感覺變強不等於真的變強:主觀清晰度、元認知信心與客觀表現
英文題名: Feeling Enhanced Is Not the Same as Being Enhanced: Subjective Clarity, Metacognitive Confidence, and Objective Performance in ADHD系列: ADHD 動態配置與認知拓撲系列,第 6 篇版本: v0.1日期: 2026-08-16作者: Neo.K(許筌崴)協作: GPT-5.6 Sol文件性質: 理論建模/認知科學命題/研究綱領文獻檢索截點: 2026-08-16
0. 醫學、藥理與證據邊界聲明
本文不是臨床研究、診斷工具、治療指南、藥理試驗或醫療建議。
本文提出的「主觀清晰度」「元認知信心」「性能估計誤差」「主觀—客觀分離」「校準漂移」「狀態增益」等概念均屬待驗證理論命題,不代表已被醫學界確認的 ADHD 病理機制。
原作者並非醫學、精神醫學、藥理學、神經科學或臨床心理專業研究者。本文不提供新的臨床、人體、神經影像、藥理或心理實驗數據;所有實證性背景均來自公開同行評審研究。
本文不使用原作者的個人用藥經驗作為一般化證據,也不提供任何自行開始、停止、增加、減少、混合或改變藥物使用方式的建議。
本文尤其不主張:
subjective enhancement = objective cognitive enhancement . \text{subjective enhancement}
=
\text{objective cognitive enhancement}. subjective enhancement = objective cognitive enhancement .
亦不主張:
confidence = accuracy . \text{confidence}
=
\text{accuracy}. confidence = accuracy .
核心限制:
felt state ≠ measured capacity ≠ task outcome . \boxed{
\text{felt state}
\neq
\text{measured capacity}
\neq
\text{task outcome}.
} felt state = measured capacity = task outcome .
摘要
ADHD 研究與 stimulant research 中經常同時出現三種不同輸出:主體感受到自己更清醒、更投入、更有精神或更「像是能做事」;主體對自己的答題、記憶或工作表現給出較高信心;而客觀任務表現則可能改善、不變,甚至在不同認知域呈現不同方向。如果把這三者都壓縮成「藥有效」或「注意力提升」,便會遺失重要的中介層。
2026 年 Current Psychology 對 140 名大學生進行研究,其中 70 名具有正式 ADHD 診斷。研究使用閱讀理解與 Raven’s Progressive Matrices,並在每題後收集 confidence ratings。ADHD 組在兩類任務中都呈現較低 performance accuracy 與較高 confidence,形成較大的 metacognitive bias;inattention 與 bias 的關聯尤其明顯,且 task modality 會調節偏差大小。另一方面,既有成人 ADHD 研究顯示 metacognition 並非所有認知域都一致受損;某些 memory-judgment 情境下,ADHD 成人仍可產生相對準確的 judgement of learning。2025 年成人 ADHD 評估研究又發現 subjective cognitive complaints 與 objective neurocognitive deficits 並不簡單對齊,支持 subjective complaint、objective performance 與 performance validity 必須分開。
Stimulant 文獻亦支持 domain specificity。2025 年一項 healthy-adult randomized double-blind placebo-controlled crossover study 發現 10 mg methylphenidate 只在 numeric working memory 等有限指標上出現改善,其他 cognition 與 visual-scanning metrics 未見廣泛提升。2026 年 Psychopharmacology 的 double-blind experimental-exam study 更發現,20 mg methylphenidate 不論在學習階段或考試階段給予,都沒有改善 factual、inferential 或 overall exam performance。與此同時,較早的 balanced-placebo 與 expectancy 研究顯示,期待收到 prescription stimulant 可以明顯改變主觀 arousal、drug effects 與 perceived enhancement,即使 objective cognition 並沒有相同幅度的變化。
基於此,本文提出「Subjective–Metacognitive–Objective Separation Hypothesis, SMOSH」。其核心不是宣稱 ADHD 必然過度自信,也不是宣稱 stimulant 只造成 placebo,而是將三層輸出分開:
Q t = subjective state quality , Q_t
=
\text{subjective state quality}, Q t = subjective state quality ,
P ^ t = metacognitive performance estimate , \widehat P_t
=
\text{metacognitive performance estimate}, P t = metacognitive performance estimate ,
P t = objective performance . P_t
=
\text{objective performance}. P t = objective performance .
並定義:
E t meta = P ^ t − P t . E_t^{\text{meta}}
=
\widehat P_t-P_t. E t meta = P t − P t .
若:
E t meta > 0 , E_t^{\text{meta}}>0, E t meta > 0 ,
表示 overestimation;
若:
E t meta < 0 , E_t^{\text{meta}}<0, E t meta < 0 ,
表示 underestimation。
本文再加入 calibration、resolution、state gain 與 transfer efficiency,提出「感覺變強」可以是真實的 phenomenological state change,同時不代表該變化已轉化成任務能力。真正的研究問題應是:
Which state changes transfer into which objective abilities, for whom, under which tasks? \boxed{
\text{Which state changes transfer into which objective abilities,
for whom, under which tasks?}
} Which state changes transfer into which objective abilities, for whom, under which tasks?
本文最後提出十一項可證偽命題與 balanced-expectancy、drug-blindness、confidence-performance calibration、cross-domain transfer、real-world exam、longitudinal within-person 等實驗綱領。
關鍵詞: ADHD、metacognition、confidence、subjective cognitive complaints、methylphenidate、expectancy、placebo、cognitive enhancement、calibration、objective performance
1. 問題:為什麼「我感覺比較強」不是一個簡單命題?
主體說:
我現在比較清醒。
和:
我現在比較專心。
以及:
我現在一定會表現比較好。
是三個不同命題。
本文首先表示:
Q t = subjective state quality . Q_t
=
\text{subjective state quality}. Q t = subjective state quality .
例如:
清晰;
鮮明;
清醒;
有動力;
有精神;
易於投入;
感覺思考順暢。
接著定義:
P ^ t = subjective estimate of performance . \widehat P_t
=
\text{subjective estimate of performance}. P t = subjective estimate of performance .
例如:
我覺得這題答對了;
我覺得我記住了;
我覺得今天考得很好;
我覺得反應更快;
我覺得遊戲操作變強。
最後:
P t = objective task performance . P_t
=
\text{objective task performance}. P t = objective task performance .
三者不能直接相等:
Q t ≠ P ^ t ≠ P t . \boxed{
Q_t
\neq
\widehat P_t
\neq
P_t.
