The development of the concept of artistic style: 
Color 

Cutting across both style and subject-
matter dimensions were three color 
categories: (1) light tones, (2) dark tones, 
and (3) green-yellow tones. 

A free classification studyl 

JOY A. FRECHTLfNG and PHILIP W. 
DA VIDSON, The George Washington 
University, Washington, D.C. 20006 

Ss, ranging in age Irom 5 years to adult, 
were asked to sort a set 01 oil paintings 
into groups 01 their own choice. The 
results indicated that (1) subject matter 
was the strongest determiner 01 children 's 
sorting behavior, (2) only the adults 
showed a high Irequency 01 groupings by. 
artistic style, and (3) color was seldom 
used as a basis lor sorting by any age 
group. These jindings were interpreted in 
terms 01 their implications lor teaching the 
artistic style concept to children. 

Walk (1967) has shown that adults can 
acquire the concept of artistic style 
through exposure to positive instances. 
Evidence suggests, however, that this 
concept is very difficuIt for the 4· to 
8-year-old child to acquire (Walk & 
Karusaitis 2). This finding cannot be 
attributed to the child's inability to find 
similarities among multidimensional 
stimuli. Kofsky & Osler (1967) reported 
that even 5-year-olds can consistently sort 
sets of cards varying in color, form, 
number, and size. Alternatively, a child's 
poor performance with the style concept 
may stern from the interfering influence of 
other stimulus variables. Artistic style can 
be characterized as a relation al concept, a 
particular combination of the more 
concrete features of a painting, such as 
subject matter, color, or form. The young 
child may be unable to perceive such 
relations and may attend only to the 
concrete features themselves. 

Studies of artistic preferences in children 
seem to support this hypothesis. Machotka 
(1963,1966) and Subes (1958) found that 
the young child prefers paintings for such 
concrete features as subject matter and 
color. It is only after the development of 
concrete operations is complete 
(cf. Inhelder & Piaget, 1955) that an 
interest in such more complex variables as 
style begins to appear. These da ta suggest 
that if the child is faced with t!1e task of 
fmding similarities among works of art, his 
sytem of classification may, indeed, be 
based upon concrete features. Such a 
tendency may have to be overcome before 
the concept of style can be learned. This 
argument might explain Walk and 
Karusaitis's2 finding that exposure to 
positive instances of style alone led to [ar 
less consistent concept acquisition in 
young children than in adults. 

Psychon. Sci., 1970, Vol. 18 (2) 

The present research reports 
developmental changes in judgment of 
similarity among paintings, assessed by 
means of a free classification task 
(cf. Kofsky & Osler, 1967). We think two 
critical questions about the development of 
the artistic style concept can be answered 
best with this paradigm: First, on what 
basis do children of different ages judge 
paintings as similar? Second, at what point 
in deveIopment does the child begin to 
respond to features of style? 

SUBJECTS 
Fifty Ss from five age groups (5-, 6-,7-, 

12-year-olds, and adults) participated in 
the experiment. Each group consisted of 
IO Ss. The children were selected from 
Washington, D.C., area private schools. The 
adults were graduate students in 
psychology at The George Washington 
University. 

STIMULI 
Sixteen 4 x 5 in. color prints of oil 

paintings were selected from a collection of 
commercial art books representing 
different style periods (Albert Skira Series, 
World Publishing Company). The paintings 
were chosen to permit three alternative 
dimensions for c1assification: artistic style, 
subject matter, and color. The three 
dimensions were divided into the following 
categories: 

Artistic Style 
Four paintings were chosen from each of 

the following style periods: 
(1) romanticism, (2) early impressionism, 
(3) late impressionism, and (4) surrealism. 

Subject Matter 
At least one painting in each style period 

represented the following subject-matter 
categories: (1) boats or water scenes with 
boats, (2) landscapes not depicting boats or 
water, (3) c10thed human figures, and 
(4) abstract themes. 

The 16 paintings and their artists, 
classified by style period, are listed in 
Table 1. 

