THE DEVELOPMENT OF THE PERCEPTION
A DISSERTATION PRESENTED TO THE GRADUATE COUNCIL
OF THE UNIVERSITY OF FLORIDA IN
PARTIAL FULFILLMENT OF THE REQUIREMENTS
FOR THE DEGREE OF DOCTOR OF PHILOSOPHY
UNIVERSITY OF FLORIDA
For Phil--the closer the better
I would like to express my appreciation to Dr. Robert
Scholes for his many contributions to this dissertation,
including his suggestions for the test designs, for the
analyses, and for the pictures, as well as his patience
in directing its progress.
I would also like to express my gratitude to Dr. Jean
Casagrande both for his comments and suggestions concerning
this dissertation and for his support and understanding
throughout my enrollment at the University of Florida.
To Dr. Paul Kotey I would like to convey my appreciation
for his reading and commenting on my dissertation as well
as for his support and advice.
I would like to thank Dr. Ira Fischler for his thorough
reading of this dissertation and for his invaluable sugges-
tions for its improvement.
To Dr. Tom Doherty I wish to express my sincere grati-
tude for his many hours of help in the analysis and inter-
pretation of the data for Experiment VI.
I would finally like to thank Steve Sledjeski of P.K.
Yonge Laboratory School for making it possible for me to
test students at that school and for providing me with
ideal testing conditions.
TABLE OF CONTENTS
ACKNOWLEDGMENTS . .
LIST OF TABLES. . .
LIST OF FIGURES . .
ABSTRACT. . .
CHAPTER I INTRODUCTION .
Notes. . .
CHAPTER II EXPERIMENTS I THROUGH V .
Experiment I . .
Experiment II. . .
Experiment III . .
Experiment IV. . .
Experiment V . .
Discussion of Experiments I Through V.
CHAPTER III EXPERIMENT VI. .
Method . .
Procedure. . .
Results. . .
CHAPTER IV DISCUSSION. . .
CHAPTER V CONCLUSION . .
APPENDIX A SENTENCES FOR EXPERIMENT I. .
APPENDIX B SENTENCES FOR EXPERIMENT IV .
APPENDIX C PICTURES FOR EXPERIMENTS I THR
APPENDIX D SENTENCES FOR EXPERIMENT VI .
REFERENCES. . .
BIOGRAPHICAL SKETCH . .
OUGH V. 96
. . iv
. . vi
. . vii
. . viii
. . 1
LIST OF TABLES
1 Mean Number of Choices of Two Pictures for
the Ambiguous Sentences. . .
2 Mean Number of Choices of Two Pictures for
the Unambiguous Sentences. . .
3 Percent of Incorrect Responses on the Four
Verb Forms by Male and Female Subjects .
4 Percent of Incorrect Responses on Subject
and Object Nominals by Male and Female Subjects.
5 Percent of Choices of Are and Look in the
Ambiguous Sentences. . .
6 Percent of Choices of Plural Verb Forms in
the Ambiguous Sentences by Male and Female
Subjects . . .
7 Percent of Incorrect Responses for the Main
Verbs Be and Look. . .
8 Percent of Incorrect Responses on Subject
and Object Nominals. . .
9 Percent of Choices of Plural Verb Forms in
the Ambiguous Sentences. . .
LIST OF FIGURES
Figure 1 Percent of incorrect responses for the
verb form Is by male and female subjects. 54
Figure 2 Percent of incorrect responses for the
verb form Are by male and female subjects 55
Figure 3 Percent of incorrect responses for the
verb form Looks by male and female subjects 56
Figure 4 Percent of incorrect responses for the
verb form Look by male and female subjects. 57
Figure 5 Percent of incorrect responses on subject
and object nominals by male and female
subjects. . . 61
Figure 6 Percent of choices of Are and Look in
the ambiguous sentences . 64
Figure 7 Percent of choices of plural verb forms
in the ambiguous sentences by male and
female subjects . . 67
Figure 8 Percent of incorrect responses for the
main verbs Be and Look. . ... 74
Figure 9 Percent of incorrect responses on subject
and object nominals . 78
Figure 10 Percent of choices of plural verb forms
in the ambiguous sentences. . ... 81
Abstract of Dissertation Presented to the Graduate Council
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy
THE DEVELOPMENT OF THE PERCEPTION
Chairman: Dr. Robert J. Scholes
Major Department: Linguistics
Sixty-nine white, native-English speakers between the
ages of eight and twelve years old were tested through picture-
verification in five related designs to determine at what age
they could perceive deep-structure ambiguity. The subjects
heard ten ambiguous sentences and twenty unambiguous sentences.
The unambiguous sentences consisted of two sentences for each
ambiguous sentence, each conveying one of the meanings of the
ambiguous sentence. Subjects chose pictures to indicate which
meanings they perceived. It was expected that the subjects
would choose one picture for each unambiguous sentence and,
when they perceived the ambiguity, choose two pictures for the
ambiguous sentences. However, the subjects consistently chose
more than one picture for both the ambiguous and unambiguous
sentences, indicating that to them all of the sentences were
Because most of the test sentences were composed of a
derived nominal, some form of the verb be, and an adjective,
it was felt that the problems were related either to the
verb be, itself, or to the plural marker which disambiguated
the nominal. A second test was designed in which twenty-four
sentences, twelve ambiguous and twelve unambiguous, were pre-
sented to 130 subjects between the ages of 7 and 12, and to
53 subjects between 18 and 29. These sentences consisted of
an ambiguous nominal, a subject nominal, or an object nominal
followed by either is and are or looks and look followed by an
adjective. The subjects were asked to circle the correct form
of the verb. For the ambiguous sentences, either form was
correct. For the subject nominal the plural form was correct,
and for the object nominal the singular form was correct. It
was concluded that subject/verb agreement in the context of
derived nominals is not fully acquired by the age of 12 years.
This finding was presented as an explanation for the inability
of the subjects in the first experiment to disambiguate the
nominal and thus perceive that twenty of the original test
sentences were unambiguous.
Attempts to understand how a child learns to speak his
native language have produced numerous anecdotal descriptions
of the child's linguistic behavior. However, systematic
descriptions of the child's acquisition processes are still
quite limited. We know the sequences of certain elements as
they appear during the first five years of the child's life.
But as the child grows older, he gains more subtle aspects of
his language. These more subtle aspects include various com-
plexities of sentence structure which the child learns to
comprehend and produce. The relationship between two sentences
which are similar in meaning and in structure can be described
with the use of transformations within the context of Trans-
formational Grammar. The Passive Transformation, for example,
states the structural relationship between sentences l.a. and
l.a. Sam Peckinpah directed that movie.
l.b. That movie was directed by Sam Peckinpah.
Although these sentences share basically the same meaning,
their structures are quite different. As part of the process
of acquiring language, a child must learn those things about
sentence structure which transformations explain. Through
various testing methods, we are able to determine that by a
certain age most normal children have a specific transformation
in their grammars. For example, by the age of nine years
most children have acquired Dative Movement, a rule which
allows the child to comprehend that sentences l.c. and l.d.
have the same meaning (Scholes, Tanis, and Turner 1976).
l.c. Tom showed the dog to the little boy.
l.d. Tom showed the little boy the dog.
The number of grammatical structures which a child must
acquire is quite large, but the number for which we know the
age of acquisition is quite small. It is important that we
be able to say when a certain structure is acquired.
One reason for this importance is that this information
provides a guideline for comparing languages. We know that
native speakers of English acquire transformations in a fixed
sequence. Presumably there is a fixed sequence for each
language. Although different languages employ a variety of
different rules, there is an overlap in their use. For
example, many languages employ a plural marker to indicate
that there is a difference between nouns in the singular and
nouns in the plural. The native English-speaking child
acquires the plural morpheme by the age of four years. The
speakers of Egyptian Arabic do not master the rules of plural
formation before the age of fifteen years (Omar 1970, quoted
in Slobin 1973). Such a comparison offers us insight into
the similarities and differences which exist among languages.
Obviously, in this case there is a vast difference in the
complexity of the rules for plural formation in English and
those for Egyptian Arabic. If enough information about the
acquisitional sequences for different languages were available,
it would be possible to determine whether there is a trend
across languages toward a universal sequence of acquisition.
Secondly, because language development is highly corre-
lated with intellectual development, a knowledge of the order
and age of acquisition of transformations would provide a
basis for determining a child's developmental progress. This
is true among normal children, but it would also give us a
reference point for studying people who are deaf or retarded
or brain-damaged or who have undergone a hemispherectomy or
are abnormal in some other respect which is related to lin-
For those people who believe that the dissolution of
language (as with aphasia) follows somewhat the reverse order
of language acquisition, the knowledge of this sequence would
improve the chances for proving or disproving this belief.
A fourth reason for knowing the age of acquisition of
certain transformations is that for people who design tests
such as reading and IQ tests, it would be far more consistent
to test children's abilities in those areas of which they are
able to comprehend the content. In other words, a child may
perform poorly on a test because he does not understand the
syntax in the instructions or in some material which he has
to read. Consequently, it may appear that the child cannot
perform a particular task when, in fact, he may be quite
capable, but he simply does not comprehend other information
that he needs in order to complete the task. The problem is
not limited to children. A student may be asked to read an
article and then to draw a particular conclusion. He may
reach the wrong conclusion, not because of faulty logic or
because of an inability to read, but rather because the syntax
of the article is too complex for him.
A specific example has occurred in the Writing Labora-
tory at the University of Florida.2 Students in the Writing
Laboratory are tested in their ability to recognize sentence
errors. On one test the students were instructed to find
the incorrect sentence from a group of three sentences. The
following sentences are two of a set of three sentences from
l.e. A wide range of choices is offered students by
educators experimenting with new programs.
l.f. Experimenting with new programs, a wide range of
choices is offered college students.
Sentence l.f. is considered grammatically incorrect because
the introductory clause is a so-called 'dangling modifier.'
However, sentence l.e., although technically correct, is
syntactically difficult to the students in the Writing Labora-
tory. The first clause is passive, the to which is the overt
marker of the dative has been deleted, and the indication
that experimenting is derived from a relative clause is also
missing. This syntactic complexity led many of the students
to choose sentence l.e. as the incorrect sentence.
On another test in the Writing Laboratory the students
were instructed to identify the errors in the sentences on
the test. They were told that there would be either sentence
fragments, run-on sentences, sentences which lack parallel
structure, or sentences with no errors. One of the items on
the test is listed here as l.g.
l.g. Although the life-support system was defective,
the problem was discovered too late. Fearing money for the
space program will end if there is a scandal.
The clause 'Fearing money for the space program will end if
there is a scandal' is a sentence fragment. Many of the stu-
dents in the Writing Laboratory had considerable trouble with
this item on the test because it made absolutely no sense to
them. One student claimed that the clause was a complete
sentence with fearing money as the subject of the verb will
end rather than money as the subject of will end. From a
purely syntactic point of view, this interpretation is
possible. However, no sentence content corresponds to such
a syntactic structure, a fact which did not bother the student
because she was unable to make sense out of it in any other
way. The meaning of this clause would have been much clearer
had there been a that complementizer between fearing and money.
