Article
Reference
On the locus of morphological effects in spoken-word recognition:
before or after lexical identification ?
GREBER, Carole, FRAUENFELDER, Ulrich Hans
Abstract
The temporal locus of morphological decomposition in spoken-word recognition was explored in three experiments in which French participants detected the initial CV (LA) or CVC (LAV) in matched monomorphemic pseudosuffixed (lavande) and polymorphemic-suffixed (lavage) carrier words. The proportion of foil trials was increased across experiments (0, 50, or 100%) to delay the moment when participants responded. For the experiment without foils and with the fastest reaction times, a similar pattern of results was obtained for the two types of carrier words. In contrast, an interaction between target type and morphological status of the carrier was obtained when the proportion of foils was higher and the detection latencies were slower.
These results point to a late processing locus of morphological decomposition.
GREBER, Carole, FRAUENFELDER, Ulrich Hans. On the locus of morphological effects in spoken-word recognition: before or after lexical identification ? Brain and Language , 1999 , vol. 68, no. 1-2, p. 46-53
DOI : 10.1006/brln.1999.2083 PMID : 10433738
Available at:
http://archive-ouverte.unige.ch/unige:83828
Disclaimer: layout of this document may differ from the published version.
1 / 1
Article ID brln.1999.2083, available online at http://www.idealibrary.com on
On the Locus of Morphological Effects in Spoken-Word Recognition: Before or After Lexical Identification?
Carole Greber*
,† and Uli H. Frauenfelder*
*Laboratoire de Psycholinguistique Expe´rimentale, Universite´ de Gene`ve, Geneva, Switzerland; and †Department of Cognitive Science, Johns Hopkins University
The temporal locus of morphological decomposition in spoken-word recognition was explored in three experiments in which French participants detected the initial CV (LA) or CVC (LAV) in matched monomorphemic pseudosuffixed (lavande) and polymorphemic-suffixed (lavage) carrier words. The proportion of foil trials was increased across experiments (0, 50, or 100%) to delay the moment when partic- ipants responded. For the experiment without foils and with the fastest reaction times, a similar pattern of results was obtained for the two types of carrier words.
In contrast, an interaction between target type and morphological status of the carrier was obtained when the proportion of foils was higher and the detection latencies were slower. These results point to a late processing locus of morphological decom- position. 1999 Academic Press
Key Words: derivational morphology; sequence detection paradigm; suffixed words; spoken-word recognition.
INTRODUCTION
How can we understand the meaning of a novel morphologically complex word like optionalization? We appear to manage this remarkable feat of recognizing neologisms by deriving their meaning from that of their compo- nent morphological parts. But what happens when we recognize more com-
Support for this research was provided by grants to the first author from the Swiss National Fund for Scientific Research (Grant 81GE-053244) and from the Dr. Medicus Fund (SBS Benevolent) and a grant to the second author from the Swiss National Fund for Scientific Research (Grant 1113-049698.96).
Address correspondence and reprint requests to Carole Greber, Department of Cognitive Science, Johns Hopkins University, Krieger Hall, 3400 N. Charles St., Baltimore, MD, 21218.
E-mail: greber@vonneumann.cog.jhu.edu.
46 0093-934X/99 $30.00
Copyright1999 by Academic Press All rights of reproduction in any form reserved.
MORPHOLOGY IN SPOKEN-WORD PROCESS
47 mon morphologically complex words like unhappy and unforgettable? Do we recognize their stems—happy and forget—and their affixes—un- and -able—separately? Does the mental lexicon store such morphological infor- mation independently? If there is morphological decomposition, when does it takes place (before or after) with respect to lexical identification?
Theories of morphological processing can be ranged along a continuum whose extremities are represented by the decompositional hypothesis (Taft & Forster, 1975) and by the Full Listing Hypothesis (Butterworth, 1983). Current theories have shifted to the middle ground and are distinguishable from each other not so much by whether morphological decomposition takes place but rather when, with respect to lexical identifi- cation, it takes place. Some models (Taft, 1979, 1994) suggest that morphemic representations are activated before the whole word is recog- nized (prelexical morphemic influence). Others (Burani & Caramazza, 1987; Laudanna, Badecker, & Caramazza, 1989, 1992; see also Chialant &
Caramazza, 1995) suppose activation of whole-word representations before morphological information becomes available (postlexical morphemic in- fluence). Still others (Frauenfelder & Schreuder, 1991; Schreuder &
Baayen, 1994) assume that morphological information becomes available either prelexically or postlexically depending on properties of the word to be recognized.
Determining the prelexical or postlexical locus of morphological decom- position requires experimental tools with certain characteristics. First, they must generate on-line measures in which listener’s responses are made dur- ing the processing of the stimulus itself. Second, they must have high tempo- ral resolution such that a precise processing moment can be studied. Finally, the experimenter should be able to shift this moment through experimental manipulation. Ideally then, the experimental task should allow the researcher to follow the time course of processing and determine which processing mo- ment is being tapped into.
