Citation
The acquisition of English adverbs

Material Information

Title:
The acquisition of English adverbs
Creator:
Paige, Joeann Chandlee, 1934-
Publication Date:
Language:
English
Physical Description:
vii, 86 leaves : ill. ; 28 cm.

Subjects

Subjects / Keywords:
Adverbs ( jstor )
Age groups ( jstor )
Cognitive development ( jstor )
Cognitive linguistics ( jstor )
Grade 1 ( jstor )
Linguistics ( jstor )
Semantics ( jstor )
Syntactics ( jstor )
Syntax ( jstor )
Verbs ( jstor )
Children -- Language ( lcsh )
Dissertations, Academic -- Linguistics -- UF ( lcsh )
English language -- Adverb ( lcsh )
Linguistics thesis Ph. D ( lcsh )
Genre:
bibliography ( marcgt )
non-fiction ( marcgt )

Notes

Thesis:
Thesis--University of Florida.
Bibliography:
Bibliography: leaves 83-85.
General Note:
Typescript.
General Note:
Vita.
Statement of Responsibility:
by Joeann Chandlee Paige.

Record Information

Source Institution:
University of Florida
Holding Location:
University of Florida
Rights Management:
The University of Florida George A. Smathers Libraries respect the intellectual property rights of others and do not claim any copyright interest in this item. This item may be protected by copyright but is made available here under a claim of fair use (17 U.S.C. §107) for non-profit research and educational purposes. Users of this work have responsibility for determining copyright status prior to reusing, publishing or reproducing this item for purposes other than what is allowed by fair use or other copyright exemptions. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder. The Smathers Libraries would like to learn more about this item and invite individuals or organizations to contact the RDS coordinator (ufdissertations@uflib.ufl.edu) with any additional information they can provide.
Resource Identifier:
022936604 ( ALEPH )
05320920 ( OCLC )

Downloads

This item has the following downloads:


Full Text


















THE ACQUISITION OF ENGLISH ADVERBS


BY

JOEANN CHANDLEE PAIGE















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


1978













ACKNOWLEDGMENTS


I am indebted to my chairman, Dr. Robert Scholes, and to the other members of my committee, Dr. Jean Casagrande, Dr. Paul Jensen, and Dr. Ira Fischler, for their advice and suggestions on all aspects of this project. I owe a special

debt of gratitude to Dr. Tom Dougherty for all the time and effort he devoted to the statistical analyses. Dr. Bill Sullivan has always found the time to discuss matters great and small, and has been a good friend as well. Ms. Hyta Mederer has also been a constant source of moral support, for which I am most grateful.

This study would not have been possible without the

enthusiastic cooperation of the teachers and administrative personnel at the P. K. Yonge Laboratory School. It was indeed a pleasure to work with a group of people who were so willing to assist me in my research, but who also ensured that the test sessions did not intrude on normal classroom activities.

Finally, I wish to thank Mrs. Anna Marie Martin, who undertook the final typing. She offered many suggestions on style and format that saved me much time and effort.


ii -















TABLE OF CONTENTS

Page

ACKNOWLEDGMENTS. . . . . . . . . . . . . . . . . . . . ii

ABSTRACT . . . . . . . . . . . . . . . . . . . . . . . v

CHAPTER I INTRODUCTION. . . . .. . . . . . . . . . . 1

General Comments on Adverbs . . . . . . . . . . . 1
General Aspects of ADV Acquisition. . . . . . . . 2
The Basic Function of ADV. . . . . . . . . . 5
Time, Tense, and Aspect. . . . . . . . . . . 6
The Syntax of ADV. . . . . . . . . . . . . . 11
The Present Study . . . . . . . . . . . . . . . . 19
Purpose. . . . . . . . . . . . . . . . . . . 20
Specifics of the Present Study . . . . . . . 20
The Questions. . . . . . . . . . . . . . . . 21
Notes . . . . . . . . . . . . . . . . . . . . . . 22

CHAPTER II THE STUDY . . . . . . . . . . . . . . . . 23

The Method. . . . . . . . . . . . . . . . . . . . 23
The Subjects. . . . . . . . . . . . . . . . . . . 24
The Test. . . . . . . . . . . . . . . . . . . . . 26
Procedure . . . . . . . . . . . . . . . . . . . . 29
Notes . . . . . . . . . . . . . . . . . . . . . . 32

CHAPTER III RESULTS . . . . . . . . . . . . . . . . . 33

Preface . . . . . . . . . . . . . . . . . . . . . 33
The Error Scores. . . . . . . . . . . . . . . . . 33
The Voluntary Data. . . .* .. . . . . . . .. .. . 47
The Relation between the Two Forms of Data. . . . 50
Summation of Results. . . . . . . . . . . . . . . 58
Note. . . . . . . . . . . . . . . . . . . . . . . 60

CHAPTER IV DISCUSSION. . . . . . . . . . . . . . . . 61

Syntax, Semantics, and the Significant Interaction 61 ADV Acquisition and Cognitive Development . . . . 69
Stages in ADV Acquisition . . . . . . . . . . . . 71
Stage 1: Unaware/ignoring ADV . . . . . . . 71
Stage 2: Partially aware/guessing at ADV. . 72


iii











Stage 3: Aware/asking about ADV . .
Stage 4: Unaware/comprehending ADV.
Notes . . . . . . . . . . . . . . . . . .

CHAPTER V SUMMARY AND CONCLUSIONS . . . . .

Review of Purpose and Questions . . . . .
General Conclusions . . . . . . . . . . .
Projections . . . . . . . . . . . . . . .

APPENDIX A Test Ss for Form 1 . . . . . . . . APPENDIX B Test Ss for Form 2 . . . . . . . . REFERENCES . . . . . . . . . . . . . . . . . . .

BIOGRAPHICAL SKETCH. . . . . . . . . . . . . . .


iv


Page

. . . . . 73
. 73 . 76

. . . . . 77 . . . . . 77
77
- . . . . 78

- . . . . 81 . . . . . 82

- - - - - 83

- . - - - 86














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 ACQUISITION OF ENGLISH ADVERBS By

Joeann Chandlee Paige

December 1978

Chairman: Robert J. Scholes
Major Department: Linguistics


The English adverbial system presents some complex problems for both the semanticist and the syntactician. There is some evidence that difficulty in linguistic description is correlated with time of acquisition. The purpose of this study was to investigate certain aspects of that proposition.

The particular syntactic and semantic dimensions of this study were very narrow. The syntactic question was: Is there a difference in time of acquisition of one-word adverbs as a function of syntactic form, i.e., morphemic items (e.g., almost) as compared to items derived by the 'basic rule' of adverb formation, i.e., ADJ+-ly=ADV (e.g., nearly). The semantic dimension covered five classes: manner, frequency, extent, relative time, and sentential. The overall question addressed in this study related to the possible relation


v








between adverb acquisition and general cognitive development.

The experimental procedure employed a newly devised

test which was a variation of the sentence-picture matching task. Since adverbs refer either to processes or whole propositions, the task was adapted to a comparison of two four-frame panels, based on familiar newspaper cartoon characters and format. The two panels differed only with respect to the applicability of the adverb in the test sentence. The subjects for the study were 75 students in Grades 1, 2, 3, and 5.

The results of the study indicated that adverb form, adverb class, and the interaction of form by class are all statistically significant, as is the other main effect of grade. There was a significant improvement between Grades 1 and 2, and between Grades 3 and 5. Because of the significant interaction between adverb class and form, only general implications for these two main effects could be discussed. Adverb form may be of some slight advantage in acquisition: -ly forms produced somewhat lower error scores overall. Semantic class is very probably a factor: sentential items were most difficult; manner items were least difficult. The middledifficulty items from the classes extent, frequency, and relative time are apparently the source of the significant interaction, and no definitive statement can be made on these items/classes from the results of this study. Other


vi









data from the study suggest a strong connection between ability to process adverbs and general level of cognitive development. Certain definable stages in adverb acquisition are proposed.


vii














CHAPTER I

INTRODUCTION


General Comments on Adverbs


The category 'adverb' is currently one of the most

challenging areas in linguistic theory. It is not hard to understand why, considering that this category covers such a diverse range of semantic concepts and syntactic structures. As Nilsen (1972) noted, in his extensive paper on the surface phenomena of English adverbs:

There seems to be a general consensus of
opinion among grammarians (no matter what
model they represent) that the most heterogeneous, and the least understood of the traditional part-of-speech categories is,
without question, the category Adverb. It
has traditionally been the miscellaneous
category into which those expressions were
placed which could not easily be handled elsewhere in the grammar. This was done
under the guise that these expressions
in some sense 'modified' the verb. (p. 179)

In the most recent major attempt to account for adverbs, Huang (1975) observed that the category adverb '. . . can be defined . . . as the part of speech satisfying the semantic definitions for adverbs but failing to exhibit the inflectional or distributional properties of nouns, verbs, and adjectives' (p. 9). Huang comes close to saying that an adverb is anything that isn't something else!


1




2



While the situation with adverbs (ADV) might not be quite that chaotic, the category does present a number of difficult grammatical problems which stem from the diversity of the semantic concepts and syntactic structures which it covers. My concern in this paper is with the question of the extent to which this 'problem area' within linguistic theory might also present problems for the language learner.

There is no strong evidence connecting difficulty in linguistic description with difficulty in acquisition, but some limited experimental evidence, plus simple intuition, suggests that such a connection exists. The wide range of semantic concepts under ADV seems to cover an equally wide range of cognitive difficulty. Similarly, it is possible, or even probable, that differences exist in the cognitive difficulty of the various syntactic structures of ADV. It would not be surprising, then, to find that there were patterns in acquisition for certain types and classes of ADV, and that these patterns were related to cognitive development.


General Aspects of ADV Acquisition

There is very little direct evidence on children and

ADVs. But from two widely disparate areas, we can say with some surety that ADV acquisition begins very early and that mastery is late.

The evidence on early acquisition comes from the data

Slobin (1971) reports on 'locative intentions' at the two-word




3


stage. That is, with appropriate environmental/circumstantial clues, an expression such as 'pot stove' is interpreted as 'the pot is on the stove'; 'doll drawer' means 'the doll is in the drawer.' Slobin also found the 'prolocatives' here and there at this early stage. Ervin-Tripp (1970) observed the locative where in the sense of both place and direction towards in very early language (at about age 3). This evidence may not be persuasive, but it is sufficient to suggest that there is some expression of the concept of ADV in very early language.

At the other end of the acquisition process, and

relevant to what goes on in between, there is the evidence from a series of studies on the language production of grade school children. These studies, related in intent, but not identical in method, were reported by Harrell (1957), Loban (1963, 1964), Hunt (1965), and O'Donnell, Griffin, and Norris (1967). The conclusions that can be drawn from these studies, relative to the point at issue here, is that there is a slow but steady increase in both the number and forms of ADV used by elementary school children with increase in grade. Apparently, the acquisition of ADV is an on-going process that continues all through the school years.

What I have said so far may seem a minimal defense of why anyone would want to look at ADV acquisition. One obvious reason is that no one has so far done so in any systematic way. In what follows, I would like to consider some more substantial reasons why the data on ADV acquisition





4


should be of more than casual interest. But first I will reiterate what I consider to be the basic questions. The first relates to the long period required for the mastery of ADV based on the data from the language production studies. Is there some inherent (cognitive) difficulty in the basic function of ADV as process modifier? Might it be that the young child is just 'not interested' in what ADV does to a sentence, and his 'need' to use ADV develops with his increasing awareness of, or concern with, the concepts that ADV conveys? Secondly, how does the semantic diversity of those various concepts covered by ADV relate to cognitive difficulty, and so to the long period of ADV acquisition? Finally, there is the question of the extent to which the complexity of the syntactic structures under ADV affects acquisition patterns. We have no firm evidence on any of these questions or issues, but in the next section, I will discuss these points as they relate to other data on linguistic and cognitive development, especially in the early school years, the age-range of interest in the study to be reported here.

The issues and items that I will consider here are for the most part at the most elemental levels of both semantics and syntax. Specifically, I will discuss the basic function of ADV; one of the most elemental of the semantic functions of ADV: time; and two aspects of elementary syntax: the ADV= ADJ+-ly rule and certain placement rules.









The Basic Function of ADV

An ADV is a 'process modifier': it sets the time frame, establishes/describes the manner, extent, etc. of a given string of words. By this definition, we really shouldn't expect very young children to be much concerned with ADV. Bronckart and Sinclair (1973), among others, have shown that young children are much more concerned with 'outcome' than with 'process,' and so we can at least assume that, until the child has reached the 'process-awareness' point in his cognitive development, ADV will play a very small role in his linguistic behavior.

There is a further aspect of this notion of 'processmodifier' that seems to relate to a well-accepted idea about cognitive development, and that is the observation that children go from concern with the concrete to an awareness of the abstract. This relates to the common conception in linguistic theory that the category ADV can be divided into 'VP-modifier' and 'proposition-modifier' (or S-ADV). While I recognize that there is considerable debate over just what is 'VP-modifier' or S-ADV, I am sure there is general agreement that such a distinction exists. The point of interest here is that 'VP-modifier' is more concrete than is S-ADV, and, therefore, should be acquired earlier. In defense of the greater complexity of S-ADV concepts, I would like to mention a distinction made by Schreiber (1970). S claims that there are two types of S-ADV, which he calls 'modal' and 'evaluative.' With 'modal' S-ADVs, the speaker is




6


expressing his 'degree of surety'; with 'evaluative'S-ADVs, he is expressing an opinion. If this interpretation is correct, and I feel that it comes very close, the child will not be able to correctly interpret S-ADVs until: 1) he reaches the point of recognizing that a comment is being made about the entire proposition; 2) in addition, he understands that this comment represents either an opinion, or just how sure the speaker is of what he is saying. I will have more to say about Schreiber's analysis in my discussion of the results of the present study. Time, Tense, and Aspect

In order to assess the relative importance of semantics in the acquisition of ADV, it seems logical to look first at TIME-ADV, for the following reasons. 1) A significant subclass of ADVs relate to time; 2) time concepts are an important part of cognitive development: and 3) there are more data on this area of linguistic development than almost any other, certainly any that relate to ADV, in the age range of interest here.

A number of investigators, notably Piaget (1971),

Barrie-Blackley (1973), Feagans (1974), and Keller-Cohen (1975), have described the child's developing time sense as it relates to general cognitive development. The very young child lives in a 'now/not-now' world. As Ervin-Tripp (1970) pointed out, before age 5, when is the hardest of the WHquestions for children to answer. She reports the following E-child exchange as an example: 'When did he eat?' -- 'Meat.'




7


Young children characteristically give concrete answers to abstract questions. They are concerned only with the 'hereand-now,' and so it seems a logical assumption that patterns in TIME-ADV acquisition would reflect this.

For the English speaking child, a sense of relative time, and, presumably, a sense or awareness of tense, apparently develops out of spatial concepts. Feagans (1974) studied the relation between temporal and spatial concepts, based on a theory first put forth by H. Clark (1970). Her conclusions were that items such as before and after were comprehended in the spatial sense before they had any meaning to the child in a temporal sense. These and other temporal connectives have been studied by Amidon and Carey (1972), Barrie-Blackley (1973), and Keller-Cohen (1975), with children in the 3-7 age range. The major conclusion of these studies, relative to the present study, is that young children do not fully understand temporal connectives. The implication is that relative time, and the associated ADVs, is in many cases incompletely understood by children in the early elementary grades.

Another significant factor in the child's comprehension of relative time is the interaction of ADV with tense and aspect. First, consider what Chafe (1973) discusses as what might be called the 'necessary TIME-ADV.' Chafe gives an amusing example, which I will loosely paraphrase. 'Dennis the Menace' announces to his parents: 'Mr. Wilson broke his arm!' The parents' immediate reaction is to rush to help




8


their neighbor--until Dennis' father asks 'When?' And Dennis says, 'When he was a little boy.' Chafe's point is that, to the child, if Dennis just learned about the incident, even though it happened long ago, it was 'new information' to him, and so that is how he stated it. The adult interpretation of such a statement is to regard it as 'in the present/ immediate past' unless some TIME-ADV locates it elsewhere. Still, both parents and child were dealing with 'past tense.' Nilsen (1972) notes that TIME-ADVs can express a 'time set' from distant past through distant future, with many interim points, and that specific items interact with tense and aspect. Obviously, the child has to become aware of the need, not only to 'locate' a statement in time (tense), but to learn just how exact that location need be, and just which ADVs can apply, given the interaction of tense and aspect.

In order to look at some practical examples of how tense, aspect, and ADV interact, I would like to consider the following sentences.

la. Mary has had mumps.

lb. Mary has had mumps before.

2a. Mary has been pregnant.

2b. Mary has been pregnant before.

3a. John is baking great pies.

3b. John is still baking great pies.

4a. Jack Nicklaus has won the Masters.

4b. Jack Nicklaus has already won the Masters.




9


In all of these S-pairs, the 'a' Ss make simple statements about past or present events. In the 'b' Ss, the ADV contributes additional implicit information, and introduces a dual time reference as well. For example, lb. implies that, if Mary is now running a fever, she could not have mumps: that disease, once contracted, provides immunity. On the other hand 2b. implies that Mary has once again 'contracted'

that particular condition. S-3a. merely makes a statement about John's present activities, but 3b. presupposes a common knowledge between speaker and hearer about John's

activities/abilities at some prior time. S-4b. can have more than one interpretation in terms of 'new information' and 'dual time reference.' 1) Jack won the trophy at an earlier time, and if he wins it this time, 'What's one more green blazer?' 2) Since Jack has achieved this pinnacle of success, what more could matter? 3) With only one hole to go, how could anyone catch up with him?

It is interesting to note that those ADVs which introduce a dual time reference can also shift emphasis from past time to present, as in lb. and 2b. In 3b., on the other hand, the shared knowledge of the past is stressed. The situation with the multiple interpretations of already is somewhat more complicated. I will discuss this ADV-item further in Chapter IV.

The major point to be made here is that words like

still, before, and already introduce a double time reference. Just when is the child ready to handle this aspect of




10


'relative time'? There is very little experimental evidence on this question, but 'durative aspect' at least seems to be a problem for the young child. Feagans (1974) found that children under 5 do not understand the durative aspect, and that even 7-year-olds had trouble with the durative notions of since and until. I think the evidence from the work of Bronckart and Sinclair (1973) is relevant here, even though they were working with French-speaking children. They found that children (under 9) tend not to use tense to locate an action, state, etc. in time but rather to connote aspect. That is, the children associated past tense with completion and present tense with duration, regardless of the absolute time frame. They also found a tendency for children below

6 to use intonation and gesture to refer to 'aspectual distinctions'; above 6, children tended to use adverbial phrases for this purpose.

I could find no specific data on the items before and

already, or on any other items in the class of relative time concepts, in the sense these notions are being discussed here. But I would strongly suspect that the child must develop well beyond the 'now/not-now' stage before he is able to comprehend items that refer to time relations between the present and the past.

One final element in the interaction of tense/aspect/ adverb deserves some comment, and that is the effect of ADV in potentially generic sentences. Smith (1964) showed that the presence of TIME-ADV in a sentence can determine whether




11


the sentence is generic, specific, or ambiguous. (Nilsen [1972] should also be credited with elaboration of this idea.) The sentences that follow are examples of the notions that Smith and Nilsen were considering.

5a. The tiger stalked the jungle.

5b. Long ago, the tiger stalked the jungle.

5c. The tiger stalked the jungle at noon on Thursday.

6a. The eagle ate small prey.

6b. The eagle ate small prey in times of famine.

6c. The eagle ate small prey at Bilbo's birthday party. For the-'a' sentences, the reference is ambiguous relative to 'generic.' But for the 'b' sentences, the TIME-ADV allows only an unambiguous generic interpretation. Similarly, for the 'c' sentences, there is an unambiguous non-generic interpretation. Until the child is able to comprehend ADV, or at least becomes aware that it can convey a significant part of the information in a statement, he will see no difference in the 'specific' versus the 'general' in such statements.


The Syntax of ADV

The points I have been discussing so far have dealt

primarily with how the semantics of ADV might present problems for the language learner. But as Slobin (1971) has noted, '. . . although one can talk about order of acquisition in terms of semantic or cognitive complexity, there is clearly a point at which formal linguistic complexity also plays a role' (p. 308). The research of C. Chomsky




12


(1969) and Kessel (1970) supports Slobin's notion. Both studies found that, for certain structures, there is a relation between formal complexity and time of acquisition. In what follows, I will argue that, even at the most elemental levels of syntax, ADV presents a major problem for formal description, but more importantly from the present perspective, that it might also be a syntactic problem for the language learner.

It would seem that the place to start any discussion of the elementary syntax of the English ADV would be with the so-called 'basic rule': ADV=ADJ+-ly. This rule certainly seems simple enough, but there are strange and inconsistent exceptions.

The rule is curious in this respect: it apparently applies, almost without exception, to all ADVs which are, in any sense, derived. It is the morphemic items to which the exceptions apply, and in no obvious fashion or systematic way. The significant point here is that this 'basic rule' applies most generally to those items that are least likely to be heard by young children: to the derived ADVs which are, for the most part, what would be considered 'advanced vocabulary.' On the other hand, those very items with which the child is likely to come into most frequent contact are the exceptions to the rule in question. What follows is a brief examination of some of the exceptions to the 'rule' which do seem to follow some sort of pattern, but not one which the child is likely to find very helpful.




13


There are certain sets of ADJ which also function as

MAN-ADVs, with no morphological change, i.e., without '+-ly.' But for no discernible reason, these items fall into 3 classes. Consider the words in Table 1.2 TABLE 1

Adjectives Which Function as Manner Adverbs Without '+-ly'

A B C


ill bright deep
left fair (unbiased) fair
right (direction) loud fast
right ('directly') quick hard
well right ('correct') high
still sharp strong
straight soft wide



For items in A, no '-ly' is possible:

7. Michael pleaded the case well (*-ly).

8. He turned the car left (*-ly).

The items in B may take '-ly' optionally:

9. The sun shone bright (-ly).

10. The wind blew soft (-ly).

There are, however, some restrictions on the items in this 'optional' category with respect to freedom of placement.

11. The sun brightly shone.

*12. The sun bright shone.

13. The boy quickly ran down the street.

*14. The boy quick ran down the street.

Apparently the optionality of '-ly' for these items applies only when they are in post-V position. (I will discuss




14


placement rules more fully in a later section.) All items in C can take '-ly,' but this results in a change in class. For the most part, MAN goes to EXT.

Just for this small set of relatively common words, which must surely become part of the child's vocabulary rather early, there is a bewildering array of exceptions to a very basic rule. There seems to be no pattern to these exceptions, and so it is an interesting experimental question as to what order and at what age they are mastered.

