diff --git a/SRC/cgedmd.f90 b/SRC/cgedmd.f90
index 9cc23e81e..67de2db1c 100644
--- a/SRC/cgedmd.f90
+++ b/SRC/cgedmd.f90
@@ -117,6 +117,8 @@
 !>    'Y' :: The data snapshots matrices X and Y are multiplied
 !>           by a diagonal matrix D so that Y*D has unit
 !>           nonzero columns (in the Euclidean 2-norm)
+!>           If all columns of Y are zero, the procedure returns
+!>           with INFO = -10.
 !>    'N' :: No data scaling.
 !>    \endverbatim
 !.....
@@ -467,7 +469,8 @@
 !>    \verbatim
 !>    INFO (output) INTEGER
 !>    -i < 0 :: On entry, the i-th argument had an
-!>              illegal value
+!>              illegal value. If JOBS == 'Y' and all columns
+!>              of Y are zero, INFO = -10.
 !>       = 0 :: Successful return.
 !>       = 1 :: Void input. Quick exit (M=0 or N=0).
 !>       = 2 :: The SVD computation of X did not converge.
@@ -833,6 +836,7 @@ SUBROUTINE CGEDMD( JOBS, JOBZ, JOBR, JOBF,  WHTSVD,   &
           ! The columns of Y will be normalized.
           ! To prevent overflows, the column norms of Y are
           ! carefully computed using CLASSQ.
+          K = 0
           DO i = 1, N
             !RWORK(i) = SCNRM2( M, Y(1,i), 1 )
             SSUM  = ONE
@@ -868,8 +872,17 @@ SUBROUTINE CGEDMD( JOBS, JOBZ, JOBR, JOBF,  WHTSVD,   &
                END IF
             ELSE
                RWORK(i) = ZERO
+               K = K + 1
             END IF
          END DO
+         IF ( K == N ) THEN
+         ! All columns of Y are zero. Return error code -10.
+         ! (the 10th input variable had an illegal value)
+         K = 0
+         INFO = -10
+         CALL XERBLA('CGEDMD',-INFO)
+         RETURN
+         END IF
          DO i = 1, N
 !           Now, apply the same scaling to the columns of X.
             IF ( RWORK(i) >  ZERO ) THEN
diff --git a/SRC/dgedmd.f90 b/SRC/dgedmd.f90
index 642e2d61b..e356ffa32 100644
--- a/SRC/dgedmd.f90
+++ b/SRC/dgedmd.f90
@@ -115,6 +115,8 @@
 !>    'Y' :: The data snapshots matrices X and Y are multiplied
 !>           by a diagonal matrix D so that Y*D has unit
 !>           nonzero columns (in the Euclidean 2-norm)
+!>           If all columns of Y are zero, the procedure returns
+!>           with INFO = -10.
 !>    'N' :: No data scaling.
 !>    \endverbatim
 !.....
@@ -501,7 +503,8 @@
 !>    \verbatim
 !>    INFO (output) INTEGER
 !>    -i < 0 :: On entry, the i-th argument had an
-!>              illegal value
+!>              illegal value. If JOBS == 'Y' and all columns
+!>              of Y are zero, INFO = -10.
 !>       = 0 :: Successful return.
 !>       = 1 :: Void input. Quick exit (M=0 or N=0).
 !>       = 2 :: The SVD computation of X did not converge.
@@ -854,6 +857,7 @@ SUBROUTINE DGEDMD( JOBS, JOBZ, JOBR, JOBF,  WHTSVD,  &
           ! The columns of Y will be normalized.
           ! To prevent overflows, the column norms of Y are
           ! carefully computed using DLASSQ.
+          K = 0
           DO i = 1, N
             !WORK(i) = DNRM2( M, Y(1,i), 1 )
             SSUM  = ONE
@@ -889,8 +893,17 @@ SUBROUTINE DGEDMD( JOBS, JOBZ, JOBR, JOBF,  WHTSVD,  &
                END IF
             ELSE
                WORK(i) = ZERO
+               K = K + 1
             END IF
          END DO
+         IF ( K == N ) THEN
+         ! All columns of Y are zero. Return error code -10.
+         ! (the 10th input variable had an illegal value)
+         K = 0
+         INFO = -10
+         CALL XERBLA('DGEDMD',-INFO)
+         RETURN
+         END IF
          DO i = 1, N
 !           Now, apply the same scaling to the columns of X.