PROCEDURE 
A free cIassification procedure was 

followed. The 16 prints were placed 
simultaneously before S in a random 
arrangement. Each S was asked to "group 
together the paintings you think belong 
together." Tbe Ss were instructed to use a 
"doesn't belong" category for those prints 
they thought could not otherwise be 
cIassified. Each S sorted the paintings only 
once. No reinforcement for choices was 
provided by Es. 

RESPONSE EVALUATION 
Each grouping made by S was scored as 

either a style, a subject-matter, or a color 
response. If a grouping failed to fit any of 
the response categories defined by the 
rules, it was scored as "uncIassifiable." 
Paintings placed by Ss in the "doesn't 
belong" category were not included in this 
scoring procedure. The following rules 
defined each response category. 

Style 
All paintings in the grouping must have 

come from the same style period. 
Subject Matter 

All paintings in the grouping must have 
shared the same theme.The original 
number of subject-matter categories was 
increased somewhat on the basis of the 
actual responses we observed. We found, 
for example, that the "cIothed human 
figures" category was frequently 
subdivided by Ss into "male" and "female" 
figures. 

There were a few cases when a grouping 
could be scored as based on either style or 
subject matter. In such cases, the S's use of 
these two dimensions in his other 
groupings determined the score for the 
ambiguous grouping. For example, two 
romantic paintings that often comprised a 

Table 1 
Stimulus Paintings and Their Artists 

Painting Artist Date 

''The raft of the 'Medusa'" Gericault 1818 
"Man and woman gazing at the moon" Friedrich 1819 
'The slave ship" Turner 1839 
"Portrait of an artist in his studio" Gericault 
"Boats moored alongside the quay " Van Gogh 1888 
"Le jardin de l'infante" Monet 1866 
"Still life with books" Van Gogh 1886 
"Camille" Monet 1866 
"Reflections on the water" Derain 1905 
"Disk" Delaunay 1912 
"Woman with a hat" Matisse 1905 
HLe lavandou" Cross 1904 
''Turin, spring" de Chirico 1914 
no title Tanguy 1927 
"Smile of my blonde" Miro 1926 
"[n the land of night" Magritte 1928 

Style 

Romantic 
Romantic 
Romantic 
Romantic 
Impressionist 
Impressionist 
Impres~ionist 

Impressionist 
Fauvist-Pointillist 
Fauvist-Cubist 
Fauvist 
Fauvist-Pointillist 
Surrealist 
Surrealist 
Surrealist 
Surrealist 

79 



separate grouping depicted boats in motion 
(Gericault's "The Raft of the 'Medusa' " 
and Turner's "The Slave Ship"). The two 
other paintings in the "boats or water 
scenes with boats" subject-matter category 
(Van Gogh's "Boats Moored Alongside the 
Quay" and Derain's "Reflections on the 
Water") depicted moored boats. If Ss put 
the Van Gogh and the Derain together as a 
separate grouping, that response was scored 
as "subject matter," since each painting 
comes from a different style period. In 
addition, we concluded that some Ss were 
subdividing the "boats, etc." 
subject-matter category into "boats in 
motion" and "moored boats." Therefore, 
when Ss grouped the Gericault with the 
Turner, and also placed the Van Gogh with 
the Derain, we scored the Gericault-Turner 
grouping as a subject-matter response. On 
the other hand, this grouping was scored as 
a style response for Ss who grouped the 
Van Gogh with other early impressionistic 
prints, and the Derain with other late 
impressionistic prints. 

Color 
All paintings in the grouping must have 

shown the same color characteristics (e.g., 
light tones, dark tones, or green-yellow 
tones). 

The groupings made by each S were 
scored by a rater trained by Es in the use 
of the rules. The rater did not otherwise 
participate in the experiment and scored 
the data without knowledge of Ss' ages. 

RESULTS 
The me an usage of stimulus dimensions 

by each age group is shown in Fig. I. Four 
major findings are indicated by these data. 
First, subject matter was a strong 
determiner of groupings at all age levels. 
The 6-, 7-, and 12-year-olds sorted the 
prints almost exclusively on the basis of 
this dimension. Second, style emerged as a 
significant basis for classification only in 
the adult groupings. Third, color was the 
least used dimension when compared with 
style and subject matter. Finally, the 
5-year-olds were the only group to give a 
large n umber of unclassifiable responses. 

