ちょうど顕著な違い
English
Share
Overview
ソース: ジョナサン ・ Flombaum 講座-ジョンズ ・ ホプキンス大学
心理物理学、心理学の部門、どの物理量を説明しようとする神経が発火と大きさの心の表現に変換されます。可知差異 (JND) に関連する 1 セットのこの領域での質問: どのくらい何かが変更を知覚するために変更する必要がありますか?このことについて直感をポンプは、実際に小さな子供が相対的に言えば、巨大な速度で成長するが、1 つほとんど日常的に起こる成長の通知を検討してください。しかし、睡眠離れてキャンプまたは祖父母が長期不在の後、子供を見るときに、子が返されるときの成長のほんの数週間は知覚よりも。それは巨大に見えることができます!日常的に起こる小さな変化が知覚するのには小さすぎるので、高さの変更は不在の後気づいただけ。しかし、ぶり、多くの小さな変更追加します。どのくらいの成長は顕著な場所を取る必要があるか。最低限は、JND です。
心理学者および神経科学者は、多くのドメインで JND を測定します。どのくらい明るい光は、気づかれること必要がありますか。どの程度大きく音がする必要がありますか。彼らは多くの場合強制選択パラダイムを用いて測定を入手します。このビデオは、サイズ、図形の領域が変更されたときに、JND を測定するための標準的なアプローチを示すについて説明します。
Procedure
Results
The graph in Figure 3 shows the proportion of time in which the comparison stimulus was chosen as a function of the size of its radius. Recall that the constant stimulus always has a 10-px radius in this experiment. This is why with a radius of 5 or 6 px the comparison is almost never chosen, and it is almost always chosen with a radius if 14 or 15 px. However, with a radius of 9 or 11 px, the comparison is difficult. Participants often make mistakes. The JND is defined as follows: The comparison size when it is chosen about 75% of the time minus its size when it is chosen 25% of the time, all divided by 2. Here, those numbers are 12 and 8, respectively. So the JND for circle radius is 2 px.
There are detailed mathematical reasons for why this is the exact calculation of a JND, having to do with statistics and the nature of normal distributions (bell curves). But looking at the graph should make the computation more intuitive. When the radius was only 1 px smaller or bigger than 10, the participant made many mistakes, performing very near 0.5, which is what she would produce if she were just guessing. But performance quickly became far more accurate with a pixel difference of 2, and it was nearly perfect with a pixel difference of 3 or larger. Figure 4 is an annotated version of Figure 3, meant to illustrate the calculation of a JND.
Figure 4. An annotated version of Figure 3.
Applications and Summary
One of the main applications of the constant stimulus approach to measuring a JND has come in neuroscience, specifically in neurophysiology studies devised to investigate how the firing of individual neurons encodes physical properties about the world. These studies usually involve a monkey with electrodes implanted in their visual cortex. The electrodes penetrate individual cells that respond to visual stimulation by firing or spiking, that is, by conducting a rapid electrical signal. In studies on using JND methods, researchers have discovered that individual neurons are noisy-they respond to the size or brightness or color of a stimulus more or less the same way every time, but with some variability. The result is that two very similar stimuli will elicit the same response some of the time. A circle with a radius of 10 px will sometimes get the same neuronal response as a circle with a radius of 9 px or a circle with a radius of 11 px. This is why JND are just-barely-noticeable: sometimes, in the brain, the relevant stimuli really do produce indistinguishable effects.
Transcript
Exactly how much does something need to change for a difference to be perceived?
Think of, for instance, young children who grow rapidly—getting taller on a daily basis. However, it’s often difficult to notice subtle changes, especially if they still struggle to reach a basketball.
Over a much longer span, their growth spurt becomes more than perceptible; in fact, the amount can seem enormous! These changes in height are only noticed after a lapse because the small day-to-day differences are too small to be perceivable.
The minimal yet perceived amount is the just-noticeable-difference, which, for this example, is the smallest amount of growth noticed.