Adding that would not have altered the fact that this clause
is a sentence fragment, but it would have helped the students
in their understanding of this item. The point is that the
students' performance on this item was meant to indicate that
they could not recognize sentence fragments, when, in fact,
the problem was one of unnecessarily complex syntax. Had
syntactic complexity been taken into consideration, the test
would have been more valid. This sort of problem should be
avoided by people who design tests.
If it is in fact the case that different dialects of the
same language have different rules or different ordering of
rules or that their speakers acquire some transformations in
different orders, then this information would be important
for the following reasons:
1. As a basis for descriptive comparison.
2. If transformation X is acquired at the age of 7 in
one dialect, but not until the age of 9 in another
dialect, then this should be recognized as a dialect
difference and not, for example, as an indication of
3. Aphasics of one dialect may possibly produce different
behavior from aphasics of another dialect; that is,
dissolution might follow a different sequence de-
pending on the aphasic's dialect.
4. Tests should be designed so as to avoid a bias in
favor of or against a particular dialect.
These four reasons are the same as the first four major rea-
sons but are applied here specifically to dialects.
Finding the age of acquisition of particular abilities
is simply one more aspect of the challenge that Noam Chomsky
set forth when he said, 'As a long-range task for general
linguistics, we might set the problem of developing an account
of this innate linguistic theory that provides a basis for
language learning' (1965:25). And an understanding or knowl-
edge of 'this innate linguistic theory' which a child uses
in order to acquire his native language would provide a fur-
ther understanding of the relationship of language to other
mental faculties. An understanding of the nature of the
child's internal grammar would provide a view into other
aspects of the mind. Since, as Chomsky put it, 'An actual
language may result only from the interaction of several
mental faculties, one being the faculty of language' (1975:
43), an increase in the understanding of one offers better
understanding of others.
It is for these reasons that we continue to search for a
systematic description of the processes which are involved in
With these ideas in mind, I set out to determine at what
age the normal, white, native English-speaking child acquires
the ability to perceive that a sentence is ambiguous at the
level of the deep structure. Ambiguity, the presence of more
than one meaning, is generally divided into three main cate-
1. Lexical ambiguity--when a single word or lexical item
has more than one meaning. In the sentence 'The young boy was
sitting by the bank' the word bank may mean either a building
where people keep money or it may mean the edge of a river.
2. Surface-structure ambiguity--when different meanings
can result depending on the way in which the words in the sen-
tence are grouped. Surface-structure ambiguity, also known as
surface-bracketing ambiguity, is defined by Fodor, Bever, and
Garrett (1974:95) as ambiguity for which there are 'alternative
constituent structure analyses.' For example, the meaning of
the sentence 'They are visiting relatives' depends on whether
the words are visiting are grouped as a single constituent or
whether the words visiting relatives are grouped as a single
constituent. The meanings of this sentence 'can be represented
by reference to alternative bracketings' (Fodor, Bever, Garrett
1974:95) of the elements of the sentence:
l.h. They (are visiting) relatives.
l.i. They are (visiting relatives).
3. Deep-structure ambiguity--when two different deep
structures undergo transformations which give them identical
surface structures. It is not possible to bracket constituents
of sentences which are ambiguous at the deep-structure level
in order to represent the differences in structure or meaning.
Consider the sentence 'The police stopped drinking on campus.'
In this sentence, it is the subject of drinking which is am-
biguous. In the deep structure there are different subjects
for drinking, but in the surface structure those subjects have
been deleted by transformations. For the meaning which is
equivalent to 'The police quit drinking on campus,' the sub-
ject of drinking was deleted by Equivalent Noun Phrase
Deletion Transformation (also known as Equi NP Deletion or,
simply, as Equi). This transformation applies when the subject
of the main clause is also the subject of a gerund or an in-
finitive. It functions to delete the subject of that gerund
or infinitive. For the meaning which is equivalent to 'The
police ended drinking on campus,' the subject was deleted by
Pro Deletion Transformation. This transformation applies when
the subject of a gerund or an infinitive is an unspecified NP
in the deep structure. Its function is to delete the subject
of that infinitive or gerund.
A behavioral difference between surface-structure am-
biguities and deep-structure ambiguities is that for the for-
mer there is a difference in intonation patterns depending on
the meaning, whereas, for the latter, the intonation pattern
remains constant regardless of meaning. The sentence 'They
are flying planes' is a surface-structure ambiguity because
it can be explained in terms of bracketing:
l.j. They (are flying) planes.
l.k. They are (flying planes).
Such a sentence can be disambiguated with an intonation pattern
comparable to the bracketing:
1.1. They are flying planes.
l.m. They are flying planes.
However, a sentence such as 'They want to stop smoking' cannot
be disambiguated through an intonation pattern. In order for
a person to perceive both meanings of a deep-structure ambiguity,
he must be able to comprehend the structures which correspond
to each meaning. In the case of 'They want to stop smoking,'
the structure which corresponds to 'They want to quit smoking'
in which they is the subject of smoking is derived through
Equi NP Deletion. The ability to comprehend this particular
meaning depends on the ability to understand a structure which
is produced through Equi NP Deletion Transformation. Similarly,
the structure which corresponds to 'They want to prevent
smoking,' in which the subject of smoking is an unidentified
or unspecified element comparable to someone or others, has
been derived through Pro Deletion Transformation. The com-
prehension of this meaning is contingent on the ability to
understand the structure of a sentence which is derived through
Pro Deletion Transformation. Consequently, the ability to
perceive the ambiguity of this sentence is based on the ability
to comprehend the structures for both meanings. Thus a child
could not perceive both meanings of a deep-structure ambiguity
unless he had acquired both transformations which are necessary
for the two meanings.
The age at which a child becomes capable of perceiving
deep-structure ambiguity should be predictable on the basis
of the age of acquisition of the transformations which are
required to generate the structures that correspond to both
meanings. For example, if a child acquires the ability to
comprehend Equi NP Deletion Transformation by the age of 4
years, but acquires Pro Deletion Transformation at the age of
9 years, then beginning only at the age of 9 years should he
be able to perceive the ambiguity of a sentence such as 'They
want to stop smoking.' This prediction is based on the
assumption that perception of deep-structure ambiguity is
determined by the ability to comprehend relevant transforma-
tions and by cognitive capacities that are developed prior to
or at the same time as these transformations. If any such
cognitive capacities are necessary for the perception of am-
biguity but have not been developed by the time that these
relevant transformations are acquired, then the transformations,
themselves, would not be sufficient for the perception of the
The assumption that the perception of deep-structure am-
biguity is determined by the ability to comprehend the relevant
transformations was the subject of this investigation. The
only other study which has investigated ambiguity from a
developmental standpoint was done by Frank Kessel (1970).
Kessel tested fifty children from kindergarten through fifth
grade (with the exception of the fourth grade) in their ability
to perceive lexical ambiguity, surface-structure ambiguity,
and deep-structure ambiguity.
In his study, subjects were asked to choose the picture
or pictures which represented the sentences which the experi-
menter read. The subjects heard twelve sentences, four for
each type of ambiguity. The four sentences with deep-structure
ambiguity which Kessel presented to his subjects were:
l.n. The eating of the chicken was sloppy.
l.o. She hit the man with glasses.
l.p. The visiting of the doctor was happening.
l.q. The shooting of the soldier was bad. (p. 26)
Kessel found that 'only after the age of 10 do children con-
sistently and spontaneously (without "lead-in" questions)
detect both meanings in surface and underlying structure am-
biguities' (p. 43). He made no attempt to determine or explain
what factors contribute to the acquisition of the ability to
1Whether a child actually acquires transformations or
whether transformations are, more appropriately, devices to
describe linguistic facts is still up for debate. For the
purpose of this paper, when I discuss the processes involved
in 'acquiring a transformation,' I am aware of the problems
which this phrase implies, but I will use this terminology for
the sake of simplicity.
2The Writing Laboratory is a one-credit course offered by
the Reading and Writing Center at the University of Florida.
It is designed to help students with such basic writing skills
as paragraph development and grammar.
3These examples are used by permission of The Writing
Laboratory. The tests from which these sentences were taken
are not professionally designed, standardized tests. They
were, instead, written within the Writing Laboratory of the
University of Florida and, like any tests which are designed
by a classroom teacher, are likely to contain errors in validity
EXPERIMENTS THROUGH V
Unlike Kessel's study, this investigation attempted to
predict perception of deep-structure ambiguity on the basis
of linguistic theory of transformations. This study attempted
to show that a child's ability to detect deep-structure am-
biguity can be predicted by his ability to handle certain
transformations. Only after a child has acquired the trans-
formations which are necessary to generate both meanings of
an ambiguous sentence, should he be able to comprehend both
meanings of that sentence.
This hypothesis was tested by presenting students with
sentences which were ambiguous at the deep-structure level.
For every ambiguous sentence there were two unambiguous sen-
tences, one for each of the meanings of the ambiguous sentences.
Not only did the unambiguous sentences convey the same meanings
as the ambiguous sentence, but they also had the same syntactic
structures which corresponded to the meanings of the ambiguous
sentence. The purpose of this paradigm was to predict behavior
on the ambiguous sentences by performance on the unambiguous
sentences which corresponded to it. Consider the following
2.a. Flying planes can be dangerous.
2.b. Flying planes is dangerous.
2.c. Flying planes are dangerous.
Sentence 2.a., which is ambiguous, contains the meanings ex-
pressed in sentences 2.b. and 2.c. Sentence 2.b. contains a
nominal which can be described as an object nominal and which
can be paraphrased as 'the act of flying planes.' The entire
nominal flying planes is the subject of this sentence. Sen-
tence 2.c., on the other hand, contains a subject nominal which
can be paraphrased as 'planes which are flying.' Since flying
functions as an adjective in sentence 2.c., only the noun
planes is the subject of the sentence.
The task which the subjects were asked to perform was to
indicate which picture or pictures corresponded to the sentence
which the experimenter read. The prediction was that the stu-
dents would choose the correct picture for each of the unam-
biguous sentences, thus indicating their comprehension of
these sentences and their structures, if they were to choose
the correct pictures for the ambiguous sentence. In other
words, the student should perform correctly on sentences 2.b.
and 2.c. if he is to get 2.a. correct.
Subjects. Forty-seven students were selected from P.K.
Yonge Laboratory School in Gainesville, Florida. There were
four groups of males and four groups of females, ranging from
8 through 11 years of age. Because of a suspected difference
in dialect, there were no black students included in the study.
All of the students were native speakers of English.
Apparatus. The test design included five pretest sentences,
used to establish that the subjects could handle the task of
pointing to one or more than one picture for each sentence that
was read, and thirty test sentences, composed of ten ambiguous
sentences and twenty unambiguous sentences. A list of these
thirty-five sentences appears in Appendix A. The twenty unam-
biguous sentences consisted of two interpretations for each
ambiguous sentence. The criteria for the ambiguous sentences
were that they be deep-structure ambiguities, that the intona-
tion pattern be the same regardless of meaning, and that they
be representable with drawings.
Of the ten ambiguous sentences, seven had as their struc-
ture nominal phrase + can be + adjective, for example, 'Visit-
ing relatives can be boring.' The unambiguous interpretations
of these seven sentences were made up of the same nominal
phrase, either the verb is or the verb are, and an adjective.