The data were analyzed with a two-way 
(age by stimulus dimension) analysis of 
variance, with repeated measures on the 
stimulus dimension variable. This statistic 
indicated a significant Age by Stimulus 
Dimension inter action (F = 2.312, 
df= 12/135, p < 0.05). A subsequent 
analysis of the simple effects, followed by 
Newman-Keuls posttests, indicated these 
significant relationships: 

Style 
The adults showed more style groupings 

than did the 5-year-olds (p < 0.05). In 
addition, the adults' style responses were 
more frequent than either their color 

80 

3.0 

'" ... 2.S ;;; 
<:) 
Cl) ... ... 

2.0 C .., 
... 
<:) 

• I.S ... -== ~ 
LO Z 

Z 
C ... .. O.S 

.•.. /' 
__ ~~''t'"~ __ """",// 

\'6-~."--- .... 6-& ...... ~-- --6 

12 

AGI 

Fig. 1. Mean usage of stimulus 
dimensions in classification of oil paintings 
as a function of age. 

(p < 0.05) or unclassifiable (p < 0.01) 
groupings. 

Subject Matter 
The 12-year-olds grouped more often by 

subject matter than did either the 
5-year-olds (p < 0.01) or the adults 
(p < 0.01). The 6-, 7-, and 12-year-olds all 
showed more subject-matter responses than 
style, color, or unclassifiable groupings 
(P< 0.05 for all comparisons). 

Color and Unc1assifiable 
Color was seldom used by any age group 

as a basis for sorting, and there were no 
significant deveiopmental differences in 
color usage. 

More unclassifiable responses were given 
by the 5-year-olds than by any other group 
(p < 0.05), and these Ss used the 
unclassifiable category more often than 
either style (p < 0.05) or color (p < 0.05)_ 

The Ss averaged approximately 4.3 
groupings per sort with about 3.2 prints 
per grouping. One-way analyses of variance 
on mean number of groupings. and mean 
grouping size data indicated no 
deveiopmental differences for either 
variable. Further, each S made an average 
of approximately 0.5 groupings of prints 
that "didn't belong," with about 4.0 prints 
per grouping. An inspection of the 
particular paintings classified as not 
belonging revealed no consistent patterns 
either between age groups or between Ss. 

DlSCUSSION 
This study used a free classification 

paradigm to ex te nd the investigation of 
aesthetic appreciation begun with 
preference experiments (Machotka, 1963, 
1966; Subes, 1958). Our results generally 
agree with the findings of these 
experiments. Subject matter provided the 
most frequent basis for classification of 

paintings in children, and increased in 
importance with age. Further , children 
seldom based their sorting on artistic style. 

We report one discrepant finding, 
however. Machotka (1963) found that his 
youngest Ss (about 5 years old) preferred 
paintings for their color more often than 
for either style or subject matter. Our 
5-year-olds not only did not sort on the 
basis of color, but showed an inability to 
classify the prints on any obvious basis. 
The most reasonable account of this 
discrepancy seems to us to He in the 
nonavailability of color as a distinct 
dimension in oil paintings. Most often, 
paintings can be characterized by their 
pattern of colors rather than by a single, 
solid color. It is possible that young 
children may prefer various individual 
paintings for the colors they display 
(cf. Machotka, 1963) but are unable to 
perceive similarities in color patterns across 
several paintings. Machotka's reslilts may 
be divergent from ours only insofar as his 
preference task required no responding to 
relations among paintings. When such 
re1ational responding is required, such as in 
the free classification task, "color" may be 
as difficult a dimension as style for young 
children. 

I n general, both preferences and 
judgment of similarity in art work proceed 
developmentaIly from attention to 
concrete stimulus aspects toward an 
appreciation of more complex relational 
features_ This evidence suggests that in 
order to teach a child to respond to artistic 
style, it may be necessary not only to 
expose Ss to positive instances of the 
concept (cf. Walk & Karusaitis 2), but also 
to provide specific training to reduce 
attention to other, preferred, concrete 
attributes. 

