TLDR: A new study uses linguistic analysis and machine learning to differentiate between Déjà Vu, Involuntary Autobiographical Memories, and Unexpected Thoughts. By examining the words people use to describe these spontaneous experiences, researchers found distinct linguistic patterns and emotional profiles for each thought type. Déjà Vu is characterized by abstract and spatial language with a neutral/positive tone, while Involuntary Autobiographical Memories are rich in personal detail and sadness. Unexpected Thoughts are marked by unpredictability and negative emotions like fear and disgust. This novel approach offers a deeper understanding of internal cognitive states and has potential applications in education and clinical assessments.
Our minds are constantly abuzz with thoughts, and a significant portion of these are spontaneous – those ideas, memories, or sensations that pop into our heads without conscious effort. These spontaneous thoughts, which can occupy up to half of our waking lives, are a fundamental part of human cognition. Traditionally, researchers have studied these experiences by asking people to describe them and rate their emotional impact or triggers. However, a recent study delves deeper, exploring the unique linguistic patterns embedded within these descriptions to uncover new insights into how our internal cognitive states manifest through language. You can read the full research paper here: A Linguistic Analysis of Spontaneous Thoughts.
The research, conducted by Videep Venkatesha, Mary Cati Poulos, Christopher Steadman, Caitlin Mills, Anne M. Cleary, and Nathaniel Blanchard, focuses on three distinct types of spontaneous thoughts: Déjà Vu, Involuntary Autobiographical Memories (IAMs), and Unexpected Thoughts (UTs). Déjà Vu is that peculiar feeling of having experienced something before, even when it’s new. IAMs are vivid personal memories that suddenly come to mind, often triggered by something in the environment. Unexpected Thoughts are surprising in their timing and content, sometimes offering new perspectives or insights.
Instead of relying solely on subjective appraisals, the researchers used advanced linguistic analysis techniques, including Term Frequency-Inverse Document Frequency (TF-IDF) and contextual embeddings from BERT, to examine how participants described these experiences. They also employed machine learning models to classify the thought types based on language alone and conducted an emotion analysis using both a BERT-based model and the Llama 3.1 large language model to understand the emotional tone of the descriptions.
Unpacking Linguistic Signatures
The study’s findings reveal that each type of spontaneous thought possesses a distinct linguistic signature. Déjà Vu experiences, for instance, are often described using abstract and spatial terms like “seemed,” “place,” and “visited.” This aligns with the understanding that Déjà Vu is a metacognitive experience, a feeling of familiarity without specific content, often tied to locations. Descriptions of Déjà Vu also tended to be more neutral or positive in emotional tone.
In contrast, Involuntary Autobiographical Memories (IAMs) are rich in personal and emotionally significant details. Words like “parents,” “girlfriend,” “mum,” “dinner,” and “party” frequently appear, reflecting their vivid, autobiographical nature. Emotionally, IAMs often carried a negative tone, particularly sadness, emphasizing their intense and personally significant character.
Unexpected Thoughts (UTs) are marked by unpredictability and cognitive disruption. Their descriptions frequently include emotionally charged words such as “unexpectedly,” “death,” “urge,” and “random.” UTs also showed a predominance of negative emotional content, including fear and sadness, highlighting their intrusive and sometimes distressing impact.
Classification and Emotional Landscape
The machine learning models achieved over 70% accuracy in classifying the thought types based purely on language, demonstrating the power of linguistic analysis in differentiating these internal states. Interestingly, Déjà Vu descriptions were more easily distinguishable, while IAMs and UTs were more frequently confused with each other. This suggests a greater phenomenological overlap between IAMs and UTs, possibly due to their shared emotionally charged and personal nature.
The emotion analysis further reinforced these distinctions. Déjà Vu descriptions were associated with more neutral and joyful language, consistent with its characterization as a curious or slightly positive experience. Both IAMs and UTs, however, tended to contain more negative emotional content, with IAMs leaning towards sadness and UTs exhibiting a broader range including fear and disgust.
Also Read:
- AI and Neurotechnology Combine to Reduce Traumatic Memories
- LEDOM: Unveiling the Power of Reverse Language Models in AI
Broader Implications
This research not only reaffirms existing theories about spontaneous thoughts but also offers a novel computational approach to studying them. By positioning language as a window into spontaneous cognition, the study opens doors for potential applications in various fields. For example, in educational settings, identifying instances of Déjà Vu could signal moments when learners are primed for exploration, while detecting Unexpected Thoughts might indicate a need to refocus attention. In clinical memory assessments, analyzing spontaneous recollections could serve as non-invasive cognitive markers.
While acknowledging limitations such as the reliance on self-reports and the focus on English-speaking participants, this study powerfully demonstrates that language carries distinct signatures aligned with existing psychological theories. It proposes that language-based analysis is a valuable complementary tool for investigating the complex and often elusive world of spontaneous cognition, offering deeper insights into how our internal mental states are expressed.
