‘How does it make you feel?’ Reclaiming subjectivity in neuroscience

‘The true method of psychological investigation is the inner observation of one’s own mind, for in no other way can we reach the direct experience of consciousness’.

—Wilhelm Wundt, 1912

Neuroscience has made extraordinary progress since Wundt wrote these words over a century ago. The mapping of the connectome has enabled a detailed visualisation of brain connectivity (Fan et al., 2022), 7T MRI now captures the brain’s architecture with unprecedented precision (Okada et al., 2022), and brain organoids allow insights into early neurodevelopment (Birtele et al., 2025). These breakthroughs have pushed the field forward, seemingly making Wundt’s words obsolete in favour of objectivity and methodological rigour.

Indeed, searching Web of Science reveals that among all articles published in the top 10 most-cited neuroscience journals over the past 20 years (2005–2025), only 1.8% included the term ‘self-report*’ in their title or abstract. This omission reflects an implicit yet, in our view, widespread belief that self-reports or subjective measures lack validity, and they are something to be avoided.

As early-career researchers, we often encounter scepticism towards self-reports, which are frequently dismissed as ‘merely subjective’, implicitly equating subjectivity with being ‘non-scientific’, while neural activity and behavioural observations are perceived as ‘real’ or objective (Ellia et al., 2021). In our experience, manuscripts are occasionally desk-rejected due to the use of self-reports as key variables, despite strong theoretical justifications for their inclusion. Therefore, when selecting optimal methods for our scientific endeavours, the temptation to abandon subjective measures in favour of objective ones can be hard to resist.

While objective measures are undeniably invaluable, we argue that in pursuit of objectivity, the individual’s subjective experience, often constituting the very essence of our field, is sometimes lost. In our research area, core constructs such as loneliness, empathy, and even psychiatric symptoms are rooted in intrapersonal experience, and inherently subjective, requiring input from one’s lived experience. For example, some individuals may live alone but still feel a strong sense of social belongingness, while others, surrounded by friends and family, may feel profoundly isolated. Indeed, neuroimaging studies have shown stronger associations between brain structure and subjective statements about loneliness (e.g. ‘No one really knows me well’) compared to objective social metrics such as number of friends (Düzel et al., 2019; Kanai et al., 2012).

Psychiatric diagnoses, too, rely heavily on subjective criteria. For instance, symptoms of major depressive disorder, such as ‘depressed mood’, ‘feelings of worthlessness’, or ‘recurrent thoughts of death’ (American Psychiatric Association (APA), 2013), are inherently rooted in the individual’s lived experience (Kyzar and Denfield, 2023), making self-report measures indispensable. Of evidence, large-scale studies combining behavioural data with neuroimaging or biomarkers found self-reports to be among the top predictors of mental health disorders (De Lacy et al., 2023), and it is an accessible, widely used, and recommended tool to support assessment of psychiatric morbidity (APA, 2013).

In these examples, neurobehavioural proxies constitute only indirect measures of the phenomena we aim to assess, whereas self-reports may outperform predictions, underscoring the value of the subjective experience. For example, although childhood maltreatment is a well-established risk factor for psychopathology, objective records of childhood maltreatment are only weakly predictive of psychopathology once subjective assessment of such events, measured using self-reports, are considered, with the latter strongly predicting both current and future psychopathology (Danese and Widom, 2020).

Subjective measures including self-reports, like all measures, are imperfect. They are prone to biases such as memory distortions and social desirability (McDonald, 2008). However, these do not invalidate them. Rather, they highlight the need for thoughtful design and integration with objective measures and complementary methods. Real-time assessment methods like Ecological Momentary Assessment reduce memory bias (Verhagen et al., 2022); combining reports from multiple informants can minimise self-reference biases (An, 2022); and leveraging AI-driven language models has the potential to accurately capture narratives of individuals’ subjective experiences (Skeggs et al., 2025). We should also strengthen psychometric rigour by adopting or developing comprehensive, validated questionnaires and continuously re-evaluating their reliability and stability.

As neuroscience continues to push the boundaries of resolution, data richness, and computational power, we argue that the subjective experience is not noise, but the signal. As such, it is time to overcome the aversion and embrace subjective measures and self-reports, not as a compromise, but as a necessary component for understanding the brain in context.

Ultimately, when it comes to the highly diverse, personal, and internal experiences studied in our field, we believe that no one is more expert than the individuals themselves. In these cases, even a thousand brain scans should not replace the simple, humble question that we as scientists should continue to ask: ‘how does it make you feel?’