How Instructions, Learning, and Expectations Shape Pain and Neurobiological Responses

Abstract

Treatment outcomes are strongly influenced by expectations, as evidenced by the placebo effect. Meta-analyses of clinical trials reveal that placebo effects are strongest in pain, indicating that psychosocial factors directly influence pain. In this review, I focus on the neural and psychological mechanisms by which instructions, learning, and expectations shape subjective pain. I address new experimental designs that help researchers tease apart the impact of these distinct processes and evaluate the evidence regarding the neural mechanisms by which these cognitive factors shape subjective pain. Studies reveal that expectations modulate pain through parallel circuits that include both pain-specific and domain-general circuits such as those involved in affect and learning. I then review how expectations, learning, and verbal instructions impact clinical outcomes, including placebo analgesia and responses to pharmacological treatments, and discuss implications for future work.

Keywords: expectations; instructions; learning; neuroscience; pain; placebo.

Figure 1

Brain mechanisms of expectancy-based pain modulation. (a) Neuroimaging studies of expectancy-based pain modulation pair predictive cues with noxious stimulation to determine how pain is influenced by threat, nociception, and decision-making. (b) Functional MRI studies that compare expectancy effects on pain with other aversive modalities (Fazeli & Büchel 2018, Horing & Büchel 2022, Sharvit et al. 2018) indicate that the anterior insula (yellow) is responsive to expectancies across domains, whereas posterior portions of the insula (red) seem to uniquely respond to pain. (c) Although dopaminergic circuits are most often studied in the context of reward and appetitive learning, these circuits are highly implicated in pain and aversive learning, as are opioidergic circuits. Mesolimbic circuits such as the medial prefrontal cortex, medial orbitofrontal/ventromedial prefrontal cortex (mOFC/VMPFC), and the ventral striatum, including the nucleus accumbens, respond to reward stimuli and generally show responses that are inversely related to pain, while the dorsal striatum shows both appetitive and aversive prediction errors (Seymour et al. 2005). Neurons in the periaqueductal gray (PAG) can facilitate or inhibit pain through opioidergic projections to the rostral ventral medulla, nucleus accumbens, and the amygdala (Fields 2006, 2018), and functional connectivity between the PAG and rostral anterior cingulate cortex (rACC) is linked to μ-opioid-based placebo analgesia (Bingel et al. 2006, Eippert et al. 2009a). Finally, the amygdala contains neurons involved in pain unpleasantness (Corder et al. 2019) and might respond uniquely to experiential threat learning (Atlas 2019), although the amygdala is also responsive to reward and reward learning (Murray 2007). All of these regions contain both μ-opioid and dopamine receptors. Panels b and c adapted from images created with Biorender.com.

Figure 2

Experimental approaches to investigate the impact of learning and instructions on pain and analgesia. (a) Error-based pain learning. By pairing neutral cues with noxious stimulation, researchers can measure how expectations develop and vary over time. Here, a visual symbol warns about the delivery of a noxious thermal stimulus via a thermode on the participant’s forearm. Quantitative models of error-driven learning capture how expectations develop as a function of predictions and prediction errors (PEs). Early in learning (pink), heat is unexpected and thus elicits an aversive PE. An appetitive PE (blue) may occur at heat offset, consistent with relief. In late learning (red), cue-outcome contingencies have been reinforced, and the PE shifts to the time of the cue, representing the prediction itself. If a stimulus is delivered that is lower than expected (right), this may generate an appetitive PE and/or a reduction in the aversive PE. Panel a adapted from images created with Biorender.com. (b) Instructed reversal paradigm. By combining verbal instructions with conditioning and contingency reversals, researchers can tease apart their independent contributions to pain, autonomic responses, and brain activation. Here, individuals are instructed that the red cue predicts more pain than the yellow cue. These contingencies are reinforced during conditioning, and then during a test phase each cue is paired with stimulation of the same intensity to test for effects of instructions and learning. Following the test, individuals are instructed that contingencies have reversed, and then the cues are both paired with the same test stimulus. If responses reverse upon instruction, they depend on higher-order knowledge, whereas if they do not, they depend on experiential learning. (c) Balanced placebo design. The balanced placebo design crosses verbal instructions (left) with treatment delivery (right) in a 2 × 2 factorial design to test whether expectancy and treatment have additive or interactive effects on clinical outcomes such as pain.