Measuring perception
Obtaining robust measures of the subject’s experience is crucial to the lab. The SENSEx team, past and present, has extensive experience both in human psychophysics, e.g. 1-9 and in the combined approach of rat psychophysics + neuronal coding + computational modeling, e.g. 7,8,10-26. The core behavior is one we have refined over the past 10 years – the judgment of stochastic, “noisy” vibrations 7-9,21,26-28. Basic algorithms extracted from tactile psychophysics, such as evidence accumulation and reaction time as a measure of confidence, align with other sensory modalities 20,29. We have developed various versions – working memory, one-back memory, reference memory. In the reference memory task, for example, the subject (rat or human) judges each trial’s vibration as belonging to one of two categories: high-intensity or low-intensity (strong or weak).
Cross-species strategy to explore single-trial choice
Parallel studies in rodents and humans – the approach taken in SENSEx – can be particularly informative about the mechanisms of perceptual decisions 30. Until some 10 years ago, neuroscientists attributed a wide range of cognitive functions to primates but not to rodents. However, as methodologies adapt to natural deportment, rodents have been found to express a surprising range of abilities. Rats hold stimuli in working memory with reliability akin to that of humans and monkeys 7, integrate separate sensory modalities to create a supramodal object representation 22,31, assess reward statistics 32, indicate their degree of confidence in their choices 33, and generalize rules extracted from a specific task to novel experiences 34. In short, rodents are ever more fulfilling their promise of expressing high-level sensory-perceptual cognition 35. Importantly, they achieve such cognition through the workings of neuronal circuits that are accessible 36, decodable 37, and manipulatable by optogenetics 38.
Typically occupying dark environments, rodents rely on tactile information generated through their facial vibrissae. Recent work emphasizes dual modes of vibrissal information acquisition: (i) generative, where the rat sweeps its whiskers forward and backwards to palpate objects 39, and (ii) receptive, where the rat keeps its whiskers still, and collects mechanical signals 7. In natural environments, vibrissal vibrations could provide signals about large predators moving above the burrow or about an arriving subway train (in rural and urban settings, respectively) 15,28 while in our studies the vibration is delivered through a computer-controlled plate.
Accumulation of evidence in vibration perception
By design, the holistic percepts emerging from the vibration depend on the accumulation of sensory evidence over time. Vibrations are low-pass filtered white noise, consisting of a stream of instantaneously varying features 7,9,20,23,28,29. The vibration’s stochastic nature precludes using a short-time-window “snapshot” to estimate the whole-vibration intensity: by central limit theorem, an ideal observer’s judgment of the input stream benefits by integrating. In fact, we see that humans and rats show better intensity judgment thresholds with increased duration up to ~500 ms. In vS1 the firing probability of vibration-sensitive neurons reports only the vibration’s recent kinematic features (last 5-30 ms) 20,23; this temporally “local” vibration code could not account for intensity, a percept built up over hundreds of ms. The explicit neuronal substrate is therefore likely to be found in integrative centers downstream from vS1. Indeed, we recently found that vS1 acts as a distribution center for perceiving stimulus intensity 26: when we optogenetically excited vS1, the intensity percept was amplified. The overall picture of vibration evidence integration (like texture, in our earlier work 24,29,39,40) resembles that seen with primate visual perception 29.