Based on 8 years of experience recruiting alcohol users into studies of comparable length
and complexity, the investigators anticipate that 92% of consented participants will be
eligible (i.e., randomized) and 97% of those individuals will complete Aim 2b. Thus, to
achieve a final sample size of 100, the investigators anticipate consenting 111
participants. This study will include 50 real time fMRI (experimental group) participants
and 50 "idealized" neurofeedback (control) participants. Participants will be randomly
assigned to experimental or control groups, stratified by AUDIT scores, SES, age, and
sex. All study procedures will take place during a single study visit. Before the imaging
session, a research assistant will aid the participant to select salient images related
to individualized future goals as well as alcohol images which elicit self-reported
craving. All participants will complete two fMRI runs within a single session. In run 1
of the scan (offline classifier training), participants will perform an image viewing
task displaying alternating blocks of these goal- and alcohol-related images. Between the
image viewing blocks we will assess reinforcer pathology with brief in-scanner measures
of delay discounting and alcohol valuation (BAAD). In total, there will be six blocks
each of the goal-oriented images, alcohol-related images, delay discounting, and BAAD (24
blocks total). Since this is a proof-of-concept experiment, the interspersed delay
discounting and BAAD blocks will enable monitoring and validating changes in the
participant's temporal window and alcohol valuation after the image viewing blocks. Only
the fMRI measures of goal- and alcohol-related image viewing will be used to build a
support vector classifier (SVC) of decreased vs. increased delay discounting. If
participants are allocated to the real time fMRI group, in run 2 the SVC model from run 1
will be used to provide participants with real-time neurofeedback while they attempt to
modulate their temporal window. The participants will see a dial with a needle on the
screen along with instructions to think of either future goals or alcohol cues. The dial
will be directly controlled by ongoing output from the SVC, updating the needle position
as participants imagine immediate (alcohol-related) or delayed (goal-related) cues.
Control participants ("idealized" neurofeedback group) will undergo the same procedures,
but in run 2 neurofeedback, will be presented with an "idealized" dial (perfect needle
movements in both alcohol and goal directions) and will be specifically instructed that
they are not controlling the interface. The investigators have used this type of sham
needle movement as a control condition in previous experiments. Based on this, it is
expected that experimental group will have increased whole brain signal to noise for
alcohol vs. goal fMRI analyses and increased recruitment of frontal-parietal networks
from enhanced visual attention to the task.