REFERENCES 
INHELDER, B_, & PlAGET, J. The growth 01 

logical thinking fTom childhood to 
adolescence: An essay on the construction of 
formal opeTational structuTes. New York: 
Basic Books, 1955. 

KOFSKY, E., & OSLER, S. Free c1assification in 
children. Child Development, 1967, 38, 
927-937. 

MACHOTKA, P. Le developpement des criteres 
esthetiques chez l'enfant. Enfance, 1963,4-5, 
357-379_ 

MACHOTKA, P. Aesthetic criteria in childhood: 
J ustifications of preference. Child 
Development, 1966,37, 877-888. 

SUBES, J. L'appreciation esthetique d'oeuvres 
par les enfants. Enfance, 1958, 11, 115-130. 

WALK, R. D. Concept formation and art: Basic 
experiment and controls. Psychonomic 
Science, 1967,9,237-238. 

NOTES 
1. This research was supported in part by 

National Institutes of Health 
Grant I-TOI-HD-203-02, NlH Predoctoral 
Fellowship MH-37,813 to the fust author, and 
NIH Predoctoral Fellowship MH-42,881 to the 
second author_ We gratefully acknowledge the 

Psychon. Sci., 1970, Vol. 18 (2) 



help of Drs. Richard D. Walk and Jacqueline J. 
Goodnow in the preparation of the manuscript. 
First authorship was determined by a flip of a 
coin. 

2. Walk, R. D., & Karusaitis, K. Artistic style 
as conccpt formation for children and adults. 
Unpublished manuscript, Thc George Washington 
University, 1969. 

Effect of R-pretraining on PAL with unlimited 
time for response! 

HIROKO BARNES and ELf SALTZ, 
Center for the Study of Cognitive 
Processes, Wayne State University, Detroit, 
Mich. 48202 

Runquist & English (1964), using 
adjectives as R terms, found that the 
effects of R pretraining on PAL 
disappeared when PAL involved 
(1) unlimited time for response and 
(2) forced availability of the R terms 
du ring PAL. They attributed the. 
disappearance of the R-pretraining effect 
to the use of the forced-availability 
technique and drew theoretical conclusions 
accordingly. The present two experiments 
indicate that, for the type of R terms used 
by Runquist and English, R·pretraining 
effects disappear under unlimited time for 
response even in the absence of forced 
availability. 

Underwood, Runquist, & Schulz (1959) 
found that having Ss leam a set of 
adjectives in free recall facilitated 
subsequent paired-associates learning 
(PAL) involving these adjectives as the 
R terms. They interpreted the data as 
showing that the R pretraining increased 
R availability. SaUz (1961) interpreted the 
data as indicating that R pretraining 
increased R differentiation and tested this 
by use of a foreed-availability technique, in 
which the R terms were present on eards 
during PAL for both pretrained and 
nonpretrained Ss. The Ss were foreed to 
find the eorrect R on the card to be seored 
as eorreet in PAL. Facilitation of PAL 
oe curred, as predicted by the 
R-differentiation theory, despite the foreed 
availability for a11 Ss. However, RUilquist & 
English (l964) argued that, in the Saltz 
(1961) study, Ss were permitted only 4 sec 
per response in PAL; facilitation could be 
due to more rapid scanning of the R cards 
by the pretrained Ss. They tested this by 
using the Saltz forced-availability 
teehnique during PAL, with adjectives as 
Rs and unlimited time to respond. Under 

Psychon. Sei., 1970, Vol. 18 (2) 

these conditions, pretraining no longer 
facilitated PAL. Saltz & Felton (1968) 
obtained the same results as did Runquist 
and English using adjectives as Rs, though 
PAL facilitation occurred when nonsense 
sy11ables were used as Rs. 

None of these studies controlled for the 
possibility that, with adjectives as the 
R terms, R-pretraining effeets disappear 
with unlimited time for response even in 
the absence of forced availability. The 
present study institutes such control 
eonditions for the dissimilar PAL adjectives 
used by Saltz & Felton (1968). 

