diff --git a/src/app.py b/src/app.py
index f5b40fe..e66c582 100644
--- a/src/app.py
+++ b/src/app.py
@@ -277,9 +277,13 @@ def loganswer(questiontype):
 
     isCorrect = data['correct']
 
-    misconceptions = ast.literal_eval(data['misconceptions'])
-    question_text = data['question_text']
-    question_options = json.dumps(data['question_options'])
+    misconceptions = ast.literal_eval(data['misconceptions'])
+    question_text = data['question_text']
+    question_frame = data.get('question_frame', '')
+    question_options_payload = {"options": data['question_options']}
+    if question_frame:
+        question_options_payload["question_frame"] = question_frame
+    question_options = json.dumps(question_options_payload)
 
     userId = getUserName()
     courseId = getUserCourse(userId)
diff --git a/src/exercisebuilder.py b/src/exercisebuilder.py
index 81e6706..289bd81 100644
--- a/src/exercisebuilder.py
+++ b/src/exercisebuilder.py
@@ -8,6 +8,7 @@
 import re
 import math
 import ltltoeng
+import ltlgloss
 from syntacticmutator import applyRandomMutationNotEquivalentTo
 
 
@@ -289,14 +290,19 @@ def formula_choice_metric(formula):
     def gen_nl_question(self, formula):
 
         as_node = ltlnode.parse_ltl_string(formula)
-        formula_eng = as_node.__to_english__()
-        if formula_eng is None or formula_eng == "":
+        gloss_result = ltlgloss.gloss_formula(as_node)
+
+        if gloss_result.best_gloss is None or gloss_result.best_gloss == "":
             return None
-        
-        formula_eng_corrected = ltltoeng.correct_grammar(formula_eng)
+
+        formula_eng_corrected = ltltoeng.correct_grammar(gloss_result.best_gloss)
         ### If there are multiple '.' in a row, replace with a single '.'
-        formula_eng_corrected = re.sub(r'\.{2,}', '.', formula_eng)
-        return formula_eng_corrected
+        formula_eng_corrected = re.sub(r'\.{2,}', '.', formula_eng_corrected)
+        return {
+            "text": formula_eng_corrected,
+            "frame": gloss_result.frame.value,
+            "candidates": {frame.value: gloss for frame, gloss in gloss_result.frame_glosses.items()},
+        }
 
 
     def get_options_with_misconceptions_as_formula(self, answer):
@@ -357,7 +363,9 @@ def build_english_to_ltl_question(self, answer):
             return None
 
