How to Easily Customize Point Shapes in ggplot2

The following demonstrates how to use the shape argument to specify the point type in a ggplot2 scatterplot:

ggplot(df, aes(x=x, y=y)) +
  geom_point(shape=19)

As noted, the default value for shape is 19 (a solid, filled circle), but you can specify any integer value ranging from 0 up to 25.

Exploring the ggplot2 Shape Index (0-25)

The numerical index system is fundamental to controlling point shapes in ggplot2. This system categorizes the available symbols into distinct groups based on their drawing properties. The shapes from 0 to 14 are generally simple outlines, while shapes 15 to 20 are solid, filled shapes. A special group, shapes 21 through 25, offers advanced control as they accept separate arguments for both border color (`color`) and fill color (`fill`), making them incredibly versatile for complex visualizations.

Understanding these indices is paramount for effective customization. Index values 0 through 5 correspond to basic geometric shapes like squares, circles, triangles, and diamonds, which are defined only by their outline. Shapes 6 through 13 introduce cross symbols, asterisks, and plus signs, often used when fill color is irrelevant. Indices 19 and 20 represent solid dots and large circles, respectively, which are robust choices for high-density plots where clarity is prioritized over internal detail.

The distinction between the different numerical ranges is crucial when planning your visualization strategy. If you need to map two separate categorical variables onto a single point—one influencing the border color and the other the internal fill color—you must select a shape between 21 and 25. These shapes (circle, square, diamond, up triangle, down triangle) are specifically designed to leverage both the `color` and `fill` aesthetics simultaneously, significantly increasing the information density of the visualization without sacrificing clarity.

To provide a comprehensive reference, the following R code generates a complete visual index of all 26 available point shapes (0 to 25) used by ggplot2, allowing easy identification of the appropriate index for any desired symbol:

library(ggplot2)

#create data frame mapping indices 0-25
df <- data.frame(x=0:25, y=0:25)

#create scatter plot showing all shapes
ggplot(df, aes(x=x, y=y)) +
  geom_point(shape=0:25, size=4)

Implementation Example 1: Using the Default Shape

While customized shapes are powerful, often the default aesthetic is sufficient for simple visualizations or for establishing a baseline reference before further customization. The default shape in ggplot2 is index 19, a solid, filled circle. When we construct a scatterplot without explicitly calling the `shape` parameter within geom_point(), the library automatically assigns this standard symbol to all data points. This is a crucial concept, as understanding the defaults allows for minimalist and efficient coding.

In this initial example, we will generate a simple scatter visualization using a synthetic dataset, deliberately omitting the `shape` argument. This ensures that the ggplot2 engine selects its standard visual representation. We will, however, specify the `size` parameter to ensure the points are clearly visible for demonstration purposes. Observe that the resulting plot exhibits the characteristic solid circle for every observation, confirming the default assignment.

This demonstration reinforces the principle that if an aesthetic is not explicitly mapped or set, ggplot2 relies on pre-defined standard values. For the geom_point() function, this default shape is highly readable and rarely needs modification unless categorical distinctions are required. The ability to rely on robust defaults simplifies the process of creating initial drafts of data visualizations in R.

The following R code demonstrates how to create a simple scatterplot using the default shape (index 19, the filled circle) for the points:

library(ggplot2)

#create data frame
df <- data.frame(x=0:25, y=0:25)

#create scatter plot with default point shape
ggplot(df, aes(x=x, y=y)) +
  geom_point(size=4)

Since we did not explicitly use the shape argument to specify a point shape, ggplot2 utilized the default shape, which is a solid, filled-in circle.

Implementation Example 2: Specifying a Custom Numeric Shape

The true power of the `shape` parameter is realized when we explicitly override the default setting to introduce a custom symbol. Choosing the appropriate shape is a design decision that impacts the clarity and interpretability of the graph. For instance, open shapes (indices 0-14) are often useful when plotting overlaid data or when aiming to reduce the visual weight of the points, allowing underlying lines or background elements to remain prominent.

In this second example, we will move beyond the default filled circle and select an alternative shape from the index. Specifically, we will choose index 2, which corresponds to an empty triangle pointing upwards. By assigning this numeric value directly to the `shape` argument within the geom_point() function, we instruct ggplot2 to render every point using this new specified symbol. This action ensures uniformity across the dataset layer.

It is important to remember that when using constant shapes (outside of `aes()`), the numeric index should be placed within the geom_point() function itself, not within the overall `aes()` mapping call. This distinction is critical: parameters set inside `aes()` are mapped to data variables, while parameters set outside `aes()` are fixed aesthetic properties for the entire geometry layer. Using an open shape like index 2 clearly demonstrates the visual contrast achievable through simple numeric selection.

