PSYCHOMETRIKA--VOL. 5, NO. 1 MARCH, 1940

P O I N T S OF N E U T R A L I T Y I N S O C I A L A T T I T U D E S O F D E L I N Q U E N T S A N D N O N - D E L I N Q U E N T S *

RUTH BISHOP

College Entrance Examination Board Princeton, New Jersey

Delinquent boys are compared with non-delinquents with re- spect to their attitudes towards a series of "good" and "bad" so- cial acts, by the use of scales having rational origins of measure- ment. A new technique, essentially an extension of Thurstone's Method of Successive Intervals, is found to give results similar to Horst's Method of Balanced Values. Significant differences in mean attitude between the two groups are not found.

T h e gene ra l p u r p o s e of th is s t u d y is to d e t e r m i n e w h e t h e r d i f fe r - ences can be d e m o n s t r a t e d b e t w e e n the a t t i t u d e s of de l inquen t s a n d non-de l inquen t s . A l t h o u g h the c r i m i n a l code is based upon the place- m e n t of ac ts on a c rude psycho log ica l scale in o r d e r of t h e i r social accep tab i l i t y , i t is qui te poss ib le t h a t the m e m b e r s of the d i f f e r en t g r o u p s in a soc ie ty do no t a g r e e as to the o r d e r in wh ich the ac t s m a y be d iv ided into " g o o d " and " b a d " f o r m s of b e h a v i o r . I t m i g h t be ex- pec ted t h a t de l inquen t and non -de l i nquen t g r o u p s would d e m o n s t r a t e d i f fe rences in t h e i r ab i l i t y to m a k e d i s t inc t ions b e t w e e n the social ac- c e p t a b i l i t y of a n y g iven p a i r of acts . F o r examp le , one of the g r o u p s m i g h t m a k e l i t t le o r no d i s t inc t ion b e t w e e n t w o ac t s a n d p lace t h e m a t a b o u t t h e s a m e place on the scale, whi le the o t h e r g r o u p m i g h t m a k e a c l ea r d i s t inc t ion b e t w e e n the ac ts , even c o n s i d e r i n g one of t h e m as b e i n g def ini te ly on the " g o o d " s ide o f the scale and the o t h e r def in i te ly on the " b a d " s ide of t he scale.

I n t he s tud ies m a d e on th i s p rob l em , a m a r k e d s i m i l a r i t y in v e r - ba l a t t i t u d e s b e t w e e n d e l i n q u e n t a n d non -de l i nquen t g r o u p s h a s been r e p o r t e d . S i m p s o n t , on the ba s i s of a se r i e s of e x p e r i m e n t s in w h i c h he used a p s y c h o p h y s i c a l r a n k o r d e r t echnique , concluded t h a t t he a t t i - t udes of college s t u d e n t s a n d school t e a c h e r s t o w a r d a g r o u p of asoc ia l

* The author wishes to thank Dr. M. W. Richardson for his invaluable en- couragement and counsel in this study. This paper is a part of a dissertation accepted by the faculty of the Department of Psychology, The University of Chicago, for the degree of Doctor of Philosophy.

t Simpson, R. M. Attitudes towards the ten commandments. J. soc. Psychol., 1933, 4, 223-230.

Simpson, R. M. Attitudes of teachers and prisoners toward seriousness of criminal acts. J. cr/m. Law Criminol., 1934, 25, 76-83.

h 3 5

36 PSYCHOMETRIKA

acts were not essentially different from the attitudes of inmates in state prisons. According to the experimental findings of Thurstone* and of Dureat, who used similar techniques, college students place crimes according to their seriousness in practically the same rank or- der as found by Simpson for a prison population.

In al! of these studies, however, the methods used have permitted the acts to be placed on :Lhe scale only in terms of the stimulus differ- ences and have not permitted direct comparisons to be made, since the point at which the delinquents and non-delinquents begin to sepa- rate acts into "good" and "bad" forms of behavior was not deter- mined. I t is entirely possible that, although the two groups place the acts in approximately the same rank order on the scale, one group may consider many more of the acts as "good" than the other, thus leading to a real difference between the attitudes of the two groups. I t therefore becomes desirable to set up scaling techniques which make it possible to locate the origin of the scale for the delinquent and non- delinquent populations separately, and hence give a basis for making direct comparisons between the attitudes of the two groups.

The purpose of this study can now be stated more precisely as the determination of possible differences in the attitudes of delin- quents and non-delinquents toward a series of social acts, when meas- ured from an absolute point of neutrality. Although the experiment- al results may be of interest in themselves, the development of new and practical techniques for this and similar problems is considered to be of major importance.