         return {
-            "question": question,
+            "question": question["text"] if isinstance(question, dict) else question,
+            "gloss_frame": question.get("frame", "") if isinstance(question, dict) else "",
+            "gloss_candidates": question.get("candidates", {}) if isinstance(question, dict) else {},
             "type": self.ENGLISHTOLTL,
             "options": options
         }
diff --git a/src/ltlgloss.py b/src/ltlgloss.py
new file mode 100644
index 0000000..cf70b44
--- /dev/null
+++ b/src/ltlgloss.py
@@ -0,0 +1,168 @@
+"""Utilities for producing cognitively-framed glosses of LTL formulae.
+
+This module takes an LTL AST node and produces three English glosses that
+highlight different cognitive frames:
+
+1. Abstract / Logical (AL): operator-forward phrasing.
+2. Deontic / Obligation (DO): emphasises guarantees the system must provide.
+3. Narrative / Story (NS): event-focused narrative phrasing.
+
+It also scores the glosses with a lightweight n-gram heuristic so callers can
+pick the best-fitting gloss while keeping track of which frame was selected.
+"""
+from __future__ import annotations
+
+import re
+from dataclasses import dataclass
+from enum import Enum
+from typing import Dict, Iterable
+
+import ltlnode
+import ltltoeng
+
+
+class CognitiveFrame(Enum):
+    ABSTRACT_LOGICAL = "AL"
+    DEONTIC_OBLIGATION = "DO"
+    NARRATIVE_STORY = "NS"
+
+
+@dataclass
+class GlossResult:
+    """Result container for gloss generation."""
+
+    best_gloss: str
+    frame: CognitiveFrame
+    frame_glosses: Dict[CognitiveFrame, str]
+
+
+_FRAME_NGRAMS: Dict[CognitiveFrame, Iterable[str]] = {
+    CognitiveFrame.ABSTRACT_LOGICAL: ("if", "implies", "always", "eventually", "until"),
+    CognitiveFrame.DEONTIC_OBLIGATION: ("must", "should", "guarantee", "ensure", "oblig"),
+    CognitiveFrame.NARRATIVE_STORY: ("over time", "story", "unfold", "event", "happen"),
+}
+
+
+def _normalize_sentence(text: str) -> str:
+    """Clean a sentence and ensure it ends with punctuation."""
+
+    cleaned = ltltoeng.correct_grammar(text.strip())
+    cleaned = re.sub(r"\s+", " ", cleaned)
+    cleaned = re.sub(r"\s+([,.;:!?])", r"\1", cleaned)
+    cleaned = re.sub(r"\(\s+", "(", cleaned)
+    cleaned = re.sub(r"\s+\)", ")", cleaned)
+
+    if cleaned and cleaned[0].isalpha():
+        cleaned = cleaned[0].upper() + cleaned[1:]
+
+    if cleaned and cleaned[-1] not in {".", "!", "?"}:
+        cleaned += "."
+
+    return cleaned
+
+
+def _normalize_base_gloss(node: ltlnode.LTLNode) -> str:
+    """Obtain a base English gloss for an LTL node."""
+
+    base = node.__to_english__()
+    if base is None:
+        return ""
+
+    return _normalize_sentence(base)
+
+
+def _gloss_do(node: ltlnode.LTLNode) -> str:
+    """Recursively build a Deontic/Obligation gloss."""
+
+    if isinstance(node, ltlnode.LiteralNode):
+        return f"{node.value} holds in the current state"
+    if isinstance(node, ltlnode.NotNode):
+        return f"the system must ensure that not ({_gloss_do(node.operand)})"
+    if isinstance(node, ltlnode.AndNode):
+        return f"the system must satisfy both ({_gloss_do(node.left)}) and ({_gloss_do(node.right)})"
+    if isinstance(node, ltlnode.OrNode):
+        return f"the system must satisfy at least one of ({_gloss_do(node.left)}) or ({_gloss_do(node.right)})"
+    if isinstance(node, ltlnode.ImpliesNode):
+        return f"whenever ({_gloss_do(node.left)}) occurs, the system is obligated to ensure ({_gloss_do(node.right)})"
+    if isinstance(node, ltlnode.GloballyNode):
+        return f"the system must ensure at all times that ({_gloss_do(node.operand)}) holds"
+    if isinstance(node, ltlnode.FinallyNode):
+        return f"the system must eventually bring about ({_gloss_do(node.operand)})"
+    if isinstance(node, ltlnode.NextNode):
+        return f"on the next step, the system must enforce ({_gloss_do(node.operand)})"
+    if isinstance(node, ltlnode.UntilNode):
+        return f"the system must keep ({_gloss_do(node.left)}) true until it achieves ({_gloss_do(node.right)})"
+
+    fallback = _normalize_base_gloss(node)
+    return fallback.rstrip(".") if fallback else ""
+
+
+def _gloss_ns(node: ltlnode.LTLNode) -> str:
+    """Recursively build a Narrative/Event gloss."""
+
+    if isinstance(node, ltlnode.LiteralNode):
+        return f"{node.value} happens"
+    if isinstance(node, ltlnode.NotNode):
+        return f"it never happens that ({_gloss_ns(node.operand)})"
+    if isinstance(node, ltlnode.AndNode):
+        return f"({_gloss_ns(node.left)}) happens while ({_gloss_ns(node.right)}) also happens"
+    if isinstance(node, ltlnode.OrNode):
+        return f"either ({_gloss_ns(node.left)}) happens or ({_gloss_ns(node.right)}) happens"
+    if isinstance(node, ltlnode.ImpliesNode):
+        return f"whenever ({_gloss_ns(node.left)}) ever happens, ({_gloss_ns(node.right)}) must also happen"
+    if isinstance(node, ltlnode.GloballyNode):
+        return f"from now on, ({_gloss_ns(node.operand)}) always holds"
+    if isinstance(node, ltlnode.FinallyNode):
+        return f"sooner or later, ({_gloss_ns(node.operand)}) will happen"
+    if isinstance(node, ltlnode.NextNode):
+        return f"right after this, ({_gloss_ns(node.operand)}) happens"
+    if isinstance(node, ltlnode.UntilNode):
+        return f"({_gloss_ns(node.left)}) keeps happening until eventually ({_gloss_ns(node.right)}) happens"
+
+    fallback = _normalize_base_gloss(node)
+    return fallback.rstrip(".") if fallback else ""
+
+
+def _build_frame_glosses(node: ltlnode.LTLNode, base_gloss: str) -> Dict[CognitiveFrame, str]:
+    """Create three frame-specific glosses from a base gloss."""
+
+    return {
+        CognitiveFrame.ABSTRACT_LOGICAL: base_gloss,
+        CognitiveFrame.DEONTIC_OBLIGATION: _normalize_sentence(_gloss_do(node)),
+        CognitiveFrame.NARRATIVE_STORY: _normalize_sentence(_gloss_ns(node)),
+    }
+
+
+def _score_gloss(frame: CognitiveFrame, gloss: str) -> int:
+    """Score a gloss using a lightweight frame-specific n-gram heuristic."""
+
+    lowered = gloss.lower()
+    return sum(lowered.count(ng) for ng in _FRAME_NGRAMS[frame])
+
+
+def gloss_formula(node: ltlnode.LTLNode) -> GlossResult:
+    """Generate and score glosses for an LTL formula.
+
+    Args:
+        node: Parsed LTL AST node.
+
+    Returns:
+        GlossResult containing the winning gloss, its frame, and all candidates.
+    """
+
+    base_gloss = _normalize_base_gloss(node)
+    if base_gloss == "":
+        return GlossResult(best_gloss="", frame=CognitiveFrame.ABSTRACT_LOGICAL, frame_glosses={})
+
+    frame_glosses = _build_frame_glosses(node, base_gloss)
+
+    best_frame = max(
+        frame_glosses,
+        key=lambda frame: (_score_gloss(frame, frame_glosses[frame]), frame.value),
+    )
+
+    return GlossResult(
+        best_gloss=frame_glosses[best_frame],
+        frame=best_frame,
+        frame_glosses=frame_glosses,
+    )
diff --git a/src/static/js/checkanswers.js b/src/static/js/checkanswers.js
index bed15ac..1e36347 100644
--- a/src/static/js/checkanswers.js
+++ b/src/static/js/checkanswers.js
@@ -11,9 +11,17 @@ function getQuestionText(parentNode) {
     return parentNode.querySelector('.actualQuestion').innerText;
 }
 
