Crowding is defined as a breakdown of object recognition and occurs when objects are too close together so that features from several objects are combined. As an illustration of crowding, one can try and identify a letter (target) to the right of another (fixation point) without looking at it directly. The distance between the target and flanking letters at which crowding occurs is called the ‘critical spacing’, i.e., how far must the flankers be from the target to allow perception of the target. Perception of the target is dependent on several factors. The objective of this experiment was to test the null hypothesis that flanker orientation (radial or tangential) had no effect on crowding when the target is placed at horizontal and vertical locations to the fixation point. The data suggested that the percentage of correct identifications was less for the radial configuration compared to the tangential configuration at all orientations with the possible exception of when the target was orientated to the left of the fixation point. Radial disadvantage under different conditions could be due to subjects have learned to read letters crowded together orientated from left to right. Data from crowding experiments may have implications for the study of age-related macular degeneration (AMD), amblyopia, and dyslexia (Levi, 2008), possible enabling a diagnostic test for dyslexia. [217 words]
Introduction
Best services for writing your paper according to Trustpilot
Crowding is defined as when objects are too close to each other and their features blend together (Pelli and Tillman, 2008). Hence, crowding impairs our ability to recognize objects with insufficient spacing and so visual acuity is reduced (Levi, 2008). Why crowding occurs is still unknown, but it seems likely that neuronal convergence is involved. It is as though the brain mechanisms responsible for recognizing letters (and other objects) in peripheral vision must pool over many spatial locations. Thus, if extra letters are displayed in the same region as the target, then all the letters are cluttered together in the brain. It is very important to understand the concept of crowding as it may have important implications for the study of age-related macular degeneration (AMD), amblyopia, and dyslexia (Levi, 2008).
Figure . Crowding – X represents a fixation target, F represents a target letter (to be identified) and both T and G represent flanking letters which cause the crowding effect. The spacing between a flanker and the target letter (e.g. between F and T) is the critical spacing.
The first step in object recognition is feature detection. The second step involves the brain combining the detected features so that the object can be recognized. As an illustration of crowding, one can try and identify a letter to the right of another without looking at it directly. This task will be easy if the letters are large and the letter to the right is not placed very far away. However, if the letter to the right is flanked by two other letters the task becomes virtually impossible or at least very difficult especially when the flanking letters are close to the target letters. The distance between the target and the flanker which crowding occurs is called the ‘critical spacing’, i.e., how far must the flankers be from the target to allow perception of the target. There are two fairly undisputed facts about crowding. First, it is not found in the very centre of the visual field (i.e. within the central fovea). Second, outside the central fovea, the critical spacing depends upon the distance of the target letter from the fixation point. As a first approximation, the critical spacing is about 50% of the eccentricity of the target. In peripheral vision, a letter that is easily recognized on its own becomes unrecognizable if surrounded by other letters. In foveal vision crowding typically only occurs over very small distances (4-6 arc min., e.g. Flom et al., 1963a; Liu & Arditi, 2000; Toet & Levi, 1992) or is reported not to occur at all (Strasburger, Harvey, & Rentschler, 1991). In contrast, crowding in peripheral vision occurs over very large distances (up to half the eccentricity of the target (Bouma, 1970; Kooi, Toet, Tripathy, & Levi, 1994; Toet & Levi, 1992)) where the retinal pooling is greater. Extensive crowding also occurs in the central visual field of strabismic amblyopes.
Critical spacing is dependent on several factors. Hence, in the normal fovea the extent of ‘crowding” is proportional to stimulus size and cannot easily be distinguished from ordinary masking (Levi et al., 2002a).This suggests that the critical spacing is proportional to the signal size, keeping the signal at the same eccentricity (zero), and both the strength and extent of foveal ”crowding” can be predicted directly by the strength and extent of masking (Levi et al., 2002a).