} Q t = P t = P t .
2. 主觀狀態可以真的改變
本文不把所有「感覺變好」都稱為幻覺。
如果藥物、睡眠、運動、咖啡因、期待或環境改變造成:
Q t + 1 > Q t , Q_{t+1}>Q_t, Q t + 1 > Q t ,
這本身就是一個真實的主觀狀態變化。
例如:
arousal ↑ , \text{arousal}\uparrow, arousal ↑ ,
motivation ↑ , \text{motivation}\uparrow, motivation ↑ ,
perceived clarity ↑ . \text{perceived clarity}\uparrow. perceived clarity ↑ .
錯誤只發生在進一步推出:
Q t ↑ ⇒ P t ↑ . Q_t\uparrow
\Rightarrow
P_t\uparrow. Q t ↑⇒ P t ↑ .
這個箭頭需要獨立驗證。
3. 第一個分離:Subjective State 不等於 Confidence
可能:
Q t ↑ Q_t\uparrow Q t ↑
但:
P ^ t \widehat P_t P t
不變。
例如一個人感覺比較清醒,卻仍知道某題很難。
反過來:
Q t ≈ constant Q_t\approx\text{constant} Q t ≈ constant
也可能有:
P ^ t ↑ \widehat P_t\uparrow P t ↑
例如:
熟悉題型;
收到正向回饋;
被告知自己能力很高;
預期藥物有效。
所以:
state feeling ≠ self-evaluation . \boxed{
\text{state feeling}
\neq
\text{self-evaluation}.
} state feeling = self-evaluation .
4. 第二個分離:Confidence 不等於 Accuracy
對 trial i i i ,令:
c i ∈ [ 0 , 1 ] c_i\in[0,1] c i ∈ [ 0 , 1 ]
表示信心,
y i ∈ { 0 , 1 } y_i\in\{0,1\} y i ∈ { 0 , 1 }
表示是否答對。
可以:
c i = 0.95 c_i=0.95 c i = 0.95
但:
y i = 0. y_i=0. y i = 0.
也可以:
c i = 0.40 c_i=0.40 c i = 0.40
但:
y i = 1. y_i=1. y i = 1.
因此 confidence 的研究不能只看平均值。
需要測:
calibration \text{calibration} calibration
與:
resolution . \text{resolution}. resolution .
5. Calibration
最簡單的 calibration bias:
B cal = 1 n ∑ i = 1 n ( c i − y i ) . B_{\text{cal}}
=
\frac1n
\sum_{i=1}^{n}
\left(
c_i-y_i
\right). B cal = n 1 i = 1 ∑ n ( c i − y i ) .
若:
B cal > 0 , B_{\text{cal}}>0, B cal > 0 ,
代表整體 overconfidence。
若:
B cal < 0 , B_{\text{cal}}<0, B cal < 0 ,
代表 underconfidence。
完美平均校準:
B cal ≈ 0. B_{\text{cal}}\approx0. B cal ≈ 0.
但:
B cal ≈ 0 B_{\text{cal}}\approx0 B cal ≈ 0
不保證 trial-by-trial 判斷精準。
6. Absolute Calibration Error
定義:
E abs = 1 n ∑ i ∣ c i − y i ∣ . E_{\text{abs}}
=
\frac1n
\sum_i
|c_i-y_i|. E abs = n 1 i ∑ ∣ c i − y i ∣.
此量避免正負誤差互相抵消。
因此:
B cal ≠ E abs . \boxed{
B_{\text{cal}}
\neq
E_{\text{abs}}.
} B cal = E abs .
一個人可以平均沒有偏差,但每一題都非常不準。
7. Brier Score
可使用:
B S = 1 n ∑ i = 1 n ( c i − y i ) 2 . BS
=
\frac1n
\sum_{i=1}^{n}
(c_i-y_i)^2. B S = n 1 i = 1 ∑ n ( c i − y i ) 2 .
較低:
B S BS B S
表示 probabilistic confidence 與實際結果更一致。
這比只問:
你覺得自己考得好不好?
更精確。
8. Resolution
Metacognitive resolution 問的是:
這個人能否區分自己哪題對、哪題錯?
定義:
R meta = Corr ( c i , y i ) . R_{\text{meta}}
=
\operatorname{Corr}
\left(
c_i,
y_i
\right). R meta = Corr ( c i , y i ) .
或使用更適合 binary outcomes 的 meta-d' 等 signal-detection measures。
因此:
calibration ≠ resolution . \boxed{
\text{calibration}
\neq
\text{resolution}.
} calibration = resolution .
一個人可以整體偏高估,但仍然知道相對而言哪些題比較可能答對。
9. 2026 大學生 ADHD 研究:較低 performance+較高 confidence
2026 年一項 Current Psychology 研究納入:
N = 140 N=140 N = 140
名大學生,其中:
N ADHD = 70 , N_{\text{ADHD}}=70, N ADHD = 70 ,
N control = 70. N_{\text{control}}=70. N control = 70.
ADHD 組需有 psychiatrist/neurologist 的正式診斷文件與大學 disability-service registration。
研究使用:
reading comprehension;
Raven's Progressive Matrices;
並在每題後收集 confidence ratings。
研究結果顯示 ADHD 組在兩類任務中:
P ADHD < P control , P_{\text{ADHD}}
<
P_{\text{control}}, P ADHD < P control ,
同時:
C ADHD > C control , C_{\text{ADHD}}
>
C_{\text{control}}, C ADHD > C control ,
形成較大的 monitoring bias。
這提供非常直接的:
P ^ ≠ P \boxed{
\widehat P
\neq
P
} P = P
實證背景。
10. Task Modality Matters
同一研究顯示 metacognitive bias 會受到 verbal/non-verbal task 特性影響。
因此本文不採:
B cal = stable ADHD constant . B_{\text{cal}}
=
\text{stable ADHD constant}. B cal = stable ADHD constant .
更合理的是:
B cal = F ( person , task , difficulty , feedback , state ) . B_{\text{cal}}
=
F
\left(
\text{person},
\text{task},
\text{difficulty},
\text{feedback},
\text{state}
\right). B cal = F ( person , task , difficulty , feedback , state ) .
所以:
metacognitive bias is context-sensitive . \boxed{
\text{metacognitive bias is context-sensitive}.
} metacognitive bias is context-sensitive .