-function getQuestionTrace(parentNode) {
-    return parentNode.querySelector('.actualQuestionTrace').innerText;
-}
+function getQuestionTrace(parentNode) {
+    return parentNode.querySelector('.actualQuestionTrace').innerText;
+}
+
+function getQuestionFrame(parentNode) {
+    let questionNode = parentNode.querySelector('.actualQuestion');
+    if (questionNode && questionNode.dataset) {
+        return questionNode.dataset.glossFrame || "";
+    }
+    return "";
+}
 
 function getQuestionOptions(parentNode) {
 
@@ -210,12 +218,13 @@ async function tracesatisfaction_mc_getfeedback(button) {
         selected_option: selected_radio.value,
         correct_option: correct_option,
         correct: correct,
-        misconceptions: selected_radio.dataset.misconceptions,
-        question_text: question_text,
-        question_options: question_options,
-        formula_for_mp_class: get_formula_for_MP_Classification(parent_node, QUESTION_TYPE),
-        exercise: getExerciseName()
-    }
+        misconceptions: selected_radio.dataset.misconceptions,
+        question_text: question_text,
+        question_options: question_options,
+        question_frame: getQuestionFrame(parent_node),
+        formula_for_mp_class: get_formula_for_MP_Classification(parent_node, QUESTION_TYPE),
+        exercise: getExerciseName()
+    }
     let response = await postFeedback(data, QUESTION_TYPE);
 }
 
