How to Easily Reorder Legend Items in ggplot2

Understanding Default Legend Behavior in ggplot2

By default, ggplot2 determines the order of legend entries based on the aesthetic mapping used (such as fill, color, or shape) and the type of variable assigned to that aesthetic. If the variable is a character string or an unordered factor, the items will usually be sorted alphabetically. If it is an ordered factor, the factor level hierarchy dictates the sequence. This automatic sorting is convenient for basic plots but becomes problematic when a specific logical or temporal flow must be maintained, such as showing “Low,” “Medium,” “High” or displaying teams in a championship rank order.

Effective data visualization requires that the legend complements the visual structure of the plot. If the visual elements (e.g., boxplots or bars) are ordered from left to right, the legend should ideally follow the same sequence to minimize cognitive load for the viewer. Therefore, mastering the techniques to manually impose a desired order is critical for producing publication-quality graphics.

Method 1: Using scale_fill_discrete() with the breaks Argument

The most direct method for overriding the default legend order in ggplot2 involves utilizing the scale functions, specifically scale_fill_discrete() (or scale_color_discrete() if using color) and specifying the desired order using the breaks argument. The breaks argument accepts a character vector that explicitly lists the category names in the exact sequence they should appear in the legend.

This technique is particularly powerful because it allows the visualization structure (e.g., the X-axis order) to remain unchanged while independently controlling the legend structure. However, it is important to ensure that the values provided to the breaks argument exactly match the category names present in the underlying data frame, otherwise the items might be excluded from the legend entirely.

You can use the following syntax to change the order of the items in a legend:

scale_fill_discrete(breaks=c('item4', 'item2', 'item1', 'item3', ...)

This function works by instructing ggplot2 to look only for the values listed within breaks and arrange them in that specified sequence within the legend key. The following examples demonstrate how to use this syntax in practice, focusing on a scenario involving multiple boxplots.

Example 1: Setting up the Initial Visualization

To illustrate the process of reordering the legend, we will first create a baseline plot. This plot uses a simple data frame containing points scored by three different teams (A, B, and C). Notice that in the initial output, ggplot2 orders the legend items alphabetically: A, B, C.

library(ggplot2)

#create data frame
df <- data.frame(team=c('A', 'A', 'A', 'A', 'B', 'B', 'B', 'B', 'C', 'C', 'C', 'C'),
                 points=c(6, 8, 13, 16, 10, 14, 19, 22, 14, 18, 24, 26))

#create multiple boxplots to visualize points scored by team
ggplot(data=df, aes(x=team, y=points, fill=team)) +
  geom_boxplot()

This initial plot uses the team variable for both the x-axis mapping and the fill aesthetic, which is responsible for generating the legend. Because ‘A’, ‘B’, and ‘C’ are simple character strings, the legend defaults to an alphabetical arrangement. Our goal in the next step is to rearrange this legend without altering the order on the x-axis.

Example 2: Applying scale_fill_discrete() to Reorder the Legend

To change the order of the items in the legend, we can append the scale_fill_discrete() function to the existing plot code. We pass a vector to the breaks argument, specifying a non-standard order, such as B, C, A. This action only modifies the legend key, leaving the rest of the plot untouched, demonstrating the precision of this technique.

library(ggplot2)

#create data frame
df <- data.frame(team=c('A', 'A', 'A', 'A', 'B', 'B', 'B', 'B', 'C', 'C', 'C', 'C'),
                 points=c(6, 8, 13, 16, 10, 14, 19, 22, 14, 18, 24, 26))

#create multiple boxplots to visualize points scored by team
ggplot(data=df, aes(x=team, y=points, fill=team)) +
  geom_boxplot() +
  scale_fill_discrete(breaks=c('B', 'C', 'A'))

Notice that the order of the items in the legend has successfully changed from the alphabetical default (A, B, C) to the specified sequence (B, C, A). This demonstrates the efficacy of the breaks argument in providing granular control over the legend display. This method is highly flexible and should be the preferred approach when only the legend order needs adjustment.

Method 2: Manipulating Factor Levels for Consistent Ordering

Although using scale_fill_discrete(breaks=...) is efficient for legend reordering, best practice in R often dictates that the underlying data structure should reflect the desired ordering. If you wish the order on the X-axis, the legend, and any subsequent statistical analysis to all follow a specific sequence, you should manipulate the factor levels of the categorical variable.