This video demonstrates a standard approach for measuring a just-noticeable-difference in shape size. Not only do we discuss the steps required to design and execute an experiment, but we also explain how to analyze the data and interpret the results describing just how small of a change in area is necessary to be perceived.
In this experiment, participants are briefly shown two different circles that vary in size and are forced to choose which one is larger.
During each trial, one is always presented with the same circumference, whereas the other is varied. This approach is referred to as the method of constant stimulus.
In this case, the constant stimulus is designed to have a radius of 10 px and located randomly on either the left or right side of the screen. In contrast, the other circle, called the comparison stimulus, will have a radius that varies between 5 and 9 and between 11 and 15 px.
Given these 10 possibilities, the comparison stimulus is shown 10 times on each side, for a total of 200 trials. The dependent variable is recorded as which stimulus was chosen to be the larger one.
Participants are expected to choose correctly if they perceived a difference in size between the two stimuli. However, when the shapes are closer in circumference and below the just-noticeable difference, performance is predicted to decline.
To begin the experiment, greet the participant in the lab. With them sitting comfortably in front of the computer, explain the task instructions: The screen will have the word “Ready?” on it until they press the space bar.
Watch as two blue stimuli appear and instruct the participant to indicate which stimulus they thought was larger by pressing the ‘L’ key for left- and ‘R’ for right-side responses. Remind them that they should guess if they are not sure which one is larger.
After answering any questions the participant might have, leave the room. Allow them to complete all of the 200 trials over a 5-min period. When they finish, return to the room and thank them for taking part in the experiment.
To analyze the data, first retrieve the programmed output file that captured each participant’s responses. Quickly glance at the data to make sure that performances were sensible—namely, that when the sizes of the comparison stimuli were 5 and 15 px, accuracy was near perfect.
Next, add a column to the output table called ‘Accuracy’ to determine whether the recorded answers are correct or not. Compare those given to the correct responses for all trials. Use the following IF statement to register a 1 when the response given was correct and 0 when it was incorrect.
Now, add another column to the table, labeled ‘Proportion of Comparison Responses’. Compare the column ‘Comparison Position’ with ‘Response’ and use a new IF statement to mark a ‘1’ when the comparison stimulus was chosen or a ‘0’ if the constant circle was chosen.
To visualize the results, make a scatter plot with the size of the comparison on the x-axis and the proportion of times it was chosen as being larger on the y-axis. Recall that the constant stimulus always had a 10-px radius, which is why stimuli with 5 or 6 px radii were almost never chosen and those with 14 or 15 were always chosen.
With a radius of 9 or 11 px, the comparison was more difficult and participants often made mistakes. In fact, performance was at chance level, suggesting that differences were not being perceived.
To calculate the just-noticeable-difference, take the comparison size that was chosen 75% of the time, in this case a radius of 12, minus the comparison size that was chosen 25% of the time—radius of 8—and divide the result by 2 for an answer of 2 px.
In other words, the radii of the circles need to differ by at least 2 px for their sizes to be accurately perceived.
Now that you are familiar with just-noticeable differences in the perception of visual objects’ sizes, let’s look at how this paradigm is used in neurophysiological studies to explore how the brain responds and in other behavioral situations, such as distinguishing between fat levels in food.
Researchers have investigated how individual neurons in the visual cortex encode the physical properties of the world, like objects’ sizes.
Using electrophysiological recording techniques that measure firing patterns in conjunction with stimuli presentation, researchers found that neurons that are sensitive to size will sometimes respond in the same way to objects that are actually different sizes.
This is why JND are just-barely-noticeable: sometimes, in the brain, the relevant stimuli really do produce indistinguishable effects.
In addition, researchers have used a just-noticeable-differences task to characterize individual thresholds for detecting fat concentrations in food.
They found that individuals with a higher body mass index required a higher just-noticeable difference, or higher threshold, before tasting fatty acids in the samples. These results could lead to new approaches to limit excess fat consumption.
You’ve just watched JoVE’s introduction to just-noticeable differences. Now you should have a good understanding of how to design and run the experiment, as well as how to analyze and assess the results.
Thanks for watching!