★ Sample & headline finding: This report analyzes only
COMPLETE families from the Shifting Attention Data set — every included family has all three
members (Father + Mother + Son). That yields 5 complete families = 15 subjects (2 ASD triads:
families 004, 005; 3 Control triads: families 003, 004, 006). Five incomplete families were excluded.
ASD and Control performed equivalently on the Focus control conditions (Visual ≈100% vs 98%;
Auditory ≈99% vs 99%) but ASD was markedly impaired on the Shift condition (≈71% vs 96%, p≈.05,
d≈1.6), with a large shift cost (29 vs 2 pts, p≈.04). The deficit is therefore specific to
attentional switching, not sustained attention — the core prediction of the study. BAPQ-based
hypotheses (H5, H6) remain PENDING because the questionnaire column in the study spreadsheet is
empty (0/42 scored).
Study Overview
CTI paradigm scored exactly per proposal: a target = any stimulus the program scored hit/miss;
accuracy = hits/(hits+misses); CTI = interval between consecutive targets, binned into the five Courchesne windows.
Note the two accuracy columns: "Rapid-Shift (Bin 1)" is the theoretically critical
measure; "All-interval" is the average that masks the effect. Compare them across conditions.
★ ASD vs Control across CTI Bins (the key contrast)
Lines = and real ASD data. The study's prediction is that Controls
stay near ceiling while the ASD group performs worse on the Shift task. Toggle Parents / Children / All groups.
How to read: The vertical distance between the green (Control) and red (ASD) lines is
the shifting deficit. In this real sample the ASD group sits ~15–22 points below Control at every interval
— the deficit is broad across the Shift task, not confined to the rapid (Bin 1) window. Controls show the
expected dip-and-recover shape; the ASD group stays flat-but-low. Note the wide ASD error bars: the deficit is
driven largely by one of the two ASD families (see the Subject × Bin heatmap below).
1. Accuracy across Cue-to-Target Intervals
Accuracy across the five cue-to-target intervals for each condition, pooled across all subjects. Hover for values; error bars are ±1 SEM. (For the ASD-vs-Control split, see the dedicated section above.)
How to read: These lines pool all subjects (ASD + Control). Bin 1 (0.4–2.5 s) is the
rapid-disengagement window. The Shift line sits below the two Focus lines across the intervals because the
Shift task is harder; the Focus controls stay near ceiling because they require no shifting. For the
diagnostic-group split, use the ASD-vs-Control section above.
2. Within-Shift Interval Profile (pooled)
Does accuracy vary across the cue-to-target intervals within the Shift task (all subjects
pooled)? Paired t-tests (Holm-corrected) compare Bin 1 to every later bin, plus adjacent-bin tests.
This is a secondary, exploratory profile — the primary diagnostic-group contrast is the ASD-vs-Control
section above. With ASD and Control pooled here, group differences are not isolated.
3. Accuracy by Condition — ASD vs Control (Rapid-Shift vs All-Interval)
Each condition shows ASD next to Control (both real data). Toggle between the
rapid-shift (Bin 1) measure and the all-interval average. Either way, the groups match on the two Focus
controls and diverge on Shift.
Why it matters: The groups are equivalent on the Focus controls (both near
ceiling) and differ on the Shift condition. In this real sample the ASD shift deficit is broad —
present across the whole shift task rather than confined to Bin 1 — so the all-interval Shift average
(≈71% ASD vs ≈96% Control) and the Bin-1 view tell the same story. The Focus-vs-Shift dissociation
(large shift cost in ASD only) is the key evidence for a switching-specific deficit.
4. Shift Accuracy by Family Role across CTI Bins
Shift accuracy by family role (sons, mothers, fathers), pooled across diagnostic groups. Error bars ±1 SEM.
Interpretation: If RSA deficit is a heritable endophenotype, parents should resemble their
affected sons — especially at Bin 1. All three roles dip at Bin 1 here. Exploratory given n=2 per role, but it
previews the parent-vs-proband contrast the full study tests.
5. Reaction Time across CTI Bins
Proposal H3 targets Bin-1 reaction time. Slowest RTs at Bin 1 that fall with longer
intervals indicate a genuine disengagement bottleneck — not a speed-accuracy trade-off.
6. Individual Subjects by Condition
Every subject plotted (color = family role). With n=6, means can mislead — this shows the raw spread.
7. Practice / Fatigue across 10 Shift Trials
Interpretation: Upward slope = learning/practice; downward = fatigue/vigilance decrement — a
confound to report. Note this uses the all-interval trial average, so it tracks general stamina, not the Bin-1 deficit.
8. Signal Detection — Hit vs False-Alarm Rate
Why it matters: High accuracy with high false alarms = button-mashing, not true detection.
Good performance sits top-left (high hits, low false alarms).
9. Subject × CTI-Bin Heatmap (Shift)
How to read: Rows = subjects (AF/AM/AS = ASD father/mother/son; CF/CM/CS = Control),
columns = CTI bins. Green = high accuracy, red/orange = poor. This is the most revealing figure: the
Control rows are almost entirely green, while the ASD deficit is concentrated in family 004
(AF004, AM004, AS004 — orange/yellow) far more than family 005 (AF005, AM005 mostly green). The group
difference is therefore real but driven heavily by one of the two ASD families — a key caveat for
interpretation. White cells = no targets fell in that bin for that subject (sparse short-interval data).
10. Full Inferential Statistics (text)
==============================================================================
INFERENTIAL STATISTICS — ASD vs CONTROL
==============================================================================
Sample: 6 ASD subjects, 9 Control subjects — COMPLETE families only
(every included family has Father + Mother + Son; 2 ASD + 3 Control complete triads).
Real between-group comparisons (proposal H1-H4) are computed below.
BAPQ-based hypotheses (H5, H6) remain PENDING: the questionnaire column in
the study spreadsheet is empty (0/42 scored).
------------------------------------------------------------------------------
(A) BETWEEN-GROUP ACCURACY by CONDITION (ASD vs Control)
------------------------------------------------------------------------------
Visual Focus : ASD= 99.7% (n=6) Control= 97.9% (n=9) Welch t= 2.66, p=0.0210, d= 1.22, Mann-Whitney p=0.0531
Auditory Focus : ASD= 98.7% (n=6) Control= 98.9% (n=9) Welch t=-0.18, p=0.8605, d=-0.11, Mann-Whitney p=0.4824
Shift : ASD= 70.7% (n=6) Control= 96.1% (n=9) Welch t=-2.46, p=0.0536, d=-1.58, Mann-Whitney p=0.0153
-> Proposal predicts groups differ on SHIFT, not on the FOCUS controls.
-> Mann-Whitney (rank-based) is reported alongside Welch t because n is small and
accuracy is non-normal/bounded; agreement between the two increases confidence.
------------------------------------------------------------------------------
(B) GROUP x CONDITION INTERACTION (shift cost: Focus - Shift)
------------------------------------------------------------------------------
Shift cost (VisualFocus - Shift): ASD=29.0 pts Control=1.8 pts
Welch t=2.66, p=0.0412, d=1.70
-> A larger ASD cost = group x condition interaction (H1): switching hurts ASD more.
------------------------------------------------------------------------------
(C) CTI-BIN ACCURACY by GROUP (Shift) — where is the deficit?
------------------------------------------------------------------------------
Bin ASD % Ctrl % Gap Welch t p
Bin 1 70.1 88.6 18.5 -1.38 0.2070
Bin 2 76.8 99.0 22.2 -2.89 0.0331 *
Bin 3 78.7 94.1 15.5 -1.08 0.3328
Bin 4 77.9 99.1 21.2 -2.24 0.0737
Bin 5 76.0 96.4 20.4 -2.28 0.0663
-> Positive gap = Control outperforms ASD at that interval.
------------------------------------------------------------------------------
(D) RAPID-SHIFT (BIN 1) — ASD vs Control, accuracy & RT
------------------------------------------------------------------------------
Bin-1 accuracy: ASD=70.1% Control=88.6% Welch t=-1.38, p=0.2070, d=-0.79
Bin-1 hit RT: ASD=491ms Control=553ms Welch t=-0.82, p=0.4370 (H3: ASD slower)
------------------------------------------------------------------------------
(E) WITHIN-GROUP CTI-BIN SHAPE (RM-ANOVA per group)
------------------------------------------------------------------------------
ASD (n=5): F(4,16)=0.04, p=0.9970, eta^2=0.009 [B1=77%, B2=79%, B3=79%, B4=80%, B5=81%]
Control (n=7): F(4,24)=2.03, p=0.1226, eta^2=0.253 [B1=92%, B2=99%, B3=94%, B4=100%, B5=98%]
-> Tests whether accuracy varies across intervals WITHIN each group.
------------------------------------------------------------------------------
(F) MODALITY CONTROL CHECK — Visual vs Auditory Focus (each group)
------------------------------------------------------------------------------
ASD : Visual=99.7% Auditory=98.7% paired t=0.70, p=0.5177
Control : Visual=97.9% Auditory=98.9% paired t=-1.59, p=0.1504
------------------------------------------------------------------------------
(G) ROBUSTNESS of the SHIFT deficit (small-n sensitivity checks)
------------------------------------------------------------------------------
Individual subjects — ASD vals: [43.7, 48.8, 52.5, 90.0, 93.2, 96.1]
Control vals: [83.1, 89.6, 96.7, 98.0, 98.6, 98.6, 100.0, 100.0, 100.0]
NOTE: the ASD distribution is bimodal — roughly half the ASD subjects perform near
control level and half are severely impaired. The group mean sits between these clusters.
Family-level means (1 value/family, removes within-family correlation):
ASD families (n=2): [48.3, 93.1]
Control families (n=3): [93.9, 95.4, 98.9]
Welch t (family means) = -1.13, p = 0.4599
CAUTION: with only 2 ASD families, the family-level test is severely underpowered and
NOT significant. The subject-level effect is large (d~1.6) and the rank-based test is
significant, but the result rests on very few independent families and one ASD family
performs at control level. Treat the Shift deficit as a strong PRELIMINARY signal, not
a confirmed group difference, until more families are added.
------------------------------------------------------------------------------
(H) IQ-MATCHING CHECK (WASI-II PRI) — proposal inclusion requirement
------------------------------------------------------------------------------
ASD PRI available: n=0
Control PRI available: n=9, mean=113.8
CANNOT verify IQ matching: PRI is missing for one group in the spreadsheet
(the complete-family ASD subjects have no PRI scores recorded). This is a gap to
close before publication — the proposal requires PRI>70 and group IQ matching.
------------------------------------------------------------------------------
(I) PENDING — require BAPQ questionnaire scores
------------------------------------------------------------------------------
H5 ASD vs Control parent BAPQ scores (t-test) : PENDING (BAPQ column empty in spreadsheet)
H6 RSA accuracy x BAPQ correlation : PENDING (BAPQ column empty in spreadsheet)
11. Subject-Level Data Table
Click a header to sort.
Data Provenance & Excluded Files
Files excluded as practice/junk/unparseable, plus whole families dropped for being
incomplete (missing a Father, Mother, or Son). Only complete triads enter the analysis.
AL p1 AF-SBDissertationA.txt: unparseable filename (test/junk/retest)