THE EMOTIONS AND THE DIMENSIONS OF DISCRIMINATION AMONG THEM
IN DAILY LIFE

by
Ilan Shalif

(=====Back to the theory
To results II. ---->
-25-

THE METHODS

The subjects

The subjects were 202 resident students (101 female and 101 male) who were not studying in the department of psychology. All had been of good eye sight and were fluent in Hebrew - the language of the verbal materials.

The materials

1) 105 photographs of facial expressions from three main sources:

a) 48 unmanipulated mixtures of emotion - the entire set of items of the Szondi test (Szondi, 1947; Szondi et al., 1959). These mixed expressions cover a very wide spectrum (not intended by Szondi) as was found by Vargha (1979) and by Shalif (1980 - unpublished).

b) 33 manipulated facial expressions of basic emotions taken from Izard (1971, 1977) and Ekman & Friesen (1975). For happiness, surprise, fear, anger, disgust and contempt (the last included for control purposes) - 4 items for each; for sadness, three items: for distress (a version of sadness?), interest and shame (the last included for control purposes) - two for each.

c) 24 manipulated mixtures of basic emotions taken from Ekman & Friesen (1975).

All the photographs were black and white and of the same size. The photographs were arranged in two parallel sets (of 105). In one of the sets, each photograph was mounted on a white card measuring 7 by 10 centimeters and all were attached to a Rolex revolving drum of telephone-number-cards. In the second set the photographs were mounted on plates, in groups of 7 to 12. The detailed list of the origin of each photograph and its place on the drum and the plates is in appendix number one.

2) 148 words in the Hebrew language, printed in alphabetic order in four columns of one page. To the left of each word was the following scale of six grades and five intermediate grades (the lines between the numbers): _1_|_2_|_3_|_4_|_5_|_6_ .

-26-

The words were taken from various studies and were translated into Hebrew. The list consisted of:

a) 96 words of indubitably emotional content to be analyzed by multidimensional scaling.

b) 35 words naming dimensions and variables found in the publications of previous research as causes for systematic changes of item content to non emotional domains.

c) 17 words of emotion that relate to the 96-word-list but differ in a systematic way (intensity or frequency), which were included for control purposes.

THE PROCEDURE

Contacting the subjects

In the end-of-academic-year examinations period (1986-7), posters were put up at students' dormitories of two neighboring universities, inviting students to take part in a doctoral research. Anonymity, fees that double the minimum and explanations were assured.

The examiner contacted students from several buildings and made efforts to include in the study a high percentage of each. (This and the content of the posters was arranged in order to minimize the effects of self selection.)

Data collecting procedure

Data was collected in the subjects' room (or an unoccupied room of a neighbor) at a table facing the wall or a window. This location was chosen for two main reasons:

1) to decrease the deviation of the emotional mood of the subjects from their mood of that day and hour;

2) to decrease the percentage of refusers - (see results).

At the beginning of the session the subject was told that anonymity was assured and that explanations would be given at the end. Then the subject was asked which was his/her dominant hand - left-handers were sat to the left of the examiner and right-handers to his right.

-27-

The first task:

First, the Rolex drum with the 105 photographs was taken out of its box (without any visible photograph). Then the subject was given a booklet of instructions opened at those for the first task. A page with 105 numbered scales like this: _1_|_2_|_3_|_4_|_5_|_6_ was given as well. At the top and at the bottom of the page, verbal labels of the six grades were written.

Then, the examiner read aloud the following instructions for the first task while the subject followed them with his eyes: "Photographs of facial expressions (some of the same persons) will be presented. You are asked to mark - after a first (short) look - one of the grades of the scale according to the measure in which the feeling or the emotion expressed in the photograph is compatible with (i.e. near or like) what you are feeling now.

The grades ofthe scale are:

1) completely incompatible - the least near to you;
2) incompatible;
3) incompatible a little more then compatible;
4) a little more compatible than not compatible;
5) compatible;
6) very compatible - the nearest.

(When it is hard to decide which of two adjacent grades to mark - one can mark the partition between them as an intermediate grade.) However, each scale must be marked as skipping is not permitted."

After the examiner finished reading the instructions, he added: "I want to emphasize that the use of grade 1 is only when (the photograph) is entirely not compatible - i.e. entirely contradicting what you are feeling." This point was added after subjects in pretests tended to give this grade to the majority of the artificial basic emotions of intense negative content.

-28-

The examiner then exposed the first photograph on the drum and at the same time asked if the instructions were understood. To subjects who delayed more than 30 seconds before marking the first item, the examiner repeated the second sentence of the instructions with emphasize on the words "after a first (short) look" This was repeated with the following items until the time for an item was less then 30 seconds.

After the first item, the examiner asked if the content of the label of the marked grade was correct (i.e. if completely incompatible or incompatible or any of the others - whatever he chose, depict him correctly - to make sure that the subject used the grades correctly). The drum was revolved to change the exposed item while the subject marked the scale for the previous one.

The second task:

At the end of photograph 105 (after about 10 to 20 minutes), the drum and the page of the 105 scales were taken away and replaced by a page containing the 148 words with their scales.

Then the examiner read the instruction for the second task which were nearly the same as those of the first task. The only difference was that here he referred to the 148 words.

The third task:

This task consisted of a Q-SORT procedure for the 105 photographs (materials - 1). In this task the subject was asked to order the items of each plate of items to prearranged grades. The grades were arranged around the median of each plate that was 3.5, the top score was 6, and the bottom was 1 (the same as in the previous tasks). In plates of 8 and 9 items there were two places for the grades 3 and 4, in the plate of 12 - two places for each grade. This procedure was followed in order to overcome subject's tendency to scale the majority of the items that are artificial expressions of negative emotions to the same lowest grade - materials: 1b, 1c. It was also intended to reconstruct the scaling procedure of Shalif, 1980 for 48 items of mixed and unmanipulated expression of materials: 1a.

