diff --git a/sklearn/model_selection/_validation.py b/sklearn/model_selection/_validation.py
index 1e5ea29740c00..1e8327330e107 100644
--- a/sklearn/model_selection/_validation.py
+++ b/sklearn/model_selection/_validation.py
@@ -23,9 +23,11 @@
 from ..utils import indexable, check_random_state, safe_indexing
 from ..utils.validation import _is_arraylike, _num_samples
 from ..utils.metaestimators import _safe_split
+from ..utils.metaestimators import if_delegate_has_method
 from ..externals.joblib import Parallel, delayed, logger
 from ..externals.six.moves import zip
 from ..metrics.scorer import check_scoring, _check_multimetric_scoring
+from ..metrics.scorer import _passthrough_scorer
 from ..exceptions import FitFailedWarning
 from ._split import check_cv
 from ..preprocessing import LabelEncoder
@@ -521,10 +523,50 @@ def _score(estimator, X_test, y_test, scorer, is_multimetric=False):
     return score
 
 
+class _MemoizedPredictEstimator:
+    def __init__(self, estimator):
+        self.estimator = estimator
+
+    def fit(self, X, y):
+        self.estimator.fit(X, y)
+
+    @if_delegate_has_method(delegate='estimator')
+    def predict(self, X):
+        if not hasattr(self, '_predictions'):
+            self._predictions = self.estimator.predict(X)
+        return self._predictions
+
+    @if_delegate_has_method(delegate='estimator')
+    def decision_function(self, X):
+        if not hasattr(self, '_decisions'):
+            self._decisions = self.estimator.decision_function(X)
+        return self._decisions
+
+    @if_delegate_has_method(delegate='estimator')
+    def predict_proba(self, X):
+        if not hasattr(self, '_probs'):
+            self._probs = self.estimator.predict_proba(X)
+        return self._probs
+
+    @if_delegate_has_method(delegate='estimator')
+    def predict_log_proba(self, X):
+        if not hasattr(self, '_log_probs'):
+            self._log_probs = self.estimator.predict_log_proba(X)
+        return self._log_probs
+
+    @if_delegate_has_method(delegate='estimator')
+    def score(self, *args, **kwargs):
+        return self.estimator.score(*args, **kwargs)
+
+
 def _multimetric_score(estimator, X_test, y_test, scorers):
     """Return a dict of score for multimetric scoring"""
     scores = {}
 
+    # Try wrapping the estimator in _MemoizedPredictEstimator
+    # If the estimator has a score, wrapping it will not do any harm
+    estimator = _MemoizedPredictEstimator(estimator)
+
     for name, scorer in scorers.items():
         if y_test is None:
             score = scorer(estimator, X_test)
diff --git a/sklearn/model_selection/tests/test_validation.py b/sklearn/model_selection/tests/test_validation.py
index dedb77026c544..a51be9c3cb6dd 100644
--- a/sklearn/model_selection/tests/test_validation.py
+++ b/sklearn/model_selection/tests/test_validation.py
@@ -38,6 +38,7 @@
 from sklearn.model_selection import learning_curve
 from sklearn.model_selection import validation_curve
 from sklearn.model_selection._validation import _check_is_permutation
+from sklearn.model_selection._validation import _multimetric_score
 
 from sklearn.datasets import make_regression
 from sklearn.datasets import load_boston
@@ -50,6 +51,7 @@
 from sklearn.metrics import precision_score
 from sklearn.metrics import r2_score
 from sklearn.metrics.scorer import check_scoring
+from sklearn.metrics.scorer import _check_multimetric_scoring
 
 from sklearn.linear_model import Ridge, LogisticRegression
 from sklearn.linear_model import PassiveAggressiveClassifier
@@ -219,6 +221,17 @@ def get_params(self, deep=False):
         return {'a': self.a, 'allow_nd': self.allow_nd}
 
 
+class CountCallPredictedEstimator:
+    def __init__(self):
+        self._n_predict_calls = 0
+        self._rng = np.random.RandomState(0)
+    def fit(self, X, y):
+        return self
+    def predict(self, X):
+        self._n_predict_calls += 1
+        return self._rng.randint(0, 2, size=X.shape[0])
+
+
 # XXX: use 2D array, since 1D X is being detected as a single sample in
 # check_consistent_length
 X = np.ones((10, 2))
@@ -1299,3 +1312,14 @@ def test_permutation_test_score_pandas():
         check_series = lambda x: isinstance(x, TargetType)
         clf = CheckingClassifier(check_X=check_df, check_y=check_series)
         permutation_test_score(clf, X_df, y_ser)
+
+
+def test_multiscore_memoizing():
+    # Check if memoizing works as expected in _multimetric_score
+    X, y = make_classification(n_samples=1000, random_state=0)
+    estimator = CountCallPredictedEstimator()
+    scorers, _ = _check_multimetric_scoring(estimator,
+                                            ['neg_mean_squared_error',
+                                             'neg_median_absolute_error'])
+    scores = _multimetric_score(estimator, X, y, scorers=scorers)
+    assert estimator._n_predict_calls == 1