Another peculiar phenomenon, similar to that just discussed, involves verb-adverb phrases. Nilsen (1972) comments that:

Sometimes there is a rather tight affinity
between the Manner Adverbials . . . and a
verb, making it so that the V-ADVm construction is functioning as a sort of a
Predicate plus Complement. (p. 81)

He gives an extensive list of such pairs, but makes no mention of how they might be subdivided as a function of the applicability of the '-ly' rule.3 In Table 2 I have listed some of the pairs from Nilsen's list which seem to function differently with regard to the rule. The items in A cannot take '-ly'; those in B can, and still remain post-V. For the items in C and D, '-ly' is possible, but with a change in meaning. For D, the '-ly' forms can remain post-V; for C, they cannot. These pairs are all, as Nilsen noted, in some sense 'locked phrases,' even 'idiomatic' to some extent (or rather, in some cases). But it is interesting to note how differently '+-ly' applies to the different subclasses.





15


TABLE 2

Verb-Adverb Phrases Which Function
as Predicate Plus Complement


A B C D


blow shut shine clear make clear work loose
sit tight stand firm make happy prove false
turn red rest easy remain loyal prove true
make fast run slow remain ignorant break even
grow tall glow bright set free continue warm
keep young lie flat cut short come clean
fall sick blow open



It is beyond the scope of this paper to try to explain these phenomenona. I mention it only as further evidence of just how difficult it must be to master the English ADV.

The one final comment I have to make on the vagaries of the '-ly' rule regards a small set of ADJs ending in 'y.' They are presented in Table 3, column A, and come from a list in Nilsen (1972, p. 80) labeled 'No Adverb Possible.' I agree with Nilsen that the items in A cannot take '+-ly,' at least, they sound very strange. And yet the items in B are also ADJ ending in 'y,' and the associated ADVs sound normal. I can find no semantic, syntactic, or phonological reason why, for example, cloudily should sound odd, but handily is common and accepted.





16


TABLE 3

Adjectives Ending in 'y' Which Can
and Cannot Take '+-ly'


A B


baggy angry
cloudy busy
dirty clumsy
dusty handy
funny happy
ugly heavy
noisy



The second area of elemental syntax that I would like to consider is the rules of ADV placement. In general, ADV placement in English is very free, as can be seen from Figure 1 on page 17. At the least, most items from most ADV classes can appear anywhere in the S. The apparent exception--the post-V position--is sometimes acceptable (for certain items) when the VP is sufficiently long, as in S-15.

15. John reads thoroughly those books which are
related to his profession.

But there are definite restrictions on ADV placement relating to morphology, ADV class, and, to a certain extent, the problem of ambiguity.

The morphological restrictions on placement relate to the items discussed in the previous section, that is, those MAN-ADVs which are of the same phonetic form as their ADJ counterparts. For example, with the items soft and bright, both forms are possible in one sentence position, but not in the other.

















Surely Quietly Often

Usually


John


surely read * the newspaper ,surely.
quietly * quietly.
often * often.
fully * fully.
usually *usually.


Fig. 1. Possible placements for English adverbs.


S-ADV MAN
FREQ EXT TIME


-J





18


16.

17.

18.

*19.

20.

21.

22.

*23.

But with

24.

*25.

*26.

*27. S-27 is,


Sam sang the song soft. Sam sang the song softly. Sam softly sang the song. Sam soft sang the song. The lamp glowed bright. The lamp glowed brightly. The lamp brightly glowed. The lamp bright glowed. items like hard, there is a different 'rule': Sam hit the nail hard. Sam hard hit the nail. Sam hit the nail hardly. Sam hardly hit the nail. of course, acceptable, but only with a change in


ADV class: from MAN to EXT.

Placement rules and restrictions which relate to ADV class and the problem of ambiguity are syntactically very interesting. Consider the following Ss:

28. I naturally read the letter. S-ADV

29. I read the letter naturally. MAN

30. I read the letter, naturally. S-ADV

31. I discussed the subject frankly. MAN

32. Frankly, I discussed the subject. S-ADV

33. I frankly discussed the subject. Ambig.: MAN/S-ADV

34. John cooks rice regularly. Ambig.: FREQ/MAN

35. John regularly cooks rice. FREQ.

There seem to be no logical reasons for the placement





19


restrictions which apply to ADV as a function of class, or to those which disambiguate. It would seem that the child simply has to memorize all such exceptions. But Nilsen has contributed what may be a valuable insight on this particular 'problem' for the language learner. He notes that there are, for certain ADV classes, positions within the S which are 'more natural.' And so, by his analysis, S-final is 'more natural' for MAN-ADV, and 'pre-V' is 'more natural' for FREQ-ADV. The trouble is that, at least for the Ss I have chosen as examples of 'the problem,' his solution doesn't seem to help very much. In comparing S-33 with S-31, it is true that it is possible to disambiguate, but S-33 is a perfectly good way of expressing the notion of S-31. With the other ambiguous S, S-34, compare it with S-36, which is perfectly acceptable.

36. John cooks rice frequently.

The point is that, what Nilsen suggests as to 'natural placement' may well be an idea worth pursuing in the area of ADV acquisition. However at this point it does not seem likely that it will be a 'hard and fast' rule. There are few such rules in the area of language.


The Present Study


In the previous section I have shown some of the

syntactic and semantic reasons why ADV is a difficult category for the English-speaking child to master. There is evidence that certain items in this category appear in very early




20


language, but that mastery of the category is very late. In sum, ADV is a complex linguistic category, for the child and for the linguist, and there is good reason to believe that the child's acquisition of the category ADV is closely related to cognitive development.


Purpose

The purpose of the present study is to make a systematic investigation of what seem to be the basic aspects of ADV acquisition. The most fundamental questions are, of course, those put forth at the beginning of this chapter:

1. To what extent does the basic function of ADV interact with cognitive development?

2. What is the relation between ADV class and cognitive difficulty?

3. What is the relation between differences in syntactic structures of ADV and cognitive difficulty? Specifics of the Present Study

Since this study is the first attempt at evaluating ADV acquisition in a systematic way, certain decisions had to be made as to how best to approach the questions presented above. It seemed that the most fundamental comparison of syntactic structures would be between one-word items that differed only in terms of 'morphemic' or 'ADJ+-ly.' Semantically, the decisions were somewhat more complex, but 5 classes were chosen for the following reasons:

MAN--There is some scant evidence that items in this




21


class appear early. There are also logical

reasons why MAN-ADVs seem to be more closely

associated with their ADJ counterparts. They

are, as a class, associated with the (presumably)

simpler concept of VP-modifier.

REL. TIME/FREQ--These items relate to the one area in

which there are some data on children and ADV,

indicating that TIME is a significant factor in

ADV acquisition in the early school years.

EXT--There is evidence that this concept grows out

of earlier concepts of TIME and SPACE, and so it

is potentially interesting in this period of

language development.

S-ADV--There is good reason to believe this class covers

the most complex of ADV concepts, and so might be

one of the last to be mastered. The Questions

Given the above criteria, the particular questions addressed in this study were:

1. Are there differences in time of acquisition as a

function of the two syntactic forms: 'morphemic' and 'ADJ+-ly'?

2. Are there differences in time of acquisition as a

function of the five semantic classes: MAN, EXT, REL. TIME, FREQ, and S-ADV?

3. Is there evidence for a relation between linguistic and cognitive development in relation to ADV acquisition?




22


Notes


The terms 'adverb' and 'adverbial' are not used consistently in the literature. Both terms can cover both items and phrases. Some writers make a distinction between adverbial phrases and one-word adverbs. In this paper the more traditional 'adverb' (ADV) will be used to refer to all forms, with the item/phrase distinction made clear where pertinent. In the study proper, only one-word ADVs are considered.
2The tables in this chapter are not inclusive. That is, I have not attempted to find all possible items which might fit any particular cell.
3Nilsen (1972, p. 81) acknowledges that this list and the observations on this relationship should be credited to Fries (1952) and Fraser (1965).














CHAPTER II

THE STUDY


The Method


This study deals with an area that is both broad and ill-defined. Even though the restrictions and specific questions stated in the previous chapter impose some limitations, certain basic decisions remain as to the best way to approach problems in this area. The first and most important is whether to look at language from the perspective of production or of comprehension. For this particular question, a brief examination of the available lists on child language, in particular, Wepman and Hass (1969) and Rinsland (1945), showed that the percentage of ADVs in child language output is so small that a production study was impractical. Moreover, a study of comprehension seemed to have two further advantages. It is somehow more realistic and it is easier to control. Comprehension data are more 'realistic' than production data in this age range in the sense that what we are concerned with here is just when a child understands the effect of ADV in a S. And a comprehension study is easier to control because the child's attention can be directed to just those linguistic situations in which ADV is critical.1


23




24


The remaining decision, then, was how best to measure

comprehension of ADV as a 'process modifier.' It seemed that the logical choice was a task that comes closest to everyday language, and, by this criterion, the picture-sentence matching task was most appropriate. This task is very similar to the 'real-life' situation, in which the child must respond to what he hears, in the way he currently (at his present acquisition level) understands it.


The Subjects


The subjects for this study were students at the P. K. Yonge Laboratory School at the University of Florida. The final analysis was made on the data from 15 first graders and 20 children in each of grades 2, 3, and 5. Eight eighth graders were also tested, since this age group could be considered 'linguistic adults.' Their data represent a 'test of the test,' that is, a demonstration that, for adults, who presumably have full command of the English ADV system, this test is a simple linguistic exercise.

There were some children tested whose data were not included in the analysis. For example, all the children whose data were included were white. A few black children were tested for social reasons, but since there is no evidence either way on the effect of the Southern black dialects on performance in language acquisition studies, it did not seem appropriate to include them in this study.2




25


There were several children I tried to test, but who, for one reason or another, were unwilling or unable to cooperate. This was clear by the end of the 6 practice items, and for them, the session ended there. This group includes a few kindergarteners: it seems that this type of test is inappropriate below age 6, and I suspect the major problem is attention span.

Finally, the male/female division is not equal for all grades. The breakdown of sex by grade is presented in Table

4.

TABLE 4

Subjects by Grade and Sex


Grade Male Female


1 6 9
2 10 10
3 10 10
5 10 10
8 3 5



The reason that there are fewer males than female subjects in Grade 1 is that the girls were generally more cooperative, and the boys more likely to be intimidated by the test situation at this age. For Grade 8, there did not seem to by any good reason to assume male/female differences would be relevant, and so no attempt was made to balance for sex in that age group.




26


The Test


As was mentioned earlier, it seemed that the best method for evaluating the present problem was a sentence-picture matching task, especially for the age range of interest here. But with the added complication of assessing 'process modifier,' some elaboration of the usual method was necessary. So instead of contrasting two or more single-frame pictures against a S, two panels of 4 frames, in the manner of newspaper cartoons, were constructed for each test S. An additional advantage of this approach was that it allowed me to use familiar cartoon characters which would be more likely to catch the childrens' interest.

Given the restrictions and the particular questions stated in Chapter I, specific items for the 5 ADV classes and 2 syntactic forms had to be selected which were arguably items a young child would hear relatively often, or could be expected to use relatively early, by some objective criterion. The choice of items was made after consulting the lists formulated by Jones and Wepman (1966), Wepman and Hass (1969), and Rinsland (1945). But all such lists are based on production, and more importantly, Wepman's lists are restricted to responses to the TAT, where the possibility of ADV occurring is limited.3 The problem with Rinsland's lists is that he makes no distinction between syntactic categories. So the final choice of items for this test was based on what these lists suggested, plus a large degree of intuition.




27


. Once the 24 ADV-items had been chosen, a S was constructed for each in a standard form: NP-ADV-VP. The exception to this format was that for morphemic-MAN items the ADV had to appear at the end of the S, as was discussed in Chapter I. Also, since S-ADVs seem most natural in S-initial position, that is where they occur in the test Ss here. The ADV items by form and class, and their position in the test Ss, are presented in Table 5. The list of test Ss can be found in Appendices A and B.

The basic form of the test was a page containing the two 4-frame cartoon panels (which will be referred to as 'panel-sets'), with the panels placed one above the other. The panels were prominently labeled A (above) and B. Each page corresponded to one of the test Ss. The sequences in the panels were variations of familiar cartoon characters, and depicted the events, actions, or situation in the corresponding S. The sole difference between the panels on a given page was the applicability of the ADV in the corresponding S. Thirty panel-sets/pages were constructed, one for each of the ADV test items, plus 6 which were used for a practice session. For the 6 practice pages, the panels differed with respect to the appropriate subject, verb, or ADJ, but never ADV. The purpose of these pages was to make certain that, before beginning the test proper, the child fully understood the task, i.e., of selecting which of the two panels best fit the S he heard.




28


TABLE 5

Adverb Items by Class and Form with Sentence Frames


Form Sentence Frames
Class ADJ+ly Morphemic


MAN carefully hard ADJ+ly:
quietly high Subj-ADV-Verb'ed-Obj
quickly fast Morphemic:
happily straight Subj-Verb'ed-Obj-ADV
EXT nearly almost Subj-ADV-Verb'ed-Obj
completely only
REL. finally already Subj-ADV-Verb'ed-Obj
TIME suddenly still
FREQ rarely seldom Subj-ADV-Verb'ed-Obj
usually often
S-ADV certainly perhaps ADV-Subj-Verb'ed-Obj
probably maybe



The Ss were presented in semi-random order. That is, they were arranged so that the same subject or verb never appeared in adjacent Ss, and the class and form of the ADV occurred in reasonably random order. The 'right answer,' i.e., the A or B panel, was truly random. Since there was some question that particular lexical items might interact with the panel-sets with which they were associated, a second form of the test was arranged by interchanging antonyms and synonyms. For example, the panel-set for rarely on Form 1 was used for usually on Form 2, since the sequence depicted in that panel-set was equally appropriate to either ADV, but the 'right' answer would be the opposite panel. This interchange was possible for only 14 of the 24 items, but that seemed sufficient to be able to answer the question of




29


possible interaction. The Ss for both forms of the test are presented in Appendices A and B. The panel sets were presented in the same order for all subjects; the ordering of the Ss was changed accordingly for the two forms.


Procedure


One of the special advantages of conducting research at the P. K. Yonge School is that each classroom has an adjacent teacher's office that is available for testing sessions. The rooms have child-size tables and chairs, and so for each session, the child and I could sit side by side, with the notebook of panel-sets before us.

At the beginning of the session, I opened the notebook to a page on which were drawn all of the cartoon characters which would appear in the panel sets. Since these were familiar characters, I asked the child if he could name them. Most children above first grade could, and seemed to enjoy doing so. Even the first graders knew at least some of the 'people.' It should be stressed, though, that this 'I.D.' page was intended only as a means of catching the child's interest. In no case in the test that followed was it critical that the child know the name of any particular character.

After discussing the characters, I told the child I

would be showing him some pictures with the 'people' in them and would be asking him questions about the pictures. I turned to the first practice page and explained that all




30


of the pages would be the same, with one panel above marked A and one below marked B. We then went through the 6 practice pages. For each page, I said, 'If I said (relevant S), which panel do you think would go with that, A or B?' No matter what his answer, I told him which was the right panel, and we discussed why. If at the end of the 6 practice pages, the child did not obviously understand the task, giving the right answer and correctly explaining why, that child was not included in the study. That is, in no case were data included if there was any question as to whether the child understood the task. At the end of the practice session, I told the child that the rest of the pages would be like the ones we had just seen, but I would no longer tell him the right answer, ' . . . because I want to know what you think is the right answer.' In all cases, no matter whether the answer was right or wrong, I responded with a 'noncommittal,' but 'positive-sounding' 'OK, let's look at the next page.'

In addition to recording which panel the child selected, I also noted any voluntary comments that were made, and, occasionally, but in no systematic manner, asked, 'Why do you think that is the right panel?' In many cases, the children offered comments or explanations without being asked.

The eighth graders were an exception to the procedure just described. For every item, they were asked to explain why they chose as they did, since, as I mentioned earlier,




31



their data were to be a measure of the validity of the test.

The overall response of the children to the test was very encouraging. It was not uncommon for a child to ask when the test was over, 'Is that all? Aren't there any more?' To the extent that 'subject involvement' relates to test reliability, this evidence reflects favorably on the results that I will report in the next chapter.





32


Notes


The question of whether comprehension precedes production is not at issue here. In fact, it may well be a nonissue. Early production of a linguistic form does not indicate full comprehension, nor does the reverse hold. For this paper, what is important is determining whether there are progressive stages, or patterns, in ADV acquisition, and what applies to comprehension should, in a general way, apply equally to production. But that is still an empirical question, and will be discussed further in Chapter IV.
2An examination of the data from those few black children indicated that they tended to score well below their white peers. I have no explanation for this, but there were strong indications that dialect differences were a factor. This, I feel, supports my early decision not to include these children in the study/analysis.
3The Thematic Apperception Test (TAT) (Murray, 1943) is a series of pictures, each depicting a specific situation, which is widely used to elicit language samples for various purposes. But since the pictures are essentially 'static,' there are few cues which could be expected to elicit ADVs.














CHAPTER III

RESULTS


Preface


The observations from this study fall into two categories: 1) the objective measures, i.e., the error scores for each test item, and 2) the voluntary comments and other responses the children made in relation to the individual test items. Only the error scores were analyzed statistically, and the results of those analyses will be considered first. The children's comments and other responses, and the relevance of these data to the error data,will be discussed later.


The Error Scores


The objective results of the study are presented in

Table 6. The mean error scores represent the proportion of trials in which an incorrect response was given. The scores are listed for each grade and each test item, with an indication of the divisions by ADV class and form. Table 7 gives the means for ADV class and form, across individual items. These are the data on which the final analyses were run.

An analysis of variance (ANOVA) was conducted to determine the effects of the several factors in this study. Since


33





34


TABLE 6

X Error Scores by Test Item by Grade


Grade
Adverb 1 2 3 5
Class/Form Test Item N=15 N=20 N=20 N=20


Manner

-ly


morph.



Extent
-ly
morph.


Rel. Time
-ly
morph.

Frequency
-ly
morph.

Sentential
-ly
morph.


carefully
quietly quickly happily hard high fast straight
nearly completely
almost only finally suddenly
already still rarely usually seldom often
certainly probably perhaps maybe


0.27 0.07 0.00 0.33 0.00
0.00 0.13 0.07
0.73 0.13 0.53 0.53
0.47 0.60
0.20 0.20 0.73 0.33 0.87 0.33
0.40 0.80 0.93 0.80


0.20
0.00 0.15
0.20 0.05 0.00 0.00 0.10
0.20 0.15 0.30
0.40 0.55
0.40 0.35 0.15 0.35 0.10 0.90
0.20 0.20 0.70 0.85 0.90


0.10 0.05 0.00 0.15 0.05 0.00 0.10 0.00 0.25
0.20 0.20 0.45 0.45 0.45 0.10 0.15
0.40 0.05 0.80 0.15
0.20 0.80 0.70 0.80


0.00 0.00 0.05 0.10 0.00 0.00 0.00 0.05 0.05
0.00 0.00 0.25
0.20 0.25 0.10 0.00 0.15 0.00
0.45 0.10 0.15 0.80 0.65 0.55





35


TABLE 7 X Error Scores for Adverb Form and Class by Grade Adverb Class


Grade Adv. Form


Man. Ext.


Rel. Time


Freq. Sen.


1 -ly
morph.
x
2 -ly
morph.
X

3 -ly
morph.


5 -ly
morph.
x

X -ly
X-morph.
XX


0.17 0.43 0.53 0.05 0.53 0.20 0.11 0.48 0.36 0.14 0.18 0.48 0.04 0.35 0.25 0.09 0.26 0.36 0.08 0.22 0.45 0.04 0.32 0.12 0.06 0.27 0.28 0.04 0.02 0.22 0.01 0.12 0.05 0.02 0.07 0.14 0.10 0.20 0.41 0.03 0.32 0.15 0.07 0.26 0.28


0.53 0.60 0.56
0.22 0.55 0.38
0.22 0.48 0.35
0.08 0.28 0.18
0.25
0.47 0.36


0.45 0.45 0.45 0.29
0.41 0.35 0.29
0.34 0.32
0.17
0.21 0.19


0.60 0.87
0.74 0.45 0.88 0.66 0.50 0.75 0.62
0.48 0.60
0.54 0.50 0.77 0.63




36


the design was unbalanced, i.e., the N for Grade 1 was less than that for the other three grades, the ANOVA was conducted by means of a general linear model: a regression analysis. In the course of this analysis, the main effects (test form, sex, grade, ADV form, and ADV class) and all possible combinations of 2-, 3-, and 4-way interactions were tested. Any main effect or interaction that was not significant was dropped from the model. Those terms that appear to be significant (a = 0.01) are reported in Table 8. As a result, the ANOVA indicated that the factors of grade, ADV form, and ADV class, as well as the interaction of form by class, are the only variables that are significantly related to the responses to the test items. But the ANOVA results say only that some means are significantly different for any variable. Therefore the Duncan Multiple Range Test was used to identify just where the real, or statistically significant differences lay. This test was run to ascertain which grades were actually different from each other, and where the differences were in the ADV form by ADV class interaction. The results of these tests, along with the graphic representations, are shown in Figures 2 and 3, respectively. It seems clear that the ability to process the ADVs included in this study improves significantly between Grades 1 and 2 and between Grades 3 and 5. The significant differences for ADV form and ADV class are complicated by the form by class interaction, and that subject requires some clarification.





37


TABLE 8

ANOVA for Grade, Adverb Form and Adverb Class


Source S.S. df m.s. F


Grade 6.19 3 2.06 29.33*
ADV Form 0.59 1 0.59 8.36*
ADV Class 25.30 4 6.32 89.87*
Form X Class 7.14 4 1.78 25.35*
Error 51.86 737 0.07 --*significant at 0.01


Although the statistical results/ANOVA indicate that both ADV form and ADV class are significant factors in the acquisition of the particular items included in this study, the significant interaction of form by class means that the differences in these main effects can only be considered as trends, and that any real differences for these two variables are masked. That is why the Duncan Test results take on special importance.

In Table 9, the individual test items are ranked by

mean score and Duncan grouping. This makes it possible to compare means, both for items within cells and for items across cells. Thus the table points up the fact that differences between means for individual items may be an important factor in the interpretation of the significant interaction of ADV form by ADV class.

Before considering the implications of that significant interaction, we can look at what seem to be the most obvious conclusions to be drawn from this table. Clearly S-ADV items





38


10 O-=-


Duncan Multiple
Range Testa


Grade
1
2
3 5


x
0.45 0.35 0.32 0.19


(4)


Group
A
B B
C


5


aR's which do not share a common grouping are significantly different; are not (- = 0.01)


letter in the those which do


Fig. 2. R errors by grade with associated Duncan
Multiple Range Test results.


90-i-


7 0...


6 0-4


0
C)

U Q)


50--


304-


20-4-.


