Prevention and treatment strategies for contextual overgeneralization

At the core of anxiety disorders lies the tendency to generalize fear from a threatening to a safe situation. A deeper understanding of the mechanisms that facilitate and restrain generalization in humans is therefore needed. Rodent studies showed that pre-exposure to a context that is similar to the threatening context enhanced generalization to the similar context. In Experiment 1 we replicated these animal findings in humans (US-expectancy). Studies on the underlying mechanisms showed that the pre-exposure representation was recalled during conditioning (due to similarity between the contexts) and the shock also became linked to the recalled representation, resulting in greater generalization. In Experiment 2 we developed a pre-exposure procedure that increased the ability to distinguish between the conditioned and pre-exposure contexts, such that presentation of the former would no longer result in recall of the latter. We then observed that overgeneralization (US-expectancy) was prevented. Pre-exposure did not affect generalization of skin conductance response or fear potentiated startle. Finally, exploratory analyses revealed that increased generalization (US-expectancy), if not prevented, could be reduced by a reminder of the conditioned context. Hence, we developed a prevention- and a treatment-strategy for overgeneralization. These findings may guide the development of new therapeutic strategies.

Test. There was a trend for an increase in differential responding between the conditioned context and the control context from the last trial of acquisition (day 2) to the test trial (day 3) (context x trial; F (1, 38) = 3.04, p = .089, η² p = .07), indicating that acquired differential ratings persisted during test. The groups did not differ in differential ratings from acquisition to test (context x trial x group; F (2, 38) < 1, p = .39, η² p = .05).

Fear potentiated startle (FPS)
Pre-exposure. Contrary to expectation we observed higher startle responding to the control context compared to the pre-exposure context (main effect context; F (1, 38) = 25.23, p < .001, η² p = .40). Note that this difference was observed regardless of type of pre-exposure context (context A, B, or X), since there were no differences between groups in startle responding during pre-exposure (context x group; context x trial x group; F s < 1.29, p s >.29, η² p <.04) (Supplementary Fig. S2A-C).
Conditioning. Differential responding changed from the first to the second trial of conditioning (context x trial; F (1, 36) = 4.89, p < .033, η² p = .12) that did not differ between groups (context x trial x group; F (2, 36) = 1.20, p > .31, η² p = .06). Startle responding to the control context was higher compared to the conditioned context on the first trial (main effect context; F (1, 36) = 7.08, p < .012, η² p = .16) and was absent on the second trial (main effect context; F (1, 36) < 1, p > .43, η² p = .02). Thus, the unexpected difference in startle responding during pre-exposure persisted on the first trial of conditioning and acquisition of startle responding was not observed

US-expectancy ratings
Conditioning. We observed context conditioning of US-expectancy ratings, evidenced by an increase in differential US-expectancy ratings from the first to the second trial of conditioning (context x trial; F (1, 32) = 102.30, p < .001, η² p = .76). The groups did not differ in acquisition of US-expectancy (context x trial x group; F (1, 32) < 1, p = .23, η² p = .05)( Supplementary Fig. S4A,B).

Test.
There was an increase in differential responding between the conditioned context A and the control context from end of acquisition (trial 2, day 2) to the test trial (day 3) (context x trial; F (1, 32) = 9.01, p = .005, η² p = .22), that did not differ between groups (context x trial x group; F (1, 32) < 1, p = .72, η² p = .00). This indicates that acquired differential ratings persisted during test.

Skin conductance response (SCR)
Pre-exposure. Responding to the pre-exposure contexts and the control context did not differ between groups (context x group; context x trial x group; F s < 1.22, p s > .31, η² p < .05) ( Supplementary Fig. S5A,B). We observed no difference in skin conductance responses to the pre-exposure contexts (B and C/X) and the control context on trials 1 to 3 of pre-exposure (main effect context; F (2, 44) < 1, p = .68, η² p = .02) ( Supplementary Fig. S5A,B).

Supplementary Figures
Experiment 1 Figure S1. Mean US-expectancy ratings to context A and the control context during conditioning (day 2) for the pre-exposure A (n = 14) (A), pre-exposure B (n = 14) (B), and pre-exposure X (n = 13) (C) groups. Error bars represent s.e.m.  Figure S2. Mean startle response during pre-exposure on day 1 (pre-exposure contexts and control context) and startle responses to context A and the control context during conditioning on day 2 for the pre-exposure A (n = 14) (A), preexposure B (n = 14) (B), and pre-exposure X (n = 11-13) (C) groups. Light grey triangles represent responses to the pre-exposure context that differed between the groups (context A, context B, or context X); grey squares refer to the control context; black diamonds represent the conditioned context A. Error bars represent s.e.m.  Figure S5. Mean skin conductance response (SCR) during pre-exposure on day 1 (pre-exposure contexts and control context) and responses to context A and the control context during conditioning on day 2 for the pre-exposure BC (n = 11-15) (A) and pre-exposure BC (n = 10-13) (B) groups. Light grey triangles symbolize responses to the pre-exposure context that differed between the groups (context C or context X), white circles represent context B that was presented in both groups; grey squares refer to the control context; black diamonds represent the conditioned context A. Error bars represent s.e.m.  Figure S6. Mean startle response during pre-exposure on day 1 (pre-exposure contexts and control context) and startle responses to context A and the control context during conditioning on day 2 for the pre-exposure BC (n = 17) (A) and preexposure BX (n = 15-17) (B) groups. Light grey triangles represent responses to the pre-exposure context that differed between the groups (context C or context X); white circles symbolize responses to context B that was presented in both groups; grey squares refer to the control context; black diamonds represent the conditioned context A. Error bars represent s.e.m.