MEMORY FOR LINGUISTIC MATERIALS

Thesis for the Degree of Ph. D.
MICHIGAN STATE UNIVERSITY
SANDRA ELAINE GRAHAM
1972

ALI!IlflflfllfljflflfllflflfllflMI!UISIIIIIIINLIII ° A __ ___

tr"?
2 Y n.
A =:

I

g**‘[ Si'f: is

U “.I 3‘” f" .r
‘- .

Winn“ *W'

This is to certify that the

thesis entitled

MEMORY FOR LINGUISTIC MATERIALS

presented by

Sandra Blaine Graha-

has been accepted towards fulfillment
of the requirements for

Jam—degree in My

Major professor

T.I<te:L5§‘\ \‘fi'nfi

7639

 

 

 

 

in”

. .‘,.;...r."'ut.~pr - V 7’! r ".z’
/ C} {a
‘ i/

t

W

8 4 ‘r' (p: ;, ¢ r
qt . ,J ‘
l 3

JAN

 

ABSTRACT
EMORY FOR LINGUISTIC MATERIALS
3?
Sandra Elaine Graham

ho experiment: were conducted to assess the
effects of surface structm‘e complexity and deep structure
complexity on the recall of linguistic materials. It was
hypothesised that surface structure complexity would influence
short term memory (81!!) in that more complex surface
structures should require more time to be constructed in 8‘11!
than. less complex surface structures. The hypothesis was
tested in a within subject design using two types of phrases:
which differed only in their surface structure complexity.
in overflow test was employed on the assumption that the
more complex phrases would require more analysis time and
that fewer of the items following the phrase would then he
recalled. 'lhe results showed no difference in recall of the
phrases or in recall of the items which followed them.

A second hypothesis considered was that the
complexity of the deep structurewould influence short term
retention of linguistic materials in that more complex deep
structures would require more analysis time in a long term

Sandra Elaine Graham
memory access loop postulated to be present in STM, and
thus require the surface structure input to be held longer
in 81!! than would a sentence with a less complex deep structure.
Deep structure complexity and presentation rate were
manipulated in a between subject 2x2 factorial design using
two types of sentences, which differed only in deep structure
complexity, and presentation rates of two and four seconds.
a 128 seconds nunber shadowing interference task was used
between presentation of a sentence and recall. The results
showed deep structure complexity but not presentation rate
to have a significant (p <.Ol) effect on the number of
sentences recalled. The results suggested that subjects were
able to obtain the same number of words from the surface
structure input for both types of sentences, but that for
these sentences labeled as more complex, it was significantly
more difficult to construct the deep structure and obtain

the meaning of the sentence.

MEMORY FOR LINQUISTIC MATERIALS

33‘

Sandra Elaine Graham

ATHESIS

Submitted to
Hichigan State University
in partial fulfillment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY
Department of Psychology

1972

/
L4 scmwmncnsm's

4610’”
The author wishes to thank her guidance committee
Dr. Gordon Weed, chairman. Dr. Abram Brch, Dr. Lester
mien. and Dr. Donald Johnson for their help and guidance
in the preparation of this thesis. m. assistance of the
late Dr. Hillia- T. Stellwagen was alse instrumental in the
preparatien ef this thesis.

ll,

TABLE OF CONTENTS

Page
HST OF TABLES 0.0.0000.00.0.0000.0.00000000000000000000.1'

INTRODUCTIO' 00......OOOOOOOOOOOOOOOOOOOOOOO0......O00.0.01
man I COOOCOOOOOOOOOOOOIOOO000......0.00.00.00.000009

lethed
Results and Discussion

mum. II 0.0.0.0....00..0....0.0.0.....0...‘..OOO...1#

Hethod
Results and Di scussien

GEEK-AL DISCUSSIOIOOOOOOOOOOOOOO...0.0.0.00000000000000002u
LIST OF REFERENCES 0.00.0.0...COO...00.00000000000000000027

APPENDIX A Transfermatienal history ef noun phrases
“led in Experiment Ieeeeeeeeeeeeeeeeeeeeeeeee29

APPENDIX B Noun phrases and strings of words used in
hperlnent I eeeeeeeeeeeeeeeeeeeeeeeeeeeeeeee32

APPENDIX C Gm pro-test {Or Experllent I e e e e e e e e e e e33
APPENDIX D &' data Experiment I eeeeeeeeeeeeeeeeeeeeeeeBu
APPENDIX E Surface and deep structures of Experiaent II

sentences eeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeee3

APPENDIX E Sentences mad 1n Experllent II e e e e e e e e e e e e e38

APPENDIX C? Mean umber of errers per sentence in
EXPOrlnent II eeeeeeeeeeeeeeeeeeeeeeeeeeeeeee39

APPme E fit! data Experiment II ......................#0

111

LIST OF TABLES

Tuble Page

1. loans and standard deviations for the number
of sentences and the number of words recalled...l7

2. Hhans and standard deviations for the number
0: Ornr' made by “Oh 00m1t10n ...............20

3. Percentages of the tetal errers of the various
618.3.. ”d. by OBOh 0°m1t1.n eeeeeeeeeeeeeeeee21

iv

IN‘IRODUCTION

The study of memory for linguistic materials has
generally been an effort to demonstrate the psychological
reality of the linguistic concepts of surface structure and
deep structure. Toward this end a number of investigators
have reported evidence which they feel demonstrate the
influence of surface structure on sentence retention (Martin
a Roberts. 1966; Martin. Roberts. 8: Collins. 1968; Roberts.
1968; and Hearing. 1970). These investigators have hypothesized
a direct relationship between the ease of recall of a sentence
and its surface structure complexity. 'Yngve' numbers
(Yngve, 1960; Martin a: Roberts. 1966). which are the
average of the number of left branches leading to a terminal
element (word) in the phrase structure tree, have been used
to assess the complexity of the surface structure. In general,
the above investigators have found that the simpler the
surface structure. the greater the probability of recalling
the sentence. Martin and Roberts (1966). Perfetti (1969 a. b).
and Wright (1969). however. have not been able to obtain
the hypothesised relationship between surface structure
complexity and sentence recall.