Grade: I


2
2


3


a6 0


4 0- -





39


Duncan Multiple Range Test a


F


0 -t-


ADV
Class S-ADV S-ADV
FREQ REL. TIME
EXT FREQ EXP


ADV
Form

morph.
-ly
morph.
-ly
morph.
-ly
-1


X Grouping


0.77 0.50
0.47 0.41 0.32 0.25 n 2)


A
B
B BC CD DE


y .r.
EL. TIME morph. 0.15 EF
IAN -ly 0.10 FG
IAN morph. 0.03 G
















0e






ADV Form:
-ly


* = morph.


ADV CLASS:


MAN


EXT


REL. TIME


FREQ


S-ADV


X's which do not share a common letter in the grouping are significantly different; those which do are not (- = 0.01).



Fig. 3. R errors by ADV form and ADV class with
associated Duncan Multiple Range Test results.


100-


90 -t-


80---


7


0 Q)


5 0 -t-


4 0 --


3 0-f-


20-


10-f-





40


TABLE 9

Comparison of Duncan Groupings with Item R's


Duncana X Error
Class-Form X's Grouping Items Scores


S-ADV-Morph. S-ADV-ly


FREQ-Morph. REL.TIME-ly EXT-Morph. FREQ-ly EXT-ly


REL.TIMEMorph. MAN-ly


MAN-Morph.


0.77 0.50

0.47 0.41 0.32 0.25


0.20 0.15 0.10


A perhaps
maybe
B probably
certainly B seldom
often
BC finally
suddenly


CD DE EF EF FG


0.03


almost only rarely usually nearly completely
already still carefully
quietly quickly happily


G hard
high fast
straight


the grouping not (a=0.01).


0.78 0.76 0.78
0.24
0.76
0.20 0.42 0.42 0.26
0.41 0.41 0.12 0.31
0.12
0.19
0.12 0.14 0.03
0.05
0.20 0.02 0.00 0.06 0.06


aR's which do not share a common letter in are significantly different; those which do are





41


were the most difficult in this test,1 and MAN, at least, morphemic MAN, were easiest. All that can reasonably be stated in terms of ADV form and class is in the sense of trends, and these are presented in Figures 4 and 5. The overall pattern in Figure 4 shows that morphemic differences are not obvious across the grades. On the other hand, Figure 5 shows some very clear differences as a function of ADV class. This apparent contradiction between statistics and graphic presentation can be explained, at least in part, by a more careful examination of Table 9.

If the Duncan groupings are compared with the item

means (see especially certainly, which I have already mentioned), it is apparent that in many cases specific item differences within class/form cells can be quite large. This argues for two tentative conclusions: first, that ADV form is not, in and of itself, a significant factor, but rather it interacts in a special way with the items in this test; second, no firm statements can be made relative to ADV class other than S-ADV and MAN. However, the possible implications of the results for the middle-difficulty items will be discussed in Chapter IV, in conjunction with data to be reported later.

During the spring of the previous school year, 16 of the third graders included in this study had been given the OtisLennon Mental Ability Test (OLMAT). The scores for this test are reported as percentile rankings relative to national norms. The test is group-administered, and not particularly sensitive,





42


100 -f-


ADV form:
* = morph.

= -ly















00


10 t


I I


Grade: 1


2


3


(4)


5


Fig. 4. Percent errors by grade as a function of
ADV form.


70H


6 0 T


0

ci) C) w)


50 t


4 0 ~--


30 -t-


2 0 -t-


I


I





43


100-- -


9 0 -t-


8 0 -4-


N,


S-ADV


REQ


RE L. T IME

E X 5

-~ 0-


MAN


I I


Grade: 1


2


3


(4)


1
5


Fig. 5. Percent errors by grade as a function of
ADV class.


70 -t-


604-


0


-p

U Q)


50-t-


40-t-


3 0 -+-


20 -1-


10~-1





44


but it does provide a reasonably reliable ranking of the child's mental ability.

An analysis of covariance (ANOCOVA) was conducted on

the two sets of scores available for this grade: the error scores from the present study and the OLMAT percentile ranks. The results are presented in Table 10. Apparently, mental aptitude is a significant factor in the ability to process the ADV items in this study for children of this age. The results indicate that, while there is no apparent effect of ADV form when mental ability is one factor in the analysis, ADV class is still a significant variable, and the interaction of ADV form by class still holds, as it did in the ANOVA.


TABLE 10

Analysis of Covariance for OLMAT Scores,
Adverb Form and Adverb Class


Source S.S. df m.s. F


OLMAT Score 0.80 1 0.80 10.46*
ADV Form 0.14 1 0.14 1.84
ADV Class 5.16 4 1.29 16.85*
Form X Class 1.99 4 0.50 6.51*
Error 11.41 149 0.08 --*significant at = 0.01


Because of this significant interaction, it is necessary to discuss the relation between the two significant main variables, ADV class and OLMAT score, separately for each of the five ADV classes. In Figure 6, regression lines have been plotted for each class as a function of OLMAT





45


score. A regression line, it should be noted, is a straight line of best fit for data points plotted on two variables, and the slope of the line reflects degree of change in one variable relative to change in the other. Figure 6 indicates only slight differences in MAN and S-ADV scores as a function of mental aptitude. This merely reinforces the ANOVA results that MAN is already 'acquired' by Grade 1, and that S-ADV is almost as much of a problem at Grade 5 as it is for the younger children. For the 'middle-difficulty' classes, the graphic representations suggest that ability to handle FREQ does not change much as a function of general mental ability, but that there is a somewhat more obvious change for EXT and REL. TIME.

It was noted in Chapter II that 8 eighth graders were tested in order to establish an 'upper-limit' for the test. Since these 13-14-year-olds could reasonably be considered 'linguistic adults,' their data were intended as a 'test of the test': that what was at issue here was really an ability to process ADVs. All of the subjects in this group were asked to explain their reasons for choosing a particular panel. Two subjects gave the correct responses and explanations for all items. One subject gave an aberrant but very interesting pattern of responses, and he will be discussed later. The remaining five subjects gave correct answers and explanations to all but one or more of the critical S-ADVs (i.e., excepting certainly, which all correctly explained). The eighth graders did, in fact, 'validate' the test as a





46


100 MAN
75
50
25

0 49 40 50 60 70 80 90 100


100 75 50 25

0l


EXT


S 00
I S S
.1


50 60 70 80 90 100


40


REL. TIME
0


50 60 70 80 90 100


100 75 50 25

0 - 40

100 75 *


50+


25
0


40


FREQ


0 * S I


50 60 70 80 90 100

- 0 S-ADV


* ''


I 1
50 60 70 80 90 100


OLMAT Percentile Scores


Fig. 6. Percent errors as a function of OLMAT
percentile score for the five classes
of ADV.


100


U1)
0

-Ii
a)
w 4J


50


0


40


I


I


I


a


Y) I


I


a


I


I


I


I





47


task involving the ability to process ADVs. The fact that so many of them had trouble with at least one of the S-ADVs is particularly interesting when compared with their comments. That issue will be treated in a later section.


The Voluntary Data


The second form of data that this study produced was

the comments and other voluntary responses that the children gave. These data were not an integral part of the experimental design, but all comments and reactions the children made during the test session were noted. For the most part, these data can be accurately described as 'voluntary responses.' However, some children were asked to explain their choices, but this was not done in any systematic way. Even for those children who were asked to explain, I did not insist that they do so.

These spontaneous or solicited responses cannot be taken as anything more than corroborating evidence for the major statistical results, and the analysis of these data was as loose as the method by which they were collected. These data are primarily 'impressionistic.' Nevertheless, there seem to be some definite and intriguing patterns in these responses, and they will be reported in terms of certain patterns or tendencies which seem to change with grade.

For Grade 1, the apparent tendency is to attend primarily to the verb. Typical comments for average-score subjects in this age group are presented in Table 11. It should be





48


TABLE 11

Sample Comments by First Graders


Snoopy carefully put the bird in the next.
'because he's lifting it up'
'because he put it in (B-3); there (A-3) blew it out' Snoopy quickly ran to kick the ball.
'because he's kicking it'
Lucy happily listened to Schroeder play.
*'because she's listening to it'
*'because she's looking at him in B and listening in A' Snoopy nearly fell off his doghouse.
*'(pointing emphatically to A-4) where it did happen' The fire completely burned Snoopy's doghouse.
'because it's burning his house' Linus only hit the ball.
*'he's just hitting the ball'
-subject repeated the S without ADV Linus finally found his blanket.
*'because he found it'
*'both the same'
Linus seldom helps his mother(A-l) 'because he's vacuuming'
*'he's doing laundry and stuff' (indicated B-1, B-2, B-4) Lucy rarely plays with Linus.
*'because (in three frames) she's playing'

* = wrong response
(A-l) refers to the first frame of panel A, etc.


noted that a response of 'they are both the same' was considered a wrong response, i.e., an indication that the subject was not processing the ADV, since the panels in a set differed only with respect to the ADV. There was also a strong tendency for these children to point to the critically wrong frame in a panel for S-ADV panel-sets and other 'implicit NEGs,' and to do so emphatically. This issue will be discussed in more detail in the next section.





49


The most prominent change in the comments by average

scorers in Grade 2, compared to Grade 1, was their tendency to mention the ADV, usually as a question after the test-S had been read to them, but also as the 'reason' for their response. Also, there were more correct explanations of right responses, occasionally with correct definitions of the ADV-item. This was still infrequent, but certainly an increase over the average first grader. But there was still a definite indication of 'verb-processing' and 'both the same' responses, especially for those items that were 'hard,' as indicated by the error score data.

The change in pattern from Grade 2 to Grade 3 is primarily one of degree. Here the children more frequently mentioned the ADV, either repeating it to themselves, asking what it meant, or using it as the reason for their responses.

By Grade 5, almost invariably, if the children made a

comment, it was to repeat or question the ADV. For the most part, they volunteered, or could give on request, good explanations for their choices. But one other notable change for this age group was that, as error scores decreased, there was a decreased likelihood of any comment being offered or mention of the ADV being made. Evidently, as 'processing ADV' becomes more automatic, 'ADV awareness' decreases.





50


The Relation between the Two Forms of Data


The two forms of data collected in this study tend to be mutually supportive, but in one area more than others. Certain items in this test can be classified as 'implicit NEG,' i.e., if that item were deleted, the test-S would go from a negative to a positive statement, as in the following:

37a. Snoopy nearly fell off the doghouse.

37b. Snoopy fell off the doghouse.

38a. Linus rarely helps his mother.

38b. Linus helps his mother.

39a. Probably Snoopy grabbed the blanket.

39b. Snoopy grabbed the blanket.

These three types of 'implicit NEG' are, of course, not equal in degree or type of negative, and that distinction will be discussed in the next chapter. But if it is true, as the comments from the first graders seem to indicate, that, early on, the strategy is to 'process the S without ADV,' then we might expect to find: 1) higher than average error scores on these items for the young children, and 2) that errors would decrease across grade. In addition, the comments of older children on these items should give some indication of the process of ADV acquisition.

The items which fall under this category are presented

in Table 12, along with mean error scores for Grades 1 through

5. The eighth graders made no errors on the non-S-ADV items, and their scores on the S-ADV items will be reported with





51


their comments. The trends in the error scores for these items are presented in Figures 7 and 8.


TABLE 12

Error Scores by Grade for 'Implicit NEG' Items



Class Item 1 2 3 5


EXT nearly 0.73 0.20 0.25 0.05
almost 0.53 0.30 0.20 0.00
FREQ rarely 0.73 0.35 0.40 0.15
seldom 0.87 0.90 0.80 0.45
S-ADV probably 0.80 0.70 0.80 0.80
perhaps 0.93 0.85 0.70 0.65
maybe 0.80 0.90 0.80 0.55



Although each of the ADV classes covered here is arguably 'implicit NEG,' each class differs subtly in 'degree' or 'type' of negative. For example, the EXT items are true NEG, i.e., S-40 is positive, while S-41 is negative:

40. Snoopy fell off the doghouse.

41. Snoopy nearly fell off the doghouse.

The FREQ items have a less obvious NEG effect, though for the linguistic decisions required in this test, they also function as implicit NEG: the correct panel-set for the test-S for these items showed an event happening once versus three times. The S-ADVs are NEG in an even more subtle sense. Since they represent 'speaker doubt' or 'degree of surety,' the statement that followed could not be judged positive. By default, the 'positive' panel was the wrong answer.




52


100-r-


seldom
















*\ -


,--Q


2


rarely almost early


3


(4)


1
5


7. X errors by grade for EXT and FREQ items.


9 0--


8 0-H-


7


0-T


6 0-i-


U)
0



-4
C)
124


5 0-t-


40-t-


30-1-


2 9-T


1 O-t-


u
Grade:


1


,... A


I


I


I


I


'


Fig.


-. -




53


100-1-


6.


90-T-


-- . .


probably perhaps

maybe


I I I


Fig. 8. X errors for S-ADV items.


N


8070-


60


N







/


0 a)


50 -


4 0~-


30-


2 0 T


10 -1~


0
Grade:


1


2


3


(4)


5


--.. .A


I


I


I





54


Because of the differences in relation to the notion of 'implicit NEG,' the S-ADVs will be discussed separately from EXT and FREQ. Only Grades 1-5 will be considered for EXT and FREQ, since all eighth graders of interest here correctly explained these items.

In Table 11, presented earlier, comments made by first graders were listed as evidence for 'processing verb.' The comments on the EXT and FREQ items listed there are consistent with the error scores for this age group and support the notion that if the ADV is omitted in processing S, there is a greater likelihood of a wrong response.

For Grades 2 and 3, the error rate drops considerably for all items except seldom. For the other items, rarely, almost, and nearly, the patterns in the voluntary responses are consistent with those mentioned earlier for all items. The most obvious change from the first grade pattern is for the children to mention the ADV, either aloud but to themselves, or as a question, or to simply ask what the word means. There seems to be some correlation between this type of response and low error scores, but, of course, this is only an impression, since these data are not amenable to statistical analysis. But there can be no question that in this age range, the voluntary comments indicate an awareness that ADV is a factor in processing the S, and this would corroborate the significant decline in errors from Grade 1 to Grade 2. For the item seldom, where the error scores do not decline appreciably until Grade 5, the outstanding




55


observation was that the children in Grades 2 and 3 took an unusually long time to respond to this test-S, whether they eventually gave a right or wrong answer. The children with high error scores tended to ask what the word meant. As with the other items being discussed here, their comments tend to reinforce the notion of 'emerging ADV awareness' in this age group, or at the least, an awareness that all elements in the string must be processed.

The comments from Grade 5 are relevant in an obverse way: they tended not to volunteer any comments on these items (the exception again is seldom, where they also tended to ask what the word meant, though, of course, less often than the younger children). The reason why this particular finding is relevant is that it supports what I have suggested earlier: once the items have been mastered, as evidenced by the error scores, the children become less 'aware' of the fact that they are 'processing ADVs,' and consequently do not mention them.

The comments on S-ADVs are less clear-cut than those for EXT and FREQ, possibly because there is no dramatic change in error scores for these items until Grade 8. But there are still some interesting implications in the comments on these items.

The changes in the comments on S-ADVs for Grades 1, 2, and 3 are not notably different from those this age group made to the other 'implicit NEG' items. In general, the comments support the high error scores as 'processing S





56


without ADV.' The children in Grade 5 mentioned the S-ADV items somewhat more often than the younger children, but there was no evidence of a greater ability to explain these items, even though the error scores dropped for this age group. In fact, this group offered fewer explanations than the younger children with higher error scores (= wrong explanations). This could be taken as evidence of 'emerging awareness' of the ADV-class, especially in relation to the comments and explanations given by the eighth graders.

Since all of the Grade 8 subjects were asked to explain their responses, their data are particularly interesting. The comments of the seven subjects who made no errors, or errors only on S-ADV items, are listed in Table 13, along with their total error scores. A consideration of this table shows that, for these three items, even at an age that is generally considered to be 'linguistically adult,' some children obviously still have problems with S-ADVs. The comment made by 'Eileen' on the item perhaps is particularly interesting: it suggests that even at this age, children are not yet ready or able to make the fine distinctions required by 'modal' ADVs such as these. The comments made by the eighth graders may be the strongest argument for the relation between the objective error scores and the voluntary comments, but the progression is also important.

There were two children included in this study whose data I would like to compare, since they made such a vivid contrast on all of the measures that I have considered so far.




57


TABLE 13

Summary of Comments by Eighth Graders on S-ADVs Item Subject Errors Comments


probably Becky
Matt
Eileen
Michele
Bill Gina
Linda

perhaps Becky
Matt

Eileen

Michele
Bill Gina
Linda

maybe Becky
Matt

Eileen Michele
Bill

Gina
Linda


A - in B he obviously did A - because you don't see it happen
*B - because A is more intentional
*hard to explain
*in A he's going through him ?Snoopy's walking behind
*looks like he meant to do it in A

A - in B she did perhaps? - can't tell if she did it in A
this is another one - don't like those 'perhaps' (items) ?she grabbed it and ran A - doesn't show her doing it
*because B-3 she's grabbing it in B she snatched it

maybe? - B (right reason) B - he maybe fell asleep, but can't say he didn't
B - in A 'Z's' show he's asleep B - in A he did maybe? - means I'm not sure, right? - B
*maybe? - A-4 - he's asleep
*A-4 he's asleep


* = wrong response ? = right response


but comment suggests wrong reason


Derek, an eighth grader, made 6 errors, twice the number of any other eighth grader. This error score would place him, by extrapolation, at about fourth grade level. What is more, he gave wrong answers to just those items for which the third grade group had high error scores. His explanations for his responses were also similar to those given by third graders.





58


His scores on the 'Eighth Grade Test' were in the 31st percentile for reading and the 14th percentile for math. By contrast, Alexander, a second grader, made only 3 errors (rarely, perhaps, and maybe) compared to a grade mean of 8.4 errors (and a fifth grade mean error score-of 4.8). I do not have an objective test score for this child, but these data, in conjunction with his comments, seem to indicate a superior mental ability. Alexander gave correct explanations for most of his responses, and so his low error score cannot be a statistical 'fluke,' especially considering those items to which he gave a wrong response.

The comparison of the data from these two children is

a strong argument for the tie between the ability to process ADV and general cognitive development. In a narrow sense, these data substantiate the claim that error scores are corroborated by the ability to explain responses. In a much broader interpretation, the data from these two children argue for an 'age-free' acquisition process, which, of course, has implications beyond this study.


Summation of Results


The statistical analyses of the error scores from this study produced the following results:

1. The form of the test is not a significant variable.

2. The sex of the subject is not significant.

3. The ability to process the ADV items included here does change significantly between Grade 1 and 2 and between Grade 3 and 5.





59


4. Because of a significant interaction between ADV

form and ADV class, it is not possible to draw firm conclusions as to the effect of these variables individually, beyond saying that morphemic-MAN items were easiest for the age groups considered here, and morphemic-S-ADVs were the most difficult. However, there are indications that ADV form is not a true factor in acquisition, but that ADV class may well be, and that the significant interaction in this study is a function of the particular items that were included.

5. On the basis of the ANOCOVA results for the third grade, mental aptitude is a definite factor in the ability to process ADVs.

The informal analysis of the voluntary comments and responses made by the children tend to corroborate the statistical results. The implications of these findings will be the subject of the next chapter.





60



Note


The item certainly needs special comment. It became clear to me during the testing sessions that the children were, for the most part, treating this item not as S-ADV, but as EXT, i.e., not as synonymous with 'I am certain' or 'I am sure,' but rather as a synonym of 'really,' 'a lot,' etc. For this reason, I think it is more reasonable to discuss the S-ADV results here in terms of only the three items which were true S-ADVs.













CHAPTER IV

DISCUSSION


Syntax, Semantics, and the Significant Interaction


Before discussing the syntactic and semantic results of the study and their possible implications, it is necessary to consider the significant interaction of form by class. No firm statements can be made about either of the main effects unless some logical explanation can be offered for the significant interaction. It is possible that factors not considered in the statistical analyses could have influenced the results, indicating that form and class interact when in fact they may not.

The significant interaction is almost certainly a function of the experimental design. By insisting on equal cell entries for the 2-form by 5-class array, and with the added limitation of 1-word ADVs, the choice of possible items to fill the cells was in several cases severely limited. The ideal would have been a selection of items of equal frequency or familiarity. But given the vagaries of language, this was not possible. The best approximation was to fill the cells with the most familiar items which met the basic criteria. This tight experimental design exacted a price: the statistically significant interaction of form and class.


61




62


A consideration of Table 14, which is a reduction of Table 9 (p. 40) demonstrates the point.


TABLE 14

Comparison of Duncan Groupings with Item X's


Duncana X Error
Class-Form X's Grouping Items Scores


FREQ-Morph. 0.47 B seldom 0.76
often 0.20
R.T.-ly 0.41 BC finally 0.42
suddenly 0.42
FREQ-ly 0.25 DE rarely 0.41
usually 0.12
R.T.-Morph. 0.15 EF already 0.19
still 0.12

aRs which do not share a common letter in the grouping are significantly different; those which do are not (- = 0.01)


Table 14 shows that FREQ-Morph. is significantly more difficult than FREQ-ly, whereas R.T.-ly is significantly more difficult than R.T.-Morph. That is, for these two classes, there is apparently an inversion of difficulty by ADV form. But the statistical analyses were run on R error scores per cell. If the individual item R's are examined, the statistical differences appear to be more a function of the pooled X's than of real differences in difficulty. For FREQ, the individual item R's are widely divergent: seldom (R = 0.76) ranks with S-ADV items (see Table 9, p. 40); often (R = 0.20) ranks close to the R.T.-Morph. item already (R = 0.19). The R for the FREQ-ly item rarely is almost




63


identical with the X's for the R.T.-ly items finally and suddenly, yet FREQ-ly and R.T.-ly are statistically different. The other FREQ-ly item, usually, has a R equal to that of the R.T.-Morph. item still.

The logical conclusions to be drawn from the statistically significant interaction of syntactic form by semantic class seems to be that 1) it is probably not 'real' but rather a function of the frequency/familiarity of individual items (to be expanded later), and 2) the major source of this significant interaction was the middle-difficulty classes: REL. TIME, FREQ, and EXT. Another rather obvious conclusion is that language does not readily fit the structured experimental designs favored by psychologists. The 'accidental gap' for the linguist can mean empty or inappropriately filled cells for the psycholinguist.

Even though the apparent form by class interaction can seemingly be explained away on logical grounds, at least for the items in this study, the other syntactic and semantic results can be legitimately discussed only in terms of trends and suggestions. But since it would be cumbersome and distracting to continually refer back to this proviso, the following discussion will assume this is understood, and will be presented in terms more positive than are strictly justified.

The syntactic question addressed in this study is very narrow. Basically, the question is: Are derived ADVs





64


acquired earlier than morphemic ADVs? The point of the question is that, since an adult speaker of English obviously has acquired the rule: ADJ+-ly=ADV, it might be the case that, in the process of acquisition as the child becomes aware of the rule, '-ly' items might be mastered earlier than morphemic items. The evidence from this study indicates that there is a small but significant difference in time of acquisition for the two forms: the '-ly' forms were less difficult than the morphemic forms. But somewhat of the same situation holds for this result as was discussed in relation to the significant interaction.

