Longitudinal effects of early psychosocial deprivation on macaque executive function: Evidence from computational modelling

Abstract

Executive function (EF) describes a group of cognitive processes underlying the organization and control of goal-directed behaviour. Environmental experience appears to play a crucial role in EF development, with early psychosocial deprivation often linked to EF impairment. However, many questions remain concerning the developmental trajectories of EF after exposure to deprivation, especially concerning specific mechanisms. Accordingly, using an 'A-not-B' paradigm and a macaque model of early psychosocial deprivation, we investigated how early deprivation influences EF development longitudinally from adolescence into early adulthood. The contribution of working memory and inhibitory control mechanisms were examined specifically via the fitting of a computational model of decision making to the choice behaviour of each individual. As predicted, peer-reared animals (i.e. those exposed to early psychosocial deprivation) performed worse than mother-reared animals across time, with the fitted model parameters yielding novel insights into the functional decomposition of group-level EF differences underlying task performance. Results indicated differential trajectories of inhibitory control and working memory development in the two groups. Such findings not only extend our knowledge of how early deprivation influences EF longitudinally, but also provide support for the utility of computational modelling to elucidate specific mechanisms linking early psychosocial deprivation to long-term poor outcomes.

Keywords: computational modelling; early psychosocial deprivation; executive function; inhibitory control; macaque; working memory.

Figure 1.

Task set-up. (a) A frontal view of the test enclosure. (b,c) Two views from above: (b) the wells are closed with the yellow arrows indicating a sliding mechanism; (c) the wells are open and a piece of food is placed in one of them.

Figure 2.

Task procedure. The series of images illustrates the sequence of a single trial. (a) Experimenter 1 shows the food to the subject, then places it in the well, and experimenter 2 blocks the vision of the subject during the delay period. (b) A ‘correct’ choice, with the subject reaching for the well where the food was hidden and then eating the food. (c) An ‘incorrect’ choice, with the subject choosing the well containing no food and experimenter 1 then highlighting where the food was actually hidden.

Figure 3.

Model architecture. The model included a decaying working memory trace of the food location (top left, the plot shows the influence of decay rate: with higher decay rates, the influence of working memory diminishes faster) and a decaying trace of the previous choice (bottom left, the plot shows the influence of decay rate: with higher decay rates, the influence of the previous choice diminishes faster). The weighted sum of these influences is used to compute the probability of choosing one side or the other using a softmax function (the plot shows the influence of the inverse softmax temperature parameter: at higher values, the decision is more deterministic).

Figure 4.

Task performance. The mother-reared group had a significantly higher percentage of correct responses (a), cumulative score (b) and maximum delay (c) in the original ‘A-not-B’ task than the peer-reared group at both time-points. Mother-reared animals also had a significantly higher percentage of correct responses in the randomized control task at the second time-point (d), and the performance of each group was not different between tasks.

Figure 5.

Fitted model parameters for the original ‘A-not-B’ task at both time-points. (a) Decay in the value of the working memory factor at the first time-point, given λ₁, and fixing w₁ = 1, for various delay durations. The solid lines represent the mean over subjects (red = mother-reared, blue = peer-reared), and the shaded areas represent the standard error. (b) As in (a), for the second time-point. (c) Decay in the value of the choice history factor rate at the first time-point, given λ₂, and fixing w₂ = 1, for various ITIs. (d) As in (c), for the second time-point.

Figure 6.

Fitted model parameters for the original and random control tasks at the second time-point. (a) Decay in the value of the working memory factor for the original ‘A-not-B’ task at the second time-point, given λ₁, and fixing w₁ = 1, for various delay durations. The solid lines represent the mean over subjects (red = mother-reared, blue = peer-reared), and the shaded areas represent s.e. (b) As in (a), for the random control task at the second time-point. (c) Decay in the value of the choice history factor rate for the original ‘A-not-B’ task at the second time-point, given λ₂, and fixing w₂ = 1, for various ITIs. (d) As in (c), for the random control task at second time-point.