This essay has been submitted by a student. This is not an example of the work written by our professional essay writers.
In the visual phenomenon known as crowding, observers identifying a target letter in peripheral vision find the task more challenging when flankers surround the target. But when flankers are removed, the task becomes easier. This suggests the 2-stage "bottom-up pooling of signals" theory; crowding occurs over a specific area in the brain where information is pooled in order to identify the target which is independent of other pooling areas. But when that region includes flankers, features of the target are combined with that of the flankers, hence leading to a jumbled up perception of the target.
Crowding occurs over various situations, and it has been suggested that the orientation (radial/tangential) of flankers or the positioning (up, down, left, right) of the target may have an effect on crowding. If this is so, as Levi et al (2009) discovered, a radial orientation of the flankers is expected to produce a larger effect on crowding. This prediction was confirmed in a group of 14 observers who correctly identified more target letters when flankers were arranged tangentially around the target, than radially, when a presentation of 48 randomly placed target letters in three different conditions were presented to them; 1) no flankers present (control), 2) flankers arranged tangentially in a specific meridian, and 3) flankers arranged radially in a specific meridian.
These results suggest that the 2 stage model of crowding should be extended to include orientation discrimination of flank letters to gain a further understanding of the processes behind crowding. (248 Words)
What is crowding?
Figure 1. Crowding. Crowding can be experienced by fixating on the round dot and trying to identify the letter 'A' when it is isolated (a), and when it is surrounded by two randomly placed flanking letters (b), where the latter would prove to be more difficult.
b.In his own words, Korte described crowding as: ''It is as if there is a pressure on both sides of the word that tends to compress it. Then the stronger, i.e. the more salient or dominant letters, are preserved and they 'squash' the weaker, i.e. the less salient letters, between them." (Korte, 1923). This extensively studied phenomenon is demonstrated in Figure. 1. When looking at an isolated letter (Figure. 1a), the letter is easily recognised in ones peripheral vision simply because the letter is quite large, and it is not placed very far away from the target. This means that the resolving power of the area surrounding the fovea was adequate enough to be able to identify the letter. However, if it is surrounded by other letters (Figure. 1b) the task of recognising the letter becomes more challenging.
Crowding appears to be more pronounced in the periphery (Bouma 1970; Toet & Levi, 1992) and is distinct from low-level feature interaction based processes such as lateral masking (Levi, Hariharan & Klein, 2002; Pelli, Palomares & Majaj, 2004; Chakravarthi & Cavanagh, 2007).
When does crowding happen?
Figure 2. Eccentricity. Critical Spacing & Flanker Letters. This is a general set-up when talking about letter recognition in crowding where the observer fixates on the fixation point and is asked to identify the target letter which is surrounded by a flanker letter on either side. Note that critical spacing should up to 0.5x the eccentricity of the target for crowding to occur.
CRITICAL SPACINGPrevious studies have proven that crowding takes place in a large variety of tasks such as letter recognition (Bouma, 1970; Flom et al., 1963; Toet & Levi, 1992), orientation discrimination (Andriessen & Bouma, 1976; Westheimer, Shimamura, &McKee, 1976); stereo acuity (Butler & Westheimer, 1978), contrast discrimination (Saarela, Sayim, Westheimer, & Herzog 2009), and face recognition (Louie, Bressler, & Whitney, 2007; Martelli, Majaj, & Pelli, 2005). In the peripheral vision, crowding is determined by eccentricity but is non dependant on the size of the target. In order to escape any crowding, the spacing of the target has to surpass the observer's critical spacing at that location in their visual field. I.e. 6mm at V1.
