). Does the size of figures affect the rate of mental rotation? Perception and Psychophysics, 44(1), 76-80.
Uecker, A and Obrzut, JE (1993). Hemisphere and gender differences in mental rotation. Brain and cognition 22, 42-50.
Shepard, RN and Cooper, LA (1982). Mental Images and their Transformations. Cambridge, MA: MIT Press.
Marshall, J (1976). George and Martha Rise and Shine. Houghton Mifflin Company, Boston.
Marshall, J (1988). George and Martha Round and Round. Houghton Mifflin Company, Boston.
Press, WH, Teukolsky, SA, Vetterling, WT, and Flannery, BP (1992). Numerical Recipes in C (second edition). Cambridge, UK: Cambridge University Press.
Take a look at the results of my study in detail, and at the analysis of the results.
February 25, 1998
Return to my home page