Reading and Writing

, Volume 25, Issue 6, pp 1239–1258

An analysis of differential response patterns on the Peabody Picture Vocabulary Test-IIIB in struggling adult readers and third-grade children

Authors

    • University of Cincinnati
  • Daphne Greenberg
    • Georgia State University
  • Rihana S. Williams
    • Emmanuel College
Article

DOI: 10.1007/s11145-011-9315-x

Cite this article as:
Pae, H.K., Greenberg, D. & Williams, R.S. Read Writ (2012) 25: 1239. doi:10.1007/s11145-011-9315-x

Abstract

This study examines the Peabody Picture Vocabulary Test-IIIB (PPVT-IIIB) performance of 130 adults identified as struggling readers, in comparison to 175 third-grade children. Response patterns to the items on the PPVT-IIIB by these two groups were investigated, focusing on items, semantic categories, and lexical features, including word length, word class, and word frequency. The score distributional properties of the two groups were different, but there were similarities and differences found in different response patterns. Analyses of word length, word class, and word frequency in the two groups’ performance resulted in counterintuitive findings for the adult participants. The struggling adult learners’ vocabulary repertoire might have been shaped by their real-life experiences rather than formal schooling. Implications and future research directions are discussed.

Keywords

Receptive vocabularyDifferential response patternsSemantic referents and categoriesLexical characteristics

The importance of vocabulary as a critical factor for literacy development has been well documented in the literature (Beck & McKeown, 1983; Lee, 2011; Pearson, Hiebert, & Kamil, 2007; Storch & Whitehurst, 2002; Tannenbaum, Torgesen, & Wagner, 2006). The National Reading Panel (2000) has indicated that the development of reading comprehension is contingent upon children’s knowledge of the word’s lexical representations and meanings because text comprehension is limited without proficient vocabulary skills. Research suggests that much of vocabulary is acquired as a byproduct of literacy activities and wide reading (Nagy & Herman, 1987). Vocabulary deficiency is a major cause of students’ academic failure, because a lack of adequate vocabulary impedes the ability to meet reading demands, and, in turn, poor comprehension skills inhibit the ability to continue vocabulary growth.

Estimating root word vocabulary growth in normative and advantaged populations, Biemiller and Slonim (2001) have found evidence for a common sequence of vocabulary acquisition in which individual words are known in a moderately fixed order in all grades from kindergarten to grade 5. They have speculated that a child’s subsequent vocabulary gains and size can be estimated based on the total number of words learned by the child at grade 2. Early language development is closely linked to the influence of parenting on vocabulary development and a child’s vocabulary exposure. A relative socio-economic status (SES) advantage is an important factor affecting a child’s vocabulary growth. According to Hart and Risley (1995), an average child from a welfare household hears on average 616 words per hour, compared to the 2,153 words a child from a professional household hears. Between the first and third grades, it is estimated that economically disadvantaged students’ vocabularies increase by about 3,000 words per year, compared to middle-class students’ vocabularies increase of about 5,000 words per year (Adams, 1990).

Vocabulary skills are also developed through instruction in school. Biemiller (2001) emphasizes the importance of schooling on acquiring vocabulary by suggesting that the majority of vocabulary is attained through direct instruction up until the sixth grade. This suggests that if students were not responsive to direct instruction during early grades, the deficient vocabulary skills in the elementary schools can be carried over to an individual’s vocabulary repertoire later in development. Incidental encounters are also important and include oral conversations, viewing various forms of media, listening to classroom lectures and discussions, and independently reading various forms of text (Bolger, Balass, Landen, & Perfetti, 2008). This suggests that responsiveness to both formal and independent learning situations is necessary for vocabulary growth.

Vocabulary knowledge can be characterized by the modality in which words are processed. Receptive vocabulary refers to words that are processed through visual or auditory modalities, while expressive vocabulary includes words that are used to externalize thoughts and ideas via output modalities (i.e., speaking or writing; Beck & McKeown, 1991). Receptive vocabulary is a set of words that an individual can assign meanings when listening to a speaker or while reading (Kamil & Hiebert, 2005). The period of receptive vocabulary growth is initiated very early in life when children are preliterate (Bates, Bretherton, & Synder, 1988) and continues to grow drastically during the schooling period because vocabulary gains become salient through written texts (Grabe, 2004). Receptive vocabulary size is typically larger than that of expressive vocabulary (Nagy & Scott, 2000; Pearson et al., 2007), because students can probe contextual meanings from unfamiliar words when listening and reading. As Pearson et al. (2007) state, “comprehension normally precedes production and … additional cueing systems (various textual and contextual aids) are available to individuals during language reception, but not during production” (p. 284). Performance on receptive vocabulary measures tends to be more highly correlated with reading comprehension skills than on expressive vocabulary measures (Oakland, De Mesquita, & Buckley, 1988). This paper focuses on receptive vocabulary.