-29-

The ordering - the scoring of the items on the plates was done by means of covering each item with white cards on each of which was printed a grade. They were given to all the subjects in a fixed order - 6, 1, 5, 2, 4, 3, (and 3.5 in plates with an odd number of items). On the card of grade 6 was written "very appropriate", on that of grade 1 was printed "very inappropriate".

The fourth task:

At the end of task three came task four that was nearly identical to the Shalif (1980) study. It consisted of ordering the 48 items of materials: 1a - the Szonti-test items, according to subject's liking and not according to appropriateness. (This use of the items is like that of the Szondi test and was included for control purposes.)

After the fourth task, the subject was given the opportunity to ask about the study and to ventilate her/his feelings. Then the examiner asked about any unusual occurrences that were observed during the session.

The computation of subjects' dimensional scores

The interpretation of the content of dimensions which are the result of multidimensional scaling is usually very subjective and open to experimental bias. In factor analysis studies, the same problem is overcome - very seldom - when subject's factor-scores are related to other variables that are not included in the factor analysis.

In this study one of the central groups of items is the 48 photographs of unmanipulated emotional expressions. As their basic emotional composition is not known, the direct interpretation of the dimensions and the directions of the mathematical solution by the aid of the 48 items is impossible.

The interpretation of the directions and the dimensions of the multidimensional scaling of the 48 (materials: 1a) items and to a large extent even that of the 96 emotional words (materials 2a) is based on correlations between subject's dimensional scores and other variables (of verbal and non verbal content).

-30-

The computation of the dimensional scores for the 48 items of task one (and afterwards those of the 96 words of task two) was done thus:

a) The output of the multidimensional analysis of the inter correlation matrix of the items (S.S.A.-I) of the 48 items (and later that of the 96 words) - the matrix of dimensional coordinates of the items - was linearly transformed so that the mean of the dimension coordinates of the items for each dimension is equal to "0".

(Different versions of multidimensional scaling techniques give the coordinate of items on the dimensions in an unstandardized way with regard to where they put the origin of the space and the range of the coordinates. The most meaningful point of a multidimensional scaling model is the centroid of the mathematical solution. Therefore, the said linear transformation was applied.)

b) The second step is the computation of each of the subjects' dimensional scores for each of the dimensions (for each item group by itself). The computation is like the multiplication of a line with a matrix . For each dimension, each subject score for an item is multiplied by the item coordinate for this dimension. The results of the multiplications are totalled and divided by the number of items. This result is the subject's score for that dimension.

As a result of this calculation, a subject who gave all the items the same score, will have the dimension score of "0" for all the dimensions of this item group. The higher a subject's item scores are for items which are clearly in one direction of the multidimensional space of the mathematical solution, the more his dimensional scores lie in that direction.

(The following two references are of similar applications of this approach: 1) Takane et al., 1977 - based on INDSCAL; 2) Shalif et al.,1981 - based on Guttman-Lingoes non metric program series, Lingoes, 1973.)

-31-

c) Afterwards the dimensional scores of the subjects for the multidimensional scaling of the 48 items of materials: 1a, and those of the 96 words of emotion of materials: 2a (that were computed as well), were related to each other (by computation of correlations) and to all other subjects' item scores and to any other score based on them.

-32-

RESULTS

The subjects

The subjects - 101 male and 101 female - were single, resident students of two near-by universities. About 90% of them were undergraduates, the others were students of preparatory classes or graduate students. About 20% of the subjects responded a poster-request for volunteers - the others were contacted by the examiner and were roommates or neighbors of the above 20%. Eighty (80%) of those approached by the examiner consented to take part in the study.

As the examiner promised anonymity to the subjects, as the tasks and the data-collecting needed his full attention, and as the study is not about the variance among people - there was no systematic data collection except that of the item-scores.

It is worth mentioning, however, that the subjects were from diverse faculties and departments. They differed in their backgrounds, involvement with the task and their moods, both before the session and during it. Some of the subjects enjoyed their tasks. Others responded with tension to different sections of the study. Some subjects had difficulty in choosing the scores for the items, others were very hasty.

The median length of time per subject was about 45 minutes, the longest was one and a half hours and the shortest about 35 minutes. The majority of the subjects were more interested in the rationale of the study than in their fee.

The items' scores

As a result of missing answers, the number of scores per item varied between 199 and 202 (the number of subjects). During the free scaling procedure, the subjects rarely used the intermediate scores. They also preferred the lower scores for most of the items of artificial facial expressions of strong and negative emotions. One can see in Table 1 on the next page, the distribution of scores of the groups of items from the free scaling procedure.

-33-

Table No. 1: The distribution of scores of the groups of items from the free scaling procedure and the percentage of each score from the groups' total. 
----------------------------------------------------------------------------
   Items    /Scores|  1 |1.5|  2 |2.5|  3 |3.5|  4 |4.5|  5 |5.5| 6  |Median
-------------|-----|----|---|----|---|----|---|----|---|----|---|----|------
1a - 48 Mixed|Freq.| 989| 35|2550|103|2440|140|2097| 73|1072| 18|  77| 3.02
expressions  |  %  |10.2|0.4|26.3|1.1|25.2|1.4|21.6|0.8|11.1|0.2| 1.8|  **
-------------|-----|----|---|----|---|----|---|----|---|----|---|----|------
1b - 33 Basic|Freq.|1841| 80|1947| 68|1112| 96| 834| 45| 491| 12| 139| 2.27
emotions     |  %  |27.6|1.2|29.2|1.0|16.7|1.4|12.5|0.7| 7.4|0.2| 2.1|  **
-------------|-----|----|---|----|---|----|---|----|---|----|---|-----------
1c - 24 Mixed|Freq.|1195| 81|1553| 71| 914| 67| 621| 31| 260|  5|  50| 2.28
basic emotio.|  %  |24.6|1.7|32.0|1.4|18.9|1.4|12.8|0.6| 5.4|0.1| 1.0|  **
-------------|-----|----|---|----|---|----|---|----|---|----|---|----|------
2 - 148 Words|Freq.|4102|210|7007|298|5082|505|6046|427|4555|140|1440| 3.12
( a + b + c )|  %  |13.8|0.7|23.5|1.0|17.0|1.7|20.3|1.4|15.3|0.5| 4.8|  **
---------------------------------------------------------------------------
**100% For that row

As the distributions of the scores of too many items is skewed, the non-parametric statistical procedures were preferred.

