You are on page 1of 6

JOURNAL OF VERBAL LEARNING AND VERBAL BEHAVIOR 10, 95-100 (1971)

Syntactical Encoding in Short-Term Memory

JOHN A. HEISEY AND CARL P. D U N C A N

Northwestern University, Evanston, llhnois 60201

Subjects learned and recalled lists of nonsense words presented without sentence structure,
or the same words embedded in key syntactical positions in actwe, passive, or interrogative
voice sentences, for three trials. On Trial 4, all Ss learned and recalled nonsense words pre-
sented in an active voice sentence. In all conditions (active, passwe, interrogative, word hst)
recall decreased, PI (proactive inhibition) increased, from Trial 1 to Trial 3. On Trial 4, the
group continuing with actwe voice showed further PI. The word hst group showed a high
degree of recovery from PI. The passive and mterrogatwe groups showed intermediate
amounts of recovery from PI. The recovery from PI when sentence structure was introduced,
or was changed from one voice to another, is taken to indicate use of syntax as an encoding
cue for recall.

Wickens (1970) developed a procedure for contents of sentences. The retention of a


studying verbal encoding which involves sentence is assumed to involve the storage of a
build-up of and release from proactive inhibi- kernel which is the encoded representation of
tion (PI). The basic method typically consists the sentence content. The encoding and decod-
of presenting Ss with a number (usually three) ing of a kernel is a function of independently
of words from the same class, then a retention retained structural information (Braine, 1963;
test, and so forth for three or four trials. For Chomsky, 1957; Mehler, 1963; Miller,
m a n y encoding classes (semantic, acoustic, 1962).
grammatical, etc.), retention decreases, PI The hypothesis of the present study is that
builds up, over trials. I f a different class of when the same sentence structure is repeatedly
words is then presented, retention of the new used to encode different sentence contents,
class is often higher than that of a control that structural encodings of sentence contents
continues with the original class; there is some should become confused and PI should result.
release from PI. I f this occurs, change to a different structure
In his review of studies using this procedure, should produce release from PI.
Wickens reported that some release from PI To minimize use of semantic information,
had been found with all but one or two of some nonsense words were used as sentence content
14 different ways of encoding or classifying in the present study. The written sentence
verbal units. In another study, Sanders and format is the indicator of sentence structure
Schroots (1969) found varying amounts of when nonsense words are the content (Epstein,
release from PI depending upon whether 1962). However, Marks and Miller (1964)
consonants were followed by digits, tones, or suggest that in such a situation, Ss either dis-
spatial positions. cover and use sentence structure or they do
An encoding category for verbal units that not. If discovery of structure is all-or-none,
has recently been of much interest is syntactic then inclusion of an additional cue to struc-
structure of sentences (Chomsky, 1957). ture should help. Thus, tree diagrams of
Active, passive, or interrogative voice of sentence structure (Jacobs & Rosenbaum,
sentences is thought to be used to encode the 1968) were used as cues in some conditions.
95
96 HEISEY AND DUNCAN

