Jul 08 2008

Smiling Babies, fMRI, Brain Modules, and Neural Networks

I have two daughters, about to turn nine and six. They are, in my completely subjective and biased assessment, the most adorable things in the universe. They evoke in me a powerful and complex set of emotions – an experience that every parent understands and no non-parent can truly appreciate.

Despite concerns about the testability of evolutionary psychological explanations, it seems obvious that such a reaction in a parent would be favored by natural selection, as would be any features in a child that provoked such a response from their parents. This in turn suggests that much of the response of a parent to their child is hard-wired in the brain and genetically determined. This doesn’t rule out cultural and learned influences, it merely suggest that a strong parenting tendency will be coded in the genes.

A recent bit of research supporting this notion was published in the latest issue of Pediatrics: What’s in a smile? Maternal brain responses to infant facial cues. The study uses fMRI, which images blood flow to the brain from which the relative activity of various brain regions can be inferred, to measure the reaction of mothers to various pictures – their child happy, neutral, and sad and another child happy, neutral, and sad. The results:

Key dopamine-associated reward-processing regions of the brain were activated when mothers viewed their own infant’s face compared with an unknown infant’s face. These included the ventral tegmental area/substantia nigra regions, the striatum, and frontal lobe regions involved in (1) emotion processing (medial prefrontal, anterior cingulate, and insula cortex), (2) cognition (dorsolateral prefrontal cortex), and (3) motor/behavioral outputs (primary motor area). Happy, but not neutral or sad own-infant faces, activated nigrostriatal brain regions interconnected by dopaminergic neurons, including the substantia nigra and dorsal putamen. A region-of-interest analysis revealed that activation in these regions was related to positive infant affect (happy > neutral > sad) for each own-unknown infant-face contrast.

In order to interpret what this study tells us I must first back up and discuss some basic principles.


I have written a great deal about fMRI studies. This is the latest state-of-the-art neuroscience research tool. But in order to interpret studies like this it is important to understand the limitations of this research tool. First, as I stated above, fMRI measures blood flow to the brain. Brain tissue that is more active because the neurons are firing will require more blood flow, and therefore blood flow can be used to infer relative brain activity. The theory is that when a subject is performing some task the fMRI will show which parts of the brain are involved in that task.

The primary limitation of this use of an fMRI is that it is impossible to know what is going on inside someone’s mind. We cannot know what a subject is actually thinking, feeling, remembering, or attending to. We can only make a rough estimate based upon one of two basic strategies: we can expose the subject to some stimuli and then image their response to that stimuli, or we can make them perform a task. This is a reasonable approach – if someone is looking at a picture, regardless of what else might be going on in their mind, at least we can know they are looking at the picture.

But all the uncontrolled bits of mental activity will act like noise concealing the signal of interest. To control for this researchers generally look at multiple trials of multiple individuals and then use statistical analysis to pick the signal out of the noise – what brain region activity do the subjects have in common? Again, this is a reasonable approach, but it is important to understand how tricky and difficult using fMRI is. Without a solid protocol and careful analysis the blotches of computer enhanced colors produced by the fMRI are little more than a Rorchach test – researchers can see whatever they want in them. It also means that results should be reproducible before we put too much faith in them. Having said that, I think fMRI can be a powerful tool when used properly to help understand how the different parts of the brain are hooked up and interact with each other.

Brain Modules vs Neural Networks

The results of fMRI studies are only meaningful when they are interpreted within a paradigm of brain function and organization. Right now neuroscientists are working to form a consistent and predictive model of brain organization. At one end of the spectrum there are those who advocate what some call the modular brain – specific anatomical parts of the brain serve specific functions. They are like modules – the fear module causes fear when activated, the anxiety module produces anxiety, and the subtraction module performs subtraction calculations, etc.

The neural networks paradigm, on the other hand, emphasizes the network of connections between different parts of the brain as primarily important to function. In this model there is no anxiety module, but rather a certain pattern of networks in a specific individual will produce anxiety. But some of the same parts of the brain active during anxiety may also take part in other networks that will produce other emotions, like happiness.

The modular paradigm works a bit better for fMRI studies. If a part of the brain lights up during an activity then researchers can conclude that that part of the brain serves the function under study. The neural network paradigm makes fMRI studies more difficult (but not impossible) to interpret – the parts of the brain that light up are involved in the network but it is not clear what they are doing.

Michael Shermer wrote an interesting article for Scientific American called The Brain is Not Modular where he argues that fMRI’s have led some scientists to over-apply the modular metaphor. In the comments section of the article neuroscientist Marco Iacoboni critiques Shermer’s article. The result in good summary of the debate over modularity.

