Why Did Methylphenidate Improve Math Productivity Without Better Working Memory? A 63-Child Crossover Study

Short answer: In a randomized, double-blind, two-period medication crossover involving 63 children with ADHD, methylphenidate was associated with improvement in mathematical schoolwork previously reported from this cohort, yet the investigators detected no statistically significant medication benefit on the specific working-memory, reaction-time, attention-lapse, interference-control or intrinsic-motivation measures tested in their follow-up analysis. A parent-reported rating of the child’s perceived academic competence increased on medication and statistically accounted for part of the relationship between treatment and math productivity. That mediation model is not proof of a biological mechanism; the trial did not measure blood methylphenidate concentrations or forensic impairment.

The original primary paper is Kortekaas-Rijlaarsdam and colleagues, Methylphenidate-Related Improvements in Math Performance Cannot Be Explained by Better Cognitive Functioning or Higher Academic Motivation: Evidence From a Randomized Controlled Trial, Journal of Attention Disorders 24(13):1824–1835 (2020, first published online in 2017), DOI 10.1177/1087054717713640. Its 12-page complete Methods, Results, two data tables, mediation analysis and discussion form the sole primary evidentiary source for the numerical findings below.

What question did the investigators actually test?

The practical question was not simply whether a prescribed stimulant can change classroom output. Previous work with this research cohort had already identified a treatment-associated increase in mathematical productivity and accuracy. This analysis asked why: would changes in executive functions, sustained attention, academic motivation or confidence-like self-perceptions explain the schoolwork outcome? The distinction matters. A child can complete more exercises without registering a proportional improvement on a short laboratory measure of digit span, visual memory, button-press timing or a questionnaire about motivation.

The investigators prespecified candidate mediating variables and examined within-child treatment differences and associations with schoolwork. Thus, three empirical steps must remain separate: whether methylphenidate changed a proposed intermediate variable, whether that variable was associated with mathematics under the regression model, and whether including it attenuated the treatment coefficient. Calling all three steps a demonstrated cause would exceed the design.

Who was studied, and which observations were independent?

The medication crossover ultimately included 63 children with diagnosed ADHD between eight and 13 years old. Sixty-five had been recruited; two did not complete the trial because of adverse events described as unrelated to the intervention. Separately, 67 typically developing children were recruited for between-group descriptive comparisons and were tested once. Accordingly, 130 children contributed to the overall comparison, but the randomized medication contrast concerns the same 63 ADHD children in both periods. The 67 comparison children were not a second randomized drug arm.

Repeating cognitive trials, collecting ratings from parents and teachers, or recording two treatment weeks does not multiply the number of independent participants. Missing observations also differed by endpoint: some math and executive tasks had isolated missing data, and a small number of parent or teacher questionnaires were unavailable. The authors described their analytic method as generalized linear mixed models with a participant random intercept to accommodate repeated within-person measurements.

How was methylphenidate administered?

Children with ADHD were already receiving clinically prescribed methylphenidate after individualized titration. The blinded experimental comparison used extended-release oral methylphenidate (Equasym XL) and visually matched placebo. Each treatment period lasted seven days; the order was randomized. At least 48 hours of medication washout was required before the first period and between weeks. On the final day of a treatment week, children were evaluated approximately 60–90 minutes after taking the assigned capsule, with assessment lasting about two hours.

The daily prescribed doses varied: 10 mg for 27%, 20 mg for 44%, 30 mg for 24%, and 40 mg for 5% of the ADHD participants. These figures are dose-allocation percentages, not observed plasma concentrations and not four separate randomized dose-finding arms. The investigators used clinically titrated equivalent doses when the child’s regular formulation differed. Equasym XL was described as having both immediate-release and delayed-release components; nominal prescribed dose therefore cannot be equated with instantaneous central nervous system exposure.

What was physically measured—and what was not?

