Lexical Acquisition in Skilled and Less Skilled Readers

Lexical acquisition ability was studied in fourth grade children through use of a paired associate design. Reading achievement predicted ability to learn words more highly than did other factors, including estimated I.Q., and shortterm memory did not. Examination of two subgroups of skilled and less skilled readers indicated that less skilled readers had more difficulty in acquiring new words. Less skilled readers made more errors and required more trials than did skilled readers. Less skilled readers also achieved lower scores on measures of short and long term recognition of the word's referents. No differences in rate of forgetting over time were found between groups. No between group differences were found in the ability to provide definitions for the newly learned words. No significant differences between groups were found on a measure of incidental word learning.

ACKNOWLEDGEMENTS r am especially grateful to Dr. Susan Brady, my major professor for her interest, guidance and support throughout the course of events which led to the completion of this study.
I also wish to thank Dr. Janet Kulberg for her valuable insights and comments.
Very special thanks are extended to research assistants, Judith Cicero, Janice Ruggeri and Linda Stoler, for their invaluable assistance in collecting data and scoring protocols. Their dedication, precision, and support were fully appreciated.
In addition, I would like to thank the Tiverton Public School System, the co-operative principles, teachers, and staff who made this work possible by contributing the use of their facilities and time.
The greatest appreciation is extended to the children who participated in this study, as well as to their parents. Background. In the study of reading disability it has become increasingly apparent that in most cases this is a language based disorder. Reading disabled children have been shown to differ from their peers on a number of diverse language skills. These include differences in phonological processing abilities (Liebermann & Shankweiler, 1985;Vellutino, 1979;Wagner & Torgesen, 1987), in syntactic abilities (Jordan, 1988;Morice & Slaugis, 1985;Vogel, 1975) and in pragmatic use of language (Donahue, 1984).
These observed differences in linguistic capabilities have led us to speculate about lexical acquisition and vocabulary knowledge in the reading disabled child. As might be expected, vocabulary differences between groups of reading disabled children and their normally achieving peers have been found (Kail & Leonard, 1986;Vellutino & Scanlon, 1987). In comparing groups that had been used for research studies, Vellutino and Scanlon (1987) noted that the reading disabled groups consistently scored lower than the non-disabled groups on measures of both productive and receptive vocabulary measures. These group differences remained even when the groups were matched on a non-verbal I.Q. measure.
At least two hypotheses can be generated concerning lexical acquisition by reading disabled students. One hypothesis states that the reading disabled child reads less than the good reader and therefore learns fewer new words. vocabulary differences are found in very young disabled readers, however, raising doubt as to whether this explanation can be sufficient.
A second hypothesis states that the observed difference in vocabulary is due to a difficulty learning new words because of a deficit in phonological processing and related linguistic abilities. This explanation has a strong theoretical and empirical basis. Learning a new word requires accurate perception, storage, and retrieval of the word. Each of these processes has been shown to differentiate good from poor readers, as will be discussed.
First, results indicate that poor readers are deficient in speech perception (e.g. Godfrey, Syrdal-Lasky, Millay, & Knox, 1981;Snowling, Goulandris, Bowlby, & Howell, 1986;Werker & Tees, 1987). In a study of speech perception using good and poor readers, it was found that the poor readers were less accurate at identifying speech sounds masked in noise.
In contrast, when environmental sounds were presented in noise, differences between the groups did not emerge (Brady, Shankweiler, & Mann, 1983). Furthermore, good and poor readers have been shown to differ in speech perception under normal conditions as well (Brady, Foggie, & Rapala, in press;Catts, 1986;Rapala & Brady, in press;Snowling, 1981).
Poor readers are found to be slower and less accurate at retrieving the correct words when performing these tasks.
In sum, reading disabled children are deficient on a variety of phonological tasks that tap perceptual, storage, and retrieval processes.
In the currently popular framework of a limited-capacity working memory system (Baddeley, 1976: Perfetti, 1986, it is plausible that deficits in any of these could lessen the efficiency and accuracy of verbal processing. This has logical consequences for tasks which are dependent on verbal processing, such as word acquisition. Indeed, a subject with severe deficiencies in phonological storage was also found to be extremely deficient in a task of learning new words for known concepts (Baddeley, Papagno, & Vallar, 1988).
In addition, in a longitudinal study of pre-readers, a positive correlation emerged between phonological short-term memory at age 4 and vocabulary gain at age 5. (Gathercole & Baddeley, 1989). There is also preliminary evidence that reading disabled children experience difficulty in learning new words. Nelson and Warrington (1980) found that a group of dyslexic children produced more errors than a control group in a task of vocabulary learning.
