THE RE-ANALYSIS OF WRONG ANSWERS AT RAVEN'S COLORED MATRICES BY ALCOHOLICS: A FINAL OVERVIEW

THE RE-ANALYSIS OF WRONG ANSWERS AT RAVEN'S COLORED MATRICES
BY ALCOHOLICS: A FINAL OVERVIEW

Renato COCCHI MD, a neurologist and a medical psychologist

Summary

This re-analysis of wrong answers given by three samples of alcoholics (67 Ss scoring <21; 90 Ss scoring 21-25; 93 Ss scoring 26-35) under detoxification at RCM brought to light new information. RCM are progressive both for their wrong answers figures and wrong answers non random groupings.

Own non random significant groupings are decreasing according to scoring sample, being maximum in the sample scoring < 21. The splitting of wrong answers non random groupings by contiguity into horizontal, vertical and diagonal contiguity appeared very interesting.

Without an analysis of each possible wrong answer all three series allow in their tables we cannot now go further.

Key words: Alcohol dependants; Raven's Colored Matrices; wrong answers; analysis.

The analysis of wrong answers given to Raven's Coloured Matrices (RCM) verges to became a more refined tool, since it could add more information to a net score.

This analysis in demented (Pola, Cocchi and Zerbi, 1988), in college students (Cocchi, 1993) or in alcoholics (Cocchi, 1993, 1995; Cocchi and Meme' 1998, Cocchi, 1999)) brought to light non random grouping of many wrong answers. According to these papers, the split of the prevalent mistakes shows the cognitive mechanisms in use to elicit these answers in a problem solving task.

Now I want to re-evaluate the wrong answers given by 3 groups of alcoholics under detoxification who scored <21 or 21-25, or 26-35. These ranges seem to make up subjects with the diagnosis of dementia, dubious dementia or no dementia according to Bertolani, De Renzi & Faglioni, 1993.

The aim of this new research is to find how different significant groups of wrong answers according to score are splitting into those 3 fields of possible intellectual functioning.

Materials and methods.

For this new research I chose the answers alcoholics did to RCM, as reported in my previous papers (Cocchi, 1993, 1995; Cocchi and Meme' 1998, Cocchi, 1999).

The <20 group of 60 alcoholics (Cocchi, 1993) had added 7 randomly chosen alcoholics from who scored 20 (Cocchi, 1999). So I could evaluate the whole range of wrong answers given by these patients to RCM. Sixty + seven alcoholics did scores < 21, ninety scored 21-25 and ninety-three scored 26-35.

Scores were given by alcoholics diagnosed as alcohol dependents according to DSM-III, R (303.92-3) or DSM-IV (F10.249). Each subject did RCM few days after admission, or when the symptoms of alcohol withdrawal cleared up.

RCM is a multiple choice test distributed into three sets of 12 tables (Series A, Ab and B), each table with an incomplete matrix of relations. To complete the pattern, a testee must select one of the six alternative answers set below the matrix, but five of them lead to a wrong result.

If we mark any possible answer with a number from one to six, from left to right, we have five figures to point the wrong answers up. The remaining one is the right answer, but its figure name varies in each table.

As for RCM scoring, a right answer makes one point, the maximum score being 36. I checked up the RCM charts where the psychologist noted the figure of the answer the subject chose. In this way we can identify each answer by a two digits figure, the first indicating the table and the second the answer chosen among six possibilities. EG. 3.6 stands for the choice 6 for the table 3.

In the previous four papers, the wrong answers were counted out and grouped table by table using to the figure name of any wrong answer. Then each table had all its wrong answers divided by five, being the ratio token in excess to the up figure. In this way each table's group of wrong answers had its theoretical number of random answers.

These random wrong answers token away from every group of a table, every remaining group shows the total of surely non random answers.

Every group of non random wrong answers had its p found out from an amount of three (p = 0.008). Now I named the wrong answers given to each table as a (scoring <21), b (scoring 21-25) or c (scoring 26-35).

Each series shows wrong answers the alcoholics did to each table, when there were significant groups of wrong answers. Of course wrong answers of subjects who scored <21 (ie. a) are over represented.

Results.

As for series A and significant non random groupings, alcoholics scoring <21 did 294 wrong answers out of 536 (54.85%) in 8 tables; who scored 21-25 did 246 wrong answers out of 540 (45.55%) in 6 tables; who scored 26-35 did 81 wrong answers out of 286 (28.32%) in 2 tables.

