Three experiments were reported in which 5"s were given a one-trial study-
test paired-associate task and instructed that either word of each pair might
be given as the cue for recall. If associations are symmetrical, the probability
of correct recall should be independent of the word given as the cue. In
Exp. I, two lists of adjective-concrete noun pairs were used. The results
showed recall to be independent of presentation order but superior when
cued with the noun. The results of Exp. II showed this symmetry effect to
be a consequence of the relative abstractness of adjectives rather than of
grammatical factors. This conclusion was supported by a further experiment
in which no asymmetry was obtained when adjectives were paired with
abstract nouns. The results were consistent with an interpretation in terms
of the differential retrieval cues provided by abstract and concrete stimuli.
The distinction between the associative is no distinction between stimulus and re-
and response learning stages in paired-asso- sponse. The 5s are presented with PA
ciate (PA) learning (Underwood & Shulz, items for one study trial and instructed that
1960) has much in common with the distinc- they will be given one (either) item from
tion drawn between the accessibility and each pair as a cue stimulus for the recall of
availability of verbal items (Handler, 1967; the remaining element (cf. Murdock,
Tulving & Pearlstone, 1966). If the cor- 1966b). If the items are high-frequency
rect response (B) is to be given to the words and are assumed therefore to be
stimulus item (A), then B must be present readily available, then the situation should
in storage (i.e., available) and the stimulus, be a favorable one for the demonstration of
A, must provide accessibility or associative associative symmetry; the probability of cor-
cues that will lead to the retrieval of B from rect recall should be independent of the par-
among the total pool of available items. ticular item given as the cue. This specific
This distinction has been used in support- hypothesis will be referred to as retrieval
ing a principle of associative symmetry symmetry.
(Asch & Ebenholtz, 1962; Horowitz, Nor- There is some evidence to suggest that
man, & Day, 1966). As formulated by for certain word combinations associations
Horowitz et al., this hypothesis states that are not established symmetrically. Kusys-
an association between two verbal units, zyn and Paivio (1966) have shown that
A-B, is symmetrical if A and B are equally the recall of adjective-noun (A-N) pairs is
available. Apparent asymmetries obtained a function of word order, performance being
from measures such as backward recall in a superior when they are presented in the
typical PA task are attributed to the differ- order N-A. Such a finding is not neces-
ential availability of the stimulus and re- sarily evidence against the principle of asso-
sponse items; the fact that 5"s are always ciative symmetry since it can be argued that
presented with the stimulus item but have to for a given order of presentation (A-N or
recall the response item from memory pro- N-A), associations are symmetrical but that
motes the relative availability of the latter. a stronger association is established for the
Consider however, a one-trial study-test N-A order. This explanation stresses the
PA task in which, for the study trial, there order of presentation as the critical variable.
1 An alternative interpretation is that the re-
This research was supported by National In-
stitute of Child Health and Human Development sults are independent of the order of pre-
Grant HD00151. sentation but are a function of the differen-
2
Now at the University of Toronto. tial retrieval cues provided by a particular
12
RETRIEVAL ASYMMETRY IN RECALL OF WORDS 13
word (A or N) given as the stimulus for a blank IBM card. At 20-sec. intervals a buzzer
recall. These alternatives are investigated sounded and 6" turned to the next card. Recall
was unpaced; S was given a deck of cards with
in Exp. I. one word typed on each and was instructed to
write in the missing member, working systematically
EXPERIMENT I through the deck, attempting only one item at a
time and never returning to a prior card.
