diff --git a/NEWS b/NEWS
index 00ed86547..6c1faf0f0 100644
--- a/NEWS
+++ b/NEWS
@@ -6,6 +6,10 @@ New in spot 1.2.5a (not yet released)
 
   * Bug fixes:
 
+    - Remove one incorrect simplification rule for PSL discovered
+      via checks on random formulaes.  (The bug was very unlikely
+      to trigger on non-random formulas, because it requires a SERE
+      with an entire subexpression that is unsatisfiable.)
     - When the automaton resulting from the translation of a positive
       formula is deterministic, ltlcross will compute its complement
       to perform additional checks against other translations of the
diff --git a/doc/tl/tl.tex b/doc/tl/tl.tex
index 784f204b2..07e2fc95d 100644
--- a/doc/tl/tl.tex
+++ b/doc/tl/tl.tex
@@ -1570,8 +1570,9 @@ Here are basic the rewritings for the weak closure and its negation:
   \nsere{r_1;r_2}&\equiv \nsere{r_1}\AND\nsere{r_2}\quad\text{if~}\varepsilon\VDash r_1\land\varepsilon\VDash r_2\\
   \sere{b;r}&\equiV b\AND\X\sere{r}&
   \nsere{b;r}&\equiV (\NOT b)\OR\X\nsere{r}\\
-  \sere{b\STAR{};r}&\equiv b\W\sere{r}&
-  \nsere{b\STAR{};r}&\equiv (\NOT b)\M\nsere{r}\\
+% These two would be correct only if $r$ is satisfiable.
+%  \sere{b\STAR{};r}&\equiv b\W\sere{r}&
+%  \nsere{b\STAR{};r}&\equiv (\NOT b)\M\nsere{r}\\
   \sere{b\STAR{\mvar{i}..\mvar{j}};r}&\equiV \underbrace{b\AND \X(b\ldots}_{\mathclap{i\text{~occurences of~}b}}\AND\X\sere{b\STAR{\mvar{0}..\mvar{j-i}}\CONCAT r})&
   \nsere{b\STAR{\mvar{i}..\mvar{j}};r}&\equiV \underbrace{(\NOT b)\OR \X((\NOT b)\ldots}_{\mathclap{i\text{~occurences of~}\NOT b}}\OR\X\nsere{b\STAR{\mvar{0}..\mvar{j-i}}\CONCAT r}) \\
   \sere{b\STAR{\mvar{i}..\mvar{j}}}&\equiV \underbrace{b\AND \X(b\AND \X(\ldots b))}_{i\text{~occurences of~}b}&
diff --git a/src/ltltest/reduc0.test b/src/ltltest/reduc0.test
index 744f654c2..296d03e0d 100755
--- a/src/ltltest/reduc0.test
+++ b/src/ltltest/reduc0.test
@@ -1,5 +1,5 @@
 #! /bin/sh
-# Copyright (C) 2013 Laboratoire de Recherche et
+# Copyright (C) 2013, 2014 Laboratoire de Recherche et
 # Développement de l'Epita (LRDE).
 #
 # This file is part of Spot, a model checking library.
@@ -24,3 +24,9 @@ set -e
 run 0 ../reduc 0 'XFa & FXa' 'XFa'
 run 0 ../reduc 0 '(Xf W 0) | X(f W 0)' 'XGf'
 