In the interest of letter recognition, crowding seems to only occur when the flanking letters are fairly close to the target letter. This distance between the flanking letters and the target letter is known as the critical spacing. In our central vision, crowding occurs only over a very small distance, roughly between 4-6 arc minutes (Flom et al.,1963a; Liu & Arditi, 2000; Toet & Levi, 1992), or is believed to not occur at all (Strasburger, Harvey, & Rentschler, 1991). In a study by Bouma (1970) he discovered that "for complete visual isolation of a letter presented at an eccentricity of É¸o, it follows that no other letters should be present (roughly) within 0.5 É¸o distance", or more simply put; the critical spacing between the target letter and the flankers should be up to half of the eccentricity of the target for crowding to have no effect. the 'Bouma law'. Critical spacing is equal for all objects. Furthermore, critical spacing at the cortex is independent of object position, and critical spacing at the visual field is proportional to object distance from fixation. The region where object spacing exceeds critical spacing is the uncrowded window. Observers cannot recognize objects that are outside this window. The uncrowded window limits how quickly people can read text and find an object in clutter.This is the point at which the retinal point spread functions of the target and the flankers are evidently separate. Eccentricity refers to the horizontal axis which is measured in degrees along the visual field. It is a measure in the visual field of how far away a given point (i.e. the target) is from the fixation point, which lies in the fovea (central vision). Generally, the aim of crowding experiments is to determine the critical spacing, and then compare it under different conditions. The critical spacing is known to be the distance at which flanker letters decrease performance. Figure 2 illustrates what eccentricity, critical spacing, and flanker letters are.
Why does crowding occur?
As to why crowding occurs is still hugely debated, but one likely reason is that neuronal convergence is involved. Mechanisms of the brain which are responsible for recognising letters, as well as other objects in the peripheral vision seem to pool over many spatial locations. Therefore, if there are extra letters present in the same region as the target letter, then all of the letters get mixed together in the brain, and so making the task of recognising the target letter a lot more difficult.
Even though crowding has been studied on a large scale using visual psychophysics for over 80 years (Bouma, 1970; Korte, 1923; Stuart & Burian, 1962), still to this day very little is known about its neural mechanisms. Various theories and ideas have been proposed based on psychophysical findings in order to explain crowding. One theory is the optics explanation by Hess, Dakin, & Kapoor, (2000); Liu & Arditi, (2000) in which crowding is due to the effect of the eye's point spread function when the stimuli are small and spaced very close together. Another theory is the receptive field theory which proposes that crowding occurs when a target and flankers fall within a single receptive field, causing pooling of those letters field (Flom, Heath, & Takahashi, 1963). Another idea associated with this theory is the crowding occurs due to a long range of horizontal connections between neurons that have similar tuning properties in the early visual cortex, for example V1 (Gilbert, 1998). Whereas on the other hand, theories about attention argue that crowding could be attributed to the coarse resolution of spatial attention (He, Cavanagh, & Intriligator, 1996, 1997; Intriligator & Cavanagh, 2001), or on unfocussed spatial attention (Strasburger, Harvey, & Rentschler, 1991; Strasburger, 2005). But none of these theories have been able to provide a reasonable explanation as to why crowding occurs in peripheral vision.
There are two general but important facts to be noted about crowding, which are: 1. crowding does not occur within the central fovea. 2. Critical spacing away from the central fovea is dependent on the distance of the target letter from the point of fixation, and as already mentioned above the critical spacing is roughly 50% of the eccentricity of the target.
In this study, we are looking to see whether crowding is effected in several conditions:
When the target letter is placed to the left or to the right (horizontal meridian) of the fixation point, or above or below (vertical meridian) the fixation point.
When the flankers are arranged in a radial or tangential orientation surrounding the target letter in each meridian.
