Vous êtes sur la page 1sur 9

Journal ol Experimental Psychology

1965, Vol. 70, No. 1, 18-26

ROUGHNESS, SMOOTHNESS, AND PREFERENCE: A STUDY


OF QUANTITATIVE RELATIONS IN
INDIVIDUAL SUBJECTS >
GOSTA EKMAN, JAN HOSMAN, AND BRITA LINDSTROM
University of Stockholm

A set of 7 surfaces were used as stimuli in a scaling experiment. For 10


Ss individual scales were obtained, by a method of ratio estimation, for
(a) roughness, (b) smoothness, and (c) preference. The main results
were: (a) For all Ss roughness was approximately a power function of
the physical stimulus variable denned as the coefficient of friction.
The exponent of the function varied greatly between Ss but exceeded 1
in almost all cases, (b) For nearly all Ss smoothness was approximately
the inverse of roughness, (c) For nearly all Ss preference was directly
proportional to smoothness.

It has been pointed out (e.g., smoothness in relation to each other


Ekman, 1961b) that psychological and to the corresponding stimulus
mechanisms may sometimes turn out variable, and (c) to study all these re-
to be rather simple when studied at lations on the level of the individual 5.
the proper level. The psychological
variable of similarity has, thus, been METHOD
demonstrated to be a simple function Subjects.Ten 5s took part in the experi-
of measurable psychological aspects ment, eight females and two males. All 5s
of the percepts entering into the were students of psychology with some pre-
comparison (Eisler, 1960; Eisler & vious experience of ratio estimation.
Stimulus materials.Seven stimulus sur-
Ekman, 1959; Ekman, Engen, Kiin- faces were used. Five stimuli (No. 1-5) were
napas, & Lindman, 1964; Ekman, pieces of sandpaper, one (No. 6) was a piece
Goude, & Waern, 1961). of cardboard, and one (No. 7) a piece of
The present paper reports a study ordinary writing paper. Several identical
in a new series of investigations of stimulus sets were prepared and used suc-
cessively in order to avoid any noticeable
psychological mechanisms, in which change in surface due to use.
a preference variable is studied as a Static coefficients of friction were experi-
function of other psychological vari- mentally determined in relation to a sample
ables. This type of problem has been of finger tips under conditions similar to those
of the main experiment. From the number of
rather neglected; one of the few in- grits used in the production of the sandpapers
stances of a similar approach was a an alternative stimulus measure, I/(No.
color preference study by Guilford of grits), was obtained which can be regarded
(1939). The particular purposes of as proportional to the diameter of the grains.
the present investigation were: (a) to The latter type of stimulus characteristic
was used by Stevens and Harris (1962) in
investigate preference as a function of their scaling of roughness and smoothness.
the subjective variables roughness and The data of the present experiments (as well
smoothness, (b) to investigate the as additional experiments covering a wider
subjective variables roughness and stimulus range) indicate, however, that the
two sets of stimulus measures are not propor-
1
This investigation was supported by the tional to each other. Under these circum-
Swedish Council for Social Science Research. stances the coefficient of friction was chosen
The work was performed while Jan Hosman, as stimulus measure.
from the University of Leyden, held a scholar- However, the physical attributes determin-
ship at the University of Stockholm. ing the degree of perceived roughness are prob-
18
SCALING ROUGHNESS, SMOOTHNESS, AND PREFERENCE 19