The validity of items and the reliability of their scores

As the procedure was very long, retesting of items was not done. As a substitution and "first approximation", the scores of the same items of the two procedures are the basis for the estimation of the lower boundary of the items' reliability.

The 48 unmanipulated mixtures

The range of Spearman correlation coefficients between the scores of the unmanipulated 48 emotional mixtures of materials: 1\a, from the two procedures, is between 0.59 (p<.001) and 0.15 (p<0.017, one tail). The median is 0.30 (p<0.001).

For each of these 48 items, the highest correlations between it and the 33 basic emotions - materials: 1b - are all significant. 46 of them are of 0.001 two tail or more, one is of 0.003 and one of 0.015 (r=-017).

One can deduce from the above findings that items of this group functioned in this study as a relatively reliable measure of the emotional feelings of the subjects.

-34-

The scores of the sub-groups of basic emotions

The scores of the items of the 9 basic emotions were added to create subgroup scores. This was done for each of the two procedures and then the matrix of correlations for all the subgroup scores were computed and are demonstrated on the following table 2. 

Table No. 2: Matrix of correlations among the subgroups of basic emotions of
each procedure and between two procedures.
-----------------------------------------------------------------------------
Free scaling procedure       |         Q-Sort procedure
-----------------------------------------------------------------------------
     | H | S | I | F | A | D | S | C | S | H | S | I | F | A | D | S | C |S
     | A | U | N | E | N | I | A | O | H | A | U | N | E | N | I | A | O |H
     | P | R | T | A | G | S | D | N | A | P | R | T | A | G | S | D | N |A
     | P | P | E | R | E | G | N | T | M | P | P | E | R | E | G | N | T |M
     | I | R | R |   | R | U | E | E | E | I | R | R |   | R | U | E | E |E
     | N | I | E |   |   | S | S | M |   | R | I | E |   |   | S | S | M |
F|   | E | S | S |   |   | T | S | P |   | E | S | S |   |   | T | S | P |
R|   | SS| E | T |   |   |   |   | T |   | SS| E | T |   |   |   |   | T |
E|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---
E|HAP  **  57  40  05  00  05 -13  21 -02|*63  40| 03 -14 -20  02 -50 -21 -36
 |SUR  57  **  50  46  39  42  25  38  13| 30 *44|-03  06 -07  05 -31 -36 -32
G|INT  40  50  **  26  22  28  24  34  31| 13  17|*13 -07 -07  00 -14 -19 -13
S|FEA  05  46  26  **  83  73  66  54  28|-10  03|-07 *30  15 -03  07 -24 -10
C|ANG  00  39  22  83  **  73  70  57  31|-22 -01|-04  20 *21  01  08 -17  00
A|DIS  05  42  28  73  73  **  67  54  33|-17 -10|-02  15  01 *21  07 -10 -04
L|SAD -13  25  24  66  70  67  **  52  39|-38 -19|-12  25  09 -02 *37 -10  04
I|CON  21  38  34  54  57  54  52  **  50|-09 -03|-07 -03  04 -02 -02*(07) 04
N|SHA -02  13  31  28  31  33  39  50  **|-29 -15|-14 -05  03 -07  24  09 *37
G|==================================| HAP| **  37| 05 -19 -28 -12 -54 -26 -37
                                   Q| SUR| 37  **| 13  00 -15 -19 -52 -36 -36
** Correlation of 1.00 between a   \| INT| 05  13| ** -13 -04 -10 -15 -11 -15
   subgroup and itself             S| FEA|-19  00|-13  **  06 -16  08 -19 -25
                                   O| ANG|-28 -15|-04  06  ** -11 -04 -03  03
 * The correlation between the     R| DIS|-12 -19|-10 -16 -11  ** -07  04 -13
   subgroup's scores of the two    T| SAD|-54 -52|-15  08 -04 -07  ** -03  28
   procedures                       | CON|-26 -36|-11 -19 -03  04 -03  **  30
                                    | SHA|-37 -36|-15 -25  03 -13  28  30  **
                                    =========================================

The correlations between the subgroups' two procedures are not homogeneous. One can see in Table two that the correlation of happiness (0.63) is the highest. It can also be seen that those of interest and contempt are very low. That of interest is barely significant (0.028 - one tail) and that of contempt - r=0.07 - is not significant at all.

It seems that for many items and subgroups of artificial basic emotions the two procedures led the subjects to focus their attention on different aspects of the items.

-35-

Content validity of the 9 sub-groups of basic emotions

The larger the inter-sub-group correlations and the smaller the correlations of the items (and the subgroup score) with other items or (subgroups), the more a subgroup score can be relied on when one interprets the content of dimensions of the multidimensional scaling.

The majority of the correlations between subgroup scores of the free scaling of the first task - as one can see in Table 3 - are significant (30 out of 36 - p<0.003, two tail or more). The median correlation is +0.385, p<0.001.
This phenomenon is mainly the result of subjects' general response set of tendency to denial to the items of the negative emotions: fear, anger, sadness, disgust; and their too willing acceptance of the three positive emotions (happiness, surprise, interest).

Those sets of response are not symmetric as the inter-correlation of the three positive emotions is relatively lower (there is more discrimination among them). Another set - that of general trend of yes-saying or no-saying - is responsible for the positive correlations between the items and the sub-groups of positive and negative emotions.

The inter-sub-group correlations of the Q-Sort procedure are much more differentiated. But even in this procedure where item scores are in competition, there are still positive correlations and with three of the items the level of significance is of p<0.001 . Only 15 of the 36 are negatives with significance of 0.05 or more.