Two techniques were used in this study. One is Horst's exten- sion of the method of paired comparisons.$ The method of paired comparisons was first suggested by Thurstonell for determining scale separations between a series of stimuli having affective value and stating the scale values of the stimuli in terms of their distance from an arbi t rary origin. Horst's procedure provides a method for locat- ing the origin (point of neutrality) in a scale and stating the scale values of the stimuli in the series in terms of their distance from this origin. The other technique is an extension of the psychophysical method of successive intervals. This method was developed in its simplest form by Thurstone, but was first described in the literature

*Thurstone, L. L. The method of paired camparisons for social values. J. abno~m. (soc.) Psychol., 1927, 20, 384-400.

-~ Durea, M. A. An experimental study of attitudes towards juvenile delin- quency. J. appl. Psychol., 1934, 17, 522-534.

$ Horst, A. P. A method for determining the absolute affective values of a series of stimulus situations. J. educ. Psychol., 1932, 23, 418-440.

II Thurstone, L. L. A law of comparative judgment. Psychol. Rev., 1927, 24, 273-286.

RUTH BISHOP 37

by Saffir.* The method of successive intervals differs essentially f rom Case V of the Law of Comparat ive Judgment , on which Horst 's meth- od is based, in tha t it is not assumed tha t the discriminal dispersion is the same for all of the stimuli.

Since the method of successive intervals, as developed by previ- ous investigators, gives only the scale separations of the stimuli in a given series, it was necessary to develop an extension of this method in order tha t the scale values be expressed in terms of their distance f rom the origin on the scale. This extension is somewhat similar in principle to tha t which Hors t developed for the method of paired com- parisons. In addition, one change was made in the calculation of the scale separations. In order to maximize the consistency of the scale values, it seemed preferable to use the centroid of the cumulative sigma values between any two successive categories for any given stimulus, ra ther than the cumulative sigma values for the successive categories. These values are given by the following equation, which is a variat ion of one of the formulas given by Kelleyt :

-- Z l ~ Z~

x : , ( l ) P2 ~ P t

where

is the centroid of a t runcated segment of a normal curve which has uni t s tandard deviation;

z, and z, are the ordinates enclosing the segment at the left and r ight respectively;

Pl and p, are the proportions of the area under the curve from the left to the ordinates with corresponding subscripts.

Hence the equations for the s tandard deviation and for the scale value of a given stimulus, using the method described by Saffir, be- come

~v~ - - , ( 2 )

( ~ Ey . - - ~ Zy~)

where

S~i : S ~ -~ , (3 )

ayj is the s tandard deviation of stimulus ] whose scale value is to be determined;

* Saffir, M. A. A comparative study of scales constructed by three psycho- physical methods. Psychometrika, 1937, 2, 179-191.

t Kelley, T. L. Statistical method. New York: Macmillan Company, 1924, equation 55, p. 101.

38 PSYCHOMETRIKA

a~ is the s tandard deviation of stimulus i whose scale value is a lready known ;

Y.:/,~ is the summation of the centroid values for the categories of st imulus i used in determining the value on the x-axis through which point (1) is passed;

~:2~ is the summation of the centroid values for the categories of st imulus i used in determining the value on the x-axis through which point (2) is passed;

~y,~ is the summation of the eentroid values for the categories of st imulus .i used in determining the value on the y-axis through which point (1) is passed;

Y~y~ is the summation of the eentroid values for the categories of s t imulus .i used in determining the value on the y-axis through which point (2) is passed;

nl is the number of categories common to stimuli i and i in- eluded in ~ c , i , ~ l j ;

"n.~ is the number of categories common to stimuli i and ] in- eluded in Y~x~, ~ y ~ ;

S,,~ is the scale value for st imulus .i ; S~ is the scale value for st imulus i .

As in the Hors t technique, the neutral i ty point for a series of stimuli, for which judgments have been obtained by the method of successive intervals, must be located by combining in some manner the subject ' s judgment about the location of the sum of any two of the items in the series with his judgment about the location on the psychological continuum of each of the stimuli considered separately. The technique developed in this paper differs f rom that of Horst , in which all of the unknowns are solved for simultaneously, in tha t the est imate of the neutra l i ty point and the calculation of the scale sepa- rat ions for the individual stimuli are divided into two separa te pro- eedures.