METHOD Table 2 shows an active sentence and its transforms. In


the word list (control) condition, the articles, the and a,
Materials the auxiliaries, did and was, and the function word, by,
Two sets o f 12 content words were used to generate were not used. The two orders o f words in the word
two sets o f four sentences, Sentence Content Sets 1 and 2 list condition matched the different word orders in
(see Table 1). The appearance o f the nonsense words in the passive vs. the active and interrogative conditions.
the sentences was limited by two rules concerning their The suffix, ed, was retained, but punctuation was not
consonant-vowel (CV) sequence: (a) within a content used, m the word list.
set the CV sequence for words with the same syntactical
Design
TABLE 1 A 4 x 2 x 2 design was used. The four structure
conditions were active, passive, and lnterrogaUve
ACTIVEDECLARATIVESENTENCESUSEDAS sentences, and word list. There were two sentence con-
BASESENTENCES tents: the two different word sets shown in Table 1.
The third variable was presence or absence o f the
Content Set 1 graphic (tree diagram) cue.
1. A sart erboked the hmm. Each S in each condition studied, and was tested for
2. The deebs plxed a ticar. retention of, a different sentence on each trial for four
3. A m w e reclled the crad. trials. O n the first three trials, the sentences that S
4. The glers cheped a kale. received were determined by the sentence-structure
condition (active, passive, or interrogative, or word
Content Set 2
list) to which he was assigned. O n the fourth trial, all
1. A nerf enleped the bott.
Ss in all conditions received an active voice sentence.
2. The koobs teged a dlson.
The content S received was from either Set 1 or
3. A pabe desaked the shum.
Set 2 in Table I. These contents were used to generate
5. The crogs shzed a sote.
the first three sentences or word lists S received. Order
o f presentation o f sentences was counterbalanced over
the first three trials. On the fourth trial, half the Ss
function (e.g., nouns) could not be identical (thus, sart,
received Sentence 4 from Set 1, half received Sentence 4
deebs, etc.), except the verb suffix, ed, and (b) between
from Set 2, exactly as they are shown in Table 1. This
content sets, each sentence was matched, word for
was the target sentence, testing release from PI.
word, in CV sequence. A b o u t half of the nonsense
words were taken from Epstein (1962); the remainder
were produced to meet the above conditions.
Sentence
TABLE 2

STRUCTURALMANIPULATIONSON A BASESENTENCE the glers~


Transformation Experimental structure

Active The glers cheped a kale.


Passive A kale was cheped by the glers.
Interrogative Did the glers chepe a kale ?
The glers cheped a kale
W o r d list glers cheped kale
kale cheped glers

The three content words in each sentence were used FIG. 1. Active sentence with corresponding tree
consistently in all sentence structure c o n d m o n s . Thus, structure above the dashed line.
the n o u n phrase, the koobs, always occurred as a n o u n
phrase in all voice transformations. It is defined as a The graphic cue was drawn on the same 5 x 8-m.
noun phrase by its lawfully constrained response to the card on which a sentence or a word list was typed.
application o f voice transformations (Jacobs & Rosen- Figure 1 shows an example. In the word list condition,
baum, 1968). the graphic cue consisted of lines drawn from a terminal
Sentence structure was manipulated by applying point marked Set (vs. Sentence in other conditions)
voice transformations to the base sentences in Table 1. directly to each o f the three nonsense words. In
SYNTACTICAL ENCODING IN SHORT-TERM MEMORY 97

