Thursday, July 18, 2013

Not so free associations


Our brains are amazingly good at associating events. Pictures, sounds, smells, actions, words are all connected together to form the basic fabric of our internal representations. Miyashita’s research group in Japan has performed a series of elegant experiments demonstrating that individual neurons in monkeys’ brains have a remarkable power to form such associations [1-3].
In these experiments, monkeys are presented with a series of abstract patterns such as the ones illustrated in the figure below. Monkeys are trained over the course of several months to memorize associations between specific pairs of such patterns. For example, they learn that pattern A goes with F and pattern B goes with J. After training, the investigators test the monkeys’ ability to remember those associations in a so-called delay-match-to-pair task, illustrated below. The monkey sees a cue image (say pattern A) and then, after a delay, is presented with patterns B and F. The monkey has to indicate (e.g. by pulling a lever or making a saccade) whether the corresponding match for cue A is B or F. Correct performance is compensated by an aliquot of their favorite juice.
            
In a recent study, the authors used a device called a “tetrode” to interrogate the responses of multiple neurons in a particular part of the monkey visual system denominated inferior temporal cortex [1]. Previous studies have shown that this area plays a critical role in visual object recognition [4, 5]. For example, monkeys who have lesions in this area find it extremely difficult to learn to discriminate among different novel object shapes. Neurons in this area of the brain are quite picky in their preferences. They respond to certain shapes but not to others. Following with the examples here, one of the neurons may respond vigorously to presentation of pattern A but only minimally to pattern B. After training, the investigators observed that the preferences of some of the neurons started to reveal the type of associations imposed at the behavioral level. In other words, a neuron would respond to A and also to F (but not to B or J) while another neuron would respond to B and J (but not to A or F). Obviously, monkeys are not born with intrinsic associations for A and F. These associations must develop during the rigorous training undergone in the laboratory.
Some of the neurons retained their individual preferences for individual items while other neurons became more associative. By looking at the exact times when neuronal pairs fired in close temporal proximity, the authors inferred presumed interactions (imagine that you are holding a conversation with a friend, by considering the sequential events in the voice patterns, one could infer that you are talking with him and not to somebody else). Interestingly, the neurons that preferred to respond to individual items (say A) seemed to directly connect to those neurons that responded to the corresponding pairs (say A and F). These observations seem to point to a hierarchy of neurons that respond to individual items and, together with other neurons, give rise to associative neurons that can hold information about joint combinations of stimuli.  

References