Comments

Wednesday, June 28, 2017

Facing the nativist facts

One common argument for innateness rests on finding some capacity very early on. So imagine that children left the womb speaking Yiddish (the langauge FL/UG makes available with all unmarked values for parameters). The claim that Yiddish was innate would (most likely) not be a hard sell. Actually, I take this back: there will always be unreconstructed Empiricists that will insist that the capacity is environmentally driven, no doubt by some angel that co-habits the womb with the kid all the while sedulously imparting Yiddish competence.

Nonetheless, early manifestation of competence is a pretty good reason for thinking that the manifest capacity rests on biologically given foundations, rather than being the reflex of environmental shaping.  This logic applies quite generally and it is interesting to collect examples of it beyond the language case. The more Eism stumbles the easier it is to ignore it in my own little domain of language.

Here is the report of paper in Current Biology that makes the argument that face recognition is pre-wired in. The evidence? Kids in utero distinguish face like images from others. Given the previous post (here) this conclusion should not be very surprising. There is good evidence that face competence relies on abstract features used to generate a face space. Moreover, these features are not extracted from exemplars and so would appear to be a pre-condition (rather than consequence) for face experience. At any rate, the present article reports on a paper that provides more evidence for this conclusion. Here’s the abstract:

It's well known that young babies are more interested in faces than other objects. Now, researchers have the first evidence that this preference for faces develops in the womb. By projecting light through the uterine wall of pregnant mothers, they found that fetuses at 34 weeks gestation will turn their heads to look at face-like images over other shapes.
Pulling this experiment off required some technical and conceptual breakthroughs: a fancy 4D ultrasound and the appreciation that light could penetrate into the uterus. This realized, the kid in utero responded to faces as infants outside the uterus respond to them. “The findings suggest that babies' preference for faces begins in the womb. There is no learning or experience after birth required.” This does not mean that face recognition is based on innate features. After all, the kid might have acquired the knowledge underlying its discriminative abilities by looking at degraded faces projected through the womb, sort of a fetus’s version Plato’s Cave. This is conceivable, but I doubt that it is believable. Here’s one reason why. It apparently takes some wattage to get the relevant facial images to the in utero kid. Decent reception requires bright lights, hence the author’s following warning:

Reid says that he discourages pregnant mothers from shining bright lights into their bellies.

So, it’s possible that the low passed filter images that the kid sees bouncing around the belly screen is what drives the face recognition capacity. But then the Easter Bunny and Santa Clause are also logically possible.

This work looks ready to push back the data at which kids capacities are cognitively set. First faces, then numbers and quantities. Reid and colleagues are rightly ambitious to push back the time line on the alter two now that faces have been studied. In my view, this kind of evidence is unnecessary as the case for substantial innate machinery was already in place absent this cool stuff (very good for parties and small talk). However, far be it from me to stop others from finding this compelling. What matters is that we dump the blank slate view so that we can examine what the biological givens are. It would be weird were there not substantial innate capacity, not that there is. The question is not whether this is true, but which possible version is.

Last point: for all you skeptics out there: note this is standard infant cognition run in a biology journal. I fail to see any difference in the logic behind this kind of work and analogous work on language. The question is what’s innate. It seems that finding out what is so is a question of biological interest, at least if the publishing venue is a clue. So, to the degree that linguists’ claims bear on the innate mental structures underlying human linguistic facility, to that degree they are doing biology. Unless of course you think that research in biology gets is bona fides via its tools; no 4D ultrasounds and bright lights no biology. But who would ever confuse a discipline with its tools?
  

1 comment:

  1. This seems like an odd position to take given the subsequent posts on statistics. The actual paper, which I recommend over the summary, is worth reading if you're at all interested in this issue.

    To begin, the fetus-face study isn't using faces at all; it's using patterns of lights, particularly three red lights organized in a triangle. If the triangle is wide side up (relative to the fetus), with two lights above one light, then it's a face. If it's apex up, with one light above two lights, then it's not. The idea is that if fetuses respond more to wide-side-up triangles than to apex-up triangles, then they have innate facial-recognition abilities.

    Well, okay. Certainly there are instances in which non-face stimuli, like electrical sockets and traffic lights, look like faces to us. Maybe it works the same way with red lights; apparently there's some evidence in support of that claim.

    But there's another problem. Specifically, each of the fetuses saw each stimulus five times. And how did they respond? Well, apparently they responded to the wide-side-up triangles an average of 1/5 times. They responded to the apex-side-up triangles an average of 0.4/5 times. This was significant using a Wilcoxon test, which the authors used for some reason even though they claim the data were normally distributed. On top of that, the determination of fetal response was not that easy; the raters making the determination had OK agreement, but not great agreement.

    In other words, the correct headline would be something like "Fetuses don't usually respond to this sort of stimulus," and consequently these data provide little or no reason to believe that face processing is innate.

    ReplyDelete