Double dissociation of sound localization and identification in the auditory cortex of cats

We have known for some time that there is a double dissociation (I will define that term in a minute) between location and identification in the visual system. Neuroscientists speak of a "where" pathway that goes from the primary visual cortex in the occipital lobe up into the parietal lobe. Lesions to this pathway produce deficits in locating objects in space using vision. There is also a "what" pathway that goes from the primary visual cortex down into the temporal lobe. Lesions to this pathway produce deficits in identifying objects using vision.

We knew that was true for vision, but there is less evidence that this division exists in hearing. Lomber and Malhotra, publishing in the journal Nature Neuroscience, show that this division does exist. They use a double dissociation experiment in cats.

What is a double dissociation?

A double dissociation is an important type of experiment in neuroscience. The idea with a double dissociation is to demonstrate the independence of two (or more) process within the brain using lesions/inactivations.

It works something like this. Say I have two cognitive processes A and B. I develop two tests to evaluate performance for A and B; we then say that I have operationalized those processes. Then I look at different brain regions to see what brain regions -- when lesioned or inactivated -- lead to deficits in A and B with respect to those tasks. When I have found the minimum regions required for those task -- i.e. the regions that cannot be subdivided further without a restoration of task performance -- I compare the regions associated with deficits in A and B. If those regions are separate in the brain, I assert that I have doubly dissociated those processes and that they are (in part...see caveats below) independent.

The data looks something like this:

You can see that lesions to one brain region result in a deficit in only one of the two processes. This is a double dissociation. A single dissociation is like the one below:

In a single dissociation, you know that process B is in degrees separable from process A, but you can only attribute process B to one brain region. Lesions to either region results in deficits in process A. This might suggest that the process A is above B in a hierarchy of processes or it might not.

To understand this better think of the ways a color TV can break and what that says about how the TV works. If you have a color TV, you could lose the sound without losing the picture. You could lose the picture without losing the sound. This suggests that the two processes are independent. They have been double dissociated. In contrast, you could lose the color in the image while still maintaining the image. You could not, on the other hand, lose the image and maintain the color. This would be a single dissociation. Color and image are separable, but image is above the color in a hierarchy.

Double dissociations are important experiments to divide the brain into different parts and associate those parts with specific functions. However, there are some important caveats to consider when you interpret a double dissociation study:

• First, the association of process A with region 1 does not imply that process A is a unitary process. It could be subdivided further, and these sub-processes might be associated with different parts of region 1.
• Second, in multi-part processes, process 1 and 2 might have common steps. There may be brain regions that result in deficits in both. When you have a double dissociation all you know is that at some point in the two processes they are independent, not that they are completely independent.
• Third, your grasp of a process is only as good as it has been operationalized. You cannot say that a brain region does "altruism" or even "vision" from a double dissociation. What you can say is that lesions to the brain resulted in a deficit in a specific task. How related your task is to the process you are attempting to operationalize matters a great deal.

An Example of a Double Dissociation: Lomber and Malhotra

Let's apply our new knowledge of double dissociations to Lomber and Malhotra. The authors show that location and identification of sounds in the cat cortex are double dissociated.

The authors selectively inactivated two regions of the cat auditory cortex: the posterior and the anterior auditory cortex. These brain regions are depicted below (Figure 1 from the paper):

Areas: AAF, anterior auditory field (dark gray); AI, primary auditory cortex; AII, second auditory cortex; dPE, dorsal posterior ectosylvian area; DZ, dorsal zone of auditory cortex; FAES, auditory field of the anterior ectosylvian sulcus; IN, insular region; iPE, intermediate posterior ectosylvian area; PAF, posterior auditory field (light gray); T, temporal region; VAF, ventral auditory field; VPAF, ventral posterior auditory field; vPE, ventral posterior ectosylvian area. Sulci (lowercase): aes, anterior ectosylvian; pes, posterior ectosylvian; ss, suprasylvian. Other abbreviations: A, anterior; D, dorsal; P, posterior; V, ventral. The areal borders shown in this figure are based on a compilation of electrophysiological mapping and cytoarchitectonic studies.

