Delayed Response Tasks for Working Memory
Rodriguez JS, Paule MG. Chapter 12: Working Memory Delayed Response Tasks in Monkeys. In: Buccafusco JJ, editor. Methods of Behavior Analysis in Neuroscience. 2nd ed. Boca Raton (FL): CRC Press/Taylor & Francis; 2009.
Paule MG, Bushnell PJ, Maurissen JP, Wenger GR, Buccafusco JJ, Chelonis JJ, Elliott R. Symposium overview: the use of delayed matching-to-sample procedures in studies of short-term memory in animals and humans. Neurotoxicol Teratol. 1998;20(5):493-502. doi:10.1016/s0892-0362(98)00013-0
Reindl E, Seed AM, Barton RA, Francis-Costa T, Kendal RL. Humans may not have a uniquely enhanced sequence memory: sequence discrimination is facilitated by causal–logical framing in humans and chimpanzees. R Soc Open Sci. 2025;12(7):250236. doi:10.1098/rsos.250236
Lind J, Enquist M, Ghirlanda S. Animal memory: A review of delayed matching-to-sample data. Behav Processes. 2015;117:90–96. doi:10.1016/j.beproc.2014.11.019
Borchert K. Delayed Matching to Sample Task (DMTS) — Technical Manual. Millisecond Software, LLC; 2015 (updated 2025). millisecond.com/library/v7/dmts/dmts/dmts.manual
Overview
Delayed response (DR) tasks are the primary behavioural paradigm for assessing working memory — the ability to hold information “online” over short periods when that information varies unpredictably in time or content. This is the type of memory devastated in Alzheimer’s disease and other dementias.
The basic structure: present a sample stimulus → withdraw it → impose a recall delay → present the sample alongside a comparison stimulus → subject must identify which was shown previously. This cycle repeats across trials with varying delays.
DR tasks are used across species — nonhuman primates, rodents, and humans — making them powerful translational tools for studying the neurobiology of memory and screening pharmacological interventions.
Key Task Variants
Delayed Matching-to-Sample (DMTS)
The subject must choose the stimulus that matches the previously presented sample. Used extensively in the NCTR Operant Test Battery with macaques. Sensitive to cholinergic, noradrenergic, and serotonergic manipulations. Performance depends on prefrontal and inferior temporal cortices.
Delayed Non-Matching-to-Sample (DNMTS)
The subject must choose the stimulus that does not match the sample. Monkeys learn DNMTS faster than DMTS due to natural novelty preference. Relies on a circuit including ventromedial/ventrolateral prefrontal cortex, rhinal cortex, thalamus, and inferotemporal cortex. Trial-unique stimuli engage the medial temporal lobe more; repetitive stimuli engage prefrontal and parietal cortices more.
Delayed Alternation
A two-choice task where correct responses alternate between trials (spatial: left/right; object: which object). Reveals distinct cortico-striatal circuits for spatial vs. nonspatial working memory.
Titrating Matching-to-Sample
Task difficulty adapts to subject performance by adjusting delay length, allowing individualized difficulty levels. Useful for equating baseline performance across subjects.
Human Applications
DMTS procedures have been applied across the human lifespan and clinical spectrum (Paule et al., 1998):
- Children: Encoding and retention processes improve rapidly in early years, plateauing before puberty
- Aging: Normative data in elderly populations; DMTS performance correlates significantly with IQ
- Clinical populations: Used to study Parkinsonism, Alzheimer’s disease, schizophrenia, and depression
- Imaging: PET studies link DMTS performance to activity in prefrontal cortex, confirming roles for these structures in mnemonic processes
- CANTAB: The Cambridge Neuropsychological Test Automated Battery includes DMTS tasks validated in neurosurgical patients and those with basal ganglia disorders
Neural Circuitry
- Prefrontal cortex (especially dorsolateral): critical for spatial working memory, response inhibition, and decision-making. Goldman-Rakic demonstrated this through lesions, electrophysiology, and imaging.
- Hippocampal formation / medial temporal lobe: important for memory consolidation, associative memory, declarative memory, and recognition memory. Role in DNMTS delay component is debated.