An examination of the individual item X's in Table 9 (p. 40) shows that the error score for 4 items, perhaps (X = 0.78), maybe (X = 0.76), probably (X = 0.78), and seldom (X = 0.76) are almost double the scores for the next most difficult items (finally and suddenly, X = 0.42). Three of the four items are morphemic. It is possible that two factors other than syntactic form could have contributed to the statistically significant difference for form. The first is level of conceptual difficulty: two of the items are S-ADVs; the second is familiarity: many children indicated they did not know the word seldom. Both of these ideas will be developed in the next section on differences between classes of ADV.

In sum, no substantial conclusions can be drawn from this study in the area of syntax. It is possible that the





65


ADJ+-ly rule is functional in the early school years, but it is almost certainly not a major factor in ADV acquisition.

The semantic question, i.e., the question of the effect of ADV class in acquisition, is the one area in which the statistical analysis of error scores seems to provide some definitive answers. The results indicate that ADV class is a significant factor in acquisition, and that the 5 semantic classes can be divided in terms of 3 levels of cognitive difficulty. In brief, MAN items are easiest/earliest acquired; FREQ, REL. TIME, and EXT are 'middle-difficulty' items; and S-ADVs are the most difficult/last acquired. The reality of these 'levels' can be supported by evidence beyond the statistical results.

Even for the youngest children, error scores on MAN items were very low. There are two factors which might account for the earlier acquisition of MAN relative to other 'process-modifier' items. The first is that the concepts covered by these MAN items are conceptually more basic than the TIME/SPACE concepts covered by the 'middle-difficulty' items. This is supported by the work Ervin-Tripp (1970) did on WH-questions. Her findings showed that children can correctly answer how questions before they can handle when questions. One small piece of evidence from this study is that some of the Grade 1 children made comments such as 'This one is quiet/quick/happy' in indicating the correct panel for the MAN-ly items.





66


The second reason involves the relation between ADJ and ADV and the notion that items are processed if and when they are understood. For the MAN items in this study, the similarity to ADJ is much stronger than for any of the other classes: all of the MAN-morph. items are phonetically identical to their ADJ counterparts. It could be that the younger children were dealing with the string containing MAN-ADV as they would a string with ADJ, and are not necessarily exhibiting an ability to process ADV. This subject will be discussed further in the final section of this chapter.

The 'middle-difficulty' items, which cover the classes REL. TIME, FREQ, and EXT, present the most difficult problems of interpretation and extrapolation. The general conclusion to be drawn from the data on these classes is that the individual item is in some way more important than either syntactic form or semantic class. For example, consider the items in the FREQ class: seldom (X = 0.76), often (i =

0.20), rarely (i = 0.41), and usually (X = 0.12). Seldom and rarely are synonymous but the X's for these items are quite different, and much higher than the R's for the other 2 items. They are, of course, different in syntactic form, but then consider the items almost (R = 0.26) and nearly (X = 0.31), also synonymous, but different in syntactic form, yet with R's that are virtually the same. It is an interesting sidelight that, during the testing sessions, some of the





67


older children, usually the fifth graders, asked, 'What does seldom/rarely mean?' When told 'not very often' their response was 'Oh, then it's (correct panel).'

This predominance of item over class or form is also evident for the following items with virtually identical X's: finally/suddenly (X = 0.42), which are REL. TIME-ly; rarely (X = 0.41), which is FREQ-ly; and only (X = 0.41), which is EXT-morph. On the basis of the data, the final conclusion for these 'middle-difficulty' items must be that 1) the items and the concepts they cover are acquired during the early school years, since the Grade 1 children made many errors on these items and the Grade 8 children made none; 2) there is probably some scale of difficulty for the concepts covered by these classes; 3) it is not possible to determine from the results of this study just what that scale might be.

The data on S-ADVs were unquestionably the most conclusive and, in many ways, the most interesting results of the study. This class of ADVs was consistently most difficult for Grades 1-5, for which there are statistical data, but it was also somewhat of a problem for some of the children in Grade 8, as indicated by their comments and their error scores. These results are not surprising: S-ADV, as a proposition modifier, was expected to be cognitively more difficult than the 'process-modifier' items. This expectation was based on general notions from language acquisition





68


studies which indicate that the child develops from specific to general concepts, but more specifically, the analysis of S-ADVs presented by Schreiber (1970), mentioned briefly in Chapter I. Schreiber's suggestions deserve further consideration in light of the results of this study. According to his analysis, S-ADVs fall into two categories: 'modal' and 'evaluative.' Modal items (e.g., apparently, clearly, obviously) indicate 'degree of surety' on the part of the speaker. Evaluative items (fortunately, oddly, naturally), on the other hand, express the speaker's opinion on the statement being made. In order to correctly process S-ADVs the child must reach the point where he is able to recognize that these items represent a 'comment on the proposition,' and that the comment represents either speaker opinion or speaker surety of the factual nature of the proposition. But there is a further complication, in relation to acquisition, from Schreiber's analysis. With modal S-ADVs, the proposition is not necessarily 'fact.' With evaluative items, it is. Consider the following Ss:

42. Presumably, Snoopy fell off the doghouse.

43. Regrettably, Snoopy fell off the doghouse.

In S-42, the event mentioned may or may not have actually occurred; in S-43, it is implicitly considered true. In sum, before the child can correctly process S-ADVs, he has to infer some rather sophisticated notions, restrictions and extrapolations (in terms of implicit information).1




69


Given Schreiber's analysis, and the results of this

study, the comments made by the eighth graders are particularly interesting (see Table 13, p. 57). The 2 children who made no errors on the test gave good explanations for all the S-ADV items. The comments of the other 5 children who gave incorrect answers to 1 or 2 of the S-ADV items seem to indicate that these children are partially aware of the special function of S-ADV (as opposed to 'process-modifier'), but that they do not yet fully understand the implication of these items.

It should be noted that all S-ADV items in this study were 'modal' by Schreiber's analysis, and so no unqualified statement can be made as to the class in general. Still it seems that once the function of S-ADV is recognized, both 'modal' and 'evaluative' ADVs should be acquired at about the same time. That is, since all S-ADVs represent 'speaker comment,' once the child becomes aware of this function, he should be able to handle both types of S-ADVs. Nevertheless it remains an interesting empirical question.


ADV Acquisition and Cognitive Development


The final question posed in Chapter I concerned the

possible relationship between ADV acquisition and cognitive development. The statistical results for ANOVA and ANOCOVA, as well as the informal analysis of the voluntary comments, relate to this question. The sum of the evidence indicates




70


that there is a strong tie between cognitive level and the ability to process ADVs.

The analysis of the error scores showed a significant

decrease in errors between Grades 1 and 2 and Grades 3 and 5. For Grade 8 there were no errors, except the few on S-ADV items. Clearly, facility with ADVs improves with age, but this does not necessarily imply a direct connection between that improved ability and cognitive development. The older children have had proportionately greater exposure to the various items. It is a question of whether the improvement is more a reflection of familiarity or of a developing awareness of the basic function of ADV. The ANOCOVA results are relevant here. They indicate a significant correlation between test scores and mental ability for one age group. Assuming all 7-year-olds have had about the same exposure to ADV, we can tentatively conclude that it is primarily level of cognitive development that is the major factor in the significant improvement across grades.

The analysis of the voluntary data provides further support for this conclusion. The children in Grade 2 and above were increasingly able to correctly explain their responses. They were demonstrating an increasing awareness of the basic function of ADV, and also of the concepts covered by the particular ADV items.

Finally, there is some outside evidence to support this conclusion. Loban (1963, 1964) found a close relationship





71


between mental ability and use of ADVs, both in number and form, for grade school children. He used a 'finish the sentence' type task involving adverbial connectives for Grades 5-9. Children in the high ability group did well even at Grade 5, whereas the low ability group were consistently poorer through Grade 9. Clearly, there is a close connection between ADV acquisition and cognitive development that deserves to be investigated in more detail.


Stages in ADV Acquisition


The major result of this study is that, if the various

forms of data are integrated, certain stages in the acquisition of ADV become apparent. The statistically significant results say that ADV acquisition relates to age (ANOVA) and to mental aptitude (ANOCOVA). The voluntary comments were even more interesting, and more illuminating, in suggesting just how ADV acquisition relates to general cognitive development. In sum, an overall view of ADV acquisition seems to emerge. This view can be presented in terms of 'stages' which relate to degree of 'awareness' and apparent strategies for processing Ss with ADV.2


Stage 1: unaware/ignoring ADV

For the youngest children in this study, the concepts

underlying many of the ADV items were evidently beyond their level of cognitive development. Therefore, the apparent strategy was to process the string without the ADV. This





72


would account for the observation that the younger children tended to quickly and enthusiastically point to the critically wrong frame for implicit NEG items: the choice is so obvious if the ADV is not considered! There is evidence that this proposed stage is present in other areas of acquisition. Fraser, Bellugi, and Brown (1963) found that young children tended to interpret a S such as the girl is pushed by the boy to mean that the girl was doing the pushing. The assumption is that, if the passive formation is not yet understood, those items in the S which make it passive are ignored, and the S is processed as active. Smith (1970) also found evidence of what she calls 'selective listening' in 3- and 4-year-olds in repetition and response studies. The conclusion was that the children repeated or responded according to those elements of the test S that they understood. Keller-Cohen (1975) reached a similar conclusion in her investigation of temporal connectives. Her subjects, who evidently did not understand the item while, substituted before, after or and then in their repetitions of test sentences containing while. There seems to be a consistent--or at least, prevalent--strategy in acquisition, across syntax and semantics, to process strings according to the current level of linguistic and/or cognitive development.


Stage 2: Partially aware/guessing at ADV

At some point, typically around the Grade 2 level, the child seems to begin to recognize that all elements in a




73


string have some significance. But since the general concepts behind many of the ADVs are still beyond his comprehension, there is a tendency to guess. There is additional evidence of this 'partial awareness' in that children at this stage often repeat the ADV aloud but to themselves or repeat it as a question. They still process the S without ADV, but without the surety (for the wrong response) exhibited by the children in the previous stage.


Stage 3: Aware/asking about ADV

At this point the child does recognize that all elements in the string are necessary for correctly processing the string. It is at this stage that the child either responds correctly or asks directly what the ADV means. Correct responses are often explained by defining or referring to the ADV.


Stage 4: Unaware/comprehending ADV

This is not truly a 'stage,' unless it can be considered the final stage, but rather some observations on typical behavior when ADV acquisition is virtually complete. The term 'unaware' reflects the fact that, once the basic function of ADV is comprehended, the ADV is rarely mentioned voluntarily, in sharp contrast to the 2 earlier stages. That is, the more complete the ability to process ADV, the less aware the child seems to be that he is doing so! It is an interesting sidelight that most of the eighth graders who made few if





74


any errors and correctly explained their responses, asked after the test session just what I was studying, and were surprised to learn that it was ADV.

There is an important point that needs to be made here, and that is, that these stages are not directly linked to chronological age, syntactic form, or semantic class. Rather they apparently relate most directly to an emerging sense of the basic function of ADV. This 'sense of ADV function' is in turn apparently correlated with cognitive development.

The contention that ADV acquisition is more closely

related to mental ability than to chronological age is supported by the ANOCOVA results for Grade 3, which showed a significant relation between errors on the test and OLMAT scores. The comparison of the error scores and voluntary comments for the gifted second grader and the exceptional eighth grader give additional support to this contention.

The statistical results of this study do not relate to the question of syntactic form in these proposed stages. The fact is that the syntactic dimensions of the study were too narrow to make any conclusive statement. But the evidence from the studies mentioned in Chapter I on the language of elementary school children, in particular, Loban (1963, 1964), suggest that children's use, and presumably comprehension, of more advanced syntactic forms increases with grade into the high school years.

For the semantic dimension of ADV, the logical conclusion seems to be that the stages would apply almost on the





75


level of individual items, but with certain predictable overlaps. The evidence that some items within a given class were more difficult than others in the same is the complicating factor. Of course, this applies primarily to TIME and EXT items; the MAN and S-ADV, in terms of statistical results and analysis of voluntary comments, fit well with this stage theory. For the TIME and EXT items, more detailed investigation is needed. But the error scores, in conjunction with voluntary comments for those items discussed under 'implicit NEG' in Chapter III do support the stages.

The evidence for the existence of the 'stages' in ADV acquisition proposed here is strong, but not conclusive. It is important to recognize that no claim has been made that a child, at whatever age, is exclusively in one stage with respect to general ADV processing. Rather, the implication is that a child--at least, beyond Stage 1--goes through the progressive stages almost on an item basis. There are, of course, indications that the less complex MAN items will be mastered first, and the S-ADV items last. But in between, the predictable patterns for ADV, based on this study, are that the typical child will be in various stages at any given point, relative to ADV class and form. It is most probable that the child handles individual ADVs at a given level dependent upon the syntactic and semantic complexity of the specific ADV item.




76



Notes


Schreiber's paper offers some interesting suggestions on syntactic complexity of S-ADV. In fact, that is the major premise of the paper. He also suggests some ways in which syntax and semantics interact for S-ADVs.

2The term 'awareness' needs to be defined as it is used in this paper. It does not imply any conscious level in language processing. Rather, it refers to a logical way of categorizing the reactions/responses of the children to this test.














CHAPTER V

SUMMARY AND CONCLUSIONS


Review of Purpose and Questions


There is ample evidence that the category ADV is a major problem in the linguistic description of English. There is also good reason to believe that data on acquisition can provide important insights for the descriptive linguist. But the data on ADV acquisition is minimal. For the most part, it must be derived from incidental evidence in studies with primary emphasis on other questions or areas of acquisition. The purpose of the present study was to look directly at ADV acquisition: to look at some of the more basic aspects of ADV in relation to acquisition, and to establish some guidelines for more detailed investigations. The elemental questions addressed in the study relate to the basic function of ADV, to the relative importance of syntax and semantics in the acquisition of ADV, and to the possible relation between ADV acquisition and cognitive development.


General Conclusions


The statistical results of the study are ambiguous in some cases, but certain general conclusions, based on an integrated interpretation of error scores and voluntary


77





78


comments, can be drawn. There may be some slight advantage of '-ly' items over morphemic ADVs in the process of acquisition. And it is possible that ADV form and class interact--at least they did statistically for the items in this study. On the other hand, it seems sure that semantic class is the most important factor in acquisition. The S-ADVs were the most difficult of the classes considered here, and MAN the least difficult. No definite statement can be made relative to EXT, FREQ, and REL. TIME.

The most conclusive statement that can be made from this study is that there is a strong connection between ADV acquisition and cognitive development. Error scores decrease with advancing grades: older children as a group are better able to process ADV than are younger children. This is not so surprising, but consider that error scores for children in the same grade are directly related to their mental aptitude scores. The decrease in error scores by grade is almost surely attributable to increase in cognitive level, and not to other factors such as increased exposure to ADV. Perhaps the most interesting conclusion drawn from the study was that there appear to be stages in ADV acquisition which relate in certain definite ways to other areas of acquisition.


Projections


This study represented a first attempt at an organized

investigation of children and ADVs. The results provided some




79


general guidelines for future research in this area. More importantly, the results suggest certain issues and specific questions in the area of ADV acquisition that deserve special attention.

Since this study was confined to 1-word ADVs, the initial question is whether the use of phrasal ADVs, which are paraphrases of the items included here, might permit a more detailed description of the development of time and space concepts. A further question relates to whether the stages that were apparent in this study can be verified and defined in more detail. That stages exist in early language is welldocumented, and so it is to be expected they are also a part of later acquisition. In fact, Chomsky (1969) found that in her ask/tell studies there was evidence of a steady progression in comprehension. Chomsky also mentioned another aspect of this 'progression' that was apparent in this study as well: that it is not linked to chronological age. This is an important consideration that is sometimes overlooked in studies of child language. Unless the N is quite large, mental age is more likely to be a critical factor than is chronological age.

A final consideration is that of methodology. The test used here worked well for the specific questions being considered. It was evident, however, that it was not appropriate for children under 6, or below first grade. It would seem that an adaption of the imitation/repetition method that has





80


been so productive with young children in other areas of acquisition should work equally well with ADVs. For older children, more information could possibly be obtained by requiring subjects to explain how the panel sets differed. This method seems especially appropriate to explain in more detail the results that Loban (1963, 1964) found: that for older children, there is an increase in both the frequency of ADV usage, but more interestingly, an increase in the different types of syntactic form of ADV.














APPENDIX A

Test Ss for Form 1


1. Charlie Brown threw the ball straight.
2. Snoopy carefully put the bird in the nest.
3. Linus only hit the ball.
4. Schroeder quietly played the piano.
5. Lucy suddenly hit Linus.
6. Probably Snoopy bumped into Linus.
7. Charlie Brown already jumped in the leaves.
8. Linus went to bed fast.
9. Charlie Brown seldom watches TV. 10. Perhaps Sally grabbed the blanket. 11. Linus usually helps his mother. 12. Snoopy almost grabbed the blanket. 13. Certainly the cat scared Snoopy. 14. Lucy rarely plays with Linus. 15. Snoopy nearly fell off his doghouse. 16. Charlie Brown flew the kite high. 17. Lucy happily listened to Schroeder play. 18. The fire completely burned Snoopy's doghouse. 19. Lucy is still talking on the phone. 20. Linus finally found his blanket. 21. Schroeder often reads books. 22. Snoopy quickly ran to kick the ball. 23. Lucy hit Linus hard. 24. Maybe Snoopy fell asleep on his doghouse.


81














APPENDIX B


Test Ss for Form 2


1. Charlie Brown threw the ball straight.
2. Snoopy carefully put the bird in the nest.
3. Linus only hit the ball.
4. Schroeder quietly played the piano.
5. Lucy finally hit Linus.
6. Perhaps Snoopy bumped into Linus.
7. Charlie Brown already jumped in the leaves.
8. Linus quickly went to bed.
9. Charlie Brown usually watches TV. 10. Probably Sally grabbed the blanket. 11. Linus seldom helps his mother. 12. Snoopy nearly grabbed the blanket. 13. Maybe the cat scared Snoopy. 14. Lucy often plays with Linus. 15. Snoopy almost fell off his doghouse. 16. Charlie Brown flew the kite high. 17. Lucy happily listened to Schroeder play. 18. The fire completely burned Snoopy's doghouse. 19. Lucy is still talking on the phone. 20. Linus suddenly found his blanket. 21. Schroeder rarely reads books. 22. Snoopy ran fast to kick the ball. 23. Lucy hit Linus hard. 24. Certainly Snoopy fell asleep on his doghouse.


82














REFERENCES


Amidon, A., and P. Carey. 1972. Why can't five-year-olds
understand before and after? J. verb. Lrng. verb.
Behav., 11.417-423.

Barrie-Blackley, Sandie. 1973. Six-year-old children's
understanding of sentences adjoined with time adverbs.
J. psycholin. Res., 2.153-165.

Bronckart, J.P., and H. Sinclair. 1973. Time, tense and
aspect. Cognition, 2.107-130.

Chafe, Wallace. 1973. Language and memory. Lg., 49.261-281.

Chomsky, Carol. 1969. The acquisition of syntax in children
from 5 to 10. Cambridge: M.I.T. Press.

Clark, E.H. 1971. On the acquisition of the meaning of
before and after. J. verb. Lrng. verb. Behav.,
10.266-275.

Clark, H. 1970. The primitive nature of children's relational concepts. Cognition and the development of
language, ed. by J.R. Hayes, 269-278. New York: Wiley.

Dingwall, William 0. 1971. A survey of linguistic science.
College Park: University of Maryland.

Ervin-Tripp, Susan. 1970. Discourse agreement: How children answer questions. Cognition and the development
of language, ed. by J.R. Hayes, 79-107. New York: Wiley.

Feagans, Lynne. 1974. Children's comprehension of some
temporal and spatial structures. Papers from the Tenth
Regional Meeting of the Chicago Linguistic Society,
139-150.

Fraser, C., U. Belluqi, and R. Brown. 1963. Control of
grammar in imitation, comprehension, and production.
J. verb. Lrng. verb. Behav., 2.121-135.

Fraser, James B. 1965. An examination of verb-particle
constructions in English. Ph.D. dissertation, M.I.T.


83






84


Fries, Charles C. 1952. The structure of English: An introduction to the construction of English sentences. New
York: Harcourt, Brace.

Harrell, Lester E., Jr. 1957. A comparison of the development of oral and written language in school-age children.
Monographs of the Society for Research in Child Development, Ser. No. 66, Vol. 22, No. 3.

Hayes, John R. 1970. Cognition and the development of
language. New York: Wiley.

Huang, Shuan-Fan. 1975. A study of adverbs. (Janua linguarum,
Series minor, 213.) The Hague: Mouton.

Hunt, Kellogg W. 1965. Grammatical structures written at
three grade levels. NCTE Res. Rep. No. 3, Champaign:
National Council of Teachers of English.

Hunt, Kellogg W. 1970. Syntactic maturity in schoolchildren
and adults. Monographs of the Society for Research in
Child Development, Ser. No. 134, Vol. 35, No. 1.

Jones, Lyle V., and Joseph M. Wepman. 1966. A spoken word
count-adults. Chicago: Language Research Associates,
Inc.

Keller-Cohen, Deborah. 1975. The acquisition of temporal
reference in pre-school children. Bloomington: The
Indiana University Linguistics Club.

Kessel, Frank S. 1970. The role of syntax in children's
comprehension from ages six to twelve. Monographs of
the Society for Research in Child Development, Ser. No.
139, Vol. 35, No. 6.

Loban, Walter D. 1964. Language ability: Grades seven,
eight, and nine. Coop. Res. Project No. 1131, Berkeley:
University of California.

Loban, Walter D. 1963. The language of elementary school
children. NCTE Res. Rep. No. 1, Champaign: National
Council of Teachers of English.

Menyuk, Paula. 1969. Sentences children use. Cambridge:
M.I.T. Press.

Murray, H.A. 1943. Thematic apperception test. Cambridge:
Harvard University Press.

Nilsen, Don L.F. 1972. English adverbials. (Janua linguarum,
Series practica, 125.) The Hague: Mouton.





85


O'Donnell, Roy C., William J. Griffin, and Raymond C. Norris.
1967. Syntax of kindergarten and elementary school .
children: A transformational analysis. NCTE Res. Rep.
No. 8, Champaign: National Council of Teachers of
English.

Piaget, Jean. 1971. The child's conception of time. New
York: Ballantine Books.

Rinsland, Henry D. 1945. A basic vocabulary of elementary
school children. New York: Macmillan.

Schreiber, Peter A. 1970. Some constraints on the formations of English sentence adverbs. Ling. Inq., l(4).83101.

Slobin, Dan I. 1971. Developmental psycholinguistics. A
survey of linguistic science, ed. by William 0. Dingwall,
298-410. College Park: University of Maryland.