ably not described exhaustively by the coef- the ratio matrices according to a
ficient of friction. Since the sandpapers are procedure described by one of the
most similar to each other in other physical
respects, the relations between the stimulus present authors (Ekman, 1958, Equa-
variable and the subjective variables may be tion 14).
most adequately studied for Stimuli No. 1-5. The psychophysical function.The
Experimental procedure.The stimuli were subjective scales of roughness and
presented in pairs of surfaces, each about
3 X 1 0 cm., covered by a screen in such a way smoothness are plotted against the
that S, without seeing the surfaces, could stimulus variable, in log-log coor-
comfortably stroke them with his two index dinates, in Fig. 1, in which there is
fingers. one diagram for each 5. Straight
There were three experimental conditions: lines represent the general trend of all
(a) roughness estimation, (b) smoothness
estimation, and (c) preference estimation. data in first approximation. For
Each S participated in all conditions, one most 5s the fit would be considerably
condition in each of three sessions on different better (and the slope of the lines a
days. Right-left position was rotated for little steeper) if only the sandpaper
each S, and the order of conditions was ro-
tated between 5s. In each condition S
stimuli were included. (A slightly
judged two random orders of 42 pairs, i.e., better fit may also be achieved by
each of the 21 different stimulus combinations introducing an additive constant [[Ek-
was judged four times by each S. man, 1958, 1961a].) It can be
In the roughness-estimation sessions 5s concluded that power functions de-
were instructed as follows: scribe the psychophysical relation in
You will be presented with pairs of sand- reasonably good approximation, at
papers. Your task is to give a judgment least for the sandpaper stimuli, for
about the roughness of the sandpapers.
Take your time just to get a good idea how
all 5s. The actual fit is most ade-
rough one paper is compared with the other. quately judged in Fig. 2, where the
Give the judgment in free numbers. For roughness functions are shown, for
instance, if the paper on the right seems all seven stimuli, in linear coordinates;
to be three times as rough as the paper on the goodness of fit of the smoothness
the left, say: the ratio is three to one. You
may also use numbers like 300 to 100, or 45, functions is much the same.
to 15, or 1 to 0.33. This result is of considerable inter-
In the smoothness-estimation sessions the est, for very little is known so far
same instruction was used with the words about the psychophysical function
"smoothness" and "smooth" substituted for for individual 5s. Almost all previous
"roughness" and "rough." conclusions have been based on group
In the preference-estimation sessions 5s
were instructed in the same way to give "a
data. One investigation by Pradhan
judgment about the character of the surface and Hoffman (1963) was in part con-
in terms of pleasantness or preference." cerned with this problem. According
The "free" form of the ratio-estimation to these authors, the power function
method used in these experiments is similar did not describe the majority of
to the form suggested by Mashhour (1964).
individual data in experiments with
weight lifting, although it did repre-
RESULTS sent the group data fairly well; hence,
Experimental data.The pairs of it was concluded that the power
numbers reported by S were trans- function might be an artifact of
formed to ratios. Arithmetic means grouping. On the other hand, recent
were computed for the four ratios ob- investigations have shown a very
tained from each 5 for each stimulus satisfactory fit of power functions
combination in each condition. In- to individual loudness data (Stevens
dividual scales were constructed from & Guirao, 1964) as well as to data
Of

Oi

p
0.2

O.o 0.0 I 0.0


-0.2 -0-1 Oo O-i -02 -0.1 Oo -0.2 -0.1 0.0 -02 -0.1 0.0 O-i -02 -0.1 0.0 0.1

o
01
a
o
3 VI
en
VI O.e Vffl DC
O)
O
06 0.6

0.4 68
r
Oi 0.2

D
C/2
0.0
-0.2 -0.1 0.0 0.1 -0.2 -01 OK 0.1 -0.2 -0.1 Oo 0.1 -02 -O.i Oo O.i -02 -0.1 00 0.1
I
Log stimulus variable
o - Smoothness - Roughness

FIG. 1. Subjective roughness (filled circles) and subjective smoothness (empty circles) plotted against the coefficient of friction,
in log-log coordinates. (Each graph represents an individual S.)
C/)
n

o

o
X
0) 0.0 0.5 10 Oo 0.0 0.5 IJD O.o 0.5 0.0 0.5 1.0
2;
c

o
a:
VI
2!
M
C/3

2!
O
"0
5S
Oo 0.5 10 Oo Os 1o Oo 05 1.0 0.0 05 l.o 0.0 0.5 W

Stimulus variable W
2!
n
FIG. 2. Subjective roughness plotted against the coefficient of friction, in linear coordinates. M
(Each graph represents an individual S.)
0.8 0.8 O.s IV
05 06 06
P
Ol 0.4 0.4

0.2 Oi 02
o
o
E Oo O.o O.o
in Oo 02 04 06 Oa 0-0 02 04 06 Os 00 0.2 Oi 0.6 09 00 02 0.6 08 0.0 0.2 0i 06 08

u
(V
in"
u
OB VI 0.8 vm 08 DC 0.8

in
06 06 0* z:
o>
o
0< Oi
a
CB

0.2 0.2 r
2
O.o Oo o
0.0 0.2 0-4 0.6 O.S 0.0 02 04 06 Oa Oo 02 0 0s 00 02 Oi 06 00 02 04 0.6 0-8

Log subjective roughness


FIG. 3. Subjective smoothness plotted against subjective roughness, in log-log coordinates.
(Each graph represents an individual S.)
iy in

I
21
C 0 2 4 6 0 2 ( 6 0 2 1 6 0 2 4 6 0 2 4 6
Gl
C/)