Adding to this confusion are the studies which
assess surface structure complexity in terms of master of

l

,’l

[I

2;
right-branches and number of self-embedded sentences contained
in a sentence (Miller a: Isard, 19614.; and Foss & Cairns,
1970). mese studies have found that the larger the number
of right branches or self-embedded sentences, that is the
more complex the surface structure, the more difficult it
becomes for subjects to recall the sentence (number of words,
meaning, and deep structure held constant. Fees and Cairns
(1970) argued that increasing sentence complexity made it
more difficult to rehearse and store linguistic materials.
They also argued that linguistic materials which were not
rehearsed and stored (unanalysed) were forgotten more rapidly
than materials which had been analysed. Fees and Cairns:
tested this hypothesis by having subjects recall right-
branching and self-embedded sentences, prefacing recall with
a list of words which were to be recalled or read out loud.
The reading or recall of the list of words was used to inter-
rupt rehearsal of the sentence. Their results supported
their hypothesis. Fees and Cairns concluded that when a
sentence is heard/read it is first analysed in term of
surface structure.

For the most part these above investigators have
held that surface structure complexity alone would account
for the observed differences in sentence memory. However.
not all investigators agree with them. Hehler (1963), Savin
and Perchoncck (1965), Rohrman (1968), and Davidson (1969)
have suggested that the likelihood of recall of a sentence

3
is directly related to the canplexity of the deep structure.

'Ihe index of complexity in all but Rohrman's study has been
the number of transformations needed to transform deep
structure into surface structure.

Rohrman (1968) used the complexity of the deep
structure base as his complexity measure. He found that the
ease of recall was directly related to the complexity of the
deep structure base and was unaffected by surface structure
complexity or number of transformations. However, Rohrman's
results have been shown to be artifical by Rohrman (1970) and
Paivio (1971). Ronrman (1970) upon further testing with his
materials demonstrated,that the transitive-intransitive
dimension had been confounded with the complexity of the
phrases in his previous study. His more complex phrases all
contained transitive verbs while the less complex ones all
contained intransitive verbs. when he controlled for deep
structure complexity and varied the type of verb, intransitive
phrases were recalled more of ten than transitive phrases.
Paivio (1971) has shown that Rohrman's eariler results (1966)
can also be accounted for in terms of imagery value.
Rohrman's shpler phrases were higher in imagery value than
his complex phrases.

levertheless, Mahler and Carey (1967) have reported
results which do argue for the psychological reality of deep
structure. Subjects were presented sentences which could
be accurately heard only fifty percent of the time. line

sentences which did not differ in deep structure or in

ll

surface structure were presented to subjects for immediate
recall. A tenth sentence was then presented which had the
same surface structure as the proceeding nine sentences but
a different deep structure. The change in the deep structure
of the final sentence was found to make its recall more
difficult. These results suggest that a set for deep
structure was established during the first nine sentences,
and argue for the psychological reality of deep structure.

In all the above studies, surface structure and
deep structure have been treated as if they were independent
concepts, and as if only one could be involved in sentence
memory. However, the medel (Chomsky, 19573 1965) in which
these concepts were first suggested treats them as related
subcomponents of the syntactic component. is such, surface
structure and deep structure should be related in any attempt
to account for sentence memory.

Miller and Chantry (1963:) have sketched a madel
of sentence memory which entails both surface structure and
deep structure. They preposed that short term memory (81!!)
processes the surface structure of a sentence symbol by '
symbol as it is received and then transmits the resulting
string to long term memory (LTH). In ether words, in am
subjects construct the terminal phrase marker which corresponds
to the surface structure and it is this phrase marker that
is transmitted to L'm. The 15TH contains a generative grammar
of the language and determines the deep structure and hence

the meaning of the string received from 3134.

5’

Neisser (1966) has proposed a medel of cognition
which postulates processes which are very similar to those
of Miller and Chomsky. Neisser proposed a two stage medal.
The initial stage of perception or cognition is a short
transient stage called iconic memory (for visual stimuli).
During this initial stage of iconic memory an image of
the visual input is constructed by the subject and briefly
stored. This image is subject to rapid decay. If it is not
processed by the next stage of active verbal memory, the
information contained in the image is lost. (a similar
process called echolic memory is postulated to handle
auditory stimuli.)

The active verbal memory is the short term store
where the information from the iconic image is synthesised
into the cognitive counterpart of the physical stimulus
and then rehearsed until the subject is required to reproduce
the verbal stimulus. For the active verbal memory to
carry out this synthesis it must have access to the long term
store where the grammar of the language with its lexicon is
stored. Otherwise, the reproduction would be merely a
simple echo response.

Miller and Chomsky's proposal lends itself to the
following interpretation in light of Neisser's model.

