A hot tip

The Guardian has taken, from time to time, to including pieces about recently published science which it thinks will interest its readers, with there being one such, reference 1, a few days ago. The recently published science turning out to be reference 2, with support from reference 3 and reference 4. References 2 and 3 both come from well-respected journals, although it has to be said that the Royal Society has started touting for publishing business, leveraging the value of its brand. Maybe volume is important, even for the posh.

I notice also that the online version of reference 1 is not quite the same as the print version. Perhaps reflecting the fact that the constraints of space are not the same online as they are on paper.

The paper at reference 2 concerns a visual illusion which does not work for babies under around six months but which does (mostly) work thereafter. This is argued to be evidence for the presently popular theory that visual perception is a mixture of feed-forward and feed-back processing. That the pixels coming in from the retina are shaped and organised by the higher brain’s knowledge that, for example, one is probably looking at a pork chop. With this feed-back processing taking rather longer to mature in the infant brain than the feed-forward sort – for which delay there appears to be plenty of other evidence, both anatomical and experimental.

The illusion may have been introduced to the world by the short article at reference 3. One has, in effect two superimposed, transparent sheets of dots, with one sheet moving up and the other sheet moving down. On the up sheet one has central red dots and peripheral green dots, on the down sheet one has central green dots and peripheral red dots. What most people actually see is one array of red dots moving up and a second array of green dots moving down. The brain has decided to override the unwanted complexity, preferring to keep things simple. Unfortunately, I was unable to get the movie of the illusion provided to work: it came in the form of a .MOV file which I think is Apple flavoured, which my Microsoft flavoured laptop would only open in something called Clipchamp, where I dare say it could be made to work, but it did not work for me. A product which Microsoft appear to have bought in and which has been loaded along with Office 365. See reference 5.

The snap above is lifted from reference 3 and it gives something of the idea, although I am not clear whether the white bars mark a clean boundary between the central and peripheral zones. The text rather suggests not, but, in any event, not important for present purposes.

Called the misbinding illusion because the binding of features (direction of movement) to things (coloured dots) has come unstuck. It is known that the processing of visual features is distributed, with one bit of brain doing one sort of feature and another bit of brain doing another, but usually the brain manages to put it all back together again into a reasonably accurate, unified percept. But, unusually, not here.

The big snag is the fact that six months old infants do not talk and the proxy used for talking is attention span. Other things being equal, infants look at unfamiliar things for longer than they look at familiar things – and this can be exploited by cunningly designed experiments. The structure of which is a familiarisation phase followed by a test phase, a test phase in which one tests whether the stimuli are familiar or not.

Added to which babies are not very good experimental subjects at the best of times. They do not do as they are told, they are very easily distracted and they are very apt to get upset for one reason or another. Maybe they need feeding or changing.

There were three experiments in all. The misbinding experiment, the results of which are snapped above, and then two control experiments to check that infants could indeed distinguish coherent from segregated movement. After exclusions, each experiment involved 40 infants, split evenly between the two age bands (5-6 months and 7-8 months) and more or less evenly between the sexes. 

It took me a while to work out the details of the three experiments, exactly what the stimuli were for the various experiments. Hopefully I got them right.

The experimenters judged when the infants were looking at the target from video tapes – which sounds a bit tricky. I assume that the eye tracking gadgetry one can use for consenting adults is not appropriate here.

The effects might consistently favour the hypothesis, but they are not large. And they involve the p-values which are themselves the subject of so much controversy these days. For which see, for example, reference 8.

The authors conclude that ‘… We found that older infants exhibited misbinding, whereas younger infants showed no such illusions … These results suggest that the development of feedback processing contributes to feature-binding…’. They generalise to observing that very young infants see the world rather differently from their elders. Or put another way, they see the raw world, warts and all. While we like to tidy things up.

And, as it happens, a nice example of one person’s conscious experience not being the same as another’s.

Equipment

The displays used in these experiments came from Cambridge Research Systems of reference 7. People who make all kinds of equipment for work of this sort.

Another experiment

Reference 4 describes another experiment of the same sort, but using a backwards masking technique – object substitution masking – rather than a visual illusion. With the line, once again, being that backwards masking requires feed-back processing of a sort not available to the very young.

Conclusions

All very plausible, despite my difficulty of working through the detail. One just hopes that the reviewers and editor have done their stuff and that the conclusions can be trusted.

And a useful reminder that infants do not arrive in the world fully baked. They need time for their brains and the wiring therein to finish growing – growing which involves a delicate interaction, a subtle dance, between internally driven growth and learning to deal with, and eventually to make use of, stimuli from the outside world.

And helpful for the Guardian to be highlighting research of reasonably general interest in this way.

References

Reference 1: Babies do not fall for illusion that fools older children, study finds: Experts says findings of dots tests are down to information processing not yet being fully developed – Nicola Davis, Guardian – 2023.

Reference 2: Infants’ visual perception without feature-binding – Shuma Tsurumi, So Kanazawa, Masami K. Yamaguchi – 2023. Article in Royal Society B. 

Reference 3: Steady-state misbinding of colour and motion – Daw-An Wu, Ryota Kanai, Shinsuke Shimojo – 2004. Short piece in Nature about the illusion. 

Reference 4: Perception of invisible masked objects in early infancy – Nakashima Y, Kanazawa S, Yamaguchi MK – 2021.

Reference 5: https://clipchamp.com/en/. 

Reference 6: http://www.scholarpedia.org/article/Object_substitution_masking. Support for reference 4 above. Another intriguing business.

Reference 7: https://www.crsltd.com/.  

Reference 8: https://psmv5.blogspot.com/2021/11/the-p-value-battle-continues.html.