We would expect to see that crowding is greater when the flankers are placed in a radial orientation in accordance to each meridian, as compared to when flankers are placed in a tangential orientation as crowding is said to not be isotropic in peripheral vision. Levi (2008) mentioned in his study that vertically arranged flankers are more damaging than horizontally arranged flankers. Toet and Levi (1992) discovered that crowding on an average extends from about 0.1x the eccentricity of the target in the tangential orientation to roughly 0.5x the eccentricity of the target in the radial orientation, and that irregularity also exists in the horizontal and vertical meridians. Crowding was also found to be much apparent when the target letter and flankers were arranged horizontally rather than vertically in a recent study (Feng, Jiang, & He, 2007), and that flankers had a much greater effect on orientation discrimination in that the amount of correct responses were reduced (He et al, 1996).
To test these predictions, we ran a study to see if letters could be recognised in the four cardinal meridians; up, down, right and left, with the flankers arranged in a tangential and radial orientation in each meridian. We compared the performance for letter recognition (percentage correct) for each observer in several different conditions as illustrated in figure 3. As a control, we presented a single target letter, not surrounded by any flankers in each meridian, which in theory each of the observers should identify correctly, leading us directly to our hypothesis:
Each predicted model was two-tailed where model 1a and 1b predicted that the two different orientations, (radial and tangential) for the up and down meridian would be opposite to that for the right and left meridian, where on the other hand model 2a and 2b predicted that the effect in each orientation (radial and tangential) would be the same for each meridian. From the table of analysis, it was seen that the experimental results followed the predicted model 2a, where the greatest amount of crowding occurred in the radial orientation for each cardinal meridian, thus we can accept our alternative experimental hypothesis and reject our null and experimental hypothesis.
(R = Radial, T = Tangential)
Figure 4? Percentage of correctly identified target letters in each meridian and orientation with error bars showing the bootstrapped standard errors when n = 14. Large error bars for the radial orientation in each meridian indicated a low confidence level for each of the values, where as overlapping that occurred between the down, right, and left meridian, meant the values were not statistically significant.
It is seen that there is quite large variation between the radial and tangential orientation when comparing the up and down meridian, but not as much between the right and left meridian. For the tangential orientation at each meridian, the majority of the target letters were correctly identified (mean average = >98%), indicating that crowding had a much lesser affect than in the tangential orientation, as compared to the radial orientation (mean average ranging between 58% - 88%).
Controls were used in each meridian where the target letter had no flankers. It was thought that each observer would have predicted every single target letter in this condition correctly, but this did not occur, which may have been due for several reasons. However, the majority of target letters were identified correctly in which we are relatively safe to say that flankers did have an effect on letter recognition in some shape or form.
Figure 5. Crowding was greater in the radial orientation than in the tangential orientation in all four of the cardinal meridians
In this study, two independent variables were compared in letter recognition. The first being whether placing the target letter in the four cardinal points had any effect on crowding, and second being if placing the flank letters in a tangential or radial orientation had any effect on crowding. Our analysis revealed that letter recognition were both qualitatively and quantitatively different under these conditions.
Analysing the first condition, to see if crowding varied in any of the four cardinal points, the 'Up' meridian had most crowding effect in comparison to the other three meridians, where the mean average of correctly identified letters was 83.95%. Coming in second place with the most amount of crowding was the 'Down' meridian where the average was 92.28%, the 'Left' meridian coming into third place with an average of 93.28%, and finally the right meridian with an average of 95.78%. This told us that crowding had a greater effect in the vertical meridians (Up & Down) than it did in the horizontal meridians (Right & Left).
One theory as to why radial orientation produced the most amount of crowding could be due to the fact that humans, mainly in the western world are taught to read from the left to the train (horizontally/tangential), thus making the task easier to recognise and distinguish the target from the flank letters as our brains have been trained and adapt to this organisation of letters.
Analysing the second condition as to whether the orientation of flank letters affected crowding was found to be true. Flankers which were arranged in a radial orientation produced a much greater effect on crowding than when arranged in a tangential orientation. This proved to be true in all four cardinal points. Crowding was greatest in the 'Up' meridian when flank letters were arranged radially (58.71%), with 'Down' coming in at second place (78.36%), followed by 'Left' (83.21%), and finally 'Right' (88.78%).