Surface structure complexity should affect both iconic
memory and active verbal memory. The more complex the surface
structure, the more difficult it should be for subjects to

form and hold the image in the iconic memory and to then

6

construct the surface structure. However, the effect of deep
structure complexity is more complicated. The deep structure
may be processed either directly by the L!!! processes or
through the active verbal memory's LTM loop. Which route
is used is most lkiely centralled by the kind of task
subjects are faced with. If subjects are required to store
the information for distant future use, the active verbal
memory could be by-passed. If, however, subjects are asked
to recall the material within a brief periOd of time, the
active verbal memory route most likely is used.

when the information is being stored for future
use, the complexity of the deep structure should affect
memory by making it more difficult to extract the information
as the complexity increases . ‘Hhen.the task requires subjects
to use the active verbal memory route, deep structure
complexity should affect memory for linguistic materials by
increasing the analysis time in the LE! loop as the complexity
of the deep structure increases. The increasing analysis
time should necessitate holding the surface structure in the
active verbal memory for increasing lengths of thme because
the surface structure is the information source for the
analysis. When rehearsal is interfered with, the effect of
deep structure complexity should become apparent.

The present study was intended to test two
hypotheses generated by the Miller-Chomsky prOposition and
Neisser's model. First, surface structure complexity should

have an effect on recall of linguistic materials when

7

presentation rates are fast because as the complexity of the
surface structure increases, more time should be required by
the active verbal memory to construct the surface structure
prior to feeding it into the LTM loop. With short exposure
times, less of the complex surface structure should be
constructed and thus there would be less available information :1}
for the meaning analysis and subsequent recall. An overflow I
test (Epstein, 1969), which requires subjects to recall the ‘*
critical material plus some extra items, was used to test this
hypochesis on the assumption that if the more complex surface
structure required more time for construction, then fewer of
the additional items should be processed through the active
verbal memory.
The second hypothesis proposed that the complexity
of the deep structure should influence reproauction of
linguistic materials when rehearsal of the surface structure
input is interfered with shortly after exposure of the
materials. A delayed recall task with a demanding intervening

task was used on the basis of two assumptions. First, it was

assumed that the nature of the task would require subjects
to process the sentence via the um loop of the active verbal

memory. Second, it was assumed that the deep structure of a

simple sentence would be processed farther than that of a
complex sentence by the time that the intervening task inter-
fered with the subjects's ability to hold the surface structure

input in the active verbal memory. Hence, recall of the

8 I
simpler sentence should be better. An interaction of
presentation rate and deep structure complexity was

predicted.

EXPERIMENT I

Method

Design and materials. The effect of surface
structure complexity or noun phrases on the immediate recall
of the phrases was tested in a within-subjects-design which
varied the complexity of the surface structure of the noun
phrases while hOlding the deep structure complexity constant.
Noun phrases of the type, 8 (simple) “yellow growling lions"
and 0 (complex) I'yellow lions growling," which have the same
deep structure and equal number of words in the surface
structure but differ in surface structure complexity, were
used. (See Appendix a.for surface structure trees and
derivational history of the noun phrases.) The ratio or the
number of noses in the surface structure phrase marker tree
to the number or terminal elements ('words') was used to
compare surface structure complexity of the noun phrases
(Miller & Chomsky, 1963). Using this metnOd, simple noun
phrases have a 2.0 ratio while complex noun phrases have a
2.66 ratio. Twelve pairs of noun phrases were used with
twelve random strings of eight words each which had Thorndike-
Lorgs frequency ratings of a and AA (see Appendix B).

Procedure. Sixteen Michigan State University
undergraduates from.intr0ductory psychOlogy classes served

as subjects in fulfilument of class requirements. Subjects
9

10
were first given a grammar pre-test. They were presented a
list of 16 three-word-strings, half of which were sentences
and half of whichmwere noun phrases. on a single sheet of
paper (see Appendix 0). Next to each string was printed
the words ”sentence" and ' phrase". For each string subjects
were asked to indicate whether they thought the string was
a sentence or a phrase by circling the apprOpriate word.
This test was to assess any tendency on the part of the
subjects to treat the noun phrases in Experiment I as sentences.

Following the pre-test, subjects were told that
a three-word-phrase followed by sight words would be flashed
on the screen. They were instructed to try to determine what
the phrase was and to then determine as many of the words as
they could. Emphasis was placed on retaining the phrase even
if this reduced the number of words they could recall and on
recalling the phrase before the words. In an effort to obtain
recall in the desired order, phrase then string;of words.
subjects wrote their recall in booklets in which they were
required to write the phrase on one page and then to turn to
the next page to write the string of words.

The noun phrases were selected such that half of
the subjects viewed six simple and six complex phrases with
only one member of each of the twelve pairs being selected.
The other half of the subjects viewed the six simple and six
complex phrases which the other subjects had not seen. (That
is, one group of subjects saw the phrase "yellow growling
lions" while the other group“saw the phrase ”yellow lions

— ll
growling".) The noun phrases were presented in a random
order on black and white slides printed in capital letters.
The strings of sight words were also presented on slides,
two words per slide. The slides were flashed at a two second
rate using a slide projector which provided a one second
effective viewing time per slide. Tbtal time for each phrase
+ string was ten seconds. Subjects were given one minute
for recall immediately following presentation of each phrase
+ string. After all the phrases had been presented subjects
were asked to attempt a final recall of the phrases.
Results and Discussion

The mean number of correct responses to the grammar
pro-test was 1h.13 which was significantly better than chance
performance, t(lS) a 5.31, p‘<}OOl. Subjects were able to
distinguish three-word-phrases from.three-word-sentences.
Hence, it seems reasonable to assume that they treated the
three-word-strings of the experiment as phrases and not as
sentences.

While correct recall of the phrases was emphasized
in the instructions, none of the subjects succeeded in recall-
ing all of the phrases during the immediate recall task. The
mean number of simple noun phrases recalled was 3.87 and the
mean number of complex noun phrases recalled was n.06. The
difference was not significant, t(lS) a 0.306, p >.05. That
recall of the phrases, with eight seconds intervening between
exposure and recall, was less than perfect may be due to

interference with rehearsal of‘the phrase produced by

12:,
processing the words of the string.