The extent of crowding is also reported to be field dependent. He et al. (1996) report that flankers have a stronger effect on orientation discrimination (i.e., reduce percent correct responses more), and the ”resolution of attention” (the minimum spacing at which observers can
select individual items (Intriligator & Cavanagh, 2001) and is coarser in the upper visual field than in the lower field. Crowding also depends strongly on target/flanker similarity. This is likely to depend on segmentation (Wilkinson,Wilson, & Ellemberg, 1997), contour integration (Field
et al., 1993; May & Hess, 2007), compulsory pooling (Parkes,Lund, Angelucci, Solomon, & Morgan, 2001), feature binding (Neri & Levi, 2006; Treisman & Schmidt, 1982) and selective attention (He et al., 1996). Thus, while it is certain that crowding occurs in the cortex, the precise location of the effect is unknown. The data suggest the possibility of a location beyond V1.
Crowding may also be dependent on the location of the target. For example the target could be placed to the left, right, below or above the fixation point. In addition, the flankers could be arranged radically or tangentially to the fixation letter. Hence, the objective of this experiment was to address the following research questions. Is there an orientation effect (vertical vs. horizontal) for the configuration of flankers (model 1)? Or is there a radial/tangential effect for the configuration of flankers (model 2)? Or do neither of these manipulations produce an effect (model 0)? By identifying which model is correct, we are helping to characterize the neural pooling process that is thought to be responsible for crowding.
Materials and Methods:
Observers
Sixteen experimental observers (10 females, 6 males age range 18 to 40) drawn from the second year undergraduate class were used in this experiment.
Equipment
The experiment was carried out using a ruler and a PC computer (screen diameter 33cm). The screen was calibrated to achieve a 5° visual angle so that the targets did not fall within the blind spot. Powerpoint software (Powerpoint 2007) was used to set up the experimental task and excel to rabulate the results.
Experiments
Two experiments were carried out, one for an observer and one for an examiner so that the observer did not remember the order of the tasks. First, a pilot experiment was carried out in which X was off centre and viewing distance was determined by the observer. This was subsequently changed in the main experiment as some observers were too close to the fixation point making the target easier to recognize while some observer were too far away making the task more difficult. In the real experiment, the viewing distance was fixed at 1 cm per degree = 57 cm. In addition, in the pilot experimen, a test time of 35 s for each slide was used but in the real experiment was < 1 second, fast enough to prevent fatigue. In viewing the target, 'Force choice' was used, i.e., even if the subject could not see the target under a particular condition, they were asked to guess the target.
Procedure
The objective of the experiment was to determine whether orientation of the flanker (radial or tangential ) affects the identification of the target. The test also included a ‘no flanker’ condition, to investigate whether it was the presence of flankers that make it difficult or easy to identify the target letter. Twenty letters were used in the experiment, excluding the letters J, I, Q, O, D and C, as these letters look too similar and may confuse the observer. The font that was chosen was Arial Bold (size 32), which was almost the same as the one used in the Bailey-Lovie chart, and is available on almost all computers. Letters were numbered 1 to 20 and then a calculator was used to generate random letters to use for the flanker and fixation target (X = central target). In the experiment, 120 slides were used, 10 of each position (i.e., up- down- horizontal- vertical). And these were presented to the subject in a random order. The experiment was not repeated so as to control addaptation and to reduce the level of fatigue experienced by the subject. The observer was asked to identify the letter and the response was recorded as how many correct and incorrect responses. These measurements therefore comprise of the ‘dependent variable’. The independent variable comprises the conditions of the experiment, i.e., the orientation of the flankers. Each task was carried out 10 times by each observer and the data were expressed as a percentage of correct responses. The means of all 16 subjects were then calculated.
Data analysis
An experiment involving two treatments or groups can be carried out in two different ways, viz., the unpaired and the paired or repeated measures method. If the objective is to compare two independent groups, then the appropriate procedure is an unpaired ‘t’ test (Armstrong and Hilton, 2011). However, the experiment described in our scenario was carried out using a paired design; i.e., the measurements are made sequentially on the same observer. Hence, the differences between the radial and tangential of the flankers for each orientation were compared using the paired sample ‘t’ test. A probability of P < 0.05 was used to judge the significance of the results.
Results
The results averaged over subjects together with an estimate of the degree of variation in the sample mean (standard error) for each set of conditions are shown in Table 1 and in Fig 1. When the target was located above the fixation point, there was a highly significant difference between the radial and tangential location of the flankers (t = 5.22, P < 0.001); i.e. significantly fewer errors were made when the flankers were orientated tangentially to the target. Similar results were obtained when the target was located below (t = 3.29, P < 0.001) and to the right (t = 2.46, P < 0.029) of the fixation point. However, when the target was orientated to the left of the fixation point, the difference beteen the radial and tangential location of the flankers did not reach significance (t = 2.16, P > 0.05). In all positions of the target, performance on the tangential flanker task was similar to that with no flankers at all.
