Developing Intelligence

Bad Science? Vindicating Cognitive Development and the Preferential Looking Paradigm

Posted by Chris Chatham on January 3, 2008
(7)
More »

Over New Year’s I had a brief discussion with a condensed-matter physicist who proclaimed that 1) “some developmental research is amazingly bad” and that 2) “they think they can tell what a baby has learned from what direction it looks,” topping it all off with 3) “you guys don’t even know what learning is!”

I won’t argue with the first point (there are bad researchers in every field, even condensed matter physics), and I’m too lazy to bother with the third (although the 2000 Nobel prize committee might disagree), but the second point – on the technique of preferential looking – I just can’t disregard. Complaints about the rigor of developmental cognitive psychology are surprisingly common (e.g.), but all too often unjustified.

Using this comment about the preferential looking paradigm as an example, such complaints can be understood in several ways:


1) Conflating accessibility with lack of sophistication.

If a scientific technique can be easily understood, it might be mistaken as the tool of an unsophisticated, naive, or stupid scientist. This logical fallacy – “affirming the consequent” – arises from a bias common to fields where the quality of your data is directly related to the money you can spend on a device to collect it.

2) “Physicist’s disdain”

This is the disdain permitted of physicists towards other scientists, all of whom are presumed to have “physics envy”, a situation where “biologists kind of envy biochemists who kind of envy chemists who kind of envy physicists. Everybody wants to wear a slightly whiter coat.”

The focus of physicist’s disdain is inconsistent. Sometimes it lingers on the lack of complex mathematics in other disciplines; other times on that apparent uselessness stigmatizing any science which hasn’t yet dreamed up a new weapon of mass destruction. Most often, it lingers on the fact that all other sciences study higher-order phenomena – phenomena which can never be truly understood without the meticulous and back-breaking lower-order work of physicists. By this view, everyone really should be a physicist, and so envy is fully justified.

3) Lack of familiarity with the technique

This is both the most obvious and the most forgivable (as well as the most likely, in the case of my friend the physicist). It turns out that most developmental cognitive psychologists are not stupid, are therefore not unaware of the higher-order level of their inquiry, and thus have gone to much trouble to validate their methods.

It also turns out the preferential looking paradigm is a particularly poor choice for complaining about a lack of rigor, since extensive work over the past 40 years has confirmed that infant gaze is a wonderful indicator of varied perceptual and cognitive phenomena. For example, the preferential looking paradigm is sufficiently sensitive to detect amblyopia and other visual disorders in newborns, and is in fact a more clinically appropriate tool than “more sophisticated psychophysical techniques” which can bore babies due to their length. By some accounts, the technique has allowed for measurements of visual acuity in infants which are as precise as those in adult subjects, and reliability is quite high. Preferential looking is now a standard to which other neonatal ophthalmological methods are compared.

The preferential looking paradigm has also proven valuable in more cognitive domains. For example, infants older than 7.5-months discriminate between visual displays where memory for simple features is not enough, but where they must be integrated into complex objects. This capacity is known as “visual binding,” and addresses a longstanding problem in cognitive neuroscience.

Preferential looking has also been productively employed outside of the visual cognitive sciences, in an impressive testament to its generality. For example, only infants older than 3.5-months prefer faces whose gender matches that of a recorded voice, suggesting audio-visual integration; infants older than 6-months will preferentially gaze towards the faces/mouths with an articulatory posture matching an audio recording of a vowel, suggesting nascent audio-motoric integration; and gradations in gaze preference demonstrate that 8-month-olds may be capable of extracting abstract rules from auditory input structured by artificial grammars.

As in condensed matter physics, and probably all sciences, some discrepancies between various measurement techniques can be expected, exacerbated by lack of agreement on the proper controls. Debate over these issues is what characterizes an intelligent and mature science, not the accessibility of its measurement techniques to laymen.

(7)
More »

Comments

  1. #1 Frederick Ross
    January 4, 2008

    Part of the reason physicists have so much trouble with other fields (speaking as a physicist now abroad in infectious disease) is that they don’t know much about statistics. Imagine you know means, standard deviations, and least squares fitting of curves. Now you’re faced with really complicated things from psychology or biology. It seems incomprehensible that you can get anything out of this.