The mean total number of words recalled after
correct simple noun phrases was 1h.0 and after correct complex
noun phrases 11.5. The difference was not significant,
t(1§) I 0.819, pj>.05. The mean number of phrases recalled
on the final recall test was 0.93 for the simple noun phrases
and 1.62 for the complex phrases. This difference was also
not significant. t(15) I 1.699. P 7.05.

The results of the present study do not permit
the acceptance of the hypothesis that surface structure
complexity affects the active verbal memory process. Both
types of phrases were recalled equally well and had similar
levels of recall for the strings of words. This failure to
find a difference, on the one hand, suggests that for these
pairs of phrases there was no difference in surface structure
complexity. The findingo, thus, do not necessarily mean that
surface structure complexity does not influence memory for
linguistic materials. Nor do they necessarily conflict with
previous findings'which have supported the hypothesis that
surface structure complexity does influence the recall of
linguistic materials. While these studies (Martin & Roberts,
1966: Martin, Roberts, & Collins, 1968; Roberts, 1968: and
Hearing, 1970) have used exposure rates which have been as
long or longer than the one employed in the present study,
they have also used more complex linguistic materials:
sentences as opposed to phrases. 0n the other hand, if the
initial assumption thatgthe phrases did differ in surface

13

structure complexity is retained, then there is no alternative
but to reject the hypothesis and conclude that surface
structure complexity does not influence reproduction of
linguistic materials.

The evidence from the first experiment does not
permit a choice to be made between these two positions.
The results of the following study do permit some conclusionsw
as to the apprOpriateness of transtormational theory to

memory for language materials.

EXPERIMENT II

The purpose of Experiment II was to assess the
effect of deep structure complexity and presentation rate
on sentence recall. The assumption was made that the delayed
recall task would require subjects to process the sentence
via the LTM loop of the active verbal memory rather than
directly processing it in the long term memory because they
were asked to attempt exact reproductions rather than to
produce a sentence of similar meaning. Given that such an
assumption holds, the hypothesis under consideration predicted
that the deep structures of simple sentences would be pro-
cessed farther than those of complex sentences by the time
that the intervening task interfered with subjects' ability
to hold the surface structure input in the active verbal
memory and, thus, affect the ability of the subjects to
reproduce the sentence. This hypothesis also assumes that
the surface structure input must be held for the time required
to analyze the deep structure or the analysis would be
incomplete. This seems reasonable in that the surface
structure is the only source of information input on which
the analysis can be based.

Method

Materials and desigg. Two types of nine-word
'lh

15
sentences were used. Both had surface structures of equal
complexity but varied in the complexity of the deep structure
bases (see Appendix E). The complexity of the surface
structures was measured by taking the ratio of the number of
nodes in the surface structure phrase marker to the number
of terminal elements ("words”). Both types of sentences had
surface structure complexity ratios of 2.0. The complexity
of the deep structures was measured by taking the ratio of
the number of nodes in the deep structure base to the nunber
of terminal elements (complex symbols) in the deep structure
base. Sentences of the form, "The women baking the bread
say it is good." (simple sentence - 8). had a deep structure
complexity ratio of 2.5. Sentences of the form, "The
explorers mapping the island found rich buried treasure."
(complex sentence -C), had a deep structure complexity
ratio of 3.1. Thus, the latter sentence was more complex
in its deep structure than the former sentence. Eight
sentences of each type were constructed (see Appendix F).

' Sentence type and presentation rate were
manipulated in a 2x2 factorial-between-subjects design. The
two sentence types were combined with presentation rates of
two seconds and four seconds to yield four conditions to
which subjects were assigned. In one condition, subjects
were presented simple sentences for two seconds per sentence
(Condition 3-2). In a second condition, subjects were presented
simple sentences for four seconds per sentence (Condition

s-h). In a third condition, subjects were presented complex

 

16

sentences for two seconds per sentence (Condition 0-2). In
the fourth condition, subjects were presented complex sentences
for four seconds per sentence (Condition C-h). All subjects
were given two trials on each sentence.

grocedure. Twenty-eight Michigan State University
undergraduates from introductory psychology classes served
as subjects in fulfillment of class requirements. Subjects
were assigned randomly to conditions with seven subjects
per condition. Subjects were tested individually. Sentences
were presented on a.memory drum.for either two or four
seconds. After each sentence was exposed, subjects were
asked to read aloud strings of random numbers which were
presented on the drum at either the two or four second rate
for 128 seconds. Each time the drum.turned two lines of 18
numbers each appeared in the window and subjects were then
asked to orally recall the sentence. Subjects were given a
second complete trial on each sentence. Protocols were scored
for the number of sentences correct, the total number of
words correctly recalled, and the kinds of errors made.
Results and Discussion

The means and standard deviations for the number
of sentences and total number of words recalled are
presented in Thble 1. The analysis of variance for number
of correct sentences revealed significant effects due to
deep structure complexity, F I 8.65, df = 1/2h, p (301,
and trials, F I 67.68, df a l/2h, p«<.001. Presentation
rate did not have a significant effect nor were any of the

17

Tible l _

Means and standard deviations for the number

of sentences and the number of words recalled.

