Heidi E. Harley – harley@ncf.edu
Wendi Fellner – wendi@wendiops.com
Barbara Losch – dakcali@yahoo.com
New College of Florida
5800 Bay Shore Road
Sarasota, FL 34243
Popular version of paper 3aAB7
Presented Wednesday morning, June 5, 2013
ICA 2013 Montreal
Dolphins have a form of biosonar, called echolocation, through which they can investigate the world using sound. To echolocate, dolphins emit sound pulses and then process the echoes that bounce back from their environment.
Figure 1. This is a spectrogram that shows the pattern of clicks a dolphin is producing as he echolocates an object. Each line is an echolocation click. The vertical axis is frequency (so you can see that a dolphin click has energy in many frequencies at once, since the click is a vertical line), and the horizontal axis is time.
Click here to listen to the echolocation trains presented in the spectrogram.
Dolphins are superior at processing sounds, and they can use echolocation to tell objects apart that vary in subtle ways, e.g., cylinders with wall thicknesses that differ by less than a centimeter. (Whitlow Au’s 1993 book, The Sonar of Dolphins, provides a very good review of dolphin echolocation.) Dolphins can also integrate information that they receive through echolocation and vision around a single object in the same way that humans can integrate information that they receive through multiple sensory systems: For example, you represent the shape, color, jangle, and texture of keys even though you’ve accessed this information from different sensory systems, i.e., vision, hearing, and touch.
In 1920, Hartridge suggested that bats produced “sound pictures” with echolocation, and dolphin echolocation has often been described as “seeing with sound”; but visual and acoustic processing are rather different from each other. For example, think about seeing a painting versus hearing a song. Both the painting and the song include patterns, but with the painting you have access to the whole pattern at once, whereas the song requires you to get information and remember it over time in order to access the pattern. When a dolphin echolocates an object, it’s possible that he’s creating a “sound picture”. However, in order to do so, he’d likely have to process the echoes that he receives from an object and calculate the time and space differences across multiple echoes returning (and not returning – think of empty spaces in and around an object) over time in order to recreate the layout of the object. Such complex processing is not impossible. After all, a similar reconstruction occurs with manmade ultrasound imagers. On the other hand, these machines produce exponentially higher rates of pulses than do dolphins and they also have an enormous memory in which to store specific returns to create an image – which they create because humans depend heavily on sight to learn about the world. In contrast, dolphins must work under the biological constraints of normal memory parameters as well as much slower echolocation pulse rates and neuronal transmission rates.
Our study investigated the likelihood that dolphins create “sound pictures” using their echolocation. We focused on shape recognition because shape is an object attribute that lends itself to a holistic representation – that is, it’s very easy to recognize a shape when you see the whole thing all at one time, whereas it can be difficult to put it together piecemeal as in a puzzle. In this study we taught a dolphin to perform a matching task in which he sensed a sample object and then received fish for choosing an object identical to that sample from among a group of three alternatives. We used multiple object sets to test the dolphin. Some of the objects in a set varied only in shape because we used the same PVC parts to create them. Other sets had objects created from multiple PVC components so their features were different.
Figure 2. The upper set of stimuli is an equal-parts set in which the same pieces of PVC were used to create each of the three objects in the set. Therefore, in this set, although shape changed across stimuli, components did not. The lower object set is an unequal-parts set in which both shape and components varied across stimuli. The small 2-inch (5.08 cm) rectangles next to each object indicate scale.
We also used some “junk” objects that varied in all kinds of ways – material, shape, size, etc. In addition, the dolphin matched the objects using either vision or echolocation. When he used vision alone, the objects were presented in air where he could see them but couldn’t echolocate them. (Dolphin echolocation doesn’t work in air because it’s designed to work in water which is much denser than air.) When he used echolocation alone, the objects were presented in water and the dolphin was trained to voluntarily wear soft latex eyecups over his eyes that he could pop off. (If he popped them off during a trial, we stopped using those objects in testing.)
We found that the dolphin’s performance was worst when he was using echolocation alone and the objects were all made of the same parts: He averaged 45% correct over 12 sets of objects; chance is 33% because there are three objects from which he could choose. He was much better with these objects when he could see them (67%), and he was better still when he viewed objects that were made from different parts (76%). He’s a good echolocator – his accuracy when he echolocated junk objects was 81% – so that doesn’t explain the differences. It’s more likely that the difference in his performance between seeing and echolocating objects is based on the way he uses vision and echolocation. Vision is good for detecting differences in shape because you can get shape information all at once when you look at an object – and echolocation really isn’t so good for shape discrimination although it’s good for other things like discriminating among objects that are made of different materials. Our current findings do not suggest that dolphins create “sound pictures” when they echolocate because they weren’t very good at using echolocation to tell objects apart that varied only in shape. Since dolphins have both vision and echolocation at their disposal and they can integrate information from both sensory systems, they likely use both – and one doesn’t completely replicate the other.