Source: Laboratory of Jonathan Flombaum—Johns Hopkins University
Human vision depends on light-sensitive neurons that are arranged in the back of the eye on a tissue called the retina. The neurons, called the rods and cones because of their shapes, are not uniformly distributed on the retina. Instead, there is a region in the center of the retina called the macula where cones are densely packed, and especially so in a central sub-region of the macula called the fovea. Outside the fovea there are virtually no cones, and rod density decreases considerably with greater distance from the fovea. Figure 1 schematizes this arrangement. This kind of arrangement is also replicated in the visual cortex: Many more cells represent stimulation at the fovea compared to the periphery.
Figure 1. Schematic depiction of the human eye and the distribution of light-sensitive receptor cells on the retina. The pupil is the opening in the front of the eye that allows light to enter. Light is then focused onto the retina, a neural tissue in the back of the eye that is made of rods and cones, light-sensitive cells. At the center of the retina is the macula, and in the center of the macula is the fovea. The graph schematizes the density of rod and cone receptors on the retina as a function of their position. Cones, which are responsible for color vision, are found almost exclusively in the fovea. Rods, which support seeing in low-light conditions, are similarly clustered more heavily near the fovea, with quickly falling densities outside the macula.
The result: We see very well in the part of space that our eyes are pointing at directly, the part of space stimulating the fovea; but we actually don't see very well in the periphery. We don't really notice it though, because our eyes move around constantly, building a representation of space from many individual fixations.
One way to study the properties of peripheral vision is with a phenomenon known as crowding.1 Crowding refers to an inability to recognize objects in clutter, and we experience crowding especially strongly when objects are shown in the periphery. Figure 2a is an example in which you should be able to experience crowding: Look at the cross in the center, and see if you can report the letter that is in the middle of the pack on the right. It is probably pretty hard. Now in Figure 2b try to report the letter on the left. Much easier! In this figure the letter is not crowded and there is no clutter around it, so it is easier to recognize.
Figure 2a. Crowded stimulus. Fixate the cross in the center, and see if you can recognize the letter in the middle of the pack on the left. It should be difficult, because the letters are in the periphery, and the central letter is crowded by the letters around it.
Figure 2b. Uncrowded stimulus. This stimulus is identical to Figure 2a, except that the letter G is uncrowded-no other letters surround it. Even while fixating the cross, the letter should be easy to recognize, even though it is just as much in the periphery as the G in Figure 2a.
This video will demonstrate how to design and implement a crowding experiment with letters as stimuli.
1. Stimuli and Design
- Trials will involve all English consonants, written in black and shown in 36 pt. Helvetica font.
- Each trial of the experiment will begin with a fixation cross presented in the center of the display for 500 ms, followed by stimuli presented to the right or left of the cross for 500 ms, and finally followed by a response screen. Figure 3 schematizes the experimental sequence.
Figure 3. Trial sequence. The sequence of events in a single trial of the experiment is as follows: The participant fixates the central cross and presses the spacebar to initiate the trial. After 500 ms, the test stimulus appears either to the right or left of fixation, including three letters. The participant should avoid making eye movements, but should try to identify the letter in the center of the group. The letters disappear after 500 ms, at which point the participant is prompted to enter a responses reporting the letter that they saw in the middle of the group.
Figure 4. Variables for manipulation. There are two crucial variables that can be manipulated in a crowding experiment. The first is called the eccentricity, referring to the distance between the fixation cross and the target stimulus, in this case, the letter in the center of a group of three. The second variable is called the inter-stimulus spacing, which refers to the distance between the target letter and each of its neighbors. In an experiment, these two variables will usually be manipulated together, by a third derived variable called the relative spacing, defined as the ratio of the relative spacing to the eccentricity.
- The stimuli presented in each trial will always come in groups of three. Two variables will be manipulated in the experiment: (1) the distance between the fixation cross and the stimulus in the middle of the group of stimuli, a measure called the eccentricity; (2) the distance between the middle stimulus in the group and each of the other two, a measure called the inter-stimulus spacing. Figure 4 shows a sample trial and identifies these two variables.