 

Mber of Sentences recalled

 

 

 

 

 

 

 

 

Condition Trial 1 __J_T:rial 2
3-2 if , 3.86 6.00
an 2.29 1.31

8-1; I 3.57 6.86
so 1.68 0.97

c-2 it 0.86 b.57
an 0.99 2.61

c-u I 2.1» 5.71
so 1.36 1.96

Total number of words Recalled

Qnditien gial 1 21.1 2
8-2 i 59 .57 6807]-
an 6.9» 2.71

3-» H 61.57 70.114
an 3.98 1.39

c-2 I 118.00 62.29
30 9.75 10.81

c-u E 56.29 68.11
SD 8.75 11.03

 

18

interactions significant. The analysis of variance for the
total number of words recalled revealed significant effects
due only to deep structure complexity, F I 5.79, df = l/2h,
p (.05, and trials, F = 107.30, a: =- 1/21. p (.001.

The hypothesis that sentences with.more complex
deep structures take longer to be analyzed by the LTM loop
of active verbal memory was supported. The results
suggest that the intervening task interfered with subjects'
ability to hold the surface structure input in active verbal
memory while the deep structure was analysed in the LTM
loop. The more complex the deep structure the longer the
analysis time required, and thus less of the complex sentence
had been analyzed when subjects' rehearsal of the input was
interfered with. The result was poorer recall of the complex
sentences.

. A main effect due to presentation rate was not
found. It had been predicted that recall would be worse at
the two second rate for both simple and complex sentences
with the effect being.more pronounced for complex sentences.
Separate analyses of variance for both types of sentences
indicated that there was no difference in the number of
simple sentences recalled at either rate nor was there a
significant difference in the number of complex sentences
recalled at either rate. This suggests that subjects were
able to process as much of the sentence as they could in
two seconds with the additional two seconds possibly provid-

ing a brief rehearsal of the processed material.

 

19

The number of words correctly recalled were
collasped over presentation rate. The mean number of words
recalled from simple sentences was 60.5 and 69.h for Trial 1
and Trial 2, respectively. The mean number of words recalled
from.complex sentences was 52.1 and 65.3 for Trial 1 and
Trial 2, respectively. On Trial 1, significantly more words
were recalled from simple sentences than from.complex
sentences, t(26) ='Z.hh. p (.025. On Trial 2, the difference
was not significant, t(26) I 1.5h, p >.05.

These results indicate that subjects were able
to determine more of the surface structure input of simple
sentences on the first presentation than of the complex
sentences. Subjects viewing simple sentences had more
information on which to base their deep structure analysis
which resulted in better recall of simple sentences on
Trial 1. On Trial 2, subjects under all conditions were
able to add to this initial information in the some amount.
However, the difference in reproduction of correct sentences
.indicates that subjects were better able to analyze the deep
structure of the siMple sentences than of the complex
sentences. Tlis suggests that the deep structure of the
more complex sentences required more time for analysis in
the LTM loop of the active verbal memory.

In an attempt to clarify the above findings, the
number and kinds of errors made were considered. The mean
number of errors per condition are presented in Table 2.

An analysis of variance revealed significant effects due to

20

Thble 2
means and standard deviations for the

number of errors made by each condition.

 

 

Qnditien Trial 1 gial 2
3-2 K 7071 3a).“
SD 3.70 3.96
3-1: i 8.00 2.11:
so 3.511 1 .6t
c-2 l 10.86 7.28
as 14.112 6.9!»
Cd. I 9.112 3.14
so 2.95 3.3“

 

Percentages of the total errors of the

21

Table 3

various classes made by each condition.“

 

Classes of errors

Deletions
Extraneeus
intrusions

Experimenmal
intrusions

Tense
Additions

Structural
changes

lumber

Trial
Trial

Trial
Trial

Trial
Trial

Trial
Trial

Trial
Trial

Trial
Trial

Trial
Trial

N» H“ n» N» NH NH up

 

Conditions
_§:2 §-8 c-z c-t
22.22 21.83 88.68 80.90
18.18 6.67 23.53 36.36
38.88 33.93 28.95 28.79
36.36 3.33 37.26 22.73
18.52 16.07 6.58 16.67
18.18 13.33 17.65 9.09
11.11 10.71 1. 2 3.03
13.68 20.00 5. 9.09
7.81 3.57 0.00 0.00
9.09 0.00 1.96 0.00
1.85 12.50 13.16 7.57
8.55 26.66 13.73 22.73
0.00 1.78 0.00 3.03
0.00 0.00 0.00 0.00

 

'30. Appendix H for raw data.

,22
deep structure complexity, F = 7.71, df = l/2h, p (2025, and
trials, F: 18.2h, df = 1/28, p.<.001. Presentation rate
was again not a significant factor nor were any of the inter-
actions significant. These results also indicate the effect
of deep structure complexity on sentence memory as more
complex sentences had more errors in recall than simple
sentences. However, they do not clarify the previous
findings in regard to the failure to find an effect due to
presentation rate.

‘The percentages of the total errors made by each
condition, which the classes of errors represent, are
presented in stle 3. The errors were categorized into seven
classes: deletions in which one or more words were missing
from the sentence; extraneous intrusions which.were the sub-
stitution of non-experimental words for words in the sentence:
intrusions of experimental words from other experimental sentences
into the sentence; tense which was a change in the tense of
the sentence or any phrase; additions which were the addition
of extra words to the sentence: structural changes which
changed the structure of the sentence in some way other than
through the addition of words; and number changes which were
changes in the number of the nouns. (More than one error
could occur per sentence with up to five errors being found
in one sentence (see Appendix 0).) Deletions and
extraneous intrusions were the most prevalent types of errors
made. This is in line with the previous findings. The higher

percentages of deletions suggests incomplete acquisition of

23

information and analysis while the extraneous intrusions
indicate attempts to nfill in" the blanks in the unfinished
constructions. Experimental intrusions were also high. This
is most likely due to confusion of sentences by the subjects.
The sentences of each type being all structurally the same
would easily permit substitution of words among sentences.
without making them structurally impossible.

