diff --git a/src/shady-css/parser.ts b/src/shady-css/parser.ts
index 8246686..9310b80 100644
--- a/src/shady-css/parser.ts
+++ b/src/shady-css/parser.ts
@@ -88,7 +88,8 @@ class Parser {
     } else if (token.is(TokenType.comment)) {
       return this.parseComment(tokenizer);
 
-    } else if (token.is(TokenType.word)) {
+    } else if (token.is(TokenType.word) ||
+               token.is(TokenType.openBrace)) {
       return this.parseDeclarationOrRuleset(tokenizer);
 
     } else if (token.is(TokenType.propertyBoundary)) {
@@ -241,13 +242,13 @@ class Parser {
    * @param tokenizer A Tokenizer node.
    */
   parseDeclarationOrRuleset(tokenizer: Tokenizer): Declaration|Ruleset|null {
-    let ruleStart = null;
-    let ruleEnd = null;
-    let colon = null;
+    if (!tokenizer.currentToken) {
+      return null;
+    }
 
-    // This code is not obviously correct. e.g. there's what looks to be a
-    // null-dereference if the declaration starts with an open brace or
-    // property boundary.. though that may be impossible.
+    let ruleStart = tokenizer.currentToken;
+    let ruleEnd = ruleStart.previous;
+    let colon = null;
 
     while (tokenizer.currentToken) {
       if (tokenizer.currentToken.is(TokenType.whitespace)) {
@@ -266,25 +267,15 @@ class Parser {
         if (tokenizer.currentToken.is(TokenType.colon)) {
           colon = tokenizer.currentToken;
         }
-
-        if (ruleStart === null) {
-          ruleStart = tokenizer.advance();
-          ruleEnd = ruleStart;
-        } else {
-          ruleEnd = tokenizer.advance();
-        }
+        ruleEnd = tokenizer.advance();
       }
     }
 
-    if (tokenizer.currentToken === null) {
-      // terminated early
-      return null;
-    }
-
     // A ruleset never contains or ends with a semi-colon.
-    if (tokenizer.currentToken.is(TokenType.propertyBoundary)) {
+    if (!tokenizer.currentToken ||
+        tokenizer.currentToken.is(TokenType.propertyBoundary)) {
       const nameRange =
-          tokenizer.getRange(ruleStart!, colon ? colon.previous : ruleEnd);
+          tokenizer.getRange(ruleStart, colon ? colon.previous : ruleEnd);
       const declarationName =
           tokenizer.cssText.slice(nameRange.start, nameRange.end);
 
@@ -298,12 +289,13 @@ class Parser {
             this.nodeFactory.expression(expressionValue, expressionRange);
       }
 
-      if (tokenizer.currentToken.is(TokenType.semicolon)) {
+      if (tokenizer.currentToken &&
+          tokenizer.currentToken.is(TokenType.semicolon)) {
         tokenizer.advance();
       }
 
       const range = tokenizer.trimRange(tokenizer.getRange(
-          ruleStart!,
+          ruleStart,
           tokenizer.currentToken && tokenizer.currentToken.previous ||
               ruleEnd));
 
@@ -313,16 +305,17 @@ class Parser {
     } else if (colon && colon === ruleEnd) {
       const rulelist = this.parseRulelist(tokenizer);
 
-      if (tokenizer.currentToken.is(TokenType.semicolon)) {
+      if (tokenizer.currentToken &&
+          tokenizer.currentToken.is(TokenType.semicolon)) {
         tokenizer.advance();
       }
 
-      const nameRange = tokenizer.getRange(ruleStart!, ruleEnd.previous);
+      const nameRange = tokenizer.getRange(ruleStart, ruleEnd.previous);
       const declarationName =
           tokenizer.cssText.slice(nameRange.start, nameRange.end);
 
       const range = tokenizer.trimRange(tokenizer.getRange(
-          ruleStart!,
+          ruleStart,
           tokenizer.currentToken && tokenizer.currentToken.previous ||
               ruleEnd));
 
@@ -330,16 +323,16 @@ class Parser {
           declarationName, rulelist, nameRange, range);
       // Otherwise, this is a ruleset:
     } else {
-      const selectorRange = tokenizer.getRange(ruleStart!, ruleEnd);
+      const selectorRange = tokenizer.getRange(ruleStart, ruleEnd);
       const selector =
           tokenizer.cssText.slice(selectorRange.start, selectorRange.end);
       const rulelist = this.parseRulelist(tokenizer);
-      const start = ruleStart!.start;
+      const start = ruleStart.start;
       let end;
       if (tokenizer.currentToken) {
         end = tokenizer.currentToken.previous ?
             tokenizer.currentToken.previous.end :
-            ruleStart!.end;
+            ruleStart.end;
       } else {
         // no current token? must have reached the end of input, so go up
         // until there
diff --git a/src/test/parser-test.ts b/src/test/parser-test.ts
index 0ed28d7..aded6c0 100644
--- a/src/test/parser-test.ts
+++ b/src/test/parser-test.ts
@@ -147,6 +147,42 @@ describe('Parser', () => {
             nodeFactory.discarded(';')
           ]));
     });
+
+    it('can parse empty selectors', () => {
+      expect(parser.parse('{ empty-a } { empty-b } empty-c { empty-d }'))
+          .to.containSubset(nodeFactory.stylesheet([
+            nodeFactory.ruleset(
+              '', nodeFactory.rulelist([
+                nodeFactory.declaration('empty-a', undefined)])),
+            nodeFactory.ruleset(
+              '', nodeFactory.rulelist([
+                nodeFactory.declaration('empty-b', undefined)])),
+            nodeFactory.ruleset(
+              'empty-c', nodeFactory.rulelist([
+                nodeFactory.declaration('empty-d', undefined)])),
+          ]));
+    });
+
+    it('can parse unclosed blocks', () => {
+      expect(parser.parse('uncl-a { uncl-b: uncl-c uncl-d'))
+          .to.containSubset(nodeFactory.stylesheet([
+            nodeFactory.ruleset(
+              'uncl-a', nodeFactory.rulelist([
+                nodeFactory.declaration(
+                  'uncl-b', nodeFactory.expression('uncl-c uncl-d'))
+              ]))
+          ]));
+    });
+
+    it('can parse unclosed blocks without a colon', () => {
+      expect(parser.parse('uncol-a { uncol-b'))
+          .to.containSubset(nodeFactory.stylesheet([
+            nodeFactory.ruleset(
+              'uncol-a', nodeFactory.rulelist([
+                nodeFactory.declaration('uncol-b', undefined)
+            ]))
+          ]));
+    });
   });
 
   describe('when extracting ranges', () => {
diff --git a/src/test/stringifier-test.ts b/src/test/stringifier-test.ts
index e6d7149..9d7834d 100644
--- a/src/test/stringifier-test.ts
+++ b/src/test/stringifier-test.ts
@@ -107,6 +107,14 @@ describe('Stringifier', () => {
       expect(cssText).to.be.eql(':root{--qux:vim;--foo:{bar:baz;};}');
     });
 
+    it('can stringify empty selectors', () => {
+      const cssText =
+          stringifier.stringify(parser.parse(
+              '{ empty-a } { empty-b } empty-c { empty-d }'));
+      expect(cssText).to.be
+          .eql('{empty-a;}{empty-b;}empty-c{empty-d;}');
+    });
+
     describe('with discarded nodes', () => {
       it('stringifies to a corrected stylesheet', () => {
         const cssText =