For each of 10 students, write two numbers: hours of video games last week and hours of sleep per night. Put games on the horizontal axis and sleep on the vertical axis. A student with 4 hours of games and 9 hours of sleep becomes the dot at (4,9); one with 15 hours of games and 6.5 hours of sleep becomes the dot at (15,6.5). Ten students make ten dots, and each dot keeps both of one student's numbers together.
Then read the cloud as a whole. Dots drifting down to the right mean more gaming goes with less sleep, a negative association; dots drifting up mean a positive one. A bend, like an arch, means the relationship is curved. A dot far from the rest, such as a student with 20 hours of games and 10 hours of sleep, is an outlier worth checking: it may be a typing mistake or a real exception.
A line graph also puts dots on two axes, but it joins them in time order because the same thing is measured again and again, like one city's temperature each hour. Scatter-plot dots are separate cases, so joining them would invent a path that does not exist. And if one variable is a category, like favorite sport, there is no number to place on an axis, so a bar graph fits that data instead.
Think of it like this. A scatter plot works like pins on a map. Each pin marks one place by two numbers, how far east and how far north. A scatter plot pins each person by their two measurements, and the pattern of pins shows how those measurements relate, just as pins crowded along a coast show where people live.
Core idea. A scatter plot shows how two numerical variables move together.