VI VI DC 8
21

0 2 1 0 2 ( 6 0 2 4 6 0 2 4 6 0 2 4 6
w
jo
M
Subjective smoothness 21
n
FIG. 4. Preference plotted against subjective smoothness, in linear coordinates. (Each graph represents an individual S.)
24 G. EKMAN, J. HOSMAN, AND B. LINDSTROM

concerning saltness and sweetness (Ek- correction is performed (Eisler, 1962;


man & Aakesson, 1964). cf., Schneider & Lane, 1963).
Whereas the individual power func- Let roughness be denoted * and
tion may be considered as approxi- smoothness y. If the relation is one
mately verified in the present experi- of simple reciprocality, then y = c/x,
ment, it is seen from the graphs in or log y = log c log x. This hy-
Fig. 1 that there is a considerable pothesis has been tested by plotting
variation with regard to the slope of smoothness against roughness, in log-
the straight lines fitted to the data. log coordinates, in Fig. 3 where each 5
In other words, the exponent of the is represented by a separate graph.
power function is a variable over It is seen that for most 5s the rela-
individuals. In the present sample tion, log y = log c log x, describes
of 10 5s, the exponent of the functions the data in fairly good approxima-
fitted to all seven stimuli varies tion; the straight lines drawn in the
between .8 and 3.5 for roughness and graphs all have a slope of 1. It may
between .8 and 3.0 for smooth- thus be concluded that scales of
ness. Although part of this variation roughness and smoothness, when con-
may be due to unreliability of data, structed by the method of free ratio
it can be concluded from inspection estimation used in this experiment,
of the graphs in Fig. 1 and, particu- tend to be inversely related for indi-
larly, Fig. 2 that other sources of vidual 5s.
variation are much more important. The preference function.One of
Reciprocality of scales.It has been the main purposes of the present
shown in previous investigations by investigation was to study preference
Torgerson (1960, 1961), Eisler (1962), as a function of the subjective vari-
Stevens and Harris (1962), and ables of roughness and smoothness.
Schneider and Lane (1963) that In Fig. 4 the preference variable is
magnitude scales tend to be inversely plotted against smoothness, in linear
related for continua defined by oppo- coordinates. It is immediately seen
site terms (e.g., loudness and soft- that the relation is one of direct
ness). In some cases the inverse proportionality.
relationship is not obtained in reason- Letting p denote preference, the
able approximation, unless a scale relations between the three subjective

0.6 K 0.6
01
01 c
f
I 0.4 O 0.4
O

in
io,
in
0.2
0)
01 o
o .
_j O.o O.o
-0.2 -0.1 O.o 0,1 0.0 0.2 0.4 0.6 1 2 3

Log stimulus variable Log roughness Smoothness

FIG. 5. Group data summarizing results. (Diagram A shows subjective roughness, filled
circles, and subjective smoothness, empty circles, plotted against the stimulus variable, in log-
log coordinates. Diagram B illustrates, in log-log coordinates, the reciprocality of roughnets
and smoothness. Diagram C shows preference as a function of smoothness, in linear coordinates.)
SCALING ROUGHNESS, SMOOTHNESS, AND PREFERENCE 25