Thus the two interpretations for 'Visiting relatives can be
boring' were 'Visiting relatives is boring' and 'Visiting
relatives are boring.' These last two sentences were not
ambiguous because the verb determined whether the nominal was
to be interpreted as a subject nominal or an object nominal.
The original sentence 'Visiting relatives can be boring' is
ambiguous because there is no overt marking for number on the
verb can be.
It was also important that the adjectives in these sen-
tences be appropriate to either interpretation of the nominal.
Hence, both the act of visiting relatives and relatives who
visit may be boring. However, the sentence 'Visiting relatives
can be generous' does not offer the same duality of interpre-
tation. It is grammatical to say 'Visiting relatives are
generous,' but the counterpart 'Visiting relatives is generous'
is ungrammatical. Similarly, 'Moving houses is easy' is ac-
ceptable, whereas 'Moving houses are easy' is not. And, con-
sequently, 'Moving houses can be easy' is not ambiguous.
Two of the three remaining ambiguous sentences were com-
posed of, among other constituents, the verb stop followed by
a gerund. The ambiguity arises because there are two inter-
pretations of the subject of the gerund. These different
interpretations occur because in the deep structure there are
different subjects for the verbs which eventually become
gerunds. Consider the sentence 'The police stopped fighting
on campus.' This sentence may mean either that the police
quit fighting on campus, in which the police is the subject
of fighting, i.e., the police were doing the fighting, or
that the police ended fighting on campus, in which the subject
of fighting is some unspecified but implied agent other than
The final sentence type on the test is also one in which
a nominal construction may be interpreted in more than one way.
The test sentence is 'The shooting of the Indian was bad.'
This nominal + of + NP construction is often ambiguous because
the NP may be both the agent of the action expressed in the
nominal or it may be the object of that action. In the case
of this sentence, the Indian may be either the shooter or the
For each sentence there was a set of four pictures for
the subjects to choose from. An example of each set of pic-
tures appears in Appendix C. For the pretest sentences, in
two cases only one of the four choices was correct, and in
three cases there were two pictures which were not identical
but which were appropriate choices for the pretest sentence.
For example, for the pretest sentence 'They are riding horses,'
there were two pictures of people riding horses, but the pic-
tures differed in their details. There were also two pictures
which were not appropriate to the sentence. The purpose of
having two different pictures which corresponded to the same
sentence was to determine that the subject could perform the
task of pointing to two pictures if they both matched the sen-
For the test sentences there were also four pictures for
each sentence. For each ambiguous sentence and its two unam-
biguous counterparts, the same four pictures were used. How-
ever, the location of the pictures within each set was ran-
domized to avoid a positional bias. The four pictures con-
sisted of one picture for each meaning of the ambiguous sen-
tence and two pictures of related but different and inappro-
All of the pictures were made into two-inch slides and
were projected onto a solid-colored wall with a Kodak Carousel
750 Slide Projector. The pictures were labeled A, B, C, or D.
The score sheet for each subject provided space for re-
cording the name, age, sex, and teacher, as well as the subject's
responses to both the pretest and test sentences. Sufficient
space was allotted for any notable comments which the subjects
may have made or for any remarks which the experimenter may
have felt necessary.
Subjects were tested individually in a quiet room which
was dark enough to allow adequate viewing of the slides. The
subject was seated in a chair approximately six feet from the
wall on which the slides were projected. The projector was
behind and to the left of the subject. The experimenter sat
to the right of the subject and controlled the projector with
a push-button, remote-control device.
The experimenter explained to each subject that she wanted
him to tell her which picture or pictures matched the sentences
which she was going to read to him. The experimenter repeated
or rephrased the instructions when necessary. On some occasions
if the subject said he did not understand, the experimenter
said, 'Let's try one and see if you understand then.' As soon
as the first slide appeared and the subject heard the first
sentence, the experimenter asked, 'Now, which picture or
pictures go with that sentence?' Almost always, if there had
been any question in the subject's mind, he would answer, 'Oh!
Now I get it,' or something comparable.
The subject indicated which picture or pictures he felt
were appropriate by naming the letter A, B, C, or D which
labelled the pictures. At no time did the task of choosing
letters to indicate which picture or pictures the subject
felt appropriate seem too difficult.
The experimenter recorded the subjects' responses on the
score sheets, taking care to record the sequence if more than
one response were given. The experimenter also recorded any
of the subjects' comments which the experimenter felt were
valuable to this study. Some of these comments offered in-
sights into the subjects' thoughts as they participated in
this experiment. For example, in the sentence 'He wants to
prevent smoking,' several subjects asked, 'What does prevent
mean? Start or Stop?' For the sentence 'Moving houses can
be frightening,' there were comments such as 'What kind of
moving?' and 'When you're moving them or when they are moving
by themselves?' Another subject, when he heard the sentence
'Shooting stars is exciting,' commented, 'That just doesn't
seem right. Is exciting? Oh! D.' One subject who had no
problem comprehending the ambiguities found them very funny.
'A dog walking This is funny, get it?' 'That is funny.
Grownups are playing.' These and other comments provided a
valuable view of what the subjects were doing and thinking and
how they performed the tasks of this experiment.
The conditions of this experiment were informal. If a
subject did not clearly understand a sentence which was read
to him, the experimenter repeated the sentence. The task of
this experiment was based on the subject's ability to compre-
hend the structure of a sentence. It was not a test of his
ability to perceive the sentence itself. Because intonation
does not affect the reading of the sentence, repetition did
not alter any of the conditions of the experiment. Also, if
the subject had any questions and if the answers to those
questions would not alter the conditions of the experiment,
the experimenter answered the subject's questions. For ex-
ample, when one subject asked, 'What does prevent mean? Start
or Stop?' the experimenter answered, 'Stop.' However, when a
subject asked, 'What kind of moving?' in reference to the
sentence 'Moving houses can be frightening,' the experimenter
answered, 'Whatever it means to you.' An answer to the last
question on the part of the experimenter would have invalidated
the test question. There would have been no way to explain
what kind of moving without providing too much information to
After each subject had responded to the test sentences,
the experimenter asked him to look at some of the pictures
which he had just seen and to indicate what the pictures
represented to him. This was done because there was some
question as to the interpretation of some of the pictures.
To make sure that the pictures represented to the subjects
what they represented to the experimenter, the experimenter
showed some of the slides again and asked, 'What is going on
in picture A [or any other picture]?' The experimenter did
this for usually five or six pictures per subject. Ultimately,
there seemed to be no problem with the subjects' interpretations.
The subjects generally interpreted them correctly. They often
verbalized their interpretations with the same words that they
had heard in the testing. During the test they may not have
chosen the picture of two adults playing Pin the Tail on the
Donkey (Appendix C, p. 97 ) for the sentence 'Playing grownups
can be funny.' But when they described that picture they
usually said, 'Two grownups playing Pin the Tail on the Donkey.'
Usually, in their descriptions of the subject nominals, the
ing form followed the noun: 'Two dogs walking,' 'A boy shooting
at the stars,' 'Two houses moving,' or 'A horse painting a
The entire testing procedure took just less than ten
minutes per subject.
The results of this experiment were quite different from
those which were anticipated. The expected pattern was for
the subjects to choose two pictures each for the first ten
sentences, indicating their comprehension of the ambiguities
of these sentences and to choose one picture for each of the
last twenty sentences, indicating their comprehension of these
sentences which were not ambiguous.
On the first ten sentences, those which were ambiguous,
there was no significant difference at the .05 level in per-
ception of ambiguity regardless of age as determined by the
Mann-Whitney U Test (Bruning and Kintz 1977:224). See Table
1. This perception of ambiguity was measured as the number of
times the subjects chose more than one picture to match the
ambiguous sentences. If a subject chose more than two pictures
for a sentence, and sometimes this happened, this still counted
as one instance of perception of ambiguity. This absence of a
significant difference means that the eight-year olds perceived
ambiguity as often as eleven-year olds did. This particular
finding differs considerably from the results which Frank Kessel
Even more surprising were the results on the unambiguous
sentences. The subjects continued to choose more than one pic-
ture per sentence even for the unambiguous sentences. This
pattern of response was the same for all four age groups as
determined by the Mann-Whitney U Test at the .05 level of
significance. See Table 2. In other words, performance did
not improve with age. Regardless of age the subjects con-
sistently chose more than one picture for the unambiguous
Frequently, the subjects chose more than one picture for
the unambiguous sentences more often than they did for the am-
biguous sentences. It appeared that the more they heard the
nominals, the more meaning they perceived for them, regardless
of the fact that the second and third times the subjects heard
the nominals, the nominals occurred in unambiguous sentences.
For example, if a subject heard 'Playing grownups can be
funny' and chose only one picture, indicating he perceived
only one meaning for the sentence, he frequently chose more
than one picture later on in the test when he heard 'Playing
grownups is funny' and 'Playing grownups are funny.' Quite
obviously, subjects who chose more than one picture for the
Mean Number of Choices of Two Pictures
for the Ambiguous Sentences
Age Mean Number of Sentences
8 3.36 11
9 3.90 10
10 4.63 11
11 5.40 15
Mean Number of Choices of Two Pictures
for the Unambiguous Sentences
Age Mean Number of Sentences
8 5.45 11
9 6.60 10
10 8.09 11
11 8.46 15
unambiguous sentences were responding to the ambiguity of the
nominals and not to the verb form is or are which should have
disambiguated the sentences.
Another indication that the subjects' responses were
determined solely by the meaning of the nominal, and not the
nominal in the context of the entire sentence, was the fact
that some subjects saw one meaning for an ambiguous sentence
and they continued to see that same meaning for both of the
unambiguous sentences. For these subjects, too, the verb
form is or are had no effect on their choice of picture.
Often during the testing procedure the subject would
repeat the sentence or part of it to himself. On those
occasions when the subject repeated just the nominal in an
unambiguous sentence and then selected two pictures, the
experimenter cautioned the subject to listen to the entire
sentence and then the experimenter repeated the sentence.
This, however, never had any effect. The subjects consis-
tently chose the same pictures again.
Sometimes the subjects heard an unambiguous sentence
but repeated the ambiguous version of it. The experimenter
responded with 'No; listen carefully. Playing grownups is
funny.' Although stressing the verb form violated the con-
ditions of the experiment because it should have affected the
subjects' responses, there was no effect whatsoever. No
amount of prompting changed the subjects' choices of pictures.
After the test was completed, the experimenter informally
asked those subjects who chose more than one picture for some
of the unambiguous sentences if they could distinguish between
two sentences from the test whose meaning depended on the verb
form is or are: 'Can you tell me the difference between
"Shooting stars is exciting" and "Shooting stars are exciting?"'
All of the subjects, some more consistently than others, were
able to make the distinction. They were not always correct
in their explanations, and they often had problems verbalizing
the difference, but it was definite that when the subjects
heard both sentences together, they were able to distinguish
Two of the subjects, a nine-year-old girl and a ten-year-
old boy demonstrated a thorough understanding of the principles
involved in both the ambiguous and the unambiguous sentences.
