diff --git a/src/select/mod.rs b/src/select/mod.rs
index 1d297e2..d643a01 100644
--- a/src/select/mod.rs
+++ b/src/select/mod.rs
@@ -2,8 +2,8 @@ use std::collections::HashSet;
 use std::fmt;
 
 use array_tool::vec::{Intersect, Union};
-use serde_json::map::Entry;
 use serde_json::{Number, Value};
+use serde_json::map::Entry;
 
 use parser::*;
 
@@ -390,6 +390,18 @@ impl<'a> Into<ExprTerm<'a>> for &Vec<&'a Value> {
     }
 }
 
+fn walk_all_with_num<'a>(vec: &[&'a Value], tmp: &mut Vec<&'a Value>, index: f64) {
+    walk(vec, tmp, &|v| if v.is_array() {
+        if let Some(item) = v.get(index as usize) {
+            Some(vec![item])
+        } else {
+            None
+        }
+    } else {
+        None
+    });
+}
+
 fn walk_all_with_str<'a>(vec: &[&'a Value], tmp: &mut Vec<&'a Value>, key: &str, is_filter: bool) {
     if is_filter {
         walk(vec, tmp, &|v| match v {
@@ -833,6 +845,16 @@ impl<'a, 'b> Selector<'a, 'b> {
         debug!("all_from_current_with_str: {}, {:?}", key, self.current);
     }
 
+    fn all_from_current_with_num(&mut self, index: f64) {
+        if let Some(current) = self.current.take() {
+            let mut tmp = Vec::new();
+
+            walk_all_with_num(&current, &mut tmp, index);
+            self.current = Some(tmp);
+        }
+        debug!("all_from_current_with_num: {}, {:?}", index, self.current);
+    }
+
     fn compute_absolute_path_filter(&mut self, token: &ParseToken) -> bool {
         if !self.selectors.is_empty() {
             match token {
@@ -895,7 +917,7 @@ impl<'a, 'b> Selector<'a, 'b> {
     }
 
     fn visit_array_eof(&mut self) {
-        if self.is_nested_array() {
+        if self.is_last_before_token_match(ParseToken::Array) {
             if let Some(Some(e)) = self.terms.pop() {
                 if let ExprTerm::String(key) = e {
                     self.next_in_filter_with_str(&key);
@@ -907,6 +929,20 @@ impl<'a, 'b> Selector<'a, 'b> {
             }
         }
 
+        if self.is_last_before_token_match(ParseToken::Leaves) {
+            self.tokens.pop();
+            self.tokens.pop();
+            if let Some(Some(e)) = self.terms.pop() {
+                if let ExprTerm::Number(n) = &e {
+                    self.all_from_current_with_num(to_f64(n));
+                    self.terms.pop();
+                    return;
+                }
+
+                self.terms.push(Some(e));
+            }
+        }
+
         if let Some(Some(e)) = self.terms.pop() {
             match e {
                 ExprTerm::Number(n) => {
@@ -934,6 +970,14 @@ impl<'a, 'b> Selector<'a, 'b> {
         self.tokens.pop();
     }
 
+    fn is_last_before_token_match(&mut self, token: ParseToken) -> bool {
+        if self.tokens.len() > 1 {
+            return &token == &self.tokens[self.tokens.len() - 2];
+        }
+
+        return false;
+    }
+
     fn visit_all(&mut self) {
         if let Some(ParseToken::Array) = self.tokens.last() {
             self.tokens.pop();
@@ -954,21 +998,6 @@ impl<'a, 'b> Selector<'a, 'b> {
         }
     }
 
-    fn is_nested_array(&mut self) -> bool {
-        let mut is_nested_array = false;
-
-        if let Some(t) = self.tokens.pop() {
-            if let ParseToken::Array = t {
-                if let Some(ParseToken::Array) = self.tokens.last() {
-                    is_nested_array = true;
-                }
-            }
-            self.tokens.push(t);
-        }
-
-        is_nested_array
-    }
-
     fn visit_key(&mut self, key: &str) {
         if let Some(ParseToken::Array) = self.tokens.last() {
             self.terms.push(Some(ExprTerm::String(key.to_string())));
diff --git a/tests/array_filter.rs b/tests/array_filter.rs
index 707336e..9ffa79e 100644
--- a/tests/array_filter.rs
+++ b/tests/array_filter.rs
@@ -216,3 +216,38 @@ fn array_multiple_key() {
         json!(["blue", "Leonor Herman"]),
     );
 }
+
+#[test]
+fn bugs40_bracket_notation_after_recursive_descent() {
+    setup();
+
+    select_and_then_compare(
+        "$..[0]",
+        json!([
+            "first",
+            {
+                "key": [
+                    "first nested",
+                    {
+                        "more": [
+                            {"nested": ["deepest", "second"]},
+                            ["more", "values"]
+                        ]
+                    }
+                ]
+            }
+        ]),
+        json!([
+           "first",
+           "first nested",
+           {
+              "nested" : [
+                 "deepest",
+                 "second"
+              ]
+           },
+           "deepest",
+           "more"
+        ]),
+    );
+}
\ No newline at end of file