Smith, Carlota S. 1964. Determiners and relative clauses
in a generative grammar of English. Lg., 40.37-52.

Smith, Carlota S. 1970. An experimental approach to children's linguistic competence. Cognition and the development of language, ed. by J.R. Hayes, 109-135. New York:
Wiley.

Wepman, Joseph M., and Wilbur Hass. 1969. A spoken word
count (children--ages 5, 6 and 7). Chicago: Language
Research Associates, Inc.














BIOGRAPHICAL SKETCH


Joeann Chandlee Paige was born on April 3, 1934, in

Wilmington, Delaware. She earned a classical diploma from the Ursuline Academy in Wilmington in June, 1951, and entered the University of Delaware in September, 1951. She completed her undergraduate work for the Bachelor of Arts degree in January, 1955, and was accepted directly into the graduate school at Delaware. In September, 1955, she entered the graduate program in Experimental Psychology at Tufts University in Medford, Massachusetts, with a two-year teaching assistantship. She received her Master of Science degree from Tufts in October, 1957, but remained at the school as assistant investigator on a government-funded project. In June of 1958, she accepted a position at the I.B.M. Research Laboratory in San Jose, California, as Associate Engineering Psychologist, a position she held for three years. She is married to Arnold Paige, a professor of Electrical Engineering at the University of Florida. They are the parents of three children.


86










I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy.




Robert J. Scholes, Chairman Professor of- Linguistics
and Speech



I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy.




'Jean Casagrande
Associate Professor of
Linguistics



I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy.




aul J. Jense
Professor of Speech



I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy.




Ira S. Fischler
Assistant Professor of
Psychology










This dissertation was submitted to the Graduate Faculty of the Program in Linguistics in the College of Liberal Arts and Sciences and to the Graduate Council, and was accepted as partial fulfillment of the requirements for the degree of Doctor of Philosophy.

December 1978




Dean, Graduate School




Full Text

PAGE 1

THE ACQUISITION OF ENGLISH ADVERBS BY JOEANN CHANDLEE PAIGE 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 1978

PAGE 2

ACKNOWLEDGMENTS I am indebted to my chairman, Dr. Robert Scholes, and to the other members of my committee, Dr. Jean Casagrande, Dr. Paul Jensen, and Dr. Ira Fischler, for their advice and suggestions on all aspects of this project. I owe a special debt of gratitude to Dr. Tom Dougherty for all the time and effort he devoted to the statistical analyses. Dr. Bill Sullivan has always found the time to discuss matters great and small, and has been a good friend as well. Ms. Hyta Mederer has also been a constant source of moral support, for which I am most grateful. This study would not have been possible without the enthusiastic cooperation of the teachers and administrative personnel at the P. K. Yonge Laboratory School. It was indeed a pleasure to work with a group of people who were so willing to assist me in my research, but who also ensured that the test sessions did not intrude on normal classroom activities . Finally, I wish to thank Mrs. Anna Marie Martin, who undertook the final typing. She offered many suggestions on style and format that saved me much time and effort. 11

PAGE 3

TABLE OF CONTENTS Page ACKNOWLEDGMENTS ii ABSTRACT v CHAPTER I INTRODUCTION Â’ 1 General Comments on Adverbs 1 General Aspects of ADV Acquisition 2 The Basic Function of ADV 5 Time, Tense, and Aspect 6 The Syntax of ADV 11 The Present Study 19 Purpose 20 Specifics of the Present Study 20 The Questions 21 Notes 22 CHAPTER II THE STUDY 2 3 The Method 2 3 The Subjects 24 The Test 26 Procedure 29 Notes 32 CHAPTER III RESULTS . 33 Preface 33 The Error Scores 33 The Voluntary Data 47 The Relation between the Two Forms of Data. ... 50 Summation of Results 58 Note 60 CHAPTER IV DISCUSSION 61 Syntax, Semantics, and the Significant Interaction 61 ADV Acquisition and Cognitive Development .... 69 Stages in ADV Acquisition 71 Stage 1: Unaware/ignoring ADV 71 Stage 2: Partially aware/guessing at ADV. . 72 iii

PAGE 4

Page Stage 3: Aware/asking about ADV 73 Stage 4: Unaware/comprehending ADV 7 3 Notes 7 6 CHAPTER V SUMMARY AND CONCLUSIONS 77 Review of Purpose and Questions 77 General Conclusions Projections 78 APPENDIX A Test Ss for Form 1 81 APPENDIX B Test Ss for Form 2 82 REFERENCES 83 BIOGRAPHICAL SKETCH 86 IV

PAGE 5

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 ACQUISITION OF ENGLISH ADVERBS By Joeann Chandlee Paige December 1978 Chairman: Robert J. Scholes Major Department: Linguistics The English adverbial system presents some complex problems for both the semanticist and the syntactician . There is some evidence that difficulty in linguistic description is correlated with time of acquisition. The purpose of this study was to investigate certain aspects of that proposition. The particular syntactic and semantic dimensions of this study were very narrow. The syntactic question was: Is there a difference in time of acquisition of one-word adverbs as a function of syntactic form, i.e., morphemic items (e.g., almost) as compared to items derived by the 'basic rule' of adverb formation, i.e., ADJ+-ly=ADV (e.g., nearly ) . The semantic dimension covered five classes: manner, frequency, extent, relative time, and sentential. The overall question addressed in this study related to the possible relation v

PAGE 6

between adverb acquisition and general cognitive development . The experimental procedure employed a newly devised test which was a variation of the sentence-picture matching task. Since adverbs refer either to processes or whole propositions, the task was adapted to a comparison of two four-frame panels, based on familiar newspaper cartoon characters and format. The two panels differed only with respect to the applicability of the adverb in the test sentence. The subjects for the study were 75 students in Grades 1 , 2 , 3 , and 5 . The results of the study indicated that adverb form, adverb class, and the interaction of form by class are all statistically significant, as is the other main effect of grade. There was a significant improvement between Grades 1 and 2, and between Grades 3 and 5. Because of the significant interaction between adverb class and form, only general implications for these two main effects could be discussed. Adverb form may be of some slight advantage in acquisition: — ly forms produced somewhat lower error scores overall. Semantic class is very probably a factor: sentential items were most difficult; manner items were least difficult. The middledifficulty items from the classes extent, frequency, and relative time are apparently the source of the significant interaction, and no definitive statement can be made on these items/classes from the results of this study. Other vi

PAGE 7

data from the study suggest a strong connection between ability to process adverbs and general level of cognitive development. Certain definable stages in adverb acquisition are proposed. vn

PAGE 8

CHAPTER I INTRODUCTION General Comments on Adverbs The category 'adverb' is currently one of the most challenging areas in linguistic theory. It is not hard to understand why, considering that this category covers such a diverse range of semantic concepts and syntactic structures. As Nilsen (1972) noted, in his extensive paper on the surface phenomena of English adverbs: There seems to be a general consensus of opinion among grammarians (no matter what model they represent) that the most heterogeneous, and the least understood of the traditional part-of-speech categories is, without question, the category Adverb. It has traditionally been the miscellaneous category into which those expressions were placed which could not easily be handled elsewhere in the grammar. This was done under the guise that these expressions in some sense 'modified' the verb. (p. 179) In the most recent major attempt to account for adverbs, Huang (197 5) observed that the category adverb 1 . . . can be defined ... as the part of speech satisfying the semantic definitions for adverbs but failing to exhibit the inflectional or distributional properties of nouns, verbs, and adjectives ' (p. 9) . Huang comes close to saying that an adverb is anything that isn't something else! 1

PAGE 9

2 While the situation with adverbs (ADV) might not be quite that chaotic, the category does present a number of difficult grammatical problems which stem from the diversity of the semantic concepts and syntactic structures which it covers.'*' My concern in this paper is with the question of the extent to which this 'problem area' within linguistic theory might also present problems for the language learner. There is no strong evidence connecting difficulty in linguistic description with difficulty in acquisition, but some limited experimental evidence, plus simple intuition, suggests that such a connection exists. The wide range of semantic concepts under ADV seems to cover an equally wide range of cognitive difficulty. Similarly, it is possible, or even probable, that differences exist in the cognitive difficulty of the various syntactic structures of ADV. It would not be surprising, then, to find that there were patterns in acquisition for certain types and classes of ADV, and that these patterns were related to cognitive development . General Aspects of ADV Acquisition There is very little direct evidence on children and ADVs . But from two widely disparate areas, we can say with some surety that ADV acquisition begins very early and that mastery is late. The evidence on early acquisition comes from the data Slobin (1971) reports on 'locative intentions' at the two-word

PAGE 10

3 stage. That is, with appropriate environmental/circumstantial clues, an expression such as 'pot stove' is interpreted as 'the pot is on the stove'; 'doll drawer' means 'the doll is in the drawer.' Slobin also found the 'prolocatives' here and there at this early stage. Ervin-Tripp (1970) observed the locative where in the sense of both place and direction towards in very early language (at about age 3). This evidence may not be persuasive, but it is sufficient to suggest that there is some expression of the concept of ADV in very early language. At the other end of the acquisition process, and relevant to what goes on in between, there is the evidence from a series of studies on the language production of grade school children. These studies, related in intent, but not identical in method, were reported by Harrell (1957), Loban (1963, 1964), Hunt (1965), and O'Donnell, Griffin, and Norris (1967). The conclusions that can be drawn from these studies, relative to the point at issue here, is that there is a slow but steady increase in both the number and forms of ADV used by elementary school children with increase in grade. Apparently, the acquisition of ADV is an on-going process that continues all through the school years. What I have said so far may seem a minimal defense of why anyone would want to look at ADV acquisition. One obvious reason is that no one has so far done so in any systematic way. In what follows, I would like to consider some more substantial reasons why the data on ADV acquisition

PAGE 11

4 should be of more than casual interest. But first I will reiterate what I consider to be the basic questions. The first relates to the long period required for the mastery of ADV based on the data from the language production studies. Is there some inherent (cognitive) difficulty in the basic function of ADV as process modifier? Might it be that the young child is just 'not interested' in what ADV does to a sentence, and his 'need' to use ADV develops with his increasing awareness of, or concern with, the concepts that ADV conveys? Secondly, how does the semantic diversity of those various concepts covered by ADV relate to cognitive difficulty, and so to the long period of ADV acquisition? Finally, there is the question of the extent to which the complexity of the syntactic structures under ADV affects acquisition patterns. We have no firm evidence on any of these questions or issues, but in the next section, I will discuss these points as they relate to other data on linguistic and cognitive development, especially in the early school years, the age-range of interest in the study to be reported here. The issues and items that I will consider here are for the most part at the most elemental levels of both semantics and syntax. Specifically, I will discuss the basic function of ADV; one of the most elemental of the semantic functions of ADV: time; and two aspects of elementary syntax: the ADV= ADJ+-ly rule and certain placement rules.

PAGE 12

5 The Basic Function of ADV An ADV is a 'process modifier': it sets the time frame, establishes/describes the manner, extent, etc. of a given string of words. By this definition, we really shouldn't expect very young children to be much concerned with ADV. Bronckart and Sinclair (1973) , among others, have shown that young children are much more concerned with 'outcome' than with 'process,' and so we can at least assume that, until the child has reached the 'process-awareness' point in his cognitive development, ADV will play a very small role in his linguistic behavior. There is a further aspect of this notion of 'processmodifier' that seems to relate to a well-accepted idea about cognitive development, and that is the observation that children go from concern with the concrete to an awareness of the abstract. This relates to the common conception in linguistic theory that the category ADV can be divided into 'VP-modif ier ' and 'proposition-modifier' (or S-ADV) . While I recognize that there is considerable debate over just what is ' VP-modif ier ' or S-ADV, I am sure there is general agreement that such a distinction exists. The point of interest here is that 'VP-modif ier ' is more concrete than is S-ADV, and, therefore, should be acquired earlier. In defense of thb greater complexity of S-ADV concepts, I would like to mention a distinction made by Schreiber (1970) . S claims that there are two types of S-ADV, which he calls 'modal' and 'evaluative.' With 'modal' S-ADVs, the speaker is

PAGE 13

6 expressing his 'degree of surety'; with ' evaluative' S-ADVs , he is expressing an opinion. If this interpretation is correct, and I feel that it comes very close, the child will not be able to correctly interpret S-ADVs until: 1) he reaches the point of recognizing that a comment is being made about the entire proposition; 2) in addition, he understands that this comment represents either an opinion, or just how sure the speaker is of what he is saying. I will have more to say about Schreiber's analysis in my discussion of the results of the present study. Time, Tense, and Aspect In order to assess the relative importance of semantics in the acquisition of ADV, it seems logical to look first at TIME -ADV, for the following reasons. 1) A significant subclass of ADVs relate to time; 2) time concepts are an important part of cognitive development; and 3) there are more data on this area of linguistic development than almost any other, certainly any that relate to ADV, in the age range of interest here. A number of investigators, notably Piaget (1971), Barrie -Black ley (1973), Feagans (1974), and Keller-Cohen (1975), have described the child's developing time sense as it relates to general cognitive development. The very young child lives in a ' now/not-now' world. As Ervin-Tripp (1970) pointed out, before age 5, when is the hardest of the WHquestions for children to answer. She reports the following E-child exchange as an example: 'When did he eat?' — 'Meat.'

PAGE 14

7 Young children characteristically give concrete answers to abstract questions. They are concerned only with the ' hereand-now , ' and so it seems a logical assumption that patterns in TIME-ADV acquisition would reflect this. For the English speaking child, a sense of relative time, and, presumably, a sense or awareness of tense, apparently develops out of spatial concepts. Feagans (1974) studied the relation between temporal and spatial concepts, based on a theory first put forth by H. Clark (1970). Her conclusions were that items such as before and after were comprehended in the spatial sense before they had any meaning to the child in a temporal sense. These and other temporal connectives have been studied by Amidon and Carey (1972) , Barrie-Blackley (1973), and Keller-Cohen (1975), with children in the 3-7 age range. The major conclusion of these studies, relative to the present study, is that young children do not fully understand temporal connectives. The implication is that relative time, and the associated ADVs, is in many cases incompletely understood by children in the early elementary grades. Another significant factor in the child's comprehension of relative time is the interaction of ADV with tense and aspect. First, consider what Chafe (1973) discusses as what might be called the 'necessary TIME-ADV.' Chafe gives an amusing example, which I will loosely paraphrase. 'Dennis the Menace' announces to his parents: 'Mr. Wilson broke his arm!' The parents' immediate reaction is to rush to help

PAGE 15

8 their neighbor — until Dennis' father asks 'When?' And Dennis says, 'When he was a little boy.' Chafe's point is that, to the child, if Dennis just learned about the incident, even though it happened long ago, it was 'new information' to him, and so that is how he stated it. The adult interpretation of such a statement is to regard it as 'in the present/ immediate past' unless some TIME-ADV locates it elsewhere. Still, both parents and child were dealing with 'past tense.' Nilsen (1972) notes that TIME-ADVs can express a 'time set' from distant past through distant future, with many interim points, and that specific items interact with tense and aspect. Obviously, the child has to become aware of the need, not only to 'locate' a statement in time (tense) , but to learn just how exact that location need be, and just which AD Vs can apply, given the interaction of tense and aspect . In order to look at some practical examples of how tense, aspect, and ADV interact, I would like to consider the following sentences. la. Mary has had mumps. lb. Mary has had mumps before. 2a. Mary has been pregnant. 2b. Mary has been pregnant before. 3a. John is baking great pies. 3b. John is still baking great pies. 4a. Jack Nicklaus has won the Masters. Jack Nicklaus has already won the Masters. 4b.

PAGE 16

9 In all of these S-pairs, the 'a' Ss make simple statements about past or present events. In the 'b' Ss, the ADV contributes additional implicit information, and introduces a dual time reference as well. For example, lb. implies that, if Mary is now running a fever, she could not have mumps: that disease, once contracted, provides immunity. On the other hand 2b. implies that Mary has once again 'contracted' that particular condition. S-3a. merely makes a statement about John's present activities, but 3b. presupposes a common knowledge between speaker and hearer about John's activities/abilities at some prior time. S-4b. can have more than one interpretation in terms of 'new information' and 'dual time reference.' 1) Jack won the trophy at an earlier time, and if he wins it this time, 'What's one more green blazer?' 2) Since Jack has achieved this pinnacle of success, what more could matter? 3) With only one hole to go, how could anyone catch up with him? It is interesting to note that those ADVs which introduce a dual time reference can also shift emphasis from past time to present, as in lb. and 2b. In 3b., on the other hand, the shared knowledge of the past is stressed. The situation with the multiple interpretations of already is somewhat more complicated. I will discuss this ADV-item further in Chapter IV. The major point to be made here is that words like still , before , and already introduce a double time reference. Just when is the child ready to handle this aspect of

PAGE 17

10 'relative time'? There is very little experimental evidence on this question, but 'durative aspect' at least seems to be a problem for the young child. Feagans (1974) found that children under 5 do not understand the durative aspect, and that even 7-year-olds had trouble with the durative notions of since and until . I think the evidence from the work of Bronckart and Sinclair (1973) is relevant here, even though they were working with French-speaking children. They found that children (under 9) tend not to use tense to locate an action, state, etc. in time but rather to connote aspect. That is, the children associated past tense with completion and present tense with duration, regardless of the absolute time frame. They also found a tendency for children below 6 to use intonation and gesture to refer to 'aspectual distinctions'; above 6, children tended to use adverbial phrases for this purpose. I could find no specific data on the items before and already, or on any other items in the class of relative time concepts, in the sense these notions are being discussed here. But I would strongly suspect that the child must develop well beyond the ' now/not-now' stage before he is able to comprehend items that refer to time relations between the present and the past. One final element in the interaction of tense/aspect/ adverb deserves some comment, and that is the effect of ADV in potentially generic sentences. Smith (1964) showed that the presence of TIME-ADV in a sentence can determine whether

PAGE 18

11 the sentence is generic, specific, or ambiguous. (Nilsen [1972] should also be credited with elaboration of this idea.) The sentences that follow are examples of the notions that Smith and Nilsen were considering. 5a. The tiger stalked the jungle. 5b. Long ago, the tiger stalked the jungle. 5c. The tiger stalked the jungle at noon on Thursday. 6a. The eagle ate small prey. 6b. The eagle ate small prey in times of famine. 6c. The eagle ate small prey at Bilbo's birthday party. For the 'a' sentences, the reference is ambiguous relative to 'generic.' But for the 'b' sentences, the TIME-ADV allows only an unambiguous generic interpretation. Similarly, for the 'c' sentences, there is an unambiguous non-generic interpretation. Until the child is able to comprehend ADV, or at least becomes aware that it can convey a significant part of the information in a statement, he will see no difference in the 'specific' versus the 'general' in such statements . The Syntax of ADV The points I have been discussing so far have dealt primarily with how the semantics of ADV might present problems for the language learner. But as Slobin (1971) has noted, '. . . although one can talk about order of acquisition in terms of semantic or cognitive complexity, there is clearly a point at which formal linguistic complexity also plays a role' (p. 308). The research of C. Chomsky

PAGE 19

12 (1969) and Kessel (1970) supports Slobin's notion. Both studies found that, for certain structures, there is a relation between formal complexity and time of acquisition. In what follows, I will argue that, even at the most elemental levels of syntax, ADV presents a major problem for formal description, but more importantly from the present perspective, that it might also be a syntactic problem for the language learner. It would seem that the place to start any discussion of the elementary syntax of the English ADV would be with the so-called 'basic rule': ADV=ADJ+-ly. This rule certainly seems simple enough, but there are strange and inconsistent exceptions . The rule is curious in this respect: it apparently applies, almost without exception, to all ADVs which are, in any sense, derived. It is the morphemic items to which the exceptions apply, and in no obvious fashion or systematic way. The significant point here is that this 'basic rule' applies most generally to those items that are least likely to be heard by young children: to the derived ADVs which are, for the most part, what would be considered 'advanced vocabulary. ' On the other hand, those very items with which the child is likely to come into most frequent contact are the exceptions to the rule in question. What follows is a brief examination of some of the exceptions to the 'rule' which do seem to follow some sort of pattern, but not one which the child is likely to find very helpful.

PAGE 20

13 There are certain sets of ADJ which also function as MAN -AD Vs , with no morphological change, i.e., without '+-ly.' But for no discernible reason, these items fall into 3 2 classes. Consider the words in Table 1. TABLE 1 Adjectives Which Function as Manner Adverbs Without '+-ly' A B C ill bright deep left fair (unbiased) fair right (direction) loud fast right ( ' directly ' ) quick hard well right ('correct') high still sharp strong straight soft wide For items in A, no Â’-lyÂ’ is possible: 7. Michael pleaded the case well (*-ly) . 8. He turned the car left (*-ly) . The items in B may take '-ly' optionally: 9. The sun shone bright (-ly) . 10. The wind blew soft (-ly) . There are, however, some restrictions on the items in this 'optional' category with respect to freedom of placement. 11. The sun brightly shone. *12. The sun bright shone. 13. The boy quickly ran down the street. *14. The boy quick ran down the street. Apparently the optionality of '-ly' for these items applies only when they are in post-V position. (I will discuss

PAGE 21

14 placement rules more fully in a later section.) All items in C can take '-ly,' but this results in a change in class. For the most part, MAN goes to EXT. Just for this small set of relatively common words, which must surely become part of the child's vocabulary rather early, there is a bewildering array of exceptions to a very basic rule. There seems to be no pattern to these exceptions, and so it is an interesting experimental question as to what order and at what age they are mastered. Another peculiar phenomenon, similar to that just discussed, involves verb-adverb phrases. Nilsen (1972) comments that : Sometimes there is a rather tight affinity between the Manner Adverbials . . . and a verb, making it so that the V-ADV m construction is functioning as a sort of a Predicate plus Complement. (p. 81) He gives an extensive list of such pairs, but makes no mention of how they might be subdivided as a function of the 3 applicability of the '-ly' rule. In Table 2 I have listed some of the pairs from Nilsen' s list which seem to function differently with regard to the rule. The items in A cannot take '-ly'; those in B can, and still remain post-V. For the items in C and D, '-ly' is possible, but with a change in meaning. For D, the '-ly' forms can remain post-V; for C, they cannot. These pairs are all, as Nilsen noted, in some sense 'locked phrases,' even 'idiomatic' to some extent (or rather, in some cases). But it is interesting to note how differently '+-ly' applies to the different subclasses.

PAGE 22

15 TABLE 2 Verb-Adverb Phrases Which Function as Predicate Plus Complement A B C D blow shut shine clear make clear work loose sit tight stand firm make happy prove false turn red rest easy remain loyal prove true make fast run slow remain ignorant break even grow tall glow bright set free continue warm keep young lie flat cut short come clean fall sick blow open It is beyond the scope of this paper to try to explain these phenomenona. I mention it only as further evidence of just how difficult it must be to master the English ADV. The one final comment I have to make on the vagaries of the ' -ly' rule regards a small set of ADJs ending in 1 y. ' They are presented in Table 3, column A, and come from a list in Nilsen (1972, p. 80) labeled 'No Adverb Possible.' I agree with Nilsen that the items in A cannot take ' H — ly, ' at least, they sound very strange. And yet the items in B are also ADJ ending in 'y,' and the associated ADVs sound normal. I can find no semantic, syntactic, or phonological reason why, for example, cloudily should sound odd, but handily is common and accepted.

