diff --git a/CHANGES.rst b/CHANGES.rst
index f2db8da89d2..ac2c375bc31 100644
--- a/CHANGES.rst
+++ b/CHANGES.rst
@@ -20,6 +20,12 @@ With thanks to Basile Pinsard for contributions.
   * MAINT: Finalize upstreaming of ANTs' interfaces to Nipype (#550)
   * MAINT: Update to Python +3.6 (#541)
 
+1.2.8 (September 03, 2020)
+==========================
+Bug-fix release in the 1.2.x series with a minor improvement of the correlations plot.
+
+* FIX: Improved control over correlations plot (#561)
+
 1.2.7 (August 12, 2020)
 =======================
 Bug-fix release in the 1.2.x series with a very minor improvement of the reportlets.
diff --git a/niworkflows/interfaces/plotting.py b/niworkflows/interfaces/plotting.py
index f23c37eabbb..8561dda7a1c 100644
--- a/niworkflows/interfaces/plotting.py
+++ b/niworkflows/interfaces/plotting.py
@@ -140,8 +140,12 @@ class _ConfoundsCorrelationPlotInputSpec(BaseInterfaceInputSpec):
         "which all correlation magnitudes "
         "should be ranked and plotted",
     )
+    columns = traits.List(
+        traits.Str,
+        desc="Filter out all regressors not found in this list."
+    )
     max_dim = traits.Int(
-        70,
+        20,
         usedefault=True,
         desc="Maximum number of regressors to include in "
         "plot. Regressors with highest magnitude of "
@@ -172,8 +176,9 @@ def _run_interface(self, runtime):
             self._results["out_file"] = self.inputs.out_file
         confounds_correlation_plot(
             confounds_file=self.inputs.confounds_file,
+            columns=self.inputs.columns if isdefined(self.inputs.columns) else None,
+            max_dim=self.inputs.max_dim,
             output_file=self._results["out_file"],
             reference=self.inputs.reference_column,
-            max_dim=self.inputs.max_dim,
         )
         return runtime
diff --git a/niworkflows/tests/test_confounds.py b/niworkflows/tests/test_confounds.py
index d11a3b442de..929a7c36193 100644
--- a/niworkflows/tests/test_confounds.py
+++ b/niworkflows/tests/test_confounds.py
@@ -160,6 +160,15 @@ def test_ConfoundsCorrelationPlot():
     """confounds correlation report test"""
     confounds_file = os.path.join(datadir, "confounds_test.tsv")
     cc_rpt = ConfoundsCorrelationPlot(
-        confounds_file=confounds_file, reference_column="a"
+        confounds_file=confounds_file, reference_column="a",
     )
     _smoke_test_report(cc_rpt, "confounds_correlation.svg")
+
+
+def test_ConfoundsCorrelationPlotColumns():
+    """confounds correlation report test"""
+    confounds_file = os.path.join(datadir, "confounds_test.tsv")
+    cc_rpt = ConfoundsCorrelationPlot(
+        confounds_file=confounds_file, reference_column="a", columns=["b", "d", "f"],
+    )
+    _smoke_test_report(cc_rpt, "confounds_correlation_cols.svg")
diff --git a/niworkflows/viz/plots.py b/niworkflows/viz/plots.py
index c5dea7282d4..dd38001b198 100644
--- a/niworkflows/viz/plots.py
+++ b/niworkflows/viz/plots.py
@@ -903,40 +903,55 @@ def compcor_variance_plot(
 
 
 def confounds_correlation_plot(
-    confounds_file, output_file=None, figure=None, reference="global_signal", max_dim=70
+    confounds_file,
+    columns=None,
+    figure=None,
+    max_dim=20,
+    output_file=None,
+    reference="global_signal",
 ):
     """
+    Generate a bar plot with the correlation of confounds.
+
     Parameters
     ----------
-    confounds_file: str
+    confounds_file: :obj:`str`
         File containing all confound regressors to be included in the
         correlation plot.
-    output_file: str or None
+    figure: figure or None
+        Existing figure on which to plot.
+    columns: :obj:`list` or :obj:`None`.
+        Select a list of columns from the dataset.
+    max_dim: :obj:`int`
+        The maximum number of regressors to be included in the output plot.
+        Reductions (e.g., CompCor) of high-dimensional data can yield so many
+        regressors that the correlation structure becomes obfuscated. This
+        criterion selects the ``max_dim`` regressors that have the largest
+        correlation magnitude with ``reference`` for inclusion in the plot.
+    output_file: :obj:`str` or :obj:`None`
         Path where the output figure should be saved. If this is not defined,
         then the plotting axes will be returned instead of the saved figure
         path.
-    figure: figure or None
-        Existing figure on which to plot.
-    reference: str
-        `confounds_correlation_plot` prepares a bar plot of the correlations
+    reference: :obj:`str`
+        ``confounds_correlation_plot`` prepares a bar plot of the correlations
         of each confound regressor with a reference column. By default, this
         is the global signal (so that collinearities with the global signal
         can readily be assessed).
-    max_dim: int
-        The maximum number of regressors to be included in the output plot.
-        Reductions (e.g., CompCor) of high-dimensional data can yield so many
-        regressors that the correlation structure becomes obfuscated. This
-        criterion selects the `max_dim` regressors that have the largest
-        correlation magnitude with `reference` for inclusion in the plot.
 
     Returns
     -------
     axes and gridspec
-        Plotting axes and gridspec. Returned only if `output_file` is None.
-    output_file: str
+        Plotting axes and gridspec. Returned only if ``output_file`` is ``None``.
+    output_file: :obj:`str`
         The file where the figure is saved.
     """
     confounds_data = pd.read_table(confounds_file)
+
+    if columns:
+        columns = set(columns)  # Drop duplicates
+        columns.add(reference)  # Make sure the reference is included
+        confounds_data = confounds_data[[el for el in columns]]
+
     confounds_data = confounds_data.loc[
         :, np.logical_not(np.isclose(confounds_data.var(skipna=True), 0))
     ]