Frozen by Punishment

“The lower animals, like man, manifestly feel pleasure and pain, happiness and misery.” — Charles Darwin in ‘The Expression of the Emotions in Man and Animals’


Training a dog to heel through jerks often turns into a frustrating and self-defeating exercise. In the face of continuous leash jerks, dogs often start moving increasingly slower and in response to the lagging dog the aversive trainer responds with even more punishment. It is a vicious circle; the slower the dog moves the more it gets punished which makes it even slower. Some sensitive dogs will even stop moving altogether, standing still as the clueless trainer continues to punish them. Aversive trainers will call these dogs stubborn, willful and dominant. Is there a better explanation than the one that blames the dog?

A 2010 paper by Kreitzer et al provides a few answers.

Regulation of parkinsonian motor behaviours by optogenetic control of basal ganglia circuitry” is really the precursor to the study discussed in the previous blog. I present them in reverse order because after reading “Distinct roles for direct and indirect pathway striatal neurons in reinforcement,” I gained deeper appreciation for the older paper.

Basal Ganglia and Direct/Indirect Pathways

Schematic of Direct and Indirect Pathways Seibyl 2012.  Quarterly Journal of Nuclear Medicine and Molecular Imaging 2012 February:56(1):4-16

Schematic of Direct and Indirect Pathways.  Seibyl 2012. Quarterly Journal of Nuclear Medicine and Molecular Imaging 2012 February:56(1):4-16

Direct and Indirect Pathways. Paton & Louie 2012. Nat Neurosciences V15 N6, p 808

Direct and Indirect Pathways.
Paton & Louie 2012. Nat Neuroscience V15 N6, p 808

The basal ganglia are a group of nuclei involved in a number of functions including motor control, muscle tone, action selection (decision-making), reinforcement learning and reward-error prediction. The largest component of basal ganglia is the striatum which receives input from all cortical regions; it also receives input from the thalamus and amygdala. The striatum is mainly (96%) composed of specialized cells called medium spiny neurons (MSN), the two types of two types of MSN differ by their target output and the type of dopamine receptor they carry.

Cells that express D1 dopamine receptor send output mostly to the substantia nigra (SNr) and the internal globus pallidus (GPi), both of which then target the thalamus; this is the direct pathway.

Cells expressing D2 receptor send inhibitory signal to the external globus pallidus (GPe). The GPe then targets subthalamic nucleus (STN) which sends out excitatory signals to  SNr and GPi increasing their inhibitory influence on the thalamus.

These two pathways converge on the thalamus which sends excitatory output to the cortex. The indirect pathway ends up inhibiting thalamus and thus the cortex gets reduced excitatory signals. The direct pathway facilitates thalamic activity through disinhibition – reduced inhibition – thereby raising cortical activity.

It is important to note that both of these pathways are always “on”. Activating one pathway doesn’t shut off the other pathway; rather it is more like one pathway exerts greater influence than the other one.

Regulation of Motor Behavior

The authors (this is still back in ’09) were interested in testing the model describing the direct and indirect pathways. To do this they targeted the dorsomedial striatum using optogenetic protocols described in the last blog.

The one major difference was the extended illumination lasting 30s instead of 1s used in the operant task in 2012. The other difference is the lack of behavioral assays; the goal was to observe changes in motor control. As in the previous case we have 3 populations; the control mice, direct-pathway D1-ChR1 mice and indirect pathway D2-ChR1 mice.

The lines show the movement and the dots indicate the location every 30s. The grey lines show the 20s of activity before illumination and the colored lines indicate activity during and following the 30s of optogenetic activation.Kravitz 2010

The lines show the movement and the dots indicate the location every 30s. The grey lines show the 20s of activity before illumination and the colored lines indicate activity during and following the 30s of optogenetic activation.
Kravitz 2010

The difference is striking. One can immediately see that the distance between RED dots (the researchers should have used a better color scheme, green for “go” and red for “stop”) is much greater than for either the grey or green/D2 dots.

The three graphs below are self-explanatory and I include them for completeness.

Opto - 05 Opto 06Opto 07.