The measured outcomes were school task performance, computer-task behavior, and questionnaire ratings. Verbal working memory was assessed by reverse digit span; visuospatial working memory required reproducing reversed sequences of locations in a grid. A modified Flanker task provided reaction-time speed, a modeled tail of slow reactions interpreted as lapses of attention, and incongruent-versus-congruent interference. Intrinsic motivation was measured with the Children’s Academic Intrinsic Motivation Inventory. Parents rated sensitivity to reward and punishment and perceived academic competence; children and teachers contributed separate competence ratings.

No collected specimen was described as being analyzed quantitatively for methylphenidate or its metabolite ritalinic acid. Specifically, the paper provides no directly assayed legal venous whole-blood concentration, no centrifuged plasma or serum drug result, and no quantitative urine, oral-fluid or hair toxicology curve linked to the cognitive tasks. The paper did not evaluate a chemically validated dosing-time algorithm, driving, an actual traffic event, aviation performance or a person-specific statutory impairment classification.

Which cognitive results were genuinely nonsignificant?

The original source reported no statistically significant beneficial active-versus-placebo effects on reverse digit span, visuospatial working memory, reaction-time speed, modeled attention lapses, or interference control. For example, the reported mean visuospatial memory index was 54.95 (SD 33.86) during placebo and 57.66 (SD 37.27) during medication, but the tested medication effect was nonsignificant. Modeled attention lapses decreased numerically from 150.33 (SD 74.72) to 135.57 (SD 88.06), yet the comparison was only marginal, indicated by p = .08, not evidence meeting the authors’ conventional threshold.

The abstract reports the nonsignificant cognition comparisons spanning p = .082–.641. One should not translate such results into proof that methylphenidate has no cognitive effects in anyone, on any task, or at any dose. A null result in one battery means the trial did not demonstrate an effect there with its particular age range, titrated regimens, period lengths, sample size and measurement methods. It can coexist with positive effects on an outcome not captured by the same task.

Did intrinsic academic motivation improve?

Not detectably in the questionnaires used here. General intrinsic motivation averaged 60.05 (SD 8.47) on placebo and 60.13 (SD 9.91) on active treatment; math motivation averaged 76.34 (SD 20.70) and 77.37 (SD 20.39), respectively. Reading motivation changed from 80.08 (SD 19.99) to 81.33 (SD 19.85). None of those comparisons was statistically significant. Parent-reported sensitivity to reward and punishment also did not show significant medication effects.

That finding is narrower than a categorical claim that students were no more willing to participate, concentrate or work. Questionnaire constructs, brief test behavior and sustained classroom engagement are not interchangeable. The researchers’ motivation scales tested particular forms of self-reported intrinsic interest and parent-rated reward sensitivity. They were not direct, moment-by-moment observations of voluntary effort on every math problem.

What did the key original measurements show?

These are means from different scales, so a difference of 1 point on a parent-rating scale cannot be treated as equivalent to a 1-point cognitive-task change. Nor does the table establish direct pharmacokinetic exposure: all numerical entries concern behavioral or rating scores. The parent-rated competence result is the clearest medication-associated change in this mediation analysis, but the opinions of different raters did not move in lockstep.

Why did parent ratings differ from child and teacher ratings?

Parents rated the child’s perceived academic competence higher during active treatment, whereas the child and teacher ratings did not show a significant drug difference. The study’s authors considered sensitivity to real-world behavioral changes and response bias as possible explanations. Parents observed the child outside school and could notice differences in engagement or confidence. They also might infer a child’s perceived competence from better symptom control or productivity. These are plausible interpretations, not separately randomized causal mechanisms.

The specific rating instrument asked each respondent to evaluate perceptions of competence; it was not a direct objective test of the child’s latent self-belief. Describing the finding as “methylphenidate objectively raised self-esteem” would distort the evidence. The proper formulation is that parents’ ratings of their child’s perceived academic competence were higher under medication than placebo in this trial, whereas other informants did not confirm a significant difference.

How did the mathematical productivity mediation model work?

A previously reported mathematics outcome from these same children improved under methylphenidate. In the present paper, mathematical productivity—not math accuracy—was statistically related to parent-rated competence after adjustment for treatment. When parent-rated competence was inserted into the productivity model, the medication coefficient decreased. A multilevel mediation calculation yielded a positive indirect-effect interval that excluded zero. That pattern is compatible with partial statistical mediation, but it does not rule out parent-report bias, unmeasured paths or a common underlying improvement in symptoms.