Neither selected all of the pictures correctly, but they both
improved as they progressed through the test. The female sub-
ject chose two pictures for the ambiguous sentences eight out
of ten times and for the unambiguous sentences only six out of
twenty times. The six choices of more than one picture for
the unambiguous sentences occurred early in that part of the
test. After the first seven unambiguous sentences, she chose
two pictures only once. When the experimenter asked her,
after the formal testing was completed, why she chose only
one picture for some of the unambiguous sentences, the subject
attempted to explain. She became quite confused in her ex-
planation, lacking the ability to articulate the procedure
she was using. However, she clearly understood the difference.
Her best attempt at an explanation was 'You just don't say it
The male subject articulated more of his thoughts as
he participated in the test. For the ambiguous sentences he
chose two pictures seven times. For the unambiguous sentences,
he chose two pictures only four times. When he heard the un-
ambiguous sentences, he often repeated them slowly and thought
carefully about his answers. His comments indicated his com-
prehension of the sentences and the tasks involved: 'Shooting
stars are exciting?' He then chose the correct picture. For
the sentence 'Shooting stars is exciting,' he commented, 'That
just doesn't seem right. Is exciting? Oh! D.' And for
'Moving houses are frightening,' he responded, 'Now I get it
--houses that are moving.' This subject's indecision seemed
to imply that he had only recently acquired or was still in
the process of acquiring the ability to handle these sentences.
Although neither of these two subjects was correct on all
of the test items, each clearly demonstrated an ability to
distinguish during the test between the ambiguous and unam-
biguous sentences. It was not necessary for these subjects
to hear both unambiguous sentences of a set together in order
for them to distinguish between the sentences. These subjects
seemed well on their way to acquiring the linguistic structures
and cognitive capacities necessary for the tasks of this ex-
periment. They were the only two subjects to display this
Because the results of this experiment were so different
from those which were hypothesized, the experimenter decided
to alter the test conditions somewhat in order to determine
the reasons for the unexpected results. Each of the changes in
the test design is considered as a separate experiment, and
each is discussed separately in the following sections.
Since the ambiguous sentences preceded the unambiguous
sentences in Experiment I, the ambiguous sentences may have
been causing the subjects to see more than one meaning for
the unambiguous sentences. For this reason, the order of
presentation of the test sentences was altered so that the
unambiguous sentences preceded the ambiguous sentences.
Subjects. Six subjects, three male and three female,
from P.K. Yonge Laboratory School in Gainesville, Florida,
participated in this experiment. There were three groups
consisting of one male and one female. The groups were
separated by age into 8, 9, and 10 years old. All subjects
were white, native speakers of English.
Apparatus. The apparatus for this experiment were the
same as those used in Experiment I. The only change in the
apparatus was that the twenty unambiguous sentences were pre-
sented before the ten ambiguous sentences. This was done be-
cause it was felt that the subjects may have seen ambiguity in
the unambiguous sentences because they had already been exposed
to the ambiguous sentences.
The procedure was identical to that of Experiment I.
Changing the order of the sentences so that the unambiguous
sentences preceded the ambiguous sentences did not alter the
subjects' interpretations of the unambiguous sentences. Five
of the six subjects saw ambiguity in eight to eleven of the
unambiguous sentences. The sixth subject saw no ambiguity in
any of the sentences. Apparently, the order of presentation,
whether that which was used in Experiment I or that used in
Experiment II, was not relevant to the subjects' perception
With the exception of the student who saw no ambiguity
in any of the thirty test sentences, there was no difference
by age or sex in the perception of ambiguity in both the
ambiguous and the unambiguous sentences. The results of this
experiment suggest that the order of presentation of these
sentences does not affect performance. That the subjects
perceived ambiguity in sentences that were theoretically un-
ambiguous still remained to be explained.
The first two experiments demonstrated four points:
1. Subjects perceived ambiguity in both the ambiguous
and the unambiguous sentences.
2. The order of presentation of test sentences may be
either ambiguous followed by unambiguous sentences or unam-
biguous followed by ambiguous sentences without any effects
in the subjects' perception of ambiguity.
3. The verb forms is and are which should have disam-
biguated the nominals did not.
4. Although the subjects perceived many of the unam-
biguous sentences as ambiguous, they were able to distinguish
between the two sentences of the pair of unambiguous sentences
when those two sentences were presented together for them to
compare. To restate this point, if the experimenter asked a
subject to explain the difference between two sentences which
differed on the surface only in the verb is or are such as
sentences 2.d. and 2.e., the subjects were almost always ca-
pable of making the necessary distinction. They did not,
however, use this information during the test situation. In-
stead, when they heard either 2.d. or 2.e., they indicated that
the sentence was ambiguous by choosing more than one picture
to match the sentence.
2.d. Moving houses is frightening.
2.e. Moving houses are frightening.
Because of these facts, it was felt that if the subjects
were exposed to sentences of the same type as those on the
test before the actual test began, and if they were given
practice in making distinctions between these sentences, this
pretest experience might encourage the students to employ this
distinction during the test.
Subjects. Five students from P.K. Yonge Laboratory
School in Gainesville, Florida, participated in this experi-
ment. Of the five, four were ten years old and one was nine.
Four of the subjects, including the nine-year-old, were female,
and one was male. All of the subjects were white, native
speakers of English.
Apparatus. The apparatus for this experiment were
identical to those in Experiment I with one exception. Four
new, theoretically unambiguous sentences were added as
additional pretest items:
2.f. Bouncing balls is fun.
2.g. Bouncing balls are fun.
2.h. Flying kites is fun.
2.i. Flying kites are fun.
The procedure in this experiment was almost identical
to that of Experiment I. The only difference was that, after
the five original pretest sentences, the experimenter pre-
sented four more pretest sentences to the subjects. The
experimenter asked each subject to listen carefully to the
two sentences which she was about to repeat. The experimenter
then asked, 'Can you tell me the difference between these two
sentences: "Bouncing balls is fun" and "Bouncing balls are
fun"?' The experimenter recorded each subject's responses
and then asked, 'Can you tell me the difference between these
two sentences: "Flying kites is fun" and "Flying kites are
fun"?' The experimenter again recorded the subject's responses
and then began the actual test. The rest of the test continued
in the same manner as in Experiment I.
In general, exposure to the sentence patterns and ex-
perience in distinguishing between pairs of sentences did not
facilitate performance on the test.
Four of the five subjects chose more than one picture
for some of the unambiguous sentences. One chose two pictures
for eight of the twenty unambiguous sentences, one chose two
for nine of the twenty, one for ten of the twenty, and one
for eleven of the twenty.
Only one subject, a ten-year-old female, chose only one
picture for each of the unambiguous sentences. She chose two
pictures for five of the ten ambiguous sentences. It would
appear that she understood the task of the experiment fairly
well, except for the fact that, of the twenty unambiguous
sentences, she chose seven incorrect pictures.
Of the four who continued to see ambiguity in the unam-
biguous sentences, three demonstrated an ability to distinguish
between the two sentences in each pair that was presented in
the pretest. That ability did not, however, appear during the
test situation, since they still chose more than one picture
several times. One subject could not explain a semantic
difference between the two sentences in each pair of unambiguous
sentences in the pretest. For him, the sentences with the
verb is had the same meaning as the sentences with are, except
that he felt those with is were ungrammatical. This feeling
did not stop him during the test from choosing two pictures
for nine of the twenty unambiguous sentences.
This test design did not help the subjects to see the
difference in the sentences with is and the sentences with are.
Because none of the previous test designs consistently
elicited a distinction between the members of the pairs of
unambiguous sentences, a new design was developed. The problem
of the verb be was eliminated by changing the structure of the
sentences which had previously had can be to I like + nominal
('I like visiting relatives'), those which had is to I like
+ infinitive ('I like to visit relatives'), and those which
had are to I like + noun + that are + verb + ing ('I like
relatives that are visiting'). It was hoped that eliminating
the verb be would eliminate the subjects' problems in dis-
tinguishing between the two types of nominals.
Subjects. Four students from P.K. Yonge Laboratory
School in Gainesville, Florida, were chosen to participate
in this experiment. There were two ten-year-old males and
two eleven-year-old females. All four subjects were white,
native speakers of English.
Apparatus. This test design included the five pretest
sentences which were used in Experiment I to assure that the
subjects could adequately perform the task of selecting more
than one picture if more than one picture was appropriate to
the sentence which they heard. The test sentences, which
appear in Appendix B, consisted of twenty-seven sentences,
nine of which were ambiguous and eighteen of which were unam-
biguous. Of the nine ambiguous sentences, six consisted of
I like + nominal, in which the nominal could be either a sub-
ject nominal or an object nominal. The other three ambiguous
sentences were the same three from the original test design
which were not built on the nominal + can be + adjective
structure. The unambiguous sentences consisted of the six
sentences from the original design which were not dependent on
the verb be and of twelve sentences which were interpretations
of the I like + nominal sentences. These twelve sentences
consisted of six I like + infinitive structures and six I like
+ noun + that are + verb + ing, as in the following set:
2.j. I like playing grownups.
2.k. I like to play grownups.
2.1. I like grownups that are playing.
As was true of the design for Experiment I, each unam-
biguous sentence represented one of the meanings contained in
the ambiguous sentences. Again, the purpose of this design
change was to eliminate the problems that arose with the verb
be. The remaining aspects of the apparatus--the slides,
projector, and score sheet--were the same as those in Experi-
This aspect was also similar to that of Experiment I.
The only difference was in the sentences that were read.
On the nine ambiguous sentences, two subjects chose two
pictures seven times, one subject chose two pictures eight
times, and one chose two pictures nine times. On the eighteen
unambiguous sentences, one subject chose two pictures five
times, one seven times, one nine times, and one ten times.
This design, like the ones that preceded it, failed to
elicit a consistent distinction between subject and object
nominals. In even the most unlikely cases, for example, 'I
like to visit relatives,' two of the subjects chose two
No matter what the method of testing, it seemed that the
subjects could not be forced into distinguishing between when
the subject-nominal interpretation was appropriate and when
the object-nominal interpretation was appropriate. For some
reason the subjects appeared to be treating all of the nominals
as surface lexical items rather than as derived structures.
Because there were no age differences in either the per-
ception of ambiguity or in performance on the unambiguous
sentences, it remained unclear as to when subjects actually
begin to make the distinctions necessary to perform well on
the unambiguous sentences. Some of the results of the pre-
ceding experiments implied that the ability to interpret the
relationships among the trio of sentences made up of an am-
biguous sentence and its two unambiguous counterparts, the
ability to distinguish between the two unambiguous sentences
in the trio, and the ability to recognize that the unambiguous
sentences were in fact unambiguous were available only to
those people who have an overt interest in language. These
abilities certainly did not seem to be a part of these native
speakers' intuitions about language.
It was decided that it would be helpful to know how
college-aged adults would perform the tasks of Experiment I in
this investigation. If they had no problems with the test
sentences, then somewhere between the ages of 12 and 18 years,
there should be a point which could be labeled 'the normal
age of acquisition' of the abilities needed to handle the
elements of this test. If adults were not able to make complete
distinctions between the unambiguous sentences and were not
sure that the unambiguous sentences were unambiguous, then
these abilities most probably belong only to those people who
consciously analyze their language.
This latter possibility seems remote because it would be
quite wasteful for a language to offer distinctions for its
speakers if the speakers do not naturally utilize them. These
distinctions should not exist for the language-conscious only,
but should be an active part of any native speaker's ability
to interpret his language. On the other hand, it is possible
that this wasteful aspect is prevalent in language. Many
people may never acquire certain transformations.