The representations of words stored in an individuals’ receptive repertoire are multifaceted (Nagy, Anderson, & Herman, 1987). The organization and retrieval of words may vary depending on semantic knowledge stored in long-term memory and the ability to make use of the connections among words. The organization and retrieval of word knowledge can also be affected by various factors relating to the words themselves or the learner. There are several lexical factors including word class (i.e., part-of-speech or syntactic categories for nouns, verbs, and adjectives), frequency, and meaning that contribute to the multidimensional nature of organization within memory. For example, nouns may be easier to learn than verbs and adjectives because their referents connect to concrete visual systems (i.e., cat or desk) and link directly to innate features of the perceptual faculty (Golinkoff & Hirsh-Pasek, 2008). Gentner (1982) claimed a universal noun primacy in children’s early vocabulary, indicating that nouns are learned before verbs because nouns denote names for entities. However, it should be noted that research conducted in other languages, such as Chinese and Korean, showed salient verb dominance in children vocabulary acquisition (Choi & Gopnik, 1995; Tardif, 1996).

The mental lexicon is also organized by frequency of occurrences within spoken and written text as well as word length. Studies of visual word recognition and reading have demonstrated that words with high frequencies are accessed more quickly than words with low frequencies in a variety of reading related tasks (for a review, see Monsell, 1991). Pinker (1987) has pointed out that uncommon words in a million-word corpus of text have weak memory entries, and, hence, they should be harder to retrieve upon request. It is speculated that the mental dictionary is organized differently on the basis of the frequency of exposures. With respect to the relationship between the word frequency and the surface form of the word, Zipf’s law has demonstrated that the most common words are short in length and uncommon words are longer (Kintsch, 1994; Leopold, 1998).

Words are interrelated within the lexicon according to their meaning. Many English words have multiple meanings (e.g., homographs: bat, a baseball vs. bat, an animal), with different functions (e.g., heteronyms: present, verb vs. present, noun) in different sentences, texts, and contexts. Word knowledge is also intertwined from the meaning of one word (e.g., mind) to that of other words in variations (e.g., mindful, mindfully, mindfulness, mindless, mindlessly, mindlessness, mindset, absent-minded, mind reading, mind’s eye, etc.). Words are also hierarchically organized according to categories which vary according to the perceptual features of the referent (e.g., fruits, colors, animals, etc.; Nelson & Nelson, 1990), temporal order, functional and spatial relations among aspects of everyday life experiences (e.g., clothing that is worn outside or animals found at the zoo; Luciarello, Kyratzis, & Nelson, 1992; Riva, Nichelli, & Devoti, 2000). Furthermore, the quality of these interconnections across syntactic classes, frequencies, lengths, and meanings is reflected in many factors, such as the speed of retrieval processes, precision of the word choice and usage, and adaptation to different modes (e.g., receptive vs. expressive) and different print purposes (e.g., formal vs. informal occasions; Dale & O’Rourke, 1986; Nagy & Scott, 2000).

The complexity of word knowledge places different demands on learners, because the simple storage of a word in memory does not guarantee the ability to use a word in reading and writing. Certain characteristics of the learner including age and schooling have been demonstrated to influence one’s flexibility with word knowledge. Estimates of vocabulary breadth suggest that children have a much smaller vocabulary size compared to adults. Anderson and Freebody (1983) estimate that elementary school children have a vocabulary size of 2,562 words compared to a college sophomore’s vocabulary of 200,000 words. For children, differences in exposure to formal instruction have been demonstrated to influence the density of vocabulary. For example, Nelson and Nelson (1990) found that kindergartners who had attended formal preschool programs provided significantly more category items in response to slot-filler and taxonomic prompts than children who had not attended preschool. However, similar findings have not been found for adults. For example, Manly, Byrd, Touradji, Sanchez, and Stern (2004) found that the number of words produced in response to taxonomic prompts did not differ between illiterate and literate adults.

Along with the importance of lexical characteristics that influence word knowledge, an understanding of the way students recognize sounds of words and retrieve semantic information of the given words also provides valuable information. An analysis of response pattern is useful because it provides windows into vocabulary processes. Examination of error patterns in vocabulary measures allows us to scrutinize how students retrieve conceptual and semantic cues to process decontextualized single words in their vocabulary repertoire. A pattern of miscues is more informative than an individual error because deviations from optimal responses are not random occurrences of mistakes or errors but show systematic patterns of retrieval and representations of a student’s vocabulary repertoire.

Although much is known about children’s vocabulary knowledge and learning, there is a paucity of research studies that have investigated struggling adult readers’ vocabulary knowledge and how they perform on vocabulary measures in relation to children’s word knowledge. Struggling adult readers provide a theoretically interesting ecological sample because their age and exposure to formal schooling trajectories predict different vocabulary densities. The current study attempted to broaden our understanding of a child–adult comparison by analyzing differential response patterns on the PPVT-Third Edition Form B (PPVT-IIIB) in struggling adult readers and third-grade children. Specifically, the objective of this study was to examine the receptive vocabulary performance on the items and categories of stimulus words in the PPVT-IIIB between struggling adult readers and young readers. Since we were interested in multiple dimensions of students’ performance on the PPVT-IIIB with respect to the connection between oral language and its meaning, the focal point of the response-pattern analysis centered on deviations from the expected response (e.g., age and grade norms), the proportion of accurate and inaccurate responses, and the pattern of the variations students produced on individual words. Four research questions were addressed in this study.
  1. 1.

