9.3.8 - How psychological knowledge changes

Models develop through tests

A psychological model changes when researchers find that its predictions no longer explain the available evidence adequately.

Cognitive psychology illustrates refinement rather than simple replacement. The multi-store model describes distinct sensory, short-term and long-term stores, with attention and rehearsal contributing to information transfer. Its short-term store has limited capacity and duration. It organised questions about capacity, duration and coding into a testable account. However, describing short-term memory as a single store leaves a problem: how can people retain some information while also using information to reason?

Baddeley and Hitch's 1974 working-memory account treated temporary memory as an active system with differentiated functions and limited resources. Concurrent-task experiments examined how retaining information affected reasoning and other activities. Performance could be disrupted without being completely abolished, motivating an account in which tasks draw on partly distinct resources. This expanded the questions being asked from 'where is information stored?' to 'how is it maintained and used?'

Baddeley's 2000 episodic buffer proposal addressed how different kinds of information might be integrated into coherent episodes and connected with long-term memory. It was an addition to a developing account, not a component that can be retroactively attributed to the original 1974 model. A newer component should make explanations or predictions more adequate; adding a label without testable consequences would not itself be progress.

Older findings may retain value after a model changes. Acoustic or semantic interference can still reveal features of memory even if the theory of stores is revised. Case studies of selective memory impairment also challenge a simple unitary account, though an individual case cannot reveal every normal memory process. Experimental and case-study evidence offer different constraints that an adequate model must address.

A reasoned comparison asks what the later account explains better, what evidence prompted it and which uncertainties remain. 'Newer' is a date, not a methodological strength. 'More detailed' helps only if the additional detail is supported and improves explanatory or predictive power.

Replication under changed ethical conditions

Ethical development can improve participant protection while changing which historical comparisons researchers can make directly.

Milgram's obedience programme examined responses to an authority instructing participants to administer what they believed were increasingly severe shocks. It made situational pressure central to the explanation of obedience, but deception and participant distress generated lasting ethical criticism. The scientific importance of a question does not make every historical procedure acceptable to repeat.

Burger's 2009 report describes a partial replication with 70 adults, stopping after the decision at 150 volts. Screening, repeated withdrawal reminders and prompt debriefing were safeguards. Continued obedience at that point supported the persistence of situational pressure. The study did not observe what its participants would do at 450 volts, so a claim about that endpoint is an inference, not a replicated finding.

This comparison contains two developments: a later test of an older phenomenon and a method modified to reduce risk. However, time period, recruitment and safeguards also changed. A difference in results cannot automatically be attributed to a historical change in society. Similarity does not prove that every underlying motive was the same either.

Replication repeats important features of a study to test whether a finding recurs. Exact duplication is often impossible or undesirable. Researchers should state which features were retained, which changed and how the changes affect comparison. A restricted replication can provide valuable evidence while answering a narrower question than its predecessor.

Social theories also change through re-interpretation. Obedience can be examined in terms of perceived legitimacy, responsibility, identification and situational support for dissent. Competing accounts must explain specific behaviour and variation, rather than being judged only by which historical researcher proposed them. Access to fuller records can reveal whether a neat textbook description omitted ambiguity, but an allegation about a classic study is not automatically an established fact.

The balanced judgement is that ethical constraints and scientific improvement can occur together. Protection may narrow direct comparison, yet transparent reporting of that limit is stronger evidence than claiming an ethically modified method is identical to the original.