In each individual test animal, the authors performed surgery on the cats brain to insert cooling electrodes bilaterally (remember that the brain has two hemispheres both of which need to be inactivated) over each of these test regions. The cooling electrodes temporally inactivated the region they are placed above. The ability to temporarily inactivate a brain region is a handy thing because it means that the animals can be their own controls and be tested for both regions.

Under conditions of cooling anterior auditory field (AAF), posterior auditory field (PAF), and uncooled controls, the animals were then tested on two tasks (for each of the two process: localization and identification).

First, the authors test the ability of the animals to orient and approach a loud burst of noise projected from different directions. The animal had to identify the direction in order to receive a reward. The data is shown below (Figure 3 in the paper, click to enlarge):

These are radial diagrams, so the performance in percentage at different angles is depicted as the length of the line away from the center. As you can see, there is a profound deficit in sound localization from the temporary inactivation of PAF.

The authors then tested the ability of the animals to discriminate between sounds. The test sounds that they used were sort of like Morse code with repetitive patterns composed of short or long beeps. The patterns were balanced for loudness and duration, so the animal really did have to identify the correct pattern by pattern alone. The cat was forced to choose between approaching two patterns only one of which was rewarded.

Here is the data from that experiment (Figure 5 in the paper, click to enlarge):

You can see that performance on this task drops to near chance when the AAF is cooled.

See how the results of these experiments follow the rubric for double dissociation that I discussed above. Temporary inactivation of PAF resulted in deficits in sound localization but not identification. Temporary inactivation of AAF resulted in deficits in sound identification but not localization. Thus, we can say that these two processes are at least partially indepedent. I say partially because we know that this separate information has to come back together at some point. First, it is derived from the same organs. Second, in order to form a complete representation of an object the separate pieces of information must eventually rejoin.

What is the significance of this work?

This big finding is that the "division of labor" (their words) that is present in the visual system is paralleled in the auditory system. They summarize thusly:

Our results demonstrate a clear division of labor in auditory cortex. Although a one-to-one relationship might not be expected between functional streams in visual cortex and functional streams in auditory cortex, our results significantly strengthen the notion that functional segregations and processing streams are a common attribute of mammalian cortical sensory systems. Specifically, the proposal that 'what' and 'where' streams may exist in auditory cortex is substantially supported. However, the spatial and pattern processing dichotomy is not the only proposed cortical processing configuration. Visual cortical processing pathways subserving perception (ventral stream) and action (dorsal stream)8 have been hypothesized. This theory emphasizes the output requirements of the dorsal and ventral pathways, rather than the input or sensory distinctions. Evidence in support of this model has been obtained from double-dissociation studies of neurological patients, as well as functional imaging studies of healthy subjects. When applied to our findings, this proposal would suggest that AAF is involved in perception and PAF is involved with action. Deactivation of AAF disrupted the perception of the gap sequences and the deactivation of PAF disrupted the action of accurately directing the head and body, and subsequent approach, to the acoustic stimulus. Therefore, although we did not specifically design our tasks to test the perception-action dichotomy, our results do support this cortical segregation. Finally, it is also important to consider that there may be more than two processing streams. These 'streamlets' or 'streams within streams' may very well be present in nonprimary auditory cortex. Given the large number of ways an acoustic pattern or object can be defined, there may well be multiple auditory object areas in the 'what' processing pathway. Furthermore, although our results substantially support the proposal that 'what' and 'where' streams may exist in auditory cortex, further verification of this will require a demonstration that other perceptual attributes that help to identify a sound, such as its pitch or timbre, can also be disrupted independently of sound localization. (Emphasis mine. Citations removed.)

An excellent paper, and an excellent example of an important experimental paradigm in neuroscience!

Hat-tip: Faculty of 1000

Lomber, S.G., Malhotra, S. (2008). Double dissociation of 'what' and 'where' processing in auditory cortex. Nature Neuroscience, 11(5), 609-616. DOI: 10.1038/nn.2108