diff --git a/cjs/shared/parse-from-string.js b/cjs/shared/parse-from-string.js
index 84a9fdf3..aac2862b 100644
--- a/cjs/shared/parse-from-string.js
+++ b/cjs/shared/parse-from-string.js
@@ -1,7 +1,7 @@
 'use strict';
 const HTMLParser2 = require('htmlparser2');
 
-const {ELEMENT_NODE, SVG_NAMESPACE} = require('./constants.js');
+const {ELEMENT_NODE, SVG_NAMESPACE, TEXT_NODE} = require('./constants.js');
 const {CUSTOM_ELEMENTS, PREV, END, VALUE} = require('./symbols.js');
 const {keys} = require('./object.js');
 
@@ -95,7 +95,11 @@ const parseFromString = (document, isHTML, markupLanguage) => {
       if (parsingCData) {
         append(node, document.createCDATASection(text), active);
       } else {
-        append(node, document.createTextNode(text), active);
+        const prev = node[END][PREV];
+        if (prev.nodeType === TEXT_NODE)
+          prev[VALUE] += text;
+        else
+          append(node, document.createTextNode(text), active);
       }
     },
 
diff --git a/esm/shared/parse-from-string.js b/esm/shared/parse-from-string.js
index 08c6fe39..14b678ae 100644
--- a/esm/shared/parse-from-string.js
+++ b/esm/shared/parse-from-string.js
@@ -1,6 +1,6 @@
 import * as HTMLParser2 from 'htmlparser2';
 
-import {ELEMENT_NODE, SVG_NAMESPACE} from './constants.js';
+import {ELEMENT_NODE, SVG_NAMESPACE, TEXT_NODE} from './constants.js';
 import {CUSTOM_ELEMENTS, PREV, END, VALUE} from './symbols.js';
 import {keys} from './object.js';
 
@@ -93,7 +93,11 @@ export const parseFromString = (document, isHTML, markupLanguage) => {
       if (parsingCData) {
         append(node, document.createCDATASection(text), active);
       } else {
-        append(node, document.createTextNode(text), active);
+        const prev = node[END][PREV];
+        if (prev.nodeType === TEXT_NODE)
+          prev[VALUE] += text;
+        else
+          append(node, document.createTextNode(text), active);
       }
     },
 
diff --git a/test/html/document.js b/test/html/document.js
index b5e15062..71d40498 100644
--- a/test/html/document.js
+++ b/test/html/document.js
@@ -98,3 +98,40 @@ const issue187 = (new DOMParser).parseFromString(`<!DOCTYPE html><html><head></h
 issue187.body.innerHTML = '<span></span>';
 let span = issue187.body.firstElementChild;
 assert(span.ownerDocument.body, issue187.body);
+
+// @issue 353: the parser must merge consecutive ontext pieces into one Text node.
+// htmlparser2 with decodeEntities:true calls ontext once per piece around each
+// character reference; adjacent pieces under the same parent must be merged into
+// a single Text node (per the HTML parsing algorithm "insert a character").
+{
+  const parse = html => (new DOMParser).parseFromString(`<!DOCTYPE html><html><body>${html}</body></html>`, 'text/html');
+
+  // entity split: "a &amp; b" -> one Text node "a & b"
+  let p = parse('<p>a &amp; b</p>').querySelector('p');
+  assert(p.childNodes.length, 1, 'entity split merged into one Text node');
+  assert(p.firstChild.data, 'a & b', 'entity split data');
+
+  // multi-entity: "x &lt; y &gt; z &#169;" -> one Text node
+  p = parse('<p>x &lt; y &gt; z &#169;</p>').querySelector('p');
+  assert(p.childNodes.length, 1, 'multi-entity merged into one Text node');
+  assert(p.firstChild.data, 'x < y > z ©', 'multi-entity data');
+
+  // plain text unchanged (still one node)
+  p = parse('<p>plain text</p>').querySelector('p');
+  assert(p.childNodes.length, 1, 'plain text stays one node');
+  assert(p.firstChild.data, 'plain text', 'plain text data');
+
+  // text after an element boundary is NOT merged across it
+  p = parse('<p>a &amp;<b>x</b> &amp; b</p>').querySelector('p');
+  assert(p.childNodes.length, 3, 'text not merged across element boundary');
+  assert(p.firstChild.data, 'a &', 'first run before element');
+  assert(p.lastChild.data, ' & b', 'last run after element');
+  assert(p.querySelector('b').childNodes.length, 1, 'element own text still one node');
+  assert(p.querySelector('b').firstChild.data, 'x', 'element own text data');
+
+  // text after a comment boundary is NOT merged across it
+  p = parse('<p>a <!-- c --> &amp; b</p>').querySelector('p');
+  assert(p.childNodes.length, 3, 'text not merged across comment boundary');
+  assert(p.firstChild.data, 'a ', 'first run before comment');
+  assert(p.lastChild.data, ' & b', 'last run after comment');
+}