diff --git a/.coveragerc b/.coveragerc
deleted file mode 100644
index 013dd202..00000000
--- a/.coveragerc
+++ /dev/null
@@ -1,2 +0,0 @@
-[report]
-show_missing = True
diff --git a/mir_eval/separation.py b/mir_eval/separation.py
index 96570eaf..0201fe40 100644
--- a/mir_eval/separation.py
+++ b/mir_eval/separation.py
@@ -1,5 +1,11 @@
 # -*- coding: utf-8 -*-
 """
+.. warning::
+    The mir_eval.separation module is deprecated in mir_eval version 0.8, and will be removed.
+    We recommend that you migrate your code to use an alternative package such as sigsep-museval
+    https://sigsep.github.io/sigsep-mus-eval/
+    
+
 Source separation algorithms attempt to extract recordings of individual
 sources from a recording of a mixture of sources.  Evaluation methods for
 source separation compare the extracted sources from reference sources and
@@ -141,6 +147,7 @@ def _any_source_silent(sources):
     )
 
 
+@util.deprecated(version="0.8", version_removed="0.9")
 def bss_eval_sources(reference_sources, estimated_sources, compute_permutation=True):
     """
     Ordering and measurement of the separation quality for estimated source
@@ -251,6 +258,7 @@ def bss_eval_sources(reference_sources, estimated_sources, compute_permutation=T
         return (sdr, sir, sar, popt)
 
 
+@util.deprecated(version="0.8", version_removed="0.9")
 def bss_eval_sources_framewise(
     reference_sources,
     estimated_sources,
@@ -362,6 +370,7 @@ def bss_eval_sources_framewise(
     return sdr, sir, sar, perm
 
 
+@util.deprecated(version="0.8", version_removed="0.9")
 def bss_eval_images(reference_sources, estimated_sources, compute_permutation=True):
     """Compute the bss_eval_images function from the
     BSS_EVAL Matlab toolbox.
@@ -496,6 +505,7 @@ def bss_eval_images(reference_sources, estimated_sources, compute_permutation=Tr
         return (sdr, isr, sir, sar, popt)
 
 
+@util.deprecated(version="0.8", version_removed="0.9")
 def bss_eval_images_framewise(
     reference_sources,
     estimated_sources,
@@ -841,6 +851,7 @@ def _safe_db(num, den):
     return 10 * np.log10(num / den)
 
 
+@util.deprecated(version="0.8", version_removed="0.9")
 def evaluate(reference_sources, estimated_sources, **kwargs):
     """Compute all metrics for the given reference and estimated signals.
 
diff --git a/mir_eval/util.py b/mir_eval/util.py
index 4c9e2326..05778279 100644
--- a/mir_eval/util.py
+++ b/mir_eval/util.py
@@ -5,6 +5,8 @@
 
 import os
 import inspect
+import warnings
+from decorator import decorator
 
 import numpy as np
 
@@ -939,3 +941,22 @@ def midi_to_hz(midi):
         Frequency/frequencies in Hz corresponding to `midi`
     """
     return 440.0 * (2.0 ** ((midi - 69.0) / 12.0))
+
+
+def deprecated(*, version, version_removed):
+    """Mark a function as deprecated.
+
+    Using the decorated (old) function will result in a warning.
+    """
+
+    def __wrapper(func, *args, **kwargs):
+        """Warn the user, and then proceed."""
+        warnings.warn(
+            f"{func.__module__}.{func.__name__}\n\tDeprecated as of mir_eval version {version}."
+            f"\n\tIt will be removed in mir_eval version {version_removed}.",
+            category=FutureWarning,
+            stacklevel=3,  # Would be 2, but the decorator adds a level
+        )
+        return func(*args, **kwargs)
+
+    return decorator(__wrapper)
diff --git a/setup.cfg b/setup.cfg
index 1426ed2c..d2973d47 100644
--- a/setup.cfg
+++ b/setup.cfg
@@ -1,6 +1,13 @@
 [tool:pytest]
 addopts = --cov-report term-missing --cov mir_eval --cov-report=xml --mpl --mpl-baseline-path=baseline_images/test_display
 
+[coverage:report]
+show_missing = True
+
+[coverage:run]
+omit =
+    separation.py
+
 [pydocstyle]
 # convention = numpy
 # Below is equivalent to numpy convention + D400 and D205
@@ -48,13 +55,13 @@ keywords = audio music mir dsp
 install_requires =
     numpy >= 1.15.4
     scipy >= 1.4.0
+    decorator
 
