diff --git a/NAMESPACE b/NAMESPACE
index ec8164f..ad65708 100644
--- a/NAMESPACE
+++ b/NAMESPACE
@@ -1,5 +1,6 @@
 # Generated by roxygen2: do not edit by hand
 
+S3method(plot,giottoTree)
 S3method(print,giotto_plot_save_param)
 export("sankeyLabel<-")
 export(FSV_show)
diff --git a/NEWS.md b/NEWS.md
index 17a8a0a..f9edcf1 100644
--- a/NEWS.md
+++ b/NEWS.md
@@ -1,6 +1,7 @@
 # GiottoVisuals 0.2.16 (in development)
 
 ## enhancements
+- `plot()` on a `giottoTree` from `Giotto::calculateClusterTree()` gains `what = c("tree", "heatmap")`. `"tree"` is the plain `hclust` plot, as before; `"heatmap"` draws the correlation matrix the tree was built from, in leaf order with its branches on both axes, so the colours and branches always agree.
 - `showClusterHeatmap()` gains `cluster_custom_order`, to lay the clusters out in a given order, such as a cluster tree's leaf order (`tree$labels[tree$order]`). The clusters are then not clustered and no dendrogram is drawn, so the plot shows the correlations in that arrangement without claiming a structure. `expression_values` also accepts any expression name, not only `"normalized"` / `"scaled"` / `"custom"`.
 - image layers in spatial plots sample with `terra::spatSample(method = "display")`, which reads from the image's overviews when it has them, instead of `"regular"`, which ignored them. On a 10x JPEG-2000 `.ome.tif`, adding a whole-image layer went from ~2.3 s to ~0.2 s. Plotted pixel values now come from the overview level, so they are averaged rather than exact.
 - plot functions accept `view` and `space`: the name of a view or coordinate-frame recipe slotted on the object. The object is pre-narrowed through `GiottoClass::resolveRecipe()` once, so every slot read in a plot shares one resolver pass. The pass is scoped to the plot's own `spat_unit` / `feat_type` (GiottoClass >= 0.7.3), so on an object whose units do not share cell IDs, such as cells and nuclei, `spatPlot2D(view = )` no longer fails with "invalid extent" from a unit the view had narrowed to nothing. Added to `spatPlot2D()`, `spatInSituPlotPoints()`, `spatInSituPlotHex()`, `spatInSituPlotDensity()`, `spatDeconvPlot()`, `spatFeatPlot2D()` and `spatCellPlot2D()`. Plots that read no coordinates take `view` only, since a space changes nothing they draw: the heatmap / violin / ridge / dot / dendrogram family and `dimPlot2D()`, `dimFeatPlot2D()`, `dimCellPlot2D()`.
diff --git a/R/plot_cluster_tree.R b/R/plot_cluster_tree.R
index a925d5d..1b1d167 100644
--- a/R/plot_cluster_tree.R
+++ b/R/plot_cluster_tree.R
@@ -111,6 +111,66 @@ showClusterDendrogram <- function(gobject,
 }
 
