The power of prediction reduces activation in the primary visual cortex

Prediction is an invaluable skill for navigating through complex environments. Somehow the brain generates predictions about perceptual inputs it's likely to receive using contextual information from recent memory. Statistical regularities are learned (e.g. movement and attack patterns of Mega Man bosses) and lead to less activation in corresponding brain areas. The brain is truly a miserly organ. "Why put in more work than I have to when I know what's gonna happen next", says the brain.

Alink and colleagues over at the Max Planck Institute in Germany decided to check out what's really going on in the brain when it's making visual predictions.

Using fMRI, the team tested whether predictability reduced responses in the human visual cortex as put forth in Rao and Ballard's 1999 model of predictive coding. They assessed the theoretical claim by looking at the response of V1 (primary visual cortex) when detecting predictive and non-predictive motions.

In the first experiment the researchers had 12 healthy participants fixate their eyes on a screen during which 9 degree bars were presented above and below the fixation point. The bars induced an upward and downward long-range apparent motion. Their motions were either predictable (bar is positioned and timed exactly on the motion trajectory of linear apparent motion between the lower and upper bar stimuli) or unpredictable (bar appears at a time at which linear apparent motion had already passed the position of the test bar). A baseline condition was also presented where apparent motion, but no bar was presented. There were 81 trials for each condition, each trial lasting 7 seconds.

The second experiment differed in that the participants saw dots rather than bars through hi-tech MRI compatible goggles with two organic light-emitting diode displays. Presenting the MR Vision 2000 Ultra:

MR Vision 2000 Ultra is an excellent way of entering the world of visual stimulation for fMRI. Featuring dual, backlit digital displays, MR Vision 2000 avoids gross pixellation, pixel dropout and color aberration endemic to fiber-optic based systems...The lightweight, headmounted display fits entirely inside the bore of the magnet and is capable of deliveringfield-sequential 3D or standard video at a resolution of 180,000 pixels. This breakthrough advance utilizes the same cutting-edge technology that was developed for our research-proven fMRI 3D Goggle systems. Now its possible to immerse your patients in breathtaking 3D video! (I can't seem to find pictures of this contraption anywhere)

The authors indeed found that in both experiments the stimulus predictability reduced activation levels in V1. The first experiment showed that stimuli with a predictable onset caused lower activation of V1 compared to identical stimuli with unpredictable onset. Experiment 2 showed that responses in V1 and the hMT/V5+ (human visual motion area) were lowest when apparent motion predicted the direction of random dot motion. Visual responses increased as the motion was made less predictable.

Unfortunately, due to crappy fMRI temporal resolution, they couldn't figure out whether reduced activations in V1 were the result of feedback, local V1 processing, or an interaction between the two.

They conclude that their findings "provide strong empirical evidence for the idea that the visual cortex actively anticipates its visual input and that such anticipation allows predictable stimuli to be processed with less neural activation at the earliest cortical relay for visual processing".

As I was reading this paper, I was reminded of the neural efficiency hypothesis where more intelligent people were found to have less activation of cortical areas when performing cognitively easy tasks compared to their less intelligent counterparts. I can't shake the idea that the power of prediction is somehow inextricably tied to intelligent life. Seems to intuitively make sense doesn't it? The better the ability to accurately predict -> the better the planning -> the less likely to get zapped by a raygun.

Bet you can't beat that pesky Bubble Man as fast as I can!

Alink, A., Schwiedrzik, C., Kohler, A., Singer, W., & Muckli, L. (2010). Stimulus Predictability Reduces Responses in Primary Visual Cortex Journal of Neuroscience, 30 (8), 2960-2966 DOI: 10.1523/JNEUROSCI.3730-10.2010

Doppelmayr M, Klimesch W, Sauseng P, Hödlmoser K, Stadler W, & Hanslmayr S (2005). Intelligence related differences in EEG-bandpower. Neuroscience letters, 381 (3), 309-13 PMID: 15896490