Learning models of PTSD: Theoretical accounts and psychobiological evidence

Abstract

Learning abnormalities have long been centrally implicated in posttraumatic psychopathology. Indeed, of all anxiety disorders, PTSD may be most clearly attributable to discrete, aversive learning events. In PTSD, such learning is acquired during the traumatic encounter and is expressed as both conditioned fear to stimuli associated with the event and more general over-reactivity-or failure to adapt-to intense, novel, or fear-related stimuli. The relatively straightforward link between PTSD and these basic, evolutionarily old, learning processes of conditioning, sensitization, and habituation affords models of PTSD comprised of fundamental, experimentally tractable mechanisms of learning that have been well characterized across a variety of mammalian species including humans. Though such learning mechanisms have featured prominently in explanatory models of psychological maladjustment to trauma for at least 90years, much of the empirical testing of these models has occurred only in the past two decades. The current review delineates the variety of theories forming this longstanding tradition of learning-based models of PTSD, details empirical evidence for such models, attempts an integrative account of results from this literature, and specifies limitations of, and future directions for, studies testing learning models of PTSD.

Keywords: Extinction; Fear-conditioning; Habituation; Overgeneralization; PTSD; Sensitization.

Figure 1

Schematic display of the competing nature of acquisition (ACQ) and extinction (EXT) memories of the conditioned stimulus (CS). In this example, ACQ of conditioned fear follows from presentations of the CS together with an electric-shock US during a morning (AM) testing session. The EXT memory is encoded later that afternoon (PM) during a testing session in which the CS is repeatedly presented in the absence of the US. These ACQ and EXT learning experiences generate two competing memories of the CS, with the former representing the CS as an anxiogenic predictor of aversive shock and the latter characterizing the CS as a benign stimulus associated with no aversive outcome. Activation of the ACQ memory of the CS elicits fear, while activation of the EXT memory of the CS elicits no fear. Successful EXT learning depends on an EXT memory of the benign CS that is strong enough to outcompete the ACQ memory for excitation, thereby inhibiting both activation of the ACQ memory and the associated fear reactivity to the CS.

Figure 2

(Adapted from Grillon et al., 2009) Average startle magnitudes in PTSD patients and healthy controls across three levels of contextual aversiveness: 1) neutral context (low threat), during which no unpleasant stimulation was delivered, 2) predictable context (moderate threat) during which unpleasant stimulation was given predictably, and (3) unpredictable context during which unpleasant stimulation could be given at any time.

Figure 3

(A) Conditioned and generalization stimuli included 10 rings of gradually increasing size with extremes serving as the conditioned danger-cue paired with electric shock (CS+) and the conditioned safety-cue unpaired with shock (CS−). For half of participants (Group 1), the largest ring was the CS+ and the smallest was the CS−, and for the other half (Group 2) this was reversed. The eight rings of intermediary size served as generalization stimuli (GSs) and created a continuum-of-similarity from CS+ to CS−. Before analysis, responses to every two intermediaries were collapsed into a single class of stimulus, leaving four classes of generalization stimuli (GS₁, GS₂, GS₃, GS₄). Collapsing was implemented to avoid an unduly large number of trials while maintaining a gradual continuum-of-size across rings. Immediately following presentation of each ring, participants were asked to rate their perceived risk for receiving a shock on a scale of 0–2 (where 0 = no risk, 1= some risk, and 2= a lot of risk) to assess the degree to which the conditioned threat value of the CS+ generalized to resembling GSs. (B) Generalization results for 13 adult PTSD patients with mixed traumatic histories (sexual abuse [n=2], physical assault [n=4], car accident [n=3], other [n=4]) and 13 age- and sex-matched healthy comparisons indicate less steep gradients of perceived risk in the PTSD group (Group × Stimulus-type interaction, p=.003). Though this suggests overgeneralization in patients, higher overall risk ratings in PTSD patients (p=.02) imply the additional influence of non-associative sensitization on the shape of response gradients in patients.