General slowing in language impairment: methodological considerations in testing the hypothesis

Area of Science:

• Psycholinguistics
• Developmental Psychology
• Cognitive Neuroscience

Background:

• The general slowing hypothesis posits that individuals with language impairment (LI) exhibit slower processing speeds across various cognitive tasks.
• Conventional statistical methods for testing this hypothesis, such as ordinary least squares regression (OLS), have faced scrutiny regarding their efficacy and interpretation.
• Response time (RT) data from diverse cognitive and language tasks provide a basis for examining processing speed differences between LI and typically developing (chronological-age-matched, CA) groups.

Purpose of the Study:

• To compare the effectiveness of ordinary least squares regression (OLS) and hierarchical linear modeling with random coefficients (HLM) in analyzing response time (RT) data to test the general slowing hypothesis in language impairment (LI).
• To investigate whether LI is characterized by general slowing across studies or by study-specific variations in processing speed.
• To assess the extent and significance of processing speed differences between LI and chronological-age-matched (CA) groups using advanced statistical modeling.

Main Methods:

• Analysis of response time (RT) data from 25 studies encompassing 20 different perceptual-motor, cognitive, and language tasks.
• Comparison of results obtained from ordinary least squares regression (OLS) and hierarchical linear modeling with random coefficients (HLM).
• Separate HLM analysis focused on specific language tasks, including picture naming and word recognition.

Main Results:

• Ordinary least squares regression (OLS) supported the general slowing hypothesis, indicating LI groups were approximately 10% slower than CA groups.
• Hierarchical linear modeling (HLM) revealed a larger average slowing of 18% but with significantly greater variability, rendering the overall slowing statistically non-significant.
• HLM identified significant study-specific slowing, suggesting processing speed differences vary across contexts rather than being a general deficit in LI. Analysis of language tasks showed minimal (2%) and non-significant slowing.

Conclusions:

• Hierarchical linear modeling (HLM) provides a more nuanced understanding of processing speed in language impairment (LI), highlighting study-specific slowing over general slowing.
• The findings challenge the conventional interpretation of response time (RT) data using ordinary least squares regression (OLS) for the general slowing hypothesis.
• Methodological limitations in analyzing response time (RT) data for assessing general slowing in LI populations are emphasized, advocating for advanced statistical approaches.