The Puliguous Effects of Context Length on Incidental Word Learning

Most of the words in a person‘s vocabulary are acquired during the course of natural reading, through a process called incidental word learning. Incidental word learning is mediated by a number of factors concerning word, context, and reader characteristics. The current study investigated the effects of one such factor, context length, using a combination of word learning and eye-tracking methodologies. Participants read novel words embedded in either sentence or paragraph contexts while their eyes were tracked. Eye movement measures, primarily gaze duration and total time, demonstrated that novel words had longer reading times than real words, and that words in sentence contexts had longer reading times than words in paragraphs. The increased processing devoted to words in sentences may result in an increased rate of learning for novel words. This effect was not verified by the posttest vocabulary assessment, which did not find significant word learning. The low rate of learning observed in this study is attributed to the absence of any benefit from partial word knowledge, information that readers possess about words that are unknown or largely unknown. Research has shown that words are easier to learn if they are partially known. This study argues that partial word knowledge is more influential and pervasive than previously considered, and that as a result, the rates of learning reported by many incidental word learning studies may be inflated. Context and Word Learning 2 The Puliguous Effects of Context Length on Incidental Word Learning Despite decades of research on vocabulary size, there is little consensus over the number of words in the English language or the average number of words in a person‘s vocabulary at any given age. Nagy and Anderson (1984) offer the largest estimates to date, identifying over 88,000 distinct words in the language and reporting the vocabulary of an average high school student to be between 25,000 and 50,000 words. The heterogeneity of these estimates reflects variation in how word knowledge is tested, the source used to represent the corpus of words, and how word is defined; Nagy and Anderson, for example, define a distinct word as a word family, a group of morphologically related words whose meanings can be easily determined if one word from the family is known. Children, even those who do relatively little reading, encounter between 16,000 and 24,000 unknown words every year (Nagy, Anderson, & Herman, 1987). Based on the estimate that the average high school senior knows 40,000 words, a child‘s vocabulary must grow by approximately 3,000 words a year (Nagy, Herman, & Anderson, 1985; Nagy & Herman, 1984). Direct Vocabulary Instruction How are these new words learned? When children first acquire language, words are learned through oral conversation. However, typical oral conversation uses only the most common words from the available corpus — even the conversations of college graduates contain fewer rare words than the average children‘s book (Beck & McKeown, 2007; Cunningham & Stanovich, 1991; Hayes & Ahrens, 1988). Another potential source of learning is direct vocabulary instruction, an activity that takes Context and Word Learning 3 place in most primary and many secondary schools (Scott, Jamieson-Noel, & Asselin, 2003). A classic example of classroom vocabulary instruction is the keyword lesson, a method in which a set of words to be learned is selected from a story that the class is reading. The words identified as keywords are usually semantically unfamiliar to students, important for comprehension of the story, and useful in a variety of contexts (Beck, McKeown, McCaslin, & Burkes, 1979). A wide range of exercises are used to teach keywords, with varying efficacy (e.g., Beck, Perfetti, & McKeown, 1982; Biemiller & Boote, 2006; Jenkins, Stein, & Wysocki, 1984; Scott et al., 2003). Instructional methods that expose students to the semantic, orthographic, phonologic, syntactic, and contextual features of words have been found to be positively correlated with vocabulary growth, and more effective than methods that focus on only a single feature (for details on two such methods, Rich Instruction and Anchored Word Instruction, see Beck & McKeown (2007) and Juel, Biancarosa, Coker, & Deffes (2003); for a general discussion, see Stahl & Fairbanks, 1986). Although direct vocabulary instruction can result in vocabulary growth, it is evidently not the sole source of vocabulary learning. Observational studies and reviews of reading programs have indicated that in most classrooms, very little time is devoted to vocabulary instruction (Beck et al., 1979; Durkin, 1979; Juel et al., 2003, Scott et al., 2003). When instruction does take place, it is often one-dimensional and superficial, concentrating primarily on the rote memorization of definitions (Scott et al., 2003). While instructional methods could be improved in many cases, even rigorous vocabulary programs only cover a few hundred words a year, an amount insufficient to account for the vocabulary growth observed during the school years Context and Word Learning 4 (Beck et al. 1982; Nagy & Herman, 1984). Additionally, vocabulary knowledge continues to grow throughout adulthood, when instances of direct vocabulary instruction are generally thought to be infrequent. If direct instruction is inadequate, how is the remaining portion of a person‘s vocabulary learned? Incidental vs. Intentional Word Learning It appears that a substantial amount of vocabulary is learned during natural reading (e.g., Nagy & Herman 1984, 1987). Although it had long been assumed that people acquire new words in this manner, research on this method of learning, which is commonly called incidental word learning, has