Methods for Studying Language Production

Lise Menn, Nan Bernstein Ratner

• Publisher: Psychology Press
• Year: 1999
• eBook ISBN: 9781135676353
• Topic: Personal Development
• Subtopic: History & Theory in Psychology

In this volume, which simultaneously honors the career contributions of Jean Berko Gleason and provides an overview of a broad and increasingly important research area, a panel of highly productive language researchers share and evaluate methods of eliciting and analyzing language production across the life span and in varying populations. Chapters address a wide variety of historical and evolving approaches to data collection for the study of morphosyntax, the lexicon, and pragmatics, both laboratory-based and naturalistic. Special concerns that arise in the study of atypical child development, aging, and second language acquisition are a focus of the discussion.

1
In the Beginning Was the Wug: Forty Years of Language-Elicitation Studies

Nan Bernstein Ratner
University of Maryland

Lise Menn
University of Colorado

All fundamental scientific innovation must marry new ways of thinking with better styles of seeing. Neither abstract theorizing nor meticulous observation can provoke a change of such magnitude all by itself.

—Gould (1998, p. 18)

This book represents some major approaches to collecting language production data from children and young adults, but it is not intended as a complete handbook, which would require a much larger volume. The use of elicited narrative in cross-linguistic comparative developmental study, for example, has been richly elaborated elsewhere, especially by Berman and Slobin (1994), and is not covered here; for narrative in cross-linguistic aphasiology, Menn and Obler (1990) may be useful. Other aspects of production and methods of eliciting specific syntactic forms have been covered in the first several chapters of McDaniel, McKee, and Cairns (1998). For priming studies, see Bock, Lobell, & Morey (1992) and other papers cited in Bock and Levelt (1994). The CHILDES data analysis (see also Sokolov & Snow, 1994) and the CHILDES Bib database can be searched for more information on how database corpora have been used (www//childes.psy.cmu.edu).

This book also is limited in its treatment of cross-cultural issues in data collection: The way people talk to children in front of strangers or while they are being recorded may differ a great deal across cultures, and even more when the stranger is a foreigner. This said, we are delighted to present a wider range of topics in language elicitation than has ever been brought together. All the contributors speak from experience, and we have all learned the art of elicitation by apprenticeship and by trial and error. We have created this book in the hopes of reducing the number of trials that you, the reader, will have to endure.

LOOKING BACKWARD: IN THE BEGINNING WAS THE WUG

Before the late 1950s language production was studied naturalistically, with rare efforts to elicit language knowledge through experimental paradigms. This was particularly true for the study of child language (Bar-Adon & Leopold, 1971). We begin our book on methods of studying language production, begins with a review of the paradigm that changed the nature of developmental psycholinguistic research: Berko Gleason’s (1958)¹ wug test. Through the use of hand-drawn figures of whimsical creatures performing odd acts, Berko elicited plurals, past tenses, and diminutives, among other morphological affixes, by presenting the now classic task: “This is a picture of a wug. Now I have two of them. I have two_____.”

The responses of Berko Gleason’s subjects’ provided the first experimental evidence that young children had productive knowledge of the morphological patterns of their language, generalizable to novel word forms. Prideaux (1985) noted:

Jean Berko, in her pioneering work in language acquisition, set the stage for a great deal of experimental work by providing both conceptual and methodological orientations which have proven very valuable and productive… [this] work constitute[d] a landmark in language acquisition research. It set the stage for a series of studies employing her basic methodology, or variations of it, which have reaped rich rewards over the …years, (pp. 8–9)

Other researchers called the paradigm “ingenious” (Clark & Clark, 1977; Maratsos, 1979) and “a great methodological contribution” (Slobin, 1973) to child language acquisition research. Uniquely, the wug was both a controlled and naturalistic method for eliciting language behavior, at least for children who are used to playing utterance-completion games with their caregivers.

The wug has been used in many ways over the years. A citation search of published research that has utilized “wug like tasks,” as they are sometimes called, reveals extensions of the technique almost immediately following its original publication. Many researchers sought to replicate Berko’s findings of morphological knowledge by young children, extending the original scope of inquiry to speakers of other languages and bilinguals,² children with varying diagnoses of specific language impairment or delay,³ children and adults with mental retardation,⁴ reading-impaired children,⁵ and children from differing socioeconomic and dialect communities.⁶ Even a cursory listing of such efforts reveals their scope and remarkable frequency.

Subsequent work with English-speaking children was able to detail more finely the progress that children make toward mastery of the individual morphemes targeted by the wug task (Anisfeld & Tucker, 1967; Derwing & Baker, 1979; Ervin, 1964; Menn & MacWhinney, 1984; see also Prideaux, 1985, for discussion of unpublished research using the wug technique). Refinement of the design also allowed researchers to evaluate the adequacy of competing accounts of how regular affixation is mastered (see Derwing & Baker, 1986).