In this experiment recall of A-N pairs While it would be desirable to obtain recall
was investigated as a function of presenta- scores for both adjective and noun cueing from the
tion order (A-N or N-A) and of the cue same 5" on the same item, such a procedure is
for recall (A or N). A third factor con- inappropriate since the two recalls are not inde-
sidered, and one which is of interest in the pendent; the first recall constitutes a possible
second learning trial which would augment per-
use of grammatically structured sequences formance on the second recall of that pair. To
such as A-N pairs, is the operation of selec- overcome this difficulty each 5"s recall was cued
tion restrictions (Chomsky, 1965; Katz & with the noun for half the items and with the
Postal, 1964). Given one word of the pair, adjective for the other half. The particular items
constituting these halves were counterbalanced
such rules limit the lexical items with which across 5"s so that each item was cued with adjec-
that word might be paired. Thus, given tive and noun for an equal number of 5"s. The
the noun "grass," the qualifying adjective order in which the words were typed on the stim-
"short" meets the selection restrictions, but ulus cards was similarly counterbalanced. Thus,
the adjective "angry" does not. Such pair- each S had 10 items in each of the four combina-
tions of conditions derivable from the presentation
ings will be termed appropriate and anomal- order (A-N or N-A) combined with the cue
ous, respectively. Some evidence for the for recall (A or N), the items constituting these
effect of selection rules in the learning of blocks being counterbalanced across 5*s such that
grammatically structured sequences has been each item was placed in each of these conditions for
reported by Rohwer (1966). an equal number of 5s. Study and recall card
decks were well shuffled for each 61 to ensure
a random sequential order of presentation.
Method Subjects.Forty undergraduate students enrolled
Lists.In order to control existing association in an introductory psychology course at Penn-
strength between the A-N pairs, restricted asso- sylvania State University served as >?s, for which
ciation norms were compiled by obtaining noun they received extra grade points. Twenty 5"s were
responses to 120 adjectives, and adjective responses assigned randomly to each list and within each list
to 120 nouns. These norms are based on single randomly assigned to one of the four blocks. Test-
responses from independent samples, i.e., each of ing was conducted in groups of between four and
70 college students instructed to respond with the six 5"s.
first word (of the specified form class) that came to
mind. Results and Discussion
These norms were used to construct two lists,
each consisting of 40 A-N pairs. All nouns were The mean recall scores for the various
highly concrete. Concreteness was established us- conditions are given in Table 1. The order
ing the rating procedure of Spreen and Shulz of presentation has no effect, F < 1, nor is
(1966); a noun was considered concrete if it its interaction with lists or cue stimulus
received a mean rating (based on 20 5"s) of 6.0
or above on the 7-point scale. List 1 contained significant, F < 1 in each case. The su-
40 appropriate A-N pairs in which neither ele-
ment elicited the other with a frequency greater TABLE 1
than one according to the restricted norms. Sam- MEAN NUMBER OF CORRECT RESPONSES FOR LISTS 1
ple items are: small apple, heavy cake, warm dress, AND 2 IN Exp. I
ordinary glove. List 2 contained the same nouns
as in List 1 but paired with adjectives such that List 1 List 2
the pairing was anomalous. Sample items from
List 2 are: busy apple, loud cake, slow dress, clever Presenta-
tion Cue Cue
glove. Order Total Total
Procedure.The one-trial study-test procedure
outlined previously was used. The 5"s were allowed A N A N
20 sec. to study each of the 40 pairs and were in- A-N 5.75 8.30 14.05 4.80 6.50 11.30
structed that either word of the pair might be N-A 5.75 8.40 14.15 5.35 6.50 11.85
given them as the cue for recalling the other mem- Total 11.50 16.70 28.20 10.15 13.00 23.15
ber. Each of the stimulus pairs was typed on
14 ROBERT S. LOCKHART
periority of recall with noun as compared form class relationship of noun qualifier.
with adjective cueing is quite substantial, These alternatives are considered in Exp. II.
-F (1, 38) = 81.4, p < .01. This difference
was obtained for both lists: for List 1, t (38) EXPERIMENT II
= 11.6, p < .01, and for List 2, t (38) = Method
7.1, p < .01. However, this difference be-
tween adjective and noun cueing is greater Word lists.Two lists of 36 pairs were con-
structed for this experiment. Thirty-six adjec-
for appropriate pairs (List 1) than for tives with abstract noun equivalents were selected.