This is a complex and rapidly evolving area of research and I do not pretend to be an expert as it is not my area of research. But here is my best shot and synthesizing what I have read. It seems that in practice most neuroscientists are somewhere in the middle between modules and networks – combining both concepts. I think this is the right approach because I think the brain combines both approaches. There does appear to be modules in the brain – we had evidence for this long before fMRI scans. As a clinical neurologist I can examine a patient with a stroke and in most cases tell you exactly the size and location of the lesion that we will see on the MRI – based solely on the deficits on exam. If a patient has an isolated Wernicke’s aphasia – an inability to understand verbal commands – then they will have a medium-sized lesion in the angular gyrus of the dominant (usually left) hemisphere. That piece of the brain serves a very specific purpose.

But while the more basic or straightforward cortical functions are highly modular, the more complex or sophisticated higher cortical functions are not. There does not appear to be a piece of the brain that provides attention or concentration, or creativity. These seem to follow more of a network model of brain organization.

It is also possible that while there are modules in the brain their function may be more abstract and the specific effect they have varies depending upon what other parts of the brain that connect to them are also active. The trick for neuroscientists may be finding the most accurate way to describe the complex and abstract function of the various brain modules. There may be a layer or two of hidden complexity to what the fMRI studies are actually telling us. It seems that the more we study the brain the more complex a puzzle it becomes, and perhaps we have not yet crossed the threshold where new information isn’t just giving us more questions.

I don’t mean to downplay the vast amount of information we already have about brain function. As a said, I put this into practice almost every day. Rather I think that fMRI scanning has given us a new window into brain function and it is revealing a new layer of depth to its complexity. The modularity debate reflects our current struggle to understand this new depth. As is typical of scientific endeavors, the debate is healthy and is likely to lead to improved models of brain organization and function.

Smiling Babies

With all this in mind, what does this new study of mothers looking at their smiling infants tell us? The abstract’s conclusion says:

When first-time mothers see their own infant’s face, an extensive brain network seems to be activated, wherein affective and cognitive information may be integrated and directed toward motor/behavioral outputs. Dopaminergic reward-related brain regions are activated specifically in response to happy, but not sad, infant faces. Understanding how a mother responds uniquely to her own infant, when smiling or crying, may be the first step in understanding the neural basis of mother-infant attachment.

It is no surprise that something as complex as a mother’s reaction to their infant is very complex – I think I could have told you that without this study. But it does give us some specific information. There is an emotional component to the reaction – one that involves the reward center of the brain. This means that smiling babies make their mother’s feel good is a way that reinforces the behavior. Some news stories of this study have likened this response to the high addicts get from crack. I think that’s a stretch. The reward system seems to be one of those multi-purpose modules where the significance of its activity can only be understood in the context of the network in which it is firing.

The study also suggests that the emotions generated by seeing one’s smiling infant also is connected to, or triggers, certain thoughts. These emotions and thoughts also connect to specific behaviors. Although the study did not say it, I strongly suspect that one of those behaviors is the smiling, waving, and cooing that most parents will give back to their smiling infants. It likely also connects to more complex behaviors, like the instinct to protect and nurture the adorable little bundles of joy.

Crying babies (as any parent can tell you) elicits a very different emotional response. This is a very negative experience – one is motivated to stop the crying as soon as possible and failure to do so quickly may result in guilt, shame, and feelings of inadequacy. It is no surprise that a crying baby elicits no reward response – must make the baby smile and laugh. Anything to make it smile… Aahhh! It’s smiling – how adorable.

12 responses so far

12 thoughts on “Smiling Babies, fMRI, Brain Modules, and Neural Networks”

  1. arthwollipot says:

    I always said that I hated kids. Then I had kids of my own, and now I say that I hate other peoples’ kids. It was very strange. Ever since my son was born I have said that there’s something weird and biological going on where one’s own children are concerned. It’s good to have this confirmed neurologically. Thanks.

  2. Niobe says:

    I wonder how this would measure between mothers and adopted infants. There is generally a huge desire for a child that could be projected onto the infant, but there isn’t the hormonal precursor that pregnancy and delivery bring along.

    Equally so, between mothers that had vaginal opposed to cesarean delivery, where the latter lacks certain hormones from being released, allegedly.

    After that, we could start mapping the brains of stepmoms 😀

  3. Novella: “I have two daughters, about to turn nine and six. They are… the most adorable things in the universe.”

    As the father of five daughters, let me assure you of a new assessment when they turn fourteen.