Critically, the paper does not establish that an external intervention which changes a parent’s rating, without affecting anything else, would cause more correct or completed mathematics exercises. Nor is a regression coefficient a chemical concentration. The following table retains the actual statistical scale and comparison rather than converting coefficients to invented percent improvement.

What were the actual mediation coefficients?

The distinction between productivity and accuracy is pivotal. Completing a larger share of available exercises and answering a greater fraction correctly are different endpoints. Here, evidence for the proposed parent-competence path attached to productivity. It did not establish equivalent mediation of accuracy. If an expert report claims this experiment proved broad improvement in “mathematical intelligence,” the actual outcome definitions and the partial mediation findings do not support that generalization.

Could expectancy or measurement method explain the pattern?

Possibly, though the design itself cannot isolate one explanation. The trial was double blind and used matched capsules, reducing obvious treatment labeling. Nevertheless, repeated parent observations and rating scales are vulnerable to changes in parental impressions. The authors reported an association between parent-rated competence and parent-rated symptom improvement, and found the former retained predictive association with math productivity when both were included. This addresses one measured overlap, not all sources of common-method bias.

Computer tasks and rating scales may also differ in reliability and in what they capture. A brief Flanker measure of very slow responses is not an exhaustive measure of how readily a child starts homework. Reverse digit span does not represent all working-memory operations needed during schoolwork. Stronger school output can be produced by changes in persistence, opportunity, rule-following or response initiation even if an experimental task shows no detectable shift. These are hypotheses for further testing, not independent new observations of the present trial.

Why does seven days of treatment matter?

Each active or placebo period lasted only one week. The observed changes therefore address relatively short-term behavior in a previously titrated group of children, not graduation rates, future occupational performance or cognitive development. The investigators noted that participants had an established history of clinical stimulant treatment; the mean reported duration of treatment before enrollment was about 30.7 months. A within-child crossover can control many stable personal differences while still leaving questions about longer-term carryover, withdrawal, expectancy and variable dose history.

The study describes at least 48 hours of washout before and between periods, but did not report a pharmacokinetic sample series demonstrating a zero methylphenidate or ritalinic-acid concentration on each crossover baseline. Without direct analyte assays, those periods should be described as protocol washouts rather than individually confirmed chemical clearance intervals. This distinction becomes especially important when someone tries to translate a clinical regimen into a forensic timeline.

What does the typically developing comparison add?

The separate group of 67 typically developing children provides context for baseline functioning. On the placebo condition, children with ADHD performed worse on visuospatial working memory and showed more modeled attention lapses than that comparison group. They also scored lower on intrinsic motivation toward mathematics and on perceived academic competence, with parental and teacher reports supporting competence-group differences. By contrast, the study did not find significant ADHD-versus-control differences on every selected task; reverse digit span, reaction-time speed and interference control were among the null between-group comparisons.

Those contrasts are descriptive, not evidence that the 67 controls received methylphenidate or that medication moved the ADHD group to an identical normative score. Group classification and clinical treatment were not randomized between ADHD and typically developing children. Baseline differences in estimated IQ, history and referral can complicate any causal comparison across those two populations.

Why the drug concentration and specimen distinction is essential

In clinical studies, an assigned oral dose is an exposure instruction. A laboratory toxicology result, by contrast, requires a precisely identified physical specimen and validated analytic procedure. Methylphenidate measured in venous whole blood would not automatically have the same concentration as methylphenidate in separated plasma or serum; urinary analytes and metabolite ratios answer still different questions. Oral-fluid screening cannot be substituted for a quantified blood concentration or an observed behavioral endpoint without suitable validation. This study did not collect and validate the paired measurements required for any of those comparisons.

Readers who need foundational definitions can consult the existing published same-site guides on biological matrices, confirmatory chemical testing, and therapeutic drug monitoring. The links provide contextual terminology only; none is a substitute for the primary trial’s actual measured variables.

