Contents lists available at ScienceDirect Medical Hypotheses journal homepage: www.elsevier.com/locate/mehy Mirror self recognition as a product of forward models; implications for delusions of body image and visual neglect ☆ Sean O'Connor 568A Jericho Road, Whittier, NC 28789, United States ARTICLE INFO Keywords: Forward models Mirror self-recognition Mirrored-self misidentification Body identity integrity disorder Phantom limbs Gender dysphoria Sensory attenuation Prediction error ABSTRACT Forward models allow individuals to learn to predict the sensory consequences of their own behavior. Social forward models have been proposed as an extension of forward models, allowing individuals to learn to predict the response of another to the individual's own behavior. This article proposes similarly that an individual who treats their reflection as another may learn to predict the behavior of their reflection, offering a new perspective on mirror self-recognition and a potential framework through which to investigate visual delusions. Specifically this article investigates this framework by considering four body image delusions; mirrored-self misidentifica- tion, body integrity disorder (BID), phantom limbs, and gender dysphoria, and two delusions associated with visual neglect; somatoparaphrenia and mirror agnosia. Introduction Forward models have been proposed as a mechanism for the central nervous system to differentiate between sensory input from the en- vironment and sensory input from self-generated behavior [1]. Pick- ering & Garrod [2] use an example of moving a hand to a target to demonstrate how a motor act is proposed to consist of two parallel processes simultaneously triggered by a motor command: “First, it (the motor command) causes the action implementer to generate the act, which in turn leads the perceptual implementer to construct a percept of the experience of moving the hand….Second, it sends an efference copy of the action command to cause the for- ward action model to generate the predicted act of moving the hand….The predicted act then causes the forward perceptual model to construct a predicted percept of the experience of moving the hand.” These two processes allow the individual to differentiate between sensory input from self-generated behavior and sensory input from the environment by comparing the predicted sensory experience with the actual sensory experience. When the prediction generated by the for- ward model = perceptual experience (also referred to as P = E) a phenomenon known as sensory attenuation occurs, dampening atten- tion to the sensory experience and allowing the individual to know that the experience is a result of their own behavior [3]. A common example used to display this sensory attenuation is that an individual cannot tickle himself. According to motor theory this is because the predicted experience of the efference copy for the tickling motor act matches the experience of tickling; thus, P = E and the tickling sensation is canceled [2]. When the predicted sensory experience and actual sensory experi- ence do not match (also referred to as P=/=E), sensory neurons fire, drawing the attention of the individual to this prediction error [3]. This prediction error corrects the forward model such that the next action can be modified. A simple example to demonstrate this is that if a forward model is used to predict where an individual’s hand will be during a grasp, and the hand is a little further to the right than pre- dicted, this prediction error is noted and the next time the individual attempted the same grasp, he would grasp a little further to the right [2]. Thus, forward models can be used for predicting the perceptual outcomes of an action and granting agency of the sensory experience by attenuating the sensory effects of self-motion. Forward models have further been extended beyond simple motor control to social control. As Wolpert, Doya, and Kawato [4] write, “We can consider a similar forward or predictive model for social interaction. In this case another person’s response to my motor commands or communicative behaviour is modelled. Again, dis- crepancies between anticipated and actual behaviour can be used to refine such a model. Therefore, by monitoring one’s own action and the response of others it is possible to learn a predictive model of the likely behaviour of someone in response to our actions.” https://doi.org/10.1016/j.mehy.2019.109292 Received 3 June 2019; Received in revised form 25 June 2019; Accepted 26 June 2019 ☆ No sources of support in the form of grants. E-mail address: spoconno@gmail.com. Medical Hypotheses 130 (2019) 109292 0306-9877/ © 2019 Elsevier Ltd. All rights reserved. T