PAGE 23

16 TABLE 3 Adjectives Ending in 'y' Which Can and Cannot Take ' +ly ' A B baggy angry cloudy busy dirty clumsy dusty handy funny happy ugly heavy noisy The second area of elemental syntax that I would like to consider is the rules of ADV placement. In general, ADV placement in English is very free, as can be seen from Figure 1 on page 17. At the least, most items from most ADV classes can appear anywhere in the S. The apparent exception--the post-V position--is sometimes acceptable (for certain items) when the VP is sufficiently long, as in S-15. 15. John reads thoroughly those books which are related to his profession. But there are definite restrictions on ADV placement relating to morphology, ADV class, and, to a certain extent, the problem of ambiguity. The morphological restrictions on placement relate to the items discussed in the previous section, that is, those MAN-ADVs which are of the same phonetic form as their ADJ counterparts. For example, with the items soft and bright , both forms are possible in one sentence position, but not in the other.

PAGE 24

17 """S >i >i >1 i — 1 H • • 1 — 1 (D -P C £>i i 1 M CD 0 fa * * * * * .. ^ read < *N >i k^I r— 1 r— 1 rH -P £ 1 — 1 w d m Ifl tj* O 44 d c n o •d >i rH rH rl -P C i — 1 aj q a w I cn S fa M E-t

PAGE 25

18 16. Sam sang the song soft. 17. Sam sang the song softly. 18. Sam softly sang the song. *19. Sam soft sang the song. 20. The lamp glowed bright. 21. The lamp glowed brightly. 22. The lamp brightly glowed. *23. The lamp bright glowed. But with items like hard , there is a different 'rule' : 24. Sam hit the nail hard. *25. Sam hard hit the nail. *26. Sam hit the nail hardly. *27. Sam hardly hit the nail. S-27 is, of course, acceptable, but only with a change in ADV class: from MAN to EXT. Placement rules and restrictions which relate to ADV class and the problem of ambiguity are syntactically very interesting 28. I . Consider the following Ss: naturally read the letter. S-ADV 29. I read the letter naturally. MAN 30. I read the letter, naturally. S-ADV 31. I discussed the subject frankly. MAN 32. Frankly, I discussed the subject. S-ADV 33. I frankly discussed the subject. Ambig. : 34. John cooks rice regularly. Ambig . : 35. John regularly cooks rice. FREQ. There seem to be no logical reasons for the placement

PAGE 26

19 restrictions which apply to ADV as a function of class, or to those which disambiguate. It would seem that the child simply has to memorize all such exceptions. But Nilsen has contributed what may be a valuable insight on this particular 'problem' for the language learner. He notes that there are, for certain ADV classes, positions within the S which are 'more natural.' And so, by his analysis, S-final is 'more natural' for MAN-ADV, and 'pre-V' is 'more natural' for FREQ-ADV. The trouble is that, at least for the Ss I have chosen as examples of 'the problem,' his solution doesn't seem to help very much. In comparing S-33 with S-31, it is true that it is possible to disambiguate, but S-33 is a perfectly good way of expressing the notion of S-31. With the other ambiguous S, S-34, compare it with S-36, which is perfectly acceptable. 36. John cooks rice frequently. The point is that, what Nilsen suggests as to 'natural placement' may well be an idea worth pursuing in the area of ADV acquisition. However at this point it does not seem likely that it will be a 'hard and fast' rule. There are few such rules in the area of language. The Present Study In the previous section I have shown some of the syntactic and semantic reasons why ADV is a difficult category for the English-speaking child to master. There is evidence that certain items in this category appear in very early

PAGE 27

20 language, but that mastery of the category is very late. In sum, ADV is a complex linguistic category, for the child and for the linguist, and there is good reason to believe that the child's acquisition of the category ADV is closely related to cognitive development. Purpose The purpose of the present study is to make a systematic investigation of what seem to be the basic aspects of ADV acquisition. The most fundamental questions are, of course, those put forth at the beginning of this chapter: 1. To what extent does the basic function of ADV interact with cognitive development? 2. What is the relation between ADV class and cognitive difficulty? 3. What is the relation between differences in syntactic structures of ADV and cognitive difficulty? Specifics of the Present Study Since this study is the first attempt at evaluating ADV acquisition in a systematic way, certain decisions had to be made as to how best to approach the questions presented above. It seemed that the most fundamental comparison of syntactic structures would be between one-word items that differed only in terms of 'morphemic' or 'ADJ+-ly.' Semantically, the decisions were somewhat more complex, but 5 classes were chosen for the following reasons: MAN--There is some scant evidence that items in this

PAGE 28

21 class appear early. There are also logical reasons why MAN-ADVs seem to be more closely associated with their ADJ counterparts. They are, as a class, associated with the (presumably) simpler concept of VP-modifier. REL. TIME/FREQ--These items relate to the one area in which there are some data on children and ADV, indicating that TIME is a significant factor in ADV acquisition in the early school years. EXT--There is evidence that this concept grows out of earlier concepts of TIME and SPACE, and so it is potentially interesting in this period of language development. S-ADV--There is good reason to believe this class covers the most complex of ADV concepts, and so might be one of the last to be mastered. The Questions Given the above criteria, the particular questions addressed in this study were: 1. Are there differences in time of acquisition as a function of the two syntactic forms: 'morphemic' and 'ADJ+-ly'? 2. Are there differences in time of acquisition as a function of the five semantic classes: MAN, EXT, REL. TIME, FREQ, and S-ADV? 3. Is there evidence for a relation between linguistic and cognitive development in relation to ADV acquisition?

PAGE 29

22 Notes The terms 'adverb' and 'adverbial' are not used consistently in the literature. Both terms can cover both items and phrases. Some writers make a distinction between adverbial phrases and one-word adverbs. In this paper the more traditional 'adverb' (ADV) will be used to refer to all forms, with the item/phrase distinction made clear where pertinent. In the study proper, only one-word ADVs are considered. 2 The tables in this chapter are not inclusive. That is, I have not attempted to find all possible items which might fit any particular cell. 3 Nilsen (1972, p. 81) acknowledges that this list and the observations on this relationship should be credited to Fries (1952) and Fraser (1965) .

PAGE 30

CHAPTER II THE STUDY The Method This study deals with an area that is both broad and ill-defined. Even though the restrictions and specific questions stated in the previous chapter impose some limitations, certain basic decisions remain as to the best way to approach problems in this area. The first and most important is whether to look at language from the perspective of production or of comprehension. For this particular question, a brief examination of the available lists on child language, in particular, Wepman and Hass (1969) and Rinsland (1945) , showed that the percentage of ADVs in child language output is so small that a production study was impractical. Moreover, a study of comprehension seemed to have two further advantages. It is somehow more realistic and it is easier to control. Comprehension data are more 'realistic' than production data in this age range in the sense that what we are concerned with here is just when a child understands the effect of ADV in a S. And a comprehension study is easier to control because the child's attention can be directed to just those linguistic situations in which ADV is critical. ^ 23

PAGE 31

24 The remaining decision, then, was how best to measure comprehension of ADV as a 'process modifier.' It seemed that the logical choice was a task that comes closest to everyday language, and, by this criterion, the picture-sentence matching task was most appropriate. This task is very similar to the 'real-life' situation, in which the child must respond to what he hears, in the way he currently (at his present acquisition level) understands it. The Subjects The subjects for this study were students at the P. K. Yonge Laboratory School at the University of Florida. The final analysis was made on the data from 15 first graders and 20 children in each of grades 2, 3, and 5. Eight eighth graders were also tested, since this age group could be considered 'linguistic adults.' Their data represent a 'test of the test,' that is, a demonstration that, for adults, who presumably have full command of the English ADV system, this test is a simple linguistic exercise. There were some children tested whose data were not included in the analysis. For example, all the children whose data were included were white. A few black children were tested for social reasons, but since there is no evidence either way on the effect of the Southern black dialects on performance in language acquisition studies, it 2 did not seem appropriate to include them in this study.

PAGE 32

25 There were several children I tried to test, but who, for one reason or another, were unwilling or unable to cooperate. This was clear by the end of the 6 practice items, and for them, the session ended there. This group includes a few kindergarteners: it seems that this type of test is inappropriate below age 6, and I suspect the major problem is attention span. Finally, the male/female division is not equal for all grades. The breakdown of sex by grade is presented in Table 4. TABLE 4 Subjects by Grade and Sex Grade Male Female 1 6 9 2 10 10 3 10 10 5 10 10 8 3 5 The reason that there are fewer males than female subjects in Grade 1 is that the girls were generally more cooperative, and the boys more likely to be intimidated by the test situation at this age. For Grade 8, there did not seem to by any good reason to assume male/female differences would be relevant, and so no attempt was made to balance for sex in that age group.

PAGE 33

26 The Test As was mentioned earlier, it seemed that the best method for evaluating the present problem was a sentence-picture matching task, especially for the age range of interest here. But with the added complication of assessing 'process modifier,' some elaboration of the usual method was necessary. So instead of contrasting two or more single-frame pictures against a S, two panels of 4 frames, in the manner of newspaper cartoons, were constructed for each test S. An additional advantage of this approach was that it allowed me to use familiar cartoon characters which would be more likely to catch the childrens' interest. Given the restrictions and the particular questions stated in Chapter I, specific items for the 5 ADV classes and 2 syntactic forms had to be selected which were arguably items a young child would hear relatively often, or could be expected to use relatively early, by some objective criterion The choice of items was made after consulting the lists formulated by Jones and Wepman (1966) , Wepman and Hass (1969) and Rinsland (1945) . But all such lists are based on production, and more importantly, Wepman ' s lists are restricted to responses to the TAT, where the possibility of ADV 3 occurring is limited. The problem with Rinsland 's lists is that he makes no distinction between syntactic categories. So the final choice of items for this test was based on what these lists suggested, plus a large degree of intuition.

PAGE 34

27 • Once the 24 ADV-items had been chosen, a S was constructed for each in a standard form: NP-ADV-VP. The exception to this format was that for morphemic-MAN items the ADV had to appear at the end of the S, as was discussed in Chapter I. Also, since S-ADVs seem most natural in S-initial position, that is where they occur in the test Ss here. The ADV items by form and class, and their position in the test Ss, are presented in Table 5. The list of test Ss can be found in Appendices A and B. The basic form of the test was a page containing the two 4-frame cartoon panels (which will be referred to as 'panel-sets'), with the panels placed one above the other. The panels were prominently labeled A (above) and B. Each page corresponded to one of the test Ss. The sequences in the panels were variations of familiar cartoon characters, and depicted the events, actions, or situation in the corresponding S. The sole difference between the panels on a given page was the applicability of the ADV in the corresponding S. Thirty panel-sets/pages were constructed, one for each of the ADV test items, plus 6 which were used for a practice session. For the 6 practice pages, the panels differed with respect to the appropriate subject, verb, or AD J , but never ADV. The purpose of these pages was to make certain that, before beginning the test proper, the child fully understood the task, i.e., of selecting which of the two panels best fit the S he heard.

PAGE 35

28 TABLE 5 Adverb Items by Class and Form with Sentence Frames Form Sentence Frames Class ADJ+ly Morphemic MAN carefully quietly quickly happily hard high fast straight ADJ+ly : Subj-ADV-Verb' ed-Obj Morphemic : Sub j -Verb ' ed-Obj -ADV EXT nearly completely almost only Sub j -ADV-Verb ' ed-Obj REL. TIME finally suddenly already still Sub j -ADV-Verb ' ed-Obj FREQ rarely usually seldom often Subj-ADV-Verb ' ed-Obj S-ADV certainly probably perhaps maybe ADV-Sub j -Verb ' ed-Obj The Ss were presented in semi-random order. That is, they were arranged so that the same subject or verb never appeared in adjacent Ss, and the class and form of the ADV occurred in reasonably random order. The 'right answer,' i.e., the A or B panel, was truly random. Since there was some question that particular lexical items might interact with the panel-sets with which they were associated, a second form of the test was arranged by interchanging antonyms and synonyms. For example, the panel-set for rarely on Form 1 was used for usually on Form 2, since the sequence depicted in that panel-set was equally appropriate to either ADV, but the 'right' answer would be the opposite panel. This interchange was possible for only 14 of the 24 items, but that seemed sufficient to be able to answer the question of

PAGE 36

29 possible interaction. The Ss for both forms of the test are presented in Appendices A and B. The panel sets were presented in the same order for all subjects; the ordering of the Ss was changed accordingly for the two forms. Procedure One of the special advantages of conducting research at the P. K. Yonge School is that each classroom has an adjacent teacher's office that is available for testing sessions. The rooms have child-size tables and chairs, and so for each session, the child and I could sit side by side, with the notebook of panel-sets before us. At the beginning of the session, I opened the notebook to a page on which were drawn all of the cartoon characters which would appear in the panel sets. Since these were familiar characters, I asked the child if he could name them. Most children above first grade could, and seemed to enjoy doing so. Even the first graders knew at least some of the 'people.' It should be stressed, though, that this 'I.D.' page was intended only as a means of catching the child's interest. In no case in the test that followed was it critical that the child know the name of any particular character. After discussing the characters, I told the child I would be showing him some pictures with the 'people' in them and would be asking him questions about the pictures. I turned to the first practice page and explained that all

PAGE 37

30 of the pages would be the same, with one panel above marked A and one below marked B. We then went through the 6 practice pages. For each page, I said, 'If I said (relevant S) , which panel do you think would go with that, A or B? ' No matter what his answer, I told him which was the right panel, and we discussed why. If at the end of the 6 practice pages, the child did not obviously understand the task, giving the right answer and correctly explaining why, that child was not included in the study. That is, in no case were data included if there was any question as to whether the child understood the task. At the end of the practice session, I told the child that the rest of the pages would be like the ones we had just seen, but I would no longer tell him the right answer, ' . . . because I want to know what you think is the right answer.' In all cases, no matter whether the answer was right or wrong, I responded with a 'noncommittal,' but 'positive-sounding' 'OK, let's look at the next page.' In addition to recording which panel the child selected, I also noted any voluntary comments that were made, and, occasionally, but in no systematic manner, asked, 'Why do you think that is the right panel?' In many cases, the children offered comments or explanations wiÂ’thout being asked . The eighth graders were an exception to the procedure just described. For every item, they were asked to explain why they chose as they did, since, as I mentioned earlier.

PAGE 38

31 their data were to be a measure of the validity of the test . The overall response of the children to the test was very encouraging. It was not uncommon for a child to ask when the test was over, 'Is that all? Aren't there any more?' To the extent that 'subject involvement' relates to test reliability, this evidence reflects favorably on the results that I will report in the next chapter.

PAGE 39

32 Notes The question of whether comprehension precedes production is not at issue here. In fact, it may well be a nonissue. Early production of a linguistic form does not indicate full comprehension, nor does the reverse hold. For this paper, what is important is determining whether there are progressive stages, or patterns, in ADV acquisition, and what applies to comprehension should, in a general way, apply equally to production. But that is still an empirical question, and will be discussed further in Chapter IV. 2 An examination of the data from those few black children indicated that they tended to score well below their white peers. I have no explanation for this, but there were strong indications that dialect differences were a factor. This, I feel, supports my early decision not to include these children in the study/analysis. ^The Thematic Apperception Test (TAT) (Murray, 1943) is a series of pictures, each depicting a specific situation, which is widely used to elicit language samples for various purposes. But since the pictures are essentially 'static,' there are few cues which could be expected to elicit ADVs.

PAGE 40

CHAPTER III RESULTS Preface The observations from this study fall into two categories: 1) the objective measures, i.e., the error scores for each test item, and 2) the voluntary comments and other responses the children made in relation to the individual test items. Only the error scores were analyzed statistically, and the results of those analyses will be considered first. The children's comments and other responses, and the relevance of these data to the error data, will be discussed later . The Error Scores The objective results of the study are presented in Table 6. The mean error scores represent the proportion of trials in which an incorrect response was given. The scores are listed for each grade and each test item, with an indication of the divisions by ADV class and form. Table 7 gives the means for ADV class and form, across individual items. These are the data on which the final analyses were run. An analysis of variance (ANOVA) was conducted to determine the effects of the several factors in this study. Since 33

PAGE 41

34 TABLE 6 X Error Scores by Test Item by Grade Adverb Class/Form Test Item Grade 1 N=15 2 N=2 0 3 N=2 0 5 N=2 0 Manner carefully 0.27 0.20 0.10 0.00 quietly 0.07 0.00 0.05 0.00 -ly quickly 0.00 0.15 0.00 0.05 happily 0.33 0.20 0.15 0. 10 hard 0.00 0.05 0.05 0.00 morph. high 0.00 0.00 0.00 0.00 fast 0.13 0.00 0.10 0.00 straight 0.07 0.10 0.00 0.05 Extent nearly 0.73 0.20 0.25 0.05 -ly completely 0.13 0.15 0.20 0.00 morph. almost 0.53 0.30 0.20 0.00 only 0.53 0.40 0.45 0.25 Rel. Time finally 0.47 0.55 0.45 0.20 -ly suddenly 0.60 0.40 0.45 0.25 morph. already 0.20 0.35 0.10 0.10 still 0.20 0.15 0.15 0.00 Frequency rarely 0.73 0.35 0.40 0.15 -ly usually 0.33 0.10 0.05 0.00 morph. seldom 0.87 0.90 0.80 0.45 often 0.33 0.20 0.15 0.10 Sentential certainly 0.40 0.20 0.20 0.15 -ly probably 0.80 0.70 0.80 0.80 morph. perhaps 0.93 0.85 0.70 0.65 maybe 0.80 0.90 0.80 0.55

PAGE 42

35 TABLE 7 X Error Scores for Adverb Form and Class by Grade Adverb Class Rel . Grade Adv. Form Man . Ext . Time Freq . Sen. XX 1 -ly 0.17 0.43 0.53 0.53 0.60 0.45 morph. 0.05 0.53 0.20 0.60 0.87 0.45 X 0.11 0.48 0.36 0.56 0.74 0.45 2 -ly 0.14 0.18 0.48 0.22 0.45 0.29 morph . 0.04 0.35 0.25 0.55 0.88 0.41 X 0.09 0.26 0.36 0.38 0.66 0.35 3 -ly 0.08 0.22 0.45 0.22 0.50 0.29 morph. 0.04 0.32 0.12 0.48 0.75 0.34 X 0. 06 0.27 0.28 0.35 0.62 0.32 5 -ly 0.04 0.02 0.22 0.08 0.48 0.17 morph. 0.01 0.12 0.05 0.28 0.60 0.21 X 0.02 0.07 0.14 0.18 0.54 0.19 X -ly 0.10 0.20 0.41 0.25 0.50 X morph. 0.03 0.32 0.15 0.47 0.77 XX 0.07 0.26 0.28 0.36 0.63

PAGE 43

36 the design was unbalanced, i.e., the N for Grade 1 was less than that for the other three grades, the ANOVA was conducted by means of a general linear model: a regression analysis. In the course of this analysis, the main effects (test form, sex, grade, ADV form, and ADV class) and all possible combinations of 2-, 3-, and 4-way interactions were tested. Any main effect or interaction that was not significant was dropped from the model. Those terms that appear to be significant (“ = 0.01) are reported in Table 8. As a result, the ANOVA indicated that the factors of grade, ADV form, and ADV class, as well as the interaction of form by class, are the only variables that are significantly related to the responses to the test items. But the ANOVA results say only that some means are significantly different for any variable. Therefore the Duncan Multiple Range Test was used to identify just where the real, or statistically significant differences lay. This test was run to ascertain which grades were actually different from each other, and where the differences were in the ADV form by ADV class interaction. The results of these tests, along with the graphic representations, are shown in Figures 2 and 3, respectively. It seems clear that the ability to process the ADVs included in this study improves significantly between Grades 1 and 2 and between Grades 3 and 5. The significant differences for ADV form and ADV class are complicated by the form by class interaction, and that subject requires some clarification.

PAGE 44

37 TABLE 8 ANOVA for Grade, Adverb Form and Adverb Class Source S.S. df m. s . F Grade 6.19 3 2.06 29.33* ADV Form 0.59 1 0.59 8.36* ADV Class 25.30 4 6.32 89.87* Form X Class 7.14 4 1.78 25.35* Error 51.86 737 0.07 — — *significant at = 0.01 Although the statistical results/ANOVA indicate that both ADV form and ADV class are significant factors in the acquisition of the particular items included in this study, the significant interaction of form by class means that the differences in these main effects can only be considered as trends, and that any real differences for these two variables are masked. That is why the Duncan Test results take on special importance. In Table 9, the individual test items are ranked by mean score and Duncan grouping. This makes it possible to compare means, both for items within cells and for items across cells. Thus the table points up the fact that differences between means for individual items may be an important factor in the interpretation of the significant interaction of ADV form by ADV class. Before considering the implications of that significant interaction, we can look at what seem to be the most obvious conclusions to be drawn from this table. Clearly S-ADV items

PAGE 45

10 9 8 7 6 5 4 3 2 \ H 38 Duncan Multiple Range Test a Grade X Group 1 0.45 A 2 0.35 3 0.32 5 0.19 C a-r. X's which do not share a common letter in the Â’ouping are significantly different; those which do e not (<* = 0.01) . 2 . X errors by grade with associated Duncan Multiple Range Test results. CO CO

PAGE 46

100 90 80 70 60 50 40 30 2 0 10 CLAS gr< ar Fig 39 Duncan Multiple Range Test' ADV Class S-ADV S-ADV FREQ ADV Form morph. -ly morph. REL. TIME -ly EXT FREQ morph. -ly X 0.77 0.50 0.47 0.41 0.32 0.25 Grouping A B B BC CD DE MAN EXT REL. TIME FREQ S-ADV X's which do not share a common letter in the ping are significantly different; those which do not (<* = 0.01). 3. X errors by ADV form and ADV class with associated Duncan Multiple Range Test results.

PAGE 47

40 TABLE 9 Comparison of Duncan Groupings with Item X's Duncan 3 X Error Class-Form X's Grouping I terns Scores S-ADV-Morph . 0.77 A perhaps 0.78 maybe 0.76 S-ADV-ly 0.50 B probably 0.78 certainly 0.24 FREQ-Morph. 0.47 B seldom 0.76 often 0.20 REL . TIME-ly 0.41 BC finally 0.42 suddenly 0.42 EXT-Morph. 0.32 CD almost 0.26 only 0.41 FREQ-ly 0.25 DE rarely 0.41 usually 0.12 EXT-ly 0.20 EF nearly 0.31 completely 0.12 REL. TIME0.15 EF already 0.19 Morph . still 0.12 MAN-ly 0.10 FG carefully 0.14 quietly 0.03 quickly 0.05 happily 0.20 MAN-Morph . 0.03 G hard 0.02 high 0.00 fast 0.06 straight 0.06 a77, ^ — — X's which do not share a common letter in the grouping are significantly different; those which do are not (°c=0.01).