As for series Ab and significant non random groupings, alcoholics scoring <21 did 464 wrong answers out of 665 (69.77%) in 10 tables; who scored 21-25 did 367 wrong answers out of 810 (45.31%) in 9 tables; who scored 26-35 did 103 wrong answers out of 465 (22.15%) in 4 tables.

As for series B and significant non random grouping,s alcoholics scoring <21 did 494 wrong answers out of 632 (78.16%) in 10 tables; who scored 21-25 did 583 wrong answers out of 900 (64.78%) in 10 tables; who scored 26-35 did 275 wrong answers out of 651 (42.24%) in 7 tables.

I reported results in tables 1-3. Beside the column of each of the six figure names, there is a second column, (n)bis, which indicates the amounts of surely non random wrong answers, from a minimum group of three (p = .008).

In tables 4-6 I split the significant groupings of wrong answers into four categories: similarity, contiguity, opposition and confabulation. Then contiguity was split into three subcategories as horizontal, vertical and diagonal contiguity.

As you can see in series A, only 11.6 wrong answers groupings are common to the 3 scoring samples.

Groupings of wrong answers 7.4; 8.6; 9.5; 10.6 are the same both for samples a and b, and groupings 11.1 and 12.6 are the same for both samples b and c. Groupings found only in sample a are 4.3, 6.2, 8.1, 11.4; only in sample b are 11.1, 11.2 and 12.1. No grouping was an exclusive of the sample c.

If we carefully watch at groupings 8.1, 11.2, 11.4 in samples a and b only one grouping of the each couple reached the significant level. The same we can see for three-samples groupings 11.1 and 12.6 for sample a.

The non significant figures of those couples and three-samples groupings are always the highest figures among the remaining non significant ones in the same table.

Table 1: groups of wrong answers, matrix by matrix, to series A and significant groups of non random wrong answers. The symbol [*] runs for the right answer.

No. of the table (no. + series)	no. of wrong answer per name figure, as done and corrected wrong answer
No. of the table (no. + series)	Tot. answers	1	1bis	2	2bis	3	3bis	4	4bis	5	5bis	6	6bis	amount
4a	67	0		*		5	3+	2		0		0		7
6a	67	5		11	7&	*		1		1		0		18
7a	67	9		6		6		23	12$	0		*		44
7b	90	1		7		7		16	10 $	1		*		32
8a	67	11	3+	*		4		2		2		18	10$	37
8b	90	6		*		0		0		0		19	14$	25
9a	67	*		6		0		3		24	16$	6		39
9b	90	*		1		1		3		17	12$	3		25
10a	67	4		6		*		5		0		28	19$	43
10b	90	0		1		*		7		0		31	23$	39
11a	67	9		10		6		17	4!	*		18	6@	60
11b	90	21	5#	21	5#	2		10		*		25	9$	79
11c	93	15	3+	6		2		9		*		29	17$	61
12a	67	19	9$	9		3		*		3		12		46
12b	90	10		12		2		*		3		19	5#	46
12c	93	4		6		0		*		0		10	6@	20

p: + = .008; ! <.002; # <.0004; @ <.00007; & <.00002; $ <.0000006

Table 2: groups of wrong answers, matrix by matrix, to series Ab and significant groups of non random wrong answers. The symbol [*] runs for the right answer.

No. of the table (no. + series)	no. of wrong answers per name figure, as done and corrected wrong answers
No. of the table (no. + series)	Tot. answers	1	1bis	2	2bis	3	3bis	4	4bis	5	5bis	6	6bis	amount
2a	67	3		6	41	1		0		*		0		10
4a	67	25	14$	6		1		20	9$	0		*		52
4b	90	17	12$	0		0		3		1		*		21
5a	67	16	9$	*		6		1		9		1		33
5b	90	2		*		8	5#	1		2		0		13
6a	67	*		27	16$	1		7		12		4		51
6b	90	*		11	7&	1		0		8	3+	1		21
7a	66	4		17	8$	*		3		13	4!	4		41
7b	90	4		10	6@	*		0		1		2		17
8a	66	5		25	13$	4		*		23	11$	1		58
8b	90	0		25	18$	2		*		7		2		36
8c	93	1		8	5#	0		*		2		0		11
9a	66	2		21	10$	17	6@	1		13		*		54
9b	90	1		4		20	13$	1		10	3+	*		36
10a	66	1		36	25$	*		2		1		12		52
10b	90	0		16	5#	*		1		1		37	26$	55
10c	93	0		2		*		0		0		11	8$	13
11a	66	0		31	20$	16	5#	1		*		6		54
11b	90	2		16	7&	23	14$	1		*		2		44
11c	93	1		3		15	10$	1		*		0		20
12a	65	30	18$	*		4		19	7&	3		3		59
12b	90	33	15$	*		5		40	22$	5		5		88
12c	93	19	7&	*		9		18	6@	7		6		59