    Plus a lot of physicists still haven’t realized that while in the end everything does reduce to particle physics, it is PHYSICALLY IMPOSSIBLE to calculate almost anything nontrivial from first principles (thanks, computer science).

    Oh, and I have to protest about the math. The math in most of physics is actually pretty trivial: mass on a spring. The great successes of modern physics have all been masses on springs and perturbations thereof. It’s annoying and long to carry out the calculations, but there’s no depth to it. But the complaint doesn’t actually mean what it sounds like. It’s a plea, and here’s the translation: “Please explain the mindset underlying what you’re doing as manipulations of a set of symbols.” Why a set of symbols? Because they’re (hopefully) unambiguous, and they let you check your own intuition by mechanical calculation, and correct your mindset when they disagree.

    For something like Newtonian mechanics, this is great. The symbols and their relation to reality aren’t in question anymore. For younger fields like quantum mechanics and thermodynamics, it’s a bit of a problem because it’s very easy to get the semantics wrong and then your symbols fail you. In really young fields like most of biology, it’s begging for trouble.

    But a lot of it’s really just arrogance fostered by the culture in physics departments. “Biology? We don’t need no stinkin’ biology.” This is partially because the physicist goes and tries to learn some biology and is inundated by endless three letter protein names and “factors,” and no one will flat out tell him what’s going on (which is usually that no one knows, but here’s some accumulated experimental facts).

  2. #2 CHCH
    January 4, 2008

    Hi Frederick – that was a very nice response – I was expecting something more vicious or at least hyperbolic!

    Having already fully offended “my friend the physicist,” this post may not be around for that long :(

  3. #3 Seth Herd
    January 4, 2008

    As an undergraduate physics major/math minor, I don’t think that those fields offer a lot that’s useful to psychologists. The desire to have things neatly explained in closed-form equations works well in physics, presumably because the physical world really does obey relatively simple laws. The brain/mind is more complex; I’m very suspicious of attempting to treat it with closed-form math, since that constrains one to relatively simple hypotheses. It’s like the old adage about searching for your keys under a light, since it’s too dark to find them where you actually lost them. The psychology papers that deal most heavily in closed-form math seem to me some of the least helpful around.

    Having experienced physics and psychology/neuroscience, I don’t feel any envy- psychology/neuroscience is both more interesting, and more useful. Your physicist friend’s comment sounds simply ignorant (unless he has something more contentful to say to back it up). People from many fields have big heads, and it’s not the worst thing- some intellectual push-and-shove can help get people to think outside of their own field’s established paths. So the useful question is whether his critiques are on target; you’ve shown that the second isn’t; the third strikes me as also interesting.

    As for ‘really knowing’ what learning is, I think you (and the field) do know what it is- the change of synaptic strengths so that the structure of the world becomes better represented in the brain.

  4. #4 David Harmon
    January 5, 2008

    Having already fully offended “my friend the physicist,” this post may not be around for that long

    Bah! If he can’t be bothered to consider that scientists outside his own field might, possibly, know what they’re doing, then he darn well deserves to be offended!

  5. #5 CHCH
    January 7, 2008

    to be clear, my friend the physicist is offended b/c the quote was taken out of context (we were laughing; it was merely a “conversational dig” rather than a serious informed opinion).

    still, it made for an interesting post, and he now understands that was the intended goal (not to indict him personally).

  6. #6 Jon Nordland
    January 23, 2008

    I for one am still skeptic. That might be because i just remember a few arguments that seems to prove that you present your argument like a conspiracy theory person. That looking-tests, like you mention, can be used to find problems WITH THE VISUAL SYSTEM seems like a poor way to prove that it works for learning-tests. Second, that looking-test have a wide application isn’t proof either. Its like saying that dividing rod works because wood as a wide range of application.

    Sorry, my English isn’t that good. I just wanted to point out that I’m scenting still. hehe.

  7. #7 CHCH
    January 23, 2008

    Jon, the basic idea behind preferential looking is valid on first principles. If kids of one age reliably discriminate between two audio/visual displays, and younger kids don’t, and these are replicable/reliable effects, it must reflect learning or development. The nature of the learning (audio-motoric integration or whatever) may be debated based on the exact stimuli, but a cross-sectional design with decent N and a significant result is hard to argue with. After establishing that pref. looking is reliable & sensitive, with reference to the visual tests, I just listed a few papers with cross-sectional designs whose results made intuitive sense.