The Index of Intrusion Control (IIC): Capturing individual variability in intentional intrusion control in the laboratory

Abstract

Intrusive memories can be downregulated using intentional memory control, as measured via the Think/No-Think paradigm. In this task, participants retrieve or suppress memories in response to an associated reminder cue. After each suppression trial, participants rate whether the association intruded into awareness. Previous research has found that repeatedly exerting intentional control over memory intrusions reduces their frequency. This decrease is often summarised with a linear index, which may miss more complex patterns characterising the temporal dynamics of intrusion control. The goal of this paper is to propose a novel metric of intrusion control that captures those dynamic changes over time as a single index. Results from a mega-analysis of published datasets revealed that the change in intrusion frequencies across time is not purely linear, but also includes non-linear dynamics that seem best captured by a log function of the number of suppression attempts. To capture those linear and non-linear dynamics, we propose the Index of Intrusion Control (IIC), which relies on the integral of intrusion changes across suppression attempts. Simulations revealed that the IIC best captured the linear and non-linear dynamics of intrusion suppression when compared with other linear or non-linear indexes of control, such as the regression slope or Spearman correlation, respectively. Our findings demonstrate how the IIC may therefore act as a more reliable metric to capture individual differences in intrusion control, and examine the role of non-linear dynamics characterizing the conscious access to unwanted memories.

Keywords: Index of Intrusion Control (IIC); Intentional memory control; Intrusions; Think/No-Think task.

Fig. 1

Variability in intrusion control across repeated suppression between studies. The violin plots display data from four studies (Davidson et al., : n = 17; Harrington et al., : n = 29; Mary et al., ; n = 73; Legrand et al., [study 2]: n = 24). The means (indicated by the black bar) and the distribution of the data demonstrate large variability within each block of the T/NT phase both within and between studies. Each graph was made for intrusion responses to neutral stimuli

Fig. 2

Annotated line graph of example intrusion frequencies over 10 blocks to demonstrate the formula used to calculate the Index of Intrusion Control. The IIC calculates the total area within the repeated measures by combining the area between each time point (crossed rectangles) and the increase or decrease between time points (white triangles)

Fig. 3

Example data presenting intrusion frequencies across 10 vs 5 blocks vs IIC for three participants from Harrington et al (2021). A The intrusion frequencies across 10 blocks. For participant 1, the slope has an increased fit (R² = .67) compared to participants 2 and 3, who have a low fit or no fit for the observed data (R² =.24 and R² =.00, respectively). B The same data condensed into five fewer blocks. For participants 1 and 2, condensing the data increases linearity and improves the fit of the slope when compared to 10 blocks. For participant 3, the slope fit does not improve. C The IIC. The area used to calculate the value is blocked out in grey

Fig. 4

Outline of the model used to simulate intrusion control. The formula (A) simulates the frequency of intrusions (I) across T/NT blocks (t) using both linear (β_linear) and log-non-linear (β_nonlinear) terms. The sampling distribution of these suppression terms is calibrated from the combined real datasets (B). Data are then simulated for 50 participants across 8 blocks (C) after adding 5% uniform noise. The various metrics of intrusion control are computed and compared to the suppression factor used to generate the data to understand which metric best captures intentional suppression of intrusions