When the observed reading group differences in vocabulary are considered in light of this research, the hypothesis that they are due to reading experience becomes less tenable. This is not to deny the probability that some of the observed difference in vocabulary may occur as a result of differences in exposure to text; children certainly expand vocabulary through reading (Jenkins, Stein, & Wysocki, 1984;Nagy, Anderson, & Herman, 1987 The process can occur with a single exposure to a word and explicit instruction or direction to notice the word need not be present (Carey, 1978).
To test the hypothesis that less skilled readers learn words differently from their skilled reader peers, children in both groups were introduced to six new words which were deliberately taught and learned, as well as to two words which were incidentally learned. These words were nonsense words which were devised specifically for this study.
Several aspects of word learning were compared between the two groups including: An age range limit of 9 years, 5 months to 10 years, 6 months was selected as a criterion for inclusion in the study. Subjects who had a high score on one measure of I.Q. (PPVT and WISC-R), but were within an average range on the other measure were therefore included. Nine potential subjects were dropped from the final analysis because of failure to meet one or more points of the above criteria. One additional subject was dropped from the final analysis because of incomplete data.
Reading groups were formed by using two subtest scores, Pictures for each word were painted on 11" x 14" white poster boards.
A list of the words and definitions appears in Appendix A.
A game was created for use in this study. This Journey Game assessed short-term recall of the words learned in both the directed and incidental learning conditions. To play the game, a small remote-controlled robot with a tape deck was used. This robot played a recording of the game script, "A Journey to Another Planet", which had been taped by a local actor. During this game the children were asked to say specific experimental words in response to specific cues  The child was introduced to the concept of the game and to the task of learning the words.
The words were put into two trial groups of three words each.
The experimenter said a word, showed the child a picture of the object, and gave the definition. The child was then asked to repeat the word. After three words had been introduced in this manner, the trial blocks began. In each trial block the examiner stated the word and definition while presenting the appropriate picture. After the three words had been presented, the examiner showed the pictures, one by one, and asked for the words. If the child made an error, the examiner supplied the correct target and retaught the word. The same group of three words was then presented in another trial block. The order of the words within each block varied to avoid having the task become either predictable or tedious.
A word was counted as learned if it was correctly produced on two successive blocks of trials. Each group of words was presented at least four times regardless of whether or not the criteria of two successful recalls for all three words had been met. A maximum of ten trials was selected as a cutoff.
Children who did not reach the criteria of two consecutive successful trials for each word were not considered for inclusion in the reading groups. (Two children -one skilled reader and one less skilled reader, were eliminated from the reading groups on this basis). After the first group of three words was presented, the experimenter taught the child the second group of three words, following the same procedure.
Definition Testing. Immediately following the training, the examiner said the words and asked the child supply the definitions. Scores were obtained by counting the number of components of the definitions which the child supplied. If the child did not correctly pair the words and definitions during testing, the examiner paired the definition components given by the child with the appropriate target. This corrective feedback was given only after all six definitions had been tested.
In addition to the six words taught as described above, two other experimental words were introduced to the child.
Both of these words were mentioned twice by the examiner, but were not deliberately taught. These two words are the stimuli for the fast mapping condition. Data from tasks using these words were collected in the Journey Game as well as in the recognition booklets. These data were analyzed separately.
A detailed description of the vocabulary training procedure appears in Appendix B.
Journey Game (Recall testing). Following the training phase children were individually introduced to the robot.
Short-term recall was then assessed in a game. The Robot described an encounter with an object on the planet in terms which closely matched the learned definitions. The child was then asked to provide the correct target word for each given definition. All eight words were used in the game. The script of the Journey to Another Planet appears in Appendix c.
Booklet. Following the game, the short-term recognition booklet was introduced. Children were asked by the examiner to mark a particular target from a variety of foils on each page.
Session 3: A repeat of the booklet task was done at an interval of between one and three weeks from the initial presentation.
In this presentation, the robot "said" the words. Intervals between Session 2 and Session 3 were evenly distributed between the reading groups.
Because of the poor performance of all children on the short-term recall task (Journey Game) in session 2, a planned long-term recall task was eliminated from Session 3. The data from the incidental learning condition were analyzed separately from that of the directed learning condition.

