Learning to avoid dangerous signals while preserving normal responses to safe stimuli is essential for everyday behavior and survival. Following identical experiences, subjects exhibit fear specificity ranging from high (specializing fear to only the dangerous stimulus) to low (generalizing fear to safe stimuli), yet the neuronal basis of fear specificity remains unknown. Here, we identified the neuronal code that underlies inter-subject variability in fear specificity using longitudinal imaging of neuronal activity before and after differential fear conditioning in the auditory cortex of mice. Neuronal activity prior to, but not after learning predicted the level of specificity following fear conditioning across subjects. Stimulus representation in auditory cortex was reorganized following conditioning. However, the reorganized neuronal activity did not relate to the specificity of learning. These results present a novel neuronal code that determines individual patterns in learning. Keywords: fear conditioning, auditory cortex, sensory systems, learning, computational model, imaging, sensory cortex, tuning curve, neurobiology, population coding.

All imaging sessions were carried out inside a single-walled acoustic isolation booth (Industrial Acoustics). Mice were placed in the imaging setup, and the head plate was secured to a custom base (eMachine Shop) serving to immobilize the head. Mice were gradually habituated to head-fixing over 3 – 5 days, 3 – 4 weeks after surgery and before imaging commenced. Imaging took place in mice aged 17.5 – 19.6 weeks (min: 12.9, max: 27.1 weeks).

We recorded changes in fluorescence of GCaMP6s/m caused by fluctuations in calcium concentration in transfected neurons of awake, head-fixed mice, using two-photon microscopy (Ultima in vivo multiphoton microscope, Bruker). The laser (940 nm, Chameleon Ti-Sapphire) power at the brain surface was kept below 30 mW. Recordings were made at 512 x 512 pixels and 13-bit resolution at ~30 frames per second.

Stimuli were generated at a sampling rate of 400 kHz using Matlab (MathWorks, USA)  and consisted of 100-ms long tone pips in the 5−32-kHz frequency range and presented at 60 – 80 dB SPL. In a single recording session, each frequency was repeated 15 – 25 times in a pseudo-random order with a 4-s inter-stimulus interval.

This dataset consists of all data necessary to reproduce figures from the manuscript. Please read the README file. Code supplied is written in Matlab. Data is supplied in Matlab files (.mat)