Morphological Aspects of Language Processing

Laurie Beth Feldman

• Publisher: Psychology Press
• Year: 2013
• ISBN:
  9781134778331
• Edition: 1
• Topic: Développement personnel
• Subtopic: Compétences de rédaction et de présentation

It is now well established that phonological -- and orthographic -- codes play a crucial role in the recognition of isolated words and in understanding the sequences of words that comprise a sentence. However, words and sentences are organized with respect to morphological as well as phonological components. It is thus unfortunate that the morpheme has received relatively little attention in the experimental literature, either from psychologists or linguists. Due to recent methodological developments, however, now is an opportune time to address morphological issues. In the experimental literature, there is a tendency to examine various psycholinguistic processes in English and then to assume that the account given applies with equal significance to English and to other languages. Written languages differ, however, in the extent to which they capture phonological as contrasted with morphological units. Moreover, with respect to the morpheme, languages differ in the principle by which morphemes are connected to form new words. This volume focuses on morphological processes in word recognition and reading with an eye toward comparing morphological processes with orthographic and phonological processes. Cross-language comparisons are examined as a tool with which to probe universal linguistic processes, and a variety of research methodologies are described. Because it makes the experimental literature in languages other than English more accessible, this book is expected to be of interest to many readers. It also directs attention to the subject of language processing in general -- an issue which is of central interest to cognitive psychologists and linguists as well as educators and clinicians.

PART 1

Visual and Orthographic Issues in Morphological Processing

1

On the Role of Spelling in Morpheme Recognition: Experimental Studies with Children and Adults^*

Bruce L. Derwing, Martha L. Smith,^† and Grace E. Wiebe

Department of Linguistics, University of Alberta

Two experiments were conducted to assess the effect of spelling knowledge on morpheme recognition (MR; i.e., skill in identifying the root morphemes of derived words). Both experiments involved a spelling test and an MR test. The expectation was that subjects who spelled the root portions of the derived words and the root words in the same way (S-spellers, e.g., know/knowledge or *no/*noladge) would be more successful on the MR test than subjects who spelled the root portions differently (D-spellers, e.g., know/*noladge). The first experiment was performed with children, so an MR test based on non-technical vocabulary had to be devised, and a six-question test battery was the result. Two of those questions (both versions of Derwing’s, 1976, “Comes From” test, i.e., “Does the [derived] word come from the [root] word?”) served as the basis for the results reported here. The second experiment involved university students familiar with basic linguistic terminology. For these subjects the MR test consisted simply of asking them whether they thought the derived word contained the same root as the isolated word. While the results showed that S-spellers generally outperformed D-spellers on both MR tests, the results were reliable on less than half of the items, and only a few of these provided unambiguous evidence in favor of the proposed directionality of the effect. The conclusion was nonetheless supported that knowledge of spelling and other education-based factors can sometimes lead to large-scale individual differences in morpheme recognition, as in many other linguistic and metalinguistic skills, and some important implications of this result on standard theoretical and methodological practice in linguistics are discussed.

INTRODUCTION

Over the past half century or so, the field of linguistics in North America has largely been dominated by what Scinto (1986) terms the “phonocentric canon.” The often unquestioned and unstated assumption of this canon is that “the voice is somehow primary and central to language and, by implication, other instantiations of language are only secondary reflections of the voice” (p. 2). Such a bias is clearly reflected in the Bloomfieldian dictum that “[W]riting is not language, but merely a way of recording language by means of visible marks” (Bloomfield, 1933, p. 21).

In other places, however, the central goal of linguistics has been characterized as that of establishing and codifying the “knowledge” that native speakers have of their language (Chomsky, 1965, and elsewhere). Clearly, not all linguistic knowledge derives from speech alone. In particular, since we live in a society where literacy is highly valued and widespread, the typical speaker can hardly avoid exposure to the written form of the language. Indeed, this experience with the written language can begin at a quite early age, as children often begin learning to read and write while they are still undergoing early linguistic and general cognitive development. Nor does one’s knowledge of language cease to grow at any particular age, as even mature adults continue to make frequent discoveries about their language and its use. Moreover, since psycholinguistic research has largely utilized subjects who are literate or in the process of becoming literate, it is of some importance that workers in this field try to sort out intuitively acquired linguistic behavior from education-derived behavior (Jaeger, 1984).¹

