The use of control groups in artificial grammar learning

Area of Science:

• Cognitive Psychology
• Linguistics
• Experimental Psychology

Background:

• Artificial grammar learning (AGL) research often uses control groups without training to assess learning.
• The validity of AGL studies relies on an additivity assumption, where non-specific variables affect experimental and control groups equally.

Purpose of the Study:

• To investigate the validity of the additivity assumption in AGL research.
• To determine if control groups without training are methodologically sound for assessing grammar learning.

Main Methods:

• Two experiments were conducted to test the additivity assumption.
• Control groups without training and with training on randomized strings were compared.
• Published AGL research using control groups without training was reanalyzed.

Main Results:

• The additivity assumption was found to be invalid in both experiments.
• Control groups without training showed biases related to non-specific features, unlike experimental groups.
• Control groups trained on randomized strings exhibited fewer biases than untrained control groups.

Conclusions:

• The use of control groups without training in AGL research is methodologically unsound.
• Biases in untrained control groups can confound the assessment of artificial grammar learning.
• Reanalysis of prior studies supports the conclusion that untrained control groups are problematic.