PAGE 48

41 were the most difficult in this test, 1 and MAN, at least, morphemic MAN, were easiest. All that can reasonably be stated in terms of ADV form and class is in the sense of trends, and these are presented in Figures 4 and 5. The overall pattern in Figure 4 shows that morphemic differences are not obvious across the grades. On the other hand. Figure 5 shows some very clear differences as a function of ADV class. This apparent contradiction between statistics and graphic presentation can be explained, at least in part, by a more careful examination of Table 9. If the Duncan groupings are compared with the item means (see especially certainly , which I have already mentioned) , it is apparent that in many cases specific item differences within class/form cells can be quite large. This argues for two tentative conclusions: first, that ADV form is not, in and of itself, a significant factor, but rather it interacts in a special way with the items in this test; second, no firm statements can be made relative to ADV class other than S-ADV and MAN. However, the possible implications of the results for the middle-difficulty items will be discussed in Chapter IV, in conjunction with data to be reported later. During the spring of the previous school year, 16 of the third graders included in this study had been given the OtisLennon Mental Ability Test (OLMAT) . The scores for this test are reported as percentile rankings relative to national norms The test is group-administered, and not particularly sensitive

PAGE 49

Percent Errors 42 Fig. 4. Percent errors by grade as a function of ADV form.

PAGE 50

100 90 80 70 60 50 40 30 ' 20 ' 10 ' 43 . 5 . Percent errors by grade as a function of ADV class.

PAGE 51

44 but it does provide a reasonably reliable ranking of the child's mental ability. An analysis of covariance (ANOCOVA) was conducted on the two sets of scores available for this grade: the error scores from the present study and the OLMAT percentile ranks. The results are presented in Table 10. Apparently, mental aptitude is a significant factor in the ability to process the ADV items in this study for children of this age. The results indicate that, while there is no apparent effect of ADV form when mental ability is one factor in the analysis, ADV class is still a significant variable, and the interaction of ADV form by class still holds, as it did in the ANOVA . TABLE 10 Analysis of Covariance for OLMAT Scores, Adverb Form and Adverb Class Source S.S. df m. s . F OLMAT Score 0.80 1 0.80 10.46* ADV Form 0.14 1 0.14 1.84 ADV Class 5.16 4 1.29 16.85* Form X Class 1.99 4 0.50 6.51* Error 11.41 149 0.08 *significant at = 0.01 Because of this significant interaction, it is necessary to discuss the relation between the two significant main variables, ADV class and OLMAT score, separately for each of the five ADV classes. In Figure 6, regression lines have been plotted for each class as a function of OLMAT

PAGE 52

45 score. A regression line, it should be noted, is a straight line of best fit for data points plotted on two variables, and the slope of the line reflects degree of change in one variable relative to change in the other. Figure 6 indicates only slight differences in MAN and S-ADV scores as a function of mental aptitude. This merely reinforces the ANOVA results that MAN is already 'acquired' by Grade 1, and that S-ADV is almost as much of a problem at Grade 5 as it is for the younger children. For the 'middle-difficulty' classes, the graphic representations suggest that ability to handle FREQ does not change much as a function of general mental ability, but that there is a somewhat more obvious change for EXT and REL. TIME. It was noted in Chapter II that 8 eighth graders were tested in order to establish an 'upper-limit' for the test. Since these 13-14-year-olds could reasonably be considered 'linguistic adults,' their data were intended as a 'test of the test' : that what was at issue here was really an ability to process ADVs. All of the subjects in this group were asked to explain their reasons for choosing a particular panel. Two subjects gave the correct responses and explanations for all items. One subject gave an aberrant but very interesting pattern of responses, and he will be discussed later. The remaining five subjects gave correct answers and explanations to all but one or more of the critical S-ADVs (i.e. , excepting certainly , which all correctly explained) . The eighth graders did, in fact, 'validate' the test as a

PAGE 53

Percent Errors 46 100 75 50 25 0 HU 40 MAN 50 60 70 80 90 100 100 75 50 25 0 100 75 50 25 0 L-Uf 40 KV440 50 60 70 80 H 1 1 1 50 60 70 80 OLMAT Percentile Scores FREQ — t 1 90 100 • S-ADV 90 100 Fig. 6. Percent errors as a function of OLMAT percentile score for the five classes of ADV.

PAGE 54

47 task involving the ability to process ADVs. The fact that so many of them had trouble with at least one of the S-ADVs is particularly interesting when compared with their comments. That issue will be treated in a later section. The Voluntary Data The second form of data that this study produced was the comments and other voluntary responses that the children gave. These data were not an integral part of the experimental design, but all comments and reactions the children made during the test session were noted. For the most part, these data can be accurately described as 'voluntary responses. ' However, some children were asked to explain their choices, but this was not done in any systematic way. Even for those children who were asked to explain, I did not insist that they do so. These spontaneous or solicited responses cannot be taken as anything more than corroborating evidence for the major statistical results, and the analysis of these data was as loose as the method by which they were collected. These data are primarily 'impressionistic.' Nevertheless, there seem to be some definite and intriguing patterns in these responses, and they will be reported in terms of certain patterns or tendencies which seem to change with grade. For Grade 1, the apparent tendency is to attend primarily to the verb. Typical comments for average-score subjects in this age group are presented in Table 11. It should be

PAGE 55

48 TABLE 11 Sample Comments by First Graders Snoopy carefully put the bird in the next. 'because he's lifting it up' 'because he put it in (B-3) ; there (A-3) blew it out' Snoopy quickly ran to kick the ball. 'because he's kicking it' Lucy happily listened to Schroeder play. * 'because she's listening to it' * 'because she's looking at him in B and listening in A' Snoopy nearly fell off his doghouse. *' (pointing emphatically to A4) where it did happen' The fire completely burned Snoopy's doghouse. 'because it's burning his house' Linus only hit the ball. * ' he ' s just hitting the ball' -subject repeated the S without ADV Linus finally found his blanket. * 'because he found it' * ' both the same' Linus seldom helps his mother( A1 ) 'because he's vacuuming' * ' he ' s doing laundry and stuff' (indicated B-l, B-2, B-4) Lucy rarely plays with Linus. *' because (in three frames) she's playing' * = wrong response ( A— 1 ) refers to the first frame of panel A, etc. noted that a response of 'they are both the same' was considered a wrong response, i.e., an indication that the subject was not processing the ADV, since the panels in a set differed only with respect to the ADV. There was also a strong tendency for these children to point to the critically wrong frame in a panel for S-ADV panel-sets and other 'implicit NEGs,' and to do so emphatically. This issue will be discussed in more detail in the next section.

PAGE 56

49 The most prominent change in the comments by average scorers in Grade 2, compared to Grade 1, was their tendency to mention the ADV, usually as a question after the test-S had been read to them, but also as the 'reason' for their response. Also, there were more correct explanations of right responses, occasionally with correct definitions of the ADV-item. This was still infrequent, but certainly an increase over the average first grader. But there was still a definite indication of 'verb-processing' and 'both the same' responses, especially for those items that were 'hard,' as indicated by the error score data. The change in pattern from Grade 2 to Grade 3 is primarily one of degree. Here the children more frequently mentioned the ADV, either repeating it to themselves, asking what it meant, or using it as the reason for their responses. By Grade 5, almost invariably, if the children made a comment, it was to repeat or question the ADV. For the most part, they volunteered, or could give on request, good explanations for their choices. But one other notable change for this age group was that, as error scores decreased, there was a decreased likelihood of any comment being offered or mention of the ADV being made. Evidently, as 'processing ADV' becomes more automatic, 'ADV awareness' decreases.

PAGE 57

50 The Relation between the Two Forms of Data The two forms of data collected in this study tend to be mutually supportive, but in one area more than others. Certain items in this test can be classified as 'implicit NEG, ' i.e., if that item were deleted, the test-S would go from a negative to a positive statement, as in the following: 37a. Snoopy nearly fell off the doghouse. 37b. Snoopy fell off the doghouse. 38a. Linus rarely helps his mother. 38b. Linus helps his mother. 39a. Probably Snoopy grabbed the blanket. 39b. Snoopy grabbed the blanket. These three types of 'implicit NEG' are, of course, not equal in degree or type of negative, and that distinction will be discussed in the next chapter. But if it is true, as the comments from the first graders seem to indicate, that, early on, the strategy is to 'process the S without ADV , ' then we might expect to find: 1) higher than average error scores on these items for the young children, and 2) that errors would decrease across grade. In addition, the comments of older children on these items should give some indication of the process of ADV acquisition. The items which fall under this category are presented in Table 12, along with mean error scores for Grades 1 through 5. The eighth graders made no errors on the non-S-ADV items, and their scores on the S-ADV items will be reported with

PAGE 58

51 their comments. The trends in the error scores for these items are presented in Figures 7 and 8. TABLE 12 Error Scores by Grade for 'Implicit NEG' Items Class Item 1 2 3 5 EXT nearly 0.73 0.20 0.25 0.05 almost 0.53 0.30 0.20 0.00 FREQ rarely 0.73 0.35 0.40 0.15 seldom 0.87 0.90 0.80 0 .45 S-ADV probably 0.80 0.70 0.80 0.80 perhaps 0.93 0.85 0.70 0.65 maybe 0.80 0.90 0.80 0.55 Although each of the ADV classes covered here is arguably 'implicit NEG,' each class differs subtly in 'degree' or 'type' of negative. For example, the EXT items are true NEG, i.e., S-40 is positive, while S-41 is negative: 40. Snoopy fell off the doghouse. 41. Snoopy nearly fell off the doghouse. The FREQ items have a less obvious NEG effect, though for the linguistic decisions required in this test, they also function as implicit NEG: the correct panel-set for the test-S for these items showed an event happening once versus three times. The S— ADVs are NEG in an even more subtle sense. Since they represent 'speaker doubt' or degree of surety,' the statement that followed could not be judged positive. By default, the 'positive' panel was the wrong answer .

PAGE 59

Percent Errors 52 Fig. 7. X errors by grade for EXT and FREQ items.

PAGE 60

100 90 80 70 60 50 40 30 20 10 0 Gr 53 2 3 5 Â’ig. 8. X errors for S-ADV items.

PAGE 61

54 Because of the differences in relation to the notion of 'implicit NEG, 1 the S-ADVs will be discussed separately from EXT and FREQ. Only Grades 1-5 will be considered for EXT and FREQ, since all eighth graders of interest here correctly explained these items. In Table 11, presented earlier, comments made by first graders were listed as evidence for 'processing verb.' The comments on the EXT and FREQ items listed there are consistent with the error scores for this age group and support the notion that if the ADV is omitted in processing S, there is a greater likelihood of a wrong response. For Grades 2 and 3, the error rate drops considerably for all items except seldom . For the other items, rarely , almost , and nearly , the patterns in the voluntary responses are consistent with those mentioned earlier for all items. The most obvious change from the first grade pattern is for the children to mention the ADV, either aloud but to themselves, or as a question, or to simply ask what the word means. There seems to be some correlation between this type of response and low error scores, but, of course, this is only an impression, since these data are not amenable to statistical analysis. But there can be no question that in this age range, the voluntary comments indicate an awareness that ADV is a factor in processing the S, and this would corroborate the significant decline in errors from Grade 1 to Grade 2. For the item seldom, where the error scores do not decline appreciably until Grade 5, the outstanding

PAGE 62

55 observation was that the children in Grades 2 and 3 took an unusually long time to respond to this test-S, whether they eventually gave a right or wrong answer. The children with high error scores tended to ask what the word meant. As with the other items being discussed here, their comments tend to reinforce the notion of 'emerging ADV awareness' in this age group, or at the least, an awareness that all elements in the string must be processed. The comments from Grade 5 are relevant in an obverse way: they tended not to volunteer any comments on these items (the exception again is seldom , where they also tended to ask what the word meant, though, of course, less often than the younger children) . The reason why this particular finding is relevant is that it supports what I have suggested earlier: once the items have been mastered, as evidenced by the error scores, the children become less 'aware' of the fact that they are 'processing ADVs,' and consequently do not mention them. The comments on S-ADVs are less clear-cut than those for EXT and FREQ, possibly because there is no dramatic change in error scores for these items until Grade 8. But there are still some interesting implications in the comments on these items. The changes in the comments on S-ADVs for Grades 1, 2, and 3 are not notably different from those this age group made to the other 'implicit NEG' items. In general, the comments support the high error scores as 'processing S

PAGE 63

56 without ADV. ' The children in Grade 5 mentioned the S-ADV items somewhat more often than the younger children, but there was no evidence of a greater ability to explain these items, even though the error scores dropped for this age group. In fact, this group offered fewer explanations than the younger children with higher error scores (= wrong explanations). This could be taken as evidence of 'emerging awareness' of the ADV-class, especially in relation to the comments and explanations given by the eighth graders. Since all of the Grade 8 subjects were asked to explain their responses, their data are particularly interesting. The comments of the seven subjects who made no errors, or errors only on S-ADV items, are listed in Table 13, along with their total error scores. A consideration of this table shows that, for these three items, even at an age that is generally considered to be 'linguistically adult,' some children obviously still have problems with S-ADVs. The comment made by 'Eileen' on the item perhaps is particularly interesting: it suggests that even at this age, children are not yet ready or able to make the fine distinctions required by 'modal' ADVs such as these. The comments made by the eighth graders may be the strongest argument for the relation between the objective error scores and the voluntary comments, but the progression is also important. There were two children included in this study whose data I would like to compare, since they made such a vivid contrast on all of the measures that I have considered so far.

PAGE 64

57 TABLE 13 Summary of Comments by Eighth Graders on S-ADVs Item Subject Errors Comments probably Becky 0 A in B he obviously did Matt 0 A because you don't see it happi Eileen 1 *B because A is more intentional Michele 1 *hard to explain Bill 1 *in A he's going through him Gina 2 TSnoopy's walking behind Linda 2 * looks like he meant to do it in A perhaps Becky 0 A in B she did Matt 0 perhaps? can't tell if she did it in A Eileen 1 this is another one don't like those 'perhaps' (items) Michele 1 ?she grabbed it and ran Bill 1 A doesn't show her doing it Gina 2 *because B-3 she's grabbing it Linda 2 in B she snatched it maybe Becky 0 maybe? B (right reason) Matt 0 B he maybe fell asleep, but can't say he didn't Eileen 1 B in A ' Z ' s ' show he's asleep Michele 1 B in A he did Bill 1 maybe? means I'm not sure, right? B Gina 2 *maybe? A-4 he's asleep Linda 2 *A-4 he's asleep * = wrong response ? = right response but comment suggests wrong reason Derek, an eighth grader , made 6 errors , twice the number of any other eighth grader. This error score would place him, by extrapolation, at about fourth grade level. What is more, he gave wrong answers to just those items for which the third grade group had high error scores. His explanations for his responses were also similar to those given by third graders.

PAGE 65

58 His scores on the 'Eighth Grade Test' were in the 31st percentile for reading and the 14th percentile for math. By contrast, Alexander, a second grader, made only 3 errors ( rarely , perhaps , and maybe ) compared to a grade mean of 8.4 errors (and a fifth grade mean error score'of 4.8). I do not have an objective test score for this child, but these data, in conjunction with his comments, seem to indicate a superior mental ability. Alexander gave correct explanations for most of his responses, and so his low error score cannot be a statistical 'fluke,' especially considering those items to which he gave a wrong response. The comparison of the data from these two children is a strong argument for the tie between the ability to process ADV and general cognitive development. In a narrow sense, these data substantiate the claim that error scores are corroborated by the ability to explain responses. In a much broader interpretation, the data from these two children argue for an 'age-free' acquisition process, which, of course, has implications beyond this study. Summation of Results The statistical analyses of the error scores from this study produced the following results: 1. The form of the test is not a significant variable. 2. The sex of the subject is not significant. 3. The ability to process the ADV items included here does change significantly between Grade 1 and 2 and between Grade 3 and 5 .

PAGE 66

59 4. Because of a significant interaction between ADV form and ADV class, it is not possible to draw firm conclusions as to the effect of these variables individually, beyond saying that morphemic-MAN items were easiest for the age groups considered here, and morphemic-S-ADVs were the most difficult. However, there are indications that ADV form is not a true factor in acquisition, but that ADV class may well be, and that the significant interaction in this study is a function of the particular items that were included . 5. On the basis of the ANOCOVA results for the third grade, mental aptitude is a definite factor in the ability to process ADVs. The informal analysis of the voluntary comments and responses made by the children tend to corroborate the statistical results. The implications of these findings will be the subject of the next chapter.

PAGE 67

60 Note The item certainly needs special comment. It became clear to me during the testing sessions that the children were, for the most part, treating this item not as S-ADV, but as EXT, i.e., not as synonymous with 'I am certain' or 'I am sure,' but rather as a synonym of 'really,' 'a lot,' etc. For this reason, I think it is more reasonable to discuss the S-ADV results here in terms of only the three items which were true S-ADVs.

PAGE 68

CHAPTER IV DISCUSSION Syntax, Semantics, and the Significant Interaction Before discussing the syntactic and semantic results of the study and their possible implications, it is necessary to consider the significant interaction of form by class. No firm statements can be made about either of the main effects unless some logical explanation can be offered for the significant interaction. It is possible that factors not considered in the statistical analyses could have influenced the results, indicating that form and class interact when in fact they may not. The significant interaction is almost certainly a function of the experimental design. By insisting on equal cell entries for the 2-form by 5-class array, and with the added limitation of 1-word ADVs, the choice of possible items to fill the cells was in several cases severely limited. The ideal would have been a selection of items of equal frequency or familiarity. But given the vagaries of language, this was not possible. The best approximation was to fill the cells with the most familiar items which met the basic criteria. This tight experimental design exacted a price: the statistically significant interaction of form and class. 61

PAGE 69

62 A consideration of Table 14, which is a reduction of Table 9 (p. 40) demonstrates the point. TABLE 14 Comparison of Duncan Groupings with Item X's Duncan 3 X Error Class-Form X's Grouping Items Scores FREQ-Morph. 0.47 B seldom 0.76 often 0.20 R.T.-ly 0.41 BC finally 0.42 suddenly 0.42 FREQ-ly 0.25 DE rarely 0.41 usually 0.12 R.T. -Morph. 0.15 EF already 0.19 still 0.12 a X's which do not share a common letter in the grouping are significantly different; those which do are not ( a = 0.01) Table 14 shows that FREQ-Morph. is significantly more difficult than FREQ-ly , whereas R.T.-ly is significantly more difficult than R.T. -Morph. That is, for these two classes, there is apparently an inversion of difficulty by ADV form. But the statistical analyses were run on X error scores per cell. If the individual item X s are examined, the statistical differences appear to be more a function of the pooled X's than of real differences in difficulty. For FREQ, the individual item X's are widely divergent: seldom (X = 0.76) ranks with S-ADV items (see Table 9, p. 40); often (X = 0.20) ranks close to the R.T. -Morph, item already (X = 0.19). The X for the FREQ-ly item rarely is almost

PAGE 70

63 identical with the X's for the R.T.-ly items finally and suddenly , yet FREQ-ly and R.T.-ly are statistically different. The other FREQ-ly item, usually , has a X equal to that of the R.T. -Morph, item still . The logical conclusions to be drawn from the statistically significant interaction of syntactic form by semantic class seems to be that 1) it is probably not 'real' but rather a function of the frequency/familiarity of individual items (to be expanded later) , and 2) the major source of this significant interaction was the middle-difficulty classes: REL. TIME, FREQ, and EXT. Another rather obvious conclusion is that language does not readily fit the structured experimental designs favored by psychologists. The 'accidental gap' for the linguist can mean empty or inappropriately filled cells for the psycholinguist. Even though the apparent form by class interaction can seemingly be explained away on logical grounds, at least for the items in this study, the other syntactic and semantic results can be legitimately discussed only in terms of trends and suggestions. But since it would be cumbersome and distracting to continually refer back to this proviso, the following discussion will assume this is understood, and will be presented in terms more positive than are strictly justified . The syntactic question addressed in this study is very narrow. Basically, the question is: Are derived ADVs

PAGE 71

64 acquired earlier than morphemic ADVs? The point of the question is that, since an adult speaker of English obviously has acquired the rule: ADJH — ly=ADV, it might be the case that, in the process of acquisition as the child becomes aware of the rule, ' -ly' items might be mastered earlier than morphemic items. The evidence from this study indicates that there is a small but significant difference in time of acquisition for the two forms: the '-ly' forms were less difficult than the morphemic forms. But somewhat of the same situation holds for this result as was discussed in relation to the significant interaction. An examination of the individual item X's in Table 9 (p. 40) shows that the error score for 4 items, perhaps (X = 0.78), maybe (X = 0.76), probably (X = 0.78), and seldom (X = 0.76) are almost double the scores for the next most difficult items ( finally and suddenly , X = 0.42). Three of the four items are morphemic. It is possible that two factors other than syntactic form could have contributed to the statistically significant difference for form. The first is level of conceptual difficulty: two of the items are S-ADVs; the second is familiarity: many children indicated they did not know the word seldom . Both of these ideas will be developed in the next section on differences between classes of ADV. In sum, no substantial conclusions can be drawn from this study in the area of syntax. It is possible that the

PAGE 72

65 ADJ+-ly rule is functional in the early school years, but it is almost certainly not a major factor in ADV acquisition. The semantic question, i.e., the question of the effect of ADV class in acquisition, is the one area in which the statistical analysis of error scores seems to provide some definitive answers. The results indicate that ADV class is a significant factor in acquisition, and that the 5 semantic classes can be divided in terms of 3 levels of cognitive difficulty. In brief, MAN items are easiest/earliest acquired; FREQ, REL. TIME, and EXT are 'middle-difficulty' items; and S-ADVs are the most difficult/last acquired. The reality of these 'levels' can be supported by evidence beyond the statistical results. Even for the youngest children, error scores on MAN items were very low. There are two factors which might account for the earlier acquisition of MAN relative to other 'process-modifier' items. The first is that the concepts covered by these MAN items are conceptually more basic than the TIME/SPACE concepts covered by the 'middle-difficulty' items. This is supported by the work Ervin-Tripp (1970) did on WH-questions . Her findings showed that children can correctly answer how questions before they can handle when questions. One small piece of evidence from this study is that some of the Grade 1 children made comments such as 'This one is quiet/quick/happy' in indicating the correct panel for the MAN-ly items.