The three positive sub-groups

Table 3 of the free-grading procedure and Table 4 of the Q-Sort one, show the matrices of correlations among the 10 items of the three positive basic emotions of the study. From those correlations one can see that the border between the contents of each subgroup is not sharply defined. Note the difference between the cohesiveness of the groups.

-36- 

Table No. 3: The matrix of correlations among the 10 items of the  sub-
groups of happiness, surprise, interest from the free grading procedure
---------------------------------**-----------------------**-----------
|Subgroup|      Happiness        ||       Surprise        ||  Interest
|--------|-----------------------||-----------------------||-----------
|Item No.|  1  |  9  | 17  | 25  ||  2  | 10  | 18  | 26  ||  8  |  24
|--------|-----|-----|-----|-----||-----|-----|-----|-----||-----|-----
| H |  1 | 1.0 | 0.64| 0.59| 0.56|| 0.51| 0.31| 0.36| 0.46|| 0.37| 0.17
| a |----|-----|-----|-----|-----||-----|-----|-----|-----||-----|-----
| p |  9 | 0.64| 1.0 | 0.78| 0.78|| 0.42| 0.42| 0.40| 0.52|| 0.40| 0.19
| p |----|-----|-----|-----|-----||-----|-----|-----|-----||-----|-----
| i | 17 | 0.59| 0.78| 1.0 | 0.80|| 0.35| 0.31| 0.41| 0.45|| 0.34| 0.18
| ne|----|-----|-----|-----|-----||-----|-----|-----|-----||-----|-----
| ss| 25 | 0.56| 0.78| 0.80| 1.0 || 0.38| 0.37| 0.43| 0.60|| 0.31| 0.22
|========|=====|=====|=====|=====||=====|=====|=====|=====||=====|=====
| S |  2 | *** | *** | *** | *** || 1.0 | 0.60| 0.68| 0.46|| 0.42| 0.15
| u |----|-----|-----|-----|-----||-----|-----|-----|-----||-----|-----
| r | 10 | *** | *** | *** | *** || 0.60| 1.0 | 0.59| 0.49|| 0.34| 0.19
| p |----|-----|-----|-----|-----||-----|-----|-----|-----||-----|-----
| r | 18 | *** | *** | *** | *** || 0.68| 0.59| 1.0 | 0.51|| 0.50| 0.30
| i |----|-----|-----|-----|-----||-----|-----|-----|-----||-----|-----
| se| 26 | *** | *** | *** | *** || 0.46| 0.49| 0.51| 1.0 || 0.40| 0.25
|========|=====|=====|=====|=====||=====|=====|=====|=====||=====|=====
|Int|  8 | *** | *** | *** | *** || *** | *** | *** | *** || 1.0 | 0.23
|ere|----|-----|-----|-----|-----||-----|-----|-----|-----||-----|-----
|st | 24 | *** | *** | *** | *** || *** | *** | *** | *** || 0.23| 1.0
=======================================================================

Table No. 4: The matrix of correlations among the 10 items of the sub-
groups of happiness, surprise, interest from the Q-SORT procedure
---------------------------------**-----------------------**-----------
|Subgroup|      Happiness        ||       Surprise        ||  Interest
|--------|-----------------------||-----------------------||-----------
|Item No.|  1  |  9  | 17  | 25  ||  2  | 10  | 18  | 26  ||  8  |  24
|--------|-----|-----|-----|-----||-----|-----|-----|-----||-----|-----
| H |  1 | 1.0 | 0.47| 0.56| 0.49|| 0.35| 0.06| 0.16| 0.35|| 0.12|0.12
| a |----|-----|-----|-----|-----||-----|-----|-----|-----||-----|----
| p |  9 | 0.47| 1.0 | 0.64| 0.67|| 0.43| 0.13| 0.24| 0.31||-0.07|0.06
| p |----|-----|-----|-----|-----||-----|-----|-----|-----||-----|----
| i | 17 | 0.56| 0.64| 1.0 | 0.63|| 0.29| 0.01|-0.03| 0.33|| 0.07|0.09
| ne|----|-----|-----|-----|-----||-----|-----|-----|-----||-----|----
| ss| 25 | 0.49| 0.67| 0.63| 1.0 || 0.42| 0.15| 0.21| 0.26|| 0.07|-.01
|========|=====|=====|=====|=====||=====|=====|=====|=====||=====|====
| S |  2 | *** | *** | *** | *** || 1.0 | 0.19| 0.40| 0.42|| 0.28|0.01
| u |----|-----|-----|-----|-----||-----|-----|-----|-----||-----|----
| r | 10 | *** | *** | *** | *** || 0.19| 1.0 | 0.36| 0.20||-0.04|-.06
| p |----|-----|-----|-----|-----||-----|-----|-----|-----||-----|----
| r | 18 | *** | *** | *** | *** || 0.40| 0.36| 1.0 | 0.26|| 0.06|-.02
| i |----|-----|-----|-----|-----||-----|-----|-----|-----||-----|----
| se| 26 | *** | *** | *** | *** || 0.42| 0.20| 0.26| 1.0 || 0.31|-.04
|========|=====|=====|=====|=====||=====|=====|=====|=====||=====|====
|Int|  8 | *** | *** | *** | *** || *** | *** | *** | *** || 1.0 |-.07
|ere|----|-----|-----|-----|-----||-----|-----|-----|-----||-----|----
|st | 24 | *** | *** | *** | *** || *** | *** | *** | *** ||-0.07| 1.0
=======================================================================

-37-

As can be seen in tables 3 and 4, there is more differentiation between those subgroups in the free scaling than in the Q-SORT - as in Tables no.8 and no.9 - which show the correlations between the subgroups and the dimensions of the scaling analysis.

The validity of happiness

The highest correlations (of the free grading - task 1) of items 1, 9, 17, are those that are between them; the next higher ones are with the four items of surprise and item 8 of interest; those of item 25 is like the others except its high correlation of 0.60 with item 26 of surprise. In the Q-SORT procedure the three highest correlations of the happiness items are those that are among themselves. It seems that this subgroup has a relatively high validity.