The following R code illustrates how to create a scatterplot using an empty triangle (shape=2) for the point shape, providing a clear visual modification from the default:

library(ggplot2)

#create data frame
df <- data.frame(x=0:25, y=0:25)

#create scatter plot with custom point shape
ggplot(df, aes(x=x, y=y)) +
  geom_point(shape=2, size=4)

Conditional Shaping: Mapping Aesthetics to Variables

The most advanced and frequently used application of the `shape` parameter involves using aesthetic mapping to assign different shapes based on the categorical values of a variable within the dataset. This technique is invaluable for multivariate visualization, allowing the reader to instantly distinguish between different groups or categories directly within the plot area. By placing the `shape` argument inside the main `aes()` call, we tell ggplot2 to look up the data variable specified and automatically assign a unique shape index to each unique level of that variable.

When ggplot2 performs this conditional mapping, it intelligently selects distinct shapes from its index (0-25) and assigns them sequentially to the factor levels present in the specified variable. It is a best practice to also map the `color` aesthetic to the same variable. This practice ensures that not only are the groups distinguishable by shape (e.g., circles vs. triangles) but also by color, doubling the visual cue and significantly improving accessibility and interpretation.

It is important to consider the limitations of this automatic assignment. Since ggplot2 only has 26 unique shapes (and some are visually similar), this technique is best suited for variables with a small number of categorical levels (ideally 6 or fewer). If the categorical variable has too many unique values, the automatically assigned shapes may repeat or become visually indistinguishable, necessitating manual scale adjustments using functions like `scale_shape_manual()`.

In the following example, we create a dataset where points belong to one of three different categories (‘A’, ‘B’, or ‘C’). We then map both the `shape` and `color` aesthetics to the `team` variable, allowing the visual attributes of the points to dynamically reflect their underlying categorical assignment. This is a powerful demonstration of aesthetic mapping in action.

library(ggplot2)

#create data frame containing categorical variable 'team'
df <- data.frame(team=c('A', 'A', 'B', 'B', 'C', 'C'),
                 points=c(8, 11, 13, 15, 19, 25),
                 assists=c(4, 8, 7, 10, 11, 7))

#create scatter plot where point shape and color are based on the 'team' variable
ggplot(df, aes(x=points, y=assists, group=team)) +
  geom_point(aes(shape=team, color=team), size=4)

Notice that the shape and color of the points in the plot are both determined by the value for the team variable, demonstrating successful conditional aesthetic mapping. Because the aesthetic was mapped, ggplot2 also automatically produced a legend on the right side of the plot to clearly show which shapes and colors correspond to which team category.

Best Practices for Choosing Point Shapes

While ggplot2 provides a wide array of 26 shapes, not all are equally effective for data visualization. Selecting the right symbol is crucial for maximizing visual clarity and ensuring accurate data interpretation. A key rule of thumb is to prioritize shapes that are easily distinguishable, particularly when they are small or densely packed. Simple geometric shapes (like circles, squares, and triangles—indices 15, 16, 17) generally perform better than highly detailed or complex symbols (like asterisks or crosses).

When comparing groups, always aim for maximum contrast. For example, pairing a solid shape (like index 19 or 15) with an open shape (like index 1 or 0) often provides better differentiation than pairing two visually similar open shapes. If you are distinguishing more than three or four groups, relying solely on shape becomes challenging for the human eye. In such cases, combining shape mapping with color mapping (as shown in Example 3) is a necessary strategy to ensure robust visual coding.

Furthermore, consider the use of shapes 21 through 25, the highly customizable symbols. These open the door to advanced multivariate plotting where the border color can represent one variable (e.g., location) and the fill color can represent a second variable (e.g., performance metric). Using these shapes effectively prevents the need for excessive layering or facetting, consolidating information into a single, cohesive visualization. Always test your chosen shapes on the final plot size to confirm they remain legible and distinct.

Summary and Further Customization

Mastering the `shape` aesthetic in ggplot2 is essential for creating compelling and informative statistical graphics in R. We have explored the fundamental application of the geom_point() function, from setting a constant numeric shape using the index 0-25, to dynamically mapping the shape aesthetic to categorical variables within the dataset. The ability to control this visual element allows for sophisticated encoding of multivariate data, significantly enhancing the analytical value of your visualizations.

For those looking to achieve even finer control over their point shapes, ggplot2 offers scale modification functions. If the automatic shape assignment is not optimal, you can manually override the selected shapes using `scale_shape_manual(values = c(15, 17, 21))`, allowing you to define exactly which index corresponds to which category level. This level of customization is invaluable for maintaining consistent visual standards across multiple related plots.

Finally, remember that point shape interacts heavily with other aesthetics, particularly `size` and `color`. Ensuring a harmonious relationship between these properties is key to high-quality data presentation. Always consult the official documentation for the geom_point() function for comprehensive details on all available parameters and scaling options.

The following tutorials explain how to perform other common operations in ggplot2:

• Tutorial on Adjusting Point Size
• Guide to Using Fill vs. Color Aesthetics
• Mapping Multiple Variables in Scatterplots