Central Tendency Effects in Temporal Interval Reproduction in Autism

Saturday, May 17, 2014
Atrium Ballroom (Marriott Marquis Atlanta)
T. Karaminis1, L. E. Neil1, G. Cappagli1, D. Aagten-Murphy2, G. M. Cicchini3, D. Burr2 and E. Pellicano1, (1)Centre for Research in Autism & Education, Institute of Education, London, United Kingdom, (2)University of Florence, Pisa, Italy, (3)Institute of Neuroscience, Consiglio Nazionale delle Ricerche, Pisa, Italy

Central tendency, the tendency of judgements of quantities (of length, duration, number, color etc.) to gravitate towards their mean value, is one of the most robust effects in perception. In the temporal domain, duration estimates of the same time intervals are shorter or longer depending on whether they are presented in the context of short or long intervals. A computational Bayesian approach has recently been shown to predict this effect in temporal reproduction. According to this model, the central tendency effect reflects the extent to which internal representations of a mean value (prior knowledge about temporal statistics) are integrated with sensory estimates (the likelihood of observations) to generate a final (posterior) perceptual judgement. The more noisy and ambiguous the sensory estimates, the more final perceptual judgements rely on prior knowledge. Priors therefore improve the efficiency of computations by reducing overall noise or error.

A recent theoretical account proposed that attenuated Bayesian priors might be responsible for the unique perceptual experience of autistic people, leading to a tendency to perceive the world more accurately rather than modulated by prior experience. We therefore predicted that children with autism should present reduced central tendency effects compared to typically developing (TD) children of similar age and ability. 


We sought to test this Bayesian model of autistic perception by comparing central tendency effects in children with autism and TD children and modelling their patterns of performance based on an existing computational model.


In an ongoing study, 13 children with autism, aged between 7 and 14 years, and 44 TD children of similar age and ability received two child-friendly tasks, modified from existing studies: (1) a “Ready, Set, Go!” temporal reproduction paradigm with two interval ranges (1006-1536 msec and 1270-1800 msec) and (2) a temporal discrimination task (comparison stimulus of 500 msec). Sixteen typical adults also took part. We compared the three groups on their central tendency effects and on the reliability and accuracy of their temporal estimates. We performed simulations with the Bayesian model to estimate the form of participants’ prior knowledge and how this compares to that required for optimal computations.


Analyses showed a general trend of increased central tendency in children with autism compared to TD children and adults, as well as less reliable and less accurate temporal judgements. Similar differences were seen between TD children and adults. Model simulations suggested that children with autism and TD children generated perceptual judgements based on representations of prior knowledge that were more broad than adults, and in a non-optimal manner.


We show here that the tools of Bayesian inference can be used to model typical children, adults, and autistic children’s temporal reproduction performance. We further demonstrate the, contrary to predictions, children with autism and TD children weight prior knowledge and sensory estimates in a similar fashion.