- When preparing the experiment, these two variables will actually be controlled by a third variable derived from their ratio. We'll call that the relative spacing and define it as the inter-stimulus spacing divided by the eccentricity. Since it is a ratio defined over two variables with the same units, note that the relative spacing does not have units.
- Now, to sequence the experiment, program a set of 200 trials.
- Each trial will belong to one of four relative spacing conditions, with values of 0.25, 0.4, 0.5, and 0.75. In pixels, the following are the values we will use for the inter-stimulus spacing and the eccentricity in each condition, respectively: (100, 400); (50, 125); (75, 150); (600, 800). But note, any values can be used so long as they produce the desired ratios.
- Randomly intermix an equal number of trials from the four conditions. On each trial, randomly choose the three consonants to display, with the restriction that no letter is repeated in a single trial.
- Also program the experiment to show the stimuli to the right of fixation in half of the trials, and to the left of fixation in the other half.
- Be sure to program the experiment to report the relevant outputs, minimally for each trial: the trial number, the three letters shown, the relative spacing, the response supplied by the participant, and whether that response was right or wrong.
2. Running the Experiment
- To run the experiment, recruit participants.
- When a participant arrives in the lab, be sure to have her complete informed consent. Then seat the participant 60 cm from the computer monitor.
- For this experiment, it important to also have a chin rest.
- Instruct the participant to place their chin in the chin rest and to resist moving during trials. Let them know that the experiment is self-paced, and that they are free to move or rest between trials as they please.
- Explain the instructions to the participant as follows:
- "In this experiment, we are interested in investigating how easily we can recognize objects shown in the visual periphery. On each trial, I'd like you to fixate on the cross in the center of the screen, and do your best not to move your eyes. When you press the spacebar to start a trial, after 500 ms, three letters will appear on the screen either to the left or right of fixation. They will only remain on the screen for 500 ms, and they may be a little hard to see. Your task is to try your best to recognize the letter in the middle of the pack, that is, the letter presented between the two others. Remember, try not to move your eyes, even when the letters appear. After they disappear, enter the letter you saw in the middle using the keyboard. If you are unsure, just guess."
- Remain in the room with the participant while she completes a few trials. Then leave her to complete the experiment.
- Data analysis in this experiment is straightforward. Simply compute average letter recognition accuracy as a function of relative spacing. So just count up the number of trials in which the participant accurately recognized the central letter in each spacing condition, and divide by 50, since there were 50 trials for each spacing condition.
We don’t always see objects in our surroundings clearly, especially if they are located in the periphery of our visual field.
Throughout the day, an individual will move their eyes to look directly at different, distinct items—like elements of a painting in a museum.
When this happens, the object being closely studied—like an apple—is positioned in the middle of the observer’s field of view, and as a result is seen clearly.
In contrast, an item far off to the side of the apple—in this instance, a dog—is located in the periphery of the visual field, and is perceived as fuzzy.
Such haziness can actually worsen in a phenomenon called visual crowding. Here, if the peripheral object is surrounded by "clutter"—like more canines a rogue artist decided to paint—it will be unrecognizable.
In this video, we explore how to investigate crowding using a letter-based approach. We not only explain how to design stimuli and collect and interpret peripheral vision data, but we also note how researchers are studying the concept in other contexts—like how it affects driving safety.
In this experiment, participants are asked to identify letters that are delivered to their peripheral visual field.
This is accomplished by first having them focus on a small fixation cross presented in the middle of a computer monitor, after which the letter stimuli appear.
During this phase, a row of three equally-spaced, capitalized English consonants—like JXW—are shown to one side of the cross, and thus are seen only with peripheral vision. Vowels are specifically excluded, as they could form short words that might interfere with data collection.
Although all of these characters are in the same size and type of font, two key variables are manipulated in these stimuli—eccentricity and inter-stimulus spacing—to better understand the factors that affect crowding.