The origin of a scale derived from successive interval da ta may be determined if the problem is set up so that there is an odd number of categories, one of the categories being defined as neutral. As in the Hors t technique, the subjects are asked to make judgments about the affeetive value of the sum of any two individual stimuli in the series, making use of the descriptive categories employed in judging the in-

n (n - - - I ) dividual stimuli. Theoretically, there are summat ion equa-

2 tions, bu t as in any psychophysieal technique, s i tuat ions obviously close together and clearly some distance f rom the point of zero affect may be omitted f rom the series since they would yield too high or too low proport ions to give stable scale values or accurate es t imates of the point of neutral i ty.

R U T H BISHOP 3 9

The summation si tuations are first placed on the same a rb i t r a ry scale as the individual stimuli, using as a unit of measurement the s tandard deviation of some arbi t rar i ly selected individual stimulus. The point of neutra l i ty may then be est imated by the equation

M

K ~ L=I M + S ~ '

where

(4)

K is the mean est imate of the point of neutral i ty for the se- r ies ;

x-L is the centroid for the category defined as neutral for the summation si tuation L , which is a combination of individ- ual si tuations i and j ;

at, is the s tandard deviation for the summation situation L ; S~, is the a rb i t r a ry scale value for the summation situation L ; S~ is the a rb i t r a ry scale value for the individual si tuation

whose s tandard deviation has been used as the unit of meas- urement in est imating the a rb i t r a ry scale values of the summation situations. Obviously, if the s tandard deviation of the same si tuation is used as the unit of measurement fo r es t imat ing the scale separat ions of both the individual and summation situations, this term will vanish in equation (4), since in a rb i t r a ry scaling procedure it is cus tomary to

assign a scale value of zero to the si tuat ion whose s tandard deviation has been used as the unit of measurement .

M is the number of est imates made of the point of neutral i ty. By a simple l inear t ransformat ion, the a rb i t r a ry scale values may

now be expressed in te rms of their distances on the scale f rom this est imated point of neutral i ty, thus:

SKi = SI, j -{- K , (5) where

S~ is the scale value of the individual st imulus .i in te rms of its distance f rom the mean est imate of the point of neutra l i ty of the scale;

$I1~ is the a rb i t r a ry scale value for individual si tuation ] ; K is the value given by equation (4).

In construct ing the experimental scale to be used in this inves- t igat ion of the differences in the verbal at t i tudes of delinquents and non-delinquents, a series of fifteen social acts were selected which seemed to be within the experience of boys of eleven years or older, and which seemed to show a fair ly even graduat ion from asocial to prosocial acts. The acts were selected f rom a longer prel iminary list on the basis of the judgments of a large number of individuals who are interested in social issues and who have some unders tanding of

40 PSYCttOMETRIKA

the problems included in this study. The exper imenta l schedule is divided into fore" parts. Pa r t s I and 11, respectively, give the data fo r the difference equations used in Hors t ' s extension of the paired com- parison method and the extension of the method of successive inter- vals described in this paper. Pa r t s I I I and IV provide the data for the summat ion equations used in the two methods. In order to correct fo r t ime and space errors , two forms of the scale were con- structed. In Form Ii, the order of the items in each section and pa i r is the reverse of the order in Form I. As near ly as possible, the sched- ule is self-administrat ive, and each subject is given as much t ime as he likes in which to make his responses.

Two groups of subjects were used in this s tudy: a group of white male delinquents and a comparable group of white male non-delin- quents or controls. A delinquent was a rb i t r a r i l y defined as a person who was confined in a r e fo rma to ry or special "school" at the t ime the schedule was given. A control was a rb i t r a r i ly defined as a person who was a t tending a public or pr iva te school, or one who was a member of some group in a local communi ty center, a t the t ime the schedule was given, and who did not have a court delinquency record as f a r as could be determined by the teachers and officials in the schools and communi ty centers. All of the subjects were selected f rom the Chica- go area, and had sixth grade or be t te r reading achievement level.

Since at t i tudes are general ly considered to be a funct ion of the total exper ient ia l s i tuat ion of an individual, it seemed desirable to a t t empt to select the respective del inquent and control populations in such a maturer tha t they would be fa i r ly comparable wi th respect to the i r d is t r ibut ion in certain variables. A difference between the pro- port ions of the two groups in any ca tegory in which a comparison was being made was regarded as significantly g rea t e r than zero when p, - - p~

3.00 .* No significant differences were found in the t r

exper imenta l propor t ions of delinquents and controls who (1) had average or be t te r ra t ings on a s tandard intelligence test , (2) had eighth grade or be t te r reading achievement tes t scores, (3) belonged to ma jo r nat ional i ty groups, (4) came f rom high or low delinquency areas as defined by Shaw and McKay. t There was, however, a sig- nificant difference between the two groups as to age and education, the delinquents being considerably older and having less educat ion than the controls. Nevertheless, this was not considered serious since