variables can now be summarized is concerned, there appear to exist two


thus: p = by = a/x, where a and b main alternatives: (a) The differences
are constants related to the arbitrary are "perceptually real," i.e., the differ-
units of measurement. ences reflect some kind of variation with
regard to sensitivity, (b) The differ-
Group data.Group scales have ences reflect the different ways in which
been obtained by computing geo- 5s handle the numbers, i.e., the differ-
metric means of individual scale ences represent systematic errors in the
values. The psychophysical function, scale construction, basically an inability
the reciprocality of smoothness and of some or all of the 5s to give ade-
roughness, and the preference func- quate quantitative reports.
tion are illustrated by group data in It appears impossible to decide, on the
Fig. 5 A-C. basis of present knowledge, between
These graphs do not add anything these alternative interpretations of the
to the results we have already ob- interindividual differences which have
been found. Similar alternative inter-
tained and illustrated. Insofar as pretations have been suggested in an
we are interested in the generality interesting paper by Jones and Marcus
over individualsof our results, the (1961), reporting a study of individual
group data convey less information. power functions for three continua.
They are presented here only as a Positive correlations between exponents
summary of results. were considered as supporting the view
that interindividual differences largely
DISCUSSION reflect differences in handling numbers.
On the other hand, the present authors
The conclusions of the present study and their co-workers, in unpublished
have all been based on individual data. studies of a larger number of continua,
This was possible partly because of the have found correlations which are low
reasonably satisfactory regularity of or moderate as compared to remarkably
these data. It should be borne in mind high reliability coefficients characteristic
that each individual scale was con- of individual scales. A further clarifica-
structed from a ratio matrix based on tion of the issue is highly desirable.
4 X 21 or altogether 84 estimates, ob- Generally speaking, the problem of
tained in the course of one session, lasting individual number behavior is basic to
about 1 hr. The complete method of vast areas of current psychophysics
ratio estimationinvolving all different (cf., Ekman, 1964a, 1964b).
combinations of stimuliappears ef- Finally, it is worthy of notice that
ficient for the construction of reasonably some of the generalities revealed by the
reliable individual scales. present study may be valid results
The considerable interindividual varia- regardless of the answer to the questions
tion with regard to the exponent of the just discussed. The power function
power function for roughness and smooth- might be the true individual psycho-
ness is reflected also in the scales of physical law, even if the exponents
preference. Basically it is a variation should not be taken at their face values.
with regard to the subjective range cor- The reciprocality of smoothness and
responding to the constant stimulus roughness might be a genuine result,
range. A given set of stimuli elicits which it was possible to demonstrate
subjective responses which cover a wide because both scales were distorted in
range in some 5s (e.g., I, V, and VII) the same way. For similar reasons, the
and a narrow range in some other 5s simple relation found between the prefer-
(e.g., Ill, VI, and VIII); these differ- ence variable and the perceptual vari-
ences are conspicuous in Fig. 1-4. As ables of smoothness and roughness might
far as the interpretation of these findings probably be accepted as a valid result.
G. EKMAN, j. HOSMAN, AND B.
REFERENCES EKMAN, G., GOUDE, G., & WAERN, Y.
Subjective similarity in two perceptual
EISLER, H. Similarity in the continuum of continua. /. exp. Psychol., 1961, 61, 222-
heaviness with some methodological and 227.
theoretical considerations. Scand. J. Psy- GUILFORD, J. P. A study in psychodynamics.
chol, 1960, 1, 69-81. Psychometrika, 1939, 4, 1-23.
EISLER, H. Empirical test of a model relating JONES, F. N., & MARCUS, M. J. The subject
magnitude and category scales. Scand. J. effect in judgments of subjective magnitude.
Psychol., 1962, 3, 88-96. /. exp. Psychol., 1961, 61, 40-44.
EISLER, H., & EKMAN, G. A mechanism of MASHHOUR, M. Psychophysical relations in
subjective similarity. Ada psychol., Amster- the perception of velocity. Stockholm:
dam, 1959, 16, 1-10. Almqvist & Wiksell, 1964.
EKMAN, G. Two generalized ratio scaling PRADHAN, P. L., & HOFFMAN, P. J. Effect
methods. J. Psychol., 1958, 45, 287-295. of spacing and range of stimuli on magni-
EKMAN, G. A simple method for fitting psy- tude estimation judgments. /. exp. Psy-
chophysical power functions. /. Psychol., chol., 1963, 66, 533-541.
1961, 51, 343-350. (a)
EKMAN, G. Some aspects of psychophysical SCHNEIDER, B., & LANE, H. Ratio scales,
research. In W. A. Rosenblith (Ed.), category scales, and variability in the
Sensory communication. New York: Wiley, production of loudness and softness. /.
1961. Pp. 35-47. (b) Acoust. Soc. Amer., 1963, 35, 1953-1961.
EKMAN, G. Is the power law a special case of STEVENS, J. C., & GUIRAO, M. Individual
Fechner's law? Percept, mot. Skills, 1964, loudness functions. /. Acoust. Soc. Amer.,
19, 730. (a) 1964, 36, 2210-2213.
EKMAN, G. A. A note on alternative psycho- STEVENS, S. S., & HARRIS, J. R. The scaling
physical laws. Report No. 175, 1964, of subjective roughness and smoothness.
University of Stockholm, Psychological /. exp. Psychol., 1962, 64, 489-494.
Laboratories, (b) TORGERSON, W. S. Quantitative judgment
EKMAN, G., & AAKESSON, C. Saltness, sweet- scales. In H. Gulliksen & S. Messick
ness and preference; A study of quantitative (Eds.), Psychological scaling: Theory and
relations in individual subjects. Report applications. New York: Wiley, 1960.
No. 177, 1964, University of Stockholm, Pp. 21-31.
Psychological Laboratories. TORGERSON, W. S. Distances and ratios in
EKMAN, G., ENGEN, T., K^NNAPAS, T.( & psychophysical scaling. Acta psychol.,
LINDMAN, R. A quantitative principle of Amsterdam, 1961, 19, 201-201.
qualitative similarity. /. exp. Psychol.,
1964, 68, 530-536. (Early publication received October 13, 1964)

Vous aimerez peut-être aussi