  4. rbstansfield says:

    When will the lay press internalize the idea that enjoyment is, per se, the activation of dopaminurgic striatal brain areas? It has little-to-nothing to do with crack. That MSNBC article says: “Turns out that seeing your own child smile actually activates the pleasure receptors in the brain typically associated with food, sex — and drug addiction.” I could easily write a similar sentence: “Turns out that MSNBC writers actually use computers which are typically associated nerds, engineers — and white collar criminals.”

    If I reach into your brain and electrically stimulate certain striatal areas, you will go weak in the knees with pleasure. If I give you something you very much enjoy, and you enjoy it, your striatal areas will activate. They are the same thing: the experience and the brain activity. It is trivially true that they correlate.

    It’s easy to blame the journalists, I guess, since most of our culture is deeply dualistic and “Your Brain” still seems to have a connotation of something kind of alien and weird. But shame on the scientists themselves, who choose medial prefrontal, anterior cingulate, and insula cortex as their regions of interest. It’s a safe bet to set your alpha high enough for those areas to make sure a smiling baby vs. crying baby contrast achieves statistical significance in anterior cingulate. But that’s hardly the interesting thing about these results.

    Dr. Novella is correct that the correct analysis is in the network, the dynamics of processing. fMRI isn’t the best tool for studying the covariance of different brain areas across time, so you can’t fault these researchers for a lack of hardcore network analysis of these data. But still, too much focus on individual “lighting up” of certain areas (especially “pleasure centers”) will get the press all hot and bothered, and it doesn’t add much to the literature.

  5. daedalus2u says:

    I would just like to point out for the nth time that what fMRI BOLD is really measuring is where NO levels in the brain reach levels sufficient to activate sGC and cause vasodilation that results in the oxyhemoglobin/deoxyhemoglobin changes that fMRI BOLD actually measures.

    The onset time, range and duration of that NO signal depends on the background level of NO.

    I was at a talk on autism at MIT recently where the speaker, Marlene Behrmann, Ph.D.


    mentioned that the “regions of interest”, that is the regions activated during measurement by fMRI BOLD, were smaller in people with high functioning autism than in normal controls. This precisely fits with my low NO hypothesis of autism, where the range of those neurogenic NO signals that are causing the vasodilation are reduced because of the lower background. They may also be reduced in magnitude, but increasing the background is something I know how to do.

  6. HCN says:

    Baby smiles do not just affect mothers. They even make people happy who have never had children of their own.

    When my first born was almost three months old my parents paid for me to visit them with the new baby. This was in early December, so the airplane was fairly empty (between holiday rushes). So to go use the restroom I had the flight attendants watch the baby in his car seat. When I came out I saw about four people hovering over the baby, and the young male flight attendant grinning ear to ear saying “He smiled at me!”.

    (side note: he was a very easy going baby, he never cried on that trip… when I left with him I heard people say “I didn’t know there was a baby on the plane!”)

    My father-in-law seemed to feel a special connection to my second son, because that child always smiled for him (hubby’s step-dad, never had kids of his own).

    Side note: Each of my children seemed to reveal a special personality trait with their first smiles.

    With my oldest his first smile was just turning his head up from nursing and giving me the sweetest heart melting smile. Even as an adult he is a sweet sweet man (who likes to eat). When he was a toddler and preschooler he actually shared (with a speech disorder his main form of communication was a big smile), in elementary school he was given an award for “acts of random kindness”, and when he was in 8th grade a young lady came to me and told me that he was a very nice kid who was the only one who helped when she dropped all her books and papers.

    My second son did not smile, he sparkled! He did not just smile with his mouth, but with his whole body. He does everything in a gung-ho style. While my oldest son’s “terrible twos” last one week, this boy’s “terrible twos” lasted from age 18 months until he was seven years old. He was easy to shop for because he was excited about everything. He is a now a bright high school kid who has two summer jobs, and big circle of friends.

    My youngest child is my daughter. Her first smile came when she was barely three weeks old… much, much too early. She also crawled early and walked early. For those of you with kids, you know that early mobility is not really desirable. It is not good when the motor skills outpace the neuro skills! Her neuro skills were pretty good… and she learned to read well early. Our shared passion is reading, our special time is going to used book stores and the Library Foundation book sales. In the 8th grade she had me buy her Dostoevsky’s “Notes From the Underground” for a book report. But she is now fourteen years old… and what DevelsAdvocate says is true (she has been more trouble than both boys put together!).

  7. Niels Kjaer says:

    Thanx for a brilliant blog!

    I have a little story to tell.