    How do struggling adult readers, compared to children, perform overall on the PPVT-IIIB?

     
  2. 2.

    How do struggling adult readers, compared to children, perform on individual PPVT-IIIB items?

     
  3. 3.

    How do struggling adult readers, compared to children, perform across various semantic categories of PPVT-IIIB words?

     
  4. 4.

    How do the lexical characteristics (i.e., word length, word class, and word frequency) influence the performance of the struggling adult readers, compared to children?

     

Due to differences in age, experience, and exposure to formal instruction, we expected the struggling adult readers and young readers to differ in the pattern of their receptive vocabulary performances and responses.

Method

Participants

The participants were 305 children and struggling adult readers; 130 adults and 175 third-grade children. Of the 130 adults, 120 (92%) adult participants were African Americans, and the rest were Caucasians. Of the 175 third-grade children, 138 (79%) were Caucasians, and 37 (21%) were African Americans. The gender distribution was more balanced than that of race: 48% males and 52% females for the child group; 27% males and 73% females for the adult group.

Struggling adult readers were recruited as part of a larger intervention study1 and read isolated words at the third- and fifth-grade levels on the Letter/Word Identification subtest of the Woodcock-Johnson Tests of Achievement-III (WJ-III; Woodcock, McGrew, & Mather, 2001). Their reading skills were at the low end–within the 10th percentile—as indicated by their standard scores of reading fluency and passage comprehension subtests of the WJ-III (standard scores = 78.80, 76.50, respectively). The mean age of the adults was 39.14 (SD = 16.10), and ranged from 16 to 73 years of age. On average, their highest grade completed in school was 9.11 (SD = 1.43), ranging from 5 to 12 years. Thirty-five percent of the participants had history of receiving welfare.

Third-grade children were enrolled in general education classrooms during the spring in a suburban school district in a moderately sized city in the Southeast of the USA. Their mean age was 9.70 (SD = .58), ranging from 8 years 5 months to 11 years 4 months. These third-grade children were recruited as part of a larger investigation related to the dimensions of vocabulary knowledge as described in Tannenbaum et al. (2006).

Measure

The PPVT-IIIB (Dunn & Dunn, 1997) was administered. The PPVT-III was designed to measure an individual’s receptive vocabulary knowledge for Standard American English, and was normed on American English speakers with an age range from 2 to 90 years old. Following the standard testing protocol, the examinee was shown four pictures while the examiner said a single stimulus word. The examinee verbally or nonverbally indicated which picture best represented the stimulus word. According to the examiner’s manual of the PPVT-IIIB, internal consistency alphas for the age groups from 2 to 90 ranged from .92 to .98 (median: .95), and split-half ranged from .86 to .96 (median: .94). The test–retest coefficients ranged from .91 to .94.

Procedure

The PPVT-IIIB was administered by trained graduate students. Prior to the testing of the struggling adult readers, testers were given extensive training by the project’s psychometrician on test administration and adult literacy sensitivity issues. Testers practiced in front of the psychometrician, and were observed by the psychometrician before they were permitted to test the adults on their own. For testing the third-grade children, testers were trained and then were required to demonstrate proficient administration before testing the participants.

Analysis plan

As a preliminary analysis, indices of skewness, kurtosis, and box plots were checked to examine performance distributional properties of the scores on the PPVT-IIIB. The scores were symmetrically distributed throughout the range of values. Since the standard scores are contingent upon age and revealed a floor effect for the adult participants, the proportions (percentages) of correct or incorrect numbers were used whenever applicable. Raw scores and the proportions (percentages) of correct or incorrect numbers are sample dependent, ordinal level measures, and as such, they are not expressed on a linear scale, and can be distorted on both ends of the scoring continuum. For comparability’s sake, the percentages were transformed into natural logarithmic values to place the data on an equally linear interval scale. Whenever statistical analyses were performed, the log values were utilized. When percentages are easier than log scores to demonstrate the differences, percentages are used for the sake of presentation.

Since this research was a descriptive study, only basic statistical analyses were run. Bivariate correlation, paired t-statistics, and chi-square techniques were performed using the natural log scales.

