# Cognitive psychology: test a processing model without overclaiming a mechanism — Study worksheet

[Study guides](https://unireadinglist.com/guides/cognitive-models/)

[Course topic: Cognitive Psychology](https://unireadinglist.com/subjects/psychology/#course-cognitive-psychology)

Sources checked: 2026-11-04

Study route and exercise: University Reading List editorial project.

For learners who can read a two-factor table and subtract means. The free computational companion needs additional neural-network and quantitative preparation; the commercial cognitive texts provide broader experimental context. All times below are invented exact means, not participant data or a significance test.

## Choose your starting point

- [ ] I need an experimental cognitive spine
  Use Reisberg to connect methods and evidence to perception, attention and memory. Record both the proposed explanation and the comparison that could distinguish it.

- [ ] I want a wide critical reference
  Use Eysenck and Keane for competing accounts across cognitive domains. Select a small cluster of explanations rather than collecting every named effect.

- [ ] I want to build a computational account
  Use the free computational neuroscience text after basic network preparation. A simulation should specify mechanisms and outputs, not turn a verbal stage label into a neural fact.

## Compare the three books

### Computational Cognitive Neuroscience

Randall C. O’Reilly; Yuko Munakata; Michael J. Frank; Thomas E. Hazy; Contributors · 5 · 2024
Free computational companion

Choose it when: You want explicit network mechanisms and simulations to complement behavioral explanations.
Use another route when: You need a gentle first survey of cognitive experiments or cannot yet follow activation, learning and network representations.
Selected reading: Fifth edition, 2024: introductory computational foundations and the relevant perception or cognitive-system simulation, using PDF/EPUB when HTML equations render poorly.
Limits and trade-offs: The CC BY 4.0 text is freely available, but a model’s behavior depends on its implementation and assumptions. It is not the same kind of experimental survey as the other two books.

#### Try before choosing

Write a mapping from one model variable to an observable response measure, and name an output that could challenge that mapping.
Check your work: A hidden unit or stage label is not automatically a measured brain structure or uniquely identified psychological process.

Your notes:

________________________________

[Official free reading](https://book.compcogneuro.org/)

[Book record and access options](https://unireadinglist.com/subjects/psychology/computational-cognitive-neuroscience-oreilly/)

### Cognition: Exploring the Science of the Mind

Daniel Reisberg · 9 · 2025
Evidence-led cognitive survey

Choose it when: You need an accessible sequence from methods to perception, attention and memory.
Use another route when: You want complete mathematical derivations or a free full-book route from a limited publisher demo.
Selected reading: Selected ninth edition: methods and neural basis first, then the relevant perception, attention or working-memory discussion by topic.
Limits and trade-offs: The 2025 print record and its publisher demo are distinct objects. A limited demo does not make every chapter freely available.

#### Try before choosing

For the four-cell worksheet, separate the observed means from the claim that clarity affects only encoding.
Check your work: Selective influence is a model assumption to defend; it is not visible just by naming the experimental factors.

Your notes:

________________________________

ISBN: 978-1-324-11478-9

[Book record and access options](https://unireadinglist.com/subjects/psychology/cognition-exploring-the-science-of-the-mind/)

### Cognitive Psychology: A Student's Handbook

Michael W. Eysenck; Mark T. Keane · 8 · 2020
Broad competing-accounts reference

Choose it when: You want to compare explanations and empirical limits across cognitive domains.
Use another route when: You are trying to read the entire reference before forming one testable contrast.
Selected reading: Eighth edition, 2020: research approaches and selected attention, perception or memory accounts, matched to the question.
Limits and trade-offs: The broad survey supports comparison but needs active selection. Its eighth-edition ISBN must not be attached to a later revision or a neighboring catalogue title.

#### Try before choosing

Construct two different nonnegative stage decompositions that fit the same four means.
Check your work: Equal predicted totals can conceal different latent stage durations; goodness of fit alone does not identify every parameter.

Your notes:

________________________________

ISBN: 9781138482234

[Book record and access options](https://unireadinglist.com/subjects/psychology/cognitive-psychology-a-students-handbook/)

## Check your preparation

### What is observed?

Your notes:

________________________________

### What is selective influence?

Your notes:

________________________________

### What does additivity imply?

Your notes:

________________________________

## A six-step study plan

- [ ] Separate measurement and account

Selected reading: Reisberg methods discussion.

What to do: Label the four observable means and the unobserved stage parameters.

Check your work: Do not report a fitted latent duration as directly measured.

Your notes:

________________________________

- [ ] Calculate simple effects

Selected reading: Selected experimental comparison.

What to do: Compute degradation costs separately for direct and reversed mapping.

Check your work: Both effects use the same millisecond unit and order of subtraction.

Your notes:

________________________________

- [ ] Fit a candidate account

Selected reading: The original worksheet.

What to do: Choose nonnegative E, S and M values that reproduce every mean.

Check your work: Check all four predictions, not just the baseline.

Your notes:

________________________________

- [ ] Test identifiability

Selected reading: Eysenck and Keane competing accounts.

What to do: Find a second, different decomposition with identical predictions.

Check your work: Distinguish fitting the total from identifying its components.

Your notes:

________________________________

- [ ] Change one cell

Selected reading: The solved follow-up.

What to do: Calculate the interaction and show the additive constraint it violates.

Check your work: Reject the specified model without claiming that a unique alternative has been proved.