In R, categorical variables are often stored as factors. The order of levels within a factor determines how those categories are treated in plotting and modeling functions. To change the order globally, you would typically use the base R function levels() or, more commonly and robustly, functions from the forcats package (part of the tidyverse), such as fct_relevel(). For example, to set the levels to B, C, A across the entire dataset, the preparatory code would look like this:

library(dplyr)
library(forcats)

# Convert 'team' to factor and relevel it
df <- df %>%
  mutate(team = fct_relevel(team, 'B', 'C', 'A'))

# Now, plotting without specifying breaks will automatically use the new factor order
ggplot(data=df, aes(x=team, y=points, fill=team)) +
  geom_boxplot()

By modifying the factor levels, the resulting ggplot2 plot would automatically display B, C, then A on the X-axis and B, C, then A in the legend, ensuring complete consistency between all visual elements derived from that variable. This is generally preferred for data consistency, whereas the breaks argument offers a quick fix exclusively for the legend.

Example 3: Customizing Legend Labels with the labels Argument

In addition to controlling the order, it is often necessary to change the descriptive text associated with the legend items without changing the underlying data values. For instance, the original data might use short codes (like ‘A’, ‘B’, ‘C’), but the legend needs to display more informative text (like ‘A Team’, ‘B Team’). This customization is achieved by adding the labels argument within the same scale_fill_discrete() function.

Crucially, when using both breaks and labels, the vectors must be of the same length and must correspond positionally. If breaks lists c('B', 'C', 'A'), then the labels vector must have three elements, where the first label corresponds to ‘B’, the second to ‘C’, and the third to ‘A’.

library(ggplot2)

#create data frame
df <- data.frame(team=c('A', 'A', 'A', 'A', 'B', 'B', 'B', 'B', 'C', 'C', 'C', 'C'),
                 points=c(6, 8, 13, 16, 10, 14, 19, 22, 14, 18, 24, 26))

#create multiple boxplots to visualize points scored by team
ggplot(data=df, aes(x=team, y=points, fill=team)) +
  geom_boxplot() +
  scale_fill_discrete(breaks=c('B', 'C', 'A'),
                      labels=c('B Team', 'C Team', 'A Team'))

Notice that the legend labels have been updated to reflect the more descriptive titles (‘B Team’, ‘C Team’, ‘A Team’) while maintaining the custom order established by the breaks argument. This combined approach offers the ultimate flexibility in legend presentation, allowing for both reordering and relabeling within a single, coherent function call.

Advanced Considerations and Best Practices

When working with complex visualizations, remember that ggplot2 provides various scale functions depending on the aesthetic being mapped and the variable type. If you are mapping colors to a discrete variable, you use scale_color_discrete(). If you are using continuous data and mapping size, you would use a continuous scale function, which manages legend entries differently. For manually ordering discrete aesthetics like fill, color, or shape, the use of the breaks argument is consistently the most reliable way to enforce a custom legend sequence.

Furthermore, always prioritize consistency. If you use the breaks argument to reorder the legend, consider whether the visual elements on the plot itself (e.g., bars or points) should also follow that order. If they should, it is highly recommended to preprocess the data by explicitly setting the factor levels before plotting, ensuring that the visual mapping is intrinsically linked to the desired sequence.

Summary of Legend Reordering Techniques

To summarize the methods for controlling legend order in ggplot2, use the following guidelines:

• For Legend-Only Reordering: Use scale_fill_discrete(breaks = c(...)) or scale_color_discrete(breaks = c(...)). This is the fastest method when the X-axis order should remain the default or an independent order.

• For Global Consistency: Preprocess the variable using factor manipulation (e.g., fct_relevel or base levels()). This ensures the order is consistent across all derived visualizations and statistical summaries.

• For Custom Labels: Use the labels argument alongside breaks within the scale function to change the text displayed in the legend key while maintaining the desired sequence.

Mastery of these techniques allows R users to transform standard, automatically generated plots into custom, highly informative data visualizations that effectively guide the audience through the intended analytical narrative. Controlling the legend order is a fundamental step toward professional data reporting.

Further ggplot2 Resources

The following tutorials explain how to perform other common operations in ggplot2, helping you advance your data visualization skills:

• Tutorial on modifying colors in ggplot2.

• Guide to adjusting axis labels and titles.

• Documentation on controlling facet wraps.