Sunday, May 17, 2009

Who Goes There?

It isn’t a pleasant memory.

We’d spent the better part of the morning looking for a lost child, hope waning, our fears growing, as we did. Lilly, as always, was doing her part; but the general gloom of a clouded-over, gray sky melded with a kind of cohesive murk within my little dog team.

You see, I was in Another State, working with people who didn’t know me, and my subordinates — local, trainee dog handlers assigned to me at the command post — didn’t like the way I was doing things and weren’t trying very hard to hide it.

Today I’ll admit, rookie handler that I was, I may have been a little bloody minded about doing things by the book. I’d been taught that search teams often missed search subjects who were on the boundaries between search tasks, and so I was intent on avoiding that by covering just a tiny bit beyond my assigned area. My two teammates might well not have batted an eyelash at that, but in this particular case it required us to cross a boggy little creek and tramp along the swampy opposite side.

Frankly, I thought at the time, and may well have been right, that they were just being lazy. Maybe I didn’t try very hard to hide that [1].

Soon after, gridding inward from the creek, I looked up to see a disturbing thing: a crumpled little body dressed in white, lying on the ground. One of my walkers must have seen it the same moment I did, because we both paced toward it — quickly but not at a run, in what for me, at least, was a moment of profound ambivalence.

And we walked up on a crumpled, white plastic bag, lying on the ground.

“I thought …” my walker began.

“Me too,” I replied to her, too relieved to say more.

But I had seen that body …

The morning of July 3, 1863, and the legendary Robert E. Lee was looking up the long, naked slope of Cemetery Ridge toward the Federal Army of the Potomac, dug in on its summit. What he saw was a thinly protected line, denuded by virtue of the fact that his opposite number, George Meade, had been pulling men out of there to protect his flanks from the brutal strikes Lee’s Army of Northern Virginia made on them the previous day.

The Confederates had swept the Federals off Seminary Ridge, to the west, on July 1; on the 2nd they’d failed to dislodge the northerners from Cemetery Ridge, but it had been very close. Now Lee could see their weakness in the center; he could feel that they were a push away from crumbling, as they had done so many times before.

Trouble was, Lee was in the decided minority among his own army. James Longstreet, one of his best generals and the man to whom he would entrust the upcoming attack, had been arguing since the previous day that it was doomed to fail. Neither of the two could have known that, technically, Lee was right, in that Meade only had about 5,000 men defending a ridgetop about to be hit by 11,000 Confederates. But Longstreet, even without the virtue of hindsight, could see that the position was so strong — Longstreet’s men would have to walk, under artillery and rifle fire, for nearly a mile in the open, before reaching the Federal line — that it wouldn’t take very many men to stop them.

“General,” historian Shelby Foote, in his magnificent tome about the Civil War, quotes Longstreet, “I have been a soldier all my life … and should know as well as anyone what soldiers can do. It is my opinion that no 15,000 men ever arrayed for battle can take that position.”

It’s emblematic that Longstreet got the number of men in his own command wrong — for one thing, a couple of the divisions he would send into combat were leant from another general’s corps; for another, he understandably hadn’t kept up with the massive casualties his own corps had been taking over the previous two days. But Lee saw it differently, and gave him a direct order. Longstreet, in turn, ordered the charge — with a voiceless nod to division commander Major General George Pickett, captured in agonizingly accurate visual detail in the movie Gettysburg. And an attack began that ended with Pickett, upon Lee telling him that he needed to gather his retreating men to defend against a possible Union counter-strike, replying in anguish: “General Lee, I have no division now.”

He was exaggerating; his division had suffered only 60 percent casualties.

Lee had seen that the Federal center was weak …

It’s happened with nearly every search dog we trained, and, almost by the calendar, at exactly one year of age. In a dusk training problem, the dog encounters the practice subject unexpectedly, because the wind is blowing the wrong way.

The dog, seeing a human — their eyesight is quite good, particularly at twilight — but not smelling him, sees … an ogre. A dog who has learned the find-refind-lead the handler back sequence, has performed the routine with unerring fidelity until this moment, not only refuses to approach. She barks that shrill but powerful panic bark you usually hear only from adolescent dogs.

You have to jolly her up, reassure her (without coddling) that she’s wrong, there is nothing to fear there, just a person, maybe someone she already knows, and it’s perfectly safe to approach. She does, and greets the subject with over-the-top affection and what looks to all the world like embarrassment. Seldom does the dog need this treatment more than once; she’ll be confidently making dusk and dark finds without a hitch.

But that first time, she sees the monster …

It’s a borderline cliché that we see what we expect to see. But Hendrikje Nienbord and Bruce Cumming from the National Eye Institute have produced new findings that suggest that the very wiring of our nervous system conspires to delude us.

The Eye Guys took monkeys and trained them to perform a task that depended on whether the center of a circular pattern displayed before them was protruding or receding; they then recorded the activity of sensory neurons from the eye that were sending visual signals up to the brain.

The general idea had been that the eye detects the light pattern, it sends signals through the sensory neurons to the higher brain, and then the brain decides whether the dot is approaching or receding. But it didn’t work out that way.

Bear with me on this one, it’s a real head-banger of a paper, very dense stuff to parse out.

* As the subtlety of the choice got harder over time, the coupling between the visual neuron’s activity and the accuracy of the monkey’s choice (in other words, the nerve saying, for example, “innie,” and the monkey making the “innie” choice) didn’t decrease, which is what you’d expect if the signal only traveled from eye to nerve to brain. Right or wrong, the nerve cell’s activity — remember, it’s supposed to be sending signals up to the brain — was more and more reflecting what the monkey was going to choose rather than what it had to be seeing.

* Increasing the reward increased the accuracy of the monkey’s choice, but decreased the coupling between nerve cell and choice. This rather upside-down result made more sense when they dissected what was happening over time: for a window of about one second, the larger reward got the monkey to focus on the image rather than its expectations. After that second, it started to see what it expected to see — again, the sensory neuron was tracking more closely to what the monkey was going to choose than what it actually saw.

* If I understand the paper correctly, the relationship between the nerve cell activity and the choice was strongly affected by what the monkey had seen previously. Again, the monkeys’ visual experience was tracking with their biases, not the image being shown to them.

Like I said, it’s not an easy one to think through: but taken together, it looks like the signal didn’t move from the eye to the nerve to the brain as much as the eye and the brain argued over what the nerve cell was going to do. And the brain often won; the monkeys were seeing what they expected to see.

As an accompanying article in that issue of Nature quotes Cumming, “In a way, the brain is tampering with the data.” The reason? As he told Nature, it may be that it’s better to have a preconception ready to act upon than to wait on the facts — and get squished in the meantime.

Still, that imperative only goes so far — maybe a second or two. The next time you have more time than that to decide something important — God forbid, of life-and death — and are sure you see something, take another look.

[1] I don’t remember if, at the time, it had occurred to me that there would be a political price to pay for my insistence. Political trouble did, in fact, come from that direction, well over a year later. Today I know that a number of factors made this come about; what I don’t know is whether their report back to their teammates that day was one of them.

No comments: