THE SCHIQ: A new method of measuring the exceptionally intelligent (child prodigies)
Vernon M Neppe MD, PhD, FRSSAf, DFAPA, BN&NP, DPM, MMed,
Director, Pacific Neuropsychiatric Institute, Seattle, WA
www.PNI.org psyche@PNI.org; 206 527 6289.
(Adj Full Professor, Dept of Psychiatry, St Louis University, St Louis, MO.
Collaborators on this research:Stevan Damjanovic, Dr Greg Grove.
This is a preliminary model for genius. This requires far more research.
Our research paper reports on a preliminary pilot study to evaluate intelligence at the higher ranges using a different technique, namely applying historical data about the accomplishments of child prodigies. Child prodigies were chosen because the extent of their advancements could relatively easily be compared with older age peers. 3 raters ranked the vignettes used very closely using historical data.
The data is described in detail in the previous section on Prodigy research
The following were key early findings from our data:
Three raters were carefully chosen. All had had major interest, personal involvement, knowledge and experience in the high IQ measurement domain, substantial knowledge in research or theory of intelligence, and were involved in evaluating high IQ individuals e.g. for admission to various high IQ groups. They ranked the estimated IQ score (standard deviation 15) for each of the 28 individual items described. Their mean IQ estimates of each item were used to compare interrater reliability though all raters recognized the need for an item IQ range.
RESULTS
Rankings of each vignette item based on means of the 3 raters A, B and C
question# |
B |
C |
A |
Mean |
most diff |
Least diff. |
rater out |
1 |
175 |
170 |
175 |
173.3 |
5 |
0 |
C |
2 |
170 |
173 |
183 |
175.3 |
13 |
3 |
A |
3 |
160 |
161 |
160 |
160.3 |
1 |
0 |
C |
4 |
145 |
163 |
153 |
153.7 |
18 |
8 |
C+ |
5 |
155 |
158 |
145 |
152.7 |
13 |
3 |
A |
6 |
170 |
156 |
162 |
162.7 |
14 |
6 |
B+ |
7 |
160 |
158 |
160 |
159.3 |
2 |
0 |
C |
8 |
180 |
170 |
168 |
172.7 |
12 |
2 |
B |
9 |
145 |
135 |
134 |
138.0 |
11 |
1 |
B |
10 |
170 |
168 |
168 |
168.7 |
2 |
0 |
B |
11 |
180 |
173 |
183 |
178.7 |
10 |
3 |
C |
12 |
152 |
155 |
150 |
152.3 |
5 |
2 |
C |
13 |
150 |
145 |
153 |
149.3 |
8 |
3 |
C |
14 |
125 |
130 |
130 |
128.3 |
5 |
0 |
B |
15 |
135 |
135 |
149 |
139.7 |
14 |
0 |
A |
16 |
120 |
128 |
140 |
129.3 |
20 |
8 |
A+ |
17 |
145 |
139 |
145 |
143.0 |
6 |
0 |
C |
18 |
130 |
154 |
153 |
145.7 |
24 |
1 |
B |
19 |
165 |
163 |
160 |
162.7 |
5 |
2 |
A |
20 |
160 |
168 |
175 |
167.7 |
15 |
7 |
B+ |
21 |
175 |
168 |
170 |
171.0 |
7 |
2 |
B |
22 |
147 |
153 |
157 |
152.3 |
10 |
4 |
B |
23 |
140 |
145 |
140 |
141.7 |
5 |
0 |
C |
24 |
150 |
144 |
148 |
147.3 |
6 |
2 |
C |
25 |
170 |
149 |
165 |
161.3 |
21 |
5 |
C |
26 |
170 |
159 |
156 |
161.7 |
14 |
3 |
B |
27 |
155 |
153 |
153 |
153.7 |
2 |
0 |
B |
28 |
150 |
160 |
153 |
154.3 |
10 |
3 |
C |
Mean |
155.3 |
154.8 |
156.7 |
155.6 |
9.9 |
2.4 |
11B12C,5A |
Provisional illustrations from this data of individual item variances:
Despite 28 items, the mean differences between the 3 raters were extraordinarily close (a range of 1.9 total; implying <0.1 IQ point per item). The intra-rater variation between 2 raters is very little
(24/28 scored ≤5 on least; 9/28 on most variation analysis).
However, several items showed more marked variations in estimate. The most difficult to estimate IQ items were the accomplishments in the very young (particularly <5 years old and even more so under 3 years) and the extremely intelligent (≥4.5d). Overall assessments of “IQ scores” were more difficult with 1ess than 4 items listed per individual prodigy. Matching the prodigy children with the correct corresponding creative and educational accomplishments in adults was complex, though rank ordering the adult-child pairs correlated well. For example, If one looks at differences of say 15 or more:
Early age: e.g. Unscrewed, taught time all very young don't correlate. The geography one was okay apparently. But this suggests estimating say under age 4 is not consistent and may be far more variable. Without the combination items #10, #11, #17, we have 25 items. Taking out all those below age 6 unless there are direct statistics about mental age elements. (#1) If range, e.g. 5 to 8 then average is 6.5. Specifics: Reading is not clear for mental age because kids are sometimes trained to read. When eliminating these early age uninterpretable items and also combination items ( #3, #4, #13 (+ #15 as based on 2 items), #16, #17 (as higher score would be #16), #25, #26 then: 18/20 scored ≤5 on the least; 7/20 on the most.
Given the paucity of comparable outside child research to validate this technique, we used these preliminary results to further correct by guesstimation applying an increased standard error for younger age groups, very exceptional intelligence, and less test items. The scores were then recalculated producing a composite IQ score called SCHIQ (standard, corrected historical IQ) for the 7 subjects. This ranged from 146 though to 179. The two individuals with the largest number of data point vignettes correlated almost exactly with their estimated IQs using other techniques (179, 169), after corrections (Scoville’s for mental/chronological age: Ferguson’s combination of scores calculation). The results of the other five subjects also appeared appropriate based on the data available.
Early data from case vignettes of scoring intelligence (excludes age <5 unless can justify milestone) produced GASE= Guesstimated Age Related Standard Error.
Mean 154.8 154.6 156.3 155.2 10.4 2.4 10B,7C,3A
The GASE variations: aged 20 or over: -3 to +7; aged 13- 20: -4 to +10; aged 11 to 12: -5 to +13; aged 9 to 10: -6 to +14; aged 7 to 8: -7 to +15; aged 5 to 6: -8 to +16; aged 4: -10 to +18; aged 3: -12 to +20; aged ±under 3: -14 to +22
Effectively with each increment below age 6 to <3 the GASE range went up by 4.
Because of this a formula was derived, not discussed here, beyond illustrating that it exists:
SCHIQ = (SHIQ) +NIQ= (ULIQ-ASEC) +NIQ= (ULIQ-ASEC) +NIQ= (MLMIQ+TADS [if TADS≥0]-ASEC) = (MLMIQ+TADS [if TADS≥0]-ASEC). Essentially, this formula corrects for just estimating means via several raters, adding in corrections for numbers of vignettes realizing the greater reliability of several ones, perceiving the difficulties estimating profoundly high intelligence because there is no comparative base possibly other than documented creative achievement, and recognizes the difficulty of predicted super intelligent estimations based on very young age, but also takes account of the limitations of testing, and the possible lack of attainment of limits based on simple case vignettes. The formula is there to be adapted, is provisional and uses guesstimations which may be incorrect.
Adjustments to SCHIQ. Complex calculations, steps left out, for illustration only.
Ss |
Vig |
max |
min |
#PAD |
PADS |
NADS |
TADS |
SHIQ≥ |
|
|
|
mean |
mean |
Sub |
base |
* |
|
MLMIQ |
SCHIQ |
D |
8 |
191.3 |
174.7 |
5 |
182.7 |
169.3 |
172.7 |
174.7 |
179 |
E |
5 |
182.5 |
164.7 |
3 |
175.7 |
162.7 |
166.2 |
168.2 |
168 |
F |
2 |
153 |
138.7 |
2 |
148.7 |
|
148.7 |
148.7 |
146 |
G |
4 |
178 |
168 |
1 |
178 |
162.3 |
163.6 |
168 |
168 |
H |
4 |
167.3 |
144.3 |
3 |
149.7 |
|
149.7 |
149.7 |
150 |
I |
2 |
169.7 |
159.7 |
2 |
161.7 |
|
159.7 |
159.7 |
156 |
J |
3 |
176.7 |
154.7 |
3 |
161.3 |
154.7 |
159.3 |
159 |
157 |
Ss: subjects
Vig: number of vignettes per subject
Max and min means: Maximum and minimum means based on raters A, B and C
Adjusted IQ follows: based on rankings of age, profoundness and early vignette corrections:
# PAD; PADS IQ score: positive differences vs NADS negative difference ;
TADS IQ = total additional difference IQ score ;
MLMIQ = Mean limiting median IQ;
SHIQ = standardized uncorrected historical IQ;
SCHIQ= standardized historical corrected IQ
a. age above was analyzed, two other adjustments were made.
b. number of items vignettes —at least 4 appears critical;
c. profound IQ to >4d SCIQ adjusted. In all, a sliding scale was used.
Item examples are based on statistical definitions illustrating using SD=15: there were 28 items—a few are below.
(There is data suggesting the IQ skills above this level of ≥4d may be up to 8x more)
IQ 100 = 1d = 50th percentile = average; IQ 115 = 1d = superior = 1 in 6.
IQ 130 = 2 d = very superior = 1 in 44; IQ 145= 3d = exceptional = 1 in 741;
IQ 153 ~= 3.5= 1 in about 5,000; IQ 160 = 4d * = 1 in about 30,000 [i]
IQ 164 = 4.27d* = 1 in about 100,000; IQ 170= 6.67 d* = I in about 500,000
IQ 175 = 5d * ~= 1 in about 3.5 million; IQ 180 = 5.33d* ~= 1 in about 20 million
IQ 190= 6d * ~= 1 in about billion. These are statistical translations of estimates.
Analysis:
This study required special psychometrists skilled in high IQ assessment. (used those involved with establishing very high IQ societies and psychometrist allocations for those societies – present and future) It achieved very strong correlations of results between psychometrists, i.e. high inter-rater reliability. It accounted for some kind of external validating features e.g. Exceptional achievement—possibly the best measure, but possibly not available with those who fail. It allowed a historical measure of very high IQ based on achievement. Measures had previously largely been limited before focussing on convergence at very high levels (of e.g. logic) but not on the divergence linked with creativity. These should strongly correlate with SCHIQ
Corrections hypothesized because of variance should be higher with skill vignettes at very young ages, at very exceptional levels (>4 and even 4.5 d), with less vignettes to build on (more vignettes make scoring easier, plus imply several examples of exceptional skills or maintained advanced skills) and preliminarily validated.
Conclusions
The empirical data preliminarily appear to represent these exceptional subjects well. This technique appears promising in measuring IQ in child prodigies who then become high achiever adults. This is so as the SCHIQ measures allow for a possibly accurate measure of extremely high intelligence using outside validators of creative, academic and other recognized achievement. A replication is in process. The SCHIQ can be potentially applied to other settings, too. We are currently developing a standardized questionnaire for this second round of testing and accumulating a new population. However, our technique certainly depends on the experience and background of the raters, implying careful choice.
ACKNOWLEDGEMENTS: I gratefully acknowledge the assistance of Stevan Damjanovic and Greg Grove for their contributions to this research.
- These very high IQ scores disputably require extremely high creativity as well as profound divergent intelligence skills. It is therefore debatable whether any current IQ tests can appropriately measure these scores, both because of rarity of the statistic, because creativity is generally not measured and because even high IQ tests measure complex divergence. This is why these individual case histories may assist.
- All these case histories are genuine and the facts have been confirmed.
- All these individuals are current.
Neppe VM: Intelligence assessment by history taking in child prodigies with adult achievement—the SCHIQ, in International Society for Intelligence Research Conference Book. Decatur, GA, 2008, p 21