Lab Talk

Finding an Alternative to Connectionism

Mainstream theory of brain function revolves around patterns of connectivity.  Could the adaptations of neurons be an alternative anchor that provides greater insight?

Today’s mainstream theory assumes that all of the computations of the brain are defined by the connectivity of the neural networks. The units of these computations are the summation and subtraction processes implemented by hyperpolarization and depolarization of cell membranes— which then locally “calculate” the result out of the inhibitory and excitatory inputs received through synapses from other cells.

The overall computation of the brain is then defined by the connectivity matrix among a large number of units. This has led to the theory of connectionism which posits that to create the functions of the brain, all you have to do is get the connections right. There is a certain pattern of connections that is necessary to perceive a grandmother; there is a pattern of connections needed to create working memory; other types of connections are needed for attention; also connectivity patterns have been discussed for making decision making, and even consciousness.

However, in previous posts I have described various challenges in the connectionist framework. I also discussed why I think the theories today don’t work well —and hence, don’t have future.

See related post The Crisis of Computational Neuroscience

Finally, I listed desirable properties of a good theory and tried to demonstrate that theories with such properties exist.  What if we were to abandon connectionism altogether? If connections are not the core aspect of computations and outcome, what other aspect might be more central? If inhibition and excitation are not the central computational mechanism for mental operations, what should we replace them with? What could be an alternative aspect of neural function that one could anchor to?

The goal would be to identify another aspect of the brain’s activity that could explain without conflict a number of brain functions such as  working memory, perception, and spontaneous activity? If we find something and build a theory around that something, this theory would need to provide explanations for multiple phenomena without suffering from the superposition problem.

Focusing on Adaptation

As I have proposed elsewhere (Nikolić, 2015), one alternative is to focus on the fast adaptations of nerve cells. A pervasive phenomenon in neuroscience is that nerve cells adapt to inputs. If you continuously stimulate a cell, be it a neuron or a receptor, they will typically not respond with a continuous level of activity. The cells will rapidly adapt or habituate to the stimulus. Sometimes they will adapt more, sometimes less. But they will adapt. These habituations are often fast as the effects can be seen as soon as after one hundred milliseconds. Sometimes fast adaptations can take longer, i.e., several 100s of milliseconds.

Figures 1 and 2 show examples of fast adaptations visible in the data that I have observed in my own studies.



Figure 1: Example of fast adaptation to visual stimulation in retina: Although the visual stimulus was presented for 500 milliseconds, the neuron spiked intensively only during the first 200-300 milliseconds after which it almost completely ceased to fire. Fast adaptation occurred reliably in all 15 repetitions of the stimulus. The data also show a commonly observed phenomenon of off-responses—the neuron fires vigorously after the stimulus has been removed—which is another phenomenon at odds with the principles of connectionism. (The image is adapted from Nikolić & Gansel, 2012.)


Figure 2. Examples of fast adaptation in visual cortex and an illustration on how much fast adaptation can vary across experiments. The extracellularly recorded multi-unit neural activity of the left reduces its firing rate to less than one third already after about 100 ms. This stark reduction occurs despite the presented grating stimuli being of optimal orientation and not being static but rather slowly drifting across the visual field. The multi-unit on the right was subjected to the same stimulation conditions but kept relatively high levels of activity over longer periods of time. (Adapted from Biederlack et al., 2006.)

The timing of fastest adaptations matches the timing of our thoughts. Our mental operations take about as long as it takes for cells to quickly adapt. Any mental operation cannot take place faster than within about 100 ms. We cannot change our minds—be it percepts, decisions, ideas, thoughts, focus of attention—any faster than that. Therefore, from that perspective fast adaptations match behavioural data better than inhibition and excitation computations do, which are much faster processes.

More importantly, one has to explore how these fast adaptations work. What can they bring to the story of how the brain functions? Some of the key questions it would have to answer are:

  • How would a nervous system perceive using fast adaptations?
  • How would it pay attention?
  • How would working memory be implemented?
  • What would it mean to make a decision in such a system?
  • Do fast adaptations learn and if so, how?
  • Would a system relying on fast adaptations need be spontaneously active?
  • Which advantages would fast adaptations bring over connectionism?
  • What empirical evidence supports the fast-adaptation hypothesis?
  • Can it all be done without a superposition problem that the connectionism suffers from?
  • What is the role of all the connections and inhibition/excitation if they are not central for mental operations?
  • How to impact or enable fast adaptation?


Can one build a brain theory around fast adaptations? I think it is possible. Stay tuned for more detail.



Biederlack, J., M. Castelo-Branco, S. Neuenschwander, D.W. Wheeler, W. Singer and D. Nikolić (2006). Brightness induction: Rate enhancement and neuronal synchronization as complementary codes. Neuron 52, 1073-1083

Nikolić, D. (2015). Practopoiesis: Or how life fosters a mind. Journal of Theoretical Biology, 373, 40-61.[]

Nikolić, D. & K. Gansel (2012). Der Kontext entscheidet. Gehirn&Geist, 5/2012: 20-23.

Further resources:

You can read here more about the hierarchy of adaptive mechanisms [].

Also, here is an introductory animated video explaining the structure of the hierarchy [].


Click here [] to see the entire blog series by Danko Nikolić

Leave a Reply