Results

Overall performance (research question 1)

The raw scores were computed as the total number of correct responses. As indicated in Table 1, the struggling adult readers performed much lower (approximately two standard deviations below the mean) than the average adult normative group, while the children showed slightly higher mean scores than that of the normative group. Specifically, the mean standard score for the adults was 73.72 (SD = 9.12), ranging from 43 to 94, while the mean standard score for the children was 101.32 (SD = 7.43), ranging from 79 to 110. There was a significant difference between the two groups in terms of the standard scores (t(129) = 16.92, p = .000), which was not surprising given that standard scores are sensitive to age. The mean difference of the standard scores was 27.6, meaning that the children performed better than the adult participants when compared to their normative reference group.
Table 1

The mean raw scores and standard scores and standard deviations on the PPVT-IIIB

 

Raw

SS

M

SD

M

SD

Struggling adult readers (n = 130)

139.45

17.74

73.72

9.11

Third-grade children (= 175)

132.99

14.06

101.32

7.43

SS standard score, M mean, SD standard deviation

As indicated in Table 1, the mean correct raw score for the adults was 139.45 (SD = 17.74; range = 93–184) and for the third-grade children was 132.99 (SD = 14.06; range = 95–169), yielding a nonsignificant difference between the two groups (t(129) = −1.79, p = .076). Since the two groups were not different in terms of their correct raw scores, an examination of the profiles of the two groups’ performance was possible.

Histograms of the frequencies of individual scores indicate that the adults had a wider range of the score distribution than that of the children, and included outliers at the lower end. The graphs depict heavy concentrations of the scores in the middle, although, compared to the children, more adults’ scores fell near the low end than their counterparts. Overall, all are assumed to be adequately represented by the midpoint (Fig. 1).
https://static-content.springer.com/image/art%3A10.1007%2Fs11145-011-9315-x/MediaObjects/11145_2011_9315_Fig1_HTML.gif
Fig. 1

Histograms of the PPVT raw scores and standard scores by group

Responses to the individual items were also compared across the two groups of the participants. Both groups performed near 100% correct on items below 72, which were expected items to be known by children age 6 or 7 years of age. The children showed a more systematic pattern than the adult counterparts with respect to item performance (see Fig. 2).
https://static-content.springer.com/image/art%3A10.1007%2Fs11145-011-9315-x/MediaObjects/11145_2011_9315_Fig2_HTML.gif
Fig. 2

Item response rate by the adult struggling readers and the third-grade children

Figure 2 displays the correct rates (percentages) on each item across the groups. Although, as a whole, there was no statistically significant difference (t = 1.58, p > .05) between the groups, some items demonstrated notable differences between the two groups. The item that showed the most discrepant rate, regardless of the direction (i.e., which group performed better on which item), between the groups was item 141 (lubricating, diff. = 56.77), followed by item 146 (embracing, diff. = 50.64). Three items (item 136, physician, diff. = 45.67; item 131, banister, diff. = 42.90; item 142, amphibian, diff. = 42.13) showed different rates between 42 and 46%. There were 11 items that showed different rates in the range between 30 and 39%, while 12 items ranged from 20 to 29%, and 21 items between 10 and 19%. Thirty-six items showed marginal differences.

Item response patterns (research question 2)

On the basis of the result of the first research question, we selected a subset of the items to be analyzed from item 73 through item 156. This decision was made for two reasons: (1) Item 73 (the first item of Set 7) was the administration starting point for ages 8–9 (the ages of most of the children), and item 156 (the last item of Set 13) was the starting point for ages 17 through adult, if a basal was established; (2) The participants correctly answered items 73 and lower at the 95% or higher rate level (adults’ mean correct percentage: Set 5: 99.29% and Set 6: 95.64%; children’s mean correct percentage: Set 5: 99.95% and Set 6: 99.19%), and most of them failed to produce correct responses beyond item 156 (adults’ mean correct percentage: Set 14: 31.60% and Set 15: 17.44%; children’s mean correct percentage: Set 14: 12.76% and Set 15: 7.05%).

As indicated earlier, the percentage of incorrect responses for each item was transformed into natural log values. When we looked at response patterns and profiles of item difficulty variability on item 73 through item 156, there was more of a non-linear trend on the lower items for the adults than for the children. As seen in Fig. 3, the children’s performance showed a consistent pattern on the lower items and a more fluctuated fashion on the higher items. In contrast, the adults’ performance was variable across the gamut.
https://static-content.springer.com/image/art%3A10.1007%2Fs11145-011-9315-x/MediaObjects/11145_2011_9315_Fig3_HTML.gif
Fig. 3

Item difficulty profiles of incorrect responses for the two groups (Item # 73–156)

Performance differences on each item by the two groups were also examined. In general, the adults showed poorer performance relative to the children in the lower end (i.e., items 73–108) of the range of items from 73 to 156. Conversely, children demonstrated poorer performance relative to the adults in the higher end (i.e., items 109–156) of the range. To illustrate this pattern, a few notable trends are discussed further. Item 74 showed the most balanced correct performance (sanding; 94% adults vs. 94% children). However, the children’s correct response accuracies exceeded those of adults for items 77 (antlers), 80 (walrus), 92 (shore), 102 (artic), 104 (pelican), 105 (pyramid), and 108 (hoof). The adults’ correct responses exceeded those of children’s for the remaining items in a batch of item 109 and higher with the exception of items 126 (archaeologist, 49% adults vs. 76% children), 129 (pod, 28% adults vs. 66% children) and 142 (amphibian; 36% adults vs. 78% children) which may represent items that are more constrained to academic text.