One place where there is prima facie evidence of interaction between knowledge derived from hearing/speaking the spoken language and seeing/writing the written one is in the area of morpheme recognition (MR; i.e., that particular aspect of general morphological knowledge that involves the identification of the minimal, meaningful components of morphologically complex words, especially shared roots). Surprisingly little experimental research seems to have been carried out in this area, considering the central role that the morpheme notion has played in most linguistic theoretical accounts² and in psychological models of speech recognition and retrieval. Yet most of what is known or believed about morphemes is actually based more on what linguists have proposed or believed about language forms, rather than on any hard evidence related to the internalized knowledge of speakers. As noted in Derwing (1973), for example, linguists make frequent assumptions about the presumed synchronic morphological relatedness of forms—the assumed morphological connectedness of the words fable and fabulous (Chomsky & Halle, 1968, p. 196) provides a classical case in point—but until quite recently little empirical work has been done to substantiate such claims. (Note that it is not the indubitable historical or etymological relationship between these words that is in question here, but rather the synchronic claim that these words are morphologically related in the minds of contemporary speakers.) It is therefore important for the psychological veracity of such morphological models as proposed in both linguistics and psychology that we understand not only which morphemes are recognized by speakers, but also the bases upon which this recognition takes place.

BACKGROUND

Most prior work on the relationship between knowledge of spelling and morphological knowledge has taken the approach that it is the second that enhances the first, i.e., either that awareness of morphological relationship aids spelling (e.g., Carlisle, 1988; Fischer, Shankweiler, & Liberman, 1985) or that morphological awareness can be an important contributory factor in improving reading skill (e.g., Leong, 1989; and both the Carlisle and Fowler chapters in this volume). This paper takes a somewhat different tack and explores spelling as one factor (of several) that influences the identification of morphological relationships in the first place. While by no means challenging or denying the significance of the prior work, we wish to emphasize the fact that the relationship between spelling and morphological knowledge is undoubtedly reciprocal, and that spelling similarities can give rise to morphological insights perhaps at least as readily as morphological awareness can lead to improvements in the skills of writing and reading words in English (cf. Templeton & Scarborough-Franks, 1985, who also emphasize the important developmental role played by orthographic information in the accessing of some derivational morphological processes). We will give an informal report below on some of the early exploratory work in this area, which focuses largely on methodological issues, followed by the description of a new experimental study that attempts to approach the problem in a more systematic way.

FACTORS IN MORPHEME RECOGNITION

Since the (mental) lexicon is generally assumed to be the place where meaning and phonological form are psychologically linked, and since morphemes are taken to be the minimal or basic elements into which words can be meaningfully segmented, the two most likely factors to affect MR skill are those of phonetic (or phonological) and semantic (or meaning) similarity between the hypothetically related forms (Derwing, 1973, pp. 122-126). Consequently, these were the two factors that were first explored in Derwing (1976). A set of 50 potentially related word pairs that illustrated a small number of high frequency derivational relationships were selected to represent the full range of variation along the dimensions of phonetic and semantic similarity. In addition to the fable/fabulous example already mentioned, this list included such word pairs as teach/teacher, dirt/dirty, quiet/quietly, wonder/wonderful, number/numerous, awe/awful, hand/handle, moon/month, holy/holiday, hand/handkerchief, cup/cupboard, lace/necklace, and puppy/dog, as well as a few non-etymologically related pairs such as ear/eerie and fry/Friday.

Three different tests were devised in this investigation: one to measure perceived similarity in sound between words or parts of words (henceforth PS, for “phonetic similarity”), one to measure the perceived similarity in meaning between them (henceforth SS, for “semantic similarity”), and a third to measure the ability to identify any shared minimal, meaningful component parts (i.e., the MR skill itself). In Derwing’s initial attempt, reasonably straightforward rating tasks were devised for the first two of these, but the third proved to be much more challenging.