The validity of surprise

In task 1, the highest three correlations of items 10 and 18 are with items of surprise; for item 2 those three include a correlation with item 1 of happiness and only the fourth is with item 26 of surprise; in item 26 the two highest correlations are with items of happiness and only afterwards those of surprise. In the Q-Sort task the findings are nearly the same.

These findings suggest that the validity of the surprise items is not very high. It seems that item 26 is a mixture of happiness and surprise - more happiness than surprise. It might be that all the items of this sub group are mixtures of surprise with the content of the items of the sub group of happiness.

The validity of interest

This subgroup consists of two items only - no. 8 and no. 24. In task 1 the eight highest correlations of item 8 are with the items of happiness and surprise (only then with item 24). For item 24 the highest six are with two items of surprise, and with items of shame, sadness, distress, and anger - one of each subgroup. Only the seventh is with item 8 of its sub group. In the Q-Sort procedure the correlation between item 8 and 24 is not significant (r=-0.07). It seems that the content validity of this subgroup is very small, if it exists at all. The content of item no. 24 seems to be the most deviant.

-38-

The validity of the Four negative basic emotions

The basic list of this study consists of the above three positive emotions and the four negative ones - namely: fear, anger, disgust, and sadness. A strong tendency of the subjects to deny the negative emotions caused the concentration of the scores of the items of those four sub groups at the lower side of the grading-scale and relatively very high correlations among the 15 items of those subgroups. Therefore, the contribution of the data of the first task to the examination of the content validity of those subgroups, is small.

Table 5 of the free grading (task 1) and Table 6 (of the Q-Sort) that follow, contain the matrices of correlations of the 15 items of those sub groups and the two items of distress (included for control). The majority of those items were graded by about 75% of the subjects with the lowest grade they chose to use (1 or 2).

***********Tables No. 5 and 6 about here********

The effect of "the-end-of-the-scale" of task 1 decreases to alarge extent the value of the data of this task - for those subgroups of artificial facial expressions of basic emotions.

The validity of fear

In the Q-Sort (table 6) the highest correlations of the items of fear are between themselves, except one which is with item 4 of anger which is apparently a mixture of fear and anger. Those findings indicate that this subgroup has good content validity.

-39- 

Table No. 5: Matrix of correlations  of the items of the subgroups fear,
anger, disgust, sadness and distress -  of the free gradings.
-----------------------------**-------------------**--------------------
|Subgroup|       Fear        ||       Anger       ||      Disgust      |
|--------|-------------------||-------------------||-------------------|
|Item No.| 3  | 14 | 19 | 31 || 4  | 12 | 20 | 28 || 5  | 15 | 21 | 32 |
|--------|----|----|----|----||----|----|----|----||----|----|----|----|
| F |  3 |1.0 |0.53|0.59|0.61||0.75|0.53|0.53|0.54||0.54|0.44|0.54|0.44|
| e |----|----|----|----|----||----|----|----|----||----|----|----|----|
| a | 14 |0.53|1.0 |0.63|0.61||0.55|0.51|0.57|0.56||0.40|0.54|0.57|0.61|
| r |----|----|----|----|----||----|----|----|----||----|----|----|----|
|   | 19 |0.59|0.63|1.0 |0.60||0.62|0.55|0.76|0.66||0.55|0.44|0.64|0.44|
|   |----|----|----|----|----||----|----|----|----||----|----|----|----|
|   | 31 |0.61|0.61|0.60|1.0 ||0.56|0.54|0.57|0.62||0.46|0.40|0.54|0.45|
|========|====|====|====|====||====|====|====|====||====|====|====|====|
| A |  4 | ** | ** | ** | ** ||1.0 |0.63|0.57|0.59||0.62|0.41|0.54|0.42|
| n |----|----|----|----|----||----|----|----|----||----|----|----|----|
| g | 12 | ** | ** | ** | ** ||0.63|1.0 |0.63|0.65||0.59|0.43|0.62|0.33|
| e |----|----|----|----|----||----|----|----|----||----|----|----|----|
| r | 20 | ** | ** | ** | ** ||0.57|0.63|1.0 |0.65||0.55|0.49|0.70|0.36|
|   |----|----|----|----|----||----|----|----|----||----|----|----|----|
|   | 28 | ** | ** | ** | ** ||0.59|0.65|0.65|1.0 ||0.48|0.41|0.59|0.41|
|========|====|====|====|====||====|====|====|====||====|====|====|====|
| D |  5 | ** | ** | ** | ** || ** | ** | ** | ** ||1.0 |0.39|0.62|0.32|
| i |----|----|----|----|----||----|----|----|----||----|----|----|----|
| s | 15 | ** | ** | ** | ** || ** | ** | ** | ** ||0.39|1.0 |0.53|0.35|
| g |----|----|----|----|----||----|----|----|----||----|----|----|----|
| u | 21 | ** | ** | ** | ** || ** | ** | ** | ** ||0.62|0.53|1.0 |0.59|
| s |----|----|----|----|----||----|----|----|----||----|----|----|----|
| t | 32 | ** | ** | ** | ** || ** | ** | ** | ** ||0.32|0.35|0.59|1.0 |
|=======================================================================
| S |  6 |0.40|0.42|0.48|0.44||0.44|0.35|0.45|0.43||0.48|0.26|0.42|0.37|
| a |----|----|----|----|----||----|----|----|----||----|----|----|----|
| d | 22 |0.50|0.46|0.55|0.42||0.59|0.54|0.58|0.48||0.50|0.40|0.53|0.46|
| ne|----|----|----|----|----||----|----|----|----||----|----|----|----|
| ss| 27 |0.33|0.38|0.46|0.38||0.42|0.33|0.42|0.46||0.36|0.30|0.38|0.37|
|=======================================================================
|dis| 11 |0.59|0.51|0.57|0.54||0.53|0.50|0.59|0.51||0.48|0.50|0.53|0.39|
|tre|----|----|----|----|----||----|----|----|----||----|----|----|----|
|ss | 30 |0.55|0.55|0.58|0.68||0.58|0.47|0.54|0.55||0.43|0.41|0.52|0.48|
========================================================================
---------**--------------**---------*
|Subgroup|| Sadness      ||Distress |
|--------||--------------||----|----|              -------**-------------
|Item No.|| 6  | 22 | 27 || 11 | 30 |              |Subgroup|| Distress |
|--------||----|----|----||----|----|              |--------||----------|
| S |  6 ||1.0 |0.57|0.49||0.53|0.48|              |Item No.||  11 | 30 |
| a |----||----|----|----||----|----|              |--------||-----|----|
| d | 22 ||0.57|1.0 |0.41||0.52|0.56|              |Dis| 11 || 1.0 |0.44|
| ne| ---||----|----|----||----|----|              |tre|----||-----|----|
| ss| 27 ||0.49|0.41|1.0 ||0.35|0.51|              |ss | 30 || 0.44|1.0 |
=====================================              ======================