    3 years ago I was “finally” diagnosed as having Autism(Asperger if you like) and AD(H)D. On the 5th attempt, I finally (5 weeks ago) found a medicine (dexamphetamine) which influenced my AD(H)D. I tested (using 25 years of experience in data analysis related to CERN) how dexamphetamine worked on me by analysing how I played a complex but extremely well understood (computer-)game called Joggle (or Boggle if you like). My conclusion was clear: When dexamphetamine worked(45 minutes to about 4 hours after intake) white noise on the bookkeeping of my active memory simply disappeared. Otherwise everything remained unchanged (number and quality of thoughts etc.).

    I then realized that AD(H)D(at least my type) has nothing to do with (my) IQ. I concluded that humans suffer (e.g. end up in jail)from AD(H)D due to a poor environment, while humans in a good environment and AD(H)D tend to end up as entrepreneurs.
    I have now learned how to slightly alter my way of thinking so that I don’t actually need the dexamphetamine.

    I then translated the MEME of IQ=100 to Autism. I concluded that my Autism is likely a change (probably genetic in my case, due to the influence of a certain protein) of the relative “weight” my brain gives to abstract/raw and empathic/processed thoughts. I theorize that this dual filtering process is related to the functioning of our mirror neurons.
    I used this knowledge to understand my life, including why I was rather ill at the age of two. I also found out that it is possible for me to change the weight of abtract/emphatic thoughts in my cognitive thinking. When I applied this as a ToM, I could suddenly understand how other humans think. I no longer feel that I am a Martian, but a human being.

    Giving a working ToM to an Asperger scientist (automatically?) leads to a ToE. My ToE explains why the rising oil prices is a very good sign, and that our world has just started undergoing a very promising 2nd order phase transition. Much to my surprise, no changes of cultural symbols are needed. Just a few simple MEMEs at the right places. Teaching autist children to “communicate” like humans will be as challenging as teaching neuraltypical adults to “think” like humans.

    My Cheshire cat is now smiling…

  8. woobegone says:

    rbstansfield : I wouldn’t be so sure that all reward processes necessarily involve dopamine – look at this recent paper finding that opiates have no effect on dopamine levels as assessed using raclopride displacement, but are rewarding


    I saw David Nutt giving a talk about these results and apparently he had great difficulty publishing them!

    But I think in general you’re quite right, there is nothing very interesting in the fact that pleasant things activate the same areas that we already know activate to other pleasant things. What would be interesting in this case would be if you discovered that there was a specific area that responded to, say, smiling babies but not smiling adults (which also activate reward areas.)

  9. rbstansfield says:

    woobegone: Thanks for that reference. Addiction is fascinating, and that reference kind of underscores my point about how the mere activation of “pleasure centers” has next to nothing to do with crack addiction.

    My (outdated, I’m sure) understanding of addiction is that the brain has equilibriated itself to incorporate the habitual ingestion/experience of the addictive substance/behavior. It doesn’t surprise me that a sample of addicts does not show dopaminurgic response to the thing they are addicted to, because their brains already expect it: in fact, their brains will start to falter without it, which is what makes addiction different from mere craving or habit. Am I hopelessly out of date on that?

  10. Kari says:

    Very interesting in light of this NPR Story I heard the other day. Seems that some sociologists believe that having children is harmful to our mental health. Not sure how to works into evolutionary psychology.

    And DevilsAdvocate is correct. My daughter just turned 14 – a world of difference than when she was 9.

  11. What happens with later stage addicts is that the amounts and frequencies of abusive drug ingestion eventually outstrips the brain’s capacity to produce dopamine fast enough to ‘keep up’. This dopamine depletion means a lessened effect experienced by the addict – who then uses more drug, unwittingly worsening the situation. Dopamine depletion can lead to depression above and beyond that already caused by the myriad of problems caused by being an addict. Cocaine abuse is a good way to deplete your dopamine levels.

    rbstanfield’s ‘outdated’ notion is not. The process is called tolerance, one of 7 symptoms of substance dependence (where 3 or more within a 12 mo period are required for diagnosis):

    Per the Diagnostic and Statistical Manual IV:

    (1) tolerance, as defined by either of the following:

    (a a need for markedly increased amounts of the substance to achieve intoxication or deisred effect

    (b) markedly diminished effect with continued use of the same amount of the substance

    Tolerance is not limited to substance abuse and dependence. Physicians frequently raise the dosages and/or frequencies of prescribed meds to compensate for lessened therapeutic efficacy due to tolerance. Typically, these are meds that have been for a considerable time.

    rbs: “which is what makes addiction different from mere craving or habit..”

    Craving is a symptom of addiction. ‘Habit’ is not a clinical term, but a common use synonym for drug abuse or addiction, sometimes used to describe a burgeoning abuse that is not yet addiction.

Leave a Reply