Points of differenee between the Saltz 
(1961) theory of response (R) 
differentiation and the Underwood, 
Runquist, & Sehultz (1959) R-availability 
theory were tested in studies by Runquist 
& English (1964) and Saltz & FeIton 
(1968). However, a erucial control group 
was ornitted from both these studies. The 
present paper reports the results for this 
control condition. . 

METHan 
The Ss were ISO students in 

introductory psychology at Wayne State 
University, 78 in Experiment 1 and 72 in 
Experiment 2. Experiment 1 used 10 
dissimilar adjectives as R terms; 
Experiment 2 used 5 dissimilar adjectives. 
These were taken from Saltz & Felton 
(1968), Experiments ) and 2, respectively. 

In each experiment, one group reeeived 
relevant pretraining and one group received 
irrelevant, as in Saltz and FeIton. 
Pretraining paralJeled that of the other 
studies cited above: On each of five trials, 
Ss were shown the adjectives (in a different 

TabJe 1 
Mean Errors in PAL as a Function of Relevant 
Pretraining (RP) Versus IrreleYant Pretraining 

(lP) at 2-Sec and Unlimited Time for 
Response Ouring P AL 

Experiment I Experiment 2 

RP IP RP IP 

2-Sec 14.62 19.20 7.94 8.94 
Unlimitcd 8.17 9.87 7.66 6.33 

order on each trial), then asked to reeal! 
them in any order. In PAL, half the Ss 
were permitted unlimited time to respond, 
and half were given 2 see to respond on 
eaeh pair. A study-test procedure was used 
in PAL. 

RESULTS 
Each experiment was designed to permit 

two orthogonal comparisons in PAL: effect 
of relevant (RP) vs irrelevant pretraining 
(IP) at 2-sec anticipation intervaI, and RP 
vs IP at unlimited time for response. 
Table) summarizes the data. In 
Experiment 1, at 2-see anticipation rate 
during PAL, the RP eondition made 
significantly fewer errors during PAL than 
the IP [F(I,74) = 5.36]. Under unlimited 
time for response, the RP vs IP difference 
shrank markedly and was no longer 
significant [F(1,74) = 1.49]. In 
Experiment 2, no significant effeets were 
found at either time intervaI. For the 2-see 
condition, F(I ,68) = 1.14; for unlimited 
time for response, F(1 ,68) = 1.00. 

CONCLUSIONS 
When adjectives are used as R terms, the 

effects of prior R pretraining on 
subsequent P AL largely dissipate if Ss are 
permitted unlimited time for response 
during P AL. This is true if P AL involves 
forced availability or if Ss attempt unaided 
recall of the correet R term for each 
stimulus. Thus, the evidence is eontrary to 
the Runquist and English hypothesis. The 
disappearance of R-pretraining effeets on 
PAL, under eonditions of forced 
availability and unlimited time for 
response, cannot readily be interpreted as 
support for a R-availability theory. Perhaps 
the task becomes so easy, under unlimited 
time for response, that differences between 
experimental eonditions cannot be 
displayed. Or, perhaps increased time for 
response reduces the effect of intralist 
interference and so makes R pretraining 
unnecessary . 

REFERENCES 
RUNQUlST, W. N_, & ENGLISH, J. M. Response 

pretraining and forced availability in 
paired-associate verbal learning. Psychonomic 
Science, 1964, I, 121-122. 

SALTZ, E. Response pretraining: Oüferentiation 
or availability. Journal of Experimental 
Psychology, 1961,62,583-587. 

SAL TZ, E., & FEL TON, M. Response pretraining 
and subsequent paired-associate learning. 
Journal of Experimental Psycbology, 1968, 
77, 258-262. 

UNDERWOOO, B. J., RUNQUIST, W. N., & 
SCHULZ, R. W. Response Jearning in 
paired-associate Iists as a function of intraIist 
similarity. Journal of Experimental 
Psychology, 1959,58,70-78. 

NOTE 
1. Tbe present study was partially supported 

by National Science Foundation Grant 
GD 16851; in addition, the study was carried out 
while tbe fust author beld aNational Science 
F oundation undergraduate research participation 
award under Grant GY 5949. 

81