11. ADHD 不等於普遍「不知道自己表現如何」
較早成人 ADHD metacognition 研究顯示,不同 cognitive domains 的 self-awareness 並不一致。
例如:
attention metacognition \text{attention metacognition} attention metacognition
可能受損,
但:
memory metacognition \text{memory metacognition} memory metacognition
在部分設計中可以較完整。
因此:
ADHD metacognitive deficit ≠ global metacognitive blindness . \boxed{
\text{ADHD metacognitive deficit}
\neq
\text{global metacognitive blindness}.
} ADHD metacognitive deficit = global metacognitive blindness .
第 9 節所見 overconfidence 不能直接推廣至所有 ADHD 個體與所有任務。
12. Subjective Cognitive Complaints 不是 Objective Deficit 的替代物
2025 年對 adult ADHD referrals 的研究比較:
subjective cognitive complaints;
objective neurocognitive performance;
performance validity;
symptom validity。
研究結論指出 subjective cognitive complaints 與 objective neurocognitive deficits 並不簡單對齊。
因此:
C complaint ≠ D objective . \boxed{
C_{\text{complaint}}
\neq
D_{\text{objective}}.
} C complaint = D objective .
這並不表示主觀困難是假的。
它表示:
subjective burden \text{subjective burden} subjective burden
與:
laboratory performance \text{laboratory performance} laboratory performance
是不同 measurement domains。
13. 主觀抱怨可能比實驗室測驗更接近日常功能,也可能受到其他狀態影響
主觀 cognitive complaint 可以整合:
長時間真實生活經驗;
睡眠;
壓力;
emotion;
task demand;
accumulated failure;
self-concept。
因此:
C complaint = F ( P daily , E , S , M , H ) . C_{\text{complaint}}
=
F
\left(
P_{\text{daily}},
E,
S,
M,
H
\right). C complaint = F ( P daily , E , S , M , H ) .
而單次 neuropsychological test:
P lab = G ( task , session , motivation , state ) . P_{\text{lab}}
=
G
\left(
\text{task},
\text{session},
\text{motivation},
\text{state}
\right). P lab = G ( task , session , motivation , state ) .
所以:
C complaint ≠ P lab C_{\text{complaint}}
\neq
P_{\text{lab}} C complaint = P lab
不必然代表其中一個量無效。
14. Performance Validity 是必要的第三變量
若測驗資料包含:
V perf V_{\text{perf}} V perf
表示 performance validity,
則:
P observed P_{\text{observed}} P observed
不能直接視為:
P true . P_{\text{true}}. P true .
同樣,
V symptom V_{\text{symptom}} V symptom
也影響 self-report interpretation。
因此:
subjective–objective discrepancy ≠ pure metacognitive bias by default . \boxed{
\text{subjective–objective discrepancy}
\neq
\text{pure metacognitive bias by default}.
} subjective–objective discrepancy = pure metacognitive bias by default .
未來研究需控制 measurement validity。
15. Stimulant 效應首先可能是一個 State Change
第 2 篇已提出:
N t → Z t , \mathbf N_t
\rightarrow
\mathbf Z_t, N t → Z t ,
其中:
Z t = ( A , R , S , V , G , F ) . \mathbf Z_t
=
(A,R,S,V,G,F). Z t = ( A , R , S , V , G , F ) .
本篇再把主觀狀態寫為:
Q t = Ω ( Z t , E t , B i ) . Q_t
=
\Omega
\left(
\mathbf Z_t,
\mathbf E_t,
\mathbf B_i
\right). Q t = Ω ( Z t , E t , B i ) .
因此 stimulant 或其他 perturbation 可以首先造成:
Δ Q t ≠ 0. \Delta Q_t\neq0. Δ Q t = 0.
這個主觀 state effect 本身並不需要 objective performance 改善才算存在。
16. 2025 Healthy-Adult MPH Study:改善是局部,不是全域
2025 年 Aitken 等人進行 randomized, double-blind, placebo-controlled crossover study:
N = 25 N=25 N = 25
名健康成人接受:
10 mg methylphenidate 10\text{ mg methylphenidate} 10 mg methylphenidate
與 placebo。
研究發現:
numeric working memory accuracy ↑ \text{numeric working memory accuracy}
\uparrow numeric working memory accuracy ↑
且 error 降低。
但是其他大部分:
cognitive outcomes;
visual scanning metrics;
沒有顯著 treatment effect。
因此:
pharmacological effect ≠ global cognitive gain . \boxed{
\text{pharmacological effect}
\neq
\text{global cognitive gain}.
} pharmacological effect = global cognitive gain .
17. Domain-Specific Gain Vector
本文定義 objective performance vector:
P t = ( P attention , P WM , P memory , P speed , P inhibition , P reasoning , P academic , … ) . \mathbf P_t
=
\left(
P_{\text{attention}},
P_{\text{WM}},
P_{\text{memory}},
P_{\text{speed}},
P_{\text{inhibition}},
P_{\text{reasoning}},
P_{\text{academic}},
\ldots
\right). P t = ( P attention , P WM , P memory , P speed , P inhibition , P reasoning , P academic , … ) .
藥物或狀態改變:
Δ P t \Delta \mathbf P_t Δ P t
不必滿足:
Δ P k > 0 ∀ k . \Delta P_k>0
\quad
\forall k. Δ P k > 0 ∀ k .
更可能:
Δ P = ( + , 0 , 0 , + , 0 , − , … ) . \Delta \mathbf P
=
(+,\ 0,\ 0,\ +,\ 0,\ -,\ldots). Δ P = ( + , 0 , 0 , + , 0 , − , … ) .
所以「認知增強」最好不要使用單一 scalar。
18. 2026 Experimental Exam Study:小型 cognitive gain 不必轉成真實任務 gain
2026 年 Sambeth 等人使用 double-blind placebo-controlled experimental exam。
健康大學生分為:
placebo–placebo;
MPH before learning;
MPH before exam。
給予:
20 mg MPH . 20\text{ mg MPH}. 20 mg MPH .
結果:
factual MCQ ≈ no improvement , \text{factual MCQ}
\approx
\text{no improvement}, factual MCQ ≈ no improvement ,
inferential MCQ ≈ no improvement , \text{inferential MCQ}
\approx
\text{no improvement}, inferential MCQ ≈ no improvement ,
open inference ≈ no improvement , \text{open inference}
\approx
\text{no improvement}, open inference ≈ no improvement ,
overall grade ≈ no improvement . \text{overall grade}
\approx
\text{no improvement}. overall grade ≈ no improvement .