 
+#' @title Plot a cluster tree
+#' @name plot.giottoTree
+#' @description Draw a `giottoTree` from `Giotto::calculateClusterTree()`,
+#' either as the tree itself or as the correlation matrix it was built from.
+#' @param x a `giottoTree`
+#' @param what `"tree"` (default) draws the dendrogram with the base `hclust`
+#' method. `"heatmap"` draws the correlation matrix recorded on the tree, in
+#' leaf order, with the tree's branches on both axes.
+#' @param ... passed to the base `hclust` plot method (e.g. `hang = -1`), or
+#' to `ComplexHeatmap::Heatmap()`
+#' @returns invisibly: `x` for `"tree"`, the `Heatmap` object for `"heatmap"`
+#' @details
+#' The heatmap shows the matrix **the branches were built from**, so the two
+#' always agree. It describes the tree, not whatever object is current: to see
+#' correlations recomputed from an object in a tree's order, use
+#' [showClusterHeatmap()] with `cluster_custom_order =
+#' tree$labels[tree$order]`, which draws no branches. For the tree with
+#' annotations and per-cluster evidence, see [plotClusterTree()].
+#' @examples
+#' g <- GiottoData::loadGiottoMini("visium")
+#' tree <- Giotto::calculateClusterTree(g, cluster_column = "leiden_clus")
+#' plot(tree, hang = -1)
+#' plot(tree, what = "heatmap")
+#' @method plot giottoTree
+#' @export
+plot.giottoTree <- function(x, what = c("tree", "heatmap"), ...) {
+    what <- match.arg(what)
+    if (identical(what, "tree")) {
+        # the plain hclust method, without `what` reaching its graphics args
+        y <- x
+        class(y) <- setdiff(class(x), "giottoTree")
+        graphics::plot(y, ...)
+        return(invisible(x))
+    }
+
+    cm <- attr(x, "cor_matrix")
+    if (is.null(cm)) {
+        stop("[plot.giottoTree] this tree carries no correlation matrix. ",
+            "To draw correlations from an object in its order, use ",
+            "`showClusterHeatmap(cluster_custom_order = ",
+            "tree$labels[tree$order])`.", call. = FALSE)
+    }
+    package_check(pkg_name = "ComplexHeatmap", repository = "Bioc")
+    p <- attr(x, "params")
+    title <- sprintf("%s correlation of cluster means%s",
+        p$cor %null% "", if (is.null(p$n_feats)) "" else
+            sprintf(", %d features", p$n_feats))
+    hm <- ComplexHeatmap::Heatmap(
+        matrix = cm[x$labels, x$labels, drop = FALSE],
+        cluster_rows = x,
+        cluster_columns = x,
+        name = "r",
+        column_title = trimws(title),
+        ...
+    )
+    ComplexHeatmap::draw(hm)
+    invisible(hm)
+}
+
+
 #' @title plotClusterTree
 #' @name plotClusterTree
 #' @description Publication figure for an annotated cluster tree: the
diff --git a/man/plot.giottoTree.Rd b/man/plot.giottoTree.Rd
new file mode 100644
index 0000000..62df40b
--- /dev/null
+++ b/man/plot.giottoTree.Rd
@@ -0,0 +1,38 @@
+% Generated by roxygen2: do not edit by hand
+% Please edit documentation in R/plot_cluster_tree.R
+\name{plot.giottoTree}
+\alias{plot.giottoTree}
+\title{Plot a cluster tree}
+\usage{
+\method{plot}{giottoTree}(x, what = c("tree", "heatmap"), ...)
+}
+\arguments{
+\item{x}{a \code{giottoTree}}
+
+\item{what}{\code{"tree"} (default) draws the dendrogram with the base \code{hclust}
+method. \code{"heatmap"} draws the correlation matrix recorded on the tree, in
+leaf order, with the tree's branches on both axes.}
+
+\item{...}{passed to the base \code{hclust} plot method (e.g. \code{hang = -1}), or
+to \code{ComplexHeatmap::Heatmap()}}
+}
+\value{
+invisibly: \code{x} for \code{"tree"}, the \code{Heatmap} object for \code{"heatmap"}
+}
+\description{
+Draw a \code{giottoTree} from \code{Giotto::calculateClusterTree()},
+either as the tree itself or as the correlation matrix it was built from.
+}
+\details{
+The heatmap shows the matrix \strong{the branches were built from}, so the two
+always agree. It describes the tree, not whatever object is current: to see
+correlations recomputed from an object in a tree's order, use
+\code{\link[=showClusterHeatmap]{showClusterHeatmap()}} with \code{cluster_custom_order = tree$labels[tree$order]}, which draws no branches. For the tree with
+annotations and per-cluster evidence, see \code{\link[=plotClusterTree]{plotClusterTree()}}.
+}
+\examples{
+g <- GiottoData::loadGiottoMini("visium")
+tree <- Giotto::calculateClusterTree(g, cluster_column = "leiden_clus")
+plot(tree, hang = -1)
+plot(tree, what = "heatmap")
+}
diff --git a/tests/testthat/test-plot-cluster-tree.R b/tests/testthat/test-plot-cluster-tree.R
index c288c80..957d8d1 100644
--- a/tests/testthat/test-plot-cluster-tree.R
+++ b/tests/testthat/test-plot-cluster-tree.R
@@ -189,3 +189,29 @@ test_that("plotClusterTree counts within the tree's view, and flags another", {
         "the tree was built with `view = \"v\"`"
     ), "clusters with no leaf")
 })
+
+
+test_that("plot(tree) draws the tree, and what = 'heatmap' its own matrix", {
+    skip_if_not_installed("ComplexHeatmap")
+    fx <- .pct_gobject()
+    tree <- .pct_tree(fx$gobject)
+    grDevices::pdf(NULL)
+    on.exit(grDevices::dev.off(), add = TRUE)
+
+    expect_no_warning(expect_identical(plot(tree, hang = -1), tree))
+
+    hm <- plot(tree, what = "heatmap")
+    expect_s4_class(hm, "Heatmap")
+    # the matrix the branches were built from, so the two cannot disagree
+    expect_identical(hm@matrix,
+        attr(tree, "cor_matrix")[tree$labels, tree$labels])
+    expect_identical(hm@row_dend_param$obj$merge, tree$merge)
+})
+
+
+test_that("a tree without its matrix says where to go instead", {
+    fx <- .pct_gobject()
+    tree <- .pct_tree(fx$gobject)
+    attr(tree, "cor_matrix") <- NULL
+    expect_error(plot(tree, what = "heatmap"), "cluster_custom_order")
+})