sentence-structure conditions, the tree diagram was the r e s p e c t i v e l y (e.g., sark f o r sart, glars f o r glers).
same for the first three trials. The active tree-diagrams A w o r d was c o u n t e d i n c o r r e c t w h e n e v e r t h e
were used on Trial 4 for all conditions.
first letter was w r o n g , o r if a n E n g l i s h w o r d
Subjects r e s u l t e d f r o m a single letter s u b s t i t u t i o n (e.g.,
The Ss were 192 students from introductory psy- pare f o r pake).
chology course, 12 in each of the 16 conditions. All
were natwe speakers of English. Recall on Trials 1-3
Procedure I n e a c h c o n d i t i o n o v e r t h e first t h r e e trials,
The E instructed S that he would be shown a group of Ss r e c a l l e d d i f f e r e n t n o n s e n s e w o r d s o n e a c h
words, including some nonsense words, on a card for a trial b u t r e p e a t e d t h e s a m e s e n t e n c e s t r u c t u r e .
brief period. Next S would be shown a number from A n a l y s i s o f v a r i a n c e o f recall o v e r t h e t h r e e
which he was to count backwards by threes. Then S
trials r e v e a l e d t h a t t h e effect o f trials w a s
would be asked to write down as many of the words, In
the same left-to-right order m which they had significant, F ( 2 , 3 5 2 ) = 58.06, p < .001. P r e -
appeared, as he could recall. The E showed S a practice sence o r a b s e n c e o f t h e g r a p h i c cue w a s n o t
set of words on a card, and gave an example of counting significant, n o r was s e n t e n c e s t r u c t u r e (active,
backwards by threes. For Ss who were to see the gra- p a s s w e , i n t e r r o g a t i v e , o r w o r d list). N o i n t e r -
phic cue, E told S that he would be shown a way in
a c t i o n was significant. I n all f o u r s t r u c t u r e
which the words might be fit together, and that this
cue would help him remember the words and their c o n d i t i o n s , recall d e c h n e d o v e r trials, as m a y
proper order. The S was told that he would not be be seen in T a b l e 3.
asked to recall the cue.
In the noncued conditions each set (trial) of words TABLE 3
was exposed for 20 sec. In the cued conditions the
MEAN NONSENSE WORDS RECALLED ON EACH
words were exposed for 5 sec., then the card containing
TRIAL BY EACH OF THE EIGHT GROUPS
the cue was exposed for 15 sec. At the end of the 20-sec.
study period, E covered all materials with a card on Trial
which was printed some two-digit numbers from 51 to
99. The S counted backward by threes from this num- Condition 1 2 3 4
ber for 15 sec. The E then gave S 30 sec. in which to
write down as many of the words as he could, and in Noncued Active 3 00 2.50 2.21 1.92
the original order The S was told that ff he could not Passive 2.96 2.62 2.12 233
remember a word, to leave a blank space and go on to Interrogative 2.96 2.62 2.29 2.67
another word At the end of the 30-sec. period the next Word list 2.87 2.62 2.38 2.83
trial was begun. Cued Active 2.92 2.38 2.25 1.79
Passive 2.79 2.46 2.17 2.17
RESULTS Interrogative 2.87 2.29 1.87 2.08
Word list 2.79 2.50 2.21 2.54
E a c h S a t t e m p t e d to recall f o u r s e n t e n c e s o r
w o r d lists, e a c h c o n t a i n i n g t h r e e n o n s e n s e
w o r d s . R e c a l l o f c o n t e n t was m e a s u r e d in t w o Recall on Trial 4
w a y s : (a) t h e n u m b e r o f n o n s e n s e w o r d s p e r O n T r i a l 4, Ss w h o h a d h a d a c t i v e s e n t e n c e
s e n t e n c e o r list c o r r e c t l y r e c a l l e d w a s scored, s t r u c t u r e o n T r i a l s 1-3 c o n t i n u e d w i t h a c t i v e
u s i n g l e n i e n t spelling c r i t e r i a ; (b) t h e n u m b e r structure. T h e Ss w h o h a d h a d p a s s i v e o r
o f n o n s e n s e w o r d letters r e c a l l e d was s c o r e d i n t e r r o g a t i v e s t r u c t u r e , o r w o r d list, c h a n g e d
as a p e r c e n t a g e o f t o t a l letters ( w h i c h v a r i e d t o a c t i v e s t r u c t u r e . T h e r e f o r e , in t h e a n a l y s i s
f r o m 14 t o 15) in all t h r e e n o n s e n s e w o r d s . I n o f v a r i a n c e f o r w o r d recall o n T r i a l 4, a c o m -
scoring number of words a misspelled word p a r i s o n was m a d e b e t w e e n t h e a c t i v e v o i c e
was allowed correct when word pronunciation c o n d i t i o n a n d the c o m b i n e d p a s s i v e , i n t e r r o -
was n o t s u b s t a n t i a l l y c h a n g e d (e.g., soat f o r gative, a n d w o r d list c o n d i t i o n s . T h i s c o m -
sote), o r w h e n a c o n s o n a n t o r a v o w e l was p a r i s o n w a s h i g h l y significant, F ( 1 , 1 7 6 ) =
replaced by another consonant or vowel, 20.99, p < .001 ; the active c o n d i t i o n s h o w e d
98 HEISEY AND DUNCAN