Almost immediately following publication of Berko Gleason’s research, clinicians noted the possible extension of the wug paradigm to formal assessment of language performance in children with suspected disorder. The Berry-Talbott Test was but one of many attempts to adapt the wug task as a tool for assessment (see Berry, 1969; more recently, Channell & Ford, 1991; Rubin, Patterson, & Kantor, 1991). However, clinical use of the wug task has been sporadic because of concerns about its relation to spontaneous language performance and the appropriateness of its normative expectations. More recently, as researchers focus attention on familial forms of specific language impairment (SLI; e.g., Gopnik & Crago, 1991), the “wug test” has been used to assess the nature of the underlying deficit in this population (Goad & Rebellati, 1994). Not only has nonsense-word affixation been tested, but the very ability of SLI children to associate the novel word wug with a variety of concepts has been appraised under varying levels of support (see Ellis Weismer & Hesketh, 1993). Similarly, the wug test has been used to assess conceptual generalization in Down syndrome children (Duffy & Wishart, 1994).

A review of citation lists of Berko Gleason’s (1958) Word article over the years gives an interesting mini-view of the evolution of developmental psycholinguistics. References to the wug task appear among the first issues of many seminal journals of our time, such as Applied Psycholinguistics and the Journal of Child Language. In the first 15 years following publication, the article was extensively cited by researchers attempting to validate its utility and extend its findings to nontypical populations. Over time, however, it became apparent that children from infinitely varied backgrounds can in fact generate appropriate morphological inflections for words they have never encountered before; thus, the fact that almost any human being can do the task (albeit some more successfully than others) became much less interesting than the question of how it is accomplished.

A number of alternative hypotheses can explain how children perform on wuglike tasks (see Prideaux, 1985, for a particularly thorough discussion). The nonsense-word paradigm continues to feed active debate on the adequacy of competing models in explaining morphological acquisition and the potential role of modularity in language learning and use (cf. Prasada & Pinker, 1993); thus, after a period of relative quiet, citations of the article burgeoned in the 1990s. The wug has become a sort of rallying point for theories and simulations of how networks learn to generalize to novel events (e.g., Rumelhart & McClelland, 1986) and whether, for instance, the ability of children to inflect is best interpreted by connectionist models or by symbolic, rule-based architectures. The original interpretation of the wug studies, that children do not learn to inflect by rote or mimicry, has been tempered over the years by proposals that some language learning may indeed occur by rote, as well as by analogy and rule (MacWhinney, 1978). As theory now grapples with the results of functional brain imaging tasks, the wug task surfaces in discussions of how best to interpret neuro-imaging data on past tense processing (e.g., the debate between Jaeger et al., 1996, and Seidenberg & Hoeffner, 1998).

Because this is a book about language-elicitation techniques, some analysis of the strengths and weaknesses of the historic technique is appropriate. Extended application of the paradigm brought with it metaanalysis of its utility, including thoughtful discussion of the role that Berko Gleason’s clever original pictured stimuli (or any pictured prompts) played in supporting children’s performance on morphological tasks, as well as the inevitable discussion of whether nonsense-word affixation tasks were legitimate windows into the process by which children learn morphology or are able to demonstrate their competence (cf. Levy, 1987; Lewis & Windsor, 1996). It has become increasingly clear that subjects of all ages and degrees of linguistic sophistication tend, for the most part, to do more poorly on nonsense-word affixation tasks than on those that utilize real words of the language (e.g., Miller & Ervin, 1964; Perez Pereira, 1989); in fact, Berko herself noted this discrepancy in comparing the responses of her young subjects to the nonsense words and selected real-word stimuli. As thesis advisor to the original study, Brown (1973) provided a careful analysis of how to relate Berko Gleason’s data to patterns of affixation in spontaneous language. He suggested important differences between controlled elicitation of grammatical affixes and spontaneous language use, including the child’s need to understand cue words in the experimental carrier phrases (e.g., “yesterday” and “whose”) and ability to focus on relevant aspects of the supporting pictorial stimuli.

Brown noted that Adam, Eve and Sarah’s spontaneous language performance did not pattern exactly like performance by older children in the original wug study. His 1973 text spends more than 10 pages considering whether these differences can be accounted for by differing criteria for judging morpheme productivity. An immediate observation was that Brown’s famous “14 morphemes” list collapsed all allomorphic variations into single morpheme classes (such as plural or past tense), whereas Berko Gleason’s careful sampling of the varying morphophonemic contexts elicited a wide range of performance from children that seemed to depress accurate production of a supposedly early acquired morpheme seen in spontaneous language. In an analysis similar to one that motivated Bryant and Anisfeld’s (1969) experimental inquiry into morphological performance, Brown noted that there are probably at least three levels of difficulty in the original wugset stimuli: those stem+affix pairs that only require the child to know a basic form of affix and to automatically apply voicing assimilation (dogs, cats; tagged, walked); those that pose an additional burden of specifying the voicing (days; played); and those that are phonologically context and morpheme specific (houses; patted). Although obligatory contexts for the more difficult forms are less frequent in children’s spontaneous speech samples than are those for the first group of affixes, Berko Gleason sampled broadly and somewhat evenly across the allomorphs, thereby depressing the children’s apparent level of competence— as well as producing differences in patterns of acquisition between her data and Brown’s subsequent data. In a discussion that foreshadowed later concern over the roles that imitation and rule abstraction might play in language development (cf. MacWhinney, 1978; Pinker, 1991), Brown also speculated that some real-word plurals might, after all, be stored as wholes, whereas others might not.