Generally, the results of this experiment indicate
that deep structure complexity does affect the processing
and subsequent recall of sentences. It is contended here
that deep structure complexity influences reproduction of
linguistic materials by influencing the amount of time needed
for analysing the sentence in the LJM loop of the active
verbal memory. It should be pointed out that this is an
observed effect for verbatim recall ever a fairly short
retention interval. Whether or not deep structure complexity
would influence the recallability of linguistic materials that
are initially processed for long term storage and use, is a

question which cannot be answered from these data.

GENERALIDISCUSSION

The results of the present studies do not un-

equivocally support surface structure and deep structure

‘i

i. E ~ Hm... J;

factors in memory for linguistic materials. They do suggest
that surface structure complexity will have little influence
on subsequent reproduction of the materials if the structures
are acceptable grammatical constructions. One major difference
between the present study and studies such as Miller and
Isard's (196h), who have found a surface structure effect,

is that their complex surface structures, which resulted in
lower recall scores, were also highly unacceptable although
grammatical constructions. In the present study, the phrases
used were grammatical and acceptable. Perhaps, then, in
regard to surface structure, acceptability of the structure
would be a better predictor than complexity.

The results also indicaterthat deep structure
complexity will influence reproduction of linguistic materials.
It is argued that this is done through the amount of time
required for construction of the deep structure in a LTM loop.
Addition of this long termhmemory loop may seem.to complicate
the active verbal memory unnecessarily as direct access to
the LTM linguistic store is also postulated. However, with-
out such a loop, any short-tenn.reproduction (reproduction

within the same experimental'session) would have to be

28

25
considered echolic as no means of attaining meaning would be
provided.

The question still remains, however, as to the
psychological usefulness of concepts such as surface structure
and deep structure. Something akin to these linguistic concepts
must exist psychologically for the distinctions made by r1

Chomsky in his well worn examples of syntactic ambiguity and I

syntacticparaphrase to be evident to speakers of human ,1
languages. The fact that such distinctions can be made by
speakers is evidence to support the need for some kind of
psychological structure which performs this function. Also
arguing for a psychological counterpart of linguistic grammar
(set of rules) is the variability and novelty found in
language which can only reasonably be account for through
the use of a set of rules — a grammar. The evidence on the
development of language also suggests the necessity of a
rule system.

All these pieces of evidence point to the need
for a psychologically real grammar which in some way conforms
to the Chomskian grammar. Why then are the results of
studies of the psychology of grammar so contradictary and
unclear? Perhaps the answer lies in the fact that for the
most part the study of language has concetrated on syntax.
‘while the surface structure and deep structure components may
be necessary, it may not be possible to understand and
observe their functioning as purely syntactic entites.

Indeed, as syntactic entities they any not exist

26
psychologically but only-have existence when considered in

regard to their influence on the meaning of linguistic
utterances. Thus, it would seem.necessary to understand the
semantic system - how meanings are acquired and stored -
before the existence and influence of such structures can

adequately be assessed.

LIB '1' OF REFERENCE

LIST OF REFERENCES

Chomsky, N. ”Syntactic Structures. The Hague: Mouton &

Co., #

F

(

Chomsky, N. Aspects of the Theory of Syntax. Cambridge, Mass.: 1
MeIe e 0"; 1 0 ’1‘

Davidson, R.E. Transitional errors and deep structure
difference. Ps chonomic Scienc , 1969, 13, 293-298.

Epstein, H. Recall of word lists following learning of
sentences and of anomalous and.random.strings.‘

Journal of VeEbal Learnigg 55d Vggbal Behavigg,
s s ‘2 e

Foss, C.J}, & Cairns, H.S. Some effects of memory limitation
upon sentence comprehension and recall. Journas
D

of Verbal Learning and Vezbal Behavigz,
. ‘ 0

Martin, 3., & Roberts, K.H. Grammatical factors in sentence
retention. Journ of V b a i a d V
Behavior, 19 , 2 -2 .

Martin, 3., & Roberts, K.H. Sentence length and sentence
retention in free learning situation. Psychonomic
Science, 1967. Q, 535-536-

Mhrtin, 8., Roberts, K.H., 8 Collins, Aefl. Short term memory
for sentences. ' our al of V bal Le ni

.VOPUSI BOhtVlOE. 9 s s ' e

Mehler, J. Some effects of grammatical transformations on the
recall of English sentences. Journa 'of Va b

Leggning and Verbal Behavigg, 3, g, 3 - .

Mehler, J., a Carey, P. .Role of surface'and base structure
in the perception of sentences. Journal of Verb
Learning and Verbal Behavior. 1967, EL 335-538.

Miller, 0.1., & Chomsky, R. Fintary model of language users.
In R. Luce, R. Buss, & E. Galanter (Eds.) Handbook

27

 

I I) [Ill 1) IIIll I!) I V

28

of Mathematical Ps cholo Vol. I. New York:
John w ey & Sons, 9 . . h19-h91.

Miller, G.A., & Isard, 8. Free recall of self-embedded English
sentences. Tnformation and Control, 1968. I;

 

 

. 292-303. .
Neisser, U.. Go nitive Ps cholo . New York: Appleton-
Century-Crofts, I955.
Paivio, A. Imagery and deep structure in the recall of ,
English nominalizations Journal of Verbal Le nin 5'
and verbal Behavior, 1971, ZQ, 1-12. m
Perfetti, C.A. Sentence retention and the depth hypothesis.