further PI while the other conditions showed Relation of Word to Voice Recall
some recovery from PI. A comparison between As shown in Table 4, in the passive voice
passive and interrogative conditions combined condition there were 25 Ss who showed
versus the word list gave F(1, 176)= 7.71, correct (active) voice recall on Trial 4, and
p < .01 ; the word list condition showed greater 19 Ss who showed voice error. For the inter-
recovery from PI than these two structure rogative condition, these subgroups contained
conditions. The passive and interrogative 31 Ss and 12 Ss, respectively. Combining these
voice conditions did not differ, F < 1. groups, there were 56 Ss who exhibited
The cue variable was significant, F(1, 176) correct voice on Trial 4 and 31 Ss who were
= 6.99, p < .01; the cued Ss recalled fewer incorrect in voice recall. The correct voice Ss
words than the noncued Ss on Trial 4. Because also recalled slightly more words on Trial 4,
of this, the means in Table 3 are given separ- a mean of 2.62 words versus 2.06 for the
ately for cued and noncued conditions on all incorrect voice Ss, t = 3.50,p < .01. (The voice
four trials even though cue was not significant correct Ss had also recalled slightly more
on Trials 1-3. words on Trial 3, 2.30 vs. 2.13 for incorrect
The only other significant term in the voice Ss, but the means do not differ signifi-
analysis was sentence content, F(1, 176)= cantly.) The results were the same for per-
8.03. Words from the test sentence in Content centage of letters recalled; voice correct and
Set 2 (Sentences 4 in Table 1) were most voice incorrect Ss did not differ on Trial 3
difficult to recall. No interaction was signifi- (80.2 and 78.0 ~ , respectively) but did differ
cant. on Trial 4 (87.3 and 61.3 ~ , t = 7.56, p < .001).
The percentage of nonsense word letters
recalled on Trials 1-3 and on Trial 4 was TABLE 4
analyzed. The results were essentially iden-
NUMBER OF as IN EACH STRUCTURE CONDITION WHO
tical with those for word recall. MADE EACH OF THREE TYPES OF RECALLOF THE VOICE
OF THE ACTIVE SENTENCE ON TRIAL 4
Syntactic Recall
Structure condition
On Trial 4 all Ss received sentences in the
active voice. Two judges scored syntactic Recall type Active Passive Interrogative Word list
recall, using three categories: (a) sentence
Voice correct 43 25 31 47
fragment, the recalled words were not suffi-
Voiceerror 0 19 12 0
cient in number to determine the voice of the Sentence
sentence, (b) voice error, the sentence was fragment 5 4 5 1
recalled in the wrong voice, shown by the
presence of did, indicating a question, or was
DISCUSSION
by, indicating the passive, (c) correct recall of
sentence voice. The two judges agreed perfectly In the first three trials, Ss learned and re-
in scoring. Table 4 shows the number of Ss in called nonsense words under eight different
each syntactic recall category. Analysis of this conditions: active, passive, or interrogative
matrix yielded X2 = 45.62, p < .001. The Ss sentence structure, or no structure (word list),
who had had passive voice sentences on Trials and presence or absence of the graphic cue. In
1-3 made only passive voice errors on the light of Wickens' (1970) review of PI in short-
active voice sentence of Trial 4, and the inter- term memory, it is not surprising to find that
rogative condition Ss made only interrogative this study also found build-up of PI over trials.
errors. Thus, syntactic recall on Trial 4 seems However, it is worth noting that the PI build-
dependent on the type of voice experienced on up did not vary as a function of any of the ex-
previous trials. perimental conditions. Although not specific-
SYNTACTICAL ENCODING IN SHORT-TERM MEMORY 99