Subjects. Seven students from the University of Florida
were chosen haphazardly to participate in this experiment.
These students ranged from 18 to 22 years in age. Three were
males and four were females. All were white, native speakers
Apparatus. The apparatus were identical to those in
The procedure was identical to that of Experiment I. The
only change was in location. Experiment I was conducted at
P.K. Yonge Laboratory School. Experiment V was conducted in
General Purpose Building A on the main campus of the University
On the whole, these adult subjects demonstrated a more
thorough comprehension of the elements of this test. The ten
ambiguous sentences presented more problems to these subjects
than the unambiguous sentences did. No subject chose two
pictures for all of the ambiguous sentences. The seven sub-
jects each heard ten ambiguous sentences, for a total of
seventy total exposures to the ambiguous sentences. Out of
these seventy exposures, the subjects correctly chose two
pictures twenty-nine times or 41 percent of the time.
Performance on the unambiguous sentences was much better.
Twenty unambiguous sentences presented to the seven subjects
provided one hundred forty exposures to the unambiguous sen-
tences. The subjects chose two pictures only fourteen times
out of the possible one hundred forty times, or for 10 percent
of the time. This was by far the fewest times that subjects
in any one test design chose two pictures for the unambiguous
sentences. Even though these subjects chose only one picture
for the unambiguous sentences more frequently than the subjects
in previous test situations, they chose the wrong picture
twenty-nine times or 21 percent of the time.
In order for subjects at a particular stage in develop-
ment to have mastered a specific aspect of their language,
they should demonstrate an ability to use that part of their
language consistently. In other words, subjects should reach
an age at which it becomes useless to test them because at
that age they show no difficulty with the test items. This
does not mean that all the subjects in that age group must
get 100 percent of the items correct. There are numerous
reasons why some errors may show up. But there should be a
consistent demonstration of proficiency with the test items.
The subjects in this test have not demonstrated such a
proficiency. Perceiving only 41 percent of the ambiguities
and seeing ambiguity in unambiguous sentences 10 percent of the
time with twenty-nine errors in the choice of pictures for the
unambiguous sentences would not seem to be a demonstration of
proficiency with these test items. It is true that these sub-
jects chose two pictures only 10 percent of the time, whereas
the subjects in the previous experiments chose two pictures
from 28 percent to 39 percent of the time.
The subjects' performances during the test and their
comments after the test often indicated different things. For
example, one subject chose two pictures for only three am-
biguous sentences and chose only one picture for each of the
unambiguous sentences. Although the test demonstrated that
this subject understood the unambiguous sentences, it did not
indicate that he perceived the ambiguity in the first ten
sentences. His comments after the test, however, showed that
he understood not only the ambiguities ('with can be, it can
have two meanings') but also the design of the test ('changing
from can be to is and are'). Although this ability to describe
the test design seems quite simple, very few of the subjects,
regardless of age, were able to explain what was happening
during the test.
Another subject's performance during the test differed
from his comments after the test. This subject chose two
pictures for seven of the ten ambiguous sentences and for
five of the twenty unambiguous sentences. Yet he commented
that all of the sentences with is were ungrammatical.
Like the younger subjects, the adults were able to dis-
tinguish between sentences with is and those with are if the
two sentences were presented together for them to compare.
They, too, did not always employ this distinction when they
Discussion of Experiments I Through V
An experiment was designed to determine the age at which
children develop the ability to perceive deep-structure am-
biguity as in the sentence 'Flying kites can be exciting.'
It was hypothesized that the ability to perceive both meanings
in this sentence would develop only after the child could com-
prehend both of the unambiguous sentences which represent the
meanings which are contained in this ambiguous sentence. These
two sentences are 'Flying kites is exciting' and 'Flying kites
are exciting.' The results of this experiment were completely
different from those which were anticipated.
There was no significant age difference in ability to
perceive the ambiguity of the ambiguous sentences. What the
subjects failed to comprehend was the fact that the unambiguous
sentences were unambiguous. These subjects repeatedly saw more
than one meaning for the unambiguous sentences.
In order to determine whether the subjects' behaviors were
the result of some feature of the design or if, in fact, the
subjects could not comprehend that these sentences were not
ambiguous and if not, why, four more designs were used. The
results from these four experiments indicated that with the
exception of the adults and regardless of test design the
subjects were unable to tell that the unambiguous sentences
were unambiguous, unless the pair of unambiguous sentences
was presented together for the subjects to compare. Only then
could the subjects explain the difference between the two sen-
tences and recognize that these sentences were not ambiguous.
This inability on the part of these subjects to analyze
the unambiguous sentences unless they were presented in pairs
suggests that the ability to perform such an analysis is
available only to those people who consciously analyze their
Although it may seem strange that a language may offer
distinctions which its speakers do not naturally use, this
happens quite often in terms of vocabulary items which people
may never learn or use and transformations which some people
may never acquire. That many people may never be able to
distinguish between the unambiguous sentences would not hinder
their ability to communicate since most ambiguities become
disambiguated in the contexts in which they are used.
The results of these five experiments raised a large
number of questions. The most obvious question was 'why did
the subjects perform so differently from the way that the
theory had led me to predict?' This question, of course,
embodied many other questions.
Since the only difference in the surface structure in the
pairs of unambiguous sentences such as 'Flying kites is ex-
citing' and 'Flying kites are exciting' lies in the verbs is
and are, it appeared that these verbs were not meaningful to
the subjects in their interpretations of the sentences. There
was no question that the subjects heard the verbs, since the
experimenter violated the conditions of the original experi-
ment to make sure that they heard these verbs. What was un-
clear was whether it was the verb be or the presence or ab-
sence of the plural marker on the verb which was causing prob-
lems for the subjects. It seemed quite possible that it was
the verb be because this verb carries no meaning.
In order to determine which of these factors was operating,
a completely new test design was developed. This design
tested both the verb be and the plural marker in the contexts
of twelve ambiguous sentences and twelve unambiguous sentences.
The subjects who participated in the first five experi-
ments demonstrated an inability to properly analyze sentences
whose structures were of the form nominal + be + adjective,
when the nominal was structurally ambiguous. The sentences
which the subjects heard were either ambiguous because the
verb form did not determine which meaning of the nominal
should be available or unambiguous because the verb form did
determine which meaning of the nominal should be available.
Sentence 3.a. is an example of the ambiguous sentence type,
and sentences 3.b. and 3.c. are examples in which the verb
form should disambiguate the nominal.
3.a. Moving houses can be dangerous.
3.b. Moving houses is dangerous.
3.c. Moving houses are dangerous.
From a descriptive standpoint, the singular form of a verb
determines that the nominal be interpreted as an object
nominal; that is, in sentence 3.b., houses is the object of
moving. The plural form of the verb determines that the
nominal be interpreted as a subject nominal; for example, in
sentence 3.c., houses is the subject of the sentence and
moving functions as an adjective modifying houses. The verb
in sentence 3.a., can be, has no inflection for number and,
consequently, provides no information to disambiguate the
Apparently, not all native speakers, even some adults,
make these distinctions between 3.b. and 3.c. unless they
have both sentences available at the same time. Most of the
subjects in the previous experiments viewed sentences of the
form of 3.b. and 3.c. as ambiguous. It appears that the
syntactic information contained in the verb was not enough to
override the ambiguity of the nominal. It was not clear
whether the problem resided in the verb be which has no meaning
or in the presence or absence of the plural marker.
In order to determine what factors were operating for
the subjects, a new experiment was designed. Twenty-four
sentences based on the nominal + verb + adjective structure
made up the test. In this design both the verb be and the
verb look were tested in both singular and plural forms. This
was done to determine if subjects would have the same problems
if the main verb in the sentence were not some form of be.
Another variable in this design was the type of nominal.
Twelve of the nominals were ambiguous and twelve were not. All
of the nominals contained a plural noun. The subjects were
asked to select the correct verb form; that is, they were to
circle is or are in twelve sentences and looks or look in
twelve sentences. For the ambiguous nominals either form of
the verb would be correct. (See sentences 3.d. and 3.e.)
However, for the unambiguous nominals there was only one
correct verb form for each sentence. (See sentences 3.f.
3.d. Sailing boats is/are dangerous.
3.e. Sailing boats looks/look dangerous.
3.f. Making cookies is/are fun.
3.g. Making cookies looks/look fun.
3.h. Roaring lions is/are scary.
3.i. Roaring lions looks/look scary.
The nominals in sentences 3.f. and 3.g. are object nominals,
and the only grammatical verb form in each sentence is the
singular form. The nominals in sentences 3.h. and 3.i., how-
ever, are subject nominals and are grammatical with only the
plural form of the verb.
Because the subjects in the previous experiments had
viewed sentences such as 3.d. as ambiguous, it was suspected
that the subjects in this experiment might do the same. It
was also expected that some subjects might choose the plural
verb form for all of the nominals because the nominal phrases
end in s and the plural noun is next to the verb. This type
of behavior constitutes a strategy which may not always pro-
duce the correct result. For example, in sentence 3.f. some
speakers may choose are because the noun cookies, which is
plural, is next to the verb. Although the word cookies is
not the subject of the sentence, it was felt that some speakers
might behave as if it were. Such a strategy would generate
the ungrammatical sentences 3.j. and 3.k.
*3.j. Making cookies are fun.
*3.k. Making cookies look fun.
This strategy applied to subject nominals also generates the
plural verb form, but in this case, it is the correct form.
If this strategy is actually being utilized, there should be
significantly more errors with object nominals than with sub-
The subjects' behavior on this test should indicate the
importance of the verb form in interpreting the nominal. If
there were not significantly more errors with object nominals
than with subject nominals, it might mean that the verb form
was irrelevant to the interpretation of nominals. This would
follow if the syntactic information contained in the verb were
not enough to override the ambiguity of the nominal.
There were two other questions which this design attempted
1. Is there a developmental point at which the verb be-
comes relevant to the interpretation of the nominal?
2. Is this linguistic behavior in any way a function of
It seems that the ability to distinguish between 'Flying
kites is fun' and 'Flying kites are fun' should not be related
to how well a person has learned his English grammar lessons.
Since the two sentences represent differences in meaning, they
should be different to all native speakers of English, who
presumably cut up the semantic spectrum along the same lines.
There is some indication, however, that these sentences have
the same meaning for some speakers, and consequently, they do
not share the same divisions of the semantic spectrum with
those speakers for whom these two sentences have different
meanings. It may happen that the only explanation for this
linguistic behavior is that, unless the subject is overtly
and consciously aware of the rules of his grammar, he does
not attend to all the grammatical technicalities of a sentence.
These subjects rely, quite probably, on context. Contextual
factors ordinarily eliminate the need to analyze a sentence
closely enough to determine whether a speaker is talking
about the act of flying kites or about kites which are flying.
It was in an attempt to sort out some of these factors
that the following experiment was run.
Subjects. One hundred eighty-three subjects participated
in this experiment. One hundred thirty of these subjects were
students between the ages of seven and twelve years old from
P.K. Yonge Laboratory School in Gainesville, Florida. The
other fifty-three subjects were students at the University of
Florida between the ages of eighteen and twenty-nine. In all
there were eighty-nine females and ninety-four males. All sub-
jects were native speakers of English. In this experiment
there were no restrictions on race.
