This study examined how input complexity affects long-term retention of statistically learned regularities in an artificial language learning task. Participants were split into two groups, one that received mainly simple input and one that received mainly complex input. The complex input group showed poorer retention after a two-week delay. Additionally, those in the complex input group relied more on surface-level fragments of the language rather than the deeper grammatical regularities, suggesting increased input complexity hinders the ability to form abstract representations of the linguistic patterns.
Analytical Profile of Coleus Forskohlii | Forskolin .pptx
How incremental stimulus complexity affects long-term retention of statistical learning
1. Input complexity affects long-term
retention of statistically learned
regularities in an artificial
language learning task
Ethan Jost, Katherine Brill-Schuetz, Kara Morgan-Short
and Morten H. Christiansen
2019
Frontiers in Human Neuroscience
2. Input complexity affects long-term
retention of statistically learned
regularities in an artificial
language learning task
Ethan Jost, Katherine Brill-Schuetz, Kara Morgan-Short
and Morten H. Christiansen
2019
Frontiers in Human Neuroscience
4. Statistical learning
• Latent distributional properties in a disorganised environment
• Very early development in humans (Kirkham et al., 2002)
• Shown by other animals (Hauser et al., 2001)
• Uses in statistics and cognitive sciences (Louwerse & Connell, 2011)
5. Statistical learning
• Latent distributional properties in a disorganised environment
• Very early development in humans (Kirkham et al., 2002)
• Shown by other animals (Hauser et al., 2001)
• Uses in statistics and cognitive sciences (Louwerse & Connell, 2011)
• Relatively implicit and unconscious (cf. textbook instruction)
6. Statistical learning
• Latent distributional properties in a disorganised environment
• Very early development in humans (Kirkham et al., 2002)
• Shown by other animals (Hauser et al., 2001)
• Uses in statistics and cognitive sciences (Louwerse & Connell, 2011)
• Relatively implicit and unconscious (cf. textbook instruction)
• SL might redress the poverty of the stimulus and indeed convey an
advantage to children during language development (Newport, 1990)
8. Consolidation and retention
• Not so many studies
• Some studies not using linguistic stimuli
• Longitudinal testing
• Weeks (Jost et al., 2019)
9. Consolidation and retention
• Not so many studies
• Some studies not using linguistic stimuli
• Longitudinal testing
• Weeks (Jost et al., 2019)
• Months (Morgan-Short et al., 2012)
• Limitation: great degree of practice obscuring effect of exposure
10. Consolidation and retention
• Not so many studies
• Some studies not using linguistic stimuli
• Longitudinal testing
• Weeks (Jost et al., 2019)
• Months (Kara Morgan-Short et al., 2012)
• Limitation: great degree of practice obscuring effect of exposure
• One year (Kobor et al., 2017; Romano et al., 2010; Smalle et al., 2017)
• Limitations of Kobor et al. (2017) and Romano et al. (2010): no language, no meaning,
simple statistical structure in stimuli, no generalisation of regularities to new items
11. Consolidation and retention
• Not so many studies
• Some studies not using linguistic stimuli
• Longitudinal testing
• Weeks (Jost et al., 2019)
• Months (Morgan-Short et al., 2012)
• Limitation: great degree of practice obscuring effect of exposure
• One year (Kobor et al., 2017; Romano et al., 2010; Smalle et al., 2017)
• Limitations of Kobor et al. (2017) and Romano et al. (2010): no language, no meaning,
simple statistical structure in stimuli, no generalisation of regularities to new items
• 17 years (Mitchell, 2006)
13. Consolidation and retention
• Sleep (reviewed by Jost et al., 2019)
• Spaced repetition (reviewed by Ullman & Lovelett, 2018)
14. Consolidation and retention
• Sleep (reviewed by Jost et al., 2019)
• Spaced repetition (reviewed by Ullman & Lovelett, 2018)
• Retrieval practice (reviewed by Ullman & Lovelett, 2018)
15. Consolidation and retention
• Sleep (reviewed by Jost et al., 2019)
• Spaced repetition (reviewed by Ullman & Lovelett, 2018)
• Retrieval practice (reviewed by Ullman & Lovelett, 2018)
• …
16. Consolidation and retention
• Sleep (reviewed by Jost et al., 2019)
• Spaced repetition (reviewed by Ullman & Lovelett, 2018)
• Retrieval practice (reviewed by Ullman & Lovelett, 2018)
• Incremental stimulus complexity (Poletiek et al., 2018)
17. • Primary direction of each of the three clusters (10 seconds):
https://www.jspsych.org/6.3/demos/jspsych-rdk-demo3.html
39. • M = 6.93
• Remember the order or the frequency of the numbers?
40. • M = 6.93
• Remember the order or the frequency of the numbers?
w 3 5 6 4 9
7 5 8 5 6
3 8 7 4 2 n
41. Consolidation and retention
• Sleep (reviewed by Jost et al., 2019)
• Spaced repetition (reviewed by Ullman & Lovelett, 2018)
• Retrieval practice (reviewed by Ullman & Lovelett, 2018)
• Incremental stimulus complexity (Poletiek et al., 2018)
• Simplication of the task
• Exogenous (e.g., by researchers)
• Endogenous (by oneself)
• Chunking
42. Chunking
‘A mechanism by which distributional regularities are used to form
discrete representations of an input’ (Isbilen et al., 2022)
43. Chunking
Smalle et al. (2018)
• Hebb-repetition task: ‘From a total of 36 different drawings, nine
drawings were associated with each set of three CV-pools. Of these nine
drawings, three drawings were associated with the pool used to
generate the filler sequences; another three drawings were associated
with the first Hebb sequence and the remaining three drawings were
associated with the second Hebb sequence.’
• Findings: children exhibited better retention than adults after 1 year.
• Limitations: no test of input complexity, no generalisation to novel items
44. Chunking
Jost et al. (2019): ‘Rapidly recoding and compressing information
by chunking may allow learners to more efficiently process input,
and to do so at higher levels of abstraction. In fact, stronger
learners may show a decreased reliance on surface-level
fragment information when tested due to the fact that they have
already used that information to internalize the higher-order
regularities, and no longer rely on them as a crutch.’
45. Jost et al. (2019)
• Retention: 2-week delay
• Hypothesis: overall retention
• Input complexity
• Hypothesis: greater complexity impaired learning
• Chunking
• Hypothesis: impaired learners less reliance on grammatical regularities
more reliance on fragment information, or chunk strength
• Smaller effect of chunk strength in Session 2
46. Jost et al. (2019)
• Implicit instruction: training input contained grammatical regularities
but participants’ attention was not explicitly drawn to the regularities.
• Simple training group (N = 23): 80% simple, 20% complex input
• Complex training group (N = 24): 80% complex, 20% simple input
54. Observations on the statistics
• Participants and items. No random slopes Violation of
independence of observations (Brauer & Curtin, 2018)
• Somewhat lax workflow of tests: number could have been smaller
• Correlations
• Calculation of p values was somewhat anti-conservative
(see Luke, 2017): namely, model comparison instead of Satterthwaite
or Kenward-Roger approximation for degrees of freedom
55. Conclusions
• Complexity of stimuli impaired retention
• Complex training group relied more on surface-level fragment
information (chunk strength) than simple training group