Semantic category profiles (research question 3)

All the items under consideration were classified into semantic categories provided by the AGS publishing (publisher of the PPVT, acquired by Pearson). The mean percentage of performance on each category was calculated for each participant and then it was transformed into the log scale. The greatest gap in the log value of correct responses was found in the category of plants. However, due to the low number of occurrences of the categories, a t test statistic for statistical differences between the two groups’ performance was not performed except for the actions and animals categories. A statistically significant difference was found between the two groups’ performances in the actions category (t(129) = 2.86, p < .01), with the adults making fewer errors than the children. Table 2 exhibits the children’ and adults’ log values of correct responses.
Table 2

Means and standard deviations of logarithmic values of correct responses by category (Item # 73–156)

Category

# Items

Adult learners (n = 130)

Third-grade children (= 175)

Mean diff.

M

SD

M

SD

Actions

23

4.37

.28

4.16

.49

.21*

Adjectives

4

4.21

.23

4.20

.21

.01

Animals

10

4.18

.26

4.32

.37

−.14

Body parts

2

4.55

.05

4.54

.03

.01

Books/money

2

3.87

.06

3.52

.83

.35

Buildings

2

4.29

.21

3.99

.35

.3

Clothing/accessories

2

4.33

.28

3.97

.79

.36

Emotions

1

4.30

na

4.23

na

.07

Food

2

4.57

.03

4.40

.17

.17

Geographical

7

4.22

.42

4.36

.48

−.14

Household

3

4.03

.40

3.87

.58

.16

Musical

1

4.13

na

4.38

na

−.25

People

4

4.28

.19

4.17

.34

.11

Plants

2

3.40

.78

4.31

.16

−.91

Shapes

4

4.30

.17

4.30

.32

0

Tools

3

4.20

.07

3.90

.57

.3

Toys/recreations

1

4.14

na

4.28

na

−.14

Vehicles

7

4.18

.38

3.93

.72

.25

Workers

4

4.16

.28

3.74

.78

.42

Minus (−) sign indicates that the children performed better than the adults in the given category

Mean diff. mean difference, M mean, SD standard deviation

* p < .01

n/a not available because this category consisted of only one item

Lexical characteristics (research question 4)

To determine whether the adults and the children differed in their vocabulary performance as a result of the lexical characteristics of words, three primary lexical characteristics were examined: word length, word class, and word frequency. First, we obtained the mean numbers of letters and syllables per word by the semantic category in order to examine the surface characteristics of the vocabulary in the PPVT-III. The categories of body part and plants contained the lowest number of letters (mean # of characters per word: 4), while the category of workers held the highest number of letters per word (mean number of characters per word: 11.25). The categories of plants and toys had one syllable each, while the action category had a mean number of 2.75 per word. There was a variation in the occurrence of complex words (Edit Central, 2009) in the category of the words. The categories of body parts, clothing, musical, plants, and toys/accessories did not include any complex words, whereas the categories of animals and workers contained four complex words (see Table 3).
Table 3

Category profiles of letters, syllables, and complex words

Category

Mean # letters

Mean # syllables

# Complex words

Complex words

Actions

8.7

2.75

2

Calculating, lubricating

Adjectives

7.75

3

3

Abundant, fictional, irregular

Animals

6.7

2.3

4

Predatory, pelican, amphibian, liberated

Body parts

4

1.5

0

n/a

Books/money

6.5

2.5

1

Currency

Buildings

7.5

2.5

1

Banister

Clothing/accessories

5

1.5

0

n/a

Emotions

10

3

1

Astonished

Food

9

3

2

Nutritious, beverage

Geographical

6.43

2.14

2

Tropical, peninsula

Household

7

2.33

1

Porcelain

Musical

7

2

0

n/a

People

7.25

2.25

1

Salutation

Plants

4

1

0

n/a

Shapes

6.25

2.25

2

Pyramid, parallel

Tools

7.67

2.67

1

Speedometer

Toys/recreations

4

1

0

n/a

Vehicles

6.53

2.14

1

Aviation

Workers

11.25

3.5

4

Archaeologist, construction, physician, agriculture

Letter mean # of letters per word, Syllable mean # of syllables per word, n/a not applicable

The number of letters per word was significantly correlated with that of syllable per words (r = .87, p < .001). However, a nonsignificant correlation was found between the number of letters and word frequency and between the number of syllables and word frequency. Interestingly, the word length did not significantly correlate with the children’s nor the adults’ vocabulary scores.

Next, we examined the word class of the subset of items, along with word frequency. The frequency of words represents the number of observations in one million words. Since the PPVT is a measure of receptive vocabulary for Standard American English (Dunn & Dunn, 1997), we used the word frequency of the Corpus of Contemporary American English (Davies, 2009). This corpus combines word frequencies from spoken and written text. The word class (i.e., part-of-speech) was classified on the basis of a word’s form, meaning, or grammatical function in context. Nouns were the highest occurrences in the item pool (49 items, 58%), followed by verbs (22 items, 26%) and adjectives (13 items, 16%, respectively). The mean frequencies were as follows: nouns: 4,398 occurrences per million (SD = 7,985; range: 101–47,303), adjectives: 2,758 occurrences (SD = 2,234; range: 523–8,256), and verbs: 1,193 occurrences (SD = 877; range: 15–3,295), respectively. There was a significant difference in frequency distributions when compared among the three different grammatical forms of words: χ2 = 25.07, df = 2, p = .000. There were significant correlations between the children’s and adults’ performance by the three word classes: r = .84 for adjectives, r = .55 for nouns, and r = .70 for verbs, all ps < .01. However, there were no significant correlations of the participants’ performance with the word class and word frequency.