ally hypothesized, it had initially seemed method for computing loss and recovery from
reasonable to expect that PI build-up might PI, the active structure (not shifted) group
proceed at different rates for the three sentence showed a loss of 37.3 ~ from Trial 1 to Trial 4.
structure groups versus the no-structure word The combined passive and interrogative
list group. This did not occur, nor did the groups recovered 41.5 ~ of this loss, while the
attempt to emphasize structure by the graphic word list group recovered 75.5 ~ of the loss.
cue have any effect in the first three trials. As This latter value for recovery is exceeded by
Wickens' (1970) summary of 14 different only one or two, and equalled by one or two,
classes of materials that have been used in of the 14 types of shift for which Wickens
short-term PI studies indicates, development computed recovery percentages. The 4 1 . 5 ~
of PI over successive trials on the same class value, for the passive and interrogative Ss who
(variously defined) is a very general pheno- experienced sentence structure shift, is about
menon. the median level of recovery reported by
After PI has been built up by a few trials on Wickens.
the same class, changing to a different class of The word list Ss did not have syntax during
materials on a final trial produces varying the first three trials and so could not use syn-
results. Partial release from PI (better reten- tax as an encoding cue. When sentence struc-
tion on the final trial compared to the im- ture was first introduced to these Ss on Trial 4,
mediately preceding trial, but rarely as good as the effect was to produce a large recovery from
on the first trial) occurs for some but not all PI. Furthermore, the fact that 47 out of these
shifts of class. More quantitatively, Wickens 48 Ss correctly recalled sentence structure on
presented a bar graph (Wickens, 1970; Fig. 7, Trial 4 (Table 4) suggests that they used syntax
p. 7) showing recovery from PI varying from as an encoding cue when it was made available.
very large amounts in semantic-type classes to The passive and interrogative conditions
statistically insignificant amounts in classes are perhaps better test cases for syntax as an
based on physical characteristics of verbal encoding cue. Since both groups had one type
units. of voice initially, then were shifted to another,
In the present study, there were three dis- it was possible to measure whether they had
tinct levels of performance on the final trial. been using structure by whether they shifted.
The lowest level (poorest retention) was shown As noted earlier, 56 Ss did shift to the active
by the active sentence structure condition, voice on Trial 4 but 31 Ss did not. These 31
which condition corresponds to the control Ss continued to use the voice they had had on
group in the experiments covered by Wickens, the first three trials. This meant that in trying
that is, Ss who continued with the same class to recall the Trial 4 sentence, they transformed
on all trials. The active structure group it from the active voice in which it had been
showed, as have other control groups, further presented to the voice in which the three pre-
PI build-up on the final trial. A higher level of vious sentences had been presented. Thus, at
retention, and some recovery from PI, was least for some Ss, syntax can be used to such a
shown by the two groups that had had passive degree that they become fixated on one type
or interrogative sentence structure on the first of voice.
three trials, when they were shifted to active
structure on the last trial. These groups did REFERENCES
not differ in amount of PI recovery. The
greatest recovery was shown by word-list Ss, BRAINE, M. O. S. On learning the grammatical order
of words Psychological Revzew, 1963, 70, 323-
who had not experienced any kind of sentence 348
structure until presented with the active CHOMSKY,N Aspects of a theory of syntax. Cambridge,
structure on the final trial. Using Wickens' Mass : M.I.T. Press, 1965.
100 HEISEY AND DUNCAN

EPSTEIN, W. The influence of syntactic structure on MILLER,G. A. Some psychological studies of grammar.
learning. American Journal of Psychology, 1962, American Psychologist, 1962, 17, 748-762.
75, 121-126. SANDERS,A. F., & SCHROOTS,J. J. F. Cogmtive cate-
JACO~S,R. A., and ROSENBAUM,P. S. English transfor- gories and memory span. III. Effects of similarity
mational grammar. Waltham, Mass.: Blaisdell on recall. Quarterly Journal of Experimental
Co., 1968. Psychology, 1969, 21, 21-28.
MARKS,A., & MILLER,G. A. The role of semantic and WlCrENS, D. D. Encoding categories of words: An
syntactic constraint m the memorization of Eng- empxncal approach to meaning. Psychological
lish sentences. Journal of Verbal Learning and Review, 1970, 77, 1-15.
Verbal Behavior, 1964, 3, 1-5.
MEI~LER,J. Effects of grammatical transformations on
the recall of English sentences. Journal of Verbal
Learning and Verbal Behavior, 1963.2, 346-350. (Received August 19, 1970)

You might also like