Comparison of Skilled and Less-Skilled Reader Groups, N=22
To ascertain if contribution made by the reading ability would result in group differences between the skilled and less skilled reading groups on the outdome variables, several Analysis of Covariance (ANCOVA) procedures were performed.
In these ANCOVAs, the ESTIQ score was used as the covariate.
Although ESTIQ did not appear to substantially influence the learning of the words in the regression analysis, there were group differences in ESTIQ found between the skilled (M= 115.5) and less skilled (M=99.5) readers. The ANCOVA procedure was selected to control for these initial group differences.  A repeated measures ANOVA, for recognition scores of groups over time (short and long-term recognition) failed to produce any significant results. There was no difference between the relative ability of the groups to recall the words over time.

Retrieval and Recognition Data (untrained words)
As no subject was able to produce any of the untrained words in the recall situation, there was no analysis for this portion of the study.
A Chi-square was preformed, using the untrained recognition scores as the outcome variable.
failed to find significant group differences, A, Table 6).

DISCUSSION
The analysis of the data indicated that a positive relationship exists between reading ability and lexical acquisition ability. Reading ability predicted more of the accountable variance than either I.Q. (measured by either the ESTIQ or the PPVT) or short-term memory (measured by DIGIT) in measures of word learning (i.e. TRIALS, ERROR). This was found using the entire sample of 56 fourth grade children.
The predicted differences were found between the skilled and less skilled readers when the groups were compared on the measures TRIALS and ERRORS. This was true even when I.Q. was statistically controlled.
These results indicate that less skilled readers do not learn the phonological labels for concepts as readily or as well as their skilled reading peers do.
In contrast, the ability to grasp the conceptual attributes of the words, as measured by the definition score, was most highly influenced by I.Q. and not by reading ability.
Both skilled and less skilled readers were incorporating the elements of each definition into a single conceptual unit, however, the less skilled readers were less able to retrieve the labels for these concepts during word learning. Group differences in the recall of the newly learned words did not emerge. It may be that good readers as well as poor readers find it difficult to retrieve newly learned words. Alternatively it may be that the words themselves were too numerous and/or complex for fourth grade children, or that the novelty of the situation hampered the children's ability to respond. Given the poor performance by all children on this measure, one or both of these alternative explanations seems to be a potential factor.
In an analysis of the recognition data, the differences in the amount of variance accounted for by either the reading measures or the ESTIQ was slight. The reading measures had only a small advantage as predictors. When I.Q. was statistically controlled, significant group differences in recognition were present. The groups were comparable, however, in the stability of their ability to recognize the words over time.
There were no differences in either the production or the recognition of the incidentally learned words. At least three factors may have influenced this outcome. First, this task was easier than recognition of the learned words. For each of these two words the choices presented were representations of the two incidental words, two learned words, and a 'wildcard' foil. The incorrect but plausible choices (seen before but not trained) was limited to one. A second possible factor is the limited range of scoring 0 -2 . And finally the small sample size may have contributed to an increased Beta level in the statistical analysis.
If there are truly no differences in the rate of incidental word learning between skilled and less skilled readers, we must question why differences emerged in the directed learning phase. It is first of all possible that any significant difference is a statistical error. It had been decided that because of the small and uneven sample size, a MANCOVA would be inappropriate for the number of dependent variables.
In addition, a high Type II probability was seen as particularly detrimental to such an exploratory investigation. The decision to run multiple ANCOVAs increased the probability of a Type I error.
A second possible explanation is that the less skilled readers have repeatedly experienced difficulty in school tasks. Although efforts were made to keep this from being an academic task by embedding it in the context of a game, the possibility must still be considered that the observed differences may have been due to expectations of failure and/or anxiety on the part of the less skilled readers.
This second explanation is able to account for the observed group differences, but it is an unlikely as an explanation to account for the overall relationship between reading ability and word learning which emerged in the analysis of the entire sample .
It may appear to the reader that the less skilled readers did not do as well when many items to remember were presented (the six trained words), but were comparable to the skilled reader group when only two items were presented. However, in this study all eight words were presented during session two.
The incidental words were mentioned before each of the two training blocks. This minimizes the possibility that the difficulty of the poor readers was due to less ability to learn multiple stimuli.
Therefore, support was found for the hypothesis which states that there are differences in lexical acquisition ability between skilled and less skilled readers. Learning of new vocabulary did differ between reading groups in a directed learning condition, and the failure to find differences in incidental learning may have been due to design or statistical problems.
Future studies are necessary to further explore the relationship of those factors which affect both reading and vocabulary acquisition. Before further studies are initiated, the nature of the errors made by the two groups must be analyzed and considered. Future investigations may then be constructed to more accurately target the underlying processes in which the two groups differ.
For instance, the children were required to repeat each item immediately after it had been introduced. If there were differences at this point it would point to possible differences in either the ability to perceive the word, or the ability to produce the words.
To more accurately assess where difficulty lies, several steps may be taken. First, tasks can be devised which compare the retrieval of a newly presented target word by the child with the child's ability to distinguish between the targets and foils. If the child is able to choose the correct word but is not able to repeat it accurately, it would suggest that perception and encoding are probably not the greatest area of difficulty. Second, the possibility exists that discrepancies in the ability to retrieve new words, relative to the ability to retrieve known words. There is evidence that less-skilled readers have greater difficulties than skilled readers in the area of lexical retrieval Rudel 1976a, 1976b).
Whether or not this can account for the group differences in learning new words however is still to be explored.
Third, additional exploration of the relationship between short-term memory and learning new words can be undertaken.
The present study used Digit Span as an estimate of short-term memory, and this variable was not a good predictor of word learning. However, there is evidence from Gathercole and Baddeley (1989) that Digit Span may not be an adequate measure of phonological memory. In addition Turner and Engle (1989) hypothesis that Digit Span may allow rehearsal and grouping strategies to circumvent working memory deficits. To more accurately assess working memory, they suggest a more complicated measure which involves a "background" task (see Turner & Engle, 1989).