                                 -40-

Table No. 6: Matrix of correlations of the items of the  subgroups fear,
 anger, disgust, sadness and distress -  of the Q-Sort ratings.
-----------------------------**-------------------**-------------------*
|Subgroup|       Fear        ||       Anger       ||      Disgust      |
|--------|-------------------||-------------------||-------------------|
|Item No.| 3  | 14 | 19 | 31 || 4  | 12 | 20 | 28 || 5  | 15 | 21 | 32 |
|--------|----|----|----|----||----|----|----|----||----|----|----|----|
| F |  3 |1.0 |0.22|0.21|0.23||-.02|0.00|0.14|0.01||-.29|0.11|-.10|0.08|
| e |----|----|----|----|----||----|----|----|----||----|----|----|----|
| a | 14 |0.22|1.0 |0.17|0.22||0.31|-.01|0.16|-.01||-.14|-.11|-.06|0.00|
| r |----|----|----|----|----||----|----|----|----||----|----|----|----|
|   | 19 |0.21|0.17|1.0 |0.24||0.13|0.10|-.21|0.10||-.09|-.05|-.19|0.09|
|   |----|----|----|----|----||----|----|----|----||----|----|----|----|
|   | 31 |0.23|0.22|0.24|1.0 ||0.11|0.11|0.02|-.11||-.02|0.03|0.03|-.04|
|========|====|====|====|====||====|====|====|====||====|====|====|====|
| A |  4 | ** | ** | ** | ** ||1.0 |0.10|0.24|0.19||-.31|-.10|-.05|-.02|
| n |----|----|----|----|----||----|----|----|----||----|----|----|----|
| g | 12 | ** | ** | ** | ** ||0.10|1.0 |0.14|0.28||0.28|-.09|0.24|-.03|
| e |----|----|----|----|----||----|----|----|----||----|----|----|----|
| r | 20 | ** | ** | ** | ** ||0.24|0.14|1.0 |0.18||-.02|-.15|-.20|-.06|
|   |----|----|----|----|----||----|----|----|----||----|----|----|----|
|   | 28 | ** | ** | ** | ** ||0.19|0.28|0.18|1.0 ||0.21|-.10|0.17|-.19|
|========|====|====|====|====||====|====|====|====||====|====|====|====|
| D |  5 | ** | ** | ** | ** || ** | ** | ** | ** ||1.0 |-.02|0.48|0.12|
| i |----|----|----|----|----||----|----|----|----||----|----|----|----|
| s | 15 | ** | ** | ** | ** || ** | ** | ** | ** ||-.02|1.0 |0.10|0.06|
| g |----|----|----|----|----||----|----|----|----||----|----|----|----|
| u | 21 | ** | ** | ** | ** || ** | ** | ** | ** ||0.48|0.10|1.0 |0.25|
| s |----|----|----|----|----||----|----|----|----||----|----|----|----|
| t | 32 | ** | ** | ** | ** || ** | ** | ** | ** ||0.12|0.06|0.25|1.0 |
|=======================================================================
| S |  6 |-.12|0.13|0.13|0.12||-.08|-.01|-.07|-.10||-.10|-.05|-.04|0.01|
| a |----|----|----|----|----||----|----|----|----||----|----|----|----|
| d | 22 |0.01|0.18|0.08|0.15||0.09|0.02|-.16|-.08||-.07|0.12|-.14|0.09|
| ne|----|----|----|----|----||----|----|----|----||----|----|----|----|
| ss| 27 |-.00|0.21|0.08|-.04||0.12|0.13|0.07|0.02||0.00|-.10|-.08|-.15|
|=======================================================================
|dis| 11 |0.07|-.12|0.20|0.13||0.08|-.05|0.06|0.16||0.01|-.03|0.00|0.05|
|tre|----|----|----|----|----||----|----|----|----||----|----|----|----|
|ss | 30 |0.11|0.27|0.15|0.12||0.19|0.24|0.15|-.01||-.01|-.06|0.08|-.28|
=========================================================================
---------**--------------**---------*
|Subgroup|| Sadness      ||Distress |
|--------||--------------||----|----|              -------**-------------
|Item No.|| 6  | 22 | 27 || 11 | 30 |              |Subgroup|| Distress |
|--------||----|----|----||----|----|              |--------||----------|
| S |  6 ||1.0 |0.44|0.42||0.37|0.29|              |Item No.||  11 | 30 |
| a |----||----|----|----||----|----|              |--------||-----|----|
| d | 22 ||0.44|1.0 |0.38||0.24|0.28|              |Dis| 11 || 1.0 |0.16|
| ne| ---||----|----|----||----|----|              |tre|----||-----|----|
| ss| 27 ||0.42|0.41|1.0 ||0.22|0.07|              |ss | 30 || 0.16|1.0 |
=====================================              ======================

-41-

The validity of anger

As seen in Table 6, item 28 is the most central to this subgroup; the correlation of item 20 with one item of fear is higher then its correlation with item 12 supposedly of this subgroup; the correlations of item 4 with items of fear are higher than with item 12 of its own group; item 12 has two high correlations with items of disgust that are higher than its correlation with item 4 and 20.

