Poker Neuroscience: On Neural Tells

In poker, a player can gain an advantage if she can detect tells, subtle behaviors from other players that communicate the strength (or weakness) of their hands. Tells are not always reliable (e.g., clumsiness could be misinterpreted as anxiety about a bad hand) so veteran players use them cautiously.

In fact, it is hard to find a behavior that, on its own, predicts what someone is thinking. It is almost always a collection of behaviors, persistent across time and situations, that suggest a psychological state. Wouldn’t you be more convinced that someone loved you if they expressed this emotion in 100 ways instead of one? Individual behaviors are rarely good tells.

Yet neuroscience commentators often claim to write about neural tells: individual neural measurements that are highly reliable indicators of psychological variables like love (e.g., the oxytocin myth). But neural tells are Procrustean beds because, as I explain below, they ought to be significantly less reliable than behavioral tells.

Levels of Analysis

Psychological variables, like love, cannot be measured directly (after all, what is love?). Scientists simply search for tells, measurable variables that ought to correlate with the psychological variable. For example, the number of poems a couple write each other can be a tell of the strength of their love. But the correlation between the psychological variable and the tell weakens when these two variables exist in different levels of analysis.

For instance, brain activity can be measured at multiple levels of analysis. At an electro-physiological level, scientists can measure voltage changes that follow neural activity. At a hemodynamic level, scientists can measure oxygen level changes in the blood around active neurons. In this example, brain activity is examined at one of two levels of analysis.

Mixed Levels…

A scientist can mix levels of analysis. She can hypothesize about electrophysiology but measure hemodynamic activity as a tell. For example, she can use functional magnetic resonance imaging (fMRI) to measure oxygen concentration changes in the brain’s blood, but hypothesize that these changes reflect fluctuations in the brain’s electrical activity.

Luckily for her, electrical and hemodynamic activity are highly correlated (Logothetis & Wandell, 2004): the presence of one very often indicates the presence of the other. In turn, hemodynamic activity is a great tell for neural activity. Unfortunately, this is not the case in most situations that involve hypotheses and variables that mix levels of analysis.

…Make Tells Less Reliable

Variables often lack one-to-one relationships across levels of analysis. For example, love (psychological level) cannot be mapped to a single action (behavioral level). No behavior by a person indicates, always and exclusively, that she feels in love. Using this logic, neuroscience commentators should not expect to find highly reliable neural tells.

Most likely, it is a collection of neural measurements that should predict behavior and psychological variables. A variable in a higher level of analysis is almost always comprised of multiple variables in a lower level of analysis. So even a single emotion like love ought to have a complicated neural basis that is constituted by more than one neural measurement.

Takeaways

• Neuroscience commentators claim to write about neural tells, individual neural measurements that are highly reliable indicators of psychological variables
• But variable relationships across levels of analysis (behavioral to psychological, neural to psychological) are not one-to-one, making neural tells unlikely to exist
• More often than not, it should be a collection of neural measurements that, together, predict psychological variables

The Human Brain Is Complex

Scientists constantly marvel at the intricacy of the human brain, “the cathedral of complexity” (Coveney & Highfield, 1995, p. 279). Here, I share some facts about the brain to make clear its complexity and, in turn, the difficulty that neuroscientists face in understanding it.

Size and Weight

The average human brain is a small, lightweight object, measuring about 73 in3 (1,200 cm3; Cosgrove, Masure, & Staley, 2007) and weighing about 3.3 lb (1.5 kg; Herculano-Houzel, 2009). But these summary statistics belie the significant variability in brain size and weight across individual people. For example, brain size differs by demographic attributes like age, race, sex, and social class (Rushton & Ankney, 1996).

Cortical Shape

The brain’s cortex is shaped in gyri (ridges) and sulci (fissures). The organization of gyri and sulci show significant variability between brains (Ono, Kubik, & Abernathey, 1990) and even between hemispheres in the same brain (Brett, Johnsrude, & Owen, 2002). Moreover, not all cortical regions show the same extent and type of variability (Thompson, Schwartz, Lin, Khan, & Toga, 1996).

Cells

The brain is composed of approximately 170 billion cells, about half of which are neurons; the rest are glial cells (Azevedo et al., 2009). Neurons communicate information with electrical and chemical signals (Purves et al., 2001). Glial cells serve many functions, from the production of neurotransmitters to maintenance of the blood-brain barrier (Oberheim, Wang, Goldman, & Nedergaard, 2006).

Connectivity

The brain’s neurons form a densely-connected network with connections in the order of 1 quadrillion (10¹⁵; Murry & Sturde, 1995). The strength and existence of many of these connections are not fixed (Sporns, Tononi, & Kötter, 2005); they change across development as a function of normal maturation and specific experiences (Holtmaat & Svoboda, 2009).

Chemistry

Neurotransmitters, of which more than 100 have been identified, help neurons communicate information (Purves et al., 2001). The amount and behavior of these chemicals varies across individual people, including systematic differences as a function of age (Rosene & Nicholson, 1999), sex (Zaidi et al., 2010), and mental health (Charney, Buxbaum, Sklar, & Nestler, 2013).