The first, eccentricity, is the distance in pixels from the central letter to the intersection of the lines in the cross, which relates to where in the periphery of the visual field stimuli are delivered; the higher the eccentricity, the more peripheral the presentation of the letters.
In contrast, inter-stimulus spacing is defined as the distance—also in pixels—between the central consonant and each of the letters that abut it. This measurement assesses how close the flankers need to be to the middle character, to provide the clutter necessary for visual crowding.
The numbers chosen for these two distances in any given stimulus are actually controlled by a third variable defined as the ratio of inter-stimulus spacing to eccentricity—called the relative spacing.
Four different values—0.25, 0.4, 0.5, and 0.75—are tested to specifically assess whether inter-stimulus spacing needs to be a certain size, in relation to eccentricity, to have a crowding effect on peripheral vision.
For example, if a letter trio has a relative spacing value of 0.75, this means that the inter-stimulus distance is three-fourths the size of the eccentricity. So, the flanking consonants would be located relatively far away from the central character.
After the stimulus disappears, a response screen is shown, which prompts participants to type in what they perceived as the central letter.
Two hundred such trials are performed, in which stimuli with different relative spacing values are equally—but randomly—presented.
Here, the dependent variable is the percentage of trials in which the middle letter is correctly identified.
Based on previous work, it is expected that participants will only be able to accurately recognize the central characters in stimuli with a relative spacing of 0.5 or greater.
Importantly, this indicates that inter-stimulus spacing needs to be at least half the size of eccentricity to prevent crowding, a stipulation known as Bouma’s rule.
Greet the participant when they arrive, and have them sign informed consent materials. Then, seat them in front of a computer monitor with a keyboard.
Place their chin in the apparatus positioned approximately 60 cm from the screen.
Continue to explain the task, emphasizing that the experiment is self-paced, and—to proceed—the spacebar must be pressed. Also note that if in any instance the participant is unsure of the identity of the central consonant, they should provide their best guess.
Then, watch as the participant performs several practice trials. For each, ensure that the fixation cross is presented for 500 ms, followed by it and the letters, for the same amount of time. Also check that the participant, when prompted, enters their responses by pressing a consonant button on the keyboard.
Once they understand the task, leave the room, and allow the participant to complete the 200 trials.
To analyze the data, for each relative spacing ratio, compute the percentage of trials in which participants correctly identified the central consonant.
Notice that, as relative spacing increased, accuracy improved. Specifically, when this ratio was 0.5, participants demonstrated a performance of 75%, and this value jumped to approximately 95% when the relative spacing was 0.75.
However, with a ratio of 0.4, participants only accurately recognized the central consonant in 20% of trials, and at 0.25 this value dropped to 5%—a frequency that roughly corresponds to chance, if the central letter was guessed randomly
Collectively, these results indicate that crowding only occurs if the inter-stimulus spacing is less than half the size of eccentricity—for example, either 25% or 40% of this distance, as tested here—an observation supporting Bouma’s rule.
Now that you know how the manipulation of eccentricity and inter-stimulus spacing can be used to study crowding, let’s take a look at other ways aspects of peripheral vision are being explored.
Crowding has also been looked at in relation to automobile safety, and whether the number of objects in an environment can influence what a driver sees.
Such work has determined that crowding—either several cars, traffic cones, or signs in an area—can cause a driver to not effectively perceive a pedestrian.
For example, a person may observe a blurry object amidst cars, and think it is a parked bike—until the item darts out into the street, and turns out to be a man running to catch a train.
Work on the limits of peripheral vision is encouraging researchers to come up with ways to improve pedestrian and driving safety, like creating clear, well-lit crosswalks.
In contrast, web developers are also applying what we know about peripheral vision to create effective pop-up ads.
Due to the effects of crowding on letters, such banners are designed to not contain a lot of text, since these words won’t be distinguishable when they appear on the side of a monitor—in an internet user’s peripheral vision.
Rather, these advertisements contain bright, moving elements that attract someone’s attention, and cause them to move their eyes and fixate on this promotion. Then—hopefully—the person will click on it and order whatever is being sold.