It seems that item 20 and especially item 4 are mixtures of anger and fear. It is also indicated that item 12 is a mixture of anger and disgust, when disgust is the dominant part and item 20 is a mixture of anger (mainly) and disgust.

The validity of disgust

The correlations of the items of this subgroup show that item 21 is the most loaded with this emotion; item 5 is a little less loaded with it; item 32 has in it a mixture of fear and is somewhat marginal to this sub-group; item 15 functions as a mixture of various emotions to which the contribution of disgust is small.

The correlations of this item with the artificial mixtures of emotions (1c of materials) of Q-Sort, reveal that only one out of seven mixtures containing disgust is significantly related to this item. Item 21 is in significant correlation with all those seven (and not with the other 17), item 5 is in correlation with six of those seven; item 32 is in a significant (positive) correlation with five of those seven - the three that contain surprise and the two that contain fear.

It can be concluded from the above findings that the content validity of this subgroup is insufficient. The interpretation of the dimensions of the scaling analysis will have to be related on the items 5 and 21 of this subgroup and on item 12 (originally of the anger subgroup).

The validity of sadness

The three items of this subgroup are (mainly according to the Q-Sort) of a relatively homogeneous group. Those three are relatively highly correlated with item 14 (fear) and the two distress items.

-42-

The validity of the two distress items

Izard (1971, 1977) presented items 11 and 27 as samples of the basic emotion distress. But item 27 is taken by him from Ekman & Friesen (1975) where it is presented as an item of the basic emotion sadness. According to them distress is, in the main, a mixture of sadness and fear and item 30 is their example for it. According to Table 5 and 6 it seems that the Ekman & Friesen version is more accurate than that of Izard and sadness is found to be nearer the concept of basic emotion than distress.

In conclusion of the test of the validity of the four negative basic emotions: it seems that the validity of those subgroups is wanting (insufficient).
However, the findings about the relevant 17 items are clear enough, and that is a sound enough base for the interpretation of dimensions content. It is not clear yet whether methodological problems of this study or those of Izard and Ekman & Friesen contributed the largest factor of deviations.

The validity of the sub-groups contempt and shame

Those two basic emotions (according to Izard, 1971, 1977) do not have enough support in research (according to Ekman et. al., 1982). They were included in the study for the purposes of control and comparison.

The validity of contempt

This subgroup is not highly cohesive. In the first task the highest correlation of item 7 is with item 14 - fear - r=0.31; for item 16 the highest is with fear no. 31 - r=0.38; for item 23 it is with anger no. 20 and with distress no. 30 - both have r=0.55; that of item 33 is with shame no.29 - r=0.37. Those correlations are higher than the inter-sub-group correlations which are demonstrated in the following Table 7.

-43- 
Table No. 7: Matrix of correlation of the subgroups  contempt  and  shame
that were included for control and comparison purposes
*----------------------------------------**------------------------------*
|Procedure|     Free grading (task one)  ||    Q-Sort (task three)       |
-----------------------------------------||-------------------**---------|
|Subgroup |      Contempt     ||  Shame  ||      Contempt     ||  Shame  |
|---------|-------------------||---------||-------------------||---------|
|Item No. | 7  | 16 | 23 | 33 || 13 | 29 || 7  | 16 | 23 | 33 || 13 | 29 |
|---------|----|----|----|----||----|----||----|----|----|----||----|----|
| C |  7  |1.0 |0.18|0.20|0.22||0.30|0.26||1.0 |0.17|0.10|0.03||0.20|0.15|
| o |-----|----|----|----|----||----|----||----|----|----|----||----|----|
| n | 16  |0.18|1.0 |0.27|0.28||0.27|0.39||0.17|1.0 |0.12|0.01||0.11|0.28|
| t |-----|----|----|----|----||----|----||----|----|----|----||----|----|
| e | 23  |0.20|0.27|1.0 |0.24||0.18|0.24||0.10|0.12|1.0 |0.20||0.15|0.05|
| m |-----|----|----|----|----||----|----||----|----|----|----||----|----|
| pt| 33  |0.22|0.28|0.24|1.0 ||0.29|0.37||0.03|0.01|0.20|1.0 ||0.12|-.05|
|=========|====|====|====|====||====|====||====|====|====|====||====|====|
| S | 13  | ** | ** | ** | ** ||1.0 |0.39|| ** | ** | ** | ** ||1.0 |0.37|
| ha|-----|----|----|----|----||----|----||----|----|----|----||----|----|
| me| 29  | ** | ** | ** | ** ||0.39|1.0 || ** | ** | ** | ** ||0.37|1.0 |
*========================================**==============================*
In the Q-Sort task: items 7 and 16 (which are photographs of males) are significantly correlated as a cluster and both have their highest correlation with shame items; item 23 and item 33 are photographs of females and represent a cluster of their own.

The significant correlations of item 7 with artificial mixtures (materials 1c) are with one of the three containing contempt (the other emotion of this mixture is anger) and with one item which is supposed to be neutral; item 16 is correlated with two of the three mixtures containing contempt, with three of the seven containing disgust and with three that contain anger; item 23 is correlated with one mixture of contempt, three out of seven of disgust; item 33 is correlated with the two "neutral" items, with one of happiness + sadness and one of happiness + contempt.

It seems that the contempt subgroup has a weak common denominator, that the loadings of the items with it is not high, and that each of them is a mixture with out any obvious content.

The validity of shame

The two items of this subgroup act as a cohesive cluster in both procedures.

-44-

The validity of the 24 artificial mixtures

The 24 items of materials: 1/c, suffered in the first task from the same set of response of the basic emotions - denial. Therefore, the examinations of the content validity of those items is mainly based on the Q-Sort task.

The two neutral items 1 and 14 (included for control on Ekman & Friesen (1975) procedure of item creations show that they are not neutral.

