Method at a glance
- Modality: behavioural (visuospatial and object working memory); no physiological recording
- Measures: The capacity to hold a cue in mind across a filled delay — the spatial (‘where’) domain in delayed response, the object (‘what’) domain in delayed non-matching to sample. In the somatic marker literature its job is negative: to establish that a patient’s decision-making deficit is not a working-memory deficit.
- Procedure: Adapted from the non-human primate delay paradigms (Goldman-Rakic; Fuster) with a distracter added, because the animal versions are too easy for humans. Delayed response: four cards appear on screen for 2 s, two face down and two face up showing red or black; the cards vanish for 1, 10, 30 or 60 s and reappear all face down; the subject selects the two that had been face up. Delayed non-matching to sample: one face-up red or black card appears for 2 s, vanishes for 1/10/30/60 s, then four cards appear (two red, two black) and the subject selects the two opposite in colour to the sample. During every delay the subject reads aloud a series of semantically meaningless sentences, to block rehearsal and load working memory. Scored as percent correct at the 10, 30 and 60 s delays; Bechara et al.’s impairment cutoff is ≤80% correct at 60 s, below which no control ever performed.
- Also called: delay task, delayed response, delayed non-matching to sample, DNMS, working memory delay task
Strengths
- Purpose-built to be the control condition the Iowa Gambling Task needs. Run in the same session on the same patients, it converts ‘VM patients decide badly’ into ‘VM patients decide badly with working memory intact’, which is the step the somatic marker argument requires and which conventional neuropsychological batteries (IQ, language, attention, Wisconsin Card Sorting) do not supply as directly.
- The verbal distracter is the design’s good part. Without it, subjects rehearse the cue verbally and the task stops taxing the visuospatial and object systems it was imported to test — which is also why left-dorsolateral patients show no impairment on it (the verbal route is already blocked for everyone).
- Two domains crossed against one lesion set. Running spatial and object versions together covers the inferior/superior dorsolateral division reported in primates (Goldman-Rakic 1987, 1992; Wilson et al. 1993) and in humans (Courtney et al. 1996), so a lesion spanning both regions cannot escape detection by falling in the untested domain.
- Delivers a clean anatomical double dissociation against the gambling task: anterior VM damage impairs deciding with delay performance intact; right dorsolateral damage impairs delay performance with gambling performance advantageous (if low-normal). See Bechara et al. (2000).
Limitations
- The dissociation it delivers is asymmetric, and the asymmetry is usually dropped in citation. Working memory can be normal or superior alongside a severe decision deficit — but decision-making is worse when working memory is impaired. Right-DL patients scored in the low-normal range on the gambling task, not the normal range. So the task shows the two functions are separable, not that decision-making is independent of working memory, and the framework itself concedes markers work partly by allocating attention and working memory.
- Results are averaged across the two procedures in the source that reports them (Bechara et al. 2000), on the stated grounds that the lesions were not restricted to inferior or superior dorsolateral regions. So the spatial/object distinction the two-task design exists to honour is collapsed before analysis, and the paper’s own figures cannot be used to check it. The disaggregated results are in Bechara et al. (1998), which is not in
raw/.- Small, unequal, and unreported samples. Nine VM and ten dorsolateral patients against 21 controls in the source experiment, with figures comparing subgroups of four. No statistics accompany any delay-task comparison in the 2000 review — the double dissociation is asserted from bar charts.
- The impairment cutoff is defined by the control sample it is applied to. ‘≤80% correct at 60 s, a cutoff below which no normal control ever performed’ is a criterion derived from the same controls the patients are compared against, which is circular if used to claim a specific rate of impairment rather than merely to sort individuals.
- Only the right-hemisphere group was impaired. All left-DL patients had normal working memory. The authors attribute this to the verbal distracter having already removed the verbal strategy, consistent with right-lateralized dorsolateral activation in the imaging literature — a plausible explanation, but it means the task’s sensitivity is hemisphere-dependent and a left-lateralized working-memory deficit would be invisible to it.
