A diffusion model explanation of the worst performance rule for reaction time and IQ
Section snippets
IQ and the diffusion model
To model how IQ affects RT, we examined how IQ might be related to one or more processing components of the diffusion model. If IQ were related to drift rate, an obvious possibility, people with higher IQ's would extract higher quality evidence from stimuli, memory, and so on. If IQ were related to boundary separation, a perhaps less likely possibility, people with higher IQ might adjust their decision criteria to require less evidence before making a decision.
To examine these two
Drift rate and the worst performance rule
Table 1 shows results for nine sets of simulated data (1000 subjects per set). For the first three, the across trial variability parameters of the model (η, sz, and st) were set to near 0.0 (0.001). In row 1 of the table, when there was no across subject variability in boundary separation (sa = 0) or the nondecision components (sTer = 0), then drift rate correlated strongly positively with accuracy, and strongly negatively with mean RT and the five RT quantiles (the 0.1, 0.3, 0.5, 0.7, and
Discussion
The results in Table 1, Table 2 show that the diffusion model generates data consistent with the worst performance rule if IQ is correlated with drift rate or if IQ is correlated with boundary separation, but only when other components of processing vary across subjects.
The two possibilities can be discriminated because the correlations between drift rate and data pattern differently than the correlations between boundary separation and data. For drift rate, the correlation with accuracy is
Conclusions
The worst performance rule refers to the empirical finding that when IQ is correlated with RT, the correlation is larger for the slower than the faster responses. The simulations presented in this article show that the diffusion model predicts the worst performance rule if IQ is manifested in drift rate or if it is manifested in boundary separation. The key for the success of the model is variability in the values of components of processing across subjects. The simulations that demonstrated
Acknowledgement
Preparation of this article was supported by NIA grant R01-AG17083 and NIMH grant R37-MH44640. Florian Schmiedek is now at Humboldt-Univeristy, Berlin.
References (31)
IQ, the worst performance rule, and Spearman's law: A reanalysis and extension
Intelligence
(2003)- et al.
Serial versus parallel processing in visual search tasks and IQ
Personality and Individual Differences
(1993) - et al.
Meta-analysis of the relationship between intelligence and inspection time
Intelligence
(2001) A test of Larson and Alderton's (1990) worst performance rule of reaction time variability
Personality and Individual Differences
(1992)- et al.
Inspection time and intelligence: A meta-analysis
Intelligence
(1989) - et al.
Reaction time variability and intelligence: A “worst performance” analysis of individual differences
Intelligence
(1990) - et al.
A diffusion model analysis of the effects of aging on recognition memory
Journal of Memory and Language
(2004) Relations of successive percentiles of reaction time distributions to cognitive variables and adult age
Intelligence
(1998)- et al.
The accumulator model of two-choice discrimination
Journal of Mathematical Psychology
(1988) - et al.
The rationale for the inspection time index
Personality and Individual Differences
(1986)
Distributions of reaction times of retardates and normals
American Journal of Mental Deficiency
Looking down on human intelligence: From psychometrics to the brain
Reaction time and psychometric g
A recruitment theory of simple behavior
Psychometrika
Response times
Cited by (60)
A randomness perspective on intelligence processes
2022, IntelligencePost-error slowing is associated with intelligence
2021, IntelligenceTo predict the future, consider the past: Revisiting Carroll (1993) as a guide to the future of intelligence research
2021, IntelligenceCitation Excerpt :Other models require the control of response times (Goldhammer, 2015) which implies lengthy testing. Diffusion models (Ratcliff, Schmiedek, & McKoon, 2008) stress the rate of information accumulation and its relation with working memory (Schmiedek, Oberauer, Wilhelm, Süß, & Wittmann, 2007) but simplified versions of this model deliver three person-parameters that should be jointly considered to understand performance. Evaluating performance on seemingly simple tasks therefore requires considering sophisticated models of response behavior (Kyllonen & Zu, 2016), which is rarely done in intelligence research and hardly ever in applied settings.
A meta-analysis of the worst performance rule
2019, IntelligenceCitation Excerpt :For the drift diffusion model account of the worst performance rule, on the other hand, the implications are more straightforward, as it would not be challenged by a negatively accelerated trend of the worst performance rule across quantiles of the reaction time distribution. In their simulation study, Ratcliff et al. (2008) demonstrated that the shape of the relationship between cognitive abilities and reaction times across quantiles depended on the trial-to-trial variability of two diffusion model parameters, boundary separation and non-decision time. Although certain specifications of trial-to-trial variability in these parameters may predict a logarithmic trend of the worst performance rule, it may be questioned whether the drift diffusion model account qualifies as a theoretical explanation of the worst performance rule if it can account for any trend of the relationship between cognitive abilities and reaction times across quantiles.