Variability in neural activity and behavior
Introduction
Behavior is variable and often unpredictable. While some external events trigger only one specific behavior, such as a reflex, most external events trigger one out of many possible behaviors in a seemingly stochastic fashion. This behavioral variability is also observed in the laboratory, and can rarely be avoided completely. The colloquial Harvard Law of Animal Behavior sums it up as follows: ‘Under carefully controlled experimental circumstances, an animal will behave as it damned well pleases.’
Just as behavior is variable, so is the activity of neurons. This variability is observed in essentially all neural recordings, and even in the absence of behavior. Over the last two decades, many studies have quantified neural variability during the performance of well-controlled behavioral tasks. These studies have thereby opened the door to explore the relationship between neural and behavioral variability. Focusing on a sensory area, the pioneering work by Britten and colleagues showed that neural activity and behavioral choice covary on a trial-by-trial basis, even if the stimulus is constant [1]. This early work suggested that random fluctuations in the processing of sensory stimuli could be at the origin of behavioral variability. More recent work has rekindled interest in this topic, showing that the answer is more elusive than originally thought, and highlighting the flexible nature of neural variability across task conditions and behavioral states.
Here, we review recent research that addresses the relationship between neural and behavioral variability. We focus on studies of perceptual decision-making and consider three different, but related aspects of the problem: What is the origin of the observed neural variability? How does neural variability impact behavior? Finally, what are the functional or adaptive roles that neural variability may play?
Section snippets
Sources of neural variability
In the classical, reductionist approach, we seek to explain changes in some target variables (say, neural activity) by changes in some other variables (say, an external stimulus). In any particular neurophysiological experiment, not all factors that influence the targeted neural activity can be controlled or monitored. Unknown influences range from thermodynamic fluctuations in the state of ion channels to the various internal states of an organism. Since we lack access to these unknown or
Controllable sources of variability
In the classic experiments of Britten and colleagues, monkeys were trained to discriminate the overall direction of motion of a random dot pattern 1, 2. More specifically, the behaviorally relevant information was provided by the fraction of dots moving in a coherent direction. However, many distinct patterns of random dot motion are compatible with a given motion coherence, and neurons in cortical area MT are extremely sensitive to these fine-grained dot patterns, especially at low coherences
Uncontrollable sources of variability
The strong coupling of widely different spatial and temporal scales in biological systems in general, and in the nervous system in particular, generates sources of variability that are difficult to control. One obvious example are microscopic fluctuations of small structural components of neurons, such as ion channels or the synaptic machinery involved in vesicle release [27]. Although the precise impact of this microscopic noise on neural variability remains unclear, it can, in principle, be
Neural versus behavioral variability: normative approaches
In the same way that the experimenter needs to determine which variables affect neural activity, neural circuits need to properly allocate responsibility to the diverse influences that can change their synaptic inputs. In solving this inference problem, the brain may use processing strategies that appear inefficient or suboptimal, as recently proposed by Beck and colleagues [47•]. Ideally, neural circuits should retain the relevant aspects of their complex and high-dimensional inputs (the
Neural versus behavioral variability: mechanistic approaches
The relation between neural and behavioral variability was first quantitatively characterized in the context of the classical random dot motion discrimination task, with the remarkable observation that trial-to-trial fluctuations in the activity of single neurons in monkey area MT evoked by stimuli of the same coherence were correlated with behavioral choice [1], a tendency referred to as choice probability (CP). This relationship between neural variability and choice was formalized in the
The functional roles of behavioral variability
Is variability a bug or a feature? Variability may be a feature if it provides adaptive benefits to the organism. A simple case of such adaptive benefits at the behavioral level is exploration. The need to explore new possibilities may benefit from a stochastic strategy. In the theory of reinforcement learning, for instance, the tradeoff between exploring something new or exploiting the already known is handled probabilistically 64, 65. Since all environments change over time, animals can
Conclusions
The relation between neural and behavioral variability is key to understanding how brains generate behavior. A lot of insight about this relation has been gained during the last decade, and yet it seems as if we merely scratched the surface. In any particular experiment, all of the effects mentioned above are likely to play a role, from uncontrolled environmental variables to noise generated within the nervous system, including contributions linked to specific processing strategies used by the
References and recommended reading
Papers of particular interest, published within the period of review, have been highlighted as:
• of special interest
•• of outstanding interest
Acknowledgements
We thank Nuno Calaim, Asma Motiwala and Claudia Feierstein for comments on the manuscript. A.R. and C.K.M acknowledge support from the Champalimaud Foundation.
