The difference between a striking result and a usable tool

Can the ability to perceive one’s own heartbeat help identify a successful trader? The question becomes seductive when a study’s outcome can be expressed as a real-world survival or performance association. But strong narrative appeal should increase methodological scrutiny rather than weaken it. A usable prediction model must identify who will be assessed, what information exists before the outcome, how the outcome will be measured, and whether the relationship survives independent validation. A retrospective association is an observation, not an operational instrument. In a 2016 Scientific Reports study, Kandasamy and colleagues examined interoceptive ability in real-world traders and reported an association with survival on a London trading floor. The setting is valuable because field data avoid some artificial features of a simplified laboratory lottery. Yet survival is shaped by hiring, assignment, capital allocation, desk rules, product specialization, firm conditions and individual experience. An observed association cannot isolate heartbeat perception as the cause of continued employment. [1]

Even the phrase “interoceptive ability” requires a measurement definition. Laboratory heartbeat-detection tasks can be influenced by task instructions, prior beliefs, knowledge of heart rate, counting strategy and measurement reliability. A score in one protocol should not be treated as an innate or permanent property without evidence of stability and construct validity. The right first question is how well the proposed measure represents the construct it names.

Physiological responses during live trading

An earlier study by Lo and Repin measured physiological variables such as skin conductance and cardiovascular responses in ten professional securities traders during live trading and linked them to market events. It found correlations between some responses and transient market conditions, with possible differences by trader experience. The direct field observation matters because it challenges the idea that professional economic decision processes are free of emotional physiology. [2] However, the same findings illustrate why identifying causation is difficult. A volatile market can affect the trader’s physiological state. A trader’s existing position and risk exposure can affect both their physiology and their response to the next price movement. The person can also adjust trading behavior in ways that influence subsequent sensations. Those feedback loops are not resolved simply by observing that two measurements vary together.

With ten participants, generalizability is necessarily limited, and the design cannot on its own justify machine-learning claims about identifying the best traders. A single highly experienced participant can contribute many observations, but the effective independent sample for cross-person predictions remains small. Proper validation must separate the train and test data by participant and, where relevant, by market period rather than randomly splitting adjacent observations that share context.

Experimental manipulation clarifies some arrows but not all

Experiments can sometimes strengthen identification by manipulating a candidate mechanism. A 2015 study investigating cortisol, testosterone and financial risk taking combined hormone measurements in an experimental asset market with pharmacological manipulations in other participants. It reported changes in risk taking and optimism under specified conditions. This is more informative about causal pathways within those experimental treatments than pure field correlation, yet it is still a substantial leap from a short laboratory trading game to long-term professional survival. [3] More generally, different levels of analysis should not be collapsed. Interoceptive performance, cortisol concentration, sympathetic arousal, self-reported anxiety, allocation size, return variance and employment tenure are distinct endpoints. A proposed framework that combines them must specify whether its intended result is explanation of a momentary choice, classification of a physiological state, or prediction of future financial outcomes. Success on one endpoint is not validation for another.

The most common hidden weakness in a high-dimensional composite score is selection. If researchers try many candidate features and report only the best association, the winner will overestimate future performance unless the feature selection itself is included inside a properly nested validation procedure. Market data also contain regime changes and serial dependence; naive cross-validation can allow temporally neighboring observations from the same event into both training and testing. The model may appear to recognize human decision quality while merely recognizing a time period.

A prospective validation program

A stronger design would enroll traders or decision-makers before the relevant performance period, preregister the physiological protocol, retain individual baseline measures and prespecify outcomes. It would compare an interoceptive metric to a simple benchmark using prior experience, product type and exposure constraints. An incremental-prediction question is more informative than asking whether the biomarker is statistically significant by itself. Missing data, attrition and participants who leave for non-performance reasons must be documented rather than silently excluded. Researchers should then distinguish discrimination from calibration. A model may rank two people correctly while assigning unreliable probabilities; it may calibrate in one training population but degrade when used for another market or job type. If the stated aim is to improve choices, a further prospective intervention trial would be needed to test whether using the measurement changes outcomes rather than merely correlates with them. These are separate levels of evidence, each with its own burden of proof.

The answer to the headline question is therefore conditional and limited. The primary studies justify asking whether physiology and interoception contain information relevant to financial decision behavior; they do not establish a clinic-ready test or a profitable trading indicator. NSI is proposed, not a clinically or financially validated prediction system. Its credibility would come from transparent protocols, precommitted hypotheses, external replication and willingness to report null results—not from treating an intriguing correlation as an institutional credential.

Scientific bibliography & sources

  1. Kandasamy et al., Interoceptive Ability Predicts Survival on a London Trading Floor, Scientific Reports (2016). https://www.nature.com/articles/srep32986
  2. Lo and Repin, The Psychophysiology of Real-Time Financial Risk Processing, NBER Working Paper 8508 (2001). https://www.nber.org/papers/w8508
  3. Cortisol and testosterone increase financial risk taking and may destabilize markets, PubMed PMID 26135946 (2015). https://pubmed.ncbi.nlm.nih.gov/26135946/

Research note

NSI is a proposed research framework, not an independently validated medical or financial prediction system. This article is research and education, not individualized medical or investment advice. Market observations are timestamped historical context, not live prices.