Short answer: A single successful recognition test does not establish that alcohol left every form of working memory unaffected. In a randomized, placebo-controlled laboratory experiment involving 90 social drinkers, the effect of alcohol differed depending on whether a person had to recognize an image shown moments earlier or correctly reject a previously unseen image. Among adults aged 55–70, the higher assigned alcohol condition was associated with lower efficiency in correctly rejecting novel images than placebo. Across both age groups, correct-rejection accuracy was also lower at the higher condition. But the results do not yield a validated test of an individual’s intoxication, testimonial accuracy, driving safety, or responsibility for an event.
The source is Boissoneault, Sklar, Prather and Nixon, “Acute Effects of Moderate Alcohol on Psychomotor, Set Shifting, and Working Memory Function in Older and Younger Social Drinkers” (Journal of Studies on Alcohol and Drugs, 2014). The central forensic lesson concerns task specificity: accurately identifying something previously seen and correctly identifying something as not previously seen are different response types. Their observed alcohol associations were not interchangeable.
What question did the experiment actually test?
The investigators compared age groups and assigned alcohol conditions on short laboratory tasks, not on eyewitness interviews or road driving. They wanted to know whether adults in an older age bracket responded differently to low and moderate alcohol conditions from adults in a younger bracket—and whether a memory result depended on speed, accuracy, or their combined measure of efficiency. They tested the distinction between a hit, an accurate recognition of a previously presented face or scene, and a correct rejection, an accurate indication that a probe image had not appeared in the cue set.
That design can reveal that one outcome changes while another does not. It cannot identify a universal impairment threshold. An age-by-dose statistical interaction means the study’s group response pattern depended on the assigned conditions; it does not mean the experiment can forecast how any named older or younger person would perform. It also does not establish that age alone caused the difference independently of every other characteristic.
Who participated, and how were the groups assigned?
The study enrolled 51 younger adults, ages 25–35, and 39 older adults, ages 55–70, for a total of 90 social drinkers. Conditions were allocated between participants rather than exposing the same individual to all three beverages on separate test days. The authors described a double-blind, placebo-controlled two-age-by-three-dose factorial design. The sample was screened for relevant medical, psychiatric, and substance-use histories; the resulting findings concern a selected group, not every adult in those age ranges.
| Age group | Placebo | Target 0.04 g/dL | Target 0.065 g/dL | Total |
|---|---|---|---|---|
| 25–35 years | 19 | 15 | 17 | 51 |
| 55–70 years | 12 | 13 | 14 | 39 |
| Both age groups | 31 | 28 | 31 | 90 |
Differences between groups matter. In particular, the researchers reported some dose-group variation in drinking history and older-group depressive symptom scores despite randomization. Their selected clinical screening and design reduce some sources of variation but do not remove every possible confound. The study also was not powered to establish credible interactions involving age, sex, and dose simultaneously.
Did researchers measure blood alcohol—or estimate a breath-based target?
The specified 0.04 and 0.065 g/dL values were assigned target breath-alcohol concentration levels in the study’s reporting convention. The amount of ethanol in each beverage was calculated with a sex-modified Widmark formula to reach the chosen target. The investigators then collected actual exhaled breath measurements using an Intoxilyzer 400PA, with checks at approximately 10, 25, 60, and 75 minutes after alcohol administration. These readings should not be relabeled as independently measured venous whole-blood gas-chromatography results.
To standardize immediate conditions, the study required four hours of fasting and provided a roughly 220-calorie snack around an hour before the beverage. Alcohol was mixed into sugar-free lemon-lime soda, and placebo drinks were presented with sensory cues intended to support blinding. Five participants, including two older adults, received a real booster when their breath reading was sufficiently below target; the remaining participants received a vehicle-only booster. Participants were tested once under their assigned condition.
The authors found the expected dose effect on breath measurements but did not find a significant difference in measured BrAC between age groups at the sampled time points. Within the active conditions, the breath readings were higher by the working-memory task than by the earlier Trail Making Test, which illustrates why time after dosing and task order belong in any interpretation. There was no paired serial venous whole-blood study, no independent biomarker of long-term drinking, and no event-time reconstruction.
How were hits and correct rejections measured?