* Guilford, J. P. Psychometric methods. New York: McGraw-Hill, 1936. Pp. 59-63.

t Shaw, Clifford & MeKay, H. D. Juvenile delinquency and urban areas. (Unpublished monograph prepared at the Institute for Juvenile Research. Chi- e-tgo, 1938.}

R U T H B1SHOP 11

there were only a few items, and no individual stimuli, which showed a significant difference in the proportions of subjects making a given response in either the delinquent or control groups when the differ- ences in these two selective factors were as large as possible.

Since the responses made in the two forms were found to be high- ly consistent, and since both the delinquent and control groups seemed to be fa i r ly homogeneous, it was decided to consider all the 251 delin- quents in one group and all the 318 controls in another for scaling purposes.*

The scale values for the delinquent and control groups, as deter- mined by the extensions of the psyehophysical methods of paired com- parisons and successive intervals previously described in this paper, are shown in Table I. The relationship between the scale values for these two groups is shown graphically in Figures 1 and 2.

S t i m u l u s

T A B L E I

C o m p a r i s o n o f Sca le V a l u e s

S a v e a d r o w n i n g c h i l d ........................ P u t o u t a f i re b e f o r e i t b e c o m e s

s e r i o u s .............................................. 3.06 H e l p a n old l a d y a c r o s s t h e s t r e e t .... 2.63 M a k e g o o d g r a d e s in schoo l ................ 2.33 Go to c h u r c h r e g u l a r l y ........................ 1.99 Te l l s t r a n g e r h o w to f ind a c e r t a i n

a d d r e s s . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.96 L i s t e n t o r a d i o . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67 G o to m o v i e . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64 W a l k f o u r a b r e a s t e n t h e s i d e w a l k .... - - .83

A b s o l u t e Sca le V a l u e s

H o p a t r u c k f o r a r i d e ........................ - - 1 . 4 0 S h o o t c r a p s (d ice ) ................................ - - 1 . 4 2 K i c k a d o g a s l e e p on t h e s i d e w a l k .... - - 2 . 3 5 S t e a l a b a n a n a f r o m a f r u i t s t a n d .... - - 2 . 4 3 S t e a l a b i cyc l e ...................................... - - 3 . 6 8 S n a t c h a p u r s e ...................................... - - 4 . 3 3

H o r s t E x t e n s i o n o f T e c h n i q u e S u c c e s s i v e I n t e r v a l s

Del . Con t . Del . Con t .

4.04 3.66 2.48 3.14

2.86 2.18 2.61 2.38 1.95 2.18 2.38 1.97 2.27 1.84 2.23 2.19

1.58 1.43 1.49 .89 1.07 1.05 .92 .90 .86

- - . 4 7 - - . 5 3 - - .33 - - 1 . 6 3 - - .89 - - 1 . 5 4 - - 1 . 4 8 - - .88 - - 1 . 2 7 - - 2 . 4 7 - - 1 . 5 3 - - 1 . 9 2 - - 2 . 1 2 - - 1 . 3 5 - - 1 . 5 9 - - 3 . 4 3 - - 2 . 5 5 - - 3 . 0 4 - - 3 . 6 2 - - 2 . 9 9 - - 3 . 1 7

In order to make a direct comparison between the estimates of the point of neutra l i ty obtained for the delinquent and control groups, it is necessary to use a method based on all scale values in the two ex-

* T h e a u t h o r w i s h e s to t h a n k D r . A. W . B r o w n a n d h i s s t u d e n t s f o r a d m i n - i s t e r i n g t h e s c h e d u l e s to t h e a d u l t d e l i n q u e n t g r o u p s .

/12 P S Y C H O M E T R I K A

4 O

DEL. 210

-40~.2.0 - 2 £ 0

~ 4 T ~ s 2.0 4.0

FIG. 1. - - Comparison of the uncorrected scale values for delinquents and controls o b t M n e d b y the extension of the method of paired comparison.