p: + = .008; ! <.002; # <.0004; @ <.00007; & <.00002; $ <.0000006

As you can see in series Ab, only 8.2, 11.3, 12.1 and 12.4 wrong answers groupings are concordant for this series.

Groupings of wrong answers 4.1; 6.2; 7.2; 9.3; 10.2 and 11.2 are the same both for samples a and b, and grouping 10.6 is the same for both samples b and c. Groupings found only in sample a are 2.2; 4.2; 5.1; 7.5; 8.5 and 9.2; only in sample b are 5.3, 6.5; 9.3; and 9.5.

. No grouping was an exclusive of the sample c. If we carefully watch at groupings 4.4; 6.5 and 9.5; in samples a and b only one grouping of the each couple reached the significant level. The same we can see for three-samples groupings 10.6, 11.2 for sample a or c. The non significant figures of those couples and three-samples groupings are always the highest figures among the remaining non significant ones in the same table.

Table 3: groups of wrong answers, matrix by matrix, to series B and significant groups of non random wrong answers. The symbol [*] runs for the right answer.

No. of the table (no. + series)	no. of wrong answers per name figure, as done and corrected wrong answers
No. of the table (no. + series)	Tot. answers	1	1bis	2	2bis	3	3bis	4	4bis	5	5bis	6	6bis	Amount
2a	64	1		10	6@	4		2		1		*		18
3a	64	*		18	13$	1		2		1		2		24
3b	90	*		5	3+	0		0		1		0		6
4b	90	2		*		1		5	3+	0		0		8
5a	63	*		16	5#	3		12		23	12$	1		55
5b	90	*		2		0		3		25	19$	0		30
6a	63	15		24	12$	*		3		6		10		58
6b	90	3		35	26$	*		1		2		4		45
6c	93	0		10	8$	*		0		0		0		10
7a	63	21	10$	19	8$	3		2		*		9		54
7b	90	18	6@	17	5#	3		3		*		19	7&	60
7c	93	6		10	4 !	0		1		*		13	7&	30
8a	63	11		3		3		20	8$	21	9$	*		58
8b	90	5		3		1		35	17$	44	26$	*		88
8c	93	1		4		7		15	5#	24	14$	*		51
9a	63	33	21$	9		3		*		11		3		59
9b	90	45	27$	7		2		*		29	11$	3		86
9c	93	28	18$	2		5		*		7		7		49
10a	63	23	12$	23	12$	*		4		1		4		55
10b	90	35	18$	36	19$	*		3		5		6		85
10c	93	15	6@	10		*		4		7		6		42
11a	63	7		24	13$	17	6@	*		5		2		55
11b	90	3		40	22$	29	11$	*		10		8		90
11c	93	7		12		18	5#	*		19	6@	7		63
12a	63	17	5#	16	4!	12		5		*		8		58
12b	90	7		22	5#	30	13$	14		*		12		85
12c	93	3		25	11$	12		9		*		19	5#	68

p: + = .008; ! <.002; # <.0004; @ <.00007; & <.00002; $ <.0000006

As you can see in series B, only 6.2, 7.2, 8.4; 8.5; 9.1; 10.1; 11.3 and 12.2 wrong answers groupings are concordant for this series.

Groupings of wrong answers 3.2; 5.5; 7.1; 10.2 and 11.2 are the same both for samples a and b, and grouping 7.6 is the same for both samples b and c. Groupings found only in sample a are 2.2; 4.4; 5.2 and 12.1; only in sample b are 4.4, 9.5 and 12.3.

