Research Help: Why do we overestimate variability when objects are more similar to each other?

[saraConnor]

Hello! I am a first year phd student in a Cog Psyc program. I know this isn't in the biology department, but humans are biological beings so I thought that gaining some insight from some biology grads would be very valuable. My research focus is in the area of ensemble perception -- which is our visual systems amazing ability to accurately and efficiently compute the mean of a set of similar objects. For example, if you are presented with a display of circles varying in size and asked what the average size of the circles are, you will estimate the mean very accurately! This is also referred to as summary statistics. 

My research focuses on the perception of variability. Through a series of experiments where I present a sequential set of 9 objects (different line orientations for one experiment), I have found a consistent bias to overestimate objects when they are more similar to each other (they vary by 1 degree) compared to when they are different from each other (they vary by 5 degrees). For example, if participants see a set of 9 lines that all vary within a range of 1 degree (low variability), they overestimate the variability in those lines by 55% where as they are pretty good at estimating the variability when the lines vary by 6 degrees (high variability). I have thought that maybe people are relying on the first few items or the last few items in a set to make their estimates, but I don't find this when I run the analysis. I also don't really find that they are using a specific group of lines to make their  estimates. It also doesn't seem like they are using a range heuristic to make their estimates. That is, taking the largest and smallest of the set to estimate the variability. 

One thing that I have been thinking about is that we overestimate things that are more similar so that we are able to notices those subtle differences. For example, imagine that you are in the rainforest where there are various shades of greens and browns, meaning that there would be a lot of variability in the colors of greens and browns. It would be more advantageous if you were able to overestimate the variability in greens when they are mostly similar because this would allow you to detect some prey or a predator. My issue with this proposal is that I don't know how to make this into a testable mechanism or theory to explain why we overestimate the variability in objects that are similar. Mostly because I'm not in evolutionary psychology. I am in a vision lab, where we use psychophysics methodologies. 

I haven't found any clear literature that hones in on why people may be biased to overestimate a sequentially presented group of objects (that is, objects that are presented one at a time) that are more similar to each other. Are there any fellow grads (maybe even psychophysicists) out there who may have a plausible mechanisms (neural, statistical, etc) that may explain this? Any help brainstorming would be amazing! Thank you for your time and interest.