@@ -239,12 +248,13 @@ async function tracesatisfaction_yn_getfeedback(button) {
         selected_option: selected_radio.value,
         correct_option: correct_option,
         correct: correct,
-        misconceptions: selected_radio.dataset.misconceptions,
-        question_text: question_text,
-        question_options: question_options,
-        formula_for_mp_class: get_formula_for_MP_Classification(parent_node, QUESTION_TYPE),
-        exercise: getExerciseName()
-    }
+        misconceptions: selected_radio.dataset.misconceptions,
+        question_text: question_text,
+        question_options: question_options,
+        question_frame: getQuestionFrame(parent_node),
+        formula_for_mp_class: get_formula_for_MP_Classification(parent_node, QUESTION_TYPE),
+        exercise: getExerciseName()
+    }
     let response = await postFeedback(data, QUESTION_TYPE);
 }
 
@@ -268,12 +278,13 @@ async function englishtoltl_getfeedback(button) {
         selected_option: selected_radio.value,
         correct_option: correct_option,
         correct: correct,
-        misconceptions: selected_radio.dataset.misconceptions,
-        question_text: question_text,
-        question_options: question_options,
-        formula_for_mp_class: get_formula_for_MP_Classification(parent_node, QUESTION_TYPE),
-        exercise: getExerciseName()
-    }
+        misconceptions: selected_radio.dataset.misconceptions,
+        question_text: question_text,
+        question_options: question_options,
+        question_frame: getQuestionFrame(parent_node),
+        formula_for_mp_class: get_formula_for_MP_Classification(parent_node, QUESTION_TYPE),
+        exercise: getExerciseName()
+    }
 
     let response = await postFeedback(data, QUESTION_TYPE);
     displayServerResponse(response);
diff --git a/src/templates/exercise.html b/src/templates/exercise.html
index 1486112..1bf1b9f 100644
--- a/src/templates/exercise.html
+++ b/src/templates/exercise.html
@@ -91,8 +91,8 @@ <h6 id="exerciseName">{{ exercise_name }}</h6>
 
 
                     {% if q.type == "tracesatisfaction_yn" %}
-                    <h5 id="{{ qindex }}question_text" class="card-title mb-2">
-                        <pre><code class="actualQuestion language-ltl"> {{ q.question}} </code></pre>
+                    <h5 id="{{ qindex }}question_text" class="card-title mb-2">
+                        <pre><code class="actualQuestion language-ltl" data-gloss-frame="{{ q.gloss_frame | default('') }}"> {{ q.question}} </code></pre>
                     </h5>
                     <pre class="mermaid" id="q{{qindex}}_yn_trace"
                         aria-label="{{q.trace}}"> {{ q.mermaid | safe}}  </pre>
@@ -101,12 +101,13 @@ <h5 id="{{ qindex }}question_text" class="card-title mb-2">
                         <pre class="actualQuestionTrace"> {{ q.trace}}</pre>
                     </div>
                     {% elif q.type == "tracesatisfaction_mc" %}
-                    <h5 id="{{ qindex }}question_text" class="card-title mb-2">
-                        <pre><code class="actualQuestion language-ltl">{{ q.question}}</code></pre>
+                    <h5 id="{{ qindex }}question_text" class="card-title mb-2">
+                        <pre><code class="actualQuestion language-ltl" data-gloss-frame="{{ q.gloss_frame | default('') }}">{{ q.question}}</code></pre>
                     </h5>
                     {% else %}
-                    <h5 id="{{ qindex }}question_text" class="card-title mb-2 actualQuestion"
-                        style="white-space: pre-wrap;">{{ q.question}}
+                    <h5 id="{{ qindex }}question_text" class="card-title mb-2 actualQuestion"
+                        data-gloss-frame="{{ q.gloss_frame | default('') }}"
+                        style="white-space: pre-wrap;">{{ q.question}}
                     </h5>
                     <div class="row mb-2 ml-2">
                         {% if q.type == "englishtoltl" %}
diff --git a/src/templates/prebuiltexercises/lightpanel.html b/src/templates/prebuiltexercises/lightpanel.html
index 5a9997a..4f0b901 100644
--- a/src/templates/prebuiltexercises/lightpanel.html
+++ b/src/templates/prebuiltexercises/lightpanel.html
@@ -74,8 +74,8 @@ <h5 class="card-title">
 