4,(

DEL

-4D -2.0 ~CON~ROL$ ) 2.0 4.0

FIG. 2. - - Comparison of the uncorrected scale values for delinquents and controls obtained by the extension of the method of successive intervals,

perimental series which are being compared. In the method described below, it is assumed that both the delinquent and control groups have used the same psychological unit of measurement in making their judgments . Hence, the dispersions of the scale values in the two ex- perimental populations may be equated simply by mult iplying each scale value m the group having the smaller dispersion by the ratio of the dispersion of the second group to this one. The difference in the points of neutrality for the two groups may then be defined as the residual resulting from the minimization of the sums of the squares of the discrepancies in the corrected scale values of the respective

RUTH BISHOP 43

stimuli for the two experimental populations. The equation for this difference is developed in the conventional manner. Let

Scj be the scale value for stimulus j determined on a control population ;

S~i be the scale value for stimulus j determined on a delinquent population;

a be the residual which results when the sums of the squares of the discrepancies in the scale values have been made a minimum ;

T be the number of stimuli in the experimental series.

Then the equation to be minimized is

T

V = E (Scj - - S~ + a )~ . (6) j = l

Expanding this equation, we have

T T T T T

V : ~ S2cj "-~ ~ S2~j . ~ T a 2 ~ 2 ~ S(9 S,lj ~- 2 a ~ S~j ~ 2 a ~ Sjj . j = l j - 1 j - l J-1 i 1

(7) Different iat ing this expression with respect to a , we have

d V r r - - 2 Ta -~ 2 ~, S~i - - 2 ~ S , ~ . (8)

d a j =-----i j=~

Set t ing this equation equal to zero, and solving for a , we have

T T

E S,Ji E Sci j = l ~=1

a -- (9) T T

This expression also may be wri t ten

a = ~ - - ~ , (10)

where

¢ is the estimated difference in the points of neutra l i ty on the two scales being compared;

xd is the mean of all the scale values in the experimental series as determined for a delinquent group;

~:c is the mean of all the scale values in the experimental series as determined for a control group.

For this par t icular sample, the dispersions of the scale values fo r the delinquent and control groups were approximately the same, the rat io between the grea ter and smaller dispersions of scale values being 1.15 for each of the methods. Neither was there a significant

4 l PSYCHOMETRIKA

difference in the 1)oint of neutral i ty for the two experimental gvoul)s, a being less than 0.1 for each of the methods.

When comparisons are made between the scale values for the tw(, experimental g'VOUl)S, as determined by either of the scaling meth- ods, the following conclusions may be drawn:

(1) There is a linear relationship between the scale values of the delinquent and control groups. From this relationship, it may be deduced that both of the groups were able to dif- ferent iate equally well between the items in the experiment- al scale.

S ~o .

p.c. - ' L ' ; -,t..,O ' Z ~

/ I -4.q

FIG. 3. - - Compar i son of the cor rec ted scale va lues for de l inquen t s ob ta ined by the ex t enshm of t im methods of successive i n t e r v a l s and p a i r e d compar i son .

4 ~

SJ.~C

,P.C. -40 ~~-20 o 2D 4 D

/ -

FIG. 4. - - Compar i son of the cor rec ted scale va lues f o r cont ro l s ob ta ined by the ex tens ion of the me thods of successive i n t e r w d s and pa i r ed campar i son .

RUTH BISHOP 45

(2) The foregoing conclusion is fur ther supported by the fact that the dispersion of the individual stimuli along the psy- chological continuum is essentially the same for the two groups.

(3) The delinquent and control groups locate the point of neu- trality at approximately the same place in the experimental series.

From Figures 3 and 4, it may be seen that the scale values given by the two techniques used in this study have a linear relationship for both the delinquent and control populations. The fit is quite close except for the item regarding church attendance. This item gave less consistent results than any of the other stimuli in the experimental series. There are at least three possible explanations for this dis- crepancy. One possibility is that the item cannot be represented adequately in one dimension*; another is that the subjects react to the item differentially according to the judgment they are asked to make; while a third possibility is that the discrepancy is simply due to a sampling error. While the present data do not permit a complete answer to this question, the third explanation possibly may be cor- rect since the item is found to satisfy a linear relationship between delinquents and controls within the same technique, and the scale values given by the two techniques for the control population are considerably more consistent than for the delinquent population. There is a possibility, however, that the result is due in part to the different kinds of judgments which the subjects were asked to make in the two techniques.

In conclusion, it may be pointed out that although no significant differences were demonstrated in the verbal attitudes of the two ex- perimental groups used in this study, the techniques would seem to be of value in attacking similar problems. Horst's technique and the extension of the method of successive intervals developed in this pa- per give essentially the same results. The latter method, however, requires considerably less time both in the collection of the experi- mental data and in the calculation of the scale values, and, for this reason, it is probably preferable for general use.

* Richardson, M. W. Multidimensional psychophysics. Psychol. Bull., 1938, 35, 659-660.