因此:
laboratory-domain improvement ⇏ academic transfer . \boxed{
\text{laboratory-domain improvement}
\not\Rightarrow
\text{academic transfer}.
} laboratory-domain improvement ⇒ academic transfer .
19. Transfer Efficiency
若低階或特定 cognitive metric 改善:
Δ P micro > 0 , \Delta P_{\text{micro}}>0, Δ P micro > 0 ,
而高階任務改善:
Δ P macro , \Delta P_{\text{macro}}, Δ P macro ,
定義:
η transfer = Δ P macro Δ P micro + ε . \eta_{\text{transfer}}
=
\frac{
\Delta P_{\text{macro}}
}{
\Delta P_{\text{micro}}+\varepsilon
}. η transfer = Δ P micro + ε Δ P macro .
如果:
η transfer ≈ 0 , \eta_{\text{transfer}}\approx0, η transfer ≈ 0 ,
表示 micro-level gain 沒有成功轉移到 macro task。
因此:
capacity gain ≠ transfer gain . \boxed{
\text{capacity gain}
\neq
\text{transfer gain}.
} capacity gain = transfer gain .
20. 真實任務通常是 multiplicative system
考試表現可能依賴:
P exam = F ( K , M , A , W M , R , S , T , E ) , P_{\text{exam}}
=
F
\left(
K,
M,
A,
WM,
R,
S,
T,
E
\right), P exam = F ( K , M , A , W M , R , S , T , E ) ,
其中:
K K K :knowledge;
M M M :memory;
A A A :attention;
W M WM W M :working memory;
R R R :reasoning;
S S S :strategy;
T T T :time management;
E E E :emotional/environmental state。
因此只提高:
W M WM W M
不必然顯著提高:
P exam . P_{\text{exam}}. P exam .
21. Expectancy 可以直接改變 Subjective State
Prescription-stimulant expectancy research 顯示,當受試者相信自己收到 stimulant 時:
Q t Q_t Q t
可以改變。
例如:
subjective arousal;
perceived drug effect;
confidence;
perceived enhancement。
這使:
expectancy → Q t \boxed{
\text{expectancy}
\rightarrow
Q_t
} expectancy → Q t
成為獨立路徑。
不需要:
active drug \text{active drug} active drug
存在。
22. Expectancy 不必改善 Objective Cognition
較早 methylphenidate balanced-expectancy study 已直接指出:
expectation → subjective arousal \text{expectation}
\rightarrow
\text{subjective arousal} expectation → subjective arousal
但:
expectation ⇏ general objective cognitive enhancement . \text{expectation}
\not\Rightarrow
\text{general objective cognitive enhancement}. expectation ⇒ general objective cognitive enhancement .
較新的 Adderall-expectancy randomized research 亦顯示 expectancy 主要改變 subjective mood/drug effects,而 objective cognitive benefit 並沒有同樣穩定出現。
因此:
expectancy gain ≠ performance gain . \boxed{
\text{expectancy gain}
\neq
\text{performance gain}.
} expectancy gain = performance gain .
23. Placebo 不能被過度簡化
2024 年 methylphenidate placebo-controlled pediatric ADHD research 發現 parent-rated symptoms 在 placebo condition 下可以出現 non-specific improvement。
但:
teacher ratings 未必同樣改善;
baseline severity 可解釋部分 regression-to-the-mean;
parent/child treatment expectation 並未在該研究中顯著預測 placebo improvement。
因此:
placebo-related change ≠ expectancy only . \boxed{
\text{placebo-related change}
\neq
\text{expectancy only}.
} placebo-related change = expectancy only .
Non-specific effect 可能包括:
expectancy + regression to mean + observer effect + context + measurement noise . \text{expectancy}
+
\text{regression to mean}
+
\text{observer effect}
+
\text{context}
+
\text{measurement noise}. expectancy + regression to mean + observer effect + context + measurement noise .
24. 「心理補償」需要更精確的名稱
本文不使用模糊的:
心理補償。
改定義:
G Q = Q post − Q pre , G_Q
=
Q_{\text{post}}
-
Q_{\text{pre}}, G Q = Q post − Q pre ,
為 subjective-state gain。
定義:
G C = P ^ post − P ^ pre , G_C
=
\widehat P_{\text{post}}
-
\widehat P_{\text{pre}}, G C = P post − P pre ,
為 confidence gain。
定義:
G P = P post − P pre , G_P
=
P_{\text{post}}
-
P_{\text{pre}}, G P = P post − P pre ,
為 objective-performance gain。
三者可以:
G Q > 0 , G_Q>0, G Q > 0 ,
G C > 0 , G_C>0, G C > 0 ,
但:
G P ≈ 0. G_P\approx0. G P ≈ 0.
這就是本文最重要的候選 dissociation。
25. Subjective Enhancement Gap
定義:
S E G = G Q − G P . SEG
=
G_Q-G_P. S E G = G Q − G P .
如果:
S E G > 0 , SEG>0, S E G > 0 ,
表示主觀增益高於客觀增益。
但注意:
Q Q Q
與:
P P P
可能不同單位。
因此正式研究應先將兩者轉成:
z -score z\text{-score} z -score
或 latent standardized metric,再計算:
S E G z . SEG_z. S E G z .
26. Metacognitive Enhancement Gap
定義:
M E G = G C − G P . MEG
=
G_C-G_P. M E G = G C − G P .
若:
M E G > 0 , MEG>0, M E G > 0 ,
表示 perceived performance improvement 超過 measured performance improvement。
這比單純說:
placebo。
更精確。
因為 active pharmacological effect 與 expectancy 可以同時存在。
27. 三層狀態空間
本文定義:
S t S M O = ( Q t , P ^ t , P t ) . \mathbf S_t^{SMO}
=
\left(
Q_t,
\widehat P_t,
P_t
\right). S t S M O = ( Q t , P t , P t ) .
可能存在至少四個典型 regime。
27.1 Concordant Improvement
Q ↑ , Q\uparrow, Q ↑ ,
P ^ ↑ , \widehat P\uparrow, P ↑ ,
P ↑ . P\uparrow. P ↑ .
27.2 Felt Enhancement Without Transfer
Q ↑ , Q\uparrow, Q ↑ ,
P ^ ↑ , \widehat P\uparrow, P ↑ ,
P ≈ 0. P\approx0. P ≈ 0.
27.3 Objective Gain Without Strong Feeling
Q ≈ 0 , Q\approx0, Q ≈ 0 ,
P ^ ≈ 0 , \widehat P\approx0, P ≈ 0 ,
P ↑ . P\uparrow. P ↑ .