Future Research
In follow-up research, the experience gained from the present study may serve as a guide. The possible inhibiting effect of equipment in the robot game (if a similar design is used) must be addressed. This can be done by giving the subjects a greater opportunity to become used to the equipment before the items to be scored are presented.
This study used only a brief introduction before the two sample items were presented. Because of time limitations the session with the robot could not be extended to assure the children's familiarity and ease with the robot, although it was evident to the investigators that this would have been desirable in some cases.
Changes in the construction of the target words need to be considered. For the present study an attempt was made to construct two and three syllable words using a wide variety of phoneme combinations.
There is a possibility that the resulting targets words were too complex for fourth-grade students. This difficulty might be addressed by matching the experimental nonsense words to common vocabulary items in the classroom curriculum, (i.e., number of syllables and types of consonant blends used).
In addition changes in the training procedures require consideration. Shorter sessions, with fewer items taught in each, might be preferable. This would allow for multiple sampling; in addition, a greater number of sampling items would result if the data from the sessions were combined.
An interesting variation in training would be to teach the words without the constant use of pictures found in the paired-associate design. This would make the task one which relies more on linguistic processes. This would diminish the differences between the conditions and also allow the difficulty level of the tasks for incidentally learned items to more closely approximate that of the taught items.
The results of this study, if they are confirmed and replicated, will be of interest in studying the linguistic factors common to reading and lexical acquisition, and in opening new avenues to explore in understanding language. In addition, there is potential benefit to educators in understanding the learning behavior of poor readers, as well as in planning and implementing the entire curriculum for the student who has a reading impairment with concomitant linguistic deficiencies.    .oo  [After this introduction to all three words, trial blocks began.
Each card was exposed -in random order -and the child was asked for the name of the pictured object. All errors made by the child were immediately corrected as follows "that is not quite correct, it's a it." If the child offered the definition instead of the word, the prompt was "yes, that's correct, but what is it called?".
The cards, words, and definitions were presented again -in random order. The child was not asked to repeat the words however, following this the cards were displayed (individually) and the child was asked to provide the correct word with the prompt "what is this?". Each trial block of three words was presented at least four times, presentations discontinued only after two consecutive correct trials had been achieved for each word and only after the fourth exposure to the cards and definitions. Even if all words were not successfully learned, there was a maxi' mum of ten trials. No child requested to stop.
Although the wording used was varied slightly (to avoid tedium) the underlined elements of the initial presentation were always included.
All errors were noted and corrected during training. If a subject commented on a feature of a target in the picture (color, shape, etc.) this was acknowledged only with a smile and a nod, or a comment such as "oh yes" without further elaboration or discussion. If a child was persistent in trying to discuss some feature the examiner responded with the comment "perhaps we will learn more about that when Robie takes the trip." Note: In a preliminary testing of the robot it was thought that male voices reproduced more clearly than female voice. It was therefore decided to have Robie be a 'male' robot, referred to as "he".
By the end of this portion of the training the children had learned all of the trained words, and had been exposed to each incidental word twice and only twice.

Definition testing
Immediately after both blocks of words had been gone thorough,the examiner asked for the definition of each word.
Example, "what is a taysum? tell ne everything you know about taysum". A record was kept so that a score could be obtained of the number of definitional elements which the child produced for each word.
At the end of this task, any incorrect pairings of words and definitions were corrected. This was done by stating the elements of the given definition given that went together with the correct target word. For example, if a child said a Biffet is a small, talking, bird the correction was put into the following terms.
"There was something that you had twisted. A Corbealyon is a small bird, a Biffet is something else." Note that if a child included an incorrect descriptor it was not repeated in the correction.
No further teaching of the words was attempted beyond this.

APPENDIX C Journey Game
In the Journey Game, the subjects heard the robot describe what he encounters on the imaginary planet. During the course of the game the robot asked the subjects to supply the six trained words as well as the two incidentally learned words.
Examiner "Now that you know the names of what Robie will find on the planet, we can play the game. Robie is going to tell you what he sees on his imaginary journey. Sometimes he may ask you questions, answer the best that you can. Even if you aren't quite sure, try to help Robie out with your very best guess." When the Robot asked as question the tape was stopped (by remote control), to give the child time to answer. If the child did not respond to the robot's questions, the examiner encouraged with smiles and nods and the prompt "take a guess" .
Every effort was made to encourage response while ensuring that the situation did not become uncomfortable or unpleasant for the child.