The 11.5 and 12.6 groupings were exclusive of the sample c. If we carefully watch at grouping 5.2 in samples a and b only one grouping of that couple reached the significant level. The same we can see for three-samples groupings 7.1, 10.2 and 11.2 for sample c. The non significant figures of those couples and three-samples groupings are always the highest figures among the remaining non significant ones in the same table.

Table 4: non random wrong answers divided into the four chosen categories.
Series A

Tab. no. / answers no./ sample	identity answers	Contiguity answers by			opposition answers	confabulation answers
Tab. no. / answers no./ sample	identity answers	horizontal	vertical	diagonal	opposition answers	confabulation answers
4.3a	3
6.2a					7
7.4a						12
7.4b						10
8.1a		3
8.6a			10
8.6b			14
9.5a					16
9.5b					12
10.6a					19
10.6b					26
11.4a			4
11.6a		6
11.6b		9
11.6c		17
11.1b						5
11.1c						3
11.2b						5
12.1a			9
12.6b		5
12.6c		6
totals	3	46	37		80	35

As you can see in table 4 the main figure of wrong answers pertains to opposition answers (39.80%), followed by horizontal contiguity answers (22.88%), vertical contiguity answers (18.41%), confabulation answers (17.41%) and similarity answers(1.49%).

These rates cannot be compared each another because the chance to wrongly answer by using one of these mechanisms is not the same in every table of the series A.

Table 5: Series Ab: non random wrong answers divided into the four chosen categories.

Tab. no. / answers no./ sample	Identity answers	contiguity answers by			opposition answers	confabulation answers
Tab. no. / answers no./ sample	Identity answers	horizontal	vertical	Diagonal	opposition answers	confabulation answers
4.1a	4(*)		14
4.1b			12
5.1a				9
5.3b			5
6.2a			16
6.2b			7
6.5b		3
7.2a			8
7.2b			6
7.5a		4
8.2a			13
8.2b			18
8.2c			5
8.5a				11
9.2a				10
9.3a			6
9.3b			13
9.5b		3
10.2a		25
10.2b		5
10.6b			26
10.6c			8
11.2a			20
11.2b			7
11.3a				5
11.3b				14
11.3c				10
12.1a			18
12.1b			15
12.1c			7
12.4a		7
12.4b		22
12.4c		6
totals	4	75	224	59

(*) by 90° rotation

As you can see in table 5, now the main figure of wrong answers pertains to vertical contiguity answers (61.88%), followed by horizontal contiguity answers (20.72%), diagonal contiguity answers (16.30%) and similarity answers (1.10%). Answers of opposition or confabulation did not find any place in this series.

Also these rates cannot be compared each another because the chance to wrongly answer by using one of these mechanisms is not the same in every table of the series Ab.

Table 6: Series B; non random wrong answers divided into the four chosen categories.

Tab. no. / answers no./ sample	identity answers	Contiguity answers by			opposition answers	Confabulation Answers
		horiz.	vertical	diagonal
2.3a	6
3.2a		13
3.2b		3
4.4b			3
5.2a		5
5.5a			12
5.5b			19
6.2a			12
6.2b			26
6.2c			8
7.1a		10
7.1b		6
7.2a			8
7.2b			5
7.2c			4
7.6b						7
7.6c						7
8.4a		8
8.4b		17
8.4c		5
8.5a			9
8.5b			26
8.5c			4
9.1a			21
9.1b			27
9.1c			18
9.5b		11
10.1a			12
10.1b			18
10.1c			6
10.2a		12
10.2b		19
11.2a		13
11.2b		22
11.3a			6
11.3b			11
11.3c			5
11.5c						6
12.1a		5
12.2a			4
12.2b			5
12.2c			11
12.3b	13
12.6c				5
Totals	19	149	280	5		20

As you can see in table 6, now the main figure of wrong answers pertains to vertical contiguity answers (59.83%), followed by horizontal contiguity answers (31.84%), confabulation answers (4.27%), similarity answers (4.06%) and diagonal contiguity answers (1.07%). Answers of opposition did not find any place in this series. These rates cannot be compared each another because the chance to wrongly answer by using one of these mechanisms is not the same in every table of the series B.

Discussion.

This reanalysis of wrong answers given by three samples of alcoholics under detoxification bring some new information on RCM. Matrices are progressively difficult by definition and the three samples have been chosen according to scoring (sample a = < 21; sample b = 21-25; sample c = 26-35). These facts drew to an increased number of non random wrong answers groupings such as 21 in series A, 35 in series Ab and 40 in series B. The same happens for all three samples concordant significant groupings of wrong answers, namely 1 in series A, 4 in series Ab and 8 in series B.

