ggplot2 scale_shape() in R: Map a Variable to Point Shapes

The scale_shape() function in ggplot2 maps a discrete (factor) variable to point shapes, so each level of the variable draws a different plotting symbol. It is the standard scale behind any chart that uses the shape aesthetic to distinguish groups.

scale_shape()                                         # default 6 shapes
scale_shape(name = "Species")                         # tune legend title
scale_shape_manual(values = c(16, 17, 15))            # pick shape codes by hand
scale_shape_manual(values = c("setosa" = 16, "virginica" = 17))  # named values
scale_shape_identity()                                # use column values as shape codes
scale_shape(solid = FALSE)                            # open (hollow) shapes only
scale_shape_discrete(name = "Group", labels = c("A","B","C"))    # custom labels

What scale_shape() does in one sentence

scale_shape() maps each level of a discrete variable to a unique point shape so groups are visually separable without color. ggplot picks shapes 16, 17, 15, 3, 7, 8 by default (filled circle, filled triangle, filled square, plus, asterisk in square, asterisk).

Syntax and arguments

scale_shape(name = waiver(), breaks = waiver(), labels = waiver(), limits = NULL, solid = TRUE, na.translate = TRUE, na.value = NA, guide = "legend"). The solid argument toggles between filled (default) and hollow shapes. There is no values argument on scale_shape() itself; to pick specific codes use scale_shape_manual().

The shape = Species mapping inside aes() tells ggplot which variable controls the symbol. scale_shape() then auto-picks one shape per factor level. With three Species, ggplot assigns 16, 17, 15.

Five common patterns

1. Default scale_shape() with a factor

The auto-assigned shapes are 16 (circle), 17 (triangle), 15 (square). They are filled by default. If you want different symbols, jump to scale_shape_manual() in pattern 2.

2. Manual shape codes with scale_shape_manual()

Use scale_shape_manual() whenever you need a specific symbol per group. The 26 R shape codes go from 0 (open square) to 25 (filled down triangle). Codes 21 to 25 are special: they accept both color (border) and fill (interior).

3. Open shapes with solid = FALSE

solid = FALSE swaps to open (hollow) shapes: 1 (open circle), 2 (open triangle), 0 (open square). Hollow shapes reveal overlap and work well on dark backgrounds. Bump stroke to make the outline readable.

4. Fill-aware shapes (21 to 25)

Shape codes 21 to 25 are the "filled with border" family. They listen to both color (border) and fill (interior). This pattern produces publication-grade points where each Species has a distinct symbol AND a distinct fill color.

5. Tidy legend with custom labels

The name argument renames the legend header. The labels argument relabels each level, useful for italicized scientific names or shortened display strings. Without labels, ggplot uses the raw factor levels.

scale_shape() vs scale_shape_manual() vs scale_shape_identity()

Common pitfalls

Pitfall 1: more than 6 levels. scale_shape() warns "The shape palette can deal with a maximum of 6 discrete values" and drops rows beyond six. Use scale_shape_manual(values = 0:nlevels(x)) or rethink the encoding.

Pitfall 2: fill ignored on shapes 16 to 20. Only codes 21 to 25 use the fill aesthetic. If you mapped fill but see no color inside the points, you picked the wrong shape family.

Pitfall 3: continuous variable passed to shape. aes(shape = mpg) errors with "A continuous variable cannot be mapped to the shape aesthetic." Cast to a factor first (factor(cut(mpg, 3))) or use color/size instead.

Try it yourself

Related ggplot2 functions

After mastering scale_shape(), look at:

• geom_point(): the main geom that uses the shape aesthetic
• scale_shape_manual(): pick shape codes per level
• scale_shape_identity(): when the data column already holds shape codes
• scale_color_brewer(): pair color with shape for accessible group encoding
• aes(): where you map shape = variable in the first place

For the full reference, see the official ggplot2 scale_shape page.

FAQ

What shape codes does ggplot2 use?

ggplot2 supports R's 26 plotting symbols, numbered 0 to 25. Codes 0 to 14 are line-drawn (no fill), 15 to 20 are solid filled, and 21 to 25 are filled-with-border (they respond to both color and fill). The default scale_shape() picks from 16, 17, 15, 3, 7, 8 in order. You can see all 26 visually with ggplot(data.frame(x = 0:25), aes(x %% 6, x %/% 6, shape = factor(x))) + geom_point(size = 5) + scale_shape_manual(values = 0:25).

How do I use more than 6 shapes in ggplot2?

The default scale_shape() palette tops out at 6 unique shapes. Pass scale_shape_manual(values = c(16, 17, 15, 3, 7, 8, 18, 19)) (or any list of valid codes 0 to 25) to extend it. If you need 7+ groups, consider whether shape is the right encoding; color or faceting usually reads better than tiny shape differences.

What is the difference between scale_shape and scale_shape_manual?

scale_shape() auto-assigns shapes from a fixed palette (max 6 groups). scale_shape_manual() lets you specify exactly which shape code each factor level gets via values = c(...). Use scale_shape() for quick exploratory plots and scale_shape_manual() for publication where the shape-to-group mapping must be deliberate.

Why are my filled shapes showing no fill color?

Only shape codes 21 to 25 listen to the fill aesthetic. Codes 16 to 20 are colored by the color aesthetic instead. If you set fill = Species and used shape 16, the fill is silently ignored. Switch to scale_shape_manual(values = c(21, 22, 24)) and the fill mapping will take effect.

Can I map a continuous variable to shape in ggplot2?

Not directly. ggplot2 errors with "A continuous variable cannot be mapped to the shape aesthetic" because shape is inherently discrete. Two workarounds: bin the variable into 3 to 6 categories with cut(x, 4) and pass the result, or use scale_shape_binned() which does the binning for you. For continuous encoding, prefer color or size.