- It measures nothing bodily. No physiological recording, no interoceptive measure, no somatic state. Its entire relevance to this wiki is as the instrument that rules a rival explanation partly out — and ‘partly’ is the operative word.
Connections
- Key studies: Bechara et al. (2000)
- Related methods: Iowa Gambling Task, Skin conductance response (SCR), Pavlovian defense conditioning
Imported into the somatic marker programme from the non-human primate working-memory literature, for one purpose: to show that the VM patients who fail the Iowa Gambling Task are not failing it because they cannot hold information in mind.
Enters this wiki with Bechara, Damasio & Damasio (2000), where the result is reported in full. The wiki had carried it only as a line on Antoine Bechara (“dissociated working memory from decision-making within human prefrontal cortex”).
Why it exists
The Somatic marker hypothesis needs the VM patient to be specifically impaired. If the same lesion degraded working memory, the gambling deficit would be explicable without any appeal to bodily signals — you cannot weigh long-run consequences you cannot hold onto. Conventional testing already showed normal IQ, language and attention, but “working memory” as measured in the primate tradition is a sharper instrument aimed at the neighbouring cortex, and the dorsolateral sector is precisely where the frontal-lobe literature had located it (Goldman-Rakic; Fuster; Milner and Petrides).
So the design is a crossing: two frontal lesion groups, two tasks, one prediction each.
| delay tasks | gambling task | |
|---|---|---|
| bilateral anterior VM (n = 9) | normal — some superior | severely impaired |
| right dorsolateral (n = 10, right-lesioned subset) | impaired | low-normal |
| left dorsolateral | normal | — |
The asymmetry, which is the part worth keeping
Cited as a double dissociation, and anatomically it is one. Functionally it is not symmetric, and the source says so plainly: “working memory and decision making were asymmetrically dependent.”
- Working memory does not need decision-making. A VM patient can be at floor on the gambling task and at ceiling on the delay tasks.
- Decision-making partly needs working memory. Right-DL patients with severe delay-task impairment did not choose badly — they chose advantageously, but their gambling performance “falls in the low normal range.”
This matters for how the wiki reads a specific criticism. Iowa Gambling Task lists working memory as the first of five rival explanations of the VM deficit catalogued by Dunn et al. (2006). The Iowa group had already measured it, five years earlier, and found it a partial contributor rather than a substitute explanation. What the design does not do — and what would be needed to retire the rival — is quantify how much of the VM deficit survives after working memory is accounted for. Nine patients, no statistics, no covariate analysis.
It also sits with the second Dunn revision on Ventromedial prefrontal cortex (VM): IGT performance is worse with damage extending into DLPFC (Bechara et al. 1998b — the same study), correlates with DLPFC resting-state activity, and large PFC lesions impair the task where discrete ones may not. The framework absorbs this by saying markers work partly through attention and working-memory allocation, at the cost Dunn et al. name: it “does make the predictions of the SMH difficult to distinguish from other, non-somatic theories.” The delay tasks are where that cost was first incurred, by the framework’s own authors, reporting it as a finding.
The anatomical byproduct
The more durable result may be the one about lesion extent rather than about working memory. All nine VM patients failed the gambling task, but they split on the delay tasks — and the split tracked anatomy:
- Abnormal gambling / abnormal delay (n = 5): lesions extending posteriorly, plausibly into the basal forebrain.
- Abnormal gambling / normal delay (n = 4): lesions confined anteriorly, sparing it.
Posterior extension toward basal forebrain is the same variable that predicts the other co-morbid deficits in this material — Ventromedial prefrontal cortex (VM) records Tranel et al. (1996), where patients with lesions reaching ACC and basal forebrain fail Pavlovian conditioning as well as gambling, and the 2000 paper separately notes that VM patients with basal-forebrain involvement do show motor impulsiveness and perseveration where anterior-only patients do not.
So “the VM patient” is at least two patients, and which one you have is legible from the posterior margin of the lesion. That is a caution about every group-level VM result in this wiki, and it is available from the Iowa material itself.