PAGE 73

66 The second reason involves the relation between ADJ and ADV and the notion that items are processed if and when they are understood. For the MAN items in this study, the similarity to ADJ is much stronger than for any of the other classes: all of the MAN-morph. items are phonetically identical to their ADJ counterparts. It could be that the younger children were dealing with the string containing MAN -ADV as they would a string with ADJ, and are not necessarily exhibiting an ability to process ADV. This subject will be discussed further in the final section of this chapter. The 'middle-difficulty' items, which cover the classes REL. TIME, FREQ, and EXT, present the most difficult problems of interpretation and extrapolation. The general conclusion to be drawn from the data on these classes is that the individual item is in some way more important than either syntactic form or semantic class. For example, consider the items in the FREQ class: seldom (X = 0.76) , often (X = 0.20), rarely (X = 0.41), and usually (X = 0.12). Seldom and rarely are synonymous but the X's for these items are quite different, and much higher than the X's for the other 2 items. They are, of course, different in syntactic form, but then consider the items almost (X = 0.26) and nearly (X = 0.31), also synonymous, but different in syntactic form, yet with X's that are virtually the same. It is an interesting sidelight that, during the testing sessions, some of the

PAGE 74

67 older children, usually the fifth graders, asked, 'What does seldom / rarely mean?' When told 'not very often' their response was 'Oh, then it's (correct panel).' This predominance of item over class or form is also evident for the following items with virtually identical X's: finally / suddenly (X = 0.42), which are REL. TIME-ly; rarely (X = 0.41), which is FREQ-ly; and only (X = 0.41), which is EXT -morph. On the basis of the data, the final conclusion for these 'middle-difficulty' items must be that 1) the items and the concepts they cover are acquired during the early school years, since the Grade 1 children made many errors on these items and the Grade 8 children made none; 2) there is probably some scale of difficulty for the concepts covered by these classes; 3) it is not possible to determine from the results of this study just what that scale might be. The data on S-ADVs were unquestionably the most conclusive and, in many ways, the most interesting results of the study. This class of ADVs was consistently most difficult for Grades 1-5, for which there are statistical data, but it was also somewhat of a problem for some of the children in Grade 8, as indicated by their comments and their error scores. These results are not surprising: S-ADV, as a proposition modifier, was expected to be cognitively more difficult than the 'process-modifier' items. This expectation was based on general notions from language acquisition

PAGE 75

68 studies which indicate that the child develops from specific to general concepts, but more specifically, the analysis of S-ADVs presented by Schreiber (1970) , mentioned briefly in Chapter I. Schreiber' s suggestions deserve further consideration in light of the results of this study. According to his analysis, S-ADVs fall into two categories: 'modal' and 'evaluative.' Modal items (e.g., apparently , clearly , obviously) indicate 'degree of surety' on the part of the speaker. Evaluative items ( fortunately , oddly , naturally ) , on the other hand, express the speaker's opinion on the statement being made. In order to correctly process S-ADVs the child must reach the point where he is able to recognize that these items represent a 'comment on the proposition,' and that the comment represents either speaker opinion or speaker surety of the factual nature of the proposition. But there is a further complication, in relation to acquisition, from Schreiber 's analysis. With modal S-ADVs, the proposition is not necessarily 'fact.' With evaluative items, it is. Consider the following Ss: 42. Presumably, Snoopy fell off the doghouse. 43. Regrettably, Snoopy fell off the doghouse. In S-42, the event mentioned may or may not have actually occurred; in S-43, it is implicitly considered true. In sum, before the child can correctly process S-ADVs, he has to infer some rather sophisticated notions, restrictions and extrapolations (in terms of implicit information).

PAGE 76

69 Given Schreiber's analysis, and the results of this study, the comments made by the eighth graders are particularly interesting (see Table 13, p. 57) . The 2 children who made no errors on the test gave good explanations for all the S-ADV items. The comments of the other 5 children who gave incorrect answers to 1 or 2 of the S-ADV items seem to indicate that these children are partially aware of the special function of S-ADV (as opposed to 'process-modifier'), but that they do not yet fully understand the implication of these items. It should be noted that all S-ADV items in this study were 'modal' by Schreiber's analysis, and so no unqualified statement can be made as to the class in general. Still it seems that once the function of S-ADV is recognized, both 'modal' and 'evaluative' ADVs should be acquired at about the same time. That is, since all S-ADVs represent 'speaker comment, ' once the child becomes aware of this function, he should be able to handle both types of S-ADVs. Nevertheless it remains an interesting empirical question. ADV Acquisition and Cognitive Development The final question posed in Chapter I concerned the possible relationship between ADV acquisition and cognitive development. The statistical results for ANOVA and ANOCOVA, as well as the informal analysis of the voluntary comments, relate to this question. The sum of the evidence indicates

PAGE 77

70 that there is a strong tie between cognitive level and the ability to process ADVs. The analysis of the error scores showed a significant decrease in errors between Grades 1 and 2 and Grades 3 and 5 For Grade 8 there were no errors, except the few on S-ADV items. Clearly, facility with ADVs improves with age, but this does not necessarily imply a direct connection between that improved ability and cognitive development. The older children have had proportionately greater exposure to the various items. It is a question of whether the improvement is more a reflection of familiarity or of a developing aware ness of the basic function of ADV. The ANOCOVA results are relevant here. They indicate a significant correlation between test scores and mental ability for one age group. Assuming all 7-year-olds have had about the same exposure to ADV, we can tentatively conclude that it is primarily level of cognitive development that is the major factor in the significant improvement across grades. The analysis of the voluntary data provides further support for this conclusion. The children in Grade 2 and above were increasingly able to correctly explain their responses. They were demonstrating an increasing awareness of the basic function of ADV, and also of the concepts covered by the particular ADV items. Finally, there is some outside evidence to support this Loban (1963, 1964) found a close relationship conclusion .

PAGE 78

71 between mental ability and use of ADVs, both in number and form, for grade school children. He used a 'finish the sentence' type task involving adverbial connectives for Grades 5-9. Children in the high ability group did well even at Grade 5, whereas the low ability group were consistently poorer through Grade 9. Clearly, there is a close connection between ADV acquisition and cognitive development that deserves to be investigated in more detail. Stages in ADV Acquisition The major result of this study is that, if the various forms of data are integrated, certain stages in the acquisition of ADV become apparent. The statistically significant results say that ADV acquisition relates to age (ANOVA) and to mental aptitude (ANOCOVA) . The voluntary comments were even more interesting, and more illuminating, in suggesting just how ADV acquisition relates to general cognitive development. In sum, an overall view of ADV acquisition seems to emerge. This view can be presented in terms of 'stages' which relate to degree of 'awareness' and apparent strategies for processing Ss with ADV. 2 Stage 1: unaware/ignoring ADV For the youngest children in this study, the concepts underlying many of the ADV items were evidently beyond their level of cognitive development. Therefore, the apparent strategy was to process the string without the ADV. This

PAGE 79

72 would account for the observation that the younger children tended to quickly and enthusiastically point to the critically wrong frame for implicit NEG items: the choice is so obvious if the ADV is not considered! There is evidence that this proposed stage is present in other areas of acquisition. Fraser, Bellugi, and Brown (1963) found that young children tended to interpret a S such as the girl is pushed by the boy to mean that the girl was doing the pushing. The assumption is that, if the passive formation is not yet understood, those items in the S which make it passive are ignored, and the S is processed as active. Smith (1970) also found evidence of what she calls ' selective listening 1 in 3and 4-year-olds in repetition and response studies. The conclusion was that the children repeated or responded according to those elements of the test S that they understood. Keller-Cohen (1975) reached a similar conclusion in her investigation of temporal connectives. Her subjects, who evidently did not understand the item while , substituted before, after or and then in their repetitions of test sentences containing while . There seems to be a consistent--or at least, prevalent — strategy in acquisition, across syntax and semantics, to process strings according to the current level of linguistic and/or cognitive development. Stage 2: Partially aware/guessing at ADV At some point, typically around the Grade 2 level, the child seems to begin to recognize that all elements in a

PAGE 80

73 string have some significance. But since the general concepts behind many of the ADVs are still beyond his comprehension, there is a tendency to guess. There is additional evidence of this 'partial awareness' in that children at this stage often repeat the ADV aloud but to themselves or repeat it as a question. They still process the S without ADV, but without the surety (for the wrong response) exhibited by the children in the previous stage. Stage 3 : Aware/asking about ADV At this point the child does recognize that all elements in the string are necessary for correctly processing the string. It is at this stage that the child either responds correctly or asks directly what the ADV means. Correct responses are often explained by defining or referring to the ADV. Stage 4: Unaware/comprehending ADV This is not truly a 'stage,' unless it can be considered the final stage, but rather some observations on typical behavior when ADV acquisition is virtually complete. The term 'unaware' reflects the fact that, once the basic function of ADV is comprehended, the ADV is rarely mentioned voluntarily, in sharp contrast to the 2 earlier stages. That is, the more complete the ability to process ADV, the less aware the child seems to be that he is doing so! It is an interesting sidelight that most of the eighth graders who made few if

PAGE 81

74 any errors and correctly explained their responses, asked after the test session just what I was studying, and were surprised to learn that it was ADV. There is an important point that needs to be made here, and that is, that these stages are not directly linked to chronological age, syntactic form, or semantic class. Rather they apparently relate most directly to an emerging sense of the basic function of ADV. This 'sense of ADV function' is in turn apparently correlated with cognitive development. The contention that ADV acquisition is more closely related to mental ability than to chronological age is supported by the ANOCOVA results for Grade 3, which showed a significant relation between errors on the test and OLMAT scores. The comparison of the error scores and voluntary comments for the gifted second grader and the exceptional eighth grader give additional support to this contention. The statistical results of this study do not relate to the question of syntactic form in these proposed stages. The fact is that the syntactic dimensions of the study were too narrow to make any conclusive statement. But the evidence from the studies mentioned in Chapter I on the language of elementary school children, in particular, Loban (1963, 1964), suggest that children's use, and presumably comprehension, of more advanced syntactic forms increases with grade into the high school years. For the semantic dimension of ADV, the logical conclusion seems to be that the stages would apply almost on the

PAGE 82

75 level of individual items, but with certain predictable overlaps. The evidence that some items within a given class were more difficult than others in the same is the complicating factor. Of course, this applies primarily to TIME and EXT items; the MAN and S-ADV, in terms of statistical results and analysis of voluntary comments, fit well with this stage theory. For the TIME and EXT items, more detailed investigation is needed. But the error scores, in conjunction with voluntary comments for those items discussed under 'implicit NEG' in Chapter III do support the stages. The evidence for the existence of the 'stages' in ADV acquisition proposed here is strong, but not conclusive. It is important to recognize that no claim has been made that a child, at whatever age, is exclusively in one stage with respect to general ADV processing. Rather, the implication is that a child--at least, beyond Stage l--goes through the progressive stages almost on an item basis. There are, of course, indications that the less complex MAN items will be mastered first, and the S-ADV items last. But in between, the predictable patterns for ADV, based on this study, are that the typical child will be in various stages at any given point, relative to ADV class and form. It is most probable that the child handles individual ADVs at a given level dependent upon the syntactic and semantic complexity of the specific ADV item.

PAGE 83

76 Notes ^Schreiber's paper offers some interesting suggestions on syntactic complexity of S-ADV. In fact, that is the major premise of the paper. He also suggests some ways in which syntax and semantics interact for S-ADVs. 2 The term 'awareness' needs to be defined as it is used in this paper. It does not imply any conscious level in language processing. Rather, it refers to a logical way of categorizing the reactions/responses of the children to this test.

PAGE 84

CHAPTER V SUMMARY AND CONCLUSIONS Review of Purpose and Questions There is ample evidence that the category ADV is a major problem in the linguistic description of English. There is also good reason to believe that data on acquisition can provide important insights for the descriptive linguist. But the data on ADV acquisition is minimal. For the most part, it must be derived from incidental evidence in studies with primary emphasis on other questions or areas of acquisition. The purpose of the present study was to look directly at ADV acquisition: to look at some of the more basic aspects of ADV in relation to acquisition, and to establish some guidelines for more detailed investigations. The elemental questions addressed in the study relate to the basic function of ADV, to the relative importance of syntax and semantics in the acquisition of ADV, and to the possible relation between ADV acquisition and cognitive development. General Conclusions The statistical results of the study are ambiguous in some cases, but certain general conclusions, based on an integrated interpretation of error scores and voluntary 77

PAGE 85

78 comments, can be drawn. There may be some slight advantage of 1 -ly' items over morphemic ADVs in the process of acquisition. And it is possible that ADV form and class interact — at least they did statistically for the items in this study. On the other hand, it seems sure that semantic class is the most important factor in acquisition. The S-ADVs were the most difficult of the classes considered here, and MAN the least difficult. No definite statement can be made relative to EXT, FREQ, and REL. TIME. The most conclusive statement that can be made from this study is that there is a strong connection between ADV acquisition and cognitive development. Error scores decrease with advancing grades: older children as a group are better able to process ADV than are younger children. This is not so surprising, but consider that error scores for children in the same grade are directly related to their mental aptitude scores. The decrease in error scores by grade is almost surely attributable to increase in cognitive level, and not to other factors such as increased exposure to ADV. Perhaps the most interesting conclusion drawn from the study was that there appear to be stages in ADV acquisition which relate in certain definite ways to other areas of acquisition. Projections This study represented a first attempt at an organized investigation of children and ADVs. The results provided some

PAGE 86

79 general guidelines for future research in this area. More importantly, the results suggest certain issues and specific questions in the area of ADV acquisition that deserve special attention. Since this study was confined to 1-word ADVs, the initial question is whether the use of phrasal ADVs, which are paraphrases of the items included here, might permit a more detailed description of the development of time and space concepts. A further question relates to whether the stages that were apparent in this study can be verified and defined in more detail. That stages exist in early language is welldocumented, and so it is to be expected they are also a part of later acquisition. In fact, Chomsky (1969) found that in her ask/tell studies there was evidence of a steady progression in comprehension. Chomsky also mentioned another aspect of this 'progression' that was apparent in this study as well: that it is not linked to chronological age. This is an important consideration that is sometimes overlooked in studies of child language. Unless the N is quite large, mental age is more likely to be a critical factor than is chronological age . A final consideration is that of methodology. The test used here worked well for the specific questions being considered. It was evident, however, that it was not appropriate for children under 6, or below first grade. It would seem that an adaption of the imitation/repetition method that has

PAGE 87

80 been so productive with young children in other areas of acquisition should work equally well with ADVs. For older children, more information could possibly be obtained by requiring subjects to explain how the panel sets differed. This method seems especially appropriate to explain in more detail the results that Loban (1963, 1964) found: that for older children, there is an increase in both the frequency of ADV usage, but more interestingly, an increase in the different types of syntactic form of ADV.

PAGE 88

APPENDIX A Test Ss for Form 1 1. Charlie Brown threw the ball straight. 2. Snoopy carefully put the bird in the nest. 3. Linus only hit the ball. 4. Schroeder quietly played the piano. 5. Lucy suddenly hit Linus. 6. Probably Snoopy bumped into Linus. 7. Charlie Brown already jumped in the leaves. 8. Linus went to bed fast. 9. Charlie Brown seldom watches TV. 10. Perhaps Sally grabbed the blanket. 11. Linus usually helps his mother. 12. Snoopy almost grabbed the blanket. 13. Certainly the cat scared Snoopy. 14. Lucy rarely plays with Linus. 15. Snoopy nearly fell off his doghouse. 16. Charlie Brown flew the kite high. 17. Lucy happily listened to Schroeder play. 18. The fire completely burned Snoopy's doghouse. 19. Lucy is still talking on the phone. 20. Linus finally found his blanket. 21. Schroeder often reads books. 22. Snoopy quickly ran to kick the ball. 23. Lucy hit Linus hard. 24. Maybe Snoopy fell asleep on his doghouse. 81

PAGE 89

APPENDIX B Test Ss for Form 2 1. Charlie Brown threw the ball straight. 2. Snoopy carefully put the bird in the nest. 3. Linus only hit the ball. 4. Schroeder quietly played the piano. 5. Lucy finally hit Linus. 6. Perhaps Snoopy bumped into Linus. 7. Charlie Brown already jumped in the leaves. 8. Linus quickly went to bed. 9. Charlie Brown usually watches TV. 10. Probably Sally grabbed the blanket. 11. Linus seldom helps his mother. 12. Snoopy nearly grabbed the blanket. 13. Maybe the cat scared Snoopy. 14. Lucy often plays with Linus. 15. Snoopy almost fell off his doghouse. 16. Charlie Brown flew the kite high. 17. Lucy happily listened to Schroeder play. 18. The fire completely burned Snoopy's doghouse 19. Lucy is still talking on the phone. 20. Linus suddenly found his blanket. 21. Schroeder rarely reads books. 22. Snoopy ran fast to kick the ball. 23. Lucy hit Linus hard. 24. Certainly Snoopy fell asleep on his doghouse 82

PAGE 90

REFERENCES Amidon, A., and P. Carey. 1972. Why can't five-year-olds understand before and after ? J. verb. Lrng . verb. Behav., 11.417-423. Barrie-Blackley , Sandie. 1973. Six-year-old children's understanding of sentences adjoined with time adverbs. J. psycholin. Res., 2.153-165. Bronckart, J.P., and H. Sinclair. 1973. Time, tense and aspect. Cognition, 2.107-130. Chafe, Wallace. 1973. Language and memory. Lg . , 49.261-281. Chomsky, Carol. 1969. The acquisition of syntax in children from 5 to 10. Cambridge: M.I.T. Press. Clark, E.H. 1971. On the acquisition of the meaning of before and after. J. verb. Lrng. verb. Behav., 10.266-275. Clark, H. 1970. The primitive nature of children's relational concepts. Cognition and the development of language, ed. by J.R. Hayes, 269-278. New York: Wiley. Dingwall, William 0. 1971. A survey of linguistic science. College Park: University of Maryland. Ervin-Tripp, Susan. 1970. Discourse agreement: How children answer questions. Cognition and the development of language, ed. by J.R. Hayes, 79-107. New York: Wiley. Feagans, Lynne. 1974. Children's comprehension of some temporal and spatial structures. Papers from the Tenth Regional Meeting of the Chicago Linguistic Society, 139-150. Fraser, C., U. Bellugi, and R. Brown. 1963. Control of grammar in imitation, comprehension, and production. J. verb. Lrng. verb. Behav., 2.121-135. Fraser, James B. 1965. An examination of verb-particle constructions in English. Ph.D. dissertation, M.I.T. 83

PAGE 91

84 Fries, Charles C. 1952. The structure of English: An introduction to the construction of English sentences. New York: Harcourt, Brace. Harrell, Lester E., Jr. 1957. A comparison of the development of oral and written language in school-age children. Monographs of the Society for Research in Child Development, Ser. No. 66, Vol . 22, No. 3. Hayes, John R. 1970. Cognition and the development of language. New York: Wiley. Huang, Shuan-Fan. 1975. A study of adverbs. (Janua linguarum, Series minor, 213.) The Hague: Mouton. Hunt, Kellogg W . 1965. Grammatical structures written at three grade levels. NCTE Res. Rep. No. 3, Champaign. National Council of Teachers of English. Hunt, Kellogg W. 1970. Syntactic maturity in schoolchildren and adults. Monographs of the Society for Research in Child Development, Ser. No. 134, Vol. 35, No. 1. Jones, Lyle V., and Joseph M. Wepman . 1966. A spoken word count-adults. Chicago: Language Research Associates, Inc . Keller-Cohen , Deborah. 1975. The acquisition of temporal reference in pre-school children. Bloomington: The Indiana University Linguistics Club. Kessel , Frank S. 1970. The role of syntax in children's comprehension from ages six to twelve. Monographs of the Society for Research in Child Development, Ser. No. 139, Vol. 35, No. 6. Loban, Walter D. 1964. Language ability: Grades seven, eight, and nine. Coop. Res. Project No. 1131, Berkeley: University of California. Loban, Walter D. 1963. The language of elementary school children. NCTE Res. Rep. No. 1, Champaign: National Council of Teachers of English. Menyuk, Paula. 1969. Sentences children use. Cambridge: M.I.T. Press. Murray, H . A . 1943. Thematic apperception test. Cambridge: Harvard University Press. Nilsen , Don L.F. 1972. English adverbials. (Janua linguarum. Series practica, 125.) The Hague: Mouton.

PAGE 92

85 O'Donnell, Roy C. , William J. Griffin, and Raymond C. Norris. 1967. Syntax of kindergarten and elementary school children: A transformational analysis. NCTE Res. Rep. No. 8, Champaign: National Council of Teachers of English. Piaget, Jean. 1971. The child's conception of time. New York: Ballantine Books. Rinsland, Henry D. 1945. A basic vocabulary of elementary school children. New York: Macmillan. Schreiber, Peter A. 1970. Some constraints on the formations of English sentence adverbs. Ling. Inq. , 1(4). 83101 . Slobin , Dan I. 1971. Developmental psycholinguistics. A survey of linguistic science, ed. by William 0. Dingwall 298-410. College Park: University of Maryland. Smith, Carlota S. 1964. Determiners and relative clauses in a generative grammar of English. Lg . , 40.37-52. Smith, Carlota S. 1970. An experimental approach to children's linguistic competence. Cognition and the develop ment of language, ed. by J.R. Hayes, 109-135. New York: Wiley. Wepman, Joseph M. , and Wilbur Hass. 1969. A spoken word count (children--ages 5, 6 and 7). Chicago: Language Research Associates, Inc.

PAGE 93

BIOGRAPHICAL SKETCH Joeann Chandlee Paige was born on April 3, 1934, in Wilmington, Delaware. She earned a classical diploma from the Ursuline Academy in Wilmington in June, 1951, and entered the University of Delaware in September, 1951. She completed her undergraduate work for the Bachelor of Arts degree in January, 1955, and was accepted directly into the graduate school at Delaware. In September, 1955, she entered the graduate program in Experimental Psychology at Tufts University in Medford, Massachusetts, with a two-year teaching assistantship . She received her Master of Science degree from Tufts in October, 1957, but remained at the school as assistant investigator on a government-funded project. In June of 1958, she accepted a position at the I.B.M. Research Laboratory in San Jose, California, as Associate Engineering Psychologist, a position she held for three years. She is married to Arnold Paige, a professor of Electrical Engineering at the University of Florida. They are the parents of three children. 86

PAGE 94

I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. i &lxd C '.( f L=tL i o Robert J. Seholes, Chairman Professor of Linguistics and Speech I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. 'Jean Casagrande--Associate Professor of Linguistics I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. ^ 7 (/L t( ( Ct'y / 0 v ,>ji (_ i ti f A ^ Paul J. Jensen ' Professor of/Speech I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. / /) 7 Ira S. Fischler Assistant Professor of Psychology

PAGE 95

This dissertation was submitted to the Graduate Faculty of the Program in Linguistics in the College of Liberal Arts and Sciences and to the Graduate Council, and was accepted as partial fulfillment of the requirements for the degree of Doctor of Philosophy. December 1978 Dean, Graduate School