27.4 Miscalibrated Decline
Q ↑ , Q\uparrow, Q ↑ ,
P ^ ↑ , \widehat P\uparrow, P ↑ ,
但:
P ↓ . P\downarrow. P ↓ .
第四種是安全上最重要的候選狀態之一。
28. Confidence 可以改變策略
若:
P ^ t ↑ \widehat P_t\uparrow P t ↑
造成:
verification effort ↓ , \text{verification effort}\downarrow, verification effort ↓ ,
則:
P t + 1 P_{t+1} P t + 1
可能反而下降。
令 verification intensity:
V t . V_t. V t .
候選:
V t = V 0 − λ P ^ t . V_t
=
V_0
-
\lambda \widehat P_t. V t = V 0 − λ P t .
若 confidence 過高:
P ^ t ≫ P t , \widehat P_t\gg P_t, P t ≫ P t ,
則:
V t ↓ . V_t\downarrow. V t ↓ .
因此 metacognitive bias 不只是 measurement outcome,也可能回饋行為。
29. Metacognitive Feedback Loop
建立:
P t → P ^ t → A t + 1 → P t + 1 . P_t
\rightarrow
\widehat P_t
\rightarrow
A_{t+1}
\rightarrow
P_{t+1}. P t → P t → A t + 1 → P t + 1 .
其中:
A t + 1 A_{t+1} A t + 1
為下一步 action/strategy。
因此:
E t meta E_t^{\text{meta}} E t meta
若長期偏差,會累積成:
過少複習;
過早結束;
過度冒險;
不尋求協助;
或相反的過度檢查。
30. Underconfidence 也同樣重要
本文不只研究:
E meta > 0. E_{\text{meta}}>0. E meta > 0.
如果:
E meta < 0 , E_{\text{meta}}<0, E meta < 0 ,
可能造成:
unnecessary checking;
avoidance;
reduced exploration;
delayed decisions;
low self-efficacy。
所以:
perfect metacognition ≠ low confidence . \boxed{
\text{perfect metacognition}
\neq
\text{low confidence}.
} perfect metacognition = low confidence .
理想狀態是:
P ^ t ≈ P t . \widehat P_t
\approx
P_t. P t ≈ P t .
31. Confidence Calibration 不等於 Self-Esteem
本文中的:
P ^ t \widehat P_t P t
是 task-specific performance estimate。
它不等於:
self-esteem , \text{self-esteem}, self-esteem ,
identity , \text{identity}, identity ,
global self-confidence . \text{global self-confidence}. global self-confidence .
因此「正向自我感」與「metacognitive calibration」不能混為一談。
32. State Clarity 也不等於 Mood
Q t Q_t Q t
可以包含:
clarity;
alertness;
engagement;
vividness。
但 mood:
M t affect M_t^{\text{affect}} M t affect
是另一維。
因此:
Q t ≠ M t affect . Q_t
\neq
M_t^{\text{affect}}. Q t = M t affect .
未來實驗需將:
mood;
arousal;
subjective clarity;
confidence;
分開量測。
33. Objective Performance 也不是單一真值
P t P_t P t
本身應是向量:
P t . \mathbf P_t. P t .
例如:
P t = ( accuracy , RT , RTV , memory , inhibition , transfer , real-world outcome ) . \mathbf P_t
=
(
\text{accuracy},
\text{RT},
\text{RTV},
\text{memory},
\text{inhibition},
\text{transfer},
\text{real-world outcome}
). P t = ( accuracy , RT , RTV , memory , inhibition , transfer , real-world outcome ) .
因此一個 treatment 可以:
R T ↓ RT\downarrow R T ↓
但:
a c c u r a c y ≈ 0 , accuracy\approx0, a cc u r a cy ≈ 0 ,
或:
W M ↑ WM\uparrow W M ↑
但:
e x a m ≈ 0. exam\approx0. e x am ≈ 0.
34. Speed–Accuracy Trade-off
若:
R T ↓ RT\downarrow R T ↓
但:
e r r o r ↑ , error\uparrow, er r or ↑ ,
不能直接叫「認知增強」。
定義:
P eff = f ( a c c u r a c y , R T ) . P_{\text{eff}}
=
f
\left(
accuracy,
RT
\right). P eff = f ( a cc u r a cy , R T ) .
不同任務可有不同成本函數。
因此:
faster ≠ better . \boxed{
\text{faster}
\neq
\text{better}.
} faster = better .
35. Real-World Transfer 是第四層
除了:
Q t , P ^ t , P t , Q_t,
\widehat P_t,
P_t, Q t , P t , P t ,
本文再加入:
R t world = real-world functional outcome . R_t^{\text{world}}
=
\text{real-world functional outcome}. R t world = real-world functional outcome .
所以最終:
Q ≠ P ^ ≠ P lab ≠ R world . \boxed{
Q
\neq
\widehat P
\neq
P_{\text{lab}}
\neq
R_{\text{world}}.
} Q = P = P lab = R world .
這正是實驗室 cognitive gain 不一定轉成 academic/occupational gain 的原因之一。
36. Transfer Chain
定義:
Q t → P ^ t → P t micro → P t macro → R t world . Q_t
\rightarrow
\widehat P_t
\rightarrow
P_t^{\text{micro}}
\rightarrow
P_t^{\text{macro}}
\rightarrow
R_t^{\text{world}}. Q t → P t → P t micro → P t macro → R t world .
每一箭頭都有 transfer coefficient:
η 1 , η 2 , η 3 , η 4 . \eta_1,\eta_2,\eta_3,\eta_4. η 1 , η 2 , η 3 , η 4 .
因此:
R t world = F ( Q , P ^ , P micro , P macro , η 1 , … , η 4 ) . R_t^{\text{world}}
=
F
\left(
Q,
\widehat P,
P_{\text{micro}},
P_{\text{macro}},
\eta_1,\ldots,\eta_4
\right). R t world = F ( Q , P , P micro , P macro , η 1 , … , η 4 ) .
任何一段:
η k ≈ 0 \eta_k\approx0 η k ≈ 0
都可能阻止上游 gain 轉化為真實生活結果。
37. ADHD-specific Hypothesis:Calibration Variability
本文不主張:
E [ E meta ] \mathbb E
\left[
E_{\text{meta}}
\right] E [ E meta ]
一定在所有 ADHD 個體中更高。
更值得測試的是:
Var [ E meta ∣ task ] . \operatorname{Var}
\left[
E_{\text{meta}}
\mid
\text{task}
\right]. Var [ E meta ∣ task ] .