Thus from the results we can conclude that the orientation of flank letters has a much greater effect on crowding (increased crowding) when comparing it to see if the direction of target letter has any effect on crowding.
Studies by Bouma (1970) found that crowding is asymmetric in peripheral vision, and that when two flankers were placed on either side of the target letter, they had a stronger effect on crowding than a single flanker. He also noted that when a single flanker was placed at an eccentric point that was much further away from the target, crowding was more evident when compared to placing a single flanker at an eccentric point that was much closer to the fovea. In another study conducted by Bouma (1973), comparisons of letter recognition between the most inner letter; the letter closest to the fovea, and the most outer letter; the most peripheral letter were made where he discovered that the most peripheral letters had a higher recognition score than the letters closer to the fovea, in both the left and right visual field. This was contrary to what common sense would suggest as one would expect the inner letters to be more visible than the outer letters, which is at a greater eccentricity.
Other factors also effect crowding where crowding seems to be more evident and widespread when the target letters and flankers are similar.
Bouma (1970) discovered that when the target letter and the flankers are the same contrast polarity, such as when they both are black or both white, the effect of crowding is greater and more extensive than if the target letter and flankers are opposite polarity where for example the target letter is black and the flankers are white. This therefore shows strong evidence which supports both low level and high level theories for crowding.
As the target letter and the flank letter were produced in the same polarity in our study, may have had an effect on our results, and so if we were to reproduce the study again, having opposite polarity for the target letter and the flank letters maybe another condition to test and see if the same results are reproduced again.
Well established properties of crowding have been discovered in peripheral vision. When two flankers are arranged radially instead of tangentially, crowding is stronger (Levi & Carney, 2009; Toet & Levi, 1992), which has been found to be true of this study. However, when comparing two flankers on either side of the target or above and below in the central field, this is not found to be true. It can therefore be concluded that this characteristic is only specific to peripheral crowding.
Even though there have been several hypothesis put forward to explain crowding, there has still been a definite account of it yet. There have been two well known theories, which are the 'bottom-up pooling of signals' hypothesis, and the 'top-down attention account of crowding'. Parkes, Lund, Angelucci, Solomon & Morgan, (2001); Levi et al., (2002); Pelli et al., (2004), suggested that crowding may reflect the compulsory pooling of signals which posits a two stage process in identifying objects. The first stage is the 'feature detection' where the features of an object are independently collected, and the second stage is the 'feature integration' stage where this information is pooled in order to identify the target. This pooling is said to occur over some area which is independent of other such pooling regions, and when that region includes flankers, the features of the target are combined with the features of the flankers, and therefore leads to a perception that is all jumbled up. On the other hand, pooling may reflect the poor resolution of attention (He, Cavanagh & Intriligator, 1996; Intriligator & Cavanagh, 2001). The theory of the attention resolution states that at a given eccentricity, attention has a minimum available size of selection region, which a slightly larger than the smallest resolvable detail at that eccentricity. This is the reason as to why when several objects are placed within this region, that being able to identify the target can no longer be independently resolved.
Previous studies combining psychophysical and fMRI (functional Magnetic Resonance Imaging) adaptation techniques were conducted to search for the cortical locus of crowding. The psychophysical experiment founded that when an observer's attention was controlled, crowding did not affect the TEAE (threshold elevation after effect), and regardless of what the contrast level of the adapting stimulus was. On the other hand, in the fMRI experiment, the adaptation of the orientation-selective fMRI in V1 was also not affected, but downstream from V1 it was found that crowding undermined the effect of adaptation in V2 and V3. This demonstrated that crowding occurs beyond V1, supporting the two-stage model of crowding (D. M. Levi, 2008). The V4 region has also been projected to be a possible locus of crowding (Levi, 2008) due to its critical role in feature integration (Desimone & Duncan, 1995) and due to the similarities between the size/anisotropy of the receptive fields and the spatial extent of crowding (Pinon, Gattass, & Sousa, 1998; Toet & Levi,1992).
