diff --git a/src/lib/study.js b/src/lib/study.js
index 850b003c..b30bbc4a 100644
--- a/src/lib/study.js
+++ b/src/lib/study.js
@@ -241,9 +241,11 @@ function getAnomalyGroupFromType(type) {
 /**
  * Structure a flat list of anomalies to the requested structure
  */
-function structureResults(structure, anomalies, crawlResults) {
+function structureResults(structure, sort, anomalies, crawlResults) {
   const levels = structure.split('/')
     .map(level => level.replace(/\s+/g, ''));
+  const sortKeys = sort.split('/')
+    .map(level => level.replace(/\s+/g, ''));
   const report = [];
 
   switch (levels[0]) {
@@ -348,10 +350,25 @@ function structureResults(structure, anomalies, crawlResults) {
       break;
   }
 
+  switch (sortKeys[0] ?? 'default') {
+    case 'name':
+      report.sort((e1, e2) => e1.name.localeCompare(e2.name));
+      break;
+    case 'title':
+      // Note: for actual anomalies, the title is the anomaly message. All
+      // other entries have title.
+      report.sort((e1, e2) => (e1.title ?? e1.message).localeCompare(e2.title ?? e2.message));
+      break;
+    case 'default':
+    default:
+      break;
+  }
+
   if (levels.length > 1) {
     const itemsStructure = levels.slice(1).join('/');
+    const itemsSort = sortKeys.slice(1).join('/');
     for (const entry of report) {
-      entry.items = structureResults(itemsStructure, entry.anomalies, crawlResults);
+      entry.items = structureResults(itemsStructure, itemsSort, entry.anomalies, crawlResults);
       delete entry.anomalies;
     }
   }
@@ -556,6 +573,7 @@ export default async function study(specs, options = {}) {
 
   const what = options.what ?? ['all'];
   const structure = options.structure ?? 'type + spec';
+  const sort = options.sort ?? 'default';
   const format = options.format ?? 'issue';
 
   if (!what.includes('all')) {
@@ -600,7 +618,7 @@ export default async function study(specs, options = {}) {
 
   // Now that we have a flat report of anomalies,
   // let's structure and serialize it as requested
-  const report = structureResults(structure, anomalies, options.crawlResults);
+  const report = structureResults(structure, sort, anomalies, options.crawlResults);
 
   // And serialize it using the right format
   const result = {
diff --git a/strudy.js b/strudy.js
index bbd4bdd2..370bdaed 100644
--- a/strudy.js
+++ b/strudy.js
@@ -68,6 +68,7 @@ program
   .option('-i, --issues <folder>', 'report issues as markdown files in the given folder')
   .option('-m, --max <max>', 'maximum number of issue files to create/update', myParseInt, 0)
   .option('-s, --spec <specs...>', 'restrict analysis to given specs')
+  .option('--sort <sort>', 'key(s) to use to sort the structured report', 'default')
   .option('--structure <structure>', 'report structure', 'type+spec')
   .option('--tr <trreport>', 'path/URL to crawl report on published specs')
   .option('--update-mode <mode>', 'what issue files to update', 'new')
@@ -136,6 +137,39 @@ Usage notes for some of the options:
   For instance:
     $ strudy inspect . --spec picture-in-picture https://w3c.github.io/mediasession/
 
+--sort <sort>
+  Specifies the key(s) to use to sort each level in the structured report.
+  Use "/" to separate levels. See --structure for details on the possible
+  report structure.
+
+  Possible keys:
+  "default"   follow the natural order of the underlying structures, e.g.
+              return specs in the order in which they appear in the initial
+              list, anomalies in extraction order (which usually follows the
+              document order)
+  "name"      sort entries by the name. For a "spec" level, the name is the
+              spec's shortname. For a "type" level, the name is the anomaly
+              type name. For a "type+spec" level, the name is the name of the
+              file that would be created if --issues is set, meaning the spec's
+              shortname completed with the anomaly type name.
+  "title"     sort entries by their title. For a "spec" level, the title is the
+              spec's title. For the final level, the title is the anomaly
+              message. Etc.
+
+  If the --sort value contains more levels than there are in the structured
+  report, additional keys are ignored. If the value contails fewer levels than
+  there are in the structured report, the default order is used for unspecified
+  levels.
+
+  For example, if the structure is "type/spec", the --sort option could be:
+  "default"           to use the default order at all levels
+  "default/title"     to use the default order for the root level, and to sort
+                      specs by title
+  "name/title/title"  to sort anomaly types by names, specs by title, and
+                      anomalies by message.
+
+  Sort is always ascending.
+
 --structure <type>
   Describes the hierarchy in the report(s) that Strudy returns. Possible values:
   "flat"            no level, report anomalies one by one
@@ -257,6 +291,7 @@ Format must be one of "json" or "markdown".`)
     const anomaliesReport = await study(edReport.results, {
       what: options.what,
       structure: options.structure,
+      sort: options.sort,
       format: options.format === 'json' ?
         'json' :
         (options.issues ? 'issue' : 'full'),
diff --git a/test/study.js b/test/study.js
index d439997b..de633d5f 100644
--- a/test/study.js
+++ b/test/study.js
@@ -191,4 +191,22 @@ describe('The main study function', function () {
 See [Dealing with the event loop](https://html.spec.whatwg.org/multipage/webappapis.html#event-loop-for-spec-authors) in the HTML specification for guidance on how to deal with algorithm sections that run *in parallel*.`
     });
   });
+
+  it('sorts entries as requested in the final report', async function() {
+    const crawlResult = [
+      populateSpec(specUrl, { error: 'Boo' }),
+      populateSpec(specUrl2, { error: 'Borked' })
+    ];
+    const report = await study(crawlResult, { structure: 'type/spec', sort: 'default/title', htmlFragments: {} });
+    assertNbAnomalies(report.results, 1);
+    assertAnomaly(report.results, 0, {
+      title: 'Crawl error',
+      content:
+`The following crawl errors occurred:
+* [Hello universe API](https://w3c.github.io/universe/)
+  * [ ] Borked
+* [Hello world API](https://w3c.github.io/world/)
+  * [ ] Boo`
+    });
+  });
 });
\ No newline at end of file