Apparatus. The apparatus consisted of one sheet of paper
for each subject with places at the top for name, age, and sex
and with twenty-four sentences, each containing the singular
and the plural forms of the verb be in the present tense or
the verb look in the present tense. All of the sentences had
a nominal + verb + adjective format. In twelve of the sen-
tences the nominal was ambiguous; that is, it could be either
a subject nominal or an object nominal (for example, 'flying
kites'). For these twelve sentences, both the singular and
the plural forms of the verb were grammatical. In the other
twelve sentences the nominals were not ambiguous; six nominals
were subject nominals and six were object nominals. The sub-
ject nominals were grammatical with only the plural verb, and
the object nominals were grammatical with only the singular
verb. Each nominal in the test was paired with the verb be
in one sentence and the verb look in another sentence. The
order of the sentences on the test was random. A copy of
this test is in Appendix D.
This test was administered in the classrooms at P.K.
Yonge and at the University of Florida by the classroom in-
structors. Each instructor passed out the test forms and
directed the students to read each sentence carefully and to
circle the correct form of the verb. No other instructions
were given. However, for one group, an adjustment had to be
made. The youngest group, the seven-year olds, was incapable
of reading the test, so their teacher read it out loud to
them. Because intonation pattern does not affect the meaning
of these sentences, no problems arose from the sentences being
Two forms of this test were administered. In one form
the order of the twenty-four sentences was the reverse of the
order for the second form. One half of the subjects received
form one and the other half received form two.
The data were analyzed using the SAS General Linear Model
Procedure computer program for analysis of variance. In the
ambiguous sentences subjects could not make an error in their
choice of verb form. However, because it was thought that
subjects might employ a strategy of selecting the plural form
of a verb because the noun which precedes the verb is always
plural, it was important to know whether the subjects chose
the singular or the plural form of the verb in the ambiguous
sentences. Therefore, in order to distinguish the responses
for the unambiguous sentences from those for the ambiguous
sentences, the verbs were divided into six categories: two
singular--is and looks, two plural--are and look, and two
categories labelled be and 1k for the ambiguous sentences.
The subjects' responses were recorded in the following manner:
for the unambiguous sentences, a correct response was coded
as 0, and an incorrect response was coded as 1. Because there
was no way to make an incorrect choice of verb in the ambiguous
sentences, selection of the singular verb form was coded as 0
and of the plural verb form was coded as 1.
The data were also analyzed from the perspective of the
nominals. There were three categories of nominals. For the
unambiguous sentences the categories were either Subject or
Object nominals. For the ambiguous sentences, the nominal
could be interpreted as either subject or object, and, there-
fore, the category for the nominals in the ambiguous sentences
was labelled Subj-Obj. The numbers in the first two categories
were the number of times the subjects chose the wrong verb form
for a particular nominal type. The numbers in the category
Subj-Obj were the number of times subjects chose a plural verb
form for the ambiguous sentences. The responses for nominals
also represent a collapsing of the verbs which were appropriate
for the particular nominal type. For example, the mean for
nine-year-old males with respect to subject nominals represents
a collapsing of the scores for nine-year-old males on the verbs
are and look because these two verb forms are the correct re-
sponses for subject nominals.
One final point, all of the subjects older than twelve
were combined into one group. The subjects actually were be-
tween the ages of eighteen and twenty-nine, but they were all
considered in one age group which was labelled 21-year olds.
Verbs in unambiguous sentences. For those sentences
which were not ambiguous, there was a three-way interaction
of sex, age, and verb which was significant at the .01 level.
The two-way interaction of sex and verb was not significant at
the .01 level. The age by verb interaction and the sex by
age interactions were significant, the former at an alpha level
of .0001 and the latter at an alpha level of .0027. Although
the main effect of sex was not significant, the main effects
of age and of verb were significant at the .0001 level.
Figures 1 through 4 indicate the percent of incorrect responses
for male and female subjects at each age interval for the four
verb forms. Figure 1 indicates the number of times subjects
chose the verb are when the correct response was is. Figure
2 indicates the number of incorrect responses for are, Figure
3 for looks, and Figure 4 for look. If the four figures are
imposed on one graph, the three-way interaction becomes evi-
dent. These scores are also listed in Table 3.
Is. When the correct response was is, the seven-year-old
male and female subjects' performances of 47 and 42 percent
respectively were not significantly different. The eight-year-
old subjects were not significantly different either from each
other or from the seven-year olds. The nine-year-old males
continued to perform in the same manner as the seven- and
eight-year olds. The nine-year-old females, however, improved
significantly, choosing the verb form are only 28 percent of
the time when they should have chosen is. At the age of ten,
female subjects performed just as the nine-year-old females.
The males, however, did worse at this age than at any other
age, choosing are 60 percent of the time when they should have
chosen is. The responsesof the eleven-year olds were almost
the reverse of those for the ten-year olds. The male subjects'
scores improved to 24 percent, while the females' scores
worsened by 28 percentage points to 57 percent. For the
twelve-year olds, there was another reversal. The males'
scores rose to 43 percent incorrect, while the females improved
by 40 percent to 17 percent incorrect. Finally, the scores
for male and female adults were nearly identical at 11 percent
and 12 percent, respectively.
Are. The subjects' behavior when the correct response
was are remained more consistent. At the ages of 7, 9, 10,
and 11, females did slightly, but not significantly, better
than males. At the ages of 8 and 21, males did slightly better.
And at the age of 12, both males and females had an identical
score of 0 percent incorrect responses.
Looks. At the age of 7, male subjects chose the verb
form look 13 percent of the times when they should have chosen
looks, whereas the females did this 18 percent of the time.
At the age of 8, both males and females did worse than the 7
year olds. The females made errors 26 percent of the time,
while the males were incorrect 44 percent of the time. At the
age of 9, the males improved to 33 percent and the females
improved to 4 percent. At the age of 10, both males and
females made errors 17 percent of the time. At 11, males
improved slightly to 12 percent and females remained almost
the same at 18 percent. At 12, the males had slightly more
errors--14 percent, and females made slightly fewer errors--8
percent. The adult males again improved slightly, this time
to 9 percent, and the females made errors 15 percent of the
time. From the age of ten to adult, none of the scores was
significantly different. Although the scores for male and
female subjects alternated in terms of which sex had fewer
errors at each of these ages, the fact that there was no
Responses on the Four
and Female Subjects
Age Sex Is Are Looks Look
7 M 47 13 13 53
7 F 42 6 18 30
8 M 49 10 44 33
8 F 46 13 26 49
9 M 45 9 33 55
9 F 28 5 4 48
10 M 60 17 17 33
10 F 29 13 17 28
11 M 24 15 12 18
11 F 57 9 18 27
12 M 43 0 14 0
12 F 17 0 8 25
21 M 11 1 9 4
21 F 12 1 15 3
Note. All scores are rounded off to the nearest
45- ..* .' **
0 -, -----
7 8 9 10 11 12 21
Figure 1. Percent of incorrect responses for the verb
form Is by male and female subjects.
10- .. -
5- *** '*.
7 8 9 10 11 12 21
Figure 2. Percent of incorrect responses for the verb
form Are by male and female subjects.
Percent of incorrect responses for the verb
form Looks by male and female subjects.
I I I 1 I 1
7 8 9 10 11 12 21
25- "" ***.
7 8 9 10 11 12 21
Figure 4. Percent of incorrect responses for the verb
form Look by male and female subjects.
significant difference in these scores indicates that there
was no genuine improvement after the age of 10.
Look. At the age of 7, male subjects chose looks 53
percent of the times that they should have chosen look. Fe-
male subjects chose looks incorrectly 30 percent of the time.
At eight, male subjects chose looks incorrectly 33 percent,
while females made errors 49 percent of the time. Nine-year-
old male and female subjects made errors 55 and 48 percent of
the time, respectively. Both male and female subjects improved
significantly by the age of ten to 33 and 28 percent, respec-
tively. At 11, males improved to 18 percent while females
stayed at 27 percent. At 12, the males again improved; this
time they made no errors. The females improved slightly to
25 percent. As adults, the male subjects made errors only
4 percent of the time and females improved to 3 percent.
Nominals in unambiguous sentences. This category is
simply a different perspective on the same data. The scores
are the combined scores for both the verbs which were gram-
matical for a particular nominal type. For example, in the
sentence 'Growling lions is/are--looks/look scary,' both are
and look are grammatical because this sentence contains a
subject nominal. These scores, then, indicate the number of
errors in choice of verb that males and females at each age
interval made for a particular type of nominal. The purpose
of looking at these data in this manner is to determine which
type of nominal presented more difficulty to the subjects.
The result of this analysis was that there was a signifi-
cant three-way interaction of sex, age, and nominal at the .01
level of significance. The two-way interactions of sex by
nominal and of age by nominal were not significant at the .01
level. The sex by age interaction was significant at the
level of .0047. The main effect of sex was not significant,
but the main effects of age and of nominal were significant
at the level of .0001. These scores are listed in Table 4
and the interaction is evident in Figure 5.
At the age of 7, male and female subjects made errors
with object nominals 30 percent of the time. For the subject
nominals, males made errors 33 percent and females 18 percent
of the time. At eight, male subjects improved on subject
nominals to 22 percent but had 46 percent incorrect with ob-
ject nominals. Females made more errors on both subject and
object nominals, 31 percent and 40 percent, respectively.
At nine, males improved with object nominals to 39 per-
cent but with subject nominals they were incorrect 32 percent
of the time. The females improved on subject nominals to 26
percent incorrect and on object nominals to 17 percent in-
correct.- At 10, the males stayed the same on object nominals
but improved with subject nominals to 25 percent incorrect.
Females improved with subject nominals to 20 percent but had
23 percent incorrect on object nominals. At the age of 11,
male subjects improved somewhat to 17 percent on subject
nominals and improved significantly to 18 percent on object
Percent of Incorrect Responses on Subject and Object
Nominals by Male and Female Subjects
Age Sex Subject Nominal Object Nominal
7 M 33 30
7 F 18 30
8 M 22 46
8 F 31 40
9 M 32 39
9 F 26 17
10 M 25 39
10 F 20 23
11 M 17 18
11 F 18 38
12 M 0 29
12 F 13 13
21 M 2 10
21 F 2 14
Note. All scores are rounded off to the nearest
g20- .* *.. \
a0- \ ".
7 8 9 10 11 12 21
Subject Nominals, Males
.****** Object Nominals, Males
--- Subject Nominals, Females
Object Nominals, Females
Figure 5. Percent of incorrect responses on subject and
object nominals by male and female subjects.
nominals. Female subjects improved just slightly to 18 percent
with subject nominals but made significantly more errors, 38
percent, with object nominals.
For the 12-year olds, there was a 29-point divergence
between the scores for the male subjects, as they had 29 per-
cent incorrect with object nominals and no errors with subject
nominals. The female subjects made errors 13 percent of the
time for both subject and object nominals. Adults were more
consistent by sex. The males had 10 percent incorrect for
the object nominals while females had 14 percent incorrect
for the object nominals, and both sexes had 2 percent incor-
rect for the subject nominals.