 
                     {% if q.type == "tracesatisfaction_yn" %}
-                    <h5 id="{{ qindex }}question_text" class="card-subtitle mb-2">Formula: 
-                        <pre><code class="actualQuestion language-ltl"> {{ q.question}} </code></pre></h5>
+                    <h5 id="{{ qindex }}question_text" class="card-subtitle mb-2">Formula:
+                        <pre><code class="actualQuestion language-ltl" data-gloss-frame="{{ q.gloss_frame | default('') }}"> {{ q.question}} </code></pre></h5>
                     <pre class="mermaid" id="q{{qindex}}_yn_trace"
                         aria-label="{{q.trace}}"> {{ q.mermaid | safe}}  </pre>
 
@@ -84,10 +84,11 @@ <h5 id="{{ qindex }}question_text" class="card-subtitle mb-2">Formula:
                         <pre class="actualQuestionTrace"> {{ q.trace}}</pre>
                     </div>
                     {% elif q.type == "tracesatisfaction_mc" %}
-                    <h5 id="{{ qindex }}question_text" class="card-subtitle mb-2">Formula: 
-                        <pre><code class="actualQuestion language-ltl">{{ q.question}}</code></pre></h5>
+                    <h5 id="{{ qindex }}question_text" class="card-subtitle mb-2">Formula:
+                        <pre><code class="actualQuestion language-ltl" data-gloss-frame="{{ q.gloss_frame | default('') }}">{{ q.question}}</code></pre></h5>
                     {% else %}
                     <h5 id="{{ qindex }}question_text" class="card-subtitle mb-2 actualQuestion"
+                        data-gloss-frame="{{ q.gloss_frame | default('') }}"
                         style="white-space: pre-wrap;">{{ q.question}}
                     </h5>
                     {% endif %}
diff --git a/src/templates/prebuiltexercises/robotrain.html b/src/templates/prebuiltexercises/robotrain.html
index 88a1bc2..f30c343 100644
--- a/src/templates/prebuiltexercises/robotrain.html
+++ b/src/templates/prebuiltexercises/robotrain.html
@@ -84,7 +84,7 @@ <h6>
 
                     {% if q.type == "tracesatisfaction_yn" %}
                     <h5 id="{{ qindex }}question_text" class="card-title mb-2">
-                        <pre><code class="actualQuestion language-ltl"> {{ q.question}} </code></pre>
+                        <pre><code class="actualQuestion language-ltl" data-gloss-frame="{{ q.gloss_frame | default('') }}"> {{ q.question}} </code></pre>
                     </h5>
                     <pre class="mermaid" id="q{{qindex}}_yn_trace"
                         aria-label="{{q.trace}}"> {{ q.mermaid | safe}}  </pre>
@@ -94,10 +94,11 @@ <h5 id="{{ qindex }}question_text" class="card-title mb-2">
                     </div>
                     {% elif q.type == "tracesatisfaction_mc" %}
                     <h5 id="{{ qindex }}question_text" class="card-title mb-2">
-                        <pre><code class="actualQuestion language-ltl">{{ q.question}}</code></pre>
+                        <pre><code class="actualQuestion language-ltl" data-gloss-frame="{{ q.gloss_frame | default('') }}">{{ q.question}}</code></pre>
                     </h5>
                     {% else %}
                     <h5 id="{{ qindex }}question_text" class="card-title mb-2 actualQuestion"
+                        data-gloss-frame="{{ q.gloss_frame | default('') }}"
                         style="white-space: pre-wrap;">{{ q.question}}
                     </h5>
                     {% endif %}