Up
Down
Right
Left
Degrees of freedom
13
13
13
13
t Stat
-5.222
-3.292
-2.46
-2.158
P (two tailed)
0.0002
0.0058
0.0287
0.0502
Preferred Orientation
T
T
T
T
Figure 2. Statistical analysis – Statistical results of the average of each direction with diagram below to ilustrate the position of the target letter (‘T’).
Theory
Up
Down
Left
1
Horizontal
T
T
R
Vertical
R
R
T
2
Radial
R
R
R
Tangential
T
T
T
3
Other
?
?
?
Figure 3. Table of predicted results – Theory one (Horizontal/ Vertical). Theory two (Radial/ Tangential). Theory three Shows any other possibilities that have not been considered so far.
Figure 4. Distribution of correct responses – The mean responses of the observers to the various conditions together with their standard errors.
Discussion
The objective of the of the experiment was to test the null hypothisis that there was no difference in the response of observers when flanking letters were arranged radially or tangentially when the target was located vertically or horizonatlly above the fixation point. The data suggested that the percentage of correct identifications was less for the radial configuration compared to the tangential configuration at all orientations with the possible exception of when the target was orientated to the left of the fixation point. Feng et al showed there was a difference between horizontal and vertical in crowding. By splitting the visual space into four, they found that flankers are more affective when positioned horizontally (Feng, Jiang, & He, 2007).
The better performance on the tangential task may be the result of our visual system being more adapted to perceive horizontal and vertical stimuli rather than radial stimuli. It is also possible that tangential stimuli are more effectively represented within the visual field compared with radial stimuli. However, in area V1 of the cerebral cortex, the cortex is divided into two kinds of slabs, one set for ocular dominance and one set for orientation (Hubel, 1988). The orientation slabs include all possible orientations of a target from vertical to horizontal and therefore the visual system does not favour any specific orientation. The better performance on the horizontal task could therefore be due the fact that we are trained to read letters close together horizontally from right to left. If this hypothesis is correct, it may be interesting to observe Chinese subjects on this task as thay would be trained to read symbols orientated from top to bottom.
Another aspect of the data appears to be that there is less difference between radial and tangential orientations of the flankers when the target is located to the left of the fixation point. This lack of significance could be due to the greater variability between the reponse of the subjects to present this task obscuring differences between the means. Nandy and Tjan (2007) observed that confusion was a key cause of crowding. However, in the fovea this does not occur due to the close receptor spacing.
In a future experiment it would be better to carry out the investigation under more controlled conditions, e.g., in a room with no windows in order to reduce glare from outside. In addition, the experiment could have been improved by presenting the slides in a different order or by using less slides but with more repeats per subject. Crowding is reduced when target and flankers differ in color (Kooi et al., 1994; Nazir, 1992) or when the target and flankers are the same color, but the target appears on a different colored background (Poder, 2007). Hence, it would also be possible to vary the colour of the target and flankers to make it easier for the subject to observe. Carrying out the experiment on more experimental subjects would also increase its statistical ‘power’ allowing smaller observer effects to be detected.
Conclusion
The objective of this experiment was to disprove the null hypothesis that flanker orientation (radial or tangential) has no effect on crowding when the target is placed at horizontal and vertical locations to the fixation point. The data suggested that the percentage of correct identifications was less for the radial configuration compared to the tangential configuration at all orientations, with the possible exception of when the target was orientated to the left of the fixation point.
Only further research will provide more information as to precisely which structures in the cortex are responsible for crowding and why. Therefore when the process of crowding is fully understood we will be able to use this concept more effectively, for example in the marketing industry and in the design of advertisements for display on posters or web pages in order to include as much information as possible but still keeping it legible for the reader. In addition, this knowledge may have implications for the study of age-related macular degeneration (AMD), amblyopia, and dyslexia (Levi, 2008), possibly enabling a more advanced diagnostic test for dyslexia.