References (73)
- et al.
Modulation of visual responses by behavioral state in mouse visual cortex
Neuron
(2010) - et al.
Walking modulates speed sensitivity in drosophila motion vision
Curr Biol
(2010) - et al.
Neural variability in premotor cortex is modulated by trial history and predicts behavioral performance
Neuron
(2013) - et al.
Differential attention-dependent response modulation across cell classes in macaque visual area v4
Neuron
(2007) - et al.
Spatial attention decorrelates intrinsic activity fluctuations in macaque area v4
Neuron
(2009) - et al.
Attention improves performance primarily by reducing interneuronal correlations
Nat Neurosci
(2009) - et al.
Model of global spontaneous activity and local structured activity during delay periods in the cerebral cortex
Cereb Cortex
(1997) - et al.
Slow dynamics and high variability in balanced cortical networks with clustered connections
Nat Neurosci
(2012) - et al.
Abrupt transitions between prefrontal neural ensemble states accompany behavioral transitions during rule learning
Neuron
(2010) - et al.
Temporal integration of optic flow, measured by contrast and coherence thresholds
Vision Res
(2001)
Representation of confidence associated with a decision by neurons in the parietal cortex
Science
Speed and accuracy of olfactory discrimination in the rat
Nat Neurosci
Decision-related activity in sensory neurons: correlations among neurons and with behavior
Annu Rev Neurosci
Humans integrate visual and haptic information in a statistically optimal fashion
Nature
Changes in the neural control of a complex motor sequence during learning
J Neurophysiol
A relationship between behavioral choice and the visual responses of neurons in macaque MT
Visual Neurosci
Neuronal correlates of a perceptual decision
Nature
Temporal precision of spike trains in extrastriate cortex of the behaving macaque monkey
Neural Comput
Precise olfactory responses tile the sniff cycle
Nat Neurosci
Olfactory bulb units: activity correlated with inhalation cycles and odor quality
Science
Selective delay activity in the medial prefrontal cortex of the rat: contribution of sensorimotor information and contingency
J Neurophysiol
Noise and the visual threshold
Nature
Matching behavior and the representation of value in the parietal cortex
Science
The relative influences of priors and sensory evidence on an oculomotor decision variable during perceptual learning
J Neurophysiol
Stimulus onset quenches neural variability: a widespread cortical phenomenon
Nat Neurosci
Cortical state and attention
Nat Rev Neurosci
Basal forebrain activation enhances cortical coding of natural scenes
Nat Neurosci
State-dependent representation of amplitude-modulated noise stimuli in rat auditory cortex
J Neurosci
The asynchronous state in cortical circuits
Science
A method for closed-loop presentation of sensory stimuli conditional on the internal brain-state of awake animals
J Neurosci Methods
Acetylcholine contributes through muscarinic receptors to attentional modulation in v1
Nature
Laminar differences in gamma and alpha coherence in the ventral stream
Proc Natl Acad Sci U S A
Synchrony and covariation of firing rates in the primary visual cortex during contour grouping
Nat Neurosci
A neuronal population measure of attention predicts behavioral performance on individual trials
J Neurosci
Optical imaging of neuronal populations during decision-making
Science
Noise in the nervous system
Nat Rev Neurosci
Cited by (131)
Energy balance regulates synchronization of photosensitive neural network
2023, Physica A: Statistical Mechanics and its ApplicationsBounded contribution of human early visual cortex to the topographic anisotropy in spatial extent perception
2024, Communications BiologyEnhancing Interoceptive Abilities and Emotional Processing: Effects of HD-tDCS
2024, Journal of Integrative Neuroscience