以及:
Var t [ E meta ] . \operatorname{Var}_t
\left[
E_{\text{meta}}
\right]. Var t [ E meta ] .
若 ADHD 是 configuration-sensitive system,可能出現:
calibration itself is state- and task-dependent . \boxed{
\text{calibration itself is state- and task-dependent}.
} calibration itself is state- and task-dependent .
38. Clinical Symptom Improvement 不等於 Cognitive Enhancement
成人 ADHD treatment 可以:
symptom score ↓ \text{symptom score}\downarrow symptom score ↓
而某些 objective cognition:
P k P_k P k
改善,
但不代表:
P j ↑ ∀ j . P_j\uparrow
\quad
\forall j. P j ↑ ∀ j .
2025 年 adult ADHD single-dose MPH challenge research 顯示,acute changes in certain clinical/Qb measures 可對兩個月 treatment response 提供資訊,但這仍是 domain-specific predictive relationship,不是 global cognitive enhancement。
因此:
clinical response ≠ global cognitive upgrade . \boxed{
\text{clinical response}
\neq
\text{global cognitive upgrade}.
} clinical response = global cognitive upgrade .
39. SMOSH 十一項核心命題
SM-H1:三層分離命題
Q ≠ P ^ ≠ P . Q
\neq
\widehat P
\neq
P. Q = P = P .
若三者在各類任務中幾乎完全同步,模型應簡化。
SM-H2:Domain-Specific Calibration 命題
E meta = F ( domain , difficulty , state ) . E_{\text{meta}}
=
F
\left(
\text{domain},
\text{difficulty},
\text{state}
\right). E meta = F ( domain , difficulty , state ) .
SM-H3:Subjective–Objective Dissociation 命題
存在:
G Q > 0 G_Q>0 G Q > 0
而:
G P ≈ 0 G_P\approx0 G P ≈ 0
的可重現情況。
SM-H4:Confidence–Performance Dissociation 命題
存在:
G C > 0 G_C>0 G C > 0
而:
G P ≈ 0 G_P\approx0 G P ≈ 0
的可重現情況。
SM-H5:Expectancy Path 命題
expectancy → Q \text{expectancy}
\rightarrow
Q expectancy → Q
與:
expectancy → P ^ \text{expectancy}
\rightarrow
\widehat P expectancy → P
可在沒有 active pharmacology 時存在。
SM-H6:Limited Transfer 命題
Δ P micro > 0 \Delta P_{\text{micro}}>0 Δ P micro > 0
不保證:
Δ P macro > 0. \Delta P_{\text{macro}}>0. Δ P macro > 0.
SM-H7:Task Interaction 命題
同一 intervention:
I I I
對不同 P k P_k P k 的效應方向與大小不相同。
SM-H8:Validity-Control 命題
subjective–objective discrepancy 的研究必須控制:
V perf , V symptom . V_{\text{perf}},
V_{\text{symptom}}. V perf , V symptom .
SM-H9:Metacognitive Feedback 命題
E meta E_{\text{meta}} E meta
可透過 strategy selection 影響下一時刻 performance。
SM-H10:Underconfidence Symmetry 命題
metacognitive dysfunction 不只包含 overconfidence,也包含 context-dependent underconfidence。
SM-H11:Incremental Value 命題
若加入:
Q , P ^ , E meta Q,
\widehat P,
E_{\text{meta}} Q , P , E meta
無法在 out-of-sample prediction 中提升對 functional outcome 的預測,SMOSH 應被簡化。
40. 實驗一:Triple Measurement
每一 trial 同時量:
subjective clarity;
confidence;
objective accuracy。
即:
( Q i , c i , y i ) . (Q_i,c_i,y_i). ( Q i , c i , y i ) .
避免以一次 global questionnaire 代替 trial-level dynamics。
41. 實驗二:Balanced Expectancy Design
建立:
2 × 2 2\times2 2 × 2
設計:
active drug/placebo;
told drug/told placebo。
形成四組:
( D + , E + ) , (D^{+},E^{+}), ( D + , E + ) ,
( D + , E − ) , (D^{+},E^{-}), ( D + , E − ) ,
( D − , E + ) , (D^{-},E^{+}), ( D − , E + ) ,
( D − , E − ) . (D^{-},E^{-}). ( D − , E − ) .
可以分離:
pharmacological effect , \text{pharmacological effect}, pharmacological effect ,
expectancy effect , \text{expectancy effect}, expectancy effect ,
interaction . \text{interaction}. interaction .
42. 實驗三:Cross-Domain Battery
同一受試者測:
P attention , P_{\text{attention}}, P attention ,
P working memory , P_{\text{working memory}}, P working memory ,
P episodic memory , P_{\text{episodic memory}}, P episodic memory ,
P reasoning , P_{\text{reasoning}}, P reasoning ,
P motor , P_{\text{motor}}, P motor ,
P academic simulation . P_{\text{academic simulation}}. P academic simulation .
同步收集:
P ^ k . \widehat P_k. P k .
如果 subjective global enhancement 很高,但只有少數 P k P_k P k 提升,直接支持 domain-specific model。
43. 實驗四:Realistic Exam Transfer
使用:
study phase;
delay;
factual questions;
inference;
open answers。
並另外收集:
Q study , Q_{\text{study}}, Q study ,
P ^ exam , \widehat P_{\text{exam}}, P exam ,
P exam . P_{\text{exam}}. P exam .
這可以直接測:
felt study quality → ? actual grade . \text{felt study quality}
\rightarrow?
\text{actual grade}. felt study quality → ? actual grade .
44. 實驗五:Metacognitive Resolution
逐題收集 confidence。
計算:
B cal , B_{\text{cal}}, B cal ,
E abs , E_{\text{abs}}, E abs ,
B S , BS, B S ,
R meta . R_{\text{meta}}. R meta .
避免只測:
你覺得自己做得如何?
45. 實驗六:Feedback Recalibration
第一 block 不給 feedback。
第二 block 提供:
correct/incorrect \text{correct/incorrect} correct / incorrect
與 calibration feedback。
觀察:
E meta , 2 − E meta , 1 . E_{\text{meta},2}
-
E_{\text{meta},1}. E meta , 2 − E meta , 1 .