Overall, the adults showed higher scores on the verb items than the adjectives and the nouns. Adults performed significantly higher than the children on the verbs: t(21) = 3.10, p = .007. There was a significant difference in the adult learners’ performance on the three word classes (i.e., adjectives, nouns, and verbs; F(2,81) = 4.00, p = .022). However, no significant difference was found for the children in the performance on the three word classes (F(2,81) = .93, p = .400). Figure 4 shows the word class (i.e., part-of-speech) profile of the correct responses for the two groups.
https://static-content.springer.com/image/art%3A10.1007%2Fs11145-011-9315-x/MediaObjects/11145_2011_9315_Fig4_HTML.gif
Fig. 4

Part-of-speech log values of correct responses for the two groups (items #73–156)

Next, analyses were performed with the restricted age equivalency (AE) group breakdown. There were minor differences in the patterns shown by the two groups. For those whose AE scores fell in the range of 8–10 years, the two groups showed the lowest performance on the noun items. The adults’ performance on the verb items was consistently higher than that of children. Compared to the adults, the children of the AE 8 and AE 10 groups showed slightly better scores on adjectives and nouns. The gaps between the two groups got smaller in the 10-year AE group than 9-year AE group (see Fig. 5).
https://static-content.springer.com/image/art%3A10.1007%2Fs11145-011-9315-x/MediaObjects/11145_2011_9315_Fig5_HTML.gif
Fig. 5

Word class performance by AE group

Lastly, an item’s word frequency was examined. It was interesting to find that the PPVT-IIIB item arrangement was not reflective of the American corpus-derived word frequency, given that test items were typically arranged by the degree of difficulty (Miller & Lee, 1993) and that many research studies have found frequency effects in reading (Monsell, 1991). We looked at the word frequency and the students’ performance on each item in the subset of items (items 73–156). As seen in Fig. 6, items 73–156 did not exhibit an ascending pattern in the word frequency and item allocation. The mean frequency was 3,305 per million (SD = 6,303), ranging from 5 (item 150, orating) to 47,303 (item 89, river). As expected due to the inconsistent pattern of the word frequencies in the item allocation, there were no significant correlations between the word frequency and the participants’ correct responses on items in this subset (r = .13, p > .05 for the children; r = −.08, p > .05 for the adults). However, there was a significant correlation between the adults’ and the children’s performance on items 73–156 (r = .60, p = .000). Among the low end of the PPVT-IIIB items chosen for analyses, item 80 (walrus) was observed the lowest frequency which was 234 observations in a million words, followed by item 75 (hyena) with 301 observations. Regarding the high frequency word items, item 76 (pair) had 19,524 observations and item 89 (river) had 47,303 observations per one million words.
https://static-content.springer.com/image/art%3A10.1007%2Fs11145-011-9315-x/MediaObjects/11145_2011_9315_Fig6_HTML.gif
Fig. 6

Word frequencies of items 73–156

Discussion

Vocabulary size is considered to be a good predictor of language and literacy competence after controlling for individual differences in terms of SES, gender, birth order, and ethnicity (Lee, 2011). Since vocabulary skills are found to be an important attribute of success in reading (Beck & McKeown, 1983; Lee, 2011; Pearson et al., 2007; Storch & Whitehurst, 2002; Tannenbaum, Torgesen, & Wagner, 2006), a wealth of research studies has highlighted the critical role of vocabulary in many different elements of text knowledge and comprehension (Hirsch & Nation, 1992; Nagy & Herman, 1987). However, previous research has focused on vocabulary skills without deeply exploring and analyzing the item characteristics, such as forms (i.e., word length), word classes, and semantics of the vocabulary under consideration. This study represents an attempt to fill the void in the vocabulary literature by comparing two populations who vary vastly in their life experiences. Specifically, this study investigated struggling adult readers’ global receptive knowledge, their knowledge of particular semantic categories, syntactic word-class categories, and performance differences as a result of an item’s lexical frequency, in comparison to third-grade children’s response patterns.

Overall performance

Results indicated that the adults slightly outperformed the third grade children in their total correct responses. However, the adults performed very poorly when compared to the PPVT-IIIB normative group. In fact, the adults showed a mean AE score of 11.40 years.