Monitoring procedures appear to have some of the properties required to study time-course issues in morphological processing. More specifically, these on-line responses can be based on either a prelexical or lexical repre- sentation (or both) depending on the details of the experimental procedure.
For example, the position of the target in the carrier word or the nature of
the filler items can determine the locus of the response. Until now only the
phoneme-monitoring task has been used to study morphological processing
(Meunier, 1997; Schriefers, Zwitserlood, & Roelofs, 1991). However,
sequence-monitoring (Frauenfelder & Kearns, 1997), which involves detect-
ing larger targets, appears to have some further advantages. In particular,
the presence of foil trials—on which there is a partial match between target
and carrier—can affect the detection latencies and hence the depth of pro-
cessing. In what follows, a sequence detection task with foil manipulations
is used.
TABLE 1
Examples of Test Items for the Three Experiments Defined in Terms of Target and Carrier Types
Carrier types
Suffixed words Monomorphemic words
Target types [lavage (washing)] [lavande (lavander)]
Syllables (LA) Syllabic congruence Syllabic congruence
Stems (LAV) Morphemic congruence Morphemic incongruence
METHOD
Stimuli
The targets and carrier items making up the test trials were identical for the three experi- ments (see examples in Table 1). Carrier items were selected as pairs, matched in word fre- quency and phonemic length. One member was a monomorphemic word with a pseudostem [lavande (lavander)] and the other a suffixed word [lav⫹age (wash⫹ing)]. No word began with a prefix or pseudoprefix. All stems were free. The CV and CVC targets corresponded either to the first syllable, -/LA/ in la/vande or in la/vage, or to the pseudostem or to the stem -/LAV/, respectively, in lavande and lavage.
The proportions of foils and distracter trials varied across the experiments: Experiment 1, 100% distracters; Experiment 2, 50% distracters and 50% foils; and Experiment 3, 100% foils.
In foil trials, the target shared some, but not all of its phonemes with the word carrier, like the target /BU/ in the carrier /bala˜s/ (balance). In distracter trials, the target did not share any phonemes with the carrier, like the target /BU/ in the carrier /kaRe/ (square).
Hypotheses
No Effect of Morphological Structure (Fig. 1c). If morphemic structure does not influence word recognition, as the Full Listing Hypothesis proposes, we should obtain systematically faster RTs for the CV than for the CVC target in both carrier conditions. This pattern is predicted because the target corresponds to the first syllable according to a syllabic account (Mehler, Dommergues, Frauenfelder, & Segui, 1981) or because there are fewer phonemes to detect according to a phonemic account (Cutler & Norris, 1986).
Early Influence of Morphological Structure (Fig. 1a). If morphemic information has a pre- lexical influence on the recognition process, as a decomposition hypothesis assumes (Taft and Forster, 1975), the detection of CVC targets should always be faster, independent of whether they occur in monomorphemic or polymorphemic words. For both carrier word types, a
‘‘blind’’ morphemic analysis is first conducted, thereby producing an effect of morphological structure even for monomorphemic words with a pseudostem. It should be noted that the RT advantage of CVC over CV targets predicted for both carriers could actually be reduced or even eliminated due to the opposing influence of the syllabic/phonemic effects. Therefore, an absence of a difference between two target types could also be consistent with this hypothesis.
Late Influence of Morphological Structure (Fig. 1b). If morphemic information has a post- lexical influence, a detection advantage for CVC targets should be observed not for the carrier with the pseudostem but only for polymorphemic carrier words. Here, the whole word is first identified and then the morphological parts of the word become available. Faster RTs for CV targets are predicted with monomorphemic words for which no morphological structure is
MORPHOLOGY IN SPOKEN-WORD PROCESS
49
Prelexical influence
Postlexical influence
No influence syllables
CV
syllables
CVC target
target
target
(pseudo-)stems
(pseudo-)stems mono poly
mono poly
mono poly
a
b
c
FIG. 1. RT pattern predicted by the three main hypotheses on the nature of the morpholog- ical effect.
represented in the lexicon. Like for the proceding hypothesis, a weaker version that takes into account the counteracting phonemic/syllabic effect would predict a noncrossover interaction between target and carrier type.
Effect of the Proportion of Foils. To examine the time course of morphological processing in greater detail, we manipulated the proportion of foils. By increasing this proportion, we expected to increase RTs, thereby allowing lexical processing to proceed further. The experi- ment with 100% foils should slow responses down the most and tap into later processing. In contrast, the experiment without foils should generate the earliest responses and is the most likely to reveal prelexical morphemic effects. Table 2 summarizes the likelihood of early and late morphological effects as a function of the foil manipulation.