The working-memory task presented four cue images—two faces and two scenes—followed by a delay and a probe. Participants had to remember either faces while ignoring scenes or scenes while ignoring faces. A probe was either previously present or new, with the source reporting a 50% probability of previous presentation per trial. The researchers measured both whether the response was correct and how long it took.
A hit is not a correct rejection. Calling an old face “old” can be correct, while calling a new face “new” requires a different discrimination. A person can show good recognition of repeated material and still be more error-prone when deciding whether unfamiliar material had appeared. The investigators separately analyzed the two response types after identifying a significant three-way age-by-dose-by-trial-type interaction, F(2,84) = 5.83, p = .004. That analysis decision is central to understanding the different patterns.
The paper calls efficiency a ratio built from accuracy and reaction time. It is a constructed performance index, not a chemically measured value, clinical diagnosis, or real-world reliability percentage. The passive-viewing control blocks were excluded from the primary behavioral analyses because accuracy approached ceiling. Any statement about “memory performance” therefore needs the named tested condition and outcome.
What happened when participants rejected previously unseen images?
The strongest specific result relevant to distinguishing new from old material was an age-group-by-dose interaction for correct-rejection efficiency: F(2,84) = 4.04, p = .02. In planned comparisons, older adults assigned the 0.065 g/dL target condition had lower efficiency than older placebo participants (p = .04). The lower target group did not differ significantly from either comparator in the combined correct-rejection efficiency analysis. Younger adults showed an opposite-direction comparison of higher efficiency at the higher condition, but the cited overall younger-group contrast was p = .07—not a statistically significant difference at .05.
Measured correct-rejection accuracy across both age groups also differed by dose, F(2,84) = 7.57, p = .001. The source reported group means of 94.8% for placebo, 93.2% for the 0.04 g/dL target, and 88.6% for the 0.065 g/dL target. Higher-dose accuracy was lower than both placebo (p = .001) and the lower target (p = .02). Those percentages describe laboratory responses grouped by assigned beverage, not the probability that any particular person’s memories, testimony, or observations are reliable.
The corresponding age-by-dose interaction in correct-rejection reaction time alone had p = .07. Thus the analysis does not support saying that alcohol definitively slowed all older participants’ correct responses or sped all younger participants’ responses. Performance efficiency incorporates both speed and accuracy, but a significant efficiency interaction should not be restated as a significant reaction-time interaction.
Did recognition of previously shown images show the same pattern?
No uniformly parallel finding was established. For hits—correct identification of previously displayed probes—the overall age-by-dose interaction in efficiency was F(2,84) = 2.53, p = .09. That is a trend reported by the authors, not a significant interaction at the conventional .05 threshold. A planned comparison among older participants assigned the 0.04 g/dL target produced higher hit efficiency than placebo (p = .03), and a related hit-accuracy comparison was reported at p = .05. However, the 0.04 versus 0.065 g/dL comparison had p = .07 for hit accuracy, and hit reaction time showed no dose-related effects in either age group.
It would therefore be misleading to interpret the isolated lower-dose comparison as evidence that moderate drinking generally improves memory in older adults. The main interaction was not significant; the design made multiple outcome-specific comparisons; the task was constrained to short face-and-scene cues; and no clinical or safety advantage was assessed. The results illustrate why favorable performance on one endpoint does not negate an unfavorable effect on another endpoint recorded in the same experiment.
Primary results at a glance
| Outcome and source measure | Original reported result | What follows—and what does not |
|---|---|---|
| Age × dose × probe type | F(2,84) = 5.83; p = .004 | Hit and correct-rejection findings warrant separate analysis; not a prediction tool. |
| Correct-rejection efficiency | Age × dose F(2,84) = 4.04; p = .02; older high target versus placebo p = .04 | Group-specific laboratory accuracy/time relationship; not a personal impairment diagnosis. |
| Correct-rejection accuracy | Placebo 94.8%; low target 93.2%; higher target 88.6%; dose F(2,84) = 7.57, p = .001 | Accuracy was worse at higher target across age groups, not a witness-error rate. |
| Correct-rejection reaction time | Age × dose F(2,84) = 2.78; p = .07 | No significant interaction for RT alone; do not call it proven slower. |
| Hit efficiency | Age × dose F(2,84) = 2.53; p = .09; selected older lower-target versus placebo p = .03 | The omnibus interaction was not significant, despite one positive planned contrast. |
| Trail Making Part B time | Age × dose F(2,89) = 2.51; p = .09; older high-target versus lower-target p = .01 | Suggestive task-specific difference; neither a significant omnibus interaction nor a driving outcome. |
Underlying primary source: Boissoneault et al. (2014), original Methods, Table 1, Results and original Figure 1–3 captions; DOI 10.15288/jsad.2014.75.870. A target in g/dL under the study’s BrAC convention must not be described as a measured laboratory whole-blood specimen.