Brown also noted another important methodological difference between spontaneous language sampling and elicitation tasks that is relevant to many concepts discussed later in this volume. Language sampling tends to isolate patterns in language from large amounts of data produced by small numbers of subjects, whereas Berko Gleason’s study paved the way for eliciting data from large numbers of subjects. Cross-sectional patterns of failure or success of individual children may thus fail to duplicate patterns of mastery that emerge during the development of a single child. Finally, the novelty of the task itself may induce changes in children’s tactics for language use that do not mirror their tendencies in natural conversation. This has also been a chronic concern of researchers over the years in comparing structured vs. naturalistic means of data collection, and numerous contributors to this volume address the concept. After analyzing the data in a number of ways, Brown (1973) was confident that Berko Gleason’s data could be reconciled with the order of acquisition data that flowed from analysis of his own Harvard children. He closed his discussion with the observation that naturalistic and elicited methods “together give us the best chance of discovering the truth” (p. 293). Many of the chapters in this volume agree with his assessment, and they show how researchers have tried to marry structured questions about language with more naturalistic contexts for their exploration.

Other researchers noted limitations in extrapolating children’s productive knowledge of inflectional morphology from wuglike tasks. Levy (1987) suggested that older children may handle the task differently from younger children, who are still acquiring the basics of their language system; her youngest subjects showed closest parallels between the nonsenseword task and affixation of real nouns. Dever (1973) noted a poor correlation between the characteristics of spontaneous speech and performance on wuglike tasks by children with mental retardation, suggesting that the technique has less utility as a predictor of spontaneous language performance in this population. Indeed, the fact that many children with cognitive impairment find the task immensely difficult sheds interesting light on the nature of their generalization difficulties (Ratner, 1998).

The wug task is most often referenced in discussions of the acquisition of inflectional morphology. However, the 1958 nonsense-word task was also used to study the abilities of children to use derivational morphemes. Children in the original study found these morphemes surprisingly difficult to use, and this observation paved the way for numerous other studies in the same area (cf. Derwing & Baker, 1986). It is also sometimes easy to forget that the wug paradigm was only part of the original set of tasks that Berko used to explore children’s morphological development. In the 1958 study, children were also asked to speculate on the origins of common English compound words; interest in children’s capacity to coin and de-compose such terms continues to be of interest to researchers today (see Clark, chap. 4, this volume).

The passage of time has left its mark on the wug. During the 1970s, the genderless wug found itself included in a critique that found too few female characters and role models in speech and language tests (Rabe & Matlin, 1978). Like many visual icons that have endured (e.g., cartoon characters or corporate logos), the wug has undergone some minor facelifts over time (see Fig. 1.1). Although (like many of us) it seems to have put on a little weight over the years, it seems to be busily teaching future generations of students of psycholinguistics. Perusal of textbooks and a web search of psycholinguistics and language acquisition syllabi show an almost universal use of assignments designed to encourage (or require) that students test the paradigm for themselves. (In one interesting version, the instructor has “morphed” the wug into Henry Moore-like statuary, a tribute to the artistry of the original.) As an enduring concept in psycholinguistic research, the wug has become a generic, like Kleenex or Xerox, a concept so basic to what we know and do that increasingly it appears in the popular literature without attribution to its origins (cf. Pinker, 1994; and press discussion by Allman, 1991, and Safire, 1994).

Fig. 1.1. The wug parade. From left to right; Berko Gleason (1958); reprint of the article in Saporta (1961); Berko Gleason & Ratner (1998).

In any field, the discovery of facts is of paramount importance in moving knowledge forward. However, the discovery of techniques is no less important: We can only find keys where there is light to search by. Perhaps no innovation other than the invention of the tape recorder has had such an indelible effect on the field of child language research.

Theories come and go, including those that account for how a child knows that more than one wug is wugs. Although, we are still learning from Berko’s insight th...

Table of contents

• Cover Page
• Title Page
• Copyright Page
• 1: In the Beginning Was the Wug: Forty Years of Language-Elicitation Studies
• I: Eliciting Knowledge Of Language
• II: Gathering Production Data in Naturalistic Settings
• III: Developmental Disorders
• IV: Adult Disorders