如果 ADHD-related bias 可被即時 feedback 快速修正,表示 metacognitive representation 不是固定缺陷。
46. 實驗七:State–Performance Time Series
每隔:
Δ t \Delta t Δ t
收集:
Q t , Q_t, Q t ,
P ^ t , \widehat P_t, P t ,
P t . P_t. P t .
建立 cross-lag model:
Q t → P t + 1 , Q_t
\rightarrow
P_{t+1}, Q t → P t + 1 ,
P t → P ^ t + 1 , P_t
\rightarrow
\widehat P_{t+1}, P t → P t + 1 ,
P ^ t → s t r a t e g y t + 1 . \widehat P_t
\rightarrow
strategy_{t+1}. P t → s t r a t e g y t + 1 .
此方法比 pre/post 平均更適合 dynamic configuration theory。
47. 失敗條件
SMOSH 應被削弱,如果:
Q Q Q 無法與 mood/arousal 分離;
confidence 無法提供超越 subjective state 的信息;
confidence 與 performance 在所有 domain 幾乎完全對齊;
ADHD group 的 metacognitive bias 在 replication 中不穩定;
task modality 不影響 calibration;
expectancy 沒有任何獨立主觀效應;
micro cognitive gain 幾乎總能高比例 transfer 至 macro outcome;
subjective–objective discrepancy 完全可由 measurement invalidity 解釋;
longitudinal state data 不優於 simple trait score。
若:
P SMOSH,out ≤ P simple,out , P_{\text{SMOSH,out}}
\leq
P_{\text{simple,out}}, P SMOSH,out ≤ P simple,out ,
則應優先保留較簡單模型。
48. 最重要的反例:不是所有 stimulant effect 都只是「感覺」
2025 healthy-adult MPH study 確實發現 numeric working memory accuracy 改善。
因此本文拒絕:
stimulant effect = placebo only . \boxed{
\text{stimulant effect}
=
\text{placebo only}.
} stimulant effect = placebo only .
藥物可以有 objective effects。
真正命題是:
objective effect is domain-specific and need not match subjective global enhancement . \boxed{
\text{objective effect is domain-specific
and need not match subjective global enhancement}.
} objective effect is domain-specific and need not match subjective global enhancement .
49. 第二個重要反例:不是所有 ADHD 都過度自信
既有 metamemory research 顯示,在合適條件下 ADHD adults 可以作出相對準確的 memory-performance judgment。
因此本文拒絕:
ADHD = global overconfidence . \boxed{
\text{ADHD}
=
\text{global overconfidence}.
} ADHD = global overconfidence .
更合理:
E meta = F ( profile , domain , task , state ) . E_{\text{meta}}
=
F
\left(
\text{profile},
\text{domain},
\text{task},
\text{state}
\right). E meta = F ( profile , domain , task , state ) .
50. 第三個反例:主觀困難可能是真實功能負擔
即使:
C complaint ≉ P lab , C_{\text{complaint}}
\not\approx
P_{\text{lab}}, C complaint ≈ P lab ,
也不能推出:
C complaint = false . C_{\text{complaint}}
=
\text{false}. C complaint = false .
因為 laboratory task 與 daily-life demands 不同。
所以:
subjective complaint ≠ objective lab deficit \boxed{
\text{subjective complaint}
\neq
\text{objective lab deficit}
} subjective complaint = objective lab deficit
與:
subjective complaint is invalid \boxed{
\text{subjective complaint is invalid}
} subjective complaint is invalid
是完全不同的命題。
51. 與前五篇整合
第 1 篇:
dynamic configuration . \text{dynamic configuration}. dynamic configuration .
第 2 篇:
N t → Z t . \mathbf N_t
\rightarrow
\mathbf Z_t. N t → Z t .
第 3 篇:
S t ≠ A t ≠ Π t ≠ O t ≠ U t ≠ G t . S_t
\neq
A_t
\neq
\Pi_t
\neq
O_t
\neq
U_t
\neq
G_t. S t = A t = Π t = O t = U t = G t .
第 4 篇:
X t = allocation-state dynamics . \mathbf X_t
=
\text{allocation-state dynamics}. X t = allocation-state dynamics .
第 5 篇:
G t = cognitive path topology . \mathcal G_t
=
\text{cognitive path topology}. G t = cognitive path topology .
第 6 篇加入:
S t S M O = ( Q t , P ^ t , P t ) . \boxed{
\mathbf S_t^{SMO}
=
\left(
Q_t,
\widehat P_t,
P_t
\right).
} S t S M O = ( Q t , P t , P t ) .
因此系列主幹更新為:
N t → Z t → Π t → X t → G t → O t → U t → G t → P t \boxed{
\mathbf N_t
\rightarrow
\mathbf Z_t
\rightarrow
\Pi_t
\rightarrow
\mathbf X_t
\rightarrow
\mathcal G_t
\rightarrow
O_t
\rightarrow
U_t
\rightarrow
G_t
\rightarrow
P_t
} N t → Z t → Π t → X t → G t → O t → U t → G t → P t
並行存在一條 metacognitive loop:
P t → P ^ t → s t r a t e g y t + 1 → P t + 1 . \boxed{
P_t
\rightarrow
\widehat P_t
\rightarrow
strategy_{t+1}
\rightarrow
P_{t+1}.
} P t → P t → s t r a t e g y t + 1 → P t + 1 .
主觀狀態:
Q t Q_t Q t
則同時受到:
Z t \mathbf Z_t Z t
與 expectancy/context 影響。
52. 本文不主張的內容
本文不主張:
ADHD 個體都會高估自己;
ADHD 個體都缺乏 metacognition;
subjective cognitive complaints 是假的;
objective neuropsychological tests 等於真實生活完整能力;
stimulant 只有 placebo effect;
stimulant 沒有 objective cognitive effect;
methylphenidate 不能改善 ADHD symptoms;
healthy-adult stimulant research 可直接外推 ADHD patients;
2026 exam study 可證明任何 ADHD 患者的用藥無效;
confidence 越低越好;
subjective clarity 越低越準;
placebo effect 等於 expectancy;
perceived enhancement 等於 deception;
本模型可用於個人藥物決策;
本文的任何數學量已成為臨床 biomarker。
53. 結論
本文最核心的修正很簡單:
feeling better ≠ estimating yourself better ≠ performing better . \boxed{
\text{feeling better}
\neq
\text{estimating yourself better}
\neq
\text{performing better}.
} feeling better = estimating yourself better = performing better .