Journal of Verbal Legzning and Vegbal thgviog. 1969. *
.3 ’ 0 ‘

Perfetti, C.A. Lexical density and phrase structure depth as
variables in sentence retention. 0 a of V b

Learning and Vegbal Behavigg, 1969. _, 7 ~72 . )
Roberts, K.H. Grammatical and associative constraints in
sentence retention. Jo nal of Verbal Learnin an
verbIl BOMVIOI‘. 1968. " O
Rohrman, N.L. The role of syntactic structure in the recall

of English nominalizations. Journal of Verbal
Learninggand Verbal Behavior, 1963, 1, 905-915.

Rohrman, N.L. More on.recall of nominalizations. Journal of
verbal Learnin and V rbal Behavior, 1970, 2, 535-536.

Savin, H., & Perchonock, E. Grammatical structures and
immediate recall of English sentences. Journal of

Verbal Learning and Verbal Behavigg, 1965, E, 3K8-353.

Thorndike, E.L., 6: Large, I. The Teacher's Word Book of 30,000
Words. New York: Thachers Co ege, Co umbia

University, Bureau of Publications, 1988.

Hearing, A.J. The storage of complex sentences. Jo as of
Verbal Learning and Verbal Behaviog, 197 . e ' 9-

‘Wright, P. Two studies of the depth hypothesis. British
Journal of Psycholggz, 1969, 69, 63-69.

Yagve, V.H. A model and an hypothesis for language structure.

Proceedings of the American Philosophical Society,
0 " 0

APPENDICES

APPENDIX A

Transformational history of noun phrases used in nperinent 1.

Example: yellow growling lions: yellow lions growling

3' -n .
”’1/ 8W ’
‘ VP °
'92 -s I
I e+Pres
up 32 V3
lion (w i
151 . "’3 in I" 853“
' +Pres VB +m8
11.1: I' 1
4"
11.“ e 1"
+21 +1, y -V
+Pl

Deep 8813108”.

29

 

NP
Relative
Transformation I
applied t. 82
lion
+16
+21

1")
.1.

+Pro
+21

His!
was

s |

I \\vr 81‘0"].
Aux +V
~11

+Pres

 

l -3
I I +Pres
lion T l8 l
4-] e
m ‘53“
Relative ' 1“'3 A“! +Pros
Transofrmation I ' -ll
applied to 31 s +Pr°l vs
which I I
+Pro
+91 which: "110'
+Dof +h'o "V
+Don +P1
+Def
«then
M? ‘- VP
2 Aux
x . | -s l
l n +P1‘0. VB
T3 N grewl
+ .
yellow 11.“ I +1598
-v +' Ithh
“’1 +Pro
Adjective Shift +P1
Transformation +Def
Opplied to 82 +Den

-5...u*‘.'-. “€11; '

jau‘ a...—

A‘ f .na—ni.
-m ‘1-71.-. 1' '~ '

 

“1!. 0 a

" 7-

 

 

.\.
I: .L...’ ii in!»

31

fl}. rm
4 l

he

VB! ‘ VB

 

 

' yellow lion growl
ye low growl lion . -V +N +1!
-V +7. . +1! +Pl +Prog
+Prog +Pl
' Relative Clause Reduction
“3”“7' Shift Transformation applied
Transformation t. 31

applied to 81

8
g—
3.—

l l a J.
llow growling lions l

yellow lions growling
ample Phrase Complex Phrase

 

 

Noun Affix and Progressive Affix i'ranssornations

applied to both structures to generate the surface
structures.

APPENDIX B:

Noun phrases and strings of words used in Experiment 1.

Simple :
Complex 8

Simple :
Complex 3

Simple :
Complex 3

Simple :
Complex 3

Simple :
Complex 8

Simple :
Complex :

Simple :
Complex 3

Simple 8
Complex 3

Simple :
Complex 8

Simple 8
Complex 3

Simple :
Complex 8

Simple :
Complex :

Noun Phrases

bright blinding lights
bright lights blinding

strict demanding parents
strict parents demanding

fierce barking dogs
fierce dogs barking

slim dancing girls
slime girls dancing

fat preaching monks
fat monks preaching

yellow growling lions
yellow lions growling

black running cats
black cats running

famous performing artists
famous artists performing

happy playing children
happy children playing

poisonous hissing snakes
poisonous snakes hissing

strong invading armies
strong armies invading

green wriggling tadpoles
green tadpoles wriggling

32

Word Strings

decline Joy garden wage
nail oak object pack

feed vote offico‘gas
need add lay kniof

hard safe up only
nest Jump future put

pause hero band fate
elect hat adopt castle

open sold what yard
rush duty kill earth

feature ready virture wave
tea said uncle able

nose holy each palace
vast dead bank teeth

guide utter visit tape
large bar keep sat

test get warmer hamle
danger 111 wide knock

half. lack equal child
empire also saw tender

wait join yield afford
dash care final bee

yet wash you editor
health park knee name

APPENDIX C

Grammar pro-test for Experiment I

SNARLIN TIGERS KILL

HOUNDS ARE SAYING
INVESTIGATIONS ARE REVEAHIG
STROM RECURRIDB ILLUSIONS
IDUD ALARHS RIMIm

PLIIm BIRDS SING

ZEALOUS EXPLORERS COLONIZIM‘
HASSES WERE CHMRING
POWERFUL REIGNING RIMS
SPOTIESS KITCHENS GLEAHIIB
DECREPIT CREAKIH} CHAIRS
WOODEN HOUSES BURN

SHALL mIS WHISTLIM

RULES ARE PRUSTRATIIE
PASSOIATE commas" IDVE
DARING HEN FL!