What did the Trail Making Test add, and what did it not add?
The Trail Making Test was a separate timed paper-and-pencil task administered earlier after beverage consumption. Part A required connecting numbers in sequence; Part B alternated numbers and letters, adding a set-shifting requirement. In the original analysis, older adults took longer overall on both parts. For Part A, the investigators reported age and dose main effects but no age-by-dose interaction. The lower target was associated with shorter completion times than placebo in the pooled analysis—again not a universal enhancement claim.
For Part B, the age-by-dose interaction was p = .09 and the main dose effect was not significant. Planned comparisons found older participants in the higher target group were slower than those assigned the lower target, with a source-reported p = .01; younger groups did not show the same planned comparison. Those results support attention to a particular task and contrast. They do not show that any participant drove badly, could not follow instructions, or would make a mistaken identification outside the laboratory.
Importantly, the Trail Making Test was completed before the working-memory task, and the researchers found higher measured breath readings later. A comparison of the two tasks is also a comparison of different cognitive demands and different post-dose time points. It cannot be used as a simple within-person demonstration that concentration alone caused the task difference.
Why are a measured concentration, performance score, and legal inference different?
There are three different propositions. First, a breath instrument produced alcohol readings under the study’s breath reporting convention. Second, experimental tasks produced directly observed completion times and responses; some associations with dose and age were statistically detectable at the group level. Third, an opinion might attempt to infer from either measure that a particular individual could not remember an event, operate a vehicle, perceive a scene, or form a legally relevant mental state. The third proposition was not tested.
No venous whole-blood samples were analyzed by forensic headspace gas chromatography to calibrate a person’s blood concentration against a reported breath value. A breath target calculated before drinking is an administrative dosing target, not a specimen. A working-memory efficiency ratio is built from behavioral observations, not an ethanol assay. A result at a single post-dose interval does not back-calculate a reliable earlier event-time BAC. No prospective individual sensitivity/specificity threshold, diagnostic cutoff, or accident-causation algorithm was developed.
For a separate explanation of where an ethanol calculation acquires its assumptions, read the site’s published research brief on what two timed blood samples do—and do not—establish about alcohol elimination. That brief discusses a different problem. It should not be mistaken for evidence about the cognitive task in the present experiment.
How should an expert handle a witness who passed one memory task?
Passing one task can be informative about performance on that task, at that moment, under the stated conditions. It is not proof of intact perception, encoding, retention, later retrieval, source monitoring, resistance to suggestion, or truthfulness. Conversely, a poor rejection score in a group experiment cannot by itself discredit an actual witness. Neither inference follows automatically from a group mean or a statistical interaction.
Relevant forensic questions begin with the specific alleged failure: Was the issue noticing a detail, distinguishing a new person from a previously seen person, holding a sequence in mind, rapidly changing rules, or later recounting what happened? A test involving four brief images and a nine-second delay is not equivalent to an uncontrolled incident. Differences in lighting, stress, duration, divided attention, and post-event influences were not measured. Those variables must be established from case evidence rather than imputed from laboratory averages.
The study also cannot assign individual causation when it has no person’s baseline performance across counterbalanced alcohol and placebo sessions. It randomized participants between conditions; a named witness in another case was not randomized in this experiment. The strongest defensible general proposition is narrower: different memory operations can show different alcohol-related statistical patterns, and a single “normal” score need not represent all operations.
What limitations constrain the age comparison?
