The construction of emotional meaning in language

Katie Hoemann, Yeasle Lee, Èvelyne Dussault, Simon Devylder, Lyle Ungar, Dirk Geeraerts, Batja MesquitaPsyArXiv 2025 · DOI

TL;DR

Emotion is not just a label we slap on an experience; it is a dynamic process of "meaning-making" that situates us within our circumstances. This paper argues that to truly understand emotion, researchers must move beyond simple emotion words (like "sad" or "angry") and instead analyze the linguistic structures of attention, construal, and appraisal that reveal how emotional meaning is constructed in real-time.

What Problem Does This Paper Address?

For decades, the psychological study of emotion has relied heavily on explicit labels—asking people to rate how "disappointed" or "frustrated" they feel. The authors argue this approach is too low-dimensional. It ignores the "how" and "why" of an emotional experience.

Two people might both say they are "disappointed," but one might be focusing on their own failed effort (internal focus), while the other focuses on a friend’s betrayal (social focus). One might view the event as a static "thing" (noun-based), while the other sees it as an unfolding process (verb-based). This paper addresses the need for a higher-dimensional, situated account of emotion that leverages the rich, ubiquitous data found in natural language.

Methods

This is a perspective paper that synthesizes theories from cognitive linguistics, computational linguistics, and social psychology. It proposes a framework for studying the construction of emotional meaning through three core linguistic processes:

  1. Attention: Identifying what features of an experience are salient. The authors suggest using top-down methods (like LIWC dictionaries to count categories like "family" or "work" words) and bottom-up methods (like topic modeling to discover themes that emerge naturally from text).
  2. Construal: Analyzing the speaker's mental vantage point. This involves looking at grammatical features:
  3. Word Class: Nouns (static entities) vs. Verbs (dynamic processes).
  4. Tense and Aspect: How "linguistic distancing" (using past tense) affects emotion regulation.
  5. Agency: Whether the speaker is the subject ("I fear dogs") or the object ("Dogs frighten me").
  6. Appraisal: Evaluating the event along dimensions of valence (pleasantness), intensity, and certainty. This can be measured via sentiment analysis, punctuation (exclamation points), or modal verbs ("might" vs. "clearly").

Datasets

As a theoretical perspective paper, it does not introduce a new dataset. However, it reviews and cites research utilizing a wide variety of sources, including:

  • Daily diaries and experience sampling reports.
  • Social media posts (Facebook, Twitter/X).
  • Transcribed psychotherapy sessions.
  • Cross-cultural samples (e.g., US English vs. Belgian Dutch vs. Hadza speakers).

Key Insights and Ideas

  • Grammar is a Window into Emotion: The way we structure a sentence reveals our perceived locus of control. Using the passive voice ("I was frightened by the dogs") makes the grammatical object ("I") more salient, potentially biasing how the speaker and listener perceive the "primary actor" in an emotional event.
  • Emotions as Processes vs. Entities: Some languages (like Hindustani) spontaneously generate more verbs for emotions, while English prefers adjectives or nouns. Framing an emotion as a process ("to anger") may be associated with beliefs that emotions are more malleable and easier to regulate.
  • The "Average American" Bias in AI: The authors warn that most current NLP and LLM tools for emotion annotation are trained on US English. This creates a risk of "averaging out" cultural nuances, such as the tendency in some African and East Asian contexts to describe emotional events using more embodied or social language.
  • Beyond Labels: Meaning occurs at the level of "usage events." What an utterance means is defined by what it is used to do in a specific social and physical context.

Key Results

The paper provides a roadmap for "constructionist" emotion research:

  • Multidimensionality: By combining attention, construal, and appraisal, researchers can move from a 1D "label" to a high-dimensional map of an emotional state.
  • Feasibility: The authors highlight that the advancement of text-analytic tools (from LIWC to BERT and generative AI) makes it increasingly possible to automate the detection of these complex linguistic features at scale.
  • Interdisciplinary Integration: The core result is a successful synthesis showing that "emotion" and "language" are not separate systems, but two systems of meaning that transduce raw experience into shared symbols.

Limitations and Future Work

  • Cultural Generalizability: Much of the cited work is Western-centric. The authors emphasize the urgent need for more cross-linguistic research to see if these linguistic markers of construal function the same way globally.
  • Contextual Blind Spots: Language alone doesn't capture everything. Future work should integrate linguistic analysis with physical location (GPS), social power dynamics, and paralinguistic cues like vocal intonation and facial expressions.
  • Automation vs. Nuance: While LLMs are getting better at qualitative labeling, some features (like "motifs" or complex narrative arcs) still often require high-effort manual annotation to ensure reliability.

Reference: Hoemann, K., et al. (2025). The construction of emotional meaning in language. [Pre-peer-review version].

Summary updated

Abstract

The experience of emotion is a form of meaning-making: it encapsulates one’s relationship to the circumstances. Certain concerns are put in focus; one sees oneself and events from a particular perspective; events are evaluated along a set of relevant dimensions. These three processes – attention, construal, and appraisal – are key to understanding how emotional meaning is assembled in context. In this perspective, we illustrate their importance for the science of emotion and propose means of studying them through language. Our interdisciplinary approach seeks to move the field beyond the words that label emotions, to a higher-dimensional, dynamical account of emotional meaning.
BibTeX
@misc{hoemann2025construction,
  author = {Katie Hoemann and Yeasle Lee and Èvelyne Dussault and Simon Devylder and Lyle Ungar and Dirk Geeraerts and Batja Mesquita},
  title = {{The construction of emotional meaning in language}},
  year = {2025},
  doi = {10.31234/osf.io/2c57n},
  howpublished = {PsyArXiv},
}

Loading related content...

Figure