Direct pathway activation:

  • Reduces freezing
  • More frequent locomotor initiation
  • Increases ambulatory time
  • Faster, vigorous fine movement

Indirect pathway activation:

  • Increases freezing time
  • Decreases motor initiation
  • Reduces ambulatory time
  • Bradykinesia

The authors summarize the findings quite nicely:

Together, these data establish a causal role for the direct pathway in decreasing freezing and increasing locomotor initiations, and for the indirect pathway in increasing freezing, decreasing locomotor initiations and inducing bradykinesia.

My Speculative Musings

In the previous blog we learned the direct and indirect pathways mediate reward and punishment. Here we have sustained activation of the same neural pathways regulating locomotor initiations.

Recalling that D2 indirect pathway mediates punishment, I posit that Millan’s dogs are not submissive but in a parkinsonian state and that this freezing effect might be partially involved in the dogs Millan punishes into “submission”.  They are not ‘calm and submissive’, just unable to initiate movement due the prolonged punishment. I will even state – though with weaker conviction because of the conflicting data – that ‘balanced’ trainers sabotage themselves because by using reward and punishment they are pitting these pathways against each other.

Like Darwin said, I may be “wondering beyond my proper bounds” in proposing something beyond the scope of the paper, yet I think I am in safe ground in saying that we can integrate these findings into the general observation that punishment suppresses behavior because it also suppresses movement. (Of course if I am wrong, I’d also like to know about it)

While this study focuses on motor behavior and I’ve selected to restrict the discussion to animal behavior, the basal ganglia is involved in various pathologies. Some of the better known basal ganglia-related disorders include Huntington’s disease, Parkinson’s disease, and dystonia. Recent studies have also found differences in size and shape in the basal ganglia of boys affected with ADHA and autism (Qiu 2009, Qiu 2010); not surprisingly, alterations in motor control and various types cognitive deficits are observed with these disorders. The research to understand the basal ganglia is ongoing as is the search for more effective interventions. [links included for those interested]

My final thought is this: whether it’s a mini-shogi expert choosing the best move or a mouse choosing to press a lever, both human and non-human animals think and make decisions in remarkably similar ways; those who excuse their punitive training methods by saying they are “treating it like an animal” don’t know what the hell they are talking about.


Related articles

Note that anatomists insist in making things difficult. Structures can have different names depending on the discipline and human/no-human distinctions are also made in anatomical names and terms of location.


Kravitz, A., Freeze, B., Parker, P., Kay, K., Thwin, M., Deisseroth, K., & Kreitzer, A. (2010). Regulation of parkinsonian motor behaviours by optogenetic control of basal ganglia circuitry Nature, 466 (7306), 622-626 DOI: 10.1038/nature09159

Qiu A, Adler M, Crocetti D, Miller MI, Mostofsky SH. (2010) Basal ganglia shapes predict social, communication, and motor dysfunctions in boys with autism spectrum disorder. J Am Acad Child Adolesc Psychiatry. 2010 Jun;49(6):539-51, 551.e1-4. doi: 10.1016/j.jaac.2010.02.012.

Qiu A, Crocetti D, Adler M, Mahone EM, Denckla MB, Miller MI, Mostofsky SH. (2009) Basal Ganglia Volume and Shape in Children With Attention Deficit Hyperactivity Disorder. Am J Psychiatry. 2009 January; 166(1): 74–82. doi:  10.1176/appi.ajp.2008.08030426

9 thoughts on “Frozen by Punishment

  1. You have come to some really interesting conclusions. I am glad you summed it up at the end because it was getting beyond my scope of understanding, but it does seem like you have got something very likely in what causes certain behaviors in dogs.

  2. I guess, though, that freezing could be necessary sometimes in the natural world and many parent animals use “punishment” to stop their offspring from being hurt. That’s not to say I think punishment is good. Learning is much quicker using positive reinforcement, but maybe there is a place for some punishment? Confusing!

    • Freezing is the primary mode of defence for young animals but as you saw in the previous blog that explores the same neural pathways in the acquisition of an operant task, the direct pathway was far superior.

  3. Pingback: The Transparent Brain | Science of Dogs

  4. Pingback: Millan is proof that Positive Reinforcement Works | Science of Dogs

  5. Pingback: Exhaustion and Emily the Pitbull | Science of Dogs

  6. Pingback: My Way is Not the Only Way | Science of Dogs

Comments are closed.