Procedure
The same procedure was used for all experiments. Each experiment started with 10 training trials and continued with 216 test trials. On each trial, participants first heard a beep and then after 250 ms they received the CV or CVC target auditorily. Finally after 400 ms they heard a word that contained (carrier word) or did not contain (filler word) this target in initial position.
They were instructed to push the reaction time button as quickly as possible when they detected the target.
TABLE 2
Probability of Early and Late Morphological Effects Ranging from least likely (⫺⫺) to Most likely (⫹⫹)
Probability of
Percentage of foils Early morphological effect Late morphological effect
0 ⫹⫹ ⫺⫺
50 ⫹ ⫹
100 ⫺⫺ ⫹⫹
The presentation order of each member of the experimental pairs was counterbalanced. Half the participants heard the monomorphemic and then the polymorphemic word; the other half heard these items in opposite order. The same counterbalancing was used for the targets. The targets and carriers were proposed by different native speakers of French to avoid any acousti- cal matching by participants.
EXPERIMENT 1: NO FOILS
Participants
Thirty native French participants from the University of Geneva took part in this experiment.
Results
The results for this experiment as well as those for the other two are pre- sented in Fig. 2. The CV syllabic targets were detected faster than CVC (pseudo)stem targets [F1(1, 29) ⫽ 9.984, p ⬍ .01; F2(1, 32) ⫽ 4.087, p ⫽ .05)]. Further, the RTs for suffixed words were faster than for monomorphe- mic words only in the by-subject analysis [F1(1, 29) ⫽ 6.441, p ⫽ .01]. The interaction between target and carrier types was not significant [F(1, 29) ⫽ 1.14, p ⬍ .1; F2(1, 29) ⬍ 1].
EXPERIMENT 2: 50% FOILS
Participants
Thirty-one French students from the University of Geneva participated in this experiment.
Results
Overall, CV syllabic targets were detected faster than CVC (pseudo)stem targets [F1(1, 30) ⫽ 22.227, p ⬍ .01 and F2(1, 28) ⫽ 13.585, p ⬍ .01].
This difference was significant in planned comparisons only for the
monomorphemic carriers [monomorphemic: F1(1, 30) ⫽ 20.579, p ⬍ .01,
F2(1, 14) ⫽ 10.513, p ⬍ .01; polymorphemic: F1(1, 30) ⫽ 2.394, p ⬎ .1,
F2(1, 14) ⫽ 3.073, p ⬎ .1]. Moreover, the advantage for polymorphemic
MORPHOLOGY IN SPOKEN-WORD PROCESS
51 items effect was only significant on the by-subject analysis [F1(1, 30) ⫽ 4.901, p ⬍ .05]. The interaction between target and carrier types was signifi- cant [F1(1, 30) ⫽ 6.203, p ⬍ .05] and [F2(1, 28) ⫽ 4.197, p ⫽ .05].
EXPERIMENT 3: 100% FOILS
Participants
Thirty-one French students from the University of Geneva participated in this experiment.
Results
Overall, CV syllabic targets were detected faster than the CVC (pseudo-) stem targets on the by-subject and by-item analyses [F1(1, 26) ⫽ 5.939, p ⬍ .05] and [F2(1, 30) ⫽ 3.284, p ⬍ 0.1]. Specific comparisons showed that this effect was only significant for monomorphemic carriers [F1(1, 26) ⫽ 12.004, p ⬍ .01; F2(1, 15) ⫽ 4.035, p ⫽ .06]. The carrier effect was not significant [F1(1, 26) ⫽ 1.342, ns and F2(1, 30) ⬍ 1]. The interaction be- tween target and carrier was significant by subject [F1(1, 26) ⫽ 5.94, p ⬍ .05] but not by item [F2(1, 30) ⫽ 1.817, ns].
Experiment 1
Experiment 2
Experiment 3
CV CVC
CV CVC
CV CVC
target
target
target
mono poly
mono poly
mono poly
a
b
c
450 440 430 420 410 400 390
620 610 600 590 580 570 560
710 700 690 680 670 660 650
429
582
681 673
705 686 617 592 438 431 412
FIG. 2. Results for the experiments containing (a) 0%, (b) 50%, and 100% of foils.
No differences on subjects’ errors were significant in any of the three experiments (error rate was less than 5%).
COMPARISON BETWEEN THE THREE EXPERIMENTS
Detection latencies differed significantly across experiments [F(2, 85) ⫽ 51.622, p ⬍ .001]. Post hoc comparisons showed significant differences be- tween the three experiments (Scheffe´ test, p ⬍ .001); participants detected targets faster in the first experiment (m ⫽ 429 ms) than in the second (m ⫽ 598 ms) and faster in the second than in the third (695 ms).