anomalous pairs (List 2 ) ; the interaction The adjective and noun were related in one of
Lists X Cue yields F (1, 38) = 6.63, p < three ways: (a) the adjective was formed by the
.05. Finally, recall (irrespective of cue) is addition of a morpheme to the noun, e.g., danger-
superior for List 1 than for List 2, F (I, dangerous; (6) the noun was formed by the addi-
tion of a morpheme to the adjective, e.g., honest-
38) = 4.31, p < .05. honesty; and (c) the form class transformation is
These results display a very pronounced effected by a change in the stem or a change in
effect of retrieval asymmetry which is clearly the final morpheme, e.g., proud-pride, confident-
independent of the order of presentation. confidence. In order to counterbalance any dif-
ferential effect of these three relationships an
Moreover, while they indicate that adher- equal number (12) of each kind was included.
ence to the selection rules governing the The two lists then were constructed by pairing each
appropriateness of A-N combinations exerts adjective anomalously with a concrete noun (List
a facilitating effect on recall and leads to 3) and pairing the same concrete noun with the
some increase in the superiority of noun abstract noun equivalent of the adjective (List 4).
Hence, these two lists differ only to the extent of
over adjective cueing, this influence is small a morpheme marking the form class of one member
when compared to the main effect across in each pair. Since the A-N pairings are anomal-
both lists. ous, the possible effects of selection restrictions have
It may be concluded, therefore, that when been minimized. Sample items from List 3 are:
agreeable pocket, angry bottle, careful butter,
S"s are presented with A-N combinations, courageous hammer, happy tree, honest piano.
such words are associatively linked in a Samples from List 4 are: agreement pocket, anger
manner which results in retrieval asym- bottle, care butter, courage hammer, happiness tree,
metry, recall being superior when cued with honesty piano.
the noun. Experiment I, however, provides Procedure.The procedure used was identical to
that of the previous experiment. The same coun-
little information concerning the basis of terbalancing of conditions was employed and ex-
tnis observed asymmetry. Two possible tended to hold for each of the sublists of 12 items
interpretations may be considered. First, representing the three kinds of adjective abstract
the effect might be attributed to the relative noun relationships.
Subjects.Forty undergraduate students enrolled
abstractness of adjectives as compared with in an introductory psychology course served as Ss.
concrete nouns. The word-order effect re- Twenty .9s were assigned randomly to each list,
ported by Kusyszyn and Paivio (1966) and within lists were assigned randomly to one of
also has been obtained when abstract nouns the counterbalancing conditions. The experiment
are paired with concrete nouns (e.g., Paivio, was conducted in groups of between four and six 5"s.
1965, 1967; Paivio & Olver, 1964). It has Results
been found that concreteness facilitates recall
with the effect being greater on the stimulus As in the previous experiment, the effect
side. The fact that adjectives and abstract of presentation order was not significant,
nouns have much in common in this respect F < 1. Table 2 gives the mean recall for
is evidenced by the high frequency with Lists 3 and 4 as a function of the cue stim-
which adjectives have abstract noun counter- ulus. Recall is superior when cued with the
parts (e.g., danger-dangerous), the form concrete noun, F (1, 38) = 16.1, p < .01.
class being marked by the presence or ab- However, neither the effect of lists nor the
sence of a final morpheme. This interpreta- interaction Lists X Cue is significant, F < 1,
tion in tefms of abstractness denies the im- and F (1, 38) = 1.9, p > .05, respectively.
portance of form class as such. Second, the The absence of an effect attributable to this
effect might be essentially a function of the interaction indicates that form class per se
RETRIEVAL ASYMMETRY IN RECALL OF WORDS IS
ROSENBERG, S. The influence of grammatical and Journal of Verbal Learning and Verbal Be-
associative habits on verbal learning. In S. havior, 1966, 5, 381-391.
Rosenberg (Ed.), Directions in psycholinguistics. UNDERWOOD, B. J., & SCHULZ, R. W. Meaningful-
New York: Macmillan, 1965. ness and verbal learning. Philadelphia: Lippin-
SPREEN, O., & SHULZ, R. W. Parameters of cott, 1960.
abstractness, meaningfulness, and pronunciability WAUGH, N. C, & NORMAN, D. A. Primary
for 329 nouns. Journal of Verbal Learning and memory. Psychological Review, 1965, 72, 89-
Verbal Behavior, 1966, S, 459-468. 104.
TULVING, E., & PEARLSTONE, Z. Availability versus
accessibility of information in memory for words. (Received December 30, 1967)