+# Two incorrect reductions.  Those used
+# to reduce respectively to a W (b && !b) and !a M (b || !b).
+# But both are wrong.  The reduction  {a*;r} = a W r  seems only
+# valid if r has a non-empty language.
+run 0 ../reduc 0 '{a[*];{b && !b}}'
+run 0 ../reduc 0 '!{a[*];{b && !b}}'
diff --git a/src/ltltest/reduccmp.test b/src/ltltest/reduccmp.test
index 0ff861728..b557ab0f1 100755
--- a/src/ltltest/reduccmp.test
+++ b/src/ltltest/reduccmp.test
@@ -355,14 +355,14 @@ for x in ../reduccmp ../reductaustr; do
      run 0 $x '{a|b*|c|d*}<>->e' '((a | c) & e) | (e M b) | (e M d)'
      run 0 $x '{{[*0] | a};b;{[*0] | a};c;e[*]} <>-> f' \
 	      '{{[*0] | a};b;{[*0] | a}} <>-> X(c & (f | X(f M e)))'
-     run 0 $x '{a;b[*];c[*];e;f*}' 'a & X(b W (c W e))'
-     run 0 $x '{a;b*;(a* && (b;c));c*}' 'a & X(b W {a[*] && {b;c}})'
+     run 0 $x '{a;b[*];c[*];e;f*}' 'a & X({b*;c*;e})'
+     run 0 $x '{a;b*;(a* && (b;c));c*}' 'a & X({b*;(a* && (b;c))})'
      run 0 $x '{a;a;b[*2..];b}' 'a & X(a & X(b & X(b & Xb)))'
      run 0 $x '!{a;a;b[*2..];b}' '!a | X(!a | X(!b | X(!b | X!b)))'
-     run 0 $x '!{a;b[*];c[*];e;f*}' '!a | X(!b M (!c M !e))'
-     run 0 $x '!{a;b*;(a* && (b;c));c*}' '!a | X(!b M !{a[*] && {b;c}})'
-     run 0 $x '{(a;c*;d)|(b;c)}' '(a & X(c W d)) | (b & Xc)'
-     run 0 $x '!{(a;c*;d)|(b;c)}' '(X(!c M !d) | !a) & (X!c | !b)'
+     run 0 $x '!{a;c[*];e;f*}' '!a | X!{c[*];e}'
+     run 0 $x '!{a;b*;(a* && (b;c));c*}' '!a | X(!{b*;(a* && (b;c))})'
+     run 0 $x '{(a;c*;d)|(b;c)}' '(a & X{c*;d}) | (b & Xc)'
+     run 0 $x '!{(a;c*;d)|(b;c)}' '(X(!{c*;d}) | !a) & (X!c | !b)'
      run 0 $x '(Xc R b) & (Xc W 0)' 'b & XGc'
      run 0 $x '{{c*|1}[*0..1]}<>-> v' '{{c[+]|1}[*0..1]}<>-> v'
      run 0 $x '{{b*;c*}[*3..5]}<>-> v' '{{b*;c*}[*0..5]} <>-> v'
diff --git a/src/ltlvisit/simplify.cc b/src/ltlvisit/simplify.cc
index 6625ab8f9..1ddf29376 100644
--- a/src/ltlvisit/simplify.cc
+++ b/src/ltlvisit/simplify.cc
@@ -1621,10 +1621,10 @@ namespace spot
 
 		  // Some term does not accept the empty word.
 		  unsigned end = mo->size() - 1;
-		  // {b₁;b₂;b₃*;e₁;f₁;e₂;f₂;e₂;e₃;e₄}
-		  //    = b₁&X(b₂&X(b₃ W {e₁;f₁;e₂;f₂}))
-		  // !{b₁;b₂;b₃*;e₁;f₁;e₂;f₂;e₂;e₃;e₄}
-		  //    = !b₁|X(!b₂|X(!b₃ M !{e₁;f₁;e₂;f₂}))
+		  // {b₁;b₂;e₁;f₁;e₂;f₂;e₂;e₃;e₄}
+		  //    = b₁&X(b₂&X({e₁;f₁;e₂;f₂}))
+		  // !{b₁;b₂;e₁;f₁;e₂;f₂;e₂;e₃;e₄}
+		  //    = !b₁|X(!b₂|X(!{e₁;f₁;e₂;f₂}))
 		  // if e denotes a term that accepts [*0]
 		  // and b denotes a Boolean formula.
 		  while (mo->nth(end)->accepts_eword())
@@ -1633,9 +1633,7 @@ namespace spot
 		  while (start <= end)
 		    {
 		      const formula* r = mo->nth(start);
-		      const bunop* es = is_KleenStar(r);
-		      if ((r->is_boolean() && !opt_.reduce_size_strictly)
-			  || (es && es->child()->is_boolean()))
+		      if (r->is_boolean() && !opt_.reduce_size_strictly)
 			++start;
 		      else
 			break;
@@ -1681,21 +1679,6 @@ namespace spot
 				tail =
 				  multop::instance(multop::And, e, tail);
 			    }
-			  // {b*;f} = b W {f}
-			  // !{b*;f} = !b M !{f}
-			  else
-			    {
-			      const bunop* es = is_KleenStar(e);
-			      assert(es);
-			      const formula* c = es->child()->clone();
-			      if (doneg)
-				tail =
-				  binop::instance(binop::M,
-						  unop::instance(unop::Not, c),
-						  tail);
-			      else
-				tail = binop::instance(binop::W, c, tail);
-			    }
 			}
 		      mo->destroy();
 		      result_ = recurse_destroy(tail);