The rate at which an individual reads depends on the amount of letters they take in on each fixation. The spacing of letters in a piece of text is usually consistent but critical spacing of the observer increases with distance from the fixation point. Beyond a specific eccentricity point, the critical spacing of the observer exceeds that of the text, and so the letters crowd each other, distorting recognition. Peripheral vision beyond this eccentricity point is said to be crowded where observers are unable to read them, and central vision within this eccentricity is said to be uncrowded where observers are able to them.
The purpose of this study was to demonstrate how the orientation of flank letters has an effect on letter recognition, and if placing the target letter in a specific direction had an effect in letter recognition as well. Not only did the direction, but the orientation show an effect in letter recognition, where the orientation of the flank letters surrounding the target letter had a much greater effect on crowding.
Spacing between the letters, letter size, and eccentricity were kept at a constant throughout the study, but how do we know these did not have an impact on the outcome of the results? In other words, will changing the letter size to a smaller font size, smaller spacing between the letters, smaller eccentricity produce the same results in comparison to larger font size, larger spacing, and larger eccentricity, or to that obtained in the study?
Changes to the methodology of this study could be made to see if the results differ to what was found. Presenting the stimuli for 3 seconds may have been too long which may have made the task much easier for the observer to identify the target letter, thus resulting in an over estimation of correctly identified letters. The observer may have had a greater opportunity to move their eyes away from the fixation point to the target letter and identifying it correctly. But on the other hand, having such a long time for each stimuli may have set in some fatigue resulting in the observer not attending to the task very well, and so identifying more target incorrectly than they should have. we would therefore propose to present the stimuli for 1 second if we were to carry out the study again instead of for 3 seconds.
Carrying out the study in exactly the same way but presenting more conditions such as different font sizes, eccentricity, and critical spacing will provide more ground as to what effects crowding, and as to how it is affected.
Some letters such as I and J could have been presented next to each other, and due to them looking very familiar this could have confused the observer in recognising the target letter. Therefore for future experiments it may be ideal to remove letters which look similar such as EF, TL, OQ etc. It may have been possible that lateral inhibition between the letters suppressed the limbs so that the observers were unable to reach the level of decision making. On the other hand, the combination of 26 uppercase letters may have produced too many interaction patterns, and so using a small set of specifically designed stimulus may help to provide more definite results.
A larger range of observers (sex, age, ethnicities) will also provide more variation in results, and may help to identify groups of people in which crowding is greater or less.
Using a chin and head rest may prove to be useful, and help observers to fixate on the fixation point with more ease.
More repetitions of the study will enable us to measure more samples, thus reducing the standard error, allowing for more and more accurate estimations of the true mean by the mean of the experimental results. As the sample size increases, or if the experiment is repeated numerous times, it will enable the mean of our results to get closer and closer to the true mean, or to the mean of the whole population, and therefore using the mean as our best estimates of the unknown.
In the future, more studies complementary to crowding and the effects of orientation and positions, as well as psychophysical studies with brain imaging, and various kinds of stimuli are needed to be carried out in order to obtain a fuller understanding of crowding.
The main findings were:
Crowding is affected in the two conditions that were observed. However, crowding had a greater effect when flank letters were place in a radial orientation in comparison to a tangential orientation, and an even greater effect when comparing it to the direction in which the target letter and flank letters are placed.
Crowding in the vertical meridian had a greater effect than in the horizontal meridian.
An interesting study would be to carry out this study in exactly the same way but using observers from a Chinese/Japanese origin where they have been brought up to read from top to bottom (vertical/radial), and to compare their results with what we have found in this study. In theory, if crowding occurs in a specific orientation because our brains have adapted to the way we recognise letters and words from left to right, then observers who read from top to bottom would experience crowding in the horizontal and tangential orientation.