Each participant attended one assigned experimental beverage condition. The between-person comparison limits what can be said about a specific person’s change from sober to drinking. Group sizes were relatively small once divided into six age-by-dose cells; older placebo, lower, and higher target groups contained 12, 13, and 14 people respectively. An interaction at the group level does not establish sensitivity or specificity for identifying impairment in one person.
The authors also identified limits from uncertain previous meals outside the four-hour fasting window, sleep and other unmeasured risk factors, variation in medication use, and a task that imposed only a modest memory load. There were too few older men in some cells for a meaningful age-by-sex-by-dose analysis. Because participants did not give confidence ratings about “old” or “new” judgments, the investigators could not perform a comprehensive signal-detection analysis. The paper did not perform a driving simulator, field-sobriety validation, courtroom recall test, or real-life eyewitness identification.
Multiple planned comparisons and outcomes should be reported with their actual p values rather than flattened into an all-or-none story. The selected lower-dose hit contrast, the nonsignificant hit interaction, the significant correct-rejection efficiency interaction, the significant correct-rejection accuracy dose effect, and the nonsignificant response-time interaction are all part of the result. Discarding the inconvenient null findings would distort the interpretation.
What records would be needed for a case-specific opinion?
A case-specific assessment would require a reliable timeline of the alleged event, any alcohol consumption, the exact collection method and time of any specimen, the laboratory or device reporting units, quality-control records, medications and relevant physiological history, and the actual tasks or conduct observed. If a memory issue is disputed, identify what was noticed, whether it was previously encountered, how soon it was tested, whether the person was answering “old” or “new,” and how the testing context differs from the case.
A group experiment cannot replace contemporaneous observations or the person’s own medical and analytical records. Likewise, a legal definition of intoxication is not supplied by a working-memory test. The study may inform scientifically disciplined questions for an expert to examine, but it does not decide guilt, negligence, fitness, or whether a particular recollection is accurate. Further topic-based reading paths are available in the site’s research library, with attribution and scope limitations explained in its editorial and research-source policy.
Frequently asked questions
Does a correct recognition response prove that alcohol did not affect memory?
No. A correct recognition of a previously presented image is a different outcome from correctly rejecting an image that was absent. The source reported different statistical patterns for these two response types. No one successful response rules out every other cognitive deficit.
Did 0.065 g/dL represent a measured venous whole-blood BAC?
No. It was the assigned target under the experiment’s BrAC reporting convention, with ethanol dose estimated using a formula. Breath was measured repeatedly, but the experiment did not establish a paired individual venous whole-blood gas-chromatography concentration.
Did low-dose alcohol significantly improve all older participants’ memory?
No. One selected lower-target-versus-placebo hit-efficiency comparison among older adults had p = .03, but the overall hit age-by-dose interaction had p = .09. There was no general cognitive-improvement finding or tested clinical benefit.
What was the strongest finding about previously unseen images?
Correct-rejection accuracy across age groups was lower in the higher target condition, with a reported dose effect F(2,84) = 7.57, p = .001. There was also an age-by-dose correct-rejection efficiency interaction F(2,84) = 4.04, p = .02. Those are task-specific group results.
Were older participants significantly slower on correct rejection because of alcohol?
The age-by-dose interaction for correct-rejection reaction time alone was p = .07, not significant at .05. It is improper to replace the significant efficiency interaction with an assertion that reaction time had a significant interaction.
Can the study estimate an individual driver’s impairment or crash responsibility?
No. The study tested laboratory cognitive responses, not actual driving, collisions, diagnostic classification, or accident mechanisms. A person-specific opinion would need independent relevant evidence and cannot be calculated from these group averages.
Can this study establish whether an eyewitness is telling the truth?
No. The study did not measure deception, credibility, or real eyewitness identification. Its primary relevance is to the need to distinguish the particular memory operation being tested and the assumptions needed to extend any result to a person.
Original study and scope
Boissoneault J, Sklar A, Prather R, Nixon SJ. Acute effects of moderate alcohol on psychomotor, set shifting, and working memory function in older and younger social drinkers. Journal of Studies on Alcohol and Drugs. 2014;75(5):870–879. DOI: 10.15288/jsad.2014.75.870. This is an educational analysis of one experimental primary study, not an opinion about an identified person, and it does not establish a legal or medical decision rule.