Item 1 is correlated to the four happiness items, with one surprise item, with one contempt item, with mixture 9 that includes in it happiness + surprise, with mixture 13 of surprise + questioning and with mixture 22 of happiness + contempt.

Item 14 is positively correlated with both shame items (the higher r=0.43), with one item of contempt, and with one item of interest. It was also positively correlated with mixtures no. 5 and no. 19 - ambiguous content of a slight anger, determination seriousness; and no. 1 - neutral.

Those correlations indicate that Ekman & Friesen's (1975) technique for creating artificial facial expression is not flawless. It seems that it cannot prevent the contribution of the models' mood and chronic facial expressions to the manipulated-artificial facial expression.

The six items that include happiness are also inconsistent. Four are correlated to all four happiness items (the higher is r=0.60), but the other two have too low correlations with the happiness items. Those two are negatively correlated with many items of facial expression of emotion - the higher is r=-0.21 p<0.002 (two tail).

The negative correlations of the above two indicate that the study includes more emotional contents than the intended 9 basic emotions.

The four items that include surprise have 14 correlations with the four surprise items - only one is not significantly positive.

-45-

The six items that include anger are not of the same quality. Only one item is correlated with all four items of anger. Three are correlated with three items of anger and two are significantly correlated to none of them.

The five items that include disgust are correlated to the two main items of disgust (no. 5 and no. 21). Four of them are correlated with item no. 12 of the anger subgroup which was found to be a mixture of disgust and anger. Only one of the five mixtures is correlated to disgust item no. 23 which is marginal to the disgust basic emotion.

The four items that include sadness are not of the same quality. Two of them are correlated with the four sadness and the two distress items, one of the mixtures is correlated with three sadness items and one of distress mixture no. 24 has only one low correlation with them (r=0.13) which is with one of the distress items.

The three items that include contempt have low correlations with the items of the basic emotion contempt. Only one mixture is correlated to three items of contempt.

In conclusion to the examinations of the validity of items 1a, 1b, 1c

The content validity of the artificial facial expressions of emotion was found wanting and not of the same quality for all basic emotion or mixtures of those. It seems that the reliance on these items is feasible but it must be done cautiously. The interpretation of the meaning of the dimensions and the directions of basic emotion in the multidimensional scaling space will have to be based on the convergence of various findings.

The directions of the basic emotions in their multidimensional -scaling space The main hypothesis of the study, dealing with facial expressions of emotion claims that the content of the main dimensions of discrimination among emotions matches the content of the seven basic emotions that comprise the primary list of this study.

-46-

I n order to examine this, the 48 items that are unmanipulated facial expressions - materials: 1a - were analyzed by multidimensional scaling. This was done by a version of the S.S.A.P.-I of the Guttman-Lingoes series (Lingoes, 1973) as adapted by Shalif, et. al. (1981). The basic procedure of this adaptation is based on the available version of the S.S.A.-I of the same source. When using it, one can include in the analysis up to 100 items, reach up to the ten-dimensional level and compute subjects` dimensional scores of a relatively unlimited number of subjects.

The coefficients of alienations of the S.S.A.I analyses
The result of the multidimensional analyses of the 48 items of material:1a is as follows: for two dimensions - the coefficient of alienation is 0.222; for three dimensions - 0.172; for four - 0.123; for five - 0.103; for six - 0.082; for seven - 0.075; for eight - 0.061; for nine - 0.058; for ten dimensions - 0.049.

The size of the second dimension is about 70% of the first dimension; the tenth dimension is about 80% of the ninth and 40% of the first.

Analysis of artificial data

An analysis of artificial data was done in order to asses the relations between relative sizes of the dimensions and coefficients of alienations, and the relative weight of the basic variables of the domain.

The simulation was done as follows: first - each of 48 "items" was assigned a point in multidimensional space using the Random procedure of the Pascal programing language; second - the matrix of distances was computed to the above configuration of points; third - the derived matrix was analyzed by the S.S.A.-I.

Four variables were changed systematically: a) the number of dimensions of the initial configuration of the 48 "items"; b) the relative sizes of the dimensions; c) the size of a random factor added to the distances of the matrix of distances among the "items"; d) the number of dimensions of the S.S.A.-I analysis.

-47-

The simulation that was found the nearest to the real data of this study is one of a ten dimensional configuration. In it the relative size of the second dimension is 70% of the first one. Each of the following dimensions is 90% of the one before it. To each distance of the matrix of the distances among the "items" was added a random number whose size was between zero and a 25% of that distance. Table 8 contains the coefficients of alienations of the multidimensional analyses of the real data, the above simulation and the result of that simulation which does not contain in it the added random factor. 
Table No. 8: The coefficients of alienations of the S.S.A.-I analyses of real
data and two simulations - with and without added random size to the matrix.
---------------------------------------------------------------------------|
|Number of the  |  coefficients of  alienations  of  the  three  versions  |
|dimensions of  |----------------------------------------------------------|
|the  analysis  |real data|I. with added random size|II. without added size|
|---------------|----------------------------------------------------------|
|       2       |  0.222  |         0.227           |         0.21         |
|       3       |  0.172  |         0.16            |         0.11         |
|       4       |  0.123  |         0.12            |         0.082        |
|       5       |  0.103  |         0.09            |         0.041        |
|       6       |  0.082  |         0.078           |                      |
|       7       |  0.075  |         0.057           |                      |
|       8       |  0.061  |                         |                      |
|       9       |  0.058  |                         |                      |
|      10       |  0.049  |         0.047           |         0.01         |
===========================================================================|
It can be concluded that the relation between the change in coefficient of alienation and the value of added dimensions is not linear. It is clear that the coefficient of alienation is not a reliable measure in the determination of the number of dimensions of the studied domain or their relative weight.

To get back ON-LINE with the Biofeedback WITHOUT instruments
The other Home Pages: http://www.netvision.net.il/php/gshalif
http://www.geocities.com/~drilanshalif
The E-Mail: gshalif@netvision.net.il
To the TOP

For suggestions and complains click on: gshalif@netvision.net.il

(=====Back to the theory
To results II. ---->