             IF ( WORK(i) >  ZERO ) THEN
diff --git a/SRC/sgedmd.f90 b/SRC/sgedmd.f90
index 68e85d821..60b177dc5 100644
--- a/SRC/sgedmd.f90
+++ b/SRC/sgedmd.f90
@@ -116,6 +116,8 @@
 !>    'Y' :: The data snapshots matrices X and Y are multiplied
 !>           by a diagonal matrix D so that Y*D has unit
 !>           nonzero columns (in the Euclidean 2-norm)
+!>           If all columns of Y are zero, the procedure returns
+!>           with INFO = -10.
 !>    'N' :: No data scaling.
 !>    \endverbatim
 !.....
@@ -501,7 +503,8 @@
 !>    \verbatim
 !>    INFO (output) INTEGER
 !>    -i < 0 :: On entry, the i-th argument had an
-!>              illegal value
+!>              illegal value. If JOBS == 'Y' and all columns
+!>              of Y are zero, INFO = -10.
 !>       = 0 :: Successful return.
 !>       = 1 :: Void input. Quick exit (M=0 or N=0).
 !>       = 2 :: The SVD computation of X did not converge.
@@ -854,6 +857,7 @@ SUBROUTINE SGEDMD( JOBS, JOBZ, JOBR, JOBF,  WHTSVD,  &
           ! The columns of Y will be normalized.
           ! To prevent overflows, the column norms of Y are
           ! carefully computed using SLASSQ.
+          K = 0
           DO i = 1, N
             !WORK(i) = DNRM2( M, Y(1,i), 1 )
             SSUM  = ONE
@@ -889,8 +893,17 @@ SUBROUTINE SGEDMD( JOBS, JOBZ, JOBR, JOBF,  WHTSVD,  &
                END IF
             ELSE
                WORK(i) = ZERO
+               K = K + 1
             END IF
          END DO
+         IF ( K == N ) THEN
+         ! All columns of Y are zero. Return error code -10.
+         ! (the 10th input variable had an illegal value)
+         K = 0
+         INFO = -10
+         CALL XERBLA('SGEDMD',-INFO)
+         RETURN
+         END IF
          DO i = 1, N
 !           Now, apply the same scaling to the columns of X.
             IF ( WORK(i) >  ZERO ) THEN
diff --git a/SRC/zgedmd.f90 b/SRC/zgedmd.f90
index 7e40a3f1b..2b84942e5 100644
--- a/SRC/zgedmd.f90
+++ b/SRC/zgedmd.f90
@@ -117,6 +117,8 @@
 !>    'Y' :: The data snapshots matrices X and Y are multiplied
 !>           by a diagonal matrix D so that Y*D has unit
 !>           nonzero columns (in the Euclidean 2-norm)
+!>           If all columns of Y are zero, the procedure returns
+!>           with INFO = -10.
 !>    'N' :: No data scaling.
 !>    \endverbatim
 !.....
@@ -467,7 +469,8 @@
 !>    \verbatim
 !>    INFO (output) INTEGER
 !>    -i < 0 :: On entry, the i-th argument had an
-!>              illegal value
+!>              illegal value. If JOBS == 'Y' and all columns
+!>              of Y are zero, INFO = -10.
 !>       = 0 :: Successful return.
 !>       = 1 :: Void input. Quick exit (M=0 or N=0).
 !>       = 2 :: The SVD computation of X did not converge.
@@ -833,6 +836,7 @@ SUBROUTINE ZGEDMD( JOBS, JOBZ, JOBR, JOBF,  WHTSVD,   &
           ! The columns of Y will be normalized.
           ! To prevent overflows, the column norms of Y are
           ! carefully computed using ZLASSQ.
+          K = 0
           DO i = 1, N
             !RWORK(i) = DZNRM2( M, Y(1,i), 1 )
             SSUM  = ONE
@@ -868,8 +872,17 @@ SUBROUTINE ZGEDMD( JOBS, JOBZ, JOBR, JOBF,  WHTSVD,   &
                END IF
             ELSE
                RWORK(i) = ZERO
+               K = K + 1
             END IF
          END DO
+         IF ( K == N ) THEN
+         ! All columns of Y are zero. Return error code -10.
+         ! (the 10th input variable had an illegal value)
+         K = 0
+         INFO = -10
+         CALL XERBLA('ZGEDMD',-INFO)
+         RETURN
+         END IF
          DO i = 1, N
 !           Now, apply the same scaling to the columns of X.
             IF ( RWORK(i) >  ZERO ) THEN