Skip to main content

Abstract

The artificial grammar learning (AGL) paradigm has been used for decades to investigate the ability to implicitly learn structured patterns in the environment (Reber, 1967).Two types of information are known to influence learning: grammaticality (the extent that sequences are consistent with the rules of the grammar) and chunk strength (surface similarity to previously observed exemplars). To further investigate the role of these two types of information on learning, we administered theAGL task to eighteen participants (18–33 years old). During Training, participants were instructed to reproduce symbol sequences generated from an artificial grammar; during Test, participants had to decide whether new sequences followed the grammar or not. Apart from grammatical, sequences could also have high or low chunk strength(CS; a measure of a test item’s surface similarity to training sequences). Thus, sequences fell into four conditions: Grammatical – High chunk strength (GH), Grammatical – Low chunk strength (GL), Non-Grammatical – High chunk strength (NGH), and Non-Grammatical – Low chunk strength (NGL)].Findings revealed a statistically significant main effect for grammaticality [F(1,17)=23.777; p=h2=.583] and CS [F(1,17)=6.026; p=.025; h2=.262]. This was qualified by a significant interaction of grammaticality x CS [F(1,17)=28.504; p=h2=.626]. For the main effect of grammaticality, there was a statistically significant difference [t(17) = 5.315; p M=19.61;SD=5.393) and NGL (M=13.44; SD=3.823); but the difference between GH (M=19.61; SD=4.767) and GL (M=21.89; SD=4.600) was non-significant [t(17) = -2.195; p =.042]. Additionally, for CS the difference between GL and NGL was statistically significant [t(17) = 6.251; pt(17) = .172; p =.865]. Consistent with previous research, both grammaticality and CS influenced participants’ learning. Testendorsements were higher for items high in CS or grammaticality (or both) and lower for items that were both non-grammatical and low in CS.

Files

File nameDate UploadedVisibilityFile size
auto_convert.pdf
19 Jul 2022
Public
214 kB

Metrics

Metadata

  • Event location
    • Nesbitt 3110

  • Event date
    • 3 November 2018

  • Date submitted

    19 July 2022