Up until now, we’ve focused on crowding in normal participants; however, researchers are also looking at whether this perceptual phenomenon is related to visual defects associated with certain diseases.
For example, some work has involved the presentation of closely-spaced letter trios to patients diagnosed with a neurodegenerative condition similar to Alzheimer’s. Importantly, these stimuli were shown in the middle of a computer monitor, and thus delivered to the center of the visual field.
Interestingly, fewer patients were able to name the central letter, compared to healthy controls.
Collectively, this work provides evidence for the expansion of crowding—normally only a problem in the periphery—into central vision, and offers a possible explanation for the reading difficulties some neurodegenerative patients experience.
You’ve just watched JoVE’s video on peripheral crowding. By now, you should understand how to manipulate letter spacing to investigate this phenomenon, and collect and interpret vision data. In addition, you should grasp how crowding is being applied to other areas, such as the design of pop-up ads.
Thanks for watching!
Figure 5 graphs accuracy as a function of relative spacing. As relative spacing got bigger, performance improved by a lot. What this means is that performance benefits when the inter-stimulus spacing is at least half as big as the eccentricity. In fact, the idea that spacing needs to be half as big as eccentricity to prevent crowding is known as Bouma's Rule, after the scientist who discovered how the ratio between inter-stimulus spacing and eccentricity controls crowding. When the ratio is 0.5, as shown in the graph, performance is usually around 75% or better. Below 0.5, accuracy is often close to chance. Note that even with a relative spacing of 0.4, performance in this experiment was less than 25%, and with a relative spacing of 0.25, it was close to random. There are 21 consonants in English, so guessing would produce the right answer nearly 5% of the time.
Figure 5. Results of the crowding experiment. Recognition accuracy was very poor, and at times close to chance in trials with a relative spacing less than 0.5. But for trials with spacing of 0.5 or greater, recognition was usually better than 75% accurate. 0.5 is generally the critical relative spacing that prevents crowding.
Now that you know the basics of running a crowding experiment, you can run an experiment to show that relative spacing is the crucial determinant of crowding. Here is how: Pick four eccentricity values, say 50, 100, 200, and 250 px. For each, identify the four inter-stimulus spacing values that will give you the relative spacing values from the previous experiment, i.e. 0.25, 0.4, 0.5, and 0.75. Now you have four different ways of producing the same relative spacing values but with different eccentricities. That's 16 conditions in total. Run an experiment with 50 trials of all 16 conditions, and plot the data as shown in the Figure 6. You should find that relative spacing is the crucial determinant of performance (as opposed to eccentricity).
Figure 6. Results of a second crowding experiment designed to contrast the effects of eccentricity and relative spacing on performance. The x-axis displays the four relative spacing values used, and the different color icons represent the different eccentricities. If eccentricity were the primary constraint on recognition, then icons with the same color would tend to group together in terms of recognition accuracy. But instead, accuracy seemed to be governed by relative spacing.
Applications and Summary
One reason that understanding crowding is important has to do with macular degeneration. Macular degeneration is a condition that mostly affects older adults, involving the degeneration of the macula, the densely populated part of the retina that includes the fovea. Macular degeneration is the leading cause of blindness in the US among people over 65. It leaves people heavily reliant on peripheral vision. Thus, research on crowding can help scientists to understand the limitations and affordances of peripheral vision, how to improve it, and generally how to design the environment to prevent crowding in important situations.
Understanding how crowding works also plays a role in how engineers, graphic designers, and web developers arrange many of the displays that people engage with on a daily basis. For example, when a pop-up or banner ad appears in your web-browser, it is often designed to catch your attention, but not be 100% readable or resolvable because of crowding-the people behind the scenes want you to move your eyes and look at the ad after it catches your attention.
1. Whitney, D., & Levi, D. M. (2011). Visual crowding: A fundamental limit on conscious perception and object recognition. Trends in cognitive sciences,15(4), 160-168.