To our knowledge, the first experimental attempt to tap the MR skill was done by Berko (1958). Though this classic paper is remembered and cited mainly for its so-called “wug test,” i.e., the use of nonce-words (such as WUG) to assess the productive ability of speakers to deal with the English morphological inflections (such as plural, past, etc.), the paper also included a short section at the end that amounted to an MR test for root morphemes in a short list of English noun compounds (such as Thanksgiving, blackboard, and football). Specifically, what Berko did was to ask her subjects why such words were given the names they were. What she was looking for were examples of so-called “etymological responses,” which would reveal an awareness of the morphological composition of these words. (Thus, as Berko notes, if a subject were to respond to the word Thanksgiving by saying something like, “Thanksgiving is called Thanksgiving because the pilgrims gave thanks,” the last two words would betray an awareness that both give and thank were intrinsic parts of the larger word.) Derwing (1976) showed, however, that without further experimental controls, this test often gave counterintuitive results for such seemingly transparent cases as dirty and quickly, where only a small number of subjects (a mere 15% and 4%, respectively) made explicit mention of the root elements dirt and quick. (What subjects typically did instead was to provide paraphrases or other appropriate circumlocutions.) Though controls can readily be conceived that might have strengthened these results (e.g., by making it clear somehow in the instructions that etymological-like responses were the kind sought), Derwing took a different tack and introduced an entirely new task, called the “Comes From” (CF) test.

In the original version of the CF test, Derwing presented examples of potentially related word pairs (such as teach/teacher, dirt/dirty, quiet/quietly, etc.) to subjects and asked them whether or not they thought that the second word “came from” the first (in some vague and unspecified etymological sense). For example, “Do you think that the word teacher ‘comes from’ the word teach?” A five-point answer scale was used, where 4=No doubt about it, 3=Probably, 2=Can’t decide, 1=Probably not and 0=No way.³

In sharp contrast with Berko’s task, the CF test gave highly satisfactory results in two important respects, for a sample of 65 linguistically naive university undergraduate subjects. First of all, its results were the intuitively correct ones for those word pairs that had transparent morphological connections (i.e., pairs such as teach/teacher, dirt/dirty, quiet/quietly). These pairs were independently rated as being highly similar on both the PS and SS scales and had average MR ratings of 3.95 or above out of a possible 4.00. By the same token, this test also gave the expected results for those other word pairs in which no clear synchronic morphological relationships are intuitive (i.e., pairs such as dog/puppy, ear/eerie, and bash/bashful). These pairs were rated as having very low PS (the first case) or SS values (the latter two) and had MR means below 1.00. Secondly, and just as importantly, the CF test also gave results that sharply contrasted the SS and MR values for word pairs such as dog/puppy and straw/strawberry, where the intuitive semantic similarity is high in the first case and low in the second (SS means of 3.74 and 0.92, respectively), but where the MR means for these same words were almost completely reversed (0.95 and 3.20, respectively). These results indicate that the CF test, as intended, was tapping something very different from mere semantic similarity. Both of these findings, then, speak to the critical validity and reliability issues that any useful experimental task must successfully address.

In general, then, the main finding from this study was that both PS and SS strongly influenced MR judgments, as revealed by the CF task, but that, of the two, the semantic factor was the more important or dominant one.

As noted in Derwing and Baker (1986), however, PS and SS were not the only factors that affected MR judgments, as measured primarily by the CF task. More specifically, the following three categories of factors also seemed to play an important role: (1) construction type: the judged morphemic analyzability of some words seemed to be influenced not merely by the phonetic and semantic transparency of its supposed root but also by such factors as the transparency and/or productivity of the supposed affix. (Thus, while the word pairs wonder/wonderful and hand/handle were rated very much the same with respect to both SS and PS in Derwing’s study, the first, with the highly productive affix, received a higher MR score than did the other); (2) orthographic similarity: the pairs break/breakfast and hand/handkerchief received higher MR ratings than the pair price/precious, though all three pairs were simi...

Table of contents

• Cover
• Halftitle
• Title
• Copyright
• Dedication
• Contents
• Contributors
• Acknowledgment
• Introduction
• Part 1 Visual and Orthographic Issues in Morphological Processing
• Part 2 Semantic Issues in Morphological Processing
• Part 3 Phonological Issues in Morphological Processing
• Part 4 Structural and Statistical Issues in Morphological Processing
• Author Index
• Subject Index