Verbs in ambiguous sentences. The scores in this cate-
gory indicate the percentage of times that the subjects chose
a plural verb form in each ambiguous sentence. The results
were that the three-way interaction of sex by age by verb and
the two-way interaction of sex by verb were not significant
at the .01 level of significance. The two-way interactions
of verb by age and of sex by age were significant to the
level of .0001. The main effect of sex was not significant
at the .01 level, whereas the main effects of verb and of age
were significant to the .0001 level. Table 5 contains a list
of the scores for the verb-by-age interaction. Figure 6 in-
dicates the percentage of plural verb forms that the subjects
chose for each main verb at each age interval for those sen-
tences which were ambiguous.
Percent of Choices of Are and Look
in the Ambiguous Sentences
Age Are Look
7 38.0 75.0
8 35.0 62.0
9 54.5 77.0
10 57.0 79.0
11 44.5 77.5
12 71.5 63.5
21 76.5 75.0
Note. All scores are rounded off to the nearest
one-half percentage point.
...*** .**** ...
O 75- *. ..
20 a i |
7 8 9 10 11 12 21
Figure 6. Percent of choices of Are and Look in
the ambiguous sentences.
There is a pattern from the age of 7 through the age of
11 for the subjects to choose the plural form at least 23 per-
cent more often for the main verb look than for the main verb
be. The differences were significant at the .01 level at the
ages of 7, 8, 10, and 11. At the ages of 12 and 21, subjects
chose slightly more plural forms for be than for look. The
differences for these two ages were not significant. Overall,
as the age of the subjects increased, the choices of plural
verb forms also increased. The choices of plural verb form
for the verb look in the ambiguous sentences were between 62
percent and 79 percent for all ages. However, for the verb
be the percentage of plural responses ranged from 35 percent
for the eight-year olds to 76.5 percent for the adults.
The scores for the age-by-sex interaction are listed in
Table 6, and Figure 7 illustrates this interaction. At the
ages of 7, 8, 11, 12, and 21, male subjects chose plural verb
forms more often than female subjects did. The females chose
more plurals at the ages of 9 and 10. Only the differences
at the ages of 10 and 11 were significantly different at the
.01 level. From the ages of 7 to 8 there was a decline in
plural choices for both sexes. From 10 to 11, there was a
decline in the plural choices for females. From 11 to 12,
there was a slight decline among the male subjects. In all
other instances, the number of choices of plural verb forms
Nominals in ambiguous sentences. Again, to look at the
nominals is simply to look at the same data from a different
Percent of Choices of Plural Verb Forms in the
Ambiguous Sentences by Male and Female Subjects
Age Male Female
7 65.0 48.0
8 49.0 47.0
9 59.5 72.5
10 61.0 75.0
11 73.5 48.0
12 68.0 66.5
21 76.0 75.0
Note. All scores are rounded off to the nearest
one-half percentage point.
I I i I I I I
7 8 9 10 11 12 21
Percent of choices of plural verb forms in the
ambiguous sentences by male and female subjects.
perspective. In this case, to look at the number of times
that the subjects chose a plural verb form is to determine
how often they interpreted the ambiguous nominal as a subject
nominal. The result was a two-way interaction of sex and age
which is significant at the .01 level. The main effect of
sex was not significant at the .01 level. However, the main
effect of age was significant at the .0001 level of signifi-
cance. The scores for the nominals in the ambiguous sentences
are the same as the scores for the age-by-sex interaction for
verbs in ambiguous sentences. Therefore, Table 6 and Figure
7 are applicable here and will not be repeated.
Because of the significant three-way interactions, it
is impossible to draw any clear conclusions about the data.
There are, however, some trends which can be discussed.
Verbs in unambiguous sentences. When the correct re-
sponse was are, the subjects had the fewest problems. The
highest percentage of errors with are was 17 percent for the
ten-year-old males. Performance with this verb form was also
the most consistent: There was considerably less fluctuation
in the scores for this verb form than for the others.
For all four verb forms, the number of errors for the
twelve-year olds and adults was smaller than the number of
errors for the seven-year olds. The greatest amount of im-
provement occurred with the verb forms look and is. The
least amount of improvement occurred with looks.
Between the ages of seven and eleven, the subjects'
performance on the two plural forms look and are never over-
lapped as there were significantly more errors with look at
the ages of seven, eight, and nine, and for the ten-year-old
males. With the exception of the twelve-year-old female sub-
jects' performance with look, all the twelve-year olds and
adults had almost identical scores on look and are, and those
scores were equal to 0 percent for the twelve-year olds and
very close to 0 percent for the adults. Consequently, the
twelve-year-old and adult subjects had fewer problems when
the correct answer was a plural verb form.
Nominals in unambiguous sentences. This perspective on
the data also indicates that there can be no automatic con-
clusions. The seven-year-old subjects performed better than
the eight-year olds in all cases except for the males with
subject nominals. After the age of eight, performance on the
object nominals was highly inconsistent for both males and
females. Females began to improve after the age of eight with
subject nominals, and they continued to do so all the way to
adults. Males did worse as nine-year olds with subject
nominals, but after that they steadily improved. At the age
of twelve they had 0 percent incorrect with subject nominals,
and as adults, their score of 2 percent incorrect was not
significantly different from 0 percent. The adults' per-
formances on object nominals of 10 percent incorrect for males
and 14 percent incorrect for females, although higher than
that of the subject nominals, were not significantly different.
All of the subjects seemed to acquire the ability to handle
subject nominals earlier and more completely than they did
Verbs and nominals in ambiguous sentences. The scores
in this category are the number of times that subjects chose
a plural verb form (are or look) in the ambiguous sentences,
or, in other words, how many times they interpreted the nominals
in the ambiguous sentences as subject nominals. Between the
ages of seven and eleven the subjects interpreted the nominals
in the ambiguous sentences as subject nominals 23 percent to
37 percent more often for those sentences with look than for
those with be. At the age of twelve and as adults, subjects
interpreted the nominals with be as subject nominals more
often than those with look. Thus the interaction of age and
verb was significant. As adults the subjects' performances
with both verbs were nearly identical. The scores for look
were high for all ages and finished at 75 percent. The scores
for be were as low as 38 percent and 35 percent for the seven-
and eight-year olds, respectively, and finished at 76.5 per-
cent for the adults. Consequently, the subjects showed a
tendency to choose the plural form more often as they got
Having considered the statistical results of this test,
we should attempt to answer those questions which were posed
at the outset of this experiment.
Since the subjects in the first five experiments perceived
so many sentences as ambiguous, would the subjects in this
experiment perceive ambiguity in this format? Not only did
the subjects not perceive ambiguity in the unambiguous sen-
tences, they did not indicate that they perceived the ambiguity
in those sentences that were ambiguous. One subject, a twenty-
two-year-old female selected one verb for each sentence but
noted (incorrectly) at the bottom of the test form that all
of the sentences were ambiguous, and, therefore, all of the
verb forms were appropriate. All other subjects selected
only one verb form for each sentence. It is, however, quite
possible that the instructions--'Circle the correct verb form
in each sentence'--influenced the subjects' performance.
The subjects in the first five experiments did not
seem to use the syntactic information available in the verbs
in the test sentences. Because the verb forms in these
sentences were forms of the verb be, a verb which is
semantically empty, it was felt that some of the problems
which the subjects encountered might be related to the
nature of the verb be. For that reason, the verb look was
included in this experiment. If the subjects had problems
with be but not with look, then there would be evidence that
the semantically empty verb be was responsible for the prob-
lems of the subjects in the first series of experiments. If
the subjects had as many or more problems with look as with
be, then their problems must reside with the syntactic in-
formation carried by the verb and not with the semantics.
The subjects in this experiment had almost identical
totals of errors with be and look. If the scores for the
verb forms is and are are collapsed, and if the scores for
looks and look are collapsed, and if the sexes are collapsed
for both main verbs, the result is one score each for the
main verbs be and look at each age interval. These scores
are listed in Table 7 and graphed in Figure 8. It was
pointed out earlier that the three-way interaction of age,
sex, and verb in the unambiguous sentences was significant.
Percent of Incorrect Responses for the
Main Verbs Be and Look
7 27.0 28.5
8 29.5 37.5
9 22.0 35.0
10 29.5 23.5
11 26.5 18.5
12 14.5 12.0
21 6.0 7.5
Note. All scores are rounded off to the nearest one-
half percentage point.
aThe scores for Be represent the collapsing of the
scores for the verb forms is and are for both males and
bThe scores for Look are the collapsed scores for
looks and look for both males and females.
Percent of incorrect responses for the main
verbs Be and Look.
Because of this significance, collapsing verbs and sexes is
not legitimate statistical procedure. However, having one
score for the main verbs makes comparison of those verbs
easier. So, keeping in mind that this is a violation of
statistical procedure and that, as a result, it obscures
a large amount of information, this information will be
considered in collapsed form.
Seven-year olds had nearly equal trouble with be and
look, with an average of 27 percent and 28.5 percent in-
correct, respectively. Eight-year-old subjects had more
errors with both verbs: slightly more with be at 29.5
percent and somewhat more with look at 37.5 percent. The
greatest discrepancy occurred at the age of nine. These
subjects improved to 22 percent with be and to 35 percent
with look, for a difference of 13 percentage points between
the two. At the age of ten, subjects had more problems with
be, averaging 29.5 percent incorrect and improved considerably
with look to an average of 23.5 percent. After the age of
ten, subjects improved with both verbs at each age interval.
The seven-, eight-, and nine-year olds and adults had more
problems with look while the ten-, eleven-, and twelve-year
olds had more problems with be. It appears that the sub-
jects began to resolve their problems earlier with look than
with be, since their scores improved from the age of eight
onward, whereas they did not improve consistently with be
until after the age of ten.
Overall, the subjects' performances with be and look
were quite similar. There seems to be no evidence that
the semantically empty nature of be was responsible for
the problems which the subjects had in this syntactic con-
The third area under investigation was how the subjects
would handle the singular and plural verb forms. It was
suggested that if the subjects had problems in their choice
of verb form, they would choose the plural verb form more
often because in every sentence, regardless of its phrase
structure, a plural noun occurred immediately before the
There are two ways to determine how the subjects handled
singular and plural markers. The first way was to look at
the subjects' performance with nominals in the unambiguous
sentences. Figure 9 represents the information presented
earlier as Figure 5, with the exception that the variable
sex has been collapsed. Again, the collapsing is a viola-
tion of statistical procedure because of a significant three-
way interaction of sex, age, and nominal, but it provides
a clearer comparison of the nominals. Figure 9 and Table 8
indicate that for every age except nine the subjects had
more problems with object nominals than with subject nominals.
This means that the subjects chose plural verb forms when
they should have chosen singular forms more often than they
chose singular when they should have chosen plural. Eight-
year olds made the most errors in choice of plural form.
Percent of Incorrect Responses on Subject and Object Nominals
Age Subject Nominals Object Nominals
7 25.5 30
8 26.5 41
9 29.0 28
10 22.5 31
11 17.5 28
12 6.0 21
21 2.0 12
Note. All scores are rounded off to the nearest one-
half percentage point. The variable sex has been collapsed.
E 40 .*'.
U 30 .****-..
0 II II
7 8 9 10 11 12 21
-__-* Subject Nominals
*....** Object Nominals
Figure 9. Percent of incorrect responses on subject
and object nominals.