DISCUSSION
These experiments indicate that morphological structure has an influence on the recognition of morphologically complex spoken words. Indeed, differ- ent patterns of results were obtained for monomorphemic words with a pseu- dostem and truly morphologically complex words. For the former, all three experiments showed a phonological/syllabic effect (CV target ⬍ pseudostem CVC target) independent of the proportion of foils. In contrast, the results for the polymorphemic carriers varied as a function of the foil manipulation.
An advantage for the detection of the CV targets was obtained only in the absence of foils. When the proportion of foils was high (50 or 100%), this advantage disappeared due to the effect of the morphological match. In sum, the fact that the morphological effect was limited to polymorphemic words and to slower RTs clearly points to a late processing locus for morphology.
Although these results suggest that morphological information becomes available postlexically, they do not allow us to decide between different mor- phological formats. Indeed, they are compatible with accounts which assume that whole words and morphemes are represented as separate entries in the lexicon (AAM model, Laudanna, Badecker, & Caramazza, 1989) or that words are represented as a network of morphemes (Fowler et al., 1985;
Schriefers et al., 1991). They are inconsistent with either extreme on the morphological continuum represented by Taft’s model and the Full Listing Hypothesis.
The present study has shown that the sequence monitoring paradigm com-
bined with the manipulation of the proportion of foils gives the researcher
some control over response speed and depth of processing. It provides a
useful tool for studying the time course of spoken-word recognition. When
applied to prefixes this procedure could provide some important insights
since it has been argued (Cole´, Beauvillain, & Segui, 1989; Meunier, 1997)
that prefixed and suffixed words are not processed the same way. Further
research is clearly necessary to determine the exact location of French mor-
phological processing on this continuum.
MORPHOLOGY IN SPOKEN-WORD PROCESS
53
REFERENCESBurani, C., & Caramazza, A. 1987. Representation and processing of derived words. Language and Cognitive Processes, 3/4(2), 217–227.
Butterworth, B. 1983. Lexical representations. In B. Butterworth (Ed.), Language production (Vol. 2, pp. 257–294). London: Academic Press.
Chialant, D., & Caramazza, A. 1995. Where is morphology and how is it represented? The case of written word recognition. In L. B. Feldman (Eds.), Morphological aspects of language processing (pp. 55–76). Hillsdale, NJ: Erlbaum.
Cole, P., Beauvillain, C., & Segui, J. 1989. On the representation and processing of prefixed and suffixed derived words: A differential frequency effect. Journal of Memory and Lan- guage, 28, 1–13.
Cutler, A., Mehler, J., Norris, D., & Segui, J. 1986. The syllable’s differing role in the segmen- tation of French and English. Journal of Memory and Language, 25, 385–400.
Fowler, C. A., Napps, S. E., & Feldman, L. B. 1985. Relations among regularly and irregularly morphologically related words in the lexicon as revealed by repetition priming. Journal of Memory and Language, 13, 241–255.
Frauenfelder, U. H., & Schreuder, R. 1991. Constraining psycholinguistic models of morpho- logical processing and representation: The role of productivity. In G. Booij & J. van Marle (Eds.), Yearbook of morphology. (pp. 665–674). Amsterdam: Kluwer.
Frauenfelder, U. H., & Kearns, R. 1997. Sequence detection. In F. Grosjean & U. H. Frau- enfelder (Eds.), A guide to spoken word recognition paradigms. East Sussex, UK: Psy- chology Press.
Laudanna, A., Badecker, W., & Caramazza, A. 1989. Priming homographic stems. Journal of Memory and Language, 28, 531–546.
Laudanna, A., Badecker, W., & Caramazza, A. 1992. Processing inflectional and derivational morphology. Journal of Memory and Language, 31, 333–348.
Mehler, J., Dommergues, J.-Y., Frauenfelder, U. H., & Segui, J. 1981. The syllable’s role in speech segmentation. Journal of Verbal Learning and Verbal Behavior, 20, 298–305.
Meunier, F. 1997. Morphologie et traıˆtement du langage parle´. The`se de doctorat, Paris V.
Schreuder, R., & Baayen, R. H. 1995. Modelling morphological processing. In L. B. Feldman (Eds.), Morphological aspects of language processing (pp. 131–156). Hillsdale, NJ:
Erlbaum.
Schriefers, H., Zwitserlood, P., & Roelofs, A. 1991. The identification of morphological com- plex spoken words: Continuous processing or decomposition? Journal of Memory and Language, 30, 26–47.
Taft, M. 1979. Recognition of affixed words and the word frequency effect. Memory and Cognition, 7(4), 263–272.
Taft, M. 1994. Interactive-activation as a framework for understanding morphological pro- cessing. Language and Cognitive Processes, 9(3), 271–294.
Taft, M., & Forster, K. I. 1975. Lexical storage and retrieval of prefixed words. Journal of Verbal Learning and Verbal Behavior, 14, 638–647.