 [options.extras_require]
 display =
     matplotlib >= 3.3.0
 tests =
     matplotlib >= 3.3.0
-    decorator
     pytest
     pytest-cov
     pytest-mpl
diff --git a/tests/test_display.py b/tests/test_display.py
index 943a6674..415e7a1a 100644
--- a/tests/test_display.py
+++ b/tests/test_display.py
@@ -47,7 +47,7 @@ def test_display_segment():
     mir_eval.display.segments(intervals, labels, text=False)
 
     # Draw a legend
-    plt.legend()
+    plt.legend(loc="upper right")
     return plt.gcf()
 
 
@@ -126,7 +126,7 @@ def test_display_labeled_intervals_compare():
     )
     mir_eval.display.labeled_intervals(est_int, est_labels, alpha=0.5, label="Estimate")
 
-    plt.legend()
+    plt.legend(loc="upper right")
     return plt.gcf()
 
 
@@ -198,7 +198,7 @@ def test_display_hierarchy_nolabel():
     # Plot reference and estimate with a common label set
     mir_eval.display.hierarchy([int0, int1], [lab0, lab1])
 
-    plt.legend()
+    plt.legend(loc="upper right")
     return plt.gcf()
 
 
@@ -218,7 +218,7 @@ def test_display_hierarchy_label():
     # Plot reference and estimate with a common label set
     mir_eval.display.hierarchy([int0, int1], [lab0, lab1], levels=["Large", "Small"])
 
-    plt.legend()
+    plt.legend(loc="upper right")
     return plt.gcf()
 
 
@@ -238,7 +238,7 @@ def test_display_pitch_hz():
     # Plot pitches on a Hz scale
     mir_eval.display.pitch(ref_times, ref_freqs, unvoiced=True, label="Reference")
     mir_eval.display.pitch(est_times, est_freqs, unvoiced=True, label="Estimate")
-    plt.legend()
+    plt.legend(loc="upper left")
     return plt.gcf()
 
 
@@ -323,7 +323,7 @@ def test_display_multipitch_midi():
     mir_eval.display.multipitch(ref_t, ref_p, midi=True, alpha=0.5, label="Reference")
     mir_eval.display.multipitch(est_t, est_p, midi=True, alpha=0.5, label="Estimate")
 
-    plt.legend()
+    plt.legend(loc="upper left")
     return plt.gcf()
 
 
@@ -392,6 +392,7 @@ def test_display_ticker_midi_zoom():
     tolerance=6,
 )
 @pytest.mark.xfail(OLD_FT, reason=f"freetype version < {FT_VERSION}", strict=False)
+@pytest.mark.skip()
 def test_display_separation():
     plt.figure()
 
@@ -410,6 +411,7 @@ def test_display_separation():
     tolerance=6,
 )
 @pytest.mark.xfail(OLD_FT, reason=f"freetype version < {FT_VERSION}", strict=False)
+@pytest.mark.skip()
 def test_display_separation_label():
     plt.figure()
 
@@ -419,7 +421,7 @@ def test_display_separation_label():
 
     mir_eval.display.separation([x0, x1, x2], fs=fs, labels=["Alice", "Bob", "Carol"])
 
-    plt.legend()
+    plt.legend(loc="upper right")
     return plt.gcf()
 
 
@@ -439,7 +441,7 @@ def test_display_events():
     # Plot both with labels
     mir_eval.display.events(beats_ref, label="reference")
     mir_eval.display.events(beats_est, label="estimate")
-    plt.legend()
+    plt.legend(loc="upper right")
     return plt.gcf()
 
 
diff --git a/tests/test_separation.py b/tests/test_separation.py
index 4ad7f553..36fdc95f 100644
--- a/tests/test_separation.py
+++ b/tests/test_separation.py
@@ -26,6 +26,9 @@
 assert len(ref_files) == len(est_files) == len(sco_files) > 0
 file_sets = list(zip(ref_files, est_files, sco_files))
 
+# Skip separation tests since deprecation
+pytest.skip(allow_module_level=True)
+
 
 @pytest.fixture
 def separation_data(request):
@@ -119,8 +122,9 @@ def test_empty_input(metric):
         # And that the metric returns empty arrays
         assert np.allclose(metric(*args), np.array([]))
 
-        assert "reference_sources is empty" in str(record[0].message)
-        assert "estimated_sources is empty" in str(record[1].message)
+        # These warning counters are now offset by 1 because of the deprecation message
+        assert "reference_sources is empty" in str(record[1].message)
+        assert "estimated_sources is empty" in str(record[2].message)
 
 
 @pytest.mark.parametrize(