33

SENTENCE
SENTENCE
SENTENCE
SENTENCE
SENTENCE
SENTENCE
SENTENCE
SENTENCE
SENTENCE
SENTENCE
SENTENCE
SENTENCE
SENTENCE
SENTENCE
SENTENCE
SENTENCE

PHRASE
PHRASE
PHRASE
PHRASE
PHRASE
PHRASE
PHRASE
PHRASE
PHRASE
PHRASE
PHRASE
PHRASE
PHRASE

PHRASE
PHRASE

.APPENDIX D

Raw data Experiment I

Grammar Pro-test

NMmber of correct identifications

WOVOU¢UNfl

\OGVOSUCNNP M

35

Immediate Recall Test

Number of Phrases Correct

Simple Complex

2 6

3
6 6
2 z
2 5
2 5
2 6
5 8
5 8
3 3
'2 2
3 u
6 2

Anunnumuumn

number of Phrses Recalled
fiums

HONI‘UOOOOONfli‘OwI‘

MEMrmfmmhlwunei

after Correct Phrases

Simple

16
22
22
10
6
a
9
18
18

12

7
18

25
O

is

Imumu

pnwmcnuunmnunnnp

Complex

18
8
1
.3
11
8
22

6
u

13
12

5
ll
20

5

5

‘APPENDIX 8‘

Surface and deep structures of Experiment II sentences.

Surface Structures

.___, -114

 

 

f und I treasure
'\ “Tr"
VB) 'P rich buried
The
IL

 

explorers I Complex Sentence
maNping t e island

 

36

37}

‘—

Deep Structures

 

+ +Art +V' bread 1 ed
Def ' b and SO

 

 

 

explorer
*5.
+ ox lorer map
+‘19 Eu’ +9
+00! +111 «tn-es
'DUI 4+Art
+Def
-Dem

Complex Sentence 1 +8g

The
The
The
The
The
The

The

The
The
The
The
The
The

(APPENDIX P

Sentences used in Experiment II

Simple

monks chanting the prayers believe they are pious.
dooters performing the operation report it was successful.
pianist playing the sonata thinks he is talented.
detective studying the trial reports it is fresh.

women baking the bread say it is good.

author writing the book says it will sell.

artist painting the portrait feels it is excellent.

police investigating the death believed it was sucide.

Complex

lawyer defending the suspect uses long wordy arguments.
friar preaching the sermon says long pious prayers.

guide carrying the camera photographs rare black swans.
people contributing the painting are famous wealthy writers.
men examining the plans build large cargo ships.

artist studying the scene paints gay colorful pictures.

lady reading the book writes charming short poems.

explorer mapping the island found rich buried treasure.

APPENDIX GT

lean number of errors per sentence in Experiment 11.

 

Conditions

8-2

3.1.

0-2

c-b

 

Number of errers/sentence
L 2 :1 "
T1131 1 1.8“ 1e28 0072 0.17 0000
M31 2 1028 0057 0000 0000 0.1“
mal 1 1.43 1eh3 0e86 0017 0000
1.11.31 2 Gel? 0072 Del“ 0.00 0.00
T2131 1 “.00 2.71 0.17 0.1“ 0.00
Trlal 2 1.28 1.15 0.86 0.17 0.00
”1‘1 1 3.00 2.39 00“ 001“ 0000
Trial 2 1.00 O. 6 0.1“ 0.00 0.00

 

39

Raw Data Experiment II

APPENDIX E

 

umber of gentences correct

 

 

 

 

 

 

 

 

0.111121»: 63-2 34 c-2 c4
Trial 1 2 1 2 1 2 1 2-
T
1' 2 3 3 7 o 7 2 6
2 7 7 3 7 o o 1 n
B 2 6 1 5 1 l1 l1 8
3 6 5 7 1 2 u 7
5 7 7 6 8 3 8 o 2
6 1 6 2 6 1 5 2 7
7 5 7 5 8 o 6 2 6
Number? of Words Correct
001111121» 7 3-2 s-b c-2 c-b
Trial 1 2 1 2 1 2 1 2
f’ 53 611 60 7o 58 71 6b 69
2 69 71 60 7o 29 no a? 65
a 58 7o 61 69 3‘23 611 62 72
52 66 63 69 53 67 72
5 69 71 70 72 9 72 £11 60
6 53 68 56 69 o 68 59 71
7 63 71 61 72 so 68 5t. 70
Number of Errors
0.111121» 8-2 s-b c-2 c-n
mu 1 2 1 2 1 2 1 2
1‘ 11 7 12 2 11 1 8 l1
2 u 1 9 2 12 20 12 5
a 8 2 11 5 1o 7 7 o
8 6 5 3 18 15 a i
5 a 1 2 o 8 o 1 1o
6 1 1. 12 3 10 h 9 1
7 6 1 5 , o 11 u 11 2

 

#1

Various Types of Errors Made

 

 

 

 

 

 

Type of error Conditions
8- 8-4 c-2 c-b
Deletions Trial 1 12 12 37 2?
Trial 2 b 1 12 8
Extraneous Trial 1 21 19 22 19
intrusions Trial 2 8 5 19 5
Experimental Trial 1 10 9 5 11
intrusions Trial 2 . h 2 9 2
Tense Trial 1 6 6 1 2
Trial 2 3 3 3 2
Additions Trial 1 b 2 O 0
Trial 2 2 0 1 0
Structural Trial 1 1 7 10 5
changes 11-1-11 2 1 b 7 5
NMmber Trial 1 O 1 O 2
Trial 2 O O 0 0

Total‘Errors

Conditions 8-2 s-h cbz c-h
Trial 1 5h 56 76 66
Trial 2 22 15 51 22

 

HICHIGRN

1111

3 2

STRTE UNIV. LIBRRRIES
|||11|\lllllMlllll1"11111111111HI
93101665143