真正需要同時測量:
Q t , Q_t, Q t ,
P ^ t , \widehat P_t, P t ,
P t . P_t. P t .
主觀狀態提升可以是真實 effect:
G Q > 0. G_Q>0. G Q > 0.
Confidence 也可以提高:
G C > 0. G_C>0. G C > 0.
但是客觀性能:
G P G_P G P
必須獨立測量。
因此:
G Q > 0 , G C > 0 , G P ≈ 0 G_Q>0,
\qquad
G_C>0,
\qquad
G_P\approx0 G Q > 0 , G C > 0 , G P ≈ 0
不是邏輯矛盾。
同樣:
G P > 0 G_P>0 G P > 0
也不表示:
G P ( k ) > 0 ∀ k . G_P(k)>0
\quad
\forall k. G P ( k ) > 0 ∀ k .
2025–2026 文獻共同支持一個更謹慎的方向:成人 ADHD 的 confidence–performance calibration 可能具有 task-specific bias;subjective cognitive complaints 與 objective neurocognitive deficits 並不一對一;methylphenidate 在部分 cognition domains 可以產生 measurable gain,但這種 gain 並不保證轉化為一般化 academic performance。
因此本篇最終提出:
cognitive enhancement should be treated as a vector with subjective, metacognitive, laboratory, and real-world components, not as a single scalar . \boxed{
\text{cognitive enhancement should be treated as a vector
with subjective, metacognitive, laboratory,
and real-world components,
not as a single scalar}.
} cognitive enhancement should be treated as a vector with subjective, metacognitive, laboratory, and real-world components, not as a single scalar .
真正可證偽的問題是:
Can subjective-state gain and metacognitive-confidence gain be dissociated from objective-performance gain, and do those dissociations predict later behavior? \boxed{
\text{Can subjective-state gain and metacognitive-confidence gain
be dissociated from objective-performance gain,
and do those dissociations predict later behavior?}
} Can subjective-state gain and metacognitive-confidence gain be dissociated from objective-performance gain, and do those dissociations predict later behavior?
如果答案是否定,SMOSH 應被簡化。
如果答案肯定,則「感覺變強」不再需要被粗暴分成「真的變強」或「只是 placebo」;它可以被更精確地表示為一個真實主觀狀態改變,其是否轉化為能力與實際成果,則由另一條可被獨立測量的因果鏈決定。
參考文獻
Elsholz, L., et al. ADHD symptomatology and metacognitive monitoring insights for college student support. Current Psychology . 2026. DOI: 10.1007/s12144-026-09164-9.
Butzbach, M., Fuermaier, A. B. M., Aschenbrenner, S., et al. Metacognition in adult ADHD: subjective and objective perspectives on self-awareness of cognitive functioning. Journal of Neural Transmission . 2021;128:1191–1204. PMID: 33464422.
Phillips, M. S., Turchmanovych-Hienkel, N., Leese, M., et al. The Relationship Between Subjective Cognitive Complaints, Invalid Symptom Reporting, and Neurocognitive Test Performance Validity Among Adults Being Evaluated for ADHD. Journal of Psychiatric Practice . 2025;31(1):13–19. DOI: 10.1097/PRA.0000000000000834.
Internalizing and somatic symptoms influence the discrepancy between subjective and objective cognitive difficulties in adults with ADHD who have valid and invalid test scores. Journal of the International Neuropsychological Society . 2024. PMID: 39291402.
Aitken, B., Downey, L. A., Rose, S., Manning, B., Arkell, T. R., Shiferaw, B., & Hayley, A. C. Acute Administration of 10 mg Methylphenidate on Cognitive Performance and Visual Scanning in Healthy Adults: Randomised, Double-Blind, Placebo-Controlled Study. Human Psychopharmacology: Clinical and Experimental . 2025;40(2):e70002. DOI: 10.1002/hup.70002.
Sambeth, A., Toth, M., & Blokland, A. Methylphenidate (Ritalin) does not improve exam performance in an experimental setting. Psychopharmacology . 2026;243(3):593–600. DOI: 10.1007/s00213-025-06864-1.
Looby, A., & Earleywine, M. Expectation to receive methylphenidate enhances subjective arousal but not cognitive performance. Experimental and Clinical Psychopharmacology . 2011. PMID: 21875224.
Looby, A., et al. Expectancy for Adderall influences subjective mood and drug effects regardless of concurrent caffeine ingestion: A randomized controlled trial. Psychopharmacology . 2023. PMID: 37740001.
Ilieva, I. P., Boland, J., & Farah, M. J. Objective and subjective cognitive enhancing effects of mixed amphetamine salts in healthy people. Neuropharmacology . 2013;64:496–505. PMID: 22884611.
Cropsey, K. L., Schiavon, S., Hendricks, P. S., et al. Mixed-amphetamine salts expectancies among college students: Is stimulant induced cognitive enhancement a placebo effect? Drug and Alcohol Dependence . 2017;178:302–309. PMID: 28686989.
Vertessen, K., et al. Placebo-related improvement with methylphenidate treatment in children with ADHD. European Child & Adolescent Psychiatry . Published online 2024. DOI: 10.1007/s00787-024-02550-3.
Pretzsch, C. M., Parlatini, V., et al. Clinical response to a single-dose methylphenidate challenge is indicative of treatment response at two months in adults with ADHD. Translational Psychiatry . 2025. DOI: 10.1038/s41398-025-03557-3.
文獻使用聲明
本文僅使用上述研究建立截至 2026-08-16 的外部實證邊界。
本文提出的 SMOSH、subjective-state variable Q t Q_t Q t 、metacognitive estimate P ^ t \widehat P_t P t 、objective-performance vector P t \mathbf P_t P t 、metacognitive error E t meta E_t^{\text{meta}} E t meta 、Subjective Enhancement Gap、Metacognitive Enhancement Gap、transfer efficiency η transfer \eta_{\text{transfer}} η transfer 與多層 transfer chain,均為本文理論構件,不應被誤認為上述研究作者的原始結論。
不同文獻包含正式 ADHD、ADHD referrals、healthy adults、college students、children,並使用 metacognitive confidence tasks、neuropsychological evaluation、methylphenidate placebo-control、experimental exam、self-report 與 clinical measures 等不同方法。本文不把它們視為單一大型實驗的直接累加證據。
狀態: v0.1,理論稿新增原始臨床/人體數據: 無醫學用途: 無下一篇: 《發展、補償與臨床可見性:為什麼 ADHD 可以到成年才被發現?》