Item response patterns

About 40% of the total items showed discrepant success rates (84 items out of 204). Since this includes items before basal and after ceiling items, the percentage may be misleading. Therefore, we only included items 73 through 156 for further analyses. The examination of PPVT-IIIB items 73 through 156 with respect to the item profiles and the participants’ response patterns demonstrated differences of item performance between the struggling adult readers and third-grade children. We observed a salient difference between the groups in this subset of items. The third-grade children followed a response pattern which is typically observed on vocabulary measures with gradually ascending item difficulty. They performed better on the lower items that were within their age range and poorer on higher items that were outside of their age range. Interestingly, the struggling adult readers showed a reverse pattern; they produced more errors on the lower items and produced more correct responses on the higher items, compared to the children’s response pattern. This was counterintuitive for the adults in that it was expected that the items were arranged in a sequential difficulty that is indexed by word length, polysemy, date of entry into the English language, and word frequency (Miller & Lee, 1993). The repertoire of struggling adult readers’ vocabulary skills may be atypical because word frequency and word difficulties are typically correlated (Pinker, 1987).

Item difficulty differences between children and adults may result from different exposures in daily milieu, social settings, surroundings, and school environments. For example, the greatest gap in the item accuracy between the two groups was found in the response to lubricating with adults having more accurate responses. The children responded more accurately than the adults to words, such as archaeologist, pod, and amphibian, words which are more likely to appear in school textbooks. The source of these kinds of differences might point to discrepancies in everyday word usage, word familiarity, and world knowledge (Anderson & Nagy, 1992). In other words, an instant retrieval to a stimulus may be shaped by the different degree of environmental and formal instruction experiences between these adults and children. Because adults are different from emergent children readers in terms of prior knowledge and real-life experience, it is expected that the struggling adults show different vocabulary knowledge from the children. What makes the results of this study important is that the profile of the struggling adult learners’ vocabulary skills is documented.

Semantic profiles

The results of the semantic categories were similar to those of the PPVT-IIIB item response patterns. The examination of semantic categories yielded evidence that adults’ real-life experiences, relative to children’s, influenced their receptive vocabulary performance. The semantic categories of the PPVT-IIIB yielded a differing response pattern between the adults and the children. The categories in which the adults tended to demonstrate stronger vocabulary knowledge seems to be directly related to daily activities, and they are comparatively weak in school-related vocabulary, whereas the children appeared to show stronger vocabulary in the areas which are considered academic constructs covered in school. The adults’ vocabulary skills seem to be related to the negative collateral consequences which result from their unsuccessful schooling and experiences. Given that the adults’ reading skills were at the 10th percentile rank, insufficient formal childhood education may be a risk factor for a future vocabulary deficit. A lack of sufficient schooling may be a lifetime barrier that is hard to overcome with other alternative experiences.

Lexical characteristics

The data did not support several previous findings related to form and syntactic class. A non-linear pattern of response accuracy did not support a general tendency that shorter words appear more frequently than longer words in human activities, and, as a result, they are learned more easily (Leopold, 1998). Therefore, the locus of the performance difference between the children and the struggling adult readers did not seem to be related to word length.

With respect to syntactic word-class category, the results of this study showed that density, depth, and width of the vocabulary, in terms of part-of-speech, did not support Gentner’s (1982) noun primacy hypothesis. According to Gentner (1982), the vocabulary repertoire of nouns is stronger than that of verbs. Because nouns denote names for entities, they are easier to learn than verbs, which entail names for variable states and actions. Both the children and the adults showed stronger vocabulary skills with adjectives and verbs than nouns. This finding is contrary to the noun primacy over verbs (Gentner, 1982). The irregularities of the performance across the syntactic word-class types as well as the different pattern of performance between the adults and children might have to do with the asymmetric distribution of words in each word class in the PPVT-IIIB. The mean difference between the adults and children was greater in the verbs than nouns. One explanation might be a wide range of word frequency in the noun target words included in the PPVT-IIIB. Although nouns share referential commonalities with terms of objects and entities, the infrequent nominal words included in the PPVT-IIIB can be hard to be learned and processed.

Among adjectives, nouns, and verbs, the adults received the lowest score on the noun stimuli, while the scores of verbs and adjectives were not different each other. A response pattern of certain word classes may have implications for the organization of mental lexicons, semantic contents, and grammatical forms. Adjectives, nouns, and verbs function differently at the level of structure (form). The noun as a lexical category is used to name or identify a person, place, thing, event, substance, trait, feature, quality, quantity, concept, and the like, referring to entities that are organized in taxonomic hierarchies or structured taxonomic relationships. It can occur as the main word in the subject of a clause or the object of a verb or a preposition. The adjective adds detail or description to the noun in front of the noun (attributive adjective; e.g., She has seen aninterestingmovie) or after the verb, be, seem, or look (predicative adjective; e.g., The movie isinteresting). The verb conveys action or a state of being. The noun may be less predictable than other grammatical categories, such as adjectives and verbs, because the noun is a member of a larger, open, lexical category than other word classes (Tardif, 1996), especially for those who have limited formal instruction. The results indicate that certain responses to word-classes or semantic attributes categories relate to the role of words in the sentence. The noun may not be a descriptor of equal weight as the adjective and the verb on a syntactic scale.

