5-choice serial reaction time task

An open-hardware replication of the 5-choice serial reaction time task (5-CSRTT), a classic mouse paradigm for sustained attention and impulsivity. The original relied on proprietary hardware and a custom language; here it runs on BeeHive plus Python.

Ingredients

BeeHive boards:

Other components:

How it works

The animal faces five nose-poke ports. On each trial, one port's yellow LED at the back lights briefly; the mouse must poke that port to earn a food pellet. Each port carries an IR LED + IR sensor pair from the IR sensor array: a poke breaks the beam, detected with microsecond precision, which is what makes accurate reaction-time and premature-response scoring possible.

Correct pokes trigger the pellet dispenser — a redesigned open-source dispenser that uses a servo instead of a stepper and is laid flat for easy 3D-printing. The same dispenser is shared with the mouse maze.

Because the whole paradigm lives in Python, behavioural variants — cue duration, inter-trial interval, punishment for premature or incorrect responses — are changed purely in code, no rewiring. In practice, self-directed training reaches stable performance in 7–10 days, versus the 3–5 months typical of the traditional setup.

Wiring

• The five IR LED + sensor pairs to the IR sensor array, which takes a mainboard data line.
• The five yellow cue LEDs to mainboard outputs (or a switch array if pins are tight).
• Pellet dispenser servo to a mainboard data line + 5 V power.

Code

This box has its own repository — BeeHive-org/5-choice-serial-reaction-time — with the full paradigm under software/5-csrtt/, driven from a host PC using the Belay library. The building blocks are small MicroPython test scripts, reproduced below with the box's actual pin assignments.

Read the five nose-poke IR sensors — a broken beam (value() == 0) is a poke (IR_test.py):

from machine import Pin
from time import sleep

pokes = [Pin(p, Pin.IN) for p in (15, 2, 16, 17, 19)]  # nose-poke IR receivers

while True:
    print([p.value() for p in pokes])
    sleep(1)

Light the yellow cue LEDs (Led_test.py) and drive the pellet-dispenser servo (Servo_test.py):

from machine import Pin, PWM
import time

cues = [Pin(p, Pin.OUT) for p in (14, 27, 25, 26, 32)]  # one per nose-poke
dispenser = PWM(Pin(18), freq=50)                        # pellet-dispenser servo

def dispense():
    dispenser.duty(23)               # 0.5 ms pulse -> one end of travel
    time.sleep(1)
    dispenser.duty(123)              # 2.4 ms pulse -> the other end
    time.sleep(1)

The complete trial logic — cueing, response scoring, the food magazine, and the staged training protocol — lives in outside2.py and serial_beehive.py.

Results / notes

Replicating the 5-CSRTT on open hardware slashes both cost and training time: self-directed training reaches stable performance in 7–10 days rather than 3–5 months, and behavioural paradigms are altered entirely in Python. The servo pellet dispenser is shared with the mouse maze.