Can this paper answer individual legal or expert-witness questions?

It can inform a limited question: in a child ADHD cohort already receiving clinical treatment, increased math output under oral methylphenidate did not require a detectable change in the selected computer-based cognitive indices or self-reported motivation, and a parent-rated competence variable partially mediated productivity in the authors’ statistical framework. This is useful in assessing whether an expert has conflated academic output, a neuropsychological test score and a biological drug assay.

The study cannot determine whether a particular student had medication present in whole blood at a specified time, was intoxicated, could safely operate a vehicle, falsely reported attention difficulties, or caused an event. It also cannot certify that medication caused a precise improvement in one child’s grades. Any individualized claim would require its own clinical, analytical, temporal and factual foundations and attention to the limits of the relevant test.

Where are the major uncertainties?

Several uncertainties are material rather than footnotes. The sample was relatively small and clinically selected; assigned doses differed by prior titration; each period was short; some tasks had missing observations; and no untreated pre-enrollment baseline or directly measured drug-concentration time course was presented. Parent ratings may be related to medication-responsive behaviors other than a child’s underlying self-perception. The selected cognition and motivation tasks did not cover every possible mechanism of school success. Statistical mediation is sensitive to model specification, measurement reliability and common causes of the mediator and outcome.

Finally, the mathematical schoolwork benefit referenced in this paper derives from an earlier analysis of the same participants. It should not be counted as independent replication. The correct evidentiary statement is that the follow-up analysis tested potential explanatory pathways for a previously reported treatment effect within one research program, not that it established a new external cohort confirming the same result.

Frequently asked forensic-science questions

1. Does the study establish that methylphenidate does not improve working memory?

No. It found no statistically significant treatment effect on the specific reverse-digit-span and visuospatial working-memory indices under the tested conditions. Different doses, populations, duration and tasks can produce different results. This trial cannot demonstrate a universal absence of working-memory effects.

2. Were 130 children randomized to methylphenidate and placebo?

No. The crossover randomized the treatment order for 63 children with ADHD. The additional 67 typically developing children were assessed once for contextual comparisons; they were not a drug-randomized trial arm.

3. Does a positive parent competence rating prove the child felt more confident?

No. It documents a higher parent-rated perception during medication. The child’s own and teachers’ scales did not show a significant corresponding difference. That divergence limits claims about an objectively demonstrated private mental state.

4. Did parent-rated competence explain every medication-related mathematics outcome?

No. The significant statistical mediation concerned math productivity, not accuracy. The treatment coefficient for productivity also remained significant after adding the rating variable, indicating at most partial mediation under the model.

5. What methylphenidate blood concentration caused the observed productivity change?

The study cannot answer that. It assigned oral clinical doses but did not chemically quantify parent methylphenidate in venous whole blood, separated plasma or serum linked to the performance times. No specimen-based concentration-response threshold was established.

6. Can improved productivity be treated as proof a child was not impaired?

No. Finishing more math exercises is neither an individualized intoxication assay nor an actual driving task. Clinical educational performance and forensic impairment are different measurement targets, and a case-specific inference needs independent evidence.

7. What is the soundest statement for a scientific report?

Specify the 63-child ADHD crossover; seven-day clinically dosed oral methylphenidate versus placebo; nonsignificant selected cognition and motivation outcomes; significant parent-rated competence and partial math-productivity mediation; lack of chemical drug assay; and the prohibition on extrapolating these aggregate findings to an individual’s legal or incident status.

Primary research and scope of interpretation

The sole original numerical source is Kortekaas-Rijlaarsdam A-F, Luman M, Sonuga-Barke E, Bet P, and Oosterlaan J, Methylphenidate-Related Improvements in Math Performance Cannot Be Explained by Better Cognitive Functioning or Higher Academic Motivation: Evidence From a Randomized Controlled Trial, Journal of Attention Disorders 24(13):1824–1835. The complete primary study is available through DOI 10.1177/1087054717713640. This article is independent critical scientific analysis; it does not report an additional clinical trial, laboratory blood assay, legal finding, or personalized medical advice.