---
title: "Syntactic Belief Update: A New Explanation for Garden Path Sentence Processing Difficulty"
source: "arXiv AI/CL/LG（论文）"
source_tier: "t2"
url: "https://arxiv.org/abs/2606.27206v1"
permalink: "https://feed.astralor.com/items/itm_81795aa59b092d1d090d7a20"
published: "2026-06-25T16:02:14.000Z"
score: 53
channel: "ai"
tags: ["Syntactic Belief Update","Generalized Rényi Divergence","Psycholinguistics","Probabilistic Syntactic Parsing"]
---

# Syntactic Belief Update: A New Explanation for Garden Path Sentence Processing Difficulty

## At a glance

Garden path sentences pose processing difficulty for humans: the sentence prefix leads the listener towards one interpretation until a critical word reveals the initial interpretation was wrong. Lexical surprisal, which usually predicts processing difficulty well, fails for these sentences. This paper proposes an alternative: actively predict a probability distribution over syntactic trees (syntactic belief) and update it after each word. If the processor is led down a garden path, beliefs become wrong and require a large update at the critical word. The update magnitude is measured by generalized Rényi divergence, which depends on lexical items but is fully independent of their probability. Syntactic Belief Update provides a better fit to human reading time data, suggesting a new research direction for purely non-lexical alternatives to surprisal in psycholinguistics.

## Key points

- Garden path processing difficulty is not well predicted by lexical surprisal.
- Proposes Syntactic Belief Update model that maintains and updates a probability distribution over syntactic trees.
- Uses generalized Rényi divergence to quantify update magnitude, independent of lexical probability.
- Model better fits human reading time data, revealing non-lexical processing mechanisms.

## Full text

Garden path sentences present a processing difficulty for humans -- the sentence prefix leads the listener towards one interpretation, until the listener hears a critical word that shows that the initial interpretation was wrong. Lexical surprisal, a measure that usually predicts sentence processing difficulty quite well, fails to provide good predictions for garden path sentences. We propose an alternative that actively predicts a probability distribution over syntactic trees (its syntactic belief) and updates that distribution after each new word. If a processor is led down a garden path, syntactic beliefs will be wrong and will require a large update at the critical word. The magnitude of the update is measured with a generalized Rényi divergence. Crucially, this metric is dependent on lexical items, but is fully independent of the probability of lexical items. This Syntactic Belief Update provides a better fit to the human reading time data on garden path sentences. This suggests a new research direction examining purely non-lexical alternatives to surprisal for psycholinguistics.

---

Verify against the original before citing; curated by Astral Feed: https://feed.astralor.com/items/itm_81795aa59b092d1d090d7a20