Thursday, 15 January 2015

The Size-Weight Illusion Induced Through Human Echolocation

Echolocation is the ability to use sound to perceive the spatial layout of your surroundings (the size and shape and distance to objects, etc). Lots of animals use it, but humans can too, with training. Some blind people have taught themselves to echolocate using self-generated sounds (e.g. clicks of the tongue or fingers) and the result can be amazing (I show this video of Daniel Kish in class sometimes; see the website for the World Access for the Blind group too).

In humans, this is considered an example of sensory substitution; using one modality to do what you would normally do with another. This ability is interesting to people because the world is full of people with damaged sensory systems (blind people, deaf people, etc) and being able to replace, say, vision with sound is one way to deal with the loss. Kish in particular is a strong advocate of echolocation over white canes for blind people because canes have a limited range. Unlike vision and sound, they can only tell you about what they are in physical contact with, and not what's 'over there'. 'Over there' is a very important place for an organism because it's where almost all of the world is, and if you can perceive it you give yourself more opportunities for activity and more time to make that activity work out. This is why Kish can ride a bike.

A recent paper (Buckingham, Milne, Byrne & Goodale, 2014; Gavin is on Twitter too) looked at whether information about object size gained via echolocation can create a size-weight illusion (SWI). I thought this was kind of a fun thing to do and so we read and critiqued this paper for lab meeting.

If you have two objects of the same actual weight but different sizes, people who perceive both the weight and size report that the smaller object is heavier than the larger object. This is typically considered a mis-perception of weight caused by inappropriately taking the size into consideration (see Figure 1).
Figure 1. When my daughter increases her volume with a jacket that doesn't weigh much, she feels much lighter than without the jacket. I have only *nearly* thrown her while picking her up :)
This is one of the most reliable and unbreakable illusions around. There are a variety of explanations for exactly why it happens (Buckingham, 2014). I personally like the very ecological idea that the SWI is a side effect of perceiving the throwability of objects, but regardless of the cause, if you have information about both the size and the weight of objects you will show the illusion.

Buckingham et al used 3 blind echolocators, 3 blind non-echolocators and 4 sighted controls. There were 6 objects; cubes that weighed either 1.5kg or 2kg and were either 15, 35 or 55 cubic centimetres). Groups first 'viewed' the objects (with vision for the controls or via tongue or finger clicks for the blind participants). They then lifted the objects by pulling on a rope attached to the boxes via a pulley and hook (this means you get no haptic perception of size to confound the results but can perceive the weight). There's a video of this procedure included with the paper.

Participants lifted the heavy, medium sized object 5 times at the start of the study and this was described as weighing '50 on a scale of 1-100'. For each test trial as described above, they rated the weight of the object on this scale using this reference. 

These ratings were normalised. For each subject, their weight judgments were expressed as a percentage of the highest rating they had given to objects in that weight set (this makes ratings for the two actual weights comparable). These ratings were then averaged over object weights and the ANOVA revealed both a main effect of object size as well as an interaction of group and size (Figure 2). 
Figure 2. Normalised and averaged ratings of weight by group and object size
They then subtracted the average ratings of large objects from that of small objects to get a measure of the size of the illusion (remember that the illusion makes small objects feel heavier; see Figure 3a). They tested each magnitude with a 1-tailed t-test to see if it was significantly different from 0; the echolocators (p<.05) and the sighted controls (p<.01) were. They then tested the three magnitudes with an ANOVA to look for differences in magnitude between the groups; there was a main effect and post-hocs showed that the sighted controls had a larger effect than the blind people and that the echolocators also showed a larger effect than the non-trained blind controls. Figure 3b shows the difference in raw judgments of weight between heavy and light objects. There was no significant difference between the groups, although with a p=.07 they were a bit lucky. They suggest the (nearly significant) greater discrimination by the blind people reflects known improvements in non-visual sensitivities as a result of being blind, although there's no way to match the judgments to the actual weights making it impossible to tell if the blind people were better at detecting the difference or simply rating it as larger.

Figure 3. a) The magnitude of the illusion for each group. b) The difference in ratings given as a function of actual weight changes
The net result of this analysis (although see below) is that the echolocating experts exhibited an SWI, although it was smaller than the illusion in the controls. 

One Tailed t-Tests
One immediate problem is the use of one tailed t-tests to test the critical condition; did the echolocators show an illusion significantly greater than 0? One tailed tests are very problematic (see this post by Andy Field for a great summary) and are not standard in social sciences for these reasons. They are basically only appropriate if it is physically impossible to get a mean difference in the other direction (which it is not in this case, and the blind controls show this too). Worse, the one tailed test is effectively a two-tailed test at twice the critical value of p; you only need a t value with a probability of <.1 to be significant. 

I used this table to check their result. The one tailed test for the echolocators resulted in t(2) = 3.4, p<.05 one tailed. For df=2, the critical value of t for the one tailed test with a 5% error rate is 2.92 (hence their result is significant). t(crit) for the two tailed test at 5%, however, is 4.303. Using the more appropriate two-tailed test, the echolocators show no significant size-weight illusion. 

The one tailed test sounds reasonable; there are clear reasons to only expect the illusion magnitude to go in one direction. But it is possible for the effect to go the other way, and the one tailed test is a less conservative test in general, and these facts make it inappropriate to use. 

One other related methodological thing. The sighted controls were able to view the boxes throughout the trial, from the beginning right up until they pulled on the rope to feel the weight. The echolocators did not do this; from the Methods it seems unlikely that they were clicking all the time, just during the exploration phase of the trial. While you can get the SWI from brief exposure to the size of an object, this methodological difference might be the reason the echolocating experts only showed a weak effect. Changing this might tighten up the result.

The authors try to make it sound quite surprising that 'echolocation-derived representations of object size influence the conscious perception of the ostensibly unrelated variable of object weight' (pg 5). But perception isn't about modality (or representations, but who's counting? :). It's about information, and to the extent that sound provides information about size, there's no reason why this shouldn't induce the size-weight illusion. Trained echolocators seem to be able to use sound to perceive their spatial layout in fairly useful detail (although the resolution of sound as a medium will affect the quality of this), so I actually think this should work and I wouldn't be all that surprised.

Hopefully the authors will get a chance to tighten the methods and analyses and do this again because it's a nice result that should work and it fits nicely with ecological notions of information being primary. It will have to be them that fixes it, though, because who else has 3 trained echolocators handy? (Note: if you are or know an echolocator in the UK send them our way :)

Buckingham G. (2014). Getting a grip on heaviness perception: A review of weight illusions and their probable causes. Experimental Brain Research, 232, 1623–1629. doi:10.1007/s00221-014-3926-9

Buckingham, G., Milne, J. L., Byrne, C. M., & Goodale, M. A. (2014). The Size-Weight Illusion Induced Through Human Echolocation. Psychological Science, doi:0956797614561267.

1 comment:

  1. Great article! I am going to forward this to Dr. Taji Huang she will get a kick out of this.