Groupings of all three series found only in sample a are 4+ 6 + 4; found only in sample b are 3 + 4 + 3, and 2 only in sample c. This suggests that lower scoring samples have more significant own wrong answers groupings according to the lower level of cognitive functioning.

The central position of the sample b (21-25 scoring) could also be more defined by comparing the figures of concordant wrong answers groupings for samples a and b to the same for samples b and c. Being these figures 15 vs.4, we can suggest that sample b.is more similar to sample a than to sample c, at least for those alcoholics.

The non significant figures of one term of many couples and three-samples groupings are always the highest figures among the remaining non significant ones in the same table. This can drive us to doubt that in some cases the criterion adopted to reach significance is perhaps too cautious.

As for spitting in the four chosen categories (similarity, contiguity, opposition, confabulation), the non balanced presence of any possible wrong answer in each table does not help us to a better analysis. The main rate of opposition wrong answers in series A could only come out from the fact that more tables of series A allow the opposition wrong answer.

Surely the three subcategories of contiguity wrong answers brought us sound information. The alcoholics can elicit different contiguity wrong answers by horizontal, vertical and diagonal contiguity. Orizontal wrong answers seem more easy to elicit in series A.

Conclusion

This re-analysis of wrong answers given by three samples of alcoholics under detoxification at RCM brought to light new information, but many data lack a defined reference frame. RCM are progressive both for their wrong answers figures and wrong answers significant non random groupings.

Own non random significant groupings are decreasing according to scoring sample, being maximum in the sample scoring < 21.

The splitting of wrong answers non random groupings by contiguity into horizontal, vertical and diagonal contiguity appeared a new and promising pathway to explore carefully.

Without an analysis of each possible wrong answer all three series allow in their tables we cannot now go further.

References

L., De Renzi E., Faglioni P.: Test di memoria non verbale di impiego diagnostico in clinica: Taratura su soggetti normali. Arch. Psicol. Neurol. Psichiat. 1993, 54: 477-486.

Cocchi R.: Meccanismi "logici" nella acquisizione del linguaggio verbale: Una ipotesi esplicativa neurofisiologica degli ipercorrettismi. Riv. Neurobiol. 1982, 28: 162-190.

Cocchi R.: Analisi delle risposte errate, date alle PM47 di Raven, Rev. 1962, Forma I, da un campione di studenti universitari. Riv. Ital. Disturbo Intellet. 1993, 6: 83-90.

Cocchi R.: Alcolisti con punteggio < 20 alle Matrici Colorate di Raven: Analisi degli errori. Riv. Ital. Disturbo Intellet. 1993, 6: 269-275.

Cocchi R.: Alcoholics scoring 21-25 at Raven's Coloured Matrices: The analysis of wrong answers. It. J. Intellect Impair. 1996, 9: 181-187.

Cocchi R: Alcoholics scoring 20 at Raven's Coloured Matrices: The analysis of wrong answers. t. J. Intellect Impair. 1996, 12: 41-47.

Cocchi R., Meme' M.: Alcoholics scoring 26-35 at Raven's Coloured Matrices: The analysis of wrong answers It. J. Intellect. Impair. 1998, 11: 173-178.

Cocchi R., Pola A., Sellerini M., Tosca P., Zerbi F.: Mirror speaking after neurosurgery. Case history. Acta Neurol. Belg. 1986, 86: 224-232.

Jacobson R.: Essais de linguistique generale. Editions de Minuit, Paris 1963.

Jacobson R., Halle M.: Fundamentals of Language. Mouton, Den Haag, 1956.

Pola A., Cocchi R., Zerbi F.: Progressive Matrices PM 47 in demented inpatients: qualitative analysis of mistakes and problem solving strategies. Ital. J. Intellect. Impair. 1988, 1: 111-118.

Pribram K.H.: Languages of the brain. Experimental paradoxes and principles in neuropsychology. Prentice-Hall, Englewood Cliff, 1971.

Printed in It. J. Intellect Impair. 1999: 12: 49-57.

Author's address: dr Renato COCCHI, via Rabbeno, 3

43100 Reggio Emilia (Italy)

renatococchi@libero.it

Italian translation

*Theoretical bases

Home Page