- Inferior temporal cortex: damage interferes with DMTS for visual stimuli
- Caudate nucleus: spatial delayed alternation activates the head (prefrontal projections); object alternation activates the body (temporal projections)
Neurotransmitter Systems
| System | Key Findings |
|---|---|
| Noradrenaline | Guanfacine (α2A agonist) and clonidine improve DR performance in young and aged monkeys |
| Acetylcholine | Nicotinic agonists (ABT-089, ABT-418) and cholinesterase inhibitors (velnacrine, physostigmine) enhance DMTS/DNMTS accuracy |
| Serotonin | 5-HT2A and 5-HT3 antagonists improve accuracy; LSD (5-HT partial agonist) degrades attention/encoding without affecting forgetting rate |
| Dopamine | D1 receptors in prefrontal cortex critical for spatial working memory; antagonists impair delayed alternation |
| GABA | Diazepam decreases DMTS accuracy; FG 7142 (GABA-A inverse agonist) impairs performance via dopamine mechanisms |
| Opioids | Morphine decreases reinforcers earned and PTC, increases latencies |
| Cannabinoids | Δ9-THC and marijuana smoke increase response latencies and decrease task completion |
Analytical Framework
Data are typically plotted as response accuracy vs. recall delay:
- Y-intercept: reflects attention, stimulus discrimination, and encoding
- Slope: rate of short-term memory decay (forgetting)
- Delay-dependent effects: treatment changes slope (rate of forgetting) — suggests memory-specific effects
- Delay-independent effects: treatment shifts intercept — suggests attentional or encoding effects
Common Confounds
- Position bias: subjects may develop side preferences in two-choice paradigms
- Ceiling/floor effects: task difficulty must be calibrated (adjust delays, stimulus complexity)
- Motor confounds: response latency measures help disentangle memory effects from motor or motivational effects
Comparative Meta-Analysis (Lind et al., 2015)
A meta-analysis of over 90 DMTS data sets from 25 species (birds, mammals, and bees) estimated two parameters from performance-vs-delay curves:
- Zero-delay performance (intercept): reflects attention, encoding, and discrimination. Little evidence of species differences — most animals achieve near-perfect accuracy at zero delay.
- Performance half-life (decay rate): related to memory span. Performance half-lives range from a few seconds to several minutes. Half-lives over 1 minute have been observed only in mammals. Primates do not stand out from other mammals.
Key findings:
- Pigeons perform worse than mammals at longer delays, but are the only extensively studied bird — other birds may bridge longer delays
- Extensive training may improve memory span (titration procedures yielded half-lives up to ~7 min in macaques), but improvements could also reflect learned coping with lower signal-to-noise ratios rather than genuine memory enhancement
- Humans can achieve errorless performance after 48 h delays (Overman & Doty, 1980), far exceeding other species
- Specialized vs. generalist memory: long-term animal memories (days to months, e.g. food caching in corvids) likely reflect specialized memory systems for ecologically significant information, not a general-purpose memory capacity. Generalist working memory for arbitrary stimuli decays rapidly in most species.
Inquisit Implementation (Borchert, 2015/2025)
A ready-to-run DMTS script is available for the Inquisit platform (Millisecond Software), implementing the procedure based on Paule et al. (1998). Designed for animal research but adaptable for human testing.
Task Parameters
| Parameter | Default | Description |
|---|---|---|
| minStimulusDuration | 5000 ms | Minimum sample presentation time |
| delay1 | 0 ms | Zero delay (tests encoding/attention) |
| delay2 | 1000 ms | Short delay |
| delay3 | 4000 ms | Medium delay |
| delay4 | 20000 ms | Long delay (tests memory decay) |
| iti | 5000 ms | Inter-trial interval |
Trial Sequence
- Sample stimulus (colored disk: red, green, or yellow) presented for minimum 5 s — subject must press to advance
- Delay interval (0, 1, 4, or 20 s)
- Two comparison disks presented — subject selects matching color
- ITI (5 s)
Design
- 3 target colors × 2 comparison colors × 2 locations × 4 delays = 48 trials per block
- 2 blocks per session (96 trials total, ~40 min duration)
- Each color presented equally often as target; each color paired equally often; target position balanced left/right
Data Output
- Summary file (
dmts_summary*.iqdat): proportion correct and mean RT per delay condition, overall accuracy - Raw trial file (
dmts_raw*.iqdat): per-trial latency, correctness, delay condition, target/comparison colors, response
The script can be modified for delayed non-matching-to-sample (DNMTS) by editing the trial.DMTS_match element. Delays can be individualised using titration procedures.
Related Concepts
- working-memory — theoretical framework
- Alzheimer’s disease — cholinergic degeneration and memory impairment
- prefrontal-cortex — central structure for executive function and working memory
- hippocampus — role in consolidation and recognition memory
- CANTAB — touchscreen-based cognitive testing battery
- sequence-discrimination-learning — comparative memory for ordered sequences