Adults made the fewest errors with both singular and plural
The second perspective on singular and plural markers
is in the subjects' performance with the ambiguous sentences.
Figure 10 and Table 9 indicate the number of times that the
subjects chose are and look in the ambiguous sentences. Sex
and verb are collapsed in this figure, and consequently, a
lot of information is obscured. However, for the purposes
of this point, this information is sufficient. Because this
figure indicates the percentage of plural verb forms which
the subjects chose at each age interval, the percentage of
singular forms chosen can be determined by subtracting the
percentage of plural choices at a particular age from 100
percent. Therefore, nine-year-old subjects chose a plural
verb form 66 percent of the time for the ambiguous sentences,
and they chose a singular form 34 percent of the time.
Figure 10 clearly indicates a tendency to choose the plural
verb form more often than the singular verb form. Only at
the age of eight did the subjects choose a singular form
more often than a plural form. There was also a tendency,
although not as consistent, for the subjects to choose plural
forms more often as they get older. Adults chose a plural
verb form 76 percent of the time for the ambiguous sentences.
Both of these perspectives indicate a tendency on the
part of the subjects to choose a plural form in the context
of the test sentences. There are at least two explanations
for this linguistic behavior. In some instances the subjects
Percent of Choices of Plural Verb Forms
in the Ambiguous Sentences
Age Plural Verb Forms
Note. All scores are rounded off to the nearest one-
half percentage point. Both sex and verb have been collapsed.
0 Plural Verb Forms
Figure 10. Percent of choices of plural verb forms
in the ambiguous sentences.
may be employing a strategy of choosing a plural verb form
to agree with the plural noun that precedes the verb, rather
than analyzing the structure of the nominal. The second
possible explanation is that the subjects chose a plural
verb form because they interpreted the nominal as a subject
nominal. The structure of a subject nominal is 'simpler'
than that of an object nominal. The problem of linguistic
simplicity is still unsolved; however, in this case there
are several reasons to believe that the meaning of a subject
nominal is more readily apparent than that of an object
First, the interpretation of the subject nominal does
not require any information which is not available on the
surface. Visiting relatives in the sense of 'relatives
that are visiting' has lost no content in the derivation
since that and are are not content words. Visiting rela-
tives in the sense of 'someone is visiting relatives' loses
the information carried by the subject of visiting. Although
that subject may be unspecified, it is still implied by the
structure. Also, the subject of visiting may be specified
by the context of the sentence or in the manner of 'Visiting
relatives is boring to Jane' in which Jane appears overtly
in the sentence as well as being the deleted subject of
visiting. Consequently, the listener must supply informa-
tion in order to interpret an object nominal, but all of the
information is available in a subject nominal.
Secondly, although the subject nominal is derived
from a deep-structure sentence, it resembles other adjective
+ noun combinations. The test sentences with subject
nominals as their subjects have the more common composition
of adjective + noun + verb + adjective, whereas the sentences
which have object nominals as their subjects have the com-
position of Iverb + objects + verb + adjective. Thus, the
listener must analyze the structure of an object nominal in
order to understand it, whereas it is possible to interpret
a subject nominal on the basis of its lexical content.
If, in fact, the interpretation of a nominal as a
subject nominal is more readily available to the subjects,
this would explain why subjects interpreted the nominals
as subject nominals more frequently than as object nominals,
and, therefore, why they chose plural verb forms more often.
Figures 9 and 10 seem to have some contradictory indi-
cations. As the subjects get older they choose a plural
verb form/interpret a nominal as a subject nominal more
often if the sentence is ambiguous. However, as subjects
get older they make fewer errors of choosing plural forms/
interpret object nominals as subject nominals less often
in the unambiguous sentences.
It may be that as they get older the subjects rely
heavily on the strategy in the ambiguous sentences, or
possibly, when there is no reason not to do so they inter-
pret the nominal as a subject nominal. And for the
unambiguous sentences the subjects' performances may improve
because, as they become older, the subjects acquire a better
understanding of the structure of the object nominal, as well
as its relationship with the verb. It may, in fact, be the
case that the subjects actually acquire Pro Deletion Trans-
formation--one transformation which is necessary for the
derivation of an object nominal--during the age span that
was tested in the experiment.
The numerous errors in subject/verb agreement make it
evident that, at least through the age of twelve, the sub-
jects do not fully understand the relationship between the
nominal type and the number of the verb. This, to some
degree, explains why many of the subjects in the first series
of experiments failed to attend to the verb is or are in
determining the meaning of the nominals in the test sentences
and, instead, perceived the sentences as ambiguous. The
only group of subjects who consistently demonstrated an
understanding of the relationship between the nominal and
the number of the verb was the adults. With the verb form
are, subjects never had many problems. And with looks, the
subjects began at about the age of ten to have fewer problems.
But for is and look, the age of understanding was somewhere
between twelve years and adults.
One final question which was considered was the role of
education in this linguistic behavior. As was mentioned
earlier, the difference between 'Moving houses is dangerous'
and 'Moving houses are dangerous' should be available to all
native speakers because it represents a difference in meaning.
However, not all subjects demonstrated a comprehension of
this difference. Could it be that this distinction is avail-
able only to those speakers who have learned the rules of
subject/verb agreement and for people such as linguists who
are highly conscious of the nuances of their language? Al-
though education was not a variable that was controlled in
these experiments, it is safe to say that education is not
a determining factor. Numerous studies have shown that
attempts to impose a rule onto a speaker's grammar before the
speaker reaches the normal stage of acquisition for that
rule confuses the speaker rather than speeding up the
acquisition process (McNeill 1966 and Braine 1971). It
is also true that the eight-year-old and older subjects in
the first five experiments were able to distinguish between
these two sentences if the sentences were presented together
for the subjects to compare. Apparently they have internalized
the information which is necessary to tell the difference
between the two sentence types if both are present. Possibly
they compare some features of the two sentences. But if
only one sentence is present so that there is nothing for
the subjects to compare that sentence with, they seem to
rely only on the lexical information in the surface structure.
This tactic leads them to comprehend two meanings in either
sentence. That is, they see the ambiguity which exists in
the nominal in sentences 4.a. and 4.b., but they do not rely
on the syntactic information of the verb to enforce a choice
between the two meanings. For this reason many of the sub-
jects in the first five experiments treated the unambiguous
4.a. Moving houses is dangerous.
4.b. Moving houses are dangerous.
test sentences as if they were ambiguous.
At some point in the acquisition process, speakers may
acquire an ability to compare sentence 4.a. with sentence
4.b. internally. That is, they may hear sentence 4.a. in
isolation and may be able to analyze it by generating sentence
4.b. and comparing the two. Possibly the subjects in these
experiments had not reached this point. They were able to
compare the two sentence types if both were presented by the
experimenter. But they were not able to generate the second
sentence internally on their own.
The problems which these subjects manifested under the
test conditions imply that the subjects would have difficulties
in normal communication. According to the results of these
experiments, on hearing the sentence 'Moving houses is
dangerous,' the subjects would comprehend two different
meanings and would be forced to guess which meaning the
speaker had intended. However, this would not, in fact, be
a problem. The context of a conversation would provide enough
information so that no choice would have to be made.
The conditions which occurred in the test situation and
which occur in the context of a conversation may differ in
that in a conversation the subjects do not need to compare
the potentially ambiguous sentence with any other sentence
because the meaning has been specified by the context. In
the test conditions if the experimenter offered no sentence
for comparison and if the subject had not acquired the ability
to generate the sentence for comparison, then the subject had
no method with which to disambiguate the nominal. A subject
at this stage would then perceive sentence 4.a. and sentence
4.b. as ambiguous.
There are two suggestions for any further investigation
or follow-up studies.
1. In experiment six the verb forms are and looks
always occurred in the second position for the subjects to
choose from (see Appendix D). It was also the case that the
subjects chose are and looks more often than they chose is
and look. It is possible that the subjects responded in this
manner because of a recency effect of selecting the verb form
which they encountered last. For this reason it is suggested
that, in any future study of this type, the order of the verb
forms not be the same throughout the test.
2. One of the test sentences in experiment six was
'Making cookies is/are looks/look fun.' Although the correct
verb forms for this sentence are is and looks, it is possible
that 'Making cookies look fun' is a grammatical phrase, as
in the sentence 'Mother always makes making cookies look fun.'
Although 'Making cookies look fun' is not grammatical as a
sentence, it and other phrases which may be interpreted as
grammatical parts of sentences could present some problems
in similar test designs.
This study indicates that a study of ambiguity or other
studies based in the theory of transformations must consider
not just a set of transformations but also the specific
lexical items--for example, the verb forms--which are in-
cluded in the study.
The first experiment in this study was based firmly in
the theory of Transformational Grammar, so much so that it
seemed almost uninteresting to attempt to prove the obvious.
It seemed unquestionable that sentence 5.a. is ambiguous
and that sentences 5.b. and 5.c. are unambiguous and represent
the meanings which are contained in sentence 5.a.
5.a. Flying kites can be exciting.
5.b. Flying kites is exciting.
5.c. Flying kites are exciting.
The question under consideration was at what stage speakers
would perceive the ambiguity of sentence 5.a. The behavior
of the subjects in these experiments contradicted the
theoretically-based assumptions. The subjects in Experiments
I through V indicated that sentences of the types of 5.a.,
5.b., and 5.c. were ambiguous. For these subjects it appeared
that the verb forms is and are were not functioning to dis-
ambiguate the nominal. This behavior was also unexpected.
Roger Brown (1973) has shown that by Stage V, which occurred
in the children he studied by the time they were 3-1/2 years
old, children have reached the criterion of 90 percent correct
usage of the copula and of the third-person singular morpheme
s. Thus, simple subject/verb agreement is acquired very
early by speakers of English.
Apparently, subject/verb agreement in the context of
sentences such as 5.b. and 5.c. was not acquired at such an
early stage. This was tested in experiment six where it was
shown that subjects as old as twelve years had problems with
subject/verb agreement, especially in the context of object
nominals as in sentence 5.b. These problems do not interfere
with communication simply because the context of a conversa-
tion serves to prevent any misinterpretation of the ambiguous
The inability on the part of these subjects to handle
subject/verb agreement competently explains why in the first
five experiments the subjects perceived the unambiguous test
sentences as ambiguous. Because they had not clearly acquired
the rules for subject/verb agreement in the context of an
ambiguous nominal, they were incapable of using that syntactic
information to disambiguate the nominal, and, consequently,
they perceived the sentences as ambiguous.
SENTENCES FOR EXPERIMENT I
1. The cat is lying on the table.
2. Riding horses can be fun.
3. The clown is holding his shoe.
4. Eating popcorn can be messy.
5. The woman is wearing a dress.
1. Playing grownups can be funny.
2. Playing grownups is funny.
3. Playing grownups are funny.
4. He wants to stop smoking.
5. He wants to quit smoking.
6. He wants to prevent smoking.
7. Shooting stars can be exciting.
8. Shooting stars is exciting.
9. Shooting stars are exciting.
10. Sailing boats can be dangerous.
11. Sailing boats is dangerous.
12. Sailing boats are dangerous.
13. Painting horses can be messy.
14. Painting horses is messy.
15. Painting horses are messy.