accumulated in only the past thirty years. For the purposes of this study, I will use the definition of incidental word learning asserted by Swanborn and de Glopper (1999): the derivation and learning of new word meanings during familiar reading circumstances. Explicitly, incidental word learning requires deriving the meanings of novel words through inferences made possible by predictive contexts, and retaining the meanings of those words in memory. These activities are done in order to comprehend the text at hand, rather than with any intent to learn new words. Incidental word learning is often compared with word learning that occurs through the intentional derivation of meaning from context. For clarity, I will refer to this paradigm as intentional word learning. In the literature, this is often (though inconsistently) labeled derivational learning or learning from context, terms that could equally apply to incidental learning; ̳intentional‘ emphasizes the conscious nature of the task in contrast with incidental learning. In studies of intentional word Context and Word Learning 5 learning, participants are instructed to either (a) derive the meanings of unknown words and provide definitions while the supporting context is still in front of them, or (b) use contextual information to actively try to learn unknown words, which requires remembering the meanings for later testing. Meaning derivation is also necessary for incidental learning, and research on the two processes has been highly interrelated. Intentional word learning is more conducive to experimental study, and therefore has been more thoroughly researched. Findings from this research are sometimes extrapolated to incidental learning. The true extent of their similarity, however, has yet to be determined. Incidental word learning is generally considered to be a much more challenging process for readers (e.g., Jenkins & Dixon, 1983; Nagy et al., 1987). Direct experimental comparison has found that words are more easily acquired intentionally than incidentally (Konopak at al., 1987). Instruction in word meaning derivation may, with difficulty, be able to improve intentional learning (Fukkink (2002) and Fukkink and de Glopper (1998) found improvement; Kranzer (1988) and Carnine, Kameenui, and Coyle (1984) did not). Incidental learning, on the other hand, does not improve with instruction in intentional derivation (Fukkink, 2002; Kranzer, 1988). Though no benefit of instruction has been observed, studies have observed that the ability to derive meaning from context partially predicts the ability to learn incidentally (Fukkink, 2002; Kranzer, 1988). Conceptually, it seems unlikely that intentional and incidental word learning are entirely disassociated. Careful comparison may be worthwhile for understanding both processes, individually and in relation to each other. Context and Word Learning 6 Influences on Incidental and Intentional Word Learning Sternberg and Powell (1983) and Jenkins and Dixon (1983) describe a number of mediating variables that might affect word learning from context. These hypothesized variables occur at the level of word, context, and participant. Today, most of these variables have been examined in the context of both intentional and incidental learning. While not all have been uniformly supported as influential factors, I will borrow their organization of the potential manipulations in word learning studies. Word-level variables. Word-level variables concern characteristics of the unknown words to be learned, and include concreteness, conceptual difficulty, and part of speech. Concrete words (e.g., dog and table) are more easily learned than abstract words (e.g., love and justice) both intentionally (Fukkink, Blok, & de Glopper, 2001) and incidentally (Schwanenflugel, Stahl, & McFalls, 1997). Similarly, words with conceptually simple meanings (i.e., words that are exact synonyms for other words) are easier to learn than words with abstruse meanings both intentionally (Daneman & Green, 1986) and incidentally (Nagy et al., 1987). The results for part of speech have been inconsistent. Very young children are much quicker to acquire nouns than other parts of speech, as are adults in simulations of child-like learning (see Piccin & Waxman (2007) for an in-depth discussion of this ̳noun advantage‘ and of the Human Simulation Paradigm). Despite these findings, studies of intentional learning have rarely found an effect of part of speech. Under certain conditions, an inhibition of noun-learning compared with verbs and adjectives has been seen (Bolger, Balass, Landen, & Perfetti, 2008). Schwanenflugel et al. (1997) likewise Context and Word Learning 7 found that children were worse at incidentally learning nouns than verbs or adjectives. They conjectured that this may occur because as children learn lowerfrequency words, the heuristics they originally used to learn nouns (i.e., nouns refer to distinct whole objects) fail, requiring an adjustment in strategy. Context-level variables. Context-level variables concern characteristics of the texts in which the unknown words are embedded, and include informativeness (constraint), conceptual difficulty, number of word occurrences, and variability. When context is more informative or constrained to predict the meaning of the word (sometimes called ̳considerate‘ context), words are easier to learn through intentional derivation (Bolger et al., 2008; Daneman & Green, 1986; Fukkink, 2002). Likewise, when misleading contexts are introduced, learning becomes more difficult (Frishkoff, Collins-Thompson, Perfetti, & Callan, 2008). The effects of context informativeness have been much less unanimously reported for incidental learning. Herman, Anderson, Pearson, and Nagy (1987) found a facilitative effect of informativeness (as did Elley (1989), for contexts that were heard rather than read), but others have failed to find any significant learning benefit from increased context informativeness (Nagy et al., 1987; Schwanenflugel et al., 1997). The counterintuitive finding that informativeness does not affect incidental word learning has yet to be fully explained. However, Nagy et al. (1987) suggest that the effect of informativeness found in other studies may actually be confounded with the effect of the conceptual difficulty of the context. They found this factor to significantly influence incidental learning: contexts containing simple concepts result in more learning than conceptually difficult contexts. The impact of context Context and Word Learning 8 conceptual difficulty may also explain the finding that intentional derivation is easier if a synonym for the unknown word is provided in the context than if derivation requires making an inference (Carnine et al., 1984). Much debate surrounds the question of how many times a word must be read in order to learn its meaning. Under some conditions, minimal exposure (a single occurrence) may be sufficient to result in a measurable degree of word learning (e.g., enough knowledge of meaning to select the correct definition from a set) for both intentional (Bolger et al., 2008) and incidental learning (Nagy et al., 1987; Nagy et al., 1985). However, the number of words learned increases and the understanding of known meanings improves with multiple exposures during both intentional (Beck et al., 1982; McKeown, Beck, Omanson, & Perfetti, 1983) and incidental learning (Biemiller & Boote, 2006; Jenkins et al., 1984). When unknown words occur multiple times in intentional learning tasks, learning is improved if the words appear in a variety of different contexts rather than in the same context multiple times (Bolger et al. 2008). An earlier study by Konopak et al. (1987) failed to find facilitative effects of multiple exposures or context variability. However, number of occurrences and variability are generally manipulated across contexts. When, as in the Konopak study, these variables are manipulated within a single context, number of occurrences and variability may be confounded with informativeness (Konopak et al. noted that when a word in their study occurred multiple times, the context in each instance was generally less informative than the context for a word which occurred only once). Beck et al. (1982) suggest that multiple encounters in varied contexts lead to a greater number of Context and Word Learning 9 retrieval cues, an outcome that should likewise benefit incidental learning. This variable has yet to be adequately tested for incidental learning, however, because as Swanborn and de Glopper (1999) explain, repeated exposure to unknown words across contexts may artificially draw attention to those words. Participant-level variables. Participant-level variables concern individual differences, primarily in reading skill. In the following studies, the classification of skilled and less skilled was generally made according to a standardized measure of comprehension or a battery of verbal ability assessments. Skilled readers learn more words than less skilled readers both intentionally (Bolger et al., 2008; Freebody & Anderson, 1983; Frishkoff et al., 2008; Kranzer, 1988) and incidentally (Herman et al., 1987; Jenkins et al., 1984; Kranzer, 1988, Swanborn & de Glopper, 2002). Vocabulary learning in turn improves the comprehension of contexts containing both taught and untaught words. Because of this relationship, the performance of skilled and less skilled readers becomes even more disparate over time (Beck et al., 1982; Jenkins et al., 1984; McKeown et al., 1983; Stahl & Fairbanks, 1986). It has been theorized that skilled and less skilled readers employ different strategies when deriving word meanings from context (McKeown, 1985). Using ERP recordings, Frishkoff, Perfetti, and Westbury (2009) found different patterns of brain activation in skilled and less skilled readers during intentional derivation. Curtis (1987) suggests that qualitative, if not strategic, differences in word knowledge may persist even for words that are well-known to both skilled and less skilled readers: when asked to define words, skilled readers tend to use abstract descriptions while less skilled readers tend to give a contextualized response. Context and Word Learning 10 The differing-strategies theory is supported by the fact that the effects of some contextual variables differ according to reading skill, including reading purpose, proximity of informative context to the unknown word, and spacing of context reading. Skilled readers learn more words incidentally when instructed to read for the purpose of answering comprehension questions, but less skilled readers do not (Swanborn & de Glopper, 2002). Less skilled readers are better at intentionally deriving word meaning when the essential informative context is in close proximity to the word to be learned than when the context is separated, while proximity of context does not affect skilled readers (Cain, Oakhill, & Lemmon, 2004). When reading multiple contexts containing a single unknown word, skilled readers learn more intentionally if the contexts are interspersed with other contexts (spaced practice) than if they are read in immediate succession (massed practice); the learning of less skilled readers is not influenced by spacing (Frishkoff et al., 2008). However, it has also been argued that inequalities in both comprehension and word learning ability can be explained by differences in global cognitive functions such as memory. Daneman & Green (1986), proponents of this theory, found that reading span, a measure of working memory, is positively correlated with the ability to deriving meaning