Your notes:

________________________________

- [ ] Connect to a simulation

Selected reading: The free computational companion.

What to do: Specify a hypothetical simulation output and a competing explanation for the same response pattern.

Check your work: A behavioral pattern, implementation and neural interpretation require separate evidence.

Your notes:

________________________________

## Worked exercise: Original response-time table: additivity does not identify stage lengths

Invented exact mean response times are: clear/direct 400 ms, degraded/direct 460 ms, clear/reversed 450 ms, degraded/reversed 510 ms. Test the selective-influence additive serial model RT=E(clarity)+S(mapping)+M. Compute factor effects and give two different nonnegative stage decompositions.

Assumptions: Clarity has two levels and affects only E; mapping has two levels and affects only S; M is constant. Durations add. These are candidate model restrictions. The table contains exact artificial means, with no sampling noise, participant data or neural measurements.

Your notes:

________________________________

## Change an assumption: Changed cell: a model constraint fails without identifying its replacement

Change only degraded/reversed from 510 to 550 ms. Keep the other three exact means. Can the same selectively influenced additive model fit all four cells? What does the failure establish?

Your notes:

________________________________

## Before moving on

- I can compute both sets of simple effects and the interaction.
- I can exhibit and check two observationally equivalent decompositions.
- I can state exactly which assumptions a changed pattern refutes.

## Answer and interpretation

### What is observed?

Only four exact mean response times in milliseconds. No distributions, errors, individual differences or sample sizes are supplied.

### What is selective influence?

Here clarity changes only encoding duration E, mapping changes only selection duration S, and the constant motor component M does not change. This is stipulated for testing, not established by the table.

### What does additivity imply?

Under the specified additive model, the effect of one factor is the same at both levels of the other. Observing that property is compatible with the model but does not uniquely prove its architecture.

### Original response-time table: additivity does not identify stage lengths

1. Degradation costs 460−400=60 ms for direct mapping and 510−450=60 ms for reversed mapping. Reversing costs 450−400=50 ms for clear stimuli and 510−460=50 ms for degraded stimuli.
2. The difference of differences is (510−450)−(460−400)=0 ms. The means satisfy this additive model’s no-interaction constraint.
3. One fit is E(clear)=150, E(degraded)=210, S(direct)=100, S(reversed)=150 and M=150 ms. The four sums are 400,460,450,510 ms.
4. A second fit is E(clear)=200, E(degraded)=260, S(direct)=50, S(reversed)=100 and M=150 ms. It predicts the same four means with different encoding and selection levels.
5. Thus the stage-level decomposition is not identified by these means. The 60 and 50 ms contrasts are fixed within the specified model, but additivity alone does not uniquely establish serial stages or a neural mechanism.

The table has a 60 ms degradation effect, a 50 ms mapping effect and zero interaction. At least two distinct latent decompositions fit exactly; absolute stage durations are not identified.

Test the result: Substitute both parameter sets into every cell. Do not infer a unique architecture from compatibility or treat exact synthetic means as statistically established population effects.

### Change an assumption: Changed cell: a model constraint fails without identifying its replacement

1. The direct-mapping degradation cost remains 60 ms, while the reversed-mapping cost becomes 550−450=100 ms. Their difference is 40 ms.
2. Equivalently, the mapping cost is 50 ms when clear and 550−460=90 ms when degraded. The interaction is again 40 ms.
3. An additive fit to the first three cells must predict the fourth as 460+450−400=510 ms, not 550 ms. No choice of E, S and constant M under the stated restrictions can eliminate that contradiction.
4. Reject this particular additive selective-influence account for the exact table. The pattern does not uniquely prove parallel processing, a brain location or one alternative model; selective influence, stage dependence or strategy assumptions might differ.

The changed table has a 40 ms interaction and cannot fit the specified additive model. It rules out that joint set of restrictions, not every serial account or all but one alternative.

Test the result: Reconstruct the fourth-cell prediction from the other three. With sampled human data, variability, design and inference would also be needed; none are supplied here.

## Common mistakes

### Model label treated as observation

A named processing stage is latent, not directly measured by a total response time.

### Fit mistaken for identification

Different parameter sets or architectures can share observable predictions.

### Interaction overinterpreted

Failure of one additive model does not uniquely establish a neural mechanism or a competing architecture.

## Sources and edition checks

- [Authors: free 2024 fifth edition and simulations](https://book.compcogneuro.org/)
  The open fifth edition supplies the contrasting computational route and simulation focus.

- [Norton: selected ninth-edition record](https://wwnorton.com/books/9781324114789)
  The exact ninth-edition record anchors the experimental survey.

- [Norton: ninth-edition limited reader demo and contents](https://nerd.wwnorton.com/nerd/364146/r/goto/cfi/210%21/4?control=control-assignment&demo=true)
  The publisher’s limited demo supports selected introductory scope, not a free-full-text claim.

- [Routledge: selected eighth-edition book](https://www.routledge.com/Cognitive-Psychology-A-Students-Handbook/Eysenck-Keane/p/book/9781138482234)
  The selected eighth-edition identity is retained.

- [Routledge:2020 eighth-edition identity and scope](https://www.routledge.com/rsc/downloads/Psychology_Spring_2025_-_GBP.pdf)
  The publisher catalogue corroborates year, ISBN and broad cognitive scope; the response-time packet is original.

These are proposed study activities, not official course assignments or a claim that every book has been read cover to cover.

[How we curate](https://unireadinglist.com/editorial/)