Because a fundamental consistency was expected between an individual’s word familiarity and the frequency with which words appear in the language corpus, we also examined the word frequencies of the PPVT-IIIB items, using the corpus-derived word frequency of the Standard American English (Davies, 2009). The results indicated no significant correlations between the word frequencies of the target words and the participants’ item performance. It was interesting to find that the PPVT-IIIB items were not reflective of the American corpus-derived word frequency, given that test items are typically arranged by the degree of difficulty and that many research studies have found frequency effects on a variety of word recognition tasks (Monsell, 1991). This may be attributable to differences in the frequency metrics used to select items on the PPVT-IIIB and the metric used for analysis in the current study. Despite sampling differences between past and recent frequency estimates, the lower items on the PPVT-IIIB might reflect commonly occurring words in elementary grade academic texts; therefore, the third-grade children had more advantages on those items than the struggling adult readers.

Pearson et al.’s (2007) analysis of the PPVT-III may provide an alternative explanation for our findings. They explain that this test is designed to cover a wide range of levels of vocabulary knowledge in a single assessment. Due to its wide range of coverage, the PPVT-III requires many “obscure” words to accommodate test-takers at the high end of the vocabulary knowledge scale (Pearson et al., 2007). This characteristic may explain the low frequency effect in the sample of this study. Another explanation for our findings is that the PPVT-III assesses oral receptive vocabulary knowledge, which requires an identification of the picture that matches the word spoken by the examiner. Since the corpus data on which this study relied are based on both oral and written frequency, performance on the oral receptive vocabulary measure using the PPVT-IIIB may have been masked by the oral and written word copora.

Findings indicated that word frequency was not correlated with the numbers of graphemes nor with the number of syllables. The word rank is typically a reverse proportion, and there is a general tendency that the word length decreases as the word frequency increases (Leopold, 1998). The results of this study did not support the Zipf’s law (i.e., the most frequent words are short words, and rare words are longer ones; Kintsch, 1994; Leopold, 1998). This finding can presumably be explained in two ways. First, as seen Fig. 6, it can be attributed to the irregularities found in the item allocation of the PPVT-IIIB. Secondly, it might have to do with the methods of measuring word lengths. Given that English is based on the phonological structure, measuring a word length by graphemes or syllables does not offer a complete account of the phonological structure of words. Consequently, counting the phonological unit may be more reflective to relative importance of different segments than graphemes and syllables. However, measuring the word length by the phonological unit can also be problematic because there is no universally accepted criterion available. Overall, this study demonstrated a differing pattern in receptive vocabularies between children and struggling adult readers. This differential prototype may result from various factors such as exposure, environment, experience, and availability of cognitive resources between the two groups. This study contributes to the field in that the results provide evidence for similarities and differences of adults’ and children’s receptive vocabulary knowledge.

Limitations and future research

Although this study attempted to identify multi-faceted dimensions of the vocabulary performance of the adults and third-grade children, several limitations should be noted. First, the participants’ race was asymmetric; the majority of the adult participants were African-American, while the majority of the third-grade children were Caucasian. Since vocabulary gain is closely related to home environment and parent vocabulary usage (Hart & Risley, 1995), the influence of SES and home culture on vocabulary learning warrants further investigations. Secondly, there was no reading measure available to directly compare the reading achievement of the two groups. The availability of the participants’ reading scores would have allowed us to look further into the interaction and interface of vocabulary and reading at different levels. Thirdly, the struggling adult readers were not compared to a control group of adult expert readers. Although the comparison of struggling adult learners to third graders who were at their similar vocabulary level offers valuable information with respect to the adults’ strengths and weaknesses, a contrast of struggling adults’ and typical adults’ vocabulary profiles would have enabled us to more fully understand the nature of struggling adult readers’ vocabulary knowledge vis-à-vis children and adults who read at “expected” levels.

Future research studies should attempt to replicate this study while correcting for the noted limitations in this study. Other research ideas are also suggested. For example, in order to identify the nature and the locus of response patterns on receptive vocabulary tests, an analysis of the foil items using different criteria, such as phonological and semantic neighborhoods to the target words, visual representation of the stimuli, and semantic relatedness of the target and foil items, would deepen the explanation of vocabulary knowledge and strategy usage. Next, a contrast of responses to words in both receptive vocabulary and expressive vocabularies with respect to error patterns could provide valuable information on the observed relationship between correct and incorrect responses in both receptive and expressive vocabulary knowledge. Finally, effects of different types of vocabulary assessment, such as discrete versus embedded, selective versus comprehensive, contextualized versus decontextualized (Pearson et al., 2007) on test-takers’ performance should be investigated. The dimension of assessment tools and method effects will offer a better understanding what vocabulary measures actually assess and what vocabulary measures can or cannot assess for adults and children.

Footnotes
1

Research supported by the Eunice Kennedy Shriver National Institute of Child Health and Human Development, the National Institute for Literacy, and the US Department of Education—grant # R01 HD43801.

 

Copyright information

© Springer Science+Business Media B.V. 2011